openkylin
/
linux
mirror of https://gitee.com/openkylin/linux.git
9
0
Fork 0
Code Issues Projects Releases Wiki Activity
linux/fs/nfs/super.c

2851 lines
70 KiB
C
Raw Normal View History

NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
/*
* linux/fs/nfs/super.c
*
* Copyright (C) 1992 Rick Sladkey
*
* nfs superblock handling functions
*
Switch to a valid email address... Signed-off-by: Alan Cox <alan@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-10-27 23:19:48 +08:00
* Modularised by Alan Cox <alan@lxorguk.ukuu.org.uk>, while hacking some
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
* experimental NFS changes. Modularisation taken straight from SYS5 fs.
*
* Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
* J.S.Peatfield@damtp.cam.ac.uk
*
* Split from inode.c by David Howells <dhowells@redhat.com>
*
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
* - superblocks are indexed on server only - all inodes, dentries, etc. associated with a
* particular server are held in the same superblock
* - NFS superblocks can have several effective roots to the dentry tree
* - directory type roots are spliced into the tree when a path from one root reaches the root
* of another (see nfs_lookup())
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
*/
#include <linux/module.h>
#include <linux/init.h>
#include <linux/time.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/string.h>
#include <linux/stat.h>
#include <linux/errno.h>
#include <linux/unistd.h>
#include <linux/sunrpc/clnt.h>
#include <linux/sunrpc/stats.h>
#include <linux/sunrpc/metrics.h>
NFS/SUNRPC: use transport protocol naming Instead of an { address family, raw IP protocol number }-tuple, use the newly-defined RPC identifier when creating clients in the upper layers. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:48:23 +08:00
#include <linux/sunrpc/xprtsock.h>
NFS: support RDMA mounts Adds hooks to the string-based NFS mount to support an "rdma" protocol option. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:49:41 +08:00
#include <linux/sunrpc/xprtrdma.h>
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
#include <linux/nfs_fs.h>
#include <linux/nfs_mount.h>
#include <linux/nfs4_mount.h>
#include <linux/lockd/bind.h>
#include <linux/smp_lock.h>
#include <linux/seq_file.h>
#include <linux/mount.h>
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace As noted in the previous patch, the NFSv4 client mount code currently has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in nfs4_path_walk() will fail with an error. This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the namespace on the server, then uses the ordinary VFS and NFS path lookup code to walk down the mount path in that namespace. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-23 03:09:14 +08:00
#include <linux/mnt_namespace.h>
#include <linux/namei.h>
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
#include <linux/nfs_idmap.h>
#include <linux/vfs.h>
#include <linux/inet.h>
NFS: Add support for AF_INET6 addresses in nfs_compare_super() Refactor nfs_compare_super() and add AF_INET6 support. Replace the generic memcmp() to document explicitly what parts of the addresses must match in this check, and make the comparison independent of the lengths of both addresses. A side benefit is both tests are more computationally efficient than a memcmp(). Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:57:38 +08:00
#include <linux/in6.h>
#include <net/ipv6.h>
NFS: handle interface identifiers in incoming IPv6 addresses Add support in the kernel NFS client's address parser for interface identifiers. IPv6 link-local addresses require an additional "interface identifier", which is a network device name or an integer that indexes the array of local network interfaces. They are suffixed to the address with a '%'. For example: fe80::215:c5ff:fe3b:e1b2%2 indicates an interface index of 2. Or fe80::215:c5ff:fe3b:e1b2%eth0 indicates that requests should be routed through the eth0 device. Without the interface ID, link-local addresses are not usable for NFS. Both the kernel NFS client mount option parser and the mount.nfs command can take either form. The mount.nfs command always passes the address through getnameinfo(3), which usually re-writes interface indices as device names. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:37:01 +08:00
#include <linux/netdevice.h>
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
#include <linux/nfs_xdr.h>
[PATCH] include/linux/nfsd/const.h: remove NFS_SUPER_MAGIC NFS_SUPER_MAGIC is already defined in include/linux/magic.h Signed-off-by: Adrian Bunk <bunk@stusta.de> Cc: Neil Brown <neilb@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-02-12 16:53:40 +08:00
#include <linux/magic.h>
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
#include <linux/parser.h>
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
#include <asm/system.h>
#include <asm/uaccess.h>
#include "nfs4_fs.h"
#include "callback.h"
#include "delegation.h"
#include "iostat.h"
#include "internal.h"
NFS: Define and create superblock-level objects Define and create superblock-level cache index objects (as managed by nfs_server structs). Each superblock object is created in a server level index object and is itself an index into which inode-level objects are inserted. Ideally there would be one superblock-level object per server, and the former would be folded into the latter; however, since the "nosharecache" option exists this isn't possible. The superblock object key is a sequence consisting of: (1) Certain superblock s_flags. (2) Various connection parameters that serve to distinguish superblocks for sget(). (3) The volume FSID. (4) The security flavour. (5) The uniquifier length. (6) The uniquifier text. This is normally an empty string, unless the fsc=xyz mount option was used to explicitly specify a uniquifier. The key blob is of variable length, depending on the length of (6). The superblock object is given no coherency data to carry in the auxiliary data permitted by the cache. It is assumed that the superblock is always coherent. This patch also adds uniquification handling such that two otherwise identical superblocks, at least one of which is marked "nosharecache", won't end up trying to share the on-disk cache. It will be possible to manually provide a uniquifier through a mount option with a later patch to avoid the error otherwise produced. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Steve Dickson <steved@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Daire Byrne <Daire.Byrne@framestore.com>
2009-04-03 23:42:42 +08:00
#include "fscache.h"
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
#define NFSDBG_FACILITY NFSDBG_VFS
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
enum {
/* Mount options that take no arguments */
Opt_soft, Opt_hard,
Opt_posix, Opt_noposix,
Opt_cto, Opt_nocto,
Opt_ac, Opt_noac,
Opt_lock, Opt_nolock,
Opt_v2, Opt_v3,
NFS: support RDMA mounts Adds hooks to the string-based NFS mount to support an "rdma" protocol option. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:49:41 +08:00
Opt_udp, Opt_tcp, Opt_rdma,
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
Opt_acl, Opt_noacl,
Opt_rdirplus, Opt_nordirplus,
NFS: Add the mount option "nosharecache" Prior to David Howell's mount changes in 2.6.18, users who mounted different directories which happened to be from the same filesystem on the server would get different super blocks, and hence could choose different mount options. As long as there were no hard linked files that crossed from one subtree to another, this was quite safe. Post the changes, if the two directories are on the same filesystem (have the same 'fsid'), they will share the same super block, and hence the same mount options. Add a flag to allow users to elect not to share the NFS super block with another mount point, even if the fsids are the same. This will allow users to set different mount options for the two different super blocks, as was previously possible. It is still up to the user to ensure that there are no cache coherency issues when doing this, however the default behaviour will be to share super blocks whenever two paths result in the same fsid. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
Opt_sharecache, Opt_nosharecache,
NFS: add "[no]resvport" mount option The standard default security setting for NFS is AUTH_SYS. An NFS client connects to NFS servers via a privileged source port and a fixed standard destination port (2049). The client sends raw uid and gid numbers to identify users making NFS requests, and the server assumes an appropriate authority on the client has vetted these values because the source port is privileged. On Linux, by default in-kernel RPC services use a privileged port in the range between 650 and 1023 to avoid using source ports of well- known IP services. Using such a small range limits the number of NFS mount points and the number of unique NFS servers to which a client can connect concurrently. An NFS client can use unprivileged source ports to expand the range of source port numbers, allowing more concurrent server connections and more NFS mount points. Servers must explicitly allow NFS connections from unprivileged ports for this to work. In the past, bumping the value of the sunrpc.max_resvport sysctl on the client would permit the NFS client to use unprivileged ports. Bumping this setting also changes the maximum port number used by other in-kernel RPC services, some of which still required a port number less than 1023. This is exacerbated by the way source port numbers are chosen by the Linux RPC client, which starts at the top of the range and works downwards. It means that bumping the maximum means all RPC services requesting a source port will likely get an unprivileged port instead of a privileged one. Changing this setting effects all NFS mount points on a client. A sysadmin could not selectively choose which mount points would use non-privileged ports and which could not. Lastly, this mechanism of expanding the limit on the number of NFS mount points was entirely undocumented. To address the need for the NFS client to use a large range of source ports without interfering with the activity of other in-kernel RPC services, we introduce a new NFS mount option. This option explicitly tells only the NFS client to use a non-privileged source port when communicating with the NFS server for one specific mount point. This new mount option is called "resvport," like the similar NFS mount option on FreeBSD and Mac OS X. A sister patch for nfs-utils will be submitted that documents this new option in nfs(5). The default setting for this new mount option requires the NFS client to use a privileged port, as before. Explicitly specifying the "noresvport" mount option allows the NFS client to use an unprivileged source port for this mount point when connecting to the NFS server port. This mount option is supported only for text-based NFS mounts. [ Sidebar: it is widely known that security mechanisms based on the use of privileged source ports are ineffective. However, the NFS client can combine the use of unprivileged ports with the use of secure authentication mechanisms, such as Kerberos. This allows a large number of connections and mount points while ensuring a useful level of security. Eventually we may change the default setting for this option depending on the security flavor used for the mount. For example, if the mount is using only AUTH_SYS, then the default setting will be "resvport;" if the mount is using a strong security flavor such as krb5, the default setting will be "noresvport." ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond.Myklebust@netapp.com: Fixed a bug whereby nfs4_init_client() was being called with incorrect arguments.] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:37 +08:00
Opt_resvport, Opt_noresvport,
NFS: Add mount options to enable local caching on NFS Add NFS mount options to allow the local caching support to be enabled. The attached patch makes it possible for the NFS filesystem to be told to make use of the network filesystem local caching service (FS-Cache). To be able to use this, a recent nfsutils package is required. There are three variant NFS mount options that can be added to a mount command to control caching for a mount. Only the last one specified takes effect: (*) Adding "fsc" will request caching. (*) Adding "fsc=<string>" will request caching and also specify a uniquifier. (*) Adding "nofsc" will disable caching. For example: mount warthog:/ /a -o fsc The cache of a particular superblock (NFS FSID) will be shared between all mounts of that volume, provided they have the same connection parameters and are not marked 'nosharecache'. Where it is otherwise impossible to distinguish superblocks because all the parameters are identical, but the 'nosharecache' option is supplied, a uniquifying string must be supplied, else only the first mount will be permitted to use the cache. If there's a key collision, then the second mount will disable caching and give a warning into the kernel log. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Steve Dickson <steved@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Daire Byrne <Daire.Byrne@framestore.com>
2009-04-03 23:42:48 +08:00
Opt_fscache, Opt_nofscache,
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
/* Mount options that take integer arguments */
Opt_port,
Opt_rsize, Opt_wsize, Opt_bsize,
Opt_timeo, Opt_retrans,
Opt_acregmin, Opt_acregmax,
Opt_acdirmin, Opt_acdirmax,
Opt_actimeo,
Opt_namelen,
Opt_mountport,
NFS: Remove support for the 'mountprog' option Remove the mount option that allows users to specify an alternate mountd program number. The client hasn't support setting an alternate mountd program number for a very long time. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-10-27 01:32:29 +08:00
Opt_mountvers,
NFS: Remove support for the 'nfsprog' option Remove the mount option that allows users to specify an alternate NFS program number. The client hasn't support setting an alternate NFS program number for a very long time. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-10-27 01:32:24 +08:00
Opt_nfsvers,
nfs41: add mount command option minorversion mount -t nfs4 -o minorversion=[0|1] specifies whether to use 4.0 or 4.1. By default, the minorversion is set to 0. Signed-off-by: Mike Sager <sager@netapp.com> [set default minorversion to 0 as per Trond and SteveD's request] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-04-01 21:21:48 +08:00
Opt_minorversion,
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
/* Mount options that take string arguments */
NFS: Remove the NIPQUAD from nfs_try_mount In the name of address family compatibility, we can't have the NIP_FMT and NIPQUAD macros in nfs_try_mount(). Instead, we can make use of an unused mount option to display the mount server's hostname. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:13 +08:00
Opt_sec, Opt_proto, Opt_mountproto, Opt_mounthost,
NFS: Add new 'mountaddr=' mount option I got the 'mounthost=' option wrong - it shouldn't look for an address value, but rather a hostname value. However, the in-kernel mount client and NFS client cannot resolve a hostname by themselves; they rely on user-land to pass in the resolved address. Create a new mount option that does take an address so that the mount program's address can be passed in. The mount hostname is now ignored by the kernel. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-12 06:01:04 +08:00
Opt_addr, Opt_mountaddr, Opt_clientaddr,
NFS: Add mount options for controlling the lookup cache Add the following NFS-specific mount options to the parser. -o lookupcache=all /* Default: cache positive & negative dentries */ -o lookupcache=pos[itive] /* Don't cache negative dentries */ -o lookupcache=none /* Strict revalidation of all dentries */ Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-16 05:58:14 +08:00
Opt_lookupcache,
NFS: Add mount options to enable local caching on NFS Add NFS mount options to allow the local caching support to be enabled. The attached patch makes it possible for the NFS filesystem to be told to make use of the network filesystem local caching service (FS-Cache). To be able to use this, a recent nfsutils package is required. There are three variant NFS mount options that can be added to a mount command to control caching for a mount. Only the last one specified takes effect: (*) Adding "fsc" will request caching. (*) Adding "fsc=<string>" will request caching and also specify a uniquifier. (*) Adding "nofsc" will disable caching. For example: mount warthog:/ /a -o fsc The cache of a particular superblock (NFS FSID) will be shared between all mounts of that volume, provided they have the same connection parameters and are not marked 'nosharecache'. Where it is otherwise impossible to distinguish superblocks because all the parameters are identical, but the 'nosharecache' option is supplied, a uniquifying string must be supplied, else only the first mount will be permitted to use the cache. If there's a key collision, then the second mount will disable caching and give a warning into the kernel log. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Steve Dickson <steved@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Daire Byrne <Daire.Byrne@framestore.com>
2009-04-03 23:42:48 +08:00
Opt_fscache_uniq,
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
/* Special mount options */
Opt_userspace, Opt_deprecated, Opt_sloppy,
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
Opt_err
};
vfs: Use const for kernel parser table This is a much better version of a previous patch to make the parser tables constant. Rather than changing the typedef, we put the "const" in all the various places where its required, allowing the __initconst exception for nfsroot which was the cause of the previous trouble. This was posted for review some time ago and I believe its been in -mm since then. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com> Cc: Alexander Viro <aviro@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-10-13 17:46:57 +08:00
static const match_table_t nfs_mount_option_tokens = {
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
{ Opt_userspace, "bg" },
{ Opt_userspace, "fg" },
NFS: Allow any value for the "retry" option The kernel NFS mount option parser should ignore the retry= mount option since it is meaningful only in user space. Today it expects a number rather than arbitrary text, so it ignores the option if the value is numeric, but chokes if there are other characters in the value. Change it to allow any text (except ",") as its value. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-13 00:37:33 +08:00
{ Opt_userspace, "retry=%s" },
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
{ Opt_sloppy, "sloppy" },
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
{ Opt_soft, "soft" },
{ Opt_hard, "hard" },
NFS: Treat "intr" and "nointr" options as deprecated Clean up: the "intr" and "nointr" mount options were recently retired. Document this in the NFS mount option parser. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-13 00:37:41 +08:00
{ Opt_deprecated, "intr" },
{ Opt_deprecated, "nointr" },
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
{ Opt_posix, "posix" },
{ Opt_noposix, "noposix" },
{ Opt_cto, "cto" },
{ Opt_nocto, "nocto" },
{ Opt_ac, "ac" },
{ Opt_noac, "noac" },
{ Opt_lock, "lock" },
{ Opt_nolock, "nolock" },
{ Opt_v2, "v2" },
{ Opt_v3, "v3" },
{ Opt_udp, "udp" },
{ Opt_tcp, "tcp" },
NFS: support RDMA mounts Adds hooks to the string-based NFS mount to support an "rdma" protocol option. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:49:41 +08:00
{ Opt_rdma, "rdma" },
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
{ Opt_acl, "acl" },
{ Opt_noacl, "noacl" },
{ Opt_rdirplus, "rdirplus" },
{ Opt_nordirplus, "nordirplus" },
NFS: Add the mount option "nosharecache" Prior to David Howell's mount changes in 2.6.18, users who mounted different directories which happened to be from the same filesystem on the server would get different super blocks, and hence could choose different mount options. As long as there were no hard linked files that crossed from one subtree to another, this was quite safe. Post the changes, if the two directories are on the same filesystem (have the same 'fsid'), they will share the same super block, and hence the same mount options. Add a flag to allow users to elect not to share the NFS super block with another mount point, even if the fsids are the same. This will allow users to set different mount options for the two different super blocks, as was previously possible. It is still up to the user to ensure that there are no cache coherency issues when doing this, however the default behaviour will be to share super blocks whenever two paths result in the same fsid. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
{ Opt_sharecache, "sharecache" },
{ Opt_nosharecache, "nosharecache" },
NFS: add "[no]resvport" mount option The standard default security setting for NFS is AUTH_SYS. An NFS client connects to NFS servers via a privileged source port and a fixed standard destination port (2049). The client sends raw uid and gid numbers to identify users making NFS requests, and the server assumes an appropriate authority on the client has vetted these values because the source port is privileged. On Linux, by default in-kernel RPC services use a privileged port in the range between 650 and 1023 to avoid using source ports of well- known IP services. Using such a small range limits the number of NFS mount points and the number of unique NFS servers to which a client can connect concurrently. An NFS client can use unprivileged source ports to expand the range of source port numbers, allowing more concurrent server connections and more NFS mount points. Servers must explicitly allow NFS connections from unprivileged ports for this to work. In the past, bumping the value of the sunrpc.max_resvport sysctl on the client would permit the NFS client to use unprivileged ports. Bumping this setting also changes the maximum port number used by other in-kernel RPC services, some of which still required a port number less than 1023. This is exacerbated by the way source port numbers are chosen by the Linux RPC client, which starts at the top of the range and works downwards. It means that bumping the maximum means all RPC services requesting a source port will likely get an unprivileged port instead of a privileged one. Changing this setting effects all NFS mount points on a client. A sysadmin could not selectively choose which mount points would use non-privileged ports and which could not. Lastly, this mechanism of expanding the limit on the number of NFS mount points was entirely undocumented. To address the need for the NFS client to use a large range of source ports without interfering with the activity of other in-kernel RPC services, we introduce a new NFS mount option. This option explicitly tells only the NFS client to use a non-privileged source port when communicating with the NFS server for one specific mount point. This new mount option is called "resvport," like the similar NFS mount option on FreeBSD and Mac OS X. A sister patch for nfs-utils will be submitted that documents this new option in nfs(5). The default setting for this new mount option requires the NFS client to use a privileged port, as before. Explicitly specifying the "noresvport" mount option allows the NFS client to use an unprivileged source port for this mount point when connecting to the NFS server port. This mount option is supported only for text-based NFS mounts. [ Sidebar: it is widely known that security mechanisms based on the use of privileged source ports are ineffective. However, the NFS client can combine the use of unprivileged ports with the use of secure authentication mechanisms, such as Kerberos. This allows a large number of connections and mount points while ensuring a useful level of security. Eventually we may change the default setting for this option depending on the security flavor used for the mount. For example, if the mount is using only AUTH_SYS, then the default setting will be "resvport;" if the mount is using a strong security flavor such as krb5, the default setting will be "noresvport." ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond.Myklebust@netapp.com: Fixed a bug whereby nfs4_init_client() was being called with incorrect arguments.] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:37 +08:00
{ Opt_resvport, "resvport" },
{ Opt_noresvport, "noresvport" },
NFS: Add mount options to enable local caching on NFS Add NFS mount options to allow the local caching support to be enabled. The attached patch makes it possible for the NFS filesystem to be told to make use of the network filesystem local caching service (FS-Cache). To be able to use this, a recent nfsutils package is required. There are three variant NFS mount options that can be added to a mount command to control caching for a mount. Only the last one specified takes effect: (*) Adding "fsc" will request caching. (*) Adding "fsc=<string>" will request caching and also specify a uniquifier. (*) Adding "nofsc" will disable caching. For example: mount warthog:/ /a -o fsc The cache of a particular superblock (NFS FSID) will be shared between all mounts of that volume, provided they have the same connection parameters and are not marked 'nosharecache'. Where it is otherwise impossible to distinguish superblocks because all the parameters are identical, but the 'nosharecache' option is supplied, a uniquifying string must be supplied, else only the first mount will be permitted to use the cache. If there's a key collision, then the second mount will disable caching and give a warning into the kernel log. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Steve Dickson <steved@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Daire Byrne <Daire.Byrne@framestore.com>
2009-04-03 23:42:48 +08:00
{ Opt_fscache, "fsc" },
{ Opt_fscache_uniq, "fsc=%s" },
{ Opt_nofscache, "nofsc" },
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
{ Opt_port, "port=%s" },
{ Opt_rsize, "rsize=%s" },
{ Opt_wsize, "wsize=%s" },
{ Opt_bsize, "bsize=%s" },
{ Opt_timeo, "timeo=%s" },
{ Opt_retrans, "retrans=%s" },
{ Opt_acregmin, "acregmin=%s" },
{ Opt_acregmax, "acregmax=%s" },
{ Opt_acdirmin, "acdirmin=%s" },
{ Opt_acdirmax, "acdirmax=%s" },
{ Opt_actimeo, "actimeo=%s" },
{ Opt_namelen, "namlen=%s" },
{ Opt_mountport, "mountport=%s" },
{ Opt_mountvers, "mountvers=%s" },
{ Opt_nfsvers, "nfsvers=%s" },
{ Opt_nfsvers, "vers=%s" },
nfs41: add mount command option minorversion mount -t nfs4 -o minorversion=[0|1] specifies whether to use 4.0 or 4.1. By default, the minorversion is set to 0. Signed-off-by: Mike Sager <sager@netapp.com> [set default minorversion to 0 as per Trond and SteveD's request] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-04-01 21:21:48 +08:00
{ Opt_minorversion, "minorversion=%u" },
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
{ Opt_sec, "sec=%s" },
{ Opt_proto, "proto=%s" },
{ Opt_mountproto, "mountproto=%s" },
{ Opt_addr, "addr=%s" },
{ Opt_clientaddr, "clientaddr=%s" },
NFS: Remove the NIPQUAD from nfs_try_mount In the name of address family compatibility, we can't have the NIP_FMT and NIPQUAD macros in nfs_try_mount(). Instead, we can make use of an unused mount option to display the mount server's hostname. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:13 +08:00
{ Opt_mounthost, "mounthost=%s" },
NFS: Add new 'mountaddr=' mount option I got the 'mounthost=' option wrong - it shouldn't look for an address value, but rather a hostname value. However, the in-kernel mount client and NFS client cannot resolve a hostname by themselves; they rely on user-land to pass in the resolved address. Create a new mount option that does take an address so that the mount program's address can be passed in. The mount hostname is now ignored by the kernel. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-12 06:01:04 +08:00
{ Opt_mountaddr, "mountaddr=%s" },
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
NFS: Add mount options for controlling the lookup cache Add the following NFS-specific mount options to the parser. -o lookupcache=all /* Default: cache positive & negative dentries */ -o lookupcache=pos[itive] /* Don't cache negative dentries */ -o lookupcache=none /* Strict revalidation of all dentries */ Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-16 05:58:14 +08:00
{ Opt_lookupcache, "lookupcache=%s" },
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
{ Opt_err, NULL }
};
enum {
NFS: support RDMA mounts Adds hooks to the string-based NFS mount to support an "rdma" protocol option. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:49:41 +08:00
Opt_xprt_udp, Opt_xprt_tcp, Opt_xprt_rdma,
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
Opt_xprt_err
};
vfs: Use const for kernel parser table This is a much better version of a previous patch to make the parser tables constant. Rather than changing the typedef, we put the "const" in all the various places where its required, allowing the __initconst exception for nfsroot which was the cause of the previous trouble. This was posted for review some time ago and I believe its been in -mm since then. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com> Cc: Alexander Viro <aviro@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-10-13 17:46:57 +08:00
static const match_table_t nfs_xprt_protocol_tokens = {
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
{ Opt_xprt_udp, "udp" },
{ Opt_xprt_tcp, "tcp" },
NFS: support RDMA mounts Adds hooks to the string-based NFS mount to support an "rdma" protocol option. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:49:41 +08:00
{ Opt_xprt_rdma, "rdma" },
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
{ Opt_xprt_err, NULL }
};
enum {
Opt_sec_none, Opt_sec_sys,
Opt_sec_krb5, Opt_sec_krb5i, Opt_sec_krb5p,
Opt_sec_lkey, Opt_sec_lkeyi, Opt_sec_lkeyp,
Opt_sec_spkm, Opt_sec_spkmi, Opt_sec_spkmp,
Opt_sec_err
};
vfs: Use const for kernel parser table This is a much better version of a previous patch to make the parser tables constant. Rather than changing the typedef, we put the "const" in all the various places where its required, allowing the __initconst exception for nfsroot which was the cause of the previous trouble. This was posted for review some time ago and I believe its been in -mm since then. Signed-off-by: Steven Whitehouse <swhiteho@redhat.com> Cc: Alexander Viro <aviro@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-10-13 17:46:57 +08:00
static const match_table_t nfs_secflavor_tokens = {
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
{ Opt_sec_none, "none" },
{ Opt_sec_none, "null" },
{ Opt_sec_sys, "sys" },
{ Opt_sec_krb5, "krb5" },
{ Opt_sec_krb5i, "krb5i" },
{ Opt_sec_krb5p, "krb5p" },
{ Opt_sec_lkey, "lkey" },
{ Opt_sec_lkeyi, "lkeyi" },
{ Opt_sec_lkeyp, "lkeyp" },
NFS: add missing spkm3 strings to mount option parser This patch adds previous missing spkm3 string values that are needed to parse mount options in the kernel.
2008-02-14 05:47:06 +08:00
{ Opt_sec_spkm, "spkm3" },
{ Opt_sec_spkmi, "spkm3i" },
{ Opt_sec_spkmp, "spkm3p" },
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
{ Opt_sec_err, NULL }
};
NFS: Add mount options for controlling the lookup cache Add the following NFS-specific mount options to the parser. -o lookupcache=all /* Default: cache positive & negative dentries */ -o lookupcache=pos[itive] /* Don't cache negative dentries */ -o lookupcache=none /* Strict revalidation of all dentries */ Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-16 05:58:14 +08:00
enum {
Opt_lookupcache_all, Opt_lookupcache_positive,
Opt_lookupcache_none,
Opt_lookupcache_err
};
static match_table_t nfs_lookupcache_tokens = {
{ Opt_lookupcache_all, "all" },
{ Opt_lookupcache_positive, "pos" },
{ Opt_lookupcache_positive, "positive" },
{ Opt_lookupcache_none, "none" },
{ Opt_lookupcache_err, NULL }
};
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
[PATCH] restore sane ->umount_begin() API Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2008-04-24 19:21:56 +08:00
static void nfs_umount_begin(struct super_block *);
Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ Conflicts: fs/nfs/inode.c fs/super.c Fix conflicts between patch 'NFS: Split fs/nfs/inode.c' and patch 'VFS: Permit filesystem to override root dentry on mount'
2006-06-24 20:41:41 +08:00
static int nfs_statfs(struct dentry *, struct kstatfs *);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
static int nfs_show_options(struct seq_file *, struct vfsmount *);
static int nfs_show_stats(struct seq_file *, struct vfsmount *);
Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ Conflicts: fs/nfs/inode.c fs/super.c Fix conflicts between patch 'NFS: Split fs/nfs/inode.c' and patch 'VFS: Permit filesystem to override root dentry on mount'
2006-06-24 20:41:41 +08:00
static int nfs_get_sb(struct file_system_type *, int, const char *, void *, struct vfsmount *);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
static int nfs_xdev_get_sb(struct file_system_type *fs_type,
Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ Conflicts: fs/nfs/inode.c fs/super.c Fix conflicts between patch 'NFS: Split fs/nfs/inode.c' and patch 'VFS: Permit filesystem to override root dentry on mount'
2006-06-24 20:41:41 +08:00
int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
static void nfs_kill_super(struct super_block *);
NFS: implement option checking when remounting NFS filesystems (resend) When remounting an NFS or NFS4 filesystem, the new NFS options are not respected, yet the remount will still return success. This patch adds a remount_fs sb op for NFS that checks any new nfs mount options against the existing ones and fails the mount if any have changed. This is only implemented for string-based mount options since doing this with binary options isn't really feasible. This is essentially the same as the original patch I sent out, but adds a check to see if the addr= option has changed. Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-11 03:38:39 +08:00
static int nfs_remount(struct super_block *sb, int *flags, char *raw_data);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
static struct file_system_type nfs_fs_type = {
.owner = THIS_MODULE,
.name = "nfs",
.get_sb = nfs_get_sb,
.kill_sb = nfs_kill_super,
[PATCH] Allow file systems to manually d_move() inside of ->rename() Some file systems want to manually d_move() the dentries involved in a rename. We can do this by making use of the FS_ODD_RENAME flag if we just have nfs_rename() unconditionally do the d_move(). While there, we rename the flag to be more descriptive. OCFS2 uses this to protect that part of the rename operation with a cluster lock. Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Andrew Morton <akpm@osdl.org>
2006-09-09 05:22:21 +08:00
.fs_flags = FS_RENAME_DOES_D_MOVE|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
};
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
struct file_system_type nfs_xdev_fs_type = {
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
.owner = THIS_MODULE,
.name = "nfs",
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
.get_sb = nfs_xdev_get_sb,
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
.kill_sb = nfs_kill_super,
[PATCH] Allow file systems to manually d_move() inside of ->rename() Some file systems want to manually d_move() the dentries involved in a rename. We can do this by making use of the FS_ODD_RENAME flag if we just have nfs_rename() unconditionally do the d_move(). While there, we rename the flag to be more descriptive. OCFS2 uses this to protect that part of the rename operation with a cluster lock. Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Andrew Morton <akpm@osdl.org>
2006-09-09 05:22:21 +08:00
.fs_flags = FS_RENAME_DOES_D_MOVE|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
};
[PATCH] Mark struct super_operations const This patch is inspired by Arjan's "Patch series to mark struct file_operations and struct inode_operations const". Compile tested with gcc & sparse. Signed-off-by: Josef 'Jeff' Sipek <jsipek@cs.sunysb.edu> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-02-12 16:55:41 +08:00
static const struct super_operations nfs_sops = {
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
.alloc_inode = nfs_alloc_inode,
.destroy_inode = nfs_destroy_inode,
.write_inode = nfs_write_inode,
.statfs = nfs_statfs,
.clear_inode = nfs_clear_inode,
.umount_begin = nfs_umount_begin,
.show_options = nfs_show_options,
.show_stats = nfs_show_stats,
NFS: implement option checking when remounting NFS filesystems (resend) When remounting an NFS or NFS4 filesystem, the new NFS options are not respected, yet the remount will still return success. This patch adds a remount_fs sb op for NFS that checks any new nfs mount options against the existing ones and fails the mount if any have changed. This is only implemented for string-based mount options since doing this with binary options isn't really feasible. This is essentially the same as the original patch I sent out, but adds a check to see if the addr= option has changed. Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-11 03:38:39 +08:00
.remount_fs = nfs_remount,
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
};
#ifdef CONFIG_NFS_V4
Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ Conflicts: fs/nfs/inode.c fs/super.c Fix conflicts between patch 'NFS: Split fs/nfs/inode.c' and patch 'VFS: Permit filesystem to override root dentry on mount'
2006-06-24 20:41:41 +08:00
static int nfs4_get_sb(struct file_system_type *fs_type,
int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt);
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace As noted in the previous patch, the NFSv4 client mount code currently has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in nfs4_path_walk() will fail with an error. This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the namespace on the server, then uses the ordinary VFS and NFS path lookup code to walk down the mount path in that namespace. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-23 03:09:14 +08:00
static int nfs4_remote_get_sb(struct file_system_type *fs_type,
int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
static int nfs4_xdev_get_sb(struct file_system_type *fs_type,
int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt);
static int nfs4_referral_get_sb(struct file_system_type *fs_type,
int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt);
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace As noted in the previous patch, the NFSv4 client mount code currently has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in nfs4_path_walk() will fail with an error. This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the namespace on the server, then uses the ordinary VFS and NFS path lookup code to walk down the mount path in that namespace. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-23 03:09:14 +08:00
static int nfs4_remote_referral_get_sb(struct file_system_type *fs_type,
int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
static void nfs4_kill_super(struct super_block *sb);
static struct file_system_type nfs4_fs_type = {
.owner = THIS_MODULE,
.name = "nfs4",
.get_sb = nfs4_get_sb,
.kill_sb = nfs4_kill_super,
[PATCH] Allow file systems to manually d_move() inside of ->rename() Some file systems want to manually d_move() the dentries involved in a rename. We can do this by making use of the FS_ODD_RENAME flag if we just have nfs_rename() unconditionally do the d_move(). While there, we rename the flag to be more descriptive. OCFS2 uses this to protect that part of the rename operation with a cluster lock. Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Andrew Morton <akpm@osdl.org>
2006-09-09 05:22:21 +08:00
.fs_flags = FS_RENAME_DOES_D_MOVE|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
};
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace As noted in the previous patch, the NFSv4 client mount code currently has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in nfs4_path_walk() will fail with an error. This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the namespace on the server, then uses the ordinary VFS and NFS path lookup code to walk down the mount path in that namespace. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-23 03:09:14 +08:00
static struct file_system_type nfs4_remote_fs_type = {
.owner = THIS_MODULE,
.name = "nfs4",
.get_sb = nfs4_remote_get_sb,
.kill_sb = nfs4_kill_super,
.fs_flags = FS_RENAME_DOES_D_MOVE|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
};
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
struct file_system_type nfs4_xdev_fs_type = {
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
.owner = THIS_MODULE,
.name = "nfs4",
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
.get_sb = nfs4_xdev_get_sb,
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
.kill_sb = nfs4_kill_super,
[PATCH] Allow file systems to manually d_move() inside of ->rename() Some file systems want to manually d_move() the dentries involved in a rename. We can do this by making use of the FS_ODD_RENAME flag if we just have nfs_rename() unconditionally do the d_move(). While there, we rename the flag to be more descriptive. OCFS2 uses this to protect that part of the rename operation with a cluster lock. Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Andrew Morton <akpm@osdl.org>
2006-09-09 05:22:21 +08:00
.fs_flags = FS_RENAME_DOES_D_MOVE|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
};
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace As noted in the previous patch, the NFSv4 client mount code currently has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in nfs4_path_walk() will fail with an error. This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the namespace on the server, then uses the ordinary VFS and NFS path lookup code to walk down the mount path in that namespace. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-23 03:09:14 +08:00
static struct file_system_type nfs4_remote_referral_fs_type = {
.owner = THIS_MODULE,
.name = "nfs4",
.get_sb = nfs4_remote_referral_get_sb,
.kill_sb = nfs4_kill_super,
.fs_flags = FS_RENAME_DOES_D_MOVE|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
};
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
struct file_system_type nfs4_referral_fs_type = {
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
.owner = THIS_MODULE,
.name = "nfs4",
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
.get_sb = nfs4_referral_get_sb,
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
.kill_sb = nfs4_kill_super,
[PATCH] Allow file systems to manually d_move() inside of ->rename() Some file systems want to manually d_move() the dentries involved in a rename. We can do this by making use of the FS_ODD_RENAME flag if we just have nfs_rename() unconditionally do the d_move(). While there, we rename the flag to be more descriptive. OCFS2 uses this to protect that part of the rename operation with a cluster lock. Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Andrew Morton <akpm@osdl.org>
2006-09-09 05:22:21 +08:00
.fs_flags = FS_RENAME_DOES_D_MOVE|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
};
[PATCH] Mark struct super_operations const This patch is inspired by Arjan's "Patch series to mark struct file_operations and struct inode_operations const". Compile tested with gcc & sparse. Signed-off-by: Josef 'Jeff' Sipek <jsipek@cs.sunysb.edu> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-02-12 16:55:41 +08:00
static const struct super_operations nfs4_sops = {
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
.alloc_inode = nfs_alloc_inode,
.destroy_inode = nfs_destroy_inode,
.write_inode = nfs_write_inode,
.statfs = nfs_statfs,
.clear_inode = nfs4_clear_inode,
.umount_begin = nfs_umount_begin,
.show_options = nfs_show_options,
.show_stats = nfs_show_stats,
NFS: implement option checking when remounting NFS filesystems (resend) When remounting an NFS or NFS4 filesystem, the new NFS options are not respected, yet the remount will still return success. This patch adds a remount_fs sb op for NFS that checks any new nfs mount options against the existing ones and fails the mount if any have changed. This is only implemented for string-based mount options since doing this with binary options isn't really feasible. This is essentially the same as the original patch I sent out, but adds a check to see if the addr= option has changed. Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-11 03:38:39 +08:00
.remount_fs = nfs_remount,
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
};
#endif
mm: clean up and kernelify shrinker registration I can never remember what the function to register to receive VM pressure is called. I have to trace down from __alloc_pages() to find it. It's called "set_shrinker()", and it needs Your Help. 1) Don't hide struct shrinker. It contains no magic. 2) Don't allocate "struct shrinker". It's not helpful. 3) Call them "register_shrinker" and "unregister_shrinker". 4) Call the function "shrink" not "shrinker". 5) Reduce the 17 lines of waffly comments to 13, but document it properly. Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Cc: David Chinner <dgc@sgi.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-17 19:03:17 +08:00
static struct shrinker acl_shrinker = {
.shrink = nfs_access_cache_shrinker,
.seeks = DEFAULT_SEEKS,
};
NFS: Add an ACCESS cache memory shrinker A pinned inode may in theory end up filling memory with cached ACCESS calls. This patch ensures that the VM may shrink away the cache in these particular cases. The shrinker works by iterating through the list of inodes on the global nfs_access_lru_list, and removing the least recently used access cache entry until it is done (or until the entire cache is empty). Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-07-25 23:28:19 +08:00
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
/*
* Register the NFS filesystems
*/
int __init register_nfs_fs(void)
{
int ret;
ret = register_filesystem(&nfs_fs_type);
if (ret < 0)
goto error_0;
ret = nfs_register_sysctl();
if (ret < 0)
goto error_1;
[PATCH] nfs: fix congestion control The current NFS client congestion logic is severly broken, it marks the backing device congested during each nfs_writepages() call but doesn't mirror this in nfs_writepage() which makes for deadlocks. Also it implements its own waitqueue. Replace this by a more regular congestion implementation that puts a cap on the number of active writeback pages and uses the bdi congestion waitqueue. Also always use an interruptible wait since it makes sense to be able to SIGKILL the process even for mounts without 'intr'. Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Acked-by: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: Christoph Lameter <clameter@engr.sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-03-17 05:38:26 +08:00
#ifdef CONFIG_NFS_V4
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
ret = register_filesystem(&nfs4_fs_type);
if (ret < 0)
goto error_2;
#endif
mm: clean up and kernelify shrinker registration I can never remember what the function to register to receive VM pressure is called. I have to trace down from __alloc_pages() to find it. It's called "set_shrinker()", and it needs Your Help. 1) Don't hide struct shrinker. It contains no magic. 2) Don't allocate "struct shrinker". It's not helpful. 3) Call them "register_shrinker" and "unregister_shrinker". 4) Call the function "shrink" not "shrinker". 5) Reduce the 17 lines of waffly comments to 13, but document it properly. Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Cc: David Chinner <dgc@sgi.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-17 19:03:17 +08:00
register_shrinker(&acl_shrinker);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
return 0;
#ifdef CONFIG_NFS_V4
error_2:
nfs_unregister_sysctl();
[PATCH] nfs: fix congestion control The current NFS client congestion logic is severly broken, it marks the backing device congested during each nfs_writepages() call but doesn't mirror this in nfs_writepage() which makes for deadlocks. Also it implements its own waitqueue. Replace this by a more regular congestion implementation that puts a cap on the number of active writeback pages and uses the bdi congestion waitqueue. Also always use an interruptible wait since it makes sense to be able to SIGKILL the process even for mounts without 'intr'. Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Acked-by: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: Christoph Lameter <clameter@engr.sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-03-17 05:38:26 +08:00
#endif
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
error_1:
unregister_filesystem(&nfs_fs_type);
error_0:
return ret;
}
/*
* Unregister the NFS filesystems
*/
void __exit unregister_nfs_fs(void)
{
mm: clean up and kernelify shrinker registration I can never remember what the function to register to receive VM pressure is called. I have to trace down from __alloc_pages() to find it. It's called "set_shrinker()", and it needs Your Help. 1) Don't hide struct shrinker. It contains no magic. 2) Don't allocate "struct shrinker". It's not helpful. 3) Call them "register_shrinker" and "unregister_shrinker". 4) Call the function "shrink" not "shrinker". 5) Reduce the 17 lines of waffly comments to 13, but document it properly. Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Cc: David Chinner <dgc@sgi.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-17 19:03:17 +08:00
unregister_shrinker(&acl_shrinker);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
#ifdef CONFIG_NFS_V4
unregister_filesystem(&nfs4_fs_type);
#endif
nfs: fix oops re sysctls and V4 support NFS unregisters sysctls only if V4 support is compiled in. However, sysctl table is not V4 specific, so unregister it always. Steps to reproduce: [build nfs.ko with CONFIG_NFS_V4=n] modrobe nfs rmmod nfs ls /proc/sys Unable to handle kernel paging request at ffffffff880661c0 RIP: [<ffffffff802af8e3>] proc_sys_readdir+0xd3/0x350 PGD 203067 PUD 207063 PMD 7e216067 PTE 0 Oops: 0000 [1] SMP CPU 1 Modules linked in: lockd nfs_acl sunrpc Pid: 3335, comm: ls Not tainted 2.6.23-rc3-bloat #2 RIP: 0010:[<ffffffff802af8e3>] [<ffffffff802af8e3>] proc_sys_readdir+0xd3/0x350 RSP: 0018:ffff81007fd93e78 EFLAGS: 00010286 RAX: ffffffff880661c0 RBX: ffffffff80466370 RCX: ffffffff880661c0 RDX: 00000000000014c0 RSI: ffff81007f3ad020 RDI: ffff81007efd8b40 RBP: 0000000000000018 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000001 R11: ffffffff802a8570 R12: ffffffff880661c0 R13: ffff81007e219640 R14: ffff81007efd8b40 R15: ffff81007ded7280 FS: 00002ba25ef03060(0000) GS:ffff81007ff81258(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: ffffffff880661c0 CR3: 000000007dfaf000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process ls (pid: 3335, threadinfo ffff81007fd92000, task ffff81007d8a0000) Stack: ffff81007f3ad150 ffffffff80283f30 ffff81007fd93f48 ffff81007efd8b40 ffff81007ee00440 0000000422222222 0000000200035593 ffffffff88037e9a 2222222222222222 ffffffff80466500 ffff81007e416400 ffff81007e219640 Call Trace: [<ffffffff80283f30>] filldir+0x0/0xf0 [<ffffffff80283f30>] filldir+0x0/0xf0 [<ffffffff802840c7>] vfs_readdir+0xa7/0xc0 [<ffffffff80284376>] sys_getdents+0x96/0xe0 [<ffffffff8020bb3e>] system_call+0x7e/0x83 Code: 41 8b 14 24 85 d2 74 dc 49 8b 44 24 08 48 85 c0 74 e7 49 3b RIP [<ffffffff802af8e3>] proc_sys_readdir+0xd3/0x350 RSP <ffff81007fd93e78> CR2: ffffffff880661c0 Kernel panic - not syncing: Fatal exception Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com> Acked-by: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: <stable@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-09-19 13:46:40 +08:00
nfs_unregister_sysctl();
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
unregister_filesystem(&nfs_fs_type);
}
NFS: Clean up nfs_sb_active/nfs_sb_deactive Instead of causing umount requests to block on server->active_wq while the asynchronous sillyrename deletes are executing, we can use the sb->s_active counter to obtain a reference to the super_block, and then release that reference in nfs_async_unlink_release(). Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-28 06:19:01 +08:00
void nfs_sb_active(struct super_block *sb)
NFS: Stop sillyname renames and unmounts from racing Added an active/deactive mechanism to the nfs_server structure allowing async operations to hold off umount until the operations are done. Signed-off-by: Steve Dickson <steved@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-11-08 17:05:04 +08:00
{
NFS: Clean up nfs_sb_active/nfs_sb_deactive Instead of causing umount requests to block on server->active_wq while the asynchronous sillyrename deletes are executing, we can use the sb->s_active counter to obtain a reference to the super_block, and then release that reference in nfs_async_unlink_release(). Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-28 06:19:01 +08:00
struct nfs_server *server = NFS_SB(sb);
NFS: Stop sillyname renames and unmounts from racing Added an active/deactive mechanism to the nfs_server structure allowing async operations to hold off umount until the operations are done. Signed-off-by: Steve Dickson <steved@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-11-08 17:05:04 +08:00
NFS: Clean up nfs_sb_active/nfs_sb_deactive Instead of causing umount requests to block on server->active_wq while the asynchronous sillyrename deletes are executing, we can use the sb->s_active counter to obtain a reference to the super_block, and then release that reference in nfs_async_unlink_release(). Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-28 06:19:01 +08:00
if (atomic_inc_return(&server->active) == 1)
atomic_inc(&sb->s_active);
NFS: Stop sillyname renames and unmounts from racing Added an active/deactive mechanism to the nfs_server structure allowing async operations to hold off umount until the operations are done. Signed-off-by: Steve Dickson <steved@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-11-08 17:05:04 +08:00
}
NFS: Clean up nfs_sb_active/nfs_sb_deactive Instead of causing umount requests to block on server->active_wq while the asynchronous sillyrename deletes are executing, we can use the sb->s_active counter to obtain a reference to the super_block, and then release that reference in nfs_async_unlink_release(). Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-28 06:19:01 +08:00
void nfs_sb_deactive(struct super_block *sb)
NFS: Stop sillyname renames and unmounts from racing Added an active/deactive mechanism to the nfs_server structure allowing async operations to hold off umount until the operations are done. Signed-off-by: Steve Dickson <steved@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-11-08 17:05:04 +08:00
{
struct nfs_server *server = NFS_SB(sb);
NFS: Clean up nfs_sb_active/nfs_sb_deactive Instead of causing umount requests to block on server->active_wq while the asynchronous sillyrename deletes are executing, we can use the sb->s_active counter to obtain a reference to the super_block, and then release that reference in nfs_async_unlink_release(). Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-28 06:19:01 +08:00
if (atomic_dec_and_test(&server->active))
deactivate_super(sb);
NFS: Stop sillyname renames and unmounts from racing Added an active/deactive mechanism to the nfs_server structure allowing async operations to hold off umount until the operations are done. Signed-off-by: Steve Dickson <steved@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-11-08 17:05:04 +08:00
}
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
/*
* Deliver file system statistics to userspace
*/
Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ Conflicts: fs/nfs/inode.c fs/super.c Fix conflicts between patch 'NFS: Split fs/nfs/inode.c' and patch 'VFS: Permit filesystem to override root dentry on mount'
2006-06-24 20:41:41 +08:00
static int nfs_statfs(struct dentry *dentry, struct kstatfs *buf)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{
NFS: Use the dentry superblock directly in nfs_statfs() Use the nominated dentry's superblock directly in the NFS statfs() op to get a file handle, rather than using s_root (which will become a dummy dentry in a future patch). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:10 +08:00
struct nfs_server *server = NFS_SB(dentry->d_sb);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
unsigned char blockbits;
unsigned long blockres;
NFS: Use the dentry superblock directly in nfs_statfs() Use the nominated dentry's superblock directly in the NFS statfs() op to get a file handle, rather than using s_root (which will become a dummy dentry in a future patch). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:10 +08:00
struct nfs_fh *fh = NFS_FH(dentry->d_inode);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
struct nfs_fattr fattr;
struct nfs_fsstat res = {
.fattr = &fattr,
};
int error;
NFS: Move rpc_ops from nfs_server to nfs_client Move the rpc_ops from the nfs_server struct to the nfs_client struct as they're common to all server records of a particular NFS protocol version. Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:12 +08:00
error = server->nfs_client->rpc_ops->statfs(server, fh, &res);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
if (error < 0)
goto out_err;
NFS: statfs error-handling fix The nfs statfs function returns a success code on error, and fills the output buffer with invalid values. The attached patch makes it return a correct error code instead. Signed-off-by: Amnon Aaronsohn <amnonaar@gmail.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> (Modified patch to reinstate the dprintk())
2007-04-10 13:05:26 +08:00
buf->f_type = NFS_SUPER_MAGIC;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
/*
* Current versions of glibc do not correctly handle the
* case where f_frsize != f_bsize. Eventually we want to
* report the value of wtmult in this field.
*/
NFS: Use the dentry superblock directly in nfs_statfs() Use the nominated dentry's superblock directly in the NFS statfs() op to get a file handle, rather than using s_root (which will become a dummy dentry in a future patch). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:10 +08:00
buf->f_frsize = dentry->d_sb->s_blocksize;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
/*
* On most *nix systems, f_blocks, f_bfree, and f_bavail
* are reported in units of f_frsize. Linux hasn't had
* an f_frsize field in its statfs struct until recently,
* thus historically Linux's sys_statfs reports these
* fields in units of f_bsize.
*/
NFS: Use the dentry superblock directly in nfs_statfs() Use the nominated dentry's superblock directly in the NFS statfs() op to get a file handle, rather than using s_root (which will become a dummy dentry in a future patch). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:10 +08:00
buf->f_bsize = dentry->d_sb->s_blocksize;
blockbits = dentry->d_sb->s_blocksize_bits;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
blockres = (1 << blockbits) - 1;
buf->f_blocks = (res.tbytes + blockres) >> blockbits;
buf->f_bfree = (res.fbytes + blockres) >> blockbits;
buf->f_bavail = (res.abytes + blockres) >> blockbits;
buf->f_files = res.tfiles;
buf->f_ffree = res.afiles;
buf->f_namelen = server->namelen;
NFS: statfs error-handling fix The nfs statfs function returns a success code on error, and fills the output buffer with invalid values. The attached patch makes it return a correct error code instead. Signed-off-by: Amnon Aaronsohn <amnonaar@gmail.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> (Modified patch to reinstate the dprintk())
2007-04-10 13:05:26 +08:00
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
return 0;
out_err:
nfs: replace remaining __FUNCTION__ occurrences __FUNCTION__ is gcc-specific, use __func__ Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-05-03 04:42:44 +08:00
dprintk("%s: statfs error = %d\n", __func__, -error);
NFS: statfs error-handling fix The nfs statfs function returns a success code on error, and fills the output buffer with invalid values. The attached patch makes it return a correct error code instead. Signed-off-by: Amnon Aaronsohn <amnonaar@gmail.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> (Modified patch to reinstate the dprintk())
2007-04-10 13:05:26 +08:00
return error;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Fix up split of fs/nfs/inode.c Fix ups for the splitting of the superblock stuff out of fs/nfs/inode.c, including: (*) Move the callback tcpport module param into callback.c. (*) Move the idmap cache timeout module param into idmap.c. (*) Changes to internal.h: (*) namespace-nfs4.c was renamed to nfs4namespace.c. (*) nfs_stat_to_errno() is in nfs2xdr.c, not nfs4xdr.c. (*) nfs4xdr.c is contingent on CONFIG_NFS_V4. (*) nfs4_path() is only uses if CONFIG_NFS_V4 is set. Plus also: (*) The sec_flavours[] table should really be const. Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:07 +08:00
/*
* Map the security flavour number to a name
*/
NFS: Display the chosen RPCSEC_GSS security flavour in /proc/mounts Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-06-09 21:34:34 +08:00
static const char *nfs_pseudoflavour_to_name(rpc_authflavor_t flavour)
{
NFS: Fix up split of fs/nfs/inode.c Fix ups for the splitting of the superblock stuff out of fs/nfs/inode.c, including: (*) Move the callback tcpport module param into callback.c. (*) Move the idmap cache timeout module param into idmap.c. (*) Changes to internal.h: (*) namespace-nfs4.c was renamed to nfs4namespace.c. (*) nfs_stat_to_errno() is in nfs2xdr.c, not nfs4xdr.c. (*) nfs4xdr.c is contingent on CONFIG_NFS_V4. (*) nfs4_path() is only uses if CONFIG_NFS_V4 is set. Plus also: (*) The sec_flavours[] table should really be const. Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:07 +08:00
static const struct {
NFS: Display the chosen RPCSEC_GSS security flavour in /proc/mounts Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-06-09 21:34:34 +08:00
rpc_authflavor_t flavour;
const char *str;
} sec_flavours[] = {
{ RPC_AUTH_NULL, "null" },
{ RPC_AUTH_UNIX, "sys" },
{ RPC_AUTH_GSS_KRB5, "krb5" },
{ RPC_AUTH_GSS_KRB5I, "krb5i" },
{ RPC_AUTH_GSS_KRB5P, "krb5p" },
{ RPC_AUTH_GSS_LKEY, "lkey" },
{ RPC_AUTH_GSS_LKEYI, "lkeyi" },
{ RPC_AUTH_GSS_LKEYP, "lkeyp" },
{ RPC_AUTH_GSS_SPKM, "spkm" },
{ RPC_AUTH_GSS_SPKMI, "spkmi" },
{ RPC_AUTH_GSS_SPKMP, "spkmp" },
NFS: Clean-up: fix a compiler warning in fs/nfs/super.c /home/cel/linux/fs/nfs/super.c: In function 'nfs_pseudoflavour_to_name': /home/cel/linux/fs/nfs/super.c:270: warning: comparison between signed and unsigned Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:40 +08:00
{ UINT_MAX, "unknown" }
NFS: Display the chosen RPCSEC_GSS security flavour in /proc/mounts Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-06-09 21:34:34 +08:00
};
int i;
NFS: Clean-up: fix a compiler warning in fs/nfs/super.c /home/cel/linux/fs/nfs/super.c: In function 'nfs_pseudoflavour_to_name': /home/cel/linux/fs/nfs/super.c:270: warning: comparison between signed and unsigned Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:40 +08:00
for (i = 0; sec_flavours[i].flavour != UINT_MAX; i++) {
NFS: Display the chosen RPCSEC_GSS security flavour in /proc/mounts Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-06-09 21:34:34 +08:00
if (sec_flavours[i].flavour == flavour)
break;
}
return sec_flavours[i].str;
}
NFS: Show most mount options via nfs_show_options() Display all mount options in /proc/mount which may be needed to reconstruct a previous mount. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Miklos Szeredi <miklos@szeredi.hu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:37 +08:00
static void nfs_show_mountd_options(struct seq_file *m, struct nfs_server *nfss,
int showdefaults)
{
struct sockaddr *sap = (struct sockaddr *)&nfss->mountd_address;
switch (sap->sa_family) {
case AF_INET: {
struct sockaddr_in *sin = (struct sockaddr_in *)sap;
fs: replace NIPQUAD() Using NIPQUAD() with NIPQUAD_FMT, %d.%d.%d.%d or %u.%u.%u.%u can be replaced with %pI4 Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-31 15:56:28 +08:00
seq_printf(m, ",mountaddr=%pI4", &sin->sin_addr.s_addr);
NFS: Show most mount options via nfs_show_options() Display all mount options in /proc/mount which may be needed to reconstruct a previous mount. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Miklos Szeredi <miklos@szeredi.hu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:37 +08:00
break;
}
case AF_INET6: {
struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sap;
net: replace %p6 with %pI6 Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-30 03:52:50 +08:00
seq_printf(m, ",mountaddr=%pI6", &sin6->sin6_addr);
NFS: Show most mount options via nfs_show_options() Display all mount options in /proc/mount which may be needed to reconstruct a previous mount. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Miklos Szeredi <miklos@szeredi.hu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:37 +08:00
break;
}
default:
if (showdefaults)
seq_printf(m, ",mountaddr=unspecified");
}
if (nfss->mountd_version || showdefaults)
seq_printf(m, ",mountvers=%u", nfss->mountd_version);
if (nfss->mountd_port || showdefaults)
seq_printf(m, ",mountport=%u", nfss->mountd_port);
switch (nfss->mountd_protocol) {
case IPPROTO_UDP:
seq_printf(m, ",mountproto=udp");
break;
case IPPROTO_TCP:
seq_printf(m, ",mountproto=tcp");
break;
default:
if (showdefaults)
seq_printf(m, ",mountproto=auto");
}
}
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
/*
* Describe the mount options in force on this server representation
*/
NFS: Show most mount options via nfs_show_options() Display all mount options in /proc/mount which may be needed to reconstruct a previous mount. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Miklos Szeredi <miklos@szeredi.hu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:37 +08:00
static void nfs_show_mount_options(struct seq_file *m, struct nfs_server *nfss,
int showdefaults)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{
NFS: Add extra const qualifiers Add some extra const qualifiers into NFS. Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:11 +08:00
static const struct proc_nfs_info {
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
int flag;
NFS: Add extra const qualifiers Add some extra const qualifiers into NFS. Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:11 +08:00
const char *str;
const char *nostr;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
} nfs_info[] = {
{ NFS_MOUNT_SOFT, ",soft", ",hard" },
NFS: Show most mount options via nfs_show_options() Display all mount options in /proc/mount which may be needed to reconstruct a previous mount. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Miklos Szeredi <miklos@szeredi.hu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:37 +08:00
{ NFS_MOUNT_POSIX, ",posix", "" },
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{ NFS_MOUNT_NOCTO, ",nocto", "" },
{ NFS_MOUNT_NOAC, ",noac", "" },
{ NFS_MOUNT_NONLM, ",nolock", "" },
{ NFS_MOUNT_NOACL, ",noacl", "" },
NFS: Added support to turn off the NFSv3 READDIRPLUS RPC. READDIRPLUS can be a performance hindrance when the client is working with large directories. In addition, some servers still have bugs in their implementations (e.g. Tru64 returns wrong values for the fsid). Add a mount flag to enable users to turn it off at mount time following the implementation in Apple's NFS client. Signed-off-by: Steve Dickson <steved@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-04-15 05:01:15 +08:00
{ NFS_MOUNT_NORDIRPLUS, ",nordirplus", "" },
NFS: add "[no]resvport" mount option The standard default security setting for NFS is AUTH_SYS. An NFS client connects to NFS servers via a privileged source port and a fixed standard destination port (2049). The client sends raw uid and gid numbers to identify users making NFS requests, and the server assumes an appropriate authority on the client has vetted these values because the source port is privileged. On Linux, by default in-kernel RPC services use a privileged port in the range between 650 and 1023 to avoid using source ports of well- known IP services. Using such a small range limits the number of NFS mount points and the number of unique NFS servers to which a client can connect concurrently. An NFS client can use unprivileged source ports to expand the range of source port numbers, allowing more concurrent server connections and more NFS mount points. Servers must explicitly allow NFS connections from unprivileged ports for this to work. In the past, bumping the value of the sunrpc.max_resvport sysctl on the client would permit the NFS client to use unprivileged ports. Bumping this setting also changes the maximum port number used by other in-kernel RPC services, some of which still required a port number less than 1023. This is exacerbated by the way source port numbers are chosen by the Linux RPC client, which starts at the top of the range and works downwards. It means that bumping the maximum means all RPC services requesting a source port will likely get an unprivileged port instead of a privileged one. Changing this setting effects all NFS mount points on a client. A sysadmin could not selectively choose which mount points would use non-privileged ports and which could not. Lastly, this mechanism of expanding the limit on the number of NFS mount points was entirely undocumented. To address the need for the NFS client to use a large range of source ports without interfering with the activity of other in-kernel RPC services, we introduce a new NFS mount option. This option explicitly tells only the NFS client to use a non-privileged source port when communicating with the NFS server for one specific mount point. This new mount option is called "resvport," like the similar NFS mount option on FreeBSD and Mac OS X. A sister patch for nfs-utils will be submitted that documents this new option in nfs(5). The default setting for this new mount option requires the NFS client to use a privileged port, as before. Explicitly specifying the "noresvport" mount option allows the NFS client to use an unprivileged source port for this mount point when connecting to the NFS server port. This mount option is supported only for text-based NFS mounts. [ Sidebar: it is widely known that security mechanisms based on the use of privileged source ports are ineffective. However, the NFS client can combine the use of unprivileged ports with the use of secure authentication mechanisms, such as Kerberos. This allows a large number of connections and mount points while ensuring a useful level of security. Eventually we may change the default setting for this option depending on the security flavor used for the mount. For example, if the mount is using only AUTH_SYS, then the default setting will be "resvport;" if the mount is using a strong security flavor such as krb5, the default setting will be "noresvport." ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond.Myklebust@netapp.com: Fixed a bug whereby nfs4_init_client() was being called with incorrect arguments.] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:37 +08:00
{ NFS_MOUNT_UNSHARED, ",nosharecache", "" },
{ NFS_MOUNT_NORESVPORT, ",noresvport", "" },
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{ 0, NULL, NULL }
};
NFS: Add extra const qualifiers Add some extra const qualifiers into NFS. Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:11 +08:00
const struct proc_nfs_info *nfs_infop;
NFS: Move rpc_ops from nfs_server to nfs_client Move the rpc_ops from the nfs_server struct to the nfs_client struct as they're common to all server records of a particular NFS protocol version. Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:12 +08:00
struct nfs_client *clp = nfss->nfs_client;
NFS: Show most mount options via nfs_show_options() Display all mount options in /proc/mount which may be needed to reconstruct a previous mount. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Miklos Szeredi <miklos@szeredi.hu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:37 +08:00
u32 version = clp->rpc_ops->version;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Show most mount options via nfs_show_options() Display all mount options in /proc/mount which may be needed to reconstruct a previous mount. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Miklos Szeredi <miklos@szeredi.hu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:37 +08:00
seq_printf(m, ",vers=%u", version);
NFS: numeric mount parameters are unsigned Clean up: use %u instead of %d when displaying NFS mount options. Nit: Fix reporting of "namlen=" option in nfs_show_mount_stats. The mount option is called "namlen" without the "e". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:08 +08:00
seq_printf(m, ",rsize=%u", nfss->rsize);
seq_printf(m, ",wsize=%u", nfss->wsize);
NFS: Show most mount options via nfs_show_options() Display all mount options in /proc/mount which may be needed to reconstruct a previous mount. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Miklos Szeredi <miklos@szeredi.hu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:37 +08:00
if (nfss->bsize != 0)
seq_printf(m, ",bsize=%u", nfss->bsize);
seq_printf(m, ",namlen=%u", nfss->namelen);
NFS: use documenting macro constants for initializing ac{reg, dir}{min, max} Clean up. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-27 05:47:12 +08:00
if (nfss->acregmin != NFS_DEF_ACREGMIN*HZ || showdefaults)
NFS: numeric mount parameters are unsigned Clean up: use %u instead of %d when displaying NFS mount options. Nit: Fix reporting of "namlen=" option in nfs_show_mount_stats. The mount option is called "namlen" without the "e". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:08 +08:00
seq_printf(m, ",acregmin=%u", nfss->acregmin/HZ);
NFS: use documenting macro constants for initializing ac{reg, dir}{min, max} Clean up. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-27 05:47:12 +08:00
if (nfss->acregmax != NFS_DEF_ACREGMAX*HZ || showdefaults)
NFS: numeric mount parameters are unsigned Clean up: use %u instead of %d when displaying NFS mount options. Nit: Fix reporting of "namlen=" option in nfs_show_mount_stats. The mount option is called "namlen" without the "e". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:08 +08:00
seq_printf(m, ",acregmax=%u", nfss->acregmax/HZ);
NFS: use documenting macro constants for initializing ac{reg, dir}{min, max} Clean up. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-27 05:47:12 +08:00
if (nfss->acdirmin != NFS_DEF_ACDIRMIN*HZ || showdefaults)
NFS: numeric mount parameters are unsigned Clean up: use %u instead of %d when displaying NFS mount options. Nit: Fix reporting of "namlen=" option in nfs_show_mount_stats. The mount option is called "namlen" without the "e". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:08 +08:00
seq_printf(m, ",acdirmin=%u", nfss->acdirmin/HZ);
NFS: use documenting macro constants for initializing ac{reg, dir}{min, max} Clean up. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-27 05:47:12 +08:00
if (nfss->acdirmax != NFS_DEF_ACDIRMAX*HZ || showdefaults)
NFS: numeric mount parameters are unsigned Clean up: use %u instead of %d when displaying NFS mount options. Nit: Fix reporting of "namlen=" option in nfs_show_mount_stats. The mount option is called "namlen" without the "e". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:08 +08:00
seq_printf(m, ",acdirmax=%u", nfss->acdirmax/HZ);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
if (nfss->flags & nfs_infop->flag)
seq_puts(m, nfs_infop->str);
else
seq_puts(m, nfs_infop->nostr);
}
NFS - print accurate transport protocol Use the per-transport strings to display the transport protocol accurately. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:48:47 +08:00
seq_printf(m, ",proto=%s",
rpc_peeraddr2str(nfss->client, RPC_DISPLAY_PROTO));
NFS: Show most mount options via nfs_show_options() Display all mount options in /proc/mount which may be needed to reconstruct a previous mount. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Miklos Szeredi <miklos@szeredi.hu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:37 +08:00
if (version == 4) {
if (nfss->port != NFS_PORT)
seq_printf(m, ",port=%u", nfss->port);
} else
if (nfss->port)
seq_printf(m, ",port=%u", nfss->port);
NFS: Support per-mountpoint timeout parameters. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-21 05:03:59 +08:00
seq_printf(m, ",timeo=%lu", 10U * nfss->client->cl_timeout->to_initval / HZ);
seq_printf(m, ",retrans=%u", nfss->client->cl_timeout->to_retries);
NFS: Display the chosen RPCSEC_GSS security flavour in /proc/mounts Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-06-09 21:34:34 +08:00
seq_printf(m, ",sec=%s", nfs_pseudoflavour_to_name(nfss->client->cl_auth->au_flavor));
NFS: Show most mount options via nfs_show_options() Display all mount options in /proc/mount which may be needed to reconstruct a previous mount. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Miklos Szeredi <miklos@szeredi.hu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:37 +08:00
if (version != 4)
nfs_show_mountd_options(m, nfss, showdefaults);
#ifdef CONFIG_NFS_V4
if (clp->rpc_ops->version == 4)
seq_printf(m, ",clientaddr=%s", clp->cl_ipaddr);
#endif
NFS: Add mount options to enable local caching on NFS Add NFS mount options to allow the local caching support to be enabled. The attached patch makes it possible for the NFS filesystem to be told to make use of the network filesystem local caching service (FS-Cache). To be able to use this, a recent nfsutils package is required. There are three variant NFS mount options that can be added to a mount command to control caching for a mount. Only the last one specified takes effect: (*) Adding "fsc" will request caching. (*) Adding "fsc=<string>" will request caching and also specify a uniquifier. (*) Adding "nofsc" will disable caching. For example: mount warthog:/ /a -o fsc The cache of a particular superblock (NFS FSID) will be shared between all mounts of that volume, provided they have the same connection parameters and are not marked 'nosharecache'. Where it is otherwise impossible to distinguish superblocks because all the parameters are identical, but the 'nosharecache' option is supplied, a uniquifying string must be supplied, else only the first mount will be permitted to use the cache. If there's a key collision, then the second mount will disable caching and give a warning into the kernel log. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Steve Dickson <steved@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Daire Byrne <Daire.Byrne@framestore.com>
2009-04-03 23:42:48 +08:00
if (nfss->options & NFS_OPTION_FSCACHE)
seq_printf(m, ",fsc");
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
/*
* Describe the mount options on this VFS mountpoint
*/
static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
{
struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
nfs_show_mount_options(m, nfss, 0);
NFS: eliminate NIPQUAD(clp->cl_addr.sin_addr) To ensure the NFS client displays IPv6 addresses properly, replace address family-specific NIPQUAD() invocations with a call to the RPC client to get a formatted string representing the remote peer's address. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:57:16 +08:00
seq_printf(m, ",addr=%s",
rpc_peeraddr2str(nfss->nfs_client->cl_rpcclient,
RPC_DISPLAY_ADDR));
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
return 0;
}
/*
* Present statistical information for this VFS mountpoint
*/
static int nfs_show_stats(struct seq_file *m, struct vfsmount *mnt)
{
int i, cpu;
struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
struct rpc_auth *auth = nfss->client->cl_auth;
struct nfs_iostats totals = { };
seq_printf(m, "statvers=%s", NFS_IOSTAT_VERS);
/*
* Display all mount option settings
*/
seq_printf(m, "\n\topts:\t");
seq_puts(m, mnt->mnt_sb->s_flags & MS_RDONLY ? "ro" : "rw");
seq_puts(m, mnt->mnt_sb->s_flags & MS_SYNCHRONOUS ? ",sync" : "");
seq_puts(m, mnt->mnt_sb->s_flags & MS_NOATIME ? ",noatime" : "");
seq_puts(m, mnt->mnt_sb->s_flags & MS_NODIRATIME ? ",nodiratime" : "");
nfs_show_mount_options(m, nfss, 1);
seq_printf(m, "\n\tage:\t%lu", (jiffies - nfss->mount_time) / HZ);
seq_printf(m, "\n\tcaps:\t");
seq_printf(m, "caps=0x%x", nfss->caps);
NFS: numeric mount parameters are unsigned Clean up: use %u instead of %d when displaying NFS mount options. Nit: Fix reporting of "namlen=" option in nfs_show_mount_stats. The mount option is called "namlen" without the "e". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:08 +08:00
seq_printf(m, ",wtmult=%u", nfss->wtmult);
seq_printf(m, ",dtsize=%u", nfss->dtsize);
seq_printf(m, ",bsize=%u", nfss->bsize);
seq_printf(m, ",namlen=%u", nfss->namelen);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
#ifdef CONFIG_NFS_V4
NFS: Remove the redundant nfs_client->cl_nfsversion We can get the same information from the rpc_ops structure instead. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-15 03:56:07 +08:00
if (nfss->nfs_client->rpc_ops->version == 4) {
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
seq_printf(m, "\n\tnfsv4:\t");
seq_printf(m, "bm0=0x%x", nfss->attr_bitmask[0]);
seq_printf(m, ",bm1=0x%x", nfss->attr_bitmask[1]);
seq_printf(m, ",acl=0x%x", nfss->acl_bitmask);
}
#endif
/*
* Display security flavor in effect for this mount
*/
NFS: numeric mount parameters are unsigned Clean up: use %u instead of %d when displaying NFS mount options. Nit: Fix reporting of "namlen=" option in nfs_show_mount_stats. The mount option is called "namlen" without the "e". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:08 +08:00
seq_printf(m, "\n\tsec:\tflavor=%u", auth->au_ops->au_flavor);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
if (auth->au_flavor)
NFS: numeric mount parameters are unsigned Clean up: use %u instead of %d when displaying NFS mount options. Nit: Fix reporting of "namlen=" option in nfs_show_mount_stats. The mount option is called "namlen" without the "e". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:08 +08:00
seq_printf(m, ",pseudoflavor=%u", auth->au_flavor);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
/*
* Display superblock I/O counters
*/
for_each_possible_cpu(cpu) {
struct nfs_iostats *stats;
preempt_disable();
stats = per_cpu_ptr(nfss->io_stats, cpu);
for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
totals.events[i] += stats->events[i];
for (i = 0; i < __NFSIOS_BYTESMAX; i++)
totals.bytes[i] += stats->bytes[i];
NFS: Add some new I/O counters for FS-Cache doing things for NFS Add some new NFS I/O counters for FS-Cache doing things for NFS. A new line is emitted into /proc/pid/mountstats if caching is enabled that looks like: fsc: <rok> <rfl> <wok> <wfl> <unc> Where <rok> is the number of pages read successfully from the cache, <rfl> is the number of failed page reads against the cache, <wok> is the number of successful page writes to the cache, <wfl> is the number of failed page writes to the cache, and <unc> is the number of NFS pages that have been disconnected from the cache. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Steve Dickson <steved@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Daire Byrne <Daire.Byrne@framestore.com>
2009-04-03 23:42:43 +08:00
#ifdef CONFIG_NFS_FSCACHE
for (i = 0; i < __NFSIOS_FSCACHEMAX; i++)
totals.fscache[i] += stats->fscache[i];
#endif
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
preempt_enable();
}
seq_printf(m, "\n\tevents:\t");
for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
seq_printf(m, "%lu ", totals.events[i]);
seq_printf(m, "\n\tbytes:\t");
for (i = 0; i < __NFSIOS_BYTESMAX; i++)
seq_printf(m, "%Lu ", totals.bytes[i]);
NFS: Add some new I/O counters for FS-Cache doing things for NFS Add some new NFS I/O counters for FS-Cache doing things for NFS. A new line is emitted into /proc/pid/mountstats if caching is enabled that looks like: fsc: <rok> <rfl> <wok> <wfl> <unc> Where <rok> is the number of pages read successfully from the cache, <rfl> is the number of failed page reads against the cache, <wok> is the number of successful page writes to the cache, <wfl> is the number of failed page writes to the cache, and <unc> is the number of NFS pages that have been disconnected from the cache. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Steve Dickson <steved@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Daire Byrne <Daire.Byrne@framestore.com>
2009-04-03 23:42:43 +08:00
#ifdef CONFIG_NFS_FSCACHE
if (nfss->options & NFS_OPTION_FSCACHE) {
seq_printf(m, "\n\tfsc:\t");
for (i = 0; i < __NFSIOS_FSCACHEMAX; i++)
seq_printf(m, "%Lu ", totals.bytes[i]);
}
#endif
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
seq_printf(m, "\n");
rpc_print_iostats(m, nfss->client);
return 0;
}
/*
* Begin unmount by attempting to remove all automounted mountpoints we added
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
* in response to xdev traversals and referrals
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
*/
[PATCH] restore sane ->umount_begin() API Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2008-04-24 19:21:56 +08:00
static void nfs_umount_begin(struct super_block *sb)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{
vfs: umount_begin BKL pushdown Push BKL down into ->umount_begin() Signed-off-by: Alessio Igor Bogani <abogani@texware.it> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2009-04-24 15:06:53 +08:00
struct nfs_server *server;
NFS: Re-enable forced umounts They disappeared some time around 2.6.18. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-06-06 07:13:47 +08:00
struct rpc_clnt *rpc;
vfs: umount_begin BKL pushdown Push BKL down into ->umount_begin() Signed-off-by: Alessio Igor Bogani <abogani@texware.it> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2009-04-24 15:06:53 +08:00
lock_kernel();
server = NFS_SB(sb);
NFS: Re-enable forced umounts They disappeared some time around 2.6.18. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-06-06 07:13:47 +08:00
/* -EIO all pending I/O */
rpc = server->client_acl;
if (!IS_ERR(rpc))
rpc_killall_tasks(rpc);
rpc = server->client;
if (!IS_ERR(rpc))
rpc_killall_tasks(rpc);
vfs: umount_begin BKL pushdown Push BKL down into ->umount_begin() Signed-off-by: Alessio Igor Bogani <abogani@texware.it> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2009-04-24 15:06:53 +08:00
unlock_kernel();
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Clean-up: Refactor IP address sanity checks in NFS client NFS and NFSv4 mounts can now share server address sanity checking. And, it provides an easy mechanism for adding IPv6 address checking at some later point. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:46 +08:00
/*
NFS: Verify IPv6 addresses properly Add support to nfs_verify_server_address for recognizing AF_INET6 addresses. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:57:45 +08:00
* Sanity-check a server address provided by the mount command.
*
* Address family must be initialized, and address must not be
* the ANY address for that family.
NFS: Clean-up: Refactor IP address sanity checks in NFS client NFS and NFSv4 mounts can now share server address sanity checking. And, it provides an easy mechanism for adding IPv6 address checking at some later point. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:46 +08:00
*/
static int nfs_verify_server_address(struct sockaddr *addr)
{
switch (addr->sa_family) {
case AF_INET: {
NFS: Verify IPv6 addresses properly Add support to nfs_verify_server_address for recognizing AF_INET6 addresses. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:57:45 +08:00
struct sockaddr_in *sa = (struct sockaddr_in *)addr;
[NET] endianness noise: INADDR_ANY Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-03-18 13:44:53 +08:00
return sa->sin_addr.s_addr != htonl(INADDR_ANY);
NFS: Verify IPv6 addresses properly Add support to nfs_verify_server_address for recognizing AF_INET6 addresses. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:57:45 +08:00
}
case AF_INET6: {
struct in6_addr *sa = &((struct sockaddr_in6 *)addr)->sin6_addr;
return !ipv6_addr_any(sa);
NFS: Clean-up: Refactor IP address sanity checks in NFS client NFS and NFSv4 mounts can now share server address sanity checking. And, it provides an easy mechanism for adding IPv6 address checking at some later point. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:46 +08:00
}
}
return 0;
}
NFS: Add string length argument to nfs_parse_server_address To make nfs_parse_server_address() more generally useful, allow it to accept input strings that are not terminated with '\0'. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:53 +08:00
static void nfs_parse_ipv4_address(char *string, size_t str_len,
struct sockaddr *sap, size_t *addr_len)
NFS: Refactor mount option address parsing into separate function Refactor the logic to parse incoming text-based IP addresses. Use the in4_pton() function instead of the older in_aton(), following the lead of the in-kernel CIFS client. Later we'll add IPv6 address parsing using the matching in6_pton() function. For now we can't allow IPv6 address parsing: we must expand the size of the address storage fields in the nfs_parsed_mount_options struct before we can parse and store IPv6 addresses. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:21 +08:00
{
NFS: Add string length argument to nfs_parse_server_address To make nfs_parse_server_address() more generally useful, allow it to accept input strings that are not terminated with '\0'. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:53 +08:00
struct sockaddr_in *sin = (struct sockaddr_in *)sap;
u8 *addr = (u8 *)&sin->sin_addr.s_addr;
NFS: Refactor mount option address parsing into separate function Refactor the logic to parse incoming text-based IP addresses. Use the in4_pton() function instead of the older in_aton(), following the lead of the in-kernel CIFS client. Later we'll add IPv6 address parsing using the matching in6_pton() function. For now we can't allow IPv6 address parsing: we must expand the size of the address storage fields in the nfs_parsed_mount_options struct before we can parse and store IPv6 addresses. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:21 +08:00
NFS: Add string length argument to nfs_parse_server_address To make nfs_parse_server_address() more generally useful, allow it to accept input strings that are not terminated with '\0'. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:53 +08:00
if (str_len <= INET_ADDRSTRLEN) {
dfprintk(MOUNT, "NFS: parsing IPv4 address %*s\n",
(int)str_len, string);
sin->sin_family = AF_INET;
*addr_len = sizeof(*sin);
if (in4_pton(string, str_len, addr, '\0', NULL))
NFS: Pull covers off IPv6 address parsing Now that the needed IPv6 infrastructure is in place, allow the NFS client's IP address parser to generate AF_INET6 addresses. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:35 +08:00
return;
NFS: Add string length argument to nfs_parse_server_address To make nfs_parse_server_address() more generally useful, allow it to accept input strings that are not terminated with '\0'. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:53 +08:00
}
sap->sa_family = AF_UNSPEC;
*addr_len = 0;
}
#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
NFS: fix nfs_parse_ip_address() corner case Bruce observed that nfs_parse_ip_address() will successfully parse an IPv6 address that looks like this: "::1%" A scope delimiter is present, but there is no scope ID following it. This is harmless, as it would simply set the scope ID to zero. However, in some cases we would like to flag this as an improperly formed address. We are now also careful to reject addresses where garbage follows the address (up to the length of the string), instead of ignoring the non-address characters; and where the scope ID is nonsense (not a valid device name, but also not numeric). Before, both of these cases would result in a harmless zero scope ID. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-10-09 03:38:10 +08:00
static int nfs_parse_ipv6_scope_id(const char *string, const size_t str_len,
const char *delim,
struct sockaddr_in6 *sin6)
NFS: handle interface identifiers in incoming IPv6 addresses Add support in the kernel NFS client's address parser for interface identifiers. IPv6 link-local addresses require an additional "interface identifier", which is a network device name or an integer that indexes the array of local network interfaces. They are suffixed to the address with a '%'. For example: fe80::215:c5ff:fe3b:e1b2%2 indicates an interface index of 2. Or fe80::215:c5ff:fe3b:e1b2%eth0 indicates that requests should be routed through the eth0 device. Without the interface ID, link-local addresses are not usable for NFS. Both the kernel NFS client mount option parser and the mount.nfs command can take either form. The mount.nfs command always passes the address through getnameinfo(3), which usually re-writes interface indices as device names. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:37:01 +08:00
{
char *p;
size_t len;
NFS: fix nfs_parse_ip_address() corner case Bruce observed that nfs_parse_ip_address() will successfully parse an IPv6 address that looks like this: "::1%" A scope delimiter is present, but there is no scope ID following it. This is harmless, as it would simply set the scope ID to zero. However, in some cases we would like to flag this as an improperly formed address. We are now also careful to reject addresses where garbage follows the address (up to the length of the string), instead of ignoring the non-address characters; and where the scope ID is nonsense (not a valid device name, but also not numeric). Before, both of these cases would result in a harmless zero scope ID. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-10-09 03:38:10 +08:00
if ((string + str_len) == delim)
return 1;
NFS: handle interface identifiers in incoming IPv6 addresses Add support in the kernel NFS client's address parser for interface identifiers. IPv6 link-local addresses require an additional "interface identifier", which is a network device name or an integer that indexes the array of local network interfaces. They are suffixed to the address with a '%'. For example: fe80::215:c5ff:fe3b:e1b2%2 indicates an interface index of 2. Or fe80::215:c5ff:fe3b:e1b2%eth0 indicates that requests should be routed through the eth0 device. Without the interface ID, link-local addresses are not usable for NFS. Both the kernel NFS client mount option parser and the mount.nfs command can take either form. The mount.nfs command always passes the address through getnameinfo(3), which usually re-writes interface indices as device names. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:37:01 +08:00
if (*delim != IPV6_SCOPE_DELIMITER)
NFS: fix nfs_parse_ip_address() corner case Bruce observed that nfs_parse_ip_address() will successfully parse an IPv6 address that looks like this: "::1%" A scope delimiter is present, but there is no scope ID following it. This is harmless, as it would simply set the scope ID to zero. However, in some cases we would like to flag this as an improperly formed address. We are now also careful to reject addresses where garbage follows the address (up to the length of the string), instead of ignoring the non-address characters; and where the scope ID is nonsense (not a valid device name, but also not numeric). Before, both of these cases would result in a harmless zero scope ID. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-10-09 03:38:10 +08:00
return 0;
if (!(ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL))
return 0;
NFS: handle interface identifiers in incoming IPv6 addresses Add support in the kernel NFS client's address parser for interface identifiers. IPv6 link-local addresses require an additional "interface identifier", which is a network device name or an integer that indexes the array of local network interfaces. They are suffixed to the address with a '%'. For example: fe80::215:c5ff:fe3b:e1b2%2 indicates an interface index of 2. Or fe80::215:c5ff:fe3b:e1b2%eth0 indicates that requests should be routed through the eth0 device. Without the interface ID, link-local addresses are not usable for NFS. Both the kernel NFS client mount option parser and the mount.nfs command can take either form. The mount.nfs command always passes the address through getnameinfo(3), which usually re-writes interface indices as device names. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:37:01 +08:00
len = (string + str_len) - delim - 1;
p = kstrndup(delim + 1, len, GFP_KERNEL);
if (p) {
unsigned long scope_id = 0;
struct net_device *dev;
dev = dev_get_by_name(&init_net, p);
if (dev != NULL) {
scope_id = dev->ifindex;
dev_put(dev);
} else {
NFS: fix nfs_parse_ip_address() corner case Bruce observed that nfs_parse_ip_address() will successfully parse an IPv6 address that looks like this: "::1%" A scope delimiter is present, but there is no scope ID following it. This is harmless, as it would simply set the scope ID to zero. However, in some cases we would like to flag this as an improperly formed address. We are now also careful to reject addresses where garbage follows the address (up to the length of the string), instead of ignoring the non-address characters; and where the scope ID is nonsense (not a valid device name, but also not numeric). Before, both of these cases would result in a harmless zero scope ID. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-10-09 03:38:10 +08:00
if (strict_strtoul(p, 10, &scope_id) == 0) {
kfree(p);
return 0;
}
NFS: handle interface identifiers in incoming IPv6 addresses Add support in the kernel NFS client's address parser for interface identifiers. IPv6 link-local addresses require an additional "interface identifier", which is a network device name or an integer that indexes the array of local network interfaces. They are suffixed to the address with a '%'. For example: fe80::215:c5ff:fe3b:e1b2%2 indicates an interface index of 2. Or fe80::215:c5ff:fe3b:e1b2%eth0 indicates that requests should be routed through the eth0 device. Without the interface ID, link-local addresses are not usable for NFS. Both the kernel NFS client mount option parser and the mount.nfs command can take either form. The mount.nfs command always passes the address through getnameinfo(3), which usually re-writes interface indices as device names. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:37:01 +08:00
}
kfree(p);
NFS: fix nfs_parse_ip_address() corner case Bruce observed that nfs_parse_ip_address() will successfully parse an IPv6 address that looks like this: "::1%" A scope delimiter is present, but there is no scope ID following it. This is harmless, as it would simply set the scope ID to zero. However, in some cases we would like to flag this as an improperly formed address. We are now also careful to reject addresses where garbage follows the address (up to the length of the string), instead of ignoring the non-address characters; and where the scope ID is nonsense (not a valid device name, but also not numeric). Before, both of these cases would result in a harmless zero scope ID. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-10-09 03:38:10 +08:00
NFS: handle interface identifiers in incoming IPv6 addresses Add support in the kernel NFS client's address parser for interface identifiers. IPv6 link-local addresses require an additional "interface identifier", which is a network device name or an integer that indexes the array of local network interfaces. They are suffixed to the address with a '%'. For example: fe80::215:c5ff:fe3b:e1b2%2 indicates an interface index of 2. Or fe80::215:c5ff:fe3b:e1b2%eth0 indicates that requests should be routed through the eth0 device. Without the interface ID, link-local addresses are not usable for NFS. Both the kernel NFS client mount option parser and the mount.nfs command can take either form. The mount.nfs command always passes the address through getnameinfo(3), which usually re-writes interface indices as device names. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:37:01 +08:00
sin6->sin6_scope_id = scope_id;
dfprintk(MOUNT, "NFS: IPv6 scope ID = %lu\n", scope_id);
NFS: fix nfs_parse_ip_address() corner case Bruce observed that nfs_parse_ip_address() will successfully parse an IPv6 address that looks like this: "::1%" A scope delimiter is present, but there is no scope ID following it. This is harmless, as it would simply set the scope ID to zero. However, in some cases we would like to flag this as an improperly formed address. We are now also careful to reject addresses where garbage follows the address (up to the length of the string), instead of ignoring the non-address characters; and where the scope ID is nonsense (not a valid device name, but also not numeric). Before, both of these cases would result in a harmless zero scope ID. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-10-09 03:38:10 +08:00
return 1;
NFS: handle interface identifiers in incoming IPv6 addresses Add support in the kernel NFS client's address parser for interface identifiers. IPv6 link-local addresses require an additional "interface identifier", which is a network device name or an integer that indexes the array of local network interfaces. They are suffixed to the address with a '%'. For example: fe80::215:c5ff:fe3b:e1b2%2 indicates an interface index of 2. Or fe80::215:c5ff:fe3b:e1b2%eth0 indicates that requests should be routed through the eth0 device. Without the interface ID, link-local addresses are not usable for NFS. Both the kernel NFS client mount option parser and the mount.nfs command can take either form. The mount.nfs command always passes the address through getnameinfo(3), which usually re-writes interface indices as device names. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:37:01 +08:00
}
NFS: fix nfs_parse_ip_address() corner case Bruce observed that nfs_parse_ip_address() will successfully parse an IPv6 address that looks like this: "::1%" A scope delimiter is present, but there is no scope ID following it. This is harmless, as it would simply set the scope ID to zero. However, in some cases we would like to flag this as an improperly formed address. We are now also careful to reject addresses where garbage follows the address (up to the length of the string), instead of ignoring the non-address characters; and where the scope ID is nonsense (not a valid device name, but also not numeric). Before, both of these cases would result in a harmless zero scope ID. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-10-09 03:38:10 +08:00
return 0;
NFS: handle interface identifiers in incoming IPv6 addresses Add support in the kernel NFS client's address parser for interface identifiers. IPv6 link-local addresses require an additional "interface identifier", which is a network device name or an integer that indexes the array of local network interfaces. They are suffixed to the address with a '%'. For example: fe80::215:c5ff:fe3b:e1b2%2 indicates an interface index of 2. Or fe80::215:c5ff:fe3b:e1b2%eth0 indicates that requests should be routed through the eth0 device. Without the interface ID, link-local addresses are not usable for NFS. Both the kernel NFS client mount option parser and the mount.nfs command can take either form. The mount.nfs command always passes the address through getnameinfo(3), which usually re-writes interface indices as device names. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:37:01 +08:00
}
NFS: Add string length argument to nfs_parse_server_address To make nfs_parse_server_address() more generally useful, allow it to accept input strings that are not terminated with '\0'. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:53 +08:00
static void nfs_parse_ipv6_address(char *string, size_t str_len,
struct sockaddr *sap, size_t *addr_len)
{
struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sap;
u8 *addr = (u8 *)&sin6->sin6_addr.in6_u;
NFS: handle interface identifiers in incoming IPv6 addresses Add support in the kernel NFS client's address parser for interface identifiers. IPv6 link-local addresses require an additional "interface identifier", which is a network device name or an integer that indexes the array of local network interfaces. They are suffixed to the address with a '%'. For example: fe80::215:c5ff:fe3b:e1b2%2 indicates an interface index of 2. Or fe80::215:c5ff:fe3b:e1b2%eth0 indicates that requests should be routed through the eth0 device. Without the interface ID, link-local addresses are not usable for NFS. Both the kernel NFS client mount option parser and the mount.nfs command can take either form. The mount.nfs command always passes the address through getnameinfo(3), which usually re-writes interface indices as device names. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:37:01 +08:00
const char *delim;
NFS: Add string length argument to nfs_parse_server_address To make nfs_parse_server_address() more generally useful, allow it to accept input strings that are not terminated with '\0'. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:53 +08:00
if (str_len <= INET6_ADDRSTRLEN) {
dfprintk(MOUNT, "NFS: parsing IPv6 address %*s\n",
(int)str_len, string);
NFS: Pull covers off IPv6 address parsing Now that the needed IPv6 infrastructure is in place, allow the NFS client's IP address parser to generate AF_INET6 addresses. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:35 +08:00
NFS: Add string length argument to nfs_parse_server_address To make nfs_parse_server_address() more generally useful, allow it to accept input strings that are not terminated with '\0'. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:53 +08:00
sin6->sin6_family = AF_INET6;
*addr_len = sizeof(*sin6);
NFS: fix nfs_parse_ip_address() corner case Bruce observed that nfs_parse_ip_address() will successfully parse an IPv6 address that looks like this: "::1%" A scope delimiter is present, but there is no scope ID following it. This is harmless, as it would simply set the scope ID to zero. However, in some cases we would like to flag this as an improperly formed address. We are now also careful to reject addresses where garbage follows the address (up to the length of the string), instead of ignoring the non-address characters; and where the scope ID is nonsense (not a valid device name, but also not numeric). Before, both of these cases would result in a harmless zero scope ID. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-10-09 03:38:10 +08:00
if (in6_pton(string, str_len, addr,
IPV6_SCOPE_DELIMITER, &delim) != 0) {
if (nfs_parse_ipv6_scope_id(string, str_len,
delim, sin6) != 0)
return;
NFS: handle interface identifiers in incoming IPv6 addresses Add support in the kernel NFS client's address parser for interface identifiers. IPv6 link-local addresses require an additional "interface identifier", which is a network device name or an integer that indexes the array of local network interfaces. They are suffixed to the address with a '%'. For example: fe80::215:c5ff:fe3b:e1b2%2 indicates an interface index of 2. Or fe80::215:c5ff:fe3b:e1b2%eth0 indicates that requests should be routed through the eth0 device. Without the interface ID, link-local addresses are not usable for NFS. Both the kernel NFS client mount option parser and the mount.nfs command can take either form. The mount.nfs command always passes the address through getnameinfo(3), which usually re-writes interface indices as device names. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:37:01 +08:00
}
NFS: Pull covers off IPv6 address parsing Now that the needed IPv6 infrastructure is in place, allow the NFS client's IP address parser to generate AF_INET6 addresses. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:35 +08:00
}
NFS: Refactor mount option address parsing into separate function Refactor the logic to parse incoming text-based IP addresses. Use the in4_pton() function instead of the older in_aton(), following the lead of the in-kernel CIFS client. Later we'll add IPv6 address parsing using the matching in6_pton() function. For now we can't allow IPv6 address parsing: we must expand the size of the address storage fields in the nfs_parsed_mount_options struct before we can parse and store IPv6 addresses. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:21 +08:00
sap->sa_family = AF_UNSPEC;
NFS: Add string length argument to nfs_parse_server_address To make nfs_parse_server_address() more generally useful, allow it to accept input strings that are not terminated with '\0'. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:53 +08:00
*addr_len = 0;
}
#else
static void nfs_parse_ipv6_address(char *string, size_t str_len,
struct sockaddr *sap, size_t *addr_len)
{
sap->sa_family = AF_UNSPEC;
*addr_len = 0;
}
#endif
/*
* Construct a sockaddr based on the contents of a string that contains
* an IP address in presentation format.
*
* If there is a problem constructing the new sockaddr, set the address
* family to AF_UNSPEC.
*/
nfs: Fix misparsing of nfsv4 fs_locations attribute The code incorrectly assumes here that the server name (or ip address) is null-terminated. This can cause referrals to fail in some cases. Also support ipv6 addresses. Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-08-21 04:10:23 +08:00
void nfs_parse_ip_address(char *string, size_t str_len,
NFS: Add string length argument to nfs_parse_server_address To make nfs_parse_server_address() more generally useful, allow it to accept input strings that are not terminated with '\0'. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:53 +08:00
struct sockaddr *sap, size_t *addr_len)
{
unsigned int i, colons;
colons = 0;
for (i = 0; i < str_len; i++)
if (string[i] == ':')
colons++;
if (colons >= 2)
nfs_parse_ipv6_address(string, str_len, sap, addr_len);
else
nfs_parse_ipv4_address(string, str_len, sap, addr_len);
NFS: Refactor mount option address parsing into separate function Refactor the logic to parse incoming text-based IP addresses. Use the in4_pton() function instead of the older in_aton(), following the lead of the in-kernel CIFS client. Later we'll add IPv6 address parsing using the matching in6_pton() function. For now we can't allow IPv6 address parsing: we must expand the size of the address storage fields in the nfs_parsed_mount_options struct before we can parse and store IPv6 addresses. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:21 +08:00
}
NFS: set transport defaults after mount option parsing is finished Move the UDP/TCP default timeo/retrans settings for text mounts to nfs_init_timeout_values(), which was were they were always being initialised (and sanity checked) for binary mounts. Document the default timeout values using appropriate #defines. Ensure that we initialise and sanity check the transport protocols that may have been specified by the user. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-03 02:43:47 +08:00
/*
* Sanity check the NFS transport protocol.
*
*/
static void nfs_validate_transport_protocol(struct nfs_parsed_mount_data *mnt)
{
switch (mnt->nfs_server.protocol) {
case XPRT_TRANSPORT_UDP:
case XPRT_TRANSPORT_TCP:
case XPRT_TRANSPORT_RDMA:
break;
default:
mnt->nfs_server.protocol = XPRT_TRANSPORT_TCP;
}
}
/*
* For text based NFSv2/v3 mounts, the mount protocol transport default
* settings should depend upon the specified NFS transport.
*/
static void nfs_set_mount_transport_protocol(struct nfs_parsed_mount_data *mnt)
{
nfs_validate_transport_protocol(mnt);
if (mnt->mount_server.protocol == XPRT_TRANSPORT_UDP ||
mnt->mount_server.protocol == XPRT_TRANSPORT_TCP)
return;
switch (mnt->nfs_server.protocol) {
case XPRT_TRANSPORT_UDP:
mnt->mount_server.protocol = XPRT_TRANSPORT_UDP;
break;
case XPRT_TRANSPORT_TCP:
case XPRT_TRANSPORT_RDMA:
mnt->mount_server.protocol = XPRT_TRANSPORT_TCP;
}
}
NFS: Refactor logic for parsing NFS security flavor mount options Clean up: Refactor the NFS mount option parsing function to extract the security flavor parsing logic into a separate function. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 04:33:38 +08:00
/*
* Parse the value of the 'sec=' option.
*
NFS: Set security flavor default for NFSv2/3 mounts like other defaults Set the default security flavor when we set the other mount option default values. After this change, only the legacy user-space mount path needs to set the NFS_MOUNT_SECFLAVOUR flag. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 04:33:46 +08:00
* The flavor_len setting is for v4 mounts.
NFS: Refactor logic for parsing NFS security flavor mount options Clean up: Refactor the NFS mount option parsing function to extract the security flavor parsing logic into a separate function. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 04:33:38 +08:00
*/
static int nfs_parse_security_flavors(char *value,
struct nfs_parsed_mount_data *mnt)
{
substring_t args[MAX_OPT_ARGS];
dfprintk(MOUNT, "NFS: parsing sec=%s option\n", value);
switch (match_token(value, nfs_secflavor_tokens, args)) {
case Opt_sec_none:
mnt->auth_flavor_len = 0;
mnt->auth_flavors[0] = RPC_AUTH_NULL;
break;
case Opt_sec_sys:
mnt->auth_flavor_len = 0;
mnt->auth_flavors[0] = RPC_AUTH_UNIX;
break;
case Opt_sec_krb5:
mnt->auth_flavor_len = 1;
mnt->auth_flavors[0] = RPC_AUTH_GSS_KRB5;
break;
case Opt_sec_krb5i:
mnt->auth_flavor_len = 1;
mnt->auth_flavors[0] = RPC_AUTH_GSS_KRB5I;
break;
case Opt_sec_krb5p:
mnt->auth_flavor_len = 1;
mnt->auth_flavors[0] = RPC_AUTH_GSS_KRB5P;
break;
case Opt_sec_lkey:
mnt->auth_flavor_len = 1;
mnt->auth_flavors[0] = RPC_AUTH_GSS_LKEY;
break;
case Opt_sec_lkeyi:
mnt->auth_flavor_len = 1;
mnt->auth_flavors[0] = RPC_AUTH_GSS_LKEYI;
break;
case Opt_sec_lkeyp:
mnt->auth_flavor_len = 1;
mnt->auth_flavors[0] = RPC_AUTH_GSS_LKEYP;
break;
case Opt_sec_spkm:
mnt->auth_flavor_len = 1;
mnt->auth_flavors[0] = RPC_AUTH_GSS_SPKM;
break;
case Opt_sec_spkmi:
mnt->auth_flavor_len = 1;
mnt->auth_flavors[0] = RPC_AUTH_GSS_SPKMI;
break;
case Opt_sec_spkmp:
mnt->auth_flavor_len = 1;
mnt->auth_flavors[0] = RPC_AUTH_GSS_SPKMP;
break;
default:
return 0;
}
return 1;
}
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
/*
* Error-check and convert a string of mount options from user space into
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
* a data structure. The whole mount string is processed; bad options are
* skipped as they are encountered. If there were no errors, return 1;
* otherwise return 0 (zero).
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
*/
static int nfs_parse_mount_options(char *raw,
struct nfs_parsed_mount_data *mnt)
{
NFS: use new LSM interfaces to explicitly set mount options NFS and SELinux worked together previously because SELinux had NFS specific knowledge built in. This design was approved by both groups back in 2004 but the recent NFS changes to use nfs_parsed_mount_data and the usage of nfs_clone_mount_data showed this to be a poor fragile solution. This patch fixes the NFS functionality regression by making use of the new LSM interfaces to allow an FS to explicitly set its own mount options. The explicit setting of mount options is done in the nfs get_sb functions which are called before the generic vfs hooks try to set mount options for filesystems which use text mount data. This does not currently support NFSv4 as that functionality did not exist in previous kernels and thus there is no regression. I will be adding the needed code, which I believe to be the exact same as the v3 code, in nfs4_get_sb for 2.6.26. Signed-off-by: Eric Paris <eparis@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: James Morris <jmorris@namei.org>
2008-03-06 03:20:18 +08:00
char *p, *string, *secdata;
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
int rc, sloppy = 0, invalid_option = 0;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
if (!raw) {
dfprintk(MOUNT, "NFS: mount options string was NULL.\n");
return 1;
}
dfprintk(MOUNT, "NFS: nfs mount opts='%s'\n", raw);
NFS: use new LSM interfaces to explicitly set mount options NFS and SELinux worked together previously because SELinux had NFS specific knowledge built in. This design was approved by both groups back in 2004 but the recent NFS changes to use nfs_parsed_mount_data and the usage of nfs_clone_mount_data showed this to be a poor fragile solution. This patch fixes the NFS functionality regression by making use of the new LSM interfaces to allow an FS to explicitly set its own mount options. The explicit setting of mount options is done in the nfs get_sb functions which are called before the generic vfs hooks try to set mount options for filesystems which use text mount data. This does not currently support NFSv4 as that functionality did not exist in previous kernels and thus there is no regression. I will be adding the needed code, which I believe to be the exact same as the v3 code, in nfs4_get_sb for 2.6.26. Signed-off-by: Eric Paris <eparis@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: James Morris <jmorris@namei.org>
2008-03-06 03:20:18 +08:00
secdata = alloc_secdata();
if (!secdata)
goto out_nomem;
rc = security_sb_copy_data(raw, secdata);
if (rc)
goto out_security_failure;
rc = security_sb_parse_opts_str(secdata, &mnt->lsm_opts);
if (rc)
goto out_security_failure;
free_secdata(secdata);
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
while ((p = strsep(&raw, ",")) != NULL) {
substring_t args[MAX_OPT_ARGS];
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
unsigned long option;
Merge branch 'devel-for-2.6.31' into for-2.6.31 Conflicts: fs/nfs/client.c fs/nfs/super.c
2009-06-18 09:13:00 +08:00
int int_option;
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
int token;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
if (!*p)
continue;
dfprintk(MOUNT, "NFS: parsing nfs mount option '%s'\n", p);
token = match_token(p, nfs_mount_option_tokens, args);
switch (token) {
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
/*
* boolean options: foo/nofoo
*/
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
case Opt_soft:
mnt->flags |= NFS_MOUNT_SOFT;
break;
case Opt_hard:
mnt->flags &= ~NFS_MOUNT_SOFT;
break;
case Opt_posix:
mnt->flags |= NFS_MOUNT_POSIX;
break;
case Opt_noposix:
mnt->flags &= ~NFS_MOUNT_POSIX;
break;
case Opt_cto:
mnt->flags &= ~NFS_MOUNT_NOCTO;
break;
case Opt_nocto:
mnt->flags |= NFS_MOUNT_NOCTO;
break;
case Opt_ac:
mnt->flags &= ~NFS_MOUNT_NOAC;
break;
case Opt_noac:
mnt->flags |= NFS_MOUNT_NOAC;
break;
case Opt_lock:
mnt->flags &= ~NFS_MOUNT_NONLM;
break;
case Opt_nolock:
mnt->flags |= NFS_MOUNT_NONLM;
break;
case Opt_v2:
mnt->flags &= ~NFS_MOUNT_VER3;
break;
case Opt_v3:
mnt->flags |= NFS_MOUNT_VER3;
break;
case Opt_udp:
mnt->flags &= ~NFS_MOUNT_TCP;
NFS/SUNRPC: use transport protocol naming Instead of an { address family, raw IP protocol number }-tuple, use the newly-defined RPC identifier when creating clients in the upper layers. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:48:23 +08:00
mnt->nfs_server.protocol = XPRT_TRANSPORT_UDP;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_tcp:
mnt->flags |= NFS_MOUNT_TCP;
NFS/SUNRPC: use transport protocol naming Instead of an { address family, raw IP protocol number }-tuple, use the newly-defined RPC identifier when creating clients in the upper layers. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:48:23 +08:00
mnt->nfs_server.protocol = XPRT_TRANSPORT_TCP;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
NFS: support RDMA mounts Adds hooks to the string-based NFS mount to support an "rdma" protocol option. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:49:41 +08:00
case Opt_rdma:
mnt->flags |= NFS_MOUNT_TCP; /* for side protocols */
mnt->nfs_server.protocol = XPRT_TRANSPORT_RDMA;
NFS: load the rpc/rdma transport module automatically When mounting an NFS/RDMA server with the "-o proto=rdma" or "-o rdma" options, attempt to dynamically load the necessary "xprtrdma" client transport module. Doing so improves usability, while avoiding a static module dependency and any unnecesary resources. Signed-off-by: Tom Talpey <tmtalpey@gmail.com> Cc: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-03-12 02:37:56 +08:00
xprt_load_transport(p);
NFS: support RDMA mounts Adds hooks to the string-based NFS mount to support an "rdma" protocol option. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:49:41 +08:00
break;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
case Opt_acl:
mnt->flags &= ~NFS_MOUNT_NOACL;
break;
case Opt_noacl:
mnt->flags |= NFS_MOUNT_NOACL;
break;
case Opt_rdirplus:
mnt->flags &= ~NFS_MOUNT_NORDIRPLUS;
break;
case Opt_nordirplus:
mnt->flags |= NFS_MOUNT_NORDIRPLUS;
break;
NFS: Add the mount option "nosharecache" Prior to David Howell's mount changes in 2.6.18, users who mounted different directories which happened to be from the same filesystem on the server would get different super blocks, and hence could choose different mount options. As long as there were no hard linked files that crossed from one subtree to another, this was quite safe. Post the changes, if the two directories are on the same filesystem (have the same 'fsid'), they will share the same super block, and hence the same mount options. Add a flag to allow users to elect not to share the NFS super block with another mount point, even if the fsids are the same. This will allow users to set different mount options for the two different super blocks, as was previously possible. It is still up to the user to ensure that there are no cache coherency issues when doing this, however the default behaviour will be to share super blocks whenever two paths result in the same fsid. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
case Opt_sharecache:
mnt->flags &= ~NFS_MOUNT_UNSHARED;
break;
case Opt_nosharecache:
mnt->flags |= NFS_MOUNT_UNSHARED;
break;
NFS: add "[no]resvport" mount option The standard default security setting for NFS is AUTH_SYS. An NFS client connects to NFS servers via a privileged source port and a fixed standard destination port (2049). The client sends raw uid and gid numbers to identify users making NFS requests, and the server assumes an appropriate authority on the client has vetted these values because the source port is privileged. On Linux, by default in-kernel RPC services use a privileged port in the range between 650 and 1023 to avoid using source ports of well- known IP services. Using such a small range limits the number of NFS mount points and the number of unique NFS servers to which a client can connect concurrently. An NFS client can use unprivileged source ports to expand the range of source port numbers, allowing more concurrent server connections and more NFS mount points. Servers must explicitly allow NFS connections from unprivileged ports for this to work. In the past, bumping the value of the sunrpc.max_resvport sysctl on the client would permit the NFS client to use unprivileged ports. Bumping this setting also changes the maximum port number used by other in-kernel RPC services, some of which still required a port number less than 1023. This is exacerbated by the way source port numbers are chosen by the Linux RPC client, which starts at the top of the range and works downwards. It means that bumping the maximum means all RPC services requesting a source port will likely get an unprivileged port instead of a privileged one. Changing this setting effects all NFS mount points on a client. A sysadmin could not selectively choose which mount points would use non-privileged ports and which could not. Lastly, this mechanism of expanding the limit on the number of NFS mount points was entirely undocumented. To address the need for the NFS client to use a large range of source ports without interfering with the activity of other in-kernel RPC services, we introduce a new NFS mount option. This option explicitly tells only the NFS client to use a non-privileged source port when communicating with the NFS server for one specific mount point. This new mount option is called "resvport," like the similar NFS mount option on FreeBSD and Mac OS X. A sister patch for nfs-utils will be submitted that documents this new option in nfs(5). The default setting for this new mount option requires the NFS client to use a privileged port, as before. Explicitly specifying the "noresvport" mount option allows the NFS client to use an unprivileged source port for this mount point when connecting to the NFS server port. This mount option is supported only for text-based NFS mounts. [ Sidebar: it is widely known that security mechanisms based on the use of privileged source ports are ineffective. However, the NFS client can combine the use of unprivileged ports with the use of secure authentication mechanisms, such as Kerberos. This allows a large number of connections and mount points while ensuring a useful level of security. Eventually we may change the default setting for this option depending on the security flavor used for the mount. For example, if the mount is using only AUTH_SYS, then the default setting will be "resvport;" if the mount is using a strong security flavor such as krb5, the default setting will be "noresvport." ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond.Myklebust@netapp.com: Fixed a bug whereby nfs4_init_client() was being called with incorrect arguments.] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:37 +08:00
case Opt_resvport:
mnt->flags &= ~NFS_MOUNT_NORESVPORT;
break;
case Opt_noresvport:
mnt->flags |= NFS_MOUNT_NORESVPORT;
break;
NFS: Add mount options to enable local caching on NFS Add NFS mount options to allow the local caching support to be enabled. The attached patch makes it possible for the NFS filesystem to be told to make use of the network filesystem local caching service (FS-Cache). To be able to use this, a recent nfsutils package is required. There are three variant NFS mount options that can be added to a mount command to control caching for a mount. Only the last one specified takes effect: (*) Adding "fsc" will request caching. (*) Adding "fsc=<string>" will request caching and also specify a uniquifier. (*) Adding "nofsc" will disable caching. For example: mount warthog:/ /a -o fsc The cache of a particular superblock (NFS FSID) will be shared between all mounts of that volume, provided they have the same connection parameters and are not marked 'nosharecache'. Where it is otherwise impossible to distinguish superblocks because all the parameters are identical, but the 'nosharecache' option is supplied, a uniquifying string must be supplied, else only the first mount will be permitted to use the cache. If there's a key collision, then the second mount will disable caching and give a warning into the kernel log. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Steve Dickson <steved@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Daire Byrne <Daire.Byrne@framestore.com>
2009-04-03 23:42:48 +08:00
case Opt_fscache:
mnt->options |= NFS_OPTION_FSCACHE;
kfree(mnt->fscache_uniq);
mnt->fscache_uniq = NULL;
break;
case Opt_nofscache:
mnt->options &= ~NFS_OPTION_FSCACHE;
kfree(mnt->fscache_uniq);
mnt->fscache_uniq = NULL;
break;
case Opt_fscache_uniq:
string = match_strdup(args);
if (!string)
goto out_nomem;
kfree(mnt->fscache_uniq);
mnt->fscache_uniq = string;
mnt->options |= NFS_OPTION_FSCACHE;
break;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
/*
* options that take numeric values
*/
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
case Opt_port:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0 || option > USHORT_MAX)
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
mnt->nfs_server.port = option;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_rsize:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0)
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
mnt->rsize = option;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_wsize:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0)
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
mnt->wsize = option;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_bsize:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0)
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
mnt->bsize = option;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_timeo:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0 || option == 0)
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
mnt->timeo = option;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_retrans:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0 || option == 0)
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
mnt->retrans = option;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_acregmin:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0)
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
mnt->acregmin = option;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_acregmax:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0)
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
mnt->acregmax = option;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_acdirmin:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0)
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
mnt->acdirmin = option;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_acdirmax:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0)
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
mnt->acdirmax = option;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_actimeo:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0)
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
mnt->acregmin = mnt->acregmax =
mnt->acdirmin = mnt->acdirmax = option;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_namelen:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0)
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
mnt->namlen = option;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_mountport:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0 || option > USHORT_MAX)
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
mnt->mount_server.port = option;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_mountvers:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0 ||
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
option < NFS_MNT_VERSION ||
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
option > NFS_MNT3_VERSION)
goto out_invalid_value;
mnt->mount_server.version = option;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_nfsvers:
NFS: More "sloppy" parsing problems Specifying "port=-5" with the kernel's current mount option parser generates "unrecognized mount option". If "sloppy" is set, this causes the mount to succeed and use the default values; the desired behavior is that, since this is a valid option with an invalid value, the mount should fail, even with "sloppy." To properly handle "sloppy" parsing, we need to distinguish between correct options with invalid values, and incorrect options. We will need to parse integer values by hand, therefore, and not rely on match_token(). For instance, these must all fail with "invalid value": port=12345678 port=-5 port=samuel and not with "unrecognized option," as they do currently. Thus, for the sake of match_token() we need to treat the values for these options as strings, and do the conversion to integers using strict_strtol(). This is basically the same solution we used for the earlier "retry=" fix (commit ecbb3845), except in this case the kernel actually has to parse the value, rather than ignore it. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:14 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
rc = strict_strtoul(string, 10, &option);
kfree(string);
if (rc != 0)
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
switch (option) {
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
case NFS2_VERSION:
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
mnt->flags &= ~NFS_MOUNT_VER3;
break;
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
case NFS3_VERSION:
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
mnt->flags |= NFS_MOUNT_VER3;
break;
default:
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
goto out_invalid_value;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
}
break;
nfs41: add mount command option minorversion mount -t nfs4 -o minorversion=[0|1] specifies whether to use 4.0 or 4.1. By default, the minorversion is set to 0. Signed-off-by: Mike Sager <sager@netapp.com> [set default minorversion to 0 as per Trond and SteveD's request] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-04-01 21:21:48 +08:00
case Opt_minorversion:
Merge branch 'devel-for-2.6.31' into for-2.6.31 Conflicts: fs/nfs/client.c fs/nfs/super.c
2009-06-18 09:13:00 +08:00
if (match_int(args, &int_option))
nfs41: add mount command option minorversion mount -t nfs4 -o minorversion=[0|1] specifies whether to use 4.0 or 4.1. By default, the minorversion is set to 0. Signed-off-by: Mike Sager <sager@netapp.com> [set default minorversion to 0 as per Trond and SteveD's request] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-04-01 21:21:48 +08:00
return 0;
Merge branch 'devel-for-2.6.31' into for-2.6.31 Conflicts: fs/nfs/client.c fs/nfs/super.c
2009-06-18 09:13:00 +08:00
if (int_option < 0 || int_option > NFS4_MAX_MINOR_VERSION)
nfs41: add mount command option minorversion mount -t nfs4 -o minorversion=[0|1] specifies whether to use 4.0 or 4.1. By default, the minorversion is set to 0. Signed-off-by: Mike Sager <sager@netapp.com> [set default minorversion to 0 as per Trond and SteveD's request] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-04-01 21:21:48 +08:00
return 0;
Merge branch 'devel-for-2.6.31' into for-2.6.31 Conflicts: fs/nfs/client.c fs/nfs/super.c
2009-06-18 09:13:00 +08:00
mnt->minorversion = int_option;
nfs41: add mount command option minorversion mount -t nfs4 -o minorversion=[0|1] specifies whether to use 4.0 or 4.1. By default, the minorversion is set to 0. Signed-off-by: Mike Sager <sager@netapp.com> [set default minorversion to 0 as per Trond and SteveD's request] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-04-01 21:21:48 +08:00
break;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
/*
* options that take text values
*/
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
case Opt_sec:
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
NFS: Refactor logic for parsing NFS security flavor mount options Clean up: Refactor the NFS mount option parsing function to extract the security flavor parsing logic into a separate function. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 04:33:38 +08:00
rc = nfs_parse_security_flavors(string, mnt);
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
kfree(string);
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
if (!rc) {
dfprintk(MOUNT, "NFS: unrecognized "
"security flavor\n");
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
return 0;
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
}
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
case Opt_proto:
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
token = match_token(string,
nfs_xprt_protocol_tokens, args);
switch (token) {
NFS: mount option parser chokes on proto= The new text-based NFS mount option parsing logic doesn't recognize any valid transport protocols due to a silly mistake in the protocol token matching logic. This prevents basic mount requests such as: mount.nfs server:/export /mnt -o proto=tcp from working with the new text-based NFS mount API. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-08-30 05:58:57 +08:00
case Opt_xprt_udp:
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
mnt->flags &= ~NFS_MOUNT_TCP;
NFS/SUNRPC: use transport protocol naming Instead of an { address family, raw IP protocol number }-tuple, use the newly-defined RPC identifier when creating clients in the upper layers. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:48:23 +08:00
mnt->nfs_server.protocol = XPRT_TRANSPORT_UDP;
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
kfree(string);
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
NFS: mount option parser chokes on proto= The new text-based NFS mount option parsing logic doesn't recognize any valid transport protocols due to a silly mistake in the protocol token matching logic. This prevents basic mount requests such as: mount.nfs server:/export /mnt -o proto=tcp from working with the new text-based NFS mount API. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-08-30 05:58:57 +08:00
case Opt_xprt_tcp:
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
mnt->flags |= NFS_MOUNT_TCP;
NFS/SUNRPC: use transport protocol naming Instead of an { address family, raw IP protocol number }-tuple, use the newly-defined RPC identifier when creating clients in the upper layers. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:48:23 +08:00
mnt->nfs_server.protocol = XPRT_TRANSPORT_TCP;
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
kfree(string);
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
NFS: support RDMA mounts Adds hooks to the string-based NFS mount to support an "rdma" protocol option. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:49:41 +08:00
case Opt_xprt_rdma:
/* vector side protocols to TCP */
mnt->flags |= NFS_MOUNT_TCP;
mnt->nfs_server.protocol = XPRT_TRANSPORT_RDMA;
NFS: load the rpc/rdma transport module automatically When mounting an NFS/RDMA server with the "-o proto=rdma" or "-o rdma" options, attempt to dynamically load the necessary "xprtrdma" client transport module. Doing so improves usability, while avoiding a static module dependency and any unnecesary resources. Signed-off-by: Tom Talpey <tmtalpey@gmail.com> Cc: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-03-12 02:37:56 +08:00
xprt_load_transport(string);
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
kfree(string);
NFS: support RDMA mounts Adds hooks to the string-based NFS mount to support an "rdma" protocol option. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:49:41 +08:00
break;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
default:
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
dfprintk(MOUNT, "NFS: unrecognized "
"transport protocol\n");
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
return 0;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
}
break;
case Opt_mountproto:
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
token = match_token(string,
nfs_xprt_protocol_tokens, args);
NFS: Fix a double free in nfs_parse_mount_options() Due to an apparent typo, commit a67d18f89f5782806135aad4ee012ff78d45aae7 (NFS: load the rpc/rdma transport module automatically) lead to the 'proto=' mount option doing a double free, while Opt_mountproto leaks a string. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-04-07 07:41:35 +08:00
kfree(string);
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
switch (token) {
NFS: mount option parser chokes on proto= The new text-based NFS mount option parsing logic doesn't recognize any valid transport protocols due to a silly mistake in the protocol token matching logic. This prevents basic mount requests such as: mount.nfs server:/export /mnt -o proto=tcp from working with the new text-based NFS mount API. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-08-30 05:58:57 +08:00
case Opt_xprt_udp:
NFS/SUNRPC: use transport protocol naming Instead of an { address family, raw IP protocol number }-tuple, use the newly-defined RPC identifier when creating clients in the upper layers. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:48:23 +08:00
mnt->mount_server.protocol = XPRT_TRANSPORT_UDP;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
NFS: mount option parser chokes on proto= The new text-based NFS mount option parsing logic doesn't recognize any valid transport protocols due to a silly mistake in the protocol token matching logic. This prevents basic mount requests such as: mount.nfs server:/export /mnt -o proto=tcp from working with the new text-based NFS mount API. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-08-30 05:58:57 +08:00
case Opt_xprt_tcp:
NFS/SUNRPC: use transport protocol naming Instead of an { address family, raw IP protocol number }-tuple, use the newly-defined RPC identifier when creating clients in the upper layers. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:48:23 +08:00
mnt->mount_server.protocol = XPRT_TRANSPORT_TCP;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
NFS: support RDMA mounts Adds hooks to the string-based NFS mount to support an "rdma" protocol option. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:49:41 +08:00
case Opt_xprt_rdma: /* not used for side protocols */
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
default:
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
dfprintk(MOUNT, "NFS: unrecognized "
"transport protocol\n");
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
return 0;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
}
break;
case Opt_addr:
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
NFS: Add string length argument to nfs_parse_server_address To make nfs_parse_server_address() more generally useful, allow it to accept input strings that are not terminated with '\0'. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:53 +08:00
nfs_parse_ip_address(string, strlen(string),
(struct sockaddr *)
&mnt->nfs_server.address,
&mnt->nfs_server.addrlen);
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
kfree(string);
break;
case Opt_clientaddr:
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
NFS: Address memory leaks in the NFS client mount option parser David Howells noticed that repeating the same mount option twice during an NFS mount request can result in orphaned memory in certain cases. Only the client_address and mount_server.hostname strings are initialized in the mount parsing loop, so those appear to be the only two pointers that might be written over by repeating a mount option. The strings in the nfs_server section of the nfs_parsed_mount_data structure are set only once after the options are parsed, thus these are not susceptible to being overwritten. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-01-17 05:38:10 +08:00
kfree(mnt->client_address);
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
mnt->client_address = string;
break;
NFS: Remove the NIPQUAD from nfs_try_mount In the name of address family compatibility, we can't have the NIP_FMT and NIPQUAD macros in nfs_try_mount(). Instead, we can make use of an unused mount option to display the mount server's hostname. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:13 +08:00
case Opt_mounthost:
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
NFS: Address memory leaks in the NFS client mount option parser David Howells noticed that repeating the same mount option twice during an NFS mount request can result in orphaned memory in certain cases. Only the client_address and mount_server.hostname strings are initialized in the mount parsing loop, so those appear to be the only two pointers that might be written over by repeating a mount option. The strings in the nfs_server section of the nfs_parsed_mount_data structure are set only once after the options are parsed, thus these are not susceptible to being overwritten. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-01-17 05:38:10 +08:00
kfree(mnt->mount_server.hostname);
NFS: Remove the NIPQUAD from nfs_try_mount In the name of address family compatibility, we can't have the NIP_FMT and NIPQUAD macros in nfs_try_mount(). Instead, we can make use of an unused mount option to display the mount server's hostname. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:13 +08:00
mnt->mount_server.hostname = string;
break;
NFS: Add new 'mountaddr=' mount option I got the 'mounthost=' option wrong - it shouldn't look for an address value, but rather a hostname value. However, the in-kernel mount client and NFS client cannot resolve a hostname by themselves; they rely on user-land to pass in the resolved address. Create a new mount option that does take an address so that the mount program's address can be passed in. The mount hostname is now ignored by the kernel. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-12 06:01:04 +08:00
case Opt_mountaddr:
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
NFS: Add string length argument to nfs_parse_server_address To make nfs_parse_server_address() more generally useful, allow it to accept input strings that are not terminated with '\0'. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:53 +08:00
nfs_parse_ip_address(string, strlen(string),
(struct sockaddr *)
&mnt->mount_server.address,
&mnt->mount_server.addrlen);
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
kfree(string);
break;
NFS: Add mount options for controlling the lookup cache Add the following NFS-specific mount options to the parser. -o lookupcache=all /* Default: cache positive & negative dentries */ -o lookupcache=pos[itive] /* Don't cache negative dentries */ -o lookupcache=none /* Strict revalidation of all dentries */ Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-16 05:58:14 +08:00
case Opt_lookupcache:
string = match_strdup(args);
if (string == NULL)
goto out_nomem;
token = match_token(string,
nfs_lookupcache_tokens, args);
kfree(string);
switch (token) {
case Opt_lookupcache_all:
mnt->flags &= ~(NFS_MOUNT_LOOKUP_CACHE_NONEG|NFS_MOUNT_LOOKUP_CACHE_NONE);
break;
case Opt_lookupcache_positive:
mnt->flags &= ~NFS_MOUNT_LOOKUP_CACHE_NONE;
mnt->flags |= NFS_MOUNT_LOOKUP_CACHE_NONEG;
break;
case Opt_lookupcache_none:
mnt->flags |= NFS_MOUNT_LOOKUP_CACHE_NONEG|NFS_MOUNT_LOOKUP_CACHE_NONE;
break;
default:
dfprintk(MOUNT, "NFS: invalid "
"lookupcache argument\n");
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
return 0;
NFS: Add mount options for controlling the lookup cache Add the following NFS-specific mount options to the parser. -o lookupcache=all /* Default: cache positive & negative dentries */ -o lookupcache=pos[itive] /* Don't cache negative dentries */ -o lookupcache=none /* Strict revalidation of all dentries */ Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-16 05:58:14 +08:00
};
break;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
/*
* Special options
*/
case Opt_sloppy:
sloppy = 1;
dfprintk(MOUNT, "NFS: relaxing parsing rules\n");
break;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
case Opt_userspace:
case Opt_deprecated:
NFS: Treat "intr" and "nointr" options as deprecated Clean up: the "intr" and "nointr" mount options were recently retired. Document this in the NFS mount option parser. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-13 00:37:41 +08:00
dfprintk(MOUNT, "NFS: ignoring mount option "
"'%s'\n", p);
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
break;
default:
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
invalid_option = 1;
NFS: Allow either strict or sloppy mount option parsing The kernel's NFS client mount option parser currently doesn't allow unrecognized or incorrect mount options. This prevents misspellings or incorrectly specified mount options from possibly causing silent data corruption. However, NFS mount options are not standardized, so different operating systems can use differently spelled mount options to support similar features, or can support mount options which no other operating system supports. "Sloppy" mount option parsing, which allows the parser to ignore any option it doesn't recognize, is needed to support automounters that often use maps that are shared between heterogenous operating systems. The legacy mount command ignores the validity of the values of mount options entirely, except for the "sec=" and "proto=" options. If an incorrect value is specified, the out-of-range value is passed to the kernel; if a value is specified that contains non-numeric characters, it appears as though the legacy mount command sets that option to zero (probably incorrect behavior in general). In any case, this sets a precedent which we will partially follow for the kernel mount option parser: + if "sloppy" is not set, the parser will be strict about both unrecognized options (same as legacy) and invalid option values (stricter than legacy) + if "sloppy" is set, the parser will ignore unrecognized options and invalid option values (same as legacy) An "invalid" option value in this case means that either the type (integer, short, or string) or sign (for integer values) of the specified value is incorrect. This patch does two things: it changes the NFS client's mount option parsing loop so that it parses the whole string instead of failing at the first unrecognized option or invalid option value. An unrecognized option or an invalid option value cause the option to be skipped. Then, the patch adds a "sloppy" mount option that allows the parsing to succeed anyway if there were any problems during parsing. When parsing a set of options is complete, if there are errors and "sloppy" was specified, return success anyway. Otherwise, only return success if there are no errors. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 07:28:02 +08:00
dfprintk(MOUNT, "NFS: unrecognized mount option "
"'%s'\n", p);
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
}
}
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
if (!sloppy && invalid_option)
return 0;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
return 1;
NFS: Invalid mount option values should always fail, even with "sloppy" Ian Kent reports: "I've noticed a couple of other regressions with the options vers and proto option of mount.nfs(8). The commands: mount -t nfs -o vers=<invalid version> <server>:/<path> /<mountpoint> mount -t nfs -o proto=<invalid proto> <server>:/<path> /<mountpoint> both immediately fail. But if the "-s" option is also used they both succeed with the mount falling back to defaults (by the look of it). In the past these failed even when the sloppy option was given, as I think they should. I believe the sloppy option is meant to allow the mount command to still function for mount options (for example in shared autofs maps) that exist on other Unix implementations but aren't present in the Linux mount.nfs(8). So, an invalid value specified for a known mount option is different to an unknown mount option and should fail appropriately." See RH bugzilla 486266. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
out_invalid_value:
printk(KERN_INFO "NFS: bad mount option value specified: %s \n", p);
return 0;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
out_nomem:
printk(KERN_INFO "NFS: not enough memory to parse option\n");
return 0;
NFS: use new LSM interfaces to explicitly set mount options NFS and SELinux worked together previously because SELinux had NFS specific knowledge built in. This design was approved by both groups back in 2004 but the recent NFS changes to use nfs_parsed_mount_data and the usage of nfs_clone_mount_data showed this to be a poor fragile solution. This patch fixes the NFS functionality regression by making use of the new LSM interfaces to allow an FS to explicitly set its own mount options. The explicit setting of mount options is done in the nfs get_sb functions which are called before the generic vfs hooks try to set mount options for filesystems which use text mount data. This does not currently support NFSv4 as that functionality did not exist in previous kernels and thus there is no regression. I will be adding the needed code, which I believe to be the exact same as the v3 code, in nfs4_get_sb for 2.6.26. Signed-off-by: Eric Paris <eparis@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: James Morris <jmorris@namei.org>
2008-03-06 03:20:18 +08:00
out_security_failure:
free_secdata(secdata);
printk(KERN_INFO "NFS: security options invalid: %d\n", rc);
return 0;
NFS: Add enums and match tables for mount option parsing This generic infrastructure works for both NFS and NFSv4 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:44 +08:00
}
NFS: Introduce generic mount client API For NFSv2 and v3 mounts, the first step is to contact the server's MOUNTD and request the file handle for the root of the mounted share. Add a function to the NFS client that handles this operation. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:49 +08:00
/*
* Use the remote server's MOUNT service to request the NFS file handle
* corresponding to the provided path.
*/
static int nfs_try_mount(struct nfs_parsed_mount_data *args,
struct nfs_fh *root_fh)
{
NFS: Update MNT and MNT3 reply decoding functions Solder xdr_stream-based XDR decoding functions into the in-kernel mountd client that are more careful about checking data types and watching for buffer overflows. The new MNT3 decoder includes support for auth-flavor list decoding. The "_sz" macro for MNT3 replies was missing the size of the file handle. I've added this back, and included the size of the auth flavor array. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
unsigned int auth_flavor_len = 0;
NFS: introduce nfs_mount_info struct for calling nfs_mount() Clean up: convert nfs_mount() to take a single data structure argument to make it simpler to add more arguments. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:35 +08:00
struct nfs_mount_request request = {
.sap = (struct sockaddr *)
&args->mount_server.address,
.dirpath = args->nfs_server.export_path,
.protocol = args->mount_server.protocol,
.fh = root_fh,
NFS: "[no]resvport" mount option changes mountd client too If the admin has specified the "noresvport" option for an NFS mount point, the kernel's NFS client uses an unprivileged source port for the main NFS transport. The kernel's mountd client should use an unprivileged port in this case as well. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:37 +08:00
.noresvport = args->flags & NFS_MOUNT_NORESVPORT,
NFS: Update MNT and MNT3 reply decoding functions Solder xdr_stream-based XDR decoding functions into the in-kernel mountd client that are more careful about checking data types and watching for buffer overflows. The new MNT3 decoder includes support for auth-flavor list decoding. The "_sz" macro for MNT3 replies was missing the size of the file handle. I've added this back, and included the size of the auth flavor array. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 09:02:13 +08:00
.auth_flav_len = &auth_flavor_len,
NFS: introduce nfs_mount_info struct for calling nfs_mount() Clean up: convert nfs_mount() to take a single data structure argument to make it simpler to add more arguments. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:35 +08:00
};
NFS: Support non-IPv4 addresses in nfs_parsed_mount_data Replace the nfs_server and mount_server address fields in the nfs_parsed_mount_data structure with a "struct sockaddr_storage" instead of a "struct sockaddr_in". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:28 +08:00
int status;
NFS: Introduce generic mount client API For NFSv2 and v3 mounts, the first step is to contact the server's MOUNTD and request the file handle for the root of the mounted share. Add a function to the NFS client that handles this operation. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:49 +08:00
if (args->mount_server.version == 0) {
if (args->flags & NFS_MOUNT_VER3)
args->mount_server.version = NFS_MNT3_VERSION;
else
args->mount_server.version = NFS_MNT_VERSION;
}
NFS: introduce nfs_mount_info struct for calling nfs_mount() Clean up: convert nfs_mount() to take a single data structure argument to make it simpler to add more arguments. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:35 +08:00
request.version = args->mount_server.version;
NFS: Introduce generic mount client API For NFSv2 and v3 mounts, the first step is to contact the server's MOUNTD and request the file handle for the root of the mounted share. Add a function to the NFS client that handles this operation. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:49 +08:00
NFS: Remove the NIPQUAD from nfs_try_mount In the name of address family compatibility, we can't have the NIP_FMT and NIPQUAD macros in nfs_try_mount(). Instead, we can make use of an unused mount option to display the mount server's hostname. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:13 +08:00
if (args->mount_server.hostname)
NFS: introduce nfs_mount_info struct for calling nfs_mount() Clean up: convert nfs_mount() to take a single data structure argument to make it simpler to add more arguments. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:35 +08:00
request.hostname = args->mount_server.hostname;
NFS: Remove the NIPQUAD from nfs_try_mount In the name of address family compatibility, we can't have the NIP_FMT and NIPQUAD macros in nfs_try_mount(). Instead, we can make use of an unused mount option to display the mount server's hostname. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:13 +08:00
else
NFS: introduce nfs_mount_info struct for calling nfs_mount() Clean up: convert nfs_mount() to take a single data structure argument to make it simpler to add more arguments. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:35 +08:00
request.hostname = args->nfs_server.hostname;
NFS: Remove the NIPQUAD from nfs_try_mount In the name of address family compatibility, we can't have the NIP_FMT and NIPQUAD macros in nfs_try_mount(). Instead, we can make use of an unused mount option to display the mount server's hostname. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:13 +08:00
NFS: Introduce generic mount client API For NFSv2 and v3 mounts, the first step is to contact the server's MOUNTD and request the file handle for the root of the mounted share. Add a function to the NFS client that handles this operation. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:49 +08:00
/*
* Construct the mount server's address.
*/
NFS: Support non-IPv4 addresses in nfs_parsed_mount_data Replace the nfs_server and mount_server address fields in the nfs_parsed_mount_data structure with a "struct sockaddr_storage" instead of a "struct sockaddr_in". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:28 +08:00
if (args->mount_server.address.ss_family == AF_UNSPEC) {
NFS: introduce nfs_mount_info struct for calling nfs_mount() Clean up: convert nfs_mount() to take a single data structure argument to make it simpler to add more arguments. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:35 +08:00
memcpy(request.sap, &args->nfs_server.address,
NFS: Support non-IPv4 addresses in nfs_parsed_mount_data Replace the nfs_server and mount_server address fields in the nfs_parsed_mount_data structure with a "struct sockaddr_storage" instead of a "struct sockaddr_in". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:28 +08:00
args->nfs_server.addrlen);
args->mount_server.addrlen = args->nfs_server.addrlen;
}
NFS: introduce nfs_mount_info struct for calling nfs_mount() Clean up: convert nfs_mount() to take a single data structure argument to make it simpler to add more arguments. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:35 +08:00
request.salen = args->mount_server.addrlen;
NFS: Support non-IPv4 addresses in nfs_parsed_mount_data Replace the nfs_server and mount_server address fields in the nfs_parsed_mount_data structure with a "struct sockaddr_storage" instead of a "struct sockaddr_in". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:28 +08:00
[NFS] [PATCH] NFS: initialize default port in kernel mount client If no mount server port number is specified, the previous change to the kernel mount client inadvertently allows the NFS server's port number to be the used as the mount server's port number. If the user specifies an NFS server port (-o port=x), the mount will fail. The fix below sets the mount server's port to 0 if no mount server port is specified by the user. Signed-off-by: James Lentini <jlentini@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-25 05:32:49 +08:00
/*
* autobind will be used if mount_server.port == 0
*/
NFS: introduce nfs_mount_info struct for calling nfs_mount() Clean up: convert nfs_mount() to take a single data structure argument to make it simpler to add more arguments. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:35 +08:00
nfs_set_port(request.sap, args->mount_server.port);
NFS: Introduce generic mount client API For NFSv2 and v3 mounts, the first step is to contact the server's MOUNTD and request the file handle for the root of the mounted share. Add a function to the NFS client that handles this operation. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:49 +08:00
/*
* Now ask the mount server to map our export path
* to a file handle.
*/
NFS: introduce nfs_mount_info struct for calling nfs_mount() Clean up: convert nfs_mount() to take a single data structure argument to make it simpler to add more arguments. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:35 +08:00
status = nfs_mount(&request);
NFS: Kernel mount client should use async bind Simplify the in-kernel mount client by using autobind instead of an explicit call to rpc_getport_sync. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-12 06:00:58 +08:00
if (status == 0)
return 0;
NFS: Introduce generic mount client API For NFSv2 and v3 mounts, the first step is to contact the server's MOUNTD and request the file handle for the root of the mounted share. Add a function to the NFS client that handles this operation. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:49 +08:00
NFS: missing newline in NFS mount debugging message Clean up. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-13 00:37:49 +08:00
dfprintk(MOUNT, "NFS: unable to mount server %s, error %d\n",
NFS: introduce nfs_mount_info struct for calling nfs_mount() Clean up: convert nfs_mount() to take a single data structure argument to make it simpler to add more arguments. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:35 +08:00
request.hostname, status);
NFS: Introduce generic mount client API For NFSv2 and v3 mounts, the first step is to contact the server's MOUNTD and request the file handle for the root of the mounted share. Add a function to the NFS client that handles this operation. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:49 +08:00
return status;
}
NFS: Support raw IPv6 address hostnames during NFS mount operation Traditionally the mount command has looked for a ":" to separate the server's hostname from the export path in the mounted on device name, like this: mount server:/export /mounted/on/dir The server's hostname is "server" and the export path is "/export". You can also substitute a specific IPv4 network address for the server hostname, like this: mount 192.168.0.55:/export /mounted/on/dir Raw IPv6 addresses present a problem, however, because they look something like this: fe80::200:5aff:fe00:30b Note the use of colons. To get around the presence of colons, copy the Solaris convention used for mounting IPv6 servers by address: wrap a raw IPv6 address with square brackets. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:45 +08:00
static int nfs_parse_simple_hostname(const char *dev_name,
char **hostname, size_t maxnamlen,
char **export_path, size_t maxpathlen)
NFS: Use common device name parsing logic for NFSv4 and NFSv2/v3 To support passing a raw IPv6 address as a server hostname, we need to expand the logic that handles splitting the passed-in device name into a server hostname and export path Start by pulling device name parsing out of the mount option validation functions and into separate helper functions. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:37 +08:00
{
size_t len;
char *colon, *comma;
colon = strchr(dev_name, ':');
if (colon == NULL)
goto out_bad_devname;
len = colon - dev_name;
if (len > maxnamlen)
goto out_hostname;
/* N.B. caller will free nfs_server.hostname in all cases */
*hostname = kstrndup(dev_name, len, GFP_KERNEL);
if (!*hostname)
goto out_nomem;
/* kill possible hostname list: not supported */
comma = strchr(*hostname, ',');
if (comma != NULL) {
if (comma == *hostname)
goto out_bad_devname;
*comma = '\0';
}
colon++;
len = strlen(colon);
if (len > maxpathlen)
goto out_path;
*export_path = kstrndup(colon, len, GFP_KERNEL);
if (!*export_path)
goto out_nomem;
dfprintk(MOUNT, "NFS: MNTPATH: '%s'\n", *export_path);
return 0;
out_bad_devname:
dfprintk(MOUNT, "NFS: device name not in host:path format\n");
return -EINVAL;
out_nomem:
dfprintk(MOUNT, "NFS: not enough memory to parse device name\n");
return -ENOMEM;
out_hostname:
dfprintk(MOUNT, "NFS: server hostname too long\n");
return -ENAMETOOLONG;
out_path:
dfprintk(MOUNT, "NFS: export pathname too long\n");
return -ENAMETOOLONG;
}
NFS: Support raw IPv6 address hostnames during NFS mount operation Traditionally the mount command has looked for a ":" to separate the server's hostname from the export path in the mounted on device name, like this: mount server:/export /mounted/on/dir The server's hostname is "server" and the export path is "/export". You can also substitute a specific IPv4 network address for the server hostname, like this: mount 192.168.0.55:/export /mounted/on/dir Raw IPv6 addresses present a problem, however, because they look something like this: fe80::200:5aff:fe00:30b Note the use of colons. To get around the presence of colons, copy the Solaris convention used for mounting IPv6 servers by address: wrap a raw IPv6 address with square brackets. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:45 +08:00
/*
* Hostname has square brackets around it because it contains one or
* more colons. We look for the first closing square bracket, and a
* colon must follow it.
*/
static int nfs_parse_protected_hostname(const char *dev_name,
char **hostname, size_t maxnamlen,
char **export_path, size_t maxpathlen)
{
size_t len;
char *start, *end;
start = (char *)(dev_name + 1);
end = strchr(start, ']');
if (end == NULL)
goto out_bad_devname;
if (*(end + 1) != ':')
goto out_bad_devname;
len = end - start;
if (len > maxnamlen)
goto out_hostname;
/* N.B. caller will free nfs_server.hostname in all cases */
*hostname = kstrndup(start, len, GFP_KERNEL);
if (*hostname == NULL)
goto out_nomem;
end += 2;
len = strlen(end);
if (len > maxpathlen)
goto out_path;
*export_path = kstrndup(end, len, GFP_KERNEL);
if (!*export_path)
goto out_nomem;
return 0;
out_bad_devname:
dfprintk(MOUNT, "NFS: device name not in host:path format\n");
return -EINVAL;
out_nomem:
dfprintk(MOUNT, "NFS: not enough memory to parse device name\n");
return -ENOMEM;
out_hostname:
dfprintk(MOUNT, "NFS: server hostname too long\n");
return -ENAMETOOLONG;
out_path:
dfprintk(MOUNT, "NFS: export pathname too long\n");
return -ENAMETOOLONG;
}
/*
* Split "dev_name" into "hostname:export_path".
*
* The leftmost colon demarks the split between the server's hostname
* and the export path. If the hostname starts with a left square
* bracket, then it may contain colons.
*
* Note: caller frees hostname and export path, even on error.
*/
static int nfs_parse_devname(const char *dev_name,
char **hostname, size_t maxnamlen,
char **export_path, size_t maxpathlen)
{
if (*dev_name == '[')
return nfs_parse_protected_hostname(dev_name,
hostname, maxnamlen,
export_path, maxpathlen);
return nfs_parse_simple_hostname(dev_name,
hostname, maxnamlen,
export_path, maxpathlen);
}
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
/*
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
* Validate the NFS2/NFS3 mount data
* - fills in the mount root filehandle
NFS: Add final pieces to support in-kernel mount option parsing Hook in final components required for supporting in-kernel mount option parsing for NFSv2 and NFSv3 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:54 +08:00
*
* For option strings, user space handles the following behaviors:
*
* + DNS: mapping server host name to IP address ("addr=" option)
*
* + failure mode: how to behave if a mount request can't be handled
* immediately ("fg/bg" option)
*
* + retry: how often to retry a mount request ("retry=" option)
*
* + breaking back: trying proto=udp after proto=tcp, v2 after v3,
* mountproto=tcp after mountproto=udp, and so on
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
*/
NFS: use in-kernel mount argument structure for nfsv[23] mounts The user-visible nfs_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv[23] in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:43:56 +08:00
static int nfs_validate_mount_data(void *options,
struct nfs_parsed_mount_data *args,
NFS: Add final pieces to support in-kernel mount option parsing Hook in final components required for supporting in-kernel mount option parsing for NFSv2 and NFSv3 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:54 +08:00
struct nfs_fh *mntfh,
const char *dev_name)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{
NFS: use in-kernel mount argument structure for nfsv[23] mounts The user-visible nfs_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv[23] in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:43:56 +08:00
struct nfs_mount_data *data = (struct nfs_mount_data *)options;
NFS: Add final pieces to support in-kernel mount option parsing Hook in final components required for supporting in-kernel mount option parsing for NFSv2 and NFSv3 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:54 +08:00
NFS: Clean up nfs_validate_mount_data Move error handling code out of the main code path. The switch statement was also improperly indented, according to Documentation/CodingStyle. This prepares nfs_validate_mount_data for the addition of option string parsing. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:56 +08:00
if (data == NULL)
goto out_no_data;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: use in-kernel mount argument structure for nfsv[23] mounts The user-visible nfs_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv[23] in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:43:56 +08:00
args->flags = (NFS_MOUNT_VER3 | NFS_MOUNT_TCP);
args->rsize = NFS_MAX_FILE_IO_SIZE;
args->wsize = NFS_MAX_FILE_IO_SIZE;
NFS: use documenting macro constants for initializing ac{reg, dir}{min, max} Clean up. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-27 05:47:12 +08:00
args->acregmin = NFS_DEF_ACREGMIN;
args->acregmax = NFS_DEF_ACREGMAX;
args->acdirmin = NFS_DEF_ACDIRMIN;
args->acdirmax = NFS_DEF_ACDIRMAX;
NFS: Save the value of the "port=" mount option During a remount based on the mount options displayed in /proc/mounts, we want to preserve the original behavior of the mount request. Let's save the original setting of the "port=" mount option in the mount's nfs_server structure. This allows us to simplify the default behavior of port setting for NFSv4 mounts: by default, NFSv2/3 mounts first try an RPC bind to determine the NFS server's port, unless the user specified the "port=" mount option; Users can force the client to skip the RPC bind by explicitly specifying "port=<value>". NFSv4, by contrast, assumes the NFS server port is 2049 and skips the RPC bind, unless the user specifies "port=". Users can force an RPC bind for NFSv4 by explicitly specifying "port=0". I added a couple of extra comments to clarify this behavior. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Miklos Szeredi <miklos@szeredi.hu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:22 +08:00
args->mount_server.port = 0; /* autobind unless user sets port */
args->nfs_server.port = 0; /* autobind unless user sets port */
NFS/SUNRPC: use transport protocol naming Instead of an { address family, raw IP protocol number }-tuple, use the newly-defined RPC identifier when creating clients in the upper layers. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:48:23 +08:00
args->nfs_server.protocol = XPRT_TRANSPORT_TCP;
NFS: Set security flavor default for NFSv2/3 mounts like other defaults Set the default security flavor when we set the other mount option default values. After this change, only the legacy user-space mount path needs to set the NFS_MOUNT_SECFLAVOUR flag. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 04:33:46 +08:00
args->auth_flavors[0] = RPC_AUTH_UNIX;
NFS: use in-kernel mount argument structure for nfsv[23] mounts The user-visible nfs_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv[23] in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:43:56 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
switch (data->version) {
NFS: Clean up nfs_validate_mount_data Move error handling code out of the main code path. The switch statement was also improperly indented, according to Documentation/CodingStyle. This prepares nfs_validate_mount_data for the addition of option string parsing. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:56 +08:00
case 1:
data->namlen = 0;
case 2:
data->bsize = 0;
case 3:
if (data->flags & NFS_MOUNT_VER3)
goto out_no_v3;
data->root.size = NFS2_FHSIZE;
memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
case 4:
if (data->flags & NFS_MOUNT_SECFLAVOUR)
goto out_no_sec;
case 5:
memset(data->context, 0, sizeof(data->context));
case 6:
NFS: Fix filehandle size comparisons in the mount code Fix a sign issue in xdr_decode_fhstatus3() Fix incorrect comparison in nfs_validate_mount_data() Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 03:21:11 +08:00
if (data->flags & NFS_MOUNT_VER3) {
if (data->root.size > NFS3_FHSIZE || data->root.size == 0)
goto out_invalid_fh;
NFS: Clean up nfs_validate_mount_data Move error handling code out of the main code path. The switch statement was also improperly indented, according to Documentation/CodingStyle. This prepares nfs_validate_mount_data for the addition of option string parsing. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:56 +08:00
mntfh->size = data->root.size;
NFS: Fix filehandle size comparisons in the mount code Fix a sign issue in xdr_decode_fhstatus3() Fix incorrect comparison in nfs_validate_mount_data() Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 03:21:11 +08:00
} else
NFS: Clean up nfs_validate_mount_data Move error handling code out of the main code path. The switch statement was also improperly indented, according to Documentation/CodingStyle. This prepares nfs_validate_mount_data for the addition of option string parsing. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:56 +08:00
mntfh->size = NFS2_FHSIZE;
memcpy(mntfh->data, data->root.data, mntfh->size);
if (mntfh->size < sizeof(mntfh->data))
memset(mntfh->data + mntfh->size, 0,
sizeof(mntfh->data) - mntfh->size);
NFS: Verify server address before invoking in-kernel mount client Re-order mount option sanity checking slightly to ensure we have a valid server address *before* trying to do the mountd RPC call. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-25 03:39:50 +08:00
NFS: use in-kernel mount argument structure for nfsv[23] mounts The user-visible nfs_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv[23] in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:43:56 +08:00
/*
* Translate to nfs_parsed_mount_data, which nfs_fill_super
* can deal with.
*/
NFS: Don't apply NFS_MOUNT_FLAGMASK to text-based mounts The point of introducing text-based mounts was to allow us to add functionality without having to worry about legacy binary mount formats. The mask should be there in order to ensure that binary formats don't start enabling features that they cannot support. There is no justification for applying it to the text mount path. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-08-16 04:59:14 +08:00
args->flags = data->flags & NFS_MOUNT_FLAGMASK;
NFS: use in-kernel mount argument structure for nfsv[23] mounts The user-visible nfs_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv[23] in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:43:56 +08:00
args->rsize = data->rsize;
args->wsize = data->wsize;
args->timeo = data->timeo;
args->retrans = data->retrans;
args->acregmin = data->acregmin;
args->acregmax = data->acregmax;
args->acdirmin = data->acdirmin;
args->acdirmax = data->acdirmax;
NFS: Support non-IPv4 addresses in nfs_parsed_mount_data Replace the nfs_server and mount_server address fields in the nfs_parsed_mount_data structure with a "struct sockaddr_storage" instead of a "struct sockaddr_in". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:28 +08:00
memcpy(&args->nfs_server.address, &data->addr,
sizeof(data->addr));
args->nfs_server.addrlen = sizeof(data->addr);
if (!nfs_verify_server_address((struct sockaddr *)
&args->nfs_server.address))
goto out_no_address;
NFS: use in-kernel mount argument structure for nfsv[23] mounts The user-visible nfs_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv[23] in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:43:56 +08:00
if (!(data->flags & NFS_MOUNT_TCP))
NFS/SUNRPC: use transport protocol naming Instead of an { address family, raw IP protocol number }-tuple, use the newly-defined RPC identifier when creating clients in the upper layers. Signed-off-by: Tom Talpey <tmt@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:48:23 +08:00
args->nfs_server.protocol = XPRT_TRANSPORT_UDP;
NFS: use in-kernel mount argument structure for nfsv[23] mounts The user-visible nfs_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv[23] in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:43:56 +08:00
/* N.B. caller will free nfs_server.hostname in all cases */
args->nfs_server.hostname = kstrdup(data->hostname, GFP_KERNEL);
args->namlen = data->namlen;
args->bsize = data->bsize;
NFS: Set security flavor default for NFSv2/3 mounts like other defaults Set the default security flavor when we set the other mount option default values. After this change, only the legacy user-space mount path needs to set the NFS_MOUNT_SECFLAVOUR flag. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 04:33:46 +08:00
if (data->flags & NFS_MOUNT_SECFLAVOUR)
args->auth_flavors[0] = data->pseudoflavor;
NFS - fix potential NULL pointer dereference v2 There is possible NULL pointer dereference if kstr[n]dup failed. So fix them for safety. Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-04-18 00:42:09 +08:00
if (!args->nfs_server.hostname)
goto out_nomem;
NFS: use new LSM interfaces to explicitly set mount options NFS and SELinux worked together previously because SELinux had NFS specific knowledge built in. This design was approved by both groups back in 2004 but the recent NFS changes to use nfs_parsed_mount_data and the usage of nfs_clone_mount_data showed this to be a poor fragile solution. This patch fixes the NFS functionality regression by making use of the new LSM interfaces to allow an FS to explicitly set its own mount options. The explicit setting of mount options is done in the nfs get_sb functions which are called before the generic vfs hooks try to set mount options for filesystems which use text mount data. This does not currently support NFSv4 as that functionality did not exist in previous kernels and thus there is no regression. I will be adding the needed code, which I believe to be the exact same as the v3 code, in nfs4_get_sb for 2.6.26. Signed-off-by: Eric Paris <eparis@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: James Morris <jmorris@namei.org>
2008-03-06 03:20:18 +08:00
/*
* The legacy version 6 binary mount data from userspace has a
* field used only to transport selinux information into the
* the kernel. To continue to support that functionality we
* have a touch of selinux knowledge here in the NFS code. The
* userspace code converted context=blah to just blah so we are
* converting back to the full string selinux understands.
*/
if (data->context[0]){
#ifdef CONFIG_SECURITY_SELINUX
int rc;
char *opts_str = kmalloc(sizeof(data->context) + 8, GFP_KERNEL);
if (!opts_str)
return -ENOMEM;
strcpy(opts_str, "context=");
data->context[NFS_MAX_CONTEXT_LEN] = '\0';
strcat(opts_str, &data->context[0]);
rc = security_sb_parse_opts_str(opts_str, &args->lsm_opts);
kfree(opts_str);
if (rc)
return rc;
#else
return -EINVAL;
#endif
}
NFS: Clean up nfs_validate_mount_data Move error handling code out of the main code path. The switch statement was also improperly indented, according to Documentation/CodingStyle. This prepares nfs_validate_mount_data for the addition of option string parsing. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:56 +08:00
break;
NFS: Add final pieces to support in-kernel mount option parsing Hook in final components required for supporting in-kernel mount option parsing for NFSv2 and NFSv3 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:54 +08:00
default: {
int status;
NFS: use in-kernel mount argument structure for nfsv[23] mounts The user-visible nfs_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv[23] in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:43:56 +08:00
if (nfs_parse_mount_options((char *)options, args) == 0)
return -EINVAL;
NFS: Add final pieces to support in-kernel mount option parsing Hook in final components required for supporting in-kernel mount option parsing for NFSv2 and NFSv3 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:54 +08:00
NFS: Verify server address before invoking in-kernel mount client Re-order mount option sanity checking slightly to ensure we have a valid server address *before* trying to do the mountd RPC call. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-25 03:39:50 +08:00
if (!nfs_verify_server_address((struct sockaddr *)
&args->nfs_server.address))
goto out_no_address;
NFS: Move the nfs_set_port() call out of nfs_parse_mount_options() The remount path does not need to set the port in the server address. Since it's not really a part of option parsing, move the nfs_set_port() call to nfs_parse_mount_options()'s callers. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-27 05:47:05 +08:00
nfs_set_port((struct sockaddr *)&args->nfs_server.address,
args->nfs_server.port);
NFS: set transport defaults after mount option parsing is finished Move the UDP/TCP default timeo/retrans settings for text mounts to nfs_init_timeout_values(), which was were they were always being initialised (and sanity checked) for binary mounts. Document the default timeout values using appropriate #defines. Ensure that we initialise and sanity check the transport protocols that may have been specified by the user. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-03 02:43:47 +08:00
nfs_set_mount_transport_protocol(args);
NFS: Use common device name parsing logic for NFSv4 and NFSv2/v3 To support passing a raw IPv6 address as a server hostname, we need to expand the logic that handles splitting the passed-in device name into a server hostname and export path Start by pulling device name parsing out of the mount option validation functions and into separate helper functions. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:37 +08:00
status = nfs_parse_devname(dev_name,
&args->nfs_server.hostname,
PAGE_SIZE,
&args->nfs_server.export_path,
NFS_MAXPATHLEN);
if (!status)
status = nfs_try_mount(args, mntfh);
NFS: Add final pieces to support in-kernel mount option parsing Hook in final components required for supporting in-kernel mount option parsing for NFSv2 and NFSv3 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:54 +08:00
NFS: Use common device name parsing logic for NFSv4 and NFSv2/v3 To support passing a raw IPv6 address as a server hostname, we need to expand the logic that handles splitting the passed-in device name into a server hostname and export path Start by pulling device name parsing out of the mount option validation functions and into separate helper functions. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:37 +08:00
kfree(args->nfs_server.export_path);
args->nfs_server.export_path = NULL;
NFS: Add final pieces to support in-kernel mount option parsing Hook in final components required for supporting in-kernel mount option parsing for NFSv2 and NFSv3 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:54 +08:00
if (status)
NFS: Return a real error code from mount(2) Don't filter the return code from the in-kernel rpcbind or NFS mount clients. Return the real error code so that callers of the new NFS text-based mount API can apply a useful retry strategy. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-08-30 05:58:59 +08:00
return status;
NFS: Add final pieces to support in-kernel mount option parsing Hook in final components required for supporting in-kernel mount option parsing for NFSv2 and NFSv3 mounts. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:54 +08:00
break;
}
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
#ifndef CONFIG_NFS_V3
NFS: use in-kernel mount argument structure for nfsv[23] mounts The user-visible nfs_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv[23] in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:43:56 +08:00
if (args->flags & NFS_MOUNT_VER3)
NFS: Clean up nfs_validate_mount_data Move error handling code out of the main code path. The switch statement was also improperly indented, according to Documentation/CodingStyle. This prepares nfs_validate_mount_data for the addition of option string parsing. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:56 +08:00
goto out_v3_not_compiled;
#endif /* !CONFIG_NFS_V3 */
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Clean up nfs_validate_mount_data Move error handling code out of the main code path. The switch statement was also improperly indented, according to Documentation/CodingStyle. This prepares nfs_validate_mount_data for the addition of option string parsing. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:56 +08:00
return 0;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Clean up nfs_validate_mount_data Move error handling code out of the main code path. The switch statement was also improperly indented, according to Documentation/CodingStyle. This prepares nfs_validate_mount_data for the addition of option string parsing. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:56 +08:00
out_no_data:
dfprintk(MOUNT, "NFS: mount program didn't pass any mount data\n");
return -EINVAL;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
NFS: Clean up nfs_validate_mount_data Move error handling code out of the main code path. The switch statement was also improperly indented, according to Documentation/CodingStyle. This prepares nfs_validate_mount_data for the addition of option string parsing. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:56 +08:00
out_no_v3:
dfprintk(MOUNT, "NFS: nfs_mount_data version %d does not support v3\n",
data->version);
return -EINVAL;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Clean up nfs_validate_mount_data Move error handling code out of the main code path. The switch statement was also improperly indented, according to Documentation/CodingStyle. This prepares nfs_validate_mount_data for the addition of option string parsing. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:56 +08:00
out_no_sec:
dfprintk(MOUNT, "NFS: nfs_mount_data version supports only AUTH_SYS\n");
return -EINVAL;
#ifndef CONFIG_NFS_V3
out_v3_not_compiled:
dfprintk(MOUNT, "NFS: NFSv3 is not compiled into kernel\n");
return -EPROTONOSUPPORT;
#endif /* !CONFIG_NFS_V3 */
NFS - fix potential NULL pointer dereference v2 There is possible NULL pointer dereference if kstr[n]dup failed. So fix them for safety. Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-04-18 00:42:09 +08:00
out_nomem:
dfprintk(MOUNT, "NFS: not enough memory to handle mount options\n");
return -ENOMEM;
NFS: Clean up nfs_validate_mount_data Move error handling code out of the main code path. The switch statement was also improperly indented, according to Documentation/CodingStyle. This prepares nfs_validate_mount_data for the addition of option string parsing. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:56 +08:00
out_no_address:
dfprintk(MOUNT, "NFS: mount program didn't pass remote address\n");
return -EINVAL;
out_invalid_fh:
dfprintk(MOUNT, "NFS: invalid root filehandle\n");
return -EINVAL;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: implement option checking when remounting NFS filesystems (resend) When remounting an NFS or NFS4 filesystem, the new NFS options are not respected, yet the remount will still return success. This patch adds a remount_fs sb op for NFS that checks any new nfs mount options against the existing ones and fails the mount if any have changed. This is only implemented for string-based mount options since doing this with binary options isn't really feasible. This is essentially the same as the original patch I sent out, but adds a check to see if the addr= option has changed. Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-11 03:38:39 +08:00
static int
nfs_compare_remount_data(struct nfs_server *nfss,
struct nfs_parsed_mount_data *data)
{
if (data->flags != nfss->flags ||
data->rsize != nfss->rsize ||
data->wsize != nfss->wsize ||
data->retrans != nfss->client->cl_timeout->to_retries ||
data->auth_flavors[0] != nfss->client->cl_auth->au_flavor ||
data->acregmin != nfss->acregmin / HZ ||
data->acregmax != nfss->acregmax / HZ ||
data->acdirmin != nfss->acdirmin / HZ ||
data->acdirmax != nfss->acdirmax / HZ ||
data->timeo != (10U * nfss->client->cl_timeout->to_initval / HZ) ||
data->nfs_server.addrlen != nfss->nfs_client->cl_addrlen ||
memcmp(&data->nfs_server.address, &nfss->nfs_client->cl_addr,
data->nfs_server.addrlen) != 0)
return -EINVAL;
return 0;
}
static int
nfs_remount(struct super_block *sb, int *flags, char *raw_data)
{
int error;
struct nfs_server *nfss = sb->s_fs_info;
struct nfs_parsed_mount_data *data;
struct nfs_mount_data *options = (struct nfs_mount_data *)raw_data;
struct nfs4_mount_data *options4 = (struct nfs4_mount_data *)raw_data;
NFS: Fix a dependency on CONFIG_NFS_V4 in nfs_remount Fix the 'nfs4_fs_type' undeclared error in nfs_remount when compiling sans NFSv4... Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: Jeff Layton <jlayton@redhat.com>
2008-06-18 04:12:00 +08:00
u32 nfsvers = nfss->nfs_client->rpc_ops->version;
NFS: implement option checking when remounting NFS filesystems (resend) When remounting an NFS or NFS4 filesystem, the new NFS options are not respected, yet the remount will still return success. This patch adds a remount_fs sb op for NFS that checks any new nfs mount options against the existing ones and fails the mount if any have changed. This is only implemented for string-based mount options since doing this with binary options isn't really feasible. This is essentially the same as the original patch I sent out, but adds a check to see if the addr= option has changed. Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-11 03:38:39 +08:00
/*
* Userspace mount programs that send binary options generally send
* them populated with default values. We have no way to know which
* ones were explicitly specified. Fall back to legacy behavior and
* just return success.
*/
nfs_remount oops when rebooting + possible fix Jeff, Trond, The commit 48b605f83c920d8daa50e43fc2c7f718e04c7bfa (NFS: implement option checking when remounting NFS filesystems (resend)) generate an Oops on my platform when rebooting while its root FS on an NFS share (NFSv3, TCP) : Unmounting local filesystems...done. Unable to handle kernel NULL pointer dereference at virtual address 00000000 pgd = c3d00000 [00000000] *pgd=a3d72031, *pte=00000000, *ppte=00000000 Internal error: Oops: 17 [#1] Modules linked in: cpufreq_powersave cpufreq_ondemand cpufreq_userspace cpufreq_conservative ext3 jbd sd_mod pata_pcmcia libata scsi_mod pcmcia loop firmware_class pxafb cfbcopyarea cfbimgblt cfbfillrect pxa2xx_cs pxa2xx_core pcmcia_core snd_pxa2xx_ac97 snd_ac97_codec ac97_bus snd_pxa2xx_pcm snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd isp116x_hcd soundcore rtc_sa1100 snd_page_alloc pxa25x_udc usbcore rtc_ds1307 rtc_core CPU: 0 Not tainted (2.6.26-03414-g33af79d-dirty #15) PC is at nfs_remount+0x40/0x264 LR is at do_remount_sb+0x158/0x194 pc : [<c00bbf54>] lr : [<c0076c40>] psr: 60000013 sp : c2dd1e70 ip : c2dd1e98 fp : c2dd1e94 r10: 00000040 r9 : c3d17000 r8 : c3c3fc40 r7 : 00000000 r6 : 00000000 r5 : c3d2b200 r4 : 00000000 r3 : 00000003 r2 : 00000000 r1 : c2dd1e9c r0 : c3c3fc00 Flags: nZCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment user Control: 0000397f Table: a3d00000 DAC: 00000015 Process mount (pid: 1462, stack limit = 0xc2dd0270) Stack: (0xc2dd1e70 to 0xc2dd2000) 1e60: 00000000 c3c3fc00 00000000 00000000 1e80: c3c3fc40 c3d17000 c2dd1ebc c2dd1e98 c0076c40 c00bbf20 c01c61e4 00000001 1ea0: c2dd1ebc 00000001 c3c3fc00 c2dd1ef0 c2dd1ee4 c2dd1ec0 c008c6d8 c0076af4 1ec0: 00000021 00000040 c2dd1ef0 c3d77000 c3eaa000 00000000 c2dd1f6c c2dd1ee8 1ee0: c008d1bc c008c5f8 00000000 c2dd0000 c3c0c320 c3805b38 c002064c 0001f820 1f00: 0001f810 00000001 00000001 00000000 c2dd0000 00000000 c2dd1f34 c2dd1f28 1f20: c005ead8 c005e6f8 c2dd1f44 c2dd1f38 c005eaf8 c005ead0 c2dd1f6c c2dd1f48 1f40: c008ae3c 00000000 c3d77000 0001f810 c0ed0021 c0020ca8 c2dd0000 00000000 1f60: c2dd1fa4 c2dd1f70 c008d2d4 c008d0bc 00000000 0001f810 c2dd1f9c c3eaa000 1f80: c3d17000 00000000 00000000 be8b6aa8 be8b6ad0 00000015 00000000 c2dd1fa8 1fa0: c0020b00 c008d254 00000000 be8b6aa8 0001f810 0001f820 0001f830 c0ed0021 1fc0: 00000000 be8b6aa8 be8b6ad0 00000015 00000000 be8b6ad0 0001f810 be8b6aa8 1fe0: 0001f810 be8b6964 0000aab8 40125124 60000010 0001f810 00000000 00000000 Backtrace: [<c00bbf14>] (nfs_remount+0x0/0x264) from [<c0076c40>] (do_remount_sb+0x158/0x194) r9:c3d17000 r8:c3c3fc40 r7:00000000 r6:00000000 r5:c3c3fc00 r4:00000000 [<c0076ae8>] (do_remount_sb+0x0/0x194) from [<c008c6d8>] (do_remount+0xec/0x118) r6:c2dd1ef0 r5:c3c3fc00 r4:00000001 [<c008c5ec>] (do_remount+0x0/0x118) from [<c008d1bc>] (do_mount+0x10c/0x198) [<c008d0b0>] (do_mount+0x0/0x198) from [<c008d2d4>] (sys_mount+0x8c/0xd4) [<c008d248>] (sys_mount+0x0/0xd4) from [<c0020b00>] (ret_fast_syscall+0x0/0x2c) r7:00000015 r6:be8b6ad0 r5:be8b6aa8 r4:00000000 Code: 0a000086 ea000006 e3530003 8a000004 (e5923000) ---[ end trace 55e1b689cf8c8a6a ]--- ------------[ cut here ]------------ WARNING: at kernel/exit.c:966 do_exit+0x3c/0x628() Modules linked in: cpufreq_powersave cpufreq_ondemand cpufreq_userspace cpufreq_conservative ext3 jbd sd_mod pata_pcmcia libata scsi_mod pcmcia loop firmware_class pxafb cfbcopyarea cfbimgblt cfbfillrect pxa2xx_cs pxa2xx_core pcmcia_core snd_pxa2xx_ac97 snd_ac97_codec ac97_bus snd_pxa2xx_pcm snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd isp116x_hcd soundcore rtc_sa1100 snd_page_alloc pxa25x_udc usbcore rtc_ds1307 rtc_core [<c0025168>] (dump_stack+0x0/0x14) from [<c0032154>] (warn_on_slowpath+0x4c/0x68) [<c0032108>] (warn_on_slowpath+0x0/0x68) from [<c003531c>] (do_exit+0x3c/0x628) r6:0000000b r5:c3c3dc80 r4:c2dd0000 [<c00352e0>] (do_exit+0x0/0x628) from [<c0025004>] (die+0x2b0/0x30c) [<c0024d54>] (die+0x0/0x30c) from [<c00270bc>] (__do_kernel_fault+0x6c/0x80) [<c0027050>] (__do_kernel_fault+0x0/0x80) from [<c00272e0>] (do_page_fault+0x210/0x230) r7:c3fa7118 r6:c3c3dc80 r5:c3d166a8 r4:00010000 [<c00270d0>] (do_page_fault+0x0/0x230) from [<c00201ec>] (do_DataAbort+0x3c/0xa0) [<c00201b0>] (do_DataAbort+0x0/0xa0) from [<c002064c>] (__dabt_svc+0x4c/0x60) Exception stack(0xc2dd1e28 to 0xc2dd1e70) 1e20: c3c3fc00 c2dd1e9c 00000000 00000003 00000000 c3d2b200 1e40: 00000000 00000000 c3c3fc40 c3d17000 00000040 c2dd1e94 c2dd1e98 c2dd1e70 1e60: c0076c40 c00bbf54 60000013 ffffffff r8:c3c3fc40 r7:00000000 r6:00000000 r5:c2dd1e5c r4:ffffffff [<c00bbf14>] (nfs_remount+0x0/0x264) from [<c0076c40>] (do_remount_sb+0x158/0x194) r9:c3d17000 r8:c3c3fc40 r7:00000000 r6:00000000 r5:c3c3fc00 r4:00000000 [<c0076ae8>] (do_remount_sb+0x0/0x194) from [<c008c6d8>] (do_remount+0xec/0x118) r6:c2dd1ef0 r5:c3c3fc00 r4:00000001 [<c008c5ec>] (do_remount+0x0/0x118) from [<c008d1bc>] (do_mount+0x10c/0x198) [<c008d0b0>] (do_mount+0x0/0x198) from [<c008d2d4>] (sys_mount+0x8c/0xd4) [<c008d248>] (sys_mount+0x0/0xd4) from [<c0020b00>] (ret_fast_syscall+0x0/0x2c) r7:00000015 r6:be8b6ad0 r5:be8b6aa8 r4:00000000 ---[ end trace 55e1b689cf8c8a6a ]--- /etc/rc6.d/S60umountroot: line 17: 1462 Segmentation fault mount $MOUNT_FORCE_OPT -n -o remount,ro -t dummytype dummydev / 2> /dev/null The new super.c:nfs_remount function doesn't check the validity of the options/options4 pointers. Unfortunately, this seems to happend. The obvious patch seems to check the pointers, and not to do anything if the happend to be NULL. Tested on an XScale PXA255 system, latest git. Regards, M. Signed-off-by: Marc Zyngier <marc.zyngier@altran.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-17 19:21:55 +08:00
if ((nfsvers == 4 && (!options4 || options4->version == 1)) ||
(nfsvers <= 3 && (!options || (options->version >= 1 &&
options->version <= 6))))
NFS: implement option checking when remounting NFS filesystems (resend) When remounting an NFS or NFS4 filesystem, the new NFS options are not respected, yet the remount will still return success. This patch adds a remount_fs sb op for NFS that checks any new nfs mount options against the existing ones and fails the mount if any have changed. This is only implemented for string-based mount options since doing this with binary options isn't really feasible. This is essentially the same as the original patch I sent out, but adds a check to see if the addr= option has changed. Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-11 03:38:39 +08:00
return 0;
data = kzalloc(sizeof(*data), GFP_KERNEL);
if (data == NULL)
return -ENOMEM;
Push BKL down into ->remount_fs() [xfs, btrfs, capifs, shmem don't need BKL, exempt] Signed-off-by: Alessio Igor Bogani <abogani@texware.it> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2009-05-12 21:10:54 +08:00
lock_kernel();
NFS: implement option checking when remounting NFS filesystems (resend) When remounting an NFS or NFS4 filesystem, the new NFS options are not respected, yet the remount will still return success. This patch adds a remount_fs sb op for NFS that checks any new nfs mount options against the existing ones and fails the mount if any have changed. This is only implemented for string-based mount options since doing this with binary options isn't really feasible. This is essentially the same as the original patch I sent out, but adds a check to see if the addr= option has changed. Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-11 03:38:39 +08:00
/* fill out struct with values from existing mount */
data->flags = nfss->flags;
data->rsize = nfss->rsize;
data->wsize = nfss->wsize;
data->retrans = nfss->client->cl_timeout->to_retries;
data->auth_flavors[0] = nfss->client->cl_auth->au_flavor;
data->acregmin = nfss->acregmin / HZ;
data->acregmax = nfss->acregmax / HZ;
data->acdirmin = nfss->acdirmin / HZ;
data->acdirmax = nfss->acdirmax / HZ;
data->timeo = 10U * nfss->client->cl_timeout->to_initval / HZ;
data->nfs_server.addrlen = nfss->nfs_client->cl_addrlen;
memcpy(&data->nfs_server.address, &nfss->nfs_client->cl_addr,
data->nfs_server.addrlen);
/* overwrite those values with any that were specified */
error = nfs_parse_mount_options((char *)options, data);
if (error < 0)
goto out;
/* compare new mount options with old ones */
error = nfs_compare_remount_data(nfss, data);
out:
kfree(data);
Push BKL down into ->remount_fs() [xfs, btrfs, capifs, shmem don't need BKL, exempt] Signed-off-by: Alessio Igor Bogani <abogani@texware.it> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2009-05-12 21:10:54 +08:00
unlock_kernel();
NFS: implement option checking when remounting NFS filesystems (resend) When remounting an NFS or NFS4 filesystem, the new NFS options are not respected, yet the remount will still return success. This patch adds a remount_fs sb op for NFS that checks any new nfs mount options against the existing ones and fails the mount if any have changed. This is only implemented for string-based mount options since doing this with binary options isn't really feasible. This is essentially the same as the original patch I sent out, but adds a check to see if the addr= option has changed. Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-11 03:38:39 +08:00
return error;
}
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
/*
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
* Initialise the common bits of the superblock
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
*/
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
static inline void nfs_initialise_sb(struct super_block *sb)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
struct nfs_server *server = NFS_SB(sb);
NFS: Eliminate client_sys in favour of cl_rpcclient Eliminate nfs_server::client_sys in favour of nfs_client::cl_rpcclient as we only really need one per server that we're talking to since it doesn't have any security on it. The retransmission management variables are also moved to the common struct as they're required to set up the cl_rpcclient connection. The NFS2/3 client and client_acl connections are thenceforth derived by cloning the cl_rpcclient connection and post-applying the authorisation flavour. The code for setting up the initial common connection has been moved to client.c as nfs_create_rpc_client(). All the NFS program definition tables are also moved there as that's where they're now required rather than super.c. Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:12 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
sb->s_magic = NFS_SUPER_MAGIC;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/* We probably want something more informative here */
snprintf(sb->s_id, sizeof(sb->s_id),
"%x:%x", MAJOR(sb->s_dev), MINOR(sb->s_dev));
NFS: Eliminate client_sys in favour of cl_rpcclient Eliminate nfs_server::client_sys in favour of nfs_client::cl_rpcclient as we only really need one per server that we're talking to since it doesn't have any security on it. The retransmission management variables are also moved to the common struct as they're required to set up the cl_rpcclient connection. The NFS2/3 client and client_acl connections are thenceforth derived by cloning the cl_rpcclient connection and post-applying the authorisation flavour. The code for setting up the initial common connection has been moved to client.c as nfs_create_rpc_client(). All the NFS program definition tables are also moved there as that's where they're now required rather than super.c. Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:12 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (sb->s_blocksize == 0)
sb->s_blocksize = nfs_block_bits(server->wsize,
&sb->s_blocksize_bits);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (server->flags & NFS_MOUNT_NOAC)
sb->s_flags |= MS_SYNCHRONOUS;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
nfs_super_set_maxbytes(sb, server->maxfilesize);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
/*
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
* Finish setting up an NFS2/3 superblock
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
*/
NFS: use in-kernel mount argument structure for nfsv[23] mounts The user-visible nfs_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv[23] in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:43:56 +08:00
static void nfs_fill_super(struct super_block *sb,
struct nfs_parsed_mount_data *data)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{
struct nfs_server *server = NFS_SB(sb);
NFS: Eliminate client_sys in favour of cl_rpcclient Eliminate nfs_server::client_sys in favour of nfs_client::cl_rpcclient as we only really need one per server that we're talking to since it doesn't have any security on it. The retransmission management variables are also moved to the common struct as they're required to set up the cl_rpcclient connection. The NFS2/3 client and client_acl connections are thenceforth derived by cloning the cl_rpcclient connection and post-applying the authorisation flavour. The code for setting up the initial common connection has been moved to client.c as nfs_create_rpc_client(). All the NFS program definition tables are also moved there as that's where they're now required rather than super.c. Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:12 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
sb->s_blocksize_bits = 0;
sb->s_blocksize = 0;
if (data->bsize)
sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (server->flags & NFS_MOUNT_VER3) {
/* The VFS shouldn't apply the umask to mode bits. We will do
* so ourselves when necessary.
*/
sb->s_flags |= MS_POSIXACL;
sb->s_time_gran = 1;
Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ Conflicts: fs/nfs/inode.c fs/super.c Fix conflicts between patch 'NFS: Split fs/nfs/inode.c' and patch 'VFS: Permit filesystem to override root dentry on mount'
2006-06-24 20:41:41 +08:00
}
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
sb->s_op = &nfs_sops;
nfs_initialise_sb(sb);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
/*
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
* Finish setting up a cloned NFS2/3 superblock
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
*/
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
static void nfs_clone_super(struct super_block *sb,
const struct super_block *old_sb)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
struct nfs_server *server = NFS_SB(sb);
sb->s_blocksize_bits = old_sb->s_blocksize_bits;
sb->s_blocksize = old_sb->s_blocksize;
sb->s_maxbytes = old_sb->s_maxbytes;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
if (server->flags & NFS_MOUNT_VER3) {
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/* The VFS shouldn't apply the umask to mode bits. We will do
* so ourselves when necessary.
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
*/
sb->s_flags |= MS_POSIXACL;
sb->s_time_gran = 1;
}
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
sb->s_op = old_sb->s_op;
nfs_initialise_sb(sb);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Error when mounting the same filesystem with different options Unless the user sets the NFS_MOUNT_NOSHAREDCACHE mount flag, we should return EBUSY if the filesystem is already mounted on a superblock that has set conflicting mount options. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
static int nfs_compare_mount_options(const struct super_block *s, const struct nfs_server *b, int flags)
{
const struct nfs_server *a = s->s_fs_info;
const struct rpc_clnt *clnt_a = a->client;
const struct rpc_clnt *clnt_b = b->client;
if ((s->s_flags & NFS_MS_MASK) != (flags & NFS_MS_MASK))
goto Ebusy;
if (a->nfs_client != b->nfs_client)
goto Ebusy;
if (a->flags != b->flags)
goto Ebusy;
if (a->wsize != b->wsize)
goto Ebusy;
if (a->rsize != b->rsize)
goto Ebusy;
if (a->acregmin != b->acregmin)
goto Ebusy;
if (a->acregmax != b->acregmax)
goto Ebusy;
if (a->acdirmin != b->acdirmin)
goto Ebusy;
if (a->acdirmax != b->acdirmax)
goto Ebusy;
if (clnt_a->cl_auth->au_flavor != clnt_b->cl_auth->au_flavor)
goto Ebusy;
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
return 1;
NFS: Error when mounting the same filesystem with different options Unless the user sets the NFS_MOUNT_NOSHAREDCACHE mount flag, we should return EBUSY if the filesystem is already mounted on a superblock that has set conflicting mount options. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
Ebusy:
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
return 0;
}
struct nfs_sb_mountdata {
struct nfs_server *server;
int mntflags;
};
static int nfs_set_super(struct super_block *s, void *data)
{
struct nfs_sb_mountdata *sb_mntdata = data;
struct nfs_server *server = sb_mntdata->server;
int ret;
s->s_flags = sb_mntdata->mntflags;
s->s_fs_info = server;
ret = set_anon_super(s, server);
if (ret == 0)
server->s_dev = s->s_dev;
return ret;
}
NFS: Add support for AF_INET6 addresses in nfs_compare_super() Refactor nfs_compare_super() and add AF_INET6 support. Replace the generic memcmp() to document explicitly what parts of the addresses must match in this check, and make the comparison independent of the lengths of both addresses. A side benefit is both tests are more computationally efficient than a memcmp(). Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:57:38 +08:00
static int nfs_compare_super_address(struct nfs_server *server1,
struct nfs_server *server2)
{
struct sockaddr *sap1, *sap2;
sap1 = (struct sockaddr *)&server1->nfs_client->cl_addr;
sap2 = (struct sockaddr *)&server2->nfs_client->cl_addr;
if (sap1->sa_family != sap2->sa_family)
return 0;
switch (sap1->sa_family) {
case AF_INET: {
struct sockaddr_in *sin1 = (struct sockaddr_in *)sap1;
struct sockaddr_in *sin2 = (struct sockaddr_in *)sap2;
if (sin1->sin_addr.s_addr != sin2->sin_addr.s_addr)
return 0;
if (sin1->sin_port != sin2->sin_port)
return 0;
break;
}
case AF_INET6: {
struct sockaddr_in6 *sin1 = (struct sockaddr_in6 *)sap1;
struct sockaddr_in6 *sin2 = (struct sockaddr_in6 *)sap2;
if (!ipv6_addr_equal(&sin1->sin6_addr, &sin2->sin6_addr))
return 0;
if (sin1->sin6_port != sin2->sin6_port)
return 0;
break;
}
default:
return 0;
}
return 1;
}
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
static int nfs_compare_super(struct super_block *sb, void *data)
{
struct nfs_sb_mountdata *sb_mntdata = data;
struct nfs_server *server = sb_mntdata->server, *old = NFS_SB(sb);
int mntflags = sb_mntdata->mntflags;
NFS: Add support for AF_INET6 addresses in nfs_compare_super() Refactor nfs_compare_super() and add AF_INET6 support. Replace the generic memcmp() to document explicitly what parts of the addresses must match in this check, and make the comparison independent of the lengths of both addresses. A side benefit is both tests are more computationally efficient than a memcmp(). Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:57:38 +08:00
if (!nfs_compare_super_address(old, server))
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
return 0;
/* Note: NFS_MOUNT_UNSHARED == NFS4_MOUNT_UNSHARED */
if (old->flags & NFS_MOUNT_UNSHARED)
return 0;
if (memcmp(&old->fsid, &server->fsid, sizeof(old->fsid)) != 0)
return 0;
return nfs_compare_mount_options(sb, server, mntflags);
NFS: Error when mounting the same filesystem with different options Unless the user sets the NFS_MOUNT_NOSHAREDCACHE mount flag, we should return EBUSY if the filesystem is already mounted on a superblock that has set conflicting mount options. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
}
mm: bdi: expose the BDI object in sysfs for NFS Register NFS' backing_dev_info under sysfs with the name "nfs-MAJOR:MINOR" Signed-off-by: Miklos Szeredi <mszeredi@suse.cz> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-30 15:54:33 +08:00
static int nfs_bdi_register(struct nfs_server *server)
{
return bdi_register_dev(&server->backing_dev_info, server->s_dev);
}
Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ Conflicts: fs/nfs/inode.c fs/super.c Fix conflicts between patch 'NFS: Split fs/nfs/inode.c' and patch 'VFS: Permit filesystem to override root dentry on mount'
2006-06-24 20:41:41 +08:00
static int nfs_get_sb(struct file_system_type *fs_type,
int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{
struct nfs_server *server = NULL;
struct super_block *s;
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
struct nfs_parsed_mount_data *data;
struct nfs_fh *mntfh;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
struct dentry *mntroot;
NFS: Add the mount option "nosharecache" Prior to David Howell's mount changes in 2.6.18, users who mounted different directories which happened to be from the same filesystem on the server would get different super blocks, and hence could choose different mount options. As long as there were no hard linked files that crossed from one subtree to another, this was quite safe. Post the changes, if the two directories are on the same filesystem (have the same 'fsid'), they will share the same super block, and hence the same mount options. Add a flag to allow users to elect not to share the NFS super block with another mount point, even if the fsids are the same. This will allow users to set different mount options for the two different super blocks, as was previously possible. It is still up to the user to ensure that there are no cache coherency issues when doing this, however the default behaviour will be to share super blocks whenever two paths result in the same fsid. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
int (*compare_super)(struct super_block *, void *) = nfs_compare_super;
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
struct nfs_sb_mountdata sb_mntdata = {
.mntflags = flags,
};
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
int error = -ENOMEM;
data = kzalloc(sizeof(*data), GFP_KERNEL);
mntfh = kzalloc(sizeof(*mntfh), GFP_KERNEL);
if (data == NULL || mntfh == NULL)
goto out_free_fh;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
security_init_mnt_opts(&data->lsm_opts);
NFS: use new LSM interfaces to explicitly set mount options NFS and SELinux worked together previously because SELinux had NFS specific knowledge built in. This design was approved by both groups back in 2004 but the recent NFS changes to use nfs_parsed_mount_data and the usage of nfs_clone_mount_data showed this to be a poor fragile solution. This patch fixes the NFS functionality regression by making use of the new LSM interfaces to allow an FS to explicitly set its own mount options. The explicit setting of mount options is done in the nfs get_sb functions which are called before the generic vfs hooks try to set mount options for filesystems which use text mount data. This does not currently support NFSv4 as that functionality did not exist in previous kernels and thus there is no regression. I will be adding the needed code, which I believe to be the exact same as the v3 code, in nfs4_get_sb for 2.6.26. Signed-off-by: Eric Paris <eparis@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: James Morris <jmorris@namei.org>
2008-03-06 03:20:18 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/* Validate the mount data */
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
error = nfs_validate_mount_data(raw_data, data, mntfh, dev_name);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (error < 0)
NFS: Clean up error handling in nfs_get_sb The error return logic in nfs_get_sb now matches nfs4_get_sb, and is more maintainable. A subsequent patch will take advantage of this simplification. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:35 +08:00
goto out;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/* Get a volume representation */
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
server = nfs_create_server(data, mntfh);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (IS_ERR(server)) {
error = PTR_ERR(server);
NFS: Clean up error handling in nfs_get_sb The error return logic in nfs_get_sb now matches nfs4_get_sb, and is more maintainable. A subsequent patch will take advantage of this simplification. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:35 +08:00
goto out;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
sb_mntdata.server = server;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Add the mount option "nosharecache" Prior to David Howell's mount changes in 2.6.18, users who mounted different directories which happened to be from the same filesystem on the server would get different super blocks, and hence could choose different mount options. As long as there were no hard linked files that crossed from one subtree to another, this was quite safe. Post the changes, if the two directories are on the same filesystem (have the same 'fsid'), they will share the same super block, and hence the same mount options. Add a flag to allow users to elect not to share the NFS super block with another mount point, even if the fsids are the same. This will allow users to set different mount options for the two different super blocks, as was previously possible. It is still up to the user to ensure that there are no cache coherency issues when doing this, however the default behaviour will be to share super blocks whenever two paths result in the same fsid. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
if (server->flags & NFS_MOUNT_UNSHARED)
compare_super = NULL;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/* Get a superblock - note that we may end up sharing one that already exists */
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
s = sget(fs_type, compare_super, nfs_set_super, &sb_mntdata);
Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ Conflicts: fs/nfs/inode.c fs/super.c Fix conflicts between patch 'NFS: Split fs/nfs/inode.c' and patch 'VFS: Permit filesystem to override root dentry on mount'
2006-06-24 20:41:41 +08:00
if (IS_ERR(s)) {
error = PTR_ERR(s);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
goto out_err_nosb;
Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ Conflicts: fs/nfs/inode.c fs/super.c Fix conflicts between patch 'NFS: Split fs/nfs/inode.c' and patch 'VFS: Permit filesystem to override root dentry on mount'
2006-06-24 20:41:41 +08:00
}
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (s->s_fs_info != server) {
nfs_free_server(server);
server = NULL;
mm: bdi: expose the BDI object in sysfs for NFS Register NFS' backing_dev_info under sysfs with the name "nfs-MAJOR:MINOR" Signed-off-by: Miklos Szeredi <mszeredi@suse.cz> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-30 15:54:33 +08:00
} else {
error = nfs_bdi_register(server);
if (error)
goto error_splat_super;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
}
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (!s->s_root) {
/* initial superblock/root creation */
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
nfs_fill_super(s, data);
NFS: Define and create superblock-level objects Define and create superblock-level cache index objects (as managed by nfs_server structs). Each superblock object is created in a server level index object and is itself an index into which inode-level objects are inserted. Ideally there would be one superblock-level object per server, and the former would be folded into the latter; however, since the "nosharecache" option exists this isn't possible. The superblock object key is a sequence consisting of: (1) Certain superblock s_flags. (2) Various connection parameters that serve to distinguish superblocks for sget(). (3) The volume FSID. (4) The security flavour. (5) The uniquifier length. (6) The uniquifier text. This is normally an empty string, unless the fsc=xyz mount option was used to explicitly specify a uniquifier. The key blob is of variable length, depending on the length of (6). The superblock object is given no coherency data to carry in the auxiliary data permitted by the cache. It is assumed that the superblock is always coherent. This patch also adds uniquification handling such that two otherwise identical superblocks, at least one of which is marked "nosharecache", won't end up trying to share the on-disk cache. It will be possible to manually provide a uniquifier through a mount option with a later patch to avoid the error otherwise produced. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Steve Dickson <steved@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Daire Byrne <Daire.Byrne@framestore.com>
2009-04-03 23:42:42 +08:00
nfs_fscache_get_super_cookie(s, data);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
}
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
mntroot = nfs_get_root(s, mntfh);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (IS_ERR(mntroot)) {
error = PTR_ERR(mntroot);
goto error_splat_super;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ Conflicts: fs/nfs/inode.c fs/super.c Fix conflicts between patch 'NFS: Split fs/nfs/inode.c' and patch 'VFS: Permit filesystem to override root dentry on mount'
2006-06-24 20:41:41 +08:00
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
error = security_sb_set_mnt_opts(s, &data->lsm_opts);
NFS: use new LSM interfaces to explicitly set mount options NFS and SELinux worked together previously because SELinux had NFS specific knowledge built in. This design was approved by both groups back in 2004 but the recent NFS changes to use nfs_parsed_mount_data and the usage of nfs_clone_mount_data showed this to be a poor fragile solution. This patch fixes the NFS functionality regression by making use of the new LSM interfaces to allow an FS to explicitly set its own mount options. The explicit setting of mount options is done in the nfs get_sb functions which are called before the generic vfs hooks try to set mount options for filesystems which use text mount data. This does not currently support NFSv4 as that functionality did not exist in previous kernels and thus there is no regression. I will be adding the needed code, which I believe to be the exact same as the v3 code, in nfs4_get_sb for 2.6.26. Signed-off-by: Eric Paris <eparis@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: James Morris <jmorris@namei.org>
2008-03-06 03:20:18 +08:00
if (error)
goto error_splat_root;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
s->s_flags |= MS_ACTIVE;
mnt->mnt_sb = s;
mnt->mnt_root = mntroot;
NFS: Clean up error handling in nfs_get_sb The error return logic in nfs_get_sb now matches nfs4_get_sb, and is more maintainable. A subsequent patch will take advantage of this simplification. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:35 +08:00
error = 0;
out:
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
kfree(data->nfs_server.hostname);
kfree(data->mount_server.hostname);
NFS: Define and create superblock-level objects Define and create superblock-level cache index objects (as managed by nfs_server structs). Each superblock object is created in a server level index object and is itself an index into which inode-level objects are inserted. Ideally there would be one superblock-level object per server, and the former would be folded into the latter; however, since the "nosharecache" option exists this isn't possible. The superblock object key is a sequence consisting of: (1) Certain superblock s_flags. (2) Various connection parameters that serve to distinguish superblocks for sget(). (3) The volume FSID. (4) The security flavour. (5) The uniquifier length. (6) The uniquifier text. This is normally an empty string, unless the fsc=xyz mount option was used to explicitly specify a uniquifier. The key blob is of variable length, depending on the length of (6). The superblock object is given no coherency data to carry in the auxiliary data permitted by the cache. It is assumed that the superblock is always coherent. This patch also adds uniquification handling such that two otherwise identical superblocks, at least one of which is marked "nosharecache", won't end up trying to share the on-disk cache. It will be possible to manually provide a uniquifier through a mount option with a later patch to avoid the error otherwise produced. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Steve Dickson <steved@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Daire Byrne <Daire.Byrne@framestore.com>
2009-04-03 23:42:42 +08:00
kfree(data->fscache_uniq);
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
security_free_mnt_opts(&data->lsm_opts);
out_free_fh:
kfree(mntfh);
kfree(data);
NFS: Clean up error handling in nfs_get_sb The error return logic in nfs_get_sb now matches nfs4_get_sb, and is more maintainable. A subsequent patch will take advantage of this simplification. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:35 +08:00
return error;
Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ Conflicts: fs/nfs/inode.c fs/super.c Fix conflicts between patch 'NFS: Split fs/nfs/inode.c' and patch 'VFS: Permit filesystem to override root dentry on mount'
2006-06-24 20:41:41 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
out_err_nosb:
nfs_free_server(server);
NFS: Clean up error handling in nfs_get_sb The error return logic in nfs_get_sb now matches nfs4_get_sb, and is more maintainable. A subsequent patch will take advantage of this simplification. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:35 +08:00
goto out;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
NFS: use new LSM interfaces to explicitly set mount options NFS and SELinux worked together previously because SELinux had NFS specific knowledge built in. This design was approved by both groups back in 2004 but the recent NFS changes to use nfs_parsed_mount_data and the usage of nfs_clone_mount_data showed this to be a poor fragile solution. This patch fixes the NFS functionality regression by making use of the new LSM interfaces to allow an FS to explicitly set its own mount options. The explicit setting of mount options is done in the nfs get_sb functions which are called before the generic vfs hooks try to set mount options for filesystems which use text mount data. This does not currently support NFSv4 as that functionality did not exist in previous kernels and thus there is no regression. I will be adding the needed code, which I believe to be the exact same as the v3 code, in nfs4_get_sb for 2.6.26. Signed-off-by: Eric Paris <eparis@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: James Morris <jmorris@namei.org>
2008-03-06 03:20:18 +08:00
error_splat_root:
dput(mntroot);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
error_splat_super:
Convert obvious places to deactivate_locked_super() Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2009-05-06 13:34:22 +08:00
deactivate_locked_super(s);
NFS: Clean up error handling in nfs_get_sb The error return logic in nfs_get_sb now matches nfs4_get_sb, and is more maintainable. A subsequent patch will take advantage of this simplification. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:35 +08:00
goto out;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/*
* Destroy an NFS2/3 superblock
*/
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
static void nfs_kill_super(struct super_block *s)
{
struct nfs_server *server = NFS_SB(s);
mm: bdi: expose the BDI object in sysfs for NFS Register NFS' backing_dev_info under sysfs with the name "nfs-MAJOR:MINOR" Signed-off-by: Miklos Szeredi <mszeredi@suse.cz> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-30 15:54:33 +08:00
bdi_unregister(&server->backing_dev_info);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
kill_anon_super(s);
NFS: Define and create superblock-level objects Define and create superblock-level cache index objects (as managed by nfs_server structs). Each superblock object is created in a server level index object and is itself an index into which inode-level objects are inserted. Ideally there would be one superblock-level object per server, and the former would be folded into the latter; however, since the "nosharecache" option exists this isn't possible. The superblock object key is a sequence consisting of: (1) Certain superblock s_flags. (2) Various connection parameters that serve to distinguish superblocks for sget(). (3) The volume FSID. (4) The security flavour. (5) The uniquifier length. (6) The uniquifier text. This is normally an empty string, unless the fsc=xyz mount option was used to explicitly specify a uniquifier. The key blob is of variable length, depending on the length of (6). The superblock object is given no coherency data to carry in the auxiliary data permitted by the cache. It is assumed that the superblock is always coherent. This patch also adds uniquification handling such that two otherwise identical superblocks, at least one of which is marked "nosharecache", won't end up trying to share the on-disk cache. It will be possible to manually provide a uniquifier through a mount option with a later patch to avoid the error otherwise produced. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Steve Dickson <steved@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Daire Byrne <Daire.Byrne@framestore.com>
2009-04-03 23:42:42 +08:00
nfs_fscache_release_super_cookie(s);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
nfs_free_server(server);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/*
* Clone an NFS2/3 server record on xdev traversal (FSID-change)
*/
static int nfs_xdev_get_sb(struct file_system_type *fs_type, int flags,
const char *dev_name, void *raw_data,
struct vfsmount *mnt)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{
struct nfs_clone_mount *data = raw_data;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
struct super_block *s;
struct nfs_server *server;
struct dentry *mntroot;
NFS: Add the mount option "nosharecache" Prior to David Howell's mount changes in 2.6.18, users who mounted different directories which happened to be from the same filesystem on the server would get different super blocks, and hence could choose different mount options. As long as there were no hard linked files that crossed from one subtree to another, this was quite safe. Post the changes, if the two directories are on the same filesystem (have the same 'fsid'), they will share the same super block, and hence the same mount options. Add a flag to allow users to elect not to share the NFS super block with another mount point, even if the fsids are the same. This will allow users to set different mount options for the two different super blocks, as was previously possible. It is still up to the user to ensure that there are no cache coherency issues when doing this, however the default behaviour will be to share super blocks whenever two paths result in the same fsid. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
int (*compare_super)(struct super_block *, void *) = nfs_compare_super;
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
struct nfs_sb_mountdata sb_mntdata = {
.mntflags = flags,
};
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
int error;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
dprintk("--> nfs_xdev_get_sb()\n");
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/* create a new volume representation */
server = nfs_clone_server(NFS_SB(data->sb), data->fh, data->fattr);
if (IS_ERR(server)) {
error = PTR_ERR(server);
goto out_err_noserver;
}
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
sb_mntdata.server = server;
NFS: Move rpc_ops from nfs_server to nfs_client Move the rpc_ops from the nfs_server struct to the nfs_client struct as they're common to all server records of a particular NFS protocol version. Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:12 +08:00
NFS: Add the mount option "nosharecache" Prior to David Howell's mount changes in 2.6.18, users who mounted different directories which happened to be from the same filesystem on the server would get different super blocks, and hence could choose different mount options. As long as there were no hard linked files that crossed from one subtree to another, this was quite safe. Post the changes, if the two directories are on the same filesystem (have the same 'fsid'), they will share the same super block, and hence the same mount options. Add a flag to allow users to elect not to share the NFS super block with another mount point, even if the fsids are the same. This will allow users to set different mount options for the two different super blocks, as was previously possible. It is still up to the user to ensure that there are no cache coherency issues when doing this, however the default behaviour will be to share super blocks whenever two paths result in the same fsid. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
if (server->flags & NFS_MOUNT_UNSHARED)
compare_super = NULL;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/* Get a superblock - note that we may end up sharing one that already exists */
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
s = sget(&nfs_fs_type, compare_super, nfs_set_super, &sb_mntdata);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (IS_ERR(s)) {
error = PTR_ERR(s);
goto out_err_nosb;
}
NFS: Eliminate client_sys in favour of cl_rpcclient Eliminate nfs_server::client_sys in favour of nfs_client::cl_rpcclient as we only really need one per server that we're talking to since it doesn't have any security on it. The retransmission management variables are also moved to the common struct as they're required to set up the cl_rpcclient connection. The NFS2/3 client and client_acl connections are thenceforth derived by cloning the cl_rpcclient connection and post-applying the authorisation flavour. The code for setting up the initial common connection has been moved to client.c as nfs_create_rpc_client(). All the NFS program definition tables are also moved there as that's where they're now required rather than super.c. Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:12 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (s->s_fs_info != server) {
nfs_free_server(server);
server = NULL;
mm: bdi: expose the BDI object in sysfs for NFS Register NFS' backing_dev_info under sysfs with the name "nfs-MAJOR:MINOR" Signed-off-by: Miklos Szeredi <mszeredi@suse.cz> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-30 15:54:33 +08:00
} else {
error = nfs_bdi_register(server);
if (error)
goto error_splat_super;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Generalise the nfs_client structure Generalise the nfs_client structure by: (1) Moving nfs_client to a more general place (nfs_fs_sb.h). (2) Renaming its maintenance routines to be non-NFS4 specific. (3) Move those maintenance routines to a new non-NFS4 specific file (client.c) and move the declarations to internal.h. (4) Make nfs_find/get_client() take a full sockaddr_in to include the port number (will be required for NFS2/3). (5) Make nfs_find/get_client() take the NFS protocol version (again will be required to differentiate NFS2, 3 & 4 client records). Also: (6) Make nfs_client construction proceed akin to inodes, marking them as under construction and providing a function to indicate completion. (7) Make nfs_get_client() wait interruptibly if it finds a client that it can share, but that client is currently being constructed. (8) Make nfs4_create_client() use (6) and (7) instead of locking cl_sem. Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:10 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (!s->s_root) {
/* initial superblock/root creation */
nfs_clone_super(s, data->sb);
}
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
mntroot = nfs_get_root(s, data->fh);
if (IS_ERR(mntroot)) {
error = PTR_ERR(mntroot);
goto error_splat_super;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Fix a possible Oops in fs/nfs/super.c Sigh... commit 4584f520e1f773082ef44ff4f8969a5d992b16ec (NFS: Fix NFS mountpoint crossing...) had a slight flaw: server can be NULL if sget() returned an existing superblock. Fix the fix by dereferencing s->s_fs_info. Thanks to Coverity/Adrian Bunk and Frank Filz for spotting the bug. (See http://bugzilla.kernel.org/show_bug.cgi?id=9647) Also add in the same namespace Oops fix for NFSv4 in both the mountpoint crossing case, and the referral case. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-01-03 02:28:57 +08:00
if (mntroot->d_inode->i_op != NFS_SB(s)->nfs_client->rpc_ops->dir_inode_ops) {
kernel BUG at fs/nfs/namespace.c:108! - can be triggered by bad server Hi Trond, I have discovered that the BUG_ON in nfs_follow_mountpoint: BUG_ON(IS_ROOT(dentry)); can be triggered by a misbehaving server. What happens is the client does a lookup and discoveres that the named directory has a different fsid, so it initiates a mount. It then performs a GETATTR on the mounted directory and gets a different fsid again (due to a bug in the NFS server). This causes nfs_follow_mountpoint to be called on the newly mounted root, which triggers the BUG_ON. To duplicate this, have a directory which contains some mountpoints, and export that directory with the "crossmnt" flag using nfs-utils 1.1.1 (or 1.1.0 I think) The GETATTR on the root of the mounted filesystem will return the information for the top exportpoint, while a lookup will return the correct information. This difference causes the NFS client to BUG. I think the best way to fix this is to trap this possibility early, so just before completing the mount in the NFS client, check that it isn't going to use nfs_mountpoint_inode_operations. As long as i_op will never change once set (is that true?), this should be adequately safe. The following patch shows a possible approach, and it works for me. i.e. when the NFS server is misbehaving, I get ESTALE on those mountpoints, while when the NFS server is working correctly, I get correct behaviour on the client. NeilBrown Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-11-01 13:50:20 +08:00
dput(mntroot);
error = -ESTALE;
goto error_splat_super;
}
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
s->s_flags |= MS_ACTIVE;
mnt->mnt_sb = s;
mnt->mnt_root = mntroot;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: use new LSM interfaces to explicitly set mount options NFS and SELinux worked together previously because SELinux had NFS specific knowledge built in. This design was approved by both groups back in 2004 but the recent NFS changes to use nfs_parsed_mount_data and the usage of nfs_clone_mount_data showed this to be a poor fragile solution. This patch fixes the NFS functionality regression by making use of the new LSM interfaces to allow an FS to explicitly set its own mount options. The explicit setting of mount options is done in the nfs get_sb functions which are called before the generic vfs hooks try to set mount options for filesystems which use text mount data. This does not currently support NFSv4 as that functionality did not exist in previous kernels and thus there is no regression. I will be adding the needed code, which I believe to be the exact same as the v3 code, in nfs4_get_sb for 2.6.26. Signed-off-by: Eric Paris <eparis@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: James Morris <jmorris@namei.org>
2008-03-06 03:20:18 +08:00
/* clone any lsm security options from the parent to the new sb */
security_sb_clone_mnt_opts(data->sb, s);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
dprintk("<-- nfs_xdev_get_sb() = 0\n");
return 0;
NFS: Generalise the nfs_client structure Generalise the nfs_client structure by: (1) Moving nfs_client to a more general place (nfs_fs_sb.h). (2) Renaming its maintenance routines to be non-NFS4 specific. (3) Move those maintenance routines to a new non-NFS4 specific file (client.c) and move the declarations to internal.h. (4) Make nfs_find/get_client() take a full sockaddr_in to include the port number (will be required for NFS2/3). (5) Make nfs_find/get_client() take the NFS protocol version (again will be required to differentiate NFS2, 3 & 4 client records). Also: (6) Make nfs_client construction proceed akin to inodes, marking them as under construction and providing a function to indicate completion. (7) Make nfs_get_client() wait interruptibly if it finds a client that it can share, but that client is currently being constructed. (8) Make nfs4_create_client() use (6) and (7) instead of locking cl_sem. Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:10 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
out_err_nosb:
nfs_free_server(server);
out_err_noserver:
dprintk("<-- nfs_xdev_get_sb() = %d [error]\n", error);
return error;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
error_splat_super:
Convert obvious places to deactivate_locked_super() Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2009-05-06 13:34:22 +08:00
deactivate_locked_super(s);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
dprintk("<-- nfs_xdev_get_sb() = %d [splat]\n", error);
return error;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
#ifdef CONFIG_NFS_V4
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
/*
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
* Finish setting up a cloned NFS4 superblock
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
*/
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
static void nfs4_clone_super(struct super_block *sb,
const struct super_block *old_sb)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
sb->s_blocksize_bits = old_sb->s_blocksize_bits;
sb->s_blocksize = old_sb->s_blocksize;
sb->s_maxbytes = old_sb->s_maxbytes;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
sb->s_time_gran = 1;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
sb->s_op = old_sb->s_op;
nfs_initialise_sb(sb);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/*
* Set up an NFS4 superblock
*/
static void nfs4_fill_super(struct super_block *sb)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
sb->s_time_gran = 1;
sb->s_op = &nfs4_sops;
nfs_initialise_sb(sb);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFSv4: Fix the 'nolock' option regression NFSv4 should just ignore the 'nolock' option. It is an NFSv2/v3 thing... This fixes the Oops in http://bugzilla.kernel.org/show_bug.cgi?id=13330 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 04:22:58 +08:00
static void nfs4_validate_mount_flags(struct nfs_parsed_mount_data *args)
{
args->flags &= ~(NFS_MOUNT_NONLM|NFS_MOUNT_NOACL|NFS_MOUNT_VER3);
}
NFS: Introduce nfs4_validate_mount_options Refactor NFSv4 mount processing to break out mount data validation in the same way it's broken out in the NFSv2/v3 mount path. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:01 +08:00
/*
* Validate NFSv4 mount options
*/
NFS: use in-kernel mount argument structure for nfsv4 mounts The user-visible nfs4_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv4 in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:44:33 +08:00
static int nfs4_validate_mount_data(void *options,
struct nfs_parsed_mount_data *args,
const char *dev_name)
NFS: Introduce nfs4_validate_mount_options Refactor NFSv4 mount processing to break out mount data validation in the same way it's broken out in the NFSv2/v3 mount path. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:01 +08:00
{
NFS: Support non-IPv4 addresses in nfs_parsed_mount_data Replace the nfs_server and mount_server address fields in the nfs_parsed_mount_data structure with a "struct sockaddr_storage" instead of a "struct sockaddr_in". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:28 +08:00
struct sockaddr_in *ap;
NFS: use in-kernel mount argument structure for nfsv4 mounts The user-visible nfs4_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv4 in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:44:33 +08:00
struct nfs4_mount_data *data = (struct nfs4_mount_data *)options;
NFS: Introduce nfs4_validate_mount_options Refactor NFSv4 mount processing to break out mount data validation in the same way it's broken out in the NFSv2/v3 mount path. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:01 +08:00
char *c;
if (data == NULL)
goto out_no_data;
NFS: use in-kernel mount argument structure for nfsv4 mounts The user-visible nfs4_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv4 in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:44:33 +08:00
args->rsize = NFS_MAX_FILE_IO_SIZE;
args->wsize = NFS_MAX_FILE_IO_SIZE;
NFS: use documenting macro constants for initializing ac{reg, dir}{min, max} Clean up. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-27 05:47:12 +08:00
args->acregmin = NFS_DEF_ACREGMIN;
args->acregmax = NFS_DEF_ACREGMAX;
args->acdirmin = NFS_DEF_ACDIRMIN;
args->acdirmax = NFS_DEF_ACDIRMAX;
NFS: Save the value of the "port=" mount option During a remount based on the mount options displayed in /proc/mounts, we want to preserve the original behavior of the mount request. Let's save the original setting of the "port=" mount option in the mount's nfs_server structure. This allows us to simplify the default behavior of port setting for NFSv4 mounts: by default, NFSv2/3 mounts first try an RPC bind to determine the NFS server's port, unless the user specified the "port=" mount option; Users can force the client to skip the RPC bind by explicitly specifying "port=<value>". NFSv4, by contrast, assumes the NFS server port is 2049 and skips the RPC bind, unless the user specifies "port=". Users can force an RPC bind for NFSv4 by explicitly specifying "port=0". I added a couple of extra comments to clarify this behavior. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Miklos Szeredi <miklos@szeredi.hu> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-03-15 02:10:22 +08:00
args->nfs_server.port = NFS_PORT; /* 2049 unless user set port= */
NFS4: Set security flavor default for NFSv4 mounts like other defaults Set the default security flavor when we set the other mount option default values for NFSv4. This cleans up the NFSv4 mount option parsing path to look like the NFSv2/v3 one. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 04:33:54 +08:00
args->auth_flavors[0] = RPC_AUTH_UNIX;
args->auth_flavor_len = 0;
nfs41: add mount command option minorversion mount -t nfs4 -o minorversion=[0|1] specifies whether to use 4.0 or 4.1. By default, the minorversion is set to 0. Signed-off-by: Mike Sager <sager@netapp.com> [set default minorversion to 0 as per Trond and SteveD's request] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-04-01 21:21:48 +08:00
args->minorversion = 0;
NFS: use in-kernel mount argument structure for nfsv4 mounts The user-visible nfs4_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv4 in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:44:33 +08:00
NFS: Introduce nfs4_validate_mount_options Refactor NFSv4 mount processing to break out mount data validation in the same way it's broken out in the NFSv2/v3 mount path. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:01 +08:00
switch (data->version) {
case 1:
NFS: Support non-IPv4 addresses in nfs_parsed_mount_data Replace the nfs_server and mount_server address fields in the nfs_parsed_mount_data structure with a "struct sockaddr_storage" instead of a "struct sockaddr_in". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:28 +08:00
ap = (struct sockaddr_in *)&args->nfs_server.address;
if (data->host_addrlen > sizeof(args->nfs_server.address))
goto out_no_address;
if (data->host_addrlen == 0)
NFS: Introduce nfs4_validate_mount_options Refactor NFSv4 mount processing to break out mount data validation in the same way it's broken out in the NFSv2/v3 mount path. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:01 +08:00
goto out_no_address;
NFS: Support non-IPv4 addresses in nfs_parsed_mount_data Replace the nfs_server and mount_server address fields in the nfs_parsed_mount_data structure with a "struct sockaddr_storage" instead of a "struct sockaddr_in". Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-12-11 03:59:28 +08:00
args->nfs_server.addrlen = data->host_addrlen;
if (copy_from_user(ap, data->host_addr, data->host_addrlen))
NFS: Introduce nfs4_validate_mount_options Refactor NFSv4 mount processing to break out mount data validation in the same way it's broken out in the NFSv2/v3 mount path. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:01 +08:00
return -EFAULT;
NFS: use in-kernel mount argument structure for nfsv4 mounts The user-visible nfs4_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv4 in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:44:33 +08:00
if (!nfs_verify_server_address((struct sockaddr *)
&args->nfs_server.address))
NFS: Introduce nfs4_validate_mount_options Refactor NFSv4 mount processing to break out mount data validation in the same way it's broken out in the NFSv2/v3 mount path. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:01 +08:00
goto out_no_address;
NFS4: Set security flavor default for NFSv4 mounts like other defaults Set the default security flavor when we set the other mount option default values for NFSv4. This cleans up the NFSv4 mount option parsing path to look like the NFSv2/v3 one. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 04:33:54 +08:00
if (data->auth_flavourlen) {
if (data->auth_flavourlen > 1)
goto out_inval_auth;
NFSv4: Fix a bug in nfs4_validate_mount_data() The previous patch introduced a bug when copying the server address. Also clarify a copy into the auth_flavours array: currently the two size calculations are equivalent, but we may decide to change the size of auth_flavors[] at some point. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-21 08:23:51 +08:00
if (copy_from_user(&args->auth_flavors[0],
NFS: use in-kernel mount argument structure for nfsv4 mounts The user-visible nfs4_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv4 in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:44:33 +08:00
data->auth_flavours,
NFSv4: Fix a bug in nfs4_validate_mount_data() The previous patch introduced a bug when copying the server address. Also clarify a copy into the auth_flavours array: currently the two size calculations are equivalent, but we may decide to change the size of auth_flavors[] at some point. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-21 08:23:51 +08:00
sizeof(args->auth_flavors[0])))
NFS: Introduce nfs4_validate_mount_options Refactor NFSv4 mount processing to break out mount data validation in the same way it's broken out in the NFSv2/v3 mount path. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:01 +08:00
return -EFAULT;
}
c = strndup_user(data->hostname.data, NFS4_MAXNAMLEN);
if (IS_ERR(c))
return PTR_ERR(c);
NFS: use in-kernel mount argument structure for nfsv4 mounts The user-visible nfs4_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv4 in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:44:33 +08:00
args->nfs_server.hostname = c;
NFS: Introduce nfs4_validate_mount_options Refactor NFSv4 mount processing to break out mount data validation in the same way it's broken out in the NFSv2/v3 mount path. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:01 +08:00
c = strndup_user(data->mnt_path.data, NFS4_MAXPATHLEN);
if (IS_ERR(c))
return PTR_ERR(c);
NFS: use in-kernel mount argument structure for nfsv4 mounts The user-visible nfs4_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv4 in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:44:33 +08:00
args->nfs_server.export_path = c;
dfprintk(MOUNT, "NFS: MNTPATH: '%s'\n", c);
NFS: Introduce nfs4_validate_mount_options Refactor NFSv4 mount processing to break out mount data validation in the same way it's broken out in the NFSv2/v3 mount path. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:01 +08:00
c = strndup_user(data->client_addr.data, 16);
if (IS_ERR(c))
return PTR_ERR(c);
NFS: use in-kernel mount argument structure for nfsv4 mounts The user-visible nfs4_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv4 in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:44:33 +08:00
args->client_address = c;
/*
* Translate to nfs_parsed_mount_data, which nfs4_fill_super
* can deal with.
*/
args->flags = data->flags & NFS4_MOUNT_FLAGMASK;
args->rsize = data->rsize;
args->wsize = data->wsize;
args->timeo = data->timeo;
args->retrans = data->retrans;
args->acregmin = data->acregmin;
args->acregmax = data->acregmax;
args->acdirmin = data->acdirmin;
args->acdirmax = data->acdirmax;
args->nfs_server.protocol = data->proto;
NFS: set transport defaults after mount option parsing is finished Move the UDP/TCP default timeo/retrans settings for text mounts to nfs_init_timeout_values(), which was were they were always being initialised (and sanity checked) for binary mounts. Document the default timeout values using appropriate #defines. Ensure that we initialise and sanity check the transport protocols that may have been specified by the user. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-03 02:43:47 +08:00
nfs_validate_transport_protocol(args);
NFS: Introduce nfs4_validate_mount_options Refactor NFSv4 mount processing to break out mount data validation in the same way it's broken out in the NFSv2/v3 mount path. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:01 +08:00
break;
NFS: Add support for mounting NFSv4 file systems with string options Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:59 +08:00
default: {
NFS: Use common device name parsing logic for NFSv4 and NFSv2/v3 To support passing a raw IPv6 address as a server hostname, we need to expand the logic that handles splitting the passed-in device name into a server hostname and export path Start by pulling device name parsing out of the mount option validation functions and into separate helper functions. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:37 +08:00
int status;
NFS: use in-kernel mount argument structure for nfsv4 mounts The user-visible nfs4_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv4 in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:44:33 +08:00
if (nfs_parse_mount_options((char *)options, args) == 0)
NFS: Add support for mounting NFSv4 file systems with string options Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:59 +08:00
return -EINVAL;
if (!nfs_verify_server_address((struct sockaddr *)
NFS: use in-kernel mount argument structure for nfsv4 mounts The user-visible nfs4_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv4 in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:44:33 +08:00
&args->nfs_server.address))
NFS: Add support for mounting NFSv4 file systems with string options Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:59 +08:00
return -EINVAL;
NFS: Move the nfs_set_port() call out of nfs_parse_mount_options() The remount path does not need to set the port in the server address. Since it's not really a part of option parsing, move the nfs_set_port() call to nfs_parse_mount_options()'s callers. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-27 05:47:05 +08:00
nfs_set_port((struct sockaddr *)&args->nfs_server.address,
args->nfs_server.port);
NFS: set transport defaults after mount option parsing is finished Move the UDP/TCP default timeo/retrans settings for text mounts to nfs_init_timeout_values(), which was were they were always being initialised (and sanity checked) for binary mounts. Document the default timeout values using appropriate #defines. Ensure that we initialise and sanity check the transport protocols that may have been specified by the user. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-07-03 02:43:47 +08:00
nfs_validate_transport_protocol(args);
NFSv4: Fix the 'nolock' option regression NFSv4 should just ignore the 'nolock' option. It is an NFSv2/v3 thing... This fixes the Oops in http://bugzilla.kernel.org/show_bug.cgi?id=13330 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-06-18 04:22:58 +08:00
nfs4_validate_mount_flags(args);
NFS4: Set security flavor default for NFSv4 mounts like other defaults Set the default security flavor when we set the other mount option default values for NFSv4. This cleans up the NFSv4 mount option parsing path to look like the NFSv2/v3 one. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-25 04:33:54 +08:00
if (args->auth_flavor_len > 1)
NFS: Add support for mounting NFSv4 file systems with string options Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:59 +08:00
goto out_inval_auth;
NFS: use in-kernel mount argument structure for nfsv4 mounts The user-visible nfs4_mount_data does not contain sufficient data to describe new mount options, and also is now a legacy structure. Replace it with the internal nfs_parsed_mount_data for nfsv4 in-kernel use. Signed-off-by: Tom Talpey <tmt@netapp.com> Acked-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-09-11 01:44:33 +08:00
if (args->client_address == NULL)
NFSv4: handle lack of clientaddr in option string If a NFSv4 mount is attempted with string based options, and the option string doesn't contain a clientaddr= option, the kernel will currently oops. Check for this situation and return a proper error. Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-18 23:28:43 +08:00
goto out_no_client_address;
NFS: Use common device name parsing logic for NFSv4 and NFSv2/v3 To support passing a raw IPv6 address as a server hostname, we need to expand the logic that handles splitting the passed-in device name into a server hostname and export path Start by pulling device name parsing out of the mount option validation functions and into separate helper functions. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-24 00:36:37 +08:00
status = nfs_parse_devname(dev_name,
&args->nfs_server.hostname,
NFS4_MAXNAMLEN,
&args->nfs_server.export_path,
NFS4_MAXPATHLEN);
if (status < 0)
return status;
NFS: Add support for mounting NFSv4 file systems with string options Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:59 +08:00
break;
}
NFS: Introduce nfs4_validate_mount_options Refactor NFSv4 mount processing to break out mount data validation in the same way it's broken out in the NFSv2/v3 mount path. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:01 +08:00
}
return 0;
out_no_data:
dfprintk(MOUNT, "NFS4: mount program didn't pass any mount data\n");
return -EINVAL;
out_inval_auth:
dfprintk(MOUNT, "NFS4: Invalid number of RPC auth flavours %d\n",
data->auth_flavourlen);
return -EINVAL;
out_no_address:
dfprintk(MOUNT, "NFS4: mount program didn't pass remote address\n");
return -EINVAL;
NFSv4: handle lack of clientaddr in option string If a NFSv4 mount is attempted with string based options, and the option string doesn't contain a clientaddr= option, the kernel will currently oops. Check for this situation and return a proper error. Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-18 23:28:43 +08:00
out_no_client_address:
dfprintk(MOUNT, "NFS4: mount program didn't pass callback address\n");
return -EINVAL;
NFS: Introduce nfs4_validate_mount_options Refactor NFSv4 mount processing to break out mount data validation in the same way it's broken out in the NFSv2/v3 mount path. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:13:01 +08:00
}
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/*
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace As noted in the previous patch, the NFSv4 client mount code currently has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in nfs4_path_walk() will fail with an error. This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the namespace on the server, then uses the ordinary VFS and NFS path lookup code to walk down the mount path in that namespace. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-23 03:09:14 +08:00
* Get the superblock for the NFS4 root partition
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
*/
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace As noted in the previous patch, the NFSv4 client mount code currently has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in nfs4_path_walk() will fail with an error. This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the namespace on the server, then uses the ordinary VFS and NFS path lookup code to walk down the mount path in that namespace. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-23 03:09:14 +08:00
static int nfs4_remote_get_sb(struct file_system_type *fs_type,
Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ Conflicts: fs/nfs/inode.c fs/super.c Fix conflicts between patch 'NFS: Split fs/nfs/inode.c' and patch 'VFS: Permit filesystem to override root dentry on mount'
2006-06-24 20:41:41 +08:00
int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace As noted in the previous patch, the NFSv4 client mount code currently has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in nfs4_path_walk() will fail with an error. This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the namespace on the server, then uses the ordinary VFS and NFS path lookup code to walk down the mount path in that namespace. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-23 03:09:14 +08:00
struct nfs_parsed_mount_data *data = raw_data;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
struct super_block *s;
struct nfs_server *server;
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
struct nfs_fh *mntfh;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
struct dentry *mntroot;
NFS: Add the mount option "nosharecache" Prior to David Howell's mount changes in 2.6.18, users who mounted different directories which happened to be from the same filesystem on the server would get different super blocks, and hence could choose different mount options. As long as there were no hard linked files that crossed from one subtree to another, this was quite safe. Post the changes, if the two directories are on the same filesystem (have the same 'fsid'), they will share the same super block, and hence the same mount options. Add a flag to allow users to elect not to share the NFS super block with another mount point, even if the fsids are the same. This will allow users to set different mount options for the two different super blocks, as was previously possible. It is still up to the user to ensure that there are no cache coherency issues when doing this, however the default behaviour will be to share super blocks whenever two paths result in the same fsid. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
int (*compare_super)(struct super_block *, void *) = nfs_compare_super;
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
struct nfs_sb_mountdata sb_mntdata = {
.mntflags = flags,
};
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
int error = -ENOMEM;
mntfh = kzalloc(sizeof(*mntfh), GFP_KERNEL);
if (data == NULL || mntfh == NULL)
goto out_free_fh;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
security_init_mnt_opts(&data->lsm_opts);
NFS: use new LSM interfaces to explicitly set mount options NFS and SELinux worked together previously because SELinux had NFS specific knowledge built in. This design was approved by both groups back in 2004 but the recent NFS changes to use nfs_parsed_mount_data and the usage of nfs_clone_mount_data showed this to be a poor fragile solution. This patch fixes the NFS functionality regression by making use of the new LSM interfaces to allow an FS to explicitly set its own mount options. The explicit setting of mount options is done in the nfs get_sb functions which are called before the generic vfs hooks try to set mount options for filesystems which use text mount data. This does not currently support NFSv4 as that functionality did not exist in previous kernels and thus there is no regression. I will be adding the needed code, which I believe to be the exact same as the v3 code, in nfs4_get_sb for 2.6.26. Signed-off-by: Eric Paris <eparis@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: James Morris <jmorris@namei.org>
2008-03-06 03:20:18 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/* Get a volume representation */
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
server = nfs4_create_server(data, mntfh);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (IS_ERR(server)) {
error = PTR_ERR(server);
NFS: Clean-up: Replace nfs_copy_user_string with strndup_user The new string utility function strndup_user can be used instead of nfs_copy_user_string, eliminating an unnecessary duplication of function. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:30 +08:00
goto out;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
sb_mntdata.server = server;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Add the mount option "nosharecache" Prior to David Howell's mount changes in 2.6.18, users who mounted different directories which happened to be from the same filesystem on the server would get different super blocks, and hence could choose different mount options. As long as there were no hard linked files that crossed from one subtree to another, this was quite safe. Post the changes, if the two directories are on the same filesystem (have the same 'fsid'), they will share the same super block, and hence the same mount options. Add a flag to allow users to elect not to share the NFS super block with another mount point, even if the fsids are the same. This will allow users to set different mount options for the two different super blocks, as was previously possible. It is still up to the user to ensure that there are no cache coherency issues when doing this, however the default behaviour will be to share super blocks whenever two paths result in the same fsid. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
if (server->flags & NFS4_MOUNT_UNSHARED)
compare_super = NULL;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/* Get a superblock - note that we may end up sharing one that already exists */
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace As noted in the previous patch, the NFSv4 client mount code currently has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in nfs4_path_walk() will fail with an error. This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the namespace on the server, then uses the ordinary VFS and NFS path lookup code to walk down the mount path in that namespace. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-23 03:09:14 +08:00
s = sget(&nfs4_fs_type, compare_super, nfs_set_super, &sb_mntdata);
Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ Conflicts: fs/nfs/inode.c fs/super.c Fix conflicts between patch 'NFS: Split fs/nfs/inode.c' and patch 'VFS: Permit filesystem to override root dentry on mount'
2006-06-24 20:41:41 +08:00
if (IS_ERR(s)) {
error = PTR_ERR(s);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
goto out_free;
Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ Conflicts: fs/nfs/inode.c fs/super.c Fix conflicts between patch 'NFS: Split fs/nfs/inode.c' and patch 'VFS: Permit filesystem to override root dentry on mount'
2006-06-24 20:41:41 +08:00
}
NFSv4: Fix a use-after-free issue with the nfs server. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-24 13:03:05 +08:00
if (s->s_fs_info != server) {
nfs_free_server(server);
server = NULL;
mm: bdi: expose the BDI object in sysfs for NFS Register NFS' backing_dev_info under sysfs with the name "nfs-MAJOR:MINOR" Signed-off-by: Miklos Szeredi <mszeredi@suse.cz> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-30 15:54:33 +08:00
} else {
error = nfs_bdi_register(server);
if (error)
goto error_splat_super;
NFSv4: Fix a use-after-free issue with the nfs server. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-24 13:03:05 +08:00
}
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (!s->s_root) {
/* initial superblock/root creation */
nfs4_fill_super(s);
NFS: Define and create superblock-level objects Define and create superblock-level cache index objects (as managed by nfs_server structs). Each superblock object is created in a server level index object and is itself an index into which inode-level objects are inserted. Ideally there would be one superblock-level object per server, and the former would be folded into the latter; however, since the "nosharecache" option exists this isn't possible. The superblock object key is a sequence consisting of: (1) Certain superblock s_flags. (2) Various connection parameters that serve to distinguish superblocks for sget(). (3) The volume FSID. (4) The security flavour. (5) The uniquifier length. (6) The uniquifier text. This is normally an empty string, unless the fsc=xyz mount option was used to explicitly specify a uniquifier. The key blob is of variable length, depending on the length of (6). The superblock object is given no coherency data to carry in the auxiliary data permitted by the cache. It is assumed that the superblock is always coherent. This patch also adds uniquification handling such that two otherwise identical superblocks, at least one of which is marked "nosharecache", won't end up trying to share the on-disk cache. It will be possible to manually provide a uniquifier through a mount option with a later patch to avoid the error otherwise produced. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Steve Dickson <steved@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Daire Byrne <Daire.Byrne@framestore.com>
2009-04-03 23:42:42 +08:00
nfs_fscache_get_super_cookie(s, data);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
}
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
mntroot = nfs4_get_root(s, mntfh);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (IS_ERR(mntroot)) {
error = PTR_ERR(mntroot);
goto error_splat_super;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
error = security_sb_set_mnt_opts(s, &data->lsm_opts);
nfs/lsm: make NFSv4 set LSM mount options NFSv3 get_sb operations call into the LSM layer to set security options passed from userspace. NFSv4 hooks were not originally added since it was reasonably late in the merge window and NFSv3 was the only thing that had regressed (v4 has never supported any LSM options) This patch makes NFSv4 call into the LSM to set security options rather than just blindly dropping them with no notice to the user as happens today. This patch was tested in a simple NFSv4 environment with the context= option and appeared to work as expected. Signed-off-by: Eric Paris <eparis@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Stephen Smalley <sds@tycho.nsa.gov> Acked-by: James Morris <jmorris@namei.org> Cc: Casey Schaufler <casey@schaufler-ca.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-05-03 04:42:42 +08:00
if (error)
goto error_splat_root;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
s->s_flags |= MS_ACTIVE;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
mnt->mnt_sb = s;
mnt->mnt_root = mntroot;
NFS: Clean-up: Replace nfs_copy_user_string with strndup_user The new string utility function strndup_user can be used instead of nfs_copy_user_string, eliminating an unnecessary duplication of function. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:30 +08:00
error = 0;
out:
NFS: Reduce the NFS mount code stack usage. This appears to fix the Oops reported in http://bugzilla.kernel.org/show_bug.cgi?id=10826 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-06-20 02:20:11 +08:00
security_free_mnt_opts(&data->lsm_opts);
out_free_fh:
kfree(mntfh);
NFS: Clean-up: Replace nfs_copy_user_string with strndup_user The new string utility function strndup_user can be used instead of nfs_copy_user_string, eliminating an unnecessary duplication of function. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:30 +08:00
return error;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
out_free:
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
nfs_free_server(server);
NFS: Clean-up: Replace nfs_copy_user_string with strndup_user The new string utility function strndup_user can be used instead of nfs_copy_user_string, eliminating an unnecessary duplication of function. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:30 +08:00
goto out;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
nfs/lsm: make NFSv4 set LSM mount options NFSv3 get_sb operations call into the LSM layer to set security options passed from userspace. NFSv4 hooks were not originally added since it was reasonably late in the merge window and NFSv3 was the only thing that had regressed (v4 has never supported any LSM options) This patch makes NFSv4 call into the LSM to set security options rather than just blindly dropping them with no notice to the user as happens today. This patch was tested in a simple NFSv4 environment with the context= option and appeared to work as expected. Signed-off-by: Eric Paris <eparis@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Stephen Smalley <sds@tycho.nsa.gov> Acked-by: James Morris <jmorris@namei.org> Cc: Casey Schaufler <casey@schaufler-ca.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-05-03 04:42:42 +08:00
error_splat_root:
dput(mntroot);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
error_splat_super:
Convert obvious places to deactivate_locked_super() Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2009-05-06 13:34:22 +08:00
deactivate_locked_super(s);
NFS: Clean-up: Replace nfs_copy_user_string with strndup_user The new string utility function strndup_user can be used instead of nfs_copy_user_string, eliminating an unnecessary duplication of function. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-07-02 00:12:30 +08:00
goto out;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace As noted in the previous patch, the NFSv4 client mount code currently has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in nfs4_path_walk() will fail with an error. This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the namespace on the server, then uses the ordinary VFS and NFS path lookup code to walk down the mount path in that namespace. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-23 03:09:14 +08:00
static struct vfsmount *nfs_do_root_mount(struct file_system_type *fs_type,
int flags, void *data, const char *hostname)
{
struct vfsmount *root_mnt;
char *root_devname;
size_t len;
len = strlen(hostname) + 3;
root_devname = kmalloc(len, GFP_KERNEL);
if (root_devname == NULL)
return ERR_PTR(-ENOMEM);
snprintf(root_devname, len, "%s:/", hostname);
root_mnt = vfs_kern_mount(fs_type, flags, root_devname, data);
kfree(root_devname);
return root_mnt;
}
static int nfs_follow_remote_path(struct vfsmount *root_mnt,
const char *export_path, struct vfsmount *mnt_target)
{
struct mnt_namespace *ns_private;
struct nameidata nd;
struct super_block *s;
int ret;
ns_private = create_mnt_ns(root_mnt);
ret = PTR_ERR(ns_private);
if (IS_ERR(ns_private))
goto out_mntput;
ret = vfs_path_lookup(root_mnt->mnt_root, root_mnt,
export_path, LOOKUP_FOLLOW, &nd);
put_mnt_ns(ns_private);
if (ret != 0)
goto out_err;
s = nd.path.mnt->mnt_sb;
atomic_inc(&s->s_active);
mnt_target->mnt_sb = s;
mnt_target->mnt_root = dget(nd.path.dentry);
path_put(&nd.path);
down_write(&s->s_umount);
return 0;
out_mntput:
mntput(root_mnt);
out_err:
return ret;
}
/*
* Get the superblock for an NFS4 mountpoint
*/
static int nfs4_get_sb(struct file_system_type *fs_type,
int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt)
{
struct nfs_parsed_mount_data *data;
char *export_path;
struct vfsmount *root_mnt;
int error = -ENOMEM;
data = kzalloc(sizeof(*data), GFP_KERNEL);
if (data == NULL)
goto out_free_data;
/* Validate the mount data */
error = nfs4_validate_mount_data(raw_data, data, dev_name);
if (error < 0)
goto out;
export_path = data->nfs_server.export_path;
data->nfs_server.export_path = "/";
root_mnt = nfs_do_root_mount(&nfs4_remote_fs_type, flags, data,
data->nfs_server.hostname);
data->nfs_server.export_path = export_path;
error = PTR_ERR(root_mnt);
if (IS_ERR(root_mnt))
goto out;
error = nfs_follow_remote_path(root_mnt, export_path, mnt);
out:
kfree(data->client_address);
kfree(data->nfs_server.export_path);
kfree(data->nfs_server.hostname);
kfree(data->fscache_uniq);
out_free_data:
kfree(data);
dprintk("<-- nfs4_get_sb() = %d%s\n", error,
error != 0 ? " [error]" : "");
return error;
}
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
static void nfs4_kill_super(struct super_block *sb)
{
struct nfs_server *server = NFS_SB(sb);
nfs41: destroy_session operation Implement the destroy_session operation conforming to http://tools.ietf.org/html/draft-ietf-nfsv4-minorversion1-26 Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Andy Adamson<andros@umich.edu> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: remove extraneous rpc_clnt pointer] Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41; NFS_CS_READY required for DESTROY_SESSION] Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: pass *session in seq_args and seq_res] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> [nfs41: fix encode_destroy_session's xdr Xcoding pointer type] Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 21:22:34 +08:00
dprintk("--> %s\n", __func__);
NFSv4: Clean up the support for returning multiple delegations Add a flag to mark delegations as requiring return, then run a garbage collector. In the future, this will allow for more flexible delegation management, where delegations may be marked for return if it turns out that they are not being referenced. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-12-24 04:21:46 +08:00
nfs_super_return_all_delegations(sb);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
kill_anon_super(sb);
nfs4_renewd_prepare_shutdown(server);
NFS: Define and create superblock-level objects Define and create superblock-level cache index objects (as managed by nfs_server structs). Each superblock object is created in a server level index object and is itself an index into which inode-level objects are inserted. Ideally there would be one superblock-level object per server, and the former would be folded into the latter; however, since the "nosharecache" option exists this isn't possible. The superblock object key is a sequence consisting of: (1) Certain superblock s_flags. (2) Various connection parameters that serve to distinguish superblocks for sget(). (3) The volume FSID. (4) The security flavour. (5) The uniquifier length. (6) The uniquifier text. This is normally an empty string, unless the fsc=xyz mount option was used to explicitly specify a uniquifier. The key blob is of variable length, depending on the length of (6). The superblock object is given no coherency data to carry in the auxiliary data permitted by the cache. It is assumed that the superblock is always coherent. This patch also adds uniquification handling such that two otherwise identical superblocks, at least one of which is marked "nosharecache", won't end up trying to share the on-disk cache. It will be possible to manually provide a uniquifier through a mount option with a later patch to avoid the error otherwise produced. Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Steve Dickson <steved@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Daire Byrne <Daire.Byrne@framestore.com>
2009-04-03 23:42:42 +08:00
nfs_fscache_release_super_cookie(sb);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
nfs_free_server(server);
nfs41: destroy_session operation Implement the destroy_session operation conforming to http://tools.ietf.org/html/draft-ietf-nfsv4-minorversion1-26 Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Andy Adamson<andros@umich.edu> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: remove extraneous rpc_clnt pointer] Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41; NFS_CS_READY required for DESTROY_SESSION] Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: pass *session in seq_args and seq_res] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> [nfs41: fix encode_destroy_session's xdr Xcoding pointer type] Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 21:22:34 +08:00
dprintk("<-- %s\n", __func__);
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
/*
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
* Clone an NFS4 server record on xdev traversal (FSID-change)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
*/
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
static int nfs4_xdev_get_sb(struct file_system_type *fs_type, int flags,
const char *dev_name, void *raw_data,
struct vfsmount *mnt)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
struct nfs_clone_mount *data = raw_data;
struct super_block *s;
struct nfs_server *server;
struct dentry *mntroot;
NFS: Add the mount option "nosharecache" Prior to David Howell's mount changes in 2.6.18, users who mounted different directories which happened to be from the same filesystem on the server would get different super blocks, and hence could choose different mount options. As long as there were no hard linked files that crossed from one subtree to another, this was quite safe. Post the changes, if the two directories are on the same filesystem (have the same 'fsid'), they will share the same super block, and hence the same mount options. Add a flag to allow users to elect not to share the NFS super block with another mount point, even if the fsids are the same. This will allow users to set different mount options for the two different super blocks, as was previously possible. It is still up to the user to ensure that there are no cache coherency issues when doing this, however the default behaviour will be to share super blocks whenever two paths result in the same fsid. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
int (*compare_super)(struct super_block *, void *) = nfs_compare_super;
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
struct nfs_sb_mountdata sb_mntdata = {
.mntflags = flags,
};
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
int error;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
dprintk("--> nfs4_xdev_get_sb()\n");
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/* create a new volume representation */
server = nfs_clone_server(NFS_SB(data->sb), data->fh, data->fattr);
if (IS_ERR(server)) {
error = PTR_ERR(server);
goto out_err_noserver;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
sb_mntdata.server = server;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Add the mount option "nosharecache" Prior to David Howell's mount changes in 2.6.18, users who mounted different directories which happened to be from the same filesystem on the server would get different super blocks, and hence could choose different mount options. As long as there were no hard linked files that crossed from one subtree to another, this was quite safe. Post the changes, if the two directories are on the same filesystem (have the same 'fsid'), they will share the same super block, and hence the same mount options. Add a flag to allow users to elect not to share the NFS super block with another mount point, even if the fsids are the same. This will allow users to set different mount options for the two different super blocks, as was previously possible. It is still up to the user to ensure that there are no cache coherency issues when doing this, however the default behaviour will be to share super blocks whenever two paths result in the same fsid. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
if (server->flags & NFS4_MOUNT_UNSHARED)
compare_super = NULL;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/* Get a superblock - note that we may end up sharing one that already exists */
NFS: use correct fs type for v4 submounts and referrals Signed-off-by: Andy Adamson<andros@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-10-17 22:44:37 +08:00
s = sget(&nfs4_fs_type, compare_super, nfs_set_super, &sb_mntdata);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (IS_ERR(s)) {
error = PTR_ERR(s);
goto out_err_nosb;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (s->s_fs_info != server) {
nfs_free_server(server);
server = NULL;
mm: bdi: expose the BDI object in sysfs for NFS Register NFS' backing_dev_info under sysfs with the name "nfs-MAJOR:MINOR" Signed-off-by: Miklos Szeredi <mszeredi@suse.cz> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-30 15:54:33 +08:00
} else {
error = nfs_bdi_register(server);
if (error)
goto error_splat_super;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
}
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (!s->s_root) {
/* initial superblock/root creation */
nfs4_clone_super(s, data->sb);
}
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
mntroot = nfs4_get_root(s, data->fh);
if (IS_ERR(mntroot)) {
error = PTR_ERR(mntroot);
goto error_splat_super;
}
NFS: Fix a possible Oops in fs/nfs/super.c Sigh... commit 4584f520e1f773082ef44ff4f8969a5d992b16ec (NFS: Fix NFS mountpoint crossing...) had a slight flaw: server can be NULL if sget() returned an existing superblock. Fix the fix by dereferencing s->s_fs_info. Thanks to Coverity/Adrian Bunk and Frank Filz for spotting the bug. (See http://bugzilla.kernel.org/show_bug.cgi?id=9647) Also add in the same namespace Oops fix for NFSv4 in both the mountpoint crossing case, and the referral case. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-01-03 02:28:57 +08:00
if (mntroot->d_inode->i_op != NFS_SB(s)->nfs_client->rpc_ops->dir_inode_ops) {
dput(mntroot);
error = -ESTALE;
goto error_splat_super;
}
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
s->s_flags |= MS_ACTIVE;
mnt->mnt_sb = s;
mnt->mnt_root = mntroot;
nfs/lsm: make NFSv4 set LSM mount options NFSv3 get_sb operations call into the LSM layer to set security options passed from userspace. NFSv4 hooks were not originally added since it was reasonably late in the merge window and NFSv3 was the only thing that had regressed (v4 has never supported any LSM options) This patch makes NFSv4 call into the LSM to set security options rather than just blindly dropping them with no notice to the user as happens today. This patch was tested in a simple NFSv4 environment with the context= option and appeared to work as expected. Signed-off-by: Eric Paris <eparis@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Stephen Smalley <sds@tycho.nsa.gov> Acked-by: James Morris <jmorris@namei.org> Cc: Casey Schaufler <casey@schaufler-ca.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-05-03 04:42:42 +08:00
security_sb_clone_mnt_opts(data->sb, s);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
dprintk("<-- nfs4_xdev_get_sb() = 0\n");
return 0;
out_err_nosb:
nfs_free_server(server);
out_err_noserver:
dprintk("<-- nfs4_xdev_get_sb() = %d [error]\n", error);
return error;
error_splat_super:
Convert obvious places to deactivate_locked_super() Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2009-05-06 13:34:22 +08:00
deactivate_locked_super(s);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
dprintk("<-- nfs4_xdev_get_sb() = %d [splat]\n", error);
return error;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace As noted in the previous patch, the NFSv4 client mount code currently has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in nfs4_path_walk() will fail with an error. This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the namespace on the server, then uses the ordinary VFS and NFS path lookup code to walk down the mount path in that namespace. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-23 03:09:14 +08:00
static int nfs4_remote_referral_get_sb(struct file_system_type *fs_type,
int flags, const char *dev_name, void *raw_data,
struct vfsmount *mnt)
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
{
struct nfs_clone_mount *data = raw_data;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
struct super_block *s;
struct nfs_server *server;
struct dentry *mntroot;
struct nfs_fh mntfh;
NFS: Add the mount option "nosharecache" Prior to David Howell's mount changes in 2.6.18, users who mounted different directories which happened to be from the same filesystem on the server would get different super blocks, and hence could choose different mount options. As long as there were no hard linked files that crossed from one subtree to another, this was quite safe. Post the changes, if the two directories are on the same filesystem (have the same 'fsid'), they will share the same super block, and hence the same mount options. Add a flag to allow users to elect not to share the NFS super block with another mount point, even if the fsids are the same. This will allow users to set different mount options for the two different super blocks, as was previously possible. It is still up to the user to ensure that there are no cache coherency issues when doing this, however the default behaviour will be to share super blocks whenever two paths result in the same fsid. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
int (*compare_super)(struct super_block *, void *) = nfs_compare_super;
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
struct nfs_sb_mountdata sb_mntdata = {
.mntflags = flags,
};
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
int error;
dprintk("--> nfs4_referral_get_sb()\n");
/* create a new volume representation */
server = nfs4_create_referral_server(data, &mntfh);
if (IS_ERR(server)) {
error = PTR_ERR(server);
goto out_err_noserver;
}
NFS: Fix the mount regression This avoids the recent NFS mount regression (returning EBUSY when mounting the same filesystem twice with different parameters). The best I can do given the constraints appears to be to have the kernel first look for a superblock that matches both the fsid and the user-specified mount options, and then spawn off a new superblock if that search fails. Note that this is not the same as specifying nosharecache everywhere since nosharecache will never attempt to match an existing superblock. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Tested-by: Hua Zhong <hzhong@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-08-31 22:45:17 +08:00
sb_mntdata.server = server;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
NFS: Add the mount option "nosharecache" Prior to David Howell's mount changes in 2.6.18, users who mounted different directories which happened to be from the same filesystem on the server would get different super blocks, and hence could choose different mount options. As long as there were no hard linked files that crossed from one subtree to another, this was quite safe. Post the changes, if the two directories are on the same filesystem (have the same 'fsid'), they will share the same super block, and hence the same mount options. Add a flag to allow users to elect not to share the NFS super block with another mount point, even if the fsids are the same. This will allow users to set different mount options for the two different super blocks, as was previously possible. It is still up to the user to ensure that there are no cache coherency issues when doing this, however the default behaviour will be to share super blocks whenever two paths result in the same fsid. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-05-17 04:53:28 +08:00
if (server->flags & NFS4_MOUNT_UNSHARED)
compare_super = NULL;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
/* Get a superblock - note that we may end up sharing one that already exists */
NFS: use correct fs type for v4 submounts and referrals Signed-off-by: Andy Adamson<andros@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-10-17 22:44:37 +08:00
s = sget(&nfs4_fs_type, compare_super, nfs_set_super, &sb_mntdata);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (IS_ERR(s)) {
error = PTR_ERR(s);
goto out_err_nosb;
}
if (s->s_fs_info != server) {
nfs_free_server(server);
server = NULL;
mm: bdi: expose the BDI object in sysfs for NFS Register NFS' backing_dev_info under sysfs with the name "nfs-MAJOR:MINOR" Signed-off-by: Miklos Szeredi <mszeredi@suse.cz> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-30 15:54:33 +08:00
} else {
error = nfs_bdi_register(server);
if (error)
goto error_splat_super;
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
}
if (!s->s_root) {
/* initial superblock/root creation */
nfs4_fill_super(s);
}
NFSv4: Fix Oops in nfs4_create_referral_server The filehandle that is passed into nfs4_create_referral_server is not initialised. The expectation is that nfs4_create_referral_server will initialise it, and return it to the caller. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2007-02-03 06:46:09 +08:00
mntroot = nfs4_get_root(s, &mntfh);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
if (IS_ERR(mntroot)) {
error = PTR_ERR(mntroot);
goto error_splat_super;
}
NFS: Fix a possible Oops in fs/nfs/super.c Sigh... commit 4584f520e1f773082ef44ff4f8969a5d992b16ec (NFS: Fix NFS mountpoint crossing...) had a slight flaw: server can be NULL if sget() returned an existing superblock. Fix the fix by dereferencing s->s_fs_info. Thanks to Coverity/Adrian Bunk and Frank Filz for spotting the bug. (See http://bugzilla.kernel.org/show_bug.cgi?id=9647) Also add in the same namespace Oops fix for NFSv4 in both the mountpoint crossing case, and the referral case. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-01-03 02:28:57 +08:00
if (mntroot->d_inode->i_op != NFS_SB(s)->nfs_client->rpc_ops->dir_inode_ops) {
dput(mntroot);
error = -ESTALE;
goto error_splat_super;
}
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
s->s_flags |= MS_ACTIVE;
mnt->mnt_sb = s;
mnt->mnt_root = mntroot;
nfs/lsm: make NFSv4 set LSM mount options NFSv3 get_sb operations call into the LSM layer to set security options passed from userspace. NFSv4 hooks were not originally added since it was reasonably late in the merge window and NFSv3 was the only thing that had regressed (v4 has never supported any LSM options) This patch makes NFSv4 call into the LSM to set security options rather than just blindly dropping them with no notice to the user as happens today. This patch was tested in a simple NFSv4 environment with the context= option and appeared to work as expected. Signed-off-by: Eric Paris <eparis@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Stephen Smalley <sds@tycho.nsa.gov> Acked-by: James Morris <jmorris@namei.org> Cc: Casey Schaufler <casey@schaufler-ca.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2008-05-03 04:42:42 +08:00
security_sb_clone_mnt_opts(data->sb, s);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
dprintk("<-- nfs4_referral_get_sb() = 0\n");
return 0;
out_err_nosb:
nfs_free_server(server);
out_err_noserver:
dprintk("<-- nfs4_referral_get_sb() = %d [error]\n", error);
return error;
error_splat_super:
Convert obvious places to deactivate_locked_super() Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2009-05-06 13:34:22 +08:00
deactivate_locked_super(s);
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
dprintk("<-- nfs4_referral_get_sb() = %d [splat]\n", error);
return error;
NFS: Split fs/nfs/inode.c As fs/nfs/inode.c is rather large, heterogenous and unwieldy, the attached patch splits it up into a number of files: (*) fs/nfs/inode.c Strictly inode specific functions. (*) fs/nfs/super.c Superblock management functions for NFS and NFS4, normal access, clones and referrals. The NFS4 superblock functions _could_ move out into a separate conditionally compiled file, but it's probably not worth it as there're so many common bits. (*) fs/nfs/namespace.c Some namespace-specific functions have been moved here. (*) fs/nfs/nfs4namespace.c NFS4-specific namespace functions (this could be merged into the previous file). This file is conditionally compiled. (*) fs/nfs/internal.h Inter-file declarations, plus a few simple utility functions moved from fs/nfs/inode.c. Additionally, all the in-.c-file externs have been moved here, and those files they were moved from now includes this file. For the most part, the functions have not been changed, only some multiplexor functions have changed significantly. I've also: (*) Added some extra banner comments above some functions. (*) Rearranged the function order within the files to be more logical and better grouped (IMO), though someone may prefer a different order. (*) Reduced the number of #ifdefs in .c files. (*) Added missing __init and __exit directives. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-06-09 21:34:33 +08:00
}
NFSv4: Replace nfs4_path_walk() with VFS path lookup in a private namespace As noted in the previous patch, the NFSv4 client mount code currently has several limitations. If the mount path contains symlinks, or referrals, or even if it just contains a '..', then the client code in nfs4_path_walk() will fail with an error. This patch replaces the nfs4_path_walk()-based lookup with a helper function that sets up a private namespace to represent the namespace on the server, then uses the ordinary VFS and NFS path lookup code to walk down the mount path in that namespace. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-23 03:09:14 +08:00
/*
* Create an NFS4 server record on referral traversal
*/
static int nfs4_referral_get_sb(struct file_system_type *fs_type,
int flags, const char *dev_name, void *raw_data,
struct vfsmount *mnt)
{
struct nfs_clone_mount *data = raw_data;
char *export_path;
struct vfsmount *root_mnt;
int error;
dprintk("--> nfs4_referral_get_sb()\n");
export_path = data->mnt_path;
data->mnt_path = "/";
root_mnt = nfs_do_root_mount(&nfs4_remote_referral_fs_type,
flags, data, data->hostname);
data->mnt_path = export_path;
error = PTR_ERR(root_mnt);
if (IS_ERR(root_mnt))
goto out;
error = nfs_follow_remote_path(root_mnt, export_path, mnt);
out:
dprintk("<-- nfs4_referral_get_sb() = %d%s\n", error,
error != 0 ? " [error]" : "");
return error;
}
NFS: Share NFS superblocks per-protocol per-server per-FSID The attached patch makes NFS share superblocks between mounts from the same server and FSID over the same protocol. It does this by creating each superblock with a false root and returning the real root dentry in the vfsmount presented by get_sb(). The root dentry set starts off as an anonymous dentry if we don't already have the dentry for its inode, otherwise it simply returns the dentry we already have. We may thus end up with several trees of dentries in the superblock, and if at some later point one of anonymous tree roots is discovered by normal filesystem activity to be located in another tree within the superblock, the anonymous root is named and materialises attached to the second tree at the appropriate point. Why do it this way? Why not pass an extra argument to the mount() syscall to indicate the subpath and then pathwalk from the server root to the desired directory? You can't guarantee this will work for two reasons: (1) The root and intervening nodes may not be accessible to the client. With NFS2 and NFS3, for instance, mountd is called on the server to get the filehandle for the tip of a path. mountd won't give us handles for anything we don't have permission to access, and so we can't set up NFS inodes for such nodes, and so can't easily set up dentries (we'd have to have ghost inodes or something). With this patch we don't actually create dentries until we get handles from the server that we can use to set up their inodes, and we don't actually bind them into the tree until we know for sure where they go. (2) Inaccessible symbolic links. If we're asked to mount two exports from the server, eg: mount warthog:/warthog/aaa/xxx /mmm mount warthog:/warthog/bbb/yyy /nnn We may not be able to access anything nearer the root than xxx and yyy, but we may find out later that /mmm/www/yyy, say, is actually the same directory as the one mounted on /nnn. What we might then find out, for example, is that /warthog/bbb was actually a symbolic link to /warthog/aaa/xxx/www, but we can't actually determine that by talking to the server until /warthog is made available by NFS. This would lead to having constructed an errneous dentry tree which we can't easily fix. We can end up with a dentry marked as a directory when it should actually be a symlink, or we could end up with an apparently hardlinked directory. With this patch we need not make assumptions about the type of a dentry for which we can't retrieve information, nor need we assume we know its place in the grand scheme of things until we actually see that place. This patch reduces the possibility of aliasing in the inode and page caches for inodes that may be accessed by more than one NFS export. It also reduces the number of superblocks required for NFS where there are many NFS exports being used from a server (home directory server + autofs for example). This in turn makes it simpler to do local caching of network filesystems, as it can then be guaranteed that there won't be links from multiple inodes in separate superblocks to the same cache file. Obviously, cache aliasing between different levels of NFS protocol could still be a problem, but at least that gives us another key to use when indexing the cache. This patch makes the following changes: (1) The server record construction/destruction has been abstracted out into its own set of functions to make things easier to get right. These have been moved into fs/nfs/client.c. All the code in fs/nfs/client.c has to do with the management of connections to servers, and doesn't touch superblocks in any way; the remaining code in fs/nfs/super.c has to do with VFS superblock management. (2) The sequence of events undertaken by NFS mount is now reordered: (a) A volume representation (struct nfs_server) is allocated. (b) A server representation (struct nfs_client) is acquired. This may be allocated or shared, and is keyed on server address, port and NFS version. (c) If allocated, the client representation is initialised. The state member variable of nfs_client is used to prevent a race during initialisation from two mounts. (d) For NFS4 a simple pathwalk is performed, walking from FH to FH to find the root filehandle for the mount (fs/nfs/getroot.c). For NFS2/3 we are given the root FH in advance. (e) The volume FSID is probed for on the root FH. (f) The volume representation is initialised from the FSINFO record retrieved on the root FH. (g) sget() is called to acquire a superblock. This may be allocated or shared, keyed on client pointer and FSID. (h) If allocated, the superblock is initialised. (i) If the superblock is shared, then the new nfs_server record is discarded. (j) The root dentry for this mount is looked up from the root FH. (k) The root dentry for this mount is assigned to the vfsmount. (3) nfs_readdir_lookup() creates dentries for each of the entries readdir() returns; this function now attaches disconnected trees from alternate roots that happen to be discovered attached to a directory being read (in the same way nfs_lookup() is made to do for lookup ops). The new d_materialise_unique() function is now used to do this, thus permitting the whole thing to be done under one set of locks, and thus avoiding any race between mount and lookup operations on the same directory. (4) The client management code uses a new debug facility: NFSDBG_CLIENT which is set by echoing 1024 to /proc/net/sunrpc/nfs_debug. (5) Clone mounts are now called xdev mounts. (6) Use the dentry passed to the statfs() op as the handle for retrieving fs statistics rather than the root dentry of the superblock (which is now a dummy). Signed-Off-By: David Howells <dhowells@redhat.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2006-08-23 08:06:13 +08:00
#endif /* CONFIG_NFS_V4 */