Dirty snapshot data needs to be flushed unconditionally. If they
were created before truncation, writeback should use old truncate
size/seq.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
When iov_iter type is ITER_PIPE, copy_page_to_iter() increases
the page's reference and add the page to a pipe_buffer. It also
set the pipe_buffer's ops to page_cache_pipe_buf_ops. The comfirm
callback in page_cache_pipe_buf_ops expects the page is from page
cache and uptodate, otherwise it return error.
For ceph_sync_read() case, pages are not from page cache. So we
can't call copy_page_to_iter() when iov_iter type is ITER_PIPE.
The fix is using iov_iter_get_pages_alloc() to allocate pages
for the pipe. (the code is similar to default_file_splice_read)
Signed-off-by: Yan, Zheng <zyan@redhat.com>
For readahead/fadvise cases, caller of ceph_readpages does not
hold buffer capability. Pages can be added to page cache while
there is no buffer capability. This can cause data integrity
issue.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
try_get_cap_refs can be used as a condition in a wait_event* calls.
This is all fine until it has to call __ceph_do_pending_vmtruncate,
which in turn acquires the i_truncate_mutex. This leads to a situation
in which a task's state is !TASK_RUNNING and at the same time it's
trying to acquire a sleeping primitive. In essence a nested sleeping
primitives are being used. This causes the following warning:
WARNING: CPU: 22 PID: 11064 at kernel/sched/core.c:7631 __might_sleep+0x9f/0xb0()
do not call blocking ops when !TASK_RUNNING; state=1 set at [<ffffffff8109447d>] prepare_to_wait_event+0x5d/0x110
ipmi_msghandler tcp_scalable ib_qib dca ib_mad ib_core ib_addr ipv6
CPU: 22 PID: 11064 Comm: fs_checker.pl Tainted: G O 4.4.20-clouder2 #6
Hardware name: Supermicro X10DRi/X10DRi, BIOS 1.1a 10/16/2015
0000000000000000 ffff8838b416fa88 ffffffff812f4409 ffff8838b416fad0
ffffffff81a034f2 ffff8838b416fac0 ffffffff81052b46 ffffffff81a0432c
0000000000000061 0000000000000000 0000000000000000 ffff88167bda54a0
Call Trace:
[<ffffffff812f4409>] dump_stack+0x67/0x9e
[<ffffffff81052b46>] warn_slowpath_common+0x86/0xc0
[<ffffffff81052bcc>] warn_slowpath_fmt+0x4c/0x50
[<ffffffff8109447d>] ? prepare_to_wait_event+0x5d/0x110
[<ffffffff8109447d>] ? prepare_to_wait_event+0x5d/0x110
[<ffffffff8107767f>] __might_sleep+0x9f/0xb0
[<ffffffff81612d30>] mutex_lock+0x20/0x40
[<ffffffffa04eea14>] __ceph_do_pending_vmtruncate+0x44/0x1a0 [ceph]
[<ffffffffa04fa692>] try_get_cap_refs+0xa2/0x320 [ceph]
[<ffffffffa04fd6f5>] ceph_get_caps+0x255/0x2b0 [ceph]
[<ffffffff81094370>] ? wait_woken+0xb0/0xb0
[<ffffffffa04f2c11>] ceph_write_iter+0x2b1/0xde0 [ceph]
[<ffffffff81613f22>] ? schedule_timeout+0x202/0x260
[<ffffffff8117f01a>] ? kmem_cache_free+0x1ea/0x200
[<ffffffff811b46ce>] ? iput+0x9e/0x230
[<ffffffff81077632>] ? __might_sleep+0x52/0xb0
[<ffffffff81156147>] ? __might_fault+0x37/0x40
[<ffffffff8119e123>] ? cp_new_stat+0x153/0x170
[<ffffffff81198cfa>] __vfs_write+0xaa/0xe0
[<ffffffff81199369>] vfs_write+0xa9/0x190
[<ffffffff811b6d01>] ? set_close_on_exec+0x31/0x70
[<ffffffff8119a056>] SyS_write+0x46/0xa0
This happens since wait_event_interruptible can interfere with the
mutex locking code, since they both fiddle with the task state.
Fix the issue by using the newly-added nested blocking infrastructure
in 61ada528de ("sched/wait: Provide infrastructure to deal with
nested blocking")
Link: https://lwn.net/Articles/628628/
Signed-off-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: Yan, Zheng <zyan@redhat.com>
The length of the reply is protocol-dependent - for cephx it's
ceph_x_authorize_reply. Nothing sensible can be passed from the
messenger layer anyway.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Sage Weil <sage@redhat.com>
don't zero on short copies; if the page was uptodate it's just plain
wrong, and if it wasn't we'll be better off just returning 0 and
buggering off.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
If .readlink == NULL implies generic_readlink().
Generated by:
to_del="\.readlink.*=.*generic_readlink"
for i in `git grep -l $to_del`; do sed -i "/$to_del"/d $i; done
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
This function sets req->r_locked_dir which is supposed to indicate to
ceph_fill_trace that the parent's i_rwsem is locked for write.
Unfortunately, there is no guarantee that the dir will be locked when
d_revalidate is called, so we really don't want ceph_fill_trace to do
any dcache manipulation from this context. Clear req->r_locked_dir since
it's clearly not safe to do that.
What we really want to know with d_revalidate is whether the dentry
still points to the same inode. ceph_fill_trace installs a pointer to
the inode in req->r_target_inode, so we can just compare that to
d_inode(dentry) to see if it's the same one after the lookup.
Also, since we aren't generally interested in the parent here, we can
switch to using a GETATTR to hint that to the MDS, which also means that
we only need to reserve one cap.
Finally, just remove the d_unhashed check. That's really outside the
purview of a filesystem's d_revalidate. If the thing became unhashed
while we're checking it, then that's up to the VFS to handle anyway.
Fixes: 200fd27c8f ("ceph: use lookup request to revalidate dentry")
Link: http://tracker.ceph.com/issues/18041
Reported-by: Donatas Abraitis <donatas.abraitis@gmail.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Splice read/write implementation changed recently. When using
generic_file_splice_read(), iov_iter with type == ITER_PIPE is
passed to filesystem's read_iter callback. But ceph_sync_read()
can't serve ITER_PIPE iov_iter correctly (ITER_PIPE iov_iter
expects pages from page cache).
Fixing ceph_sync_read() requires a big patch. So use default
splice read callback for now.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Fixes the following sparse warning:
fs/ceph/xattr.c:19:28: warning:
symbol 'ceph_other_xattr_handler' was not declared. Should it be static?
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
fs/ceph/super.c: In function ‘ceph_real_mount’:
fs/ceph/super.c:818: warning: ‘root’ may be used uninitialized in this function
If s_root is already valid, dentry pointer root is never initialized,
and returned by ceph_real_mount(). This will cause a crash later when
the caller dereferences the pointer.
Fixes: ce2728aaa8 ("ceph: avoid accessing / when mounting a subpath")
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Yan, Zheng <zyan@redhat.com>
In case __ceph_do_getattr returns an error and the retry_op in
ceph_read_iter is not READ_INLINE, then it's possible to invoke
__free_page on a page which is NULL, this naturally leads to a crash.
This can happen when, for example, a process waiting on a MDS reply
receives sigterm.
Fix this by explicitly checking whether the page is set or not.
Cc: stable@vger.kernel.org # 3.19+
Signed-off-by: Nikolay Borisov <kernel@kyup.com>
Reviewed-by: Yan, Zheng <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Pull more vfs updates from Al Viro:
">rename2() work from Miklos + current_time() from Deepa"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
fs: Replace current_fs_time() with current_time()
fs: Replace CURRENT_TIME_SEC with current_time() for inode timestamps
fs: Replace CURRENT_TIME with current_time() for inode timestamps
fs: proc: Delete inode time initializations in proc_alloc_inode()
vfs: Add current_time() api
vfs: add note about i_op->rename changes to porting
fs: rename "rename2" i_op to "rename"
vfs: remove unused i_op->rename
fs: make remaining filesystems use .rename2
libfs: support RENAME_NOREPLACE in simple_rename()
fs: support RENAME_NOREPLACE for local filesystems
ncpfs: fix unused variable warning
Pull vfs xattr updates from Al Viro:
"xattr stuff from Andreas
This completes the switch to xattr_handler ->get()/->set() from
->getxattr/->setxattr/->removexattr"
* 'work.xattr' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
vfs: Remove {get,set,remove}xattr inode operations
xattr: Stop calling {get,set,remove}xattr inode operations
vfs: Check for the IOP_XATTR flag in listxattr
xattr: Add __vfs_{get,set,remove}xattr helpers
libfs: Use IOP_XATTR flag for empty directory handling
vfs: Use IOP_XATTR flag for bad-inode handling
vfs: Add IOP_XATTR inode operations flag
vfs: Move xattr_resolve_name to the front of fs/xattr.c
ecryptfs: Switch to generic xattr handlers
sockfs: Get rid of getxattr iop
sockfs: getxattr: Fail with -EOPNOTSUPP for invalid attribute names
kernfs: Switch to generic xattr handlers
hfs: Switch to generic xattr handlers
jffs2: Remove jffs2_{get,set,remove}xattr macros
xattr: Remove unnecessary NULL attribute name check
with maintenance operations offloaded to userspace (Douglas Fuller,
Mike Christie and myself). Another block device bullet is a series
fixing up layering error paths (myself).
On the filesystem side, we've got patches that improve our handling of
buffered vs dio write races (Neil Brown) and a few assorted fixes from
Zheng. Also included a couple of random cleanups and a minor CRUSH
update.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
iQEcBAABCAAGBQJX+PjZAAoJEEp/3jgCEfOLVuoH/RwtFLIb6/KZUYtBOrVVrTun
kReRlfq2xKYrGGtyQEqSuz7fBdwT1LVCVcL8kC4GFD4R67o+tNMAr6PfM/7pZABj
HRoRLgSZ9FLw4W5n0VpBIznih75QUbCdXiTCtH9eorMHU5q1YpTvVHHlF9W9Pm2I
eNGnBWpGyHVeiK66mpUCH+EQKQ4GkAVD9rneTNqLHgq2yotHkVl1j258+DL6JRGs
OBoh3RmNQaGOAS37Lss8erCSusAGEcAeGV6ubuK2lFUKyR41EkD3I0xkhNSPe+CD
RifFcpVziIeTu//cLgl0nnHGtmUytD7HgJubaPthArKIOen9ZDAfEkgI0o+JI2A=
=45O7
-----END PGP SIGNATURE-----
Merge tag 'ceph-for-4.9-rc1' of git://github.com/ceph/ceph-client
Pull Ceph updates from Ilya Dryomov:
"The big ticket item here is support for rbd exclusive-lock feature,
with maintenance operations offloaded to userspace (Douglas Fuller,
Mike Christie and myself). Another block device bullet is a series
fixing up layering error paths (myself).
On the filesystem side, we've got patches that improve our handling of
buffered vs dio write races (Neil Brown) and a few assorted fixes from
Zheng. Also included a couple of random cleanups and a minor CRUSH
update"
* tag 'ceph-for-4.9-rc1' of git://github.com/ceph/ceph-client: (39 commits)
crush: remove redundant local variable
crush: don't normalize input of crush_ln iteratively
libceph: ceph_build_auth() doesn't need ceph_auth_build_hello()
libceph: use CEPH_AUTH_UNKNOWN in ceph_auth_build_hello()
ceph: fix description for rsize and rasize mount options
rbd: use kmalloc_array() in rbd_header_from_disk()
ceph: use list_move instead of list_del/list_add
ceph: handle CEPH_SESSION_REJECT message
ceph: avoid accessing / when mounting a subpath
ceph: fix mandatory flock check
ceph: remove warning when ceph_releasepage() is called on dirty page
ceph: ignore error from invalidate_inode_pages2_range() in direct write
ceph: fix error handling of start_read()
rbd: add rbd_obj_request_error() helper
rbd: img_data requests don't own their page array
rbd: don't call rbd_osd_req_format_read() for !img_data requests
rbd: rework rbd_img_obj_exists_submit() error paths
rbd: don't crash or leak on errors in rbd_img_obj_parent_read_full_callback()
rbd: move bumping img_request refcount into rbd_obj_request_submit()
rbd: mark the original request as done if stat request fails
...
These inode operations are no longer used; remove them.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
If O_DIRECT writes are racing with buffered writes, then
the call to invalidate_inode_pages2_range() can call ceph_releasepage()
on dirty pages.
Most filesystems hold inode_lock() across O_DIRECT writes so they do not
suffer this race, but cephfs deliberately drops the lock, and opens a window
for the race.
This race can be triggered with the generic/036 test from the xfstests
test suite. It doesn't happen every time, but it does happen often.
As the possibilty is expected, remove the warning, and instead include
the PageDirty() status in the debug message.
Signed-off-by: NeilBrown <neilb@suse.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: Yan, Zheng <zyan@redhat.com>
This call can fail if there are dirty pages. The preceding call to
filemap_write_and_wait_range() will normally remove dirty pages, but
as inode_lock() is not held over calls to ceph_direct_read_write(), it
could race with non-direct writes and pages could be dirtied
immediately after filemap_write_and_wait_range() returns
If there are dirty pages, they will be removed by the subsequent call
to truncate_inode_pages_range(), so having them here is not a problem.
If the 'ret' value is left holding an error, then in the async IO case
(aio_req is not NULL) the loop that would normally call
ceph_osdc_start_request() will see the error in 'ret' and abort all
requests. This doesn't seem like correct behaviour.
So use separate 'ret2' instead of overloading 'ret'.
Signed-off-by: NeilBrown <neilb@suse.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: Yan, Zheng <zyan@redhat.com>
If start_page() fails to add a page to page cache or fails to send
OSD request. It should cal put_page() (instead of free_page()) for
relevant pages.
Besides, start_page() need to cancel fscache readpage if it fails
to send OSD request.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Reported-by: Zhi Zhang <zhang.david2011@gmail.com>
current_fs_time() uses struct super_block* as an argument.
As per Linus's suggestion, this is changed to take struct
inode* as a parameter instead. This is because the function
is primarily meant for vfs inode timestamps.
Also the function was renamed as per Arnd's suggestion.
Change all calls to current_fs_time() to use the new
current_time() function instead. current_fs_time() will be
deleted.
Signed-off-by: Deepa Dinamani <deepa.kernel@gmail.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
This is trivial to do:
- add flags argument to foo_rename()
- check if flags is zero
- assign foo_rename() to .rename2 instead of .rename
This doesn't mean it's impossible to support RENAME_NOREPLACE for these
filesystems, but it is not trivial, like for local filesystems.
RENAME_NOREPLACE must guarantee atomicity (i.e. it shouldn't be possible
for a file to be created on one host while it is overwritten by rename on
another host).
Filesystems converted:
9p, afs, ceph, coda, ecryptfs, kernfs, lustre, ncpfs, nfs, ocfs2, orangefs.
After this, we can get rid of the duplicate interfaces for rename.
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: David Howells <dhowells@redhat.com> [AFS]
Acked-by: Mike Marshall <hubcap@omnibond.com>
Cc: Eric Van Hensbergen <ericvh@gmail.com>
Cc: Ilya Dryomov <idryomov@gmail.com>
Cc: Jan Harkes <jaharkes@cs.cmu.edu>
Cc: Tyler Hicks <tyhicks@canonical.com>
Cc: Oleg Drokin <oleg.drokin@intel.com>
Cc: Trond Myklebust <trond.myklebust@primarydata.com>
Cc: Mark Fasheh <mfasheh@suse.com>
inode_change_ok() will be resposible for clearing capabilities and IMA
extended attributes and as such will need dentry. Give it as an argument
to inode_change_ok() instead of an inode. Also rename inode_change_ok()
to setattr_prepare() to better relect that it does also some
modifications in addition to checks.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jan Kara <jack@suse.cz>
To avoid clearing of capabilities or security related extended
attributes too early, inode_change_ok() will need to take dentry instead
of inode. ceph_setattr() has the dentry easily available but
__ceph_setattr() is also called from ceph_set_acl() where dentry is not
easily available. Luckily that call path does not need inode_change_ok()
to be called anyway. So reorganize functions a bit so that
inode_change_ok() is called only from paths where dentry is available.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Jan Kara <jack@suse.cz>
When file permissions are modified via chmod(2) and the user is not in
the owning group or capable of CAP_FSETID, the setgid bit is cleared in
inode_change_ok(). Setting a POSIX ACL via setxattr(2) sets the file
permissions as well as the new ACL, but doesn't clear the setgid bit in
a similar way; this allows to bypass the check in chmod(2). Fix that.
References: CVE-2016-7097
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Commit f3c4ebe65e ("ceph: using hash value to compose dentry offset")
modified "if (fpos_frag(new_pos) != fi->frag)" to "if (fi->frag |=
fpos_frag(new_pos))" in need_reset_readdir(), thus replacing a
comparison operator with an assignment one.
This looks like a typo which is reported by clang when building the
kernel with some warning flags:
fs/ceph/dir.c:600:22: error: using the result of an assignment as a
condition without parentheses [-Werror,-Wparentheses]
} else if (fi->frag |= fpos_frag(new_pos)) {
~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~
fs/ceph/dir.c:600:22: note: place parentheses around the assignment
to silence this warning
} else if (fi->frag |= fpos_frag(new_pos)) {
^
( )
fs/ceph/dir.c:600:22: note: use '!=' to turn this compound
assignment into an inequality comparison
} else if (fi->frag |= fpos_frag(new_pos)) {
^~
!=
Fixes: f3c4ebe65e ("ceph: using hash value to compose dentry offset")
Signed-off-by: Nicolas Iooss <nicolas.iooss_linux@m4x.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
pathbase is the base inode; set it to 0 if we've got no path.
Coverity-id: 146348
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Alex Elder <elder@linaro.org>
* RADOS namespace support in libceph and CephFS (Zheng Yan and myself).
The stopgaps added in 4.5 to deny access to inodes in namespaces are
removed and CEPH_FEATURE_FS_FILE_LAYOUT_V2 feature bit is now fully
supported.
* A large rework of the MDS cap flushing code (Zheng Yan).
* Handle some of ->d_revalidate() in RCU mode (Jeff Layton). We were
overly pessimistic before, bailing at the first sight of LOOKUP_RCU.
On top of that we've got a few CephFS bug fixes, a couple of cleanups
and Arnd's workaround for a weird genksyms issue.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
iQEcBAABCAAGBQJXoKLJAAoJEEp/3jgCEfOLDTUIAIcctpKUiNBokc95mQaXYl34
j7lPIaD0/Ur7JPt4nMdtlywYJYSVV2c+SglHztj/+fv0G4bWbLVEFRruh9SwKIci
PzttcmycIAqSn1f5gBZwyQbGuffd/F0EnBj7fFjcukt01i3s1ZQ7t4XtLGtAV0Ts
aIfFtx9SqWig57Z1OZqNgnhnOoh6IqNbic3FL5Hvdl5N5pFbBcQho6Vzoa5O1osH
URG6RmCcO4nykfSoxiivE7UZ+CImsXHkRD7rupBuIjqjZ8wvmZqQF5qxnkb9Dw2F
IkNhrHkTSIiv4EsNPLAETTnFSozrL1nEykKr2FBW+ti8nxNcav+8FgVapqLvFIw=
=gQ0/
-----END PGP SIGNATURE-----
Merge tag 'ceph-for-4.8-rc1' of git://github.com/ceph/ceph-client
Pull Ceph updates from Ilya Dryomov:
"The highlights are:
- RADOS namespace support in libceph and CephFS (Zheng Yan and
myself). The stopgaps added in 4.5 to deny access to inodes in
namespaces are removed and CEPH_FEATURE_FS_FILE_LAYOUT_V2 feature
bit is now fully supported
- A large rework of the MDS cap flushing code (Zheng Yan)
- Handle some of ->d_revalidate() in RCU mode (Jeff Layton). We were
overly pessimistic before, bailing at the first sight of LOOKUP_RCU
On top of that we've got a few CephFS bug fixes, a couple of cleanups
and Arnd's workaround for a weird genksyms issue"
* tag 'ceph-for-4.8-rc1' of git://github.com/ceph/ceph-client: (34 commits)
ceph: fix symbol versioning for ceph_monc_do_statfs
ceph: Correctly return NXIO errors from ceph_llseek
ceph: Mark the file cache as unreclaimable
ceph: optimize cap flush waiting
ceph: cleanup ceph_flush_snaps()
ceph: kick cap flushes before sending other cap message
ceph: introduce an inode flag to indicates if snapflush is needed
ceph: avoid sending duplicated cap flush message
ceph: unify cap flush and snapcap flush
ceph: use list instead of rbtree to track cap flushes
ceph: update types of some local varibles
ceph: include 'follows' of pending snapflush in cap reconnect message
ceph: update cap reconnect message to version 3
ceph: mount non-default filesystem by name
libceph: fsmap.user subscription support
ceph: handle LOOKUP_RCU in ceph_d_revalidate
ceph: allow dentry_lease_is_valid to work under RCU walk
ceph: clear d_fsinfo pointer under d_lock
ceph: remove ceph_mdsc_lease_release
ceph: don't use ->d_time
...
This changes the vfs dentry hashing to mix in the parent pointer at the
_beginning_ of the hash, rather than at the end.
That actually improves both the hash and the code generation, because we
can move more of the computation to the "static" part of the dcache
setup, and do less at lookup runtime.
It turns out that a lot of other hash users also really wanted to mix in
a base pointer as a 'salt' for the hash, and so the slightly extended
interface ends up working well for other cases too.
Users that want a string hash that is purely about the string pass in a
'salt' pointer of NULL.
* merge branch 'salted-string-hash':
fs/dcache.c: Save one 32-bit multiply in dcache lookup
vfs: make the string hashes salt the hash
ceph_llseek does not correctly return NXIO errors because the 'out' path
always returns 'offset'.
Fixes: 06222e491e ("fs: handle SEEK_HOLE/SEEK_DATA properly in all fs's that define their own llseek")
Signed-off-by: Phil Turnbull <phil.turnbull@oracle.com>
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Ceph creates multiple caches with the SLAB_RECLAIMABLE flag set, so
that it can satisfy its internal needs. Inspecting the code shows that
most of the caches are indeed reclaimable since they are directly
related to the generic inode/dentry shrinkers. However, one of the
cache used to satisfy struct file is not reclaimable since its
entries are freed only when the last reference to the file is
dropped. If a heavily loaded node opens a lot of files it can
introduce non-trivial discrepancies between memory shown as reclaimable
and what is actually reclaimed when drop_caches is used.
Fix this by removing the reclaimable flag for the file's cache.
Signed-off-by: Nikolay Borisov <n.borisov.lkml@gmail.com>
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Add a 'wake' flag to ceph_cap_flush struct, which indicates if there
is someone waiting for it to finish. When getting flush ack message,
we check the 'wake' flag in corresponding ceph_cap_flush struct to
decide if we should wake up waiters. One corner case is that the
acked cap flush has 'wake' flags is set, but it is not the first one
on the flushing list. We do not wake up waiters in this case, set
'wake' flags of preceding ceph_cap_flush struct instead
Signed-off-by: Yan, Zheng <zyan@redhat.com>
This patch devide __ceph_flush_snaps() into two stags. In the first
stage, __ceph_flush_snaps() assign snapcaps flush TIDs and add them
to cap flush lists. __ceph_flush_snaps() keeps holding the
i_ceph_lock in this stagge. So inode's auth cap can not change. In
the second stage, __ceph_flush_snaps() send flushsnap cap messages.
i_ceph_lock is unlocked before sending each cap message. If auth cap
changes in the middle, __ceph_flush_snaps() just stops. This is OK
because kick_flushing_inode_caps() will re-send flushsnap cap messages
to inode's new auth MDS.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
make ceph_kick_flushing_caps() ignore inodes whose cap flushes
have already been re-sent by ceph_early_kick_flushing_caps()
Signed-off-by: Yan, Zheng <zyan@redhat.com>
This patch includes following changes
- Assign flush tid to snapcap flush
- Remove session's s_cap_snaps_flushing list. Add inode to session's
s_cap_flushing list instead. Inode is removed from the list when
there is no pending snapcap flush or cap flush.
- make __kick_flushing_caps() re-send both snapcap flushes and cap
flushes.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
We don't have requirement of searching cap flush by TID. In most cases,
we just need to know TID of the oldest cap flush. List is ideal for this
usage.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
To mount non-default filesytem, user currently needs to provide mds
namespace ID. This is inconvenience.
This patch makes user be able to mount filesystem by name. If user
wants to mount non-default filesystem. Client first subscribes to
fsmap.user. Subscribe to mdsmap.<ID> after getting ID of filesystem.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
We can now handle the snapshot cases under RCU, as well as the
non-snapshot case when we don't need to queue up a lease renewal
allow LOOKUP_RCU walks to proceed under those conditions.
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: Yan, Zheng <zyan@redhat.com>
Under rcuwalk, we need to take extra care when dereferencing d_parent.
We want to do that once and pass a pointer to dentry_lease_is_valid.
Also, we must ensure that that function can handle the case where we're
racing with d_release. Check whether "di" is NULL under the d_lock, and
just return 0 if so.
Finally, we still need to kick off a renewal job if the lease is getting
close to expiration. If that's the case, then just drop out of rcuwalk
mode since that could block.
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: Yan, Zheng <zyan@redhat.com>
To check for a valid dentry lease, we need to get at the
ceph_dentry_info. Under rcuwalk though, we may end up with a dentry that
is on its way to destruction. Since we need to take the d_lock in
dentry_lease_is_valid already, we can just ensure that we clear the
d_fsinfo pointer out under the same lock before destroying it.
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: Yan, Zheng <zyan@redhat.com>
This patch adds codes that decode pool namespace information in
cap message and request reply. Pool namespace is saved in i_layout,
it will be passed to libceph when doing read/write.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Add pool namesapce pointer to struct ceph_file_layout and struct
ceph_object_locator. Pool namespace is used by when mapping object
to PG, it's also used when composing OSD request.
The namespace pointer in struct ceph_file_layout is RCU protected.
So libceph can read namespace without taking lock.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
[idryomov@gmail.com: ceph_oloc_destroy(), misc minor changes]
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Define new ceph_file_layout structure and rename old ceph_file_layout
to ceph_file_layout_legacy. This is preparation for adding namespace
to ceph_file_layout structure.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
An on-stack oid in ceph_ioctl_get_dataloc() is not initialized,
resulting in a WARN and a NULL pointer dereference later on. We will
have more of these on-stack in the future, so fix it with a convenience
macro.
Fixes: d30291b985 ("libceph: variable-sized ceph_object_id")
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
->atomic_open() can be given an in-lookup dentry *or* a negative one
found in dcache. Use d_in_lookup() to tell one from another, rather
than d_unhashed().
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Pull vfs fixes from Al Viro:
"Tmpfs readdir throughput regression fix (this cycle) + some -stable
fodder all over the place.
One missing bit is Miklos' tonight locks.c fix - NFS folks had already
grabbed that one by the time I woke up ;-)"
[ The locks.c fix came through the nfsd tree just moments ago ]
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
namespace: update event counter when umounting a deleted dentry
9p: use file_dentry()
ceph: fix d_obtain_alias() misuses
lockless next_positive()
libfs.c: new helper - next_positive()
dcache_{readdir,dir_lseek}(): don't bother with nested ->d_lock
We always mixed in the parent pointer into the dentry name hash, but we
did it late at lookup time. It turns out that we can simplify that
lookup-time action by salting the hash with the parent pointer early
instead of late.
A few other users of our string hashes also wanted to mix in their own
pointers into the hash, and those are updated to use the same mechanism.
Hash users that don't have any particular initial salt can just use the
NULL pointer as a no-salt.
Cc: Vegard Nossum <vegard.nossum@oracle.com>
Cc: George Spelvin <linux@sciencehorizons.net>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There are several issues in fscache revalidation code.
- In ceph_revalidate_work(), fscache_invalidate() is called when
fscache_check_consistency() return 0. This is complete wrong
because 0 means cache is valid.
- Handle_cap_grant() calls ceph_queue_revalidate() if client
already has CAP_FILE_CACHE. This code is confusing. Client
should revalidate the cache each time it got CAP_FILE_CACHE
anew.
- In Handle_cap_grant(), fscache_invalidate() is called if MDS
revokes CAP_FILE_CACHE. This is inconsistency with the case
that inode get evicted. In the later case, the cache is not
discarded. Client may use the cache when inode is reloaded.
This patch moves the fscache revalidation into ceph_get_caps().
Client revalidates the cache after it gets CAP_FILE_CACHE.
i_rdcache_gen should keep constance while CAP_FILE_CACHE is
used. If i_fscache_gen is not equal to i_rdcache_gen, client
needs to check cache's consistency.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
All other filesystems do not add dirty pages to fscache. They all
disable fscache when inode is opened for write. Only ceph adds
dirty pages to fscache, but the code is buggy.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
For the benefit of every single caller, take osdc instead of map.
Also, now that osdc->osdmap can't ever be NULL, drop the check.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Pull vfs fixes from Al Viro:
"Followups to the parallel lookup work:
- update docs
- restore killability of the places that used to take ->i_mutex
killably now that we have down_write_killable() merged
- Additionally, it turns out that I missed a prerequisite for
security_d_instantiate() stuff - ->getxattr() wasn't the only thing
that could be called before dentry is attached to inode; with smack
we needed the same treatment applied to ->setxattr() as well"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
switch ->setxattr() to passing dentry and inode separately
switch xattr_handler->set() to passing dentry and inode separately
restore killability of old mutex_lock_killable(&inode->i_mutex) users
add down_write_killable_nested()
update D/f/directory-locking
Pull Ceph updates from Sage Weil:
"This changeset has a few main parts:
- Ilya has finished a huge refactoring effort to sync up the
client-side logic in libceph with the user-space client code, which
has evolved significantly over the last couple years, with lots of
additional behaviors (e.g., how requests are handled when cluster
is full and transitions from full to non-full).
This structure of the code is more closely aligned with userspace
now such that it will be much easier to maintain going forward when
behavior changes take place. There are some locking improvements
bundled in as well.
- Zheng adds multi-filesystem support (multiple namespaces within the
same Ceph cluster)
- Zheng has changed the readdir offsets and directory enumeration so
that dentry offsets are hash-based and therefore stable across
directory fragmentation events on the MDS.
- Zheng has a smorgasbord of bug fixes across fs/ceph"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client: (71 commits)
ceph: fix wake_up_session_cb()
ceph: don't use truncate_pagecache() to invalidate read cache
ceph: SetPageError() for writeback pages if writepages fails
ceph: handle interrupted ceph_writepage()
ceph: make ceph_update_writeable_page() uninterruptible
libceph: make ceph_osdc_wait_request() uninterruptible
ceph: handle -EAGAIN returned by ceph_update_writeable_page()
ceph: make fault/page_mkwrite return VM_FAULT_OOM for -ENOMEM
ceph: block non-fatal signals for fault/page_mkwrite
ceph: make logical calculation functions return bool
ceph: tolerate bad i_size for symlink inode
ceph: improve fragtree change detection
ceph: keep leaf frag when updating fragtree
ceph: fix dir_auth check in ceph_fill_dirfrag()
ceph: don't assume frag tree splits in mds reply are sorted
ceph: fix inode reference leak
ceph: using hash value to compose dentry offset
ceph: don't forbid marking directory complete after forward seek
ceph: record 'offset' for each entry of readdir result
ceph: define 'end/complete' in readdir reply as bit flags
...
We should reset i_requested_max_size before waking the waiters.
(zero i_requested_max_size make waiter re-request the max size)
Signed-off-by: Yan, Zheng <zyan@redhat.com>
truncate_pagecache() drops dirty pages, it's dangerous to use it
to invalidate read cache. Besides, we shouldn't start invalidating
read cache while there are buffer writers. Because buffer writers
may add dirty pages later.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
writepage() can be interrupted when it's called by direct memory
reclaimer (the direct memory relaimer is killed). To avoid lossing
data, we redirty the page.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
ceph_update_writeable_page() is used by ceph_write_begin(). It beaks
atomicity of write operation if it's interruptible.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
when ceph_update_writeable_page() return -EAGAIN, caller should
lock the page and call ceph_update_writeable_page() again.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Fault and page_mkwrite are supposed to be uninterruptable. But they
call ceph functions that are interruptible. So they should block
signals before calling functions that are interruptible
Signed-off-by: Yan, Zheng <zyan@redhat.com>
This patch makes serverl logical caculation functions return bool to
improve readability due to these particular functions only using 0/1
as their return value.
No functional change.
Signed-off-by: Zhang Zhuoyu <zhangzhuoyu@cmss.chinamobile.com>
Nodes in i_fragtree are sorted according to ceph_compare_frag().
It means frag node in i_fragtree always follow its direct parent
node. To check if a leaf node is valid, we just need to check if
it's child of previous split node.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
The algorithm that updates i_fragtree relies on that the frag tree
splits in mds reply are of the same order of i_fragtree. This is not
true because current MDS encodes frag tree splits in ascending order
of (unsigned)frag_t. But nodes in i_fragtree are sorted according to
ceph_frag_compare().
The fix is sort the frag tree splits first, then updates i_fragtree.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
If MDS sorts dentries in dirfrag in hash order, we use hash value to
compose dentry offset. dentry offset is:
(0xff << 52) | ((24 bits hash) << 28) |
(the nth entry hash hash collision)
This offset is stable across directory fragmentation. This alos means
there is no need to reset readdir offset if directory get fragmented
in the middle of readdir.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Forward seek within same frag does not update fi->last_name, it will
not affect contents of later readdir reply. So there is no need to
forbid marking directory complete
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Jeff Layton