linux/fs/nfs/nfs4state.c

2534 lines
66 KiB
C
Raw Normal View History

/*
* fs/nfs/nfs4state.c
*
* Client-side XDR for NFSv4.
*
* Copyright (c) 2002 The Regents of the University of Michigan.
* All rights reserved.
*
* Kendrick Smith <kmsmith@umich.edu>
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
* DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* Implementation of the NFSv4 state model. For the time being,
* this is minimal, but will be made much more complex in a
* subsequent patch.
*/
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/fs.h>
#include <linux/nfs_fs.h>
#include <linux/kthread.h>
#include <linux/module.h>
#include <linux/random.h>
#include <linux/ratelimit.h>
#include <linux/workqueue.h>
#include <linux/bitops.h>
NFS: Cache state owners after files are closed Servers have a finite amount of memory to store NFSv4 open and lock owners. Moreover, servers may have a difficult time determining when they can reap their state owner table, thanks to gray areas in the NFSv4 protocol specification. Thus clients should be careful to reuse state owners when possible. Currently Linux is not too careful. When a user has closed all her files on one mount point, the state owner's reference count goes to zero, and it is released. The next OPEN allocates a new one. A workload that serially opens and closes files can run through a large number of open owners this way. When a state owner's reference count goes to zero, slap it onto a free list for that nfs_server, with an expiry time. Garbage collect before looking for a state owner. This makes state owners for active users available for re-use. Now that there can be unused state owners remaining at umount time, purge the state owner free list when a server is destroyed. Also be sure not to reclaim unused state owners during state recovery. This change has benefits for the client as well. For some workloads, this approach drops the number of OPEN_CONFIRM calls from the same as the number of OPEN calls, down to just one. This reduces wire traffic and thus open(2) latency. Before this patch, untarring a kernel source tarball shows the OPEN_CONFIRM call counter steadily increasing through the test. With the patch, the OPEN_CONFIRM count remains at 1 throughout the entire untar. As long as the expiry time is kept short, I don't think garbage collection should be terribly expensive, although it does bounce the clp->cl_lock around a bit. [ At some point we should rationalize the use of the nfs_server ->destroy method. ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond: Fixed a garbage collection race and a few efficiency issues] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-12-07 05:13:48 +08:00
#include <linux/jiffies.h>
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
#include <linux/sunrpc/clnt.h>
#include "nfs4_fs.h"
#include "callback.h"
#include "delegation.h"
#include "internal.h"
#include "nfs4idmap.h"
#include "nfs4session.h"
#include "pnfs.h"
#include "netns.h"
#define NFSDBG_FACILITY NFSDBG_STATE
#define OPENOWNER_POOL_SIZE 8
const nfs4_stateid zero_stateid = {
{ .data = { 0 } },
.type = NFS4_SPECIAL_STATEID_TYPE,
};
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
static DEFINE_MUTEX(nfs_clid_init_mutex);
int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
{
struct nfs4_setclientid_res clid = {
.clientid = clp->cl_clientid,
.confirm = clp->cl_confirm,
};
unsigned short port;
int status;
struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
goto do_confirm;
port = nn->nfs_callback_tcpport;
if (clp->cl_addr.ss_family == AF_INET6)
port = nn->nfs_callback_tcpport6;
status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
if (status != 0)
goto out;
clp->cl_clientid = clid.clientid;
clp->cl_confirm = clid.confirm;
set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
do_confirm:
status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
if (status != 0)
goto out;
clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
nfs4_schedule_state_renewal(clp);
out:
return status;
}
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
/**
* nfs40_discover_server_trunking - Detect server IP address trunking (mv0)
*
* @clp: nfs_client under test
* @result: OUT: found nfs_client, or clp
* @cred: credential to use for trunking test
*
* Returns zero, a negative errno, or a negative NFS4ERR status.
* If zero is returned, an nfs_client pointer is planted in
* "result".
*
* Note: The returned client may not yet be marked ready.
*/
int nfs40_discover_server_trunking(struct nfs_client *clp,
struct nfs_client **result,
struct rpc_cred *cred)
{
struct nfs4_setclientid_res clid = {
.clientid = clp->cl_clientid,
.confirm = clp->cl_confirm,
};
struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
unsigned short port;
int status;
port = nn->nfs_callback_tcpport;
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
if (clp->cl_addr.ss_family == AF_INET6)
port = nn->nfs_callback_tcpport6;
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
if (status != 0)
goto out;
clp->cl_clientid = clid.clientid;
clp->cl_confirm = clid.confirm;
status = nfs40_walk_client_list(clp, result, cred);
if (status == 0) {
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
/* Sustain the lease, even if it's empty. If the clientid4
* goes stale it's of no use for trunking discovery. */
nfs4_schedule_state_renewal(*result);
}
out:
return status;
}
struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
{
struct rpc_cred *cred = NULL;
if (clp->cl_machine_cred != NULL)
cred = get_rpccred(clp->cl_machine_cred);
return cred;
}
NFS: Use root's credential for lease management when keytab is missing Commit 05f4c350 "NFS: Discover NFSv4 server trunking when mounting" Fri Sep 14 17:24:32 2012 introduced Uniform Client String support, which forces our NFS client to establish a client ID immediately during a mount operation rather than waiting until a user wants to open a file. Normally machine credentials (eg. from a keytab) are used to perform a mount operation that is protected by Kerberos. Before 05fc350, SETCLIENTID used a machine credential, or fell back to a regular user's credential if no keytab is available. On clients that don't have a keytab, performing SETCLIENTID early means there's no user credential to fall back on, since no regular user has kinit'd yet. 05f4c350 seems to have broken the ability to mount with sec=krb5 on clients that don't have a keytab in kernels 3.7 - 3.10. To address this regression, commit 4edaa308 (NFS: Use "krb5i" to establish NFSv4 state whenever possible), Sat Mar 16 15:56:20 2013, was merged in 3.10. This commit forces the NFS client to fall back to AUTH_SYS for lease management operations if no keytab is available. Neil Brown noticed that, since root is required to kinit to do a sec=krb5 mount when a client doesn't have a keytab, we can try to use root's Kerberos credential before AUTH_SYS. Now, when determining a principal and flavor to use for lease management, the NFS client tries in this order: 1. Flavor: AUTH_GSS, krb5i Principal: service principal (via keytab) 2. Flavor: AUTH_GSS, krb5i Principal: user principal established for UID 0 (via kinit) 3. Flavor: AUTH_SYS Principal: UID 0 / GID 0 Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2013-07-25 00:28:28 +08:00
static void nfs4_root_machine_cred(struct nfs_client *clp)
{
struct rpc_cred *cred, *new;
new = rpc_lookup_machine_cred(NULL);
spin_lock(&clp->cl_lock);
cred = clp->cl_machine_cred;
clp->cl_machine_cred = new;
spin_unlock(&clp->cl_lock);
if (cred != NULL)
put_rpccred(cred);
}
static struct rpc_cred *
nfs4_get_renew_cred_server_locked(struct nfs_server *server)
{
struct rpc_cred *cred = NULL;
struct nfs4_state_owner *sp;
struct rb_node *pos;
for (pos = rb_first(&server->state_owners);
pos != NULL;
pos = rb_next(pos)) {
sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
if (list_empty(&sp->so_states))
continue;
cred = get_rpccred(sp->so_cred);
break;
}
return cred;
}
/**
* nfs4_get_renew_cred_locked - Acquire credential for a renew operation
* @clp: client state handle
*
* Returns an rpc_cred with reference count bumped, or NULL.
* Caller must hold clp->cl_lock.
*/
struct rpc_cred *nfs4_get_renew_cred_locked(struct nfs_client *clp)
{
struct rpc_cred *cred = NULL;
struct nfs_server *server;
/* Use machine credentials if available */
cred = nfs4_get_machine_cred_locked(clp);
if (cred != NULL)
goto out;
rcu_read_lock();
list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
cred = nfs4_get_renew_cred_server_locked(server);
if (cred != NULL)
break;
}
rcu_read_unlock();
out:
return cred;
}
static void nfs4_end_drain_slot_table(struct nfs4_slot_table *tbl)
{
if (test_and_clear_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state)) {
spin_lock(&tbl->slot_tbl_lock);
nfs41_wake_slot_table(tbl);
spin_unlock(&tbl->slot_tbl_lock);
}
}
static void nfs4_end_drain_session(struct nfs_client *clp)
{
struct nfs4_session *ses = clp->cl_session;
if (clp->cl_slot_tbl) {
nfs4_end_drain_slot_table(clp->cl_slot_tbl);
return;
}
if (ses != NULL) {
nfs4_end_drain_slot_table(&ses->bc_slot_table);
nfs4_end_drain_slot_table(&ses->fc_slot_table);
}
}
static int nfs4_drain_slot_tbl(struct nfs4_slot_table *tbl)
{
set_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state);
spin_lock(&tbl->slot_tbl_lock);
if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
reinit_completion(&tbl->complete);
spin_unlock(&tbl->slot_tbl_lock);
return wait_for_completion_interruptible(&tbl->complete);
}
spin_unlock(&tbl->slot_tbl_lock);
return 0;
}
static int nfs4_begin_drain_session(struct nfs_client *clp)
{
struct nfs4_session *ses = clp->cl_session;
int ret = 0;
if (clp->cl_slot_tbl)
return nfs4_drain_slot_tbl(clp->cl_slot_tbl);
/* back channel */
ret = nfs4_drain_slot_tbl(&ses->bc_slot_table);
if (ret)
return ret;
/* fore channel */
return nfs4_drain_slot_tbl(&ses->fc_slot_table);
}
#if defined(CONFIG_NFS_V4_1)
static int nfs41_setup_state_renewal(struct nfs_client *clp)
{
int status;
struct nfs_fsinfo fsinfo;
unsigned long now;
if (!test_bit(NFS_CS_CHECK_LEASE_TIME, &clp->cl_res_state)) {
nfs4_schedule_state_renewal(clp);
return 0;
}
now = jiffies;
status = nfs4_proc_get_lease_time(clp, &fsinfo);
if (status == 0) {
nfs4_set_lease_period(clp, fsinfo.lease_time * HZ, now);
nfs4_schedule_state_renewal(clp);
}
return status;
}
static void nfs41_finish_session_reset(struct nfs_client *clp)
{
clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
/* create_session negotiated new slot table */
clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
nfs41_setup_state_renewal(clp);
}
int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
{
int status;
if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
goto do_confirm;
status = nfs4_proc_exchange_id(clp, cred);
if (status != 0)
goto out;
set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
do_confirm:
status = nfs4_proc_create_session(clp, cred);
if (status != 0)
goto out;
nfs41_finish_session_reset(clp);
nfs_mark_client_ready(clp, NFS_CS_READY);
out:
return status;
}
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
/**
* nfs41_discover_server_trunking - Detect server IP address trunking (mv1)
*
* @clp: nfs_client under test
* @result: OUT: found nfs_client, or clp
* @cred: credential to use for trunking test
*
* Returns NFS4_OK, a negative errno, or a negative NFS4ERR status.
* If NFS4_OK is returned, an nfs_client pointer is planted in
* "result".
*
* Note: The returned client may not yet be marked ready.
*/
int nfs41_discover_server_trunking(struct nfs_client *clp,
struct nfs_client **result,
struct rpc_cred *cred)
{
int status;
status = nfs4_proc_exchange_id(clp, cred);
if (status != NFS4_OK)
return status;
status = nfs41_walk_client_list(clp, result, cred);
if (status < 0)
return status;
if (clp != *result)
return 0;
/* Purge state if the client id was established in a prior instance */
if (clp->cl_exchange_flags & EXCHGID4_FLAG_CONFIRMED_R)
set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
else
set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
nfs4_schedule_state_manager(clp);
status = nfs_wait_client_init_complete(clp);
if (status < 0)
nfs_put_client(clp);
return status;
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
}
#endif /* CONFIG_NFS_V4_1 */
/**
* nfs4_get_clid_cred - Acquire credential for a setclientid operation
* @clp: client state handle
*
* Returns an rpc_cred with reference count bumped, or NULL.
*/
struct rpc_cred *nfs4_get_clid_cred(struct nfs_client *clp)
{
struct rpc_cred *cred;
spin_lock(&clp->cl_lock);
cred = nfs4_get_machine_cred_locked(clp);
spin_unlock(&clp->cl_lock);
return cred;
}
static struct nfs4_state_owner *
nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
{
struct rb_node **p = &server->state_owners.rb_node,
*parent = NULL;
struct nfs4_state_owner *sp;
while (*p != NULL) {
parent = *p;
sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
if (cred < sp->so_cred)
p = &parent->rb_left;
else if (cred > sp->so_cred)
p = &parent->rb_right;
else {
NFS: Cache state owners after files are closed Servers have a finite amount of memory to store NFSv4 open and lock owners. Moreover, servers may have a difficult time determining when they can reap their state owner table, thanks to gray areas in the NFSv4 protocol specification. Thus clients should be careful to reuse state owners when possible. Currently Linux is not too careful. When a user has closed all her files on one mount point, the state owner's reference count goes to zero, and it is released. The next OPEN allocates a new one. A workload that serially opens and closes files can run through a large number of open owners this way. When a state owner's reference count goes to zero, slap it onto a free list for that nfs_server, with an expiry time. Garbage collect before looking for a state owner. This makes state owners for active users available for re-use. Now that there can be unused state owners remaining at umount time, purge the state owner free list when a server is destroyed. Also be sure not to reclaim unused state owners during state recovery. This change has benefits for the client as well. For some workloads, this approach drops the number of OPEN_CONFIRM calls from the same as the number of OPEN calls, down to just one. This reduces wire traffic and thus open(2) latency. Before this patch, untarring a kernel source tarball shows the OPEN_CONFIRM call counter steadily increasing through the test. With the patch, the OPEN_CONFIRM count remains at 1 throughout the entire untar. As long as the expiry time is kept short, I don't think garbage collection should be terribly expensive, although it does bounce the clp->cl_lock around a bit. [ At some point we should rationalize the use of the nfs_server ->destroy method. ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond: Fixed a garbage collection race and a few efficiency issues] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-12-07 05:13:48 +08:00
if (!list_empty(&sp->so_lru))
list_del_init(&sp->so_lru);
atomic_inc(&sp->so_count);
return sp;
}
}
return NULL;
}
static struct nfs4_state_owner *
nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
{
struct nfs_server *server = new->so_server;
struct rb_node **p = &server->state_owners.rb_node,
*parent = NULL;
struct nfs4_state_owner *sp;
int err;
while (*p != NULL) {
parent = *p;
sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
if (new->so_cred < sp->so_cred)
p = &parent->rb_left;
else if (new->so_cred > sp->so_cred)
p = &parent->rb_right;
else {
NFS: Cache state owners after files are closed Servers have a finite amount of memory to store NFSv4 open and lock owners. Moreover, servers may have a difficult time determining when they can reap their state owner table, thanks to gray areas in the NFSv4 protocol specification. Thus clients should be careful to reuse state owners when possible. Currently Linux is not too careful. When a user has closed all her files on one mount point, the state owner's reference count goes to zero, and it is released. The next OPEN allocates a new one. A workload that serially opens and closes files can run through a large number of open owners this way. When a state owner's reference count goes to zero, slap it onto a free list for that nfs_server, with an expiry time. Garbage collect before looking for a state owner. This makes state owners for active users available for re-use. Now that there can be unused state owners remaining at umount time, purge the state owner free list when a server is destroyed. Also be sure not to reclaim unused state owners during state recovery. This change has benefits for the client as well. For some workloads, this approach drops the number of OPEN_CONFIRM calls from the same as the number of OPEN calls, down to just one. This reduces wire traffic and thus open(2) latency. Before this patch, untarring a kernel source tarball shows the OPEN_CONFIRM call counter steadily increasing through the test. With the patch, the OPEN_CONFIRM count remains at 1 throughout the entire untar. As long as the expiry time is kept short, I don't think garbage collection should be terribly expensive, although it does bounce the clp->cl_lock around a bit. [ At some point we should rationalize the use of the nfs_server ->destroy method. ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond: Fixed a garbage collection race and a few efficiency issues] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-12-07 05:13:48 +08:00
if (!list_empty(&sp->so_lru))
list_del_init(&sp->so_lru);
atomic_inc(&sp->so_count);
return sp;
}
}
err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
if (err)
return ERR_PTR(err);
rb_link_node(&new->so_server_node, parent, p);
rb_insert_color(&new->so_server_node, &server->state_owners);
return new;
}
static void
nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
{
struct nfs_server *server = sp->so_server;
if (!RB_EMPTY_NODE(&sp->so_server_node))
rb_erase(&sp->so_server_node, &server->state_owners);
ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
}
static void
nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
{
sc->create_time = ktime_get();
sc->flags = 0;
sc->counter = 0;
spin_lock_init(&sc->lock);
INIT_LIST_HEAD(&sc->list);
rpc_init_wait_queue(&sc->wait, "Seqid_waitqueue");
}
static void
nfs4_destroy_seqid_counter(struct nfs_seqid_counter *sc)
{
rpc_destroy_wait_queue(&sc->wait);
}
/*
* nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
* create a new state_owner.
*
*/
static struct nfs4_state_owner *
nfs4_alloc_state_owner(struct nfs_server *server,
struct rpc_cred *cred,
gfp_t gfp_flags)
{
struct nfs4_state_owner *sp;
sp = kzalloc(sizeof(*sp), gfp_flags);
if (!sp)
return NULL;
sp->so_server = server;
sp->so_cred = get_rpccred(cred);
spin_lock_init(&sp->so_lock);
INIT_LIST_HEAD(&sp->so_states);
nfs4_init_seqid_counter(&sp->so_seqid);
atomic_set(&sp->so_count, 1);
NFS: Cache state owners after files are closed Servers have a finite amount of memory to store NFSv4 open and lock owners. Moreover, servers may have a difficult time determining when they can reap their state owner table, thanks to gray areas in the NFSv4 protocol specification. Thus clients should be careful to reuse state owners when possible. Currently Linux is not too careful. When a user has closed all her files on one mount point, the state owner's reference count goes to zero, and it is released. The next OPEN allocates a new one. A workload that serially opens and closes files can run through a large number of open owners this way. When a state owner's reference count goes to zero, slap it onto a free list for that nfs_server, with an expiry time. Garbage collect before looking for a state owner. This makes state owners for active users available for re-use. Now that there can be unused state owners remaining at umount time, purge the state owner free list when a server is destroyed. Also be sure not to reclaim unused state owners during state recovery. This change has benefits for the client as well. For some workloads, this approach drops the number of OPEN_CONFIRM calls from the same as the number of OPEN calls, down to just one. This reduces wire traffic and thus open(2) latency. Before this patch, untarring a kernel source tarball shows the OPEN_CONFIRM call counter steadily increasing through the test. With the patch, the OPEN_CONFIRM count remains at 1 throughout the entire untar. As long as the expiry time is kept short, I don't think garbage collection should be terribly expensive, although it does bounce the clp->cl_lock around a bit. [ At some point we should rationalize the use of the nfs_server ->destroy method. ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond: Fixed a garbage collection race and a few efficiency issues] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-12-07 05:13:48 +08:00
INIT_LIST_HEAD(&sp->so_lru);
seqcount_init(&sp->so_reclaim_seqcount);
mutex_init(&sp->so_delegreturn_mutex);
return sp;
}
static void
nfs4_reset_state_owner(struct nfs4_state_owner *sp)
{
/* This state_owner is no longer usable, but must
* remain in place so that state recovery can find it
* and the opens associated with it.
* It may also be used for new 'open' request to
* return a delegation to the server.
* So update the 'create_time' so that it looks like
* a new state_owner. This will cause the server to
* request an OPEN_CONFIRM to start a new sequence.
*/
sp->so_seqid.create_time = ktime_get();
}
NFS: Cache state owners after files are closed Servers have a finite amount of memory to store NFSv4 open and lock owners. Moreover, servers may have a difficult time determining when they can reap their state owner table, thanks to gray areas in the NFSv4 protocol specification. Thus clients should be careful to reuse state owners when possible. Currently Linux is not too careful. When a user has closed all her files on one mount point, the state owner's reference count goes to zero, and it is released. The next OPEN allocates a new one. A workload that serially opens and closes files can run through a large number of open owners this way. When a state owner's reference count goes to zero, slap it onto a free list for that nfs_server, with an expiry time. Garbage collect before looking for a state owner. This makes state owners for active users available for re-use. Now that there can be unused state owners remaining at umount time, purge the state owner free list when a server is destroyed. Also be sure not to reclaim unused state owners during state recovery. This change has benefits for the client as well. For some workloads, this approach drops the number of OPEN_CONFIRM calls from the same as the number of OPEN calls, down to just one. This reduces wire traffic and thus open(2) latency. Before this patch, untarring a kernel source tarball shows the OPEN_CONFIRM call counter steadily increasing through the test. With the patch, the OPEN_CONFIRM count remains at 1 throughout the entire untar. As long as the expiry time is kept short, I don't think garbage collection should be terribly expensive, although it does bounce the clp->cl_lock around a bit. [ At some point we should rationalize the use of the nfs_server ->destroy method. ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond: Fixed a garbage collection race and a few efficiency issues] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-12-07 05:13:48 +08:00
static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
{
nfs4_destroy_seqid_counter(&sp->so_seqid);
NFS: Cache state owners after files are closed Servers have a finite amount of memory to store NFSv4 open and lock owners. Moreover, servers may have a difficult time determining when they can reap their state owner table, thanks to gray areas in the NFSv4 protocol specification. Thus clients should be careful to reuse state owners when possible. Currently Linux is not too careful. When a user has closed all her files on one mount point, the state owner's reference count goes to zero, and it is released. The next OPEN allocates a new one. A workload that serially opens and closes files can run through a large number of open owners this way. When a state owner's reference count goes to zero, slap it onto a free list for that nfs_server, with an expiry time. Garbage collect before looking for a state owner. This makes state owners for active users available for re-use. Now that there can be unused state owners remaining at umount time, purge the state owner free list when a server is destroyed. Also be sure not to reclaim unused state owners during state recovery. This change has benefits for the client as well. For some workloads, this approach drops the number of OPEN_CONFIRM calls from the same as the number of OPEN calls, down to just one. This reduces wire traffic and thus open(2) latency. Before this patch, untarring a kernel source tarball shows the OPEN_CONFIRM call counter steadily increasing through the test. With the patch, the OPEN_CONFIRM count remains at 1 throughout the entire untar. As long as the expiry time is kept short, I don't think garbage collection should be terribly expensive, although it does bounce the clp->cl_lock around a bit. [ At some point we should rationalize the use of the nfs_server ->destroy method. ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond: Fixed a garbage collection race and a few efficiency issues] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-12-07 05:13:48 +08:00
put_rpccred(sp->so_cred);
kfree(sp);
}
static void nfs4_gc_state_owners(struct nfs_server *server)
{
struct nfs_client *clp = server->nfs_client;
struct nfs4_state_owner *sp, *tmp;
unsigned long time_min, time_max;
LIST_HEAD(doomed);
spin_lock(&clp->cl_lock);
time_max = jiffies;
time_min = (long)time_max - (long)clp->cl_lease_time;
list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
/* NB: LRU is sorted so that oldest is at the head */
if (time_in_range(sp->so_expires, time_min, time_max))
break;
list_move(&sp->so_lru, &doomed);
nfs4_remove_state_owner_locked(sp);
}
spin_unlock(&clp->cl_lock);
list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
list_del(&sp->so_lru);
nfs4_free_state_owner(sp);
}
}
/**
* nfs4_get_state_owner - Look up a state owner given a credential
* @server: nfs_server to search
* @cred: RPC credential to match
*
* Returns a pointer to an instantiated nfs4_state_owner struct, or NULL.
*/
struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server,
struct rpc_cred *cred,
gfp_t gfp_flags)
{
struct nfs_client *clp = server->nfs_client;
struct nfs4_state_owner *sp, *new;
spin_lock(&clp->cl_lock);
sp = nfs4_find_state_owner_locked(server, cred);
spin_unlock(&clp->cl_lock);
if (sp != NULL)
NFS: Cache state owners after files are closed Servers have a finite amount of memory to store NFSv4 open and lock owners. Moreover, servers may have a difficult time determining when they can reap their state owner table, thanks to gray areas in the NFSv4 protocol specification. Thus clients should be careful to reuse state owners when possible. Currently Linux is not too careful. When a user has closed all her files on one mount point, the state owner's reference count goes to zero, and it is released. The next OPEN allocates a new one. A workload that serially opens and closes files can run through a large number of open owners this way. When a state owner's reference count goes to zero, slap it onto a free list for that nfs_server, with an expiry time. Garbage collect before looking for a state owner. This makes state owners for active users available for re-use. Now that there can be unused state owners remaining at umount time, purge the state owner free list when a server is destroyed. Also be sure not to reclaim unused state owners during state recovery. This change has benefits for the client as well. For some workloads, this approach drops the number of OPEN_CONFIRM calls from the same as the number of OPEN calls, down to just one. This reduces wire traffic and thus open(2) latency. Before this patch, untarring a kernel source tarball shows the OPEN_CONFIRM call counter steadily increasing through the test. With the patch, the OPEN_CONFIRM count remains at 1 throughout the entire untar. As long as the expiry time is kept short, I don't think garbage collection should be terribly expensive, although it does bounce the clp->cl_lock around a bit. [ At some point we should rationalize the use of the nfs_server ->destroy method. ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond: Fixed a garbage collection race and a few efficiency issues] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-12-07 05:13:48 +08:00
goto out;
new = nfs4_alloc_state_owner(server, cred, gfp_flags);
if (new == NULL)
NFS: Cache state owners after files are closed Servers have a finite amount of memory to store NFSv4 open and lock owners. Moreover, servers may have a difficult time determining when they can reap their state owner table, thanks to gray areas in the NFSv4 protocol specification. Thus clients should be careful to reuse state owners when possible. Currently Linux is not too careful. When a user has closed all her files on one mount point, the state owner's reference count goes to zero, and it is released. The next OPEN allocates a new one. A workload that serially opens and closes files can run through a large number of open owners this way. When a state owner's reference count goes to zero, slap it onto a free list for that nfs_server, with an expiry time. Garbage collect before looking for a state owner. This makes state owners for active users available for re-use. Now that there can be unused state owners remaining at umount time, purge the state owner free list when a server is destroyed. Also be sure not to reclaim unused state owners during state recovery. This change has benefits for the client as well. For some workloads, this approach drops the number of OPEN_CONFIRM calls from the same as the number of OPEN calls, down to just one. This reduces wire traffic and thus open(2) latency. Before this patch, untarring a kernel source tarball shows the OPEN_CONFIRM call counter steadily increasing through the test. With the patch, the OPEN_CONFIRM count remains at 1 throughout the entire untar. As long as the expiry time is kept short, I don't think garbage collection should be terribly expensive, although it does bounce the clp->cl_lock around a bit. [ At some point we should rationalize the use of the nfs_server ->destroy method. ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond: Fixed a garbage collection race and a few efficiency issues] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-12-07 05:13:48 +08:00
goto out;
do {
if (ida_pre_get(&server->openowner_id, gfp_flags) == 0)
break;
spin_lock(&clp->cl_lock);
sp = nfs4_insert_state_owner_locked(new);
spin_unlock(&clp->cl_lock);
} while (sp == ERR_PTR(-EAGAIN));
if (sp != new)
nfs4_free_state_owner(new);
NFS: Cache state owners after files are closed Servers have a finite amount of memory to store NFSv4 open and lock owners. Moreover, servers may have a difficult time determining when they can reap their state owner table, thanks to gray areas in the NFSv4 protocol specification. Thus clients should be careful to reuse state owners when possible. Currently Linux is not too careful. When a user has closed all her files on one mount point, the state owner's reference count goes to zero, and it is released. The next OPEN allocates a new one. A workload that serially opens and closes files can run through a large number of open owners this way. When a state owner's reference count goes to zero, slap it onto a free list for that nfs_server, with an expiry time. Garbage collect before looking for a state owner. This makes state owners for active users available for re-use. Now that there can be unused state owners remaining at umount time, purge the state owner free list when a server is destroyed. Also be sure not to reclaim unused state owners during state recovery. This change has benefits for the client as well. For some workloads, this approach drops the number of OPEN_CONFIRM calls from the same as the number of OPEN calls, down to just one. This reduces wire traffic and thus open(2) latency. Before this patch, untarring a kernel source tarball shows the OPEN_CONFIRM call counter steadily increasing through the test. With the patch, the OPEN_CONFIRM count remains at 1 throughout the entire untar. As long as the expiry time is kept short, I don't think garbage collection should be terribly expensive, although it does bounce the clp->cl_lock around a bit. [ At some point we should rationalize the use of the nfs_server ->destroy method. ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond: Fixed a garbage collection race and a few efficiency issues] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-12-07 05:13:48 +08:00
out:
nfs4_gc_state_owners(server);
return sp;
}
/**
* nfs4_put_state_owner - Release a nfs4_state_owner
* @sp: state owner data to release
*
* Note that we keep released state owners on an LRU
* list.
* This caches valid state owners so that they can be
* reused, to avoid the OPEN_CONFIRM on minor version 0.
* It also pins the uniquifier of dropped state owners for
* a while, to ensure that those state owner names are
* never reused.
*/
void nfs4_put_state_owner(struct nfs4_state_owner *sp)
{
NFS: Cache state owners after files are closed Servers have a finite amount of memory to store NFSv4 open and lock owners. Moreover, servers may have a difficult time determining when they can reap their state owner table, thanks to gray areas in the NFSv4 protocol specification. Thus clients should be careful to reuse state owners when possible. Currently Linux is not too careful. When a user has closed all her files on one mount point, the state owner's reference count goes to zero, and it is released. The next OPEN allocates a new one. A workload that serially opens and closes files can run through a large number of open owners this way. When a state owner's reference count goes to zero, slap it onto a free list for that nfs_server, with an expiry time. Garbage collect before looking for a state owner. This makes state owners for active users available for re-use. Now that there can be unused state owners remaining at umount time, purge the state owner free list when a server is destroyed. Also be sure not to reclaim unused state owners during state recovery. This change has benefits for the client as well. For some workloads, this approach drops the number of OPEN_CONFIRM calls from the same as the number of OPEN calls, down to just one. This reduces wire traffic and thus open(2) latency. Before this patch, untarring a kernel source tarball shows the OPEN_CONFIRM call counter steadily increasing through the test. With the patch, the OPEN_CONFIRM count remains at 1 throughout the entire untar. As long as the expiry time is kept short, I don't think garbage collection should be terribly expensive, although it does bounce the clp->cl_lock around a bit. [ At some point we should rationalize the use of the nfs_server ->destroy method. ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond: Fixed a garbage collection race and a few efficiency issues] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-12-07 05:13:48 +08:00
struct nfs_server *server = sp->so_server;
struct nfs_client *clp = server->nfs_client;
if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
return;
NFS: Cache state owners after files are closed Servers have a finite amount of memory to store NFSv4 open and lock owners. Moreover, servers may have a difficult time determining when they can reap their state owner table, thanks to gray areas in the NFSv4 protocol specification. Thus clients should be careful to reuse state owners when possible. Currently Linux is not too careful. When a user has closed all her files on one mount point, the state owner's reference count goes to zero, and it is released. The next OPEN allocates a new one. A workload that serially opens and closes files can run through a large number of open owners this way. When a state owner's reference count goes to zero, slap it onto a free list for that nfs_server, with an expiry time. Garbage collect before looking for a state owner. This makes state owners for active users available for re-use. Now that there can be unused state owners remaining at umount time, purge the state owner free list when a server is destroyed. Also be sure not to reclaim unused state owners during state recovery. This change has benefits for the client as well. For some workloads, this approach drops the number of OPEN_CONFIRM calls from the same as the number of OPEN calls, down to just one. This reduces wire traffic and thus open(2) latency. Before this patch, untarring a kernel source tarball shows the OPEN_CONFIRM call counter steadily increasing through the test. With the patch, the OPEN_CONFIRM count remains at 1 throughout the entire untar. As long as the expiry time is kept short, I don't think garbage collection should be terribly expensive, although it does bounce the clp->cl_lock around a bit. [ At some point we should rationalize the use of the nfs_server ->destroy method. ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond: Fixed a garbage collection race and a few efficiency issues] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-12-07 05:13:48 +08:00
sp->so_expires = jiffies;
list_add_tail(&sp->so_lru, &server->state_owners_lru);
spin_unlock(&clp->cl_lock);
NFS: Cache state owners after files are closed Servers have a finite amount of memory to store NFSv4 open and lock owners. Moreover, servers may have a difficult time determining when they can reap their state owner table, thanks to gray areas in the NFSv4 protocol specification. Thus clients should be careful to reuse state owners when possible. Currently Linux is not too careful. When a user has closed all her files on one mount point, the state owner's reference count goes to zero, and it is released. The next OPEN allocates a new one. A workload that serially opens and closes files can run through a large number of open owners this way. When a state owner's reference count goes to zero, slap it onto a free list for that nfs_server, with an expiry time. Garbage collect before looking for a state owner. This makes state owners for active users available for re-use. Now that there can be unused state owners remaining at umount time, purge the state owner free list when a server is destroyed. Also be sure not to reclaim unused state owners during state recovery. This change has benefits for the client as well. For some workloads, this approach drops the number of OPEN_CONFIRM calls from the same as the number of OPEN calls, down to just one. This reduces wire traffic and thus open(2) latency. Before this patch, untarring a kernel source tarball shows the OPEN_CONFIRM call counter steadily increasing through the test. With the patch, the OPEN_CONFIRM count remains at 1 throughout the entire untar. As long as the expiry time is kept short, I don't think garbage collection should be terribly expensive, although it does bounce the clp->cl_lock around a bit. [ At some point we should rationalize the use of the nfs_server ->destroy method. ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond: Fixed a garbage collection race and a few efficiency issues] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-12-07 05:13:48 +08:00
}
/**
* nfs4_purge_state_owners - Release all cached state owners
* @server: nfs_server with cached state owners to release
*
* Called at umount time. Remaining state owners will be on
* the LRU with ref count of zero.
*/
void nfs4_purge_state_owners(struct nfs_server *server)
{
struct nfs_client *clp = server->nfs_client;
struct nfs4_state_owner *sp, *tmp;
LIST_HEAD(doomed);
spin_lock(&clp->cl_lock);
list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
list_move(&sp->so_lru, &doomed);
nfs4_remove_state_owner_locked(sp);
}
spin_unlock(&clp->cl_lock);
NFS: Cache state owners after files are closed Servers have a finite amount of memory to store NFSv4 open and lock owners. Moreover, servers may have a difficult time determining when they can reap their state owner table, thanks to gray areas in the NFSv4 protocol specification. Thus clients should be careful to reuse state owners when possible. Currently Linux is not too careful. When a user has closed all her files on one mount point, the state owner's reference count goes to zero, and it is released. The next OPEN allocates a new one. A workload that serially opens and closes files can run through a large number of open owners this way. When a state owner's reference count goes to zero, slap it onto a free list for that nfs_server, with an expiry time. Garbage collect before looking for a state owner. This makes state owners for active users available for re-use. Now that there can be unused state owners remaining at umount time, purge the state owner free list when a server is destroyed. Also be sure not to reclaim unused state owners during state recovery. This change has benefits for the client as well. For some workloads, this approach drops the number of OPEN_CONFIRM calls from the same as the number of OPEN calls, down to just one. This reduces wire traffic and thus open(2) latency. Before this patch, untarring a kernel source tarball shows the OPEN_CONFIRM call counter steadily increasing through the test. With the patch, the OPEN_CONFIRM count remains at 1 throughout the entire untar. As long as the expiry time is kept short, I don't think garbage collection should be terribly expensive, although it does bounce the clp->cl_lock around a bit. [ At some point we should rationalize the use of the nfs_server ->destroy method. ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond: Fixed a garbage collection race and a few efficiency issues] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-12-07 05:13:48 +08:00
list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
list_del(&sp->so_lru);
nfs4_free_state_owner(sp);
}
}
static struct nfs4_state *
nfs4_alloc_open_state(void)
{
struct nfs4_state *state;
state = kzalloc(sizeof(*state), GFP_NOFS);
if (!state)
return NULL;
atomic_set(&state->count, 1);
INIT_LIST_HEAD(&state->lock_states);
spin_lock_init(&state->state_lock);
seqlock_init(&state->seqlock);
return state;
}
void
nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
{
if (state->state == fmode)
return;
/* NB! List reordering - see the reclaim code for why. */
if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
if (fmode & FMODE_WRITE)
list_move(&state->open_states, &state->owner->so_states);
else
list_move_tail(&state->open_states, &state->owner->so_states);
}
state->state = fmode;
}
static struct nfs4_state *
__nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
{
struct nfs_inode *nfsi = NFS_I(inode);
struct nfs4_state *state;
list_for_each_entry(state, &nfsi->open_states, inode_states) {
if (state->owner != owner)
continue;
if (!nfs4_valid_open_stateid(state))
continue;
if (atomic_inc_not_zero(&state->count))
return state;
}
return NULL;
}
static void
nfs4_free_open_state(struct nfs4_state *state)
{
kfree(state);
}
struct nfs4_state *
nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
{
struct nfs4_state *state, *new;
struct nfs_inode *nfsi = NFS_I(inode);
spin_lock(&inode->i_lock);
state = __nfs4_find_state_byowner(inode, owner);
spin_unlock(&inode->i_lock);
if (state)
goto out;
new = nfs4_alloc_open_state();
spin_lock(&owner->so_lock);
spin_lock(&inode->i_lock);
state = __nfs4_find_state_byowner(inode, owner);
if (state == NULL && new != NULL) {
state = new;
state->owner = owner;
atomic_inc(&owner->so_count);
list_add(&state->inode_states, &nfsi->open_states);
ihold(inode);
state->inode = inode;
spin_unlock(&inode->i_lock);
/* Note: The reclaim code dictates that we add stateless
* and read-only stateids to the end of the list */
list_add_tail(&state->open_states, &owner->so_states);
spin_unlock(&owner->so_lock);
} else {
spin_unlock(&inode->i_lock);
spin_unlock(&owner->so_lock);
if (new)
nfs4_free_open_state(new);
}
out:
return state;
}
void nfs4_put_open_state(struct nfs4_state *state)
{
struct inode *inode = state->inode;
struct nfs4_state_owner *owner = state->owner;
if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
return;
spin_lock(&inode->i_lock);
list_del(&state->inode_states);
list_del(&state->open_states);
spin_unlock(&inode->i_lock);
spin_unlock(&owner->so_lock);
iput(inode);
nfs4_free_open_state(state);
nfs4_put_state_owner(owner);
}
/*
* Close the current file.
*/
static void __nfs4_close(struct nfs4_state *state,
fmode_t fmode, gfp_t gfp_mask, int wait)
{
struct nfs4_state_owner *owner = state->owner;
int call_close = 0;
fmode_t newstate;
atomic_inc(&owner->so_count);
/* Protect against nfs4_find_state() */
spin_lock(&owner->so_lock);
switch (fmode & (FMODE_READ | FMODE_WRITE)) {
case FMODE_READ:
state->n_rdonly--;
break;
case FMODE_WRITE:
state->n_wronly--;
break;
case FMODE_READ|FMODE_WRITE:
state->n_rdwr--;
}
newstate = FMODE_READ|FMODE_WRITE;
if (state->n_rdwr == 0) {
if (state->n_rdonly == 0) {
newstate &= ~FMODE_READ;
call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
}
if (state->n_wronly == 0) {
newstate &= ~FMODE_WRITE;
call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
}
if (newstate == 0)
clear_bit(NFS_DELEGATED_STATE, &state->flags);
}
nfs4_state_set_mode_locked(state, newstate);
spin_unlock(&owner->so_lock);
if (!call_close) {
nfs4_put_open_state(state);
nfs4_put_state_owner(owner);
} else
nfs4_do_close(state, gfp_mask, wait);
}
void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
{
__nfs4_close(state, fmode, GFP_NOFS, 0);
}
void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
{
__nfs4_close(state, fmode, GFP_KERNEL, 1);
}
/*
* Search the state->lock_states for an existing lock_owner
* that is compatible with either of the given owners.
* If the second is non-zero, then the first refers to a Posix-lock
* owner (current->files) and the second refers to a flock/OFD
* owner (struct file*). In that case, prefer a match for the first
* owner.
* If both sorts of locks are held on the one file we cannot know
* which stateid was intended to be used, so a "correct" choice cannot
* be made. Failing that, a "consistent" choice is preferable. The
* consistent choice we make is to prefer the first owner, that of a
* Posix lock.
*/
static struct nfs4_lock_state *
__nfs4_find_lock_state(struct nfs4_state *state,
fl_owner_t fl_owner, fl_owner_t fl_owner2)
{
struct nfs4_lock_state *pos, *ret = NULL;
list_for_each_entry(pos, &state->lock_states, ls_locks) {
if (pos->ls_owner == fl_owner) {
ret = pos;
break;
}
if (pos->ls_owner == fl_owner2)
ret = pos;
}
if (ret)
atomic_inc(&ret->ls_count);
return ret;
}
/*
* Return a compatible lock_state. If no initialized lock_state structure
* exists, return an uninitialized one.
*
*/
static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
{
struct nfs4_lock_state *lsp;
struct nfs_server *server = state->owner->so_server;
lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
if (lsp == NULL)
return NULL;
nfs4_init_seqid_counter(&lsp->ls_seqid);
atomic_set(&lsp->ls_count, 1);
lsp->ls_state = state;
lsp->ls_owner = fl_owner;
lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
if (lsp->ls_seqid.owner_id < 0)
goto out_free;
INIT_LIST_HEAD(&lsp->ls_locks);
return lsp;
out_free:
kfree(lsp);
return NULL;
}
void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
{
ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
nfs4_destroy_seqid_counter(&lsp->ls_seqid);
kfree(lsp);
}
/*
* Return a compatible lock_state. If no initialized lock_state structure
* exists, return an uninitialized one.
*
*/
static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
{
struct nfs4_lock_state *lsp, *new = NULL;
for(;;) {
spin_lock(&state->state_lock);
lsp = __nfs4_find_lock_state(state, owner, NULL);
if (lsp != NULL)
break;
if (new != NULL) {
list_add(&new->ls_locks, &state->lock_states);
set_bit(LK_STATE_IN_USE, &state->flags);
lsp = new;
new = NULL;
break;
}
spin_unlock(&state->state_lock);
new = nfs4_alloc_lock_state(state, owner);
if (new == NULL)
return NULL;
}
spin_unlock(&state->state_lock);
if (new != NULL)
nfs4_free_lock_state(state->owner->so_server, new);
return lsp;
}
/*
* Release reference to lock_state, and free it if we see that
* it is no longer in use
*/
void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
{
struct nfs_server *server;
struct nfs4_state *state;
if (lsp == NULL)
return;
state = lsp->ls_state;
if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
return;
list_del(&lsp->ls_locks);
if (list_empty(&state->lock_states))
clear_bit(LK_STATE_IN_USE, &state->flags);
spin_unlock(&state->state_lock);
nfs: revert "nfs4: queue free_lock_state job submission to nfsiod" This reverts commit 49a4bda22e186c4d0eb07f4a36b5b1a378f9398d. Christoph reported an oops due to the above commit: generic/089 242s ...[ 2187.041239] general protection fault: 0000 [#1] SMP [ 2187.042899] Modules linked in: [ 2187.044000] CPU: 0 PID: 11913 Comm: kworker/0:1 Not tainted 3.16.0-rc6+ #1151 [ 2187.044287] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2007 [ 2187.044287] Workqueue: nfsiod free_lock_state_work [ 2187.044287] task: ffff880072b50cd0 ti: ffff88007a4ec000 task.ti: ffff88007a4ec000 [ 2187.044287] RIP: 0010:[<ffffffff81361ca6>] [<ffffffff81361ca6>] free_lock_state_work+0x16/0x30 [ 2187.044287] RSP: 0018:ffff88007a4efd58 EFLAGS: 00010296 [ 2187.044287] RAX: 6b6b6b6b6b6b6b6b RBX: ffff88007a947ac0 RCX: 8000000000000000 [ 2187.044287] RDX: ffffffff826af9e0 RSI: ffff88007b093c00 RDI: ffff88007b093db8 [ 2187.044287] RBP: ffff88007a4efd58 R08: ffffffff832d3e10 R09: 000001c40efc0000 [ 2187.044287] R10: 0000000000000000 R11: 0000000000059e30 R12: ffff88007fc13240 [ 2187.044287] R13: ffff88007fc18b00 R14: ffff88007b093db8 R15: 0000000000000000 [ 2187.044287] FS: 0000000000000000(0000) GS:ffff88007fc00000(0000) knlGS:0000000000000000 [ 2187.044287] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b [ 2187.044287] CR2: 00007f93ec33fb80 CR3: 0000000079dc2000 CR4: 00000000000006f0 [ 2187.044287] Stack: [ 2187.044287] ffff88007a4efdd8 ffffffff810cc877 ffffffff810cc80d ffff88007fc13258 [ 2187.044287] 000000007a947af0 0000000000000000 ffffffff8353ccc8 ffffffff82b6f3d0 [ 2187.044287] 0000000000000000 ffffffff82267679 ffff88007a4efdd8 ffff88007fc13240 [ 2187.044287] Call Trace: [ 2187.044287] [<ffffffff810cc877>] process_one_work+0x1c7/0x490 [ 2187.044287] [<ffffffff810cc80d>] ? process_one_work+0x15d/0x490 [ 2187.044287] [<ffffffff810cd569>] worker_thread+0x119/0x4f0 [ 2187.044287] [<ffffffff810fbbad>] ? trace_hardirqs_on+0xd/0x10 [ 2187.044287] [<ffffffff810cd450>] ? init_pwq+0x190/0x190 [ 2187.044287] [<ffffffff810d3c6f>] kthread+0xdf/0x100 [ 2187.044287] [<ffffffff810d3b90>] ? __init_kthread_worker+0x70/0x70 [ 2187.044287] [<ffffffff81d9873c>] ret_from_fork+0x7c/0xb0 [ 2187.044287] [<ffffffff810d3b90>] ? __init_kthread_worker+0x70/0x70 [ 2187.044287] Code: 0f 1f 44 00 00 31 c0 5d c3 66 66 66 2e 0f 1f 84 00 00 00 00 00 55 48 8d b7 48 fe ff ff 48 8b 87 58 fe ff ff 48 89 e5 48 8b 40 30 <48> 8b 00 48 8b 10 48 89 c7 48 8b 92 90 03 00 00 ff 52 28 5d c3 [ 2187.044287] RIP [<ffffffff81361ca6>] free_lock_state_work+0x16/0x30 [ 2187.044287] RSP <ffff88007a4efd58> [ 2187.103626] ---[ end trace 0f11326d28e5d8fa ]--- The original reason for this patch was because the fl_release_private operation couldn't sleep. With commit ed9814d85810 (locks: defer freeing locks in locks_delete_lock until after i_lock has been dropped), this is no longer a problem so we can revert this patch. Reported-by: Christoph Hellwig <hch@infradead.org> Signed-off-by: Jeff Layton <jlayton@primarydata.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Tested-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2014-09-08 20:26:01 +08:00
server = state->owner->so_server;
if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
struct nfs_client *clp = server->nfs_client;
clp->cl_mvops->free_lock_state(server, lsp);
} else
nfs4_free_lock_state(server, lsp);
}
static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
{
struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
dst->fl_u.nfs4_fl.owner = lsp;
atomic_inc(&lsp->ls_count);
}
static void nfs4_fl_release_lock(struct file_lock *fl)
{
nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
}
static const struct file_lock_operations nfs4_fl_lock_ops = {
.fl_copy_lock = nfs4_fl_copy_lock,
.fl_release_private = nfs4_fl_release_lock,
};
int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
{
struct nfs4_lock_state *lsp;
if (fl->fl_ops != NULL)
return 0;
lsp = nfs4_get_lock_state(state, fl->fl_owner);
if (lsp == NULL)
return -ENOMEM;
fl->fl_u.nfs4_fl.owner = lsp;
fl->fl_ops = &nfs4_fl_lock_ops;
return 0;
}
static int nfs4_copy_lock_stateid(nfs4_stateid *dst,
struct nfs4_state *state,
const struct nfs_lock_context *l_ctx)
{
struct nfs4_lock_state *lsp;
fl_owner_t fl_owner, fl_flock_owner;
int ret = -ENOENT;
if (l_ctx == NULL)
goto out;
if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
goto out;
fl_owner = l_ctx->lockowner;
fl_flock_owner = l_ctx->open_context->flock_owner;
spin_lock(&state->state_lock);
lsp = __nfs4_find_lock_state(state, fl_owner, fl_flock_owner);
NFSv4: Don't try to recover NFSv4 locks when they are lost. When an NFSv4 client loses contact with the server it can lose any locks that it holds. Currently when it reconnects to the server it simply tries to reclaim those locks. This might succeed even though some other client has held and released a lock in the mean time. So the first client might think the file is unchanged, but it isn't. This isn't good. If, when recovery happens, the locks cannot be claimed because some other client still holds the lock, then we get a message in the kernel logs, but the client can still write. So two clients can both think they have a lock and can both write at the same time. This is equally not good. There was a patch a while ago http://comments.gmane.org/gmane.linux.nfs/41917 which tried to address some of this, but it didn't seem to go anywhere. That patch would also send a signal to the process. That might be useful but for now this patch just causes writes to fail. For NFSv4 (unlike v2/v3) there is a strong link between the lock and the write request so we can fairly easily fail any IO of the lock is gone. While some applications might not expect this, it is still safer than allowing the write to succeed. Because this is a fairly big change in behaviour a module parameter, "recover_locks", is introduced which defaults to true (the current behaviour) but can be set to "false" to tell the client not to try to recover things that were lost. Signed-off-by: NeilBrown <neilb@suse.de> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2013-09-04 15:04:49 +08:00
if (lsp && test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
ret = -EIO;
else if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
nfs4_stateid_copy(dst, &lsp->ls_stateid);
ret = 0;
}
spin_unlock(&state->state_lock);
nfs4_put_lock_state(lsp);
out:
return ret;
}
static void nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
{
const nfs4_stateid *src;
int seq;
do {
src = &zero_stateid;
seq = read_seqbegin(&state->seqlock);
if (test_bit(NFS_OPEN_STATE, &state->flags))
src = &state->open_stateid;
nfs4_stateid_copy(dst, src);
} while (read_seqretry(&state->seqlock, seq));
}
/*
* Byte-range lock aware utility to initialize the stateid of read/write
* requests.
*/
int nfs4_select_rw_stateid(struct nfs4_state *state,
fmode_t fmode, const struct nfs_lock_context *l_ctx,
nfs4_stateid *dst, struct rpc_cred **cred)
{
int ret;
if (!nfs4_valid_open_stateid(state))
return -EIO;
if (cred != NULL)
*cred = NULL;
ret = nfs4_copy_lock_stateid(dst, state, l_ctx);
NFSv4: Don't try to recover NFSv4 locks when they are lost. When an NFSv4 client loses contact with the server it can lose any locks that it holds. Currently when it reconnects to the server it simply tries to reclaim those locks. This might succeed even though some other client has held and released a lock in the mean time. So the first client might think the file is unchanged, but it isn't. This isn't good. If, when recovery happens, the locks cannot be claimed because some other client still holds the lock, then we get a message in the kernel logs, but the client can still write. So two clients can both think they have a lock and can both write at the same time. This is equally not good. There was a patch a while ago http://comments.gmane.org/gmane.linux.nfs/41917 which tried to address some of this, but it didn't seem to go anywhere. That patch would also send a signal to the process. That might be useful but for now this patch just causes writes to fail. For NFSv4 (unlike v2/v3) there is a strong link between the lock and the write request so we can fairly easily fail any IO of the lock is gone. While some applications might not expect this, it is still safer than allowing the write to succeed. Because this is a fairly big change in behaviour a module parameter, "recover_locks", is introduced which defaults to true (the current behaviour) but can be set to "false" to tell the client not to try to recover things that were lost. Signed-off-by: NeilBrown <neilb@suse.de> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2013-09-04 15:04:49 +08:00
if (ret == -EIO)
/* A lost lock - don't even consider delegations */
goto out;
/* returns true if delegation stateid found and copied */
if (nfs4_copy_delegation_stateid(state->inode, fmode, dst, cred)) {
ret = 0;
goto out;
}
if (ret != -ENOENT)
NFSv4: Don't try to recover NFSv4 locks when they are lost. When an NFSv4 client loses contact with the server it can lose any locks that it holds. Currently when it reconnects to the server it simply tries to reclaim those locks. This might succeed even though some other client has held and released a lock in the mean time. So the first client might think the file is unchanged, but it isn't. This isn't good. If, when recovery happens, the locks cannot be claimed because some other client still holds the lock, then we get a message in the kernel logs, but the client can still write. So two clients can both think they have a lock and can both write at the same time. This is equally not good. There was a patch a while ago http://comments.gmane.org/gmane.linux.nfs/41917 which tried to address some of this, but it didn't seem to go anywhere. That patch would also send a signal to the process. That might be useful but for now this patch just causes writes to fail. For NFSv4 (unlike v2/v3) there is a strong link between the lock and the write request so we can fairly easily fail any IO of the lock is gone. While some applications might not expect this, it is still safer than allowing the write to succeed. Because this is a fairly big change in behaviour a module parameter, "recover_locks", is introduced which defaults to true (the current behaviour) but can be set to "false" to tell the client not to try to recover things that were lost. Signed-off-by: NeilBrown <neilb@suse.de> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2013-09-04 15:04:49 +08:00
/* nfs4_copy_delegation_stateid() didn't over-write
* dst, so it still has the lock stateid which we now
* choose to use.
*/
goto out;
nfs4_copy_open_stateid(dst, state);
ret = 0;
out:
if (nfs_server_capable(state->inode, NFS_CAP_STATEID_NFSV41))
dst->seqid = 0;
return ret;
}
struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
{
struct nfs_seqid *new;
new = kmalloc(sizeof(*new), gfp_mask);
if (new == NULL)
return ERR_PTR(-ENOMEM);
new->sequence = counter;
INIT_LIST_HEAD(&new->list);
new->task = NULL;
return new;
}
void nfs_release_seqid(struct nfs_seqid *seqid)
{
struct nfs_seqid_counter *sequence;
if (seqid == NULL || list_empty(&seqid->list))
return;
sequence = seqid->sequence;
spin_lock(&sequence->lock);
list_del_init(&seqid->list);
if (!list_empty(&sequence->list)) {
struct nfs_seqid *next;
next = list_first_entry(&sequence->list,
struct nfs_seqid, list);
rpc_wake_up_queued_task(&sequence->wait, next->task);
}
spin_unlock(&sequence->lock);
}
void nfs_free_seqid(struct nfs_seqid *seqid)
{
nfs_release_seqid(seqid);
kfree(seqid);
}
/*
* Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
* failed with a seqid incrementing error -
* see comments nfs4.h:seqid_mutating_error()
*/
static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
{
switch (status) {
case 0:
break;
case -NFS4ERR_BAD_SEQID:
if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
return;
pr_warn_ratelimited("NFS: v4 server returned a bad"
" sequence-id error on an"
" unconfirmed sequence %p!\n",
seqid->sequence);
case -NFS4ERR_STALE_CLIENTID:
case -NFS4ERR_STALE_STATEID:
case -NFS4ERR_BAD_STATEID:
case -NFS4ERR_BADXDR:
case -NFS4ERR_RESOURCE:
case -NFS4ERR_NOFILEHANDLE:
case -NFS4ERR_MOVED:
/* Non-seqid mutating errors */
return;
};
/*
* Note: no locking needed as we are guaranteed to be first
* on the sequence list
*/
seqid->sequence->counter++;
}
void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
{
struct nfs4_state_owner *sp;
if (seqid == NULL)
return;
sp = container_of(seqid->sequence, struct nfs4_state_owner, so_seqid);
if (status == -NFS4ERR_BAD_SEQID)
nfs4_reset_state_owner(sp);
if (!nfs4_has_session(sp->so_server->nfs_client))
nfs_increment_seqid(status, seqid);
}
/*
* Increment the seqid if the LOCK/LOCKU succeeded, or
* failed with a seqid incrementing error -
* see comments nfs4.h:seqid_mutating_error()
*/
void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
{
if (seqid != NULL)
nfs_increment_seqid(status, seqid);
}
int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
{
struct nfs_seqid_counter *sequence;
int status = 0;
if (seqid == NULL)
goto out;
sequence = seqid->sequence;
spin_lock(&sequence->lock);
seqid->task = task;
if (list_empty(&seqid->list))
list_add_tail(&seqid->list, &sequence->list);
if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
goto unlock;
rpc_sleep_on(&sequence->wait, task, NULL);
status = -EAGAIN;
unlock:
spin_unlock(&sequence->lock);
out:
return status;
}
static int nfs4_run_state_manager(void *);
static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
{
smp_mb__before_atomic();
clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
smp_mb__after_atomic();
wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
rpc_wake_up(&clp->cl_rpcwaitq);
}
/*
* Schedule the nfs_client asynchronous state management routine
*/
void nfs4_schedule_state_manager(struct nfs_client *clp)
{
struct task_struct *task;
char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
return;
__module_get(THIS_MODULE);
atomic_inc(&clp->cl_count);
/* The rcu_read_lock() is not strictly necessary, as the state
* manager is the only thread that ever changes the rpc_xprt
* after it's initialized. At this point, we're single threaded. */
rcu_read_lock();
snprintf(buf, sizeof(buf), "%s-manager",
rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR));
rcu_read_unlock();
task = kthread_run(nfs4_run_state_manager, clp, "%s", buf);
if (IS_ERR(task)) {
printk(KERN_ERR "%s: kthread_run: %ld\n",
__func__, PTR_ERR(task));
nfs4_clear_state_manager_bit(clp);
nfs_put_client(clp);
module_put(THIS_MODULE);
}
}
/*
* Schedule a lease recovery attempt
*/
void nfs4_schedule_lease_recovery(struct nfs_client *clp)
{
if (!clp)
return;
if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
dprintk("%s: scheduling lease recovery for server %s\n", __func__,
clp->cl_hostname);
nfs4_schedule_state_manager(clp);
}
EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
/**
* nfs4_schedule_migration_recovery - trigger migration recovery
*
* @server: FSID that is migrating
*
* Returns zero if recovery has started, otherwise a negative NFS4ERR
* value is returned.
*/
int nfs4_schedule_migration_recovery(const struct nfs_server *server)
{
struct nfs_client *clp = server->nfs_client;
if (server->fh_expire_type != NFS4_FH_PERSISTENT) {
pr_err("NFS: volatile file handles not supported (server %s)\n",
clp->cl_hostname);
return -NFS4ERR_IO;
}
if (test_bit(NFS_MIG_FAILED, &server->mig_status))
return -NFS4ERR_IO;
dprintk("%s: scheduling migration recovery for (%llx:%llx) on %s\n",
__func__,
(unsigned long long)server->fsid.major,
(unsigned long long)server->fsid.minor,
clp->cl_hostname);
set_bit(NFS_MIG_IN_TRANSITION,
&((struct nfs_server *)server)->mig_status);
set_bit(NFS4CLNT_MOVED, &clp->cl_state);
nfs4_schedule_state_manager(clp);
return 0;
}
EXPORT_SYMBOL_GPL(nfs4_schedule_migration_recovery);
/**
* nfs4_schedule_lease_moved_recovery - start lease-moved recovery
*
* @clp: server to check for moved leases
*
*/
void nfs4_schedule_lease_moved_recovery(struct nfs_client *clp)
{
dprintk("%s: scheduling lease-moved recovery for client ID %llx on %s\n",
__func__, clp->cl_clientid, clp->cl_hostname);
set_bit(NFS4CLNT_LEASE_MOVED, &clp->cl_state);
nfs4_schedule_state_manager(clp);
}
EXPORT_SYMBOL_GPL(nfs4_schedule_lease_moved_recovery);
int nfs4_wait_clnt_recover(struct nfs_client *clp)
{
int res;
might_sleep();
atomic_inc(&clp->cl_count);
sched: Remove proliferation of wait_on_bit() action functions The current "wait_on_bit" interface requires an 'action' function to be provided which does the actual waiting. There are over 20 such functions, many of them identical. Most cases can be satisfied by one of just two functions, one which uses io_schedule() and one which just uses schedule(). So: Rename wait_on_bit and wait_on_bit_lock to wait_on_bit_action and wait_on_bit_lock_action to make it explicit that they need an action function. Introduce new wait_on_bit{,_lock} and wait_on_bit{,_lock}_io which are *not* given an action function but implicitly use a standard one. The decision to error-out if a signal is pending is now made based on the 'mode' argument rather than being encoded in the action function. All instances of the old wait_on_bit and wait_on_bit_lock which can use the new version have been changed accordingly and their action functions have been discarded. wait_on_bit{_lock} does not return any specific error code in the event of a signal so the caller must check for non-zero and interpolate their own error code as appropriate. The wait_on_bit() call in __fscache_wait_on_invalidate() was ambiguous as it specified TASK_UNINTERRUPTIBLE but used fscache_wait_bit_interruptible as an action function. David Howells confirms this should be uniformly "uninterruptible" The main remaining user of wait_on_bit{,_lock}_action is NFS which needs to use a freezer-aware schedule() call. A comment in fs/gfs2/glock.c notes that having multiple 'action' functions is useful as they display differently in the 'wchan' field of 'ps'. (and /proc/$PID/wchan). As the new bit_wait{,_io} functions are tagged "__sched", they will not show up at all, but something higher in the stack. So the distinction will still be visible, only with different function names (gds2_glock_wait versus gfs2_glock_dq_wait in the gfs2/glock.c case). Since first version of this patch (against 3.15) two new action functions appeared, on in NFS and one in CIFS. CIFS also now uses an action function that makes the same freezer aware schedule call as NFS. Signed-off-by: NeilBrown <neilb@suse.de> Acked-by: David Howells <dhowells@redhat.com> (fscache, keys) Acked-by: Steven Whitehouse <swhiteho@redhat.com> (gfs2) Acked-by: Peter Zijlstra <peterz@infradead.org> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Steve French <sfrench@samba.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Link: http://lkml.kernel.org/r/20140707051603.28027.72349.stgit@notabene.brown Signed-off-by: Ingo Molnar <mingo@kernel.org>
2014-07-07 13:16:04 +08:00
res = wait_on_bit_action(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
nfs_wait_bit_killable, TASK_KILLABLE);
if (res)
goto out;
if (clp->cl_cons_state < 0)
res = clp->cl_cons_state;
out:
nfs_put_client(clp);
return res;
}
int nfs4_client_recover_expired_lease(struct nfs_client *clp)
{
unsigned int loop;
int ret;
for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
ret = nfs4_wait_clnt_recover(clp);
if (ret != 0)
break;
if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
!test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
break;
nfs4_schedule_state_manager(clp);
ret = -EIO;
}
return ret;
}
/*
* nfs40_handle_cb_pathdown - return all delegations after NFS4ERR_CB_PATH_DOWN
* @clp: client to process
*
* Set the NFS4CLNT_LEASE_EXPIRED state in order to force a
* resend of the SETCLIENTID and hence re-establish the
* callback channel. Then return all existing delegations.
*/
static void nfs40_handle_cb_pathdown(struct nfs_client *clp)
{
set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
nfs_expire_all_delegations(clp);
dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
clp->cl_hostname);
}
void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
{
nfs40_handle_cb_pathdown(clp);
nfs4_schedule_state_manager(clp);
}
static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
{
if (!nfs4_valid_open_stateid(state))
return 0;
set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
/* Don't recover state that expired before the reboot */
if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
return 0;
}
set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
return 1;
}
int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
{
if (!nfs4_valid_open_stateid(state))
return 0;
set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
return 1;
}
int nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
{
struct nfs_client *clp = server->nfs_client;
if (!nfs4_state_mark_reclaim_nograce(clp, state))
return -EBADF;
dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
clp->cl_hostname);
nfs4_schedule_state_manager(clp);
return 0;
}
EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
static struct nfs4_lock_state *
nfs_state_find_lock_state_by_stateid(struct nfs4_state *state,
const nfs4_stateid *stateid)
{
struct nfs4_lock_state *pos;
list_for_each_entry(pos, &state->lock_states, ls_locks) {
if (!test_bit(NFS_LOCK_INITIALIZED, &pos->ls_flags))
continue;
if (nfs4_stateid_match_other(&pos->ls_stateid, stateid))
return pos;
}
return NULL;
}
static bool nfs_state_lock_state_matches_stateid(struct nfs4_state *state,
const nfs4_stateid *stateid)
{
bool found = false;
if (test_bit(LK_STATE_IN_USE, &state->flags)) {
spin_lock(&state->state_lock);
if (nfs_state_find_lock_state_by_stateid(state, stateid))
found = true;
spin_unlock(&state->state_lock);
}
return found;
}
void nfs_inode_find_state_and_recover(struct inode *inode,
const nfs4_stateid *stateid)
{
struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
struct nfs_inode *nfsi = NFS_I(inode);
struct nfs_open_context *ctx;
struct nfs4_state *state;
bool found = false;
spin_lock(&inode->i_lock);
list_for_each_entry(ctx, &nfsi->open_files, list) {
state = ctx->state;
if (state == NULL)
continue;
if (nfs4_stateid_match_other(&state->stateid, stateid) &&
nfs4_state_mark_reclaim_nograce(clp, state)) {
found = true;
continue;
}
if (nfs_state_lock_state_matches_stateid(state, stateid) &&
nfs4_state_mark_reclaim_nograce(clp, state))
found = true;
}
spin_unlock(&inode->i_lock);
nfs_inode_find_delegation_state_and_recover(inode, stateid);
if (found)
nfs4_schedule_state_manager(clp);
}
static void nfs4_state_mark_open_context_bad(struct nfs4_state *state)
{
struct inode *inode = state->inode;
struct nfs_inode *nfsi = NFS_I(inode);
struct nfs_open_context *ctx;
spin_lock(&inode->i_lock);
list_for_each_entry(ctx, &nfsi->open_files, list) {
if (ctx->state != state)
continue;
set_bit(NFS_CONTEXT_BAD, &ctx->flags);
}
spin_unlock(&inode->i_lock);
}
static void nfs4_state_mark_recovery_failed(struct nfs4_state *state, int error)
{
set_bit(NFS_STATE_RECOVERY_FAILED, &state->flags);
nfs4_state_mark_open_context_bad(state);
}
static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
{
struct inode *inode = state->inode;
struct nfs_inode *nfsi = NFS_I(inode);
struct file_lock *fl;
int status = 0;
struct file_lock_context *flctx = inode->i_flctx;
struct list_head *list;
if (flctx == NULL)
return 0;
list = &flctx->flc_posix;
/* Guard against delegation returns and new lock/unlock calls */
down_write(&nfsi->rwsem);
spin_lock(&flctx->flc_lock);
restart:
list_for_each_entry(fl, list, fl_list) {
if (nfs_file_open_context(fl->fl_file)->state != state)
continue;
spin_unlock(&flctx->flc_lock);
status = ops->recover_lock(state, fl);
switch (status) {
case 0:
break;
case -ESTALE:
case -NFS4ERR_ADMIN_REVOKED:
case -NFS4ERR_STALE_STATEID:
case -NFS4ERR_BAD_STATEID:
case -NFS4ERR_EXPIRED:
case -NFS4ERR_NO_GRACE:
case -NFS4ERR_STALE_CLIENTID:
case -NFS4ERR_BADSESSION:
case -NFS4ERR_BADSLOT:
case -NFS4ERR_BAD_HIGH_SLOT:
case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
goto out;
default:
pr_err("NFS: %s: unhandled error %d\n",
__func__, status);
case -ENOMEM:
case -NFS4ERR_DENIED:
case -NFS4ERR_RECLAIM_BAD:
case -NFS4ERR_RECLAIM_CONFLICT:
/* kill_proc(fl->fl_pid, SIGLOST, 1); */
status = 0;
}
spin_lock(&flctx->flc_lock);
}
if (list == &flctx->flc_posix) {
list = &flctx->flc_flock;
goto restart;
}
spin_unlock(&flctx->flc_lock);
out:
up_write(&nfsi->rwsem);
return status;
}
static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
{
struct nfs4_state *state;
struct nfs4_lock_state *lock;
int status = 0;
/* Note: we rely on the sp->so_states list being ordered
* so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
* states first.
* This is needed to ensure that the server won't give us any
* read delegations that we have to return if, say, we are
* recovering after a network partition or a reboot from a
* server that doesn't support a grace period.
*/
spin_lock(&sp->so_lock);
raw_write_seqcount_begin(&sp->so_reclaim_seqcount);
restart:
list_for_each_entry(state, &sp->so_states, open_states) {
if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
continue;
if (!nfs4_valid_open_stateid(state))
continue;
if (state->state == 0)
continue;
atomic_inc(&state->count);
spin_unlock(&sp->so_lock);
status = ops->recover_open(sp, state);
if (status >= 0) {
status = nfs4_reclaim_locks(state, ops);
if (status >= 0) {
if (!test_bit(NFS_DELEGATED_STATE, &state->flags)) {
spin_lock(&state->state_lock);
list_for_each_entry(lock, &state->lock_states, ls_locks) {
if (!test_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags))
pr_warn_ratelimited("NFS: "
"%s: Lock reclaim "
"failed!\n", __func__);
}
spin_unlock(&state->state_lock);
}
clear_bit(NFS_STATE_RECLAIM_NOGRACE,
&state->flags);
nfs4: Fix potential use after free of state in nfs4_do_reclaim. Commit e8d975e73e5f ("fixing infinite OPEN loop in 4.0 stateid recovery") introduced access to state after it was just potentially freed by nfs4_put_open_state leading to a random data corruption somewhere. BUG: unable to handle kernel paging request at ffff88004941ee40 IP: [<ffffffff813baf01>] nfs4_do_reclaim+0x461/0x740 PGD 3501067 PUD 3504067 PMD 6ff37067 PTE 800000004941e060 Oops: 0002 [#1] SMP DEBUG_PAGEALLOC Modules linked in: loop rpcsec_gss_krb5 acpi_cpufreq tpm_tis joydev i2c_piix4 pcspkr tpm virtio_console nfsd ttm drm_kms_helper syscopyarea sysfillrect sysimgblt fb_sys_fops floppy serio_raw virtio_blk drm CPU: 6 PID: 2161 Comm: 192.168.10.253- Not tainted 4.7.0-rc1-vm-nfs+ #112 Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011 task: ffff8800463dcd00 ti: ffff88003ff48000 task.ti: ffff88003ff48000 RIP: 0010:[<ffffffff813baf01>] [<ffffffff813baf01>] nfs4_do_reclaim+0x461/0x740 RSP: 0018:ffff88003ff4bd68 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffffffff81a49900 RCX: 00000000000000e8 RDX: 00000000000000e8 RSI: ffff8800418b9930 RDI: ffff880040c96c88 RBP: ffff88003ff4bdf8 R08: 0000000000000001 R09: 0000000000000000 R10: 0000000000000000 R11: 0000000000000000 R12: ffff880040c96c98 R13: ffff88004941ee20 R14: ffff88004941ee40 R15: ffff88004941ee00 FS: 0000000000000000(0000) GS:ffff88006d000000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: ffff88004941ee40 CR3: 0000000060b0b000 CR4: 00000000000006e0 Stack: ffffffff813baad5 ffff8800463dcd00 ffff880000000001 ffffffff810e6b68 ffff880043ddbc88 ffff8800418b9800 ffff8800418b98c8 ffff88004941ee48 ffff880040c96c90 ffff880040c96c00 ffff880040c96c20 ffff880040c96c40 Call Trace: [<ffffffff813baad5>] ? nfs4_do_reclaim+0x35/0x740 [<ffffffff810e6b68>] ? trace_hardirqs_on_caller+0x128/0x1b0 [<ffffffff813bb7cd>] nfs4_run_state_manager+0x5ed/0xa40 [<ffffffff813bb1e0>] ? nfs4_do_reclaim+0x740/0x740 [<ffffffff813bb1e0>] ? nfs4_do_reclaim+0x740/0x740 [<ffffffff810af0d1>] kthread+0x101/0x120 [<ffffffff810e6b68>] ? trace_hardirqs_on_caller+0x128/0x1b0 [<ffffffff818843af>] ret_from_fork+0x1f/0x40 [<ffffffff810aefd0>] ? kthread_create_on_node+0x250/0x250 Code: 65 80 4c 8b b5 78 ff ff ff e8 fc 88 4c 00 48 8b 7d 88 e8 13 67 d2 ff 49 8b 47 40 a8 02 0f 84 d3 01 00 00 4c 89 ff e8 7f f9 ff ff <f0> 41 80 26 7f 48 8b 7d c8 e8 b1 84 4c 00 e9 39 fd ff ff 3d e6 RIP [<ffffffff813baf01>] nfs4_do_reclaim+0x461/0x740 RSP <ffff88003ff4bd68> CR2: ffff88004941ee40 Signed-off-by: Oleg Drokin <green@linuxhacker.ru> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com> Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
2016-06-18 04:48:23 +08:00
nfs4_put_open_state(state);
spin_lock(&sp->so_lock);
goto restart;
}
}
switch (status) {
default:
printk(KERN_ERR "NFS: %s: unhandled error %d\n",
__func__, status);
case -ENOENT:
case -ENOMEM:
case -EACCES:
case -EROFS:
case -EIO:
case -ESTALE:
/* Open state on this file cannot be recovered */
nfs4_state_mark_recovery_failed(state, status);
break;
case -EAGAIN:
ssleep(1);
case -NFS4ERR_ADMIN_REVOKED:
case -NFS4ERR_STALE_STATEID:
case -NFS4ERR_OLD_STATEID:
case -NFS4ERR_BAD_STATEID:
case -NFS4ERR_RECLAIM_BAD:
case -NFS4ERR_RECLAIM_CONFLICT:
nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
break;
case -NFS4ERR_EXPIRED:
case -NFS4ERR_NO_GRACE:
nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
case -NFS4ERR_STALE_CLIENTID:
case -NFS4ERR_BADSESSION:
case -NFS4ERR_BADSLOT:
case -NFS4ERR_BAD_HIGH_SLOT:
case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
goto out_err;
}
nfs4_put_open_state(state);
spin_lock(&sp->so_lock);
goto restart;
}
raw_write_seqcount_end(&sp->so_reclaim_seqcount);
spin_unlock(&sp->so_lock);
return 0;
out_err:
nfs4_put_open_state(state);
spin_lock(&sp->so_lock);
raw_write_seqcount_end(&sp->so_reclaim_seqcount);
spin_unlock(&sp->so_lock);
return status;
}
static void nfs4_clear_open_state(struct nfs4_state *state)
{
struct nfs4_lock_state *lock;
clear_bit(NFS_DELEGATED_STATE, &state->flags);
clear_bit(NFS_O_RDONLY_STATE, &state->flags);
clear_bit(NFS_O_WRONLY_STATE, &state->flags);
clear_bit(NFS_O_RDWR_STATE, &state->flags);
spin_lock(&state->state_lock);
list_for_each_entry(lock, &state->lock_states, ls_locks) {
lock->ls_seqid.flags = 0;
clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
}
spin_unlock(&state->state_lock);
}
static void nfs4_reset_seqids(struct nfs_server *server,
int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
{
struct nfs_client *clp = server->nfs_client;
struct nfs4_state_owner *sp;
struct rb_node *pos;
struct nfs4_state *state;
spin_lock(&clp->cl_lock);
for (pos = rb_first(&server->state_owners);
pos != NULL;
pos = rb_next(pos)) {
sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
sp->so_seqid.flags = 0;
spin_lock(&sp->so_lock);
list_for_each_entry(state, &sp->so_states, open_states) {
if (mark_reclaim(clp, state))
nfs4_clear_open_state(state);
}
spin_unlock(&sp->so_lock);
}
spin_unlock(&clp->cl_lock);
}
static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp,
int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
{
struct nfs_server *server;
rcu_read_lock();
list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
nfs4_reset_seqids(server, mark_reclaim);
rcu_read_unlock();
}
static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
{
/* Mark all delegations for reclaim */
nfs_delegation_mark_reclaim(clp);
nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
}
static int nfs4_reclaim_complete(struct nfs_client *clp,
const struct nfs4_state_recovery_ops *ops,
struct rpc_cred *cred)
{
/* Notify the server we're done reclaiming our state */
if (ops->reclaim_complete)
return ops->reclaim_complete(clp, cred);
return 0;
}
static void nfs4_clear_reclaim_server(struct nfs_server *server)
{
struct nfs_client *clp = server->nfs_client;
struct nfs4_state_owner *sp;
struct rb_node *pos;
struct nfs4_state *state;
spin_lock(&clp->cl_lock);
for (pos = rb_first(&server->state_owners);
pos != NULL;
pos = rb_next(pos)) {
sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
spin_lock(&sp->so_lock);
list_for_each_entry(state, &sp->so_states, open_states) {
if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
&state->flags))
continue;
nfs4_state_mark_reclaim_nograce(clp, state);
}
spin_unlock(&sp->so_lock);
}
spin_unlock(&clp->cl_lock);
}
static int nfs4_state_clear_reclaim_reboot(struct nfs_client *clp)
{
struct nfs_server *server;
if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
return 0;
rcu_read_lock();
list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
nfs4_clear_reclaim_server(server);
rcu_read_unlock();
nfs_delegation_reap_unclaimed(clp);
return 1;
}
static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
{
const struct nfs4_state_recovery_ops *ops;
struct rpc_cred *cred;
int err;
if (!nfs4_state_clear_reclaim_reboot(clp))
return;
ops = clp->cl_mvops->reboot_recovery_ops;
cred = nfs4_get_clid_cred(clp);
err = nfs4_reclaim_complete(clp, ops, cred);
put_rpccred(cred);
if (err == -NFS4ERR_CONN_NOT_BOUND_TO_SESSION)
set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
}
static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
{
nfs_mark_test_expired_all_delegations(clp);
nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
}
static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
{
switch (error) {
case 0:
break;
case -NFS4ERR_CB_PATH_DOWN:
nfs40_handle_cb_pathdown(clp);
break;
case -NFS4ERR_NO_GRACE:
nfs4_state_end_reclaim_reboot(clp);
break;
case -NFS4ERR_STALE_CLIENTID:
set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
nfs4_state_start_reclaim_reboot(clp);
break;
case -NFS4ERR_EXPIRED:
set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
nfs4_state_start_reclaim_nograce(clp);
break;
case -NFS4ERR_BADSESSION:
case -NFS4ERR_BADSLOT:
case -NFS4ERR_BAD_HIGH_SLOT:
case -NFS4ERR_DEADSESSION:
case -NFS4ERR_SEQ_FALSE_RETRY:
case -NFS4ERR_SEQ_MISORDERED:
set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
/* Zero session reset errors */
break;
case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
break;
default:
dprintk("%s: failed to handle error %d for server %s\n",
__func__, error, clp->cl_hostname);
return error;
}
dprintk("%s: handled error %d for server %s\n", __func__, error,
clp->cl_hostname);
return 0;
}
static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
{
struct nfs4_state_owner *sp;
struct nfs_server *server;
struct rb_node *pos;
int status = 0;
restart:
rcu_read_lock();
list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
NFS: Cache state owners after files are closed Servers have a finite amount of memory to store NFSv4 open and lock owners. Moreover, servers may have a difficult time determining when they can reap their state owner table, thanks to gray areas in the NFSv4 protocol specification. Thus clients should be careful to reuse state owners when possible. Currently Linux is not too careful. When a user has closed all her files on one mount point, the state owner's reference count goes to zero, and it is released. The next OPEN allocates a new one. A workload that serially opens and closes files can run through a large number of open owners this way. When a state owner's reference count goes to zero, slap it onto a free list for that nfs_server, with an expiry time. Garbage collect before looking for a state owner. This makes state owners for active users available for re-use. Now that there can be unused state owners remaining at umount time, purge the state owner free list when a server is destroyed. Also be sure not to reclaim unused state owners during state recovery. This change has benefits for the client as well. For some workloads, this approach drops the number of OPEN_CONFIRM calls from the same as the number of OPEN calls, down to just one. This reduces wire traffic and thus open(2) latency. Before this patch, untarring a kernel source tarball shows the OPEN_CONFIRM call counter steadily increasing through the test. With the patch, the OPEN_CONFIRM count remains at 1 throughout the entire untar. As long as the expiry time is kept short, I don't think garbage collection should be terribly expensive, although it does bounce the clp->cl_lock around a bit. [ At some point we should rationalize the use of the nfs_server ->destroy method. ] Signed-off-by: Chuck Lever <chuck.lever@oracle.com> [Trond: Fixed a garbage collection race and a few efficiency issues] Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-12-07 05:13:48 +08:00
nfs4_purge_state_owners(server);
spin_lock(&clp->cl_lock);
for (pos = rb_first(&server->state_owners);
pos != NULL;
pos = rb_next(pos)) {
sp = rb_entry(pos,
struct nfs4_state_owner, so_server_node);
if (!test_and_clear_bit(ops->owner_flag_bit,
&sp->so_flags))
continue;
if (!atomic_inc_not_zero(&sp->so_count))
continue;
spin_unlock(&clp->cl_lock);
rcu_read_unlock();
status = nfs4_reclaim_open_state(sp, ops);
if (status < 0) {
set_bit(ops->owner_flag_bit, &sp->so_flags);
nfs4_put_state_owner(sp);
status = nfs4_recovery_handle_error(clp, status);
return (status != 0) ? status : -EAGAIN;
}
nfs4_put_state_owner(sp);
goto restart;
}
spin_unlock(&clp->cl_lock);
}
rcu_read_unlock();
return 0;
}
static int nfs4_check_lease(struct nfs_client *clp)
{
struct rpc_cred *cred;
const struct nfs4_state_maintenance_ops *ops =
clp->cl_mvops->state_renewal_ops;
int status;
/* Is the client already known to have an expired lease? */
if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
return 0;
spin_lock(&clp->cl_lock);
cred = ops->get_state_renewal_cred_locked(clp);
spin_unlock(&clp->cl_lock);
if (cred == NULL) {
cred = nfs4_get_clid_cred(clp);
status = -ENOKEY;
if (cred == NULL)
goto out;
}
status = ops->renew_lease(clp, cred);
put_rpccred(cred);
if (status == -ETIMEDOUT) {
set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
return 0;
}
out:
return nfs4_recovery_handle_error(clp, status);
}
/* Set NFS4CLNT_LEASE_EXPIRED and reclaim reboot state for all v4.0 errors
* and for recoverable errors on EXCHANGE_ID for v4.1
*/
static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
{
switch (status) {
case -NFS4ERR_SEQ_MISORDERED:
if (test_and_set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state))
return -ESERVERFAULT;
/* Lease confirmation error: retry after purging the lease */
ssleep(1);
clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
break;
case -NFS4ERR_STALE_CLIENTID:
clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
nfs4_state_start_reclaim_reboot(clp);
break;
case -NFS4ERR_CLID_INUSE:
pr_err("NFS: Server %s reports our clientid is in use\n",
clp->cl_hostname);
nfs_mark_client_ready(clp, -EPERM);
clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
return -EPERM;
case -EACCES:
case -NFS4ERR_DELAY:
case -ETIMEDOUT:
case -EAGAIN:
ssleep(1);
break;
case -NFS4ERR_MINOR_VERS_MISMATCH:
if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
nfs_mark_client_ready(clp, -EPROTONOSUPPORT);
dprintk("%s: exit with error %d for server %s\n",
__func__, -EPROTONOSUPPORT, clp->cl_hostname);
return -EPROTONOSUPPORT;
case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
* in nfs4_exchange_id */
default:
dprintk("%s: exit with error %d for server %s\n", __func__,
status, clp->cl_hostname);
return status;
}
set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
dprintk("%s: handled error %d for server %s\n", __func__, status,
clp->cl_hostname);
return 0;
}
static int nfs4_establish_lease(struct nfs_client *clp)
{
struct rpc_cred *cred;
const struct nfs4_state_recovery_ops *ops =
clp->cl_mvops->reboot_recovery_ops;
int status;
nfs4_begin_drain_session(clp);
cred = nfs4_get_clid_cred(clp);
if (cred == NULL)
return -ENOENT;
status = ops->establish_clid(clp, cred);
put_rpccred(cred);
if (status != 0)
return status;
pnfs_destroy_all_layouts(clp);
return 0;
}
/*
* Returns zero or a negative errno. NFS4ERR values are converted
* to local errno values.
*/
static int nfs4_reclaim_lease(struct nfs_client *clp)
{
int status;
status = nfs4_establish_lease(clp);
if (status < 0)
return nfs4_handle_reclaim_lease_error(clp, status);
if (test_and_clear_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state))
nfs4_state_start_reclaim_nograce(clp);
if (!test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
return 0;
}
static int nfs4_purge_lease(struct nfs_client *clp)
{
int status;
status = nfs4_establish_lease(clp);
if (status < 0)
return nfs4_handle_reclaim_lease_error(clp, status);
clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
nfs4_state_start_reclaim_nograce(clp);
return 0;
}
/*
* Try remote migration of one FSID from a source server to a
* destination server. The source server provides a list of
* potential destinations.
*
* Returns zero or a negative NFS4ERR status code.
*/
static int nfs4_try_migration(struct nfs_server *server, struct rpc_cred *cred)
{
struct nfs_client *clp = server->nfs_client;
struct nfs4_fs_locations *locations = NULL;
struct inode *inode;
struct page *page;
int status, result;
dprintk("--> %s: FSID %llx:%llx on \"%s\"\n", __func__,
(unsigned long long)server->fsid.major,
(unsigned long long)server->fsid.minor,
clp->cl_hostname);
result = 0;
page = alloc_page(GFP_KERNEL);
locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
if (page == NULL || locations == NULL) {
dprintk("<-- %s: no memory\n", __func__);
goto out;
}
inode = d_inode(server->super->s_root);
result = nfs4_proc_get_locations(inode, locations, page, cred);
if (result) {
dprintk("<-- %s: failed to retrieve fs_locations: %d\n",
__func__, result);
goto out;
}
result = -NFS4ERR_NXIO;
if (!(locations->fattr.valid & NFS_ATTR_FATTR_V4_LOCATIONS)) {
dprintk("<-- %s: No fs_locations data, migration skipped\n",
__func__);
goto out;
}
nfs4_begin_drain_session(clp);
status = nfs4_replace_transport(server, locations);
if (status != 0) {
dprintk("<-- %s: failed to replace transport: %d\n",
__func__, status);
goto out;
}
result = 0;
dprintk("<-- %s: migration succeeded\n", __func__);
out:
if (page != NULL)
__free_page(page);
kfree(locations);
if (result) {
pr_err("NFS: migration recovery failed (server %s)\n",
clp->cl_hostname);
set_bit(NFS_MIG_FAILED, &server->mig_status);
}
return result;
}
/*
* Returns zero or a negative NFS4ERR status code.
*/
static int nfs4_handle_migration(struct nfs_client *clp)
{
const struct nfs4_state_maintenance_ops *ops =
clp->cl_mvops->state_renewal_ops;
struct nfs_server *server;
struct rpc_cred *cred;
dprintk("%s: migration reported on \"%s\"\n", __func__,
clp->cl_hostname);
spin_lock(&clp->cl_lock);
cred = ops->get_state_renewal_cred_locked(clp);
spin_unlock(&clp->cl_lock);
if (cred == NULL)
return -NFS4ERR_NOENT;
clp->cl_mig_gen++;
restart:
rcu_read_lock();
list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
int status;
if (server->mig_gen == clp->cl_mig_gen)
continue;
server->mig_gen = clp->cl_mig_gen;
if (!test_and_clear_bit(NFS_MIG_IN_TRANSITION,
&server->mig_status))
continue;
rcu_read_unlock();
status = nfs4_try_migration(server, cred);
if (status < 0) {
put_rpccred(cred);
return status;
}
goto restart;
}
rcu_read_unlock();
put_rpccred(cred);
return 0;
}
/*
* Test each nfs_server on the clp's cl_superblocks list to see
* if it's moved to another server. Stop when the server no longer
* returns NFS4ERR_LEASE_MOVED.
*/
static int nfs4_handle_lease_moved(struct nfs_client *clp)
{
const struct nfs4_state_maintenance_ops *ops =
clp->cl_mvops->state_renewal_ops;
struct nfs_server *server;
struct rpc_cred *cred;
dprintk("%s: lease moved reported on \"%s\"\n", __func__,
clp->cl_hostname);
spin_lock(&clp->cl_lock);
cred = ops->get_state_renewal_cred_locked(clp);
spin_unlock(&clp->cl_lock);
if (cred == NULL)
return -NFS4ERR_NOENT;
clp->cl_mig_gen++;
restart:
rcu_read_lock();
list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
struct inode *inode;
int status;
if (server->mig_gen == clp->cl_mig_gen)
continue;
server->mig_gen = clp->cl_mig_gen;
rcu_read_unlock();
inode = d_inode(server->super->s_root);
status = nfs4_proc_fsid_present(inode, cred);
if (status != -NFS4ERR_MOVED)
goto restart; /* wasn't this one */
if (nfs4_try_migration(server, cred) == -NFS4ERR_LEASE_MOVED)
goto restart; /* there are more */
goto out;
}
rcu_read_unlock();
out:
put_rpccred(cred);
return 0;
}
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
/**
* nfs4_discover_server_trunking - Detect server IP address trunking
*
* @clp: nfs_client under test
* @result: OUT: found nfs_client, or clp
*
* Returns zero or a negative errno. If zero is returned,
* an nfs_client pointer is planted in "result".
*
* Note: since we are invoked in process context, and
* not from inside the state manager, we cannot use
* nfs4_handle_reclaim_lease_error().
*/
int nfs4_discover_server_trunking(struct nfs_client *clp,
struct nfs_client **result)
{
const struct nfs4_state_recovery_ops *ops =
clp->cl_mvops->reboot_recovery_ops;
struct rpc_clnt *clnt;
struct rpc_cred *cred;
int i, status;
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
dprintk("NFS: %s: testing '%s'\n", __func__, clp->cl_hostname);
clnt = clp->cl_rpcclient;
i = 0;
mutex_lock(&nfs_clid_init_mutex);
again:
status = -ENOENT;
cred = nfs4_get_clid_cred(clp);
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
if (cred == NULL)
goto out_unlock;
status = ops->detect_trunking(clp, result, cred);
put_rpccred(cred);
switch (status) {
case 0:
break;
case -ETIMEDOUT:
if (clnt->cl_softrtry)
break;
case -NFS4ERR_DELAY:
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
case -EAGAIN:
ssleep(1);
case -NFS4ERR_STALE_CLIENTID:
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
dprintk("NFS: %s after status %d, retrying\n",
__func__, status);
goto again;
case -EACCES:
NFS: Use root's credential for lease management when keytab is missing Commit 05f4c350 "NFS: Discover NFSv4 server trunking when mounting" Fri Sep 14 17:24:32 2012 introduced Uniform Client String support, which forces our NFS client to establish a client ID immediately during a mount operation rather than waiting until a user wants to open a file. Normally machine credentials (eg. from a keytab) are used to perform a mount operation that is protected by Kerberos. Before 05fc350, SETCLIENTID used a machine credential, or fell back to a regular user's credential if no keytab is available. On clients that don't have a keytab, performing SETCLIENTID early means there's no user credential to fall back on, since no regular user has kinit'd yet. 05f4c350 seems to have broken the ability to mount with sec=krb5 on clients that don't have a keytab in kernels 3.7 - 3.10. To address this regression, commit 4edaa308 (NFS: Use "krb5i" to establish NFSv4 state whenever possible), Sat Mar 16 15:56:20 2013, was merged in 3.10. This commit forces the NFS client to fall back to AUTH_SYS for lease management operations if no keytab is available. Neil Brown noticed that, since root is required to kinit to do a sec=krb5 mount when a client doesn't have a keytab, we can try to use root's Kerberos credential before AUTH_SYS. Now, when determining a principal and flavor to use for lease management, the NFS client tries in this order: 1. Flavor: AUTH_GSS, krb5i Principal: service principal (via keytab) 2. Flavor: AUTH_GSS, krb5i Principal: user principal established for UID 0 (via kinit) 3. Flavor: AUTH_SYS Principal: UID 0 / GID 0 Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2013-07-25 00:28:28 +08:00
if (i++ == 0) {
nfs4_root_machine_cred(clp);
goto again;
}
if (clnt->cl_auth->au_flavor == RPC_AUTH_UNIX)
break;
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
case -NFS4ERR_CLID_INUSE:
case -NFS4ERR_WRONGSEC:
/* No point in retrying if we already used RPC_AUTH_UNIX */
if (clnt->cl_auth->au_flavor == RPC_AUTH_UNIX) {
status = -EPERM;
break;
}
clnt = rpc_clone_client_set_auth(clnt, RPC_AUTH_UNIX);
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
if (IS_ERR(clnt)) {
status = PTR_ERR(clnt);
break;
}
/* Note: this is safe because we haven't yet marked the
* client as ready, so we are the only user of
* clp->cl_rpcclient
*/
clnt = xchg(&clp->cl_rpcclient, clnt);
rpc_shutdown_client(clnt);
clnt = clp->cl_rpcclient;
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
goto again;
case -NFS4ERR_MINOR_VERS_MISMATCH:
status = -EPROTONOSUPPORT;
break;
case -EKEYEXPIRED:
case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
* in nfs4_exchange_id */
status = -EKEYEXPIRED;
break;
default:
pr_warn("NFS: %s unhandled error %d. Exiting with error EIO\n",
__func__, status);
status = -EIO;
NFS: Discover NFSv4 server trunking when mounting "Server trunking" is a fancy named for a multi-homed NFS server. Trunking might occur if a client sends NFS requests for a single workload to multiple network interfaces on the same server. There are some implications for NFSv4 state management that make it useful for a client to know if a single NFSv4 server instance is multi-homed. (Note this is only a consideration for NFSv4, not for legacy versions of NFS, which are stateless). If a client cares about server trunking, no NFSv4 operations can proceed until that client determines who it is talking to. Thus server IP trunking discovery must be done when the client first encounters an unfamiliar server IP address. The nfs_get_client() function walks the nfs_client_list and matches on server IP address. The outcome of that walk tells us immediately if we have an unfamiliar server IP address. It invokes nfs_init_client() in this case. Thus, nfs4_init_client() is a good spot to perform trunking discovery. Discovery requires a client to establish a fresh client ID, so our client will now send SETCLIENTID or EXCHANGE_ID as the first NFS operation after a successful ping, rather than waiting for an application to perform an operation that requires NFSv4 state. The exact process for detecting trunking is different for NFSv4.0 and NFSv4.1, so a minorversion-specific init_client callout method is introduced. CLID_INUSE recovery is important for the trunking discovery process. CLID_INUSE is a sign the server recognizes the client's nfs_client_id4 id string, but the client is using the wrong principal this time for the SETCLIENTID operation. The SETCLIENTID must be retried with a series of different principals until one works, and then the rest of trunking discovery can proceed. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2012-09-15 05:24:32 +08:00
}
out_unlock:
mutex_unlock(&nfs_clid_init_mutex);
dprintk("NFS: %s: status = %d\n", __func__, status);
return status;
}
nfs41: add session setup to the state manager At mount, nfs_alloc_client sets the cl_state NFS4CLNT_LEASE_EXPIRED bit and nfs4_alloc_session sets the NFS4CLNT_SESSION_SETUP bit, so both bits are set when nfs4_lookup_root calls nfs4_recover_expired_lease which schedules the nfs4_state_manager and waits for it to complete. Place the session setup after the clientid establishment in nfs4_state_manager so that the session is setup right after the clientid has been established without rescheduling the state manager. Unlike nfsv4.0, the nfs_client struct is not ready to use until the session has been established. Postpone marking the nfs_client struct to NFS_CS_READY until after a successful CREATE_SESSION call so that other threads cannot use the client until the session is established. If the EXCHANGE_ID call fails and the session has not been setup (the NFS4CLNT_SESSION_SETUP bit is set), mark the client with the error and return. If the session setup CREATE_SESSION call fails with NFS4ERR_STALE_CLIENTID which could occur due to server reboot or network partition inbetween the EXCHANGE_ID and CREATE_SESSION call, reset the NFS4CLNT_LEASE_EXPIRED and NFS4CLNT_SESSION_SETUP bits and try again. If the CREATE_SESSION call fails with other errors, mark the client with the error and return. Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: NFS_CS_SESSION_SETUP cl_cons_state for back channel setup] On session setup, the CREATE_SESSION reply races with the server back channel probe which needs to succeed to setup the back channel. Set a new cl_cons_state NFS_CS_SESSION_SETUP just prior to the CREATE_SESSION call and add it as a valid state to nfs_find_client so that the client back channel can find the nfs_client struct and won't drop the server backchannel probe. Use a new cl_cons_state so that NFSv4.0 back channel behaviour which only sets NFS_CS_READY is unchanged. Adjust waiting on the nfs_client_active_wq accordingly. Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: rename NFS_CS_SESSION_SETUP to NFS_CS_SESSION_INITING] Signed-off-by: Andy Adamson <andros@netapp.com> [nfs41: set NFS_CL_SESSION_INITING in alloc_session] Signed-off-by: Andy Adamson <andros@netapp.com> [nfs41: move session setup into a function] Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [moved nfs4_proc_create_session declaration here] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-04-01 21:22:38 +08:00
#ifdef CONFIG_NFS_V4_1
void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
{
struct nfs_client *clp = session->clp;
switch (err) {
default:
set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
break;
case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
}
nfs4_schedule_state_manager(clp);
}
EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
void nfs41_notify_server(struct nfs_client *clp)
{
/* Use CHECK_LEASE to ping the server with a SEQUENCE */
set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
nfs4_schedule_state_manager(clp);
}
static void nfs4_reset_all_state(struct nfs_client *clp)
{
if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
nfs4_state_start_reclaim_nograce(clp);
dprintk("%s: scheduling reset of all state for server %s!\n",
__func__, clp->cl_hostname);
nfs4_schedule_state_manager(clp);
}
}
static void nfs41_handle_server_reboot(struct nfs_client *clp)
{
if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
nfs4_state_start_reclaim_reboot(clp);
dprintk("%s: server %s rebooted!\n", __func__,
clp->cl_hostname);
nfs4_schedule_state_manager(clp);
}
}
static void nfs41_handle_all_state_revoked(struct nfs_client *clp)
{
nfs4_reset_all_state(clp);
dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
}
static void nfs41_handle_some_state_revoked(struct nfs_client *clp)
{
nfs4_state_start_reclaim_nograce(clp);
nfs4_schedule_state_manager(clp);
dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
}
static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
{
/* FIXME: For now, we destroy all layouts. */
pnfs_destroy_all_layouts(clp);
/* FIXME: For now, we test all delegations+open state+locks. */
nfs41_handle_some_state_revoked(clp);
dprintk("%s: Recallable state revoked on server %s!\n", __func__,
clp->cl_hostname);
}
static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
{
set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
nfs4_schedule_state_manager(clp);
dprintk("%s: server %s declared a backchannel fault\n", __func__,
clp->cl_hostname);
}
static void nfs41_handle_cb_path_down(struct nfs_client *clp)
{
if (test_and_set_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
&clp->cl_state) == 0)
nfs4_schedule_state_manager(clp);
}
void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags,
bool recovery)
{
if (!flags)
return;
dprintk("%s: \"%s\" (client ID %llx) flags=0x%08x\n",
__func__, clp->cl_hostname, clp->cl_clientid, flags);
/*
* If we're called from the state manager thread, then assume we're
* already handling the RECLAIM_NEEDED and/or STATE_REVOKED.
* Those flags are expected to remain set until we're done
* recovering (see RFC5661, section 18.46.3).
*/
if (recovery)
goto out_recovery;
if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
nfs41_handle_server_reboot(clp);
if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED))
nfs41_handle_all_state_revoked(clp);
if (flags & (SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
SEQ4_STATUS_ADMIN_STATE_REVOKED))
nfs41_handle_some_state_revoked(clp);
if (flags & SEQ4_STATUS_LEASE_MOVED)
nfs4_schedule_lease_moved_recovery(clp);
if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
nfs41_handle_recallable_state_revoked(clp);
out_recovery:
if (flags & SEQ4_STATUS_BACKCHANNEL_FAULT)
nfs41_handle_backchannel_fault(clp);
else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
SEQ4_STATUS_CB_PATH_DOWN_SESSION))
nfs41_handle_cb_path_down(clp);
}
static int nfs4_reset_session(struct nfs_client *clp)
{
struct rpc_cred *cred;
int status;
if (!nfs4_has_session(clp))
return 0;
nfs4_begin_drain_session(clp);
cred = nfs4_get_clid_cred(clp);
status = nfs4_proc_destroy_session(clp->cl_session, cred);
switch (status) {
case 0:
case -NFS4ERR_BADSESSION:
case -NFS4ERR_DEADSESSION:
break;
case -NFS4ERR_BACK_CHAN_BUSY:
case -NFS4ERR_DELAY:
set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
status = 0;
ssleep(1);
goto out;
default:
status = nfs4_recovery_handle_error(clp, status);
goto out;
}
memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
status = nfs4_proc_create_session(clp, cred);
if (status) {
dprintk("%s: session reset failed with status %d for server %s!\n",
__func__, status, clp->cl_hostname);
status = nfs4_handle_reclaim_lease_error(clp, status);
goto out;
}
nfs41_finish_session_reset(clp);
dprintk("%s: session reset was successful for server %s!\n",
__func__, clp->cl_hostname);
out:
if (cred)
put_rpccred(cred);
return status;
}
nfs41: add session setup to the state manager At mount, nfs_alloc_client sets the cl_state NFS4CLNT_LEASE_EXPIRED bit and nfs4_alloc_session sets the NFS4CLNT_SESSION_SETUP bit, so both bits are set when nfs4_lookup_root calls nfs4_recover_expired_lease which schedules the nfs4_state_manager and waits for it to complete. Place the session setup after the clientid establishment in nfs4_state_manager so that the session is setup right after the clientid has been established without rescheduling the state manager. Unlike nfsv4.0, the nfs_client struct is not ready to use until the session has been established. Postpone marking the nfs_client struct to NFS_CS_READY until after a successful CREATE_SESSION call so that other threads cannot use the client until the session is established. If the EXCHANGE_ID call fails and the session has not been setup (the NFS4CLNT_SESSION_SETUP bit is set), mark the client with the error and return. If the session setup CREATE_SESSION call fails with NFS4ERR_STALE_CLIENTID which could occur due to server reboot or network partition inbetween the EXCHANGE_ID and CREATE_SESSION call, reset the NFS4CLNT_LEASE_EXPIRED and NFS4CLNT_SESSION_SETUP bits and try again. If the CREATE_SESSION call fails with other errors, mark the client with the error and return. Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: NFS_CS_SESSION_SETUP cl_cons_state for back channel setup] On session setup, the CREATE_SESSION reply races with the server back channel probe which needs to succeed to setup the back channel. Set a new cl_cons_state NFS_CS_SESSION_SETUP just prior to the CREATE_SESSION call and add it as a valid state to nfs_find_client so that the client back channel can find the nfs_client struct and won't drop the server backchannel probe. Use a new cl_cons_state so that NFSv4.0 back channel behaviour which only sets NFS_CS_READY is unchanged. Adjust waiting on the nfs_client_active_wq accordingly. Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: rename NFS_CS_SESSION_SETUP to NFS_CS_SESSION_INITING] Signed-off-by: Andy Adamson <andros@netapp.com> [nfs41: set NFS_CL_SESSION_INITING in alloc_session] Signed-off-by: Andy Adamson <andros@netapp.com> [nfs41: move session setup into a function] Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [moved nfs4_proc_create_session declaration here] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-04-01 21:22:38 +08:00
static int nfs4_bind_conn_to_session(struct nfs_client *clp)
{
struct rpc_cred *cred;
int ret;
if (!nfs4_has_session(clp))
return 0;
nfs4_begin_drain_session(clp);
cred = nfs4_get_clid_cred(clp);
ret = nfs4_proc_bind_conn_to_session(clp, cred);
if (cred)
put_rpccred(cred);
clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
switch (ret) {
case 0:
dprintk("%s: bind_conn_to_session was successful for server %s!\n",
__func__, clp->cl_hostname);
break;
case -NFS4ERR_DELAY:
ssleep(1);
set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
break;
default:
return nfs4_recovery_handle_error(clp, ret);
}
return 0;
}
nfs41: add session setup to the state manager At mount, nfs_alloc_client sets the cl_state NFS4CLNT_LEASE_EXPIRED bit and nfs4_alloc_session sets the NFS4CLNT_SESSION_SETUP bit, so both bits are set when nfs4_lookup_root calls nfs4_recover_expired_lease which schedules the nfs4_state_manager and waits for it to complete. Place the session setup after the clientid establishment in nfs4_state_manager so that the session is setup right after the clientid has been established without rescheduling the state manager. Unlike nfsv4.0, the nfs_client struct is not ready to use until the session has been established. Postpone marking the nfs_client struct to NFS_CS_READY until after a successful CREATE_SESSION call so that other threads cannot use the client until the session is established. If the EXCHANGE_ID call fails and the session has not been setup (the NFS4CLNT_SESSION_SETUP bit is set), mark the client with the error and return. If the session setup CREATE_SESSION call fails with NFS4ERR_STALE_CLIENTID which could occur due to server reboot or network partition inbetween the EXCHANGE_ID and CREATE_SESSION call, reset the NFS4CLNT_LEASE_EXPIRED and NFS4CLNT_SESSION_SETUP bits and try again. If the CREATE_SESSION call fails with other errors, mark the client with the error and return. Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: NFS_CS_SESSION_SETUP cl_cons_state for back channel setup] On session setup, the CREATE_SESSION reply races with the server back channel probe which needs to succeed to setup the back channel. Set a new cl_cons_state NFS_CS_SESSION_SETUP just prior to the CREATE_SESSION call and add it as a valid state to nfs_find_client so that the client back channel can find the nfs_client struct and won't drop the server backchannel probe. Use a new cl_cons_state so that NFSv4.0 back channel behaviour which only sets NFS_CS_READY is unchanged. Adjust waiting on the nfs_client_active_wq accordingly. Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: rename NFS_CS_SESSION_SETUP to NFS_CS_SESSION_INITING] Signed-off-by: Andy Adamson <andros@netapp.com> [nfs41: set NFS_CL_SESSION_INITING in alloc_session] Signed-off-by: Andy Adamson <andros@netapp.com> [nfs41: move session setup into a function] Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [moved nfs4_proc_create_session declaration here] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-04-01 21:22:38 +08:00
#else /* CONFIG_NFS_V4_1 */
static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
static int nfs4_bind_conn_to_session(struct nfs_client *clp)
{
return 0;
}
nfs41: add session setup to the state manager At mount, nfs_alloc_client sets the cl_state NFS4CLNT_LEASE_EXPIRED bit and nfs4_alloc_session sets the NFS4CLNT_SESSION_SETUP bit, so both bits are set when nfs4_lookup_root calls nfs4_recover_expired_lease which schedules the nfs4_state_manager and waits for it to complete. Place the session setup after the clientid establishment in nfs4_state_manager so that the session is setup right after the clientid has been established without rescheduling the state manager. Unlike nfsv4.0, the nfs_client struct is not ready to use until the session has been established. Postpone marking the nfs_client struct to NFS_CS_READY until after a successful CREATE_SESSION call so that other threads cannot use the client until the session is established. If the EXCHANGE_ID call fails and the session has not been setup (the NFS4CLNT_SESSION_SETUP bit is set), mark the client with the error and return. If the session setup CREATE_SESSION call fails with NFS4ERR_STALE_CLIENTID which could occur due to server reboot or network partition inbetween the EXCHANGE_ID and CREATE_SESSION call, reset the NFS4CLNT_LEASE_EXPIRED and NFS4CLNT_SESSION_SETUP bits and try again. If the CREATE_SESSION call fails with other errors, mark the client with the error and return. Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: NFS_CS_SESSION_SETUP cl_cons_state for back channel setup] On session setup, the CREATE_SESSION reply races with the server back channel probe which needs to succeed to setup the back channel. Set a new cl_cons_state NFS_CS_SESSION_SETUP just prior to the CREATE_SESSION call and add it as a valid state to nfs_find_client so that the client back channel can find the nfs_client struct and won't drop the server backchannel probe. Use a new cl_cons_state so that NFSv4.0 back channel behaviour which only sets NFS_CS_READY is unchanged. Adjust waiting on the nfs_client_active_wq accordingly. Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: rename NFS_CS_SESSION_SETUP to NFS_CS_SESSION_INITING] Signed-off-by: Andy Adamson <andros@netapp.com> [nfs41: set NFS_CL_SESSION_INITING in alloc_session] Signed-off-by: Andy Adamson <andros@netapp.com> [nfs41: move session setup into a function] Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [moved nfs4_proc_create_session declaration here] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-04-01 21:22:38 +08:00
#endif /* CONFIG_NFS_V4_1 */
static void nfs4_state_manager(struct nfs_client *clp)
{
int status = 0;
const char *section = "", *section_sep = "";
/* Ensure exclusive access to NFSv4 state */
do {
if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
section = "purge state";
status = nfs4_purge_lease(clp);
if (status < 0)
goto out_error;
continue;
}
if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
section = "lease expired";
/* We're going to have to re-establish a clientid */
status = nfs4_reclaim_lease(clp);
if (status < 0)
goto out_error;
continue;
}
/* Initialize or reset the session */
if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
section = "reset session";
status = nfs4_reset_session(clp);
if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
continue;
if (status < 0)
nfs41: add session setup to the state manager At mount, nfs_alloc_client sets the cl_state NFS4CLNT_LEASE_EXPIRED bit and nfs4_alloc_session sets the NFS4CLNT_SESSION_SETUP bit, so both bits are set when nfs4_lookup_root calls nfs4_recover_expired_lease which schedules the nfs4_state_manager and waits for it to complete. Place the session setup after the clientid establishment in nfs4_state_manager so that the session is setup right after the clientid has been established without rescheduling the state manager. Unlike nfsv4.0, the nfs_client struct is not ready to use until the session has been established. Postpone marking the nfs_client struct to NFS_CS_READY until after a successful CREATE_SESSION call so that other threads cannot use the client until the session is established. If the EXCHANGE_ID call fails and the session has not been setup (the NFS4CLNT_SESSION_SETUP bit is set), mark the client with the error and return. If the session setup CREATE_SESSION call fails with NFS4ERR_STALE_CLIENTID which could occur due to server reboot or network partition inbetween the EXCHANGE_ID and CREATE_SESSION call, reset the NFS4CLNT_LEASE_EXPIRED and NFS4CLNT_SESSION_SETUP bits and try again. If the CREATE_SESSION call fails with other errors, mark the client with the error and return. Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: NFS_CS_SESSION_SETUP cl_cons_state for back channel setup] On session setup, the CREATE_SESSION reply races with the server back channel probe which needs to succeed to setup the back channel. Set a new cl_cons_state NFS_CS_SESSION_SETUP just prior to the CREATE_SESSION call and add it as a valid state to nfs_find_client so that the client back channel can find the nfs_client struct and won't drop the server backchannel probe. Use a new cl_cons_state so that NFSv4.0 back channel behaviour which only sets NFS_CS_READY is unchanged. Adjust waiting on the nfs_client_active_wq accordingly. Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: rename NFS_CS_SESSION_SETUP to NFS_CS_SESSION_INITING] Signed-off-by: Andy Adamson <andros@netapp.com> [nfs41: set NFS_CL_SESSION_INITING in alloc_session] Signed-off-by: Andy Adamson <andros@netapp.com> [nfs41: move session setup into a function] Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [moved nfs4_proc_create_session declaration here] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-04-01 21:22:38 +08:00
goto out_error;
}
/* Send BIND_CONN_TO_SESSION */
if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
&clp->cl_state)) {
section = "bind conn to session";
status = nfs4_bind_conn_to_session(clp);
if (status < 0)
goto out_error;
continue;
}
if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
section = "check lease";
status = nfs4_check_lease(clp);
if (status < 0)
goto out_error;
continue;
}
if (test_and_clear_bit(NFS4CLNT_MOVED, &clp->cl_state)) {
section = "migration";
status = nfs4_handle_migration(clp);
if (status < 0)
goto out_error;
}
if (test_and_clear_bit(NFS4CLNT_LEASE_MOVED, &clp->cl_state)) {
section = "lease moved";
status = nfs4_handle_lease_moved(clp);
if (status < 0)
goto out_error;
}
/* First recover reboot state... */
if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
section = "reclaim reboot";
status = nfs4_do_reclaim(clp,
clp->cl_mvops->reboot_recovery_ops);
if (status == -EAGAIN)
continue;
if (status < 0)
goto out_error;
nfs4_state_end_reclaim_reboot(clp);
}
/* Detect expired delegations... */
if (test_and_clear_bit(NFS4CLNT_DELEGATION_EXPIRED, &clp->cl_state)) {
section = "detect expired delegations";
nfs_reap_expired_delegations(clp);
continue;
}
/* Now recover expired state... */
if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
section = "reclaim nograce";
status = nfs4_do_reclaim(clp,
clp->cl_mvops->nograce_recovery_ops);
if (status == -EAGAIN)
continue;
if (status < 0)
goto out_error;
}
nfs4_end_drain_session(clp);
if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
nfs_client_return_marked_delegations(clp);
continue;
}
nfs4_clear_state_manager_bit(clp);
/* Did we race with an attempt to give us more work? */
if (clp->cl_state == 0)
break;
if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
break;
} while (atomic_read(&clp->cl_count) > 1);
return;
out_error:
if (strlen(section))
section_sep = ": ";
pr_warn_ratelimited("NFS: state manager%s%s failed on NFSv4 server %s"
" with error %d\n", section_sep, section,
clp->cl_hostname, -status);
ssleep(1);
nfs4_end_drain_session(clp);
nfs4_clear_state_manager_bit(clp);
}
static int nfs4_run_state_manager(void *ptr)
{
struct nfs_client *clp = ptr;
allow_signal(SIGKILL);
nfs4_state_manager(clp);
nfs_put_client(clp);
module_put_and_exit(0);
return 0;
}
/*
* Local variables:
* c-basic-offset: 8
* End:
*/