4 * Client-side XDR for NFSv4.
6 * Copyright (c) 2002 The Regents of the University of Michigan.
9 * Kendrick Smith <kmsmith@umich.edu>
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its
21 * contributors may be used to endorse or promote products derived
22 * from this software without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
25 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
26 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
27 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
31 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
32 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
33 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
34 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36 * Implementation of the NFSv4 state model. For the time being,
37 * this is minimal, but will be made much more complex in a
41 #include <linux/config.h>
42 #include <linux/slab.h>
43 #include <linux/smp_lock.h>
44 #include <linux/nfs_fs.h>
45 #include <linux/nfs_idmap.h>
46 #include <linux/workqueue.h>
47 #include <linux/bitops.h>
51 #include "delegation.h"
53 #define OPENOWNER_POOL_SIZE 8
55 const nfs4_stateid zero_stateid
;
57 static DEFINE_SPINLOCK(state_spinlock
);
58 static LIST_HEAD(nfs4_clientid_list
);
60 static void nfs4_recover_state(void *);
63 init_nfsv4_state(struct nfs_server
*server
)
65 server
->nfs4_state
= NULL
;
66 INIT_LIST_HEAD(&server
->nfs4_siblings
);
70 destroy_nfsv4_state(struct nfs_server
*server
)
72 if (server
->mnt_path
) {
73 kfree(server
->mnt_path
);
74 server
->mnt_path
= NULL
;
76 if (server
->nfs4_state
) {
77 nfs4_put_client(server
->nfs4_state
);
78 server
->nfs4_state
= NULL
;
83 * nfs4_get_client(): returns an empty client structure
84 * nfs4_put_client(): drops reference to client structure
86 * Since these are allocated/deallocated very rarely, we don't
87 * bother putting them in a slab cache...
89 static struct nfs4_client
*
90 nfs4_alloc_client(struct in_addr
*addr
)
92 struct nfs4_client
*clp
;
94 if (nfs_callback_up() < 0)
96 if ((clp
= kmalloc(sizeof(*clp
), GFP_KERNEL
)) == NULL
) {
100 memset(clp
, 0, sizeof(*clp
));
101 memcpy(&clp
->cl_addr
, addr
, sizeof(clp
->cl_addr
));
102 init_rwsem(&clp
->cl_sem
);
103 INIT_LIST_HEAD(&clp
->cl_delegations
);
104 INIT_LIST_HEAD(&clp
->cl_state_owners
);
105 INIT_LIST_HEAD(&clp
->cl_unused
);
106 spin_lock_init(&clp
->cl_lock
);
107 atomic_set(&clp
->cl_count
, 1);
108 INIT_WORK(&clp
->cl_recoverd
, nfs4_recover_state
, clp
);
109 INIT_WORK(&clp
->cl_renewd
, nfs4_renew_state
, clp
);
110 INIT_LIST_HEAD(&clp
->cl_superblocks
);
111 init_waitqueue_head(&clp
->cl_waitq
);
112 rpc_init_wait_queue(&clp
->cl_rpcwaitq
, "NFS4 client");
113 clp
->cl_rpcclient
= ERR_PTR(-EINVAL
);
114 clp
->cl_boot_time
= CURRENT_TIME
;
115 clp
->cl_state
= 1 << NFS4CLNT_OK
;
120 nfs4_free_client(struct nfs4_client
*clp
)
122 struct nfs4_state_owner
*sp
;
124 while (!list_empty(&clp
->cl_unused
)) {
125 sp
= list_entry(clp
->cl_unused
.next
,
126 struct nfs4_state_owner
,
128 list_del(&sp
->so_list
);
131 BUG_ON(!list_empty(&clp
->cl_state_owners
));
133 put_rpccred(clp
->cl_cred
);
134 nfs_idmap_delete(clp
);
135 if (!IS_ERR(clp
->cl_rpcclient
))
136 rpc_shutdown_client(clp
->cl_rpcclient
);
141 static struct nfs4_client
*__nfs4_find_client(struct in_addr
*addr
)
143 struct nfs4_client
*clp
;
144 list_for_each_entry(clp
, &nfs4_clientid_list
, cl_servers
) {
145 if (memcmp(&clp
->cl_addr
, addr
, sizeof(clp
->cl_addr
)) == 0) {
146 atomic_inc(&clp
->cl_count
);
153 struct nfs4_client
*nfs4_find_client(struct in_addr
*addr
)
155 struct nfs4_client
*clp
;
156 spin_lock(&state_spinlock
);
157 clp
= __nfs4_find_client(addr
);
158 spin_unlock(&state_spinlock
);
163 nfs4_get_client(struct in_addr
*addr
)
165 struct nfs4_client
*clp
, *new = NULL
;
167 spin_lock(&state_spinlock
);
169 clp
= __nfs4_find_client(addr
);
174 list_add(&clp
->cl_servers
, &nfs4_clientid_list
);
178 spin_unlock(&state_spinlock
);
179 new = nfs4_alloc_client(addr
);
180 spin_lock(&state_spinlock
);
184 spin_unlock(&state_spinlock
);
186 nfs4_free_client(new);
191 nfs4_put_client(struct nfs4_client
*clp
)
193 if (!atomic_dec_and_lock(&clp
->cl_count
, &state_spinlock
))
195 list_del(&clp
->cl_servers
);
196 spin_unlock(&state_spinlock
);
197 BUG_ON(!list_empty(&clp
->cl_superblocks
));
198 wake_up_all(&clp
->cl_waitq
);
199 rpc_wake_up(&clp
->cl_rpcwaitq
);
200 nfs4_kill_renewd(clp
);
201 nfs4_free_client(clp
);
204 static int __nfs4_init_client(struct nfs4_client
*clp
)
206 int status
= nfs4_proc_setclientid(clp
, NFS4_CALLBACK
, nfs_callback_tcpport
);
208 status
= nfs4_proc_setclientid_confirm(clp
);
210 nfs4_schedule_state_renewal(clp
);
214 int nfs4_init_client(struct nfs4_client
*clp
)
216 return nfs4_map_errors(__nfs4_init_client(clp
));
220 nfs4_alloc_lockowner_id(struct nfs4_client
*clp
)
222 return clp
->cl_lockowner_id
++;
225 static struct nfs4_state_owner
*
226 nfs4_client_grab_unused(struct nfs4_client
*clp
, struct rpc_cred
*cred
)
228 struct nfs4_state_owner
*sp
= NULL
;
230 if (!list_empty(&clp
->cl_unused
)) {
231 sp
= list_entry(clp
->cl_unused
.next
, struct nfs4_state_owner
, so_list
);
232 atomic_inc(&sp
->so_count
);
234 list_move(&sp
->so_list
, &clp
->cl_state_owners
);
240 static struct nfs4_state_owner
*
241 nfs4_find_state_owner(struct nfs4_client
*clp
, struct rpc_cred
*cred
)
243 struct nfs4_state_owner
*sp
, *res
= NULL
;
245 list_for_each_entry(sp
, &clp
->cl_state_owners
, so_list
) {
246 if (sp
->so_cred
!= cred
)
248 atomic_inc(&sp
->so_count
);
249 /* Move to the head of the list */
250 list_move(&sp
->so_list
, &clp
->cl_state_owners
);
258 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
259 * create a new state_owner.
262 static struct nfs4_state_owner
*
263 nfs4_alloc_state_owner(void)
265 struct nfs4_state_owner
*sp
;
267 sp
= kzalloc(sizeof(*sp
),GFP_KERNEL
);
270 spin_lock_init(&sp
->so_lock
);
271 INIT_LIST_HEAD(&sp
->so_states
);
272 INIT_LIST_HEAD(&sp
->so_delegations
);
273 rpc_init_wait_queue(&sp
->so_sequence
.wait
, "Seqid_waitqueue");
274 sp
->so_seqid
.sequence
= &sp
->so_sequence
;
275 spin_lock_init(&sp
->so_sequence
.lock
);
276 INIT_LIST_HEAD(&sp
->so_sequence
.list
);
277 atomic_set(&sp
->so_count
, 1);
282 nfs4_drop_state_owner(struct nfs4_state_owner
*sp
)
284 struct nfs4_client
*clp
= sp
->so_client
;
285 spin_lock(&clp
->cl_lock
);
286 list_del_init(&sp
->so_list
);
287 spin_unlock(&clp
->cl_lock
);
291 * Note: must be called with clp->cl_sem held in order to prevent races
292 * with reboot recovery!
294 struct nfs4_state_owner
*nfs4_get_state_owner(struct nfs_server
*server
, struct rpc_cred
*cred
)
296 struct nfs4_client
*clp
= server
->nfs4_state
;
297 struct nfs4_state_owner
*sp
, *new;
300 new = nfs4_alloc_state_owner();
301 spin_lock(&clp
->cl_lock
);
302 sp
= nfs4_find_state_owner(clp
, cred
);
304 sp
= nfs4_client_grab_unused(clp
, cred
);
305 if (sp
== NULL
&& new != NULL
) {
306 list_add(&new->so_list
, &clp
->cl_state_owners
);
307 new->so_client
= clp
;
308 new->so_id
= nfs4_alloc_lockowner_id(clp
);
313 spin_unlock(&clp
->cl_lock
);
323 * Must be called with clp->cl_sem held in order to avoid races
324 * with state recovery...
326 void nfs4_put_state_owner(struct nfs4_state_owner
*sp
)
328 struct nfs4_client
*clp
= sp
->so_client
;
329 struct rpc_cred
*cred
= sp
->so_cred
;
331 if (!atomic_dec_and_lock(&sp
->so_count
, &clp
->cl_lock
))
333 if (clp
->cl_nunused
>= OPENOWNER_POOL_SIZE
)
335 if (list_empty(&sp
->so_list
))
337 list_move(&sp
->so_list
, &clp
->cl_unused
);
339 spin_unlock(&clp
->cl_lock
);
344 list_del(&sp
->so_list
);
345 spin_unlock(&clp
->cl_lock
);
350 static struct nfs4_state
*
351 nfs4_alloc_open_state(void)
353 struct nfs4_state
*state
;
355 state
= kmalloc(sizeof(*state
), GFP_KERNEL
);
362 memset(state
->stateid
.data
, 0, sizeof(state
->stateid
.data
));
363 atomic_set(&state
->count
, 1);
364 INIT_LIST_HEAD(&state
->lock_states
);
365 spin_lock_init(&state
->state_lock
);
370 nfs4_state_set_mode_locked(struct nfs4_state
*state
, mode_t mode
)
372 if (state
->state
== mode
)
374 /* NB! List reordering - see the reclaim code for why. */
375 if ((mode
& FMODE_WRITE
) != (state
->state
& FMODE_WRITE
)) {
376 if (mode
& FMODE_WRITE
)
377 list_move(&state
->open_states
, &state
->owner
->so_states
);
379 list_move_tail(&state
->open_states
, &state
->owner
->so_states
);
382 list_del_init(&state
->inode_states
);
386 static struct nfs4_state
*
387 __nfs4_find_state(struct inode
*inode
, struct rpc_cred
*cred
, mode_t mode
)
389 struct nfs_inode
*nfsi
= NFS_I(inode
);
390 struct nfs4_state
*state
;
392 mode
&= (FMODE_READ
|FMODE_WRITE
);
393 list_for_each_entry(state
, &nfsi
->open_states
, inode_states
) {
394 if (state
->owner
->so_cred
!= cred
)
396 if ((state
->state
& mode
) != mode
)
398 atomic_inc(&state
->count
);
399 if (mode
& FMODE_READ
)
401 if (mode
& FMODE_WRITE
)
408 static struct nfs4_state
*
409 __nfs4_find_state_byowner(struct inode
*inode
, struct nfs4_state_owner
*owner
)
411 struct nfs_inode
*nfsi
= NFS_I(inode
);
412 struct nfs4_state
*state
;
414 list_for_each_entry(state
, &nfsi
->open_states
, inode_states
) {
415 /* Is this in the process of being freed? */
416 if (state
->state
== 0)
418 if (state
->owner
== owner
) {
419 atomic_inc(&state
->count
);
427 nfs4_find_state(struct inode
*inode
, struct rpc_cred
*cred
, mode_t mode
)
429 struct nfs4_state
*state
;
431 spin_lock(&inode
->i_lock
);
432 state
= __nfs4_find_state(inode
, cred
, mode
);
433 spin_unlock(&inode
->i_lock
);
438 nfs4_free_open_state(struct nfs4_state
*state
)
444 nfs4_get_open_state(struct inode
*inode
, struct nfs4_state_owner
*owner
)
446 struct nfs4_state
*state
, *new;
447 struct nfs_inode
*nfsi
= NFS_I(inode
);
449 spin_lock(&inode
->i_lock
);
450 state
= __nfs4_find_state_byowner(inode
, owner
);
451 spin_unlock(&inode
->i_lock
);
454 new = nfs4_alloc_open_state();
455 spin_lock(&owner
->so_lock
);
456 spin_lock(&inode
->i_lock
);
457 state
= __nfs4_find_state_byowner(inode
, owner
);
458 if (state
== NULL
&& new != NULL
) {
460 state
->owner
= owner
;
461 atomic_inc(&owner
->so_count
);
462 list_add(&state
->inode_states
, &nfsi
->open_states
);
463 state
->inode
= igrab(inode
);
464 spin_unlock(&inode
->i_lock
);
465 /* Note: The reclaim code dictates that we add stateless
466 * and read-only stateids to the end of the list */
467 list_add_tail(&state
->open_states
, &owner
->so_states
);
468 spin_unlock(&owner
->so_lock
);
470 spin_unlock(&inode
->i_lock
);
471 spin_unlock(&owner
->so_lock
);
473 nfs4_free_open_state(new);
480 * Beware! Caller must be holding exactly one
481 * reference to clp->cl_sem!
483 void nfs4_put_open_state(struct nfs4_state
*state
)
485 struct inode
*inode
= state
->inode
;
486 struct nfs4_state_owner
*owner
= state
->owner
;
488 if (!atomic_dec_and_lock(&state
->count
, &owner
->so_lock
))
490 spin_lock(&inode
->i_lock
);
491 if (!list_empty(&state
->inode_states
))
492 list_del(&state
->inode_states
);
493 list_del(&state
->open_states
);
494 spin_unlock(&inode
->i_lock
);
495 spin_unlock(&owner
->so_lock
);
497 nfs4_free_open_state(state
);
498 nfs4_put_state_owner(owner
);
502 * Close the current file.
504 void nfs4_close_state(struct nfs4_state
*state
, mode_t mode
)
506 struct inode
*inode
= state
->inode
;
507 struct nfs4_state_owner
*owner
= state
->owner
;
508 int oldstate
, newstate
= 0;
510 atomic_inc(&owner
->so_count
);
511 /* Protect against nfs4_find_state() */
512 spin_lock(&owner
->so_lock
);
513 spin_lock(&inode
->i_lock
);
514 if (mode
& FMODE_READ
)
516 if (mode
& FMODE_WRITE
)
518 oldstate
= newstate
= state
->state
;
519 if (state
->nreaders
== 0)
520 newstate
&= ~FMODE_READ
;
521 if (state
->nwriters
== 0)
522 newstate
&= ~FMODE_WRITE
;
523 if (test_bit(NFS_DELEGATED_STATE
, &state
->flags
)) {
524 nfs4_state_set_mode_locked(state
, newstate
);
527 spin_unlock(&inode
->i_lock
);
528 spin_unlock(&owner
->so_lock
);
530 if (oldstate
!= newstate
&& nfs4_do_close(inode
, state
) == 0)
532 nfs4_put_open_state(state
);
533 nfs4_put_state_owner(owner
);
537 * Search the state->lock_states for an existing lock_owner
538 * that is compatible with current->files
540 static struct nfs4_lock_state
*
541 __nfs4_find_lock_state(struct nfs4_state
*state
, fl_owner_t fl_owner
)
543 struct nfs4_lock_state
*pos
;
544 list_for_each_entry(pos
, &state
->lock_states
, ls_locks
) {
545 if (pos
->ls_owner
!= fl_owner
)
547 atomic_inc(&pos
->ls_count
);
554 * Return a compatible lock_state. If no initialized lock_state structure
555 * exists, return an uninitialized one.
558 static struct nfs4_lock_state
*nfs4_alloc_lock_state(struct nfs4_state
*state
, fl_owner_t fl_owner
)
560 struct nfs4_lock_state
*lsp
;
561 struct nfs4_client
*clp
= state
->owner
->so_client
;
563 lsp
= kzalloc(sizeof(*lsp
), GFP_KERNEL
);
566 lsp
->ls_seqid
.sequence
= &state
->owner
->so_sequence
;
567 atomic_set(&lsp
->ls_count
, 1);
568 lsp
->ls_owner
= fl_owner
;
569 spin_lock(&clp
->cl_lock
);
570 lsp
->ls_id
= nfs4_alloc_lockowner_id(clp
);
571 spin_unlock(&clp
->cl_lock
);
572 INIT_LIST_HEAD(&lsp
->ls_locks
);
577 * Return a compatible lock_state. If no initialized lock_state structure
578 * exists, return an uninitialized one.
580 * The caller must be holding clp->cl_sem
582 static struct nfs4_lock_state
*nfs4_get_lock_state(struct nfs4_state
*state
, fl_owner_t owner
)
584 struct nfs4_lock_state
*lsp
, *new = NULL
;
587 spin_lock(&state
->state_lock
);
588 lsp
= __nfs4_find_lock_state(state
, owner
);
592 new->ls_state
= state
;
593 list_add(&new->ls_locks
, &state
->lock_states
);
594 set_bit(LK_STATE_IN_USE
, &state
->flags
);
599 spin_unlock(&state
->state_lock
);
600 new = nfs4_alloc_lock_state(state
, owner
);
604 spin_unlock(&state
->state_lock
);
610 * Release reference to lock_state, and free it if we see that
611 * it is no longer in use
613 void nfs4_put_lock_state(struct nfs4_lock_state
*lsp
)
615 struct nfs4_state
*state
;
619 state
= lsp
->ls_state
;
620 if (!atomic_dec_and_lock(&lsp
->ls_count
, &state
->state_lock
))
622 list_del(&lsp
->ls_locks
);
623 if (list_empty(&state
->lock_states
))
624 clear_bit(LK_STATE_IN_USE
, &state
->flags
);
625 spin_unlock(&state
->state_lock
);
629 static void nfs4_fl_copy_lock(struct file_lock
*dst
, struct file_lock
*src
)
631 struct nfs4_lock_state
*lsp
= src
->fl_u
.nfs4_fl
.owner
;
633 dst
->fl_u
.nfs4_fl
.owner
= lsp
;
634 atomic_inc(&lsp
->ls_count
);
637 static void nfs4_fl_release_lock(struct file_lock
*fl
)
639 nfs4_put_lock_state(fl
->fl_u
.nfs4_fl
.owner
);
642 static struct file_lock_operations nfs4_fl_lock_ops
= {
643 .fl_copy_lock
= nfs4_fl_copy_lock
,
644 .fl_release_private
= nfs4_fl_release_lock
,
647 int nfs4_set_lock_state(struct nfs4_state
*state
, struct file_lock
*fl
)
649 struct nfs4_lock_state
*lsp
;
651 if (fl
->fl_ops
!= NULL
)
653 lsp
= nfs4_get_lock_state(state
, fl
->fl_owner
);
656 fl
->fl_u
.nfs4_fl
.owner
= lsp
;
657 fl
->fl_ops
= &nfs4_fl_lock_ops
;
662 * Byte-range lock aware utility to initialize the stateid of read/write
665 void nfs4_copy_stateid(nfs4_stateid
*dst
, struct nfs4_state
*state
, fl_owner_t fl_owner
)
667 struct nfs4_lock_state
*lsp
;
669 memcpy(dst
, &state
->stateid
, sizeof(*dst
));
670 if (test_bit(LK_STATE_IN_USE
, &state
->flags
) == 0)
673 spin_lock(&state
->state_lock
);
674 lsp
= __nfs4_find_lock_state(state
, fl_owner
);
675 if (lsp
!= NULL
&& (lsp
->ls_flags
& NFS_LOCK_INITIALIZED
) != 0)
676 memcpy(dst
, &lsp
->ls_stateid
, sizeof(*dst
));
677 spin_unlock(&state
->state_lock
);
678 nfs4_put_lock_state(lsp
);
681 struct nfs_seqid
*nfs_alloc_seqid(struct nfs_seqid_counter
*counter
)
683 struct nfs_seqid
*new;
685 new = kmalloc(sizeof(*new), GFP_KERNEL
);
687 new->sequence
= counter
;
688 INIT_LIST_HEAD(&new->list
);
693 void nfs_free_seqid(struct nfs_seqid
*seqid
)
695 struct rpc_sequence
*sequence
= seqid
->sequence
->sequence
;
697 if (!list_empty(&seqid
->list
)) {
698 spin_lock(&sequence
->lock
);
699 list_del(&seqid
->list
);
700 spin_unlock(&sequence
->lock
);
702 rpc_wake_up_next(&sequence
->wait
);
707 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
708 * failed with a seqid incrementing error -
709 * see comments nfs_fs.h:seqid_mutating_error()
711 static inline void nfs_increment_seqid(int status
, struct nfs_seqid
*seqid
)
716 case -NFS4ERR_BAD_SEQID
:
717 case -NFS4ERR_STALE_CLIENTID
:
718 case -NFS4ERR_STALE_STATEID
:
719 case -NFS4ERR_BAD_STATEID
:
720 case -NFS4ERR_BADXDR
:
721 case -NFS4ERR_RESOURCE
:
722 case -NFS4ERR_NOFILEHANDLE
:
723 /* Non-seqid mutating errors */
727 * Note: no locking needed as we are guaranteed to be first
728 * on the sequence list
730 seqid
->sequence
->counter
++;
733 void nfs_increment_open_seqid(int status
, struct nfs_seqid
*seqid
)
735 if (status
== -NFS4ERR_BAD_SEQID
) {
736 struct nfs4_state_owner
*sp
= container_of(seqid
->sequence
,
737 struct nfs4_state_owner
, so_seqid
);
738 nfs4_drop_state_owner(sp
);
740 return nfs_increment_seqid(status
, seqid
);
744 * Increment the seqid if the LOCK/LOCKU succeeded, or
745 * failed with a seqid incrementing error -
746 * see comments nfs_fs.h:seqid_mutating_error()
748 void nfs_increment_lock_seqid(int status
, struct nfs_seqid
*seqid
)
750 return nfs_increment_seqid(status
, seqid
);
753 int nfs_wait_on_sequence(struct nfs_seqid
*seqid
, struct rpc_task
*task
)
755 struct rpc_sequence
*sequence
= seqid
->sequence
->sequence
;
758 if (sequence
->list
.next
== &seqid
->list
)
760 spin_lock(&sequence
->lock
);
761 if (!list_empty(&sequence
->list
)) {
762 rpc_sleep_on(&sequence
->wait
, task
, NULL
, NULL
);
765 list_add(&seqid
->list
, &sequence
->list
);
766 spin_unlock(&sequence
->lock
);
771 static int reclaimer(void *);
772 struct reclaimer_args
{
773 struct nfs4_client
*clp
;
774 struct completion complete
;
778 * State recovery routine
781 nfs4_recover_state(void *data
)
783 struct nfs4_client
*clp
= (struct nfs4_client
*)data
;
784 struct reclaimer_args args
= {
789 init_completion(&args
.complete
);
791 if (kernel_thread(reclaimer
, &args
, CLONE_KERNEL
) < 0)
792 goto out_failed_clear
;
793 wait_for_completion(&args
.complete
);
796 set_bit(NFS4CLNT_OK
, &clp
->cl_state
);
797 wake_up_all(&clp
->cl_waitq
);
798 rpc_wake_up(&clp
->cl_rpcwaitq
);
802 * Schedule a state recovery attempt
805 nfs4_schedule_state_recovery(struct nfs4_client
*clp
)
809 if (test_and_clear_bit(NFS4CLNT_OK
, &clp
->cl_state
))
810 schedule_work(&clp
->cl_recoverd
);
813 static int nfs4_reclaim_locks(struct nfs4_state_recovery_ops
*ops
, struct nfs4_state
*state
)
815 struct inode
*inode
= state
->inode
;
816 struct file_lock
*fl
;
819 for (fl
= inode
->i_flock
; fl
!= 0; fl
= fl
->fl_next
) {
820 if (!(fl
->fl_flags
& FL_POSIX
))
822 if (((struct nfs_open_context
*)fl
->fl_file
->private_data
)->state
!= state
)
824 status
= ops
->recover_lock(state
, fl
);
829 printk(KERN_ERR
"%s: unhandled error %d. Zeroing state\n",
830 __FUNCTION__
, status
);
831 case -NFS4ERR_EXPIRED
:
832 case -NFS4ERR_NO_GRACE
:
833 case -NFS4ERR_RECLAIM_BAD
:
834 case -NFS4ERR_RECLAIM_CONFLICT
:
835 /* kill_proc(fl->fl_owner, SIGLOST, 1); */
837 case -NFS4ERR_STALE_CLIENTID
:
846 static int nfs4_reclaim_open_state(struct nfs4_state_recovery_ops
*ops
, struct nfs4_state_owner
*sp
)
848 struct nfs4_state
*state
;
849 struct nfs4_lock_state
*lock
;
852 /* Note: we rely on the sp->so_states list being ordered
853 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
855 * This is needed to ensure that the server won't give us any
856 * read delegations that we have to return if, say, we are
857 * recovering after a network partition or a reboot from a
858 * server that doesn't support a grace period.
860 list_for_each_entry(state
, &sp
->so_states
, open_states
) {
861 if (state
->state
== 0)
863 status
= ops
->recover_open(sp
, state
);
865 status
= nfs4_reclaim_locks(ops
, state
);
868 list_for_each_entry(lock
, &state
->lock_states
, ls_locks
) {
869 if (!(lock
->ls_flags
& NFS_LOCK_INITIALIZED
))
870 printk("%s: Lock reclaim failed!\n",
877 printk(KERN_ERR
"%s: unhandled error %d. Zeroing state\n",
878 __FUNCTION__
, status
);
880 case -NFS4ERR_RECLAIM_BAD
:
881 case -NFS4ERR_RECLAIM_CONFLICT
:
883 * Open state on this file cannot be recovered
884 * All we can do is revert to using the zero stateid.
886 memset(state
->stateid
.data
, 0,
887 sizeof(state
->stateid
.data
));
888 /* Mark the file as being 'closed' */
891 case -NFS4ERR_EXPIRED
:
892 case -NFS4ERR_NO_GRACE
:
893 case -NFS4ERR_STALE_CLIENTID
:
902 static void nfs4_state_mark_reclaim(struct nfs4_client
*clp
)
904 struct nfs4_state_owner
*sp
;
905 struct nfs4_state
*state
;
906 struct nfs4_lock_state
*lock
;
908 /* Reset all sequence ids to zero */
909 list_for_each_entry(sp
, &clp
->cl_state_owners
, so_list
) {
910 sp
->so_seqid
.counter
= 0;
911 sp
->so_seqid
.flags
= 0;
912 spin_lock(&sp
->so_lock
);
913 list_for_each_entry(state
, &sp
->so_states
, open_states
) {
914 list_for_each_entry(lock
, &state
->lock_states
, ls_locks
) {
915 lock
->ls_seqid
.counter
= 0;
916 lock
->ls_seqid
.flags
= 0;
917 lock
->ls_flags
&= ~NFS_LOCK_INITIALIZED
;
920 spin_unlock(&sp
->so_lock
);
924 static int reclaimer(void *ptr
)
926 struct reclaimer_args
*args
= (struct reclaimer_args
*)ptr
;
927 struct nfs4_client
*clp
= args
->clp
;
928 struct nfs4_state_owner
*sp
;
929 struct nfs4_state_recovery_ops
*ops
;
932 daemonize("%u.%u.%u.%u-reclaim", NIPQUAD(clp
->cl_addr
));
933 allow_signal(SIGKILL
);
935 atomic_inc(&clp
->cl_count
);
936 complete(&args
->complete
);
938 /* Ensure exclusive access to NFSv4 state */
940 down_write(&clp
->cl_sem
);
941 /* Are there any NFS mounts out there? */
942 if (list_empty(&clp
->cl_superblocks
))
945 status
= nfs4_proc_renew(clp
);
948 case -NFS4ERR_CB_PATH_DOWN
:
950 case -NFS4ERR_STALE_CLIENTID
:
951 case -NFS4ERR_LEASE_MOVED
:
952 ops
= &nfs4_reboot_recovery_ops
;
955 ops
= &nfs4_network_partition_recovery_ops
;
957 nfs4_state_mark_reclaim(clp
);
958 status
= __nfs4_init_client(clp
);
961 /* Mark all delegations for reclaim */
962 nfs_delegation_mark_reclaim(clp
);
963 /* Note: list is protected by exclusive lock on cl->cl_sem */
964 list_for_each_entry(sp
, &clp
->cl_state_owners
, so_list
) {
965 status
= nfs4_reclaim_open_state(ops
, sp
);
967 if (status
== -NFS4ERR_NO_GRACE
) {
968 ops
= &nfs4_network_partition_recovery_ops
;
969 status
= nfs4_reclaim_open_state(ops
, sp
);
971 if (status
== -NFS4ERR_STALE_CLIENTID
)
973 if (status
== -NFS4ERR_EXPIRED
)
977 nfs_delegation_reap_unclaimed(clp
);
979 set_bit(NFS4CLNT_OK
, &clp
->cl_state
);
980 up_write(&clp
->cl_sem
);
982 wake_up_all(&clp
->cl_waitq
);
983 rpc_wake_up(&clp
->cl_rpcwaitq
);
984 if (status
== -NFS4ERR_CB_PATH_DOWN
)
985 nfs_handle_cb_pathdown(clp
);
986 nfs4_put_client(clp
);
989 printk(KERN_WARNING
"Error: state recovery failed on NFSv4 server %u.%u.%u.%u with error %d\n",
990 NIPQUAD(clp
->cl_addr
.s_addr
), -status
);