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/kernel.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/kthread.h>
47 #include <linux/module.h>
48 #include <linux/random.h>
49 #include <linux/workqueue.h>
50 #include <linux/bitops.h>
54 #include "delegation.h"
57 #define OPENOWNER_POOL_SIZE 8
59 const nfs4_stateid zero_stateid
;
61 static LIST_HEAD(nfs4_clientid_list
);
63 static int nfs4_init_client(struct nfs_client
*clp
, struct rpc_cred
*cred
)
65 int status
= nfs4_proc_setclientid(clp
, NFS4_CALLBACK
,
66 nfs_callback_tcpport
, cred
);
68 status
= nfs4_proc_setclientid_confirm(clp
, cred
);
70 nfs4_schedule_state_renewal(clp
);
74 struct rpc_cred
*nfs4_get_renew_cred(struct nfs_client
*clp
)
76 struct nfs4_state_owner
*sp
;
78 struct rpc_cred
*cred
= NULL
;
80 for (pos
= rb_first(&clp
->cl_state_owners
); pos
!= NULL
; pos
= rb_next(pos
)) {
81 sp
= rb_entry(pos
, struct nfs4_state_owner
, so_client_node
);
82 if (list_empty(&sp
->so_states
))
84 cred
= get_rpccred(sp
->so_cred
);
90 static struct rpc_cred
*nfs4_get_setclientid_cred(struct nfs_client
*clp
)
92 struct nfs4_state_owner
*sp
;
95 pos
= rb_first(&clp
->cl_state_owners
);
97 sp
= rb_entry(pos
, struct nfs4_state_owner
, so_client_node
);
98 return get_rpccred(sp
->so_cred
);
103 static void nfs_alloc_unique_id(struct rb_root
*root
, struct nfs_unique_id
*new,
104 __u64 minval
, int maxbits
)
106 struct rb_node
**p
, *parent
;
107 struct nfs_unique_id
*pos
;
111 mask
= (1ULL << maxbits
) - 1ULL;
113 /* Ensure distribution is more or less flat */
114 get_random_bytes(&new->id
, sizeof(new->id
));
116 if (new->id
< minval
)
124 pos
= rb_entry(parent
, struct nfs_unique_id
, rb_node
);
126 if (new->id
< pos
->id
)
128 else if (new->id
> pos
->id
)
133 rb_link_node(&new->rb_node
, parent
, p
);
134 rb_insert_color(&new->rb_node
, root
);
139 if (new->id
< minval
|| (new->id
& mask
) != new->id
) {
143 parent
= rb_next(parent
);
146 pos
= rb_entry(parent
, struct nfs_unique_id
, rb_node
);
147 if (new->id
< pos
->id
)
153 static void nfs_free_unique_id(struct rb_root
*root
, struct nfs_unique_id
*id
)
155 rb_erase(&id
->rb_node
, root
);
158 static struct nfs4_state_owner
*
159 nfs4_find_state_owner(struct nfs_server
*server
, struct rpc_cred
*cred
)
161 struct nfs_client
*clp
= server
->nfs_client
;
162 struct rb_node
**p
= &clp
->cl_state_owners
.rb_node
,
164 struct nfs4_state_owner
*sp
, *res
= NULL
;
168 sp
= rb_entry(parent
, struct nfs4_state_owner
, so_client_node
);
170 if (server
< sp
->so_server
) {
171 p
= &parent
->rb_left
;
174 if (server
> sp
->so_server
) {
175 p
= &parent
->rb_right
;
178 if (cred
< sp
->so_cred
)
179 p
= &parent
->rb_left
;
180 else if (cred
> sp
->so_cred
)
181 p
= &parent
->rb_right
;
183 atomic_inc(&sp
->so_count
);
191 static struct nfs4_state_owner
*
192 nfs4_insert_state_owner(struct nfs_client
*clp
, struct nfs4_state_owner
*new)
194 struct rb_node
**p
= &clp
->cl_state_owners
.rb_node
,
196 struct nfs4_state_owner
*sp
;
200 sp
= rb_entry(parent
, struct nfs4_state_owner
, so_client_node
);
202 if (new->so_server
< sp
->so_server
) {
203 p
= &parent
->rb_left
;
206 if (new->so_server
> sp
->so_server
) {
207 p
= &parent
->rb_right
;
210 if (new->so_cred
< sp
->so_cred
)
211 p
= &parent
->rb_left
;
212 else if (new->so_cred
> sp
->so_cred
)
213 p
= &parent
->rb_right
;
215 atomic_inc(&sp
->so_count
);
219 nfs_alloc_unique_id(&clp
->cl_openowner_id
, &new->so_owner_id
, 1, 64);
220 rb_link_node(&new->so_client_node
, parent
, p
);
221 rb_insert_color(&new->so_client_node
, &clp
->cl_state_owners
);
226 nfs4_remove_state_owner(struct nfs_client
*clp
, struct nfs4_state_owner
*sp
)
228 if (!RB_EMPTY_NODE(&sp
->so_client_node
))
229 rb_erase(&sp
->so_client_node
, &clp
->cl_state_owners
);
230 nfs_free_unique_id(&clp
->cl_openowner_id
, &sp
->so_owner_id
);
234 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
235 * create a new state_owner.
238 static struct nfs4_state_owner
*
239 nfs4_alloc_state_owner(void)
241 struct nfs4_state_owner
*sp
;
243 sp
= kzalloc(sizeof(*sp
),GFP_KERNEL
);
246 spin_lock_init(&sp
->so_lock
);
247 INIT_LIST_HEAD(&sp
->so_states
);
248 INIT_LIST_HEAD(&sp
->so_delegations
);
249 rpc_init_wait_queue(&sp
->so_sequence
.wait
, "Seqid_waitqueue");
250 sp
->so_seqid
.sequence
= &sp
->so_sequence
;
251 spin_lock_init(&sp
->so_sequence
.lock
);
252 INIT_LIST_HEAD(&sp
->so_sequence
.list
);
253 atomic_set(&sp
->so_count
, 1);
258 nfs4_drop_state_owner(struct nfs4_state_owner
*sp
)
260 if (!RB_EMPTY_NODE(&sp
->so_client_node
)) {
261 struct nfs_client
*clp
= sp
->so_client
;
263 spin_lock(&clp
->cl_lock
);
264 rb_erase(&sp
->so_client_node
, &clp
->cl_state_owners
);
265 RB_CLEAR_NODE(&sp
->so_client_node
);
266 spin_unlock(&clp
->cl_lock
);
271 * Note: must be called with clp->cl_sem held in order to prevent races
272 * with reboot recovery!
274 struct nfs4_state_owner
*nfs4_get_state_owner(struct nfs_server
*server
, struct rpc_cred
*cred
)
276 struct nfs_client
*clp
= server
->nfs_client
;
277 struct nfs4_state_owner
*sp
, *new;
279 spin_lock(&clp
->cl_lock
);
280 sp
= nfs4_find_state_owner(server
, cred
);
281 spin_unlock(&clp
->cl_lock
);
284 new = nfs4_alloc_state_owner();
287 new->so_client
= clp
;
288 new->so_server
= server
;
290 spin_lock(&clp
->cl_lock
);
291 sp
= nfs4_insert_state_owner(clp
, new);
292 spin_unlock(&clp
->cl_lock
);
301 * Must be called with clp->cl_sem held in order to avoid races
302 * with state recovery...
304 void nfs4_put_state_owner(struct nfs4_state_owner
*sp
)
306 struct nfs_client
*clp
= sp
->so_client
;
307 struct rpc_cred
*cred
= sp
->so_cred
;
309 if (!atomic_dec_and_lock(&sp
->so_count
, &clp
->cl_lock
))
311 nfs4_remove_state_owner(clp
, sp
);
312 spin_unlock(&clp
->cl_lock
);
317 static struct nfs4_state
*
318 nfs4_alloc_open_state(void)
320 struct nfs4_state
*state
;
322 state
= kzalloc(sizeof(*state
), GFP_KERNEL
);
325 atomic_set(&state
->count
, 1);
326 INIT_LIST_HEAD(&state
->lock_states
);
327 spin_lock_init(&state
->state_lock
);
328 seqlock_init(&state
->seqlock
);
333 nfs4_state_set_mode_locked(struct nfs4_state
*state
, mode_t mode
)
335 if (state
->state
== mode
)
337 /* NB! List reordering - see the reclaim code for why. */
338 if ((mode
& FMODE_WRITE
) != (state
->state
& FMODE_WRITE
)) {
339 if (mode
& FMODE_WRITE
)
340 list_move(&state
->open_states
, &state
->owner
->so_states
);
342 list_move_tail(&state
->open_states
, &state
->owner
->so_states
);
345 list_del_init(&state
->inode_states
);
349 static struct nfs4_state
*
350 __nfs4_find_state_byowner(struct inode
*inode
, struct nfs4_state_owner
*owner
)
352 struct nfs_inode
*nfsi
= NFS_I(inode
);
353 struct nfs4_state
*state
;
355 list_for_each_entry(state
, &nfsi
->open_states
, inode_states
) {
356 if (state
->owner
!= owner
)
358 if (atomic_inc_not_zero(&state
->count
))
365 nfs4_free_open_state(struct nfs4_state
*state
)
371 nfs4_get_open_state(struct inode
*inode
, struct nfs4_state_owner
*owner
)
373 struct nfs4_state
*state
, *new;
374 struct nfs_inode
*nfsi
= NFS_I(inode
);
376 spin_lock(&inode
->i_lock
);
377 state
= __nfs4_find_state_byowner(inode
, owner
);
378 spin_unlock(&inode
->i_lock
);
381 new = nfs4_alloc_open_state();
382 spin_lock(&owner
->so_lock
);
383 spin_lock(&inode
->i_lock
);
384 state
= __nfs4_find_state_byowner(inode
, owner
);
385 if (state
== NULL
&& new != NULL
) {
387 state
->owner
= owner
;
388 atomic_inc(&owner
->so_count
);
389 list_add(&state
->inode_states
, &nfsi
->open_states
);
390 state
->inode
= igrab(inode
);
391 spin_unlock(&inode
->i_lock
);
392 /* Note: The reclaim code dictates that we add stateless
393 * and read-only stateids to the end of the list */
394 list_add_tail(&state
->open_states
, &owner
->so_states
);
395 spin_unlock(&owner
->so_lock
);
397 spin_unlock(&inode
->i_lock
);
398 spin_unlock(&owner
->so_lock
);
400 nfs4_free_open_state(new);
407 * Beware! Caller must be holding exactly one
408 * reference to clp->cl_sem!
410 void nfs4_put_open_state(struct nfs4_state
*state
)
412 struct inode
*inode
= state
->inode
;
413 struct nfs4_state_owner
*owner
= state
->owner
;
415 if (!atomic_dec_and_lock(&state
->count
, &owner
->so_lock
))
417 spin_lock(&inode
->i_lock
);
418 if (!list_empty(&state
->inode_states
))
419 list_del(&state
->inode_states
);
420 list_del(&state
->open_states
);
421 spin_unlock(&inode
->i_lock
);
422 spin_unlock(&owner
->so_lock
);
424 nfs4_free_open_state(state
);
425 nfs4_put_state_owner(owner
);
429 * Close the current file.
431 void nfs4_close_state(struct path
*path
, struct nfs4_state
*state
, mode_t mode
)
433 struct nfs4_state_owner
*owner
= state
->owner
;
437 atomic_inc(&owner
->so_count
);
438 /* Protect against nfs4_find_state() */
439 spin_lock(&owner
->so_lock
);
440 switch (mode
& (FMODE_READ
| FMODE_WRITE
)) {
447 case FMODE_READ
|FMODE_WRITE
:
450 newstate
= FMODE_READ
|FMODE_WRITE
;
451 if (state
->n_rdwr
== 0) {
452 if (state
->n_rdonly
== 0) {
453 newstate
&= ~FMODE_READ
;
454 call_close
|= test_bit(NFS_O_RDONLY_STATE
, &state
->flags
);
455 call_close
|= test_bit(NFS_O_RDWR_STATE
, &state
->flags
);
457 if (state
->n_wronly
== 0) {
458 newstate
&= ~FMODE_WRITE
;
459 call_close
|= test_bit(NFS_O_WRONLY_STATE
, &state
->flags
);
460 call_close
|= test_bit(NFS_O_RDWR_STATE
, &state
->flags
);
463 clear_bit(NFS_DELEGATED_STATE
, &state
->flags
);
465 nfs4_state_set_mode_locked(state
, newstate
);
466 spin_unlock(&owner
->so_lock
);
469 nfs4_put_open_state(state
);
470 nfs4_put_state_owner(owner
);
472 nfs4_do_close(path
, state
);
476 * Search the state->lock_states for an existing lock_owner
477 * that is compatible with current->files
479 static struct nfs4_lock_state
*
480 __nfs4_find_lock_state(struct nfs4_state
*state
, fl_owner_t fl_owner
)
482 struct nfs4_lock_state
*pos
;
483 list_for_each_entry(pos
, &state
->lock_states
, ls_locks
) {
484 if (pos
->ls_owner
!= fl_owner
)
486 atomic_inc(&pos
->ls_count
);
493 * Return a compatible lock_state. If no initialized lock_state structure
494 * exists, return an uninitialized one.
497 static struct nfs4_lock_state
*nfs4_alloc_lock_state(struct nfs4_state
*state
, fl_owner_t fl_owner
)
499 struct nfs4_lock_state
*lsp
;
500 struct nfs_client
*clp
= state
->owner
->so_client
;
502 lsp
= kzalloc(sizeof(*lsp
), GFP_KERNEL
);
505 lsp
->ls_seqid
.sequence
= &state
->owner
->so_sequence
;
506 atomic_set(&lsp
->ls_count
, 1);
507 lsp
->ls_owner
= fl_owner
;
508 spin_lock(&clp
->cl_lock
);
509 nfs_alloc_unique_id(&clp
->cl_lockowner_id
, &lsp
->ls_id
, 1, 64);
510 spin_unlock(&clp
->cl_lock
);
511 INIT_LIST_HEAD(&lsp
->ls_locks
);
515 static void nfs4_free_lock_state(struct nfs4_lock_state
*lsp
)
517 struct nfs_client
*clp
= lsp
->ls_state
->owner
->so_client
;
519 spin_lock(&clp
->cl_lock
);
520 nfs_free_unique_id(&clp
->cl_lockowner_id
, &lsp
->ls_id
);
521 spin_unlock(&clp
->cl_lock
);
526 * Return a compatible lock_state. If no initialized lock_state structure
527 * exists, return an uninitialized one.
529 * The caller must be holding clp->cl_sem
531 static struct nfs4_lock_state
*nfs4_get_lock_state(struct nfs4_state
*state
, fl_owner_t owner
)
533 struct nfs4_lock_state
*lsp
, *new = NULL
;
536 spin_lock(&state
->state_lock
);
537 lsp
= __nfs4_find_lock_state(state
, owner
);
541 new->ls_state
= state
;
542 list_add(&new->ls_locks
, &state
->lock_states
);
543 set_bit(LK_STATE_IN_USE
, &state
->flags
);
548 spin_unlock(&state
->state_lock
);
549 new = nfs4_alloc_lock_state(state
, owner
);
553 spin_unlock(&state
->state_lock
);
555 nfs4_free_lock_state(new);
560 * Release reference to lock_state, and free it if we see that
561 * it is no longer in use
563 void nfs4_put_lock_state(struct nfs4_lock_state
*lsp
)
565 struct nfs4_state
*state
;
569 state
= lsp
->ls_state
;
570 if (!atomic_dec_and_lock(&lsp
->ls_count
, &state
->state_lock
))
572 list_del(&lsp
->ls_locks
);
573 if (list_empty(&state
->lock_states
))
574 clear_bit(LK_STATE_IN_USE
, &state
->flags
);
575 spin_unlock(&state
->state_lock
);
576 nfs4_free_lock_state(lsp
);
579 static void nfs4_fl_copy_lock(struct file_lock
*dst
, struct file_lock
*src
)
581 struct nfs4_lock_state
*lsp
= src
->fl_u
.nfs4_fl
.owner
;
583 dst
->fl_u
.nfs4_fl
.owner
= lsp
;
584 atomic_inc(&lsp
->ls_count
);
587 static void nfs4_fl_release_lock(struct file_lock
*fl
)
589 nfs4_put_lock_state(fl
->fl_u
.nfs4_fl
.owner
);
592 static struct file_lock_operations nfs4_fl_lock_ops
= {
593 .fl_copy_lock
= nfs4_fl_copy_lock
,
594 .fl_release_private
= nfs4_fl_release_lock
,
597 int nfs4_set_lock_state(struct nfs4_state
*state
, struct file_lock
*fl
)
599 struct nfs4_lock_state
*lsp
;
601 if (fl
->fl_ops
!= NULL
)
603 lsp
= nfs4_get_lock_state(state
, fl
->fl_owner
);
606 fl
->fl_u
.nfs4_fl
.owner
= lsp
;
607 fl
->fl_ops
= &nfs4_fl_lock_ops
;
612 * Byte-range lock aware utility to initialize the stateid of read/write
615 void nfs4_copy_stateid(nfs4_stateid
*dst
, struct nfs4_state
*state
, fl_owner_t fl_owner
)
617 struct nfs4_lock_state
*lsp
;
621 seq
= read_seqbegin(&state
->seqlock
);
622 memcpy(dst
, &state
->stateid
, sizeof(*dst
));
623 } while (read_seqretry(&state
->seqlock
, seq
));
624 if (test_bit(LK_STATE_IN_USE
, &state
->flags
) == 0)
627 spin_lock(&state
->state_lock
);
628 lsp
= __nfs4_find_lock_state(state
, fl_owner
);
629 if (lsp
!= NULL
&& (lsp
->ls_flags
& NFS_LOCK_INITIALIZED
) != 0)
630 memcpy(dst
, &lsp
->ls_stateid
, sizeof(*dst
));
631 spin_unlock(&state
->state_lock
);
632 nfs4_put_lock_state(lsp
);
635 struct nfs_seqid
*nfs_alloc_seqid(struct nfs_seqid_counter
*counter
)
637 struct rpc_sequence
*sequence
= counter
->sequence
;
638 struct nfs_seqid
*new;
640 new = kmalloc(sizeof(*new), GFP_KERNEL
);
642 new->sequence
= counter
;
643 spin_lock(&sequence
->lock
);
644 list_add_tail(&new->list
, &sequence
->list
);
645 spin_unlock(&sequence
->lock
);
650 void nfs_free_seqid(struct nfs_seqid
*seqid
)
652 struct rpc_sequence
*sequence
= seqid
->sequence
->sequence
;
654 spin_lock(&sequence
->lock
);
655 list_del(&seqid
->list
);
656 spin_unlock(&sequence
->lock
);
657 rpc_wake_up(&sequence
->wait
);
662 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
663 * failed with a seqid incrementing error -
664 * see comments nfs_fs.h:seqid_mutating_error()
666 static void nfs_increment_seqid(int status
, struct nfs_seqid
*seqid
)
671 case -NFS4ERR_BAD_SEQID
:
672 if (seqid
->sequence
->flags
& NFS_SEQID_CONFIRMED
)
674 printk(KERN_WARNING
"NFS: v4 server returned a bad"
675 "sequence-id error on an"
676 "unconfirmed sequence %p!\n",
678 case -NFS4ERR_STALE_CLIENTID
:
679 case -NFS4ERR_STALE_STATEID
:
680 case -NFS4ERR_BAD_STATEID
:
681 case -NFS4ERR_BADXDR
:
682 case -NFS4ERR_RESOURCE
:
683 case -NFS4ERR_NOFILEHANDLE
:
684 /* Non-seqid mutating errors */
688 * Note: no locking needed as we are guaranteed to be first
689 * on the sequence list
691 seqid
->sequence
->counter
++;
694 void nfs_increment_open_seqid(int status
, struct nfs_seqid
*seqid
)
696 if (status
== -NFS4ERR_BAD_SEQID
) {
697 struct nfs4_state_owner
*sp
= container_of(seqid
->sequence
,
698 struct nfs4_state_owner
, so_seqid
);
699 nfs4_drop_state_owner(sp
);
701 nfs_increment_seqid(status
, seqid
);
705 * Increment the seqid if the LOCK/LOCKU succeeded, or
706 * failed with a seqid incrementing error -
707 * see comments nfs_fs.h:seqid_mutating_error()
709 void nfs_increment_lock_seqid(int status
, struct nfs_seqid
*seqid
)
711 nfs_increment_seqid(status
, seqid
);
714 int nfs_wait_on_sequence(struct nfs_seqid
*seqid
, struct rpc_task
*task
)
716 struct rpc_sequence
*sequence
= seqid
->sequence
->sequence
;
719 if (sequence
->list
.next
== &seqid
->list
)
721 spin_lock(&sequence
->lock
);
722 if (sequence
->list
.next
!= &seqid
->list
) {
723 rpc_sleep_on(&sequence
->wait
, task
, NULL
, NULL
);
726 spin_unlock(&sequence
->lock
);
731 static int reclaimer(void *);
733 static inline void nfs4_clear_recover_bit(struct nfs_client
*clp
)
735 smp_mb__before_clear_bit();
736 clear_bit(NFS4CLNT_STATE_RECOVER
, &clp
->cl_state
);
737 smp_mb__after_clear_bit();
738 wake_up_bit(&clp
->cl_state
, NFS4CLNT_STATE_RECOVER
);
739 rpc_wake_up(&clp
->cl_rpcwaitq
);
743 * State recovery routine
745 static void nfs4_recover_state(struct nfs_client
*clp
)
747 struct task_struct
*task
;
749 __module_get(THIS_MODULE
);
750 atomic_inc(&clp
->cl_count
);
751 task
= kthread_run(reclaimer
, clp
, "%u.%u.%u.%u-reclaim",
752 NIPQUAD(clp
->cl_addr
.sin_addr
));
755 nfs4_clear_recover_bit(clp
);
757 module_put(THIS_MODULE
);
761 * Schedule a state recovery attempt
763 void nfs4_schedule_state_recovery(struct nfs_client
*clp
)
767 if (test_and_set_bit(NFS4CLNT_STATE_RECOVER
, &clp
->cl_state
) == 0)
768 nfs4_recover_state(clp
);
771 static int nfs4_reclaim_locks(struct nfs4_state_recovery_ops
*ops
, struct nfs4_state
*state
)
773 struct inode
*inode
= state
->inode
;
774 struct file_lock
*fl
;
777 for (fl
= inode
->i_flock
; fl
!= 0; fl
= fl
->fl_next
) {
778 if (!(fl
->fl_flags
& (FL_POSIX
|FL_FLOCK
)))
780 if (((struct nfs_open_context
*)fl
->fl_file
->private_data
)->state
!= state
)
782 status
= ops
->recover_lock(state
, fl
);
787 printk(KERN_ERR
"%s: unhandled error %d. Zeroing state\n",
788 __FUNCTION__
, status
);
789 case -NFS4ERR_EXPIRED
:
790 case -NFS4ERR_NO_GRACE
:
791 case -NFS4ERR_RECLAIM_BAD
:
792 case -NFS4ERR_RECLAIM_CONFLICT
:
793 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
795 case -NFS4ERR_STALE_CLIENTID
:
804 static int nfs4_reclaim_open_state(struct nfs4_state_recovery_ops
*ops
, struct nfs4_state_owner
*sp
)
806 struct nfs4_state
*state
;
807 struct nfs4_lock_state
*lock
;
810 /* Note: we rely on the sp->so_states list being ordered
811 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
813 * This is needed to ensure that the server won't give us any
814 * read delegations that we have to return if, say, we are
815 * recovering after a network partition or a reboot from a
816 * server that doesn't support a grace period.
818 list_for_each_entry(state
, &sp
->so_states
, open_states
) {
819 if (state
->state
== 0)
821 status
= ops
->recover_open(sp
, state
);
823 status
= nfs4_reclaim_locks(ops
, state
);
826 list_for_each_entry(lock
, &state
->lock_states
, ls_locks
) {
827 if (!(lock
->ls_flags
& NFS_LOCK_INITIALIZED
))
828 printk("%s: Lock reclaim failed!\n",
835 printk(KERN_ERR
"%s: unhandled error %d. Zeroing state\n",
836 __FUNCTION__
, status
);
838 case -NFS4ERR_RECLAIM_BAD
:
839 case -NFS4ERR_RECLAIM_CONFLICT
:
841 * Open state on this file cannot be recovered
842 * All we can do is revert to using the zero stateid.
844 memset(state
->stateid
.data
, 0,
845 sizeof(state
->stateid
.data
));
846 /* Mark the file as being 'closed' */
849 case -NFS4ERR_EXPIRED
:
850 case -NFS4ERR_NO_GRACE
:
851 case -NFS4ERR_STALE_CLIENTID
:
860 static void nfs4_state_mark_reclaim(struct nfs_client
*clp
)
862 struct nfs4_state_owner
*sp
;
864 struct nfs4_state
*state
;
865 struct nfs4_lock_state
*lock
;
867 /* Reset all sequence ids to zero */
868 for (pos
= rb_first(&clp
->cl_state_owners
); pos
!= NULL
; pos
= rb_next(pos
)) {
869 sp
= rb_entry(pos
, struct nfs4_state_owner
, so_client_node
);
870 sp
->so_seqid
.counter
= 0;
871 sp
->so_seqid
.flags
= 0;
872 spin_lock(&sp
->so_lock
);
873 list_for_each_entry(state
, &sp
->so_states
, open_states
) {
874 clear_bit(NFS_DELEGATED_STATE
, &state
->flags
);
875 clear_bit(NFS_O_RDONLY_STATE
, &state
->flags
);
876 clear_bit(NFS_O_WRONLY_STATE
, &state
->flags
);
877 clear_bit(NFS_O_RDWR_STATE
, &state
->flags
);
878 list_for_each_entry(lock
, &state
->lock_states
, ls_locks
) {
879 lock
->ls_seqid
.counter
= 0;
880 lock
->ls_seqid
.flags
= 0;
881 lock
->ls_flags
&= ~NFS_LOCK_INITIALIZED
;
884 spin_unlock(&sp
->so_lock
);
888 static int reclaimer(void *ptr
)
890 struct nfs_client
*clp
= ptr
;
891 struct nfs4_state_owner
*sp
;
893 struct nfs4_state_recovery_ops
*ops
;
894 struct rpc_cred
*cred
;
897 allow_signal(SIGKILL
);
899 /* Ensure exclusive access to NFSv4 state */
901 down_write(&clp
->cl_sem
);
902 /* Are there any NFS mounts out there? */
903 if (list_empty(&clp
->cl_superblocks
))
906 ops
= &nfs4_network_partition_recovery_ops
;
907 /* Are there any open files on this volume? */
908 cred
= nfs4_get_renew_cred(clp
);
910 /* Yes there are: try to renew the old lease */
911 status
= nfs4_proc_renew(clp
, cred
);
914 case -NFS4ERR_CB_PATH_DOWN
:
917 case -NFS4ERR_STALE_CLIENTID
:
918 case -NFS4ERR_LEASE_MOVED
:
919 ops
= &nfs4_reboot_recovery_ops
;
922 /* "reboot" to ensure we clear all state on the server */
923 clp
->cl_boot_time
= CURRENT_TIME
;
924 cred
= nfs4_get_setclientid_cred(clp
);
926 /* We're going to have to re-establish a clientid */
927 nfs4_state_mark_reclaim(clp
);
930 status
= nfs4_init_client(clp
, cred
);
935 /* Mark all delegations for reclaim */
936 nfs_delegation_mark_reclaim(clp
);
937 /* Note: list is protected by exclusive lock on cl->cl_sem */
938 for (pos
= rb_first(&clp
->cl_state_owners
); pos
!= NULL
; pos
= rb_next(pos
)) {
939 sp
= rb_entry(pos
, struct nfs4_state_owner
, so_client_node
);
940 status
= nfs4_reclaim_open_state(ops
, sp
);
942 if (status
== -NFS4ERR_NO_GRACE
) {
943 ops
= &nfs4_network_partition_recovery_ops
;
944 status
= nfs4_reclaim_open_state(ops
, sp
);
946 if (status
== -NFS4ERR_STALE_CLIENTID
)
948 if (status
== -NFS4ERR_EXPIRED
)
952 nfs_delegation_reap_unclaimed(clp
);
954 up_write(&clp
->cl_sem
);
956 if (status
== -NFS4ERR_CB_PATH_DOWN
)
957 nfs_handle_cb_pathdown(clp
);
958 nfs4_clear_recover_bit(clp
);
960 module_put_and_exit(0);
963 printk(KERN_WARNING
"Error: state recovery failed on NFSv4 server %u.%u.%u.%u with error %d\n",
964 NIPQUAD(clp
->cl_addr
.sin_addr
), -status
);
965 set_bit(NFS4CLNT_LEASE_EXPIRED
, &clp
->cl_state
);