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 int nfs4_init_clientid(struct nfs_client
*clp
, struct rpc_cred
*cred
)
68 port
= nfs_callback_tcpport
;
69 if (clp
->cl_addr
.ss_family
== AF_INET6
)
70 port
= nfs_callback_tcpport6
;
72 status
= nfs4_proc_setclientid(clp
, NFS4_CALLBACK
, port
, cred
);
74 status
= nfs4_proc_setclientid_confirm(clp
, cred
);
76 nfs4_schedule_state_renewal(clp
);
80 struct rpc_cred
*nfs4_get_machine_cred_locked(struct nfs_client
*clp
)
82 struct rpc_cred
*cred
= NULL
;
84 if (clp
->cl_machine_cred
!= NULL
)
85 cred
= get_rpccred(clp
->cl_machine_cred
);
89 static void nfs4_clear_machine_cred(struct nfs_client
*clp
)
91 struct rpc_cred
*cred
;
93 spin_lock(&clp
->cl_lock
);
94 cred
= clp
->cl_machine_cred
;
95 clp
->cl_machine_cred
= NULL
;
96 spin_unlock(&clp
->cl_lock
);
101 struct rpc_cred
*nfs4_get_renew_cred_locked(struct nfs_client
*clp
)
103 struct nfs4_state_owner
*sp
;
105 struct rpc_cred
*cred
= NULL
;
107 for (pos
= rb_first(&clp
->cl_state_owners
); pos
!= NULL
; pos
= rb_next(pos
)) {
108 sp
= rb_entry(pos
, struct nfs4_state_owner
, so_client_node
);
109 if (list_empty(&sp
->so_states
))
111 cred
= get_rpccred(sp
->so_cred
);
117 #if defined(CONFIG_NFS_V4_1)
119 struct rpc_cred
*nfs4_get_exchange_id_cred(struct nfs_client
*clp
)
121 struct rpc_cred
*cred
;
123 spin_lock(&clp
->cl_lock
);
124 cred
= nfs4_get_machine_cred_locked(clp
);
125 spin_unlock(&clp
->cl_lock
);
129 #endif /* CONFIG_NFS_V4_1 */
131 struct rpc_cred
*nfs4_get_setclientid_cred(struct nfs_client
*clp
)
133 struct nfs4_state_owner
*sp
;
135 struct rpc_cred
*cred
;
137 spin_lock(&clp
->cl_lock
);
138 cred
= nfs4_get_machine_cred_locked(clp
);
141 pos
= rb_first(&clp
->cl_state_owners
);
143 sp
= rb_entry(pos
, struct nfs4_state_owner
, so_client_node
);
144 cred
= get_rpccred(sp
->so_cred
);
147 spin_unlock(&clp
->cl_lock
);
151 static void nfs_alloc_unique_id(struct rb_root
*root
, struct nfs_unique_id
*new,
152 __u64 minval
, int maxbits
)
154 struct rb_node
**p
, *parent
;
155 struct nfs_unique_id
*pos
;
159 mask
= (1ULL << maxbits
) - 1ULL;
161 /* Ensure distribution is more or less flat */
162 get_random_bytes(&new->id
, sizeof(new->id
));
164 if (new->id
< minval
)
172 pos
= rb_entry(parent
, struct nfs_unique_id
, rb_node
);
174 if (new->id
< pos
->id
)
176 else if (new->id
> pos
->id
)
181 rb_link_node(&new->rb_node
, parent
, p
);
182 rb_insert_color(&new->rb_node
, root
);
187 if (new->id
< minval
|| (new->id
& mask
) != new->id
) {
191 parent
= rb_next(parent
);
194 pos
= rb_entry(parent
, struct nfs_unique_id
, rb_node
);
195 if (new->id
< pos
->id
)
201 static void nfs_free_unique_id(struct rb_root
*root
, struct nfs_unique_id
*id
)
203 rb_erase(&id
->rb_node
, root
);
206 static struct nfs4_state_owner
*
207 nfs4_find_state_owner(struct nfs_server
*server
, struct rpc_cred
*cred
)
209 struct nfs_client
*clp
= server
->nfs_client
;
210 struct rb_node
**p
= &clp
->cl_state_owners
.rb_node
,
212 struct nfs4_state_owner
*sp
, *res
= NULL
;
216 sp
= rb_entry(parent
, struct nfs4_state_owner
, so_client_node
);
218 if (server
< sp
->so_server
) {
219 p
= &parent
->rb_left
;
222 if (server
> sp
->so_server
) {
223 p
= &parent
->rb_right
;
226 if (cred
< sp
->so_cred
)
227 p
= &parent
->rb_left
;
228 else if (cred
> sp
->so_cred
)
229 p
= &parent
->rb_right
;
231 atomic_inc(&sp
->so_count
);
239 static struct nfs4_state_owner
*
240 nfs4_insert_state_owner(struct nfs_client
*clp
, struct nfs4_state_owner
*new)
242 struct rb_node
**p
= &clp
->cl_state_owners
.rb_node
,
244 struct nfs4_state_owner
*sp
;
248 sp
= rb_entry(parent
, struct nfs4_state_owner
, so_client_node
);
250 if (new->so_server
< sp
->so_server
) {
251 p
= &parent
->rb_left
;
254 if (new->so_server
> sp
->so_server
) {
255 p
= &parent
->rb_right
;
258 if (new->so_cred
< sp
->so_cred
)
259 p
= &parent
->rb_left
;
260 else if (new->so_cred
> sp
->so_cred
)
261 p
= &parent
->rb_right
;
263 atomic_inc(&sp
->so_count
);
267 nfs_alloc_unique_id(&clp
->cl_openowner_id
, &new->so_owner_id
, 1, 64);
268 rb_link_node(&new->so_client_node
, parent
, p
);
269 rb_insert_color(&new->so_client_node
, &clp
->cl_state_owners
);
274 nfs4_remove_state_owner(struct nfs_client
*clp
, struct nfs4_state_owner
*sp
)
276 if (!RB_EMPTY_NODE(&sp
->so_client_node
))
277 rb_erase(&sp
->so_client_node
, &clp
->cl_state_owners
);
278 nfs_free_unique_id(&clp
->cl_openowner_id
, &sp
->so_owner_id
);
282 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
283 * create a new state_owner.
286 static struct nfs4_state_owner
*
287 nfs4_alloc_state_owner(void)
289 struct nfs4_state_owner
*sp
;
291 sp
= kzalloc(sizeof(*sp
),GFP_KERNEL
);
294 spin_lock_init(&sp
->so_lock
);
295 INIT_LIST_HEAD(&sp
->so_states
);
296 INIT_LIST_HEAD(&sp
->so_delegations
);
297 rpc_init_wait_queue(&sp
->so_sequence
.wait
, "Seqid_waitqueue");
298 sp
->so_seqid
.sequence
= &sp
->so_sequence
;
299 spin_lock_init(&sp
->so_sequence
.lock
);
300 INIT_LIST_HEAD(&sp
->so_sequence
.list
);
301 atomic_set(&sp
->so_count
, 1);
306 nfs4_drop_state_owner(struct nfs4_state_owner
*sp
)
308 if (!RB_EMPTY_NODE(&sp
->so_client_node
)) {
309 struct nfs_client
*clp
= sp
->so_client
;
311 spin_lock(&clp
->cl_lock
);
312 rb_erase(&sp
->so_client_node
, &clp
->cl_state_owners
);
313 RB_CLEAR_NODE(&sp
->so_client_node
);
314 spin_unlock(&clp
->cl_lock
);
318 struct nfs4_state_owner
*nfs4_get_state_owner(struct nfs_server
*server
, struct rpc_cred
*cred
)
320 struct nfs_client
*clp
= server
->nfs_client
;
321 struct nfs4_state_owner
*sp
, *new;
323 spin_lock(&clp
->cl_lock
);
324 sp
= nfs4_find_state_owner(server
, cred
);
325 spin_unlock(&clp
->cl_lock
);
328 new = nfs4_alloc_state_owner();
331 new->so_client
= clp
;
332 new->so_server
= server
;
334 spin_lock(&clp
->cl_lock
);
335 sp
= nfs4_insert_state_owner(clp
, new);
336 spin_unlock(&clp
->cl_lock
);
340 rpc_destroy_wait_queue(&new->so_sequence
.wait
);
346 void nfs4_put_state_owner(struct nfs4_state_owner
*sp
)
348 struct nfs_client
*clp
= sp
->so_client
;
349 struct rpc_cred
*cred
= sp
->so_cred
;
351 if (!atomic_dec_and_lock(&sp
->so_count
, &clp
->cl_lock
))
353 nfs4_remove_state_owner(clp
, sp
);
354 spin_unlock(&clp
->cl_lock
);
355 rpc_destroy_wait_queue(&sp
->so_sequence
.wait
);
360 static struct nfs4_state
*
361 nfs4_alloc_open_state(void)
363 struct nfs4_state
*state
;
365 state
= kzalloc(sizeof(*state
), GFP_KERNEL
);
368 atomic_set(&state
->count
, 1);
369 INIT_LIST_HEAD(&state
->lock_states
);
370 spin_lock_init(&state
->state_lock
);
371 seqlock_init(&state
->seqlock
);
376 nfs4_state_set_mode_locked(struct nfs4_state
*state
, fmode_t fmode
)
378 if (state
->state
== fmode
)
380 /* NB! List reordering - see the reclaim code for why. */
381 if ((fmode
& FMODE_WRITE
) != (state
->state
& FMODE_WRITE
)) {
382 if (fmode
& FMODE_WRITE
)
383 list_move(&state
->open_states
, &state
->owner
->so_states
);
385 list_move_tail(&state
->open_states
, &state
->owner
->so_states
);
387 state
->state
= fmode
;
390 static struct nfs4_state
*
391 __nfs4_find_state_byowner(struct inode
*inode
, struct nfs4_state_owner
*owner
)
393 struct nfs_inode
*nfsi
= NFS_I(inode
);
394 struct nfs4_state
*state
;
396 list_for_each_entry(state
, &nfsi
->open_states
, inode_states
) {
397 if (state
->owner
!= owner
)
399 if (atomic_inc_not_zero(&state
->count
))
406 nfs4_free_open_state(struct nfs4_state
*state
)
412 nfs4_get_open_state(struct inode
*inode
, struct nfs4_state_owner
*owner
)
414 struct nfs4_state
*state
, *new;
415 struct nfs_inode
*nfsi
= NFS_I(inode
);
417 spin_lock(&inode
->i_lock
);
418 state
= __nfs4_find_state_byowner(inode
, owner
);
419 spin_unlock(&inode
->i_lock
);
422 new = nfs4_alloc_open_state();
423 spin_lock(&owner
->so_lock
);
424 spin_lock(&inode
->i_lock
);
425 state
= __nfs4_find_state_byowner(inode
, owner
);
426 if (state
== NULL
&& new != NULL
) {
428 state
->owner
= owner
;
429 atomic_inc(&owner
->so_count
);
430 list_add(&state
->inode_states
, &nfsi
->open_states
);
431 state
->inode
= igrab(inode
);
432 spin_unlock(&inode
->i_lock
);
433 /* Note: The reclaim code dictates that we add stateless
434 * and read-only stateids to the end of the list */
435 list_add_tail(&state
->open_states
, &owner
->so_states
);
436 spin_unlock(&owner
->so_lock
);
438 spin_unlock(&inode
->i_lock
);
439 spin_unlock(&owner
->so_lock
);
441 nfs4_free_open_state(new);
447 void nfs4_put_open_state(struct nfs4_state
*state
)
449 struct inode
*inode
= state
->inode
;
450 struct nfs4_state_owner
*owner
= state
->owner
;
452 if (!atomic_dec_and_lock(&state
->count
, &owner
->so_lock
))
454 spin_lock(&inode
->i_lock
);
455 list_del(&state
->inode_states
);
456 list_del(&state
->open_states
);
457 spin_unlock(&inode
->i_lock
);
458 spin_unlock(&owner
->so_lock
);
460 nfs4_free_open_state(state
);
461 nfs4_put_state_owner(owner
);
465 * Close the current file.
467 static void __nfs4_close(struct path
*path
, struct nfs4_state
*state
, fmode_t fmode
, int wait
)
469 struct nfs4_state_owner
*owner
= state
->owner
;
473 atomic_inc(&owner
->so_count
);
474 /* Protect against nfs4_find_state() */
475 spin_lock(&owner
->so_lock
);
476 switch (fmode
& (FMODE_READ
| FMODE_WRITE
)) {
483 case FMODE_READ
|FMODE_WRITE
:
486 newstate
= FMODE_READ
|FMODE_WRITE
;
487 if (state
->n_rdwr
== 0) {
488 if (state
->n_rdonly
== 0) {
489 newstate
&= ~FMODE_READ
;
490 call_close
|= test_bit(NFS_O_RDONLY_STATE
, &state
->flags
);
491 call_close
|= test_bit(NFS_O_RDWR_STATE
, &state
->flags
);
493 if (state
->n_wronly
== 0) {
494 newstate
&= ~FMODE_WRITE
;
495 call_close
|= test_bit(NFS_O_WRONLY_STATE
, &state
->flags
);
496 call_close
|= test_bit(NFS_O_RDWR_STATE
, &state
->flags
);
499 clear_bit(NFS_DELEGATED_STATE
, &state
->flags
);
501 nfs4_state_set_mode_locked(state
, newstate
);
502 spin_unlock(&owner
->so_lock
);
505 nfs4_put_open_state(state
);
506 nfs4_put_state_owner(owner
);
508 nfs4_do_close(path
, state
, wait
);
511 void nfs4_close_state(struct path
*path
, struct nfs4_state
*state
, fmode_t fmode
)
513 __nfs4_close(path
, state
, fmode
, 0);
516 void nfs4_close_sync(struct path
*path
, struct nfs4_state
*state
, fmode_t fmode
)
518 __nfs4_close(path
, state
, fmode
, 1);
522 * Search the state->lock_states for an existing lock_owner
523 * that is compatible with current->files
525 static struct nfs4_lock_state
*
526 __nfs4_find_lock_state(struct nfs4_state
*state
, fl_owner_t fl_owner
)
528 struct nfs4_lock_state
*pos
;
529 list_for_each_entry(pos
, &state
->lock_states
, ls_locks
) {
530 if (pos
->ls_owner
!= fl_owner
)
532 atomic_inc(&pos
->ls_count
);
539 * Return a compatible lock_state. If no initialized lock_state structure
540 * exists, return an uninitialized one.
543 static struct nfs4_lock_state
*nfs4_alloc_lock_state(struct nfs4_state
*state
, fl_owner_t fl_owner
)
545 struct nfs4_lock_state
*lsp
;
546 struct nfs_client
*clp
= state
->owner
->so_client
;
548 lsp
= kzalloc(sizeof(*lsp
), GFP_KERNEL
);
551 rpc_init_wait_queue(&lsp
->ls_sequence
.wait
, "lock_seqid_waitqueue");
552 spin_lock_init(&lsp
->ls_sequence
.lock
);
553 INIT_LIST_HEAD(&lsp
->ls_sequence
.list
);
554 lsp
->ls_seqid
.sequence
= &lsp
->ls_sequence
;
555 atomic_set(&lsp
->ls_count
, 1);
556 lsp
->ls_state
= state
;
557 lsp
->ls_owner
= fl_owner
;
558 spin_lock(&clp
->cl_lock
);
559 nfs_alloc_unique_id(&clp
->cl_lockowner_id
, &lsp
->ls_id
, 1, 64);
560 spin_unlock(&clp
->cl_lock
);
561 INIT_LIST_HEAD(&lsp
->ls_locks
);
565 static void nfs4_free_lock_state(struct nfs4_lock_state
*lsp
)
567 struct nfs_client
*clp
= lsp
->ls_state
->owner
->so_client
;
569 spin_lock(&clp
->cl_lock
);
570 nfs_free_unique_id(&clp
->cl_lockowner_id
, &lsp
->ls_id
);
571 spin_unlock(&clp
->cl_lock
);
572 rpc_destroy_wait_queue(&lsp
->ls_sequence
.wait
);
577 * Return a compatible lock_state. If no initialized lock_state structure
578 * exists, return an uninitialized one.
581 static struct nfs4_lock_state
*nfs4_get_lock_state(struct nfs4_state
*state
, fl_owner_t owner
)
583 struct nfs4_lock_state
*lsp
, *new = NULL
;
586 spin_lock(&state
->state_lock
);
587 lsp
= __nfs4_find_lock_state(state
, owner
);
591 list_add(&new->ls_locks
, &state
->lock_states
);
592 set_bit(LK_STATE_IN_USE
, &state
->flags
);
597 spin_unlock(&state
->state_lock
);
598 new = nfs4_alloc_lock_state(state
, owner
);
602 spin_unlock(&state
->state_lock
);
604 nfs4_free_lock_state(new);
609 * Release reference to lock_state, and free it if we see that
610 * it is no longer in use
612 void nfs4_put_lock_state(struct nfs4_lock_state
*lsp
)
614 struct nfs4_state
*state
;
618 state
= lsp
->ls_state
;
619 if (!atomic_dec_and_lock(&lsp
->ls_count
, &state
->state_lock
))
621 list_del(&lsp
->ls_locks
);
622 if (list_empty(&state
->lock_states
))
623 clear_bit(LK_STATE_IN_USE
, &state
->flags
);
624 spin_unlock(&state
->state_lock
);
625 nfs4_free_lock_state(lsp
);
628 static void nfs4_fl_copy_lock(struct file_lock
*dst
, struct file_lock
*src
)
630 struct nfs4_lock_state
*lsp
= src
->fl_u
.nfs4_fl
.owner
;
632 dst
->fl_u
.nfs4_fl
.owner
= lsp
;
633 atomic_inc(&lsp
->ls_count
);
636 static void nfs4_fl_release_lock(struct file_lock
*fl
)
638 nfs4_put_lock_state(fl
->fl_u
.nfs4_fl
.owner
);
641 static const struct file_lock_operations nfs4_fl_lock_ops
= {
642 .fl_copy_lock
= nfs4_fl_copy_lock
,
643 .fl_release_private
= nfs4_fl_release_lock
,
646 int nfs4_set_lock_state(struct nfs4_state
*state
, struct file_lock
*fl
)
648 struct nfs4_lock_state
*lsp
;
650 if (fl
->fl_ops
!= NULL
)
652 lsp
= nfs4_get_lock_state(state
, fl
->fl_owner
);
655 fl
->fl_u
.nfs4_fl
.owner
= lsp
;
656 fl
->fl_ops
= &nfs4_fl_lock_ops
;
661 * Byte-range lock aware utility to initialize the stateid of read/write
664 void nfs4_copy_stateid(nfs4_stateid
*dst
, struct nfs4_state
*state
, fl_owner_t fl_owner
)
666 struct nfs4_lock_state
*lsp
;
670 seq
= read_seqbegin(&state
->seqlock
);
671 memcpy(dst
, &state
->stateid
, sizeof(*dst
));
672 } while (read_seqretry(&state
->seqlock
, seq
));
673 if (test_bit(LK_STATE_IN_USE
, &state
->flags
) == 0)
676 spin_lock(&state
->state_lock
);
677 lsp
= __nfs4_find_lock_state(state
, fl_owner
);
678 if (lsp
!= NULL
&& (lsp
->ls_flags
& NFS_LOCK_INITIALIZED
) != 0)
679 memcpy(dst
, &lsp
->ls_stateid
, sizeof(*dst
));
680 spin_unlock(&state
->state_lock
);
681 nfs4_put_lock_state(lsp
);
684 struct nfs_seqid
*nfs_alloc_seqid(struct nfs_seqid_counter
*counter
)
686 struct nfs_seqid
*new;
688 new = kmalloc(sizeof(*new), GFP_KERNEL
);
690 new->sequence
= counter
;
691 INIT_LIST_HEAD(&new->list
);
696 void nfs_free_seqid(struct nfs_seqid
*seqid
)
698 if (!list_empty(&seqid
->list
)) {
699 struct rpc_sequence
*sequence
= seqid
->sequence
->sequence
;
701 spin_lock(&sequence
->lock
);
702 list_del(&seqid
->list
);
703 spin_unlock(&sequence
->lock
);
704 rpc_wake_up(&sequence
->wait
);
710 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
711 * failed with a seqid incrementing error -
712 * see comments nfs_fs.h:seqid_mutating_error()
714 static void nfs_increment_seqid(int status
, struct nfs_seqid
*seqid
)
716 BUG_ON(list_first_entry(&seqid
->sequence
->sequence
->list
, struct nfs_seqid
, list
) != seqid
);
720 case -NFS4ERR_BAD_SEQID
:
721 if (seqid
->sequence
->flags
& NFS_SEQID_CONFIRMED
)
723 printk(KERN_WARNING
"NFS: v4 server returned a bad"
724 " sequence-id error on an"
725 " unconfirmed sequence %p!\n",
727 case -NFS4ERR_STALE_CLIENTID
:
728 case -NFS4ERR_STALE_STATEID
:
729 case -NFS4ERR_BAD_STATEID
:
730 case -NFS4ERR_BADXDR
:
731 case -NFS4ERR_RESOURCE
:
732 case -NFS4ERR_NOFILEHANDLE
:
733 /* Non-seqid mutating errors */
737 * Note: no locking needed as we are guaranteed to be first
738 * on the sequence list
740 seqid
->sequence
->counter
++;
743 void nfs_increment_open_seqid(int status
, struct nfs_seqid
*seqid
)
745 struct nfs4_state_owner
*sp
= container_of(seqid
->sequence
,
746 struct nfs4_state_owner
, so_seqid
);
747 struct nfs_server
*server
= sp
->so_server
;
749 if (status
== -NFS4ERR_BAD_SEQID
)
750 nfs4_drop_state_owner(sp
);
751 if (!nfs4_has_session(server
->nfs_client
))
752 nfs_increment_seqid(status
, seqid
);
756 * Increment the seqid if the LOCK/LOCKU succeeded, or
757 * failed with a seqid incrementing error -
758 * see comments nfs_fs.h:seqid_mutating_error()
760 void nfs_increment_lock_seqid(int status
, struct nfs_seqid
*seqid
)
762 nfs_increment_seqid(status
, seqid
);
765 int nfs_wait_on_sequence(struct nfs_seqid
*seqid
, struct rpc_task
*task
)
767 struct rpc_sequence
*sequence
= seqid
->sequence
->sequence
;
770 spin_lock(&sequence
->lock
);
771 if (list_empty(&seqid
->list
))
772 list_add_tail(&seqid
->list
, &sequence
->list
);
773 if (list_first_entry(&sequence
->list
, struct nfs_seqid
, list
) == seqid
)
775 rpc_sleep_on(&sequence
->wait
, task
, NULL
);
778 spin_unlock(&sequence
->lock
);
782 static int nfs4_run_state_manager(void *);
784 static void nfs4_clear_state_manager_bit(struct nfs_client
*clp
)
786 smp_mb__before_clear_bit();
787 clear_bit(NFS4CLNT_MANAGER_RUNNING
, &clp
->cl_state
);
788 smp_mb__after_clear_bit();
789 wake_up_bit(&clp
->cl_state
, NFS4CLNT_MANAGER_RUNNING
);
790 rpc_wake_up(&clp
->cl_rpcwaitq
);
794 * Schedule the nfs_client asynchronous state management routine
796 void nfs4_schedule_state_manager(struct nfs_client
*clp
)
798 struct task_struct
*task
;
800 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING
, &clp
->cl_state
) != 0)
802 __module_get(THIS_MODULE
);
803 atomic_inc(&clp
->cl_count
);
804 task
= kthread_run(nfs4_run_state_manager
, clp
, "%s-manager",
805 rpc_peeraddr2str(clp
->cl_rpcclient
,
809 nfs4_clear_state_manager_bit(clp
);
811 module_put(THIS_MODULE
);
815 * Schedule a state recovery attempt
817 void nfs4_schedule_state_recovery(struct nfs_client
*clp
)
821 if (!test_bit(NFS4CLNT_LEASE_EXPIRED
, &clp
->cl_state
))
822 set_bit(NFS4CLNT_CHECK_LEASE
, &clp
->cl_state
);
823 nfs4_schedule_state_manager(clp
);
826 static int nfs4_state_mark_reclaim_reboot(struct nfs_client
*clp
, struct nfs4_state
*state
)
829 set_bit(NFS_STATE_RECLAIM_REBOOT
, &state
->flags
);
830 /* Don't recover state that expired before the reboot */
831 if (test_bit(NFS_STATE_RECLAIM_NOGRACE
, &state
->flags
)) {
832 clear_bit(NFS_STATE_RECLAIM_REBOOT
, &state
->flags
);
835 set_bit(NFS_OWNER_RECLAIM_REBOOT
, &state
->owner
->so_flags
);
836 set_bit(NFS4CLNT_RECLAIM_REBOOT
, &clp
->cl_state
);
840 int nfs4_state_mark_reclaim_nograce(struct nfs_client
*clp
, struct nfs4_state
*state
)
842 set_bit(NFS_STATE_RECLAIM_NOGRACE
, &state
->flags
);
843 clear_bit(NFS_STATE_RECLAIM_REBOOT
, &state
->flags
);
844 set_bit(NFS_OWNER_RECLAIM_NOGRACE
, &state
->owner
->so_flags
);
845 set_bit(NFS4CLNT_RECLAIM_NOGRACE
, &clp
->cl_state
);
849 static int nfs4_reclaim_locks(struct nfs4_state
*state
, const struct nfs4_state_recovery_ops
*ops
)
851 struct inode
*inode
= state
->inode
;
852 struct nfs_inode
*nfsi
= NFS_I(inode
);
853 struct file_lock
*fl
;
856 if (inode
->i_flock
== NULL
)
859 /* Guard against delegation returns and new lock/unlock calls */
860 down_write(&nfsi
->rwsem
);
861 /* Protect inode->i_flock using the BKL */
863 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
864 if (!(fl
->fl_flags
& (FL_POSIX
|FL_FLOCK
)))
866 if (nfs_file_open_context(fl
->fl_file
)->state
!= state
)
869 status
= ops
->recover_lock(state
, fl
);
874 case -NFS4ERR_ADMIN_REVOKED
:
875 case -NFS4ERR_STALE_STATEID
:
876 case -NFS4ERR_BAD_STATEID
:
877 case -NFS4ERR_EXPIRED
:
878 case -NFS4ERR_NO_GRACE
:
879 case -NFS4ERR_STALE_CLIENTID
:
880 case -NFS4ERR_BADSESSION
:
881 case -NFS4ERR_BADSLOT
:
882 case -NFS4ERR_BAD_HIGH_SLOT
:
883 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION
:
886 printk(KERN_ERR
"%s: unhandled error %d. Zeroing state\n",
889 case -NFS4ERR_DENIED
:
890 case -NFS4ERR_RECLAIM_BAD
:
891 case -NFS4ERR_RECLAIM_CONFLICT
:
892 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
899 up_write(&nfsi
->rwsem
);
903 static int nfs4_reclaim_open_state(struct nfs4_state_owner
*sp
, const struct nfs4_state_recovery_ops
*ops
)
905 struct nfs4_state
*state
;
906 struct nfs4_lock_state
*lock
;
909 /* Note: we rely on the sp->so_states list being ordered
910 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
912 * This is needed to ensure that the server won't give us any
913 * read delegations that we have to return if, say, we are
914 * recovering after a network partition or a reboot from a
915 * server that doesn't support a grace period.
918 spin_lock(&sp
->so_lock
);
919 list_for_each_entry(state
, &sp
->so_states
, open_states
) {
920 if (!test_and_clear_bit(ops
->state_flag_bit
, &state
->flags
))
922 if (state
->state
== 0)
924 atomic_inc(&state
->count
);
925 spin_unlock(&sp
->so_lock
);
926 status
= ops
->recover_open(sp
, state
);
928 status
= nfs4_reclaim_locks(state
, ops
);
930 list_for_each_entry(lock
, &state
->lock_states
, ls_locks
) {
931 if (!(lock
->ls_flags
& NFS_LOCK_INITIALIZED
))
932 printk("%s: Lock reclaim failed!\n",
935 nfs4_put_open_state(state
);
941 printk(KERN_ERR
"%s: unhandled error %d. Zeroing state\n",
947 * Open state on this file cannot be recovered
948 * All we can do is revert to using the zero stateid.
950 memset(state
->stateid
.data
, 0,
951 sizeof(state
->stateid
.data
));
952 /* Mark the file as being 'closed' */
955 case -NFS4ERR_ADMIN_REVOKED
:
956 case -NFS4ERR_STALE_STATEID
:
957 case -NFS4ERR_BAD_STATEID
:
958 case -NFS4ERR_RECLAIM_BAD
:
959 case -NFS4ERR_RECLAIM_CONFLICT
:
960 nfs4_state_mark_reclaim_nograce(sp
->so_client
, state
);
962 case -NFS4ERR_EXPIRED
:
963 case -NFS4ERR_NO_GRACE
:
964 nfs4_state_mark_reclaim_nograce(sp
->so_client
, state
);
965 case -NFS4ERR_STALE_CLIENTID
:
966 case -NFS4ERR_BADSESSION
:
967 case -NFS4ERR_BADSLOT
:
968 case -NFS4ERR_BAD_HIGH_SLOT
:
969 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION
:
972 nfs4_put_open_state(state
);
975 spin_unlock(&sp
->so_lock
);
978 nfs4_put_open_state(state
);
982 static void nfs4_clear_open_state(struct nfs4_state
*state
)
984 struct nfs4_lock_state
*lock
;
986 clear_bit(NFS_DELEGATED_STATE
, &state
->flags
);
987 clear_bit(NFS_O_RDONLY_STATE
, &state
->flags
);
988 clear_bit(NFS_O_WRONLY_STATE
, &state
->flags
);
989 clear_bit(NFS_O_RDWR_STATE
, &state
->flags
);
990 list_for_each_entry(lock
, &state
->lock_states
, ls_locks
) {
991 lock
->ls_seqid
.flags
= 0;
992 lock
->ls_flags
&= ~NFS_LOCK_INITIALIZED
;
996 static void nfs4_state_mark_reclaim_helper(struct nfs_client
*clp
, int (*mark_reclaim
)(struct nfs_client
*clp
, struct nfs4_state
*state
))
998 struct nfs4_state_owner
*sp
;
1000 struct nfs4_state
*state
;
1002 /* Reset all sequence ids to zero */
1003 for (pos
= rb_first(&clp
->cl_state_owners
); pos
!= NULL
; pos
= rb_next(pos
)) {
1004 sp
= rb_entry(pos
, struct nfs4_state_owner
, so_client_node
);
1005 sp
->so_seqid
.flags
= 0;
1006 spin_lock(&sp
->so_lock
);
1007 list_for_each_entry(state
, &sp
->so_states
, open_states
) {
1008 if (mark_reclaim(clp
, state
))
1009 nfs4_clear_open_state(state
);
1011 spin_unlock(&sp
->so_lock
);
1015 static void nfs4_state_start_reclaim_reboot(struct nfs_client
*clp
)
1017 /* Mark all delegations for reclaim */
1018 nfs_delegation_mark_reclaim(clp
);
1019 nfs4_state_mark_reclaim_helper(clp
, nfs4_state_mark_reclaim_reboot
);
1022 static void nfs4_state_end_reclaim_reboot(struct nfs_client
*clp
)
1024 struct nfs4_state_owner
*sp
;
1025 struct rb_node
*pos
;
1026 struct nfs4_state
*state
;
1028 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT
, &clp
->cl_state
))
1031 for (pos
= rb_first(&clp
->cl_state_owners
); pos
!= NULL
; pos
= rb_next(pos
)) {
1032 sp
= rb_entry(pos
, struct nfs4_state_owner
, so_client_node
);
1033 spin_lock(&sp
->so_lock
);
1034 list_for_each_entry(state
, &sp
->so_states
, open_states
) {
1035 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT
, &state
->flags
))
1037 nfs4_state_mark_reclaim_nograce(clp
, state
);
1039 spin_unlock(&sp
->so_lock
);
1042 nfs_delegation_reap_unclaimed(clp
);
1045 static void nfs_delegation_clear_all(struct nfs_client
*clp
)
1047 nfs_delegation_mark_reclaim(clp
);
1048 nfs_delegation_reap_unclaimed(clp
);
1051 static void nfs4_state_start_reclaim_nograce(struct nfs_client
*clp
)
1053 nfs_delegation_clear_all(clp
);
1054 nfs4_state_mark_reclaim_helper(clp
, nfs4_state_mark_reclaim_nograce
);
1057 static int nfs4_recovery_handle_error(struct nfs_client
*clp
, int error
)
1060 case -NFS4ERR_CB_PATH_DOWN
:
1061 nfs_handle_cb_pathdown(clp
);
1063 case -NFS4ERR_NO_GRACE
:
1064 nfs4_state_end_reclaim_reboot(clp
);
1066 case -NFS4ERR_STALE_CLIENTID
:
1067 case -NFS4ERR_LEASE_MOVED
:
1068 set_bit(NFS4CLNT_LEASE_EXPIRED
, &clp
->cl_state
);
1069 nfs4_state_end_reclaim_reboot(clp
);
1070 nfs4_state_start_reclaim_reboot(clp
);
1072 case -NFS4ERR_EXPIRED
:
1073 set_bit(NFS4CLNT_LEASE_EXPIRED
, &clp
->cl_state
);
1074 nfs4_state_start_reclaim_nograce(clp
);
1075 case -NFS4ERR_BADSESSION
:
1076 case -NFS4ERR_BADSLOT
:
1077 case -NFS4ERR_BAD_HIGH_SLOT
:
1078 case -NFS4ERR_DEADSESSION
:
1079 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION
:
1080 case -NFS4ERR_SEQ_FALSE_RETRY
:
1081 case -NFS4ERR_SEQ_MISORDERED
:
1082 set_bit(NFS4CLNT_SESSION_SETUP
, &clp
->cl_state
);
1087 static int nfs4_do_reclaim(struct nfs_client
*clp
, const struct nfs4_state_recovery_ops
*ops
)
1089 struct rb_node
*pos
;
1093 spin_lock(&clp
->cl_lock
);
1094 for (pos
= rb_first(&clp
->cl_state_owners
); pos
!= NULL
; pos
= rb_next(pos
)) {
1095 struct nfs4_state_owner
*sp
= rb_entry(pos
, struct nfs4_state_owner
, so_client_node
);
1096 if (!test_and_clear_bit(ops
->owner_flag_bit
, &sp
->so_flags
))
1098 atomic_inc(&sp
->so_count
);
1099 spin_unlock(&clp
->cl_lock
);
1100 status
= nfs4_reclaim_open_state(sp
, ops
);
1102 set_bit(ops
->owner_flag_bit
, &sp
->so_flags
);
1103 nfs4_put_state_owner(sp
);
1104 return nfs4_recovery_handle_error(clp
, status
);
1106 nfs4_put_state_owner(sp
);
1109 spin_unlock(&clp
->cl_lock
);
1113 static int nfs4_check_lease(struct nfs_client
*clp
)
1115 struct rpc_cred
*cred
;
1116 struct nfs4_state_maintenance_ops
*ops
=
1117 nfs4_state_renewal_ops
[clp
->cl_minorversion
];
1118 int status
= -NFS4ERR_EXPIRED
;
1120 /* Is the client already known to have an expired lease? */
1121 if (test_bit(NFS4CLNT_LEASE_EXPIRED
, &clp
->cl_state
))
1123 spin_lock(&clp
->cl_lock
);
1124 cred
= ops
->get_state_renewal_cred_locked(clp
);
1125 spin_unlock(&clp
->cl_lock
);
1127 cred
= nfs4_get_setclientid_cred(clp
);
1131 status
= ops
->renew_lease(clp
, cred
);
1134 return nfs4_recovery_handle_error(clp
, status
);
1137 static int nfs4_reclaim_lease(struct nfs_client
*clp
)
1139 struct rpc_cred
*cred
;
1140 struct nfs4_state_recovery_ops
*ops
=
1141 nfs4_reboot_recovery_ops
[clp
->cl_minorversion
];
1142 int status
= -ENOENT
;
1144 cred
= ops
->get_clid_cred(clp
);
1146 status
= ops
->establish_clid(clp
, cred
);
1148 /* Handle case where the user hasn't set up machine creds */
1149 if (status
== -EACCES
&& cred
== clp
->cl_machine_cred
) {
1150 nfs4_clear_machine_cred(clp
);
1153 if (status
== -NFS4ERR_MINOR_VERS_MISMATCH
)
1154 status
= -EPROTONOSUPPORT
;
1159 #ifdef CONFIG_NFS_V4_1
1160 static void nfs4_session_recovery_handle_error(struct nfs_client
*clp
, int err
)
1163 case -NFS4ERR_STALE_CLIENTID
:
1164 set_bit(NFS4CLNT_LEASE_EXPIRED
, &clp
->cl_state
);
1165 set_bit(NFS4CLNT_SESSION_SETUP
, &clp
->cl_state
);
1169 static int nfs4_reset_session(struct nfs_client
*clp
)
1173 status
= nfs4_proc_destroy_session(clp
->cl_session
);
1174 if (status
&& status
!= -NFS4ERR_BADSESSION
&&
1175 status
!= -NFS4ERR_DEADSESSION
) {
1176 nfs4_session_recovery_handle_error(clp
, status
);
1180 memset(clp
->cl_session
->sess_id
.data
, 0, NFS4_MAX_SESSIONID_LEN
);
1181 status
= nfs4_proc_create_session(clp
, 1);
1183 nfs4_session_recovery_handle_error(clp
, status
);
1186 /* Wake up the next rpc task even on error */
1187 rpc_wake_up_next(&clp
->cl_session
->fc_slot_table
.slot_tbl_waitq
);
1191 static int nfs4_initialize_session(struct nfs_client
*clp
)
1195 status
= nfs4_proc_create_session(clp
, 0);
1197 nfs_mark_client_ready(clp
, NFS_CS_READY
);
1198 } else if (status
== -NFS4ERR_STALE_CLIENTID
) {
1199 set_bit(NFS4CLNT_LEASE_EXPIRED
, &clp
->cl_state
);
1200 set_bit(NFS4CLNT_SESSION_SETUP
, &clp
->cl_state
);
1202 nfs_mark_client_ready(clp
, status
);
1206 #else /* CONFIG_NFS_V4_1 */
1207 static int nfs4_reset_session(struct nfs_client
*clp
) { return 0; }
1208 static int nfs4_initialize_session(struct nfs_client
*clp
) { return 0; }
1209 #endif /* CONFIG_NFS_V4_1 */
1211 /* Set NFS4CLNT_LEASE_EXPIRED for all v4.0 errors and for recoverable errors
1212 * on EXCHANGE_ID for v4.1
1214 static void nfs4_set_lease_expired(struct nfs_client
*clp
, int status
)
1216 if (nfs4_has_session(clp
)) {
1218 case -NFS4ERR_DELAY
:
1219 case -NFS4ERR_CLID_INUSE
:
1223 case -NFS4ERR_NOT_SAME
: /* FixMe: implement recovery
1224 * in nfs4_exchange_id */
1229 set_bit(NFS4CLNT_LEASE_EXPIRED
, &clp
->cl_state
);
1232 static void nfs4_state_manager(struct nfs_client
*clp
)
1236 /* Ensure exclusive access to NFSv4 state */
1238 if (test_and_clear_bit(NFS4CLNT_LEASE_EXPIRED
, &clp
->cl_state
)) {
1239 /* We're going to have to re-establish a clientid */
1240 status
= nfs4_reclaim_lease(clp
);
1242 nfs4_set_lease_expired(clp
, status
);
1243 if (test_bit(NFS4CLNT_LEASE_EXPIRED
,
1246 if (clp
->cl_cons_state
==
1247 NFS_CS_SESSION_INITING
)
1248 nfs_mark_client_ready(clp
, status
);
1251 clear_bit(NFS4CLNT_CHECK_LEASE
, &clp
->cl_state
);
1254 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE
, &clp
->cl_state
)) {
1255 status
= nfs4_check_lease(clp
);
1256 if (test_bit(NFS4CLNT_LEASE_EXPIRED
, &clp
->cl_state
))
1258 if (status
< 0 && status
!= -NFS4ERR_CB_PATH_DOWN
)
1262 /* Initialize or reset the session */
1263 if (test_and_clear_bit(NFS4CLNT_SESSION_SETUP
, &clp
->cl_state
)
1264 && nfs4_has_session(clp
)) {
1265 if (clp
->cl_cons_state
== NFS_CS_SESSION_INITING
)
1266 status
= nfs4_initialize_session(clp
);
1268 status
= nfs4_reset_session(clp
);
1269 if (test_bit(NFS4CLNT_LEASE_EXPIRED
, &clp
->cl_state
))
1275 /* First recover reboot state... */
1276 if (test_bit(NFS4CLNT_RECLAIM_REBOOT
, &clp
->cl_state
)) {
1277 status
= nfs4_do_reclaim(clp
,
1278 nfs4_reboot_recovery_ops
[clp
->cl_minorversion
]);
1279 if (test_bit(NFS4CLNT_LEASE_EXPIRED
, &clp
->cl_state
) ||
1280 test_bit(NFS4CLNT_SESSION_SETUP
, &clp
->cl_state
))
1282 nfs4_state_end_reclaim_reboot(clp
);
1283 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE
, &clp
->cl_state
))
1289 /* Now recover expired state... */
1290 if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE
, &clp
->cl_state
)) {
1291 status
= nfs4_do_reclaim(clp
,
1292 nfs4_nograce_recovery_ops
[clp
->cl_minorversion
]);
1293 if (test_bit(NFS4CLNT_LEASE_EXPIRED
, &clp
->cl_state
) ||
1294 test_bit(NFS4CLNT_SESSION_SETUP
, &clp
->cl_state
) ||
1295 test_bit(NFS4CLNT_RECLAIM_REBOOT
, &clp
->cl_state
))
1301 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN
, &clp
->cl_state
)) {
1302 nfs_client_return_marked_delegations(clp
);
1306 nfs4_clear_state_manager_bit(clp
);
1307 /* Did we race with an attempt to give us more work? */
1308 if (clp
->cl_state
== 0)
1310 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING
, &clp
->cl_state
) != 0)
1315 printk(KERN_WARNING
"Error: state manager failed on NFSv4 server %s"
1316 " with error %d\n", clp
->cl_hostname
, -status
);
1317 nfs4_clear_state_manager_bit(clp
);
1320 static int nfs4_run_state_manager(void *ptr
)
1322 struct nfs_client
*clp
= ptr
;
1324 allow_signal(SIGKILL
);
1325 nfs4_state_manager(clp
);
1326 nfs_put_client(clp
);
1327 module_put_and_exit(0);