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 static int nfs41_setup_state_renewal(struct nfs_client *clp)
122 struct nfs_fsinfo fsinfo;
124 status = nfs4_proc_get_lease_time(clp, &fsinfo);
126 /* Update lease time and schedule renewal */
127 spin_lock(&clp->cl_lock);
128 clp->cl_lease_time = fsinfo.lease_time * HZ;
129 clp->cl_last_renewal = jiffies;
130 spin_unlock(&clp->cl_lock);
132 nfs4_schedule_state_renewal(clp);
138 int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
142 status = nfs4_proc_exchange_id(clp, cred);
145 status = nfs4_proc_create_session(clp);
148 nfs41_setup_state_renewal(clp);
149 nfs_mark_client_ready(clp, NFS_CS_READY);
154 struct rpc_cred *nfs4_get_exchange_id_cred(struct nfs_client *clp)
156 struct rpc_cred *cred;
158 spin_lock(&clp->cl_lock);
159 cred = nfs4_get_machine_cred_locked(clp);
160 spin_unlock(&clp->cl_lock);
164 #endif /* CONFIG_NFS_V4_1 */
166 struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
168 struct nfs4_state_owner *sp;
170 struct rpc_cred *cred;
172 spin_lock(&clp->cl_lock);
173 cred = nfs4_get_machine_cred_locked(clp);
176 pos = rb_first(&clp->cl_state_owners);
178 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
179 cred = get_rpccred(sp->so_cred);
182 spin_unlock(&clp->cl_lock);
186 static void nfs_alloc_unique_id(struct rb_root *root, struct nfs_unique_id *new,
187 __u64 minval, int maxbits)
189 struct rb_node **p, *parent;
190 struct nfs_unique_id *pos;
194 mask = (1ULL << maxbits) - 1ULL;
196 /* Ensure distribution is more or less flat */
197 get_random_bytes(&new->id, sizeof(new->id));
199 if (new->id < minval)
207 pos = rb_entry(parent, struct nfs_unique_id, rb_node);
209 if (new->id < pos->id)
211 else if (new->id > pos->id)
216 rb_link_node(&new->rb_node, parent, p);
217 rb_insert_color(&new->rb_node, root);
222 if (new->id < minval || (new->id & mask) != new->id) {
226 parent = rb_next(parent);
229 pos = rb_entry(parent, struct nfs_unique_id, rb_node);
230 if (new->id < pos->id)
236 static void nfs_free_unique_id(struct rb_root *root, struct nfs_unique_id *id)
238 rb_erase(&id->rb_node, root);
241 static struct nfs4_state_owner *
242 nfs4_find_state_owner(struct nfs_server *server, struct rpc_cred *cred)
244 struct nfs_client *clp = server->nfs_client;
245 struct rb_node **p = &clp->cl_state_owners.rb_node,
247 struct nfs4_state_owner *sp, *res = NULL;
251 sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
253 if (server < sp->so_server) {
254 p = &parent->rb_left;
257 if (server > sp->so_server) {
258 p = &parent->rb_right;
261 if (cred < sp->so_cred)
262 p = &parent->rb_left;
263 else if (cred > sp->so_cred)
264 p = &parent->rb_right;
266 atomic_inc(&sp->so_count);
274 static struct nfs4_state_owner *
275 nfs4_insert_state_owner(struct nfs_client *clp, struct nfs4_state_owner *new)
277 struct rb_node **p = &clp->cl_state_owners.rb_node,
279 struct nfs4_state_owner *sp;
283 sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
285 if (new->so_server < sp->so_server) {
286 p = &parent->rb_left;
289 if (new->so_server > sp->so_server) {
290 p = &parent->rb_right;
293 if (new->so_cred < sp->so_cred)
294 p = &parent->rb_left;
295 else if (new->so_cred > sp->so_cred)
296 p = &parent->rb_right;
298 atomic_inc(&sp->so_count);
302 nfs_alloc_unique_id(&clp->cl_openowner_id, &new->so_owner_id, 1, 64);
303 rb_link_node(&new->so_client_node, parent, p);
304 rb_insert_color(&new->so_client_node, &clp->cl_state_owners);
309 nfs4_remove_state_owner(struct nfs_client *clp, struct nfs4_state_owner *sp)
311 if (!RB_EMPTY_NODE(&sp->so_client_node))
312 rb_erase(&sp->so_client_node, &clp->cl_state_owners);
313 nfs_free_unique_id(&clp->cl_openowner_id, &sp->so_owner_id);
317 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
318 * create a new state_owner.
321 static struct nfs4_state_owner *
322 nfs4_alloc_state_owner(void)
324 struct nfs4_state_owner *sp;
326 sp = kzalloc(sizeof(*sp),GFP_KERNEL);
329 spin_lock_init(&sp->so_lock);
330 INIT_LIST_HEAD(&sp->so_states);
331 INIT_LIST_HEAD(&sp->so_delegations);
332 rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue");
333 sp->so_seqid.sequence = &sp->so_sequence;
334 spin_lock_init(&sp->so_sequence.lock);
335 INIT_LIST_HEAD(&sp->so_sequence.list);
336 atomic_set(&sp->so_count, 1);
341 nfs4_drop_state_owner(struct nfs4_state_owner *sp)
343 if (!RB_EMPTY_NODE(&sp->so_client_node)) {
344 struct nfs_client *clp = sp->so_client;
346 spin_lock(&clp->cl_lock);
347 rb_erase(&sp->so_client_node, &clp->cl_state_owners);
348 RB_CLEAR_NODE(&sp->so_client_node);
349 spin_unlock(&clp->cl_lock);
353 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct rpc_cred *cred)
355 struct nfs_client *clp = server->nfs_client;
356 struct nfs4_state_owner *sp, *new;
358 spin_lock(&clp->cl_lock);
359 sp = nfs4_find_state_owner(server, cred);
360 spin_unlock(&clp->cl_lock);
363 new = nfs4_alloc_state_owner();
366 new->so_client = clp;
367 new->so_server = server;
369 spin_lock(&clp->cl_lock);
370 sp = nfs4_insert_state_owner(clp, new);
371 spin_unlock(&clp->cl_lock);
375 rpc_destroy_wait_queue(&new->so_sequence.wait);
381 void nfs4_put_state_owner(struct nfs4_state_owner *sp)
383 struct nfs_client *clp = sp->so_client;
384 struct rpc_cred *cred = sp->so_cred;
386 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
388 nfs4_remove_state_owner(clp, sp);
389 spin_unlock(&clp->cl_lock);
390 rpc_destroy_wait_queue(&sp->so_sequence.wait);
395 static struct nfs4_state *
396 nfs4_alloc_open_state(void)
398 struct nfs4_state *state;
400 state = kzalloc(sizeof(*state), GFP_KERNEL);
403 atomic_set(&state->count, 1);
404 INIT_LIST_HEAD(&state->lock_states);
405 spin_lock_init(&state->state_lock);
406 seqlock_init(&state->seqlock);
411 nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
413 if (state->state == fmode)
415 /* NB! List reordering - see the reclaim code for why. */
416 if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
417 if (fmode & FMODE_WRITE)
418 list_move(&state->open_states, &state->owner->so_states);
420 list_move_tail(&state->open_states, &state->owner->so_states);
422 state->state = fmode;
425 static struct nfs4_state *
426 __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
428 struct nfs_inode *nfsi = NFS_I(inode);
429 struct nfs4_state *state;
431 list_for_each_entry(state, &nfsi->open_states, inode_states) {
432 if (state->owner != owner)
434 if (atomic_inc_not_zero(&state->count))
441 nfs4_free_open_state(struct nfs4_state *state)
447 nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
449 struct nfs4_state *state, *new;
450 struct nfs_inode *nfsi = NFS_I(inode);
452 spin_lock(&inode->i_lock);
453 state = __nfs4_find_state_byowner(inode, owner);
454 spin_unlock(&inode->i_lock);
457 new = nfs4_alloc_open_state();
458 spin_lock(&owner->so_lock);
459 spin_lock(&inode->i_lock);
460 state = __nfs4_find_state_byowner(inode, owner);
461 if (state == NULL && new != NULL) {
463 state->owner = owner;
464 atomic_inc(&owner->so_count);
465 list_add(&state->inode_states, &nfsi->open_states);
466 state->inode = igrab(inode);
467 spin_unlock(&inode->i_lock);
468 /* Note: The reclaim code dictates that we add stateless
469 * and read-only stateids to the end of the list */
470 list_add_tail(&state->open_states, &owner->so_states);
471 spin_unlock(&owner->so_lock);
473 spin_unlock(&inode->i_lock);
474 spin_unlock(&owner->so_lock);
476 nfs4_free_open_state(new);
482 void nfs4_put_open_state(struct nfs4_state *state)
484 struct inode *inode = state->inode;
485 struct nfs4_state_owner *owner = state->owner;
487 if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
489 spin_lock(&inode->i_lock);
490 list_del(&state->inode_states);
491 list_del(&state->open_states);
492 spin_unlock(&inode->i_lock);
493 spin_unlock(&owner->so_lock);
495 nfs4_free_open_state(state);
496 nfs4_put_state_owner(owner);
500 * Close the current file.
502 static void __nfs4_close(struct path *path, struct nfs4_state *state, fmode_t fmode, int wait)
504 struct nfs4_state_owner *owner = state->owner;
508 atomic_inc(&owner->so_count);
509 /* Protect against nfs4_find_state() */
510 spin_lock(&owner->so_lock);
511 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
518 case FMODE_READ|FMODE_WRITE:
521 newstate = FMODE_READ|FMODE_WRITE;
522 if (state->n_rdwr == 0) {
523 if (state->n_rdonly == 0) {
524 newstate &= ~FMODE_READ;
525 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
526 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
528 if (state->n_wronly == 0) {
529 newstate &= ~FMODE_WRITE;
530 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
531 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
534 clear_bit(NFS_DELEGATED_STATE, &state->flags);
536 nfs4_state_set_mode_locked(state, newstate);
537 spin_unlock(&owner->so_lock);
540 nfs4_put_open_state(state);
541 nfs4_put_state_owner(owner);
543 nfs4_do_close(path, state, wait);
546 void nfs4_close_state(struct path *path, struct nfs4_state *state, fmode_t fmode)
548 __nfs4_close(path, state, fmode, 0);
551 void nfs4_close_sync(struct path *path, struct nfs4_state *state, fmode_t fmode)
553 __nfs4_close(path, state, fmode, 1);
557 * Search the state->lock_states for an existing lock_owner
558 * that is compatible with current->files
560 static struct nfs4_lock_state *
561 __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
563 struct nfs4_lock_state *pos;
564 list_for_each_entry(pos, &state->lock_states, ls_locks) {
565 if (pos->ls_owner != fl_owner)
567 atomic_inc(&pos->ls_count);
574 * Return a compatible lock_state. If no initialized lock_state structure
575 * exists, return an uninitialized one.
578 static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
580 struct nfs4_lock_state *lsp;
581 struct nfs_client *clp = state->owner->so_client;
583 lsp = kzalloc(sizeof(*lsp), GFP_KERNEL);
586 rpc_init_wait_queue(&lsp->ls_sequence.wait, "lock_seqid_waitqueue");
587 spin_lock_init(&lsp->ls_sequence.lock);
588 INIT_LIST_HEAD(&lsp->ls_sequence.list);
589 lsp->ls_seqid.sequence = &lsp->ls_sequence;
590 atomic_set(&lsp->ls_count, 1);
591 lsp->ls_state = state;
592 lsp->ls_owner = fl_owner;
593 spin_lock(&clp->cl_lock);
594 nfs_alloc_unique_id(&clp->cl_lockowner_id, &lsp->ls_id, 1, 64);
595 spin_unlock(&clp->cl_lock);
596 INIT_LIST_HEAD(&lsp->ls_locks);
600 static void nfs4_free_lock_state(struct nfs4_lock_state *lsp)
602 struct nfs_client *clp = lsp->ls_state->owner->so_client;
604 spin_lock(&clp->cl_lock);
605 nfs_free_unique_id(&clp->cl_lockowner_id, &lsp->ls_id);
606 spin_unlock(&clp->cl_lock);
607 rpc_destroy_wait_queue(&lsp->ls_sequence.wait);
612 * Return a compatible lock_state. If no initialized lock_state structure
613 * exists, return an uninitialized one.
616 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
618 struct nfs4_lock_state *lsp, *new = NULL;
621 spin_lock(&state->state_lock);
622 lsp = __nfs4_find_lock_state(state, owner);
626 list_add(&new->ls_locks, &state->lock_states);
627 set_bit(LK_STATE_IN_USE, &state->flags);
632 spin_unlock(&state->state_lock);
633 new = nfs4_alloc_lock_state(state, owner);
637 spin_unlock(&state->state_lock);
639 nfs4_free_lock_state(new);
644 * Release reference to lock_state, and free it if we see that
645 * it is no longer in use
647 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
649 struct nfs4_state *state;
653 state = lsp->ls_state;
654 if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
656 list_del(&lsp->ls_locks);
657 if (list_empty(&state->lock_states))
658 clear_bit(LK_STATE_IN_USE, &state->flags);
659 spin_unlock(&state->state_lock);
660 nfs4_free_lock_state(lsp);
663 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
665 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
667 dst->fl_u.nfs4_fl.owner = lsp;
668 atomic_inc(&lsp->ls_count);
671 static void nfs4_fl_release_lock(struct file_lock *fl)
673 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
676 static const struct file_lock_operations nfs4_fl_lock_ops = {
677 .fl_copy_lock = nfs4_fl_copy_lock,
678 .fl_release_private = nfs4_fl_release_lock,
681 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
683 struct nfs4_lock_state *lsp;
685 if (fl->fl_ops != NULL)
687 lsp = nfs4_get_lock_state(state, fl->fl_owner);
690 fl->fl_u.nfs4_fl.owner = lsp;
691 fl->fl_ops = &nfs4_fl_lock_ops;
696 * Byte-range lock aware utility to initialize the stateid of read/write
699 void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner)
701 struct nfs4_lock_state *lsp;
705 seq = read_seqbegin(&state->seqlock);
706 memcpy(dst, &state->stateid, sizeof(*dst));
707 } while (read_seqretry(&state->seqlock, seq));
708 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
711 spin_lock(&state->state_lock);
712 lsp = __nfs4_find_lock_state(state, fl_owner);
713 if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
714 memcpy(dst, &lsp->ls_stateid, sizeof(*dst));
715 spin_unlock(&state->state_lock);
716 nfs4_put_lock_state(lsp);
719 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter)
721 struct nfs_seqid *new;
723 new = kmalloc(sizeof(*new), GFP_KERNEL);
725 new->sequence = counter;
726 INIT_LIST_HEAD(&new->list);
731 void nfs_free_seqid(struct nfs_seqid *seqid)
733 if (!list_empty(&seqid->list)) {
734 struct rpc_sequence *sequence = seqid->sequence->sequence;
736 spin_lock(&sequence->lock);
737 list_del(&seqid->list);
738 spin_unlock(&sequence->lock);
739 rpc_wake_up(&sequence->wait);
745 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
746 * failed with a seqid incrementing error -
747 * see comments nfs_fs.h:seqid_mutating_error()
749 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
751 BUG_ON(list_first_entry(&seqid->sequence->sequence->list, struct nfs_seqid, list) != seqid);
755 case -NFS4ERR_BAD_SEQID:
756 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
758 printk(KERN_WARNING "NFS: v4 server returned a bad"
759 " sequence-id error on an"
760 " unconfirmed sequence %p!\n",
762 case -NFS4ERR_STALE_CLIENTID:
763 case -NFS4ERR_STALE_STATEID:
764 case -NFS4ERR_BAD_STATEID:
765 case -NFS4ERR_BADXDR:
766 case -NFS4ERR_RESOURCE:
767 case -NFS4ERR_NOFILEHANDLE:
768 /* Non-seqid mutating errors */
772 * Note: no locking needed as we are guaranteed to be first
773 * on the sequence list
775 seqid->sequence->counter++;
778 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
780 struct nfs4_state_owner *sp = container_of(seqid->sequence,
781 struct nfs4_state_owner, so_seqid);
782 struct nfs_server *server = sp->so_server;
784 if (status == -NFS4ERR_BAD_SEQID)
785 nfs4_drop_state_owner(sp);
786 if (!nfs4_has_session(server->nfs_client))
787 nfs_increment_seqid(status, seqid);
791 * Increment the seqid if the LOCK/LOCKU succeeded, or
792 * failed with a seqid incrementing error -
793 * see comments nfs_fs.h:seqid_mutating_error()
795 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
797 nfs_increment_seqid(status, seqid);
800 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
802 struct rpc_sequence *sequence = seqid->sequence->sequence;
805 spin_lock(&sequence->lock);
806 if (list_empty(&seqid->list))
807 list_add_tail(&seqid->list, &sequence->list);
808 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
810 rpc_sleep_on(&sequence->wait, task, NULL);
813 spin_unlock(&sequence->lock);
817 static int nfs4_run_state_manager(void *);
819 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
821 smp_mb__before_clear_bit();
822 clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
823 smp_mb__after_clear_bit();
824 wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
825 rpc_wake_up(&clp->cl_rpcwaitq);
829 * Schedule the nfs_client asynchronous state management routine
831 void nfs4_schedule_state_manager(struct nfs_client *clp)
833 struct task_struct *task;
835 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
837 __module_get(THIS_MODULE);
838 atomic_inc(&clp->cl_count);
839 task = kthread_run(nfs4_run_state_manager, clp, "%s-manager",
840 rpc_peeraddr2str(clp->cl_rpcclient,
844 nfs4_clear_state_manager_bit(clp);
846 module_put(THIS_MODULE);
850 * Schedule a state recovery attempt
852 void nfs4_schedule_state_recovery(struct nfs_client *clp)
856 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
857 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
858 nfs4_schedule_state_manager(clp);
861 static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
864 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
865 /* Don't recover state that expired before the reboot */
866 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
867 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
870 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
871 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
875 int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
877 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
878 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
879 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
880 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
884 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
886 struct inode *inode = state->inode;
887 struct nfs_inode *nfsi = NFS_I(inode);
888 struct file_lock *fl;
891 if (inode->i_flock == NULL)
894 /* Guard against delegation returns and new lock/unlock calls */
895 down_write(&nfsi->rwsem);
896 /* Protect inode->i_flock using the BKL */
898 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
899 if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
901 if (nfs_file_open_context(fl->fl_file)->state != state)
904 status = ops->recover_lock(state, fl);
909 case -NFS4ERR_ADMIN_REVOKED:
910 case -NFS4ERR_STALE_STATEID:
911 case -NFS4ERR_BAD_STATEID:
912 case -NFS4ERR_EXPIRED:
913 case -NFS4ERR_NO_GRACE:
914 case -NFS4ERR_STALE_CLIENTID:
915 case -NFS4ERR_BADSESSION:
916 case -NFS4ERR_BADSLOT:
917 case -NFS4ERR_BAD_HIGH_SLOT:
918 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
921 printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
924 case -NFS4ERR_DENIED:
925 case -NFS4ERR_RECLAIM_BAD:
926 case -NFS4ERR_RECLAIM_CONFLICT:
927 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
934 up_write(&nfsi->rwsem);
938 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
940 struct nfs4_state *state;
941 struct nfs4_lock_state *lock;
944 /* Note: we rely on the sp->so_states list being ordered
945 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
947 * This is needed to ensure that the server won't give us any
948 * read delegations that we have to return if, say, we are
949 * recovering after a network partition or a reboot from a
950 * server that doesn't support a grace period.
953 spin_lock(&sp->so_lock);
954 list_for_each_entry(state, &sp->so_states, open_states) {
955 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
957 if (state->state == 0)
959 atomic_inc(&state->count);
960 spin_unlock(&sp->so_lock);
961 status = ops->recover_open(sp, state);
963 status = nfs4_reclaim_locks(state, ops);
965 list_for_each_entry(lock, &state->lock_states, ls_locks) {
966 if (!(lock->ls_flags & NFS_LOCK_INITIALIZED))
967 printk("%s: Lock reclaim failed!\n",
970 nfs4_put_open_state(state);
976 printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
982 * Open state on this file cannot be recovered
983 * All we can do is revert to using the zero stateid.
985 memset(state->stateid.data, 0,
986 sizeof(state->stateid.data));
987 /* Mark the file as being 'closed' */
990 case -NFS4ERR_ADMIN_REVOKED:
991 case -NFS4ERR_STALE_STATEID:
992 case -NFS4ERR_BAD_STATEID:
993 case -NFS4ERR_RECLAIM_BAD:
994 case -NFS4ERR_RECLAIM_CONFLICT:
995 nfs4_state_mark_reclaim_nograce(sp->so_client, state);
997 case -NFS4ERR_EXPIRED:
998 case -NFS4ERR_NO_GRACE:
999 nfs4_state_mark_reclaim_nograce(sp->so_client, state);
1000 case -NFS4ERR_STALE_CLIENTID:
1001 case -NFS4ERR_BADSESSION:
1002 case -NFS4ERR_BADSLOT:
1003 case -NFS4ERR_BAD_HIGH_SLOT:
1004 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1007 nfs4_put_open_state(state);
1010 spin_unlock(&sp->so_lock);
1013 nfs4_put_open_state(state);
1017 static void nfs4_clear_open_state(struct nfs4_state *state)
1019 struct nfs4_lock_state *lock;
1021 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1022 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1023 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1024 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1025 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1026 lock->ls_seqid.flags = 0;
1027 lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
1031 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp, int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1033 struct nfs4_state_owner *sp;
1034 struct rb_node *pos;
1035 struct nfs4_state *state;
1037 /* Reset all sequence ids to zero */
1038 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1039 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1040 sp->so_seqid.flags = 0;
1041 spin_lock(&sp->so_lock);
1042 list_for_each_entry(state, &sp->so_states, open_states) {
1043 if (mark_reclaim(clp, state))
1044 nfs4_clear_open_state(state);
1046 spin_unlock(&sp->so_lock);
1050 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1052 /* Mark all delegations for reclaim */
1053 nfs_delegation_mark_reclaim(clp);
1054 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1057 static void nfs4_reclaim_complete(struct nfs_client *clp,
1058 const struct nfs4_state_recovery_ops *ops)
1060 /* Notify the server we're done reclaiming our state */
1061 if (ops->reclaim_complete)
1062 (void)ops->reclaim_complete(clp);
1065 static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
1067 struct nfs4_state_owner *sp;
1068 struct rb_node *pos;
1069 struct nfs4_state *state;
1071 nfs4_reclaim_complete(clp,
1072 nfs4_reboot_recovery_ops[clp->cl_minorversion]);
1074 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1077 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1078 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1079 spin_lock(&sp->so_lock);
1080 list_for_each_entry(state, &sp->so_states, open_states) {
1081 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags))
1083 nfs4_state_mark_reclaim_nograce(clp, state);
1085 spin_unlock(&sp->so_lock);
1088 nfs_delegation_reap_unclaimed(clp);
1091 static void nfs_delegation_clear_all(struct nfs_client *clp)
1093 nfs_delegation_mark_reclaim(clp);
1094 nfs_delegation_reap_unclaimed(clp);
1097 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1099 nfs_delegation_clear_all(clp);
1100 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1103 static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1106 case -NFS4ERR_CB_PATH_DOWN:
1107 nfs_handle_cb_pathdown(clp);
1109 case -NFS4ERR_NO_GRACE:
1110 nfs4_state_end_reclaim_reboot(clp);
1112 case -NFS4ERR_STALE_CLIENTID:
1113 case -NFS4ERR_LEASE_MOVED:
1114 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1115 nfs4_state_end_reclaim_reboot(clp);
1116 nfs4_state_start_reclaim_reboot(clp);
1118 case -NFS4ERR_EXPIRED:
1119 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1120 nfs4_state_start_reclaim_nograce(clp);
1122 case -NFS4ERR_BADSESSION:
1123 case -NFS4ERR_BADSLOT:
1124 case -NFS4ERR_BAD_HIGH_SLOT:
1125 case -NFS4ERR_DEADSESSION:
1126 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1127 case -NFS4ERR_SEQ_FALSE_RETRY:
1128 case -NFS4ERR_SEQ_MISORDERED:
1129 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1130 /* Zero session reset errors */
1136 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1138 struct rb_node *pos;
1142 spin_lock(&clp->cl_lock);
1143 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1144 struct nfs4_state_owner *sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1145 if (!test_and_clear_bit(ops->owner_flag_bit, &sp->so_flags))
1147 atomic_inc(&sp->so_count);
1148 spin_unlock(&clp->cl_lock);
1149 status = nfs4_reclaim_open_state(sp, ops);
1151 set_bit(ops->owner_flag_bit, &sp->so_flags);
1152 nfs4_put_state_owner(sp);
1153 return nfs4_recovery_handle_error(clp, status);
1155 nfs4_put_state_owner(sp);
1158 spin_unlock(&clp->cl_lock);
1162 static int nfs4_check_lease(struct nfs_client *clp)
1164 struct rpc_cred *cred;
1165 struct nfs4_state_maintenance_ops *ops =
1166 nfs4_state_renewal_ops[clp->cl_minorversion];
1167 int status = -NFS4ERR_EXPIRED;
1169 /* Is the client already known to have an expired lease? */
1170 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1172 spin_lock(&clp->cl_lock);
1173 cred = ops->get_state_renewal_cred_locked(clp);
1174 spin_unlock(&clp->cl_lock);
1176 cred = nfs4_get_setclientid_cred(clp);
1180 status = ops->renew_lease(clp, cred);
1183 return nfs4_recovery_handle_error(clp, status);
1186 static int nfs4_reclaim_lease(struct nfs_client *clp)
1188 struct rpc_cred *cred;
1189 struct nfs4_state_recovery_ops *ops =
1190 nfs4_reboot_recovery_ops[clp->cl_minorversion];
1191 int status = -ENOENT;
1193 cred = ops->get_clid_cred(clp);
1195 status = ops->establish_clid(clp, cred);
1197 /* Handle case where the user hasn't set up machine creds */
1198 if (status == -EACCES && cred == clp->cl_machine_cred) {
1199 nfs4_clear_machine_cred(clp);
1202 if (status == -NFS4ERR_MINOR_VERS_MISMATCH)
1203 status = -EPROTONOSUPPORT;
1208 #ifdef CONFIG_NFS_V4_1
1209 void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
1213 else if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED) {
1214 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1215 nfs4_state_start_reclaim_reboot(clp);
1216 nfs4_schedule_state_recovery(clp);
1217 } else if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
1218 SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
1219 SEQ4_STATUS_ADMIN_STATE_REVOKED |
1220 SEQ4_STATUS_RECALLABLE_STATE_REVOKED |
1221 SEQ4_STATUS_LEASE_MOVED)) {
1222 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1223 nfs4_state_start_reclaim_nograce(clp);
1224 nfs4_schedule_state_recovery(clp);
1225 } else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
1226 SEQ4_STATUS_BACKCHANNEL_FAULT |
1227 SEQ4_STATUS_CB_PATH_DOWN_SESSION))
1228 nfs_expire_all_delegations(clp);
1231 static void nfs4_session_recovery_handle_error(struct nfs_client *clp, int err)
1234 case -NFS4ERR_STALE_CLIENTID:
1235 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1239 static int nfs4_reset_session(struct nfs_client *clp)
1241 struct nfs4_session *ses = clp->cl_session;
1242 struct nfs4_slot_table *tbl = &ses->fc_slot_table;
1245 spin_lock(&tbl->slot_tbl_lock);
1246 set_bit(NFS4CLNT_SESSION_DRAINING, &clp->cl_state);
1247 if (tbl->highest_used_slotid != -1) {
1248 INIT_COMPLETION(ses->complete);
1249 spin_unlock(&tbl->slot_tbl_lock);
1250 status = wait_for_completion_interruptible(&ses->complete);
1251 if (status) /* -ERESTARTSYS */
1254 spin_unlock(&tbl->slot_tbl_lock);
1257 status = nfs4_proc_destroy_session(clp->cl_session);
1258 if (status && status != -NFS4ERR_BADSESSION &&
1259 status != -NFS4ERR_DEADSESSION) {
1260 nfs4_session_recovery_handle_error(clp, status);
1264 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
1265 status = nfs4_proc_create_session(clp);
1267 nfs4_session_recovery_handle_error(clp, status);
1270 /* Wake up the next rpc task even on error */
1271 clear_bit(NFS4CLNT_SESSION_DRAINING, &clp->cl_state);
1272 rpc_wake_up(&clp->cl_session->fc_slot_table.slot_tbl_waitq);
1274 nfs41_setup_state_renewal(clp);
1278 #else /* CONFIG_NFS_V4_1 */
1279 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
1280 #endif /* CONFIG_NFS_V4_1 */
1282 /* Set NFS4CLNT_LEASE_EXPIRED for all v4.0 errors and for recoverable errors
1283 * on EXCHANGE_ID for v4.1
1285 static void nfs4_set_lease_expired(struct nfs_client *clp, int status)
1287 if (nfs4_has_session(clp)) {
1289 case -NFS4ERR_DELAY:
1290 case -NFS4ERR_CLID_INUSE:
1294 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1295 * in nfs4_exchange_id */
1300 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1303 static void nfs4_state_manager(struct nfs_client *clp)
1307 /* Ensure exclusive access to NFSv4 state */
1309 if (test_and_clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
1310 /* We're going to have to re-establish a clientid */
1311 status = nfs4_reclaim_lease(clp);
1313 nfs4_set_lease_expired(clp, status);
1314 if (test_bit(NFS4CLNT_LEASE_EXPIRED,
1317 if (clp->cl_cons_state ==
1318 NFS_CS_SESSION_INITING)
1319 nfs_mark_client_ready(clp, status);
1322 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1325 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
1326 status = nfs4_check_lease(clp);
1327 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1329 if (status < 0 && status != -NFS4ERR_CB_PATH_DOWN)
1333 /* Initialize or reset the session */
1334 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)
1335 && nfs4_has_session(clp)) {
1336 status = nfs4_reset_session(clp);
1337 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1343 /* First recover reboot state... */
1344 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
1345 status = nfs4_do_reclaim(clp,
1346 nfs4_reboot_recovery_ops[clp->cl_minorversion]);
1347 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1348 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
1350 nfs4_state_end_reclaim_reboot(clp);
1351 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
1357 /* Now recover expired state... */
1358 if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
1359 status = nfs4_do_reclaim(clp,
1360 nfs4_nograce_recovery_ops[clp->cl_minorversion]);
1361 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1362 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
1363 test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1369 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
1370 nfs_client_return_marked_delegations(clp);
1374 nfs4_clear_state_manager_bit(clp);
1375 /* Did we race with an attempt to give us more work? */
1376 if (clp->cl_state == 0)
1378 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1383 printk(KERN_WARNING "Error: state manager failed on NFSv4 server %s"
1384 " with error %d\n", clp->cl_hostname, -status);
1385 nfs4_clear_state_manager_bit(clp);
1388 static int nfs4_run_state_manager(void *ptr)
1390 struct nfs_client *clp = ptr;
1392 allow_signal(SIGKILL);
1393 nfs4_state_manager(clp);
1394 nfs_put_client(clp);
1395 module_put_and_exit(0);