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 int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
123 status = nfs4_proc_exchange_id(clp, cred);
125 /* create session schedules state renewal upon success */
126 status = nfs4_proc_create_session(clp, 0);
128 nfs_mark_client_ready(clp, NFS_CS_READY);
132 struct rpc_cred *nfs4_get_exchange_id_cred(struct nfs_client *clp)
134 struct rpc_cred *cred;
136 spin_lock(&clp->cl_lock);
137 cred = nfs4_get_machine_cred_locked(clp);
138 spin_unlock(&clp->cl_lock);
142 #endif /* CONFIG_NFS_V4_1 */
144 struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
146 struct nfs4_state_owner *sp;
148 struct rpc_cred *cred;
150 spin_lock(&clp->cl_lock);
151 cred = nfs4_get_machine_cred_locked(clp);
154 pos = rb_first(&clp->cl_state_owners);
156 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
157 cred = get_rpccred(sp->so_cred);
160 spin_unlock(&clp->cl_lock);
164 static void nfs_alloc_unique_id(struct rb_root *root, struct nfs_unique_id *new,
165 __u64 minval, int maxbits)
167 struct rb_node **p, *parent;
168 struct nfs_unique_id *pos;
172 mask = (1ULL << maxbits) - 1ULL;
174 /* Ensure distribution is more or less flat */
175 get_random_bytes(&new->id, sizeof(new->id));
177 if (new->id < minval)
185 pos = rb_entry(parent, struct nfs_unique_id, rb_node);
187 if (new->id < pos->id)
189 else if (new->id > pos->id)
194 rb_link_node(&new->rb_node, parent, p);
195 rb_insert_color(&new->rb_node, root);
200 if (new->id < minval || (new->id & mask) != new->id) {
204 parent = rb_next(parent);
207 pos = rb_entry(parent, struct nfs_unique_id, rb_node);
208 if (new->id < pos->id)
214 static void nfs_free_unique_id(struct rb_root *root, struct nfs_unique_id *id)
216 rb_erase(&id->rb_node, root);
219 static struct nfs4_state_owner *
220 nfs4_find_state_owner(struct nfs_server *server, struct rpc_cred *cred)
222 struct nfs_client *clp = server->nfs_client;
223 struct rb_node **p = &clp->cl_state_owners.rb_node,
225 struct nfs4_state_owner *sp, *res = NULL;
229 sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
231 if (server < sp->so_server) {
232 p = &parent->rb_left;
235 if (server > sp->so_server) {
236 p = &parent->rb_right;
239 if (cred < sp->so_cred)
240 p = &parent->rb_left;
241 else if (cred > sp->so_cred)
242 p = &parent->rb_right;
244 atomic_inc(&sp->so_count);
252 static struct nfs4_state_owner *
253 nfs4_insert_state_owner(struct nfs_client *clp, struct nfs4_state_owner *new)
255 struct rb_node **p = &clp->cl_state_owners.rb_node,
257 struct nfs4_state_owner *sp;
261 sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
263 if (new->so_server < sp->so_server) {
264 p = &parent->rb_left;
267 if (new->so_server > sp->so_server) {
268 p = &parent->rb_right;
271 if (new->so_cred < sp->so_cred)
272 p = &parent->rb_left;
273 else if (new->so_cred > sp->so_cred)
274 p = &parent->rb_right;
276 atomic_inc(&sp->so_count);
280 nfs_alloc_unique_id(&clp->cl_openowner_id, &new->so_owner_id, 1, 64);
281 rb_link_node(&new->so_client_node, parent, p);
282 rb_insert_color(&new->so_client_node, &clp->cl_state_owners);
287 nfs4_remove_state_owner(struct nfs_client *clp, struct nfs4_state_owner *sp)
289 if (!RB_EMPTY_NODE(&sp->so_client_node))
290 rb_erase(&sp->so_client_node, &clp->cl_state_owners);
291 nfs_free_unique_id(&clp->cl_openowner_id, &sp->so_owner_id);
295 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
296 * create a new state_owner.
299 static struct nfs4_state_owner *
300 nfs4_alloc_state_owner(void)
302 struct nfs4_state_owner *sp;
304 sp = kzalloc(sizeof(*sp),GFP_KERNEL);
307 spin_lock_init(&sp->so_lock);
308 INIT_LIST_HEAD(&sp->so_states);
309 INIT_LIST_HEAD(&sp->so_delegations);
310 rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue");
311 sp->so_seqid.sequence = &sp->so_sequence;
312 spin_lock_init(&sp->so_sequence.lock);
313 INIT_LIST_HEAD(&sp->so_sequence.list);
314 atomic_set(&sp->so_count, 1);
319 nfs4_drop_state_owner(struct nfs4_state_owner *sp)
321 if (!RB_EMPTY_NODE(&sp->so_client_node)) {
322 struct nfs_client *clp = sp->so_client;
324 spin_lock(&clp->cl_lock);
325 rb_erase(&sp->so_client_node, &clp->cl_state_owners);
326 RB_CLEAR_NODE(&sp->so_client_node);
327 spin_unlock(&clp->cl_lock);
331 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct rpc_cred *cred)
333 struct nfs_client *clp = server->nfs_client;
334 struct nfs4_state_owner *sp, *new;
336 spin_lock(&clp->cl_lock);
337 sp = nfs4_find_state_owner(server, cred);
338 spin_unlock(&clp->cl_lock);
341 new = nfs4_alloc_state_owner();
344 new->so_client = clp;
345 new->so_server = server;
347 spin_lock(&clp->cl_lock);
348 sp = nfs4_insert_state_owner(clp, new);
349 spin_unlock(&clp->cl_lock);
353 rpc_destroy_wait_queue(&new->so_sequence.wait);
359 void nfs4_put_state_owner(struct nfs4_state_owner *sp)
361 struct nfs_client *clp = sp->so_client;
362 struct rpc_cred *cred = sp->so_cred;
364 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
366 nfs4_remove_state_owner(clp, sp);
367 spin_unlock(&clp->cl_lock);
368 rpc_destroy_wait_queue(&sp->so_sequence.wait);
373 static struct nfs4_state *
374 nfs4_alloc_open_state(void)
376 struct nfs4_state *state;
378 state = kzalloc(sizeof(*state), GFP_KERNEL);
381 atomic_set(&state->count, 1);
382 INIT_LIST_HEAD(&state->lock_states);
383 spin_lock_init(&state->state_lock);
384 seqlock_init(&state->seqlock);
389 nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
391 if (state->state == fmode)
393 /* NB! List reordering - see the reclaim code for why. */
394 if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
395 if (fmode & FMODE_WRITE)
396 list_move(&state->open_states, &state->owner->so_states);
398 list_move_tail(&state->open_states, &state->owner->so_states);
400 state->state = fmode;
403 static struct nfs4_state *
404 __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
406 struct nfs_inode *nfsi = NFS_I(inode);
407 struct nfs4_state *state;
409 list_for_each_entry(state, &nfsi->open_states, inode_states) {
410 if (state->owner != owner)
412 if (atomic_inc_not_zero(&state->count))
419 nfs4_free_open_state(struct nfs4_state *state)
425 nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
427 struct nfs4_state *state, *new;
428 struct nfs_inode *nfsi = NFS_I(inode);
430 spin_lock(&inode->i_lock);
431 state = __nfs4_find_state_byowner(inode, owner);
432 spin_unlock(&inode->i_lock);
435 new = nfs4_alloc_open_state();
436 spin_lock(&owner->so_lock);
437 spin_lock(&inode->i_lock);
438 state = __nfs4_find_state_byowner(inode, owner);
439 if (state == NULL && new != NULL) {
441 state->owner = owner;
442 atomic_inc(&owner->so_count);
443 list_add(&state->inode_states, &nfsi->open_states);
444 state->inode = igrab(inode);
445 spin_unlock(&inode->i_lock);
446 /* Note: The reclaim code dictates that we add stateless
447 * and read-only stateids to the end of the list */
448 list_add_tail(&state->open_states, &owner->so_states);
449 spin_unlock(&owner->so_lock);
451 spin_unlock(&inode->i_lock);
452 spin_unlock(&owner->so_lock);
454 nfs4_free_open_state(new);
460 void nfs4_put_open_state(struct nfs4_state *state)
462 struct inode *inode = state->inode;
463 struct nfs4_state_owner *owner = state->owner;
465 if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
467 spin_lock(&inode->i_lock);
468 list_del(&state->inode_states);
469 list_del(&state->open_states);
470 spin_unlock(&inode->i_lock);
471 spin_unlock(&owner->so_lock);
473 nfs4_free_open_state(state);
474 nfs4_put_state_owner(owner);
478 * Close the current file.
480 static void __nfs4_close(struct path *path, struct nfs4_state *state, fmode_t fmode, int wait)
482 struct nfs4_state_owner *owner = state->owner;
486 atomic_inc(&owner->so_count);
487 /* Protect against nfs4_find_state() */
488 spin_lock(&owner->so_lock);
489 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
496 case FMODE_READ|FMODE_WRITE:
499 newstate = FMODE_READ|FMODE_WRITE;
500 if (state->n_rdwr == 0) {
501 if (state->n_rdonly == 0) {
502 newstate &= ~FMODE_READ;
503 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
504 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
506 if (state->n_wronly == 0) {
507 newstate &= ~FMODE_WRITE;
508 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
509 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
512 clear_bit(NFS_DELEGATED_STATE, &state->flags);
514 nfs4_state_set_mode_locked(state, newstate);
515 spin_unlock(&owner->so_lock);
518 nfs4_put_open_state(state);
519 nfs4_put_state_owner(owner);
521 nfs4_do_close(path, state, wait);
524 void nfs4_close_state(struct path *path, struct nfs4_state *state, fmode_t fmode)
526 __nfs4_close(path, state, fmode, 0);
529 void nfs4_close_sync(struct path *path, struct nfs4_state *state, fmode_t fmode)
531 __nfs4_close(path, state, fmode, 1);
535 * Search the state->lock_states for an existing lock_owner
536 * that is compatible with current->files
538 static struct nfs4_lock_state *
539 __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
541 struct nfs4_lock_state *pos;
542 list_for_each_entry(pos, &state->lock_states, ls_locks) {
543 if (pos->ls_owner != fl_owner)
545 atomic_inc(&pos->ls_count);
552 * Return a compatible lock_state. If no initialized lock_state structure
553 * exists, return an uninitialized one.
556 static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
558 struct nfs4_lock_state *lsp;
559 struct nfs_client *clp = state->owner->so_client;
561 lsp = kzalloc(sizeof(*lsp), GFP_KERNEL);
564 rpc_init_wait_queue(&lsp->ls_sequence.wait, "lock_seqid_waitqueue");
565 spin_lock_init(&lsp->ls_sequence.lock);
566 INIT_LIST_HEAD(&lsp->ls_sequence.list);
567 lsp->ls_seqid.sequence = &lsp->ls_sequence;
568 atomic_set(&lsp->ls_count, 1);
569 lsp->ls_state = state;
570 lsp->ls_owner = fl_owner;
571 spin_lock(&clp->cl_lock);
572 nfs_alloc_unique_id(&clp->cl_lockowner_id, &lsp->ls_id, 1, 64);
573 spin_unlock(&clp->cl_lock);
574 INIT_LIST_HEAD(&lsp->ls_locks);
578 static void nfs4_free_lock_state(struct nfs4_lock_state *lsp)
580 struct nfs_client *clp = lsp->ls_state->owner->so_client;
582 spin_lock(&clp->cl_lock);
583 nfs_free_unique_id(&clp->cl_lockowner_id, &lsp->ls_id);
584 spin_unlock(&clp->cl_lock);
585 rpc_destroy_wait_queue(&lsp->ls_sequence.wait);
590 * Return a compatible lock_state. If no initialized lock_state structure
591 * exists, return an uninitialized one.
594 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
596 struct nfs4_lock_state *lsp, *new = NULL;
599 spin_lock(&state->state_lock);
600 lsp = __nfs4_find_lock_state(state, owner);
604 list_add(&new->ls_locks, &state->lock_states);
605 set_bit(LK_STATE_IN_USE, &state->flags);
610 spin_unlock(&state->state_lock);
611 new = nfs4_alloc_lock_state(state, owner);
615 spin_unlock(&state->state_lock);
617 nfs4_free_lock_state(new);
622 * Release reference to lock_state, and free it if we see that
623 * it is no longer in use
625 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
627 struct nfs4_state *state;
631 state = lsp->ls_state;
632 if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
634 list_del(&lsp->ls_locks);
635 if (list_empty(&state->lock_states))
636 clear_bit(LK_STATE_IN_USE, &state->flags);
637 spin_unlock(&state->state_lock);
638 nfs4_free_lock_state(lsp);
641 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
643 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
645 dst->fl_u.nfs4_fl.owner = lsp;
646 atomic_inc(&lsp->ls_count);
649 static void nfs4_fl_release_lock(struct file_lock *fl)
651 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
654 static const struct file_lock_operations nfs4_fl_lock_ops = {
655 .fl_copy_lock = nfs4_fl_copy_lock,
656 .fl_release_private = nfs4_fl_release_lock,
659 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
661 struct nfs4_lock_state *lsp;
663 if (fl->fl_ops != NULL)
665 lsp = nfs4_get_lock_state(state, fl->fl_owner);
668 fl->fl_u.nfs4_fl.owner = lsp;
669 fl->fl_ops = &nfs4_fl_lock_ops;
674 * Byte-range lock aware utility to initialize the stateid of read/write
677 void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner)
679 struct nfs4_lock_state *lsp;
683 seq = read_seqbegin(&state->seqlock);
684 memcpy(dst, &state->stateid, sizeof(*dst));
685 } while (read_seqretry(&state->seqlock, seq));
686 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
689 spin_lock(&state->state_lock);
690 lsp = __nfs4_find_lock_state(state, fl_owner);
691 if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
692 memcpy(dst, &lsp->ls_stateid, sizeof(*dst));
693 spin_unlock(&state->state_lock);
694 nfs4_put_lock_state(lsp);
697 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter)
699 struct nfs_seqid *new;
701 new = kmalloc(sizeof(*new), GFP_KERNEL);
703 new->sequence = counter;
704 INIT_LIST_HEAD(&new->list);
709 void nfs_free_seqid(struct nfs_seqid *seqid)
711 if (!list_empty(&seqid->list)) {
712 struct rpc_sequence *sequence = seqid->sequence->sequence;
714 spin_lock(&sequence->lock);
715 list_del(&seqid->list);
716 spin_unlock(&sequence->lock);
717 rpc_wake_up(&sequence->wait);
723 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
724 * failed with a seqid incrementing error -
725 * see comments nfs_fs.h:seqid_mutating_error()
727 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
729 BUG_ON(list_first_entry(&seqid->sequence->sequence->list, struct nfs_seqid, list) != seqid);
733 case -NFS4ERR_BAD_SEQID:
734 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
736 printk(KERN_WARNING "NFS: v4 server returned a bad"
737 " sequence-id error on an"
738 " unconfirmed sequence %p!\n",
740 case -NFS4ERR_STALE_CLIENTID:
741 case -NFS4ERR_STALE_STATEID:
742 case -NFS4ERR_BAD_STATEID:
743 case -NFS4ERR_BADXDR:
744 case -NFS4ERR_RESOURCE:
745 case -NFS4ERR_NOFILEHANDLE:
746 /* Non-seqid mutating errors */
750 * Note: no locking needed as we are guaranteed to be first
751 * on the sequence list
753 seqid->sequence->counter++;
756 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
758 struct nfs4_state_owner *sp = container_of(seqid->sequence,
759 struct nfs4_state_owner, so_seqid);
760 struct nfs_server *server = sp->so_server;
762 if (status == -NFS4ERR_BAD_SEQID)
763 nfs4_drop_state_owner(sp);
764 if (!nfs4_has_session(server->nfs_client))
765 nfs_increment_seqid(status, seqid);
769 * Increment the seqid if the LOCK/LOCKU succeeded, or
770 * failed with a seqid incrementing error -
771 * see comments nfs_fs.h:seqid_mutating_error()
773 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
775 nfs_increment_seqid(status, seqid);
778 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
780 struct rpc_sequence *sequence = seqid->sequence->sequence;
783 spin_lock(&sequence->lock);
784 if (list_empty(&seqid->list))
785 list_add_tail(&seqid->list, &sequence->list);
786 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
788 rpc_sleep_on(&sequence->wait, task, NULL);
791 spin_unlock(&sequence->lock);
795 static int nfs4_run_state_manager(void *);
797 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
799 smp_mb__before_clear_bit();
800 clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
801 smp_mb__after_clear_bit();
802 wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
803 rpc_wake_up(&clp->cl_rpcwaitq);
807 * Schedule the nfs_client asynchronous state management routine
809 void nfs4_schedule_state_manager(struct nfs_client *clp)
811 struct task_struct *task;
813 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
815 __module_get(THIS_MODULE);
816 atomic_inc(&clp->cl_count);
817 task = kthread_run(nfs4_run_state_manager, clp, "%s-manager",
818 rpc_peeraddr2str(clp->cl_rpcclient,
822 nfs4_clear_state_manager_bit(clp);
824 module_put(THIS_MODULE);
828 * Schedule a state recovery attempt
830 void nfs4_schedule_state_recovery(struct nfs_client *clp)
834 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
835 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
836 nfs4_schedule_state_manager(clp);
839 static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
842 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
843 /* Don't recover state that expired before the reboot */
844 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
845 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
848 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
849 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
853 int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
855 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
856 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
857 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
858 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
862 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
864 struct inode *inode = state->inode;
865 struct nfs_inode *nfsi = NFS_I(inode);
866 struct file_lock *fl;
869 if (inode->i_flock == NULL)
872 /* Guard against delegation returns and new lock/unlock calls */
873 down_write(&nfsi->rwsem);
874 /* Protect inode->i_flock using the BKL */
876 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
877 if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
879 if (nfs_file_open_context(fl->fl_file)->state != state)
882 status = ops->recover_lock(state, fl);
887 case -NFS4ERR_ADMIN_REVOKED:
888 case -NFS4ERR_STALE_STATEID:
889 case -NFS4ERR_BAD_STATEID:
890 case -NFS4ERR_EXPIRED:
891 case -NFS4ERR_NO_GRACE:
892 case -NFS4ERR_STALE_CLIENTID:
893 case -NFS4ERR_BADSESSION:
894 case -NFS4ERR_BADSLOT:
895 case -NFS4ERR_BAD_HIGH_SLOT:
896 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
899 printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
902 case -NFS4ERR_DENIED:
903 case -NFS4ERR_RECLAIM_BAD:
904 case -NFS4ERR_RECLAIM_CONFLICT:
905 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
912 up_write(&nfsi->rwsem);
916 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
918 struct nfs4_state *state;
919 struct nfs4_lock_state *lock;
922 /* Note: we rely on the sp->so_states list being ordered
923 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
925 * This is needed to ensure that the server won't give us any
926 * read delegations that we have to return if, say, we are
927 * recovering after a network partition or a reboot from a
928 * server that doesn't support a grace period.
931 spin_lock(&sp->so_lock);
932 list_for_each_entry(state, &sp->so_states, open_states) {
933 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
935 if (state->state == 0)
937 atomic_inc(&state->count);
938 spin_unlock(&sp->so_lock);
939 status = ops->recover_open(sp, state);
941 status = nfs4_reclaim_locks(state, ops);
943 list_for_each_entry(lock, &state->lock_states, ls_locks) {
944 if (!(lock->ls_flags & NFS_LOCK_INITIALIZED))
945 printk("%s: Lock reclaim failed!\n",
948 nfs4_put_open_state(state);
954 printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
960 * Open state on this file cannot be recovered
961 * All we can do is revert to using the zero stateid.
963 memset(state->stateid.data, 0,
964 sizeof(state->stateid.data));
965 /* Mark the file as being 'closed' */
968 case -NFS4ERR_ADMIN_REVOKED:
969 case -NFS4ERR_STALE_STATEID:
970 case -NFS4ERR_BAD_STATEID:
971 case -NFS4ERR_RECLAIM_BAD:
972 case -NFS4ERR_RECLAIM_CONFLICT:
973 nfs4_state_mark_reclaim_nograce(sp->so_client, state);
975 case -NFS4ERR_EXPIRED:
976 case -NFS4ERR_NO_GRACE:
977 nfs4_state_mark_reclaim_nograce(sp->so_client, state);
978 case -NFS4ERR_STALE_CLIENTID:
979 case -NFS4ERR_BADSESSION:
980 case -NFS4ERR_BADSLOT:
981 case -NFS4ERR_BAD_HIGH_SLOT:
982 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
985 nfs4_put_open_state(state);
988 spin_unlock(&sp->so_lock);
991 nfs4_put_open_state(state);
995 static void nfs4_clear_open_state(struct nfs4_state *state)
997 struct nfs4_lock_state *lock;
999 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1000 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1001 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1002 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1003 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1004 lock->ls_seqid.flags = 0;
1005 lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
1009 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp, int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1011 struct nfs4_state_owner *sp;
1012 struct rb_node *pos;
1013 struct nfs4_state *state;
1015 /* Reset all sequence ids to zero */
1016 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1017 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1018 sp->so_seqid.flags = 0;
1019 spin_lock(&sp->so_lock);
1020 list_for_each_entry(state, &sp->so_states, open_states) {
1021 if (mark_reclaim(clp, state))
1022 nfs4_clear_open_state(state);
1024 spin_unlock(&sp->so_lock);
1028 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1030 /* Mark all delegations for reclaim */
1031 nfs_delegation_mark_reclaim(clp);
1032 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1035 static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
1037 struct nfs4_state_owner *sp;
1038 struct rb_node *pos;
1039 struct nfs4_state *state;
1041 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1044 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1045 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1046 spin_lock(&sp->so_lock);
1047 list_for_each_entry(state, &sp->so_states, open_states) {
1048 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags))
1050 nfs4_state_mark_reclaim_nograce(clp, state);
1052 spin_unlock(&sp->so_lock);
1055 nfs_delegation_reap_unclaimed(clp);
1058 static void nfs_delegation_clear_all(struct nfs_client *clp)
1060 nfs_delegation_mark_reclaim(clp);
1061 nfs_delegation_reap_unclaimed(clp);
1064 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1066 nfs_delegation_clear_all(clp);
1067 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1070 static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1073 case -NFS4ERR_CB_PATH_DOWN:
1074 nfs_handle_cb_pathdown(clp);
1076 case -NFS4ERR_NO_GRACE:
1077 nfs4_state_end_reclaim_reboot(clp);
1079 case -NFS4ERR_STALE_CLIENTID:
1080 case -NFS4ERR_LEASE_MOVED:
1081 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1082 nfs4_state_end_reclaim_reboot(clp);
1083 nfs4_state_start_reclaim_reboot(clp);
1085 case -NFS4ERR_EXPIRED:
1086 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1087 nfs4_state_start_reclaim_nograce(clp);
1088 case -NFS4ERR_BADSESSION:
1089 case -NFS4ERR_BADSLOT:
1090 case -NFS4ERR_BAD_HIGH_SLOT:
1091 case -NFS4ERR_DEADSESSION:
1092 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1093 case -NFS4ERR_SEQ_FALSE_RETRY:
1094 case -NFS4ERR_SEQ_MISORDERED:
1095 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1100 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1102 struct rb_node *pos;
1106 spin_lock(&clp->cl_lock);
1107 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1108 struct nfs4_state_owner *sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1109 if (!test_and_clear_bit(ops->owner_flag_bit, &sp->so_flags))
1111 atomic_inc(&sp->so_count);
1112 spin_unlock(&clp->cl_lock);
1113 status = nfs4_reclaim_open_state(sp, ops);
1115 set_bit(ops->owner_flag_bit, &sp->so_flags);
1116 nfs4_put_state_owner(sp);
1117 return nfs4_recovery_handle_error(clp, status);
1119 nfs4_put_state_owner(sp);
1122 spin_unlock(&clp->cl_lock);
1126 static int nfs4_check_lease(struct nfs_client *clp)
1128 struct rpc_cred *cred;
1129 struct nfs4_state_maintenance_ops *ops =
1130 nfs4_state_renewal_ops[clp->cl_minorversion];
1131 int status = -NFS4ERR_EXPIRED;
1133 /* Is the client already known to have an expired lease? */
1134 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1136 spin_lock(&clp->cl_lock);
1137 cred = ops->get_state_renewal_cred_locked(clp);
1138 spin_unlock(&clp->cl_lock);
1140 cred = nfs4_get_setclientid_cred(clp);
1144 status = ops->renew_lease(clp, cred);
1147 return nfs4_recovery_handle_error(clp, status);
1150 static int nfs4_reclaim_lease(struct nfs_client *clp)
1152 struct rpc_cred *cred;
1153 struct nfs4_state_recovery_ops *ops =
1154 nfs4_reboot_recovery_ops[clp->cl_minorversion];
1155 int status = -ENOENT;
1157 cred = ops->get_clid_cred(clp);
1159 status = ops->establish_clid(clp, cred);
1161 /* Handle case where the user hasn't set up machine creds */
1162 if (status == -EACCES && cred == clp->cl_machine_cred) {
1163 nfs4_clear_machine_cred(clp);
1166 if (status == -NFS4ERR_MINOR_VERS_MISMATCH)
1167 status = -EPROTONOSUPPORT;
1172 #ifdef CONFIG_NFS_V4_1
1173 static void nfs4_session_recovery_handle_error(struct nfs_client *clp, int err)
1176 case -NFS4ERR_STALE_CLIENTID:
1177 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1181 static int nfs4_reset_session(struct nfs_client *clp)
1185 status = nfs4_proc_destroy_session(clp->cl_session);
1186 if (status && status != -NFS4ERR_BADSESSION &&
1187 status != -NFS4ERR_DEADSESSION) {
1188 nfs4_session_recovery_handle_error(clp, status);
1192 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
1193 status = nfs4_proc_create_session(clp, 1);
1195 nfs4_session_recovery_handle_error(clp, status);
1198 /* Wake up the next rpc task even on error */
1199 rpc_wake_up_next(&clp->cl_session->fc_slot_table.slot_tbl_waitq);
1203 #else /* CONFIG_NFS_V4_1 */
1204 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
1205 #endif /* CONFIG_NFS_V4_1 */
1207 /* Set NFS4CLNT_LEASE_EXPIRED for all v4.0 errors and for recoverable errors
1208 * on EXCHANGE_ID for v4.1
1210 static void nfs4_set_lease_expired(struct nfs_client *clp, int status)
1212 if (nfs4_has_session(clp)) {
1214 case -NFS4ERR_DELAY:
1215 case -NFS4ERR_CLID_INUSE:
1219 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1220 * in nfs4_exchange_id */
1225 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1228 static void nfs4_state_manager(struct nfs_client *clp)
1232 /* Ensure exclusive access to NFSv4 state */
1234 if (test_and_clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
1235 /* We're going to have to re-establish a clientid */
1236 status = nfs4_reclaim_lease(clp);
1238 nfs4_set_lease_expired(clp, status);
1239 if (test_bit(NFS4CLNT_LEASE_EXPIRED,
1242 if (clp->cl_cons_state ==
1243 NFS_CS_SESSION_INITING)
1244 nfs_mark_client_ready(clp, status);
1247 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1250 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
1251 status = nfs4_check_lease(clp);
1252 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1254 if (status < 0 && status != -NFS4ERR_CB_PATH_DOWN)
1258 /* Initialize or reset the session */
1259 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)
1260 && nfs4_has_session(clp)) {
1261 status = nfs4_reset_session(clp);
1262 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1268 /* First recover reboot state... */
1269 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
1270 status = nfs4_do_reclaim(clp,
1271 nfs4_reboot_recovery_ops[clp->cl_minorversion]);
1272 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1273 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
1275 nfs4_state_end_reclaim_reboot(clp);
1276 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
1282 /* Now recover expired state... */
1283 if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
1284 status = nfs4_do_reclaim(clp,
1285 nfs4_nograce_recovery_ops[clp->cl_minorversion]);
1286 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1287 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
1288 test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1294 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
1295 nfs_client_return_marked_delegations(clp);
1299 nfs4_clear_state_manager_bit(clp);
1300 /* Did we race with an attempt to give us more work? */
1301 if (clp->cl_state == 0)
1303 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1308 printk(KERN_WARNING "Error: state manager failed on NFSv4 server %s"
1309 " with error %d\n", clp->cl_hostname, -status);
1310 nfs4_clear_state_manager_bit(clp);
1313 static int nfs4_run_state_manager(void *ptr)
1315 struct nfs_client *clp = ptr;
1317 allow_signal(SIGKILL);
1318 nfs4_state_manager(clp);
1319 nfs_put_client(clp);
1320 module_put_and_exit(0);