4 * Client-side XDR for NFSv4.
6 * Copyright (c) 2002 The Regents of the University of Michigan.
9 * Kendrick Smith <kmsmith@umich.edu>
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its
21 * contributors may be used to endorse or promote products derived
22 * from this software without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
25 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
26 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
27 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
31 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
32 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
33 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
34 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36 * Implementation of the NFSv4 state model. For the time being,
37 * this is minimal, but will be made much more complex in a
41 #include <linux/kernel.h>
42 #include <linux/slab.h>
43 #include <linux/smp_lock.h>
44 #include <linux/nfs_fs.h>
45 #include <linux/nfs_idmap.h>
46 #include <linux/kthread.h>
47 #include <linux/module.h>
48 #include <linux/random.h>
49 #include <linux/workqueue.h>
50 #include <linux/bitops.h>
54 #include "delegation.h"
57 #define OPENOWNER_POOL_SIZE 8
59 const nfs4_stateid zero_stateid;
61 static LIST_HEAD(nfs4_clientid_list);
63 static int nfs4_init_client(struct nfs_client *clp, struct rpc_cred *cred)
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 struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
119 struct nfs4_state_owner *sp;
121 struct rpc_cred *cred;
123 spin_lock(&clp->cl_lock);
124 cred = nfs4_get_machine_cred_locked(clp);
127 pos = rb_first(&clp->cl_state_owners);
129 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
130 cred = get_rpccred(sp->so_cred);
133 spin_unlock(&clp->cl_lock);
137 static void nfs_alloc_unique_id(struct rb_root *root, struct nfs_unique_id *new,
138 __u64 minval, int maxbits)
140 struct rb_node **p, *parent;
141 struct nfs_unique_id *pos;
145 mask = (1ULL << maxbits) - 1ULL;
147 /* Ensure distribution is more or less flat */
148 get_random_bytes(&new->id, sizeof(new->id));
150 if (new->id < minval)
158 pos = rb_entry(parent, struct nfs_unique_id, rb_node);
160 if (new->id < pos->id)
162 else if (new->id > pos->id)
167 rb_link_node(&new->rb_node, parent, p);
168 rb_insert_color(&new->rb_node, root);
173 if (new->id < minval || (new->id & mask) != new->id) {
177 parent = rb_next(parent);
180 pos = rb_entry(parent, struct nfs_unique_id, rb_node);
181 if (new->id < pos->id)
187 static void nfs_free_unique_id(struct rb_root *root, struct nfs_unique_id *id)
189 rb_erase(&id->rb_node, root);
192 static struct nfs4_state_owner *
193 nfs4_find_state_owner(struct nfs_server *server, struct rpc_cred *cred)
195 struct nfs_client *clp = server->nfs_client;
196 struct rb_node **p = &clp->cl_state_owners.rb_node,
198 struct nfs4_state_owner *sp, *res = NULL;
202 sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
204 if (server < sp->so_server) {
205 p = &parent->rb_left;
208 if (server > sp->so_server) {
209 p = &parent->rb_right;
212 if (cred < sp->so_cred)
213 p = &parent->rb_left;
214 else if (cred > sp->so_cred)
215 p = &parent->rb_right;
217 atomic_inc(&sp->so_count);
225 static struct nfs4_state_owner *
226 nfs4_insert_state_owner(struct nfs_client *clp, struct nfs4_state_owner *new)
228 struct rb_node **p = &clp->cl_state_owners.rb_node,
230 struct nfs4_state_owner *sp;
234 sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
236 if (new->so_server < sp->so_server) {
237 p = &parent->rb_left;
240 if (new->so_server > sp->so_server) {
241 p = &parent->rb_right;
244 if (new->so_cred < sp->so_cred)
245 p = &parent->rb_left;
246 else if (new->so_cred > sp->so_cred)
247 p = &parent->rb_right;
249 atomic_inc(&sp->so_count);
253 nfs_alloc_unique_id(&clp->cl_openowner_id, &new->so_owner_id, 1, 64);
254 rb_link_node(&new->so_client_node, parent, p);
255 rb_insert_color(&new->so_client_node, &clp->cl_state_owners);
260 nfs4_remove_state_owner(struct nfs_client *clp, struct nfs4_state_owner *sp)
262 if (!RB_EMPTY_NODE(&sp->so_client_node))
263 rb_erase(&sp->so_client_node, &clp->cl_state_owners);
264 nfs_free_unique_id(&clp->cl_openowner_id, &sp->so_owner_id);
268 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
269 * create a new state_owner.
272 static struct nfs4_state_owner *
273 nfs4_alloc_state_owner(void)
275 struct nfs4_state_owner *sp;
277 sp = kzalloc(sizeof(*sp),GFP_KERNEL);
280 spin_lock_init(&sp->so_lock);
281 INIT_LIST_HEAD(&sp->so_states);
282 INIT_LIST_HEAD(&sp->so_delegations);
283 rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue");
284 sp->so_seqid.sequence = &sp->so_sequence;
285 spin_lock_init(&sp->so_sequence.lock);
286 INIT_LIST_HEAD(&sp->so_sequence.list);
287 atomic_set(&sp->so_count, 1);
292 nfs4_drop_state_owner(struct nfs4_state_owner *sp)
294 if (!RB_EMPTY_NODE(&sp->so_client_node)) {
295 struct nfs_client *clp = sp->so_client;
297 spin_lock(&clp->cl_lock);
298 rb_erase(&sp->so_client_node, &clp->cl_state_owners);
299 RB_CLEAR_NODE(&sp->so_client_node);
300 spin_unlock(&clp->cl_lock);
304 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct rpc_cred *cred)
306 struct nfs_client *clp = server->nfs_client;
307 struct nfs4_state_owner *sp, *new;
309 spin_lock(&clp->cl_lock);
310 sp = nfs4_find_state_owner(server, cred);
311 spin_unlock(&clp->cl_lock);
314 new = nfs4_alloc_state_owner();
317 new->so_client = clp;
318 new->so_server = server;
320 spin_lock(&clp->cl_lock);
321 sp = nfs4_insert_state_owner(clp, new);
322 spin_unlock(&clp->cl_lock);
326 rpc_destroy_wait_queue(&new->so_sequence.wait);
332 void nfs4_put_state_owner(struct nfs4_state_owner *sp)
334 struct nfs_client *clp = sp->so_client;
335 struct rpc_cred *cred = sp->so_cred;
337 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
339 nfs4_remove_state_owner(clp, sp);
340 spin_unlock(&clp->cl_lock);
341 rpc_destroy_wait_queue(&sp->so_sequence.wait);
346 static struct nfs4_state *
347 nfs4_alloc_open_state(void)
349 struct nfs4_state *state;
351 state = kzalloc(sizeof(*state), GFP_KERNEL);
354 atomic_set(&state->count, 1);
355 INIT_LIST_HEAD(&state->lock_states);
356 spin_lock_init(&state->state_lock);
357 seqlock_init(&state->seqlock);
362 nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
364 if (state->state == fmode)
366 /* NB! List reordering - see the reclaim code for why. */
367 if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
368 if (fmode & FMODE_WRITE)
369 list_move(&state->open_states, &state->owner->so_states);
371 list_move_tail(&state->open_states, &state->owner->so_states);
373 state->state = fmode;
376 static struct nfs4_state *
377 __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
379 struct nfs_inode *nfsi = NFS_I(inode);
380 struct nfs4_state *state;
382 list_for_each_entry(state, &nfsi->open_states, inode_states) {
383 if (state->owner != owner)
385 if (atomic_inc_not_zero(&state->count))
392 nfs4_free_open_state(struct nfs4_state *state)
398 nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
400 struct nfs4_state *state, *new;
401 struct nfs_inode *nfsi = NFS_I(inode);
403 spin_lock(&inode->i_lock);
404 state = __nfs4_find_state_byowner(inode, owner);
405 spin_unlock(&inode->i_lock);
408 new = nfs4_alloc_open_state();
409 spin_lock(&owner->so_lock);
410 spin_lock(&inode->i_lock);
411 state = __nfs4_find_state_byowner(inode, owner);
412 if (state == NULL && new != NULL) {
414 state->owner = owner;
415 atomic_inc(&owner->so_count);
416 list_add(&state->inode_states, &nfsi->open_states);
417 state->inode = igrab(inode);
418 spin_unlock(&inode->i_lock);
419 /* Note: The reclaim code dictates that we add stateless
420 * and read-only stateids to the end of the list */
421 list_add_tail(&state->open_states, &owner->so_states);
422 spin_unlock(&owner->so_lock);
424 spin_unlock(&inode->i_lock);
425 spin_unlock(&owner->so_lock);
427 nfs4_free_open_state(new);
433 void nfs4_put_open_state(struct nfs4_state *state)
435 struct inode *inode = state->inode;
436 struct nfs4_state_owner *owner = state->owner;
438 if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
440 spin_lock(&inode->i_lock);
441 list_del(&state->inode_states);
442 list_del(&state->open_states);
443 spin_unlock(&inode->i_lock);
444 spin_unlock(&owner->so_lock);
446 nfs4_free_open_state(state);
447 nfs4_put_state_owner(owner);
451 * Close the current file.
453 static void __nfs4_close(struct path *path, struct nfs4_state *state, fmode_t fmode, int wait)
455 struct nfs4_state_owner *owner = state->owner;
459 atomic_inc(&owner->so_count);
460 /* Protect against nfs4_find_state() */
461 spin_lock(&owner->so_lock);
462 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
469 case FMODE_READ|FMODE_WRITE:
472 newstate = FMODE_READ|FMODE_WRITE;
473 if (state->n_rdwr == 0) {
474 if (state->n_rdonly == 0) {
475 newstate &= ~FMODE_READ;
476 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
477 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
479 if (state->n_wronly == 0) {
480 newstate &= ~FMODE_WRITE;
481 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
482 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
485 clear_bit(NFS_DELEGATED_STATE, &state->flags);
487 nfs4_state_set_mode_locked(state, newstate);
488 spin_unlock(&owner->so_lock);
491 nfs4_put_open_state(state);
492 nfs4_put_state_owner(owner);
494 nfs4_do_close(path, state, wait);
497 void nfs4_close_state(struct path *path, struct nfs4_state *state, fmode_t fmode)
499 __nfs4_close(path, state, fmode, 0);
502 void nfs4_close_sync(struct path *path, struct nfs4_state *state, fmode_t fmode)
504 __nfs4_close(path, state, fmode, 1);
508 * Search the state->lock_states for an existing lock_owner
509 * that is compatible with current->files
511 static struct nfs4_lock_state *
512 __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
514 struct nfs4_lock_state *pos;
515 list_for_each_entry(pos, &state->lock_states, ls_locks) {
516 if (pos->ls_owner != fl_owner)
518 atomic_inc(&pos->ls_count);
525 * Return a compatible lock_state. If no initialized lock_state structure
526 * exists, return an uninitialized one.
529 static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
531 struct nfs4_lock_state *lsp;
532 struct nfs_client *clp = state->owner->so_client;
534 lsp = kzalloc(sizeof(*lsp), GFP_KERNEL);
537 rpc_init_wait_queue(&lsp->ls_sequence.wait, "lock_seqid_waitqueue");
538 spin_lock_init(&lsp->ls_sequence.lock);
539 INIT_LIST_HEAD(&lsp->ls_sequence.list);
540 lsp->ls_seqid.sequence = &lsp->ls_sequence;
541 atomic_set(&lsp->ls_count, 1);
542 lsp->ls_owner = fl_owner;
543 spin_lock(&clp->cl_lock);
544 nfs_alloc_unique_id(&clp->cl_lockowner_id, &lsp->ls_id, 1, 64);
545 spin_unlock(&clp->cl_lock);
546 INIT_LIST_HEAD(&lsp->ls_locks);
550 static void nfs4_free_lock_state(struct nfs4_lock_state *lsp)
552 struct nfs_client *clp = lsp->ls_state->owner->so_client;
554 spin_lock(&clp->cl_lock);
555 nfs_free_unique_id(&clp->cl_lockowner_id, &lsp->ls_id);
556 spin_unlock(&clp->cl_lock);
557 rpc_destroy_wait_queue(&lsp->ls_sequence.wait);
562 * Return a compatible lock_state. If no initialized lock_state structure
563 * exists, return an uninitialized one.
566 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
568 struct nfs4_lock_state *lsp, *new = NULL;
571 spin_lock(&state->state_lock);
572 lsp = __nfs4_find_lock_state(state, owner);
576 new->ls_state = state;
577 list_add(&new->ls_locks, &state->lock_states);
578 set_bit(LK_STATE_IN_USE, &state->flags);
583 spin_unlock(&state->state_lock);
584 new = nfs4_alloc_lock_state(state, owner);
588 spin_unlock(&state->state_lock);
590 nfs4_free_lock_state(new);
595 * Release reference to lock_state, and free it if we see that
596 * it is no longer in use
598 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
600 struct nfs4_state *state;
604 state = lsp->ls_state;
605 if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
607 list_del(&lsp->ls_locks);
608 if (list_empty(&state->lock_states))
609 clear_bit(LK_STATE_IN_USE, &state->flags);
610 spin_unlock(&state->state_lock);
611 nfs4_free_lock_state(lsp);
614 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
616 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
618 dst->fl_u.nfs4_fl.owner = lsp;
619 atomic_inc(&lsp->ls_count);
622 static void nfs4_fl_release_lock(struct file_lock *fl)
624 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
627 static struct file_lock_operations nfs4_fl_lock_ops = {
628 .fl_copy_lock = nfs4_fl_copy_lock,
629 .fl_release_private = nfs4_fl_release_lock,
632 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
634 struct nfs4_lock_state *lsp;
636 if (fl->fl_ops != NULL)
638 lsp = nfs4_get_lock_state(state, fl->fl_owner);
641 fl->fl_u.nfs4_fl.owner = lsp;
642 fl->fl_ops = &nfs4_fl_lock_ops;
647 * Byte-range lock aware utility to initialize the stateid of read/write
650 void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner)
652 struct nfs4_lock_state *lsp;
656 seq = read_seqbegin(&state->seqlock);
657 memcpy(dst, &state->stateid, sizeof(*dst));
658 } while (read_seqretry(&state->seqlock, seq));
659 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
662 spin_lock(&state->state_lock);
663 lsp = __nfs4_find_lock_state(state, fl_owner);
664 if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
665 memcpy(dst, &lsp->ls_stateid, sizeof(*dst));
666 spin_unlock(&state->state_lock);
667 nfs4_put_lock_state(lsp);
670 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter)
672 struct nfs_seqid *new;
674 new = kmalloc(sizeof(*new), GFP_KERNEL);
676 new->sequence = counter;
677 INIT_LIST_HEAD(&new->list);
682 void nfs_free_seqid(struct nfs_seqid *seqid)
684 if (!list_empty(&seqid->list)) {
685 struct rpc_sequence *sequence = seqid->sequence->sequence;
687 spin_lock(&sequence->lock);
688 list_del(&seqid->list);
689 spin_unlock(&sequence->lock);
690 rpc_wake_up(&sequence->wait);
696 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
697 * failed with a seqid incrementing error -
698 * see comments nfs_fs.h:seqid_mutating_error()
700 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
702 BUG_ON(list_first_entry(&seqid->sequence->sequence->list, struct nfs_seqid, list) != seqid);
706 case -NFS4ERR_BAD_SEQID:
707 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
709 printk(KERN_WARNING "NFS: v4 server returned a bad"
710 " sequence-id error on an"
711 " unconfirmed sequence %p!\n",
713 case -NFS4ERR_STALE_CLIENTID:
714 case -NFS4ERR_STALE_STATEID:
715 case -NFS4ERR_BAD_STATEID:
716 case -NFS4ERR_BADXDR:
717 case -NFS4ERR_RESOURCE:
718 case -NFS4ERR_NOFILEHANDLE:
719 /* Non-seqid mutating errors */
723 * Note: no locking needed as we are guaranteed to be first
724 * on the sequence list
726 seqid->sequence->counter++;
729 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
731 if (status == -NFS4ERR_BAD_SEQID) {
732 struct nfs4_state_owner *sp = container_of(seqid->sequence,
733 struct nfs4_state_owner, so_seqid);
734 nfs4_drop_state_owner(sp);
736 nfs_increment_seqid(status, seqid);
740 * Increment the seqid if the LOCK/LOCKU succeeded, or
741 * failed with a seqid incrementing error -
742 * see comments nfs_fs.h:seqid_mutating_error()
744 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
746 nfs_increment_seqid(status, seqid);
749 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
751 struct rpc_sequence *sequence = seqid->sequence->sequence;
754 spin_lock(&sequence->lock);
755 if (list_empty(&seqid->list))
756 list_add_tail(&seqid->list, &sequence->list);
757 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
759 rpc_sleep_on(&sequence->wait, task, NULL);
762 spin_unlock(&sequence->lock);
766 static int nfs4_run_state_manager(void *);
768 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
770 smp_mb__before_clear_bit();
771 clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
772 smp_mb__after_clear_bit();
773 wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
774 rpc_wake_up(&clp->cl_rpcwaitq);
778 * Schedule the nfs_client asynchronous state management routine
780 void nfs4_schedule_state_manager(struct nfs_client *clp)
782 struct task_struct *task;
784 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
786 __module_get(THIS_MODULE);
787 atomic_inc(&clp->cl_count);
788 task = kthread_run(nfs4_run_state_manager, clp, "%s-manager",
789 rpc_peeraddr2str(clp->cl_rpcclient,
793 nfs4_clear_state_manager_bit(clp);
795 module_put(THIS_MODULE);
799 * Schedule a state recovery attempt
801 void nfs4_schedule_state_recovery(struct nfs_client *clp)
805 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
806 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
807 nfs4_schedule_state_manager(clp);
810 static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
813 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
814 /* Don't recover state that expired before the reboot */
815 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
816 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
819 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
820 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
824 int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
826 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
827 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
828 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
829 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
833 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
835 struct inode *inode = state->inode;
836 struct nfs_inode *nfsi = NFS_I(inode);
837 struct file_lock *fl;
840 down_write(&nfsi->rwsem);
841 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
842 if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
844 if (nfs_file_open_context(fl->fl_file)->state != state)
846 status = ops->recover_lock(state, fl);
851 printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
853 case -NFS4ERR_EXPIRED:
854 case -NFS4ERR_NO_GRACE:
855 case -NFS4ERR_RECLAIM_BAD:
856 case -NFS4ERR_RECLAIM_CONFLICT:
857 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
859 case -NFS4ERR_STALE_CLIENTID:
863 up_write(&nfsi->rwsem);
866 up_write(&nfsi->rwsem);
870 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
872 struct nfs4_state *state;
873 struct nfs4_lock_state *lock;
876 /* Note: we rely on the sp->so_states list being ordered
877 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
879 * This is needed to ensure that the server won't give us any
880 * read delegations that we have to return if, say, we are
881 * recovering after a network partition or a reboot from a
882 * server that doesn't support a grace period.
885 spin_lock(&sp->so_lock);
886 list_for_each_entry(state, &sp->so_states, open_states) {
887 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
889 if (state->state == 0)
891 atomic_inc(&state->count);
892 spin_unlock(&sp->so_lock);
893 status = ops->recover_open(sp, state);
895 status = nfs4_reclaim_locks(state, ops);
897 list_for_each_entry(lock, &state->lock_states, ls_locks) {
898 if (!(lock->ls_flags & NFS_LOCK_INITIALIZED))
899 printk("%s: Lock reclaim failed!\n",
902 nfs4_put_open_state(state);
908 printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
913 * Open state on this file cannot be recovered
914 * All we can do is revert to using the zero stateid.
916 memset(state->stateid.data, 0,
917 sizeof(state->stateid.data));
918 /* Mark the file as being 'closed' */
921 case -NFS4ERR_RECLAIM_BAD:
922 case -NFS4ERR_RECLAIM_CONFLICT:
923 nfs4_state_mark_reclaim_nograce(sp->so_client, state);
925 case -NFS4ERR_EXPIRED:
926 case -NFS4ERR_NO_GRACE:
927 nfs4_state_mark_reclaim_nograce(sp->so_client, state);
928 case -NFS4ERR_STALE_CLIENTID:
931 nfs4_put_open_state(state);
934 spin_unlock(&sp->so_lock);
937 nfs4_put_open_state(state);
941 static void nfs4_clear_open_state(struct nfs4_state *state)
943 struct nfs4_lock_state *lock;
945 clear_bit(NFS_DELEGATED_STATE, &state->flags);
946 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
947 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
948 clear_bit(NFS_O_RDWR_STATE, &state->flags);
949 list_for_each_entry(lock, &state->lock_states, ls_locks) {
950 lock->ls_seqid.flags = 0;
951 lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
955 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp, int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
957 struct nfs4_state_owner *sp;
959 struct nfs4_state *state;
961 /* Reset all sequence ids to zero */
962 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
963 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
964 sp->so_seqid.flags = 0;
965 spin_lock(&sp->so_lock);
966 list_for_each_entry(state, &sp->so_states, open_states) {
967 if (mark_reclaim(clp, state))
968 nfs4_clear_open_state(state);
970 spin_unlock(&sp->so_lock);
974 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
976 /* Mark all delegations for reclaim */
977 nfs_delegation_mark_reclaim(clp);
978 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
981 static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
983 struct nfs4_state_owner *sp;
985 struct nfs4_state *state;
987 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
990 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
991 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
992 spin_lock(&sp->so_lock);
993 list_for_each_entry(state, &sp->so_states, open_states) {
994 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags))
996 nfs4_state_mark_reclaim_nograce(clp, state);
998 spin_unlock(&sp->so_lock);
1001 nfs_delegation_reap_unclaimed(clp);
1004 static void nfs_delegation_clear_all(struct nfs_client *clp)
1006 nfs_delegation_mark_reclaim(clp);
1007 nfs_delegation_reap_unclaimed(clp);
1010 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1012 nfs_delegation_clear_all(clp);
1013 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1016 static void nfs4_state_end_reclaim_nograce(struct nfs_client *clp)
1018 clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
1021 static void nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1024 case -NFS4ERR_CB_PATH_DOWN:
1025 nfs_handle_cb_pathdown(clp);
1027 case -NFS4ERR_STALE_CLIENTID:
1028 case -NFS4ERR_LEASE_MOVED:
1029 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1030 nfs4_state_start_reclaim_reboot(clp);
1032 case -NFS4ERR_EXPIRED:
1033 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1034 nfs4_state_start_reclaim_nograce(clp);
1035 case -NFS4ERR_BADSESSION:
1036 case -NFS4ERR_BADSLOT:
1037 case -NFS4ERR_BAD_HIGH_SLOT:
1038 case -NFS4ERR_DEADSESSION:
1039 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1040 case -NFS4ERR_SEQ_FALSE_RETRY:
1041 case -NFS4ERR_SEQ_MISORDERED:
1042 set_bit(NFS4CLNT_SESSION_SETUP, &clp->cl_state);
1046 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1048 struct rb_node *pos;
1052 spin_lock(&clp->cl_lock);
1053 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1054 struct nfs4_state_owner *sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1055 if (!test_and_clear_bit(ops->owner_flag_bit, &sp->so_flags))
1057 atomic_inc(&sp->so_count);
1058 spin_unlock(&clp->cl_lock);
1059 status = nfs4_reclaim_open_state(sp, ops);
1061 set_bit(ops->owner_flag_bit, &sp->so_flags);
1062 nfs4_put_state_owner(sp);
1063 nfs4_recovery_handle_error(clp, status);
1066 nfs4_put_state_owner(sp);
1069 spin_unlock(&clp->cl_lock);
1073 static int nfs4_check_lease(struct nfs_client *clp)
1075 struct rpc_cred *cred;
1076 struct nfs4_state_maintenance_ops *ops =
1077 nfs4_state_renewal_ops[clp->cl_minorversion];
1078 int status = -NFS4ERR_EXPIRED;
1080 /* Is the client already known to have an expired lease? */
1081 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1083 spin_lock(&clp->cl_lock);
1084 cred = ops->get_state_renewal_cred_locked(clp);
1085 spin_unlock(&clp->cl_lock);
1087 cred = nfs4_get_setclientid_cred(clp);
1091 status = ops->renew_lease(clp, cred);
1094 nfs4_recovery_handle_error(clp, status);
1098 static int nfs4_reclaim_lease(struct nfs_client *clp)
1100 struct rpc_cred *cred;
1101 int status = -ENOENT;
1103 cred = nfs4_get_setclientid_cred(clp);
1105 status = nfs4_init_client(clp, cred);
1107 /* Handle case where the user hasn't set up machine creds */
1108 if (status == -EACCES && cred == clp->cl_machine_cred) {
1109 nfs4_clear_machine_cred(clp);
1112 if (status == -NFS4ERR_MINOR_VERS_MISMATCH)
1113 status = -EPROTONOSUPPORT;
1118 #ifdef CONFIG_NFS_V4_1
1119 static void nfs4_session_recovery_handle_error(struct nfs_client *clp, int err)
1122 case -NFS4ERR_STALE_CLIENTID:
1123 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1124 set_bit(NFS4CLNT_SESSION_SETUP, &clp->cl_state);
1128 static int nfs4_reset_session(struct nfs_client *clp)
1132 status = nfs4_proc_destroy_session(clp->cl_session);
1133 if (status && status != -NFS4ERR_BADSESSION &&
1134 status != -NFS4ERR_DEADSESSION) {
1135 nfs4_session_recovery_handle_error(clp, status);
1139 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
1140 status = nfs4_proc_create_session(clp, 1);
1142 nfs4_session_recovery_handle_error(clp, status);
1145 /* Wake up the next rpc task even on error */
1146 rpc_wake_up_next(&clp->cl_session->fc_slot_table.slot_tbl_waitq);
1150 static int nfs4_initialize_session(struct nfs_client *clp)
1154 status = nfs4_proc_create_session(clp, 0);
1156 nfs_mark_client_ready(clp, NFS_CS_READY);
1157 } else if (status == -NFS4ERR_STALE_CLIENTID) {
1158 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1159 set_bit(NFS4CLNT_SESSION_SETUP, &clp->cl_state);
1161 nfs_mark_client_ready(clp, status);
1165 #else /* CONFIG_NFS_V4_1 */
1166 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
1167 static int nfs4_initialize_session(struct nfs_client *clp) { return 0; }
1168 #endif /* CONFIG_NFS_V4_1 */
1170 static void nfs4_state_manager(struct nfs_client *clp)
1174 /* Ensure exclusive access to NFSv4 state */
1176 if (test_and_clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
1177 /* We're going to have to re-establish a clientid */
1178 status = nfs4_reclaim_lease(clp);
1180 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1181 if (status == -EAGAIN)
1183 if (clp->cl_cons_state ==
1184 NFS_CS_SESSION_INITING)
1185 nfs_mark_client_ready(clp, status);
1188 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1191 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
1192 status = nfs4_check_lease(clp);
1196 /* Initialize or reset the session */
1197 if (nfs4_has_session(clp) &&
1198 test_and_clear_bit(NFS4CLNT_SESSION_SETUP, &clp->cl_state)) {
1199 if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
1200 status = nfs4_initialize_session(clp);
1202 status = nfs4_reset_session(clp);
1204 if (status == -NFS4ERR_STALE_CLIENTID)
1209 /* First recover reboot state... */
1210 if (test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
1211 status = nfs4_do_reclaim(clp, &nfs4_reboot_recovery_ops);
1212 if (status == -NFS4ERR_STALE_CLIENTID)
1214 if (test_bit(NFS4CLNT_SESSION_SETUP, &clp->cl_state))
1216 nfs4_state_end_reclaim_reboot(clp);
1220 /* Now recover expired state... */
1221 if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
1222 status = nfs4_do_reclaim(clp, &nfs4_nograce_recovery_ops);
1224 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
1225 if (status == -NFS4ERR_STALE_CLIENTID)
1227 if (status == -NFS4ERR_EXPIRED)
1229 if (test_bit(NFS4CLNT_SESSION_SETUP,
1234 nfs4_state_end_reclaim_nograce(clp);
1238 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
1239 nfs_client_return_marked_delegations(clp);
1243 nfs4_clear_state_manager_bit(clp);
1244 /* Did we race with an attempt to give us more work? */
1245 if (clp->cl_state == 0)
1247 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1252 printk(KERN_WARNING "Error: state manager failed on NFSv4 server %s"
1253 " with error %d\n", clp->cl_hostname, -status);
1254 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1255 nfs4_state_end_reclaim_reboot(clp);
1256 nfs4_clear_state_manager_bit(clp);
1259 static int nfs4_run_state_manager(void *ptr)
1261 struct nfs_client *clp = ptr;
1263 allow_signal(SIGKILL);
1264 nfs4_state_manager(clp);
1265 nfs_put_client(clp);
1266 module_put_and_exit(0);