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)
65 struct nfs4_setclientid_res clid;
69 port = nfs_callback_tcpport;
70 if (clp->cl_addr.ss_family == AF_INET6)
71 port = nfs_callback_tcpport6;
73 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
76 status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
79 clp->cl_clientid = clid.clientid;
80 nfs4_schedule_state_renewal(clp);
85 struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
87 struct rpc_cred *cred = NULL;
89 if (clp->cl_machine_cred != NULL)
90 cred = get_rpccred(clp->cl_machine_cred);
94 static void nfs4_clear_machine_cred(struct nfs_client *clp)
96 struct rpc_cred *cred;
98 spin_lock(&clp->cl_lock);
99 cred = clp->cl_machine_cred;
100 clp->cl_machine_cred = NULL;
101 spin_unlock(&clp->cl_lock);
106 struct rpc_cred *nfs4_get_renew_cred_locked(struct nfs_client *clp)
108 struct nfs4_state_owner *sp;
110 struct rpc_cred *cred = NULL;
112 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
113 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
114 if (list_empty(&sp->so_states))
116 cred = get_rpccred(sp->so_cred);
122 #if defined(CONFIG_NFS_V4_1)
124 static int nfs41_setup_state_renewal(struct nfs_client *clp)
127 struct nfs_fsinfo fsinfo;
129 status = nfs4_proc_get_lease_time(clp, &fsinfo);
131 /* Update lease time and schedule renewal */
132 spin_lock(&clp->cl_lock);
133 clp->cl_lease_time = fsinfo.lease_time * HZ;
134 clp->cl_last_renewal = jiffies;
135 spin_unlock(&clp->cl_lock);
137 nfs4_schedule_state_renewal(clp);
143 static void nfs4_end_drain_session(struct nfs_client *clp)
145 struct nfs4_session *ses = clp->cl_session;
150 if (test_and_clear_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
151 spin_lock(&ses->fc_slot_table.slot_tbl_lock);
152 max_slots = ses->fc_slot_table.max_slots;
153 while (max_slots--) {
154 struct rpc_task *task;
156 task = rpc_wake_up_next(&ses->fc_slot_table.
160 rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
162 spin_unlock(&ses->fc_slot_table.slot_tbl_lock);
166 static int nfs4_begin_drain_session(struct nfs_client *clp)
168 struct nfs4_session *ses = clp->cl_session;
169 struct nfs4_slot_table *tbl = &ses->fc_slot_table;
171 spin_lock(&tbl->slot_tbl_lock);
172 set_bit(NFS4_SESSION_DRAINING, &ses->session_state);
173 if (tbl->highest_used_slotid != -1) {
174 INIT_COMPLETION(ses->complete);
175 spin_unlock(&tbl->slot_tbl_lock);
176 return wait_for_completion_interruptible(&ses->complete);
178 spin_unlock(&tbl->slot_tbl_lock);
182 int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
186 nfs4_begin_drain_session(clp);
187 status = nfs4_proc_exchange_id(clp, cred);
190 status = nfs4_proc_create_session(clp);
193 nfs41_setup_state_renewal(clp);
194 nfs_mark_client_ready(clp, NFS_CS_READY);
199 struct rpc_cred *nfs4_get_exchange_id_cred(struct nfs_client *clp)
201 struct rpc_cred *cred;
203 spin_lock(&clp->cl_lock);
204 cred = nfs4_get_machine_cred_locked(clp);
205 spin_unlock(&clp->cl_lock);
209 #endif /* CONFIG_NFS_V4_1 */
211 struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
213 struct nfs4_state_owner *sp;
215 struct rpc_cred *cred;
217 spin_lock(&clp->cl_lock);
218 cred = nfs4_get_machine_cred_locked(clp);
221 pos = rb_first(&clp->cl_state_owners);
223 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
224 cred = get_rpccred(sp->so_cred);
227 spin_unlock(&clp->cl_lock);
231 static void nfs_alloc_unique_id(struct rb_root *root, struct nfs_unique_id *new,
232 __u64 minval, int maxbits)
234 struct rb_node **p, *parent;
235 struct nfs_unique_id *pos;
239 mask = (1ULL << maxbits) - 1ULL;
241 /* Ensure distribution is more or less flat */
242 get_random_bytes(&new->id, sizeof(new->id));
244 if (new->id < minval)
252 pos = rb_entry(parent, struct nfs_unique_id, rb_node);
254 if (new->id < pos->id)
256 else if (new->id > pos->id)
261 rb_link_node(&new->rb_node, parent, p);
262 rb_insert_color(&new->rb_node, root);
267 if (new->id < minval || (new->id & mask) != new->id) {
271 parent = rb_next(parent);
274 pos = rb_entry(parent, struct nfs_unique_id, rb_node);
275 if (new->id < pos->id)
281 static void nfs_free_unique_id(struct rb_root *root, struct nfs_unique_id *id)
283 rb_erase(&id->rb_node, root);
286 static struct nfs4_state_owner *
287 nfs4_find_state_owner(struct nfs_server *server, struct rpc_cred *cred)
289 struct nfs_client *clp = server->nfs_client;
290 struct rb_node **p = &clp->cl_state_owners.rb_node,
292 struct nfs4_state_owner *sp, *res = NULL;
296 sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
298 if (server < sp->so_server) {
299 p = &parent->rb_left;
302 if (server > sp->so_server) {
303 p = &parent->rb_right;
306 if (cred < sp->so_cred)
307 p = &parent->rb_left;
308 else if (cred > sp->so_cred)
309 p = &parent->rb_right;
311 atomic_inc(&sp->so_count);
319 static struct nfs4_state_owner *
320 nfs4_insert_state_owner(struct nfs_client *clp, struct nfs4_state_owner *new)
322 struct rb_node **p = &clp->cl_state_owners.rb_node,
324 struct nfs4_state_owner *sp;
328 sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
330 if (new->so_server < sp->so_server) {
331 p = &parent->rb_left;
334 if (new->so_server > sp->so_server) {
335 p = &parent->rb_right;
338 if (new->so_cred < sp->so_cred)
339 p = &parent->rb_left;
340 else if (new->so_cred > sp->so_cred)
341 p = &parent->rb_right;
343 atomic_inc(&sp->so_count);
347 nfs_alloc_unique_id(&clp->cl_openowner_id, &new->so_owner_id, 1, 64);
348 rb_link_node(&new->so_client_node, parent, p);
349 rb_insert_color(&new->so_client_node, &clp->cl_state_owners);
354 nfs4_remove_state_owner(struct nfs_client *clp, struct nfs4_state_owner *sp)
356 if (!RB_EMPTY_NODE(&sp->so_client_node))
357 rb_erase(&sp->so_client_node, &clp->cl_state_owners);
358 nfs_free_unique_id(&clp->cl_openowner_id, &sp->so_owner_id);
362 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
363 * create a new state_owner.
366 static struct nfs4_state_owner *
367 nfs4_alloc_state_owner(void)
369 struct nfs4_state_owner *sp;
371 sp = kzalloc(sizeof(*sp),GFP_NOFS);
374 spin_lock_init(&sp->so_lock);
375 INIT_LIST_HEAD(&sp->so_states);
376 rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue");
377 sp->so_seqid.sequence = &sp->so_sequence;
378 spin_lock_init(&sp->so_sequence.lock);
379 INIT_LIST_HEAD(&sp->so_sequence.list);
380 atomic_set(&sp->so_count, 1);
385 nfs4_drop_state_owner(struct nfs4_state_owner *sp)
387 if (!RB_EMPTY_NODE(&sp->so_client_node)) {
388 struct nfs_client *clp = sp->so_server->nfs_client;
390 spin_lock(&clp->cl_lock);
391 rb_erase(&sp->so_client_node, &clp->cl_state_owners);
392 RB_CLEAR_NODE(&sp->so_client_node);
393 spin_unlock(&clp->cl_lock);
397 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct rpc_cred *cred)
399 struct nfs_client *clp = server->nfs_client;
400 struct nfs4_state_owner *sp, *new;
402 spin_lock(&clp->cl_lock);
403 sp = nfs4_find_state_owner(server, cred);
404 spin_unlock(&clp->cl_lock);
407 new = nfs4_alloc_state_owner();
410 new->so_server = server;
412 spin_lock(&clp->cl_lock);
413 sp = nfs4_insert_state_owner(clp, new);
414 spin_unlock(&clp->cl_lock);
418 rpc_destroy_wait_queue(&new->so_sequence.wait);
424 void nfs4_put_state_owner(struct nfs4_state_owner *sp)
426 struct nfs_client *clp = sp->so_server->nfs_client;
427 struct rpc_cred *cred = sp->so_cred;
429 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
431 nfs4_remove_state_owner(clp, sp);
432 spin_unlock(&clp->cl_lock);
433 rpc_destroy_wait_queue(&sp->so_sequence.wait);
438 static struct nfs4_state *
439 nfs4_alloc_open_state(void)
441 struct nfs4_state *state;
443 state = kzalloc(sizeof(*state), GFP_NOFS);
446 atomic_set(&state->count, 1);
447 INIT_LIST_HEAD(&state->lock_states);
448 spin_lock_init(&state->state_lock);
449 seqlock_init(&state->seqlock);
454 nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
456 if (state->state == fmode)
458 /* NB! List reordering - see the reclaim code for why. */
459 if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
460 if (fmode & FMODE_WRITE)
461 list_move(&state->open_states, &state->owner->so_states);
463 list_move_tail(&state->open_states, &state->owner->so_states);
465 state->state = fmode;
468 static struct nfs4_state *
469 __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
471 struct nfs_inode *nfsi = NFS_I(inode);
472 struct nfs4_state *state;
474 list_for_each_entry(state, &nfsi->open_states, inode_states) {
475 if (state->owner != owner)
477 if (atomic_inc_not_zero(&state->count))
484 nfs4_free_open_state(struct nfs4_state *state)
490 nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
492 struct nfs4_state *state, *new;
493 struct nfs_inode *nfsi = NFS_I(inode);
495 spin_lock(&inode->i_lock);
496 state = __nfs4_find_state_byowner(inode, owner);
497 spin_unlock(&inode->i_lock);
500 new = nfs4_alloc_open_state();
501 spin_lock(&owner->so_lock);
502 spin_lock(&inode->i_lock);
503 state = __nfs4_find_state_byowner(inode, owner);
504 if (state == NULL && new != NULL) {
506 state->owner = owner;
507 atomic_inc(&owner->so_count);
508 list_add(&state->inode_states, &nfsi->open_states);
509 state->inode = igrab(inode);
510 spin_unlock(&inode->i_lock);
511 /* Note: The reclaim code dictates that we add stateless
512 * and read-only stateids to the end of the list */
513 list_add_tail(&state->open_states, &owner->so_states);
514 spin_unlock(&owner->so_lock);
516 spin_unlock(&inode->i_lock);
517 spin_unlock(&owner->so_lock);
519 nfs4_free_open_state(new);
525 void nfs4_put_open_state(struct nfs4_state *state)
527 struct inode *inode = state->inode;
528 struct nfs4_state_owner *owner = state->owner;
530 if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
532 spin_lock(&inode->i_lock);
533 list_del(&state->inode_states);
534 list_del(&state->open_states);
535 spin_unlock(&inode->i_lock);
536 spin_unlock(&owner->so_lock);
538 nfs4_free_open_state(state);
539 nfs4_put_state_owner(owner);
543 * Close the current file.
545 static void __nfs4_close(struct path *path, struct nfs4_state *state,
546 fmode_t fmode, gfp_t gfp_mask, int wait)
548 struct nfs4_state_owner *owner = state->owner;
552 atomic_inc(&owner->so_count);
553 /* Protect against nfs4_find_state() */
554 spin_lock(&owner->so_lock);
555 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
562 case FMODE_READ|FMODE_WRITE:
565 newstate = FMODE_READ|FMODE_WRITE;
566 if (state->n_rdwr == 0) {
567 if (state->n_rdonly == 0) {
568 newstate &= ~FMODE_READ;
569 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
570 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
572 if (state->n_wronly == 0) {
573 newstate &= ~FMODE_WRITE;
574 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
575 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
578 clear_bit(NFS_DELEGATED_STATE, &state->flags);
580 nfs4_state_set_mode_locked(state, newstate);
581 spin_unlock(&owner->so_lock);
584 nfs4_put_open_state(state);
585 nfs4_put_state_owner(owner);
587 nfs4_do_close(path, state, gfp_mask, wait);
590 void nfs4_close_state(struct path *path, struct nfs4_state *state, fmode_t fmode)
592 __nfs4_close(path, state, fmode, GFP_NOFS, 0);
595 void nfs4_close_sync(struct path *path, struct nfs4_state *state, fmode_t fmode)
597 __nfs4_close(path, state, fmode, GFP_KERNEL, 1);
601 * Search the state->lock_states for an existing lock_owner
602 * that is compatible with current->files
604 static struct nfs4_lock_state *
605 __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
607 struct nfs4_lock_state *pos;
608 list_for_each_entry(pos, &state->lock_states, ls_locks) {
609 if (pos->ls_owner != fl_owner)
611 atomic_inc(&pos->ls_count);
618 * Return a compatible lock_state. If no initialized lock_state structure
619 * exists, return an uninitialized one.
622 static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
624 struct nfs4_lock_state *lsp;
625 struct nfs_client *clp = state->owner->so_server->nfs_client;
627 lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
630 rpc_init_wait_queue(&lsp->ls_sequence.wait, "lock_seqid_waitqueue");
631 spin_lock_init(&lsp->ls_sequence.lock);
632 INIT_LIST_HEAD(&lsp->ls_sequence.list);
633 lsp->ls_seqid.sequence = &lsp->ls_sequence;
634 atomic_set(&lsp->ls_count, 1);
635 lsp->ls_state = state;
636 lsp->ls_owner = fl_owner;
637 spin_lock(&clp->cl_lock);
638 nfs_alloc_unique_id(&clp->cl_lockowner_id, &lsp->ls_id, 1, 64);
639 spin_unlock(&clp->cl_lock);
640 INIT_LIST_HEAD(&lsp->ls_locks);
644 static void nfs4_free_lock_state(struct nfs4_lock_state *lsp)
646 struct nfs_client *clp = lsp->ls_state->owner->so_server->nfs_client;
648 spin_lock(&clp->cl_lock);
649 nfs_free_unique_id(&clp->cl_lockowner_id, &lsp->ls_id);
650 spin_unlock(&clp->cl_lock);
651 rpc_destroy_wait_queue(&lsp->ls_sequence.wait);
656 * Return a compatible lock_state. If no initialized lock_state structure
657 * exists, return an uninitialized one.
660 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
662 struct nfs4_lock_state *lsp, *new = NULL;
665 spin_lock(&state->state_lock);
666 lsp = __nfs4_find_lock_state(state, owner);
670 list_add(&new->ls_locks, &state->lock_states);
671 set_bit(LK_STATE_IN_USE, &state->flags);
676 spin_unlock(&state->state_lock);
677 new = nfs4_alloc_lock_state(state, owner);
681 spin_unlock(&state->state_lock);
683 nfs4_free_lock_state(new);
688 * Release reference to lock_state, and free it if we see that
689 * it is no longer in use
691 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
693 struct nfs4_state *state;
697 state = lsp->ls_state;
698 if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
700 list_del(&lsp->ls_locks);
701 if (list_empty(&state->lock_states))
702 clear_bit(LK_STATE_IN_USE, &state->flags);
703 spin_unlock(&state->state_lock);
704 if (lsp->ls_flags & NFS_LOCK_INITIALIZED)
705 nfs4_release_lockowner(lsp);
706 nfs4_free_lock_state(lsp);
709 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
711 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
713 dst->fl_u.nfs4_fl.owner = lsp;
714 atomic_inc(&lsp->ls_count);
717 static void nfs4_fl_release_lock(struct file_lock *fl)
719 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
722 static const struct file_lock_operations nfs4_fl_lock_ops = {
723 .fl_copy_lock = nfs4_fl_copy_lock,
724 .fl_release_private = nfs4_fl_release_lock,
727 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
729 struct nfs4_lock_state *lsp;
731 if (fl->fl_ops != NULL)
733 lsp = nfs4_get_lock_state(state, fl->fl_owner);
736 fl->fl_u.nfs4_fl.owner = lsp;
737 fl->fl_ops = &nfs4_fl_lock_ops;
742 * Byte-range lock aware utility to initialize the stateid of read/write
745 void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner)
747 struct nfs4_lock_state *lsp;
751 seq = read_seqbegin(&state->seqlock);
752 memcpy(dst, &state->stateid, sizeof(*dst));
753 } while (read_seqretry(&state->seqlock, seq));
754 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
757 spin_lock(&state->state_lock);
758 lsp = __nfs4_find_lock_state(state, fl_owner);
759 if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
760 memcpy(dst, &lsp->ls_stateid, sizeof(*dst));
761 spin_unlock(&state->state_lock);
762 nfs4_put_lock_state(lsp);
765 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
767 struct nfs_seqid *new;
769 new = kmalloc(sizeof(*new), gfp_mask);
771 new->sequence = counter;
772 INIT_LIST_HEAD(&new->list);
777 void nfs_release_seqid(struct nfs_seqid *seqid)
779 if (!list_empty(&seqid->list)) {
780 struct rpc_sequence *sequence = seqid->sequence->sequence;
782 spin_lock(&sequence->lock);
783 list_del_init(&seqid->list);
784 spin_unlock(&sequence->lock);
785 rpc_wake_up(&sequence->wait);
789 void nfs_free_seqid(struct nfs_seqid *seqid)
791 nfs_release_seqid(seqid);
796 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
797 * failed with a seqid incrementing error -
798 * see comments nfs_fs.h:seqid_mutating_error()
800 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
802 BUG_ON(list_first_entry(&seqid->sequence->sequence->list, struct nfs_seqid, list) != seqid);
806 case -NFS4ERR_BAD_SEQID:
807 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
809 printk(KERN_WARNING "NFS: v4 server returned a bad"
810 " sequence-id error on an"
811 " unconfirmed sequence %p!\n",
813 case -NFS4ERR_STALE_CLIENTID:
814 case -NFS4ERR_STALE_STATEID:
815 case -NFS4ERR_BAD_STATEID:
816 case -NFS4ERR_BADXDR:
817 case -NFS4ERR_RESOURCE:
818 case -NFS4ERR_NOFILEHANDLE:
819 /* Non-seqid mutating errors */
823 * Note: no locking needed as we are guaranteed to be first
824 * on the sequence list
826 seqid->sequence->counter++;
829 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
831 struct nfs4_state_owner *sp = container_of(seqid->sequence,
832 struct nfs4_state_owner, so_seqid);
833 struct nfs_server *server = sp->so_server;
835 if (status == -NFS4ERR_BAD_SEQID)
836 nfs4_drop_state_owner(sp);
837 if (!nfs4_has_session(server->nfs_client))
838 nfs_increment_seqid(status, seqid);
842 * Increment the seqid if the LOCK/LOCKU succeeded, or
843 * failed with a seqid incrementing error -
844 * see comments nfs_fs.h:seqid_mutating_error()
846 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
848 nfs_increment_seqid(status, seqid);
851 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
853 struct rpc_sequence *sequence = seqid->sequence->sequence;
856 spin_lock(&sequence->lock);
857 if (list_empty(&seqid->list))
858 list_add_tail(&seqid->list, &sequence->list);
859 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
861 rpc_sleep_on(&sequence->wait, task, NULL);
864 spin_unlock(&sequence->lock);
868 static int nfs4_run_state_manager(void *);
870 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
872 smp_mb__before_clear_bit();
873 clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
874 smp_mb__after_clear_bit();
875 wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
876 rpc_wake_up(&clp->cl_rpcwaitq);
880 * Schedule the nfs_client asynchronous state management routine
882 void nfs4_schedule_state_manager(struct nfs_client *clp)
884 struct task_struct *task;
886 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
888 __module_get(THIS_MODULE);
889 atomic_inc(&clp->cl_count);
890 task = kthread_run(nfs4_run_state_manager, clp, "%s-manager",
891 rpc_peeraddr2str(clp->cl_rpcclient,
895 nfs4_clear_state_manager_bit(clp);
897 module_put(THIS_MODULE);
901 * Schedule a state recovery attempt
903 void nfs4_schedule_state_recovery(struct nfs_client *clp)
907 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
908 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
909 nfs4_schedule_state_manager(clp);
912 int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
915 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
916 /* Don't recover state that expired before the reboot */
917 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
918 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
921 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
922 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
926 int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
928 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
929 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
930 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
931 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
935 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
937 struct inode *inode = state->inode;
938 struct nfs_inode *nfsi = NFS_I(inode);
939 struct file_lock *fl;
942 if (inode->i_flock == NULL)
945 /* Guard against delegation returns and new lock/unlock calls */
946 down_write(&nfsi->rwsem);
947 /* Protect inode->i_flock using the BKL */
949 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
950 if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
952 if (nfs_file_open_context(fl->fl_file)->state != state)
955 status = ops->recover_lock(state, fl);
960 case -NFS4ERR_ADMIN_REVOKED:
961 case -NFS4ERR_STALE_STATEID:
962 case -NFS4ERR_BAD_STATEID:
963 case -NFS4ERR_EXPIRED:
964 case -NFS4ERR_NO_GRACE:
965 case -NFS4ERR_STALE_CLIENTID:
966 case -NFS4ERR_BADSESSION:
967 case -NFS4ERR_BADSLOT:
968 case -NFS4ERR_BAD_HIGH_SLOT:
969 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
972 printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
975 case -NFS4ERR_DENIED:
976 case -NFS4ERR_RECLAIM_BAD:
977 case -NFS4ERR_RECLAIM_CONFLICT:
978 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
985 up_write(&nfsi->rwsem);
989 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
991 struct nfs4_state *state;
992 struct nfs4_lock_state *lock;
995 /* Note: we rely on the sp->so_states list being ordered
996 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
998 * This is needed to ensure that the server won't give us any
999 * read delegations that we have to return if, say, we are
1000 * recovering after a network partition or a reboot from a
1001 * server that doesn't support a grace period.
1004 spin_lock(&sp->so_lock);
1005 list_for_each_entry(state, &sp->so_states, open_states) {
1006 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
1008 if (state->state == 0)
1010 atomic_inc(&state->count);
1011 spin_unlock(&sp->so_lock);
1012 status = ops->recover_open(sp, state);
1014 status = nfs4_reclaim_locks(state, ops);
1016 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1017 if (!(lock->ls_flags & NFS_LOCK_INITIALIZED))
1018 printk("%s: Lock reclaim failed!\n",
1021 nfs4_put_open_state(state);
1027 printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
1033 * Open state on this file cannot be recovered
1034 * All we can do is revert to using the zero stateid.
1036 memset(state->stateid.data, 0,
1037 sizeof(state->stateid.data));
1038 /* Mark the file as being 'closed' */
1041 case -NFS4ERR_ADMIN_REVOKED:
1042 case -NFS4ERR_STALE_STATEID:
1043 case -NFS4ERR_BAD_STATEID:
1044 case -NFS4ERR_RECLAIM_BAD:
1045 case -NFS4ERR_RECLAIM_CONFLICT:
1046 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1048 case -NFS4ERR_EXPIRED:
1049 case -NFS4ERR_NO_GRACE:
1050 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1051 case -NFS4ERR_STALE_CLIENTID:
1052 case -NFS4ERR_BADSESSION:
1053 case -NFS4ERR_BADSLOT:
1054 case -NFS4ERR_BAD_HIGH_SLOT:
1055 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1058 nfs4_put_open_state(state);
1061 spin_unlock(&sp->so_lock);
1064 nfs4_put_open_state(state);
1068 static void nfs4_clear_open_state(struct nfs4_state *state)
1070 struct nfs4_lock_state *lock;
1072 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1073 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1074 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1075 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1076 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1077 lock->ls_seqid.flags = 0;
1078 lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
1082 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp, int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1084 struct nfs4_state_owner *sp;
1085 struct rb_node *pos;
1086 struct nfs4_state *state;
1088 /* Reset all sequence ids to zero */
1089 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1090 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1091 sp->so_seqid.flags = 0;
1092 spin_lock(&sp->so_lock);
1093 list_for_each_entry(state, &sp->so_states, open_states) {
1094 if (mark_reclaim(clp, state))
1095 nfs4_clear_open_state(state);
1097 spin_unlock(&sp->so_lock);
1101 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1103 /* Mark all delegations for reclaim */
1104 nfs_delegation_mark_reclaim(clp);
1105 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1108 static void nfs4_reclaim_complete(struct nfs_client *clp,
1109 const struct nfs4_state_recovery_ops *ops)
1111 /* Notify the server we're done reclaiming our state */
1112 if (ops->reclaim_complete)
1113 (void)ops->reclaim_complete(clp);
1116 static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
1118 struct nfs4_state_owner *sp;
1119 struct rb_node *pos;
1120 struct nfs4_state *state;
1122 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1125 nfs4_reclaim_complete(clp, clp->cl_mvops->reboot_recovery_ops);
1127 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1128 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1129 spin_lock(&sp->so_lock);
1130 list_for_each_entry(state, &sp->so_states, open_states) {
1131 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags))
1133 nfs4_state_mark_reclaim_nograce(clp, state);
1135 spin_unlock(&sp->so_lock);
1138 nfs_delegation_reap_unclaimed(clp);
1141 static void nfs_delegation_clear_all(struct nfs_client *clp)
1143 nfs_delegation_mark_reclaim(clp);
1144 nfs_delegation_reap_unclaimed(clp);
1147 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1149 nfs_delegation_clear_all(clp);
1150 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1153 static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1156 case -NFS4ERR_CB_PATH_DOWN:
1157 nfs_handle_cb_pathdown(clp);
1159 case -NFS4ERR_NO_GRACE:
1160 nfs4_state_end_reclaim_reboot(clp);
1162 case -NFS4ERR_STALE_CLIENTID:
1163 case -NFS4ERR_LEASE_MOVED:
1164 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1165 nfs4_state_end_reclaim_reboot(clp);
1166 nfs4_state_start_reclaim_reboot(clp);
1168 case -NFS4ERR_EXPIRED:
1169 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1170 nfs4_state_start_reclaim_nograce(clp);
1172 case -NFS4ERR_BADSESSION:
1173 case -NFS4ERR_BADSLOT:
1174 case -NFS4ERR_BAD_HIGH_SLOT:
1175 case -NFS4ERR_DEADSESSION:
1176 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1177 case -NFS4ERR_SEQ_FALSE_RETRY:
1178 case -NFS4ERR_SEQ_MISORDERED:
1179 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1180 /* Zero session reset errors */
1186 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1188 struct rb_node *pos;
1192 spin_lock(&clp->cl_lock);
1193 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1194 struct nfs4_state_owner *sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1195 if (!test_and_clear_bit(ops->owner_flag_bit, &sp->so_flags))
1197 atomic_inc(&sp->so_count);
1198 spin_unlock(&clp->cl_lock);
1199 status = nfs4_reclaim_open_state(sp, ops);
1201 set_bit(ops->owner_flag_bit, &sp->so_flags);
1202 nfs4_put_state_owner(sp);
1203 return nfs4_recovery_handle_error(clp, status);
1205 nfs4_put_state_owner(sp);
1208 spin_unlock(&clp->cl_lock);
1212 static int nfs4_check_lease(struct nfs_client *clp)
1214 struct rpc_cred *cred;
1215 const struct nfs4_state_maintenance_ops *ops =
1216 clp->cl_mvops->state_renewal_ops;
1217 int status = -NFS4ERR_EXPIRED;
1219 /* Is the client already known to have an expired lease? */
1220 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1222 spin_lock(&clp->cl_lock);
1223 cred = ops->get_state_renewal_cred_locked(clp);
1224 spin_unlock(&clp->cl_lock);
1226 cred = nfs4_get_setclientid_cred(clp);
1230 status = ops->renew_lease(clp, cred);
1233 return nfs4_recovery_handle_error(clp, status);
1236 static int nfs4_reclaim_lease(struct nfs_client *clp)
1238 struct rpc_cred *cred;
1239 const struct nfs4_state_recovery_ops *ops =
1240 clp->cl_mvops->reboot_recovery_ops;
1241 int status = -ENOENT;
1243 cred = ops->get_clid_cred(clp);
1245 status = ops->establish_clid(clp, cred);
1247 /* Handle case where the user hasn't set up machine creds */
1248 if (status == -EACCES && cred == clp->cl_machine_cred) {
1249 nfs4_clear_machine_cred(clp);
1252 if (status == -NFS4ERR_MINOR_VERS_MISMATCH)
1253 status = -EPROTONOSUPPORT;
1258 #ifdef CONFIG_NFS_V4_1
1259 void nfs41_handle_recall_slot(struct nfs_client *clp)
1261 set_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
1262 nfs4_schedule_state_recovery(clp);
1265 static void nfs4_reset_all_state(struct nfs_client *clp)
1267 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1268 clp->cl_boot_time = CURRENT_TIME;
1269 nfs4_state_start_reclaim_nograce(clp);
1270 nfs4_schedule_state_recovery(clp);
1274 static void nfs41_handle_server_reboot(struct nfs_client *clp)
1276 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1277 nfs4_state_start_reclaim_reboot(clp);
1278 nfs4_schedule_state_recovery(clp);
1282 static void nfs41_handle_state_revoked(struct nfs_client *clp)
1285 nfs4_reset_all_state(clp);
1288 static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
1290 /* This will need to handle layouts too */
1291 nfs_expire_all_delegations(clp);
1294 static void nfs41_handle_cb_path_down(struct nfs_client *clp)
1296 nfs_expire_all_delegations(clp);
1297 if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
1298 nfs4_schedule_state_recovery(clp);
1301 void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
1305 else if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
1306 nfs41_handle_server_reboot(clp);
1307 else if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
1308 SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
1309 SEQ4_STATUS_ADMIN_STATE_REVOKED |
1310 SEQ4_STATUS_LEASE_MOVED))
1311 nfs41_handle_state_revoked(clp);
1312 else if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
1313 nfs41_handle_recallable_state_revoked(clp);
1314 else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
1315 SEQ4_STATUS_BACKCHANNEL_FAULT |
1316 SEQ4_STATUS_CB_PATH_DOWN_SESSION))
1317 nfs41_handle_cb_path_down(clp);
1320 static int nfs4_reset_session(struct nfs_client *clp)
1324 nfs4_begin_drain_session(clp);
1325 status = nfs4_proc_destroy_session(clp->cl_session);
1326 if (status && status != -NFS4ERR_BADSESSION &&
1327 status != -NFS4ERR_DEADSESSION) {
1328 status = nfs4_recovery_handle_error(clp, status);
1332 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
1333 status = nfs4_proc_create_session(clp);
1335 status = nfs4_recovery_handle_error(clp, status);
1338 /* create_session negotiated new slot table */
1339 clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
1341 /* Let the state manager reestablish state */
1342 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1343 nfs41_setup_state_renewal(clp);
1348 static int nfs4_recall_slot(struct nfs_client *clp)
1350 struct nfs4_slot_table *fc_tbl = &clp->cl_session->fc_slot_table;
1351 struct nfs4_channel_attrs *fc_attrs = &clp->cl_session->fc_attrs;
1352 struct nfs4_slot *new, *old;
1355 nfs4_begin_drain_session(clp);
1356 new = kmalloc(fc_tbl->target_max_slots * sizeof(struct nfs4_slot),
1361 spin_lock(&fc_tbl->slot_tbl_lock);
1362 for (i = 0; i < fc_tbl->target_max_slots; i++)
1363 new[i].seq_nr = fc_tbl->slots[i].seq_nr;
1364 old = fc_tbl->slots;
1365 fc_tbl->slots = new;
1366 fc_tbl->max_slots = fc_tbl->target_max_slots;
1367 fc_tbl->target_max_slots = 0;
1368 fc_attrs->max_reqs = fc_tbl->max_slots;
1369 spin_unlock(&fc_tbl->slot_tbl_lock);
1372 nfs4_end_drain_session(clp);
1376 #else /* CONFIG_NFS_V4_1 */
1377 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
1378 static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
1379 static int nfs4_recall_slot(struct nfs_client *clp) { return 0; }
1380 #endif /* CONFIG_NFS_V4_1 */
1382 /* Set NFS4CLNT_LEASE_EXPIRED for all v4.0 errors and for recoverable errors
1383 * on EXCHANGE_ID for v4.1
1385 static void nfs4_set_lease_expired(struct nfs_client *clp, int status)
1387 if (nfs4_has_session(clp)) {
1389 case -NFS4ERR_DELAY:
1390 case -NFS4ERR_CLID_INUSE:
1395 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1396 * in nfs4_exchange_id */
1401 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1404 static void nfs4_state_manager(struct nfs_client *clp)
1408 /* Ensure exclusive access to NFSv4 state */
1410 if (test_and_clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
1411 /* We're going to have to re-establish a clientid */
1412 status = nfs4_reclaim_lease(clp);
1414 nfs4_set_lease_expired(clp, status);
1415 if (test_bit(NFS4CLNT_LEASE_EXPIRED,
1418 if (clp->cl_cons_state ==
1419 NFS_CS_SESSION_INITING)
1420 nfs_mark_client_ready(clp, status);
1423 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1424 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1427 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
1428 status = nfs4_check_lease(clp);
1429 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1431 if (status < 0 && status != -NFS4ERR_CB_PATH_DOWN)
1435 /* Initialize or reset the session */
1436 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)
1437 && nfs4_has_session(clp)) {
1438 status = nfs4_reset_session(clp);
1439 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1445 /* First recover reboot state... */
1446 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
1447 status = nfs4_do_reclaim(clp,
1448 clp->cl_mvops->reboot_recovery_ops);
1449 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1450 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
1452 nfs4_state_end_reclaim_reboot(clp);
1453 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
1459 /* Now recover expired state... */
1460 if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
1461 status = nfs4_do_reclaim(clp,
1462 clp->cl_mvops->nograce_recovery_ops);
1463 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1464 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
1465 test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1471 nfs4_end_drain_session(clp);
1472 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
1473 nfs_client_return_marked_delegations(clp);
1476 /* Recall session slots */
1477 if (test_and_clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state)
1478 && nfs4_has_session(clp)) {
1479 status = nfs4_recall_slot(clp);
1486 nfs4_clear_state_manager_bit(clp);
1487 /* Did we race with an attempt to give us more work? */
1488 if (clp->cl_state == 0)
1490 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1495 printk(KERN_WARNING "Error: state manager failed on NFSv4 server %s"
1496 " with error %d\n", clp->cl_hostname, -status);
1497 nfs4_end_drain_session(clp);
1498 nfs4_clear_state_manager_bit(clp);
1501 static int nfs4_run_state_manager(void *ptr)
1503 struct nfs_client *clp = ptr;
1505 allow_signal(SIGKILL);
1506 nfs4_state_manager(clp);
1507 nfs_put_client(clp);
1508 module_put_and_exit(0);