4 * Client-side XDR for NFSv4.
6 * Copyright (c) 2002 The Regents of the University of Michigan.
9 * Kendrick Smith <kmsmith@umich.edu>
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its
21 * contributors may be used to endorse or promote products derived
22 * from this software without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
25 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
26 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
27 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
31 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
32 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
33 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
34 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36 * Implementation of the NFSv4 state model. For the time being,
37 * this is minimal, but will be made much more complex in a
41 #include <linux/kernel.h>
42 #include <linux/slab.h>
43 #include <linux/smp_lock.h>
44 #include <linux/nfs_fs.h>
45 #include <linux/nfs_idmap.h>
46 #include <linux/kthread.h>
47 #include <linux/module.h>
48 #include <linux/random.h>
49 #include <linux/workqueue.h>
50 #include <linux/bitops.h>
54 #include "delegation.h"
57 #define OPENOWNER_POOL_SIZE 8
59 const nfs4_stateid zero_stateid;
61 static LIST_HEAD(nfs4_clientid_list);
63 int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
68 port = nfs_callback_tcpport;
69 if (clp->cl_addr.ss_family == AF_INET6)
70 port = nfs_callback_tcpport6;
72 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred);
74 status = nfs4_proc_setclientid_confirm(clp, cred);
76 nfs4_schedule_state_renewal(clp);
80 struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
82 struct rpc_cred *cred = NULL;
84 if (clp->cl_machine_cred != NULL)
85 cred = get_rpccred(clp->cl_machine_cred);
89 static void nfs4_clear_machine_cred(struct nfs_client *clp)
91 struct rpc_cred *cred;
93 spin_lock(&clp->cl_lock);
94 cred = clp->cl_machine_cred;
95 clp->cl_machine_cred = NULL;
96 spin_unlock(&clp->cl_lock);
101 struct rpc_cred *nfs4_get_renew_cred_locked(struct nfs_client *clp)
103 struct nfs4_state_owner *sp;
105 struct rpc_cred *cred = NULL;
107 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
108 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
109 if (list_empty(&sp->so_states))
111 cred = get_rpccred(sp->so_cred);
117 #if defined(CONFIG_NFS_V4_1)
119 static int nfs41_setup_state_renewal(struct nfs_client *clp)
122 struct nfs_fsinfo fsinfo;
124 status = nfs4_proc_get_lease_time(clp, &fsinfo);
126 /* Update lease time and schedule renewal */
127 spin_lock(&clp->cl_lock);
128 clp->cl_lease_time = fsinfo.lease_time * HZ;
129 clp->cl_last_renewal = jiffies;
130 spin_unlock(&clp->cl_lock);
132 nfs4_schedule_state_renewal(clp);
138 static void nfs41_end_drain_session(struct nfs_client *clp,
139 struct nfs4_session *ses)
143 if (test_and_clear_bit(NFS4CLNT_SESSION_DRAINING, &clp->cl_state)) {
144 spin_lock(&ses->fc_slot_table.slot_tbl_lock);
145 max_slots = ses->fc_slot_table.max_slots;
146 while (max_slots--) {
147 struct rpc_task *task;
149 task = rpc_wake_up_next(&ses->fc_slot_table.
153 rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
155 spin_unlock(&ses->fc_slot_table.slot_tbl_lock);
159 static int nfs41_begin_drain_session(struct nfs_client *clp,
160 struct nfs4_session *ses)
162 struct nfs4_slot_table *tbl = &ses->fc_slot_table;
164 spin_lock(&tbl->slot_tbl_lock);
165 set_bit(NFS4CLNT_SESSION_DRAINING, &clp->cl_state);
166 if (tbl->highest_used_slotid != -1) {
167 INIT_COMPLETION(ses->complete);
168 spin_unlock(&tbl->slot_tbl_lock);
169 return wait_for_completion_interruptible(&ses->complete);
171 spin_unlock(&tbl->slot_tbl_lock);
175 int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
179 status = nfs41_begin_drain_session(clp, clp->cl_session);
182 status = nfs4_proc_exchange_id(clp, cred);
185 status = nfs4_proc_create_session(clp);
188 nfs41_end_drain_session(clp, clp->cl_session);
189 nfs41_setup_state_renewal(clp);
190 nfs_mark_client_ready(clp, NFS_CS_READY);
195 struct rpc_cred *nfs4_get_exchange_id_cred(struct nfs_client *clp)
197 struct rpc_cred *cred;
199 spin_lock(&clp->cl_lock);
200 cred = nfs4_get_machine_cred_locked(clp);
201 spin_unlock(&clp->cl_lock);
205 #endif /* CONFIG_NFS_V4_1 */
207 struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
209 struct nfs4_state_owner *sp;
211 struct rpc_cred *cred;
213 spin_lock(&clp->cl_lock);
214 cred = nfs4_get_machine_cred_locked(clp);
217 pos = rb_first(&clp->cl_state_owners);
219 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
220 cred = get_rpccred(sp->so_cred);
223 spin_unlock(&clp->cl_lock);
227 static void nfs_alloc_unique_id(struct rb_root *root, struct nfs_unique_id *new,
228 __u64 minval, int maxbits)
230 struct rb_node **p, *parent;
231 struct nfs_unique_id *pos;
235 mask = (1ULL << maxbits) - 1ULL;
237 /* Ensure distribution is more or less flat */
238 get_random_bytes(&new->id, sizeof(new->id));
240 if (new->id < minval)
248 pos = rb_entry(parent, struct nfs_unique_id, rb_node);
250 if (new->id < pos->id)
252 else if (new->id > pos->id)
257 rb_link_node(&new->rb_node, parent, p);
258 rb_insert_color(&new->rb_node, root);
263 if (new->id < minval || (new->id & mask) != new->id) {
267 parent = rb_next(parent);
270 pos = rb_entry(parent, struct nfs_unique_id, rb_node);
271 if (new->id < pos->id)
277 static void nfs_free_unique_id(struct rb_root *root, struct nfs_unique_id *id)
279 rb_erase(&id->rb_node, root);
282 static struct nfs4_state_owner *
283 nfs4_find_state_owner(struct nfs_server *server, struct rpc_cred *cred)
285 struct nfs_client *clp = server->nfs_client;
286 struct rb_node **p = &clp->cl_state_owners.rb_node,
288 struct nfs4_state_owner *sp, *res = NULL;
292 sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
294 if (server < sp->so_server) {
295 p = &parent->rb_left;
298 if (server > sp->so_server) {
299 p = &parent->rb_right;
302 if (cred < sp->so_cred)
303 p = &parent->rb_left;
304 else if (cred > sp->so_cred)
305 p = &parent->rb_right;
307 atomic_inc(&sp->so_count);
315 static struct nfs4_state_owner *
316 nfs4_insert_state_owner(struct nfs_client *clp, struct nfs4_state_owner *new)
318 struct rb_node **p = &clp->cl_state_owners.rb_node,
320 struct nfs4_state_owner *sp;
324 sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
326 if (new->so_server < sp->so_server) {
327 p = &parent->rb_left;
330 if (new->so_server > sp->so_server) {
331 p = &parent->rb_right;
334 if (new->so_cred < sp->so_cred)
335 p = &parent->rb_left;
336 else if (new->so_cred > sp->so_cred)
337 p = &parent->rb_right;
339 atomic_inc(&sp->so_count);
343 nfs_alloc_unique_id(&clp->cl_openowner_id, &new->so_owner_id, 1, 64);
344 rb_link_node(&new->so_client_node, parent, p);
345 rb_insert_color(&new->so_client_node, &clp->cl_state_owners);
350 nfs4_remove_state_owner(struct nfs_client *clp, struct nfs4_state_owner *sp)
352 if (!RB_EMPTY_NODE(&sp->so_client_node))
353 rb_erase(&sp->so_client_node, &clp->cl_state_owners);
354 nfs_free_unique_id(&clp->cl_openowner_id, &sp->so_owner_id);
358 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
359 * create a new state_owner.
362 static struct nfs4_state_owner *
363 nfs4_alloc_state_owner(void)
365 struct nfs4_state_owner *sp;
367 sp = kzalloc(sizeof(*sp),GFP_KERNEL);
370 spin_lock_init(&sp->so_lock);
371 INIT_LIST_HEAD(&sp->so_states);
372 INIT_LIST_HEAD(&sp->so_delegations);
373 rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue");
374 sp->so_seqid.sequence = &sp->so_sequence;
375 spin_lock_init(&sp->so_sequence.lock);
376 INIT_LIST_HEAD(&sp->so_sequence.list);
377 atomic_set(&sp->so_count, 1);
382 nfs4_drop_state_owner(struct nfs4_state_owner *sp)
384 if (!RB_EMPTY_NODE(&sp->so_client_node)) {
385 struct nfs_client *clp = sp->so_client;
387 spin_lock(&clp->cl_lock);
388 rb_erase(&sp->so_client_node, &clp->cl_state_owners);
389 RB_CLEAR_NODE(&sp->so_client_node);
390 spin_unlock(&clp->cl_lock);
394 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct rpc_cred *cred)
396 struct nfs_client *clp = server->nfs_client;
397 struct nfs4_state_owner *sp, *new;
399 spin_lock(&clp->cl_lock);
400 sp = nfs4_find_state_owner(server, cred);
401 spin_unlock(&clp->cl_lock);
404 new = nfs4_alloc_state_owner();
407 new->so_client = clp;
408 new->so_server = server;
410 spin_lock(&clp->cl_lock);
411 sp = nfs4_insert_state_owner(clp, new);
412 spin_unlock(&clp->cl_lock);
416 rpc_destroy_wait_queue(&new->so_sequence.wait);
422 void nfs4_put_state_owner(struct nfs4_state_owner *sp)
424 struct nfs_client *clp = sp->so_client;
425 struct rpc_cred *cred = sp->so_cred;
427 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
429 nfs4_remove_state_owner(clp, sp);
430 spin_unlock(&clp->cl_lock);
431 rpc_destroy_wait_queue(&sp->so_sequence.wait);
436 static struct nfs4_state *
437 nfs4_alloc_open_state(void)
439 struct nfs4_state *state;
441 state = kzalloc(sizeof(*state), GFP_KERNEL);
444 atomic_set(&state->count, 1);
445 INIT_LIST_HEAD(&state->lock_states);
446 spin_lock_init(&state->state_lock);
447 seqlock_init(&state->seqlock);
452 nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
454 if (state->state == fmode)
456 /* NB! List reordering - see the reclaim code for why. */
457 if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
458 if (fmode & FMODE_WRITE)
459 list_move(&state->open_states, &state->owner->so_states);
461 list_move_tail(&state->open_states, &state->owner->so_states);
463 state->state = fmode;
466 static struct nfs4_state *
467 __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
469 struct nfs_inode *nfsi = NFS_I(inode);
470 struct nfs4_state *state;
472 list_for_each_entry(state, &nfsi->open_states, inode_states) {
473 if (state->owner != owner)
475 if (atomic_inc_not_zero(&state->count))
482 nfs4_free_open_state(struct nfs4_state *state)
488 nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
490 struct nfs4_state *state, *new;
491 struct nfs_inode *nfsi = NFS_I(inode);
493 spin_lock(&inode->i_lock);
494 state = __nfs4_find_state_byowner(inode, owner);
495 spin_unlock(&inode->i_lock);
498 new = nfs4_alloc_open_state();
499 spin_lock(&owner->so_lock);
500 spin_lock(&inode->i_lock);
501 state = __nfs4_find_state_byowner(inode, owner);
502 if (state == NULL && new != NULL) {
504 state->owner = owner;
505 atomic_inc(&owner->so_count);
506 list_add(&state->inode_states, &nfsi->open_states);
507 state->inode = igrab(inode);
508 spin_unlock(&inode->i_lock);
509 /* Note: The reclaim code dictates that we add stateless
510 * and read-only stateids to the end of the list */
511 list_add_tail(&state->open_states, &owner->so_states);
512 spin_unlock(&owner->so_lock);
514 spin_unlock(&inode->i_lock);
515 spin_unlock(&owner->so_lock);
517 nfs4_free_open_state(new);
523 void nfs4_put_open_state(struct nfs4_state *state)
525 struct inode *inode = state->inode;
526 struct nfs4_state_owner *owner = state->owner;
528 if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
530 spin_lock(&inode->i_lock);
531 list_del(&state->inode_states);
532 list_del(&state->open_states);
533 spin_unlock(&inode->i_lock);
534 spin_unlock(&owner->so_lock);
536 nfs4_free_open_state(state);
537 nfs4_put_state_owner(owner);
541 * Close the current file.
543 static void __nfs4_close(struct path *path, struct nfs4_state *state, fmode_t fmode, int wait)
545 struct nfs4_state_owner *owner = state->owner;
549 atomic_inc(&owner->so_count);
550 /* Protect against nfs4_find_state() */
551 spin_lock(&owner->so_lock);
552 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
559 case FMODE_READ|FMODE_WRITE:
562 newstate = FMODE_READ|FMODE_WRITE;
563 if (state->n_rdwr == 0) {
564 if (state->n_rdonly == 0) {
565 newstate &= ~FMODE_READ;
566 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
567 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
569 if (state->n_wronly == 0) {
570 newstate &= ~FMODE_WRITE;
571 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
572 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
575 clear_bit(NFS_DELEGATED_STATE, &state->flags);
577 nfs4_state_set_mode_locked(state, newstate);
578 spin_unlock(&owner->so_lock);
581 nfs4_put_open_state(state);
582 nfs4_put_state_owner(owner);
584 nfs4_do_close(path, state, wait);
587 void nfs4_close_state(struct path *path, struct nfs4_state *state, fmode_t fmode)
589 __nfs4_close(path, state, fmode, 0);
592 void nfs4_close_sync(struct path *path, struct nfs4_state *state, fmode_t fmode)
594 __nfs4_close(path, state, fmode, 1);
598 * Search the state->lock_states for an existing lock_owner
599 * that is compatible with current->files
601 static struct nfs4_lock_state *
602 __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
604 struct nfs4_lock_state *pos;
605 list_for_each_entry(pos, &state->lock_states, ls_locks) {
606 if (pos->ls_owner != fl_owner)
608 atomic_inc(&pos->ls_count);
615 * Return a compatible lock_state. If no initialized lock_state structure
616 * exists, return an uninitialized one.
619 static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
621 struct nfs4_lock_state *lsp;
622 struct nfs_client *clp = state->owner->so_client;
624 lsp = kzalloc(sizeof(*lsp), GFP_KERNEL);
627 rpc_init_wait_queue(&lsp->ls_sequence.wait, "lock_seqid_waitqueue");
628 spin_lock_init(&lsp->ls_sequence.lock);
629 INIT_LIST_HEAD(&lsp->ls_sequence.list);
630 lsp->ls_seqid.sequence = &lsp->ls_sequence;
631 atomic_set(&lsp->ls_count, 1);
632 lsp->ls_state = state;
633 lsp->ls_owner = fl_owner;
634 spin_lock(&clp->cl_lock);
635 nfs_alloc_unique_id(&clp->cl_lockowner_id, &lsp->ls_id, 1, 64);
636 spin_unlock(&clp->cl_lock);
637 INIT_LIST_HEAD(&lsp->ls_locks);
641 static void nfs4_free_lock_state(struct nfs4_lock_state *lsp)
643 struct nfs_client *clp = lsp->ls_state->owner->so_client;
645 spin_lock(&clp->cl_lock);
646 nfs_free_unique_id(&clp->cl_lockowner_id, &lsp->ls_id);
647 spin_unlock(&clp->cl_lock);
648 rpc_destroy_wait_queue(&lsp->ls_sequence.wait);
653 * Return a compatible lock_state. If no initialized lock_state structure
654 * exists, return an uninitialized one.
657 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
659 struct nfs4_lock_state *lsp, *new = NULL;
662 spin_lock(&state->state_lock);
663 lsp = __nfs4_find_lock_state(state, owner);
667 list_add(&new->ls_locks, &state->lock_states);
668 set_bit(LK_STATE_IN_USE, &state->flags);
673 spin_unlock(&state->state_lock);
674 new = nfs4_alloc_lock_state(state, owner);
678 spin_unlock(&state->state_lock);
680 nfs4_free_lock_state(new);
685 * Release reference to lock_state, and free it if we see that
686 * it is no longer in use
688 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
690 struct nfs4_state *state;
694 state = lsp->ls_state;
695 if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
697 list_del(&lsp->ls_locks);
698 if (list_empty(&state->lock_states))
699 clear_bit(LK_STATE_IN_USE, &state->flags);
700 spin_unlock(&state->state_lock);
701 nfs4_free_lock_state(lsp);
704 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
706 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
708 dst->fl_u.nfs4_fl.owner = lsp;
709 atomic_inc(&lsp->ls_count);
712 static void nfs4_fl_release_lock(struct file_lock *fl)
714 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
717 static const struct file_lock_operations nfs4_fl_lock_ops = {
718 .fl_copy_lock = nfs4_fl_copy_lock,
719 .fl_release_private = nfs4_fl_release_lock,
722 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
724 struct nfs4_lock_state *lsp;
726 if (fl->fl_ops != NULL)
728 lsp = nfs4_get_lock_state(state, fl->fl_owner);
731 fl->fl_u.nfs4_fl.owner = lsp;
732 fl->fl_ops = &nfs4_fl_lock_ops;
737 * Byte-range lock aware utility to initialize the stateid of read/write
740 void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner)
742 struct nfs4_lock_state *lsp;
746 seq = read_seqbegin(&state->seqlock);
747 memcpy(dst, &state->stateid, sizeof(*dst));
748 } while (read_seqretry(&state->seqlock, seq));
749 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
752 spin_lock(&state->state_lock);
753 lsp = __nfs4_find_lock_state(state, fl_owner);
754 if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
755 memcpy(dst, &lsp->ls_stateid, sizeof(*dst));
756 spin_unlock(&state->state_lock);
757 nfs4_put_lock_state(lsp);
760 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter)
762 struct nfs_seqid *new;
764 new = kmalloc(sizeof(*new), GFP_KERNEL);
766 new->sequence = counter;
767 INIT_LIST_HEAD(&new->list);
772 void nfs_free_seqid(struct nfs_seqid *seqid)
774 if (!list_empty(&seqid->list)) {
775 struct rpc_sequence *sequence = seqid->sequence->sequence;
777 spin_lock(&sequence->lock);
778 list_del(&seqid->list);
779 spin_unlock(&sequence->lock);
780 rpc_wake_up(&sequence->wait);
786 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
787 * failed with a seqid incrementing error -
788 * see comments nfs_fs.h:seqid_mutating_error()
790 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
792 BUG_ON(list_first_entry(&seqid->sequence->sequence->list, struct nfs_seqid, list) != seqid);
796 case -NFS4ERR_BAD_SEQID:
797 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
799 printk(KERN_WARNING "NFS: v4 server returned a bad"
800 " sequence-id error on an"
801 " unconfirmed sequence %p!\n",
803 case -NFS4ERR_STALE_CLIENTID:
804 case -NFS4ERR_STALE_STATEID:
805 case -NFS4ERR_BAD_STATEID:
806 case -NFS4ERR_BADXDR:
807 case -NFS4ERR_RESOURCE:
808 case -NFS4ERR_NOFILEHANDLE:
809 /* Non-seqid mutating errors */
813 * Note: no locking needed as we are guaranteed to be first
814 * on the sequence list
816 seqid->sequence->counter++;
819 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
821 struct nfs4_state_owner *sp = container_of(seqid->sequence,
822 struct nfs4_state_owner, so_seqid);
823 struct nfs_server *server = sp->so_server;
825 if (status == -NFS4ERR_BAD_SEQID)
826 nfs4_drop_state_owner(sp);
827 if (!nfs4_has_session(server->nfs_client))
828 nfs_increment_seqid(status, seqid);
832 * Increment the seqid if the LOCK/LOCKU succeeded, or
833 * failed with a seqid incrementing error -
834 * see comments nfs_fs.h:seqid_mutating_error()
836 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
838 nfs_increment_seqid(status, seqid);
841 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
843 struct rpc_sequence *sequence = seqid->sequence->sequence;
846 spin_lock(&sequence->lock);
847 if (list_empty(&seqid->list))
848 list_add_tail(&seqid->list, &sequence->list);
849 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
851 rpc_sleep_on(&sequence->wait, task, NULL);
854 spin_unlock(&sequence->lock);
858 static int nfs4_run_state_manager(void *);
860 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
862 smp_mb__before_clear_bit();
863 clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
864 smp_mb__after_clear_bit();
865 wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
866 rpc_wake_up(&clp->cl_rpcwaitq);
870 * Schedule the nfs_client asynchronous state management routine
872 void nfs4_schedule_state_manager(struct nfs_client *clp)
874 struct task_struct *task;
876 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
878 __module_get(THIS_MODULE);
879 atomic_inc(&clp->cl_count);
880 task = kthread_run(nfs4_run_state_manager, clp, "%s-manager",
881 rpc_peeraddr2str(clp->cl_rpcclient,
885 nfs4_clear_state_manager_bit(clp);
887 module_put(THIS_MODULE);
891 * Schedule a state recovery attempt
893 void nfs4_schedule_state_recovery(struct nfs_client *clp)
897 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
898 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
899 nfs4_schedule_state_manager(clp);
902 static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
905 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
906 /* Don't recover state that expired before the reboot */
907 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
908 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
911 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
912 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
916 int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
918 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
919 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
920 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
921 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
925 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
927 struct inode *inode = state->inode;
928 struct nfs_inode *nfsi = NFS_I(inode);
929 struct file_lock *fl;
932 if (inode->i_flock == NULL)
935 /* Guard against delegation returns and new lock/unlock calls */
936 down_write(&nfsi->rwsem);
937 /* Protect inode->i_flock using the BKL */
939 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
940 if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
942 if (nfs_file_open_context(fl->fl_file)->state != state)
945 status = ops->recover_lock(state, fl);
950 case -NFS4ERR_ADMIN_REVOKED:
951 case -NFS4ERR_STALE_STATEID:
952 case -NFS4ERR_BAD_STATEID:
953 case -NFS4ERR_EXPIRED:
954 case -NFS4ERR_NO_GRACE:
955 case -NFS4ERR_STALE_CLIENTID:
956 case -NFS4ERR_BADSESSION:
957 case -NFS4ERR_BADSLOT:
958 case -NFS4ERR_BAD_HIGH_SLOT:
959 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
962 printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
965 case -NFS4ERR_DENIED:
966 case -NFS4ERR_RECLAIM_BAD:
967 case -NFS4ERR_RECLAIM_CONFLICT:
968 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
975 up_write(&nfsi->rwsem);
979 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
981 struct nfs4_state *state;
982 struct nfs4_lock_state *lock;
985 /* Note: we rely on the sp->so_states list being ordered
986 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
988 * This is needed to ensure that the server won't give us any
989 * read delegations that we have to return if, say, we are
990 * recovering after a network partition or a reboot from a
991 * server that doesn't support a grace period.
994 spin_lock(&sp->so_lock);
995 list_for_each_entry(state, &sp->so_states, open_states) {
996 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
998 if (state->state == 0)
1000 atomic_inc(&state->count);
1001 spin_unlock(&sp->so_lock);
1002 status = ops->recover_open(sp, state);
1004 status = nfs4_reclaim_locks(state, ops);
1006 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1007 if (!(lock->ls_flags & NFS_LOCK_INITIALIZED))
1008 printk("%s: Lock reclaim failed!\n",
1011 nfs4_put_open_state(state);
1017 printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
1023 * Open state on this file cannot be recovered
1024 * All we can do is revert to using the zero stateid.
1026 memset(state->stateid.data, 0,
1027 sizeof(state->stateid.data));
1028 /* Mark the file as being 'closed' */
1031 case -NFS4ERR_ADMIN_REVOKED:
1032 case -NFS4ERR_STALE_STATEID:
1033 case -NFS4ERR_BAD_STATEID:
1034 case -NFS4ERR_RECLAIM_BAD:
1035 case -NFS4ERR_RECLAIM_CONFLICT:
1036 nfs4_state_mark_reclaim_nograce(sp->so_client, state);
1038 case -NFS4ERR_EXPIRED:
1039 case -NFS4ERR_NO_GRACE:
1040 nfs4_state_mark_reclaim_nograce(sp->so_client, state);
1041 case -NFS4ERR_STALE_CLIENTID:
1042 case -NFS4ERR_BADSESSION:
1043 case -NFS4ERR_BADSLOT:
1044 case -NFS4ERR_BAD_HIGH_SLOT:
1045 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1048 nfs4_put_open_state(state);
1051 spin_unlock(&sp->so_lock);
1054 nfs4_put_open_state(state);
1058 static void nfs4_clear_open_state(struct nfs4_state *state)
1060 struct nfs4_lock_state *lock;
1062 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1063 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1064 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1065 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1066 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1067 lock->ls_seqid.flags = 0;
1068 lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
1072 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp, int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1074 struct nfs4_state_owner *sp;
1075 struct rb_node *pos;
1076 struct nfs4_state *state;
1078 /* Reset all sequence ids to zero */
1079 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1080 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1081 sp->so_seqid.flags = 0;
1082 spin_lock(&sp->so_lock);
1083 list_for_each_entry(state, &sp->so_states, open_states) {
1084 if (mark_reclaim(clp, state))
1085 nfs4_clear_open_state(state);
1087 spin_unlock(&sp->so_lock);
1091 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1093 /* Mark all delegations for reclaim */
1094 nfs_delegation_mark_reclaim(clp);
1095 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1098 static void nfs4_reclaim_complete(struct nfs_client *clp,
1099 const struct nfs4_state_recovery_ops *ops)
1101 /* Notify the server we're done reclaiming our state */
1102 if (ops->reclaim_complete)
1103 (void)ops->reclaim_complete(clp);
1106 static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
1108 struct nfs4_state_owner *sp;
1109 struct rb_node *pos;
1110 struct nfs4_state *state;
1112 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1115 nfs4_reclaim_complete(clp,
1116 nfs4_reboot_recovery_ops[clp->cl_minorversion]);
1118 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1119 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1120 spin_lock(&sp->so_lock);
1121 list_for_each_entry(state, &sp->so_states, open_states) {
1122 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags))
1124 nfs4_state_mark_reclaim_nograce(clp, state);
1126 spin_unlock(&sp->so_lock);
1129 nfs_delegation_reap_unclaimed(clp);
1132 static void nfs_delegation_clear_all(struct nfs_client *clp)
1134 nfs_delegation_mark_reclaim(clp);
1135 nfs_delegation_reap_unclaimed(clp);
1138 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1140 nfs_delegation_clear_all(clp);
1141 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1144 static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1147 case -NFS4ERR_CB_PATH_DOWN:
1148 nfs_handle_cb_pathdown(clp);
1150 case -NFS4ERR_NO_GRACE:
1151 nfs4_state_end_reclaim_reboot(clp);
1153 case -NFS4ERR_STALE_CLIENTID:
1154 case -NFS4ERR_LEASE_MOVED:
1155 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1156 nfs4_state_end_reclaim_reboot(clp);
1157 nfs4_state_start_reclaim_reboot(clp);
1159 case -NFS4ERR_EXPIRED:
1160 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1161 nfs4_state_start_reclaim_nograce(clp);
1163 case -NFS4ERR_BADSESSION:
1164 case -NFS4ERR_BADSLOT:
1165 case -NFS4ERR_BAD_HIGH_SLOT:
1166 case -NFS4ERR_DEADSESSION:
1167 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1168 case -NFS4ERR_SEQ_FALSE_RETRY:
1169 case -NFS4ERR_SEQ_MISORDERED:
1170 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1171 /* Zero session reset errors */
1177 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1179 struct rb_node *pos;
1183 spin_lock(&clp->cl_lock);
1184 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1185 struct nfs4_state_owner *sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1186 if (!test_and_clear_bit(ops->owner_flag_bit, &sp->so_flags))
1188 atomic_inc(&sp->so_count);
1189 spin_unlock(&clp->cl_lock);
1190 status = nfs4_reclaim_open_state(sp, ops);
1192 set_bit(ops->owner_flag_bit, &sp->so_flags);
1193 nfs4_put_state_owner(sp);
1194 return nfs4_recovery_handle_error(clp, status);
1196 nfs4_put_state_owner(sp);
1199 spin_unlock(&clp->cl_lock);
1203 static int nfs4_check_lease(struct nfs_client *clp)
1205 struct rpc_cred *cred;
1206 struct nfs4_state_maintenance_ops *ops =
1207 nfs4_state_renewal_ops[clp->cl_minorversion];
1208 int status = -NFS4ERR_EXPIRED;
1210 /* Is the client already known to have an expired lease? */
1211 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1213 spin_lock(&clp->cl_lock);
1214 cred = ops->get_state_renewal_cred_locked(clp);
1215 spin_unlock(&clp->cl_lock);
1217 cred = nfs4_get_setclientid_cred(clp);
1221 status = ops->renew_lease(clp, cred);
1224 return nfs4_recovery_handle_error(clp, status);
1227 static int nfs4_reclaim_lease(struct nfs_client *clp)
1229 struct rpc_cred *cred;
1230 struct nfs4_state_recovery_ops *ops =
1231 nfs4_reboot_recovery_ops[clp->cl_minorversion];
1232 int status = -ENOENT;
1234 cred = ops->get_clid_cred(clp);
1236 status = ops->establish_clid(clp, cred);
1238 /* Handle case where the user hasn't set up machine creds */
1239 if (status == -EACCES && cred == clp->cl_machine_cred) {
1240 nfs4_clear_machine_cred(clp);
1243 if (status == -NFS4ERR_MINOR_VERS_MISMATCH)
1244 status = -EPROTONOSUPPORT;
1249 #ifdef CONFIG_NFS_V4_1
1250 void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
1254 else if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED) {
1255 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1256 nfs4_state_start_reclaim_reboot(clp);
1257 nfs4_schedule_state_recovery(clp);
1258 } else if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
1259 SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
1260 SEQ4_STATUS_ADMIN_STATE_REVOKED |
1261 SEQ4_STATUS_RECALLABLE_STATE_REVOKED |
1262 SEQ4_STATUS_LEASE_MOVED)) {
1263 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1264 nfs4_state_start_reclaim_nograce(clp);
1265 nfs4_schedule_state_recovery(clp);
1266 } else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
1267 SEQ4_STATUS_BACKCHANNEL_FAULT |
1268 SEQ4_STATUS_CB_PATH_DOWN_SESSION))
1269 nfs_expire_all_delegations(clp);
1272 static int nfs4_reset_session(struct nfs_client *clp)
1274 struct nfs4_session *ses = clp->cl_session;
1277 status = nfs41_begin_drain_session(clp, ses);
1281 status = nfs4_proc_destroy_session(clp->cl_session);
1282 if (status && status != -NFS4ERR_BADSESSION &&
1283 status != -NFS4ERR_DEADSESSION) {
1284 status = nfs4_recovery_handle_error(clp, status);
1288 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
1289 status = nfs4_proc_create_session(clp);
1291 status = nfs4_recovery_handle_error(clp, status);
1295 * Let the state manager reestablish state
1296 * without waking other tasks yet.
1298 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
1299 /* Wake up the next rpc task */
1300 nfs41_end_drain_session(clp, ses);
1302 nfs41_setup_state_renewal(clp);
1307 #else /* CONFIG_NFS_V4_1 */
1308 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
1309 #endif /* CONFIG_NFS_V4_1 */
1311 /* Set NFS4CLNT_LEASE_EXPIRED for all v4.0 errors and for recoverable errors
1312 * on EXCHANGE_ID for v4.1
1314 static void nfs4_set_lease_expired(struct nfs_client *clp, int status)
1316 if (nfs4_has_session(clp)) {
1318 case -NFS4ERR_DELAY:
1319 case -NFS4ERR_CLID_INUSE:
1323 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1324 * in nfs4_exchange_id */
1329 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1332 static void nfs4_state_manager(struct nfs_client *clp)
1336 /* Ensure exclusive access to NFSv4 state */
1338 if (test_and_clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
1339 /* We're going to have to re-establish a clientid */
1340 status = nfs4_reclaim_lease(clp);
1342 nfs4_set_lease_expired(clp, status);
1343 if (test_bit(NFS4CLNT_LEASE_EXPIRED,
1346 if (clp->cl_cons_state ==
1347 NFS_CS_SESSION_INITING)
1348 nfs_mark_client_ready(clp, status);
1351 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1352 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1355 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
1356 status = nfs4_check_lease(clp);
1357 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1359 if (status < 0 && status != -NFS4ERR_CB_PATH_DOWN)
1363 /* Initialize or reset the session */
1364 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)
1365 && nfs4_has_session(clp)) {
1366 status = nfs4_reset_session(clp);
1367 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1373 /* First recover reboot state... */
1374 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
1375 status = nfs4_do_reclaim(clp,
1376 nfs4_reboot_recovery_ops[clp->cl_minorversion]);
1377 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1378 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
1380 nfs4_state_end_reclaim_reboot(clp);
1381 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
1387 /* Now recover expired state... */
1388 if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
1389 status = nfs4_do_reclaim(clp,
1390 nfs4_nograce_recovery_ops[clp->cl_minorversion]);
1391 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1392 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
1393 test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1399 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
1400 nfs_client_return_marked_delegations(clp);
1404 nfs4_clear_state_manager_bit(clp);
1405 /* Did we race with an attempt to give us more work? */
1406 if (clp->cl_state == 0)
1408 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1413 printk(KERN_WARNING "Error: state manager failed on NFSv4 server %s"
1414 " with error %d\n", clp->cl_hostname, -status);
1415 nfs4_clear_state_manager_bit(clp);
1418 static int nfs4_run_state_manager(void *ptr)
1420 struct nfs_client *clp = ptr;
1422 allow_signal(SIGKILL);
1423 nfs4_state_manager(clp);
1424 nfs_put_client(clp);
1425 module_put_and_exit(0);