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;
148 if (test_and_clear_bit(NFS4CLNT_SESSION_DRAINING, &clp->cl_state)) {
149 spin_lock(&ses->fc_slot_table.slot_tbl_lock);
150 max_slots = ses->fc_slot_table.max_slots;
151 while (max_slots--) {
152 struct rpc_task *task;
154 task = rpc_wake_up_next(&ses->fc_slot_table.
158 rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
160 spin_unlock(&ses->fc_slot_table.slot_tbl_lock);
164 static int nfs4_begin_drain_session(struct nfs_client *clp)
166 struct nfs4_session *ses = clp->cl_session;
167 struct nfs4_slot_table *tbl = &ses->fc_slot_table;
169 spin_lock(&tbl->slot_tbl_lock);
170 set_bit(NFS4CLNT_SESSION_DRAINING, &clp->cl_state);
171 if (tbl->highest_used_slotid != -1) {
172 INIT_COMPLETION(ses->complete);
173 spin_unlock(&tbl->slot_tbl_lock);
174 return wait_for_completion_interruptible(&ses->complete);
176 spin_unlock(&tbl->slot_tbl_lock);
180 int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
184 nfs4_begin_drain_session(clp);
185 status = nfs4_proc_exchange_id(clp, cred);
188 status = nfs4_proc_create_session(clp);
191 nfs41_setup_state_renewal(clp);
192 nfs_mark_client_ready(clp, NFS_CS_READY);
197 struct rpc_cred *nfs4_get_exchange_id_cred(struct nfs_client *clp)
199 struct rpc_cred *cred;
201 spin_lock(&clp->cl_lock);
202 cred = nfs4_get_machine_cred_locked(clp);
203 spin_unlock(&clp->cl_lock);
207 #endif /* CONFIG_NFS_V4_1 */
209 struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
211 struct nfs4_state_owner *sp;
213 struct rpc_cred *cred;
215 spin_lock(&clp->cl_lock);
216 cred = nfs4_get_machine_cred_locked(clp);
219 pos = rb_first(&clp->cl_state_owners);
221 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
222 cred = get_rpccred(sp->so_cred);
225 spin_unlock(&clp->cl_lock);
229 static void nfs_alloc_unique_id(struct rb_root *root, struct nfs_unique_id *new,
230 __u64 minval, int maxbits)
232 struct rb_node **p, *parent;
233 struct nfs_unique_id *pos;
237 mask = (1ULL << maxbits) - 1ULL;
239 /* Ensure distribution is more or less flat */
240 get_random_bytes(&new->id, sizeof(new->id));
242 if (new->id < minval)
250 pos = rb_entry(parent, struct nfs_unique_id, rb_node);
252 if (new->id < pos->id)
254 else if (new->id > pos->id)
259 rb_link_node(&new->rb_node, parent, p);
260 rb_insert_color(&new->rb_node, root);
265 if (new->id < minval || (new->id & mask) != new->id) {
269 parent = rb_next(parent);
272 pos = rb_entry(parent, struct nfs_unique_id, rb_node);
273 if (new->id < pos->id)
279 static void nfs_free_unique_id(struct rb_root *root, struct nfs_unique_id *id)
281 rb_erase(&id->rb_node, root);
284 static struct nfs4_state_owner *
285 nfs4_find_state_owner(struct nfs_server *server, struct rpc_cred *cred)
287 struct nfs_client *clp = server->nfs_client;
288 struct rb_node **p = &clp->cl_state_owners.rb_node,
290 struct nfs4_state_owner *sp, *res = NULL;
294 sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
296 if (server < sp->so_server) {
297 p = &parent->rb_left;
300 if (server > sp->so_server) {
301 p = &parent->rb_right;
304 if (cred < sp->so_cred)
305 p = &parent->rb_left;
306 else if (cred > sp->so_cred)
307 p = &parent->rb_right;
309 atomic_inc(&sp->so_count);
317 static struct nfs4_state_owner *
318 nfs4_insert_state_owner(struct nfs_client *clp, struct nfs4_state_owner *new)
320 struct rb_node **p = &clp->cl_state_owners.rb_node,
322 struct nfs4_state_owner *sp;
326 sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
328 if (new->so_server < sp->so_server) {
329 p = &parent->rb_left;
332 if (new->so_server > sp->so_server) {
333 p = &parent->rb_right;
336 if (new->so_cred < sp->so_cred)
337 p = &parent->rb_left;
338 else if (new->so_cred > sp->so_cred)
339 p = &parent->rb_right;
341 atomic_inc(&sp->so_count);
345 nfs_alloc_unique_id(&clp->cl_openowner_id, &new->so_owner_id, 1, 64);
346 rb_link_node(&new->so_client_node, parent, p);
347 rb_insert_color(&new->so_client_node, &clp->cl_state_owners);
352 nfs4_remove_state_owner(struct nfs_client *clp, struct nfs4_state_owner *sp)
354 if (!RB_EMPTY_NODE(&sp->so_client_node))
355 rb_erase(&sp->so_client_node, &clp->cl_state_owners);
356 nfs_free_unique_id(&clp->cl_openowner_id, &sp->so_owner_id);
360 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
361 * create a new state_owner.
364 static struct nfs4_state_owner *
365 nfs4_alloc_state_owner(void)
367 struct nfs4_state_owner *sp;
369 sp = kzalloc(sizeof(*sp),GFP_KERNEL);
372 spin_lock_init(&sp->so_lock);
373 INIT_LIST_HEAD(&sp->so_states);
374 INIT_LIST_HEAD(&sp->so_delegations);
375 rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue");
376 sp->so_seqid.sequence = &sp->so_sequence;
377 spin_lock_init(&sp->so_sequence.lock);
378 INIT_LIST_HEAD(&sp->so_sequence.list);
379 atomic_set(&sp->so_count, 1);
384 nfs4_drop_state_owner(struct nfs4_state_owner *sp)
386 if (!RB_EMPTY_NODE(&sp->so_client_node)) {
387 struct nfs_client *clp = sp->so_client;
389 spin_lock(&clp->cl_lock);
390 rb_erase(&sp->so_client_node, &clp->cl_state_owners);
391 RB_CLEAR_NODE(&sp->so_client_node);
392 spin_unlock(&clp->cl_lock);
396 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct rpc_cred *cred)
398 struct nfs_client *clp = server->nfs_client;
399 struct nfs4_state_owner *sp, *new;
401 spin_lock(&clp->cl_lock);
402 sp = nfs4_find_state_owner(server, cred);
403 spin_unlock(&clp->cl_lock);
406 new = nfs4_alloc_state_owner();
409 new->so_client = clp;
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_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_KERNEL);
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, fmode_t fmode, int wait)
547 struct nfs4_state_owner *owner = state->owner;
551 atomic_inc(&owner->so_count);
552 /* Protect against nfs4_find_state() */
553 spin_lock(&owner->so_lock);
554 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
561 case FMODE_READ|FMODE_WRITE:
564 newstate = FMODE_READ|FMODE_WRITE;
565 if (state->n_rdwr == 0) {
566 if (state->n_rdonly == 0) {
567 newstate &= ~FMODE_READ;
568 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
569 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
571 if (state->n_wronly == 0) {
572 newstate &= ~FMODE_WRITE;
573 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
574 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
577 clear_bit(NFS_DELEGATED_STATE, &state->flags);
579 nfs4_state_set_mode_locked(state, newstate);
580 spin_unlock(&owner->so_lock);
583 nfs4_put_open_state(state);
584 nfs4_put_state_owner(owner);
586 nfs4_do_close(path, state, wait);
589 void nfs4_close_state(struct path *path, struct nfs4_state *state, fmode_t fmode)
591 __nfs4_close(path, state, fmode, 0);
594 void nfs4_close_sync(struct path *path, struct nfs4_state *state, fmode_t fmode)
596 __nfs4_close(path, state, fmode, 1);
600 * Search the state->lock_states for an existing lock_owner
601 * that is compatible with current->files
603 static struct nfs4_lock_state *
604 __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
606 struct nfs4_lock_state *pos;
607 list_for_each_entry(pos, &state->lock_states, ls_locks) {
608 if (pos->ls_owner != fl_owner)
610 atomic_inc(&pos->ls_count);
617 * Return a compatible lock_state. If no initialized lock_state structure
618 * exists, return an uninitialized one.
621 static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
623 struct nfs4_lock_state *lsp;
624 struct nfs_client *clp = state->owner->so_client;
626 lsp = kzalloc(sizeof(*lsp), GFP_KERNEL);
629 rpc_init_wait_queue(&lsp->ls_sequence.wait, "lock_seqid_waitqueue");
630 spin_lock_init(&lsp->ls_sequence.lock);
631 INIT_LIST_HEAD(&lsp->ls_sequence.list);
632 lsp->ls_seqid.sequence = &lsp->ls_sequence;
633 atomic_set(&lsp->ls_count, 1);
634 lsp->ls_state = state;
635 lsp->ls_owner = fl_owner;
636 spin_lock(&clp->cl_lock);
637 nfs_alloc_unique_id(&clp->cl_lockowner_id, &lsp->ls_id, 1, 64);
638 spin_unlock(&clp->cl_lock);
639 INIT_LIST_HEAD(&lsp->ls_locks);
643 static void nfs4_free_lock_state(struct nfs4_lock_state *lsp)
645 struct nfs_client *clp = lsp->ls_state->owner->so_client;
647 spin_lock(&clp->cl_lock);
648 nfs_free_unique_id(&clp->cl_lockowner_id, &lsp->ls_id);
649 spin_unlock(&clp->cl_lock);
650 rpc_destroy_wait_queue(&lsp->ls_sequence.wait);
655 * Return a compatible lock_state. If no initialized lock_state structure
656 * exists, return an uninitialized one.
659 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
661 struct nfs4_lock_state *lsp, *new = NULL;
664 spin_lock(&state->state_lock);
665 lsp = __nfs4_find_lock_state(state, owner);
669 list_add(&new->ls_locks, &state->lock_states);
670 set_bit(LK_STATE_IN_USE, &state->flags);
675 spin_unlock(&state->state_lock);
676 new = nfs4_alloc_lock_state(state, owner);
680 spin_unlock(&state->state_lock);
682 nfs4_free_lock_state(new);
687 * Release reference to lock_state, and free it if we see that
688 * it is no longer in use
690 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
692 struct nfs4_state *state;
696 state = lsp->ls_state;
697 if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
699 list_del(&lsp->ls_locks);
700 if (list_empty(&state->lock_states))
701 clear_bit(LK_STATE_IN_USE, &state->flags);
702 spin_unlock(&state->state_lock);
703 nfs4_free_lock_state(lsp);
706 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
708 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
710 dst->fl_u.nfs4_fl.owner = lsp;
711 atomic_inc(&lsp->ls_count);
714 static void nfs4_fl_release_lock(struct file_lock *fl)
716 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
719 static const struct file_lock_operations nfs4_fl_lock_ops = {
720 .fl_copy_lock = nfs4_fl_copy_lock,
721 .fl_release_private = nfs4_fl_release_lock,
724 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
726 struct nfs4_lock_state *lsp;
728 if (fl->fl_ops != NULL)
730 lsp = nfs4_get_lock_state(state, fl->fl_owner);
733 fl->fl_u.nfs4_fl.owner = lsp;
734 fl->fl_ops = &nfs4_fl_lock_ops;
739 * Byte-range lock aware utility to initialize the stateid of read/write
742 void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner)
744 struct nfs4_lock_state *lsp;
748 seq = read_seqbegin(&state->seqlock);
749 memcpy(dst, &state->stateid, sizeof(*dst));
750 } while (read_seqretry(&state->seqlock, seq));
751 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
754 spin_lock(&state->state_lock);
755 lsp = __nfs4_find_lock_state(state, fl_owner);
756 if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
757 memcpy(dst, &lsp->ls_stateid, sizeof(*dst));
758 spin_unlock(&state->state_lock);
759 nfs4_put_lock_state(lsp);
762 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter)
764 struct nfs_seqid *new;
766 new = kmalloc(sizeof(*new), GFP_KERNEL);
768 new->sequence = counter;
769 INIT_LIST_HEAD(&new->list);
774 void nfs_release_seqid(struct nfs_seqid *seqid)
776 if (!list_empty(&seqid->list)) {
777 struct rpc_sequence *sequence = seqid->sequence->sequence;
779 spin_lock(&sequence->lock);
780 list_del_init(&seqid->list);
781 spin_unlock(&sequence->lock);
782 rpc_wake_up(&sequence->wait);
786 void nfs_free_seqid(struct nfs_seqid *seqid)
788 nfs_release_seqid(seqid);
793 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
794 * failed with a seqid incrementing error -
795 * see comments nfs_fs.h:seqid_mutating_error()
797 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
799 BUG_ON(list_first_entry(&seqid->sequence->sequence->list, struct nfs_seqid, list) != seqid);
803 case -NFS4ERR_BAD_SEQID:
804 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
806 printk(KERN_WARNING "NFS: v4 server returned a bad"
807 " sequence-id error on an"
808 " unconfirmed sequence %p!\n",
810 case -NFS4ERR_STALE_CLIENTID:
811 case -NFS4ERR_STALE_STATEID:
812 case -NFS4ERR_BAD_STATEID:
813 case -NFS4ERR_BADXDR:
814 case -NFS4ERR_RESOURCE:
815 case -NFS4ERR_NOFILEHANDLE:
816 /* Non-seqid mutating errors */
820 * Note: no locking needed as we are guaranteed to be first
821 * on the sequence list
823 seqid->sequence->counter++;
826 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
828 struct nfs4_state_owner *sp = container_of(seqid->sequence,
829 struct nfs4_state_owner, so_seqid);
830 struct nfs_server *server = sp->so_server;
832 if (status == -NFS4ERR_BAD_SEQID)
833 nfs4_drop_state_owner(sp);
834 if (!nfs4_has_session(server->nfs_client))
835 nfs_increment_seqid(status, seqid);
839 * Increment the seqid if the LOCK/LOCKU succeeded, or
840 * failed with a seqid incrementing error -
841 * see comments nfs_fs.h:seqid_mutating_error()
843 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
845 nfs_increment_seqid(status, seqid);
848 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
850 struct rpc_sequence *sequence = seqid->sequence->sequence;
853 spin_lock(&sequence->lock);
854 if (list_empty(&seqid->list))
855 list_add_tail(&seqid->list, &sequence->list);
856 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
858 rpc_sleep_on(&sequence->wait, task, NULL);
861 spin_unlock(&sequence->lock);
865 static int nfs4_run_state_manager(void *);
867 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
869 smp_mb__before_clear_bit();
870 clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
871 smp_mb__after_clear_bit();
872 wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
873 rpc_wake_up(&clp->cl_rpcwaitq);
877 * Schedule the nfs_client asynchronous state management routine
879 void nfs4_schedule_state_manager(struct nfs_client *clp)
881 struct task_struct *task;
883 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
885 __module_get(THIS_MODULE);
886 atomic_inc(&clp->cl_count);
887 task = kthread_run(nfs4_run_state_manager, clp, "%s-manager",
888 rpc_peeraddr2str(clp->cl_rpcclient,
892 nfs4_clear_state_manager_bit(clp);
894 module_put(THIS_MODULE);
898 * Schedule a state recovery attempt
900 void nfs4_schedule_state_recovery(struct nfs_client *clp)
904 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
905 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
906 nfs4_schedule_state_manager(clp);
909 int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
912 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
913 /* Don't recover state that expired before the reboot */
914 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
915 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
918 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
919 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
923 int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
925 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
926 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
927 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
928 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
932 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
934 struct inode *inode = state->inode;
935 struct nfs_inode *nfsi = NFS_I(inode);
936 struct file_lock *fl;
939 if (inode->i_flock == NULL)
942 /* Guard against delegation returns and new lock/unlock calls */
943 down_write(&nfsi->rwsem);
944 /* Protect inode->i_flock using the BKL */
946 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
947 if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
949 if (nfs_file_open_context(fl->fl_file)->state != state)
952 status = ops->recover_lock(state, fl);
957 case -NFS4ERR_ADMIN_REVOKED:
958 case -NFS4ERR_STALE_STATEID:
959 case -NFS4ERR_BAD_STATEID:
960 case -NFS4ERR_EXPIRED:
961 case -NFS4ERR_NO_GRACE:
962 case -NFS4ERR_STALE_CLIENTID:
963 case -NFS4ERR_BADSESSION:
964 case -NFS4ERR_BADSLOT:
965 case -NFS4ERR_BAD_HIGH_SLOT:
966 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
969 printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
972 case -NFS4ERR_DENIED:
973 case -NFS4ERR_RECLAIM_BAD:
974 case -NFS4ERR_RECLAIM_CONFLICT:
975 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
982 up_write(&nfsi->rwsem);
986 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
988 struct nfs4_state *state;
989 struct nfs4_lock_state *lock;
992 /* Note: we rely on the sp->so_states list being ordered
993 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
995 * This is needed to ensure that the server won't give us any
996 * read delegations that we have to return if, say, we are
997 * recovering after a network partition or a reboot from a
998 * server that doesn't support a grace period.
1001 spin_lock(&sp->so_lock);
1002 list_for_each_entry(state, &sp->so_states, open_states) {
1003 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
1005 if (state->state == 0)
1007 atomic_inc(&state->count);
1008 spin_unlock(&sp->so_lock);
1009 status = ops->recover_open(sp, state);
1011 status = nfs4_reclaim_locks(state, ops);
1013 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1014 if (!(lock->ls_flags & NFS_LOCK_INITIALIZED))
1015 printk("%s: Lock reclaim failed!\n",
1018 nfs4_put_open_state(state);
1024 printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
1030 * Open state on this file cannot be recovered
1031 * All we can do is revert to using the zero stateid.
1033 memset(state->stateid.data, 0,
1034 sizeof(state->stateid.data));
1035 /* Mark the file as being 'closed' */
1038 case -NFS4ERR_ADMIN_REVOKED:
1039 case -NFS4ERR_STALE_STATEID:
1040 case -NFS4ERR_BAD_STATEID:
1041 case -NFS4ERR_RECLAIM_BAD:
1042 case -NFS4ERR_RECLAIM_CONFLICT:
1043 nfs4_state_mark_reclaim_nograce(sp->so_client, state);
1045 case -NFS4ERR_EXPIRED:
1046 case -NFS4ERR_NO_GRACE:
1047 nfs4_state_mark_reclaim_nograce(sp->so_client, state);
1048 case -NFS4ERR_STALE_CLIENTID:
1049 case -NFS4ERR_BADSESSION:
1050 case -NFS4ERR_BADSLOT:
1051 case -NFS4ERR_BAD_HIGH_SLOT:
1052 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1055 nfs4_put_open_state(state);
1058 spin_unlock(&sp->so_lock);
1061 nfs4_put_open_state(state);
1065 static void nfs4_clear_open_state(struct nfs4_state *state)
1067 struct nfs4_lock_state *lock;
1069 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1070 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1071 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1072 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1073 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1074 lock->ls_seqid.flags = 0;
1075 lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
1079 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp, int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1081 struct nfs4_state_owner *sp;
1082 struct rb_node *pos;
1083 struct nfs4_state *state;
1085 /* Reset all sequence ids to zero */
1086 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1087 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1088 sp->so_seqid.flags = 0;
1089 spin_lock(&sp->so_lock);
1090 list_for_each_entry(state, &sp->so_states, open_states) {
1091 if (mark_reclaim(clp, state))
1092 nfs4_clear_open_state(state);
1094 spin_unlock(&sp->so_lock);
1098 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1100 /* Mark all delegations for reclaim */
1101 nfs_delegation_mark_reclaim(clp);
1102 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1105 static void nfs4_reclaim_complete(struct nfs_client *clp,
1106 const struct nfs4_state_recovery_ops *ops)
1108 /* Notify the server we're done reclaiming our state */
1109 if (ops->reclaim_complete)
1110 (void)ops->reclaim_complete(clp);
1113 static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
1115 struct nfs4_state_owner *sp;
1116 struct rb_node *pos;
1117 struct nfs4_state *state;
1119 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1122 nfs4_reclaim_complete(clp,
1123 nfs4_reboot_recovery_ops[clp->cl_minorversion]);
1125 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1126 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1127 spin_lock(&sp->so_lock);
1128 list_for_each_entry(state, &sp->so_states, open_states) {
1129 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags))
1131 nfs4_state_mark_reclaim_nograce(clp, state);
1133 spin_unlock(&sp->so_lock);
1136 nfs_delegation_reap_unclaimed(clp);
1139 static void nfs_delegation_clear_all(struct nfs_client *clp)
1141 nfs_delegation_mark_reclaim(clp);
1142 nfs_delegation_reap_unclaimed(clp);
1145 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1147 nfs_delegation_clear_all(clp);
1148 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1151 static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1154 case -NFS4ERR_CB_PATH_DOWN:
1155 nfs_handle_cb_pathdown(clp);
1157 case -NFS4ERR_NO_GRACE:
1158 nfs4_state_end_reclaim_reboot(clp);
1160 case -NFS4ERR_STALE_CLIENTID:
1161 case -NFS4ERR_LEASE_MOVED:
1162 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1163 nfs4_state_end_reclaim_reboot(clp);
1164 nfs4_state_start_reclaim_reboot(clp);
1166 case -NFS4ERR_EXPIRED:
1167 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1168 nfs4_state_start_reclaim_nograce(clp);
1170 case -NFS4ERR_BADSESSION:
1171 case -NFS4ERR_BADSLOT:
1172 case -NFS4ERR_BAD_HIGH_SLOT:
1173 case -NFS4ERR_DEADSESSION:
1174 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1175 case -NFS4ERR_SEQ_FALSE_RETRY:
1176 case -NFS4ERR_SEQ_MISORDERED:
1177 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1178 /* Zero session reset errors */
1184 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1186 struct rb_node *pos;
1190 spin_lock(&clp->cl_lock);
1191 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
1192 struct nfs4_state_owner *sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1193 if (!test_and_clear_bit(ops->owner_flag_bit, &sp->so_flags))
1195 atomic_inc(&sp->so_count);
1196 spin_unlock(&clp->cl_lock);
1197 status = nfs4_reclaim_open_state(sp, ops);
1199 set_bit(ops->owner_flag_bit, &sp->so_flags);
1200 nfs4_put_state_owner(sp);
1201 return nfs4_recovery_handle_error(clp, status);
1203 nfs4_put_state_owner(sp);
1206 spin_unlock(&clp->cl_lock);
1210 static int nfs4_check_lease(struct nfs_client *clp)
1212 struct rpc_cred *cred;
1213 struct nfs4_state_maintenance_ops *ops =
1214 nfs4_state_renewal_ops[clp->cl_minorversion];
1215 int status = -NFS4ERR_EXPIRED;
1217 /* Is the client already known to have an expired lease? */
1218 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1220 spin_lock(&clp->cl_lock);
1221 cred = ops->get_state_renewal_cred_locked(clp);
1222 spin_unlock(&clp->cl_lock);
1224 cred = nfs4_get_setclientid_cred(clp);
1228 status = ops->renew_lease(clp, cred);
1231 return nfs4_recovery_handle_error(clp, status);
1234 static int nfs4_reclaim_lease(struct nfs_client *clp)
1236 struct rpc_cred *cred;
1237 struct nfs4_state_recovery_ops *ops =
1238 nfs4_reboot_recovery_ops[clp->cl_minorversion];
1239 int status = -ENOENT;
1241 cred = ops->get_clid_cred(clp);
1243 status = ops->establish_clid(clp, cred);
1245 /* Handle case where the user hasn't set up machine creds */
1246 if (status == -EACCES && cred == clp->cl_machine_cred) {
1247 nfs4_clear_machine_cred(clp);
1250 if (status == -NFS4ERR_MINOR_VERS_MISMATCH)
1251 status = -EPROTONOSUPPORT;
1256 #ifdef CONFIG_NFS_V4_1
1257 void nfs41_handle_recall_slot(struct nfs_client *clp)
1259 set_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
1260 nfs4_schedule_state_recovery(clp);
1263 static void nfs4_reset_all_state(struct nfs_client *clp)
1265 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1266 clp->cl_boot_time = CURRENT_TIME;
1267 nfs4_state_start_reclaim_nograce(clp);
1268 nfs4_schedule_state_recovery(clp);
1272 static void nfs41_handle_server_reboot(struct nfs_client *clp)
1274 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1275 nfs4_state_start_reclaim_reboot(clp);
1276 nfs4_schedule_state_recovery(clp);
1280 static void nfs41_handle_state_revoked(struct nfs_client *clp)
1283 nfs4_reset_all_state(clp);
1286 static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
1288 /* This will need to handle layouts too */
1289 nfs_expire_all_delegations(clp);
1292 static void nfs41_handle_cb_path_down(struct nfs_client *clp)
1294 nfs_expire_all_delegations(clp);
1295 if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
1296 nfs4_schedule_state_recovery(clp);
1299 void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
1303 else if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
1304 nfs41_handle_server_reboot(clp);
1305 else if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
1306 SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
1307 SEQ4_STATUS_ADMIN_STATE_REVOKED |
1308 SEQ4_STATUS_LEASE_MOVED))
1309 nfs41_handle_state_revoked(clp);
1310 else if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
1311 nfs41_handle_recallable_state_revoked(clp);
1312 else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
1313 SEQ4_STATUS_BACKCHANNEL_FAULT |
1314 SEQ4_STATUS_CB_PATH_DOWN_SESSION))
1315 nfs41_handle_cb_path_down(clp);
1318 static int nfs4_reset_session(struct nfs_client *clp)
1322 nfs4_begin_drain_session(clp);
1323 status = nfs4_proc_destroy_session(clp->cl_session);
1324 if (status && status != -NFS4ERR_BADSESSION &&
1325 status != -NFS4ERR_DEADSESSION) {
1326 status = nfs4_recovery_handle_error(clp, status);
1330 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
1331 status = nfs4_proc_create_session(clp);
1333 status = nfs4_recovery_handle_error(clp, status);
1336 /* create_session negotiated new slot table */
1337 clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
1339 /* Let the state manager reestablish state */
1340 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1341 nfs41_setup_state_renewal(clp);
1346 static int nfs4_recall_slot(struct nfs_client *clp)
1348 struct nfs4_slot_table *fc_tbl = &clp->cl_session->fc_slot_table;
1349 struct nfs4_channel_attrs *fc_attrs = &clp->cl_session->fc_attrs;
1350 struct nfs4_slot *new, *old;
1353 nfs4_begin_drain_session(clp);
1354 new = kmalloc(fc_tbl->target_max_slots * sizeof(struct nfs4_slot),
1359 spin_lock(&fc_tbl->slot_tbl_lock);
1360 for (i = 0; i < fc_tbl->target_max_slots; i++)
1361 new[i].seq_nr = fc_tbl->slots[i].seq_nr;
1362 old = fc_tbl->slots;
1363 fc_tbl->slots = new;
1364 fc_tbl->max_slots = fc_tbl->target_max_slots;
1365 fc_tbl->target_max_slots = 0;
1366 fc_attrs->max_reqs = fc_tbl->max_slots;
1367 spin_unlock(&fc_tbl->slot_tbl_lock);
1370 nfs4_end_drain_session(clp);
1374 #else /* CONFIG_NFS_V4_1 */
1375 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
1376 static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
1377 static int nfs4_recall_slot(struct nfs_client *clp) { return 0; }
1378 #endif /* CONFIG_NFS_V4_1 */
1380 /* Set NFS4CLNT_LEASE_EXPIRED for all v4.0 errors and for recoverable errors
1381 * on EXCHANGE_ID for v4.1
1383 static void nfs4_set_lease_expired(struct nfs_client *clp, int status)
1385 if (nfs4_has_session(clp)) {
1387 case -NFS4ERR_DELAY:
1388 case -NFS4ERR_CLID_INUSE:
1393 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1394 * in nfs4_exchange_id */
1399 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1402 static void nfs4_state_manager(struct nfs_client *clp)
1406 /* Ensure exclusive access to NFSv4 state */
1408 if (test_and_clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
1409 /* We're going to have to re-establish a clientid */
1410 status = nfs4_reclaim_lease(clp);
1412 nfs4_set_lease_expired(clp, status);
1413 if (test_bit(NFS4CLNT_LEASE_EXPIRED,
1416 if (clp->cl_cons_state ==
1417 NFS_CS_SESSION_INITING)
1418 nfs_mark_client_ready(clp, status);
1421 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1422 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1425 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
1426 status = nfs4_check_lease(clp);
1427 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1429 if (status < 0 && status != -NFS4ERR_CB_PATH_DOWN)
1433 /* Initialize or reset the session */
1434 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)
1435 && nfs4_has_session(clp)) {
1436 status = nfs4_reset_session(clp);
1437 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1443 /* First recover reboot state... */
1444 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
1445 status = nfs4_do_reclaim(clp,
1446 nfs4_reboot_recovery_ops[clp->cl_minorversion]);
1447 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1448 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
1450 nfs4_state_end_reclaim_reboot(clp);
1451 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
1457 /* Now recover expired state... */
1458 if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
1459 status = nfs4_do_reclaim(clp,
1460 nfs4_nograce_recovery_ops[clp->cl_minorversion]);
1461 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1462 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
1463 test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1469 nfs4_end_drain_session(clp);
1470 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
1471 nfs_client_return_marked_delegations(clp);
1474 /* Recall session slots */
1475 if (test_and_clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state)
1476 && nfs4_has_session(clp)) {
1477 status = nfs4_recall_slot(clp);
1484 nfs4_clear_state_manager_bit(clp);
1485 /* Did we race with an attempt to give us more work? */
1486 if (clp->cl_state == 0)
1488 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1493 printk(KERN_WARNING "Error: state manager failed on NFSv4 server %s"
1494 " with error %d\n", clp->cl_hostname, -status);
1495 nfs4_end_drain_session(clp);
1496 nfs4_clear_state_manager_bit(clp);
1499 static int nfs4_run_state_manager(void *ptr)
1501 struct nfs_client *clp = ptr;
1503 allow_signal(SIGKILL);
1504 nfs4_state_manager(clp);
1505 nfs_put_client(clp);
1506 module_put_and_exit(0);