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>
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/ratelimit.h>
50 #include <linux/workqueue.h>
51 #include <linux/bitops.h>
52 #include <linux/jiffies.h>
54 #include <linux/sunrpc/clnt.h>
58 #include "delegation.h"
60 #include "nfs4session.h"
64 #define NFSDBG_FACILITY NFSDBG_STATE
66 #define OPENOWNER_POOL_SIZE 8
68 const nfs4_stateid zero_stateid;
69 static DEFINE_MUTEX(nfs_clid_init_mutex);
71 int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
73 struct nfs4_setclientid_res clid = {
74 .clientid = clp->cl_clientid,
75 .confirm = clp->cl_confirm,
79 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
81 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
83 port = nn->nfs_callback_tcpport;
84 if (clp->cl_addr.ss_family == AF_INET6)
85 port = nn->nfs_callback_tcpport6;
87 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
90 clp->cl_clientid = clid.clientid;
91 clp->cl_confirm = clid.confirm;
92 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
94 status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
97 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
98 nfs4_schedule_state_renewal(clp);
104 * nfs40_discover_server_trunking - Detect server IP address trunking (mv0)
106 * @clp: nfs_client under test
107 * @result: OUT: found nfs_client, or clp
108 * @cred: credential to use for trunking test
110 * Returns zero, a negative errno, or a negative NFS4ERR status.
111 * If zero is returned, an nfs_client pointer is planted in
114 * Note: The returned client may not yet be marked ready.
116 int nfs40_discover_server_trunking(struct nfs_client *clp,
117 struct nfs_client **result,
118 struct rpc_cred *cred)
120 struct nfs4_setclientid_res clid = {
121 .clientid = clp->cl_clientid,
122 .confirm = clp->cl_confirm,
124 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
128 port = nn->nfs_callback_tcpport;
129 if (clp->cl_addr.ss_family == AF_INET6)
130 port = nn->nfs_callback_tcpport6;
132 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
135 clp->cl_clientid = clid.clientid;
136 clp->cl_confirm = clid.confirm;
138 status = nfs40_walk_client_list(clp, result, cred);
140 /* Sustain the lease, even if it's empty. If the clientid4
141 * goes stale it's of no use for trunking discovery. */
142 nfs4_schedule_state_renewal(*result);
148 struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
150 struct rpc_cred *cred = NULL;
152 if (clp->cl_machine_cred != NULL)
153 cred = get_rpccred(clp->cl_machine_cred);
157 static struct rpc_cred *
158 nfs4_get_renew_cred_server_locked(struct nfs_server *server)
160 struct rpc_cred *cred = NULL;
161 struct nfs4_state_owner *sp;
164 for (pos = rb_first(&server->state_owners);
166 pos = rb_next(pos)) {
167 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
168 if (list_empty(&sp->so_states))
170 cred = get_rpccred(sp->so_cred);
177 * nfs4_get_renew_cred_locked - Acquire credential for a renew operation
178 * @clp: client state handle
180 * Returns an rpc_cred with reference count bumped, or NULL.
181 * Caller must hold clp->cl_lock.
183 struct rpc_cred *nfs4_get_renew_cred_locked(struct nfs_client *clp)
185 struct rpc_cred *cred = NULL;
186 struct nfs_server *server;
188 /* Use machine credentials if available */
189 cred = nfs4_get_machine_cred_locked(clp);
194 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
195 cred = nfs4_get_renew_cred_server_locked(server);
205 #if defined(CONFIG_NFS_V4_1)
207 static int nfs41_setup_state_renewal(struct nfs_client *clp)
210 struct nfs_fsinfo fsinfo;
212 if (!test_bit(NFS_CS_CHECK_LEASE_TIME, &clp->cl_res_state)) {
213 nfs4_schedule_state_renewal(clp);
217 status = nfs4_proc_get_lease_time(clp, &fsinfo);
219 /* Update lease time and schedule renewal */
220 spin_lock(&clp->cl_lock);
221 clp->cl_lease_time = fsinfo.lease_time * HZ;
222 clp->cl_last_renewal = jiffies;
223 spin_unlock(&clp->cl_lock);
225 nfs4_schedule_state_renewal(clp);
232 * Back channel returns NFS4ERR_DELAY for new requests when
233 * NFS4_SESSION_DRAINING is set so there is no work to be done when draining
236 static void nfs4_end_drain_session(struct nfs_client *clp)
238 struct nfs4_session *ses = clp->cl_session;
239 struct nfs4_slot_table *tbl;
243 tbl = &ses->fc_slot_table;
244 if (test_and_clear_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
245 spin_lock(&tbl->slot_tbl_lock);
246 nfs41_wake_slot_table(tbl);
247 spin_unlock(&tbl->slot_tbl_lock);
252 * Signal state manager thread if session fore channel is drained
254 void nfs4_session_drain_complete(struct nfs4_session *session,
255 struct nfs4_slot_table *tbl)
257 if (nfs4_session_draining(session))
258 complete(&tbl->complete);
261 static int nfs4_wait_on_slot_tbl(struct nfs4_slot_table *tbl)
263 spin_lock(&tbl->slot_tbl_lock);
264 if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
265 INIT_COMPLETION(tbl->complete);
266 spin_unlock(&tbl->slot_tbl_lock);
267 return wait_for_completion_interruptible(&tbl->complete);
269 spin_unlock(&tbl->slot_tbl_lock);
273 static int nfs4_begin_drain_session(struct nfs_client *clp)
275 struct nfs4_session *ses = clp->cl_session;
278 set_bit(NFS4_SESSION_DRAINING, &ses->session_state);
280 ret = nfs4_wait_on_slot_tbl(&ses->bc_slot_table);
284 return nfs4_wait_on_slot_tbl(&ses->fc_slot_table);
287 static void nfs41_finish_session_reset(struct nfs_client *clp)
289 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
290 clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
291 /* create_session negotiated new slot table */
292 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
293 nfs41_setup_state_renewal(clp);
296 int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
300 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
302 nfs4_begin_drain_session(clp);
303 status = nfs4_proc_exchange_id(clp, cred);
306 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
308 status = nfs4_proc_create_session(clp, cred);
311 nfs41_finish_session_reset(clp);
312 nfs_mark_client_ready(clp, NFS_CS_READY);
318 * nfs41_discover_server_trunking - Detect server IP address trunking (mv1)
320 * @clp: nfs_client under test
321 * @result: OUT: found nfs_client, or clp
322 * @cred: credential to use for trunking test
324 * Returns NFS4_OK, a negative errno, or a negative NFS4ERR status.
325 * If NFS4_OK is returned, an nfs_client pointer is planted in
328 * Note: The returned client may not yet be marked ready.
330 int nfs41_discover_server_trunking(struct nfs_client *clp,
331 struct nfs_client **result,
332 struct rpc_cred *cred)
336 status = nfs4_proc_exchange_id(clp, cred);
337 if (status != NFS4_OK)
339 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
341 return nfs41_walk_client_list(clp, result, cred);
344 struct rpc_cred *nfs4_get_exchange_id_cred(struct nfs_client *clp)
346 struct rpc_cred *cred;
348 spin_lock(&clp->cl_lock);
349 cred = nfs4_get_machine_cred_locked(clp);
350 spin_unlock(&clp->cl_lock);
354 #endif /* CONFIG_NFS_V4_1 */
356 static struct rpc_cred *
357 nfs4_get_setclientid_cred_server(struct nfs_server *server)
359 struct nfs_client *clp = server->nfs_client;
360 struct rpc_cred *cred = NULL;
361 struct nfs4_state_owner *sp;
364 spin_lock(&clp->cl_lock);
365 pos = rb_first(&server->state_owners);
367 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
368 cred = get_rpccred(sp->so_cred);
370 spin_unlock(&clp->cl_lock);
375 * nfs4_get_setclientid_cred - Acquire credential for a setclientid operation
376 * @clp: client state handle
378 * Returns an rpc_cred with reference count bumped, or NULL.
380 struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
382 struct nfs_server *server;
383 struct rpc_cred *cred;
385 spin_lock(&clp->cl_lock);
386 cred = nfs4_get_machine_cred_locked(clp);
387 spin_unlock(&clp->cl_lock);
392 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
393 cred = nfs4_get_setclientid_cred_server(server);
403 static struct nfs4_state_owner *
404 nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
406 struct rb_node **p = &server->state_owners.rb_node,
408 struct nfs4_state_owner *sp;
412 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
414 if (cred < sp->so_cred)
415 p = &parent->rb_left;
416 else if (cred > sp->so_cred)
417 p = &parent->rb_right;
419 if (!list_empty(&sp->so_lru))
420 list_del_init(&sp->so_lru);
421 atomic_inc(&sp->so_count);
428 static struct nfs4_state_owner *
429 nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
431 struct nfs_server *server = new->so_server;
432 struct rb_node **p = &server->state_owners.rb_node,
434 struct nfs4_state_owner *sp;
439 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
441 if (new->so_cred < sp->so_cred)
442 p = &parent->rb_left;
443 else if (new->so_cred > sp->so_cred)
444 p = &parent->rb_right;
446 if (!list_empty(&sp->so_lru))
447 list_del_init(&sp->so_lru);
448 atomic_inc(&sp->so_count);
452 err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
455 rb_link_node(&new->so_server_node, parent, p);
456 rb_insert_color(&new->so_server_node, &server->state_owners);
461 nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
463 struct nfs_server *server = sp->so_server;
465 if (!RB_EMPTY_NODE(&sp->so_server_node))
466 rb_erase(&sp->so_server_node, &server->state_owners);
467 ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
471 nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
473 sc->create_time = ktime_get();
476 spin_lock_init(&sc->lock);
477 INIT_LIST_HEAD(&sc->list);
478 rpc_init_wait_queue(&sc->wait, "Seqid_waitqueue");
482 nfs4_destroy_seqid_counter(struct nfs_seqid_counter *sc)
484 rpc_destroy_wait_queue(&sc->wait);
488 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
489 * create a new state_owner.
492 static struct nfs4_state_owner *
493 nfs4_alloc_state_owner(struct nfs_server *server,
494 struct rpc_cred *cred,
497 struct nfs4_state_owner *sp;
499 sp = kzalloc(sizeof(*sp), gfp_flags);
502 sp->so_server = server;
503 sp->so_cred = get_rpccred(cred);
504 spin_lock_init(&sp->so_lock);
505 INIT_LIST_HEAD(&sp->so_states);
506 nfs4_init_seqid_counter(&sp->so_seqid);
507 atomic_set(&sp->so_count, 1);
508 INIT_LIST_HEAD(&sp->so_lru);
509 seqcount_init(&sp->so_reclaim_seqcount);
510 mutex_init(&sp->so_delegreturn_mutex);
515 nfs4_drop_state_owner(struct nfs4_state_owner *sp)
517 struct rb_node *rb_node = &sp->so_server_node;
519 if (!RB_EMPTY_NODE(rb_node)) {
520 struct nfs_server *server = sp->so_server;
521 struct nfs_client *clp = server->nfs_client;
523 spin_lock(&clp->cl_lock);
524 if (!RB_EMPTY_NODE(rb_node)) {
525 rb_erase(rb_node, &server->state_owners);
526 RB_CLEAR_NODE(rb_node);
528 spin_unlock(&clp->cl_lock);
532 static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
534 nfs4_destroy_seqid_counter(&sp->so_seqid);
535 put_rpccred(sp->so_cred);
539 static void nfs4_gc_state_owners(struct nfs_server *server)
541 struct nfs_client *clp = server->nfs_client;
542 struct nfs4_state_owner *sp, *tmp;
543 unsigned long time_min, time_max;
546 spin_lock(&clp->cl_lock);
548 time_min = (long)time_max - (long)clp->cl_lease_time;
549 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
550 /* NB: LRU is sorted so that oldest is at the head */
551 if (time_in_range(sp->so_expires, time_min, time_max))
553 list_move(&sp->so_lru, &doomed);
554 nfs4_remove_state_owner_locked(sp);
556 spin_unlock(&clp->cl_lock);
558 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
559 list_del(&sp->so_lru);
560 nfs4_free_state_owner(sp);
565 * nfs4_get_state_owner - Look up a state owner given a credential
566 * @server: nfs_server to search
567 * @cred: RPC credential to match
569 * Returns a pointer to an instantiated nfs4_state_owner struct, or NULL.
571 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server,
572 struct rpc_cred *cred,
575 struct nfs_client *clp = server->nfs_client;
576 struct nfs4_state_owner *sp, *new;
578 spin_lock(&clp->cl_lock);
579 sp = nfs4_find_state_owner_locked(server, cred);
580 spin_unlock(&clp->cl_lock);
583 new = nfs4_alloc_state_owner(server, cred, gfp_flags);
587 if (ida_pre_get(&server->openowner_id, gfp_flags) == 0)
589 spin_lock(&clp->cl_lock);
590 sp = nfs4_insert_state_owner_locked(new);
591 spin_unlock(&clp->cl_lock);
592 } while (sp == ERR_PTR(-EAGAIN));
594 nfs4_free_state_owner(new);
596 nfs4_gc_state_owners(server);
601 * nfs4_put_state_owner - Release a nfs4_state_owner
602 * @sp: state owner data to release
604 * Note that we keep released state owners on an LRU
606 * This caches valid state owners so that they can be
607 * reused, to avoid the OPEN_CONFIRM on minor version 0.
608 * It also pins the uniquifier of dropped state owners for
609 * a while, to ensure that those state owner names are
612 void nfs4_put_state_owner(struct nfs4_state_owner *sp)
614 struct nfs_server *server = sp->so_server;
615 struct nfs_client *clp = server->nfs_client;
617 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
620 sp->so_expires = jiffies;
621 list_add_tail(&sp->so_lru, &server->state_owners_lru);
622 spin_unlock(&clp->cl_lock);
626 * nfs4_purge_state_owners - Release all cached state owners
627 * @server: nfs_server with cached state owners to release
629 * Called at umount time. Remaining state owners will be on
630 * the LRU with ref count of zero.
632 void nfs4_purge_state_owners(struct nfs_server *server)
634 struct nfs_client *clp = server->nfs_client;
635 struct nfs4_state_owner *sp, *tmp;
638 spin_lock(&clp->cl_lock);
639 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
640 list_move(&sp->so_lru, &doomed);
641 nfs4_remove_state_owner_locked(sp);
643 spin_unlock(&clp->cl_lock);
645 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
646 list_del(&sp->so_lru);
647 nfs4_free_state_owner(sp);
651 static struct nfs4_state *
652 nfs4_alloc_open_state(void)
654 struct nfs4_state *state;
656 state = kzalloc(sizeof(*state), GFP_NOFS);
659 atomic_set(&state->count, 1);
660 INIT_LIST_HEAD(&state->lock_states);
661 spin_lock_init(&state->state_lock);
662 seqlock_init(&state->seqlock);
667 nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
669 if (state->state == fmode)
671 /* NB! List reordering - see the reclaim code for why. */
672 if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
673 if (fmode & FMODE_WRITE)
674 list_move(&state->open_states, &state->owner->so_states);
676 list_move_tail(&state->open_states, &state->owner->so_states);
678 state->state = fmode;
681 static struct nfs4_state *
682 __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
684 struct nfs_inode *nfsi = NFS_I(inode);
685 struct nfs4_state *state;
687 list_for_each_entry(state, &nfsi->open_states, inode_states) {
688 if (state->owner != owner)
690 if (!nfs4_valid_open_stateid(state))
692 if (atomic_inc_not_zero(&state->count))
699 nfs4_free_open_state(struct nfs4_state *state)
705 nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
707 struct nfs4_state *state, *new;
708 struct nfs_inode *nfsi = NFS_I(inode);
710 spin_lock(&inode->i_lock);
711 state = __nfs4_find_state_byowner(inode, owner);
712 spin_unlock(&inode->i_lock);
715 new = nfs4_alloc_open_state();
716 spin_lock(&owner->so_lock);
717 spin_lock(&inode->i_lock);
718 state = __nfs4_find_state_byowner(inode, owner);
719 if (state == NULL && new != NULL) {
721 state->owner = owner;
722 atomic_inc(&owner->so_count);
723 list_add(&state->inode_states, &nfsi->open_states);
725 state->inode = inode;
726 spin_unlock(&inode->i_lock);
727 /* Note: The reclaim code dictates that we add stateless
728 * and read-only stateids to the end of the list */
729 list_add_tail(&state->open_states, &owner->so_states);
730 spin_unlock(&owner->so_lock);
732 spin_unlock(&inode->i_lock);
733 spin_unlock(&owner->so_lock);
735 nfs4_free_open_state(new);
741 void nfs4_put_open_state(struct nfs4_state *state)
743 struct inode *inode = state->inode;
744 struct nfs4_state_owner *owner = state->owner;
746 if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
748 spin_lock(&inode->i_lock);
749 list_del(&state->inode_states);
750 list_del(&state->open_states);
751 spin_unlock(&inode->i_lock);
752 spin_unlock(&owner->so_lock);
754 nfs4_free_open_state(state);
755 nfs4_put_state_owner(owner);
759 * Close the current file.
761 static void __nfs4_close(struct nfs4_state *state,
762 fmode_t fmode, gfp_t gfp_mask, int wait)
764 struct nfs4_state_owner *owner = state->owner;
768 atomic_inc(&owner->so_count);
769 /* Protect against nfs4_find_state() */
770 spin_lock(&owner->so_lock);
771 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
778 case FMODE_READ|FMODE_WRITE:
781 newstate = FMODE_READ|FMODE_WRITE;
782 if (state->n_rdwr == 0) {
783 if (state->n_rdonly == 0) {
784 newstate &= ~FMODE_READ;
785 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
786 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
788 if (state->n_wronly == 0) {
789 newstate &= ~FMODE_WRITE;
790 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
791 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
794 clear_bit(NFS_DELEGATED_STATE, &state->flags);
796 nfs4_state_set_mode_locked(state, newstate);
797 spin_unlock(&owner->so_lock);
800 nfs4_put_open_state(state);
801 nfs4_put_state_owner(owner);
803 nfs4_do_close(state, gfp_mask, wait);
806 void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
808 __nfs4_close(state, fmode, GFP_NOFS, 0);
811 void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
813 __nfs4_close(state, fmode, GFP_KERNEL, 1);
817 * Search the state->lock_states for an existing lock_owner
818 * that is compatible with current->files
820 static struct nfs4_lock_state *
821 __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type)
823 struct nfs4_lock_state *pos;
824 list_for_each_entry(pos, &state->lock_states, ls_locks) {
825 if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
827 switch (pos->ls_owner.lo_type) {
828 case NFS4_POSIX_LOCK_TYPE:
829 if (pos->ls_owner.lo_u.posix_owner != fl_owner)
832 case NFS4_FLOCK_LOCK_TYPE:
833 if (pos->ls_owner.lo_u.flock_owner != fl_pid)
836 atomic_inc(&pos->ls_count);
843 * Return a compatible lock_state. If no initialized lock_state structure
844 * exists, return an uninitialized one.
847 static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type)
849 struct nfs4_lock_state *lsp;
850 struct nfs_server *server = state->owner->so_server;
852 lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
855 nfs4_init_seqid_counter(&lsp->ls_seqid);
856 atomic_set(&lsp->ls_count, 1);
857 lsp->ls_state = state;
858 lsp->ls_owner.lo_type = type;
859 switch (lsp->ls_owner.lo_type) {
860 case NFS4_FLOCK_LOCK_TYPE:
861 lsp->ls_owner.lo_u.flock_owner = fl_pid;
863 case NFS4_POSIX_LOCK_TYPE:
864 lsp->ls_owner.lo_u.posix_owner = fl_owner;
869 lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
870 if (lsp->ls_seqid.owner_id < 0)
872 INIT_LIST_HEAD(&lsp->ls_locks);
879 void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
881 ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
882 nfs4_destroy_seqid_counter(&lsp->ls_seqid);
887 * Return a compatible lock_state. If no initialized lock_state structure
888 * exists, return an uninitialized one.
891 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner, pid_t pid, unsigned int type)
893 struct nfs4_lock_state *lsp, *new = NULL;
896 spin_lock(&state->state_lock);
897 lsp = __nfs4_find_lock_state(state, owner, pid, type);
901 list_add(&new->ls_locks, &state->lock_states);
902 set_bit(LK_STATE_IN_USE, &state->flags);
907 spin_unlock(&state->state_lock);
908 new = nfs4_alloc_lock_state(state, owner, pid, type);
912 spin_unlock(&state->state_lock);
914 nfs4_free_lock_state(state->owner->so_server, new);
919 * Release reference to lock_state, and free it if we see that
920 * it is no longer in use
922 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
924 struct nfs_server *server;
925 struct nfs4_state *state;
929 state = lsp->ls_state;
930 if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
932 list_del(&lsp->ls_locks);
933 if (list_empty(&state->lock_states))
934 clear_bit(LK_STATE_IN_USE, &state->flags);
935 spin_unlock(&state->state_lock);
936 server = state->owner->so_server;
937 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
938 struct nfs_client *clp = server->nfs_client;
940 clp->cl_mvops->free_lock_state(server, lsp);
942 nfs4_free_lock_state(server, lsp);
945 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
947 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
949 dst->fl_u.nfs4_fl.owner = lsp;
950 atomic_inc(&lsp->ls_count);
953 static void nfs4_fl_release_lock(struct file_lock *fl)
955 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
958 static const struct file_lock_operations nfs4_fl_lock_ops = {
959 .fl_copy_lock = nfs4_fl_copy_lock,
960 .fl_release_private = nfs4_fl_release_lock,
963 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
965 struct nfs4_lock_state *lsp;
967 if (fl->fl_ops != NULL)
969 if (fl->fl_flags & FL_POSIX)
970 lsp = nfs4_get_lock_state(state, fl->fl_owner, 0, NFS4_POSIX_LOCK_TYPE);
971 else if (fl->fl_flags & FL_FLOCK)
972 lsp = nfs4_get_lock_state(state, NULL, fl->fl_pid,
973 NFS4_FLOCK_LOCK_TYPE);
978 fl->fl_u.nfs4_fl.owner = lsp;
979 fl->fl_ops = &nfs4_fl_lock_ops;
983 static int nfs4_copy_lock_stateid(nfs4_stateid *dst,
984 struct nfs4_state *state,
985 const struct nfs_lockowner *lockowner)
987 struct nfs4_lock_state *lsp;
993 if (lockowner == NULL)
996 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
999 fl_owner = lockowner->l_owner;
1000 fl_pid = lockowner->l_pid;
1001 spin_lock(&state->state_lock);
1002 lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
1003 if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
1004 nfs4_stateid_copy(dst, &lsp->ls_stateid);
1007 if (!list_empty(&lsp->ls_seqid.list))
1010 spin_unlock(&state->state_lock);
1011 nfs4_put_lock_state(lsp);
1016 static int nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
1018 const nfs4_stateid *src;
1023 src = &zero_stateid;
1024 seq = read_seqbegin(&state->seqlock);
1025 if (test_bit(NFS_OPEN_STATE, &state->flags))
1026 src = &state->open_stateid;
1027 nfs4_stateid_copy(dst, src);
1030 if (!list_empty(&state->owner->so_seqid.list))
1032 } while (read_seqretry(&state->seqlock, seq));
1037 * Byte-range lock aware utility to initialize the stateid of read/write
1040 int nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state,
1041 fmode_t fmode, const struct nfs_lockowner *lockowner)
1044 if (nfs4_copy_delegation_stateid(dst, state->inode, fmode))
1046 ret = nfs4_copy_lock_stateid(dst, state, lockowner);
1049 ret = nfs4_copy_open_stateid(dst, state);
1051 if (nfs_server_capable(state->inode, NFS_CAP_STATEID_NFSV41))
1056 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
1058 struct nfs_seqid *new;
1060 new = kmalloc(sizeof(*new), gfp_mask);
1062 new->sequence = counter;
1063 INIT_LIST_HEAD(&new->list);
1069 void nfs_release_seqid(struct nfs_seqid *seqid)
1071 struct nfs_seqid_counter *sequence;
1073 if (list_empty(&seqid->list))
1075 sequence = seqid->sequence;
1076 spin_lock(&sequence->lock);
1077 list_del_init(&seqid->list);
1078 if (!list_empty(&sequence->list)) {
1079 struct nfs_seqid *next;
1081 next = list_first_entry(&sequence->list,
1082 struct nfs_seqid, list);
1083 rpc_wake_up_queued_task(&sequence->wait, next->task);
1085 spin_unlock(&sequence->lock);
1088 void nfs_free_seqid(struct nfs_seqid *seqid)
1090 nfs_release_seqid(seqid);
1095 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
1096 * failed with a seqid incrementing error -
1097 * see comments nfs_fs.h:seqid_mutating_error()
1099 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
1104 case -NFS4ERR_BAD_SEQID:
1105 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
1107 pr_warn_ratelimited("NFS: v4 server returned a bad"
1108 " sequence-id error on an"
1109 " unconfirmed sequence %p!\n",
1111 case -NFS4ERR_STALE_CLIENTID:
1112 case -NFS4ERR_STALE_STATEID:
1113 case -NFS4ERR_BAD_STATEID:
1114 case -NFS4ERR_BADXDR:
1115 case -NFS4ERR_RESOURCE:
1116 case -NFS4ERR_NOFILEHANDLE:
1117 /* Non-seqid mutating errors */
1121 * Note: no locking needed as we are guaranteed to be first
1122 * on the sequence list
1124 seqid->sequence->counter++;
1127 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
1129 struct nfs4_state_owner *sp = container_of(seqid->sequence,
1130 struct nfs4_state_owner, so_seqid);
1131 struct nfs_server *server = sp->so_server;
1133 if (status == -NFS4ERR_BAD_SEQID)
1134 nfs4_drop_state_owner(sp);
1135 if (!nfs4_has_session(server->nfs_client))
1136 nfs_increment_seqid(status, seqid);
1140 * Increment the seqid if the LOCK/LOCKU succeeded, or
1141 * failed with a seqid incrementing error -
1142 * see comments nfs_fs.h:seqid_mutating_error()
1144 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
1146 nfs_increment_seqid(status, seqid);
1149 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
1151 struct nfs_seqid_counter *sequence = seqid->sequence;
1154 spin_lock(&sequence->lock);
1156 if (list_empty(&seqid->list))
1157 list_add_tail(&seqid->list, &sequence->list);
1158 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
1160 rpc_sleep_on(&sequence->wait, task, NULL);
1163 spin_unlock(&sequence->lock);
1167 static int nfs4_run_state_manager(void *);
1169 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
1171 smp_mb__before_clear_bit();
1172 clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
1173 smp_mb__after_clear_bit();
1174 wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
1175 rpc_wake_up(&clp->cl_rpcwaitq);
1179 * Schedule the nfs_client asynchronous state management routine
1181 void nfs4_schedule_state_manager(struct nfs_client *clp)
1183 struct task_struct *task;
1184 char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
1186 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1188 __module_get(THIS_MODULE);
1189 atomic_inc(&clp->cl_count);
1191 /* The rcu_read_lock() is not strictly necessary, as the state
1192 * manager is the only thread that ever changes the rpc_xprt
1193 * after it's initialized. At this point, we're single threaded. */
1195 snprintf(buf, sizeof(buf), "%s-manager",
1196 rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR));
1198 task = kthread_run(nfs4_run_state_manager, clp, buf);
1200 printk(KERN_ERR "%s: kthread_run: %ld\n",
1201 __func__, PTR_ERR(task));
1202 nfs4_clear_state_manager_bit(clp);
1203 nfs_put_client(clp);
1204 module_put(THIS_MODULE);
1209 * Schedule a lease recovery attempt
1211 void nfs4_schedule_lease_recovery(struct nfs_client *clp)
1215 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1216 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1217 dprintk("%s: scheduling lease recovery for server %s\n", __func__,
1219 nfs4_schedule_state_manager(clp);
1221 EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
1223 int nfs4_wait_clnt_recover(struct nfs_client *clp)
1229 res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
1230 nfs_wait_bit_killable, TASK_KILLABLE);
1234 if (clp->cl_cons_state < 0)
1235 return clp->cl_cons_state;
1239 int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1244 for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1245 ret = nfs4_wait_clnt_recover(clp);
1248 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1249 !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1251 nfs4_schedule_state_manager(clp);
1258 * nfs40_handle_cb_pathdown - return all delegations after NFS4ERR_CB_PATH_DOWN
1259 * @clp: client to process
1261 * Set the NFS4CLNT_LEASE_EXPIRED state in order to force a
1262 * resend of the SETCLIENTID and hence re-establish the
1263 * callback channel. Then return all existing delegations.
1265 static void nfs40_handle_cb_pathdown(struct nfs_client *clp)
1267 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1268 nfs_expire_all_delegations(clp);
1269 dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
1273 void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
1275 nfs40_handle_cb_pathdown(clp);
1276 nfs4_schedule_state_manager(clp);
1279 static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1282 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1283 /* Don't recover state that expired before the reboot */
1284 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
1285 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1288 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1289 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1293 static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1295 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
1296 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1297 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1298 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
1302 int nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
1304 struct nfs_client *clp = server->nfs_client;
1306 if (!nfs4_valid_open_stateid(state))
1308 nfs4_state_mark_reclaim_nograce(clp, state);
1309 dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
1311 nfs4_schedule_state_manager(clp);
1314 EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1316 void nfs_inode_find_state_and_recover(struct inode *inode,
1317 const nfs4_stateid *stateid)
1319 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1320 struct nfs_inode *nfsi = NFS_I(inode);
1321 struct nfs_open_context *ctx;
1322 struct nfs4_state *state;
1325 spin_lock(&inode->i_lock);
1326 list_for_each_entry(ctx, &nfsi->open_files, list) {
1330 if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
1332 if (!nfs4_stateid_match(&state->stateid, stateid))
1334 nfs4_state_mark_reclaim_nograce(clp, state);
1337 spin_unlock(&inode->i_lock);
1339 nfs4_schedule_state_manager(clp);
1342 static void nfs4_state_mark_open_context_bad(struct nfs4_state *state)
1344 struct inode *inode = state->inode;
1345 struct nfs_inode *nfsi = NFS_I(inode);
1346 struct nfs_open_context *ctx;
1348 spin_lock(&inode->i_lock);
1349 list_for_each_entry(ctx, &nfsi->open_files, list) {
1350 if (ctx->state != state)
1352 set_bit(NFS_CONTEXT_BAD, &ctx->flags);
1354 spin_unlock(&inode->i_lock);
1357 static void nfs4_state_mark_recovery_failed(struct nfs4_state *state, int error)
1359 set_bit(NFS_STATE_RECOVERY_FAILED, &state->flags);
1360 nfs4_state_mark_open_context_bad(state);
1364 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
1366 struct inode *inode = state->inode;
1367 struct nfs_inode *nfsi = NFS_I(inode);
1368 struct file_lock *fl;
1371 if (inode->i_flock == NULL)
1374 /* Guard against delegation returns and new lock/unlock calls */
1375 down_write(&nfsi->rwsem);
1376 /* Protect inode->i_flock using the BKL */
1378 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1379 if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
1381 if (nfs_file_open_context(fl->fl_file)->state != state)
1384 status = ops->recover_lock(state, fl);
1389 case -NFS4ERR_ADMIN_REVOKED:
1390 case -NFS4ERR_STALE_STATEID:
1391 case -NFS4ERR_BAD_STATEID:
1392 case -NFS4ERR_EXPIRED:
1393 case -NFS4ERR_NO_GRACE:
1394 case -NFS4ERR_STALE_CLIENTID:
1395 case -NFS4ERR_BADSESSION:
1396 case -NFS4ERR_BADSLOT:
1397 case -NFS4ERR_BAD_HIGH_SLOT:
1398 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1401 printk(KERN_ERR "NFS: %s: unhandled error %d. "
1402 "Zeroing state\n", __func__, status);
1404 case -NFS4ERR_DENIED:
1405 case -NFS4ERR_RECLAIM_BAD:
1406 case -NFS4ERR_RECLAIM_CONFLICT:
1407 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1414 up_write(&nfsi->rwsem);
1418 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
1420 struct nfs4_state *state;
1421 struct nfs4_lock_state *lock;
1424 /* Note: we rely on the sp->so_states list being ordered
1425 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
1427 * This is needed to ensure that the server won't give us any
1428 * read delegations that we have to return if, say, we are
1429 * recovering after a network partition or a reboot from a
1430 * server that doesn't support a grace period.
1432 spin_lock(&sp->so_lock);
1433 write_seqcount_begin(&sp->so_reclaim_seqcount);
1435 list_for_each_entry(state, &sp->so_states, open_states) {
1436 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
1438 if (!nfs4_valid_open_stateid(state))
1440 if (state->state == 0)
1442 atomic_inc(&state->count);
1443 spin_unlock(&sp->so_lock);
1444 status = ops->recover_open(sp, state);
1446 status = nfs4_reclaim_locks(state, ops);
1448 spin_lock(&state->state_lock);
1449 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1450 if (!test_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags))
1451 pr_warn_ratelimited("NFS: "
1453 "failed!\n", __func__);
1455 spin_unlock(&state->state_lock);
1456 nfs4_put_open_state(state);
1457 spin_lock(&sp->so_lock);
1463 printk(KERN_ERR "NFS: %s: unhandled error %d. "
1464 "Zeroing state\n", __func__, status);
1469 * Open state on this file cannot be recovered
1470 * All we can do is revert to using the zero stateid.
1472 nfs4_state_mark_recovery_failed(state, status);
1476 case -NFS4ERR_ADMIN_REVOKED:
1477 case -NFS4ERR_STALE_STATEID:
1478 case -NFS4ERR_BAD_STATEID:
1479 case -NFS4ERR_RECLAIM_BAD:
1480 case -NFS4ERR_RECLAIM_CONFLICT:
1481 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1483 case -NFS4ERR_EXPIRED:
1484 case -NFS4ERR_NO_GRACE:
1485 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1486 case -NFS4ERR_STALE_CLIENTID:
1487 case -NFS4ERR_BADSESSION:
1488 case -NFS4ERR_BADSLOT:
1489 case -NFS4ERR_BAD_HIGH_SLOT:
1490 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1493 nfs4_put_open_state(state);
1494 spin_lock(&sp->so_lock);
1497 write_seqcount_end(&sp->so_reclaim_seqcount);
1498 spin_unlock(&sp->so_lock);
1501 nfs4_put_open_state(state);
1502 spin_lock(&sp->so_lock);
1503 write_seqcount_end(&sp->so_reclaim_seqcount);
1504 spin_unlock(&sp->so_lock);
1508 static void nfs4_clear_open_state(struct nfs4_state *state)
1510 struct nfs4_lock_state *lock;
1512 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1513 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1514 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1515 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1516 spin_lock(&state->state_lock);
1517 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1518 lock->ls_seqid.flags = 0;
1519 clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
1521 spin_unlock(&state->state_lock);
1524 static void nfs4_reset_seqids(struct nfs_server *server,
1525 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1527 struct nfs_client *clp = server->nfs_client;
1528 struct nfs4_state_owner *sp;
1529 struct rb_node *pos;
1530 struct nfs4_state *state;
1532 spin_lock(&clp->cl_lock);
1533 for (pos = rb_first(&server->state_owners);
1535 pos = rb_next(pos)) {
1536 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1537 sp->so_seqid.flags = 0;
1538 spin_lock(&sp->so_lock);
1539 list_for_each_entry(state, &sp->so_states, open_states) {
1540 if (mark_reclaim(clp, state))
1541 nfs4_clear_open_state(state);
1543 spin_unlock(&sp->so_lock);
1545 spin_unlock(&clp->cl_lock);
1548 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp,
1549 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1551 struct nfs_server *server;
1554 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1555 nfs4_reset_seqids(server, mark_reclaim);
1559 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1561 /* Mark all delegations for reclaim */
1562 nfs_delegation_mark_reclaim(clp);
1563 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1566 static void nfs4_reclaim_complete(struct nfs_client *clp,
1567 const struct nfs4_state_recovery_ops *ops)
1569 /* Notify the server we're done reclaiming our state */
1570 if (ops->reclaim_complete)
1571 (void)ops->reclaim_complete(clp);
1574 static void nfs4_clear_reclaim_server(struct nfs_server *server)
1576 struct nfs_client *clp = server->nfs_client;
1577 struct nfs4_state_owner *sp;
1578 struct rb_node *pos;
1579 struct nfs4_state *state;
1581 spin_lock(&clp->cl_lock);
1582 for (pos = rb_first(&server->state_owners);
1584 pos = rb_next(pos)) {
1585 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1586 spin_lock(&sp->so_lock);
1587 list_for_each_entry(state, &sp->so_states, open_states) {
1588 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
1591 nfs4_state_mark_reclaim_nograce(clp, state);
1593 spin_unlock(&sp->so_lock);
1595 spin_unlock(&clp->cl_lock);
1598 static int nfs4_state_clear_reclaim_reboot(struct nfs_client *clp)
1600 struct nfs_server *server;
1602 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1606 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1607 nfs4_clear_reclaim_server(server);
1610 nfs_delegation_reap_unclaimed(clp);
1614 static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
1616 if (!nfs4_state_clear_reclaim_reboot(clp))
1618 nfs4_reclaim_complete(clp, clp->cl_mvops->reboot_recovery_ops);
1621 static void nfs_delegation_clear_all(struct nfs_client *clp)
1623 nfs_delegation_mark_reclaim(clp);
1624 nfs_delegation_reap_unclaimed(clp);
1627 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1629 nfs_delegation_clear_all(clp);
1630 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1633 static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1638 case -NFS4ERR_CB_PATH_DOWN:
1639 nfs40_handle_cb_pathdown(clp);
1641 case -NFS4ERR_NO_GRACE:
1642 nfs4_state_end_reclaim_reboot(clp);
1644 case -NFS4ERR_STALE_CLIENTID:
1645 case -NFS4ERR_LEASE_MOVED:
1646 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1647 nfs4_state_clear_reclaim_reboot(clp);
1648 nfs4_state_start_reclaim_reboot(clp);
1650 case -NFS4ERR_EXPIRED:
1651 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1652 nfs4_state_start_reclaim_nograce(clp);
1654 case -NFS4ERR_BADSESSION:
1655 case -NFS4ERR_BADSLOT:
1656 case -NFS4ERR_BAD_HIGH_SLOT:
1657 case -NFS4ERR_DEADSESSION:
1658 case -NFS4ERR_SEQ_FALSE_RETRY:
1659 case -NFS4ERR_SEQ_MISORDERED:
1660 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1661 /* Zero session reset errors */
1663 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1664 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1667 dprintk("%s: failed to handle error %d for server %s\n",
1668 __func__, error, clp->cl_hostname);
1671 dprintk("%s: handled error %d for server %s\n", __func__, error,
1676 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1678 struct nfs4_state_owner *sp;
1679 struct nfs_server *server;
1680 struct rb_node *pos;
1685 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1686 nfs4_purge_state_owners(server);
1687 spin_lock(&clp->cl_lock);
1688 for (pos = rb_first(&server->state_owners);
1690 pos = rb_next(pos)) {
1692 struct nfs4_state_owner, so_server_node);
1693 if (!test_and_clear_bit(ops->owner_flag_bit,
1696 atomic_inc(&sp->so_count);
1697 spin_unlock(&clp->cl_lock);
1700 status = nfs4_reclaim_open_state(sp, ops);
1702 set_bit(ops->owner_flag_bit, &sp->so_flags);
1703 nfs4_put_state_owner(sp);
1704 return nfs4_recovery_handle_error(clp, status);
1707 nfs4_put_state_owner(sp);
1710 spin_unlock(&clp->cl_lock);
1716 static int nfs4_check_lease(struct nfs_client *clp)
1718 struct rpc_cred *cred;
1719 const struct nfs4_state_maintenance_ops *ops =
1720 clp->cl_mvops->state_renewal_ops;
1723 /* Is the client already known to have an expired lease? */
1724 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1726 spin_lock(&clp->cl_lock);
1727 cred = ops->get_state_renewal_cred_locked(clp);
1728 spin_unlock(&clp->cl_lock);
1730 cred = nfs4_get_setclientid_cred(clp);
1735 status = ops->renew_lease(clp, cred);
1737 if (status == -ETIMEDOUT) {
1738 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1742 return nfs4_recovery_handle_error(clp, status);
1745 /* Set NFS4CLNT_LEASE_EXPIRED and reclaim reboot state for all v4.0 errors
1746 * and for recoverable errors on EXCHANGE_ID for v4.1
1748 static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
1751 case -NFS4ERR_SEQ_MISORDERED:
1752 if (test_and_set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state))
1753 return -ESERVERFAULT;
1754 /* Lease confirmation error: retry after purging the lease */
1756 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1758 case -NFS4ERR_STALE_CLIENTID:
1759 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1760 nfs4_state_clear_reclaim_reboot(clp);
1761 nfs4_state_start_reclaim_reboot(clp);
1763 case -NFS4ERR_CLID_INUSE:
1764 pr_err("NFS: Server %s reports our clientid is in use\n",
1766 nfs_mark_client_ready(clp, -EPERM);
1767 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1770 case -NFS4ERR_DELAY:
1776 case -NFS4ERR_MINOR_VERS_MISMATCH:
1777 if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
1778 nfs_mark_client_ready(clp, -EPROTONOSUPPORT);
1779 dprintk("%s: exit with error %d for server %s\n",
1780 __func__, -EPROTONOSUPPORT, clp->cl_hostname);
1781 return -EPROTONOSUPPORT;
1782 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1783 * in nfs4_exchange_id */
1785 dprintk("%s: exit with error %d for server %s\n", __func__,
1786 status, clp->cl_hostname);
1789 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1790 dprintk("%s: handled error %d for server %s\n", __func__, status,
1795 static int nfs4_establish_lease(struct nfs_client *clp)
1797 struct rpc_cred *cred;
1798 const struct nfs4_state_recovery_ops *ops =
1799 clp->cl_mvops->reboot_recovery_ops;
1802 cred = ops->get_clid_cred(clp);
1805 status = ops->establish_clid(clp, cred);
1809 pnfs_destroy_all_layouts(clp);
1814 * Returns zero or a negative errno. NFS4ERR values are converted
1815 * to local errno values.
1817 static int nfs4_reclaim_lease(struct nfs_client *clp)
1821 status = nfs4_establish_lease(clp);
1823 return nfs4_handle_reclaim_lease_error(clp, status);
1824 if (test_and_clear_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state))
1825 nfs4_state_start_reclaim_nograce(clp);
1826 if (!test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
1827 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1828 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1829 clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1833 static int nfs4_purge_lease(struct nfs_client *clp)
1837 status = nfs4_establish_lease(clp);
1839 return nfs4_handle_reclaim_lease_error(clp, status);
1840 clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1841 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1842 nfs4_state_start_reclaim_nograce(clp);
1847 * nfs4_discover_server_trunking - Detect server IP address trunking
1849 * @clp: nfs_client under test
1850 * @result: OUT: found nfs_client, or clp
1852 * Returns zero or a negative errno. If zero is returned,
1853 * an nfs_client pointer is planted in "result".
1855 * Note: since we are invoked in process context, and
1856 * not from inside the state manager, we cannot use
1857 * nfs4_handle_reclaim_lease_error().
1859 int nfs4_discover_server_trunking(struct nfs_client *clp,
1860 struct nfs_client **result)
1862 const struct nfs4_state_recovery_ops *ops =
1863 clp->cl_mvops->reboot_recovery_ops;
1864 struct rpc_clnt *clnt;
1865 struct rpc_cred *cred;
1868 dprintk("NFS: %s: testing '%s'\n", __func__, clp->cl_hostname);
1870 clnt = clp->cl_rpcclient;
1873 mutex_lock(&nfs_clid_init_mutex);
1876 cred = ops->get_clid_cred(clp);
1880 status = ops->detect_trunking(clp, result, cred);
1885 case -NFS4ERR_DELAY:
1889 case -NFS4ERR_STALE_CLIENTID:
1890 dprintk("NFS: %s after status %d, retrying\n",
1896 case -NFS4ERR_CLID_INUSE:
1897 case -NFS4ERR_WRONGSEC:
1898 clnt = rpc_clone_client_set_auth(clnt, RPC_AUTH_UNIX);
1900 status = PTR_ERR(clnt);
1903 /* Note: this is safe because we haven't yet marked the
1904 * client as ready, so we are the only user of
1907 clnt = xchg(&clp->cl_rpcclient, clnt);
1908 rpc_shutdown_client(clnt);
1909 clnt = clp->cl_rpcclient;
1912 case -NFS4ERR_MINOR_VERS_MISMATCH:
1913 status = -EPROTONOSUPPORT;
1917 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1918 * in nfs4_exchange_id */
1919 status = -EKEYEXPIRED;
1922 pr_warn("NFS: %s unhandled error %d. Exiting with error EIO\n",
1928 mutex_unlock(&nfs_clid_init_mutex);
1929 dprintk("NFS: %s: status = %d\n", __func__, status);
1933 #ifdef CONFIG_NFS_V4_1
1934 void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
1936 struct nfs_client *clp = session->clp;
1940 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1942 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1943 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1945 nfs4_schedule_lease_recovery(clp);
1947 EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
1949 static void nfs41_ping_server(struct nfs_client *clp)
1951 /* Use CHECK_LEASE to ping the server with a SEQUENCE */
1952 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1953 nfs4_schedule_state_manager(clp);
1956 void nfs41_server_notify_target_slotid_update(struct nfs_client *clp)
1958 nfs41_ping_server(clp);
1961 void nfs41_server_notify_highest_slotid_update(struct nfs_client *clp)
1963 nfs41_ping_server(clp);
1966 static void nfs4_reset_all_state(struct nfs_client *clp)
1968 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1969 set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1970 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1971 nfs4_state_start_reclaim_nograce(clp);
1972 dprintk("%s: scheduling reset of all state for server %s!\n",
1973 __func__, clp->cl_hostname);
1974 nfs4_schedule_state_manager(clp);
1978 static void nfs41_handle_server_reboot(struct nfs_client *clp)
1980 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1981 nfs4_state_start_reclaim_reboot(clp);
1982 dprintk("%s: server %s rebooted!\n", __func__,
1984 nfs4_schedule_state_manager(clp);
1988 static void nfs41_handle_state_revoked(struct nfs_client *clp)
1990 nfs4_reset_all_state(clp);
1991 dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
1994 static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
1996 /* This will need to handle layouts too */
1997 nfs_expire_all_delegations(clp);
1998 dprintk("%s: Recallable state revoked on server %s!\n", __func__,
2002 static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
2004 nfs_expire_all_delegations(clp);
2005 if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
2006 nfs4_schedule_state_manager(clp);
2007 dprintk("%s: server %s declared a backchannel fault\n", __func__,
2011 static void nfs41_handle_cb_path_down(struct nfs_client *clp)
2013 if (test_and_set_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2014 &clp->cl_state) == 0)
2015 nfs4_schedule_state_manager(clp);
2018 void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
2023 dprintk("%s: \"%s\" (client ID %llx) flags=0x%08x\n",
2024 __func__, clp->cl_hostname, clp->cl_clientid, flags);
2026 if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
2027 nfs41_handle_server_reboot(clp);
2028 if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
2029 SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
2030 SEQ4_STATUS_ADMIN_STATE_REVOKED |
2031 SEQ4_STATUS_LEASE_MOVED))
2032 nfs41_handle_state_revoked(clp);
2033 if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
2034 nfs41_handle_recallable_state_revoked(clp);
2035 if (flags & SEQ4_STATUS_BACKCHANNEL_FAULT)
2036 nfs41_handle_backchannel_fault(clp);
2037 else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
2038 SEQ4_STATUS_CB_PATH_DOWN_SESSION))
2039 nfs41_handle_cb_path_down(clp);
2042 static int nfs4_reset_session(struct nfs_client *clp)
2044 struct rpc_cred *cred;
2047 if (!nfs4_has_session(clp))
2049 nfs4_begin_drain_session(clp);
2050 cred = nfs4_get_exchange_id_cred(clp);
2051 status = nfs4_proc_destroy_session(clp->cl_session, cred);
2054 case -NFS4ERR_BADSESSION:
2055 case -NFS4ERR_DEADSESSION:
2057 case -NFS4ERR_BACK_CHAN_BUSY:
2058 case -NFS4ERR_DELAY:
2059 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
2064 status = nfs4_recovery_handle_error(clp, status);
2068 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
2069 status = nfs4_proc_create_session(clp, cred);
2071 dprintk("%s: session reset failed with status %d for server %s!\n",
2072 __func__, status, clp->cl_hostname);
2073 status = nfs4_handle_reclaim_lease_error(clp, status);
2076 nfs41_finish_session_reset(clp);
2077 dprintk("%s: session reset was successful for server %s!\n",
2078 __func__, clp->cl_hostname);
2085 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2087 struct rpc_cred *cred;
2090 if (!nfs4_has_session(clp))
2092 nfs4_begin_drain_session(clp);
2093 cred = nfs4_get_exchange_id_cred(clp);
2094 ret = nfs4_proc_bind_conn_to_session(clp, cred);
2097 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2100 dprintk("%s: bind_conn_to_session was successful for server %s!\n",
2101 __func__, clp->cl_hostname);
2103 case -NFS4ERR_DELAY:
2105 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2108 return nfs4_recovery_handle_error(clp, ret);
2112 #else /* CONFIG_NFS_V4_1 */
2113 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
2114 static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
2116 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2120 #endif /* CONFIG_NFS_V4_1 */
2122 static void nfs4_state_manager(struct nfs_client *clp)
2125 const char *section = "", *section_sep = "";
2127 /* Ensure exclusive access to NFSv4 state */
2129 if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
2130 section = "purge state";
2131 status = nfs4_purge_lease(clp);
2137 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
2138 section = "lease expired";
2139 /* We're going to have to re-establish a clientid */
2140 status = nfs4_reclaim_lease(clp);
2146 /* Initialize or reset the session */
2147 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
2148 section = "reset session";
2149 status = nfs4_reset_session(clp);
2150 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
2156 /* Send BIND_CONN_TO_SESSION */
2157 if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2159 section = "bind conn to session";
2160 status = nfs4_bind_conn_to_session(clp);
2166 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
2167 section = "check lease";
2168 status = nfs4_check_lease(clp);
2174 /* First recover reboot state... */
2175 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
2176 section = "reclaim reboot";
2177 status = nfs4_do_reclaim(clp,
2178 clp->cl_mvops->reboot_recovery_ops);
2179 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2180 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
2182 nfs4_state_end_reclaim_reboot(clp);
2183 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
2189 /* Now recover expired state... */
2190 if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
2191 section = "reclaim nograce";
2192 status = nfs4_do_reclaim(clp,
2193 clp->cl_mvops->nograce_recovery_ops);
2194 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2195 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
2196 test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
2202 nfs4_end_drain_session(clp);
2203 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
2204 nfs_client_return_marked_delegations(clp);
2208 nfs4_clear_state_manager_bit(clp);
2209 /* Did we race with an attempt to give us more work? */
2210 if (clp->cl_state == 0)
2212 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
2214 } while (atomic_read(&clp->cl_count) > 1);
2217 if (strlen(section))
2219 pr_warn_ratelimited("NFS: state manager%s%s failed on NFSv4 server %s"
2220 " with error %d\n", section_sep, section,
2221 clp->cl_hostname, -status);
2223 nfs4_end_drain_session(clp);
2224 nfs4_clear_state_manager_bit(clp);
2227 static int nfs4_run_state_manager(void *ptr)
2229 struct nfs_client *clp = ptr;
2231 allow_signal(SIGKILL);
2232 nfs4_state_manager(clp);
2233 nfs_put_client(clp);
2234 module_put_and_exit(0);