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/kthread.h>
46 #include <linux/module.h>
47 #include <linux/random.h>
48 #include <linux/ratelimit.h>
49 #include <linux/workqueue.h>
50 #include <linux/bitops.h>
51 #include <linux/jiffies.h>
52 #include <linux/sched/mm.h>
54 #include <linux/sunrpc/clnt.h>
58 #include "delegation.h"
60 #include "nfs4idmap.h"
61 #include "nfs4session.h"
65 #define NFSDBG_FACILITY NFSDBG_STATE
67 #define OPENOWNER_POOL_SIZE 8
69 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp);
71 const nfs4_stateid zero_stateid = {
73 .type = NFS4_SPECIAL_STATEID_TYPE,
75 const nfs4_stateid invalid_stateid = {
77 /* Funky initialiser keeps older gcc versions happy */
78 .data = { 0xff, 0xff, 0xff, 0xff, 0 },
80 .type = NFS4_INVALID_STATEID_TYPE,
83 const nfs4_stateid current_stateid = {
85 /* Funky initialiser keeps older gcc versions happy */
86 .data = { 0x0, 0x0, 0x0, 0x1, 0 },
88 .type = NFS4_SPECIAL_STATEID_TYPE,
91 static DEFINE_MUTEX(nfs_clid_init_mutex);
93 static int nfs4_setup_state_renewal(struct nfs_client *clp)
96 struct nfs_fsinfo fsinfo;
98 if (!test_bit(NFS_CS_CHECK_LEASE_TIME, &clp->cl_res_state)) {
99 nfs4_schedule_state_renewal(clp);
103 status = nfs4_proc_get_lease_time(clp, &fsinfo);
105 nfs4_set_lease_period(clp, fsinfo.lease_time * HZ);
106 nfs4_schedule_state_renewal(clp);
112 int nfs4_init_clientid(struct nfs_client *clp, const struct cred *cred)
114 struct nfs4_setclientid_res clid = {
115 .clientid = clp->cl_clientid,
116 .confirm = clp->cl_confirm,
120 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
122 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
124 port = nn->nfs_callback_tcpport;
125 if (clp->cl_addr.ss_family == AF_INET6)
126 port = nn->nfs_callback_tcpport6;
128 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
131 clp->cl_clientid = clid.clientid;
132 clp->cl_confirm = clid.confirm;
133 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
135 status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
138 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
139 nfs4_setup_state_renewal(clp);
145 * nfs40_discover_server_trunking - Detect server IP address trunking (mv0)
147 * @clp: nfs_client under test
148 * @result: OUT: found nfs_client, or clp
149 * @cred: credential to use for trunking test
151 * Returns zero, a negative errno, or a negative NFS4ERR status.
152 * If zero is returned, an nfs_client pointer is planted in
155 * Note: The returned client may not yet be marked ready.
157 int nfs40_discover_server_trunking(struct nfs_client *clp,
158 struct nfs_client **result,
159 const struct cred *cred)
161 struct nfs4_setclientid_res clid = {
162 .clientid = clp->cl_clientid,
163 .confirm = clp->cl_confirm,
165 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
169 port = nn->nfs_callback_tcpport;
170 if (clp->cl_addr.ss_family == AF_INET6)
171 port = nn->nfs_callback_tcpport6;
173 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
176 clp->cl_clientid = clid.clientid;
177 clp->cl_confirm = clid.confirm;
179 status = nfs40_walk_client_list(clp, result, cred);
181 /* Sustain the lease, even if it's empty. If the clientid4
182 * goes stale it's of no use for trunking discovery. */
183 nfs4_schedule_state_renewal(*result);
185 /* If the client state need to recover, do it. */
187 nfs4_schedule_state_manager(clp);
193 const struct cred *nfs4_get_machine_cred(struct nfs_client *clp)
195 return get_cred(rpc_machine_cred());
198 static void nfs4_root_machine_cred(struct nfs_client *clp)
201 /* Force root creds instead of machine */
202 clp->cl_principal = NULL;
203 clp->cl_rpcclient->cl_principal = NULL;
206 static const struct cred *
207 nfs4_get_renew_cred_server_locked(struct nfs_server *server)
209 const struct cred *cred = NULL;
210 struct nfs4_state_owner *sp;
213 for (pos = rb_first(&server->state_owners);
215 pos = rb_next(pos)) {
216 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
217 if (list_empty(&sp->so_states))
219 cred = get_cred(sp->so_cred);
226 * nfs4_get_renew_cred - Acquire credential for a renew operation
227 * @clp: client state handle
229 * Returns an rpc_cred with reference count bumped, or NULL.
230 * Caller must hold clp->cl_lock.
232 const struct cred *nfs4_get_renew_cred(struct nfs_client *clp)
234 const struct cred *cred = NULL;
235 struct nfs_server *server;
237 /* Use machine credentials if available */
238 cred = nfs4_get_machine_cred(clp);
242 spin_lock(&clp->cl_lock);
244 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
245 cred = nfs4_get_renew_cred_server_locked(server);
250 spin_unlock(&clp->cl_lock);
256 static void nfs4_end_drain_slot_table(struct nfs4_slot_table *tbl)
258 if (test_and_clear_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state)) {
259 spin_lock(&tbl->slot_tbl_lock);
260 nfs41_wake_slot_table(tbl);
261 spin_unlock(&tbl->slot_tbl_lock);
265 static void nfs4_end_drain_session(struct nfs_client *clp)
267 struct nfs4_session *ses = clp->cl_session;
269 if (clp->cl_slot_tbl) {
270 nfs4_end_drain_slot_table(clp->cl_slot_tbl);
275 nfs4_end_drain_slot_table(&ses->bc_slot_table);
276 nfs4_end_drain_slot_table(&ses->fc_slot_table);
280 static int nfs4_drain_slot_tbl(struct nfs4_slot_table *tbl)
282 set_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state);
283 spin_lock(&tbl->slot_tbl_lock);
284 if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
285 reinit_completion(&tbl->complete);
286 spin_unlock(&tbl->slot_tbl_lock);
287 return wait_for_completion_interruptible(&tbl->complete);
289 spin_unlock(&tbl->slot_tbl_lock);
293 static int nfs4_begin_drain_session(struct nfs_client *clp)
295 struct nfs4_session *ses = clp->cl_session;
298 if (clp->cl_slot_tbl)
299 return nfs4_drain_slot_tbl(clp->cl_slot_tbl);
302 ret = nfs4_drain_slot_tbl(&ses->bc_slot_table);
306 return nfs4_drain_slot_tbl(&ses->fc_slot_table);
309 #if defined(CONFIG_NFS_V4_1)
311 static void nfs41_finish_session_reset(struct nfs_client *clp)
313 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
314 clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
315 /* create_session negotiated new slot table */
316 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
317 nfs4_setup_state_renewal(clp);
320 int nfs41_init_clientid(struct nfs_client *clp, const struct cred *cred)
324 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
326 status = nfs4_proc_exchange_id(clp, cred);
329 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
331 status = nfs4_proc_create_session(clp, cred);
334 if (!(clp->cl_exchange_flags & EXCHGID4_FLAG_CONFIRMED_R))
335 nfs4_state_start_reclaim_reboot(clp);
336 nfs41_finish_session_reset(clp);
337 nfs_mark_client_ready(clp, NFS_CS_READY);
343 * nfs41_discover_server_trunking - Detect server IP address trunking (mv1)
345 * @clp: nfs_client under test
346 * @result: OUT: found nfs_client, or clp
347 * @cred: credential to use for trunking test
349 * Returns NFS4_OK, a negative errno, or a negative NFS4ERR status.
350 * If NFS4_OK is returned, an nfs_client pointer is planted in
353 * Note: The returned client may not yet be marked ready.
355 int nfs41_discover_server_trunking(struct nfs_client *clp,
356 struct nfs_client **result,
357 const struct cred *cred)
361 status = nfs4_proc_exchange_id(clp, cred);
362 if (status != NFS4_OK)
365 status = nfs41_walk_client_list(clp, result, cred);
372 * Purge state if the client id was established in a prior
373 * instance and the client id could not have arrived on the
374 * server via Transparent State Migration.
376 if (clp->cl_exchange_flags & EXCHGID4_FLAG_CONFIRMED_R) {
377 if (!test_bit(NFS_CS_TSM_POSSIBLE, &clp->cl_flags))
378 set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
380 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
382 nfs4_schedule_state_manager(clp);
383 status = nfs_wait_client_init_complete(clp);
389 #endif /* CONFIG_NFS_V4_1 */
392 * nfs4_get_clid_cred - Acquire credential for a setclientid operation
393 * @clp: client state handle
395 * Returns a cred with reference count bumped, or NULL.
397 const struct cred *nfs4_get_clid_cred(struct nfs_client *clp)
399 const struct cred *cred;
401 cred = nfs4_get_machine_cred(clp);
405 static struct nfs4_state_owner *
406 nfs4_find_state_owner_locked(struct nfs_server *server, const struct cred *cred)
408 struct rb_node **p = &server->state_owners.rb_node,
410 struct nfs4_state_owner *sp;
415 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
416 cmp = cred_fscmp(cred, sp->so_cred);
419 p = &parent->rb_left;
421 p = &parent->rb_right;
423 if (!list_empty(&sp->so_lru))
424 list_del_init(&sp->so_lru);
425 atomic_inc(&sp->so_count);
432 static struct nfs4_state_owner *
433 nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
435 struct nfs_server *server = new->so_server;
436 struct rb_node **p = &server->state_owners.rb_node,
438 struct nfs4_state_owner *sp;
443 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
444 cmp = cred_fscmp(new->so_cred, sp->so_cred);
447 p = &parent->rb_left;
449 p = &parent->rb_right;
451 if (!list_empty(&sp->so_lru))
452 list_del_init(&sp->so_lru);
453 atomic_inc(&sp->so_count);
457 rb_link_node(&new->so_server_node, parent, p);
458 rb_insert_color(&new->so_server_node, &server->state_owners);
463 nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
465 struct nfs_server *server = sp->so_server;
467 if (!RB_EMPTY_NODE(&sp->so_server_node))
468 rb_erase(&sp->so_server_node, &server->state_owners);
472 nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
474 sc->create_time = ktime_get();
477 spin_lock_init(&sc->lock);
478 INIT_LIST_HEAD(&sc->list);
479 rpc_init_wait_queue(&sc->wait, "Seqid_waitqueue");
483 nfs4_destroy_seqid_counter(struct nfs_seqid_counter *sc)
485 rpc_destroy_wait_queue(&sc->wait);
489 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
490 * create a new state_owner.
493 static struct nfs4_state_owner *
494 nfs4_alloc_state_owner(struct nfs_server *server,
495 const struct cred *cred,
498 struct nfs4_state_owner *sp;
500 sp = kzalloc(sizeof(*sp), gfp_flags);
503 sp->so_seqid.owner_id = ida_simple_get(&server->openowner_id, 0, 0,
505 if (sp->so_seqid.owner_id < 0) {
509 sp->so_server = server;
510 sp->so_cred = get_cred(cred);
511 spin_lock_init(&sp->so_lock);
512 INIT_LIST_HEAD(&sp->so_states);
513 nfs4_init_seqid_counter(&sp->so_seqid);
514 atomic_set(&sp->so_count, 1);
515 INIT_LIST_HEAD(&sp->so_lru);
516 seqcount_init(&sp->so_reclaim_seqcount);
517 mutex_init(&sp->so_delegreturn_mutex);
522 nfs4_reset_state_owner(struct nfs4_state_owner *sp)
524 /* This state_owner is no longer usable, but must
525 * remain in place so that state recovery can find it
526 * and the opens associated with it.
527 * It may also be used for new 'open' request to
528 * return a delegation to the server.
529 * So update the 'create_time' so that it looks like
530 * a new state_owner. This will cause the server to
531 * request an OPEN_CONFIRM to start a new sequence.
533 sp->so_seqid.create_time = ktime_get();
536 static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
538 nfs4_destroy_seqid_counter(&sp->so_seqid);
539 put_cred(sp->so_cred);
540 ida_simple_remove(&sp->so_server->openowner_id, sp->so_seqid.owner_id);
544 static void nfs4_gc_state_owners(struct nfs_server *server)
546 struct nfs_client *clp = server->nfs_client;
547 struct nfs4_state_owner *sp, *tmp;
548 unsigned long time_min, time_max;
551 spin_lock(&clp->cl_lock);
553 time_min = (long)time_max - (long)clp->cl_lease_time;
554 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
555 /* NB: LRU is sorted so that oldest is at the head */
556 if (time_in_range(sp->so_expires, time_min, time_max))
558 list_move(&sp->so_lru, &doomed);
559 nfs4_remove_state_owner_locked(sp);
561 spin_unlock(&clp->cl_lock);
563 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
564 list_del(&sp->so_lru);
565 nfs4_free_state_owner(sp);
570 * nfs4_get_state_owner - Look up a state owner given a credential
571 * @server: nfs_server to search
572 * @cred: RPC credential to match
573 * @gfp_flags: allocation mode
575 * Returns a pointer to an instantiated nfs4_state_owner struct, or NULL.
577 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server,
578 const struct cred *cred,
581 struct nfs_client *clp = server->nfs_client;
582 struct nfs4_state_owner *sp, *new;
584 spin_lock(&clp->cl_lock);
585 sp = nfs4_find_state_owner_locked(server, cred);
586 spin_unlock(&clp->cl_lock);
589 new = nfs4_alloc_state_owner(server, cred, gfp_flags);
592 spin_lock(&clp->cl_lock);
593 sp = nfs4_insert_state_owner_locked(new);
594 spin_unlock(&clp->cl_lock);
596 nfs4_free_state_owner(new);
598 nfs4_gc_state_owners(server);
603 * nfs4_put_state_owner - Release a nfs4_state_owner
604 * @sp: state owner data to release
606 * Note that we keep released state owners on an LRU
608 * This caches valid state owners so that they can be
609 * reused, to avoid the OPEN_CONFIRM on minor version 0.
610 * It also pins the uniquifier of dropped state owners for
611 * a while, to ensure that those state owner names are
614 void nfs4_put_state_owner(struct nfs4_state_owner *sp)
616 struct nfs_server *server = sp->so_server;
617 struct nfs_client *clp = server->nfs_client;
619 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
622 sp->so_expires = jiffies;
623 list_add_tail(&sp->so_lru, &server->state_owners_lru);
624 spin_unlock(&clp->cl_lock);
628 * nfs4_purge_state_owners - Release all cached state owners
629 * @server: nfs_server with cached state owners to release
630 * @head: resulting list of state owners
632 * Called at umount time. Remaining state owners will be on
633 * the LRU with ref count of zero.
634 * Note that the state owners are not freed, but are added
635 * to the list @head, which can later be used as an argument
636 * to nfs4_free_state_owners.
638 void nfs4_purge_state_owners(struct nfs_server *server, struct list_head *head)
640 struct nfs_client *clp = server->nfs_client;
641 struct nfs4_state_owner *sp, *tmp;
643 spin_lock(&clp->cl_lock);
644 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
645 list_move(&sp->so_lru, head);
646 nfs4_remove_state_owner_locked(sp);
648 spin_unlock(&clp->cl_lock);
652 * nfs4_purge_state_owners - Release all cached state owners
653 * @head: resulting list of state owners
655 * Frees a list of state owners that was generated by
656 * nfs4_purge_state_owners
658 void nfs4_free_state_owners(struct list_head *head)
660 struct nfs4_state_owner *sp, *tmp;
662 list_for_each_entry_safe(sp, tmp, head, so_lru) {
663 list_del(&sp->so_lru);
664 nfs4_free_state_owner(sp);
668 static struct nfs4_state *
669 nfs4_alloc_open_state(void)
671 struct nfs4_state *state;
673 state = kzalloc(sizeof(*state), GFP_NOFS);
676 refcount_set(&state->count, 1);
677 INIT_LIST_HEAD(&state->lock_states);
678 spin_lock_init(&state->state_lock);
679 seqlock_init(&state->seqlock);
680 init_waitqueue_head(&state->waitq);
685 nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
687 if (state->state == fmode)
689 /* NB! List reordering - see the reclaim code for why. */
690 if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
691 if (fmode & FMODE_WRITE)
692 list_move(&state->open_states, &state->owner->so_states);
694 list_move_tail(&state->open_states, &state->owner->so_states);
696 state->state = fmode;
699 static struct nfs4_state *
700 __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
702 struct nfs_inode *nfsi = NFS_I(inode);
703 struct nfs4_state *state;
705 list_for_each_entry_rcu(state, &nfsi->open_states, inode_states) {
706 if (state->owner != owner)
708 if (!nfs4_valid_open_stateid(state))
710 if (refcount_inc_not_zero(&state->count))
717 nfs4_free_open_state(struct nfs4_state *state)
719 kfree_rcu(state, rcu_head);
723 nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
725 struct nfs4_state *state, *new;
726 struct nfs_inode *nfsi = NFS_I(inode);
729 state = __nfs4_find_state_byowner(inode, owner);
733 new = nfs4_alloc_open_state();
734 spin_lock(&owner->so_lock);
735 spin_lock(&inode->i_lock);
736 state = __nfs4_find_state_byowner(inode, owner);
737 if (state == NULL && new != NULL) {
739 state->owner = owner;
740 atomic_inc(&owner->so_count);
742 state->inode = inode;
743 list_add_rcu(&state->inode_states, &nfsi->open_states);
744 spin_unlock(&inode->i_lock);
745 /* Note: The reclaim code dictates that we add stateless
746 * and read-only stateids to the end of the list */
747 list_add_tail(&state->open_states, &owner->so_states);
748 spin_unlock(&owner->so_lock);
750 spin_unlock(&inode->i_lock);
751 spin_unlock(&owner->so_lock);
753 nfs4_free_open_state(new);
759 void nfs4_put_open_state(struct nfs4_state *state)
761 struct inode *inode = state->inode;
762 struct nfs4_state_owner *owner = state->owner;
764 if (!refcount_dec_and_lock(&state->count, &owner->so_lock))
766 spin_lock(&inode->i_lock);
767 list_del_rcu(&state->inode_states);
768 list_del(&state->open_states);
769 spin_unlock(&inode->i_lock);
770 spin_unlock(&owner->so_lock);
772 nfs4_free_open_state(state);
773 nfs4_put_state_owner(owner);
777 * Close the current file.
779 static void __nfs4_close(struct nfs4_state *state,
780 fmode_t fmode, gfp_t gfp_mask, int wait)
782 struct nfs4_state_owner *owner = state->owner;
786 atomic_inc(&owner->so_count);
787 /* Protect against nfs4_find_state() */
788 spin_lock(&owner->so_lock);
789 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
796 case FMODE_READ|FMODE_WRITE:
799 newstate = FMODE_READ|FMODE_WRITE;
800 if (state->n_rdwr == 0) {
801 if (state->n_rdonly == 0) {
802 newstate &= ~FMODE_READ;
803 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
804 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
806 if (state->n_wronly == 0) {
807 newstate &= ~FMODE_WRITE;
808 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
809 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
812 clear_bit(NFS_DELEGATED_STATE, &state->flags);
814 nfs4_state_set_mode_locked(state, newstate);
815 spin_unlock(&owner->so_lock);
818 nfs4_put_open_state(state);
819 nfs4_put_state_owner(owner);
821 nfs4_do_close(state, gfp_mask, wait);
824 void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
826 __nfs4_close(state, fmode, GFP_NOFS, 0);
829 void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
831 __nfs4_close(state, fmode, GFP_KERNEL, 1);
835 * Search the state->lock_states for an existing lock_owner
836 * that is compatible with either of the given owners.
837 * If the second is non-zero, then the first refers to a Posix-lock
838 * owner (current->files) and the second refers to a flock/OFD
839 * owner (struct file*). In that case, prefer a match for the first
841 * If both sorts of locks are held on the one file we cannot know
842 * which stateid was intended to be used, so a "correct" choice cannot
843 * be made. Failing that, a "consistent" choice is preferable. The
844 * consistent choice we make is to prefer the first owner, that of a
847 static struct nfs4_lock_state *
848 __nfs4_find_lock_state(struct nfs4_state *state,
849 fl_owner_t fl_owner, fl_owner_t fl_owner2)
851 struct nfs4_lock_state *pos, *ret = NULL;
852 list_for_each_entry(pos, &state->lock_states, ls_locks) {
853 if (pos->ls_owner == fl_owner) {
857 if (pos->ls_owner == fl_owner2)
861 refcount_inc(&ret->ls_count);
866 * Return a compatible lock_state. If no initialized lock_state structure
867 * exists, return an uninitialized one.
870 static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
872 struct nfs4_lock_state *lsp;
873 struct nfs_server *server = state->owner->so_server;
875 lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
878 nfs4_init_seqid_counter(&lsp->ls_seqid);
879 refcount_set(&lsp->ls_count, 1);
880 lsp->ls_state = state;
881 lsp->ls_owner = fl_owner;
882 lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
883 if (lsp->ls_seqid.owner_id < 0)
885 INIT_LIST_HEAD(&lsp->ls_locks);
892 void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
894 ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
895 nfs4_destroy_seqid_counter(&lsp->ls_seqid);
900 * Return a compatible lock_state. If no initialized lock_state structure
901 * exists, return an uninitialized one.
904 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
906 struct nfs4_lock_state *lsp, *new = NULL;
909 spin_lock(&state->state_lock);
910 lsp = __nfs4_find_lock_state(state, owner, NULL);
914 list_add(&new->ls_locks, &state->lock_states);
915 set_bit(LK_STATE_IN_USE, &state->flags);
920 spin_unlock(&state->state_lock);
921 new = nfs4_alloc_lock_state(state, owner);
925 spin_unlock(&state->state_lock);
927 nfs4_free_lock_state(state->owner->so_server, new);
932 * Release reference to lock_state, and free it if we see that
933 * it is no longer in use
935 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
937 struct nfs_server *server;
938 struct nfs4_state *state;
942 state = lsp->ls_state;
943 if (!refcount_dec_and_lock(&lsp->ls_count, &state->state_lock))
945 list_del(&lsp->ls_locks);
946 if (list_empty(&state->lock_states))
947 clear_bit(LK_STATE_IN_USE, &state->flags);
948 spin_unlock(&state->state_lock);
949 server = state->owner->so_server;
950 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
951 struct nfs_client *clp = server->nfs_client;
953 clp->cl_mvops->free_lock_state(server, lsp);
955 nfs4_free_lock_state(server, lsp);
958 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
960 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
962 dst->fl_u.nfs4_fl.owner = lsp;
963 refcount_inc(&lsp->ls_count);
966 static void nfs4_fl_release_lock(struct file_lock *fl)
968 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
971 static const struct file_lock_operations nfs4_fl_lock_ops = {
972 .fl_copy_lock = nfs4_fl_copy_lock,
973 .fl_release_private = nfs4_fl_release_lock,
976 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
978 struct nfs4_lock_state *lsp;
980 if (fl->fl_ops != NULL)
982 lsp = nfs4_get_lock_state(state, fl->fl_owner);
985 fl->fl_u.nfs4_fl.owner = lsp;
986 fl->fl_ops = &nfs4_fl_lock_ops;
990 static int nfs4_copy_lock_stateid(nfs4_stateid *dst,
991 struct nfs4_state *state,
992 const struct nfs_lock_context *l_ctx)
994 struct nfs4_lock_state *lsp;
995 fl_owner_t fl_owner, fl_flock_owner;
1001 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
1004 fl_owner = l_ctx->lockowner;
1005 fl_flock_owner = l_ctx->open_context->flock_owner;
1007 spin_lock(&state->state_lock);
1008 lsp = __nfs4_find_lock_state(state, fl_owner, fl_flock_owner);
1009 if (lsp && test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
1011 else if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
1012 nfs4_stateid_copy(dst, &lsp->ls_stateid);
1015 spin_unlock(&state->state_lock);
1016 nfs4_put_lock_state(lsp);
1021 bool nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
1024 const nfs4_stateid *src;
1029 src = &zero_stateid;
1030 seq = read_seqbegin(&state->seqlock);
1031 if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1032 src = &state->open_stateid;
1035 nfs4_stateid_copy(dst, src);
1036 } while (read_seqretry(&state->seqlock, seq));
1041 * Byte-range lock aware utility to initialize the stateid of read/write
1044 int nfs4_select_rw_stateid(struct nfs4_state *state,
1045 fmode_t fmode, const struct nfs_lock_context *l_ctx,
1046 nfs4_stateid *dst, const struct cred **cred)
1050 if (!nfs4_valid_open_stateid(state))
1054 ret = nfs4_copy_lock_stateid(dst, state, l_ctx);
1056 /* A lost lock - don't even consider delegations */
1058 /* returns true if delegation stateid found and copied */
1059 if (nfs4_copy_delegation_stateid(state->inode, fmode, dst, cred)) {
1064 /* nfs4_copy_delegation_stateid() didn't over-write
1065 * dst, so it still has the lock stateid which we now
1069 ret = nfs4_copy_open_stateid(dst, state) ? 0 : -EAGAIN;
1071 if (nfs_server_capable(state->inode, NFS_CAP_STATEID_NFSV41))
1076 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
1078 struct nfs_seqid *new;
1080 new = kmalloc(sizeof(*new), gfp_mask);
1082 return ERR_PTR(-ENOMEM);
1083 new->sequence = counter;
1084 INIT_LIST_HEAD(&new->list);
1089 void nfs_release_seqid(struct nfs_seqid *seqid)
1091 struct nfs_seqid_counter *sequence;
1093 if (seqid == NULL || list_empty(&seqid->list))
1095 sequence = seqid->sequence;
1096 spin_lock(&sequence->lock);
1097 list_del_init(&seqid->list);
1098 if (!list_empty(&sequence->list)) {
1099 struct nfs_seqid *next;
1101 next = list_first_entry(&sequence->list,
1102 struct nfs_seqid, list);
1103 rpc_wake_up_queued_task(&sequence->wait, next->task);
1105 spin_unlock(&sequence->lock);
1108 void nfs_free_seqid(struct nfs_seqid *seqid)
1110 nfs_release_seqid(seqid);
1115 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
1116 * failed with a seqid incrementing error -
1117 * see comments nfs4.h:seqid_mutating_error()
1119 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
1124 case -NFS4ERR_BAD_SEQID:
1125 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
1127 pr_warn_ratelimited("NFS: v4 server returned a bad"
1128 " sequence-id error on an"
1129 " unconfirmed sequence %p!\n",
1131 case -NFS4ERR_STALE_CLIENTID:
1132 case -NFS4ERR_STALE_STATEID:
1133 case -NFS4ERR_BAD_STATEID:
1134 case -NFS4ERR_BADXDR:
1135 case -NFS4ERR_RESOURCE:
1136 case -NFS4ERR_NOFILEHANDLE:
1137 case -NFS4ERR_MOVED:
1138 /* Non-seqid mutating errors */
1142 * Note: no locking needed as we are guaranteed to be first
1143 * on the sequence list
1145 seqid->sequence->counter++;
1148 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
1150 struct nfs4_state_owner *sp;
1155 sp = container_of(seqid->sequence, struct nfs4_state_owner, so_seqid);
1156 if (status == -NFS4ERR_BAD_SEQID)
1157 nfs4_reset_state_owner(sp);
1158 if (!nfs4_has_session(sp->so_server->nfs_client))
1159 nfs_increment_seqid(status, seqid);
1163 * Increment the seqid if the LOCK/LOCKU succeeded, or
1164 * failed with a seqid incrementing error -
1165 * see comments nfs4.h:seqid_mutating_error()
1167 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
1170 nfs_increment_seqid(status, seqid);
1173 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
1175 struct nfs_seqid_counter *sequence;
1180 sequence = seqid->sequence;
1181 spin_lock(&sequence->lock);
1183 if (list_empty(&seqid->list))
1184 list_add_tail(&seqid->list, &sequence->list);
1185 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
1187 rpc_sleep_on(&sequence->wait, task, NULL);
1190 spin_unlock(&sequence->lock);
1195 static int nfs4_run_state_manager(void *);
1197 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
1199 smp_mb__before_atomic();
1200 clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
1201 smp_mb__after_atomic();
1202 wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
1203 rpc_wake_up(&clp->cl_rpcwaitq);
1207 * Schedule the nfs_client asynchronous state management routine
1209 void nfs4_schedule_state_manager(struct nfs_client *clp)
1211 struct task_struct *task;
1212 char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
1214 set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
1215 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1217 __module_get(THIS_MODULE);
1218 refcount_inc(&clp->cl_count);
1220 /* The rcu_read_lock() is not strictly necessary, as the state
1221 * manager is the only thread that ever changes the rpc_xprt
1222 * after it's initialized. At this point, we're single threaded. */
1224 snprintf(buf, sizeof(buf), "%s-manager",
1225 rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR));
1227 task = kthread_run(nfs4_run_state_manager, clp, "%s", buf);
1229 printk(KERN_ERR "%s: kthread_run: %ld\n",
1230 __func__, PTR_ERR(task));
1231 if (!nfs_client_init_is_complete(clp))
1232 nfs_mark_client_ready(clp, PTR_ERR(task));
1233 nfs4_clear_state_manager_bit(clp);
1234 nfs_put_client(clp);
1235 module_put(THIS_MODULE);
1240 * Schedule a lease recovery attempt
1242 void nfs4_schedule_lease_recovery(struct nfs_client *clp)
1246 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1247 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1248 dprintk("%s: scheduling lease recovery for server %s\n", __func__,
1250 nfs4_schedule_state_manager(clp);
1252 EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
1255 * nfs4_schedule_migration_recovery - trigger migration recovery
1257 * @server: FSID that is migrating
1259 * Returns zero if recovery has started, otherwise a negative NFS4ERR
1260 * value is returned.
1262 int nfs4_schedule_migration_recovery(const struct nfs_server *server)
1264 struct nfs_client *clp = server->nfs_client;
1266 if (server->fh_expire_type != NFS4_FH_PERSISTENT) {
1267 pr_err("NFS: volatile file handles not supported (server %s)\n",
1272 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
1275 dprintk("%s: scheduling migration recovery for (%llx:%llx) on %s\n",
1277 (unsigned long long)server->fsid.major,
1278 (unsigned long long)server->fsid.minor,
1281 set_bit(NFS_MIG_IN_TRANSITION,
1282 &((struct nfs_server *)server)->mig_status);
1283 set_bit(NFS4CLNT_MOVED, &clp->cl_state);
1285 nfs4_schedule_state_manager(clp);
1288 EXPORT_SYMBOL_GPL(nfs4_schedule_migration_recovery);
1291 * nfs4_schedule_lease_moved_recovery - start lease-moved recovery
1293 * @clp: server to check for moved leases
1296 void nfs4_schedule_lease_moved_recovery(struct nfs_client *clp)
1298 dprintk("%s: scheduling lease-moved recovery for client ID %llx on %s\n",
1299 __func__, clp->cl_clientid, clp->cl_hostname);
1301 set_bit(NFS4CLNT_LEASE_MOVED, &clp->cl_state);
1302 nfs4_schedule_state_manager(clp);
1304 EXPORT_SYMBOL_GPL(nfs4_schedule_lease_moved_recovery);
1306 int nfs4_wait_clnt_recover(struct nfs_client *clp)
1312 refcount_inc(&clp->cl_count);
1313 res = wait_on_bit_action(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
1314 nfs_wait_bit_killable, TASK_KILLABLE);
1317 if (clp->cl_cons_state < 0)
1318 res = clp->cl_cons_state;
1320 nfs_put_client(clp);
1324 int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1329 for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1330 ret = nfs4_wait_clnt_recover(clp);
1333 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1334 !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1336 nfs4_schedule_state_manager(clp);
1343 * nfs40_handle_cb_pathdown - return all delegations after NFS4ERR_CB_PATH_DOWN
1344 * @clp: client to process
1346 * Set the NFS4CLNT_LEASE_EXPIRED state in order to force a
1347 * resend of the SETCLIENTID and hence re-establish the
1348 * callback channel. Then return all existing delegations.
1350 static void nfs40_handle_cb_pathdown(struct nfs_client *clp)
1352 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1353 nfs_expire_all_delegations(clp);
1354 dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
1358 void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
1360 nfs40_handle_cb_pathdown(clp);
1361 nfs4_schedule_state_manager(clp);
1364 static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1367 if (!nfs4_valid_open_stateid(state))
1369 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1370 /* Don't recover state that expired before the reboot */
1371 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
1372 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1375 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1376 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1380 int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1382 if (!nfs4_valid_open_stateid(state))
1384 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
1385 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1386 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1387 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
1391 int nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
1393 struct nfs_client *clp = server->nfs_client;
1395 if (!nfs4_state_mark_reclaim_nograce(clp, state))
1397 nfs_inode_find_delegation_state_and_recover(state->inode,
1399 dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
1401 nfs4_schedule_state_manager(clp);
1404 EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1406 static struct nfs4_lock_state *
1407 nfs_state_find_lock_state_by_stateid(struct nfs4_state *state,
1408 const nfs4_stateid *stateid)
1410 struct nfs4_lock_state *pos;
1412 list_for_each_entry(pos, &state->lock_states, ls_locks) {
1413 if (!test_bit(NFS_LOCK_INITIALIZED, &pos->ls_flags))
1415 if (nfs4_stateid_match_other(&pos->ls_stateid, stateid))
1421 static bool nfs_state_lock_state_matches_stateid(struct nfs4_state *state,
1422 const nfs4_stateid *stateid)
1426 if (test_bit(LK_STATE_IN_USE, &state->flags)) {
1427 spin_lock(&state->state_lock);
1428 if (nfs_state_find_lock_state_by_stateid(state, stateid))
1430 spin_unlock(&state->state_lock);
1435 void nfs_inode_find_state_and_recover(struct inode *inode,
1436 const nfs4_stateid *stateid)
1438 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1439 struct nfs_inode *nfsi = NFS_I(inode);
1440 struct nfs_open_context *ctx;
1441 struct nfs4_state *state;
1445 list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
1449 if (nfs4_stateid_match_other(&state->stateid, stateid) &&
1450 nfs4_state_mark_reclaim_nograce(clp, state)) {
1454 if (nfs4_stateid_match_other(&state->open_stateid, stateid) &&
1455 nfs4_state_mark_reclaim_nograce(clp, state)) {
1459 if (nfs_state_lock_state_matches_stateid(state, stateid) &&
1460 nfs4_state_mark_reclaim_nograce(clp, state))
1465 nfs_inode_find_delegation_state_and_recover(inode, stateid);
1467 nfs4_schedule_state_manager(clp);
1470 static void nfs4_state_mark_open_context_bad(struct nfs4_state *state, int err)
1472 struct inode *inode = state->inode;
1473 struct nfs_inode *nfsi = NFS_I(inode);
1474 struct nfs_open_context *ctx;
1477 list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
1478 if (ctx->state != state)
1480 set_bit(NFS_CONTEXT_BAD, &ctx->flags);
1481 pr_warn("NFSv4: state recovery failed for open file %pd2, "
1482 "error = %d\n", ctx->dentry, err);
1487 static void nfs4_state_mark_recovery_failed(struct nfs4_state *state, int error)
1489 set_bit(NFS_STATE_RECOVERY_FAILED, &state->flags);
1490 nfs4_state_mark_open_context_bad(state, error);
1494 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
1496 struct inode *inode = state->inode;
1497 struct nfs_inode *nfsi = NFS_I(inode);
1498 struct file_lock *fl;
1499 struct nfs4_lock_state *lsp;
1501 struct file_lock_context *flctx = inode->i_flctx;
1502 struct list_head *list;
1507 list = &flctx->flc_posix;
1509 /* Guard against delegation returns and new lock/unlock calls */
1510 down_write(&nfsi->rwsem);
1511 spin_lock(&flctx->flc_lock);
1513 list_for_each_entry(fl, list, fl_list) {
1514 if (nfs_file_open_context(fl->fl_file)->state != state)
1516 spin_unlock(&flctx->flc_lock);
1517 status = ops->recover_lock(state, fl);
1523 case -NFS4ERR_ADMIN_REVOKED:
1524 case -NFS4ERR_STALE_STATEID:
1525 case -NFS4ERR_BAD_STATEID:
1526 case -NFS4ERR_EXPIRED:
1527 case -NFS4ERR_NO_GRACE:
1528 case -NFS4ERR_STALE_CLIENTID:
1529 case -NFS4ERR_BADSESSION:
1530 case -NFS4ERR_BADSLOT:
1531 case -NFS4ERR_BAD_HIGH_SLOT:
1532 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1535 pr_err("NFS: %s: unhandled error %d\n",
1539 case -NFS4ERR_DENIED:
1540 case -NFS4ERR_RECLAIM_BAD:
1541 case -NFS4ERR_RECLAIM_CONFLICT:
1542 lsp = fl->fl_u.nfs4_fl.owner;
1544 set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
1547 spin_lock(&flctx->flc_lock);
1549 if (list == &flctx->flc_posix) {
1550 list = &flctx->flc_flock;
1553 spin_unlock(&flctx->flc_lock);
1555 up_write(&nfsi->rwsem);
1559 #ifdef CONFIG_NFS_V4_2
1560 static void nfs42_complete_copies(struct nfs4_state_owner *sp, struct nfs4_state *state)
1562 struct nfs4_copy_state *copy;
1564 if (!test_bit(NFS_CLNT_DST_SSC_COPY_STATE, &state->flags))
1567 spin_lock(&sp->so_server->nfs_client->cl_lock);
1568 list_for_each_entry(copy, &sp->so_server->ss_copies, copies) {
1569 if (!nfs4_stateid_match_other(&state->stateid, ©->parent_state->stateid))
1572 complete(©->completion);
1575 spin_unlock(&sp->so_server->nfs_client->cl_lock);
1577 #else /* !CONFIG_NFS_V4_2 */
1578 static inline void nfs42_complete_copies(struct nfs4_state_owner *sp,
1579 struct nfs4_state *state)
1582 #endif /* CONFIG_NFS_V4_2 */
1584 static int __nfs4_reclaim_open_state(struct nfs4_state_owner *sp, struct nfs4_state *state,
1585 const struct nfs4_state_recovery_ops *ops)
1587 struct nfs4_lock_state *lock;
1590 status = ops->recover_open(sp, state);
1594 status = nfs4_reclaim_locks(state, ops);
1598 if (!test_bit(NFS_DELEGATED_STATE, &state->flags)) {
1599 spin_lock(&state->state_lock);
1600 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1601 if (!test_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags))
1602 pr_warn_ratelimited("NFS: %s: Lock reclaim failed!\n", __func__);
1604 spin_unlock(&state->state_lock);
1607 nfs42_complete_copies(sp, state);
1608 clear_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
1612 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
1614 struct nfs4_state *state;
1615 unsigned int loop = 0;
1618 /* Note: we rely on the sp->so_states list being ordered
1619 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
1621 * This is needed to ensure that the server won't give us any
1622 * read delegations that we have to return if, say, we are
1623 * recovering after a network partition or a reboot from a
1624 * server that doesn't support a grace period.
1626 spin_lock(&sp->so_lock);
1627 raw_write_seqcount_begin(&sp->so_reclaim_seqcount);
1629 list_for_each_entry(state, &sp->so_states, open_states) {
1630 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
1632 if (!nfs4_valid_open_stateid(state))
1634 if (state->state == 0)
1636 refcount_inc(&state->count);
1637 spin_unlock(&sp->so_lock);
1638 status = __nfs4_reclaim_open_state(sp, state, ops);
1646 printk(KERN_ERR "NFS: %s: unhandled error %d\n", __func__, status);
1654 /* Open state on this file cannot be recovered */
1655 nfs4_state_mark_recovery_failed(state, status);
1660 set_bit(ops->state_flag_bit, &state->flags);
1664 case -NFS4ERR_ADMIN_REVOKED:
1665 case -NFS4ERR_STALE_STATEID:
1666 case -NFS4ERR_OLD_STATEID:
1667 case -NFS4ERR_BAD_STATEID:
1668 case -NFS4ERR_RECLAIM_BAD:
1669 case -NFS4ERR_RECLAIM_CONFLICT:
1670 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1672 case -NFS4ERR_EXPIRED:
1673 case -NFS4ERR_NO_GRACE:
1674 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1676 case -NFS4ERR_STALE_CLIENTID:
1677 case -NFS4ERR_BADSESSION:
1678 case -NFS4ERR_BADSLOT:
1679 case -NFS4ERR_BAD_HIGH_SLOT:
1680 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1684 nfs4_put_open_state(state);
1685 spin_lock(&sp->so_lock);
1688 raw_write_seqcount_end(&sp->so_reclaim_seqcount);
1689 spin_unlock(&sp->so_lock);
1692 nfs4_put_open_state(state);
1693 spin_lock(&sp->so_lock);
1694 raw_write_seqcount_end(&sp->so_reclaim_seqcount);
1695 spin_unlock(&sp->so_lock);
1699 static void nfs4_clear_open_state(struct nfs4_state *state)
1701 struct nfs4_lock_state *lock;
1703 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1704 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1705 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1706 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1707 spin_lock(&state->state_lock);
1708 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1709 lock->ls_seqid.flags = 0;
1710 clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
1712 spin_unlock(&state->state_lock);
1715 static void nfs4_reset_seqids(struct nfs_server *server,
1716 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1718 struct nfs_client *clp = server->nfs_client;
1719 struct nfs4_state_owner *sp;
1720 struct rb_node *pos;
1721 struct nfs4_state *state;
1723 spin_lock(&clp->cl_lock);
1724 for (pos = rb_first(&server->state_owners);
1726 pos = rb_next(pos)) {
1727 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1728 sp->so_seqid.flags = 0;
1729 spin_lock(&sp->so_lock);
1730 list_for_each_entry(state, &sp->so_states, open_states) {
1731 if (mark_reclaim(clp, state))
1732 nfs4_clear_open_state(state);
1734 spin_unlock(&sp->so_lock);
1736 spin_unlock(&clp->cl_lock);
1739 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp,
1740 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1742 struct nfs_server *server;
1745 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1746 nfs4_reset_seqids(server, mark_reclaim);
1750 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1752 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1753 /* Mark all delegations for reclaim */
1754 nfs_delegation_mark_reclaim(clp);
1755 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1758 static int nfs4_reclaim_complete(struct nfs_client *clp,
1759 const struct nfs4_state_recovery_ops *ops,
1760 const struct cred *cred)
1762 /* Notify the server we're done reclaiming our state */
1763 if (ops->reclaim_complete)
1764 return ops->reclaim_complete(clp, cred);
1768 static void nfs4_clear_reclaim_server(struct nfs_server *server)
1770 struct nfs_client *clp = server->nfs_client;
1771 struct nfs4_state_owner *sp;
1772 struct rb_node *pos;
1773 struct nfs4_state *state;
1775 spin_lock(&clp->cl_lock);
1776 for (pos = rb_first(&server->state_owners);
1778 pos = rb_next(pos)) {
1779 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1780 spin_lock(&sp->so_lock);
1781 list_for_each_entry(state, &sp->so_states, open_states) {
1782 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
1785 nfs4_state_mark_reclaim_nograce(clp, state);
1787 spin_unlock(&sp->so_lock);
1789 spin_unlock(&clp->cl_lock);
1792 static int nfs4_state_clear_reclaim_reboot(struct nfs_client *clp)
1794 struct nfs_server *server;
1796 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1800 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1801 nfs4_clear_reclaim_server(server);
1804 nfs_delegation_reap_unclaimed(clp);
1808 static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
1810 const struct nfs4_state_recovery_ops *ops;
1811 const struct cred *cred;
1814 if (!nfs4_state_clear_reclaim_reboot(clp))
1816 ops = clp->cl_mvops->reboot_recovery_ops;
1817 cred = nfs4_get_clid_cred(clp);
1818 err = nfs4_reclaim_complete(clp, ops, cred);
1820 if (err == -NFS4ERR_CONN_NOT_BOUND_TO_SESSION)
1821 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1824 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1826 nfs_mark_test_expired_all_delegations(clp);
1827 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1830 static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1835 case -NFS4ERR_CB_PATH_DOWN:
1836 nfs40_handle_cb_pathdown(clp);
1838 case -NFS4ERR_NO_GRACE:
1839 nfs4_state_end_reclaim_reboot(clp);
1841 case -NFS4ERR_STALE_CLIENTID:
1842 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1843 nfs4_state_start_reclaim_reboot(clp);
1845 case -NFS4ERR_EXPIRED:
1846 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1847 nfs4_state_start_reclaim_nograce(clp);
1849 case -NFS4ERR_BADSESSION:
1850 case -NFS4ERR_BADSLOT:
1851 case -NFS4ERR_BAD_HIGH_SLOT:
1852 case -NFS4ERR_DEADSESSION:
1853 case -NFS4ERR_SEQ_FALSE_RETRY:
1854 case -NFS4ERR_SEQ_MISORDERED:
1855 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1856 /* Zero session reset errors */
1858 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1859 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1862 dprintk("%s: failed to handle error %d for server %s\n",
1863 __func__, error, clp->cl_hostname);
1866 dprintk("%s: handled error %d for server %s\n", __func__, error,
1871 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1873 struct nfs4_state_owner *sp;
1874 struct nfs_server *server;
1875 struct rb_node *pos;
1881 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1882 nfs4_purge_state_owners(server, &freeme);
1883 spin_lock(&clp->cl_lock);
1884 for (pos = rb_first(&server->state_owners);
1886 pos = rb_next(pos)) {
1888 struct nfs4_state_owner, so_server_node);
1889 if (!test_and_clear_bit(ops->owner_flag_bit,
1892 if (!atomic_inc_not_zero(&sp->so_count))
1894 spin_unlock(&clp->cl_lock);
1897 status = nfs4_reclaim_open_state(sp, ops);
1899 set_bit(ops->owner_flag_bit, &sp->so_flags);
1900 nfs4_put_state_owner(sp);
1901 status = nfs4_recovery_handle_error(clp, status);
1902 return (status != 0) ? status : -EAGAIN;
1905 nfs4_put_state_owner(sp);
1908 spin_unlock(&clp->cl_lock);
1911 nfs4_free_state_owners(&freeme);
1915 static int nfs4_check_lease(struct nfs_client *clp)
1917 const struct cred *cred;
1918 const struct nfs4_state_maintenance_ops *ops =
1919 clp->cl_mvops->state_renewal_ops;
1922 /* Is the client already known to have an expired lease? */
1923 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1925 cred = ops->get_state_renewal_cred(clp);
1927 cred = nfs4_get_clid_cred(clp);
1932 status = ops->renew_lease(clp, cred);
1934 if (status == -ETIMEDOUT) {
1935 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1939 return nfs4_recovery_handle_error(clp, status);
1942 /* Set NFS4CLNT_LEASE_EXPIRED and reclaim reboot state for all v4.0 errors
1943 * and for recoverable errors on EXCHANGE_ID for v4.1
1945 static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
1948 case -NFS4ERR_SEQ_MISORDERED:
1949 if (test_and_set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state))
1950 return -ESERVERFAULT;
1951 /* Lease confirmation error: retry after purging the lease */
1953 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1955 case -NFS4ERR_STALE_CLIENTID:
1956 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1957 nfs4_state_start_reclaim_reboot(clp);
1959 case -NFS4ERR_CLID_INUSE:
1960 pr_err("NFS: Server %s reports our clientid is in use\n",
1962 nfs_mark_client_ready(clp, -EPERM);
1963 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1966 case -NFS4ERR_DELAY:
1971 case -NFS4ERR_MINOR_VERS_MISMATCH:
1972 if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
1973 nfs_mark_client_ready(clp, -EPROTONOSUPPORT);
1974 dprintk("%s: exit with error %d for server %s\n",
1975 __func__, -EPROTONOSUPPORT, clp->cl_hostname);
1976 return -EPROTONOSUPPORT;
1977 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1978 * in nfs4_exchange_id */
1980 dprintk("%s: exit with error %d for server %s\n", __func__,
1981 status, clp->cl_hostname);
1984 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1985 dprintk("%s: handled error %d for server %s\n", __func__, status,
1990 static int nfs4_establish_lease(struct nfs_client *clp)
1992 const struct cred *cred;
1993 const struct nfs4_state_recovery_ops *ops =
1994 clp->cl_mvops->reboot_recovery_ops;
1997 status = nfs4_begin_drain_session(clp);
2000 cred = nfs4_get_clid_cred(clp);
2003 status = ops->establish_clid(clp, cred);
2007 pnfs_destroy_all_layouts(clp);
2012 * Returns zero or a negative errno. NFS4ERR values are converted
2013 * to local errno values.
2015 static int nfs4_reclaim_lease(struct nfs_client *clp)
2019 status = nfs4_establish_lease(clp);
2021 return nfs4_handle_reclaim_lease_error(clp, status);
2022 if (test_and_clear_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state))
2023 nfs4_state_start_reclaim_nograce(clp);
2024 if (!test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
2025 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
2026 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
2027 clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
2031 static int nfs4_purge_lease(struct nfs_client *clp)
2035 status = nfs4_establish_lease(clp);
2037 return nfs4_handle_reclaim_lease_error(clp, status);
2038 clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
2039 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
2040 nfs4_state_start_reclaim_nograce(clp);
2045 * Try remote migration of one FSID from a source server to a
2046 * destination server. The source server provides a list of
2047 * potential destinations.
2049 * Returns zero or a negative NFS4ERR status code.
2051 static int nfs4_try_migration(struct nfs_server *server, const struct cred *cred)
2053 struct nfs_client *clp = server->nfs_client;
2054 struct nfs4_fs_locations *locations = NULL;
2055 struct inode *inode;
2059 dprintk("--> %s: FSID %llx:%llx on \"%s\"\n", __func__,
2060 (unsigned long long)server->fsid.major,
2061 (unsigned long long)server->fsid.minor,
2065 page = alloc_page(GFP_KERNEL);
2066 locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2067 if (page == NULL || locations == NULL) {
2068 dprintk("<-- %s: no memory\n", __func__);
2072 inode = d_inode(server->super->s_root);
2073 result = nfs4_proc_get_locations(inode, locations, page, cred);
2075 dprintk("<-- %s: failed to retrieve fs_locations: %d\n",
2080 result = -NFS4ERR_NXIO;
2081 if (!locations->nlocations)
2084 if (!(locations->fattr.valid & NFS_ATTR_FATTR_V4_LOCATIONS)) {
2085 dprintk("<-- %s: No fs_locations data, migration skipped\n",
2090 status = nfs4_begin_drain_session(clp);
2096 status = nfs4_replace_transport(server, locations);
2098 dprintk("<-- %s: failed to replace transport: %d\n",
2104 dprintk("<-- %s: migration succeeded\n", __func__);
2111 pr_err("NFS: migration recovery failed (server %s)\n",
2113 set_bit(NFS_MIG_FAILED, &server->mig_status);
2119 * Returns zero or a negative NFS4ERR status code.
2121 static int nfs4_handle_migration(struct nfs_client *clp)
2123 const struct nfs4_state_maintenance_ops *ops =
2124 clp->cl_mvops->state_renewal_ops;
2125 struct nfs_server *server;
2126 const struct cred *cred;
2128 dprintk("%s: migration reported on \"%s\"\n", __func__,
2131 cred = ops->get_state_renewal_cred(clp);
2133 return -NFS4ERR_NOENT;
2138 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
2141 if (server->mig_gen == clp->cl_mig_gen)
2143 server->mig_gen = clp->cl_mig_gen;
2145 if (!test_and_clear_bit(NFS_MIG_IN_TRANSITION,
2146 &server->mig_status))
2150 status = nfs4_try_migration(server, cred);
2163 * Test each nfs_server on the clp's cl_superblocks list to see
2164 * if it's moved to another server. Stop when the server no longer
2165 * returns NFS4ERR_LEASE_MOVED.
2167 static int nfs4_handle_lease_moved(struct nfs_client *clp)
2169 const struct nfs4_state_maintenance_ops *ops =
2170 clp->cl_mvops->state_renewal_ops;
2171 struct nfs_server *server;
2172 const struct cred *cred;
2174 dprintk("%s: lease moved reported on \"%s\"\n", __func__,
2177 cred = ops->get_state_renewal_cred(clp);
2179 return -NFS4ERR_NOENT;
2184 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
2185 struct inode *inode;
2188 if (server->mig_gen == clp->cl_mig_gen)
2190 server->mig_gen = clp->cl_mig_gen;
2194 inode = d_inode(server->super->s_root);
2195 status = nfs4_proc_fsid_present(inode, cred);
2196 if (status != -NFS4ERR_MOVED)
2197 goto restart; /* wasn't this one */
2198 if (nfs4_try_migration(server, cred) == -NFS4ERR_LEASE_MOVED)
2199 goto restart; /* there are more */
2210 * nfs4_discover_server_trunking - Detect server IP address trunking
2212 * @clp: nfs_client under test
2213 * @result: OUT: found nfs_client, or clp
2215 * Returns zero or a negative errno. If zero is returned,
2216 * an nfs_client pointer is planted in "result".
2218 * Note: since we are invoked in process context, and
2219 * not from inside the state manager, we cannot use
2220 * nfs4_handle_reclaim_lease_error().
2222 int nfs4_discover_server_trunking(struct nfs_client *clp,
2223 struct nfs_client **result)
2225 const struct nfs4_state_recovery_ops *ops =
2226 clp->cl_mvops->reboot_recovery_ops;
2227 struct rpc_clnt *clnt;
2228 const struct cred *cred;
2231 dprintk("NFS: %s: testing '%s'\n", __func__, clp->cl_hostname);
2233 clnt = clp->cl_rpcclient;
2236 mutex_lock(&nfs_clid_init_mutex);
2239 cred = nfs4_get_clid_cred(clp);
2243 status = ops->detect_trunking(clp, result, cred);
2251 if (clnt->cl_softrtry)
2254 case -NFS4ERR_DELAY:
2258 case -NFS4ERR_STALE_CLIENTID:
2259 dprintk("NFS: %s after status %d, retrying\n",
2264 nfs4_root_machine_cred(clp);
2267 if (clnt->cl_auth->au_flavor == RPC_AUTH_UNIX)
2270 case -NFS4ERR_CLID_INUSE:
2271 case -NFS4ERR_WRONGSEC:
2272 /* No point in retrying if we already used RPC_AUTH_UNIX */
2273 if (clnt->cl_auth->au_flavor == RPC_AUTH_UNIX) {
2277 clnt = rpc_clone_client_set_auth(clnt, RPC_AUTH_UNIX);
2279 status = PTR_ERR(clnt);
2282 /* Note: this is safe because we haven't yet marked the
2283 * client as ready, so we are the only user of
2286 clnt = xchg(&clp->cl_rpcclient, clnt);
2287 rpc_shutdown_client(clnt);
2288 clnt = clp->cl_rpcclient;
2291 case -NFS4ERR_MINOR_VERS_MISMATCH:
2292 status = -EPROTONOSUPPORT;
2296 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
2297 * in nfs4_exchange_id */
2298 status = -EKEYEXPIRED;
2301 pr_warn("NFS: %s unhandled error %d. Exiting with error EIO\n",
2307 mutex_unlock(&nfs_clid_init_mutex);
2308 dprintk("NFS: %s: status = %d\n", __func__, status);
2312 #ifdef CONFIG_NFS_V4_1
2313 void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
2315 struct nfs_client *clp = session->clp;
2319 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
2321 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2322 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2324 nfs4_schedule_state_manager(clp);
2326 EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
2328 void nfs41_notify_server(struct nfs_client *clp)
2330 /* Use CHECK_LEASE to ping the server with a SEQUENCE */
2331 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
2332 nfs4_schedule_state_manager(clp);
2335 static void nfs4_reset_all_state(struct nfs_client *clp)
2337 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
2338 set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
2339 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
2340 nfs4_state_start_reclaim_nograce(clp);
2341 dprintk("%s: scheduling reset of all state for server %s!\n",
2342 __func__, clp->cl_hostname);
2343 nfs4_schedule_state_manager(clp);
2347 static void nfs41_handle_server_reboot(struct nfs_client *clp)
2349 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
2350 nfs4_state_start_reclaim_reboot(clp);
2351 dprintk("%s: server %s rebooted!\n", __func__,
2353 nfs4_schedule_state_manager(clp);
2357 static void nfs41_handle_all_state_revoked(struct nfs_client *clp)
2359 nfs4_reset_all_state(clp);
2360 dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
2363 static void nfs41_handle_some_state_revoked(struct nfs_client *clp)
2365 nfs4_state_start_reclaim_nograce(clp);
2366 nfs4_schedule_state_manager(clp);
2368 dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
2371 static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
2373 /* FIXME: For now, we destroy all layouts. */
2374 pnfs_destroy_all_layouts(clp);
2375 nfs_test_expired_all_delegations(clp);
2376 dprintk("%s: Recallable state revoked on server %s!\n", __func__,
2380 static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
2382 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
2383 nfs4_schedule_state_manager(clp);
2385 dprintk("%s: server %s declared a backchannel fault\n", __func__,
2389 static void nfs41_handle_cb_path_down(struct nfs_client *clp)
2391 if (test_and_set_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2392 &clp->cl_state) == 0)
2393 nfs4_schedule_state_manager(clp);
2396 void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags,
2402 dprintk("%s: \"%s\" (client ID %llx) flags=0x%08x\n",
2403 __func__, clp->cl_hostname, clp->cl_clientid, flags);
2405 * If we're called from the state manager thread, then assume we're
2406 * already handling the RECLAIM_NEEDED and/or STATE_REVOKED.
2407 * Those flags are expected to remain set until we're done
2408 * recovering (see RFC5661, section 18.46.3).
2413 if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
2414 nfs41_handle_server_reboot(clp);
2415 if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED))
2416 nfs41_handle_all_state_revoked(clp);
2417 if (flags & (SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
2418 SEQ4_STATUS_ADMIN_STATE_REVOKED))
2419 nfs41_handle_some_state_revoked(clp);
2420 if (flags & SEQ4_STATUS_LEASE_MOVED)
2421 nfs4_schedule_lease_moved_recovery(clp);
2422 if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
2423 nfs41_handle_recallable_state_revoked(clp);
2425 if (flags & SEQ4_STATUS_BACKCHANNEL_FAULT)
2426 nfs41_handle_backchannel_fault(clp);
2427 else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
2428 SEQ4_STATUS_CB_PATH_DOWN_SESSION))
2429 nfs41_handle_cb_path_down(clp);
2432 static int nfs4_reset_session(struct nfs_client *clp)
2434 const struct cred *cred;
2437 if (!nfs4_has_session(clp))
2439 status = nfs4_begin_drain_session(clp);
2442 cred = nfs4_get_clid_cred(clp);
2443 status = nfs4_proc_destroy_session(clp->cl_session, cred);
2446 case -NFS4ERR_BADSESSION:
2447 case -NFS4ERR_DEADSESSION:
2449 case -NFS4ERR_BACK_CHAN_BUSY:
2450 case -NFS4ERR_DELAY:
2451 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
2456 status = nfs4_recovery_handle_error(clp, status);
2460 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
2461 status = nfs4_proc_create_session(clp, cred);
2463 dprintk("%s: session reset failed with status %d for server %s!\n",
2464 __func__, status, clp->cl_hostname);
2465 status = nfs4_handle_reclaim_lease_error(clp, status);
2468 nfs41_finish_session_reset(clp);
2469 dprintk("%s: session reset was successful for server %s!\n",
2470 __func__, clp->cl_hostname);
2476 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2478 const struct cred *cred;
2481 if (!nfs4_has_session(clp))
2483 ret = nfs4_begin_drain_session(clp);
2486 cred = nfs4_get_clid_cred(clp);
2487 ret = nfs4_proc_bind_conn_to_session(clp, cred);
2489 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2492 dprintk("%s: bind_conn_to_session was successful for server %s!\n",
2493 __func__, clp->cl_hostname);
2495 case -NFS4ERR_DELAY:
2497 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2500 return nfs4_recovery_handle_error(clp, ret);
2504 #else /* CONFIG_NFS_V4_1 */
2505 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
2507 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2511 #endif /* CONFIG_NFS_V4_1 */
2513 static void nfs4_state_manager(struct nfs_client *clp)
2515 unsigned int memflags;
2517 const char *section = "", *section_sep = "";
2520 * State recovery can deadlock if the direct reclaim code tries
2521 * start NFS writeback. So ensure memory allocations are all
2524 memflags = memalloc_nofs_save();
2526 /* Ensure exclusive access to NFSv4 state */
2528 clear_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
2529 if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
2530 section = "purge state";
2531 status = nfs4_purge_lease(clp);
2537 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
2538 section = "lease expired";
2539 /* We're going to have to re-establish a clientid */
2540 status = nfs4_reclaim_lease(clp);
2546 /* Initialize or reset the session */
2547 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
2548 section = "reset session";
2549 status = nfs4_reset_session(clp);
2550 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
2556 /* Send BIND_CONN_TO_SESSION */
2557 if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2559 section = "bind conn to session";
2560 status = nfs4_bind_conn_to_session(clp);
2566 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
2567 section = "check lease";
2568 status = nfs4_check_lease(clp);
2574 if (test_and_clear_bit(NFS4CLNT_MOVED, &clp->cl_state)) {
2575 section = "migration";
2576 status = nfs4_handle_migration(clp);
2581 if (test_and_clear_bit(NFS4CLNT_LEASE_MOVED, &clp->cl_state)) {
2582 section = "lease moved";
2583 status = nfs4_handle_lease_moved(clp);
2588 /* First recover reboot state... */
2589 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
2590 section = "reclaim reboot";
2591 status = nfs4_do_reclaim(clp,
2592 clp->cl_mvops->reboot_recovery_ops);
2593 if (status == -EAGAIN)
2597 nfs4_state_end_reclaim_reboot(clp);
2601 /* Detect expired delegations... */
2602 if (test_and_clear_bit(NFS4CLNT_DELEGATION_EXPIRED, &clp->cl_state)) {
2603 section = "detect expired delegations";
2604 nfs_reap_expired_delegations(clp);
2608 /* Now recover expired state... */
2609 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
2610 section = "reclaim nograce";
2611 status = nfs4_do_reclaim(clp,
2612 clp->cl_mvops->nograce_recovery_ops);
2613 if (status == -EAGAIN)
2617 clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
2620 memalloc_nofs_restore(memflags);
2621 nfs4_end_drain_session(clp);
2622 nfs4_clear_state_manager_bit(clp);
2624 if (!test_and_set_bit(NFS4CLNT_DELEGRETURN_RUNNING, &clp->cl_state)) {
2625 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
2626 nfs_client_return_marked_delegations(clp);
2627 set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
2629 clear_bit(NFS4CLNT_DELEGRETURN_RUNNING, &clp->cl_state);
2632 /* Did we race with an attempt to give us more work? */
2633 if (!test_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state))
2635 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
2637 memflags = memalloc_nofs_save();
2638 } while (refcount_read(&clp->cl_count) > 1 && !signalled());
2642 if (strlen(section))
2644 pr_warn_ratelimited("NFS: state manager%s%s failed on NFSv4 server %s"
2645 " with error %d\n", section_sep, section,
2646 clp->cl_hostname, -status);
2649 memalloc_nofs_restore(memflags);
2650 nfs4_end_drain_session(clp);
2651 nfs4_clear_state_manager_bit(clp);
2654 static int nfs4_run_state_manager(void *ptr)
2656 struct nfs_client *clp = ptr;
2658 allow_signal(SIGKILL);
2659 nfs4_state_manager(clp);
2660 nfs_put_client(clp);
2661 module_put_and_exit(0);