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"
64 #include "nfs4trace.h"
66 #define NFSDBG_FACILITY NFSDBG_STATE
68 #define OPENOWNER_POOL_SIZE 8
70 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp);
72 const nfs4_stateid zero_stateid = {
74 .type = NFS4_SPECIAL_STATEID_TYPE,
76 const nfs4_stateid invalid_stateid = {
78 /* Funky initialiser keeps older gcc versions happy */
79 .data = { 0xff, 0xff, 0xff, 0xff, 0 },
81 .type = NFS4_INVALID_STATEID_TYPE,
84 const nfs4_stateid current_stateid = {
86 /* Funky initialiser keeps older gcc versions happy */
87 .data = { 0x0, 0x0, 0x0, 0x1, 0 },
89 .type = NFS4_SPECIAL_STATEID_TYPE,
92 static DEFINE_MUTEX(nfs_clid_init_mutex);
94 static int nfs4_setup_state_renewal(struct nfs_client *clp)
97 struct nfs_fsinfo fsinfo;
99 if (!test_bit(NFS_CS_CHECK_LEASE_TIME, &clp->cl_res_state)) {
100 nfs4_schedule_state_renewal(clp);
104 status = nfs4_proc_get_lease_time(clp, &fsinfo);
106 nfs4_set_lease_period(clp, fsinfo.lease_time * HZ);
107 nfs4_schedule_state_renewal(clp);
113 int nfs4_init_clientid(struct nfs_client *clp, const struct cred *cred)
115 struct nfs4_setclientid_res clid = {
116 .clientid = clp->cl_clientid,
117 .confirm = clp->cl_confirm,
121 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
123 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
125 port = nn->nfs_callback_tcpport;
126 if (clp->cl_addr.ss_family == AF_INET6)
127 port = nn->nfs_callback_tcpport6;
129 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
132 clp->cl_clientid = clid.clientid;
133 clp->cl_confirm = clid.confirm;
134 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
136 status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
139 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
140 nfs4_setup_state_renewal(clp);
146 * nfs40_discover_server_trunking - Detect server IP address trunking (mv0)
148 * @clp: nfs_client under test
149 * @result: OUT: found nfs_client, or clp
150 * @cred: credential to use for trunking test
152 * Returns zero, a negative errno, or a negative NFS4ERR status.
153 * If zero is returned, an nfs_client pointer is planted in
156 * Note: The returned client may not yet be marked ready.
158 int nfs40_discover_server_trunking(struct nfs_client *clp,
159 struct nfs_client **result,
160 const struct cred *cred)
162 struct nfs4_setclientid_res clid = {
163 .clientid = clp->cl_clientid,
164 .confirm = clp->cl_confirm,
166 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
170 port = nn->nfs_callback_tcpport;
171 if (clp->cl_addr.ss_family == AF_INET6)
172 port = nn->nfs_callback_tcpport6;
174 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
177 clp->cl_clientid = clid.clientid;
178 clp->cl_confirm = clid.confirm;
180 status = nfs40_walk_client_list(clp, result, cred);
182 /* Sustain the lease, even if it's empty. If the clientid4
183 * goes stale it's of no use for trunking discovery. */
184 nfs4_schedule_state_renewal(*result);
186 /* If the client state need to recover, do it. */
188 nfs4_schedule_state_manager(clp);
194 const struct cred *nfs4_get_machine_cred(struct nfs_client *clp)
196 return get_cred(rpc_machine_cred());
199 static void nfs4_root_machine_cred(struct nfs_client *clp)
202 /* Force root creds instead of machine */
203 clp->cl_principal = NULL;
204 clp->cl_rpcclient->cl_principal = NULL;
207 static const struct cred *
208 nfs4_get_renew_cred_server_locked(struct nfs_server *server)
210 const struct cred *cred = NULL;
211 struct nfs4_state_owner *sp;
214 for (pos = rb_first(&server->state_owners);
216 pos = rb_next(pos)) {
217 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
218 if (list_empty(&sp->so_states))
220 cred = get_cred(sp->so_cred);
227 * nfs4_get_renew_cred - Acquire credential for a renew operation
228 * @clp: client state handle
230 * Returns an rpc_cred with reference count bumped, or NULL.
231 * Caller must hold clp->cl_lock.
233 const struct cred *nfs4_get_renew_cred(struct nfs_client *clp)
235 const struct cred *cred = NULL;
236 struct nfs_server *server;
238 /* Use machine credentials if available */
239 cred = nfs4_get_machine_cred(clp);
243 spin_lock(&clp->cl_lock);
245 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
246 cred = nfs4_get_renew_cred_server_locked(server);
251 spin_unlock(&clp->cl_lock);
257 static void nfs4_end_drain_slot_table(struct nfs4_slot_table *tbl)
259 if (test_and_clear_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state)) {
260 spin_lock(&tbl->slot_tbl_lock);
261 nfs41_wake_slot_table(tbl);
262 spin_unlock(&tbl->slot_tbl_lock);
266 static void nfs4_end_drain_session(struct nfs_client *clp)
268 struct nfs4_session *ses = clp->cl_session;
270 if (clp->cl_slot_tbl) {
271 nfs4_end_drain_slot_table(clp->cl_slot_tbl);
276 nfs4_end_drain_slot_table(&ses->bc_slot_table);
277 nfs4_end_drain_slot_table(&ses->fc_slot_table);
281 static int nfs4_drain_slot_tbl(struct nfs4_slot_table *tbl)
283 set_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state);
284 spin_lock(&tbl->slot_tbl_lock);
285 if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
286 reinit_completion(&tbl->complete);
287 spin_unlock(&tbl->slot_tbl_lock);
288 return wait_for_completion_interruptible(&tbl->complete);
290 spin_unlock(&tbl->slot_tbl_lock);
294 static int nfs4_begin_drain_session(struct nfs_client *clp)
296 struct nfs4_session *ses = clp->cl_session;
299 if (clp->cl_slot_tbl)
300 return nfs4_drain_slot_tbl(clp->cl_slot_tbl);
303 ret = nfs4_drain_slot_tbl(&ses->bc_slot_table);
307 return nfs4_drain_slot_tbl(&ses->fc_slot_table);
310 #if defined(CONFIG_NFS_V4_1)
312 static void nfs41_finish_session_reset(struct nfs_client *clp)
314 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
315 clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
316 /* create_session negotiated new slot table */
317 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
318 nfs4_setup_state_renewal(clp);
321 int nfs41_init_clientid(struct nfs_client *clp, const struct cred *cred)
325 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
327 status = nfs4_proc_exchange_id(clp, cred);
330 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
332 status = nfs4_proc_create_session(clp, cred);
335 if (!(clp->cl_exchange_flags & EXCHGID4_FLAG_CONFIRMED_R))
336 nfs4_state_start_reclaim_reboot(clp);
337 nfs41_finish_session_reset(clp);
338 nfs_mark_client_ready(clp, NFS_CS_READY);
344 * nfs41_discover_server_trunking - Detect server IP address trunking (mv1)
346 * @clp: nfs_client under test
347 * @result: OUT: found nfs_client, or clp
348 * @cred: credential to use for trunking test
350 * Returns NFS4_OK, a negative errno, or a negative NFS4ERR status.
351 * If NFS4_OK is returned, an nfs_client pointer is planted in
354 * Note: The returned client may not yet be marked ready.
356 int nfs41_discover_server_trunking(struct nfs_client *clp,
357 struct nfs_client **result,
358 const struct cred *cred)
362 status = nfs4_proc_exchange_id(clp, cred);
363 if (status != NFS4_OK)
366 status = nfs41_walk_client_list(clp, result, cred);
373 * Purge state if the client id was established in a prior
374 * instance and the client id could not have arrived on the
375 * server via Transparent State Migration.
377 if (clp->cl_exchange_flags & EXCHGID4_FLAG_CONFIRMED_R) {
378 if (!test_bit(NFS_CS_TSM_POSSIBLE, &clp->cl_flags))
379 set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
381 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
383 nfs4_schedule_state_manager(clp);
384 status = nfs_wait_client_init_complete(clp);
390 #endif /* CONFIG_NFS_V4_1 */
393 * nfs4_get_clid_cred - Acquire credential for a setclientid operation
394 * @clp: client state handle
396 * Returns a cred with reference count bumped, or NULL.
398 const struct cred *nfs4_get_clid_cred(struct nfs_client *clp)
400 const struct cred *cred;
402 cred = nfs4_get_machine_cred(clp);
406 static struct nfs4_state_owner *
407 nfs4_find_state_owner_locked(struct nfs_server *server, const struct cred *cred)
409 struct rb_node **p = &server->state_owners.rb_node,
411 struct nfs4_state_owner *sp;
416 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
417 cmp = cred_fscmp(cred, sp->so_cred);
420 p = &parent->rb_left;
422 p = &parent->rb_right;
424 if (!list_empty(&sp->so_lru))
425 list_del_init(&sp->so_lru);
426 atomic_inc(&sp->so_count);
433 static struct nfs4_state_owner *
434 nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
436 struct nfs_server *server = new->so_server;
437 struct rb_node **p = &server->state_owners.rb_node,
439 struct nfs4_state_owner *sp;
444 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
445 cmp = cred_fscmp(new->so_cred, sp->so_cred);
448 p = &parent->rb_left;
450 p = &parent->rb_right;
452 if (!list_empty(&sp->so_lru))
453 list_del_init(&sp->so_lru);
454 atomic_inc(&sp->so_count);
458 rb_link_node(&new->so_server_node, parent, p);
459 rb_insert_color(&new->so_server_node, &server->state_owners);
464 nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
466 struct nfs_server *server = sp->so_server;
468 if (!RB_EMPTY_NODE(&sp->so_server_node))
469 rb_erase(&sp->so_server_node, &server->state_owners);
473 nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
475 sc->create_time = ktime_get();
478 spin_lock_init(&sc->lock);
479 INIT_LIST_HEAD(&sc->list);
480 rpc_init_wait_queue(&sc->wait, "Seqid_waitqueue");
484 nfs4_destroy_seqid_counter(struct nfs_seqid_counter *sc)
486 rpc_destroy_wait_queue(&sc->wait);
490 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
491 * create a new state_owner.
494 static struct nfs4_state_owner *
495 nfs4_alloc_state_owner(struct nfs_server *server,
496 const struct cred *cred,
499 struct nfs4_state_owner *sp;
501 sp = kzalloc(sizeof(*sp), gfp_flags);
504 sp->so_seqid.owner_id = ida_alloc(&server->openowner_id, gfp_flags);
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_spinlock_init(&sp->so_reclaim_seqcount, &sp->so_lock);
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_free(&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_free_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_KERNEL_ACCOUNT);
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);
771 nfs4_inode_return_delegation_on_close(inode);
773 nfs4_free_open_state(state);
774 nfs4_put_state_owner(owner);
778 * Close the current file.
780 static void __nfs4_close(struct nfs4_state *state,
781 fmode_t fmode, gfp_t gfp_mask, int wait)
783 struct nfs4_state_owner *owner = state->owner;
787 atomic_inc(&owner->so_count);
788 /* Protect against nfs4_find_state() */
789 spin_lock(&owner->so_lock);
790 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
797 case FMODE_READ|FMODE_WRITE:
800 newstate = FMODE_READ|FMODE_WRITE;
801 if (state->n_rdwr == 0) {
802 if (state->n_rdonly == 0) {
803 newstate &= ~FMODE_READ;
804 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
805 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
807 if (state->n_wronly == 0) {
808 newstate &= ~FMODE_WRITE;
809 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
810 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
813 clear_bit(NFS_DELEGATED_STATE, &state->flags);
815 nfs4_state_set_mode_locked(state, newstate);
816 spin_unlock(&owner->so_lock);
819 nfs4_put_open_state(state);
820 nfs4_put_state_owner(owner);
822 nfs4_do_close(state, gfp_mask, wait);
825 void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
827 __nfs4_close(state, fmode, GFP_KERNEL, 0);
830 void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
832 __nfs4_close(state, fmode, GFP_KERNEL, 1);
836 * Search the state->lock_states for an existing lock_owner
837 * that is compatible with either of the given owners.
838 * If the second is non-zero, then the first refers to a Posix-lock
839 * owner (current->files) and the second refers to a flock/OFD
840 * owner (struct file*). In that case, prefer a match for the first
842 * If both sorts of locks are held on the one file we cannot know
843 * which stateid was intended to be used, so a "correct" choice cannot
844 * be made. Failing that, a "consistent" choice is preferable. The
845 * consistent choice we make is to prefer the first owner, that of a
848 static struct nfs4_lock_state *
849 __nfs4_find_lock_state(struct nfs4_state *state,
850 fl_owner_t fl_owner, fl_owner_t fl_owner2)
852 struct nfs4_lock_state *pos, *ret = NULL;
853 list_for_each_entry(pos, &state->lock_states, ls_locks) {
854 if (pos->ls_owner == fl_owner) {
858 if (pos->ls_owner == fl_owner2)
862 refcount_inc(&ret->ls_count);
867 * Return a compatible lock_state. If no initialized lock_state structure
868 * exists, return an uninitialized one.
871 static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
873 struct nfs4_lock_state *lsp;
874 struct nfs_server *server = state->owner->so_server;
876 lsp = kzalloc(sizeof(*lsp), GFP_KERNEL_ACCOUNT);
879 nfs4_init_seqid_counter(&lsp->ls_seqid);
880 refcount_set(&lsp->ls_count, 1);
881 lsp->ls_state = state;
882 lsp->ls_owner = fl_owner;
883 lsp->ls_seqid.owner_id = ida_alloc(&server->lockowner_id, GFP_KERNEL_ACCOUNT);
884 if (lsp->ls_seqid.owner_id < 0)
886 INIT_LIST_HEAD(&lsp->ls_locks);
893 void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
895 ida_free(&server->lockowner_id, lsp->ls_seqid.owner_id);
896 nfs4_destroy_seqid_counter(&lsp->ls_seqid);
901 * Return a compatible lock_state. If no initialized lock_state structure
902 * exists, return an uninitialized one.
905 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
907 struct nfs4_lock_state *lsp, *new = NULL;
910 spin_lock(&state->state_lock);
911 lsp = __nfs4_find_lock_state(state, owner, NULL);
915 list_add(&new->ls_locks, &state->lock_states);
916 set_bit(LK_STATE_IN_USE, &state->flags);
921 spin_unlock(&state->state_lock);
922 new = nfs4_alloc_lock_state(state, owner);
926 spin_unlock(&state->state_lock);
928 nfs4_free_lock_state(state->owner->so_server, new);
933 * Release reference to lock_state, and free it if we see that
934 * it is no longer in use
936 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
938 struct nfs_server *server;
939 struct nfs4_state *state;
943 state = lsp->ls_state;
944 if (!refcount_dec_and_lock(&lsp->ls_count, &state->state_lock))
946 list_del(&lsp->ls_locks);
947 if (list_empty(&state->lock_states))
948 clear_bit(LK_STATE_IN_USE, &state->flags);
949 spin_unlock(&state->state_lock);
950 server = state->owner->so_server;
951 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
952 struct nfs_client *clp = server->nfs_client;
954 clp->cl_mvops->free_lock_state(server, lsp);
956 nfs4_free_lock_state(server, lsp);
959 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
961 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
963 dst->fl_u.nfs4_fl.owner = lsp;
964 refcount_inc(&lsp->ls_count);
967 static void nfs4_fl_release_lock(struct file_lock *fl)
969 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
972 static const struct file_lock_operations nfs4_fl_lock_ops = {
973 .fl_copy_lock = nfs4_fl_copy_lock,
974 .fl_release_private = nfs4_fl_release_lock,
977 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
979 struct nfs4_lock_state *lsp;
981 if (fl->fl_ops != NULL)
983 lsp = nfs4_get_lock_state(state, fl->fl_owner);
986 fl->fl_u.nfs4_fl.owner = lsp;
987 fl->fl_ops = &nfs4_fl_lock_ops;
991 static int nfs4_copy_lock_stateid(nfs4_stateid *dst,
992 struct nfs4_state *state,
993 const struct nfs_lock_context *l_ctx)
995 struct nfs4_lock_state *lsp;
996 fl_owner_t fl_owner, fl_flock_owner;
1002 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
1005 fl_owner = l_ctx->lockowner;
1006 fl_flock_owner = l_ctx->open_context->flock_owner;
1008 spin_lock(&state->state_lock);
1009 lsp = __nfs4_find_lock_state(state, fl_owner, fl_flock_owner);
1010 if (lsp && test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
1012 else if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
1013 nfs4_stateid_copy(dst, &lsp->ls_stateid);
1016 spin_unlock(&state->state_lock);
1017 nfs4_put_lock_state(lsp);
1022 bool nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
1025 const nfs4_stateid *src;
1030 src = &zero_stateid;
1031 seq = read_seqbegin(&state->seqlock);
1032 if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1033 src = &state->open_stateid;
1036 nfs4_stateid_copy(dst, src);
1037 } while (read_seqretry(&state->seqlock, seq));
1042 * Byte-range lock aware utility to initialize the stateid of read/write
1045 int nfs4_select_rw_stateid(struct nfs4_state *state,
1046 fmode_t fmode, const struct nfs_lock_context *l_ctx,
1047 nfs4_stateid *dst, const struct cred **cred)
1051 if (!nfs4_valid_open_stateid(state))
1055 ret = nfs4_copy_lock_stateid(dst, state, l_ctx);
1057 /* A lost lock - don't even consider delegations */
1059 /* returns true if delegation stateid found and copied */
1060 if (nfs4_copy_delegation_stateid(state->inode, fmode, dst, cred)) {
1065 /* nfs4_copy_delegation_stateid() didn't over-write
1066 * dst, so it still has the lock stateid which we now
1070 ret = nfs4_copy_open_stateid(dst, state) ? 0 : -EAGAIN;
1072 if (nfs_server_capable(state->inode, NFS_CAP_STATEID_NFSV41))
1077 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
1079 struct nfs_seqid *new;
1081 new = kmalloc(sizeof(*new), gfp_mask);
1083 return ERR_PTR(-ENOMEM);
1084 new->sequence = counter;
1085 INIT_LIST_HEAD(&new->list);
1090 void nfs_release_seqid(struct nfs_seqid *seqid)
1092 struct nfs_seqid_counter *sequence;
1094 if (seqid == NULL || list_empty(&seqid->list))
1096 sequence = seqid->sequence;
1097 spin_lock(&sequence->lock);
1098 list_del_init(&seqid->list);
1099 if (!list_empty(&sequence->list)) {
1100 struct nfs_seqid *next;
1102 next = list_first_entry(&sequence->list,
1103 struct nfs_seqid, list);
1104 rpc_wake_up_queued_task(&sequence->wait, next->task);
1106 spin_unlock(&sequence->lock);
1109 void nfs_free_seqid(struct nfs_seqid *seqid)
1111 nfs_release_seqid(seqid);
1116 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
1117 * failed with a seqid incrementing error -
1118 * see comments nfs4.h:seqid_mutating_error()
1120 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
1125 case -NFS4ERR_BAD_SEQID:
1126 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
1128 pr_warn_ratelimited("NFS: v4 server returned a bad"
1129 " sequence-id error on an"
1130 " unconfirmed sequence %p!\n",
1133 case -NFS4ERR_STALE_CLIENTID:
1134 case -NFS4ERR_STALE_STATEID:
1135 case -NFS4ERR_BAD_STATEID:
1136 case -NFS4ERR_BADXDR:
1137 case -NFS4ERR_RESOURCE:
1138 case -NFS4ERR_NOFILEHANDLE:
1139 case -NFS4ERR_MOVED:
1140 /* Non-seqid mutating errors */
1144 * Note: no locking needed as we are guaranteed to be first
1145 * on the sequence list
1147 seqid->sequence->counter++;
1150 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
1152 struct nfs4_state_owner *sp;
1157 sp = container_of(seqid->sequence, struct nfs4_state_owner, so_seqid);
1158 if (status == -NFS4ERR_BAD_SEQID)
1159 nfs4_reset_state_owner(sp);
1160 if (!nfs4_has_session(sp->so_server->nfs_client))
1161 nfs_increment_seqid(status, seqid);
1165 * Increment the seqid if the LOCK/LOCKU succeeded, or
1166 * failed with a seqid incrementing error -
1167 * see comments nfs4.h:seqid_mutating_error()
1169 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
1172 nfs_increment_seqid(status, seqid);
1175 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
1177 struct nfs_seqid_counter *sequence;
1182 sequence = seqid->sequence;
1183 spin_lock(&sequence->lock);
1185 if (list_empty(&seqid->list))
1186 list_add_tail(&seqid->list, &sequence->list);
1187 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
1189 rpc_sleep_on(&sequence->wait, task, NULL);
1192 spin_unlock(&sequence->lock);
1197 static int nfs4_run_state_manager(void *);
1199 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
1201 clear_and_wake_up_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
1202 rpc_wake_up(&clp->cl_rpcwaitq);
1206 * Schedule the nfs_client asynchronous state management routine
1208 void nfs4_schedule_state_manager(struct nfs_client *clp)
1210 struct task_struct *task;
1211 char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
1212 struct rpc_clnt *clnt = clp->cl_rpcclient;
1213 bool swapon = false;
1215 set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
1217 if (atomic_read(&clnt->cl_swapper)) {
1218 swapon = !test_and_set_bit(NFS4CLNT_MANAGER_AVAILABLE,
1221 wake_up_var(&clp->cl_state);
1226 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1229 __module_get(THIS_MODULE);
1230 refcount_inc(&clp->cl_count);
1232 /* The rcu_read_lock() is not strictly necessary, as the state
1233 * manager is the only thread that ever changes the rpc_xprt
1234 * after it's initialized. At this point, we're single threaded. */
1236 snprintf(buf, sizeof(buf), "%s-manager",
1237 rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR));
1239 task = kthread_run(nfs4_run_state_manager, clp, "%s", buf);
1241 printk(KERN_ERR "%s: kthread_run: %ld\n",
1242 __func__, PTR_ERR(task));
1243 if (!nfs_client_init_is_complete(clp))
1244 nfs_mark_client_ready(clp, PTR_ERR(task));
1246 clear_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state);
1247 nfs4_clear_state_manager_bit(clp);
1248 nfs_put_client(clp);
1249 module_put(THIS_MODULE);
1254 * Schedule a lease recovery attempt
1256 void nfs4_schedule_lease_recovery(struct nfs_client *clp)
1260 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1261 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1262 dprintk("%s: scheduling lease recovery for server %s\n", __func__,
1264 nfs4_schedule_state_manager(clp);
1266 EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
1269 * nfs4_schedule_migration_recovery - trigger migration recovery
1271 * @server: FSID that is migrating
1273 * Returns zero if recovery has started, otherwise a negative NFS4ERR
1274 * value is returned.
1276 int nfs4_schedule_migration_recovery(const struct nfs_server *server)
1278 struct nfs_client *clp = server->nfs_client;
1280 if (server->fh_expire_type != NFS4_FH_PERSISTENT) {
1281 pr_err("NFS: volatile file handles not supported (server %s)\n",
1286 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
1289 dprintk("%s: scheduling migration recovery for (%llx:%llx) on %s\n",
1291 (unsigned long long)server->fsid.major,
1292 (unsigned long long)server->fsid.minor,
1295 set_bit(NFS_MIG_IN_TRANSITION,
1296 &((struct nfs_server *)server)->mig_status);
1297 set_bit(NFS4CLNT_MOVED, &clp->cl_state);
1299 nfs4_schedule_state_manager(clp);
1302 EXPORT_SYMBOL_GPL(nfs4_schedule_migration_recovery);
1305 * nfs4_schedule_lease_moved_recovery - start lease-moved recovery
1307 * @clp: server to check for moved leases
1310 void nfs4_schedule_lease_moved_recovery(struct nfs_client *clp)
1312 dprintk("%s: scheduling lease-moved recovery for client ID %llx on %s\n",
1313 __func__, clp->cl_clientid, clp->cl_hostname);
1315 set_bit(NFS4CLNT_LEASE_MOVED, &clp->cl_state);
1316 nfs4_schedule_state_manager(clp);
1318 EXPORT_SYMBOL_GPL(nfs4_schedule_lease_moved_recovery);
1320 int nfs4_wait_clnt_recover(struct nfs_client *clp)
1326 refcount_inc(&clp->cl_count);
1327 res = wait_on_bit_action(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
1328 nfs_wait_bit_killable,
1329 TASK_KILLABLE|TASK_FREEZABLE_UNSAFE);
1332 if (clp->cl_cons_state < 0)
1333 res = clp->cl_cons_state;
1335 nfs_put_client(clp);
1339 int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1344 for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1345 ret = nfs4_wait_clnt_recover(clp);
1348 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1349 !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1351 nfs4_schedule_state_manager(clp);
1358 * nfs40_handle_cb_pathdown - return all delegations after NFS4ERR_CB_PATH_DOWN
1359 * @clp: client to process
1361 * Set the NFS4CLNT_LEASE_EXPIRED state in order to force a
1362 * resend of the SETCLIENTID and hence re-establish the
1363 * callback channel. Then return all existing delegations.
1365 static void nfs40_handle_cb_pathdown(struct nfs_client *clp)
1367 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1368 nfs_expire_all_delegations(clp);
1369 dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
1373 void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
1375 nfs40_handle_cb_pathdown(clp);
1376 nfs4_schedule_state_manager(clp);
1379 static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1382 if (!nfs4_valid_open_stateid(state))
1384 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1385 /* Don't recover state that expired before the reboot */
1386 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
1387 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1390 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1391 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1395 int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1397 if (!nfs4_valid_open_stateid(state))
1399 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
1400 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1401 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1402 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
1406 int nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
1408 struct nfs_client *clp = server->nfs_client;
1410 if (!nfs4_state_mark_reclaim_nograce(clp, state))
1412 nfs_inode_find_delegation_state_and_recover(state->inode,
1414 dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
1416 nfs4_schedule_state_manager(clp);
1419 EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1421 static struct nfs4_lock_state *
1422 nfs_state_find_lock_state_by_stateid(struct nfs4_state *state,
1423 const nfs4_stateid *stateid)
1425 struct nfs4_lock_state *pos;
1427 list_for_each_entry(pos, &state->lock_states, ls_locks) {
1428 if (!test_bit(NFS_LOCK_INITIALIZED, &pos->ls_flags))
1430 if (nfs4_stateid_match_or_older(&pos->ls_stateid, stateid))
1436 static bool nfs_state_lock_state_matches_stateid(struct nfs4_state *state,
1437 const nfs4_stateid *stateid)
1441 if (test_bit(LK_STATE_IN_USE, &state->flags)) {
1442 spin_lock(&state->state_lock);
1443 if (nfs_state_find_lock_state_by_stateid(state, stateid))
1445 spin_unlock(&state->state_lock);
1450 void nfs_inode_find_state_and_recover(struct inode *inode,
1451 const nfs4_stateid *stateid)
1453 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1454 struct nfs_inode *nfsi = NFS_I(inode);
1455 struct nfs_open_context *ctx;
1456 struct nfs4_state *state;
1460 list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
1464 if (nfs4_stateid_match_or_older(&state->stateid, stateid) &&
1465 nfs4_state_mark_reclaim_nograce(clp, state)) {
1469 if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1470 nfs4_stateid_match_or_older(&state->open_stateid, stateid) &&
1471 nfs4_state_mark_reclaim_nograce(clp, state)) {
1475 if (nfs_state_lock_state_matches_stateid(state, stateid) &&
1476 nfs4_state_mark_reclaim_nograce(clp, state))
1481 nfs_inode_find_delegation_state_and_recover(inode, stateid);
1483 nfs4_schedule_state_manager(clp);
1486 static void nfs4_state_mark_open_context_bad(struct nfs4_state *state, int err)
1488 struct inode *inode = state->inode;
1489 struct nfs_inode *nfsi = NFS_I(inode);
1490 struct nfs_open_context *ctx;
1493 list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
1494 if (ctx->state != state)
1496 set_bit(NFS_CONTEXT_BAD, &ctx->flags);
1497 pr_warn("NFSv4: state recovery failed for open file %pd2, "
1498 "error = %d\n", ctx->dentry, err);
1503 static void nfs4_state_mark_recovery_failed(struct nfs4_state *state, int error)
1505 set_bit(NFS_STATE_RECOVERY_FAILED, &state->flags);
1506 nfs4_state_mark_open_context_bad(state, error);
1510 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
1512 struct inode *inode = state->inode;
1513 struct nfs_inode *nfsi = NFS_I(inode);
1514 struct file_lock *fl;
1515 struct nfs4_lock_state *lsp;
1517 struct file_lock_context *flctx = inode->i_flctx;
1518 struct list_head *list;
1523 list = &flctx->flc_posix;
1525 /* Guard against delegation returns and new lock/unlock calls */
1526 down_write(&nfsi->rwsem);
1527 spin_lock(&flctx->flc_lock);
1529 list_for_each_entry(fl, list, fl_list) {
1530 if (nfs_file_open_context(fl->fl_file)->state != state)
1532 spin_unlock(&flctx->flc_lock);
1533 status = ops->recover_lock(state, fl);
1539 case -NFS4ERR_ADMIN_REVOKED:
1540 case -NFS4ERR_STALE_STATEID:
1541 case -NFS4ERR_BAD_STATEID:
1542 case -NFS4ERR_EXPIRED:
1543 case -NFS4ERR_NO_GRACE:
1544 case -NFS4ERR_STALE_CLIENTID:
1545 case -NFS4ERR_BADSESSION:
1546 case -NFS4ERR_BADSLOT:
1547 case -NFS4ERR_BAD_HIGH_SLOT:
1548 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1551 pr_err("NFS: %s: unhandled error %d\n",
1555 case -NFS4ERR_DENIED:
1556 case -NFS4ERR_RECLAIM_BAD:
1557 case -NFS4ERR_RECLAIM_CONFLICT:
1558 lsp = fl->fl_u.nfs4_fl.owner;
1560 set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
1563 spin_lock(&flctx->flc_lock);
1565 if (list == &flctx->flc_posix) {
1566 list = &flctx->flc_flock;
1569 spin_unlock(&flctx->flc_lock);
1571 up_write(&nfsi->rwsem);
1575 #ifdef CONFIG_NFS_V4_2
1576 static void nfs42_complete_copies(struct nfs4_state_owner *sp, struct nfs4_state *state)
1578 struct nfs4_copy_state *copy;
1580 if (!test_bit(NFS_CLNT_DST_SSC_COPY_STATE, &state->flags) &&
1581 !test_bit(NFS_CLNT_SRC_SSC_COPY_STATE, &state->flags))
1584 spin_lock(&sp->so_server->nfs_client->cl_lock);
1585 list_for_each_entry(copy, &sp->so_server->ss_copies, copies) {
1586 if ((test_bit(NFS_CLNT_DST_SSC_COPY_STATE, &state->flags) &&
1587 !nfs4_stateid_match_other(&state->stateid,
1588 ©->parent_dst_state->stateid)))
1591 if (test_and_clear_bit(NFS_CLNT_DST_SSC_COPY_STATE,
1593 clear_bit(NFS_CLNT_SRC_SSC_COPY_STATE, &state->flags);
1594 complete(©->completion);
1597 list_for_each_entry(copy, &sp->so_server->ss_copies, src_copies) {
1598 if ((test_bit(NFS_CLNT_SRC_SSC_COPY_STATE, &state->flags) &&
1599 !nfs4_stateid_match_other(&state->stateid,
1600 ©->parent_src_state->stateid)))
1603 if (test_and_clear_bit(NFS_CLNT_DST_SSC_COPY_STATE,
1605 complete(©->completion);
1607 spin_unlock(&sp->so_server->nfs_client->cl_lock);
1609 #else /* !CONFIG_NFS_V4_2 */
1610 static inline void nfs42_complete_copies(struct nfs4_state_owner *sp,
1611 struct nfs4_state *state)
1614 #endif /* CONFIG_NFS_V4_2 */
1616 static int __nfs4_reclaim_open_state(struct nfs4_state_owner *sp, struct nfs4_state *state,
1617 const struct nfs4_state_recovery_ops *ops,
1620 struct nfs4_lock_state *lock;
1623 status = ops->recover_open(sp, state);
1627 status = nfs4_reclaim_locks(state, ops);
1631 if (!test_bit(NFS_DELEGATED_STATE, &state->flags)) {
1632 spin_lock(&state->state_lock);
1633 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1634 trace_nfs4_state_lock_reclaim(state, lock);
1635 if (!test_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags))
1638 spin_unlock(&state->state_lock);
1641 nfs42_complete_copies(sp, state);
1642 clear_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
1646 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp,
1647 const struct nfs4_state_recovery_ops *ops,
1650 struct nfs4_state *state;
1651 unsigned int loop = 0;
1653 #ifdef CONFIG_NFS_V4_2
1654 bool found_ssc_copy_state = false;
1655 #endif /* CONFIG_NFS_V4_2 */
1657 /* Note: we rely on the sp->so_states list being ordered
1658 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
1660 * This is needed to ensure that the server won't give us any
1661 * read delegations that we have to return if, say, we are
1662 * recovering after a network partition or a reboot from a
1663 * server that doesn't support a grace period.
1665 spin_lock(&sp->so_lock);
1666 raw_write_seqcount_begin(&sp->so_reclaim_seqcount);
1668 list_for_each_entry(state, &sp->so_states, open_states) {
1669 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
1671 if (!nfs4_valid_open_stateid(state))
1673 if (state->state == 0)
1675 #ifdef CONFIG_NFS_V4_2
1676 if (test_bit(NFS_SRV_SSC_COPY_STATE, &state->flags)) {
1677 nfs4_state_mark_recovery_failed(state, -EIO);
1678 found_ssc_copy_state = true;
1681 #endif /* CONFIG_NFS_V4_2 */
1682 refcount_inc(&state->count);
1683 spin_unlock(&sp->so_lock);
1684 status = __nfs4_reclaim_open_state(sp, state, ops, lost_locks);
1692 printk(KERN_ERR "NFS: %s: unhandled error %d\n", __func__, status);
1700 /* Open state on this file cannot be recovered */
1701 nfs4_state_mark_recovery_failed(state, status);
1706 set_bit(ops->state_flag_bit, &state->flags);
1710 case -NFS4ERR_ADMIN_REVOKED:
1711 case -NFS4ERR_STALE_STATEID:
1712 case -NFS4ERR_OLD_STATEID:
1713 case -NFS4ERR_BAD_STATEID:
1714 case -NFS4ERR_RECLAIM_BAD:
1715 case -NFS4ERR_RECLAIM_CONFLICT:
1716 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1718 case -NFS4ERR_EXPIRED:
1719 case -NFS4ERR_NO_GRACE:
1720 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1722 case -NFS4ERR_STALE_CLIENTID:
1723 case -NFS4ERR_BADSESSION:
1724 case -NFS4ERR_BADSLOT:
1725 case -NFS4ERR_BAD_HIGH_SLOT:
1726 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1730 nfs4_put_open_state(state);
1731 spin_lock(&sp->so_lock);
1734 raw_write_seqcount_end(&sp->so_reclaim_seqcount);
1735 spin_unlock(&sp->so_lock);
1736 #ifdef CONFIG_NFS_V4_2
1737 if (found_ssc_copy_state)
1739 #endif /* CONFIG_NFS_V4_2 */
1742 nfs4_put_open_state(state);
1743 spin_lock(&sp->so_lock);
1744 raw_write_seqcount_end(&sp->so_reclaim_seqcount);
1745 spin_unlock(&sp->so_lock);
1749 static void nfs4_clear_open_state(struct nfs4_state *state)
1751 struct nfs4_lock_state *lock;
1753 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1754 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1755 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1756 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1757 spin_lock(&state->state_lock);
1758 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1759 lock->ls_seqid.flags = 0;
1760 clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
1762 spin_unlock(&state->state_lock);
1765 static void nfs4_reset_seqids(struct nfs_server *server,
1766 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1768 struct nfs_client *clp = server->nfs_client;
1769 struct nfs4_state_owner *sp;
1770 struct rb_node *pos;
1771 struct nfs4_state *state;
1773 spin_lock(&clp->cl_lock);
1774 for (pos = rb_first(&server->state_owners);
1776 pos = rb_next(pos)) {
1777 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1778 sp->so_seqid.flags = 0;
1779 spin_lock(&sp->so_lock);
1780 list_for_each_entry(state, &sp->so_states, open_states) {
1781 if (mark_reclaim(clp, state))
1782 nfs4_clear_open_state(state);
1784 spin_unlock(&sp->so_lock);
1786 spin_unlock(&clp->cl_lock);
1789 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp,
1790 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1792 struct nfs_server *server;
1795 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1796 nfs4_reset_seqids(server, mark_reclaim);
1800 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1802 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1803 /* Mark all delegations for reclaim */
1804 nfs_delegation_mark_reclaim(clp);
1805 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1808 static int nfs4_reclaim_complete(struct nfs_client *clp,
1809 const struct nfs4_state_recovery_ops *ops,
1810 const struct cred *cred)
1812 /* Notify the server we're done reclaiming our state */
1813 if (ops->reclaim_complete)
1814 return ops->reclaim_complete(clp, cred);
1818 static void nfs4_clear_reclaim_server(struct nfs_server *server)
1820 struct nfs_client *clp = server->nfs_client;
1821 struct nfs4_state_owner *sp;
1822 struct rb_node *pos;
1823 struct nfs4_state *state;
1825 spin_lock(&clp->cl_lock);
1826 for (pos = rb_first(&server->state_owners);
1828 pos = rb_next(pos)) {
1829 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1830 spin_lock(&sp->so_lock);
1831 list_for_each_entry(state, &sp->so_states, open_states) {
1832 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
1835 nfs4_state_mark_reclaim_nograce(clp, state);
1837 spin_unlock(&sp->so_lock);
1839 spin_unlock(&clp->cl_lock);
1842 static int nfs4_state_clear_reclaim_reboot(struct nfs_client *clp)
1844 struct nfs_server *server;
1846 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1850 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1851 nfs4_clear_reclaim_server(server);
1854 nfs_delegation_reap_unclaimed(clp);
1858 static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
1860 const struct nfs4_state_recovery_ops *ops;
1861 const struct cred *cred;
1864 if (!nfs4_state_clear_reclaim_reboot(clp))
1866 ops = clp->cl_mvops->reboot_recovery_ops;
1867 cred = nfs4_get_clid_cred(clp);
1868 err = nfs4_reclaim_complete(clp, ops, cred);
1870 if (err == -NFS4ERR_CONN_NOT_BOUND_TO_SESSION)
1871 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1874 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1876 nfs_mark_test_expired_all_delegations(clp);
1877 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1880 static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1885 case -NFS4ERR_CB_PATH_DOWN:
1886 nfs40_handle_cb_pathdown(clp);
1888 case -NFS4ERR_NO_GRACE:
1889 nfs4_state_end_reclaim_reboot(clp);
1891 case -NFS4ERR_STALE_CLIENTID:
1892 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1893 nfs4_state_start_reclaim_reboot(clp);
1895 case -NFS4ERR_EXPIRED:
1896 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1897 nfs4_state_start_reclaim_nograce(clp);
1899 case -NFS4ERR_BADSESSION:
1900 case -NFS4ERR_BADSLOT:
1901 case -NFS4ERR_BAD_HIGH_SLOT:
1902 case -NFS4ERR_DEADSESSION:
1903 case -NFS4ERR_SEQ_FALSE_RETRY:
1904 case -NFS4ERR_SEQ_MISORDERED:
1905 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1906 /* Zero session reset errors */
1908 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1909 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1912 dprintk("%s: failed to handle error %d for server %s\n",
1913 __func__, error, clp->cl_hostname);
1916 dprintk("%s: handled error %d for server %s\n", __func__, error,
1921 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1923 struct nfs4_state_owner *sp;
1924 struct nfs_server *server;
1925 struct rb_node *pos;
1932 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1933 nfs4_purge_state_owners(server, &freeme);
1934 spin_lock(&clp->cl_lock);
1935 for (pos = rb_first(&server->state_owners);
1937 pos = rb_next(pos)) {
1939 struct nfs4_state_owner, so_server_node);
1940 if (!test_and_clear_bit(ops->owner_flag_bit,
1943 if (!atomic_inc_not_zero(&sp->so_count))
1945 spin_unlock(&clp->cl_lock);
1948 status = nfs4_reclaim_open_state(sp, ops, &lost_locks);
1951 pr_warn("NFS: %s: lost %d locks\n",
1952 clp->cl_hostname, lost_locks);
1953 set_bit(ops->owner_flag_bit, &sp->so_flags);
1954 nfs4_put_state_owner(sp);
1955 status = nfs4_recovery_handle_error(clp, status);
1956 return (status != 0) ? status : -EAGAIN;
1959 nfs4_put_state_owner(sp);
1962 spin_unlock(&clp->cl_lock);
1965 nfs4_free_state_owners(&freeme);
1967 pr_warn("NFS: %s: lost %d locks\n",
1968 clp->cl_hostname, lost_locks);
1972 static int nfs4_check_lease(struct nfs_client *clp)
1974 const struct cred *cred;
1975 const struct nfs4_state_maintenance_ops *ops =
1976 clp->cl_mvops->state_renewal_ops;
1979 /* Is the client already known to have an expired lease? */
1980 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1982 cred = ops->get_state_renewal_cred(clp);
1984 cred = nfs4_get_clid_cred(clp);
1989 status = ops->renew_lease(clp, cred);
1991 if (status == -ETIMEDOUT) {
1992 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1996 return nfs4_recovery_handle_error(clp, status);
1999 /* Set NFS4CLNT_LEASE_EXPIRED and reclaim reboot state for all v4.0 errors
2000 * and for recoverable errors on EXCHANGE_ID for v4.1
2002 static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
2005 case -NFS4ERR_SEQ_MISORDERED:
2006 if (test_and_set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state))
2007 return -ESERVERFAULT;
2008 /* Lease confirmation error: retry after purging the lease */
2010 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
2012 case -NFS4ERR_STALE_CLIENTID:
2013 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
2014 nfs4_state_start_reclaim_reboot(clp);
2016 case -NFS4ERR_CLID_INUSE:
2017 pr_err("NFS: Server %s reports our clientid is in use\n",
2019 nfs_mark_client_ready(clp, -EPERM);
2020 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
2023 case -NFS4ERR_DELAY:
2028 case -NFS4ERR_MINOR_VERS_MISMATCH:
2029 if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
2030 nfs_mark_client_ready(clp, -EPROTONOSUPPORT);
2031 dprintk("%s: exit with error %d for server %s\n",
2032 __func__, -EPROTONOSUPPORT, clp->cl_hostname);
2033 return -EPROTONOSUPPORT;
2035 if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
2036 nfs_mark_client_ready(clp, -EIO);
2038 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
2039 * in nfs4_exchange_id */
2041 dprintk("%s: exit with error %d for server %s\n", __func__,
2042 status, clp->cl_hostname);
2045 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
2046 dprintk("%s: handled error %d for server %s\n", __func__, status,
2051 static int nfs4_establish_lease(struct nfs_client *clp)
2053 const struct cred *cred;
2054 const struct nfs4_state_recovery_ops *ops =
2055 clp->cl_mvops->reboot_recovery_ops;
2058 status = nfs4_begin_drain_session(clp);
2061 cred = nfs4_get_clid_cred(clp);
2064 status = ops->establish_clid(clp, cred);
2068 pnfs_destroy_all_layouts(clp);
2073 * Returns zero or a negative errno. NFS4ERR values are converted
2074 * to local errno values.
2076 static int nfs4_reclaim_lease(struct nfs_client *clp)
2080 status = nfs4_establish_lease(clp);
2082 return nfs4_handle_reclaim_lease_error(clp, status);
2083 if (test_and_clear_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state))
2084 nfs4_state_start_reclaim_nograce(clp);
2085 if (!test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
2086 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
2087 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
2088 clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
2092 static int nfs4_purge_lease(struct nfs_client *clp)
2096 status = nfs4_establish_lease(clp);
2098 return nfs4_handle_reclaim_lease_error(clp, status);
2099 clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
2100 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
2101 nfs4_state_start_reclaim_nograce(clp);
2106 * Try remote migration of one FSID from a source server to a
2107 * destination server. The source server provides a list of
2108 * potential destinations.
2110 * Returns zero or a negative NFS4ERR status code.
2112 static int nfs4_try_migration(struct nfs_server *server, const struct cred *cred)
2114 struct nfs_client *clp = server->nfs_client;
2115 struct nfs4_fs_locations *locations = NULL;
2116 struct inode *inode;
2120 dprintk("--> %s: FSID %llx:%llx on \"%s\"\n", __func__,
2121 (unsigned long long)server->fsid.major,
2122 (unsigned long long)server->fsid.minor,
2126 page = alloc_page(GFP_KERNEL);
2127 locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2128 if (page == NULL || locations == NULL) {
2129 dprintk("<-- %s: no memory\n", __func__);
2132 locations->fattr = nfs_alloc_fattr();
2133 if (locations->fattr == NULL) {
2134 dprintk("<-- %s: no memory\n", __func__);
2138 inode = d_inode(server->super->s_root);
2139 result = nfs4_proc_get_locations(server, NFS_FH(inode), locations,
2142 dprintk("<-- %s: failed to retrieve fs_locations: %d\n",
2147 result = -NFS4ERR_NXIO;
2148 if (!locations->nlocations)
2151 if (!(locations->fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)) {
2152 dprintk("<-- %s: No fs_locations data, migration skipped\n",
2157 status = nfs4_begin_drain_session(clp);
2163 status = nfs4_replace_transport(server, locations);
2165 dprintk("<-- %s: failed to replace transport: %d\n",
2171 dprintk("<-- %s: migration succeeded\n", __func__);
2176 if (locations != NULL)
2177 kfree(locations->fattr);
2180 pr_err("NFS: migration recovery failed (server %s)\n",
2182 set_bit(NFS_MIG_FAILED, &server->mig_status);
2188 * Returns zero or a negative NFS4ERR status code.
2190 static int nfs4_handle_migration(struct nfs_client *clp)
2192 const struct nfs4_state_maintenance_ops *ops =
2193 clp->cl_mvops->state_renewal_ops;
2194 struct nfs_server *server;
2195 const struct cred *cred;
2197 dprintk("%s: migration reported on \"%s\"\n", __func__,
2200 cred = ops->get_state_renewal_cred(clp);
2202 return -NFS4ERR_NOENT;
2207 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
2210 if (server->mig_gen == clp->cl_mig_gen)
2212 server->mig_gen = clp->cl_mig_gen;
2214 if (!test_and_clear_bit(NFS_MIG_IN_TRANSITION,
2215 &server->mig_status))
2219 status = nfs4_try_migration(server, cred);
2232 * Test each nfs_server on the clp's cl_superblocks list to see
2233 * if it's moved to another server. Stop when the server no longer
2234 * returns NFS4ERR_LEASE_MOVED.
2236 static int nfs4_handle_lease_moved(struct nfs_client *clp)
2238 const struct nfs4_state_maintenance_ops *ops =
2239 clp->cl_mvops->state_renewal_ops;
2240 struct nfs_server *server;
2241 const struct cred *cred;
2243 dprintk("%s: lease moved reported on \"%s\"\n", __func__,
2246 cred = ops->get_state_renewal_cred(clp);
2248 return -NFS4ERR_NOENT;
2253 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
2254 struct inode *inode;
2257 if (server->mig_gen == clp->cl_mig_gen)
2259 server->mig_gen = clp->cl_mig_gen;
2263 inode = d_inode(server->super->s_root);
2264 status = nfs4_proc_fsid_present(inode, cred);
2265 if (status != -NFS4ERR_MOVED)
2266 goto restart; /* wasn't this one */
2267 if (nfs4_try_migration(server, cred) == -NFS4ERR_LEASE_MOVED)
2268 goto restart; /* there are more */
2279 * nfs4_discover_server_trunking - Detect server IP address trunking
2281 * @clp: nfs_client under test
2282 * @result: OUT: found nfs_client, or clp
2284 * Returns zero or a negative errno. If zero is returned,
2285 * an nfs_client pointer is planted in "result".
2287 * Note: since we are invoked in process context, and
2288 * not from inside the state manager, we cannot use
2289 * nfs4_handle_reclaim_lease_error().
2291 int nfs4_discover_server_trunking(struct nfs_client *clp,
2292 struct nfs_client **result)
2294 const struct nfs4_state_recovery_ops *ops =
2295 clp->cl_mvops->reboot_recovery_ops;
2296 struct rpc_clnt *clnt;
2297 const struct cred *cred;
2300 dprintk("NFS: %s: testing '%s'\n", __func__, clp->cl_hostname);
2302 clnt = clp->cl_rpcclient;
2305 mutex_lock(&nfs_clid_init_mutex);
2308 cred = nfs4_get_clid_cred(clp);
2312 status = ops->detect_trunking(clp, result, cred);
2320 if (clnt->cl_softrtry)
2323 case -NFS4ERR_DELAY:
2327 case -NFS4ERR_STALE_CLIENTID:
2328 dprintk("NFS: %s after status %d, retrying\n",
2333 nfs4_root_machine_cred(clp);
2336 if (clnt->cl_auth->au_flavor == RPC_AUTH_UNIX)
2339 case -NFS4ERR_CLID_INUSE:
2340 case -NFS4ERR_WRONGSEC:
2341 /* No point in retrying if we already used RPC_AUTH_UNIX */
2342 if (clnt->cl_auth->au_flavor == RPC_AUTH_UNIX) {
2346 clnt = rpc_clone_client_set_auth(clnt, RPC_AUTH_UNIX);
2348 status = PTR_ERR(clnt);
2351 /* Note: this is safe because we haven't yet marked the
2352 * client as ready, so we are the only user of
2355 clnt = xchg(&clp->cl_rpcclient, clnt);
2356 rpc_shutdown_client(clnt);
2357 clnt = clp->cl_rpcclient;
2360 case -NFS4ERR_MINOR_VERS_MISMATCH:
2361 status = -EPROTONOSUPPORT;
2365 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
2366 * in nfs4_exchange_id */
2367 status = -EKEYEXPIRED;
2370 pr_warn("NFS: %s unhandled error %d. Exiting with error EIO\n",
2376 mutex_unlock(&nfs_clid_init_mutex);
2377 dprintk("NFS: %s: status = %d\n", __func__, status);
2381 #ifdef CONFIG_NFS_V4_1
2382 void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
2384 struct nfs_client *clp = session->clp;
2388 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
2390 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2391 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2393 nfs4_schedule_state_manager(clp);
2395 EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
2397 void nfs41_notify_server(struct nfs_client *clp)
2399 /* Use CHECK_LEASE to ping the server with a SEQUENCE */
2400 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
2401 nfs4_schedule_state_manager(clp);
2404 static void nfs4_reset_all_state(struct nfs_client *clp)
2406 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
2407 set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
2408 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
2409 nfs4_state_start_reclaim_nograce(clp);
2410 dprintk("%s: scheduling reset of all state for server %s!\n",
2411 __func__, clp->cl_hostname);
2412 nfs4_schedule_state_manager(clp);
2416 static void nfs41_handle_server_reboot(struct nfs_client *clp)
2418 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
2419 nfs4_state_start_reclaim_reboot(clp);
2420 dprintk("%s: server %s rebooted!\n", __func__,
2422 nfs4_schedule_state_manager(clp);
2426 static void nfs41_handle_all_state_revoked(struct nfs_client *clp)
2428 nfs4_reset_all_state(clp);
2429 dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
2432 static void nfs41_handle_some_state_revoked(struct nfs_client *clp)
2434 nfs4_state_start_reclaim_nograce(clp);
2435 nfs4_schedule_state_manager(clp);
2437 dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
2440 static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
2442 /* FIXME: For now, we destroy all layouts. */
2443 pnfs_destroy_all_layouts(clp);
2444 nfs_test_expired_all_delegations(clp);
2445 dprintk("%s: Recallable state revoked on server %s!\n", __func__,
2449 static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
2451 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
2452 nfs4_schedule_state_manager(clp);
2454 dprintk("%s: server %s declared a backchannel fault\n", __func__,
2458 static void nfs41_handle_cb_path_down(struct nfs_client *clp)
2460 if (test_and_set_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2461 &clp->cl_state) == 0)
2462 nfs4_schedule_state_manager(clp);
2465 void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags,
2471 dprintk("%s: \"%s\" (client ID %llx) flags=0x%08x\n",
2472 __func__, clp->cl_hostname, clp->cl_clientid, flags);
2474 * If we're called from the state manager thread, then assume we're
2475 * already handling the RECLAIM_NEEDED and/or STATE_REVOKED.
2476 * Those flags are expected to remain set until we're done
2477 * recovering (see RFC5661, section 18.46.3).
2482 if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
2483 nfs41_handle_server_reboot(clp);
2484 if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED))
2485 nfs41_handle_all_state_revoked(clp);
2486 if (flags & (SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
2487 SEQ4_STATUS_ADMIN_STATE_REVOKED))
2488 nfs41_handle_some_state_revoked(clp);
2489 if (flags & SEQ4_STATUS_LEASE_MOVED)
2490 nfs4_schedule_lease_moved_recovery(clp);
2491 if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
2492 nfs41_handle_recallable_state_revoked(clp);
2494 if (flags & SEQ4_STATUS_BACKCHANNEL_FAULT)
2495 nfs41_handle_backchannel_fault(clp);
2496 else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
2497 SEQ4_STATUS_CB_PATH_DOWN_SESSION))
2498 nfs41_handle_cb_path_down(clp);
2501 static int nfs4_reset_session(struct nfs_client *clp)
2503 const struct cred *cred;
2506 if (!nfs4_has_session(clp))
2508 status = nfs4_begin_drain_session(clp);
2511 cred = nfs4_get_clid_cred(clp);
2512 status = nfs4_proc_destroy_session(clp->cl_session, cred);
2515 case -NFS4ERR_BADSESSION:
2516 case -NFS4ERR_DEADSESSION:
2518 case -NFS4ERR_BACK_CHAN_BUSY:
2519 case -NFS4ERR_DELAY:
2520 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
2525 status = nfs4_recovery_handle_error(clp, status);
2529 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
2530 status = nfs4_proc_create_session(clp, cred);
2532 dprintk("%s: session reset failed with status %d for server %s!\n",
2533 __func__, status, clp->cl_hostname);
2534 status = nfs4_handle_reclaim_lease_error(clp, status);
2537 nfs41_finish_session_reset(clp);
2538 dprintk("%s: session reset was successful for server %s!\n",
2539 __func__, clp->cl_hostname);
2545 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2547 const struct cred *cred;
2550 if (!nfs4_has_session(clp))
2552 ret = nfs4_begin_drain_session(clp);
2555 cred = nfs4_get_clid_cred(clp);
2556 ret = nfs4_proc_bind_conn_to_session(clp, cred);
2558 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2561 dprintk("%s: bind_conn_to_session was successful for server %s!\n",
2562 __func__, clp->cl_hostname);
2564 case -NFS4ERR_DELAY:
2566 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2569 return nfs4_recovery_handle_error(clp, ret);
2574 static void nfs4_layoutreturn_any_run(struct nfs_client *clp)
2578 if (test_and_clear_bit(NFS4CLNT_RECALL_ANY_LAYOUT_READ, &clp->cl_state))
2579 iomode += IOMODE_READ;
2580 if (test_and_clear_bit(NFS4CLNT_RECALL_ANY_LAYOUT_RW, &clp->cl_state))
2581 iomode += IOMODE_RW;
2582 /* Note: IOMODE_READ + IOMODE_RW == IOMODE_ANY */
2584 pnfs_layout_return_unused_byclid(clp, iomode);
2585 set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
2588 #else /* CONFIG_NFS_V4_1 */
2589 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
2591 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2596 static void nfs4_layoutreturn_any_run(struct nfs_client *clp)
2599 #endif /* CONFIG_NFS_V4_1 */
2601 static void nfs4_state_manager(struct nfs_client *clp)
2603 unsigned int memflags;
2605 const char *section = "", *section_sep = "";
2608 * State recovery can deadlock if the direct reclaim code tries
2609 * start NFS writeback. So ensure memory allocations are all
2612 memflags = memalloc_nofs_save();
2614 /* Ensure exclusive access to NFSv4 state */
2616 trace_nfs4_state_mgr(clp);
2617 clear_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
2618 if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
2619 section = "purge state";
2620 status = nfs4_purge_lease(clp);
2626 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
2627 section = "lease expired";
2628 /* We're going to have to re-establish a clientid */
2629 status = nfs4_reclaim_lease(clp);
2635 /* Initialize or reset the session */
2636 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
2637 section = "reset session";
2638 status = nfs4_reset_session(clp);
2639 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
2645 /* Send BIND_CONN_TO_SESSION */
2646 if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2648 section = "bind conn to session";
2649 status = nfs4_bind_conn_to_session(clp);
2655 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
2656 section = "check lease";
2657 status = nfs4_check_lease(clp);
2663 if (test_and_clear_bit(NFS4CLNT_MOVED, &clp->cl_state)) {
2664 section = "migration";
2665 status = nfs4_handle_migration(clp);
2670 if (test_and_clear_bit(NFS4CLNT_LEASE_MOVED, &clp->cl_state)) {
2671 section = "lease moved";
2672 status = nfs4_handle_lease_moved(clp);
2677 /* First recover reboot state... */
2678 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
2679 section = "reclaim reboot";
2680 status = nfs4_do_reclaim(clp,
2681 clp->cl_mvops->reboot_recovery_ops);
2682 if (status == -EAGAIN)
2686 nfs4_state_end_reclaim_reboot(clp);
2690 /* Detect expired delegations... */
2691 if (test_and_clear_bit(NFS4CLNT_DELEGATION_EXPIRED, &clp->cl_state)) {
2692 section = "detect expired delegations";
2693 nfs_reap_expired_delegations(clp);
2697 /* Now recover expired state... */
2698 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
2699 section = "reclaim nograce";
2700 status = nfs4_do_reclaim(clp,
2701 clp->cl_mvops->nograce_recovery_ops);
2702 if (status == -EAGAIN)
2706 clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
2709 memalloc_nofs_restore(memflags);
2710 nfs4_end_drain_session(clp);
2711 nfs4_clear_state_manager_bit(clp);
2713 if (test_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state) &&
2714 !test_and_set_bit(NFS4CLNT_MANAGER_RUNNING,
2716 memflags = memalloc_nofs_save();
2720 if (!test_and_set_bit(NFS4CLNT_RECALL_RUNNING, &clp->cl_state)) {
2721 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
2722 nfs_client_return_marked_delegations(clp);
2723 set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
2725 nfs4_layoutreturn_any_run(clp);
2726 clear_bit(NFS4CLNT_RECALL_RUNNING, &clp->cl_state);
2731 } while (refcount_read(&clp->cl_count) > 1 && !signalled());
2735 if (strlen(section))
2737 trace_nfs4_state_mgr_failed(clp, section, status);
2738 pr_warn_ratelimited("NFS: state manager%s%s failed on NFSv4 server %s"
2739 " with error %d\n", section_sep, section,
2740 clp->cl_hostname, -status);
2743 memalloc_nofs_restore(memflags);
2744 nfs4_end_drain_session(clp);
2745 nfs4_clear_state_manager_bit(clp);
2748 static int nfs4_run_state_manager(void *ptr)
2750 struct nfs_client *clp = ptr;
2751 struct rpc_clnt *cl = clp->cl_rpcclient;
2753 while (cl != cl->cl_parent)
2756 allow_signal(SIGKILL);
2758 nfs4_state_manager(clp);
2760 if (test_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state) &&
2761 !test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state)) {
2762 wait_var_event_interruptible(&clp->cl_state,
2763 test_bit(NFS4CLNT_RUN_MANAGER,
2765 if (!atomic_read(&cl->cl_swapper))
2766 clear_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state);
2767 if (refcount_read(&clp->cl_count) > 1 && !signalled() &&
2768 !test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state))
2770 /* Either no longer a swapper, or were signalled */
2771 clear_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state);
2774 if (refcount_read(&clp->cl_count) > 1 && !signalled() &&
2775 test_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state) &&
2776 !test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state))
2779 nfs_put_client(clp);
2780 module_put_and_kthread_exit(0);