4 * Client-side XDR for NFSv4.
6 * Copyright (c) 2002 The Regents of the University of Michigan.
9 * Kendrick Smith <kmsmith@umich.edu>
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its
21 * contributors may be used to endorse or promote products derived
22 * from this software without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
25 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
26 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
27 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
31 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
32 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
33 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
34 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36 * Implementation of the NFSv4 state model. For the time being,
37 * this is minimal, but will be made much more complex in a
41 #include <linux/kernel.h>
42 #include <linux/slab.h>
44 #include <linux/nfs_fs.h>
45 #include <linux/nfs_idmap.h>
46 #include <linux/kthread.h>
47 #include <linux/module.h>
48 #include <linux/random.h>
49 #include <linux/ratelimit.h>
50 #include <linux/workqueue.h>
51 #include <linux/bitops.h>
52 #include <linux/jiffies.h>
54 #include <linux/sunrpc/clnt.h>
58 #include "delegation.h"
63 #define NFSDBG_FACILITY NFSDBG_STATE
65 #define OPENOWNER_POOL_SIZE 8
67 const nfs4_stateid zero_stateid;
68 static DEFINE_MUTEX(nfs_clid_init_mutex);
69 static LIST_HEAD(nfs4_clientid_list);
71 int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
73 struct nfs4_setclientid_res clid = {
74 .clientid = clp->cl_clientid,
75 .confirm = clp->cl_confirm,
79 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
81 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
83 port = nn->nfs_callback_tcpport;
84 if (clp->cl_addr.ss_family == AF_INET6)
85 port = nn->nfs_callback_tcpport6;
87 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
90 clp->cl_clientid = clid.clientid;
91 clp->cl_confirm = clid.confirm;
92 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
94 status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
97 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
98 nfs4_schedule_state_renewal(clp);
104 * nfs40_discover_server_trunking - Detect server IP address trunking (mv0)
106 * @clp: nfs_client under test
107 * @result: OUT: found nfs_client, or clp
108 * @cred: credential to use for trunking test
110 * Returns zero, a negative errno, or a negative NFS4ERR status.
111 * If zero is returned, an nfs_client pointer is planted in
114 * Note: The returned client may not yet be marked ready.
116 int nfs40_discover_server_trunking(struct nfs_client *clp,
117 struct nfs_client **result,
118 struct rpc_cred *cred)
120 struct nfs4_setclientid_res clid = {
121 .clientid = clp->cl_clientid,
122 .confirm = clp->cl_confirm,
124 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
128 port = nn->nfs_callback_tcpport;
129 if (clp->cl_addr.ss_family == AF_INET6)
130 port = nn->nfs_callback_tcpport6;
132 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
135 clp->cl_clientid = clid.clientid;
136 clp->cl_confirm = clid.confirm;
138 status = nfs40_walk_client_list(clp, result, cred);
140 case -NFS4ERR_STALE_CLIENTID:
141 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
143 /* Sustain the lease, even if it's empty. If the clientid4
144 * goes stale it's of no use for trunking discovery. */
145 nfs4_schedule_state_renewal(*result);
153 struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
155 struct rpc_cred *cred = NULL;
157 if (clp->cl_machine_cred != NULL)
158 cred = get_rpccred(clp->cl_machine_cred);
162 static void nfs4_clear_machine_cred(struct nfs_client *clp)
164 struct rpc_cred *cred;
166 spin_lock(&clp->cl_lock);
167 cred = clp->cl_machine_cred;
168 clp->cl_machine_cred = NULL;
169 spin_unlock(&clp->cl_lock);
174 static struct rpc_cred *
175 nfs4_get_renew_cred_server_locked(struct nfs_server *server)
177 struct rpc_cred *cred = NULL;
178 struct nfs4_state_owner *sp;
181 for (pos = rb_first(&server->state_owners);
183 pos = rb_next(pos)) {
184 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
185 if (list_empty(&sp->so_states))
187 cred = get_rpccred(sp->so_cred);
194 * nfs4_get_renew_cred_locked - Acquire credential for a renew operation
195 * @clp: client state handle
197 * Returns an rpc_cred with reference count bumped, or NULL.
198 * Caller must hold clp->cl_lock.
200 struct rpc_cred *nfs4_get_renew_cred_locked(struct nfs_client *clp)
202 struct rpc_cred *cred = NULL;
203 struct nfs_server *server;
205 /* Use machine credentials if available */
206 cred = nfs4_get_machine_cred_locked(clp);
211 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
212 cred = nfs4_get_renew_cred_server_locked(server);
222 #if defined(CONFIG_NFS_V4_1)
224 static int nfs41_setup_state_renewal(struct nfs_client *clp)
227 struct nfs_fsinfo fsinfo;
229 if (!test_bit(NFS_CS_CHECK_LEASE_TIME, &clp->cl_res_state)) {
230 nfs4_schedule_state_renewal(clp);
234 status = nfs4_proc_get_lease_time(clp, &fsinfo);
236 /* Update lease time and schedule renewal */
237 spin_lock(&clp->cl_lock);
238 clp->cl_lease_time = fsinfo.lease_time * HZ;
239 clp->cl_last_renewal = jiffies;
240 spin_unlock(&clp->cl_lock);
242 nfs4_schedule_state_renewal(clp);
249 * Back channel returns NFS4ERR_DELAY for new requests when
250 * NFS4_SESSION_DRAINING is set so there is no work to be done when draining
253 static void nfs4_end_drain_session(struct nfs_client *clp)
255 struct nfs4_session *ses = clp->cl_session;
256 struct nfs4_slot_table *tbl;
261 tbl = &ses->fc_slot_table;
262 if (test_and_clear_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
263 spin_lock(&tbl->slot_tbl_lock);
264 for (i = 0; i <= tbl->max_slotid; i++) {
265 if (rpc_wake_up_first(&tbl->slot_tbl_waitq,
266 nfs4_set_task_privileged,
270 spin_unlock(&tbl->slot_tbl_lock);
274 static int nfs4_wait_on_slot_tbl(struct nfs4_slot_table *tbl)
276 spin_lock(&tbl->slot_tbl_lock);
277 if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
278 INIT_COMPLETION(tbl->complete);
279 spin_unlock(&tbl->slot_tbl_lock);
280 return wait_for_completion_interruptible(&tbl->complete);
282 spin_unlock(&tbl->slot_tbl_lock);
286 static int nfs4_begin_drain_session(struct nfs_client *clp)
288 struct nfs4_session *ses = clp->cl_session;
291 set_bit(NFS4_SESSION_DRAINING, &ses->session_state);
293 ret = nfs4_wait_on_slot_tbl(&ses->bc_slot_table);
297 return nfs4_wait_on_slot_tbl(&ses->fc_slot_table);
300 static void nfs41_finish_session_reset(struct nfs_client *clp)
302 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
303 clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
304 /* create_session negotiated new slot table */
305 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
306 nfs41_setup_state_renewal(clp);
309 int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
313 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
315 nfs4_begin_drain_session(clp);
316 status = nfs4_proc_exchange_id(clp, cred);
319 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
321 status = nfs4_proc_create_session(clp, cred);
324 nfs41_finish_session_reset(clp);
325 nfs_mark_client_ready(clp, NFS_CS_READY);
331 * nfs41_discover_server_trunking - Detect server IP address trunking (mv1)
333 * @clp: nfs_client under test
334 * @result: OUT: found nfs_client, or clp
335 * @cred: credential to use for trunking test
337 * Returns NFS4_OK, a negative errno, or a negative NFS4ERR status.
338 * If NFS4_OK is returned, an nfs_client pointer is planted in
341 * Note: The returned client may not yet be marked ready.
343 int nfs41_discover_server_trunking(struct nfs_client *clp,
344 struct nfs_client **result,
345 struct rpc_cred *cred)
349 status = nfs4_proc_exchange_id(clp, cred);
350 if (status != NFS4_OK)
352 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
354 return nfs41_walk_client_list(clp, result, cred);
357 struct rpc_cred *nfs4_get_exchange_id_cred(struct nfs_client *clp)
359 struct rpc_cred *cred;
361 spin_lock(&clp->cl_lock);
362 cred = nfs4_get_machine_cred_locked(clp);
363 spin_unlock(&clp->cl_lock);
367 #endif /* CONFIG_NFS_V4_1 */
369 static struct rpc_cred *
370 nfs4_get_setclientid_cred_server(struct nfs_server *server)
372 struct nfs_client *clp = server->nfs_client;
373 struct rpc_cred *cred = NULL;
374 struct nfs4_state_owner *sp;
377 spin_lock(&clp->cl_lock);
378 pos = rb_first(&server->state_owners);
380 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
381 cred = get_rpccred(sp->so_cred);
383 spin_unlock(&clp->cl_lock);
388 * nfs4_get_setclientid_cred - Acquire credential for a setclientid operation
389 * @clp: client state handle
391 * Returns an rpc_cred with reference count bumped, or NULL.
393 struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
395 struct nfs_server *server;
396 struct rpc_cred *cred;
398 spin_lock(&clp->cl_lock);
399 cred = nfs4_get_machine_cred_locked(clp);
400 spin_unlock(&clp->cl_lock);
405 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
406 cred = nfs4_get_setclientid_cred_server(server);
416 static struct nfs4_state_owner *
417 nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
419 struct rb_node **p = &server->state_owners.rb_node,
421 struct nfs4_state_owner *sp;
425 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
427 if (cred < sp->so_cred)
428 p = &parent->rb_left;
429 else if (cred > sp->so_cred)
430 p = &parent->rb_right;
432 if (!list_empty(&sp->so_lru))
433 list_del_init(&sp->so_lru);
434 atomic_inc(&sp->so_count);
441 static struct nfs4_state_owner *
442 nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
444 struct nfs_server *server = new->so_server;
445 struct rb_node **p = &server->state_owners.rb_node,
447 struct nfs4_state_owner *sp;
452 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
454 if (new->so_cred < sp->so_cred)
455 p = &parent->rb_left;
456 else if (new->so_cred > sp->so_cred)
457 p = &parent->rb_right;
459 if (!list_empty(&sp->so_lru))
460 list_del_init(&sp->so_lru);
461 atomic_inc(&sp->so_count);
465 err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
468 rb_link_node(&new->so_server_node, parent, p);
469 rb_insert_color(&new->so_server_node, &server->state_owners);
474 nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
476 struct nfs_server *server = sp->so_server;
478 if (!RB_EMPTY_NODE(&sp->so_server_node))
479 rb_erase(&sp->so_server_node, &server->state_owners);
480 ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
484 nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
486 sc->create_time = ktime_get();
489 spin_lock_init(&sc->lock);
490 INIT_LIST_HEAD(&sc->list);
491 rpc_init_wait_queue(&sc->wait, "Seqid_waitqueue");
495 nfs4_destroy_seqid_counter(struct nfs_seqid_counter *sc)
497 rpc_destroy_wait_queue(&sc->wait);
501 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
502 * create a new state_owner.
505 static struct nfs4_state_owner *
506 nfs4_alloc_state_owner(struct nfs_server *server,
507 struct rpc_cred *cred,
510 struct nfs4_state_owner *sp;
512 sp = kzalloc(sizeof(*sp), gfp_flags);
515 sp->so_server = server;
516 sp->so_cred = get_rpccred(cred);
517 spin_lock_init(&sp->so_lock);
518 INIT_LIST_HEAD(&sp->so_states);
519 nfs4_init_seqid_counter(&sp->so_seqid);
520 atomic_set(&sp->so_count, 1);
521 INIT_LIST_HEAD(&sp->so_lru);
526 nfs4_drop_state_owner(struct nfs4_state_owner *sp)
528 struct rb_node *rb_node = &sp->so_server_node;
530 if (!RB_EMPTY_NODE(rb_node)) {
531 struct nfs_server *server = sp->so_server;
532 struct nfs_client *clp = server->nfs_client;
534 spin_lock(&clp->cl_lock);
535 if (!RB_EMPTY_NODE(rb_node)) {
536 rb_erase(rb_node, &server->state_owners);
537 RB_CLEAR_NODE(rb_node);
539 spin_unlock(&clp->cl_lock);
543 static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
545 nfs4_destroy_seqid_counter(&sp->so_seqid);
546 put_rpccred(sp->so_cred);
550 static void nfs4_gc_state_owners(struct nfs_server *server)
552 struct nfs_client *clp = server->nfs_client;
553 struct nfs4_state_owner *sp, *tmp;
554 unsigned long time_min, time_max;
557 spin_lock(&clp->cl_lock);
559 time_min = (long)time_max - (long)clp->cl_lease_time;
560 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
561 /* NB: LRU is sorted so that oldest is at the head */
562 if (time_in_range(sp->so_expires, time_min, time_max))
564 list_move(&sp->so_lru, &doomed);
565 nfs4_remove_state_owner_locked(sp);
567 spin_unlock(&clp->cl_lock);
569 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
570 list_del(&sp->so_lru);
571 nfs4_free_state_owner(sp);
576 * nfs4_get_state_owner - Look up a state owner given a credential
577 * @server: nfs_server to search
578 * @cred: RPC credential to match
580 * Returns a pointer to an instantiated nfs4_state_owner struct, or NULL.
582 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server,
583 struct rpc_cred *cred,
586 struct nfs_client *clp = server->nfs_client;
587 struct nfs4_state_owner *sp, *new;
589 spin_lock(&clp->cl_lock);
590 sp = nfs4_find_state_owner_locked(server, cred);
591 spin_unlock(&clp->cl_lock);
594 new = nfs4_alloc_state_owner(server, cred, gfp_flags);
598 if (ida_pre_get(&server->openowner_id, gfp_flags) == 0)
600 spin_lock(&clp->cl_lock);
601 sp = nfs4_insert_state_owner_locked(new);
602 spin_unlock(&clp->cl_lock);
603 } while (sp == ERR_PTR(-EAGAIN));
605 nfs4_free_state_owner(new);
607 nfs4_gc_state_owners(server);
612 * nfs4_put_state_owner - Release a nfs4_state_owner
613 * @sp: state owner data to release
615 * Note that we keep released state owners on an LRU
617 * This caches valid state owners so that they can be
618 * reused, to avoid the OPEN_CONFIRM on minor version 0.
619 * It also pins the uniquifier of dropped state owners for
620 * a while, to ensure that those state owner names are
623 void nfs4_put_state_owner(struct nfs4_state_owner *sp)
625 struct nfs_server *server = sp->so_server;
626 struct nfs_client *clp = server->nfs_client;
628 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
631 sp->so_expires = jiffies;
632 list_add_tail(&sp->so_lru, &server->state_owners_lru);
633 spin_unlock(&clp->cl_lock);
637 * nfs4_purge_state_owners - Release all cached state owners
638 * @server: nfs_server with cached state owners to release
640 * Called at umount time. Remaining state owners will be on
641 * the LRU with ref count of zero.
643 void nfs4_purge_state_owners(struct nfs_server *server)
645 struct nfs_client *clp = server->nfs_client;
646 struct nfs4_state_owner *sp, *tmp;
649 spin_lock(&clp->cl_lock);
650 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
651 list_move(&sp->so_lru, &doomed);
652 nfs4_remove_state_owner_locked(sp);
654 spin_unlock(&clp->cl_lock);
656 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
657 list_del(&sp->so_lru);
658 nfs4_free_state_owner(sp);
662 static struct nfs4_state *
663 nfs4_alloc_open_state(void)
665 struct nfs4_state *state;
667 state = kzalloc(sizeof(*state), GFP_NOFS);
670 atomic_set(&state->count, 1);
671 INIT_LIST_HEAD(&state->lock_states);
672 spin_lock_init(&state->state_lock);
673 seqlock_init(&state->seqlock);
678 nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
680 if (state->state == fmode)
682 /* NB! List reordering - see the reclaim code for why. */
683 if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
684 if (fmode & FMODE_WRITE)
685 list_move(&state->open_states, &state->owner->so_states);
687 list_move_tail(&state->open_states, &state->owner->so_states);
689 state->state = fmode;
692 static struct nfs4_state *
693 __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
695 struct nfs_inode *nfsi = NFS_I(inode);
696 struct nfs4_state *state;
698 list_for_each_entry(state, &nfsi->open_states, inode_states) {
699 if (state->owner != owner)
701 if (atomic_inc_not_zero(&state->count))
708 nfs4_free_open_state(struct nfs4_state *state)
714 nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
716 struct nfs4_state *state, *new;
717 struct nfs_inode *nfsi = NFS_I(inode);
719 spin_lock(&inode->i_lock);
720 state = __nfs4_find_state_byowner(inode, owner);
721 spin_unlock(&inode->i_lock);
724 new = nfs4_alloc_open_state();
725 spin_lock(&owner->so_lock);
726 spin_lock(&inode->i_lock);
727 state = __nfs4_find_state_byowner(inode, owner);
728 if (state == NULL && new != NULL) {
730 state->owner = owner;
731 atomic_inc(&owner->so_count);
732 list_add(&state->inode_states, &nfsi->open_states);
734 state->inode = inode;
735 spin_unlock(&inode->i_lock);
736 /* Note: The reclaim code dictates that we add stateless
737 * and read-only stateids to the end of the list */
738 list_add_tail(&state->open_states, &owner->so_states);
739 spin_unlock(&owner->so_lock);
741 spin_unlock(&inode->i_lock);
742 spin_unlock(&owner->so_lock);
744 nfs4_free_open_state(new);
750 void nfs4_put_open_state(struct nfs4_state *state)
752 struct inode *inode = state->inode;
753 struct nfs4_state_owner *owner = state->owner;
755 if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
757 spin_lock(&inode->i_lock);
758 list_del(&state->inode_states);
759 list_del(&state->open_states);
760 spin_unlock(&inode->i_lock);
761 spin_unlock(&owner->so_lock);
763 nfs4_free_open_state(state);
764 nfs4_put_state_owner(owner);
768 * Close the current file.
770 static void __nfs4_close(struct nfs4_state *state,
771 fmode_t fmode, gfp_t gfp_mask, int wait)
773 struct nfs4_state_owner *owner = state->owner;
777 atomic_inc(&owner->so_count);
778 /* Protect against nfs4_find_state() */
779 spin_lock(&owner->so_lock);
780 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
787 case FMODE_READ|FMODE_WRITE:
790 newstate = FMODE_READ|FMODE_WRITE;
791 if (state->n_rdwr == 0) {
792 if (state->n_rdonly == 0) {
793 newstate &= ~FMODE_READ;
794 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
795 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
797 if (state->n_wronly == 0) {
798 newstate &= ~FMODE_WRITE;
799 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
800 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
803 clear_bit(NFS_DELEGATED_STATE, &state->flags);
805 nfs4_state_set_mode_locked(state, newstate);
806 spin_unlock(&owner->so_lock);
809 nfs4_put_open_state(state);
810 nfs4_put_state_owner(owner);
812 nfs4_do_close(state, gfp_mask, wait);
815 void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
817 __nfs4_close(state, fmode, GFP_NOFS, 0);
820 void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
822 __nfs4_close(state, fmode, GFP_KERNEL, 1);
826 * Search the state->lock_states for an existing lock_owner
827 * that is compatible with current->files
829 static struct nfs4_lock_state *
830 __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type)
832 struct nfs4_lock_state *pos;
833 list_for_each_entry(pos, &state->lock_states, ls_locks) {
834 if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
836 switch (pos->ls_owner.lo_type) {
837 case NFS4_POSIX_LOCK_TYPE:
838 if (pos->ls_owner.lo_u.posix_owner != fl_owner)
841 case NFS4_FLOCK_LOCK_TYPE:
842 if (pos->ls_owner.lo_u.flock_owner != fl_pid)
845 atomic_inc(&pos->ls_count);
852 * Return a compatible lock_state. If no initialized lock_state structure
853 * exists, return an uninitialized one.
856 static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type)
858 struct nfs4_lock_state *lsp;
859 struct nfs_server *server = state->owner->so_server;
861 lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
864 nfs4_init_seqid_counter(&lsp->ls_seqid);
865 atomic_set(&lsp->ls_count, 1);
866 lsp->ls_state = state;
867 lsp->ls_owner.lo_type = type;
868 switch (lsp->ls_owner.lo_type) {
869 case NFS4_FLOCK_LOCK_TYPE:
870 lsp->ls_owner.lo_u.flock_owner = fl_pid;
872 case NFS4_POSIX_LOCK_TYPE:
873 lsp->ls_owner.lo_u.posix_owner = fl_owner;
878 lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
879 if (lsp->ls_seqid.owner_id < 0)
881 INIT_LIST_HEAD(&lsp->ls_locks);
888 void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
890 ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
891 nfs4_destroy_seqid_counter(&lsp->ls_seqid);
896 * Return a compatible lock_state. If no initialized lock_state structure
897 * exists, return an uninitialized one.
900 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner, pid_t pid, unsigned int type)
902 struct nfs4_lock_state *lsp, *new = NULL;
905 spin_lock(&state->state_lock);
906 lsp = __nfs4_find_lock_state(state, owner, pid, type);
910 list_add(&new->ls_locks, &state->lock_states);
911 set_bit(LK_STATE_IN_USE, &state->flags);
916 spin_unlock(&state->state_lock);
917 new = nfs4_alloc_lock_state(state, owner, pid, type);
921 spin_unlock(&state->state_lock);
923 nfs4_free_lock_state(state->owner->so_server, new);
928 * Release reference to lock_state, and free it if we see that
929 * it is no longer in use
931 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
933 struct nfs4_state *state;
937 state = lsp->ls_state;
938 if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
940 list_del(&lsp->ls_locks);
941 if (list_empty(&state->lock_states))
942 clear_bit(LK_STATE_IN_USE, &state->flags);
943 spin_unlock(&state->state_lock);
944 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
945 if (nfs4_release_lockowner(lsp) == 0)
948 nfs4_free_lock_state(lsp->ls_state->owner->so_server, lsp);
951 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
953 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
955 dst->fl_u.nfs4_fl.owner = lsp;
956 atomic_inc(&lsp->ls_count);
959 static void nfs4_fl_release_lock(struct file_lock *fl)
961 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
964 static const struct file_lock_operations nfs4_fl_lock_ops = {
965 .fl_copy_lock = nfs4_fl_copy_lock,
966 .fl_release_private = nfs4_fl_release_lock,
969 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
971 struct nfs4_lock_state *lsp;
973 if (fl->fl_ops != NULL)
975 if (fl->fl_flags & FL_POSIX)
976 lsp = nfs4_get_lock_state(state, fl->fl_owner, 0, NFS4_POSIX_LOCK_TYPE);
977 else if (fl->fl_flags & FL_FLOCK)
978 lsp = nfs4_get_lock_state(state, NULL, fl->fl_pid,
979 NFS4_FLOCK_LOCK_TYPE);
984 fl->fl_u.nfs4_fl.owner = lsp;
985 fl->fl_ops = &nfs4_fl_lock_ops;
989 static bool nfs4_copy_lock_stateid(nfs4_stateid *dst, struct nfs4_state *state,
990 const struct nfs_lockowner *lockowner)
992 struct nfs4_lock_state *lsp;
998 if (lockowner == NULL)
1001 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
1004 fl_owner = lockowner->l_owner;
1005 fl_pid = lockowner->l_pid;
1006 spin_lock(&state->state_lock);
1007 lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
1008 if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
1009 nfs4_stateid_copy(dst, &lsp->ls_stateid);
1012 spin_unlock(&state->state_lock);
1013 nfs4_put_lock_state(lsp);
1018 static void nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
1023 seq = read_seqbegin(&state->seqlock);
1024 nfs4_stateid_copy(dst, &state->stateid);
1025 } while (read_seqretry(&state->seqlock, seq));
1029 * Byte-range lock aware utility to initialize the stateid of read/write
1032 void nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state,
1033 fmode_t fmode, const struct nfs_lockowner *lockowner)
1035 if (nfs4_copy_delegation_stateid(dst, state->inode, fmode))
1037 if (nfs4_copy_lock_stateid(dst, state, lockowner))
1039 nfs4_copy_open_stateid(dst, state);
1042 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
1044 struct nfs_seqid *new;
1046 new = kmalloc(sizeof(*new), gfp_mask);
1048 new->sequence = counter;
1049 INIT_LIST_HEAD(&new->list);
1055 void nfs_release_seqid(struct nfs_seqid *seqid)
1057 struct nfs_seqid_counter *sequence;
1059 if (list_empty(&seqid->list))
1061 sequence = seqid->sequence;
1062 spin_lock(&sequence->lock);
1063 list_del_init(&seqid->list);
1064 if (!list_empty(&sequence->list)) {
1065 struct nfs_seqid *next;
1067 next = list_first_entry(&sequence->list,
1068 struct nfs_seqid, list);
1069 rpc_wake_up_queued_task(&sequence->wait, next->task);
1071 spin_unlock(&sequence->lock);
1074 void nfs_free_seqid(struct nfs_seqid *seqid)
1076 nfs_release_seqid(seqid);
1081 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
1082 * failed with a seqid incrementing error -
1083 * see comments nfs_fs.h:seqid_mutating_error()
1085 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
1090 case -NFS4ERR_BAD_SEQID:
1091 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
1093 pr_warn_ratelimited("NFS: v4 server returned a bad"
1094 " sequence-id error on an"
1095 " unconfirmed sequence %p!\n",
1097 case -NFS4ERR_STALE_CLIENTID:
1098 case -NFS4ERR_STALE_STATEID:
1099 case -NFS4ERR_BAD_STATEID:
1100 case -NFS4ERR_BADXDR:
1101 case -NFS4ERR_RESOURCE:
1102 case -NFS4ERR_NOFILEHANDLE:
1103 /* Non-seqid mutating errors */
1107 * Note: no locking needed as we are guaranteed to be first
1108 * on the sequence list
1110 seqid->sequence->counter++;
1113 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
1115 struct nfs4_state_owner *sp = container_of(seqid->sequence,
1116 struct nfs4_state_owner, so_seqid);
1117 struct nfs_server *server = sp->so_server;
1119 if (status == -NFS4ERR_BAD_SEQID)
1120 nfs4_drop_state_owner(sp);
1121 if (!nfs4_has_session(server->nfs_client))
1122 nfs_increment_seqid(status, seqid);
1126 * Increment the seqid if the LOCK/LOCKU succeeded, or
1127 * failed with a seqid incrementing error -
1128 * see comments nfs_fs.h:seqid_mutating_error()
1130 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
1132 nfs_increment_seqid(status, seqid);
1135 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
1137 struct nfs_seqid_counter *sequence = seqid->sequence;
1140 spin_lock(&sequence->lock);
1142 if (list_empty(&seqid->list))
1143 list_add_tail(&seqid->list, &sequence->list);
1144 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
1146 rpc_sleep_on(&sequence->wait, task, NULL);
1149 spin_unlock(&sequence->lock);
1153 static int nfs4_run_state_manager(void *);
1155 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
1157 smp_mb__before_clear_bit();
1158 clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
1159 smp_mb__after_clear_bit();
1160 wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
1161 rpc_wake_up(&clp->cl_rpcwaitq);
1165 * Schedule the nfs_client asynchronous state management routine
1167 void nfs4_schedule_state_manager(struct nfs_client *clp)
1169 struct task_struct *task;
1170 char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
1172 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1174 __module_get(THIS_MODULE);
1175 atomic_inc(&clp->cl_count);
1177 /* The rcu_read_lock() is not strictly necessary, as the state
1178 * manager is the only thread that ever changes the rpc_xprt
1179 * after it's initialized. At this point, we're single threaded. */
1181 snprintf(buf, sizeof(buf), "%s-manager",
1182 rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR));
1184 task = kthread_run(nfs4_run_state_manager, clp, buf);
1186 printk(KERN_ERR "%s: kthread_run: %ld\n",
1187 __func__, PTR_ERR(task));
1188 nfs4_clear_state_manager_bit(clp);
1189 nfs_put_client(clp);
1190 module_put(THIS_MODULE);
1195 * Schedule a lease recovery attempt
1197 void nfs4_schedule_lease_recovery(struct nfs_client *clp)
1201 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1202 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1203 dprintk("%s: scheduling lease recovery for server %s\n", __func__,
1205 nfs4_schedule_state_manager(clp);
1207 EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
1210 * nfs40_handle_cb_pathdown - return all delegations after NFS4ERR_CB_PATH_DOWN
1211 * @clp: client to process
1213 * Set the NFS4CLNT_LEASE_EXPIRED state in order to force a
1214 * resend of the SETCLIENTID and hence re-establish the
1215 * callback channel. Then return all existing delegations.
1217 static void nfs40_handle_cb_pathdown(struct nfs_client *clp)
1219 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1220 nfs_expire_all_delegations(clp);
1221 dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
1225 void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
1227 nfs40_handle_cb_pathdown(clp);
1228 nfs4_schedule_state_manager(clp);
1231 static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1234 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1235 /* Don't recover state that expired before the reboot */
1236 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
1237 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1240 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1241 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1245 static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1247 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
1248 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1249 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1250 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
1254 void nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
1256 struct nfs_client *clp = server->nfs_client;
1258 nfs4_state_mark_reclaim_nograce(clp, state);
1259 dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
1261 nfs4_schedule_state_manager(clp);
1263 EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1265 void nfs_inode_find_state_and_recover(struct inode *inode,
1266 const nfs4_stateid *stateid)
1268 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1269 struct nfs_inode *nfsi = NFS_I(inode);
1270 struct nfs_open_context *ctx;
1271 struct nfs4_state *state;
1274 spin_lock(&inode->i_lock);
1275 list_for_each_entry(ctx, &nfsi->open_files, list) {
1279 if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
1281 if (!nfs4_stateid_match(&state->stateid, stateid))
1283 nfs4_state_mark_reclaim_nograce(clp, state);
1286 spin_unlock(&inode->i_lock);
1288 nfs4_schedule_state_manager(clp);
1292 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
1294 struct inode *inode = state->inode;
1295 struct nfs_inode *nfsi = NFS_I(inode);
1296 struct file_lock *fl;
1299 if (inode->i_flock == NULL)
1302 /* Guard against delegation returns and new lock/unlock calls */
1303 down_write(&nfsi->rwsem);
1304 /* Protect inode->i_flock using the BKL */
1306 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1307 if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
1309 if (nfs_file_open_context(fl->fl_file)->state != state)
1312 status = ops->recover_lock(state, fl);
1317 case -NFS4ERR_ADMIN_REVOKED:
1318 case -NFS4ERR_STALE_STATEID:
1319 case -NFS4ERR_BAD_STATEID:
1320 case -NFS4ERR_EXPIRED:
1321 case -NFS4ERR_NO_GRACE:
1322 case -NFS4ERR_STALE_CLIENTID:
1323 case -NFS4ERR_BADSESSION:
1324 case -NFS4ERR_BADSLOT:
1325 case -NFS4ERR_BAD_HIGH_SLOT:
1326 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1329 printk(KERN_ERR "NFS: %s: unhandled error %d. "
1330 "Zeroing state\n", __func__, status);
1332 case -NFS4ERR_DENIED:
1333 case -NFS4ERR_RECLAIM_BAD:
1334 case -NFS4ERR_RECLAIM_CONFLICT:
1335 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1342 up_write(&nfsi->rwsem);
1346 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
1348 struct nfs4_state *state;
1349 struct nfs4_lock_state *lock;
1352 /* Note: we rely on the sp->so_states list being ordered
1353 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
1355 * This is needed to ensure that the server won't give us any
1356 * read delegations that we have to return if, say, we are
1357 * recovering after a network partition or a reboot from a
1358 * server that doesn't support a grace period.
1361 spin_lock(&sp->so_lock);
1362 list_for_each_entry(state, &sp->so_states, open_states) {
1363 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
1365 if (state->state == 0)
1367 atomic_inc(&state->count);
1368 spin_unlock(&sp->so_lock);
1369 status = ops->recover_open(sp, state);
1371 status = nfs4_reclaim_locks(state, ops);
1373 spin_lock(&state->state_lock);
1374 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1375 if (!test_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags))
1376 pr_warn_ratelimited("NFS: "
1378 "failed!\n", __func__);
1380 spin_unlock(&state->state_lock);
1381 nfs4_put_open_state(state);
1387 printk(KERN_ERR "NFS: %s: unhandled error %d. "
1388 "Zeroing state\n", __func__, status);
1393 * Open state on this file cannot be recovered
1394 * All we can do is revert to using the zero stateid.
1396 memset(&state->stateid, 0,
1397 sizeof(state->stateid));
1398 /* Mark the file as being 'closed' */
1403 * User RPCSEC_GSS context has expired.
1404 * We cannot recover this stateid now, so
1405 * skip it and allow recovery thread to
1409 case -NFS4ERR_ADMIN_REVOKED:
1410 case -NFS4ERR_STALE_STATEID:
1411 case -NFS4ERR_BAD_STATEID:
1412 case -NFS4ERR_RECLAIM_BAD:
1413 case -NFS4ERR_RECLAIM_CONFLICT:
1414 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1416 case -NFS4ERR_EXPIRED:
1417 case -NFS4ERR_NO_GRACE:
1418 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1419 case -NFS4ERR_STALE_CLIENTID:
1420 case -NFS4ERR_BADSESSION:
1421 case -NFS4ERR_BADSLOT:
1422 case -NFS4ERR_BAD_HIGH_SLOT:
1423 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1426 nfs4_put_open_state(state);
1429 spin_unlock(&sp->so_lock);
1432 nfs4_put_open_state(state);
1436 static void nfs4_clear_open_state(struct nfs4_state *state)
1438 struct nfs4_lock_state *lock;
1440 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1441 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1442 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1443 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1444 spin_lock(&state->state_lock);
1445 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1446 lock->ls_seqid.flags = 0;
1447 clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
1449 spin_unlock(&state->state_lock);
1452 static void nfs4_reset_seqids(struct nfs_server *server,
1453 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1455 struct nfs_client *clp = server->nfs_client;
1456 struct nfs4_state_owner *sp;
1457 struct rb_node *pos;
1458 struct nfs4_state *state;
1460 spin_lock(&clp->cl_lock);
1461 for (pos = rb_first(&server->state_owners);
1463 pos = rb_next(pos)) {
1464 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1465 sp->so_seqid.flags = 0;
1466 spin_lock(&sp->so_lock);
1467 list_for_each_entry(state, &sp->so_states, open_states) {
1468 if (mark_reclaim(clp, state))
1469 nfs4_clear_open_state(state);
1471 spin_unlock(&sp->so_lock);
1473 spin_unlock(&clp->cl_lock);
1476 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp,
1477 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1479 struct nfs_server *server;
1482 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1483 nfs4_reset_seqids(server, mark_reclaim);
1487 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1489 /* Mark all delegations for reclaim */
1490 nfs_delegation_mark_reclaim(clp);
1491 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1494 static void nfs4_reclaim_complete(struct nfs_client *clp,
1495 const struct nfs4_state_recovery_ops *ops)
1497 /* Notify the server we're done reclaiming our state */
1498 if (ops->reclaim_complete)
1499 (void)ops->reclaim_complete(clp);
1502 static void nfs4_clear_reclaim_server(struct nfs_server *server)
1504 struct nfs_client *clp = server->nfs_client;
1505 struct nfs4_state_owner *sp;
1506 struct rb_node *pos;
1507 struct nfs4_state *state;
1509 spin_lock(&clp->cl_lock);
1510 for (pos = rb_first(&server->state_owners);
1512 pos = rb_next(pos)) {
1513 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1514 spin_lock(&sp->so_lock);
1515 list_for_each_entry(state, &sp->so_states, open_states) {
1516 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
1519 nfs4_state_mark_reclaim_nograce(clp, state);
1521 spin_unlock(&sp->so_lock);
1523 spin_unlock(&clp->cl_lock);
1526 static int nfs4_state_clear_reclaim_reboot(struct nfs_client *clp)
1528 struct nfs_server *server;
1530 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1534 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1535 nfs4_clear_reclaim_server(server);
1538 nfs_delegation_reap_unclaimed(clp);
1542 static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
1544 if (!nfs4_state_clear_reclaim_reboot(clp))
1546 nfs4_reclaim_complete(clp, clp->cl_mvops->reboot_recovery_ops);
1549 static void nfs_delegation_clear_all(struct nfs_client *clp)
1551 nfs_delegation_mark_reclaim(clp);
1552 nfs_delegation_reap_unclaimed(clp);
1555 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1557 nfs_delegation_clear_all(clp);
1558 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1561 static void nfs4_warn_keyexpired(const char *s)
1563 printk_ratelimited(KERN_WARNING "Error: state manager"
1564 " encountered RPCSEC_GSS session"
1565 " expired against NFSv4 server %s.\n",
1569 static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1574 case -NFS4ERR_CB_PATH_DOWN:
1575 nfs40_handle_cb_pathdown(clp);
1577 case -NFS4ERR_NO_GRACE:
1578 nfs4_state_end_reclaim_reboot(clp);
1580 case -NFS4ERR_STALE_CLIENTID:
1581 case -NFS4ERR_LEASE_MOVED:
1582 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1583 nfs4_state_clear_reclaim_reboot(clp);
1584 nfs4_state_start_reclaim_reboot(clp);
1586 case -NFS4ERR_EXPIRED:
1587 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1588 nfs4_state_start_reclaim_nograce(clp);
1590 case -NFS4ERR_BADSESSION:
1591 case -NFS4ERR_BADSLOT:
1592 case -NFS4ERR_BAD_HIGH_SLOT:
1593 case -NFS4ERR_DEADSESSION:
1594 case -NFS4ERR_SEQ_FALSE_RETRY:
1595 case -NFS4ERR_SEQ_MISORDERED:
1596 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1597 /* Zero session reset errors */
1599 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1600 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1603 /* Nothing we can do */
1604 nfs4_warn_keyexpired(clp->cl_hostname);
1607 dprintk("%s: failed to handle error %d for server %s\n",
1608 __func__, error, clp->cl_hostname);
1611 dprintk("%s: handled error %d for server %s\n", __func__, error,
1616 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1618 struct nfs4_state_owner *sp;
1619 struct nfs_server *server;
1620 struct rb_node *pos;
1625 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1626 nfs4_purge_state_owners(server);
1627 spin_lock(&clp->cl_lock);
1628 for (pos = rb_first(&server->state_owners);
1630 pos = rb_next(pos)) {
1632 struct nfs4_state_owner, so_server_node);
1633 if (!test_and_clear_bit(ops->owner_flag_bit,
1636 atomic_inc(&sp->so_count);
1637 spin_unlock(&clp->cl_lock);
1640 status = nfs4_reclaim_open_state(sp, ops);
1642 set_bit(ops->owner_flag_bit, &sp->so_flags);
1643 nfs4_put_state_owner(sp);
1644 return nfs4_recovery_handle_error(clp, status);
1647 nfs4_put_state_owner(sp);
1650 spin_unlock(&clp->cl_lock);
1656 static int nfs4_check_lease(struct nfs_client *clp)
1658 struct rpc_cred *cred;
1659 const struct nfs4_state_maintenance_ops *ops =
1660 clp->cl_mvops->state_renewal_ops;
1663 /* Is the client already known to have an expired lease? */
1664 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1666 spin_lock(&clp->cl_lock);
1667 cred = ops->get_state_renewal_cred_locked(clp);
1668 spin_unlock(&clp->cl_lock);
1670 cred = nfs4_get_setclientid_cred(clp);
1675 status = ops->renew_lease(clp, cred);
1678 return nfs4_recovery_handle_error(clp, status);
1681 /* Set NFS4CLNT_LEASE_EXPIRED and reclaim reboot state for all v4.0 errors
1682 * and for recoverable errors on EXCHANGE_ID for v4.1
1684 static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
1687 case -NFS4ERR_SEQ_MISORDERED:
1688 if (test_and_set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state))
1689 return -ESERVERFAULT;
1690 /* Lease confirmation error: retry after purging the lease */
1692 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1694 case -NFS4ERR_STALE_CLIENTID:
1695 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1696 nfs4_state_clear_reclaim_reboot(clp);
1697 nfs4_state_start_reclaim_reboot(clp);
1699 case -NFS4ERR_CLID_INUSE:
1700 pr_err("NFS: Server %s reports our clientid is in use\n",
1702 nfs_mark_client_ready(clp, -EPERM);
1703 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1706 if (clp->cl_machine_cred == NULL)
1708 /* Handle case where the user hasn't set up machine creds */
1709 nfs4_clear_machine_cred(clp);
1710 case -NFS4ERR_DELAY:
1716 case -NFS4ERR_MINOR_VERS_MISMATCH:
1717 if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
1718 nfs_mark_client_ready(clp, -EPROTONOSUPPORT);
1719 dprintk("%s: exit with error %d for server %s\n",
1720 __func__, -EPROTONOSUPPORT, clp->cl_hostname);
1721 return -EPROTONOSUPPORT;
1723 nfs4_warn_keyexpired(clp->cl_hostname);
1724 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1725 * in nfs4_exchange_id */
1727 dprintk("%s: exit with error %d for server %s\n", __func__,
1728 status, clp->cl_hostname);
1731 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1732 dprintk("%s: handled error %d for server %s\n", __func__, status,
1737 static int nfs4_establish_lease(struct nfs_client *clp)
1739 struct rpc_cred *cred;
1740 const struct nfs4_state_recovery_ops *ops =
1741 clp->cl_mvops->reboot_recovery_ops;
1744 cred = ops->get_clid_cred(clp);
1747 status = ops->establish_clid(clp, cred);
1751 pnfs_destroy_all_layouts(clp);
1756 * Returns zero or a negative errno. NFS4ERR values are converted
1757 * to local errno values.
1759 static int nfs4_reclaim_lease(struct nfs_client *clp)
1763 status = nfs4_establish_lease(clp);
1765 return nfs4_handle_reclaim_lease_error(clp, status);
1766 if (test_and_clear_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state))
1767 nfs4_state_start_reclaim_nograce(clp);
1768 if (!test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
1769 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1770 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1771 clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1775 static int nfs4_purge_lease(struct nfs_client *clp)
1779 status = nfs4_establish_lease(clp);
1781 return nfs4_handle_reclaim_lease_error(clp, status);
1782 clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1783 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1784 nfs4_state_start_reclaim_nograce(clp);
1789 * nfs4_discover_server_trunking - Detect server IP address trunking
1791 * @clp: nfs_client under test
1792 * @result: OUT: found nfs_client, or clp
1794 * Returns zero or a negative errno. If zero is returned,
1795 * an nfs_client pointer is planted in "result".
1797 * Note: since we are invoked in process context, and
1798 * not from inside the state manager, we cannot use
1799 * nfs4_handle_reclaim_lease_error().
1801 int nfs4_discover_server_trunking(struct nfs_client *clp,
1802 struct nfs_client **result)
1804 const struct nfs4_state_recovery_ops *ops =
1805 clp->cl_mvops->reboot_recovery_ops;
1806 rpc_authflavor_t *flavors, flav, save;
1807 struct rpc_clnt *clnt;
1808 struct rpc_cred *cred;
1811 dprintk("NFS: %s: testing '%s'\n", __func__, clp->cl_hostname);
1813 len = NFS_MAX_SECFLAVORS;
1814 flavors = kcalloc(len, sizeof(*flavors), GFP_KERNEL);
1815 if (flavors == NULL) {
1819 len = rpcauth_list_flavors(flavors, len);
1824 clnt = clp->cl_rpcclient;
1825 save = clnt->cl_auth->au_flavor;
1828 mutex_lock(&nfs_clid_init_mutex);
1831 cred = ops->get_clid_cred(clp);
1835 status = ops->detect_trunking(clp, result, cred);
1842 if (clp->cl_machine_cred == NULL)
1844 /* Handle case where the user hasn't set up machine creds */
1845 nfs4_clear_machine_cred(clp);
1846 case -NFS4ERR_DELAY:
1850 dprintk("NFS: %s after status %d, retrying\n",
1854 case -NFS4ERR_CLID_INUSE:
1855 case -NFS4ERR_WRONGSEC:
1860 flav = flavors[i++];
1862 flav = flavors[i++];
1863 clnt = rpc_clone_client_set_auth(clnt, flav);
1865 status = PTR_ERR(clnt);
1868 clp->cl_rpcclient = clnt;
1871 case -NFS4ERR_MINOR_VERS_MISMATCH:
1872 status = -EPROTONOSUPPORT;
1876 nfs4_warn_keyexpired(clp->cl_hostname);
1877 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1878 * in nfs4_exchange_id */
1879 status = -EKEYEXPIRED;
1883 mutex_unlock(&nfs_clid_init_mutex);
1887 dprintk("NFS: %s: status = %d\n", __func__, status);
1891 #ifdef CONFIG_NFS_V4_1
1892 void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
1894 struct nfs_client *clp = session->clp;
1898 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1900 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1901 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1903 nfs4_schedule_lease_recovery(clp);
1905 EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
1907 static void nfs4_reset_all_state(struct nfs_client *clp)
1909 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1910 set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1911 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1912 nfs4_state_start_reclaim_nograce(clp);
1913 dprintk("%s: scheduling reset of all state for server %s!\n",
1914 __func__, clp->cl_hostname);
1915 nfs4_schedule_state_manager(clp);
1919 static void nfs41_handle_server_reboot(struct nfs_client *clp)
1921 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1922 nfs4_state_start_reclaim_reboot(clp);
1923 dprintk("%s: server %s rebooted!\n", __func__,
1925 nfs4_schedule_state_manager(clp);
1929 static void nfs41_handle_state_revoked(struct nfs_client *clp)
1931 nfs4_reset_all_state(clp);
1932 dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
1935 static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
1937 /* This will need to handle layouts too */
1938 nfs_expire_all_delegations(clp);
1939 dprintk("%s: Recallable state revoked on server %s!\n", __func__,
1943 static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
1945 nfs_expire_all_delegations(clp);
1946 if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
1947 nfs4_schedule_state_manager(clp);
1948 dprintk("%s: server %s declared a backchannel fault\n", __func__,
1952 static void nfs41_handle_cb_path_down(struct nfs_client *clp)
1954 if (test_and_set_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
1955 &clp->cl_state) == 0)
1956 nfs4_schedule_state_manager(clp);
1959 void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
1964 dprintk("%s: \"%s\" (client ID %llx) flags=0x%08x\n",
1965 __func__, clp->cl_hostname, clp->cl_clientid, flags);
1967 if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
1968 nfs41_handle_server_reboot(clp);
1969 if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
1970 SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
1971 SEQ4_STATUS_ADMIN_STATE_REVOKED |
1972 SEQ4_STATUS_LEASE_MOVED))
1973 nfs41_handle_state_revoked(clp);
1974 if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
1975 nfs41_handle_recallable_state_revoked(clp);
1976 if (flags & SEQ4_STATUS_BACKCHANNEL_FAULT)
1977 nfs41_handle_backchannel_fault(clp);
1978 else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
1979 SEQ4_STATUS_CB_PATH_DOWN_SESSION))
1980 nfs41_handle_cb_path_down(clp);
1983 static int nfs4_reset_session(struct nfs_client *clp)
1985 struct rpc_cred *cred;
1988 if (!nfs4_has_session(clp))
1990 nfs4_begin_drain_session(clp);
1991 cred = nfs4_get_exchange_id_cred(clp);
1992 status = nfs4_proc_destroy_session(clp->cl_session, cred);
1993 if (status && status != -NFS4ERR_BADSESSION &&
1994 status != -NFS4ERR_DEADSESSION) {
1995 status = nfs4_recovery_handle_error(clp, status);
1999 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
2000 status = nfs4_proc_create_session(clp, cred);
2002 dprintk("%s: session reset failed with status %d for server %s!\n",
2003 __func__, status, clp->cl_hostname);
2004 status = nfs4_handle_reclaim_lease_error(clp, status);
2007 nfs41_finish_session_reset(clp);
2008 dprintk("%s: session reset was successful for server %s!\n",
2009 __func__, clp->cl_hostname);
2016 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2018 struct rpc_cred *cred;
2021 if (!nfs4_has_session(clp))
2023 nfs4_begin_drain_session(clp);
2024 cred = nfs4_get_exchange_id_cred(clp);
2025 ret = nfs4_proc_bind_conn_to_session(clp, cred);
2028 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2031 dprintk("%s: bind_conn_to_session was successful for server %s!\n",
2032 __func__, clp->cl_hostname);
2034 case -NFS4ERR_DELAY:
2036 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2039 return nfs4_recovery_handle_error(clp, ret);
2043 #else /* CONFIG_NFS_V4_1 */
2044 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
2045 static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
2047 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2051 #endif /* CONFIG_NFS_V4_1 */
2053 static void nfs4_state_manager(struct nfs_client *clp)
2056 const char *section = "", *section_sep = "";
2058 /* Ensure exclusive access to NFSv4 state */
2060 if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
2061 section = "purge state";
2062 status = nfs4_purge_lease(clp);
2068 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
2069 section = "lease expired";
2070 /* We're going to have to re-establish a clientid */
2071 status = nfs4_reclaim_lease(clp);
2077 /* Initialize or reset the session */
2078 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
2079 section = "reset session";
2080 status = nfs4_reset_session(clp);
2081 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
2087 /* Send BIND_CONN_TO_SESSION */
2088 if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2090 section = "bind conn to session";
2091 status = nfs4_bind_conn_to_session(clp);
2097 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
2098 section = "check lease";
2099 status = nfs4_check_lease(clp);
2105 /* First recover reboot state... */
2106 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
2107 section = "reclaim reboot";
2108 status = nfs4_do_reclaim(clp,
2109 clp->cl_mvops->reboot_recovery_ops);
2110 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2111 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
2113 nfs4_state_end_reclaim_reboot(clp);
2114 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
2120 /* Now recover expired state... */
2121 if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
2122 section = "reclaim nograce";
2123 status = nfs4_do_reclaim(clp,
2124 clp->cl_mvops->nograce_recovery_ops);
2125 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2126 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
2127 test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
2133 nfs4_end_drain_session(clp);
2134 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
2135 nfs_client_return_marked_delegations(clp);
2139 nfs4_clear_state_manager_bit(clp);
2140 /* Did we race with an attempt to give us more work? */
2141 if (clp->cl_state == 0)
2143 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
2145 } while (atomic_read(&clp->cl_count) > 1);
2148 if (strlen(section))
2150 pr_warn_ratelimited("NFS: state manager%s%s failed on NFSv4 server %s"
2151 " with error %d\n", section_sep, section,
2152 clp->cl_hostname, -status);
2154 nfs4_end_drain_session(clp);
2155 nfs4_clear_state_manager_bit(clp);
2158 static int nfs4_run_state_manager(void *ptr)
2160 struct nfs_client *clp = ptr;
2162 allow_signal(SIGKILL);
2163 nfs4_state_manager(clp);
2164 nfs_put_client(clp);
2165 module_put_and_exit(0);