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_RECALL_SLOT, &clp->cl_state);
306 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
307 nfs41_setup_state_renewal(clp);
310 int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
314 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
316 nfs4_begin_drain_session(clp);
317 status = nfs4_proc_exchange_id(clp, cred);
320 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
322 status = nfs4_proc_create_session(clp, cred);
325 nfs41_finish_session_reset(clp);
326 nfs_mark_client_ready(clp, NFS_CS_READY);
332 * nfs41_discover_server_trunking - Detect server IP address trunking (mv1)
334 * @clp: nfs_client under test
335 * @result: OUT: found nfs_client, or clp
336 * @cred: credential to use for trunking test
338 * Returns NFS4_OK, a negative errno, or a negative NFS4ERR status.
339 * If NFS4_OK is returned, an nfs_client pointer is planted in
342 * Note: The returned client may not yet be marked ready.
344 int nfs41_discover_server_trunking(struct nfs_client *clp,
345 struct nfs_client **result,
346 struct rpc_cred *cred)
350 status = nfs4_proc_exchange_id(clp, cred);
351 if (status != NFS4_OK)
353 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
355 return nfs41_walk_client_list(clp, result, cred);
358 struct rpc_cred *nfs4_get_exchange_id_cred(struct nfs_client *clp)
360 struct rpc_cred *cred;
362 spin_lock(&clp->cl_lock);
363 cred = nfs4_get_machine_cred_locked(clp);
364 spin_unlock(&clp->cl_lock);
368 #endif /* CONFIG_NFS_V4_1 */
370 static struct rpc_cred *
371 nfs4_get_setclientid_cred_server(struct nfs_server *server)
373 struct nfs_client *clp = server->nfs_client;
374 struct rpc_cred *cred = NULL;
375 struct nfs4_state_owner *sp;
378 spin_lock(&clp->cl_lock);
379 pos = rb_first(&server->state_owners);
381 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
382 cred = get_rpccred(sp->so_cred);
384 spin_unlock(&clp->cl_lock);
389 * nfs4_get_setclientid_cred - Acquire credential for a setclientid operation
390 * @clp: client state handle
392 * Returns an rpc_cred with reference count bumped, or NULL.
394 struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
396 struct nfs_server *server;
397 struct rpc_cred *cred;
399 spin_lock(&clp->cl_lock);
400 cred = nfs4_get_machine_cred_locked(clp);
401 spin_unlock(&clp->cl_lock);
406 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
407 cred = nfs4_get_setclientid_cred_server(server);
417 static struct nfs4_state_owner *
418 nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
420 struct rb_node **p = &server->state_owners.rb_node,
422 struct nfs4_state_owner *sp;
426 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
428 if (cred < sp->so_cred)
429 p = &parent->rb_left;
430 else if (cred > sp->so_cred)
431 p = &parent->rb_right;
433 if (!list_empty(&sp->so_lru))
434 list_del_init(&sp->so_lru);
435 atomic_inc(&sp->so_count);
442 static struct nfs4_state_owner *
443 nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
445 struct nfs_server *server = new->so_server;
446 struct rb_node **p = &server->state_owners.rb_node,
448 struct nfs4_state_owner *sp;
453 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
455 if (new->so_cred < sp->so_cred)
456 p = &parent->rb_left;
457 else if (new->so_cred > sp->so_cred)
458 p = &parent->rb_right;
460 if (!list_empty(&sp->so_lru))
461 list_del_init(&sp->so_lru);
462 atomic_inc(&sp->so_count);
466 err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
469 rb_link_node(&new->so_server_node, parent, p);
470 rb_insert_color(&new->so_server_node, &server->state_owners);
475 nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
477 struct nfs_server *server = sp->so_server;
479 if (!RB_EMPTY_NODE(&sp->so_server_node))
480 rb_erase(&sp->so_server_node, &server->state_owners);
481 ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
485 nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
487 sc->create_time = ktime_get();
490 spin_lock_init(&sc->lock);
491 INIT_LIST_HEAD(&sc->list);
492 rpc_init_wait_queue(&sc->wait, "Seqid_waitqueue");
496 nfs4_destroy_seqid_counter(struct nfs_seqid_counter *sc)
498 rpc_destroy_wait_queue(&sc->wait);
502 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
503 * create a new state_owner.
506 static struct nfs4_state_owner *
507 nfs4_alloc_state_owner(struct nfs_server *server,
508 struct rpc_cred *cred,
511 struct nfs4_state_owner *sp;
513 sp = kzalloc(sizeof(*sp), gfp_flags);
516 sp->so_server = server;
517 sp->so_cred = get_rpccred(cred);
518 spin_lock_init(&sp->so_lock);
519 INIT_LIST_HEAD(&sp->so_states);
520 nfs4_init_seqid_counter(&sp->so_seqid);
521 atomic_set(&sp->so_count, 1);
522 INIT_LIST_HEAD(&sp->so_lru);
527 nfs4_drop_state_owner(struct nfs4_state_owner *sp)
529 struct rb_node *rb_node = &sp->so_server_node;
531 if (!RB_EMPTY_NODE(rb_node)) {
532 struct nfs_server *server = sp->so_server;
533 struct nfs_client *clp = server->nfs_client;
535 spin_lock(&clp->cl_lock);
536 if (!RB_EMPTY_NODE(rb_node)) {
537 rb_erase(rb_node, &server->state_owners);
538 RB_CLEAR_NODE(rb_node);
540 spin_unlock(&clp->cl_lock);
544 static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
546 nfs4_destroy_seqid_counter(&sp->so_seqid);
547 put_rpccred(sp->so_cred);
551 static void nfs4_gc_state_owners(struct nfs_server *server)
553 struct nfs_client *clp = server->nfs_client;
554 struct nfs4_state_owner *sp, *tmp;
555 unsigned long time_min, time_max;
558 spin_lock(&clp->cl_lock);
560 time_min = (long)time_max - (long)clp->cl_lease_time;
561 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
562 /* NB: LRU is sorted so that oldest is at the head */
563 if (time_in_range(sp->so_expires, time_min, time_max))
565 list_move(&sp->so_lru, &doomed);
566 nfs4_remove_state_owner_locked(sp);
568 spin_unlock(&clp->cl_lock);
570 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
571 list_del(&sp->so_lru);
572 nfs4_free_state_owner(sp);
577 * nfs4_get_state_owner - Look up a state owner given a credential
578 * @server: nfs_server to search
579 * @cred: RPC credential to match
581 * Returns a pointer to an instantiated nfs4_state_owner struct, or NULL.
583 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server,
584 struct rpc_cred *cred,
587 struct nfs_client *clp = server->nfs_client;
588 struct nfs4_state_owner *sp, *new;
590 spin_lock(&clp->cl_lock);
591 sp = nfs4_find_state_owner_locked(server, cred);
592 spin_unlock(&clp->cl_lock);
595 new = nfs4_alloc_state_owner(server, cred, gfp_flags);
599 if (ida_pre_get(&server->openowner_id, gfp_flags) == 0)
601 spin_lock(&clp->cl_lock);
602 sp = nfs4_insert_state_owner_locked(new);
603 spin_unlock(&clp->cl_lock);
604 } while (sp == ERR_PTR(-EAGAIN));
606 nfs4_free_state_owner(new);
608 nfs4_gc_state_owners(server);
613 * nfs4_put_state_owner - Release a nfs4_state_owner
614 * @sp: state owner data to release
616 * Note that we keep released state owners on an LRU
618 * This caches valid state owners so that they can be
619 * reused, to avoid the OPEN_CONFIRM on minor version 0.
620 * It also pins the uniquifier of dropped state owners for
621 * a while, to ensure that those state owner names are
624 void nfs4_put_state_owner(struct nfs4_state_owner *sp)
626 struct nfs_server *server = sp->so_server;
627 struct nfs_client *clp = server->nfs_client;
629 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
632 sp->so_expires = jiffies;
633 list_add_tail(&sp->so_lru, &server->state_owners_lru);
634 spin_unlock(&clp->cl_lock);
638 * nfs4_purge_state_owners - Release all cached state owners
639 * @server: nfs_server with cached state owners to release
641 * Called at umount time. Remaining state owners will be on
642 * the LRU with ref count of zero.
644 void nfs4_purge_state_owners(struct nfs_server *server)
646 struct nfs_client *clp = server->nfs_client;
647 struct nfs4_state_owner *sp, *tmp;
650 spin_lock(&clp->cl_lock);
651 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
652 list_move(&sp->so_lru, &doomed);
653 nfs4_remove_state_owner_locked(sp);
655 spin_unlock(&clp->cl_lock);
657 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
658 list_del(&sp->so_lru);
659 nfs4_free_state_owner(sp);
663 static struct nfs4_state *
664 nfs4_alloc_open_state(void)
666 struct nfs4_state *state;
668 state = kzalloc(sizeof(*state), GFP_NOFS);
671 atomic_set(&state->count, 1);
672 INIT_LIST_HEAD(&state->lock_states);
673 spin_lock_init(&state->state_lock);
674 seqlock_init(&state->seqlock);
679 nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
681 if (state->state == fmode)
683 /* NB! List reordering - see the reclaim code for why. */
684 if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
685 if (fmode & FMODE_WRITE)
686 list_move(&state->open_states, &state->owner->so_states);
688 list_move_tail(&state->open_states, &state->owner->so_states);
690 state->state = fmode;
693 static struct nfs4_state *
694 __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
696 struct nfs_inode *nfsi = NFS_I(inode);
697 struct nfs4_state *state;
699 list_for_each_entry(state, &nfsi->open_states, inode_states) {
700 if (state->owner != owner)
702 if (atomic_inc_not_zero(&state->count))
709 nfs4_free_open_state(struct nfs4_state *state)
715 nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
717 struct nfs4_state *state, *new;
718 struct nfs_inode *nfsi = NFS_I(inode);
720 spin_lock(&inode->i_lock);
721 state = __nfs4_find_state_byowner(inode, owner);
722 spin_unlock(&inode->i_lock);
725 new = nfs4_alloc_open_state();
726 spin_lock(&owner->so_lock);
727 spin_lock(&inode->i_lock);
728 state = __nfs4_find_state_byowner(inode, owner);
729 if (state == NULL && new != NULL) {
731 state->owner = owner;
732 atomic_inc(&owner->so_count);
733 list_add(&state->inode_states, &nfsi->open_states);
735 state->inode = inode;
736 spin_unlock(&inode->i_lock);
737 /* Note: The reclaim code dictates that we add stateless
738 * and read-only stateids to the end of the list */
739 list_add_tail(&state->open_states, &owner->so_states);
740 spin_unlock(&owner->so_lock);
742 spin_unlock(&inode->i_lock);
743 spin_unlock(&owner->so_lock);
745 nfs4_free_open_state(new);
751 void nfs4_put_open_state(struct nfs4_state *state)
753 struct inode *inode = state->inode;
754 struct nfs4_state_owner *owner = state->owner;
756 if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
758 spin_lock(&inode->i_lock);
759 list_del(&state->inode_states);
760 list_del(&state->open_states);
761 spin_unlock(&inode->i_lock);
762 spin_unlock(&owner->so_lock);
764 nfs4_free_open_state(state);
765 nfs4_put_state_owner(owner);
769 * Close the current file.
771 static void __nfs4_close(struct nfs4_state *state,
772 fmode_t fmode, gfp_t gfp_mask, int wait)
774 struct nfs4_state_owner *owner = state->owner;
778 atomic_inc(&owner->so_count);
779 /* Protect against nfs4_find_state() */
780 spin_lock(&owner->so_lock);
781 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
788 case FMODE_READ|FMODE_WRITE:
791 newstate = FMODE_READ|FMODE_WRITE;
792 if (state->n_rdwr == 0) {
793 if (state->n_rdonly == 0) {
794 newstate &= ~FMODE_READ;
795 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
796 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
798 if (state->n_wronly == 0) {
799 newstate &= ~FMODE_WRITE;
800 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
801 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
804 clear_bit(NFS_DELEGATED_STATE, &state->flags);
806 nfs4_state_set_mode_locked(state, newstate);
807 spin_unlock(&owner->so_lock);
810 nfs4_put_open_state(state);
811 nfs4_put_state_owner(owner);
813 nfs4_do_close(state, gfp_mask, wait);
816 void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
818 __nfs4_close(state, fmode, GFP_NOFS, 0);
821 void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
823 __nfs4_close(state, fmode, GFP_KERNEL, 1);
827 * Search the state->lock_states for an existing lock_owner
828 * that is compatible with current->files
830 static struct nfs4_lock_state *
831 __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type)
833 struct nfs4_lock_state *pos;
834 list_for_each_entry(pos, &state->lock_states, ls_locks) {
835 if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
837 switch (pos->ls_owner.lo_type) {
838 case NFS4_POSIX_LOCK_TYPE:
839 if (pos->ls_owner.lo_u.posix_owner != fl_owner)
842 case NFS4_FLOCK_LOCK_TYPE:
843 if (pos->ls_owner.lo_u.flock_owner != fl_pid)
846 atomic_inc(&pos->ls_count);
853 * Return a compatible lock_state. If no initialized lock_state structure
854 * exists, return an uninitialized one.
857 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)
859 struct nfs4_lock_state *lsp;
860 struct nfs_server *server = state->owner->so_server;
862 lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
865 nfs4_init_seqid_counter(&lsp->ls_seqid);
866 atomic_set(&lsp->ls_count, 1);
867 lsp->ls_state = state;
868 lsp->ls_owner.lo_type = type;
869 switch (lsp->ls_owner.lo_type) {
870 case NFS4_FLOCK_LOCK_TYPE:
871 lsp->ls_owner.lo_u.flock_owner = fl_pid;
873 case NFS4_POSIX_LOCK_TYPE:
874 lsp->ls_owner.lo_u.posix_owner = fl_owner;
879 lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
880 if (lsp->ls_seqid.owner_id < 0)
882 INIT_LIST_HEAD(&lsp->ls_locks);
889 void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
891 ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
892 nfs4_destroy_seqid_counter(&lsp->ls_seqid);
897 * Return a compatible lock_state. If no initialized lock_state structure
898 * exists, return an uninitialized one.
901 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner, pid_t pid, unsigned int type)
903 struct nfs4_lock_state *lsp, *new = NULL;
906 spin_lock(&state->state_lock);
907 lsp = __nfs4_find_lock_state(state, owner, pid, type);
911 list_add(&new->ls_locks, &state->lock_states);
912 set_bit(LK_STATE_IN_USE, &state->flags);
917 spin_unlock(&state->state_lock);
918 new = nfs4_alloc_lock_state(state, owner, pid, type);
922 spin_unlock(&state->state_lock);
924 nfs4_free_lock_state(state->owner->so_server, new);
929 * Release reference to lock_state, and free it if we see that
930 * it is no longer in use
932 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
934 struct nfs4_state *state;
938 state = lsp->ls_state;
939 if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
941 list_del(&lsp->ls_locks);
942 if (list_empty(&state->lock_states))
943 clear_bit(LK_STATE_IN_USE, &state->flags);
944 spin_unlock(&state->state_lock);
945 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
946 if (nfs4_release_lockowner(lsp) == 0)
949 nfs4_free_lock_state(lsp->ls_state->owner->so_server, lsp);
952 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
954 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
956 dst->fl_u.nfs4_fl.owner = lsp;
957 atomic_inc(&lsp->ls_count);
960 static void nfs4_fl_release_lock(struct file_lock *fl)
962 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
965 static const struct file_lock_operations nfs4_fl_lock_ops = {
966 .fl_copy_lock = nfs4_fl_copy_lock,
967 .fl_release_private = nfs4_fl_release_lock,
970 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
972 struct nfs4_lock_state *lsp;
974 if (fl->fl_ops != NULL)
976 if (fl->fl_flags & FL_POSIX)
977 lsp = nfs4_get_lock_state(state, fl->fl_owner, 0, NFS4_POSIX_LOCK_TYPE);
978 else if (fl->fl_flags & FL_FLOCK)
979 lsp = nfs4_get_lock_state(state, NULL, fl->fl_pid,
980 NFS4_FLOCK_LOCK_TYPE);
985 fl->fl_u.nfs4_fl.owner = lsp;
986 fl->fl_ops = &nfs4_fl_lock_ops;
990 static bool nfs4_copy_lock_stateid(nfs4_stateid *dst, struct nfs4_state *state,
991 const struct nfs_lockowner *lockowner)
993 struct nfs4_lock_state *lsp;
999 if (lockowner == NULL)
1002 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
1005 fl_owner = lockowner->l_owner;
1006 fl_pid = lockowner->l_pid;
1007 spin_lock(&state->state_lock);
1008 lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
1009 if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
1010 nfs4_stateid_copy(dst, &lsp->ls_stateid);
1013 spin_unlock(&state->state_lock);
1014 nfs4_put_lock_state(lsp);
1019 static void nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
1024 seq = read_seqbegin(&state->seqlock);
1025 nfs4_stateid_copy(dst, &state->stateid);
1026 } while (read_seqretry(&state->seqlock, seq));
1030 * Byte-range lock aware utility to initialize the stateid of read/write
1033 void nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state,
1034 fmode_t fmode, const struct nfs_lockowner *lockowner)
1036 if (nfs4_copy_delegation_stateid(dst, state->inode, fmode))
1038 if (nfs4_copy_lock_stateid(dst, state, lockowner))
1040 nfs4_copy_open_stateid(dst, state);
1043 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
1045 struct nfs_seqid *new;
1047 new = kmalloc(sizeof(*new), gfp_mask);
1049 new->sequence = counter;
1050 INIT_LIST_HEAD(&new->list);
1056 void nfs_release_seqid(struct nfs_seqid *seqid)
1058 struct nfs_seqid_counter *sequence;
1060 if (list_empty(&seqid->list))
1062 sequence = seqid->sequence;
1063 spin_lock(&sequence->lock);
1064 list_del_init(&seqid->list);
1065 if (!list_empty(&sequence->list)) {
1066 struct nfs_seqid *next;
1068 next = list_first_entry(&sequence->list,
1069 struct nfs_seqid, list);
1070 rpc_wake_up_queued_task(&sequence->wait, next->task);
1072 spin_unlock(&sequence->lock);
1075 void nfs_free_seqid(struct nfs_seqid *seqid)
1077 nfs_release_seqid(seqid);
1082 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
1083 * failed with a seqid incrementing error -
1084 * see comments nfs_fs.h:seqid_mutating_error()
1086 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
1091 case -NFS4ERR_BAD_SEQID:
1092 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
1094 pr_warn_ratelimited("NFS: v4 server returned a bad"
1095 " sequence-id error on an"
1096 " unconfirmed sequence %p!\n",
1098 case -NFS4ERR_STALE_CLIENTID:
1099 case -NFS4ERR_STALE_STATEID:
1100 case -NFS4ERR_BAD_STATEID:
1101 case -NFS4ERR_BADXDR:
1102 case -NFS4ERR_RESOURCE:
1103 case -NFS4ERR_NOFILEHANDLE:
1104 /* Non-seqid mutating errors */
1108 * Note: no locking needed as we are guaranteed to be first
1109 * on the sequence list
1111 seqid->sequence->counter++;
1114 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
1116 struct nfs4_state_owner *sp = container_of(seqid->sequence,
1117 struct nfs4_state_owner, so_seqid);
1118 struct nfs_server *server = sp->so_server;
1120 if (status == -NFS4ERR_BAD_SEQID)
1121 nfs4_drop_state_owner(sp);
1122 if (!nfs4_has_session(server->nfs_client))
1123 nfs_increment_seqid(status, seqid);
1127 * Increment the seqid if the LOCK/LOCKU succeeded, or
1128 * failed with a seqid incrementing error -
1129 * see comments nfs_fs.h:seqid_mutating_error()
1131 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
1133 nfs_increment_seqid(status, seqid);
1136 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
1138 struct nfs_seqid_counter *sequence = seqid->sequence;
1141 spin_lock(&sequence->lock);
1143 if (list_empty(&seqid->list))
1144 list_add_tail(&seqid->list, &sequence->list);
1145 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
1147 rpc_sleep_on(&sequence->wait, task, NULL);
1150 spin_unlock(&sequence->lock);
1154 static int nfs4_run_state_manager(void *);
1156 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
1158 smp_mb__before_clear_bit();
1159 clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
1160 smp_mb__after_clear_bit();
1161 wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
1162 rpc_wake_up(&clp->cl_rpcwaitq);
1166 * Schedule the nfs_client asynchronous state management routine
1168 void nfs4_schedule_state_manager(struct nfs_client *clp)
1170 struct task_struct *task;
1171 char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
1173 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1175 __module_get(THIS_MODULE);
1176 atomic_inc(&clp->cl_count);
1178 /* The rcu_read_lock() is not strictly necessary, as the state
1179 * manager is the only thread that ever changes the rpc_xprt
1180 * after it's initialized. At this point, we're single threaded. */
1182 snprintf(buf, sizeof(buf), "%s-manager",
1183 rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR));
1185 task = kthread_run(nfs4_run_state_manager, clp, buf);
1187 printk(KERN_ERR "%s: kthread_run: %ld\n",
1188 __func__, PTR_ERR(task));
1189 nfs4_clear_state_manager_bit(clp);
1190 nfs_put_client(clp);
1191 module_put(THIS_MODULE);
1196 * Schedule a lease recovery attempt
1198 void nfs4_schedule_lease_recovery(struct nfs_client *clp)
1202 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1203 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1204 dprintk("%s: scheduling lease recovery for server %s\n", __func__,
1206 nfs4_schedule_state_manager(clp);
1208 EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
1211 * nfs40_handle_cb_pathdown - return all delegations after NFS4ERR_CB_PATH_DOWN
1212 * @clp: client to process
1214 * Set the NFS4CLNT_LEASE_EXPIRED state in order to force a
1215 * resend of the SETCLIENTID and hence re-establish the
1216 * callback channel. Then return all existing delegations.
1218 static void nfs40_handle_cb_pathdown(struct nfs_client *clp)
1220 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1221 nfs_expire_all_delegations(clp);
1222 dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
1226 void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
1228 nfs40_handle_cb_pathdown(clp);
1229 nfs4_schedule_state_manager(clp);
1232 static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1235 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1236 /* Don't recover state that expired before the reboot */
1237 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
1238 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1241 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1242 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1246 static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1248 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
1249 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1250 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1251 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
1255 void nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
1257 struct nfs_client *clp = server->nfs_client;
1259 nfs4_state_mark_reclaim_nograce(clp, state);
1260 dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
1262 nfs4_schedule_state_manager(clp);
1264 EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1266 void nfs_inode_find_state_and_recover(struct inode *inode,
1267 const nfs4_stateid *stateid)
1269 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1270 struct nfs_inode *nfsi = NFS_I(inode);
1271 struct nfs_open_context *ctx;
1272 struct nfs4_state *state;
1275 spin_lock(&inode->i_lock);
1276 list_for_each_entry(ctx, &nfsi->open_files, list) {
1280 if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
1282 if (!nfs4_stateid_match(&state->stateid, stateid))
1284 nfs4_state_mark_reclaim_nograce(clp, state);
1287 spin_unlock(&inode->i_lock);
1289 nfs4_schedule_state_manager(clp);
1293 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
1295 struct inode *inode = state->inode;
1296 struct nfs_inode *nfsi = NFS_I(inode);
1297 struct file_lock *fl;
1300 if (inode->i_flock == NULL)
1303 /* Guard against delegation returns and new lock/unlock calls */
1304 down_write(&nfsi->rwsem);
1305 /* Protect inode->i_flock using the BKL */
1307 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1308 if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
1310 if (nfs_file_open_context(fl->fl_file)->state != state)
1313 status = ops->recover_lock(state, fl);
1318 case -NFS4ERR_ADMIN_REVOKED:
1319 case -NFS4ERR_STALE_STATEID:
1320 case -NFS4ERR_BAD_STATEID:
1321 case -NFS4ERR_EXPIRED:
1322 case -NFS4ERR_NO_GRACE:
1323 case -NFS4ERR_STALE_CLIENTID:
1324 case -NFS4ERR_BADSESSION:
1325 case -NFS4ERR_BADSLOT:
1326 case -NFS4ERR_BAD_HIGH_SLOT:
1327 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1330 printk(KERN_ERR "NFS: %s: unhandled error %d. "
1331 "Zeroing state\n", __func__, status);
1333 case -NFS4ERR_DENIED:
1334 case -NFS4ERR_RECLAIM_BAD:
1335 case -NFS4ERR_RECLAIM_CONFLICT:
1336 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1343 up_write(&nfsi->rwsem);
1347 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
1349 struct nfs4_state *state;
1350 struct nfs4_lock_state *lock;
1353 /* Note: we rely on the sp->so_states list being ordered
1354 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
1356 * This is needed to ensure that the server won't give us any
1357 * read delegations that we have to return if, say, we are
1358 * recovering after a network partition or a reboot from a
1359 * server that doesn't support a grace period.
1362 spin_lock(&sp->so_lock);
1363 list_for_each_entry(state, &sp->so_states, open_states) {
1364 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
1366 if (state->state == 0)
1368 atomic_inc(&state->count);
1369 spin_unlock(&sp->so_lock);
1370 status = ops->recover_open(sp, state);
1372 status = nfs4_reclaim_locks(state, ops);
1374 spin_lock(&state->state_lock);
1375 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1376 if (!test_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags))
1377 pr_warn_ratelimited("NFS: "
1379 "failed!\n", __func__);
1381 spin_unlock(&state->state_lock);
1382 nfs4_put_open_state(state);
1388 printk(KERN_ERR "NFS: %s: unhandled error %d. "
1389 "Zeroing state\n", __func__, status);
1394 * Open state on this file cannot be recovered
1395 * All we can do is revert to using the zero stateid.
1397 memset(&state->stateid, 0,
1398 sizeof(state->stateid));
1399 /* Mark the file as being 'closed' */
1404 * User RPCSEC_GSS context has expired.
1405 * We cannot recover this stateid now, so
1406 * skip it and allow recovery thread to
1410 case -NFS4ERR_ADMIN_REVOKED:
1411 case -NFS4ERR_STALE_STATEID:
1412 case -NFS4ERR_BAD_STATEID:
1413 case -NFS4ERR_RECLAIM_BAD:
1414 case -NFS4ERR_RECLAIM_CONFLICT:
1415 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1417 case -NFS4ERR_EXPIRED:
1418 case -NFS4ERR_NO_GRACE:
1419 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1420 case -NFS4ERR_STALE_CLIENTID:
1421 case -NFS4ERR_BADSESSION:
1422 case -NFS4ERR_BADSLOT:
1423 case -NFS4ERR_BAD_HIGH_SLOT:
1424 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1427 nfs4_put_open_state(state);
1430 spin_unlock(&sp->so_lock);
1433 nfs4_put_open_state(state);
1437 static void nfs4_clear_open_state(struct nfs4_state *state)
1439 struct nfs4_lock_state *lock;
1441 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1442 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1443 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1444 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1445 spin_lock(&state->state_lock);
1446 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1447 lock->ls_seqid.flags = 0;
1448 clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
1450 spin_unlock(&state->state_lock);
1453 static void nfs4_reset_seqids(struct nfs_server *server,
1454 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1456 struct nfs_client *clp = server->nfs_client;
1457 struct nfs4_state_owner *sp;
1458 struct rb_node *pos;
1459 struct nfs4_state *state;
1461 spin_lock(&clp->cl_lock);
1462 for (pos = rb_first(&server->state_owners);
1464 pos = rb_next(pos)) {
1465 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1466 sp->so_seqid.flags = 0;
1467 spin_lock(&sp->so_lock);
1468 list_for_each_entry(state, &sp->so_states, open_states) {
1469 if (mark_reclaim(clp, state))
1470 nfs4_clear_open_state(state);
1472 spin_unlock(&sp->so_lock);
1474 spin_unlock(&clp->cl_lock);
1477 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp,
1478 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1480 struct nfs_server *server;
1483 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1484 nfs4_reset_seqids(server, mark_reclaim);
1488 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1490 /* Mark all delegations for reclaim */
1491 nfs_delegation_mark_reclaim(clp);
1492 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1495 static void nfs4_reclaim_complete(struct nfs_client *clp,
1496 const struct nfs4_state_recovery_ops *ops)
1498 /* Notify the server we're done reclaiming our state */
1499 if (ops->reclaim_complete)
1500 (void)ops->reclaim_complete(clp);
1503 static void nfs4_clear_reclaim_server(struct nfs_server *server)
1505 struct nfs_client *clp = server->nfs_client;
1506 struct nfs4_state_owner *sp;
1507 struct rb_node *pos;
1508 struct nfs4_state *state;
1510 spin_lock(&clp->cl_lock);
1511 for (pos = rb_first(&server->state_owners);
1513 pos = rb_next(pos)) {
1514 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1515 spin_lock(&sp->so_lock);
1516 list_for_each_entry(state, &sp->so_states, open_states) {
1517 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
1520 nfs4_state_mark_reclaim_nograce(clp, state);
1522 spin_unlock(&sp->so_lock);
1524 spin_unlock(&clp->cl_lock);
1527 static int nfs4_state_clear_reclaim_reboot(struct nfs_client *clp)
1529 struct nfs_server *server;
1531 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1535 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1536 nfs4_clear_reclaim_server(server);
1539 nfs_delegation_reap_unclaimed(clp);
1543 static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
1545 if (!nfs4_state_clear_reclaim_reboot(clp))
1547 nfs4_reclaim_complete(clp, clp->cl_mvops->reboot_recovery_ops);
1550 static void nfs_delegation_clear_all(struct nfs_client *clp)
1552 nfs_delegation_mark_reclaim(clp);
1553 nfs_delegation_reap_unclaimed(clp);
1556 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1558 nfs_delegation_clear_all(clp);
1559 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1562 static void nfs4_warn_keyexpired(const char *s)
1564 printk_ratelimited(KERN_WARNING "Error: state manager"
1565 " encountered RPCSEC_GSS session"
1566 " expired against NFSv4 server %s.\n",
1570 static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1575 case -NFS4ERR_CB_PATH_DOWN:
1576 nfs40_handle_cb_pathdown(clp);
1578 case -NFS4ERR_NO_GRACE:
1579 nfs4_state_end_reclaim_reboot(clp);
1581 case -NFS4ERR_STALE_CLIENTID:
1582 case -NFS4ERR_LEASE_MOVED:
1583 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1584 nfs4_state_clear_reclaim_reboot(clp);
1585 nfs4_state_start_reclaim_reboot(clp);
1587 case -NFS4ERR_EXPIRED:
1588 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1589 nfs4_state_start_reclaim_nograce(clp);
1591 case -NFS4ERR_BADSESSION:
1592 case -NFS4ERR_BADSLOT:
1593 case -NFS4ERR_BAD_HIGH_SLOT:
1594 case -NFS4ERR_DEADSESSION:
1595 case -NFS4ERR_SEQ_FALSE_RETRY:
1596 case -NFS4ERR_SEQ_MISORDERED:
1597 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1598 /* Zero session reset errors */
1600 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1601 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1604 /* Nothing we can do */
1605 nfs4_warn_keyexpired(clp->cl_hostname);
1608 dprintk("%s: failed to handle error %d for server %s\n",
1609 __func__, error, clp->cl_hostname);
1612 dprintk("%s: handled error %d for server %s\n", __func__, error,
1617 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1619 struct nfs4_state_owner *sp;
1620 struct nfs_server *server;
1621 struct rb_node *pos;
1626 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1627 nfs4_purge_state_owners(server);
1628 spin_lock(&clp->cl_lock);
1629 for (pos = rb_first(&server->state_owners);
1631 pos = rb_next(pos)) {
1633 struct nfs4_state_owner, so_server_node);
1634 if (!test_and_clear_bit(ops->owner_flag_bit,
1637 atomic_inc(&sp->so_count);
1638 spin_unlock(&clp->cl_lock);
1641 status = nfs4_reclaim_open_state(sp, ops);
1643 set_bit(ops->owner_flag_bit, &sp->so_flags);
1644 nfs4_put_state_owner(sp);
1645 return nfs4_recovery_handle_error(clp, status);
1648 nfs4_put_state_owner(sp);
1651 spin_unlock(&clp->cl_lock);
1657 static int nfs4_check_lease(struct nfs_client *clp)
1659 struct rpc_cred *cred;
1660 const struct nfs4_state_maintenance_ops *ops =
1661 clp->cl_mvops->state_renewal_ops;
1664 /* Is the client already known to have an expired lease? */
1665 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1667 spin_lock(&clp->cl_lock);
1668 cred = ops->get_state_renewal_cred_locked(clp);
1669 spin_unlock(&clp->cl_lock);
1671 cred = nfs4_get_setclientid_cred(clp);
1676 status = ops->renew_lease(clp, cred);
1679 return nfs4_recovery_handle_error(clp, status);
1682 /* Set NFS4CLNT_LEASE_EXPIRED and reclaim reboot state for all v4.0 errors
1683 * and for recoverable errors on EXCHANGE_ID for v4.1
1685 static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
1688 case -NFS4ERR_SEQ_MISORDERED:
1689 if (test_and_set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state))
1690 return -ESERVERFAULT;
1691 /* Lease confirmation error: retry after purging the lease */
1693 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1695 case -NFS4ERR_STALE_CLIENTID:
1696 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1697 nfs4_state_clear_reclaim_reboot(clp);
1698 nfs4_state_start_reclaim_reboot(clp);
1700 case -NFS4ERR_CLID_INUSE:
1701 pr_err("NFS: Server %s reports our clientid is in use\n",
1703 nfs_mark_client_ready(clp, -EPERM);
1704 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1707 if (clp->cl_machine_cred == NULL)
1709 /* Handle case where the user hasn't set up machine creds */
1710 nfs4_clear_machine_cred(clp);
1711 case -NFS4ERR_DELAY:
1717 case -NFS4ERR_MINOR_VERS_MISMATCH:
1718 if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
1719 nfs_mark_client_ready(clp, -EPROTONOSUPPORT);
1720 dprintk("%s: exit with error %d for server %s\n",
1721 __func__, -EPROTONOSUPPORT, clp->cl_hostname);
1722 return -EPROTONOSUPPORT;
1724 nfs4_warn_keyexpired(clp->cl_hostname);
1725 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1726 * in nfs4_exchange_id */
1728 dprintk("%s: exit with error %d for server %s\n", __func__,
1729 status, clp->cl_hostname);
1732 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1733 dprintk("%s: handled error %d for server %s\n", __func__, status,
1738 static int nfs4_establish_lease(struct nfs_client *clp)
1740 struct rpc_cred *cred;
1741 const struct nfs4_state_recovery_ops *ops =
1742 clp->cl_mvops->reboot_recovery_ops;
1745 cred = ops->get_clid_cred(clp);
1748 status = ops->establish_clid(clp, cred);
1752 pnfs_destroy_all_layouts(clp);
1757 * Returns zero or a negative errno. NFS4ERR values are converted
1758 * to local errno values.
1760 static int nfs4_reclaim_lease(struct nfs_client *clp)
1764 status = nfs4_establish_lease(clp);
1766 return nfs4_handle_reclaim_lease_error(clp, status);
1767 if (test_and_clear_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state))
1768 nfs4_state_start_reclaim_nograce(clp);
1769 if (!test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
1770 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1771 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1772 clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1776 static int nfs4_purge_lease(struct nfs_client *clp)
1780 status = nfs4_establish_lease(clp);
1782 return nfs4_handle_reclaim_lease_error(clp, status);
1783 clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1784 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1785 nfs4_state_start_reclaim_nograce(clp);
1790 * nfs4_discover_server_trunking - Detect server IP address trunking
1792 * @clp: nfs_client under test
1793 * @result: OUT: found nfs_client, or clp
1795 * Returns zero or a negative errno. If zero is returned,
1796 * an nfs_client pointer is planted in "result".
1798 * Note: since we are invoked in process context, and
1799 * not from inside the state manager, we cannot use
1800 * nfs4_handle_reclaim_lease_error().
1802 int nfs4_discover_server_trunking(struct nfs_client *clp,
1803 struct nfs_client **result)
1805 const struct nfs4_state_recovery_ops *ops =
1806 clp->cl_mvops->reboot_recovery_ops;
1807 rpc_authflavor_t *flavors, flav, save;
1808 struct rpc_clnt *clnt;
1809 struct rpc_cred *cred;
1812 dprintk("NFS: %s: testing '%s'\n", __func__, clp->cl_hostname);
1814 len = NFS_MAX_SECFLAVORS;
1815 flavors = kcalloc(len, sizeof(*flavors), GFP_KERNEL);
1816 if (flavors == NULL) {
1820 len = rpcauth_list_flavors(flavors, len);
1825 clnt = clp->cl_rpcclient;
1826 save = clnt->cl_auth->au_flavor;
1829 mutex_lock(&nfs_clid_init_mutex);
1832 cred = ops->get_clid_cred(clp);
1836 status = ops->detect_trunking(clp, result, cred);
1843 if (clp->cl_machine_cred == NULL)
1845 /* Handle case where the user hasn't set up machine creds */
1846 nfs4_clear_machine_cred(clp);
1847 case -NFS4ERR_DELAY:
1851 dprintk("NFS: %s after status %d, retrying\n",
1855 case -NFS4ERR_CLID_INUSE:
1856 case -NFS4ERR_WRONGSEC:
1861 flav = flavors[i++];
1863 flav = flavors[i++];
1864 clnt = rpc_clone_client_set_auth(clnt, flav);
1866 status = PTR_ERR(clnt);
1869 clp->cl_rpcclient = clnt;
1872 case -NFS4ERR_MINOR_VERS_MISMATCH:
1873 status = -EPROTONOSUPPORT;
1877 nfs4_warn_keyexpired(clp->cl_hostname);
1878 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1879 * in nfs4_exchange_id */
1880 status = -EKEYEXPIRED;
1884 mutex_unlock(&nfs_clid_init_mutex);
1888 dprintk("NFS: %s: status = %d\n", __func__, status);
1892 #ifdef CONFIG_NFS_V4_1
1893 void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
1895 struct nfs_client *clp = session->clp;
1899 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1901 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1902 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1904 nfs4_schedule_lease_recovery(clp);
1906 EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
1908 void nfs41_handle_recall_slot(struct nfs_client *clp)
1910 set_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
1911 dprintk("%s: scheduling slot recall for server %s\n", __func__,
1913 nfs4_schedule_state_manager(clp);
1916 static void nfs4_reset_all_state(struct nfs_client *clp)
1918 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1919 set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1920 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1921 nfs4_state_start_reclaim_nograce(clp);
1922 dprintk("%s: scheduling reset of all state for server %s!\n",
1923 __func__, clp->cl_hostname);
1924 nfs4_schedule_state_manager(clp);
1928 static void nfs41_handle_server_reboot(struct nfs_client *clp)
1930 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1931 nfs4_state_start_reclaim_reboot(clp);
1932 dprintk("%s: server %s rebooted!\n", __func__,
1934 nfs4_schedule_state_manager(clp);
1938 static void nfs41_handle_state_revoked(struct nfs_client *clp)
1940 nfs4_reset_all_state(clp);
1941 dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
1944 static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
1946 /* This will need to handle layouts too */
1947 nfs_expire_all_delegations(clp);
1948 dprintk("%s: Recallable state revoked on server %s!\n", __func__,
1952 static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
1954 nfs_expire_all_delegations(clp);
1955 if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
1956 nfs4_schedule_state_manager(clp);
1957 dprintk("%s: server %s declared a backchannel fault\n", __func__,
1961 static void nfs41_handle_cb_path_down(struct nfs_client *clp)
1963 if (test_and_set_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
1964 &clp->cl_state) == 0)
1965 nfs4_schedule_state_manager(clp);
1968 void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
1973 dprintk("%s: \"%s\" (client ID %llx) flags=0x%08x\n",
1974 __func__, clp->cl_hostname, clp->cl_clientid, flags);
1976 if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
1977 nfs41_handle_server_reboot(clp);
1978 if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
1979 SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
1980 SEQ4_STATUS_ADMIN_STATE_REVOKED |
1981 SEQ4_STATUS_LEASE_MOVED))
1982 nfs41_handle_state_revoked(clp);
1983 if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
1984 nfs41_handle_recallable_state_revoked(clp);
1985 if (flags & SEQ4_STATUS_BACKCHANNEL_FAULT)
1986 nfs41_handle_backchannel_fault(clp);
1987 else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
1988 SEQ4_STATUS_CB_PATH_DOWN_SESSION))
1989 nfs41_handle_cb_path_down(clp);
1992 static int nfs4_reset_session(struct nfs_client *clp)
1994 struct rpc_cred *cred;
1997 if (!nfs4_has_session(clp))
1999 nfs4_begin_drain_session(clp);
2000 cred = nfs4_get_exchange_id_cred(clp);
2001 status = nfs4_proc_destroy_session(clp->cl_session, cred);
2002 if (status && status != -NFS4ERR_BADSESSION &&
2003 status != -NFS4ERR_DEADSESSION) {
2004 status = nfs4_recovery_handle_error(clp, status);
2008 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
2009 status = nfs4_proc_create_session(clp, cred);
2011 dprintk("%s: session reset failed with status %d for server %s!\n",
2012 __func__, status, clp->cl_hostname);
2013 status = nfs4_handle_reclaim_lease_error(clp, status);
2016 nfs41_finish_session_reset(clp);
2017 dprintk("%s: session reset was successful for server %s!\n",
2018 __func__, clp->cl_hostname);
2025 static int nfs4_recall_slot(struct nfs_client *clp)
2027 struct nfs4_slot_table *fc_tbl;
2028 struct nfs4_slot *new, *old;
2031 if (!nfs4_has_session(clp))
2033 nfs4_begin_drain_session(clp);
2034 fc_tbl = &clp->cl_session->fc_slot_table;
2035 new = nfs4_alloc_slots(fc_tbl, fc_tbl->target_highest_slotid + 1, GFP_NOFS);
2039 spin_lock(&fc_tbl->slot_tbl_lock);
2040 for (i = 0; i <= fc_tbl->target_highest_slotid; i++)
2041 new[i].seq_nr = fc_tbl->slots[i].seq_nr;
2042 old = fc_tbl->slots;
2043 fc_tbl->slots = new;
2044 fc_tbl->max_slots = fc_tbl->target_highest_slotid + 1;
2045 fc_tbl->max_slotid = fc_tbl->target_highest_slotid;
2046 clp->cl_session->fc_attrs.max_reqs = fc_tbl->max_slots;
2047 spin_unlock(&fc_tbl->slot_tbl_lock);
2053 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2055 struct rpc_cred *cred;
2058 if (!nfs4_has_session(clp))
2060 nfs4_begin_drain_session(clp);
2061 cred = nfs4_get_exchange_id_cred(clp);
2062 ret = nfs4_proc_bind_conn_to_session(clp, cred);
2065 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2068 dprintk("%s: bind_conn_to_session was successful for server %s!\n",
2069 __func__, clp->cl_hostname);
2071 case -NFS4ERR_DELAY:
2073 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2076 return nfs4_recovery_handle_error(clp, ret);
2080 #else /* CONFIG_NFS_V4_1 */
2081 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
2082 static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
2083 static int nfs4_recall_slot(struct nfs_client *clp) { return 0; }
2085 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2089 #endif /* CONFIG_NFS_V4_1 */
2091 static void nfs4_state_manager(struct nfs_client *clp)
2094 const char *section = "", *section_sep = "";
2096 /* Ensure exclusive access to NFSv4 state */
2098 if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
2099 section = "purge state";
2100 status = nfs4_purge_lease(clp);
2106 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
2107 section = "lease expired";
2108 /* We're going to have to re-establish a clientid */
2109 status = nfs4_reclaim_lease(clp);
2115 /* Initialize or reset the session */
2116 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
2117 section = "reset session";
2118 status = nfs4_reset_session(clp);
2119 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
2125 /* Send BIND_CONN_TO_SESSION */
2126 if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2128 section = "bind conn to session";
2129 status = nfs4_bind_conn_to_session(clp);
2135 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
2136 section = "check lease";
2137 status = nfs4_check_lease(clp);
2143 /* Recall session slots */
2144 if (test_and_clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state)) {
2145 section = "recall slot";
2146 status = nfs4_recall_slot(clp);
2152 /* First recover reboot state... */
2153 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
2154 section = "reclaim reboot";
2155 status = nfs4_do_reclaim(clp,
2156 clp->cl_mvops->reboot_recovery_ops);
2157 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2158 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
2160 nfs4_state_end_reclaim_reboot(clp);
2161 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
2167 /* Now recover expired state... */
2168 if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
2169 section = "reclaim nograce";
2170 status = nfs4_do_reclaim(clp,
2171 clp->cl_mvops->nograce_recovery_ops);
2172 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2173 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
2174 test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
2180 nfs4_end_drain_session(clp);
2181 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
2182 nfs_client_return_marked_delegations(clp);
2186 nfs4_clear_state_manager_bit(clp);
2187 /* Did we race with an attempt to give us more work? */
2188 if (clp->cl_state == 0)
2190 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
2192 } while (atomic_read(&clp->cl_count) > 1);
2195 if (strlen(section))
2197 pr_warn_ratelimited("NFS: state manager%s%s failed on NFSv4 server %s"
2198 " with error %d\n", section_sep, section,
2199 clp->cl_hostname, -status);
2201 nfs4_end_drain_session(clp);
2202 nfs4_clear_state_manager_bit(clp);
2205 static int nfs4_run_state_manager(void *ptr)
2207 struct nfs_client *clp = ptr;
2209 allow_signal(SIGKILL);
2210 nfs4_state_manager(clp);
2211 nfs_put_client(clp);
2212 module_put_and_exit(0);