4 * Client-side XDR for NFSv4.
6 * Copyright (c) 2002 The Regents of the University of Michigan.
9 * Kendrick Smith <kmsmith@umich.edu>
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its
21 * contributors may be used to endorse or promote products derived
22 * from this software without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
25 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
26 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
27 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
31 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
32 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
33 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
34 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36 * Implementation of the NFSv4 state model. For the time being,
37 * this is minimal, but will be made much more complex in a
41 #include <linux/kernel.h>
42 #include <linux/slab.h>
44 #include <linux/nfs_fs.h>
45 #include <linux/nfs_idmap.h>
46 #include <linux/kthread.h>
47 #include <linux/module.h>
48 #include <linux/random.h>
49 #include <linux/ratelimit.h>
50 #include <linux/workqueue.h>
51 #include <linux/bitops.h>
52 #include <linux/jiffies.h>
54 #include <linux/sunrpc/clnt.h>
58 #include "delegation.h"
60 #include "nfs4session.h"
64 #define NFSDBG_FACILITY NFSDBG_STATE
66 #define OPENOWNER_POOL_SIZE 8
68 const nfs4_stateid zero_stateid;
69 static DEFINE_MUTEX(nfs_clid_init_mutex);
70 static LIST_HEAD(nfs4_clientid_list);
72 int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
74 struct nfs4_setclientid_res clid = {
75 .clientid = clp->cl_clientid,
76 .confirm = clp->cl_confirm,
80 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
82 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
84 port = nn->nfs_callback_tcpport;
85 if (clp->cl_addr.ss_family == AF_INET6)
86 port = nn->nfs_callback_tcpport6;
88 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
91 clp->cl_clientid = clid.clientid;
92 clp->cl_confirm = clid.confirm;
93 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
95 status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
98 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
99 nfs4_schedule_state_renewal(clp);
105 * nfs40_discover_server_trunking - Detect server IP address trunking (mv0)
107 * @clp: nfs_client under test
108 * @result: OUT: found nfs_client, or clp
109 * @cred: credential to use for trunking test
111 * Returns zero, a negative errno, or a negative NFS4ERR status.
112 * If zero is returned, an nfs_client pointer is planted in
115 * Note: The returned client may not yet be marked ready.
117 int nfs40_discover_server_trunking(struct nfs_client *clp,
118 struct nfs_client **result,
119 struct rpc_cred *cred)
121 struct nfs4_setclientid_res clid = {
122 .clientid = clp->cl_clientid,
123 .confirm = clp->cl_confirm,
125 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
129 port = nn->nfs_callback_tcpport;
130 if (clp->cl_addr.ss_family == AF_INET6)
131 port = nn->nfs_callback_tcpport6;
133 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
136 clp->cl_clientid = clid.clientid;
137 clp->cl_confirm = clid.confirm;
139 status = nfs40_walk_client_list(clp, result, cred);
141 case -NFS4ERR_STALE_CLIENTID:
142 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
144 /* Sustain the lease, even if it's empty. If the clientid4
145 * goes stale it's of no use for trunking discovery. */
146 nfs4_schedule_state_renewal(*result);
154 struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
156 struct rpc_cred *cred = NULL;
158 if (clp->cl_machine_cred != NULL)
159 cred = get_rpccred(clp->cl_machine_cred);
163 static void nfs4_clear_machine_cred(struct nfs_client *clp)
165 struct rpc_cred *cred;
167 spin_lock(&clp->cl_lock);
168 cred = clp->cl_machine_cred;
169 clp->cl_machine_cred = NULL;
170 spin_unlock(&clp->cl_lock);
175 static struct rpc_cred *
176 nfs4_get_renew_cred_server_locked(struct nfs_server *server)
178 struct rpc_cred *cred = NULL;
179 struct nfs4_state_owner *sp;
182 for (pos = rb_first(&server->state_owners);
184 pos = rb_next(pos)) {
185 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
186 if (list_empty(&sp->so_states))
188 cred = get_rpccred(sp->so_cred);
195 * nfs4_get_renew_cred_locked - Acquire credential for a renew operation
196 * @clp: client state handle
198 * Returns an rpc_cred with reference count bumped, or NULL.
199 * Caller must hold clp->cl_lock.
201 struct rpc_cred *nfs4_get_renew_cred_locked(struct nfs_client *clp)
203 struct rpc_cred *cred = NULL;
204 struct nfs_server *server;
206 /* Use machine credentials if available */
207 cred = nfs4_get_machine_cred_locked(clp);
212 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
213 cred = nfs4_get_renew_cred_server_locked(server);
223 #if defined(CONFIG_NFS_V4_1)
225 static int nfs41_setup_state_renewal(struct nfs_client *clp)
228 struct nfs_fsinfo fsinfo;
230 if (!test_bit(NFS_CS_CHECK_LEASE_TIME, &clp->cl_res_state)) {
231 nfs4_schedule_state_renewal(clp);
235 status = nfs4_proc_get_lease_time(clp, &fsinfo);
237 /* Update lease time and schedule renewal */
238 spin_lock(&clp->cl_lock);
239 clp->cl_lease_time = fsinfo.lease_time * HZ;
240 clp->cl_last_renewal = jiffies;
241 spin_unlock(&clp->cl_lock);
243 nfs4_schedule_state_renewal(clp);
250 * Back channel returns NFS4ERR_DELAY for new requests when
251 * NFS4_SESSION_DRAINING is set so there is no work to be done when draining
254 static void nfs4_end_drain_session(struct nfs_client *clp)
256 struct nfs4_session *ses = clp->cl_session;
257 struct nfs4_slot_table *tbl;
262 tbl = &ses->fc_slot_table;
263 if (test_and_clear_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
264 spin_lock(&tbl->slot_tbl_lock);
265 for (i = 0; i <= tbl->max_slotid; i++) {
266 if (rpc_wake_up_first(&tbl->slot_tbl_waitq,
267 nfs4_set_task_privileged,
271 spin_unlock(&tbl->slot_tbl_lock);
276 * Signal state manager thread if session fore channel is drained
278 void nfs4_session_drain_complete(struct nfs4_session *session,
279 struct nfs4_slot_table *tbl)
281 if (nfs4_session_draining(session))
282 complete(&tbl->complete);
285 static int nfs4_wait_on_slot_tbl(struct nfs4_slot_table *tbl)
287 spin_lock(&tbl->slot_tbl_lock);
288 if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
289 INIT_COMPLETION(tbl->complete);
290 spin_unlock(&tbl->slot_tbl_lock);
291 return wait_for_completion_interruptible(&tbl->complete);
293 spin_unlock(&tbl->slot_tbl_lock);
297 static int nfs4_begin_drain_session(struct nfs_client *clp)
299 struct nfs4_session *ses = clp->cl_session;
302 set_bit(NFS4_SESSION_DRAINING, &ses->session_state);
304 ret = nfs4_wait_on_slot_tbl(&ses->bc_slot_table);
308 return nfs4_wait_on_slot_tbl(&ses->fc_slot_table);
311 static void nfs41_finish_session_reset(struct nfs_client *clp)
313 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
314 clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
315 /* create_session negotiated new slot table */
316 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
317 nfs41_setup_state_renewal(clp);
320 int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
324 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
326 nfs4_begin_drain_session(clp);
327 status = nfs4_proc_exchange_id(clp, cred);
330 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
332 status = nfs4_proc_create_session(clp, cred);
335 nfs41_finish_session_reset(clp);
336 nfs_mark_client_ready(clp, NFS_CS_READY);
342 * nfs41_discover_server_trunking - Detect server IP address trunking (mv1)
344 * @clp: nfs_client under test
345 * @result: OUT: found nfs_client, or clp
346 * @cred: credential to use for trunking test
348 * Returns NFS4_OK, a negative errno, or a negative NFS4ERR status.
349 * If NFS4_OK is returned, an nfs_client pointer is planted in
352 * Note: The returned client may not yet be marked ready.
354 int nfs41_discover_server_trunking(struct nfs_client *clp,
355 struct nfs_client **result,
356 struct rpc_cred *cred)
360 status = nfs4_proc_exchange_id(clp, cred);
361 if (status != NFS4_OK)
363 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
365 return nfs41_walk_client_list(clp, result, cred);
368 struct rpc_cred *nfs4_get_exchange_id_cred(struct nfs_client *clp)
370 struct rpc_cred *cred;
372 spin_lock(&clp->cl_lock);
373 cred = nfs4_get_machine_cred_locked(clp);
374 spin_unlock(&clp->cl_lock);
378 #endif /* CONFIG_NFS_V4_1 */
380 static struct rpc_cred *
381 nfs4_get_setclientid_cred_server(struct nfs_server *server)
383 struct nfs_client *clp = server->nfs_client;
384 struct rpc_cred *cred = NULL;
385 struct nfs4_state_owner *sp;
388 spin_lock(&clp->cl_lock);
389 pos = rb_first(&server->state_owners);
391 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
392 cred = get_rpccred(sp->so_cred);
394 spin_unlock(&clp->cl_lock);
399 * nfs4_get_setclientid_cred - Acquire credential for a setclientid operation
400 * @clp: client state handle
402 * Returns an rpc_cred with reference count bumped, or NULL.
404 struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
406 struct nfs_server *server;
407 struct rpc_cred *cred;
409 spin_lock(&clp->cl_lock);
410 cred = nfs4_get_machine_cred_locked(clp);
411 spin_unlock(&clp->cl_lock);
416 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
417 cred = nfs4_get_setclientid_cred_server(server);
427 static struct nfs4_state_owner *
428 nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
430 struct rb_node **p = &server->state_owners.rb_node,
432 struct nfs4_state_owner *sp;
436 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
438 if (cred < sp->so_cred)
439 p = &parent->rb_left;
440 else if (cred > sp->so_cred)
441 p = &parent->rb_right;
443 if (!list_empty(&sp->so_lru))
444 list_del_init(&sp->so_lru);
445 atomic_inc(&sp->so_count);
452 static struct nfs4_state_owner *
453 nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
455 struct nfs_server *server = new->so_server;
456 struct rb_node **p = &server->state_owners.rb_node,
458 struct nfs4_state_owner *sp;
463 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
465 if (new->so_cred < sp->so_cred)
466 p = &parent->rb_left;
467 else if (new->so_cred > sp->so_cred)
468 p = &parent->rb_right;
470 if (!list_empty(&sp->so_lru))
471 list_del_init(&sp->so_lru);
472 atomic_inc(&sp->so_count);
476 err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
479 rb_link_node(&new->so_server_node, parent, p);
480 rb_insert_color(&new->so_server_node, &server->state_owners);
485 nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
487 struct nfs_server *server = sp->so_server;
489 if (!RB_EMPTY_NODE(&sp->so_server_node))
490 rb_erase(&sp->so_server_node, &server->state_owners);
491 ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
495 nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
497 sc->create_time = ktime_get();
500 spin_lock_init(&sc->lock);
501 INIT_LIST_HEAD(&sc->list);
502 rpc_init_wait_queue(&sc->wait, "Seqid_waitqueue");
506 nfs4_destroy_seqid_counter(struct nfs_seqid_counter *sc)
508 rpc_destroy_wait_queue(&sc->wait);
512 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
513 * create a new state_owner.
516 static struct nfs4_state_owner *
517 nfs4_alloc_state_owner(struct nfs_server *server,
518 struct rpc_cred *cred,
521 struct nfs4_state_owner *sp;
523 sp = kzalloc(sizeof(*sp), gfp_flags);
526 sp->so_server = server;
527 sp->so_cred = get_rpccred(cred);
528 spin_lock_init(&sp->so_lock);
529 INIT_LIST_HEAD(&sp->so_states);
530 nfs4_init_seqid_counter(&sp->so_seqid);
531 atomic_set(&sp->so_count, 1);
532 INIT_LIST_HEAD(&sp->so_lru);
537 nfs4_drop_state_owner(struct nfs4_state_owner *sp)
539 struct rb_node *rb_node = &sp->so_server_node;
541 if (!RB_EMPTY_NODE(rb_node)) {
542 struct nfs_server *server = sp->so_server;
543 struct nfs_client *clp = server->nfs_client;
545 spin_lock(&clp->cl_lock);
546 if (!RB_EMPTY_NODE(rb_node)) {
547 rb_erase(rb_node, &server->state_owners);
548 RB_CLEAR_NODE(rb_node);
550 spin_unlock(&clp->cl_lock);
554 static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
556 nfs4_destroy_seqid_counter(&sp->so_seqid);
557 put_rpccred(sp->so_cred);
561 static void nfs4_gc_state_owners(struct nfs_server *server)
563 struct nfs_client *clp = server->nfs_client;
564 struct nfs4_state_owner *sp, *tmp;
565 unsigned long time_min, time_max;
568 spin_lock(&clp->cl_lock);
570 time_min = (long)time_max - (long)clp->cl_lease_time;
571 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
572 /* NB: LRU is sorted so that oldest is at the head */
573 if (time_in_range(sp->so_expires, time_min, time_max))
575 list_move(&sp->so_lru, &doomed);
576 nfs4_remove_state_owner_locked(sp);
578 spin_unlock(&clp->cl_lock);
580 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
581 list_del(&sp->so_lru);
582 nfs4_free_state_owner(sp);
587 * nfs4_get_state_owner - Look up a state owner given a credential
588 * @server: nfs_server to search
589 * @cred: RPC credential to match
591 * Returns a pointer to an instantiated nfs4_state_owner struct, or NULL.
593 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server,
594 struct rpc_cred *cred,
597 struct nfs_client *clp = server->nfs_client;
598 struct nfs4_state_owner *sp, *new;
600 spin_lock(&clp->cl_lock);
601 sp = nfs4_find_state_owner_locked(server, cred);
602 spin_unlock(&clp->cl_lock);
605 new = nfs4_alloc_state_owner(server, cred, gfp_flags);
609 if (ida_pre_get(&server->openowner_id, gfp_flags) == 0)
611 spin_lock(&clp->cl_lock);
612 sp = nfs4_insert_state_owner_locked(new);
613 spin_unlock(&clp->cl_lock);
614 } while (sp == ERR_PTR(-EAGAIN));
616 nfs4_free_state_owner(new);
618 nfs4_gc_state_owners(server);
623 * nfs4_put_state_owner - Release a nfs4_state_owner
624 * @sp: state owner data to release
626 * Note that we keep released state owners on an LRU
628 * This caches valid state owners so that they can be
629 * reused, to avoid the OPEN_CONFIRM on minor version 0.
630 * It also pins the uniquifier of dropped state owners for
631 * a while, to ensure that those state owner names are
634 void nfs4_put_state_owner(struct nfs4_state_owner *sp)
636 struct nfs_server *server = sp->so_server;
637 struct nfs_client *clp = server->nfs_client;
639 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
642 sp->so_expires = jiffies;
643 list_add_tail(&sp->so_lru, &server->state_owners_lru);
644 spin_unlock(&clp->cl_lock);
648 * nfs4_purge_state_owners - Release all cached state owners
649 * @server: nfs_server with cached state owners to release
651 * Called at umount time. Remaining state owners will be on
652 * the LRU with ref count of zero.
654 void nfs4_purge_state_owners(struct nfs_server *server)
656 struct nfs_client *clp = server->nfs_client;
657 struct nfs4_state_owner *sp, *tmp;
660 spin_lock(&clp->cl_lock);
661 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
662 list_move(&sp->so_lru, &doomed);
663 nfs4_remove_state_owner_locked(sp);
665 spin_unlock(&clp->cl_lock);
667 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
668 list_del(&sp->so_lru);
669 nfs4_free_state_owner(sp);
673 static struct nfs4_state *
674 nfs4_alloc_open_state(void)
676 struct nfs4_state *state;
678 state = kzalloc(sizeof(*state), GFP_NOFS);
681 atomic_set(&state->count, 1);
682 INIT_LIST_HEAD(&state->lock_states);
683 spin_lock_init(&state->state_lock);
684 seqlock_init(&state->seqlock);
689 nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
691 if (state->state == fmode)
693 /* NB! List reordering - see the reclaim code for why. */
694 if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
695 if (fmode & FMODE_WRITE)
696 list_move(&state->open_states, &state->owner->so_states);
698 list_move_tail(&state->open_states, &state->owner->so_states);
700 state->state = fmode;
703 static struct nfs4_state *
704 __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
706 struct nfs_inode *nfsi = NFS_I(inode);
707 struct nfs4_state *state;
709 list_for_each_entry(state, &nfsi->open_states, inode_states) {
710 if (state->owner != owner)
712 if (atomic_inc_not_zero(&state->count))
719 nfs4_free_open_state(struct nfs4_state *state)
725 nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
727 struct nfs4_state *state, *new;
728 struct nfs_inode *nfsi = NFS_I(inode);
730 spin_lock(&inode->i_lock);
731 state = __nfs4_find_state_byowner(inode, owner);
732 spin_unlock(&inode->i_lock);
735 new = nfs4_alloc_open_state();
736 spin_lock(&owner->so_lock);
737 spin_lock(&inode->i_lock);
738 state = __nfs4_find_state_byowner(inode, owner);
739 if (state == NULL && new != NULL) {
741 state->owner = owner;
742 atomic_inc(&owner->so_count);
743 list_add(&state->inode_states, &nfsi->open_states);
745 state->inode = inode;
746 spin_unlock(&inode->i_lock);
747 /* Note: The reclaim code dictates that we add stateless
748 * and read-only stateids to the end of the list */
749 list_add_tail(&state->open_states, &owner->so_states);
750 spin_unlock(&owner->so_lock);
752 spin_unlock(&inode->i_lock);
753 spin_unlock(&owner->so_lock);
755 nfs4_free_open_state(new);
761 void nfs4_put_open_state(struct nfs4_state *state)
763 struct inode *inode = state->inode;
764 struct nfs4_state_owner *owner = state->owner;
766 if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
768 spin_lock(&inode->i_lock);
769 list_del(&state->inode_states);
770 list_del(&state->open_states);
771 spin_unlock(&inode->i_lock);
772 spin_unlock(&owner->so_lock);
774 nfs4_free_open_state(state);
775 nfs4_put_state_owner(owner);
779 * Close the current file.
781 static void __nfs4_close(struct nfs4_state *state,
782 fmode_t fmode, gfp_t gfp_mask, int wait)
784 struct nfs4_state_owner *owner = state->owner;
788 atomic_inc(&owner->so_count);
789 /* Protect against nfs4_find_state() */
790 spin_lock(&owner->so_lock);
791 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
798 case FMODE_READ|FMODE_WRITE:
801 newstate = FMODE_READ|FMODE_WRITE;
802 if (state->n_rdwr == 0) {
803 if (state->n_rdonly == 0) {
804 newstate &= ~FMODE_READ;
805 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
806 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
808 if (state->n_wronly == 0) {
809 newstate &= ~FMODE_WRITE;
810 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
811 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
814 clear_bit(NFS_DELEGATED_STATE, &state->flags);
816 nfs4_state_set_mode_locked(state, newstate);
817 spin_unlock(&owner->so_lock);
820 nfs4_put_open_state(state);
821 nfs4_put_state_owner(owner);
823 nfs4_do_close(state, gfp_mask, wait);
826 void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
828 __nfs4_close(state, fmode, GFP_NOFS, 0);
831 void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
833 __nfs4_close(state, fmode, GFP_KERNEL, 1);
837 * Search the state->lock_states for an existing lock_owner
838 * that is compatible with current->files
840 static struct nfs4_lock_state *
841 __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type)
843 struct nfs4_lock_state *pos;
844 list_for_each_entry(pos, &state->lock_states, ls_locks) {
845 if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
847 switch (pos->ls_owner.lo_type) {
848 case NFS4_POSIX_LOCK_TYPE:
849 if (pos->ls_owner.lo_u.posix_owner != fl_owner)
852 case NFS4_FLOCK_LOCK_TYPE:
853 if (pos->ls_owner.lo_u.flock_owner != fl_pid)
856 atomic_inc(&pos->ls_count);
863 * Return a compatible lock_state. If no initialized lock_state structure
864 * exists, return an uninitialized one.
867 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)
869 struct nfs4_lock_state *lsp;
870 struct nfs_server *server = state->owner->so_server;
872 lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
875 nfs4_init_seqid_counter(&lsp->ls_seqid);
876 atomic_set(&lsp->ls_count, 1);
877 lsp->ls_state = state;
878 lsp->ls_owner.lo_type = type;
879 switch (lsp->ls_owner.lo_type) {
880 case NFS4_FLOCK_LOCK_TYPE:
881 lsp->ls_owner.lo_u.flock_owner = fl_pid;
883 case NFS4_POSIX_LOCK_TYPE:
884 lsp->ls_owner.lo_u.posix_owner = fl_owner;
889 lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
890 if (lsp->ls_seqid.owner_id < 0)
892 INIT_LIST_HEAD(&lsp->ls_locks);
899 void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
901 ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
902 nfs4_destroy_seqid_counter(&lsp->ls_seqid);
907 * Return a compatible lock_state. If no initialized lock_state structure
908 * exists, return an uninitialized one.
911 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner, pid_t pid, unsigned int type)
913 struct nfs4_lock_state *lsp, *new = NULL;
916 spin_lock(&state->state_lock);
917 lsp = __nfs4_find_lock_state(state, owner, pid, type);
921 list_add(&new->ls_locks, &state->lock_states);
922 set_bit(LK_STATE_IN_USE, &state->flags);
927 spin_unlock(&state->state_lock);
928 new = nfs4_alloc_lock_state(state, owner, pid, type);
932 spin_unlock(&state->state_lock);
934 nfs4_free_lock_state(state->owner->so_server, new);
939 * Release reference to lock_state, and free it if we see that
940 * it is no longer in use
942 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
944 struct nfs4_state *state;
948 state = lsp->ls_state;
949 if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
951 list_del(&lsp->ls_locks);
952 if (list_empty(&state->lock_states))
953 clear_bit(LK_STATE_IN_USE, &state->flags);
954 spin_unlock(&state->state_lock);
955 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
956 if (nfs4_release_lockowner(lsp) == 0)
959 nfs4_free_lock_state(lsp->ls_state->owner->so_server, lsp);
962 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
964 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
966 dst->fl_u.nfs4_fl.owner = lsp;
967 atomic_inc(&lsp->ls_count);
970 static void nfs4_fl_release_lock(struct file_lock *fl)
972 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
975 static const struct file_lock_operations nfs4_fl_lock_ops = {
976 .fl_copy_lock = nfs4_fl_copy_lock,
977 .fl_release_private = nfs4_fl_release_lock,
980 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
982 struct nfs4_lock_state *lsp;
984 if (fl->fl_ops != NULL)
986 if (fl->fl_flags & FL_POSIX)
987 lsp = nfs4_get_lock_state(state, fl->fl_owner, 0, NFS4_POSIX_LOCK_TYPE);
988 else if (fl->fl_flags & FL_FLOCK)
989 lsp = nfs4_get_lock_state(state, NULL, fl->fl_pid,
990 NFS4_FLOCK_LOCK_TYPE);
995 fl->fl_u.nfs4_fl.owner = lsp;
996 fl->fl_ops = &nfs4_fl_lock_ops;
1000 static bool nfs4_copy_lock_stateid(nfs4_stateid *dst, struct nfs4_state *state,
1001 const struct nfs_lockowner *lockowner)
1003 struct nfs4_lock_state *lsp;
1004 fl_owner_t fl_owner;
1009 if (lockowner == NULL)
1012 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
1015 fl_owner = lockowner->l_owner;
1016 fl_pid = lockowner->l_pid;
1017 spin_lock(&state->state_lock);
1018 lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
1019 if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
1020 nfs4_stateid_copy(dst, &lsp->ls_stateid);
1023 spin_unlock(&state->state_lock);
1024 nfs4_put_lock_state(lsp);
1029 static void nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
1034 seq = read_seqbegin(&state->seqlock);
1035 nfs4_stateid_copy(dst, &state->stateid);
1036 } while (read_seqretry(&state->seqlock, seq));
1040 * Byte-range lock aware utility to initialize the stateid of read/write
1043 void nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state,
1044 fmode_t fmode, const struct nfs_lockowner *lockowner)
1046 if (nfs4_copy_delegation_stateid(dst, state->inode, fmode))
1048 if (nfs4_copy_lock_stateid(dst, state, lockowner))
1050 nfs4_copy_open_stateid(dst, state);
1053 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
1055 struct nfs_seqid *new;
1057 new = kmalloc(sizeof(*new), gfp_mask);
1059 new->sequence = counter;
1060 INIT_LIST_HEAD(&new->list);
1066 void nfs_release_seqid(struct nfs_seqid *seqid)
1068 struct nfs_seqid_counter *sequence;
1070 if (list_empty(&seqid->list))
1072 sequence = seqid->sequence;
1073 spin_lock(&sequence->lock);
1074 list_del_init(&seqid->list);
1075 if (!list_empty(&sequence->list)) {
1076 struct nfs_seqid *next;
1078 next = list_first_entry(&sequence->list,
1079 struct nfs_seqid, list);
1080 rpc_wake_up_queued_task(&sequence->wait, next->task);
1082 spin_unlock(&sequence->lock);
1085 void nfs_free_seqid(struct nfs_seqid *seqid)
1087 nfs_release_seqid(seqid);
1092 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
1093 * failed with a seqid incrementing error -
1094 * see comments nfs_fs.h:seqid_mutating_error()
1096 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
1101 case -NFS4ERR_BAD_SEQID:
1102 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
1104 pr_warn_ratelimited("NFS: v4 server returned a bad"
1105 " sequence-id error on an"
1106 " unconfirmed sequence %p!\n",
1108 case -NFS4ERR_STALE_CLIENTID:
1109 case -NFS4ERR_STALE_STATEID:
1110 case -NFS4ERR_BAD_STATEID:
1111 case -NFS4ERR_BADXDR:
1112 case -NFS4ERR_RESOURCE:
1113 case -NFS4ERR_NOFILEHANDLE:
1114 /* Non-seqid mutating errors */
1118 * Note: no locking needed as we are guaranteed to be first
1119 * on the sequence list
1121 seqid->sequence->counter++;
1124 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
1126 struct nfs4_state_owner *sp = container_of(seqid->sequence,
1127 struct nfs4_state_owner, so_seqid);
1128 struct nfs_server *server = sp->so_server;
1130 if (status == -NFS4ERR_BAD_SEQID)
1131 nfs4_drop_state_owner(sp);
1132 if (!nfs4_has_session(server->nfs_client))
1133 nfs_increment_seqid(status, seqid);
1137 * Increment the seqid if the LOCK/LOCKU succeeded, or
1138 * failed with a seqid incrementing error -
1139 * see comments nfs_fs.h:seqid_mutating_error()
1141 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
1143 nfs_increment_seqid(status, seqid);
1146 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
1148 struct nfs_seqid_counter *sequence = seqid->sequence;
1151 spin_lock(&sequence->lock);
1153 if (list_empty(&seqid->list))
1154 list_add_tail(&seqid->list, &sequence->list);
1155 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
1157 rpc_sleep_on(&sequence->wait, task, NULL);
1160 spin_unlock(&sequence->lock);
1164 static int nfs4_run_state_manager(void *);
1166 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
1168 smp_mb__before_clear_bit();
1169 clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
1170 smp_mb__after_clear_bit();
1171 wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
1172 rpc_wake_up(&clp->cl_rpcwaitq);
1176 * Schedule the nfs_client asynchronous state management routine
1178 void nfs4_schedule_state_manager(struct nfs_client *clp)
1180 struct task_struct *task;
1181 char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
1183 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1185 __module_get(THIS_MODULE);
1186 atomic_inc(&clp->cl_count);
1188 /* The rcu_read_lock() is not strictly necessary, as the state
1189 * manager is the only thread that ever changes the rpc_xprt
1190 * after it's initialized. At this point, we're single threaded. */
1192 snprintf(buf, sizeof(buf), "%s-manager",
1193 rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR));
1195 task = kthread_run(nfs4_run_state_manager, clp, buf);
1197 printk(KERN_ERR "%s: kthread_run: %ld\n",
1198 __func__, PTR_ERR(task));
1199 nfs4_clear_state_manager_bit(clp);
1200 nfs_put_client(clp);
1201 module_put(THIS_MODULE);
1206 * Schedule a lease recovery attempt
1208 void nfs4_schedule_lease_recovery(struct nfs_client *clp)
1212 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1213 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1214 dprintk("%s: scheduling lease recovery for server %s\n", __func__,
1216 nfs4_schedule_state_manager(clp);
1218 EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
1220 int nfs4_wait_clnt_recover(struct nfs_client *clp)
1226 res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
1227 nfs_wait_bit_killable, TASK_KILLABLE);
1231 if (clp->cl_cons_state < 0)
1232 return clp->cl_cons_state;
1236 int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1241 for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1242 ret = nfs4_wait_clnt_recover(clp);
1245 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1246 !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1248 nfs4_schedule_state_manager(clp);
1255 * nfs40_handle_cb_pathdown - return all delegations after NFS4ERR_CB_PATH_DOWN
1256 * @clp: client to process
1258 * Set the NFS4CLNT_LEASE_EXPIRED state in order to force a
1259 * resend of the SETCLIENTID and hence re-establish the
1260 * callback channel. Then return all existing delegations.
1262 static void nfs40_handle_cb_pathdown(struct nfs_client *clp)
1264 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1265 nfs_expire_all_delegations(clp);
1266 dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
1270 void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
1272 nfs40_handle_cb_pathdown(clp);
1273 nfs4_schedule_state_manager(clp);
1276 static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1279 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1280 /* Don't recover state that expired before the reboot */
1281 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
1282 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1285 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1286 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1290 static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1292 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
1293 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1294 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1295 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
1299 void nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
1301 struct nfs_client *clp = server->nfs_client;
1303 nfs4_state_mark_reclaim_nograce(clp, state);
1304 dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
1306 nfs4_schedule_state_manager(clp);
1308 EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1310 void nfs_inode_find_state_and_recover(struct inode *inode,
1311 const nfs4_stateid *stateid)
1313 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1314 struct nfs_inode *nfsi = NFS_I(inode);
1315 struct nfs_open_context *ctx;
1316 struct nfs4_state *state;
1319 spin_lock(&inode->i_lock);
1320 list_for_each_entry(ctx, &nfsi->open_files, list) {
1324 if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
1326 if (!nfs4_stateid_match(&state->stateid, stateid))
1328 nfs4_state_mark_reclaim_nograce(clp, state);
1331 spin_unlock(&inode->i_lock);
1333 nfs4_schedule_state_manager(clp);
1337 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
1339 struct inode *inode = state->inode;
1340 struct nfs_inode *nfsi = NFS_I(inode);
1341 struct file_lock *fl;
1344 if (inode->i_flock == NULL)
1347 /* Guard against delegation returns and new lock/unlock calls */
1348 down_write(&nfsi->rwsem);
1349 /* Protect inode->i_flock using the BKL */
1351 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1352 if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
1354 if (nfs_file_open_context(fl->fl_file)->state != state)
1357 status = ops->recover_lock(state, fl);
1362 case -NFS4ERR_ADMIN_REVOKED:
1363 case -NFS4ERR_STALE_STATEID:
1364 case -NFS4ERR_BAD_STATEID:
1365 case -NFS4ERR_EXPIRED:
1366 case -NFS4ERR_NO_GRACE:
1367 case -NFS4ERR_STALE_CLIENTID:
1368 case -NFS4ERR_BADSESSION:
1369 case -NFS4ERR_BADSLOT:
1370 case -NFS4ERR_BAD_HIGH_SLOT:
1371 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1374 printk(KERN_ERR "NFS: %s: unhandled error %d. "
1375 "Zeroing state\n", __func__, status);
1377 case -NFS4ERR_DENIED:
1378 case -NFS4ERR_RECLAIM_BAD:
1379 case -NFS4ERR_RECLAIM_CONFLICT:
1380 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1387 up_write(&nfsi->rwsem);
1391 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
1393 struct nfs4_state *state;
1394 struct nfs4_lock_state *lock;
1397 /* Note: we rely on the sp->so_states list being ordered
1398 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
1400 * This is needed to ensure that the server won't give us any
1401 * read delegations that we have to return if, say, we are
1402 * recovering after a network partition or a reboot from a
1403 * server that doesn't support a grace period.
1406 spin_lock(&sp->so_lock);
1407 list_for_each_entry(state, &sp->so_states, open_states) {
1408 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
1410 if (state->state == 0)
1412 atomic_inc(&state->count);
1413 spin_unlock(&sp->so_lock);
1414 status = ops->recover_open(sp, state);
1416 status = nfs4_reclaim_locks(state, ops);
1418 spin_lock(&state->state_lock);
1419 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1420 if (!test_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags))
1421 pr_warn_ratelimited("NFS: "
1423 "failed!\n", __func__);
1425 spin_unlock(&state->state_lock);
1426 nfs4_put_open_state(state);
1432 printk(KERN_ERR "NFS: %s: unhandled error %d. "
1433 "Zeroing state\n", __func__, status);
1438 * Open state on this file cannot be recovered
1439 * All we can do is revert to using the zero stateid.
1441 memset(&state->stateid, 0,
1442 sizeof(state->stateid));
1443 /* Mark the file as being 'closed' */
1448 * User RPCSEC_GSS context has expired.
1449 * We cannot recover this stateid now, so
1450 * skip it and allow recovery thread to
1454 case -NFS4ERR_ADMIN_REVOKED:
1455 case -NFS4ERR_STALE_STATEID:
1456 case -NFS4ERR_BAD_STATEID:
1457 case -NFS4ERR_RECLAIM_BAD:
1458 case -NFS4ERR_RECLAIM_CONFLICT:
1459 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1461 case -NFS4ERR_EXPIRED:
1462 case -NFS4ERR_NO_GRACE:
1463 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1464 case -NFS4ERR_STALE_CLIENTID:
1465 case -NFS4ERR_BADSESSION:
1466 case -NFS4ERR_BADSLOT:
1467 case -NFS4ERR_BAD_HIGH_SLOT:
1468 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1471 nfs4_put_open_state(state);
1474 spin_unlock(&sp->so_lock);
1477 nfs4_put_open_state(state);
1481 static void nfs4_clear_open_state(struct nfs4_state *state)
1483 struct nfs4_lock_state *lock;
1485 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1486 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1487 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1488 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1489 spin_lock(&state->state_lock);
1490 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1491 lock->ls_seqid.flags = 0;
1492 clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
1494 spin_unlock(&state->state_lock);
1497 static void nfs4_reset_seqids(struct nfs_server *server,
1498 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1500 struct nfs_client *clp = server->nfs_client;
1501 struct nfs4_state_owner *sp;
1502 struct rb_node *pos;
1503 struct nfs4_state *state;
1505 spin_lock(&clp->cl_lock);
1506 for (pos = rb_first(&server->state_owners);
1508 pos = rb_next(pos)) {
1509 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1510 sp->so_seqid.flags = 0;
1511 spin_lock(&sp->so_lock);
1512 list_for_each_entry(state, &sp->so_states, open_states) {
1513 if (mark_reclaim(clp, state))
1514 nfs4_clear_open_state(state);
1516 spin_unlock(&sp->so_lock);
1518 spin_unlock(&clp->cl_lock);
1521 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp,
1522 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1524 struct nfs_server *server;
1527 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1528 nfs4_reset_seqids(server, mark_reclaim);
1532 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1534 /* Mark all delegations for reclaim */
1535 nfs_delegation_mark_reclaim(clp);
1536 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1539 static void nfs4_reclaim_complete(struct nfs_client *clp,
1540 const struct nfs4_state_recovery_ops *ops)
1542 /* Notify the server we're done reclaiming our state */
1543 if (ops->reclaim_complete)
1544 (void)ops->reclaim_complete(clp);
1547 static void nfs4_clear_reclaim_server(struct nfs_server *server)
1549 struct nfs_client *clp = server->nfs_client;
1550 struct nfs4_state_owner *sp;
1551 struct rb_node *pos;
1552 struct nfs4_state *state;
1554 spin_lock(&clp->cl_lock);
1555 for (pos = rb_first(&server->state_owners);
1557 pos = rb_next(pos)) {
1558 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1559 spin_lock(&sp->so_lock);
1560 list_for_each_entry(state, &sp->so_states, open_states) {
1561 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
1564 nfs4_state_mark_reclaim_nograce(clp, state);
1566 spin_unlock(&sp->so_lock);
1568 spin_unlock(&clp->cl_lock);
1571 static int nfs4_state_clear_reclaim_reboot(struct nfs_client *clp)
1573 struct nfs_server *server;
1575 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1579 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1580 nfs4_clear_reclaim_server(server);
1583 nfs_delegation_reap_unclaimed(clp);
1587 static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
1589 if (!nfs4_state_clear_reclaim_reboot(clp))
1591 nfs4_reclaim_complete(clp, clp->cl_mvops->reboot_recovery_ops);
1594 static void nfs_delegation_clear_all(struct nfs_client *clp)
1596 nfs_delegation_mark_reclaim(clp);
1597 nfs_delegation_reap_unclaimed(clp);
1600 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1602 nfs_delegation_clear_all(clp);
1603 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1606 static void nfs4_warn_keyexpired(const char *s)
1608 printk_ratelimited(KERN_WARNING "Error: state manager"
1609 " encountered RPCSEC_GSS session"
1610 " expired against NFSv4 server %s.\n",
1614 static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1619 case -NFS4ERR_CB_PATH_DOWN:
1620 nfs40_handle_cb_pathdown(clp);
1622 case -NFS4ERR_NO_GRACE:
1623 nfs4_state_end_reclaim_reboot(clp);
1625 case -NFS4ERR_STALE_CLIENTID:
1626 case -NFS4ERR_LEASE_MOVED:
1627 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1628 nfs4_state_clear_reclaim_reboot(clp);
1629 nfs4_state_start_reclaim_reboot(clp);
1631 case -NFS4ERR_EXPIRED:
1632 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1633 nfs4_state_start_reclaim_nograce(clp);
1635 case -NFS4ERR_BADSESSION:
1636 case -NFS4ERR_BADSLOT:
1637 case -NFS4ERR_BAD_HIGH_SLOT:
1638 case -NFS4ERR_DEADSESSION:
1639 case -NFS4ERR_SEQ_FALSE_RETRY:
1640 case -NFS4ERR_SEQ_MISORDERED:
1641 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1642 /* Zero session reset errors */
1644 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1645 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1648 /* Nothing we can do */
1649 nfs4_warn_keyexpired(clp->cl_hostname);
1652 dprintk("%s: failed to handle error %d for server %s\n",
1653 __func__, error, clp->cl_hostname);
1656 dprintk("%s: handled error %d for server %s\n", __func__, error,
1661 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1663 struct nfs4_state_owner *sp;
1664 struct nfs_server *server;
1665 struct rb_node *pos;
1670 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1671 nfs4_purge_state_owners(server);
1672 spin_lock(&clp->cl_lock);
1673 for (pos = rb_first(&server->state_owners);
1675 pos = rb_next(pos)) {
1677 struct nfs4_state_owner, so_server_node);
1678 if (!test_and_clear_bit(ops->owner_flag_bit,
1681 atomic_inc(&sp->so_count);
1682 spin_unlock(&clp->cl_lock);
1685 status = nfs4_reclaim_open_state(sp, ops);
1687 set_bit(ops->owner_flag_bit, &sp->so_flags);
1688 nfs4_put_state_owner(sp);
1689 return nfs4_recovery_handle_error(clp, status);
1692 nfs4_put_state_owner(sp);
1695 spin_unlock(&clp->cl_lock);
1701 static int nfs4_check_lease(struct nfs_client *clp)
1703 struct rpc_cred *cred;
1704 const struct nfs4_state_maintenance_ops *ops =
1705 clp->cl_mvops->state_renewal_ops;
1708 /* Is the client already known to have an expired lease? */
1709 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1711 spin_lock(&clp->cl_lock);
1712 cred = ops->get_state_renewal_cred_locked(clp);
1713 spin_unlock(&clp->cl_lock);
1715 cred = nfs4_get_setclientid_cred(clp);
1720 status = ops->renew_lease(clp, cred);
1723 return nfs4_recovery_handle_error(clp, status);
1726 /* Set NFS4CLNT_LEASE_EXPIRED and reclaim reboot state for all v4.0 errors
1727 * and for recoverable errors on EXCHANGE_ID for v4.1
1729 static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
1732 case -NFS4ERR_SEQ_MISORDERED:
1733 if (test_and_set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state))
1734 return -ESERVERFAULT;
1735 /* Lease confirmation error: retry after purging the lease */
1737 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1739 case -NFS4ERR_STALE_CLIENTID:
1740 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1741 nfs4_state_clear_reclaim_reboot(clp);
1742 nfs4_state_start_reclaim_reboot(clp);
1744 case -NFS4ERR_CLID_INUSE:
1745 pr_err("NFS: Server %s reports our clientid is in use\n",
1747 nfs_mark_client_ready(clp, -EPERM);
1748 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1751 if (clp->cl_machine_cred == NULL)
1753 /* Handle case where the user hasn't set up machine creds */
1754 nfs4_clear_machine_cred(clp);
1755 case -NFS4ERR_DELAY:
1761 case -NFS4ERR_MINOR_VERS_MISMATCH:
1762 if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
1763 nfs_mark_client_ready(clp, -EPROTONOSUPPORT);
1764 dprintk("%s: exit with error %d for server %s\n",
1765 __func__, -EPROTONOSUPPORT, clp->cl_hostname);
1766 return -EPROTONOSUPPORT;
1768 nfs4_warn_keyexpired(clp->cl_hostname);
1769 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1770 * in nfs4_exchange_id */
1772 dprintk("%s: exit with error %d for server %s\n", __func__,
1773 status, clp->cl_hostname);
1776 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1777 dprintk("%s: handled error %d for server %s\n", __func__, status,
1782 static int nfs4_establish_lease(struct nfs_client *clp)
1784 struct rpc_cred *cred;
1785 const struct nfs4_state_recovery_ops *ops =
1786 clp->cl_mvops->reboot_recovery_ops;
1789 cred = ops->get_clid_cred(clp);
1792 status = ops->establish_clid(clp, cred);
1796 pnfs_destroy_all_layouts(clp);
1801 * Returns zero or a negative errno. NFS4ERR values are converted
1802 * to local errno values.
1804 static int nfs4_reclaim_lease(struct nfs_client *clp)
1808 status = nfs4_establish_lease(clp);
1810 return nfs4_handle_reclaim_lease_error(clp, status);
1811 if (test_and_clear_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state))
1812 nfs4_state_start_reclaim_nograce(clp);
1813 if (!test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
1814 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1815 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1816 clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1820 static int nfs4_purge_lease(struct nfs_client *clp)
1824 status = nfs4_establish_lease(clp);
1826 return nfs4_handle_reclaim_lease_error(clp, status);
1827 clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1828 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1829 nfs4_state_start_reclaim_nograce(clp);
1834 * nfs4_discover_server_trunking - Detect server IP address trunking
1836 * @clp: nfs_client under test
1837 * @result: OUT: found nfs_client, or clp
1839 * Returns zero or a negative errno. If zero is returned,
1840 * an nfs_client pointer is planted in "result".
1842 * Note: since we are invoked in process context, and
1843 * not from inside the state manager, we cannot use
1844 * nfs4_handle_reclaim_lease_error().
1846 int nfs4_discover_server_trunking(struct nfs_client *clp,
1847 struct nfs_client **result)
1849 const struct nfs4_state_recovery_ops *ops =
1850 clp->cl_mvops->reboot_recovery_ops;
1851 rpc_authflavor_t *flavors, flav, save;
1852 struct rpc_clnt *clnt;
1853 struct rpc_cred *cred;
1856 dprintk("NFS: %s: testing '%s'\n", __func__, clp->cl_hostname);
1858 len = NFS_MAX_SECFLAVORS;
1859 flavors = kcalloc(len, sizeof(*flavors), GFP_KERNEL);
1860 if (flavors == NULL) {
1864 len = rpcauth_list_flavors(flavors, len);
1869 clnt = clp->cl_rpcclient;
1870 save = clnt->cl_auth->au_flavor;
1873 mutex_lock(&nfs_clid_init_mutex);
1876 cred = ops->get_clid_cred(clp);
1880 status = ops->detect_trunking(clp, result, cred);
1887 if (clp->cl_machine_cred == NULL)
1889 /* Handle case where the user hasn't set up machine creds */
1890 nfs4_clear_machine_cred(clp);
1891 case -NFS4ERR_DELAY:
1895 dprintk("NFS: %s after status %d, retrying\n",
1899 case -NFS4ERR_CLID_INUSE:
1900 case -NFS4ERR_WRONGSEC:
1905 flav = flavors[i++];
1907 flav = flavors[i++];
1908 clnt = rpc_clone_client_set_auth(clnt, flav);
1910 status = PTR_ERR(clnt);
1913 clp->cl_rpcclient = clnt;
1916 case -NFS4ERR_MINOR_VERS_MISMATCH:
1917 status = -EPROTONOSUPPORT;
1921 nfs4_warn_keyexpired(clp->cl_hostname);
1922 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1923 * in nfs4_exchange_id */
1924 status = -EKEYEXPIRED;
1928 mutex_unlock(&nfs_clid_init_mutex);
1932 dprintk("NFS: %s: status = %d\n", __func__, status);
1936 #ifdef CONFIG_NFS_V4_1
1937 void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
1939 struct nfs_client *clp = session->clp;
1943 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1945 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1946 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1948 nfs4_schedule_lease_recovery(clp);
1950 EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
1952 static void nfs41_ping_server(struct nfs_client *clp)
1954 /* Use CHECK_LEASE to ping the server with a SEQUENCE */
1955 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1956 nfs4_schedule_state_manager(clp);
1959 void nfs41_server_notify_target_slotid_update(struct nfs_client *clp)
1961 nfs41_ping_server(clp);
1964 static void nfs4_reset_all_state(struct nfs_client *clp)
1966 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1967 set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1968 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1969 nfs4_state_start_reclaim_nograce(clp);
1970 dprintk("%s: scheduling reset of all state for server %s!\n",
1971 __func__, clp->cl_hostname);
1972 nfs4_schedule_state_manager(clp);
1976 static void nfs41_handle_server_reboot(struct nfs_client *clp)
1978 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1979 nfs4_state_start_reclaim_reboot(clp);
1980 dprintk("%s: server %s rebooted!\n", __func__,
1982 nfs4_schedule_state_manager(clp);
1986 static void nfs41_handle_state_revoked(struct nfs_client *clp)
1988 nfs4_reset_all_state(clp);
1989 dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
1992 static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
1994 /* This will need to handle layouts too */
1995 nfs_expire_all_delegations(clp);
1996 dprintk("%s: Recallable state revoked on server %s!\n", __func__,
2000 static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
2002 nfs_expire_all_delegations(clp);
2003 if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
2004 nfs4_schedule_state_manager(clp);
2005 dprintk("%s: server %s declared a backchannel fault\n", __func__,
2009 static void nfs41_handle_cb_path_down(struct nfs_client *clp)
2011 if (test_and_set_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2012 &clp->cl_state) == 0)
2013 nfs4_schedule_state_manager(clp);
2016 void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
2021 dprintk("%s: \"%s\" (client ID %llx) flags=0x%08x\n",
2022 __func__, clp->cl_hostname, clp->cl_clientid, flags);
2024 if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
2025 nfs41_handle_server_reboot(clp);
2026 if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
2027 SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
2028 SEQ4_STATUS_ADMIN_STATE_REVOKED |
2029 SEQ4_STATUS_LEASE_MOVED))
2030 nfs41_handle_state_revoked(clp);
2031 if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
2032 nfs41_handle_recallable_state_revoked(clp);
2033 if (flags & SEQ4_STATUS_BACKCHANNEL_FAULT)
2034 nfs41_handle_backchannel_fault(clp);
2035 else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
2036 SEQ4_STATUS_CB_PATH_DOWN_SESSION))
2037 nfs41_handle_cb_path_down(clp);
2040 static int nfs4_reset_session(struct nfs_client *clp)
2042 struct rpc_cred *cred;
2045 if (!nfs4_has_session(clp))
2047 nfs4_begin_drain_session(clp);
2048 cred = nfs4_get_exchange_id_cred(clp);
2049 status = nfs4_proc_destroy_session(clp->cl_session, cred);
2050 if (status && status != -NFS4ERR_BADSESSION &&
2051 status != -NFS4ERR_DEADSESSION) {
2052 status = nfs4_recovery_handle_error(clp, status);
2056 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
2057 status = nfs4_proc_create_session(clp, cred);
2059 dprintk("%s: session reset failed with status %d for server %s!\n",
2060 __func__, status, clp->cl_hostname);
2061 status = nfs4_handle_reclaim_lease_error(clp, status);
2064 nfs41_finish_session_reset(clp);
2065 dprintk("%s: session reset was successful for server %s!\n",
2066 __func__, clp->cl_hostname);
2073 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2075 struct rpc_cred *cred;
2078 if (!nfs4_has_session(clp))
2080 nfs4_begin_drain_session(clp);
2081 cred = nfs4_get_exchange_id_cred(clp);
2082 ret = nfs4_proc_bind_conn_to_session(clp, cred);
2085 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2088 dprintk("%s: bind_conn_to_session was successful for server %s!\n",
2089 __func__, clp->cl_hostname);
2091 case -NFS4ERR_DELAY:
2093 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2096 return nfs4_recovery_handle_error(clp, ret);
2100 #else /* CONFIG_NFS_V4_1 */
2101 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
2102 static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
2104 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2108 #endif /* CONFIG_NFS_V4_1 */
2110 static void nfs4_state_manager(struct nfs_client *clp)
2113 const char *section = "", *section_sep = "";
2115 /* Ensure exclusive access to NFSv4 state */
2117 if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
2118 section = "purge state";
2119 status = nfs4_purge_lease(clp);
2125 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
2126 section = "lease expired";
2127 /* We're going to have to re-establish a clientid */
2128 status = nfs4_reclaim_lease(clp);
2134 /* Initialize or reset the session */
2135 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
2136 section = "reset session";
2137 status = nfs4_reset_session(clp);
2138 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
2144 /* Send BIND_CONN_TO_SESSION */
2145 if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2147 section = "bind conn to session";
2148 status = nfs4_bind_conn_to_session(clp);
2154 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
2155 section = "check lease";
2156 status = nfs4_check_lease(clp);
2162 /* First recover reboot state... */
2163 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
2164 section = "reclaim reboot";
2165 status = nfs4_do_reclaim(clp,
2166 clp->cl_mvops->reboot_recovery_ops);
2167 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2168 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
2170 nfs4_state_end_reclaim_reboot(clp);
2171 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
2177 /* Now recover expired state... */
2178 if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
2179 section = "reclaim nograce";
2180 status = nfs4_do_reclaim(clp,
2181 clp->cl_mvops->nograce_recovery_ops);
2182 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2183 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
2184 test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
2190 nfs4_end_drain_session(clp);
2191 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
2192 nfs_client_return_marked_delegations(clp);
2196 nfs4_clear_state_manager_bit(clp);
2197 /* Did we race with an attempt to give us more work? */
2198 if (clp->cl_state == 0)
2200 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
2202 } while (atomic_read(&clp->cl_count) > 1);
2205 if (strlen(section))
2207 pr_warn_ratelimited("NFS: state manager%s%s failed on NFSv4 server %s"
2208 " with error %d\n", section_sep, section,
2209 clp->cl_hostname, -status);
2211 nfs4_end_drain_session(clp);
2212 nfs4_clear_state_manager_bit(clp);
2215 static int nfs4_run_state_manager(void *ptr)
2217 struct nfs_client *clp = ptr;
2219 allow_signal(SIGKILL);
2220 nfs4_state_manager(clp);
2221 nfs_put_client(clp);
2222 module_put_and_exit(0);