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);
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 /* Sustain the lease, even if it's empty. If the clientid4
141 * goes stale it's of no use for trunking discovery. */
142 nfs4_schedule_state_renewal(*result);
148 struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
150 struct rpc_cred *cred = NULL;
152 if (clp->cl_machine_cred != NULL)
153 cred = get_rpccred(clp->cl_machine_cred);
157 static void nfs4_root_machine_cred(struct nfs_client *clp)
159 struct rpc_cred *cred, *new;
161 new = rpc_lookup_machine_cred(NULL);
162 spin_lock(&clp->cl_lock);
163 cred = clp->cl_machine_cred;
164 clp->cl_machine_cred = new;
165 spin_unlock(&clp->cl_lock);
170 static struct rpc_cred *
171 nfs4_get_renew_cred_server_locked(struct nfs_server *server)
173 struct rpc_cred *cred = NULL;
174 struct nfs4_state_owner *sp;
177 for (pos = rb_first(&server->state_owners);
179 pos = rb_next(pos)) {
180 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
181 if (list_empty(&sp->so_states))
183 cred = get_rpccred(sp->so_cred);
190 * nfs4_get_renew_cred_locked - Acquire credential for a renew operation
191 * @clp: client state handle
193 * Returns an rpc_cred with reference count bumped, or NULL.
194 * Caller must hold clp->cl_lock.
196 struct rpc_cred *nfs4_get_renew_cred_locked(struct nfs_client *clp)
198 struct rpc_cred *cred = NULL;
199 struct nfs_server *server;
201 /* Use machine credentials if available */
202 cred = nfs4_get_machine_cred_locked(clp);
207 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
208 cred = nfs4_get_renew_cred_server_locked(server);
218 #if defined(CONFIG_NFS_V4_1)
220 static int nfs41_setup_state_renewal(struct nfs_client *clp)
223 struct nfs_fsinfo fsinfo;
225 if (!test_bit(NFS_CS_CHECK_LEASE_TIME, &clp->cl_res_state)) {
226 nfs4_schedule_state_renewal(clp);
230 status = nfs4_proc_get_lease_time(clp, &fsinfo);
232 /* Update lease time and schedule renewal */
233 spin_lock(&clp->cl_lock);
234 clp->cl_lease_time = fsinfo.lease_time * HZ;
235 clp->cl_last_renewal = jiffies;
236 spin_unlock(&clp->cl_lock);
238 nfs4_schedule_state_renewal(clp);
244 static void nfs4_end_drain_slot_table(struct nfs4_slot_table *tbl)
246 if (test_and_clear_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state)) {
247 spin_lock(&tbl->slot_tbl_lock);
248 nfs41_wake_slot_table(tbl);
249 spin_unlock(&tbl->slot_tbl_lock);
253 static void nfs4_end_drain_session(struct nfs_client *clp)
255 struct nfs4_session *ses = clp->cl_session;
258 nfs4_end_drain_slot_table(&ses->bc_slot_table);
259 nfs4_end_drain_slot_table(&ses->fc_slot_table);
264 * Signal state manager thread if session fore channel is drained
266 void nfs4_slot_tbl_drain_complete(struct nfs4_slot_table *tbl)
268 if (nfs4_slot_tbl_draining(tbl))
269 complete(&tbl->complete);
272 static int nfs4_drain_slot_tbl(struct nfs4_slot_table *tbl)
274 set_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state);
275 spin_lock(&tbl->slot_tbl_lock);
276 if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
277 INIT_COMPLETION(tbl->complete);
278 spin_unlock(&tbl->slot_tbl_lock);
279 return wait_for_completion_interruptible(&tbl->complete);
281 spin_unlock(&tbl->slot_tbl_lock);
285 static int nfs4_begin_drain_session(struct nfs_client *clp)
287 struct nfs4_session *ses = clp->cl_session;
291 ret = nfs4_drain_slot_tbl(&ses->bc_slot_table);
295 return nfs4_drain_slot_tbl(&ses->fc_slot_table);
298 static void nfs41_finish_session_reset(struct nfs_client *clp)
300 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
301 clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
302 /* create_session negotiated new slot table */
303 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
304 nfs41_setup_state_renewal(clp);
307 int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
311 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
313 nfs4_begin_drain_session(clp);
314 status = nfs4_proc_exchange_id(clp, cred);
317 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
319 status = nfs4_proc_create_session(clp, cred);
322 nfs41_finish_session_reset(clp);
323 nfs_mark_client_ready(clp, NFS_CS_READY);
329 * nfs41_discover_server_trunking - Detect server IP address trunking (mv1)
331 * @clp: nfs_client under test
332 * @result: OUT: found nfs_client, or clp
333 * @cred: credential to use for trunking test
335 * Returns NFS4_OK, a negative errno, or a negative NFS4ERR status.
336 * If NFS4_OK is returned, an nfs_client pointer is planted in
339 * Note: The returned client may not yet be marked ready.
341 int nfs41_discover_server_trunking(struct nfs_client *clp,
342 struct nfs_client **result,
343 struct rpc_cred *cred)
347 status = nfs4_proc_exchange_id(clp, cred);
348 if (status != NFS4_OK)
350 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
352 return nfs41_walk_client_list(clp, result, cred);
355 #endif /* CONFIG_NFS_V4_1 */
358 * nfs4_get_clid_cred - Acquire credential for a setclientid operation
359 * @clp: client state handle
361 * Returns an rpc_cred with reference count bumped, or NULL.
363 struct rpc_cred *nfs4_get_clid_cred(struct nfs_client *clp)
365 struct rpc_cred *cred;
367 spin_lock(&clp->cl_lock);
368 cred = nfs4_get_machine_cred_locked(clp);
369 spin_unlock(&clp->cl_lock);
373 static struct nfs4_state_owner *
374 nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
376 struct rb_node **p = &server->state_owners.rb_node,
378 struct nfs4_state_owner *sp;
382 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
384 if (cred < sp->so_cred)
385 p = &parent->rb_left;
386 else if (cred > sp->so_cred)
387 p = &parent->rb_right;
389 if (!list_empty(&sp->so_lru))
390 list_del_init(&sp->so_lru);
391 atomic_inc(&sp->so_count);
398 static struct nfs4_state_owner *
399 nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
401 struct nfs_server *server = new->so_server;
402 struct rb_node **p = &server->state_owners.rb_node,
404 struct nfs4_state_owner *sp;
409 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
411 if (new->so_cred < sp->so_cred)
412 p = &parent->rb_left;
413 else if (new->so_cred > sp->so_cred)
414 p = &parent->rb_right;
416 if (!list_empty(&sp->so_lru))
417 list_del_init(&sp->so_lru);
418 atomic_inc(&sp->so_count);
422 err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
425 rb_link_node(&new->so_server_node, parent, p);
426 rb_insert_color(&new->so_server_node, &server->state_owners);
431 nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
433 struct nfs_server *server = sp->so_server;
435 if (!RB_EMPTY_NODE(&sp->so_server_node))
436 rb_erase(&sp->so_server_node, &server->state_owners);
437 ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
441 nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
443 sc->create_time = ktime_get();
446 spin_lock_init(&sc->lock);
447 INIT_LIST_HEAD(&sc->list);
448 rpc_init_wait_queue(&sc->wait, "Seqid_waitqueue");
452 nfs4_destroy_seqid_counter(struct nfs_seqid_counter *sc)
454 rpc_destroy_wait_queue(&sc->wait);
458 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
459 * create a new state_owner.
462 static struct nfs4_state_owner *
463 nfs4_alloc_state_owner(struct nfs_server *server,
464 struct rpc_cred *cred,
467 struct nfs4_state_owner *sp;
469 sp = kzalloc(sizeof(*sp), gfp_flags);
472 sp->so_server = server;
473 sp->so_cred = get_rpccred(cred);
474 spin_lock_init(&sp->so_lock);
475 INIT_LIST_HEAD(&sp->so_states);
476 nfs4_init_seqid_counter(&sp->so_seqid);
477 atomic_set(&sp->so_count, 1);
478 INIT_LIST_HEAD(&sp->so_lru);
479 seqcount_init(&sp->so_reclaim_seqcount);
480 mutex_init(&sp->so_delegreturn_mutex);
485 nfs4_drop_state_owner(struct nfs4_state_owner *sp)
487 struct rb_node *rb_node = &sp->so_server_node;
489 if (!RB_EMPTY_NODE(rb_node)) {
490 struct nfs_server *server = sp->so_server;
491 struct nfs_client *clp = server->nfs_client;
493 spin_lock(&clp->cl_lock);
494 if (!RB_EMPTY_NODE(rb_node)) {
495 rb_erase(rb_node, &server->state_owners);
496 RB_CLEAR_NODE(rb_node);
498 spin_unlock(&clp->cl_lock);
502 static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
504 nfs4_destroy_seqid_counter(&sp->so_seqid);
505 put_rpccred(sp->so_cred);
509 static void nfs4_gc_state_owners(struct nfs_server *server)
511 struct nfs_client *clp = server->nfs_client;
512 struct nfs4_state_owner *sp, *tmp;
513 unsigned long time_min, time_max;
516 spin_lock(&clp->cl_lock);
518 time_min = (long)time_max - (long)clp->cl_lease_time;
519 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
520 /* NB: LRU is sorted so that oldest is at the head */
521 if (time_in_range(sp->so_expires, time_min, time_max))
523 list_move(&sp->so_lru, &doomed);
524 nfs4_remove_state_owner_locked(sp);
526 spin_unlock(&clp->cl_lock);
528 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
529 list_del(&sp->so_lru);
530 nfs4_free_state_owner(sp);
535 * nfs4_get_state_owner - Look up a state owner given a credential
536 * @server: nfs_server to search
537 * @cred: RPC credential to match
539 * Returns a pointer to an instantiated nfs4_state_owner struct, or NULL.
541 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server,
542 struct rpc_cred *cred,
545 struct nfs_client *clp = server->nfs_client;
546 struct nfs4_state_owner *sp, *new;
548 spin_lock(&clp->cl_lock);
549 sp = nfs4_find_state_owner_locked(server, cred);
550 spin_unlock(&clp->cl_lock);
553 new = nfs4_alloc_state_owner(server, cred, gfp_flags);
557 if (ida_pre_get(&server->openowner_id, gfp_flags) == 0)
559 spin_lock(&clp->cl_lock);
560 sp = nfs4_insert_state_owner_locked(new);
561 spin_unlock(&clp->cl_lock);
562 } while (sp == ERR_PTR(-EAGAIN));
564 nfs4_free_state_owner(new);
566 nfs4_gc_state_owners(server);
571 * nfs4_put_state_owner - Release a nfs4_state_owner
572 * @sp: state owner data to release
574 * Note that we keep released state owners on an LRU
576 * This caches valid state owners so that they can be
577 * reused, to avoid the OPEN_CONFIRM on minor version 0.
578 * It also pins the uniquifier of dropped state owners for
579 * a while, to ensure that those state owner names are
582 void nfs4_put_state_owner(struct nfs4_state_owner *sp)
584 struct nfs_server *server = sp->so_server;
585 struct nfs_client *clp = server->nfs_client;
587 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
590 sp->so_expires = jiffies;
591 list_add_tail(&sp->so_lru, &server->state_owners_lru);
592 spin_unlock(&clp->cl_lock);
596 * nfs4_purge_state_owners - Release all cached state owners
597 * @server: nfs_server with cached state owners to release
599 * Called at umount time. Remaining state owners will be on
600 * the LRU with ref count of zero.
602 void nfs4_purge_state_owners(struct nfs_server *server)
604 struct nfs_client *clp = server->nfs_client;
605 struct nfs4_state_owner *sp, *tmp;
608 spin_lock(&clp->cl_lock);
609 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
610 list_move(&sp->so_lru, &doomed);
611 nfs4_remove_state_owner_locked(sp);
613 spin_unlock(&clp->cl_lock);
615 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
616 list_del(&sp->so_lru);
617 nfs4_free_state_owner(sp);
621 static struct nfs4_state *
622 nfs4_alloc_open_state(void)
624 struct nfs4_state *state;
626 state = kzalloc(sizeof(*state), GFP_NOFS);
629 atomic_set(&state->count, 1);
630 INIT_LIST_HEAD(&state->lock_states);
631 spin_lock_init(&state->state_lock);
632 seqlock_init(&state->seqlock);
637 nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
639 if (state->state == fmode)
641 /* NB! List reordering - see the reclaim code for why. */
642 if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
643 if (fmode & FMODE_WRITE)
644 list_move(&state->open_states, &state->owner->so_states);
646 list_move_tail(&state->open_states, &state->owner->so_states);
648 state->state = fmode;
651 static struct nfs4_state *
652 __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
654 struct nfs_inode *nfsi = NFS_I(inode);
655 struct nfs4_state *state;
657 list_for_each_entry(state, &nfsi->open_states, inode_states) {
658 if (state->owner != owner)
660 if (!nfs4_valid_open_stateid(state))
662 if (atomic_inc_not_zero(&state->count))
669 nfs4_free_open_state(struct nfs4_state *state)
675 nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
677 struct nfs4_state *state, *new;
678 struct nfs_inode *nfsi = NFS_I(inode);
680 spin_lock(&inode->i_lock);
681 state = __nfs4_find_state_byowner(inode, owner);
682 spin_unlock(&inode->i_lock);
685 new = nfs4_alloc_open_state();
686 spin_lock(&owner->so_lock);
687 spin_lock(&inode->i_lock);
688 state = __nfs4_find_state_byowner(inode, owner);
689 if (state == NULL && new != NULL) {
691 state->owner = owner;
692 atomic_inc(&owner->so_count);
693 list_add(&state->inode_states, &nfsi->open_states);
695 state->inode = inode;
696 spin_unlock(&inode->i_lock);
697 /* Note: The reclaim code dictates that we add stateless
698 * and read-only stateids to the end of the list */
699 list_add_tail(&state->open_states, &owner->so_states);
700 spin_unlock(&owner->so_lock);
702 spin_unlock(&inode->i_lock);
703 spin_unlock(&owner->so_lock);
705 nfs4_free_open_state(new);
711 void nfs4_put_open_state(struct nfs4_state *state)
713 struct inode *inode = state->inode;
714 struct nfs4_state_owner *owner = state->owner;
716 if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
718 spin_lock(&inode->i_lock);
719 list_del(&state->inode_states);
720 list_del(&state->open_states);
721 spin_unlock(&inode->i_lock);
722 spin_unlock(&owner->so_lock);
724 nfs4_free_open_state(state);
725 nfs4_put_state_owner(owner);
729 * Close the current file.
731 static void __nfs4_close(struct nfs4_state *state,
732 fmode_t fmode, gfp_t gfp_mask, int wait)
734 struct nfs4_state_owner *owner = state->owner;
738 atomic_inc(&owner->so_count);
739 /* Protect against nfs4_find_state() */
740 spin_lock(&owner->so_lock);
741 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
748 case FMODE_READ|FMODE_WRITE:
751 newstate = FMODE_READ|FMODE_WRITE;
752 if (state->n_rdwr == 0) {
753 if (state->n_rdonly == 0) {
754 newstate &= ~FMODE_READ;
755 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
756 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
758 if (state->n_wronly == 0) {
759 newstate &= ~FMODE_WRITE;
760 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
761 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
764 clear_bit(NFS_DELEGATED_STATE, &state->flags);
766 nfs4_state_set_mode_locked(state, newstate);
767 spin_unlock(&owner->so_lock);
770 nfs4_put_open_state(state);
771 nfs4_put_state_owner(owner);
773 nfs4_do_close(state, gfp_mask, wait);
776 void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
778 __nfs4_close(state, fmode, GFP_NOFS, 0);
781 void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
783 __nfs4_close(state, fmode, GFP_KERNEL, 1);
787 * Search the state->lock_states for an existing lock_owner
788 * that is compatible with current->files
790 static struct nfs4_lock_state *
791 __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type)
793 struct nfs4_lock_state *pos;
794 list_for_each_entry(pos, &state->lock_states, ls_locks) {
795 if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
797 switch (pos->ls_owner.lo_type) {
798 case NFS4_POSIX_LOCK_TYPE:
799 if (pos->ls_owner.lo_u.posix_owner != fl_owner)
802 case NFS4_FLOCK_LOCK_TYPE:
803 if (pos->ls_owner.lo_u.flock_owner != fl_pid)
806 atomic_inc(&pos->ls_count);
813 * Return a compatible lock_state. If no initialized lock_state structure
814 * exists, return an uninitialized one.
817 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)
819 struct nfs4_lock_state *lsp;
820 struct nfs_server *server = state->owner->so_server;
822 lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
825 nfs4_init_seqid_counter(&lsp->ls_seqid);
826 atomic_set(&lsp->ls_count, 1);
827 lsp->ls_state = state;
828 lsp->ls_owner.lo_type = type;
829 switch (lsp->ls_owner.lo_type) {
830 case NFS4_FLOCK_LOCK_TYPE:
831 lsp->ls_owner.lo_u.flock_owner = fl_pid;
833 case NFS4_POSIX_LOCK_TYPE:
834 lsp->ls_owner.lo_u.posix_owner = fl_owner;
839 lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
840 if (lsp->ls_seqid.owner_id < 0)
842 INIT_LIST_HEAD(&lsp->ls_locks);
849 void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
851 ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
852 nfs4_destroy_seqid_counter(&lsp->ls_seqid);
857 * Return a compatible lock_state. If no initialized lock_state structure
858 * exists, return an uninitialized one.
861 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner, pid_t pid, unsigned int type)
863 struct nfs4_lock_state *lsp, *new = NULL;
866 spin_lock(&state->state_lock);
867 lsp = __nfs4_find_lock_state(state, owner, pid, type);
871 list_add(&new->ls_locks, &state->lock_states);
872 set_bit(LK_STATE_IN_USE, &state->flags);
877 spin_unlock(&state->state_lock);
878 new = nfs4_alloc_lock_state(state, owner, pid, type);
882 spin_unlock(&state->state_lock);
884 nfs4_free_lock_state(state->owner->so_server, new);
889 * Release reference to lock_state, and free it if we see that
890 * it is no longer in use
892 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
894 struct nfs_server *server;
895 struct nfs4_state *state;
899 state = lsp->ls_state;
900 if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
902 list_del(&lsp->ls_locks);
903 if (list_empty(&state->lock_states))
904 clear_bit(LK_STATE_IN_USE, &state->flags);
905 spin_unlock(&state->state_lock);
906 server = state->owner->so_server;
907 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
908 struct nfs_client *clp = server->nfs_client;
910 clp->cl_mvops->free_lock_state(server, lsp);
912 nfs4_free_lock_state(server, lsp);
915 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
917 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
919 dst->fl_u.nfs4_fl.owner = lsp;
920 atomic_inc(&lsp->ls_count);
923 static void nfs4_fl_release_lock(struct file_lock *fl)
925 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
928 static const struct file_lock_operations nfs4_fl_lock_ops = {
929 .fl_copy_lock = nfs4_fl_copy_lock,
930 .fl_release_private = nfs4_fl_release_lock,
933 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
935 struct nfs4_lock_state *lsp;
937 if (fl->fl_ops != NULL)
939 if (fl->fl_flags & FL_POSIX)
940 lsp = nfs4_get_lock_state(state, fl->fl_owner, 0, NFS4_POSIX_LOCK_TYPE);
941 else if (fl->fl_flags & FL_FLOCK)
942 lsp = nfs4_get_lock_state(state, NULL, fl->fl_pid,
943 NFS4_FLOCK_LOCK_TYPE);
948 fl->fl_u.nfs4_fl.owner = lsp;
949 fl->fl_ops = &nfs4_fl_lock_ops;
953 static int nfs4_copy_lock_stateid(nfs4_stateid *dst,
954 struct nfs4_state *state,
955 const struct nfs_lockowner *lockowner)
957 struct nfs4_lock_state *lsp;
963 if (lockowner == NULL)
966 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
969 fl_owner = lockowner->l_owner;
970 fl_pid = lockowner->l_pid;
971 spin_lock(&state->state_lock);
972 lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
973 if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
974 nfs4_stateid_copy(dst, &lsp->ls_stateid);
977 if (!list_empty(&lsp->ls_seqid.list))
980 spin_unlock(&state->state_lock);
981 nfs4_put_lock_state(lsp);
986 static int nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
988 const nfs4_stateid *src;
994 seq = read_seqbegin(&state->seqlock);
995 if (test_bit(NFS_OPEN_STATE, &state->flags))
996 src = &state->open_stateid;
997 nfs4_stateid_copy(dst, src);
1000 if (!list_empty(&state->owner->so_seqid.list))
1002 } while (read_seqretry(&state->seqlock, seq));
1007 * Byte-range lock aware utility to initialize the stateid of read/write
1010 int nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state,
1011 fmode_t fmode, const struct nfs_lockowner *lockowner)
1014 if (nfs4_copy_delegation_stateid(dst, state->inode, fmode))
1016 ret = nfs4_copy_lock_stateid(dst, state, lockowner);
1019 ret = nfs4_copy_open_stateid(dst, state);
1021 if (nfs_server_capable(state->inode, NFS_CAP_STATEID_NFSV41))
1026 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
1028 struct nfs_seqid *new;
1030 new = kmalloc(sizeof(*new), gfp_mask);
1032 new->sequence = counter;
1033 INIT_LIST_HEAD(&new->list);
1039 void nfs_release_seqid(struct nfs_seqid *seqid)
1041 struct nfs_seqid_counter *sequence;
1043 if (list_empty(&seqid->list))
1045 sequence = seqid->sequence;
1046 spin_lock(&sequence->lock);
1047 list_del_init(&seqid->list);
1048 if (!list_empty(&sequence->list)) {
1049 struct nfs_seqid *next;
1051 next = list_first_entry(&sequence->list,
1052 struct nfs_seqid, list);
1053 rpc_wake_up_queued_task(&sequence->wait, next->task);
1055 spin_unlock(&sequence->lock);
1058 void nfs_free_seqid(struct nfs_seqid *seqid)
1060 nfs_release_seqid(seqid);
1065 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
1066 * failed with a seqid incrementing error -
1067 * see comments nfs_fs.h:seqid_mutating_error()
1069 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
1074 case -NFS4ERR_BAD_SEQID:
1075 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
1077 pr_warn_ratelimited("NFS: v4 server returned a bad"
1078 " sequence-id error on an"
1079 " unconfirmed sequence %p!\n",
1081 case -NFS4ERR_STALE_CLIENTID:
1082 case -NFS4ERR_STALE_STATEID:
1083 case -NFS4ERR_BAD_STATEID:
1084 case -NFS4ERR_BADXDR:
1085 case -NFS4ERR_RESOURCE:
1086 case -NFS4ERR_NOFILEHANDLE:
1087 /* Non-seqid mutating errors */
1091 * Note: no locking needed as we are guaranteed to be first
1092 * on the sequence list
1094 seqid->sequence->counter++;
1097 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
1099 struct nfs4_state_owner *sp = container_of(seqid->sequence,
1100 struct nfs4_state_owner, so_seqid);
1101 struct nfs_server *server = sp->so_server;
1103 if (status == -NFS4ERR_BAD_SEQID)
1104 nfs4_drop_state_owner(sp);
1105 if (!nfs4_has_session(server->nfs_client))
1106 nfs_increment_seqid(status, seqid);
1110 * Increment the seqid if the LOCK/LOCKU succeeded, or
1111 * failed with a seqid incrementing error -
1112 * see comments nfs_fs.h:seqid_mutating_error()
1114 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
1116 nfs_increment_seqid(status, seqid);
1119 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
1121 struct nfs_seqid_counter *sequence = seqid->sequence;
1124 spin_lock(&sequence->lock);
1126 if (list_empty(&seqid->list))
1127 list_add_tail(&seqid->list, &sequence->list);
1128 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
1130 rpc_sleep_on(&sequence->wait, task, NULL);
1133 spin_unlock(&sequence->lock);
1137 static int nfs4_run_state_manager(void *);
1139 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
1141 smp_mb__before_clear_bit();
1142 clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
1143 smp_mb__after_clear_bit();
1144 wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
1145 rpc_wake_up(&clp->cl_rpcwaitq);
1149 * Schedule the nfs_client asynchronous state management routine
1151 void nfs4_schedule_state_manager(struct nfs_client *clp)
1153 struct task_struct *task;
1154 char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
1156 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1158 __module_get(THIS_MODULE);
1159 atomic_inc(&clp->cl_count);
1161 /* The rcu_read_lock() is not strictly necessary, as the state
1162 * manager is the only thread that ever changes the rpc_xprt
1163 * after it's initialized. At this point, we're single threaded. */
1165 snprintf(buf, sizeof(buf), "%s-manager",
1166 rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR));
1168 task = kthread_run(nfs4_run_state_manager, clp, "%s", buf);
1170 printk(KERN_ERR "%s: kthread_run: %ld\n",
1171 __func__, PTR_ERR(task));
1172 nfs4_clear_state_manager_bit(clp);
1173 nfs_put_client(clp);
1174 module_put(THIS_MODULE);
1179 * Schedule a lease recovery attempt
1181 void nfs4_schedule_lease_recovery(struct nfs_client *clp)
1185 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1186 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1187 dprintk("%s: scheduling lease recovery for server %s\n", __func__,
1189 nfs4_schedule_state_manager(clp);
1191 EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
1193 int nfs4_wait_clnt_recover(struct nfs_client *clp)
1199 res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
1200 nfs_wait_bit_killable, TASK_KILLABLE);
1204 if (clp->cl_cons_state < 0)
1205 return clp->cl_cons_state;
1209 int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1214 for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1215 ret = nfs4_wait_clnt_recover(clp);
1218 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1219 !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1221 nfs4_schedule_state_manager(clp);
1228 * nfs40_handle_cb_pathdown - return all delegations after NFS4ERR_CB_PATH_DOWN
1229 * @clp: client to process
1231 * Set the NFS4CLNT_LEASE_EXPIRED state in order to force a
1232 * resend of the SETCLIENTID and hence re-establish the
1233 * callback channel. Then return all existing delegations.
1235 static void nfs40_handle_cb_pathdown(struct nfs_client *clp)
1237 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1238 nfs_expire_all_delegations(clp);
1239 dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
1243 void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
1245 nfs40_handle_cb_pathdown(clp);
1246 nfs4_schedule_state_manager(clp);
1249 static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1252 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1253 /* Don't recover state that expired before the reboot */
1254 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
1255 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1258 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1259 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1263 static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1265 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
1266 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1267 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1268 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
1272 int nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
1274 struct nfs_client *clp = server->nfs_client;
1276 if (!nfs4_valid_open_stateid(state))
1278 nfs4_state_mark_reclaim_nograce(clp, state);
1279 dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
1281 nfs4_schedule_state_manager(clp);
1284 EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1286 void nfs_inode_find_state_and_recover(struct inode *inode,
1287 const nfs4_stateid *stateid)
1289 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1290 struct nfs_inode *nfsi = NFS_I(inode);
1291 struct nfs_open_context *ctx;
1292 struct nfs4_state *state;
1295 spin_lock(&inode->i_lock);
1296 list_for_each_entry(ctx, &nfsi->open_files, list) {
1300 if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
1302 if (!nfs4_stateid_match(&state->stateid, stateid))
1304 nfs4_state_mark_reclaim_nograce(clp, state);
1307 spin_unlock(&inode->i_lock);
1309 nfs4_schedule_state_manager(clp);
1312 static void nfs4_state_mark_open_context_bad(struct nfs4_state *state)
1314 struct inode *inode = state->inode;
1315 struct nfs_inode *nfsi = NFS_I(inode);
1316 struct nfs_open_context *ctx;
1318 spin_lock(&inode->i_lock);
1319 list_for_each_entry(ctx, &nfsi->open_files, list) {
1320 if (ctx->state != state)
1322 set_bit(NFS_CONTEXT_BAD, &ctx->flags);
1324 spin_unlock(&inode->i_lock);
1327 static void nfs4_state_mark_recovery_failed(struct nfs4_state *state, int error)
1329 set_bit(NFS_STATE_RECOVERY_FAILED, &state->flags);
1330 nfs4_state_mark_open_context_bad(state);
1334 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
1336 struct inode *inode = state->inode;
1337 struct nfs_inode *nfsi = NFS_I(inode);
1338 struct file_lock *fl;
1341 if (inode->i_flock == NULL)
1344 /* Guard against delegation returns and new lock/unlock calls */
1345 down_write(&nfsi->rwsem);
1346 /* Protect inode->i_flock using the BKL */
1347 spin_lock(&inode->i_lock);
1348 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1349 if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
1351 if (nfs_file_open_context(fl->fl_file)->state != state)
1353 spin_unlock(&inode->i_lock);
1354 status = ops->recover_lock(state, fl);
1359 case -NFS4ERR_ADMIN_REVOKED:
1360 case -NFS4ERR_STALE_STATEID:
1361 case -NFS4ERR_BAD_STATEID:
1362 case -NFS4ERR_EXPIRED:
1363 case -NFS4ERR_NO_GRACE:
1364 case -NFS4ERR_STALE_CLIENTID:
1365 case -NFS4ERR_BADSESSION:
1366 case -NFS4ERR_BADSLOT:
1367 case -NFS4ERR_BAD_HIGH_SLOT:
1368 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1371 printk(KERN_ERR "NFS: %s: unhandled error %d. "
1372 "Zeroing state\n", __func__, status);
1374 case -NFS4ERR_DENIED:
1375 case -NFS4ERR_RECLAIM_BAD:
1376 case -NFS4ERR_RECLAIM_CONFLICT:
1377 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1380 spin_lock(&inode->i_lock);
1382 spin_unlock(&inode->i_lock);
1384 up_write(&nfsi->rwsem);
1388 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
1390 struct nfs4_state *state;
1391 struct nfs4_lock_state *lock;
1394 /* Note: we rely on the sp->so_states list being ordered
1395 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
1397 * This is needed to ensure that the server won't give us any
1398 * read delegations that we have to return if, say, we are
1399 * recovering after a network partition or a reboot from a
1400 * server that doesn't support a grace period.
1402 spin_lock(&sp->so_lock);
1403 write_seqcount_begin(&sp->so_reclaim_seqcount);
1405 list_for_each_entry(state, &sp->so_states, open_states) {
1406 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
1408 if (!nfs4_valid_open_stateid(state))
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 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0) {
1419 spin_lock(&state->state_lock);
1420 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1421 if (!test_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags))
1422 pr_warn_ratelimited("NFS: "
1424 "failed!\n", __func__);
1426 spin_unlock(&state->state_lock);
1428 nfs4_put_open_state(state);
1429 spin_lock(&sp->so_lock);
1435 printk(KERN_ERR "NFS: %s: unhandled error %d. "
1436 "Zeroing state\n", __func__, status);
1441 * Open state on this file cannot be recovered
1442 * All we can do is revert to using the zero stateid.
1444 nfs4_state_mark_recovery_failed(state, status);
1448 case -NFS4ERR_ADMIN_REVOKED:
1449 case -NFS4ERR_STALE_STATEID:
1450 case -NFS4ERR_BAD_STATEID:
1451 case -NFS4ERR_RECLAIM_BAD:
1452 case -NFS4ERR_RECLAIM_CONFLICT:
1453 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1455 case -NFS4ERR_EXPIRED:
1456 case -NFS4ERR_NO_GRACE:
1457 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1458 case -NFS4ERR_STALE_CLIENTID:
1459 case -NFS4ERR_BADSESSION:
1460 case -NFS4ERR_BADSLOT:
1461 case -NFS4ERR_BAD_HIGH_SLOT:
1462 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1465 nfs4_put_open_state(state);
1466 spin_lock(&sp->so_lock);
1469 write_seqcount_end(&sp->so_reclaim_seqcount);
1470 spin_unlock(&sp->so_lock);
1473 nfs4_put_open_state(state);
1474 spin_lock(&sp->so_lock);
1475 write_seqcount_end(&sp->so_reclaim_seqcount);
1476 spin_unlock(&sp->so_lock);
1480 static void nfs4_clear_open_state(struct nfs4_state *state)
1482 struct nfs4_lock_state *lock;
1484 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1485 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1486 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1487 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1488 spin_lock(&state->state_lock);
1489 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1490 lock->ls_seqid.flags = 0;
1491 clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
1493 spin_unlock(&state->state_lock);
1496 static void nfs4_reset_seqids(struct nfs_server *server,
1497 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1499 struct nfs_client *clp = server->nfs_client;
1500 struct nfs4_state_owner *sp;
1501 struct rb_node *pos;
1502 struct nfs4_state *state;
1504 spin_lock(&clp->cl_lock);
1505 for (pos = rb_first(&server->state_owners);
1507 pos = rb_next(pos)) {
1508 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1509 sp->so_seqid.flags = 0;
1510 spin_lock(&sp->so_lock);
1511 list_for_each_entry(state, &sp->so_states, open_states) {
1512 if (mark_reclaim(clp, state))
1513 nfs4_clear_open_state(state);
1515 spin_unlock(&sp->so_lock);
1517 spin_unlock(&clp->cl_lock);
1520 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp,
1521 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1523 struct nfs_server *server;
1526 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1527 nfs4_reset_seqids(server, mark_reclaim);
1531 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1533 /* Mark all delegations for reclaim */
1534 nfs_delegation_mark_reclaim(clp);
1535 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1538 static void nfs4_reclaim_complete(struct nfs_client *clp,
1539 const struct nfs4_state_recovery_ops *ops,
1540 struct rpc_cred *cred)
1542 /* Notify the server we're done reclaiming our state */
1543 if (ops->reclaim_complete)
1544 (void)ops->reclaim_complete(clp, cred);
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 const struct nfs4_state_recovery_ops *ops;
1590 struct rpc_cred *cred;
1592 if (!nfs4_state_clear_reclaim_reboot(clp))
1594 ops = clp->cl_mvops->reboot_recovery_ops;
1595 cred = nfs4_get_clid_cred(clp);
1596 nfs4_reclaim_complete(clp, ops, cred);
1600 static void nfs_delegation_clear_all(struct nfs_client *clp)
1602 nfs_delegation_mark_reclaim(clp);
1603 nfs_delegation_reap_unclaimed(clp);
1606 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1608 nfs_delegation_clear_all(clp);
1609 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1612 static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1617 case -NFS4ERR_CB_PATH_DOWN:
1618 nfs40_handle_cb_pathdown(clp);
1620 case -NFS4ERR_NO_GRACE:
1621 nfs4_state_end_reclaim_reboot(clp);
1623 case -NFS4ERR_STALE_CLIENTID:
1624 case -NFS4ERR_LEASE_MOVED:
1625 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1626 nfs4_state_clear_reclaim_reboot(clp);
1627 nfs4_state_start_reclaim_reboot(clp);
1629 case -NFS4ERR_EXPIRED:
1630 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1631 nfs4_state_start_reclaim_nograce(clp);
1633 case -NFS4ERR_BADSESSION:
1634 case -NFS4ERR_BADSLOT:
1635 case -NFS4ERR_BAD_HIGH_SLOT:
1636 case -NFS4ERR_DEADSESSION:
1637 case -NFS4ERR_SEQ_FALSE_RETRY:
1638 case -NFS4ERR_SEQ_MISORDERED:
1639 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1640 /* Zero session reset errors */
1642 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1643 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1646 dprintk("%s: failed to handle error %d for server %s\n",
1647 __func__, error, clp->cl_hostname);
1650 dprintk("%s: handled error %d for server %s\n", __func__, error,
1655 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1657 struct nfs4_state_owner *sp;
1658 struct nfs_server *server;
1659 struct rb_node *pos;
1664 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1665 nfs4_purge_state_owners(server);
1666 spin_lock(&clp->cl_lock);
1667 for (pos = rb_first(&server->state_owners);
1669 pos = rb_next(pos)) {
1671 struct nfs4_state_owner, so_server_node);
1672 if (!test_and_clear_bit(ops->owner_flag_bit,
1675 atomic_inc(&sp->so_count);
1676 spin_unlock(&clp->cl_lock);
1679 status = nfs4_reclaim_open_state(sp, ops);
1681 set_bit(ops->owner_flag_bit, &sp->so_flags);
1682 nfs4_put_state_owner(sp);
1683 return nfs4_recovery_handle_error(clp, status);
1686 nfs4_put_state_owner(sp);
1689 spin_unlock(&clp->cl_lock);
1695 static int nfs4_check_lease(struct nfs_client *clp)
1697 struct rpc_cred *cred;
1698 const struct nfs4_state_maintenance_ops *ops =
1699 clp->cl_mvops->state_renewal_ops;
1702 /* Is the client already known to have an expired lease? */
1703 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1705 spin_lock(&clp->cl_lock);
1706 cred = ops->get_state_renewal_cred_locked(clp);
1707 spin_unlock(&clp->cl_lock);
1709 cred = nfs4_get_clid_cred(clp);
1714 status = ops->renew_lease(clp, cred);
1716 if (status == -ETIMEDOUT) {
1717 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1721 return nfs4_recovery_handle_error(clp, status);
1724 /* Set NFS4CLNT_LEASE_EXPIRED and reclaim reboot state for all v4.0 errors
1725 * and for recoverable errors on EXCHANGE_ID for v4.1
1727 static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
1730 case -NFS4ERR_SEQ_MISORDERED:
1731 if (test_and_set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state))
1732 return -ESERVERFAULT;
1733 /* Lease confirmation error: retry after purging the lease */
1735 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1737 case -NFS4ERR_STALE_CLIENTID:
1738 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1739 nfs4_state_clear_reclaim_reboot(clp);
1740 nfs4_state_start_reclaim_reboot(clp);
1742 case -NFS4ERR_CLID_INUSE:
1743 pr_err("NFS: Server %s reports our clientid is in use\n",
1745 nfs_mark_client_ready(clp, -EPERM);
1746 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1749 case -NFS4ERR_DELAY:
1755 case -NFS4ERR_MINOR_VERS_MISMATCH:
1756 if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
1757 nfs_mark_client_ready(clp, -EPROTONOSUPPORT);
1758 dprintk("%s: exit with error %d for server %s\n",
1759 __func__, -EPROTONOSUPPORT, clp->cl_hostname);
1760 return -EPROTONOSUPPORT;
1761 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1762 * in nfs4_exchange_id */
1764 dprintk("%s: exit with error %d for server %s\n", __func__,
1765 status, clp->cl_hostname);
1768 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1769 dprintk("%s: handled error %d for server %s\n", __func__, status,
1774 static int nfs4_establish_lease(struct nfs_client *clp)
1776 struct rpc_cred *cred;
1777 const struct nfs4_state_recovery_ops *ops =
1778 clp->cl_mvops->reboot_recovery_ops;
1781 cred = nfs4_get_clid_cred(clp);
1784 status = ops->establish_clid(clp, cred);
1788 pnfs_destroy_all_layouts(clp);
1793 * Returns zero or a negative errno. NFS4ERR values are converted
1794 * to local errno values.
1796 static int nfs4_reclaim_lease(struct nfs_client *clp)
1800 status = nfs4_establish_lease(clp);
1802 return nfs4_handle_reclaim_lease_error(clp, status);
1803 if (test_and_clear_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state))
1804 nfs4_state_start_reclaim_nograce(clp);
1805 if (!test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
1806 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1807 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1808 clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1812 static int nfs4_purge_lease(struct nfs_client *clp)
1816 status = nfs4_establish_lease(clp);
1818 return nfs4_handle_reclaim_lease_error(clp, status);
1819 clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1820 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1821 nfs4_state_start_reclaim_nograce(clp);
1826 * nfs4_discover_server_trunking - Detect server IP address trunking
1828 * @clp: nfs_client under test
1829 * @result: OUT: found nfs_client, or clp
1831 * Returns zero or a negative errno. If zero is returned,
1832 * an nfs_client pointer is planted in "result".
1834 * Note: since we are invoked in process context, and
1835 * not from inside the state manager, we cannot use
1836 * nfs4_handle_reclaim_lease_error().
1838 int nfs4_discover_server_trunking(struct nfs_client *clp,
1839 struct nfs_client **result)
1841 const struct nfs4_state_recovery_ops *ops =
1842 clp->cl_mvops->reboot_recovery_ops;
1843 struct rpc_clnt *clnt;
1844 struct rpc_cred *cred;
1847 dprintk("NFS: %s: testing '%s'\n", __func__, clp->cl_hostname);
1849 clnt = clp->cl_rpcclient;
1852 mutex_lock(&nfs_clid_init_mutex);
1855 cred = nfs4_get_clid_cred(clp);
1859 status = ops->detect_trunking(clp, result, cred);
1864 case -NFS4ERR_DELAY:
1868 case -NFS4ERR_STALE_CLIENTID:
1869 dprintk("NFS: %s after status %d, retrying\n",
1874 nfs4_root_machine_cred(clp);
1879 case -NFS4ERR_CLID_INUSE:
1880 case -NFS4ERR_WRONGSEC:
1881 clnt = rpc_clone_client_set_auth(clnt, RPC_AUTH_UNIX);
1883 status = PTR_ERR(clnt);
1886 /* Note: this is safe because we haven't yet marked the
1887 * client as ready, so we are the only user of
1890 clnt = xchg(&clp->cl_rpcclient, clnt);
1891 rpc_shutdown_client(clnt);
1892 clnt = clp->cl_rpcclient;
1895 case -NFS4ERR_MINOR_VERS_MISMATCH:
1896 status = -EPROTONOSUPPORT;
1900 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1901 * in nfs4_exchange_id */
1902 status = -EKEYEXPIRED;
1905 pr_warn("NFS: %s unhandled error %d. Exiting with error EIO\n",
1911 mutex_unlock(&nfs_clid_init_mutex);
1912 dprintk("NFS: %s: status = %d\n", __func__, status);
1916 #ifdef CONFIG_NFS_V4_1
1917 void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
1919 struct nfs_client *clp = session->clp;
1923 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1925 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1926 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1928 nfs4_schedule_lease_recovery(clp);
1930 EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
1932 static void nfs41_ping_server(struct nfs_client *clp)
1934 /* Use CHECK_LEASE to ping the server with a SEQUENCE */
1935 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1936 nfs4_schedule_state_manager(clp);
1939 void nfs41_server_notify_target_slotid_update(struct nfs_client *clp)
1941 nfs41_ping_server(clp);
1944 void nfs41_server_notify_highest_slotid_update(struct nfs_client *clp)
1946 nfs41_ping_server(clp);
1949 static void nfs4_reset_all_state(struct nfs_client *clp)
1951 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1952 set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1953 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1954 nfs4_state_start_reclaim_nograce(clp);
1955 dprintk("%s: scheduling reset of all state for server %s!\n",
1956 __func__, clp->cl_hostname);
1957 nfs4_schedule_state_manager(clp);
1961 static void nfs41_handle_server_reboot(struct nfs_client *clp)
1963 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1964 nfs4_state_start_reclaim_reboot(clp);
1965 dprintk("%s: server %s rebooted!\n", __func__,
1967 nfs4_schedule_state_manager(clp);
1971 static void nfs41_handle_state_revoked(struct nfs_client *clp)
1973 nfs4_reset_all_state(clp);
1974 dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
1977 static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
1979 /* This will need to handle layouts too */
1980 nfs_expire_all_delegations(clp);
1981 dprintk("%s: Recallable state revoked on server %s!\n", __func__,
1985 static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
1987 nfs_expire_all_delegations(clp);
1988 if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
1989 nfs4_schedule_state_manager(clp);
1990 dprintk("%s: server %s declared a backchannel fault\n", __func__,
1994 static void nfs41_handle_cb_path_down(struct nfs_client *clp)
1996 if (test_and_set_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
1997 &clp->cl_state) == 0)
1998 nfs4_schedule_state_manager(clp);
2001 void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
2006 dprintk("%s: \"%s\" (client ID %llx) flags=0x%08x\n",
2007 __func__, clp->cl_hostname, clp->cl_clientid, flags);
2009 if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
2010 nfs41_handle_server_reboot(clp);
2011 if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
2012 SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
2013 SEQ4_STATUS_ADMIN_STATE_REVOKED |
2014 SEQ4_STATUS_LEASE_MOVED))
2015 nfs41_handle_state_revoked(clp);
2016 if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
2017 nfs41_handle_recallable_state_revoked(clp);
2018 if (flags & SEQ4_STATUS_BACKCHANNEL_FAULT)
2019 nfs41_handle_backchannel_fault(clp);
2020 else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
2021 SEQ4_STATUS_CB_PATH_DOWN_SESSION))
2022 nfs41_handle_cb_path_down(clp);
2025 static int nfs4_reset_session(struct nfs_client *clp)
2027 struct rpc_cred *cred;
2030 if (!nfs4_has_session(clp))
2032 nfs4_begin_drain_session(clp);
2033 cred = nfs4_get_clid_cred(clp);
2034 status = nfs4_proc_destroy_session(clp->cl_session, cred);
2037 case -NFS4ERR_BADSESSION:
2038 case -NFS4ERR_DEADSESSION:
2040 case -NFS4ERR_BACK_CHAN_BUSY:
2041 case -NFS4ERR_DELAY:
2042 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
2047 status = nfs4_recovery_handle_error(clp, status);
2051 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
2052 status = nfs4_proc_create_session(clp, cred);
2054 dprintk("%s: session reset failed with status %d for server %s!\n",
2055 __func__, status, clp->cl_hostname);
2056 status = nfs4_handle_reclaim_lease_error(clp, status);
2059 nfs41_finish_session_reset(clp);
2060 dprintk("%s: session reset was successful for server %s!\n",
2061 __func__, clp->cl_hostname);
2068 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2070 struct rpc_cred *cred;
2073 if (!nfs4_has_session(clp))
2075 nfs4_begin_drain_session(clp);
2076 cred = nfs4_get_clid_cred(clp);
2077 ret = nfs4_proc_bind_conn_to_session(clp, cred);
2080 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2083 dprintk("%s: bind_conn_to_session was successful for server %s!\n",
2084 __func__, clp->cl_hostname);
2086 case -NFS4ERR_DELAY:
2088 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2091 return nfs4_recovery_handle_error(clp, ret);
2095 #else /* CONFIG_NFS_V4_1 */
2096 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
2097 static void nfs4_end_drain_session(struct nfs_client *clp) { }
2099 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2103 #endif /* CONFIG_NFS_V4_1 */
2105 static void nfs4_state_manager(struct nfs_client *clp)
2108 const char *section = "", *section_sep = "";
2110 /* Ensure exclusive access to NFSv4 state */
2112 if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
2113 section = "purge state";
2114 status = nfs4_purge_lease(clp);
2120 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
2121 section = "lease expired";
2122 /* We're going to have to re-establish a clientid */
2123 status = nfs4_reclaim_lease(clp);
2129 /* Initialize or reset the session */
2130 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
2131 section = "reset session";
2132 status = nfs4_reset_session(clp);
2133 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
2139 /* Send BIND_CONN_TO_SESSION */
2140 if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2142 section = "bind conn to session";
2143 status = nfs4_bind_conn_to_session(clp);
2149 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
2150 section = "check lease";
2151 status = nfs4_check_lease(clp);
2157 /* First recover reboot state... */
2158 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
2159 section = "reclaim reboot";
2160 status = nfs4_do_reclaim(clp,
2161 clp->cl_mvops->reboot_recovery_ops);
2162 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2163 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
2165 nfs4_state_end_reclaim_reboot(clp);
2166 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
2172 /* Now recover expired state... */
2173 if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
2174 section = "reclaim nograce";
2175 status = nfs4_do_reclaim(clp,
2176 clp->cl_mvops->nograce_recovery_ops);
2177 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2178 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
2179 test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
2185 nfs4_end_drain_session(clp);
2186 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
2187 nfs_client_return_marked_delegations(clp);
2191 nfs4_clear_state_manager_bit(clp);
2192 /* Did we race with an attempt to give us more work? */
2193 if (clp->cl_state == 0)
2195 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
2197 } while (atomic_read(&clp->cl_count) > 1);
2200 if (strlen(section))
2202 pr_warn_ratelimited("NFS: state manager%s%s failed on NFSv4 server %s"
2203 " with error %d\n", section_sep, section,
2204 clp->cl_hostname, -status);
2206 nfs4_end_drain_session(clp);
2207 nfs4_clear_state_manager_bit(clp);
2210 static int nfs4_run_state_manager(void *ptr)
2212 struct nfs_client *clp = ptr;
2214 allow_signal(SIGKILL);
2215 nfs4_state_manager(clp);
2216 nfs_put_client(clp);
2217 module_put_and_exit(0);