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 static void nfs4_end_drain_slot_table(struct nfs4_slot_table *tbl)
220 if (test_and_clear_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state)) {
221 spin_lock(&tbl->slot_tbl_lock);
222 nfs41_wake_slot_table(tbl);
223 spin_unlock(&tbl->slot_tbl_lock);
227 static void nfs4_end_drain_session(struct nfs_client *clp)
229 struct nfs4_session *ses = clp->cl_session;
231 if (clp->cl_slot_tbl) {
232 nfs4_end_drain_slot_table(clp->cl_slot_tbl);
237 nfs4_end_drain_slot_table(&ses->bc_slot_table);
238 nfs4_end_drain_slot_table(&ses->fc_slot_table);
242 #if defined(CONFIG_NFS_V4_1)
244 static int nfs4_drain_slot_tbl(struct nfs4_slot_table *tbl)
246 set_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state);
247 spin_lock(&tbl->slot_tbl_lock);
248 if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
249 INIT_COMPLETION(tbl->complete);
250 spin_unlock(&tbl->slot_tbl_lock);
251 return wait_for_completion_interruptible(&tbl->complete);
253 spin_unlock(&tbl->slot_tbl_lock);
257 static int nfs4_begin_drain_session(struct nfs_client *clp)
259 struct nfs4_session *ses = clp->cl_session;
262 if (clp->cl_slot_tbl)
263 return nfs4_drain_slot_tbl(clp->cl_slot_tbl);
266 ret = nfs4_drain_slot_tbl(&ses->bc_slot_table);
270 return nfs4_drain_slot_tbl(&ses->fc_slot_table);
273 static int nfs41_setup_state_renewal(struct nfs_client *clp)
276 struct nfs_fsinfo fsinfo;
278 if (!test_bit(NFS_CS_CHECK_LEASE_TIME, &clp->cl_res_state)) {
279 nfs4_schedule_state_renewal(clp);
283 status = nfs4_proc_get_lease_time(clp, &fsinfo);
285 /* Update lease time and schedule renewal */
286 spin_lock(&clp->cl_lock);
287 clp->cl_lease_time = fsinfo.lease_time * HZ;
288 clp->cl_last_renewal = jiffies;
289 spin_unlock(&clp->cl_lock);
291 nfs4_schedule_state_renewal(clp);
297 static void nfs41_finish_session_reset(struct nfs_client *clp)
299 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
300 clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
301 /* create_session negotiated new slot table */
302 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
303 nfs41_setup_state_renewal(clp);
306 int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
310 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
312 nfs4_begin_drain_session(clp);
313 status = nfs4_proc_exchange_id(clp, cred);
316 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
318 status = nfs4_proc_create_session(clp, cred);
321 nfs41_finish_session_reset(clp);
322 nfs_mark_client_ready(clp, NFS_CS_READY);
328 * nfs41_discover_server_trunking - Detect server IP address trunking (mv1)
330 * @clp: nfs_client under test
331 * @result: OUT: found nfs_client, or clp
332 * @cred: credential to use for trunking test
334 * Returns NFS4_OK, a negative errno, or a negative NFS4ERR status.
335 * If NFS4_OK is returned, an nfs_client pointer is planted in
338 * Note: The returned client may not yet be marked ready.
340 int nfs41_discover_server_trunking(struct nfs_client *clp,
341 struct nfs_client **result,
342 struct rpc_cred *cred)
346 status = nfs4_proc_exchange_id(clp, cred);
347 if (status != NFS4_OK)
349 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
351 return nfs41_walk_client_list(clp, result, cred);
354 #endif /* CONFIG_NFS_V4_1 */
357 * nfs4_get_clid_cred - Acquire credential for a setclientid operation
358 * @clp: client state handle
360 * Returns an rpc_cred with reference count bumped, or NULL.
362 struct rpc_cred *nfs4_get_clid_cred(struct nfs_client *clp)
364 struct rpc_cred *cred;
366 spin_lock(&clp->cl_lock);
367 cred = nfs4_get_machine_cred_locked(clp);
368 spin_unlock(&clp->cl_lock);
372 static struct nfs4_state_owner *
373 nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
375 struct rb_node **p = &server->state_owners.rb_node,
377 struct nfs4_state_owner *sp;
381 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
383 if (cred < sp->so_cred)
384 p = &parent->rb_left;
385 else if (cred > sp->so_cred)
386 p = &parent->rb_right;
388 if (!list_empty(&sp->so_lru))
389 list_del_init(&sp->so_lru);
390 atomic_inc(&sp->so_count);
397 static struct nfs4_state_owner *
398 nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
400 struct nfs_server *server = new->so_server;
401 struct rb_node **p = &server->state_owners.rb_node,
403 struct nfs4_state_owner *sp;
408 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
410 if (new->so_cred < sp->so_cred)
411 p = &parent->rb_left;
412 else if (new->so_cred > sp->so_cred)
413 p = &parent->rb_right;
415 if (!list_empty(&sp->so_lru))
416 list_del_init(&sp->so_lru);
417 atomic_inc(&sp->so_count);
421 err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
424 rb_link_node(&new->so_server_node, parent, p);
425 rb_insert_color(&new->so_server_node, &server->state_owners);
430 nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
432 struct nfs_server *server = sp->so_server;
434 if (!RB_EMPTY_NODE(&sp->so_server_node))
435 rb_erase(&sp->so_server_node, &server->state_owners);
436 ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
440 nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
442 sc->create_time = ktime_get();
445 spin_lock_init(&sc->lock);
446 INIT_LIST_HEAD(&sc->list);
447 rpc_init_wait_queue(&sc->wait, "Seqid_waitqueue");
451 nfs4_destroy_seqid_counter(struct nfs_seqid_counter *sc)
453 rpc_destroy_wait_queue(&sc->wait);
457 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
458 * create a new state_owner.
461 static struct nfs4_state_owner *
462 nfs4_alloc_state_owner(struct nfs_server *server,
463 struct rpc_cred *cred,
466 struct nfs4_state_owner *sp;
468 sp = kzalloc(sizeof(*sp), gfp_flags);
471 sp->so_server = server;
472 sp->so_cred = get_rpccred(cred);
473 spin_lock_init(&sp->so_lock);
474 INIT_LIST_HEAD(&sp->so_states);
475 nfs4_init_seqid_counter(&sp->so_seqid);
476 atomic_set(&sp->so_count, 1);
477 INIT_LIST_HEAD(&sp->so_lru);
478 seqcount_init(&sp->so_reclaim_seqcount);
479 mutex_init(&sp->so_delegreturn_mutex);
484 nfs4_drop_state_owner(struct nfs4_state_owner *sp)
486 struct rb_node *rb_node = &sp->so_server_node;
488 if (!RB_EMPTY_NODE(rb_node)) {
489 struct nfs_server *server = sp->so_server;
490 struct nfs_client *clp = server->nfs_client;
492 spin_lock(&clp->cl_lock);
493 if (!RB_EMPTY_NODE(rb_node)) {
494 rb_erase(rb_node, &server->state_owners);
495 RB_CLEAR_NODE(rb_node);
497 spin_unlock(&clp->cl_lock);
501 static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
503 nfs4_destroy_seqid_counter(&sp->so_seqid);
504 put_rpccred(sp->so_cred);
508 static void nfs4_gc_state_owners(struct nfs_server *server)
510 struct nfs_client *clp = server->nfs_client;
511 struct nfs4_state_owner *sp, *tmp;
512 unsigned long time_min, time_max;
515 spin_lock(&clp->cl_lock);
517 time_min = (long)time_max - (long)clp->cl_lease_time;
518 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
519 /* NB: LRU is sorted so that oldest is at the head */
520 if (time_in_range(sp->so_expires, time_min, time_max))
522 list_move(&sp->so_lru, &doomed);
523 nfs4_remove_state_owner_locked(sp);
525 spin_unlock(&clp->cl_lock);
527 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
528 list_del(&sp->so_lru);
529 nfs4_free_state_owner(sp);
534 * nfs4_get_state_owner - Look up a state owner given a credential
535 * @server: nfs_server to search
536 * @cred: RPC credential to match
538 * Returns a pointer to an instantiated nfs4_state_owner struct, or NULL.
540 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server,
541 struct rpc_cred *cred,
544 struct nfs_client *clp = server->nfs_client;
545 struct nfs4_state_owner *sp, *new;
547 spin_lock(&clp->cl_lock);
548 sp = nfs4_find_state_owner_locked(server, cred);
549 spin_unlock(&clp->cl_lock);
552 new = nfs4_alloc_state_owner(server, cred, gfp_flags);
556 if (ida_pre_get(&server->openowner_id, gfp_flags) == 0)
558 spin_lock(&clp->cl_lock);
559 sp = nfs4_insert_state_owner_locked(new);
560 spin_unlock(&clp->cl_lock);
561 } while (sp == ERR_PTR(-EAGAIN));
563 nfs4_free_state_owner(new);
565 nfs4_gc_state_owners(server);
570 * nfs4_put_state_owner - Release a nfs4_state_owner
571 * @sp: state owner data to release
573 * Note that we keep released state owners on an LRU
575 * This caches valid state owners so that they can be
576 * reused, to avoid the OPEN_CONFIRM on minor version 0.
577 * It also pins the uniquifier of dropped state owners for
578 * a while, to ensure that those state owner names are
581 void nfs4_put_state_owner(struct nfs4_state_owner *sp)
583 struct nfs_server *server = sp->so_server;
584 struct nfs_client *clp = server->nfs_client;
586 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
589 sp->so_expires = jiffies;
590 list_add_tail(&sp->so_lru, &server->state_owners_lru);
591 spin_unlock(&clp->cl_lock);
595 * nfs4_purge_state_owners - Release all cached state owners
596 * @server: nfs_server with cached state owners to release
598 * Called at umount time. Remaining state owners will be on
599 * the LRU with ref count of zero.
601 void nfs4_purge_state_owners(struct nfs_server *server)
603 struct nfs_client *clp = server->nfs_client;
604 struct nfs4_state_owner *sp, *tmp;
607 spin_lock(&clp->cl_lock);
608 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
609 list_move(&sp->so_lru, &doomed);
610 nfs4_remove_state_owner_locked(sp);
612 spin_unlock(&clp->cl_lock);
614 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
615 list_del(&sp->so_lru);
616 nfs4_free_state_owner(sp);
620 static struct nfs4_state *
621 nfs4_alloc_open_state(void)
623 struct nfs4_state *state;
625 state = kzalloc(sizeof(*state), GFP_NOFS);
628 atomic_set(&state->count, 1);
629 INIT_LIST_HEAD(&state->lock_states);
630 spin_lock_init(&state->state_lock);
631 seqlock_init(&state->seqlock);
636 nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
638 if (state->state == fmode)
640 /* NB! List reordering - see the reclaim code for why. */
641 if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
642 if (fmode & FMODE_WRITE)
643 list_move(&state->open_states, &state->owner->so_states);
645 list_move_tail(&state->open_states, &state->owner->so_states);
647 state->state = fmode;
650 static struct nfs4_state *
651 __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
653 struct nfs_inode *nfsi = NFS_I(inode);
654 struct nfs4_state *state;
656 list_for_each_entry(state, &nfsi->open_states, inode_states) {
657 if (state->owner != owner)
659 if (!nfs4_valid_open_stateid(state))
661 if (atomic_inc_not_zero(&state->count))
668 nfs4_free_open_state(struct nfs4_state *state)
674 nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
676 struct nfs4_state *state, *new;
677 struct nfs_inode *nfsi = NFS_I(inode);
679 spin_lock(&inode->i_lock);
680 state = __nfs4_find_state_byowner(inode, owner);
681 spin_unlock(&inode->i_lock);
684 new = nfs4_alloc_open_state();
685 spin_lock(&owner->so_lock);
686 spin_lock(&inode->i_lock);
687 state = __nfs4_find_state_byowner(inode, owner);
688 if (state == NULL && new != NULL) {
690 state->owner = owner;
691 atomic_inc(&owner->so_count);
692 list_add(&state->inode_states, &nfsi->open_states);
694 state->inode = inode;
695 spin_unlock(&inode->i_lock);
696 /* Note: The reclaim code dictates that we add stateless
697 * and read-only stateids to the end of the list */
698 list_add_tail(&state->open_states, &owner->so_states);
699 spin_unlock(&owner->so_lock);
701 spin_unlock(&inode->i_lock);
702 spin_unlock(&owner->so_lock);
704 nfs4_free_open_state(new);
710 void nfs4_put_open_state(struct nfs4_state *state)
712 struct inode *inode = state->inode;
713 struct nfs4_state_owner *owner = state->owner;
715 if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
717 spin_lock(&inode->i_lock);
718 list_del(&state->inode_states);
719 list_del(&state->open_states);
720 spin_unlock(&inode->i_lock);
721 spin_unlock(&owner->so_lock);
723 nfs4_free_open_state(state);
724 nfs4_put_state_owner(owner);
728 * Close the current file.
730 static void __nfs4_close(struct nfs4_state *state,
731 fmode_t fmode, gfp_t gfp_mask, int wait)
733 struct nfs4_state_owner *owner = state->owner;
737 atomic_inc(&owner->so_count);
738 /* Protect against nfs4_find_state() */
739 spin_lock(&owner->so_lock);
740 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
747 case FMODE_READ|FMODE_WRITE:
750 newstate = FMODE_READ|FMODE_WRITE;
751 if (state->n_rdwr == 0) {
752 if (state->n_rdonly == 0) {
753 newstate &= ~FMODE_READ;
754 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
755 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
757 if (state->n_wronly == 0) {
758 newstate &= ~FMODE_WRITE;
759 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
760 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
763 clear_bit(NFS_DELEGATED_STATE, &state->flags);
765 nfs4_state_set_mode_locked(state, newstate);
766 spin_unlock(&owner->so_lock);
769 nfs4_put_open_state(state);
770 nfs4_put_state_owner(owner);
772 nfs4_do_close(state, gfp_mask, wait);
775 void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
777 __nfs4_close(state, fmode, GFP_NOFS, 0);
780 void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
782 __nfs4_close(state, fmode, GFP_KERNEL, 1);
786 * Search the state->lock_states for an existing lock_owner
787 * that is compatible with current->files
789 static struct nfs4_lock_state *
790 __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type)
792 struct nfs4_lock_state *pos;
793 list_for_each_entry(pos, &state->lock_states, ls_locks) {
794 if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
796 switch (pos->ls_owner.lo_type) {
797 case NFS4_POSIX_LOCK_TYPE:
798 if (pos->ls_owner.lo_u.posix_owner != fl_owner)
801 case NFS4_FLOCK_LOCK_TYPE:
802 if (pos->ls_owner.lo_u.flock_owner != fl_pid)
805 atomic_inc(&pos->ls_count);
812 * Return a compatible lock_state. If no initialized lock_state structure
813 * exists, return an uninitialized one.
816 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)
818 struct nfs4_lock_state *lsp;
819 struct nfs_server *server = state->owner->so_server;
821 lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
824 nfs4_init_seqid_counter(&lsp->ls_seqid);
825 atomic_set(&lsp->ls_count, 1);
826 lsp->ls_state = state;
827 lsp->ls_owner.lo_type = type;
828 switch (lsp->ls_owner.lo_type) {
829 case NFS4_FLOCK_LOCK_TYPE:
830 lsp->ls_owner.lo_u.flock_owner = fl_pid;
832 case NFS4_POSIX_LOCK_TYPE:
833 lsp->ls_owner.lo_u.posix_owner = fl_owner;
838 lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
839 if (lsp->ls_seqid.owner_id < 0)
841 INIT_LIST_HEAD(&lsp->ls_locks);
848 void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
850 ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
851 nfs4_destroy_seqid_counter(&lsp->ls_seqid);
856 * Return a compatible lock_state. If no initialized lock_state structure
857 * exists, return an uninitialized one.
860 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner, pid_t pid, unsigned int type)
862 struct nfs4_lock_state *lsp, *new = NULL;
865 spin_lock(&state->state_lock);
866 lsp = __nfs4_find_lock_state(state, owner, pid, type);
870 list_add(&new->ls_locks, &state->lock_states);
871 set_bit(LK_STATE_IN_USE, &state->flags);
876 spin_unlock(&state->state_lock);
877 new = nfs4_alloc_lock_state(state, owner, pid, type);
881 spin_unlock(&state->state_lock);
883 nfs4_free_lock_state(state->owner->so_server, new);
888 * Release reference to lock_state, and free it if we see that
889 * it is no longer in use
891 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
893 struct nfs_server *server;
894 struct nfs4_state *state;
898 state = lsp->ls_state;
899 if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
901 list_del(&lsp->ls_locks);
902 if (list_empty(&state->lock_states))
903 clear_bit(LK_STATE_IN_USE, &state->flags);
904 spin_unlock(&state->state_lock);
905 server = state->owner->so_server;
906 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
907 struct nfs_client *clp = server->nfs_client;
909 clp->cl_mvops->free_lock_state(server, lsp);
911 nfs4_free_lock_state(server, lsp);
914 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
916 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
918 dst->fl_u.nfs4_fl.owner = lsp;
919 atomic_inc(&lsp->ls_count);
922 static void nfs4_fl_release_lock(struct file_lock *fl)
924 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
927 static const struct file_lock_operations nfs4_fl_lock_ops = {
928 .fl_copy_lock = nfs4_fl_copy_lock,
929 .fl_release_private = nfs4_fl_release_lock,
932 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
934 struct nfs4_lock_state *lsp;
936 if (fl->fl_ops != NULL)
938 if (fl->fl_flags & FL_POSIX)
939 lsp = nfs4_get_lock_state(state, fl->fl_owner, 0, NFS4_POSIX_LOCK_TYPE);
940 else if (fl->fl_flags & FL_FLOCK)
941 lsp = nfs4_get_lock_state(state, NULL, fl->fl_pid,
942 NFS4_FLOCK_LOCK_TYPE);
947 fl->fl_u.nfs4_fl.owner = lsp;
948 fl->fl_ops = &nfs4_fl_lock_ops;
952 static int nfs4_copy_lock_stateid(nfs4_stateid *dst,
953 struct nfs4_state *state,
954 const struct nfs_lockowner *lockowner)
956 struct nfs4_lock_state *lsp;
962 if (lockowner == NULL)
965 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
968 fl_owner = lockowner->l_owner;
969 fl_pid = lockowner->l_pid;
970 spin_lock(&state->state_lock);
971 lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
972 if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
973 nfs4_stateid_copy(dst, &lsp->ls_stateid);
976 if (!list_empty(&lsp->ls_seqid.list))
979 spin_unlock(&state->state_lock);
980 nfs4_put_lock_state(lsp);
985 static int nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
987 const nfs4_stateid *src;
993 seq = read_seqbegin(&state->seqlock);
994 if (test_bit(NFS_OPEN_STATE, &state->flags))
995 src = &state->open_stateid;
996 nfs4_stateid_copy(dst, src);
999 if (!list_empty(&state->owner->so_seqid.list))
1001 } while (read_seqretry(&state->seqlock, seq));
1006 * Byte-range lock aware utility to initialize the stateid of read/write
1009 int nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state,
1010 fmode_t fmode, const struct nfs_lockowner *lockowner)
1013 if (nfs4_copy_delegation_stateid(dst, state->inode, fmode))
1015 ret = nfs4_copy_lock_stateid(dst, state, lockowner);
1018 ret = nfs4_copy_open_stateid(dst, state);
1020 if (nfs_server_capable(state->inode, NFS_CAP_STATEID_NFSV41))
1025 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
1027 struct nfs_seqid *new;
1029 new = kmalloc(sizeof(*new), gfp_mask);
1031 new->sequence = counter;
1032 INIT_LIST_HEAD(&new->list);
1038 void nfs_release_seqid(struct nfs_seqid *seqid)
1040 struct nfs_seqid_counter *sequence;
1042 if (list_empty(&seqid->list))
1044 sequence = seqid->sequence;
1045 spin_lock(&sequence->lock);
1046 list_del_init(&seqid->list);
1047 if (!list_empty(&sequence->list)) {
1048 struct nfs_seqid *next;
1050 next = list_first_entry(&sequence->list,
1051 struct nfs_seqid, list);
1052 rpc_wake_up_queued_task(&sequence->wait, next->task);
1054 spin_unlock(&sequence->lock);
1057 void nfs_free_seqid(struct nfs_seqid *seqid)
1059 nfs_release_seqid(seqid);
1064 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
1065 * failed with a seqid incrementing error -
1066 * see comments nfs_fs.h:seqid_mutating_error()
1068 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
1073 case -NFS4ERR_BAD_SEQID:
1074 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
1076 pr_warn_ratelimited("NFS: v4 server returned a bad"
1077 " sequence-id error on an"
1078 " unconfirmed sequence %p!\n",
1080 case -NFS4ERR_STALE_CLIENTID:
1081 case -NFS4ERR_STALE_STATEID:
1082 case -NFS4ERR_BAD_STATEID:
1083 case -NFS4ERR_BADXDR:
1084 case -NFS4ERR_RESOURCE:
1085 case -NFS4ERR_NOFILEHANDLE:
1086 /* Non-seqid mutating errors */
1090 * Note: no locking needed as we are guaranteed to be first
1091 * on the sequence list
1093 seqid->sequence->counter++;
1096 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
1098 struct nfs4_state_owner *sp = container_of(seqid->sequence,
1099 struct nfs4_state_owner, so_seqid);
1100 struct nfs_server *server = sp->so_server;
1102 if (status == -NFS4ERR_BAD_SEQID)
1103 nfs4_drop_state_owner(sp);
1104 if (!nfs4_has_session(server->nfs_client))
1105 nfs_increment_seqid(status, seqid);
1109 * Increment the seqid if the LOCK/LOCKU succeeded, or
1110 * failed with a seqid incrementing error -
1111 * see comments nfs_fs.h:seqid_mutating_error()
1113 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
1115 nfs_increment_seqid(status, seqid);
1118 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
1120 struct nfs_seqid_counter *sequence = seqid->sequence;
1123 spin_lock(&sequence->lock);
1125 if (list_empty(&seqid->list))
1126 list_add_tail(&seqid->list, &sequence->list);
1127 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
1129 rpc_sleep_on(&sequence->wait, task, NULL);
1132 spin_unlock(&sequence->lock);
1136 static int nfs4_run_state_manager(void *);
1138 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
1140 smp_mb__before_clear_bit();
1141 clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
1142 smp_mb__after_clear_bit();
1143 wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
1144 rpc_wake_up(&clp->cl_rpcwaitq);
1148 * Schedule the nfs_client asynchronous state management routine
1150 void nfs4_schedule_state_manager(struct nfs_client *clp)
1152 struct task_struct *task;
1153 char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
1155 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1157 __module_get(THIS_MODULE);
1158 atomic_inc(&clp->cl_count);
1160 /* The rcu_read_lock() is not strictly necessary, as the state
1161 * manager is the only thread that ever changes the rpc_xprt
1162 * after it's initialized. At this point, we're single threaded. */
1164 snprintf(buf, sizeof(buf), "%s-manager",
1165 rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR));
1167 task = kthread_run(nfs4_run_state_manager, clp, "%s", buf);
1169 printk(KERN_ERR "%s: kthread_run: %ld\n",
1170 __func__, PTR_ERR(task));
1171 nfs4_clear_state_manager_bit(clp);
1172 nfs_put_client(clp);
1173 module_put(THIS_MODULE);
1178 * Schedule a lease recovery attempt
1180 void nfs4_schedule_lease_recovery(struct nfs_client *clp)
1184 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1185 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1186 dprintk("%s: scheduling lease recovery for server %s\n", __func__,
1188 nfs4_schedule_state_manager(clp);
1190 EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
1192 int nfs4_wait_clnt_recover(struct nfs_client *clp)
1198 res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
1199 nfs_wait_bit_killable, TASK_KILLABLE);
1203 if (clp->cl_cons_state < 0)
1204 return clp->cl_cons_state;
1208 int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1213 for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1214 ret = nfs4_wait_clnt_recover(clp);
1217 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1218 !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1220 nfs4_schedule_state_manager(clp);
1227 * nfs40_handle_cb_pathdown - return all delegations after NFS4ERR_CB_PATH_DOWN
1228 * @clp: client to process
1230 * Set the NFS4CLNT_LEASE_EXPIRED state in order to force a
1231 * resend of the SETCLIENTID and hence re-establish the
1232 * callback channel. Then return all existing delegations.
1234 static void nfs40_handle_cb_pathdown(struct nfs_client *clp)
1236 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1237 nfs_expire_all_delegations(clp);
1238 dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
1242 void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
1244 nfs40_handle_cb_pathdown(clp);
1245 nfs4_schedule_state_manager(clp);
1248 static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1251 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1252 /* Don't recover state that expired before the reboot */
1253 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
1254 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1257 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1258 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1262 static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1264 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
1265 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1266 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1267 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
1271 int nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
1273 struct nfs_client *clp = server->nfs_client;
1275 if (!nfs4_valid_open_stateid(state))
1277 nfs4_state_mark_reclaim_nograce(clp, state);
1278 dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
1280 nfs4_schedule_state_manager(clp);
1283 EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1285 void nfs_inode_find_state_and_recover(struct inode *inode,
1286 const nfs4_stateid *stateid)
1288 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1289 struct nfs_inode *nfsi = NFS_I(inode);
1290 struct nfs_open_context *ctx;
1291 struct nfs4_state *state;
1294 spin_lock(&inode->i_lock);
1295 list_for_each_entry(ctx, &nfsi->open_files, list) {
1299 if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
1301 if (!nfs4_stateid_match(&state->stateid, stateid))
1303 nfs4_state_mark_reclaim_nograce(clp, state);
1306 spin_unlock(&inode->i_lock);
1308 nfs4_schedule_state_manager(clp);
1311 static void nfs4_state_mark_open_context_bad(struct nfs4_state *state)
1313 struct inode *inode = state->inode;
1314 struct nfs_inode *nfsi = NFS_I(inode);
1315 struct nfs_open_context *ctx;
1317 spin_lock(&inode->i_lock);
1318 list_for_each_entry(ctx, &nfsi->open_files, list) {
1319 if (ctx->state != state)
1321 set_bit(NFS_CONTEXT_BAD, &ctx->flags);
1323 spin_unlock(&inode->i_lock);
1326 static void nfs4_state_mark_recovery_failed(struct nfs4_state *state, int error)
1328 set_bit(NFS_STATE_RECOVERY_FAILED, &state->flags);
1329 nfs4_state_mark_open_context_bad(state);
1333 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
1335 struct inode *inode = state->inode;
1336 struct nfs_inode *nfsi = NFS_I(inode);
1337 struct file_lock *fl;
1340 if (inode->i_flock == NULL)
1343 /* Guard against delegation returns and new lock/unlock calls */
1344 down_write(&nfsi->rwsem);
1345 /* Protect inode->i_flock using the BKL */
1346 spin_lock(&inode->i_lock);
1347 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1348 if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
1350 if (nfs_file_open_context(fl->fl_file)->state != state)
1352 spin_unlock(&inode->i_lock);
1353 status = ops->recover_lock(state, fl);
1358 case -NFS4ERR_ADMIN_REVOKED:
1359 case -NFS4ERR_STALE_STATEID:
1360 case -NFS4ERR_BAD_STATEID:
1361 case -NFS4ERR_EXPIRED:
1362 case -NFS4ERR_NO_GRACE:
1363 case -NFS4ERR_STALE_CLIENTID:
1364 case -NFS4ERR_BADSESSION:
1365 case -NFS4ERR_BADSLOT:
1366 case -NFS4ERR_BAD_HIGH_SLOT:
1367 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1370 printk(KERN_ERR "NFS: %s: unhandled error %d. "
1371 "Zeroing state\n", __func__, status);
1373 case -NFS4ERR_DENIED:
1374 case -NFS4ERR_RECLAIM_BAD:
1375 case -NFS4ERR_RECLAIM_CONFLICT:
1376 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1379 spin_lock(&inode->i_lock);
1381 spin_unlock(&inode->i_lock);
1383 up_write(&nfsi->rwsem);
1387 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
1389 struct nfs4_state *state;
1390 struct nfs4_lock_state *lock;
1393 /* Note: we rely on the sp->so_states list being ordered
1394 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
1396 * This is needed to ensure that the server won't give us any
1397 * read delegations that we have to return if, say, we are
1398 * recovering after a network partition or a reboot from a
1399 * server that doesn't support a grace period.
1401 spin_lock(&sp->so_lock);
1402 write_seqcount_begin(&sp->so_reclaim_seqcount);
1404 list_for_each_entry(state, &sp->so_states, open_states) {
1405 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
1407 if (!nfs4_valid_open_stateid(state))
1409 if (state->state == 0)
1411 atomic_inc(&state->count);
1412 spin_unlock(&sp->so_lock);
1413 status = ops->recover_open(sp, state);
1415 status = nfs4_reclaim_locks(state, ops);
1417 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0) {
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);
1427 nfs4_put_open_state(state);
1428 spin_lock(&sp->so_lock);
1434 printk(KERN_ERR "NFS: %s: unhandled error %d. "
1435 "Zeroing state\n", __func__, status);
1440 * Open state on this file cannot be recovered
1441 * All we can do is revert to using the zero stateid.
1443 nfs4_state_mark_recovery_failed(state, status);
1447 case -NFS4ERR_ADMIN_REVOKED:
1448 case -NFS4ERR_STALE_STATEID:
1449 case -NFS4ERR_BAD_STATEID:
1450 case -NFS4ERR_RECLAIM_BAD:
1451 case -NFS4ERR_RECLAIM_CONFLICT:
1452 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1454 case -NFS4ERR_EXPIRED:
1455 case -NFS4ERR_NO_GRACE:
1456 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1457 case -NFS4ERR_STALE_CLIENTID:
1458 case -NFS4ERR_BADSESSION:
1459 case -NFS4ERR_BADSLOT:
1460 case -NFS4ERR_BAD_HIGH_SLOT:
1461 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1464 nfs4_put_open_state(state);
1465 spin_lock(&sp->so_lock);
1468 write_seqcount_end(&sp->so_reclaim_seqcount);
1469 spin_unlock(&sp->so_lock);
1472 nfs4_put_open_state(state);
1473 spin_lock(&sp->so_lock);
1474 write_seqcount_end(&sp->so_reclaim_seqcount);
1475 spin_unlock(&sp->so_lock);
1479 static void nfs4_clear_open_state(struct nfs4_state *state)
1481 struct nfs4_lock_state *lock;
1483 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1484 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1485 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1486 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1487 spin_lock(&state->state_lock);
1488 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1489 lock->ls_seqid.flags = 0;
1490 clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
1492 spin_unlock(&state->state_lock);
1495 static void nfs4_reset_seqids(struct nfs_server *server,
1496 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1498 struct nfs_client *clp = server->nfs_client;
1499 struct nfs4_state_owner *sp;
1500 struct rb_node *pos;
1501 struct nfs4_state *state;
1503 spin_lock(&clp->cl_lock);
1504 for (pos = rb_first(&server->state_owners);
1506 pos = rb_next(pos)) {
1507 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1508 sp->so_seqid.flags = 0;
1509 spin_lock(&sp->so_lock);
1510 list_for_each_entry(state, &sp->so_states, open_states) {
1511 if (mark_reclaim(clp, state))
1512 nfs4_clear_open_state(state);
1514 spin_unlock(&sp->so_lock);
1516 spin_unlock(&clp->cl_lock);
1519 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp,
1520 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1522 struct nfs_server *server;
1525 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1526 nfs4_reset_seqids(server, mark_reclaim);
1530 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1532 /* Mark all delegations for reclaim */
1533 nfs_delegation_mark_reclaim(clp);
1534 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1537 static void nfs4_reclaim_complete(struct nfs_client *clp,
1538 const struct nfs4_state_recovery_ops *ops,
1539 struct rpc_cred *cred)
1541 /* Notify the server we're done reclaiming our state */
1542 if (ops->reclaim_complete)
1543 (void)ops->reclaim_complete(clp, cred);
1546 static void nfs4_clear_reclaim_server(struct nfs_server *server)
1548 struct nfs_client *clp = server->nfs_client;
1549 struct nfs4_state_owner *sp;
1550 struct rb_node *pos;
1551 struct nfs4_state *state;
1553 spin_lock(&clp->cl_lock);
1554 for (pos = rb_first(&server->state_owners);
1556 pos = rb_next(pos)) {
1557 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1558 spin_lock(&sp->so_lock);
1559 list_for_each_entry(state, &sp->so_states, open_states) {
1560 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
1563 nfs4_state_mark_reclaim_nograce(clp, state);
1565 spin_unlock(&sp->so_lock);
1567 spin_unlock(&clp->cl_lock);
1570 static int nfs4_state_clear_reclaim_reboot(struct nfs_client *clp)
1572 struct nfs_server *server;
1574 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1578 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1579 nfs4_clear_reclaim_server(server);
1582 nfs_delegation_reap_unclaimed(clp);
1586 static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
1588 const struct nfs4_state_recovery_ops *ops;
1589 struct rpc_cred *cred;
1591 if (!nfs4_state_clear_reclaim_reboot(clp))
1593 ops = clp->cl_mvops->reboot_recovery_ops;
1594 cred = nfs4_get_clid_cred(clp);
1595 nfs4_reclaim_complete(clp, ops, cred);
1599 static void nfs_delegation_clear_all(struct nfs_client *clp)
1601 nfs_delegation_mark_reclaim(clp);
1602 nfs_delegation_reap_unclaimed(clp);
1605 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1607 nfs_delegation_clear_all(clp);
1608 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1611 static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1616 case -NFS4ERR_CB_PATH_DOWN:
1617 nfs40_handle_cb_pathdown(clp);
1619 case -NFS4ERR_NO_GRACE:
1620 nfs4_state_end_reclaim_reboot(clp);
1622 case -NFS4ERR_STALE_CLIENTID:
1623 case -NFS4ERR_LEASE_MOVED:
1624 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1625 nfs4_state_clear_reclaim_reboot(clp);
1626 nfs4_state_start_reclaim_reboot(clp);
1628 case -NFS4ERR_EXPIRED:
1629 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1630 nfs4_state_start_reclaim_nograce(clp);
1632 case -NFS4ERR_BADSESSION:
1633 case -NFS4ERR_BADSLOT:
1634 case -NFS4ERR_BAD_HIGH_SLOT:
1635 case -NFS4ERR_DEADSESSION:
1636 case -NFS4ERR_SEQ_FALSE_RETRY:
1637 case -NFS4ERR_SEQ_MISORDERED:
1638 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1639 /* Zero session reset errors */
1641 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1642 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1645 dprintk("%s: failed to handle error %d for server %s\n",
1646 __func__, error, clp->cl_hostname);
1649 dprintk("%s: handled error %d for server %s\n", __func__, error,
1654 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1656 struct nfs4_state_owner *sp;
1657 struct nfs_server *server;
1658 struct rb_node *pos;
1663 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1664 nfs4_purge_state_owners(server);
1665 spin_lock(&clp->cl_lock);
1666 for (pos = rb_first(&server->state_owners);
1668 pos = rb_next(pos)) {
1670 struct nfs4_state_owner, so_server_node);
1671 if (!test_and_clear_bit(ops->owner_flag_bit,
1674 atomic_inc(&sp->so_count);
1675 spin_unlock(&clp->cl_lock);
1678 status = nfs4_reclaim_open_state(sp, ops);
1680 set_bit(ops->owner_flag_bit, &sp->so_flags);
1681 nfs4_put_state_owner(sp);
1682 return nfs4_recovery_handle_error(clp, status);
1685 nfs4_put_state_owner(sp);
1688 spin_unlock(&clp->cl_lock);
1694 static int nfs4_check_lease(struct nfs_client *clp)
1696 struct rpc_cred *cred;
1697 const struct nfs4_state_maintenance_ops *ops =
1698 clp->cl_mvops->state_renewal_ops;
1701 /* Is the client already known to have an expired lease? */
1702 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1704 spin_lock(&clp->cl_lock);
1705 cred = ops->get_state_renewal_cred_locked(clp);
1706 spin_unlock(&clp->cl_lock);
1708 cred = nfs4_get_clid_cred(clp);
1713 status = ops->renew_lease(clp, cred);
1715 if (status == -ETIMEDOUT) {
1716 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1720 return nfs4_recovery_handle_error(clp, status);
1723 /* Set NFS4CLNT_LEASE_EXPIRED and reclaim reboot state for all v4.0 errors
1724 * and for recoverable errors on EXCHANGE_ID for v4.1
1726 static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
1729 case -NFS4ERR_SEQ_MISORDERED:
1730 if (test_and_set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state))
1731 return -ESERVERFAULT;
1732 /* Lease confirmation error: retry after purging the lease */
1734 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1736 case -NFS4ERR_STALE_CLIENTID:
1737 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1738 nfs4_state_clear_reclaim_reboot(clp);
1739 nfs4_state_start_reclaim_reboot(clp);
1741 case -NFS4ERR_CLID_INUSE:
1742 pr_err("NFS: Server %s reports our clientid is in use\n",
1744 nfs_mark_client_ready(clp, -EPERM);
1745 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1748 case -NFS4ERR_DELAY:
1754 case -NFS4ERR_MINOR_VERS_MISMATCH:
1755 if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
1756 nfs_mark_client_ready(clp, -EPROTONOSUPPORT);
1757 dprintk("%s: exit with error %d for server %s\n",
1758 __func__, -EPROTONOSUPPORT, clp->cl_hostname);
1759 return -EPROTONOSUPPORT;
1760 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1761 * in nfs4_exchange_id */
1763 dprintk("%s: exit with error %d for server %s\n", __func__,
1764 status, clp->cl_hostname);
1767 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1768 dprintk("%s: handled error %d for server %s\n", __func__, status,
1773 static int nfs4_establish_lease(struct nfs_client *clp)
1775 struct rpc_cred *cred;
1776 const struct nfs4_state_recovery_ops *ops =
1777 clp->cl_mvops->reboot_recovery_ops;
1780 cred = nfs4_get_clid_cred(clp);
1783 status = ops->establish_clid(clp, cred);
1787 pnfs_destroy_all_layouts(clp);
1792 * Returns zero or a negative errno. NFS4ERR values are converted
1793 * to local errno values.
1795 static int nfs4_reclaim_lease(struct nfs_client *clp)
1799 status = nfs4_establish_lease(clp);
1801 return nfs4_handle_reclaim_lease_error(clp, status);
1802 if (test_and_clear_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state))
1803 nfs4_state_start_reclaim_nograce(clp);
1804 if (!test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
1805 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1806 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1807 clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1811 static int nfs4_purge_lease(struct nfs_client *clp)
1815 status = nfs4_establish_lease(clp);
1817 return nfs4_handle_reclaim_lease_error(clp, status);
1818 clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1819 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1820 nfs4_state_start_reclaim_nograce(clp);
1825 * nfs4_discover_server_trunking - Detect server IP address trunking
1827 * @clp: nfs_client under test
1828 * @result: OUT: found nfs_client, or clp
1830 * Returns zero or a negative errno. If zero is returned,
1831 * an nfs_client pointer is planted in "result".
1833 * Note: since we are invoked in process context, and
1834 * not from inside the state manager, we cannot use
1835 * nfs4_handle_reclaim_lease_error().
1837 int nfs4_discover_server_trunking(struct nfs_client *clp,
1838 struct nfs_client **result)
1840 const struct nfs4_state_recovery_ops *ops =
1841 clp->cl_mvops->reboot_recovery_ops;
1842 struct rpc_clnt *clnt;
1843 struct rpc_cred *cred;
1846 dprintk("NFS: %s: testing '%s'\n", __func__, clp->cl_hostname);
1848 clnt = clp->cl_rpcclient;
1851 mutex_lock(&nfs_clid_init_mutex);
1854 cred = nfs4_get_clid_cred(clp);
1858 status = ops->detect_trunking(clp, result, cred);
1863 case -NFS4ERR_DELAY:
1867 case -NFS4ERR_STALE_CLIENTID:
1868 dprintk("NFS: %s after status %d, retrying\n",
1873 nfs4_root_machine_cred(clp);
1878 case -NFS4ERR_CLID_INUSE:
1879 case -NFS4ERR_WRONGSEC:
1880 clnt = rpc_clone_client_set_auth(clnt, RPC_AUTH_UNIX);
1882 status = PTR_ERR(clnt);
1885 /* Note: this is safe because we haven't yet marked the
1886 * client as ready, so we are the only user of
1889 clnt = xchg(&clp->cl_rpcclient, clnt);
1890 rpc_shutdown_client(clnt);
1891 clnt = clp->cl_rpcclient;
1894 case -NFS4ERR_MINOR_VERS_MISMATCH:
1895 status = -EPROTONOSUPPORT;
1899 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1900 * in nfs4_exchange_id */
1901 status = -EKEYEXPIRED;
1904 pr_warn("NFS: %s unhandled error %d. Exiting with error EIO\n",
1910 mutex_unlock(&nfs_clid_init_mutex);
1911 dprintk("NFS: %s: status = %d\n", __func__, status);
1915 #ifdef CONFIG_NFS_V4_1
1916 void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
1918 struct nfs_client *clp = session->clp;
1922 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1924 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1925 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1927 nfs4_schedule_lease_recovery(clp);
1929 EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
1931 static void nfs41_ping_server(struct nfs_client *clp)
1933 /* Use CHECK_LEASE to ping the server with a SEQUENCE */
1934 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1935 nfs4_schedule_state_manager(clp);
1938 void nfs41_server_notify_target_slotid_update(struct nfs_client *clp)
1940 nfs41_ping_server(clp);
1943 void nfs41_server_notify_highest_slotid_update(struct nfs_client *clp)
1945 nfs41_ping_server(clp);
1948 static void nfs4_reset_all_state(struct nfs_client *clp)
1950 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1951 set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1952 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1953 nfs4_state_start_reclaim_nograce(clp);
1954 dprintk("%s: scheduling reset of all state for server %s!\n",
1955 __func__, clp->cl_hostname);
1956 nfs4_schedule_state_manager(clp);
1960 static void nfs41_handle_server_reboot(struct nfs_client *clp)
1962 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1963 nfs4_state_start_reclaim_reboot(clp);
1964 dprintk("%s: server %s rebooted!\n", __func__,
1966 nfs4_schedule_state_manager(clp);
1970 static void nfs41_handle_state_revoked(struct nfs_client *clp)
1972 nfs4_reset_all_state(clp);
1973 dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
1976 static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
1978 /* This will need to handle layouts too */
1979 nfs_expire_all_delegations(clp);
1980 dprintk("%s: Recallable state revoked on server %s!\n", __func__,
1984 static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
1986 nfs_expire_all_delegations(clp);
1987 if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
1988 nfs4_schedule_state_manager(clp);
1989 dprintk("%s: server %s declared a backchannel fault\n", __func__,
1993 static void nfs41_handle_cb_path_down(struct nfs_client *clp)
1995 if (test_and_set_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
1996 &clp->cl_state) == 0)
1997 nfs4_schedule_state_manager(clp);
2000 void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
2005 dprintk("%s: \"%s\" (client ID %llx) flags=0x%08x\n",
2006 __func__, clp->cl_hostname, clp->cl_clientid, flags);
2008 if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
2009 nfs41_handle_server_reboot(clp);
2010 if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
2011 SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
2012 SEQ4_STATUS_ADMIN_STATE_REVOKED |
2013 SEQ4_STATUS_LEASE_MOVED))
2014 nfs41_handle_state_revoked(clp);
2015 if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
2016 nfs41_handle_recallable_state_revoked(clp);
2017 if (flags & SEQ4_STATUS_BACKCHANNEL_FAULT)
2018 nfs41_handle_backchannel_fault(clp);
2019 else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
2020 SEQ4_STATUS_CB_PATH_DOWN_SESSION))
2021 nfs41_handle_cb_path_down(clp);
2024 static int nfs4_reset_session(struct nfs_client *clp)
2026 struct rpc_cred *cred;
2029 if (!nfs4_has_session(clp))
2031 nfs4_begin_drain_session(clp);
2032 cred = nfs4_get_clid_cred(clp);
2033 status = nfs4_proc_destroy_session(clp->cl_session, cred);
2036 case -NFS4ERR_BADSESSION:
2037 case -NFS4ERR_DEADSESSION:
2039 case -NFS4ERR_BACK_CHAN_BUSY:
2040 case -NFS4ERR_DELAY:
2041 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
2046 status = nfs4_recovery_handle_error(clp, status);
2050 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
2051 status = nfs4_proc_create_session(clp, cred);
2053 dprintk("%s: session reset failed with status %d for server %s!\n",
2054 __func__, status, clp->cl_hostname);
2055 status = nfs4_handle_reclaim_lease_error(clp, status);
2058 nfs41_finish_session_reset(clp);
2059 dprintk("%s: session reset was successful for server %s!\n",
2060 __func__, clp->cl_hostname);
2067 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2069 struct rpc_cred *cred;
2072 if (!nfs4_has_session(clp))
2074 nfs4_begin_drain_session(clp);
2075 cred = nfs4_get_clid_cred(clp);
2076 ret = nfs4_proc_bind_conn_to_session(clp, cred);
2079 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2082 dprintk("%s: bind_conn_to_session was successful for server %s!\n",
2083 __func__, clp->cl_hostname);
2085 case -NFS4ERR_DELAY:
2087 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2090 return nfs4_recovery_handle_error(clp, ret);
2094 #else /* CONFIG_NFS_V4_1 */
2095 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
2097 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2101 #endif /* CONFIG_NFS_V4_1 */
2103 static void nfs4_state_manager(struct nfs_client *clp)
2106 const char *section = "", *section_sep = "";
2108 /* Ensure exclusive access to NFSv4 state */
2110 if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
2111 section = "purge state";
2112 status = nfs4_purge_lease(clp);
2118 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
2119 section = "lease expired";
2120 /* We're going to have to re-establish a clientid */
2121 status = nfs4_reclaim_lease(clp);
2127 /* Initialize or reset the session */
2128 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
2129 section = "reset session";
2130 status = nfs4_reset_session(clp);
2131 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
2137 /* Send BIND_CONN_TO_SESSION */
2138 if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2140 section = "bind conn to session";
2141 status = nfs4_bind_conn_to_session(clp);
2147 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
2148 section = "check lease";
2149 status = nfs4_check_lease(clp);
2155 /* First recover reboot state... */
2156 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
2157 section = "reclaim reboot";
2158 status = nfs4_do_reclaim(clp,
2159 clp->cl_mvops->reboot_recovery_ops);
2160 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2161 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
2163 nfs4_state_end_reclaim_reboot(clp);
2164 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
2170 /* Now recover expired state... */
2171 if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
2172 section = "reclaim nograce";
2173 status = nfs4_do_reclaim(clp,
2174 clp->cl_mvops->nograce_recovery_ops);
2175 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2176 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
2177 test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
2183 nfs4_end_drain_session(clp);
2184 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
2185 nfs_client_return_marked_delegations(clp);
2189 nfs4_clear_state_manager_bit(clp);
2190 /* Did we race with an attempt to give us more work? */
2191 if (clp->cl_state == 0)
2193 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
2195 } while (atomic_read(&clp->cl_count) > 1);
2198 if (strlen(section))
2200 pr_warn_ratelimited("NFS: state manager%s%s failed on NFSv4 server %s"
2201 " with error %d\n", section_sep, section,
2202 clp->cl_hostname, -status);
2204 nfs4_end_drain_session(clp);
2205 nfs4_clear_state_manager_bit(clp);
2208 static int nfs4_run_state_manager(void *ptr)
2210 struct nfs_client *clp = ptr;
2212 allow_signal(SIGKILL);
2213 nfs4_state_manager(clp);
2214 nfs_put_client(clp);
2215 module_put_and_exit(0);