4 * Client-side XDR for NFSv4.
6 * Copyright (c) 2002 The Regents of the University of Michigan.
9 * Kendrick Smith <kmsmith@umich.edu>
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its
21 * contributors may be used to endorse or promote products derived
22 * from this software without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
25 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
26 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
27 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
31 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
32 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
33 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
34 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36 * Implementation of the NFSv4 state model. For the time being,
37 * this is minimal, but will be made much more complex in a
41 #include <linux/kernel.h>
42 #include <linux/slab.h>
44 #include <linux/nfs_fs.h>
45 #include <linux/nfs_idmap.h>
46 #include <linux/kthread.h>
47 #include <linux/module.h>
48 #include <linux/random.h>
49 #include <linux/ratelimit.h>
50 #include <linux/workqueue.h>
51 #include <linux/bitops.h>
52 #include <linux/jiffies.h>
54 #include <linux/sunrpc/clnt.h>
58 #include "delegation.h"
63 #define NFSDBG_FACILITY NFSDBG_STATE
65 #define OPENOWNER_POOL_SIZE 8
67 const nfs4_stateid zero_stateid;
68 static DEFINE_MUTEX(nfs_clid_init_mutex);
69 static LIST_HEAD(nfs4_clientid_list);
71 int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
73 struct nfs4_setclientid_res clid = {
74 .clientid = clp->cl_clientid,
75 .confirm = clp->cl_confirm,
79 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
81 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
83 port = nn->nfs_callback_tcpport;
84 if (clp->cl_addr.ss_family == AF_INET6)
85 port = nn->nfs_callback_tcpport6;
87 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
90 clp->cl_clientid = clid.clientid;
91 clp->cl_confirm = clid.confirm;
92 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
94 status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
97 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
98 nfs4_schedule_state_renewal(clp);
104 * nfs40_discover_server_trunking - Detect server IP address trunking (mv0)
106 * @clp: nfs_client under test
107 * @result: OUT: found nfs_client, or clp
108 * @cred: credential to use for trunking test
110 * Returns zero, a negative errno, or a negative NFS4ERR status.
111 * If zero is returned, an nfs_client pointer is planted in
114 * Note: The returned client may not yet be marked ready.
116 int nfs40_discover_server_trunking(struct nfs_client *clp,
117 struct nfs_client **result,
118 struct rpc_cred *cred)
120 struct nfs4_setclientid_res clid = {
121 .clientid = clp->cl_clientid,
122 .confirm = clp->cl_confirm,
124 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
128 port = nn->nfs_callback_tcpport;
129 if (clp->cl_addr.ss_family == AF_INET6)
130 port = nn->nfs_callback_tcpport6;
132 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
135 clp->cl_clientid = clid.clientid;
136 clp->cl_confirm = clid.confirm;
138 status = nfs40_walk_client_list(clp, result, cred);
140 case -NFS4ERR_STALE_CLIENTID:
141 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
143 /* Sustain the lease, even if it's empty. If the clientid4
144 * goes stale it's of no use for trunking discovery. */
145 nfs4_schedule_state_renewal(*result);
153 struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
155 struct rpc_cred *cred = NULL;
157 if (clp->cl_machine_cred != NULL)
158 cred = get_rpccred(clp->cl_machine_cred);
162 static void nfs4_clear_machine_cred(struct nfs_client *clp)
164 struct rpc_cred *cred;
166 spin_lock(&clp->cl_lock);
167 cred = clp->cl_machine_cred;
168 clp->cl_machine_cred = NULL;
169 spin_unlock(&clp->cl_lock);
174 static struct rpc_cred *
175 nfs4_get_renew_cred_server_locked(struct nfs_server *server)
177 struct rpc_cred *cred = NULL;
178 struct nfs4_state_owner *sp;
181 for (pos = rb_first(&server->state_owners);
183 pos = rb_next(pos)) {
184 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
185 if (list_empty(&sp->so_states))
187 cred = get_rpccred(sp->so_cred);
194 * nfs4_get_renew_cred_locked - Acquire credential for a renew operation
195 * @clp: client state handle
197 * Returns an rpc_cred with reference count bumped, or NULL.
198 * Caller must hold clp->cl_lock.
200 struct rpc_cred *nfs4_get_renew_cred_locked(struct nfs_client *clp)
202 struct rpc_cred *cred = NULL;
203 struct nfs_server *server;
205 /* Use machine credentials if available */
206 cred = nfs4_get_machine_cred_locked(clp);
211 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
212 cred = nfs4_get_renew_cred_server_locked(server);
222 #if defined(CONFIG_NFS_V4_1)
224 static int nfs41_setup_state_renewal(struct nfs_client *clp)
227 struct nfs_fsinfo fsinfo;
229 if (!test_bit(NFS_CS_CHECK_LEASE_TIME, &clp->cl_res_state)) {
230 nfs4_schedule_state_renewal(clp);
234 status = nfs4_proc_get_lease_time(clp, &fsinfo);
236 /* Update lease time and schedule renewal */
237 spin_lock(&clp->cl_lock);
238 clp->cl_lease_time = fsinfo.lease_time * HZ;
239 clp->cl_last_renewal = jiffies;
240 spin_unlock(&clp->cl_lock);
242 nfs4_schedule_state_renewal(clp);
249 * Back channel returns NFS4ERR_DELAY for new requests when
250 * NFS4_SESSION_DRAINING is set so there is no work to be done when draining
253 static void nfs4_end_drain_session(struct nfs_client *clp)
255 struct nfs4_session *ses = clp->cl_session;
256 struct nfs4_slot_table *tbl;
261 tbl = &ses->fc_slot_table;
262 if (test_and_clear_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
263 spin_lock(&tbl->slot_tbl_lock);
264 max_slots = tbl->max_slots;
265 while (max_slots--) {
266 if (rpc_wake_up_first(&tbl->slot_tbl_waitq,
267 nfs4_set_task_privileged,
271 spin_unlock(&tbl->slot_tbl_lock);
275 static int nfs4_wait_on_slot_tbl(struct nfs4_slot_table *tbl)
277 spin_lock(&tbl->slot_tbl_lock);
278 if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
279 INIT_COMPLETION(tbl->complete);
280 spin_unlock(&tbl->slot_tbl_lock);
281 return wait_for_completion_interruptible(&tbl->complete);
283 spin_unlock(&tbl->slot_tbl_lock);
287 static int nfs4_begin_drain_session(struct nfs_client *clp)
289 struct nfs4_session *ses = clp->cl_session;
292 set_bit(NFS4_SESSION_DRAINING, &ses->session_state);
294 ret = nfs4_wait_on_slot_tbl(&ses->bc_slot_table);
298 return nfs4_wait_on_slot_tbl(&ses->fc_slot_table);
301 static void nfs41_finish_session_reset(struct nfs_client *clp)
303 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
304 clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
305 /* create_session negotiated new slot table */
306 clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
307 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
308 nfs41_setup_state_renewal(clp);
311 int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
315 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
317 nfs4_begin_drain_session(clp);
318 status = nfs4_proc_exchange_id(clp, cred);
321 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
323 status = nfs4_proc_create_session(clp, cred);
326 nfs41_finish_session_reset(clp);
327 nfs_mark_client_ready(clp, NFS_CS_READY);
333 * nfs41_discover_server_trunking - Detect server IP address trunking (mv1)
335 * @clp: nfs_client under test
336 * @result: OUT: found nfs_client, or clp
337 * @cred: credential to use for trunking test
339 * Returns NFS4_OK, a negative errno, or a negative NFS4ERR status.
340 * If NFS4_OK is returned, an nfs_client pointer is planted in
343 * Note: The returned client may not yet be marked ready.
345 int nfs41_discover_server_trunking(struct nfs_client *clp,
346 struct nfs_client **result,
347 struct rpc_cred *cred)
351 status = nfs4_proc_exchange_id(clp, cred);
352 if (status != NFS4_OK)
354 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
356 return nfs41_walk_client_list(clp, result, cred);
359 struct rpc_cred *nfs4_get_exchange_id_cred(struct nfs_client *clp)
361 struct rpc_cred *cred;
363 spin_lock(&clp->cl_lock);
364 cred = nfs4_get_machine_cred_locked(clp);
365 spin_unlock(&clp->cl_lock);
369 #endif /* CONFIG_NFS_V4_1 */
371 static struct rpc_cred *
372 nfs4_get_setclientid_cred_server(struct nfs_server *server)
374 struct nfs_client *clp = server->nfs_client;
375 struct rpc_cred *cred = NULL;
376 struct nfs4_state_owner *sp;
379 spin_lock(&clp->cl_lock);
380 pos = rb_first(&server->state_owners);
382 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
383 cred = get_rpccred(sp->so_cred);
385 spin_unlock(&clp->cl_lock);
390 * nfs4_get_setclientid_cred - Acquire credential for a setclientid operation
391 * @clp: client state handle
393 * Returns an rpc_cred with reference count bumped, or NULL.
395 struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
397 struct nfs_server *server;
398 struct rpc_cred *cred;
400 spin_lock(&clp->cl_lock);
401 cred = nfs4_get_machine_cred_locked(clp);
402 spin_unlock(&clp->cl_lock);
407 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
408 cred = nfs4_get_setclientid_cred_server(server);
418 static struct nfs4_state_owner *
419 nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
421 struct rb_node **p = &server->state_owners.rb_node,
423 struct nfs4_state_owner *sp;
427 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
429 if (cred < sp->so_cred)
430 p = &parent->rb_left;
431 else if (cred > sp->so_cred)
432 p = &parent->rb_right;
434 if (!list_empty(&sp->so_lru))
435 list_del_init(&sp->so_lru);
436 atomic_inc(&sp->so_count);
443 static struct nfs4_state_owner *
444 nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
446 struct nfs_server *server = new->so_server;
447 struct rb_node **p = &server->state_owners.rb_node,
449 struct nfs4_state_owner *sp;
454 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
456 if (new->so_cred < sp->so_cred)
457 p = &parent->rb_left;
458 else if (new->so_cred > sp->so_cred)
459 p = &parent->rb_right;
461 if (!list_empty(&sp->so_lru))
462 list_del_init(&sp->so_lru);
463 atomic_inc(&sp->so_count);
467 err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
470 rb_link_node(&new->so_server_node, parent, p);
471 rb_insert_color(&new->so_server_node, &server->state_owners);
476 nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
478 struct nfs_server *server = sp->so_server;
480 if (!RB_EMPTY_NODE(&sp->so_server_node))
481 rb_erase(&sp->so_server_node, &server->state_owners);
482 ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
486 nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
488 sc->create_time = ktime_get();
491 spin_lock_init(&sc->lock);
492 INIT_LIST_HEAD(&sc->list);
493 rpc_init_wait_queue(&sc->wait, "Seqid_waitqueue");
497 nfs4_destroy_seqid_counter(struct nfs_seqid_counter *sc)
499 rpc_destroy_wait_queue(&sc->wait);
503 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
504 * create a new state_owner.
507 static struct nfs4_state_owner *
508 nfs4_alloc_state_owner(struct nfs_server *server,
509 struct rpc_cred *cred,
512 struct nfs4_state_owner *sp;
514 sp = kzalloc(sizeof(*sp), gfp_flags);
517 sp->so_server = server;
518 sp->so_cred = get_rpccred(cred);
519 spin_lock_init(&sp->so_lock);
520 INIT_LIST_HEAD(&sp->so_states);
521 nfs4_init_seqid_counter(&sp->so_seqid);
522 atomic_set(&sp->so_count, 1);
523 INIT_LIST_HEAD(&sp->so_lru);
528 nfs4_drop_state_owner(struct nfs4_state_owner *sp)
530 struct rb_node *rb_node = &sp->so_server_node;
532 if (!RB_EMPTY_NODE(rb_node)) {
533 struct nfs_server *server = sp->so_server;
534 struct nfs_client *clp = server->nfs_client;
536 spin_lock(&clp->cl_lock);
537 if (!RB_EMPTY_NODE(rb_node)) {
538 rb_erase(rb_node, &server->state_owners);
539 RB_CLEAR_NODE(rb_node);
541 spin_unlock(&clp->cl_lock);
545 static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
547 nfs4_destroy_seqid_counter(&sp->so_seqid);
548 put_rpccred(sp->so_cred);
552 static void nfs4_gc_state_owners(struct nfs_server *server)
554 struct nfs_client *clp = server->nfs_client;
555 struct nfs4_state_owner *sp, *tmp;
556 unsigned long time_min, time_max;
559 spin_lock(&clp->cl_lock);
561 time_min = (long)time_max - (long)clp->cl_lease_time;
562 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
563 /* NB: LRU is sorted so that oldest is at the head */
564 if (time_in_range(sp->so_expires, time_min, time_max))
566 list_move(&sp->so_lru, &doomed);
567 nfs4_remove_state_owner_locked(sp);
569 spin_unlock(&clp->cl_lock);
571 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
572 list_del(&sp->so_lru);
573 nfs4_free_state_owner(sp);
578 * nfs4_get_state_owner - Look up a state owner given a credential
579 * @server: nfs_server to search
580 * @cred: RPC credential to match
582 * Returns a pointer to an instantiated nfs4_state_owner struct, or NULL.
584 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server,
585 struct rpc_cred *cred,
588 struct nfs_client *clp = server->nfs_client;
589 struct nfs4_state_owner *sp, *new;
591 spin_lock(&clp->cl_lock);
592 sp = nfs4_find_state_owner_locked(server, cred);
593 spin_unlock(&clp->cl_lock);
596 new = nfs4_alloc_state_owner(server, cred, gfp_flags);
600 if (ida_pre_get(&server->openowner_id, gfp_flags) == 0)
602 spin_lock(&clp->cl_lock);
603 sp = nfs4_insert_state_owner_locked(new);
604 spin_unlock(&clp->cl_lock);
605 } while (sp == ERR_PTR(-EAGAIN));
607 nfs4_free_state_owner(new);
609 nfs4_gc_state_owners(server);
614 * nfs4_put_state_owner - Release a nfs4_state_owner
615 * @sp: state owner data to release
617 * Note that we keep released state owners on an LRU
619 * This caches valid state owners so that they can be
620 * reused, to avoid the OPEN_CONFIRM on minor version 0.
621 * It also pins the uniquifier of dropped state owners for
622 * a while, to ensure that those state owner names are
625 void nfs4_put_state_owner(struct nfs4_state_owner *sp)
627 struct nfs_server *server = sp->so_server;
628 struct nfs_client *clp = server->nfs_client;
630 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
633 sp->so_expires = jiffies;
634 list_add_tail(&sp->so_lru, &server->state_owners_lru);
635 spin_unlock(&clp->cl_lock);
639 * nfs4_purge_state_owners - Release all cached state owners
640 * @server: nfs_server with cached state owners to release
642 * Called at umount time. Remaining state owners will be on
643 * the LRU with ref count of zero.
645 void nfs4_purge_state_owners(struct nfs_server *server)
647 struct nfs_client *clp = server->nfs_client;
648 struct nfs4_state_owner *sp, *tmp;
651 spin_lock(&clp->cl_lock);
652 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
653 list_move(&sp->so_lru, &doomed);
654 nfs4_remove_state_owner_locked(sp);
656 spin_unlock(&clp->cl_lock);
658 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
659 list_del(&sp->so_lru);
660 nfs4_free_state_owner(sp);
664 static struct nfs4_state *
665 nfs4_alloc_open_state(void)
667 struct nfs4_state *state;
669 state = kzalloc(sizeof(*state), GFP_NOFS);
672 atomic_set(&state->count, 1);
673 INIT_LIST_HEAD(&state->lock_states);
674 spin_lock_init(&state->state_lock);
675 seqlock_init(&state->seqlock);
680 nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
682 if (state->state == fmode)
684 /* NB! List reordering - see the reclaim code for why. */
685 if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
686 if (fmode & FMODE_WRITE)
687 list_move(&state->open_states, &state->owner->so_states);
689 list_move_tail(&state->open_states, &state->owner->so_states);
691 state->state = fmode;
694 static struct nfs4_state *
695 __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
697 struct nfs_inode *nfsi = NFS_I(inode);
698 struct nfs4_state *state;
700 list_for_each_entry(state, &nfsi->open_states, inode_states) {
701 if (state->owner != owner)
703 if (atomic_inc_not_zero(&state->count))
710 nfs4_free_open_state(struct nfs4_state *state)
716 nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
718 struct nfs4_state *state, *new;
719 struct nfs_inode *nfsi = NFS_I(inode);
721 spin_lock(&inode->i_lock);
722 state = __nfs4_find_state_byowner(inode, owner);
723 spin_unlock(&inode->i_lock);
726 new = nfs4_alloc_open_state();
727 spin_lock(&owner->so_lock);
728 spin_lock(&inode->i_lock);
729 state = __nfs4_find_state_byowner(inode, owner);
730 if (state == NULL && new != NULL) {
732 state->owner = owner;
733 atomic_inc(&owner->so_count);
734 list_add(&state->inode_states, &nfsi->open_states);
736 state->inode = inode;
737 spin_unlock(&inode->i_lock);
738 /* Note: The reclaim code dictates that we add stateless
739 * and read-only stateids to the end of the list */
740 list_add_tail(&state->open_states, &owner->so_states);
741 spin_unlock(&owner->so_lock);
743 spin_unlock(&inode->i_lock);
744 spin_unlock(&owner->so_lock);
746 nfs4_free_open_state(new);
752 void nfs4_put_open_state(struct nfs4_state *state)
754 struct inode *inode = state->inode;
755 struct nfs4_state_owner *owner = state->owner;
757 if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
759 spin_lock(&inode->i_lock);
760 list_del(&state->inode_states);
761 list_del(&state->open_states);
762 spin_unlock(&inode->i_lock);
763 spin_unlock(&owner->so_lock);
765 nfs4_free_open_state(state);
766 nfs4_put_state_owner(owner);
770 * Close the current file.
772 static void __nfs4_close(struct nfs4_state *state,
773 fmode_t fmode, gfp_t gfp_mask, int wait)
775 struct nfs4_state_owner *owner = state->owner;
779 atomic_inc(&owner->so_count);
780 /* Protect against nfs4_find_state() */
781 spin_lock(&owner->so_lock);
782 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
789 case FMODE_READ|FMODE_WRITE:
792 newstate = FMODE_READ|FMODE_WRITE;
793 if (state->n_rdwr == 0) {
794 if (state->n_rdonly == 0) {
795 newstate &= ~FMODE_READ;
796 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
797 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
799 if (state->n_wronly == 0) {
800 newstate &= ~FMODE_WRITE;
801 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
802 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
805 clear_bit(NFS_DELEGATED_STATE, &state->flags);
807 nfs4_state_set_mode_locked(state, newstate);
808 spin_unlock(&owner->so_lock);
811 nfs4_put_open_state(state);
812 nfs4_put_state_owner(owner);
814 nfs4_do_close(state, gfp_mask, wait);
817 void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
819 __nfs4_close(state, fmode, GFP_NOFS, 0);
822 void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
824 __nfs4_close(state, fmode, GFP_KERNEL, 1);
828 * Search the state->lock_states for an existing lock_owner
829 * that is compatible with current->files
831 static struct nfs4_lock_state *
832 __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type)
834 struct nfs4_lock_state *pos;
835 list_for_each_entry(pos, &state->lock_states, ls_locks) {
836 if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
838 switch (pos->ls_owner.lo_type) {
839 case NFS4_POSIX_LOCK_TYPE:
840 if (pos->ls_owner.lo_u.posix_owner != fl_owner)
843 case NFS4_FLOCK_LOCK_TYPE:
844 if (pos->ls_owner.lo_u.flock_owner != fl_pid)
847 atomic_inc(&pos->ls_count);
854 * Return a compatible lock_state. If no initialized lock_state structure
855 * exists, return an uninitialized one.
858 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)
860 struct nfs4_lock_state *lsp;
861 struct nfs_server *server = state->owner->so_server;
863 lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
866 nfs4_init_seqid_counter(&lsp->ls_seqid);
867 atomic_set(&lsp->ls_count, 1);
868 lsp->ls_state = state;
869 lsp->ls_owner.lo_type = type;
870 switch (lsp->ls_owner.lo_type) {
871 case NFS4_FLOCK_LOCK_TYPE:
872 lsp->ls_owner.lo_u.flock_owner = fl_pid;
874 case NFS4_POSIX_LOCK_TYPE:
875 lsp->ls_owner.lo_u.posix_owner = fl_owner;
880 lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
881 if (lsp->ls_seqid.owner_id < 0)
883 INIT_LIST_HEAD(&lsp->ls_locks);
890 void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
892 ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
893 nfs4_destroy_seqid_counter(&lsp->ls_seqid);
898 * Return a compatible lock_state. If no initialized lock_state structure
899 * exists, return an uninitialized one.
902 static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner, pid_t pid, unsigned int type)
904 struct nfs4_lock_state *lsp, *new = NULL;
907 spin_lock(&state->state_lock);
908 lsp = __nfs4_find_lock_state(state, owner, pid, type);
912 list_add(&new->ls_locks, &state->lock_states);
913 set_bit(LK_STATE_IN_USE, &state->flags);
918 spin_unlock(&state->state_lock);
919 new = nfs4_alloc_lock_state(state, owner, pid, type);
923 spin_unlock(&state->state_lock);
925 nfs4_free_lock_state(state->owner->so_server, new);
930 * Release reference to lock_state, and free it if we see that
931 * it is no longer in use
933 void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
935 struct nfs4_state *state;
939 state = lsp->ls_state;
940 if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
942 list_del(&lsp->ls_locks);
943 if (list_empty(&state->lock_states))
944 clear_bit(LK_STATE_IN_USE, &state->flags);
945 spin_unlock(&state->state_lock);
946 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
947 if (nfs4_release_lockowner(lsp) == 0)
950 nfs4_free_lock_state(lsp->ls_state->owner->so_server, lsp);
953 static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
955 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
957 dst->fl_u.nfs4_fl.owner = lsp;
958 atomic_inc(&lsp->ls_count);
961 static void nfs4_fl_release_lock(struct file_lock *fl)
963 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
966 static const struct file_lock_operations nfs4_fl_lock_ops = {
967 .fl_copy_lock = nfs4_fl_copy_lock,
968 .fl_release_private = nfs4_fl_release_lock,
971 int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
973 struct nfs4_lock_state *lsp;
975 if (fl->fl_ops != NULL)
977 if (fl->fl_flags & FL_POSIX)
978 lsp = nfs4_get_lock_state(state, fl->fl_owner, 0, NFS4_POSIX_LOCK_TYPE);
979 else if (fl->fl_flags & FL_FLOCK)
980 lsp = nfs4_get_lock_state(state, NULL, fl->fl_pid,
981 NFS4_FLOCK_LOCK_TYPE);
986 fl->fl_u.nfs4_fl.owner = lsp;
987 fl->fl_ops = &nfs4_fl_lock_ops;
991 static bool nfs4_copy_lock_stateid(nfs4_stateid *dst, struct nfs4_state *state,
992 const struct nfs_lockowner *lockowner)
994 struct nfs4_lock_state *lsp;
1000 if (lockowner == NULL)
1003 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
1006 fl_owner = lockowner->l_owner;
1007 fl_pid = lockowner->l_pid;
1008 spin_lock(&state->state_lock);
1009 lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
1010 if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
1011 nfs4_stateid_copy(dst, &lsp->ls_stateid);
1014 spin_unlock(&state->state_lock);
1015 nfs4_put_lock_state(lsp);
1020 static void nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
1025 seq = read_seqbegin(&state->seqlock);
1026 nfs4_stateid_copy(dst, &state->stateid);
1027 } while (read_seqretry(&state->seqlock, seq));
1031 * Byte-range lock aware utility to initialize the stateid of read/write
1034 void nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state,
1035 fmode_t fmode, const struct nfs_lockowner *lockowner)
1037 if (nfs4_copy_delegation_stateid(dst, state->inode, fmode))
1039 if (nfs4_copy_lock_stateid(dst, state, lockowner))
1041 nfs4_copy_open_stateid(dst, state);
1044 struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
1046 struct nfs_seqid *new;
1048 new = kmalloc(sizeof(*new), gfp_mask);
1050 new->sequence = counter;
1051 INIT_LIST_HEAD(&new->list);
1057 void nfs_release_seqid(struct nfs_seqid *seqid)
1059 struct nfs_seqid_counter *sequence;
1061 if (list_empty(&seqid->list))
1063 sequence = seqid->sequence;
1064 spin_lock(&sequence->lock);
1065 list_del_init(&seqid->list);
1066 if (!list_empty(&sequence->list)) {
1067 struct nfs_seqid *next;
1069 next = list_first_entry(&sequence->list,
1070 struct nfs_seqid, list);
1071 rpc_wake_up_queued_task(&sequence->wait, next->task);
1073 spin_unlock(&sequence->lock);
1076 void nfs_free_seqid(struct nfs_seqid *seqid)
1078 nfs_release_seqid(seqid);
1083 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
1084 * failed with a seqid incrementing error -
1085 * see comments nfs_fs.h:seqid_mutating_error()
1087 static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
1092 case -NFS4ERR_BAD_SEQID:
1093 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
1095 pr_warn_ratelimited("NFS: v4 server returned a bad"
1096 " sequence-id error on an"
1097 " unconfirmed sequence %p!\n",
1099 case -NFS4ERR_STALE_CLIENTID:
1100 case -NFS4ERR_STALE_STATEID:
1101 case -NFS4ERR_BAD_STATEID:
1102 case -NFS4ERR_BADXDR:
1103 case -NFS4ERR_RESOURCE:
1104 case -NFS4ERR_NOFILEHANDLE:
1105 /* Non-seqid mutating errors */
1109 * Note: no locking needed as we are guaranteed to be first
1110 * on the sequence list
1112 seqid->sequence->counter++;
1115 void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
1117 struct nfs4_state_owner *sp = container_of(seqid->sequence,
1118 struct nfs4_state_owner, so_seqid);
1119 struct nfs_server *server = sp->so_server;
1121 if (status == -NFS4ERR_BAD_SEQID)
1122 nfs4_drop_state_owner(sp);
1123 if (!nfs4_has_session(server->nfs_client))
1124 nfs_increment_seqid(status, seqid);
1128 * Increment the seqid if the LOCK/LOCKU succeeded, or
1129 * failed with a seqid incrementing error -
1130 * see comments nfs_fs.h:seqid_mutating_error()
1132 void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
1134 nfs_increment_seqid(status, seqid);
1137 int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
1139 struct nfs_seqid_counter *sequence = seqid->sequence;
1142 spin_lock(&sequence->lock);
1144 if (list_empty(&seqid->list))
1145 list_add_tail(&seqid->list, &sequence->list);
1146 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
1148 rpc_sleep_on(&sequence->wait, task, NULL);
1151 spin_unlock(&sequence->lock);
1155 static int nfs4_run_state_manager(void *);
1157 static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
1159 smp_mb__before_clear_bit();
1160 clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
1161 smp_mb__after_clear_bit();
1162 wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
1163 rpc_wake_up(&clp->cl_rpcwaitq);
1167 * Schedule the nfs_client asynchronous state management routine
1169 void nfs4_schedule_state_manager(struct nfs_client *clp)
1171 struct task_struct *task;
1172 char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
1174 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1176 __module_get(THIS_MODULE);
1177 atomic_inc(&clp->cl_count);
1179 /* The rcu_read_lock() is not strictly necessary, as the state
1180 * manager is the only thread that ever changes the rpc_xprt
1181 * after it's initialized. At this point, we're single threaded. */
1183 snprintf(buf, sizeof(buf), "%s-manager",
1184 rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR));
1186 task = kthread_run(nfs4_run_state_manager, clp, buf);
1188 printk(KERN_ERR "%s: kthread_run: %ld\n",
1189 __func__, PTR_ERR(task));
1190 nfs4_clear_state_manager_bit(clp);
1191 nfs_put_client(clp);
1192 module_put(THIS_MODULE);
1197 * Schedule a lease recovery attempt
1199 void nfs4_schedule_lease_recovery(struct nfs_client *clp)
1203 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1204 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1205 dprintk("%s: scheduling lease recovery for server %s\n", __func__,
1207 nfs4_schedule_state_manager(clp);
1209 EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
1212 * nfs40_handle_cb_pathdown - return all delegations after NFS4ERR_CB_PATH_DOWN
1213 * @clp: client to process
1215 * Set the NFS4CLNT_LEASE_EXPIRED state in order to force a
1216 * resend of the SETCLIENTID and hence re-establish the
1217 * callback channel. Then return all existing delegations.
1219 static void nfs40_handle_cb_pathdown(struct nfs_client *clp)
1221 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1222 nfs_expire_all_delegations(clp);
1223 dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
1227 void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
1229 nfs40_handle_cb_pathdown(clp);
1230 nfs4_schedule_state_manager(clp);
1233 static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1236 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1237 /* Don't recover state that expired before the reboot */
1238 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
1239 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1242 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1243 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1247 static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1249 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
1250 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1251 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1252 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
1256 void nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
1258 struct nfs_client *clp = server->nfs_client;
1260 nfs4_state_mark_reclaim_nograce(clp, state);
1261 dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
1263 nfs4_schedule_state_manager(clp);
1265 EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1267 void nfs_inode_find_state_and_recover(struct inode *inode,
1268 const nfs4_stateid *stateid)
1270 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1271 struct nfs_inode *nfsi = NFS_I(inode);
1272 struct nfs_open_context *ctx;
1273 struct nfs4_state *state;
1276 spin_lock(&inode->i_lock);
1277 list_for_each_entry(ctx, &nfsi->open_files, list) {
1281 if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
1283 if (!nfs4_stateid_match(&state->stateid, stateid))
1285 nfs4_state_mark_reclaim_nograce(clp, state);
1288 spin_unlock(&inode->i_lock);
1290 nfs4_schedule_state_manager(clp);
1294 static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
1296 struct inode *inode = state->inode;
1297 struct nfs_inode *nfsi = NFS_I(inode);
1298 struct file_lock *fl;
1301 if (inode->i_flock == NULL)
1304 /* Guard against delegation returns and new lock/unlock calls */
1305 down_write(&nfsi->rwsem);
1306 /* Protect inode->i_flock using the BKL */
1308 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1309 if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
1311 if (nfs_file_open_context(fl->fl_file)->state != state)
1314 status = ops->recover_lock(state, fl);
1319 case -NFS4ERR_ADMIN_REVOKED:
1320 case -NFS4ERR_STALE_STATEID:
1321 case -NFS4ERR_BAD_STATEID:
1322 case -NFS4ERR_EXPIRED:
1323 case -NFS4ERR_NO_GRACE:
1324 case -NFS4ERR_STALE_CLIENTID:
1325 case -NFS4ERR_BADSESSION:
1326 case -NFS4ERR_BADSLOT:
1327 case -NFS4ERR_BAD_HIGH_SLOT:
1328 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1331 printk(KERN_ERR "NFS: %s: unhandled error %d. "
1332 "Zeroing state\n", __func__, status);
1334 case -NFS4ERR_DENIED:
1335 case -NFS4ERR_RECLAIM_BAD:
1336 case -NFS4ERR_RECLAIM_CONFLICT:
1337 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1344 up_write(&nfsi->rwsem);
1348 static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
1350 struct nfs4_state *state;
1351 struct nfs4_lock_state *lock;
1354 /* Note: we rely on the sp->so_states list being ordered
1355 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
1357 * This is needed to ensure that the server won't give us any
1358 * read delegations that we have to return if, say, we are
1359 * recovering after a network partition or a reboot from a
1360 * server that doesn't support a grace period.
1363 spin_lock(&sp->so_lock);
1364 list_for_each_entry(state, &sp->so_states, open_states) {
1365 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
1367 if (state->state == 0)
1369 atomic_inc(&state->count);
1370 spin_unlock(&sp->so_lock);
1371 status = ops->recover_open(sp, state);
1373 status = nfs4_reclaim_locks(state, ops);
1375 spin_lock(&state->state_lock);
1376 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1377 if (!test_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags))
1378 pr_warn_ratelimited("NFS: "
1380 "failed!\n", __func__);
1382 spin_unlock(&state->state_lock);
1383 nfs4_put_open_state(state);
1389 printk(KERN_ERR "NFS: %s: unhandled error %d. "
1390 "Zeroing state\n", __func__, status);
1395 * Open state on this file cannot be recovered
1396 * All we can do is revert to using the zero stateid.
1398 memset(&state->stateid, 0,
1399 sizeof(state->stateid));
1400 /* Mark the file as being 'closed' */
1405 * User RPCSEC_GSS context has expired.
1406 * We cannot recover this stateid now, so
1407 * skip it and allow recovery thread to
1411 case -NFS4ERR_ADMIN_REVOKED:
1412 case -NFS4ERR_STALE_STATEID:
1413 case -NFS4ERR_BAD_STATEID:
1414 case -NFS4ERR_RECLAIM_BAD:
1415 case -NFS4ERR_RECLAIM_CONFLICT:
1416 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1418 case -NFS4ERR_EXPIRED:
1419 case -NFS4ERR_NO_GRACE:
1420 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1421 case -NFS4ERR_STALE_CLIENTID:
1422 case -NFS4ERR_BADSESSION:
1423 case -NFS4ERR_BADSLOT:
1424 case -NFS4ERR_BAD_HIGH_SLOT:
1425 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1428 nfs4_put_open_state(state);
1431 spin_unlock(&sp->so_lock);
1434 nfs4_put_open_state(state);
1438 static void nfs4_clear_open_state(struct nfs4_state *state)
1440 struct nfs4_lock_state *lock;
1442 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1443 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1444 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1445 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1446 spin_lock(&state->state_lock);
1447 list_for_each_entry(lock, &state->lock_states, ls_locks) {
1448 lock->ls_seqid.flags = 0;
1449 clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
1451 spin_unlock(&state->state_lock);
1454 static void nfs4_reset_seqids(struct nfs_server *server,
1455 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1457 struct nfs_client *clp = server->nfs_client;
1458 struct nfs4_state_owner *sp;
1459 struct rb_node *pos;
1460 struct nfs4_state *state;
1462 spin_lock(&clp->cl_lock);
1463 for (pos = rb_first(&server->state_owners);
1465 pos = rb_next(pos)) {
1466 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1467 sp->so_seqid.flags = 0;
1468 spin_lock(&sp->so_lock);
1469 list_for_each_entry(state, &sp->so_states, open_states) {
1470 if (mark_reclaim(clp, state))
1471 nfs4_clear_open_state(state);
1473 spin_unlock(&sp->so_lock);
1475 spin_unlock(&clp->cl_lock);
1478 static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp,
1479 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1481 struct nfs_server *server;
1484 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1485 nfs4_reset_seqids(server, mark_reclaim);
1489 static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1491 /* Mark all delegations for reclaim */
1492 nfs_delegation_mark_reclaim(clp);
1493 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1496 static void nfs4_reclaim_complete(struct nfs_client *clp,
1497 const struct nfs4_state_recovery_ops *ops)
1499 /* Notify the server we're done reclaiming our state */
1500 if (ops->reclaim_complete)
1501 (void)ops->reclaim_complete(clp);
1504 static void nfs4_clear_reclaim_server(struct nfs_server *server)
1506 struct nfs_client *clp = server->nfs_client;
1507 struct nfs4_state_owner *sp;
1508 struct rb_node *pos;
1509 struct nfs4_state *state;
1511 spin_lock(&clp->cl_lock);
1512 for (pos = rb_first(&server->state_owners);
1514 pos = rb_next(pos)) {
1515 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1516 spin_lock(&sp->so_lock);
1517 list_for_each_entry(state, &sp->so_states, open_states) {
1518 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
1521 nfs4_state_mark_reclaim_nograce(clp, state);
1523 spin_unlock(&sp->so_lock);
1525 spin_unlock(&clp->cl_lock);
1528 static int nfs4_state_clear_reclaim_reboot(struct nfs_client *clp)
1530 struct nfs_server *server;
1532 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1536 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1537 nfs4_clear_reclaim_server(server);
1540 nfs_delegation_reap_unclaimed(clp);
1544 static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
1546 if (!nfs4_state_clear_reclaim_reboot(clp))
1548 nfs4_reclaim_complete(clp, clp->cl_mvops->reboot_recovery_ops);
1551 static void nfs_delegation_clear_all(struct nfs_client *clp)
1553 nfs_delegation_mark_reclaim(clp);
1554 nfs_delegation_reap_unclaimed(clp);
1557 static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1559 nfs_delegation_clear_all(clp);
1560 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1563 static void nfs4_warn_keyexpired(const char *s)
1565 printk_ratelimited(KERN_WARNING "Error: state manager"
1566 " encountered RPCSEC_GSS session"
1567 " expired against NFSv4 server %s.\n",
1571 static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1576 case -NFS4ERR_CB_PATH_DOWN:
1577 nfs40_handle_cb_pathdown(clp);
1579 case -NFS4ERR_NO_GRACE:
1580 nfs4_state_end_reclaim_reboot(clp);
1582 case -NFS4ERR_STALE_CLIENTID:
1583 case -NFS4ERR_LEASE_MOVED:
1584 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1585 nfs4_state_clear_reclaim_reboot(clp);
1586 nfs4_state_start_reclaim_reboot(clp);
1588 case -NFS4ERR_EXPIRED:
1589 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1590 nfs4_state_start_reclaim_nograce(clp);
1592 case -NFS4ERR_BADSESSION:
1593 case -NFS4ERR_BADSLOT:
1594 case -NFS4ERR_BAD_HIGH_SLOT:
1595 case -NFS4ERR_DEADSESSION:
1596 case -NFS4ERR_SEQ_FALSE_RETRY:
1597 case -NFS4ERR_SEQ_MISORDERED:
1598 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1599 /* Zero session reset errors */
1601 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1602 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1605 /* Nothing we can do */
1606 nfs4_warn_keyexpired(clp->cl_hostname);
1609 dprintk("%s: failed to handle error %d for server %s\n",
1610 __func__, error, clp->cl_hostname);
1613 dprintk("%s: handled error %d for server %s\n", __func__, error,
1618 static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1620 struct nfs4_state_owner *sp;
1621 struct nfs_server *server;
1622 struct rb_node *pos;
1627 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1628 nfs4_purge_state_owners(server);
1629 spin_lock(&clp->cl_lock);
1630 for (pos = rb_first(&server->state_owners);
1632 pos = rb_next(pos)) {
1634 struct nfs4_state_owner, so_server_node);
1635 if (!test_and_clear_bit(ops->owner_flag_bit,
1638 atomic_inc(&sp->so_count);
1639 spin_unlock(&clp->cl_lock);
1642 status = nfs4_reclaim_open_state(sp, ops);
1644 set_bit(ops->owner_flag_bit, &sp->so_flags);
1645 nfs4_put_state_owner(sp);
1646 return nfs4_recovery_handle_error(clp, status);
1649 nfs4_put_state_owner(sp);
1652 spin_unlock(&clp->cl_lock);
1658 static int nfs4_check_lease(struct nfs_client *clp)
1660 struct rpc_cred *cred;
1661 const struct nfs4_state_maintenance_ops *ops =
1662 clp->cl_mvops->state_renewal_ops;
1665 /* Is the client already known to have an expired lease? */
1666 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1668 spin_lock(&clp->cl_lock);
1669 cred = ops->get_state_renewal_cred_locked(clp);
1670 spin_unlock(&clp->cl_lock);
1672 cred = nfs4_get_setclientid_cred(clp);
1677 status = ops->renew_lease(clp, cred);
1680 return nfs4_recovery_handle_error(clp, status);
1683 /* Set NFS4CLNT_LEASE_EXPIRED and reclaim reboot state for all v4.0 errors
1684 * and for recoverable errors on EXCHANGE_ID for v4.1
1686 static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
1689 case -NFS4ERR_SEQ_MISORDERED:
1690 if (test_and_set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state))
1691 return -ESERVERFAULT;
1692 /* Lease confirmation error: retry after purging the lease */
1694 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1696 case -NFS4ERR_STALE_CLIENTID:
1697 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1698 nfs4_state_clear_reclaim_reboot(clp);
1699 nfs4_state_start_reclaim_reboot(clp);
1701 case -NFS4ERR_CLID_INUSE:
1702 pr_err("NFS: Server %s reports our clientid is in use\n",
1704 nfs_mark_client_ready(clp, -EPERM);
1705 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1708 if (clp->cl_machine_cred == NULL)
1710 /* Handle case where the user hasn't set up machine creds */
1711 nfs4_clear_machine_cred(clp);
1712 case -NFS4ERR_DELAY:
1718 case -NFS4ERR_MINOR_VERS_MISMATCH:
1719 if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
1720 nfs_mark_client_ready(clp, -EPROTONOSUPPORT);
1721 dprintk("%s: exit with error %d for server %s\n",
1722 __func__, -EPROTONOSUPPORT, clp->cl_hostname);
1723 return -EPROTONOSUPPORT;
1725 nfs4_warn_keyexpired(clp->cl_hostname);
1726 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1727 * in nfs4_exchange_id */
1729 dprintk("%s: exit with error %d for server %s\n", __func__,
1730 status, clp->cl_hostname);
1733 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1734 dprintk("%s: handled error %d for server %s\n", __func__, status,
1739 static int nfs4_establish_lease(struct nfs_client *clp)
1741 struct rpc_cred *cred;
1742 const struct nfs4_state_recovery_ops *ops =
1743 clp->cl_mvops->reboot_recovery_ops;
1746 cred = ops->get_clid_cred(clp);
1749 status = ops->establish_clid(clp, cred);
1753 pnfs_destroy_all_layouts(clp);
1758 * Returns zero or a negative errno. NFS4ERR values are converted
1759 * to local errno values.
1761 static int nfs4_reclaim_lease(struct nfs_client *clp)
1765 status = nfs4_establish_lease(clp);
1767 return nfs4_handle_reclaim_lease_error(clp, status);
1768 if (test_and_clear_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state))
1769 nfs4_state_start_reclaim_nograce(clp);
1770 if (!test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
1771 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1772 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1773 clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1777 static int nfs4_purge_lease(struct nfs_client *clp)
1781 status = nfs4_establish_lease(clp);
1783 return nfs4_handle_reclaim_lease_error(clp, status);
1784 clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1785 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1786 nfs4_state_start_reclaim_nograce(clp);
1791 * nfs4_discover_server_trunking - Detect server IP address trunking
1793 * @clp: nfs_client under test
1794 * @result: OUT: found nfs_client, or clp
1796 * Returns zero or a negative errno. If zero is returned,
1797 * an nfs_client pointer is planted in "result".
1799 * Note: since we are invoked in process context, and
1800 * not from inside the state manager, we cannot use
1801 * nfs4_handle_reclaim_lease_error().
1803 int nfs4_discover_server_trunking(struct nfs_client *clp,
1804 struct nfs_client **result)
1806 const struct nfs4_state_recovery_ops *ops =
1807 clp->cl_mvops->reboot_recovery_ops;
1808 rpc_authflavor_t *flavors, flav, save;
1809 struct rpc_clnt *clnt;
1810 struct rpc_cred *cred;
1813 dprintk("NFS: %s: testing '%s'\n", __func__, clp->cl_hostname);
1815 len = NFS_MAX_SECFLAVORS;
1816 flavors = kcalloc(len, sizeof(*flavors), GFP_KERNEL);
1817 if (flavors == NULL) {
1821 len = rpcauth_list_flavors(flavors, len);
1826 clnt = clp->cl_rpcclient;
1827 save = clnt->cl_auth->au_flavor;
1830 mutex_lock(&nfs_clid_init_mutex);
1833 cred = ops->get_clid_cred(clp);
1837 status = ops->detect_trunking(clp, result, cred);
1844 if (clp->cl_machine_cred == NULL)
1846 /* Handle case where the user hasn't set up machine creds */
1847 nfs4_clear_machine_cred(clp);
1848 case -NFS4ERR_DELAY:
1852 dprintk("NFS: %s after status %d, retrying\n",
1856 case -NFS4ERR_CLID_INUSE:
1857 case -NFS4ERR_WRONGSEC:
1862 flav = flavors[i++];
1864 flav = flavors[i++];
1865 clnt = rpc_clone_client_set_auth(clnt, flav);
1867 status = PTR_ERR(clnt);
1870 clp->cl_rpcclient = clnt;
1873 case -NFS4ERR_MINOR_VERS_MISMATCH:
1874 status = -EPROTONOSUPPORT;
1878 nfs4_warn_keyexpired(clp->cl_hostname);
1879 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
1880 * in nfs4_exchange_id */
1881 status = -EKEYEXPIRED;
1885 mutex_unlock(&nfs_clid_init_mutex);
1889 dprintk("NFS: %s: status = %d\n", __func__, status);
1893 #ifdef CONFIG_NFS_V4_1
1894 void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
1896 struct nfs_client *clp = session->clp;
1900 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1902 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1903 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1905 nfs4_schedule_lease_recovery(clp);
1907 EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
1909 void nfs41_handle_recall_slot(struct nfs_client *clp)
1911 set_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
1912 dprintk("%s: scheduling slot recall for server %s\n", __func__,
1914 nfs4_schedule_state_manager(clp);
1917 static void nfs4_reset_all_state(struct nfs_client *clp)
1919 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1920 set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1921 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1922 nfs4_state_start_reclaim_nograce(clp);
1923 dprintk("%s: scheduling reset of all state for server %s!\n",
1924 __func__, clp->cl_hostname);
1925 nfs4_schedule_state_manager(clp);
1929 static void nfs41_handle_server_reboot(struct nfs_client *clp)
1931 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1932 nfs4_state_start_reclaim_reboot(clp);
1933 dprintk("%s: server %s rebooted!\n", __func__,
1935 nfs4_schedule_state_manager(clp);
1939 static void nfs41_handle_state_revoked(struct nfs_client *clp)
1941 nfs4_reset_all_state(clp);
1942 dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
1945 static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
1947 /* This will need to handle layouts too */
1948 nfs_expire_all_delegations(clp);
1949 dprintk("%s: Recallable state revoked on server %s!\n", __func__,
1953 static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
1955 nfs_expire_all_delegations(clp);
1956 if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
1957 nfs4_schedule_state_manager(clp);
1958 dprintk("%s: server %s declared a backchannel fault\n", __func__,
1962 static void nfs41_handle_cb_path_down(struct nfs_client *clp)
1964 if (test_and_set_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
1965 &clp->cl_state) == 0)
1966 nfs4_schedule_state_manager(clp);
1969 void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
1974 dprintk("%s: \"%s\" (client ID %llx) flags=0x%08x\n",
1975 __func__, clp->cl_hostname, clp->cl_clientid, flags);
1977 if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
1978 nfs41_handle_server_reboot(clp);
1979 if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
1980 SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
1981 SEQ4_STATUS_ADMIN_STATE_REVOKED |
1982 SEQ4_STATUS_LEASE_MOVED))
1983 nfs41_handle_state_revoked(clp);
1984 if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
1985 nfs41_handle_recallable_state_revoked(clp);
1986 if (flags & SEQ4_STATUS_BACKCHANNEL_FAULT)
1987 nfs41_handle_backchannel_fault(clp);
1988 else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
1989 SEQ4_STATUS_CB_PATH_DOWN_SESSION))
1990 nfs41_handle_cb_path_down(clp);
1993 static int nfs4_reset_session(struct nfs_client *clp)
1995 struct rpc_cred *cred;
1998 if (!nfs4_has_session(clp))
2000 nfs4_begin_drain_session(clp);
2001 cred = nfs4_get_exchange_id_cred(clp);
2002 status = nfs4_proc_destroy_session(clp->cl_session, cred);
2003 if (status && status != -NFS4ERR_BADSESSION &&
2004 status != -NFS4ERR_DEADSESSION) {
2005 status = nfs4_recovery_handle_error(clp, status);
2009 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
2010 status = nfs4_proc_create_session(clp, cred);
2012 dprintk("%s: session reset failed with status %d for server %s!\n",
2013 __func__, status, clp->cl_hostname);
2014 status = nfs4_handle_reclaim_lease_error(clp, status);
2017 nfs41_finish_session_reset(clp);
2018 dprintk("%s: session reset was successful for server %s!\n",
2019 __func__, clp->cl_hostname);
2026 static int nfs4_recall_slot(struct nfs_client *clp)
2028 struct nfs4_slot_table *fc_tbl;
2029 struct nfs4_slot *new, *old;
2032 if (!nfs4_has_session(clp))
2034 nfs4_begin_drain_session(clp);
2035 fc_tbl = &clp->cl_session->fc_slot_table;
2036 new = nfs4_alloc_slots(fc_tbl, fc_tbl->target_max_slots, GFP_NOFS);
2040 spin_lock(&fc_tbl->slot_tbl_lock);
2041 for (i = 0; i < fc_tbl->target_max_slots; i++)
2042 new[i].seq_nr = fc_tbl->slots[i].seq_nr;
2043 old = fc_tbl->slots;
2044 fc_tbl->slots = new;
2045 fc_tbl->max_slots = fc_tbl->target_max_slots;
2046 fc_tbl->target_max_slots = 0;
2047 clp->cl_session->fc_attrs.max_reqs = fc_tbl->max_slots;
2048 spin_unlock(&fc_tbl->slot_tbl_lock);
2054 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2056 struct rpc_cred *cred;
2059 if (!nfs4_has_session(clp))
2061 nfs4_begin_drain_session(clp);
2062 cred = nfs4_get_exchange_id_cred(clp);
2063 ret = nfs4_proc_bind_conn_to_session(clp, cred);
2066 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2069 dprintk("%s: bind_conn_to_session was successful for server %s!\n",
2070 __func__, clp->cl_hostname);
2072 case -NFS4ERR_DELAY:
2074 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2077 return nfs4_recovery_handle_error(clp, ret);
2081 #else /* CONFIG_NFS_V4_1 */
2082 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
2083 static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
2084 static int nfs4_recall_slot(struct nfs_client *clp) { return 0; }
2086 static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2090 #endif /* CONFIG_NFS_V4_1 */
2092 static void nfs4_state_manager(struct nfs_client *clp)
2095 const char *section = "", *section_sep = "";
2097 /* Ensure exclusive access to NFSv4 state */
2099 if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
2100 section = "purge state";
2101 status = nfs4_purge_lease(clp);
2107 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
2108 section = "lease expired";
2109 /* We're going to have to re-establish a clientid */
2110 status = nfs4_reclaim_lease(clp);
2116 /* Initialize or reset the session */
2117 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
2118 section = "reset session";
2119 status = nfs4_reset_session(clp);
2120 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
2126 /* Send BIND_CONN_TO_SESSION */
2127 if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2129 section = "bind conn to session";
2130 status = nfs4_bind_conn_to_session(clp);
2136 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
2137 section = "check lease";
2138 status = nfs4_check_lease(clp);
2144 /* Recall session slots */
2145 if (test_and_clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state)) {
2146 section = "recall slot";
2147 status = nfs4_recall_slot(clp);
2153 /* First recover reboot state... */
2154 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
2155 section = "reclaim reboot";
2156 status = nfs4_do_reclaim(clp,
2157 clp->cl_mvops->reboot_recovery_ops);
2158 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2159 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
2161 nfs4_state_end_reclaim_reboot(clp);
2162 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
2168 /* Now recover expired state... */
2169 if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
2170 section = "reclaim nograce";
2171 status = nfs4_do_reclaim(clp,
2172 clp->cl_mvops->nograce_recovery_ops);
2173 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2174 test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
2175 test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
2181 nfs4_end_drain_session(clp);
2182 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
2183 nfs_client_return_marked_delegations(clp);
2187 nfs4_clear_state_manager_bit(clp);
2188 /* Did we race with an attempt to give us more work? */
2189 if (clp->cl_state == 0)
2191 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
2193 } while (atomic_read(&clp->cl_count) > 1);
2196 if (strlen(section))
2198 pr_warn_ratelimited("NFS: state manager%s%s failed on NFSv4 server %s"
2199 " with error %d\n", section_sep, section,
2200 clp->cl_hostname, -status);
2202 nfs4_end_drain_session(clp);
2203 nfs4_clear_state_manager_bit(clp);
2206 static int nfs4_run_state_manager(void *ptr)
2208 struct nfs_client *clp = ptr;
2210 allow_signal(SIGKILL);
2211 nfs4_state_manager(clp);
2212 nfs_put_client(clp);
2213 module_put_and_exit(0);