]> git.ipfire.org Git - thirdparty/kernel/stable.git/blame - fs/nfs/nfs4state.c
Merge tag 'kvm-x86-misc-6.7' of https://github.com/kvm-x86/linux into HEAD
[thirdparty/kernel/stable.git] / fs / nfs / nfs4state.c
CommitLineData
1da177e4
LT
1/*
2 * fs/nfs/nfs4state.c
3 *
4 * Client-side XDR for NFSv4.
5 *
6 * Copyright (c) 2002 The Regents of the University of Michigan.
7 * All rights reserved.
8 *
9 * Kendrick Smith <kmsmith@umich.edu>
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 *
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.
23 *
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.
35 *
36 * Implementation of the NFSv4 state model. For the time being,
37 * this is minimal, but will be made much more complex in a
38 * subsequent patch.
39 */
40
6f43ddcc 41#include <linux/kernel.h>
1da177e4 42#include <linux/slab.h>
b89f4321 43#include <linux/fs.h>
1da177e4 44#include <linux/nfs_fs.h>
5043e900
TM
45#include <linux/kthread.h>
46#include <linux/module.h>
9f958ab8 47#include <linux/random.h>
8c7597f6 48#include <linux/ratelimit.h>
1da177e4
LT
49#include <linux/workqueue.h>
50#include <linux/bitops.h>
0aaaf5c4 51#include <linux/jiffies.h>
3e17898a 52#include <linux/sched/mm.h>
1da177e4 53
05f4c350
CL
54#include <linux/sunrpc/clnt.h>
55
4ce79717 56#include "nfs4_fs.h"
1da177e4
LT
57#include "callback.h"
58#include "delegation.h"
24c8dbbb 59#include "internal.h"
40c64c26 60#include "nfs4idmap.h"
76e697ba 61#include "nfs4session.h"
974cec8c 62#include "pnfs.h"
bbe0a3aa 63#include "netns.h"
511ba52e 64#include "nfs4trace.h"
1da177e4 65
e3c0fb7e
CL
66#define NFSDBG_FACILITY NFSDBG_STATE
67
1da177e4
LT
68#define OPENOWNER_POOL_SIZE 8
69
40882deb
TM
70static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp);
71
93b717fd 72const nfs4_stateid zero_stateid = {
e0714ec4 73 { .data = { 0 } },
93b717fd
TM
74 .type = NFS4_SPECIAL_STATEID_TYPE,
75};
fcd8843c
TM
76const nfs4_stateid invalid_stateid = {
77 {
445f288d
TM
78 /* Funky initialiser keeps older gcc versions happy */
79 .data = { 0xff, 0xff, 0xff, 0xff, 0 },
fcd8843c
TM
80 },
81 .type = NFS4_INVALID_STATEID_TYPE,
82};
83
2409a976
FI
84const nfs4_stateid current_stateid = {
85 {
86 /* Funky initialiser keeps older gcc versions happy */
87 .data = { 0x0, 0x0, 0x0, 0x1, 0 },
88 },
89 .type = NFS4_SPECIAL_STATEID_TYPE,
90};
91
05f4c350 92static DEFINE_MUTEX(nfs_clid_init_mutex);
1da177e4 93
5b596830
DB
94static int nfs4_setup_state_renewal(struct nfs_client *clp)
95{
96 int status;
97 struct nfs_fsinfo fsinfo;
5b596830
DB
98
99 if (!test_bit(NFS_CS_CHECK_LEASE_TIME, &clp->cl_res_state)) {
100 nfs4_schedule_state_renewal(clp);
101 return 0;
102 }
103
5b596830
DB
104 status = nfs4_proc_get_lease_time(clp, &fsinfo);
105 if (status == 0) {
7dc2993a 106 nfs4_set_lease_period(clp, fsinfo.lease_time * HZ);
5b596830
DB
107 nfs4_schedule_state_renewal(clp);
108 }
109
110 return status;
111}
112
a52458b4 113int nfs4_init_clientid(struct nfs_client *clp, const struct cred *cred)
1da177e4 114{
fd954ae1
TM
115 struct nfs4_setclientid_res clid = {
116 .clientid = clp->cl_clientid,
117 .confirm = clp->cl_confirm,
118 };
f738f517
CL
119 unsigned short port;
120 int status;
bbe0a3aa 121 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
f738f517 122
fd954ae1
TM
123 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
124 goto do_confirm;
bbe0a3aa 125 port = nn->nfs_callback_tcpport;
f738f517 126 if (clp->cl_addr.ss_family == AF_INET6)
29dcc16a 127 port = nn->nfs_callback_tcpport6;
f738f517 128
bb8b27e5
TM
129 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
130 if (status != 0)
131 goto out;
fd954ae1
TM
132 clp->cl_clientid = clid.clientid;
133 clp->cl_confirm = clid.confirm;
134 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
135do_confirm:
bb8b27e5
TM
136 status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
137 if (status != 0)
138 goto out;
fd954ae1 139 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
5b596830 140 nfs4_setup_state_renewal(clp);
bb8b27e5 141out:
1da177e4
LT
142 return status;
143}
144
05f4c350
CL
145/**
146 * nfs40_discover_server_trunking - Detect server IP address trunking (mv0)
147 *
148 * @clp: nfs_client under test
149 * @result: OUT: found nfs_client, or clp
150 * @cred: credential to use for trunking test
151 *
152 * Returns zero, a negative errno, or a negative NFS4ERR status.
153 * If zero is returned, an nfs_client pointer is planted in
154 * "result".
155 *
156 * Note: The returned client may not yet be marked ready.
157 */
158int nfs40_discover_server_trunking(struct nfs_client *clp,
159 struct nfs_client **result,
a52458b4 160 const struct cred *cred)
05f4c350
CL
161{
162 struct nfs4_setclientid_res clid = {
163 .clientid = clp->cl_clientid,
164 .confirm = clp->cl_confirm,
165 };
9f62387d 166 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
05f4c350
CL
167 unsigned short port;
168 int status;
169
9f62387d 170 port = nn->nfs_callback_tcpport;
05f4c350 171 if (clp->cl_addr.ss_family == AF_INET6)
9f62387d 172 port = nn->nfs_callback_tcpport6;
05f4c350
CL
173
174 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
175 if (status != 0)
176 goto out;
177 clp->cl_clientid = clid.clientid;
178 clp->cl_confirm = clid.confirm;
179
180 status = nfs40_walk_client_list(clp, result, cred);
202c312d 181 if (status == 0) {
05f4c350
CL
182 /* Sustain the lease, even if it's empty. If the clientid4
183 * goes stale it's of no use for trunking discovery. */
184 nfs4_schedule_state_renewal(*result);
f02f3755
Z
185
186 /* If the client state need to recover, do it. */
187 if (clp->cl_state)
188 nfs4_schedule_state_manager(clp);
05f4c350 189 }
05f4c350
CL
190out:
191 return status;
192}
193
a52458b4 194const struct cred *nfs4_get_machine_cred(struct nfs_client *clp)
a2b2bb88 195{
a52458b4 196 return get_cred(rpc_machine_cred());
a2b2bb88
TM
197}
198
d688f7b8
CL
199static void nfs4_root_machine_cred(struct nfs_client *clp)
200{
d688f7b8 201
5e16923b
N
202 /* Force root creds instead of machine */
203 clp->cl_principal = NULL;
204 clp->cl_rpcclient->cl_principal = NULL;
d688f7b8
CL
205}
206
a52458b4 207static const struct cred *
24d292b8 208nfs4_get_renew_cred_server_locked(struct nfs_server *server)
b4454fe1 209{
a52458b4 210 const struct cred *cred = NULL;
b4454fe1 211 struct nfs4_state_owner *sp;
9f958ab8 212 struct rb_node *pos;
b4454fe1 213
24d292b8
CL
214 for (pos = rb_first(&server->state_owners);
215 pos != NULL;
216 pos = rb_next(pos)) {
217 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
b4454fe1
TM
218 if (list_empty(&sp->so_states))
219 continue;
a52458b4 220 cred = get_cred(sp->so_cred);
b4454fe1
TM
221 break;
222 }
223 return cred;
224}
225
24d292b8 226/**
f15e1e8b 227 * nfs4_get_renew_cred - Acquire credential for a renew operation
24d292b8
CL
228 * @clp: client state handle
229 *
230 * Returns an rpc_cred with reference count bumped, or NULL.
231 * Caller must hold clp->cl_lock.
232 */
a52458b4 233const struct cred *nfs4_get_renew_cred(struct nfs_client *clp)
24d292b8 234{
a52458b4 235 const struct cred *cred = NULL;
24d292b8
CL
236 struct nfs_server *server;
237
e49a29bd 238 /* Use machine credentials if available */
f15e1e8b 239 cred = nfs4_get_machine_cred(clp);
e49a29bd
SP
240 if (cred != NULL)
241 goto out;
242
f15e1e8b 243 spin_lock(&clp->cl_lock);
24d292b8
CL
244 rcu_read_lock();
245 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
246 cred = nfs4_get_renew_cred_server_locked(server);
247 if (cred != NULL)
248 break;
249 }
250 rcu_read_unlock();
f15e1e8b 251 spin_unlock(&clp->cl_lock);
e49a29bd
SP
252
253out:
24d292b8
CL
254 return cred;
255}
256
62f288a0 257static void nfs4_end_drain_slot_table(struct nfs4_slot_table *tbl)
9dfdf404 258{
774d5f14 259 if (test_and_clear_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state)) {
961a828d 260 spin_lock(&tbl->slot_tbl_lock);
b75ad4cd 261 nfs41_wake_slot_table(tbl);
961a828d 262 spin_unlock(&tbl->slot_tbl_lock);
689cf5c1 263 }
9dfdf404
RL
264}
265
62f288a0
AA
266static void nfs4_end_drain_session(struct nfs_client *clp)
267{
268 struct nfs4_session *ses = clp->cl_session;
269
2cf8bca8
CL
270 if (clp->cl_slot_tbl) {
271 nfs4_end_drain_slot_table(clp->cl_slot_tbl);
272 return;
273 }
274
62f288a0
AA
275 if (ses != NULL) {
276 nfs4_end_drain_slot_table(&ses->bc_slot_table);
277 nfs4_end_drain_slot_table(&ses->fc_slot_table);
278 }
279}
280
774d5f14 281static int nfs4_drain_slot_tbl(struct nfs4_slot_table *tbl)
9dfdf404 282{
774d5f14 283 set_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state);
9dfdf404 284 spin_lock(&tbl->slot_tbl_lock);
b6bf6e7d 285 if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
16735d02 286 reinit_completion(&tbl->complete);
9dfdf404 287 spin_unlock(&tbl->slot_tbl_lock);
42acd021 288 return wait_for_completion_interruptible(&tbl->complete);
9dfdf404
RL
289 }
290 spin_unlock(&tbl->slot_tbl_lock);
291 return 0;
292}
293
42acd021
AA
294static int nfs4_begin_drain_session(struct nfs_client *clp)
295{
296 struct nfs4_session *ses = clp->cl_session;
8aafd2fd 297 int ret;
42acd021 298
2cf8bca8
CL
299 if (clp->cl_slot_tbl)
300 return nfs4_drain_slot_tbl(clp->cl_slot_tbl);
301
42acd021 302 /* back channel */
774d5f14 303 ret = nfs4_drain_slot_tbl(&ses->bc_slot_table);
42acd021
AA
304 if (ret)
305 return ret;
306 /* fore channel */
774d5f14 307 return nfs4_drain_slot_tbl(&ses->fc_slot_table);
42acd021
AA
308}
309
c9fdeb28
CL
310#if defined(CONFIG_NFS_V4_1)
311
bda197f5
TM
312static void nfs41_finish_session_reset(struct nfs_client *clp)
313{
314 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
315 clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
316 /* create_session negotiated new slot table */
bda197f5 317 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
ea51efaa 318 nfs4_setup_state_renewal(clp);
bda197f5
TM
319}
320
a52458b4 321int nfs41_init_clientid(struct nfs_client *clp, const struct cred *cred)
4d643d1d
AA
322{
323 int status;
324
fd954ae1
TM
325 if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
326 goto do_confirm;
4d643d1d 327 status = nfs4_proc_exchange_id(clp, cred);
9430fb6b
RL
328 if (status != 0)
329 goto out;
fd954ae1
TM
330 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
331do_confirm:
848f5bda 332 status = nfs4_proc_create_session(clp, cred);
9430fb6b
RL
333 if (status != 0)
334 goto out;
40882deb
TM
335 if (!(clp->cl_exchange_flags & EXCHGID4_FLAG_CONFIRMED_R))
336 nfs4_state_start_reclaim_reboot(clp);
bda197f5 337 nfs41_finish_session_reset(clp);
9430fb6b
RL
338 nfs_mark_client_ready(clp, NFS_CS_READY);
339out:
4d643d1d
AA
340 return status;
341}
342
05f4c350
CL
343/**
344 * nfs41_discover_server_trunking - Detect server IP address trunking (mv1)
345 *
346 * @clp: nfs_client under test
347 * @result: OUT: found nfs_client, or clp
348 * @cred: credential to use for trunking test
349 *
350 * Returns NFS4_OK, a negative errno, or a negative NFS4ERR status.
351 * If NFS4_OK is returned, an nfs_client pointer is planted in
352 * "result".
353 *
354 * Note: The returned client may not yet be marked ready.
355 */
356int nfs41_discover_server_trunking(struct nfs_client *clp,
357 struct nfs_client **result,
a52458b4 358 const struct cred *cred)
05f4c350
CL
359{
360 int status;
361
362 status = nfs4_proc_exchange_id(clp, cred);
363 if (status != NFS4_OK)
364 return status;
365
48d66b97
TM
366 status = nfs41_walk_client_list(clp, result, cred);
367 if (status < 0)
368 return status;
369 if (clp != *result)
370 return 0;
371
8dcbec6d
CL
372 /*
373 * Purge state if the client id was established in a prior
374 * instance and the client id could not have arrived on the
375 * server via Transparent State Migration.
376 */
377 if (clp->cl_exchange_flags & EXCHGID4_FLAG_CONFIRMED_R) {
378 if (!test_bit(NFS_CS_TSM_POSSIBLE, &clp->cl_flags))
379 set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
380 else
381 set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
382 }
48d66b97
TM
383 nfs4_schedule_state_manager(clp);
384 status = nfs_wait_client_init_complete(clp);
385 if (status < 0)
386 nfs_put_client(clp);
387 return status;
05f4c350
CL
388}
389
b4b82607
AA
390#endif /* CONFIG_NFS_V4_1 */
391
24d292b8 392/**
73d8bde5 393 * nfs4_get_clid_cred - Acquire credential for a setclientid operation
24d292b8
CL
394 * @clp: client state handle
395 *
a52458b4 396 * Returns a cred with reference count bumped, or NULL.
24d292b8 397 */
a52458b4 398const struct cred *nfs4_get_clid_cred(struct nfs_client *clp)
24d292b8 399{
a52458b4 400 const struct cred *cred;
286d7d6a 401
f15e1e8b 402 cred = nfs4_get_machine_cred(clp);
a2b2bb88 403 return cred;
286d7d6a
TM
404}
405
1da177e4 406static struct nfs4_state_owner *
a52458b4 407nfs4_find_state_owner_locked(struct nfs_server *server, const struct cred *cred)
1da177e4 408{
24d292b8 409 struct rb_node **p = &server->state_owners.rb_node,
9f958ab8 410 *parent = NULL;
414adf14 411 struct nfs4_state_owner *sp;
a52458b4 412 int cmp;
1da177e4 413
9f958ab8
TM
414 while (*p != NULL) {
415 parent = *p;
24d292b8 416 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
a52458b4 417 cmp = cred_fscmp(cred, sp->so_cred);
9f958ab8 418
a52458b4 419 if (cmp < 0)
9f958ab8 420 p = &parent->rb_left;
a52458b4 421 else if (cmp > 0)
9f958ab8
TM
422 p = &parent->rb_right;
423 else {
0aaaf5c4
CL
424 if (!list_empty(&sp->so_lru))
425 list_del_init(&sp->so_lru);
9f958ab8 426 atomic_inc(&sp->so_count);
414adf14 427 return sp;
9f958ab8 428 }
1da177e4 429 }
414adf14 430 return NULL;
1da177e4
LT
431}
432
9f958ab8 433static struct nfs4_state_owner *
24d292b8 434nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
9f958ab8 435{
24d292b8
CL
436 struct nfs_server *server = new->so_server;
437 struct rb_node **p = &server->state_owners.rb_node,
9f958ab8
TM
438 *parent = NULL;
439 struct nfs4_state_owner *sp;
a52458b4 440 int cmp;
9f958ab8
TM
441
442 while (*p != NULL) {
443 parent = *p;
24d292b8 444 sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
a52458b4 445 cmp = cred_fscmp(new->so_cred, sp->so_cred);
9f958ab8 446
a52458b4 447 if (cmp < 0)
9f958ab8 448 p = &parent->rb_left;
a52458b4 449 else if (cmp > 0)
9f958ab8
TM
450 p = &parent->rb_right;
451 else {
0aaaf5c4
CL
452 if (!list_empty(&sp->so_lru))
453 list_del_init(&sp->so_lru);
9f958ab8
TM
454 atomic_inc(&sp->so_count);
455 return sp;
456 }
457 }
24d292b8
CL
458 rb_link_node(&new->so_server_node, parent, p);
459 rb_insert_color(&new->so_server_node, &server->state_owners);
9f958ab8
TM
460 return new;
461}
462
463static void
24d292b8 464nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
9f958ab8 465{
24d292b8
CL
466 struct nfs_server *server = sp->so_server;
467
468 if (!RB_EMPTY_NODE(&sp->so_server_node))
469 rb_erase(&sp->so_server_node, &server->state_owners);
9f958ab8
TM
470}
471
7ba127ab
TM
472static void
473nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
474{
95b72eb0 475 sc->create_time = ktime_get();
7ba127ab
TM
476 sc->flags = 0;
477 sc->counter = 0;
478 spin_lock_init(&sc->lock);
479 INIT_LIST_HEAD(&sc->list);
480 rpc_init_wait_queue(&sc->wait, "Seqid_waitqueue");
481}
482
483static void
484nfs4_destroy_seqid_counter(struct nfs_seqid_counter *sc)
485{
486 rpc_destroy_wait_queue(&sc->wait);
487}
488
1da177e4
LT
489/*
490 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
491 * create a new state_owner.
492 *
493 */
494static struct nfs4_state_owner *
d1e284d5 495nfs4_alloc_state_owner(struct nfs_server *server,
a52458b4 496 const struct cred *cred,
d1e284d5 497 gfp_t gfp_flags)
1da177e4
LT
498{
499 struct nfs4_state_owner *sp;
500
d1e284d5 501 sp = kzalloc(sizeof(*sp), gfp_flags);
1da177e4
LT
502 if (!sp)
503 return NULL;
724e2df9 504 sp->so_seqid.owner_id = ida_alloc(&server->openowner_id, gfp_flags);
aae5730e
MW
505 if (sp->so_seqid.owner_id < 0) {
506 kfree(sp);
507 return NULL;
508 }
d1e284d5 509 sp->so_server = server;
a52458b4 510 sp->so_cred = get_cred(cred);
ec073428 511 spin_lock_init(&sp->so_lock);
1da177e4 512 INIT_LIST_HEAD(&sp->so_states);
7ba127ab 513 nfs4_init_seqid_counter(&sp->so_seqid);
1da177e4 514 atomic_set(&sp->so_count, 1);
0aaaf5c4 515 INIT_LIST_HEAD(&sp->so_lru);
76246c92 516 seqcount_spinlock_init(&sp->so_reclaim_seqcount, &sp->so_lock);
65b62a29 517 mutex_init(&sp->so_delegreturn_mutex);
1da177e4
LT
518 return sp;
519}
520
1d2e88e7 521static void
86cfb041 522nfs4_reset_state_owner(struct nfs4_state_owner *sp)
1da177e4 523{
86cfb041
N
524 /* This state_owner is no longer usable, but must
525 * remain in place so that state recovery can find it
526 * and the opens associated with it.
527 * It may also be used for new 'open' request to
528 * return a delegation to the server.
529 * So update the 'create_time' so that it looks like
530 * a new state_owner. This will cause the server to
531 * request an OPEN_CONFIRM to start a new sequence.
532 */
533 sp->so_seqid.create_time = ktime_get();
1da177e4
LT
534}
535
0aaaf5c4
CL
536static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
537{
7ba127ab 538 nfs4_destroy_seqid_counter(&sp->so_seqid);
a52458b4 539 put_cred(sp->so_cred);
724e2df9 540 ida_free(&sp->so_server->openowner_id, sp->so_seqid.owner_id);
0aaaf5c4
CL
541 kfree(sp);
542}
543
544static void nfs4_gc_state_owners(struct nfs_server *server)
545{
546 struct nfs_client *clp = server->nfs_client;
547 struct nfs4_state_owner *sp, *tmp;
548 unsigned long time_min, time_max;
549 LIST_HEAD(doomed);
550
551 spin_lock(&clp->cl_lock);
552 time_max = jiffies;
553 time_min = (long)time_max - (long)clp->cl_lease_time;
554 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
555 /* NB: LRU is sorted so that oldest is at the head */
556 if (time_in_range(sp->so_expires, time_min, time_max))
557 break;
558 list_move(&sp->so_lru, &doomed);
559 nfs4_remove_state_owner_locked(sp);
560 }
561 spin_unlock(&clp->cl_lock);
562
563 list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
564 list_del(&sp->so_lru);
565 nfs4_free_state_owner(sp);
566 }
567}
568
24d292b8
CL
569/**
570 * nfs4_get_state_owner - Look up a state owner given a credential
571 * @server: nfs_server to search
572 * @cred: RPC credential to match
302fad7b 573 * @gfp_flags: allocation mode
24d292b8
CL
574 *
575 * Returns a pointer to an instantiated nfs4_state_owner struct, or NULL.
576 */
577struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server,
a52458b4 578 const struct cred *cred,
d1e284d5 579 gfp_t gfp_flags)
1da177e4 580{
7539bbab 581 struct nfs_client *clp = server->nfs_client;
1da177e4
LT
582 struct nfs4_state_owner *sp, *new;
583
1da177e4 584 spin_lock(&clp->cl_lock);
24d292b8 585 sp = nfs4_find_state_owner_locked(server, cred);
1da177e4 586 spin_unlock(&clp->cl_lock);
1da177e4 587 if (sp != NULL)
0aaaf5c4 588 goto out;
d1e284d5 589 new = nfs4_alloc_state_owner(server, cred, gfp_flags);
9f958ab8 590 if (new == NULL)
0aaaf5c4 591 goto out;
aae5730e
MW
592 spin_lock(&clp->cl_lock);
593 sp = nfs4_insert_state_owner_locked(new);
594 spin_unlock(&clp->cl_lock);
d1e284d5
TM
595 if (sp != new)
596 nfs4_free_state_owner(new);
0aaaf5c4
CL
597out:
598 nfs4_gc_state_owners(server);
9f958ab8 599 return sp;
1da177e4
LT
600}
601
24d292b8
CL
602/**
603 * nfs4_put_state_owner - Release a nfs4_state_owner
604 * @sp: state owner data to release
7bf97bc2
TM
605 *
606 * Note that we keep released state owners on an LRU
607 * list.
608 * This caches valid state owners so that they can be
609 * reused, to avoid the OPEN_CONFIRM on minor version 0.
610 * It also pins the uniquifier of dropped state owners for
611 * a while, to ensure that those state owner names are
612 * never reused.
24d292b8 613 */
1da177e4
LT
614void nfs4_put_state_owner(struct nfs4_state_owner *sp)
615{
0aaaf5c4
CL
616 struct nfs_server *server = sp->so_server;
617 struct nfs_client *clp = server->nfs_client;
1da177e4
LT
618
619 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
620 return;
0aaaf5c4 621
7bf97bc2
TM
622 sp->so_expires = jiffies;
623 list_add_tail(&sp->so_lru, &server->state_owners_lru);
624 spin_unlock(&clp->cl_lock);
0aaaf5c4
CL
625}
626
627/**
628 * nfs4_purge_state_owners - Release all cached state owners
629 * @server: nfs_server with cached state owners to release
c77e2283 630 * @head: resulting list of state owners
0aaaf5c4
CL
631 *
632 * Called at umount time. Remaining state owners will be on
633 * the LRU with ref count of zero.
c77e2283
TM
634 * Note that the state owners are not freed, but are added
635 * to the list @head, which can later be used as an argument
636 * to nfs4_free_state_owners.
0aaaf5c4 637 */
c77e2283 638void nfs4_purge_state_owners(struct nfs_server *server, struct list_head *head)
0aaaf5c4
CL
639{
640 struct nfs_client *clp = server->nfs_client;
641 struct nfs4_state_owner *sp, *tmp;
0aaaf5c4
CL
642
643 spin_lock(&clp->cl_lock);
644 list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
c77e2283 645 list_move(&sp->so_lru, head);
0aaaf5c4
CL
646 nfs4_remove_state_owner_locked(sp);
647 }
1da177e4 648 spin_unlock(&clp->cl_lock);
c77e2283 649}
0aaaf5c4 650
c77e2283 651/**
6453bcd0 652 * nfs4_free_state_owners - Release all cached state owners
c77e2283
TM
653 * @head: resulting list of state owners
654 *
655 * Frees a list of state owners that was generated by
656 * nfs4_purge_state_owners
657 */
658void nfs4_free_state_owners(struct list_head *head)
659{
660 struct nfs4_state_owner *sp, *tmp;
661
662 list_for_each_entry_safe(sp, tmp, head, so_lru) {
0aaaf5c4
CL
663 list_del(&sp->so_lru);
664 nfs4_free_state_owner(sp);
665 }
1da177e4
LT
666}
667
668static struct nfs4_state *
669nfs4_alloc_open_state(void)
670{
671 struct nfs4_state *state;
672
9c00fd9a 673 state = kzalloc(sizeof(*state), GFP_KERNEL_ACCOUNT);
1da177e4
LT
674 if (!state)
675 return NULL;
ace9fad4 676 refcount_set(&state->count, 1);
1da177e4 677 INIT_LIST_HEAD(&state->lock_states);
8d0a8a9d 678 spin_lock_init(&state->state_lock);
8bda4e4c 679 seqlock_init(&state->seqlock);
c9399f21 680 init_waitqueue_head(&state->waitq);
1da177e4
LT
681 return state;
682}
683
4cecb76f 684void
dc0b027d 685nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
4cecb76f 686{
dc0b027d 687 if (state->state == fmode)
4cecb76f
TM
688 return;
689 /* NB! List reordering - see the reclaim code for why. */
dc0b027d
TM
690 if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
691 if (fmode & FMODE_WRITE)
4cecb76f
TM
692 list_move(&state->open_states, &state->owner->so_states);
693 else
694 list_move_tail(&state->open_states, &state->owner->so_states);
695 }
dc0b027d 696 state->state = fmode;
4cecb76f
TM
697}
698
1da177e4
LT
699static struct nfs4_state *
700__nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
701{
702 struct nfs_inode *nfsi = NFS_I(inode);
703 struct nfs4_state *state;
704
9ae075fd 705 list_for_each_entry_rcu(state, &nfsi->open_states, inode_states) {
1c816efa 706 if (state->owner != owner)
1da177e4 707 continue;
5d422301
TM
708 if (!nfs4_valid_open_stateid(state))
709 continue;
ace9fad4 710 if (refcount_inc_not_zero(&state->count))
1da177e4 711 return state;
1da177e4
LT
712 }
713 return NULL;
714}
715
1da177e4
LT
716static void
717nfs4_free_open_state(struct nfs4_state *state)
718{
9ae075fd 719 kfree_rcu(state, rcu_head);
1da177e4
LT
720}
721
722struct nfs4_state *
723nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
724{
725 struct nfs4_state *state, *new;
726 struct nfs_inode *nfsi = NFS_I(inode);
727
9ae075fd 728 rcu_read_lock();
1da177e4 729 state = __nfs4_find_state_byowner(inode, owner);
9ae075fd 730 rcu_read_unlock();
1da177e4
LT
731 if (state)
732 goto out;
733 new = nfs4_alloc_open_state();
ec073428 734 spin_lock(&owner->so_lock);
1da177e4
LT
735 spin_lock(&inode->i_lock);
736 state = __nfs4_find_state_byowner(inode, owner);
737 if (state == NULL && new != NULL) {
738 state = new;
1da177e4
LT
739 state->owner = owner;
740 atomic_inc(&owner->so_count);
0444d76a
DC
741 ihold(inode);
742 state->inode = inode;
29fe8399 743 list_add_rcu(&state->inode_states, &nfsi->open_states);
1da177e4 744 spin_unlock(&inode->i_lock);
ec073428
TM
745 /* Note: The reclaim code dictates that we add stateless
746 * and read-only stateids to the end of the list */
747 list_add_tail(&state->open_states, &owner->so_states);
748 spin_unlock(&owner->so_lock);
1da177e4
LT
749 } else {
750 spin_unlock(&inode->i_lock);
ec073428 751 spin_unlock(&owner->so_lock);
1da177e4
LT
752 if (new)
753 nfs4_free_open_state(new);
754 }
755out:
756 return state;
757}
758
1da177e4
LT
759void nfs4_put_open_state(struct nfs4_state *state)
760{
761 struct inode *inode = state->inode;
762 struct nfs4_state_owner *owner = state->owner;
763
ace9fad4 764 if (!refcount_dec_and_lock(&state->count, &owner->so_lock))
1da177e4 765 return;
ec073428 766 spin_lock(&inode->i_lock);
9ae075fd 767 list_del_rcu(&state->inode_states);
1da177e4 768 list_del(&state->open_states);
ec073428
TM
769 spin_unlock(&inode->i_lock);
770 spin_unlock(&owner->so_lock);
b7b7dac6 771 nfs4_inode_return_delegation_on_close(inode);
1da177e4 772 iput(inode);
1da177e4
LT
773 nfs4_free_open_state(state);
774 nfs4_put_state_owner(owner);
775}
776
777/*
83c9d41e 778 * Close the current file.
1da177e4 779 */
643168c2 780static void __nfs4_close(struct nfs4_state *state,
8535b2be 781 fmode_t fmode, gfp_t gfp_mask, int wait)
1da177e4 782{
1da177e4 783 struct nfs4_state_owner *owner = state->owner;
003707c7 784 int call_close = 0;
dc0b027d 785 fmode_t newstate;
1da177e4
LT
786
787 atomic_inc(&owner->so_count);
1da177e4 788 /* Protect against nfs4_find_state() */
ec073428 789 spin_lock(&owner->so_lock);
dc0b027d 790 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
e7616923
TM
791 case FMODE_READ:
792 state->n_rdonly--;
793 break;
794 case FMODE_WRITE:
795 state->n_wronly--;
796 break;
797 case FMODE_READ|FMODE_WRITE:
798 state->n_rdwr--;
799 }
003707c7 800 newstate = FMODE_READ|FMODE_WRITE;
e7616923 801 if (state->n_rdwr == 0) {
003707c7 802 if (state->n_rdonly == 0) {
e7616923 803 newstate &= ~FMODE_READ;
003707c7
TM
804 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
805 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
806 }
807 if (state->n_wronly == 0) {
e7616923 808 newstate &= ~FMODE_WRITE;
003707c7
TM
809 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
810 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
811 }
812 if (newstate == 0)
813 clear_bit(NFS_DELEGATED_STATE, &state->flags);
e7616923 814 }
003707c7 815 nfs4_state_set_mode_locked(state, newstate);
ec073428 816 spin_unlock(&owner->so_lock);
4cecb76f 817
003707c7 818 if (!call_close) {
b39e625b
TM
819 nfs4_put_open_state(state);
820 nfs4_put_state_owner(owner);
1f7977c1
TM
821 } else
822 nfs4_do_close(state, gfp_mask, wait);
a49c3c77
TM
823}
824
643168c2 825void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
a49c3c77 826{
da48f267 827 __nfs4_close(state, fmode, GFP_KERNEL, 0);
a49c3c77
TM
828}
829
643168c2 830void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
a49c3c77 831{
643168c2 832 __nfs4_close(state, fmode, GFP_KERNEL, 1);
1da177e4
LT
833}
834
835/*
836 * Search the state->lock_states for an existing lock_owner
3f8f2548
N
837 * that is compatible with either of the given owners.
838 * If the second is non-zero, then the first refers to a Posix-lock
839 * owner (current->files) and the second refers to a flock/OFD
840 * owner (struct file*). In that case, prefer a match for the first
841 * owner.
842 * If both sorts of locks are held on the one file we cannot know
843 * which stateid was intended to be used, so a "correct" choice cannot
844 * be made. Failing that, a "consistent" choice is preferable. The
845 * consistent choice we make is to prefer the first owner, that of a
846 * Posix lock.
1da177e4
LT
847 */
848static struct nfs4_lock_state *
8d424431
N
849__nfs4_find_lock_state(struct nfs4_state *state,
850 fl_owner_t fl_owner, fl_owner_t fl_owner2)
1da177e4 851{
3f8f2548 852 struct nfs4_lock_state *pos, *ret = NULL;
1da177e4 853 list_for_each_entry(pos, &state->lock_states, ls_locks) {
3f8f2548
N
854 if (pos->ls_owner == fl_owner) {
855 ret = pos;
856 break;
857 }
858 if (pos->ls_owner == fl_owner2)
859 ret = pos;
1da177e4 860 }
3f8f2548 861 if (ret)
194bc1f4 862 refcount_inc(&ret->ls_count);
3f8f2548 863 return ret;
1da177e4
LT
864}
865
1da177e4
LT
866/*
867 * Return a compatible lock_state. If no initialized lock_state structure
868 * exists, return an uninitialized one.
869 *
1da177e4 870 */
8003d3c4 871static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
1da177e4
LT
872{
873 struct nfs4_lock_state *lsp;
24d292b8 874 struct nfs_server *server = state->owner->so_server;
1da177e4 875
9c00fd9a 876 lsp = kzalloc(sizeof(*lsp), GFP_KERNEL_ACCOUNT);
1da177e4
LT
877 if (lsp == NULL)
878 return NULL;
7ba127ab 879 nfs4_init_seqid_counter(&lsp->ls_seqid);
194bc1f4 880 refcount_set(&lsp->ls_count, 1);
b64aec8d 881 lsp->ls_state = state;
8003d3c4 882 lsp->ls_owner = fl_owner;
724e2df9 883 lsp->ls_seqid.owner_id = ida_alloc(&server->lockowner_id, GFP_KERNEL_ACCOUNT);
48c22eb2 884 if (lsp->ls_seqid.owner_id < 0)
d2d7ce28 885 goto out_free;
8d0a8a9d 886 INIT_LIST_HEAD(&lsp->ls_locks);
1da177e4 887 return lsp;
d2d7ce28
TM
888out_free:
889 kfree(lsp);
890 return NULL;
1da177e4
LT
891}
892
5ae67c4f 893void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9f958ab8 894{
724e2df9 895 ida_free(&server->lockowner_id, lsp->ls_seqid.owner_id);
7ba127ab 896 nfs4_destroy_seqid_counter(&lsp->ls_seqid);
9f958ab8
TM
897 kfree(lsp);
898}
899
1da177e4
LT
900/*
901 * Return a compatible lock_state. If no initialized lock_state structure
902 * exists, return an uninitialized one.
903 *
1da177e4 904 */
8003d3c4 905static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
1da177e4 906{
8d0a8a9d 907 struct nfs4_lock_state *lsp, *new = NULL;
1da177e4 908
8d0a8a9d
TM
909 for(;;) {
910 spin_lock(&state->state_lock);
68e33bd6 911 lsp = __nfs4_find_lock_state(state, owner, NULL);
8d0a8a9d
TM
912 if (lsp != NULL)
913 break;
914 if (new != NULL) {
8d0a8a9d
TM
915 list_add(&new->ls_locks, &state->lock_states);
916 set_bit(LK_STATE_IN_USE, &state->flags);
917 lsp = new;
918 new = NULL;
919 break;
920 }
921 spin_unlock(&state->state_lock);
8003d3c4 922 new = nfs4_alloc_lock_state(state, owner);
8d0a8a9d
TM
923 if (new == NULL)
924 return NULL;
925 }
926 spin_unlock(&state->state_lock);
9f958ab8 927 if (new != NULL)
5ae67c4f 928 nfs4_free_lock_state(state->owner->so_server, new);
1da177e4
LT
929 return lsp;
930}
931
932/*
8d0a8a9d
TM
933 * Release reference to lock_state, and free it if we see that
934 * it is no longer in use
1da177e4 935 */
faf5f49c 936void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
1da177e4 937{
c8b2d0bf 938 struct nfs_server *server;
8d0a8a9d 939 struct nfs4_state *state;
1da177e4 940
8d0a8a9d
TM
941 if (lsp == NULL)
942 return;
943 state = lsp->ls_state;
194bc1f4 944 if (!refcount_dec_and_lock(&lsp->ls_count, &state->state_lock))
8d0a8a9d
TM
945 return;
946 list_del(&lsp->ls_locks);
947 if (list_empty(&state->lock_states))
948 clear_bit(LK_STATE_IN_USE, &state->flags);
949 spin_unlock(&state->state_lock);
0c0e0d3c
JL
950 server = state->owner->so_server;
951 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
952 struct nfs_client *clp = server->nfs_client;
953
954 clp->cl_mvops->free_lock_state(server, lsp);
955 } else
c8b2d0bf 956 nfs4_free_lock_state(server, lsp);
1da177e4
LT
957}
958
8d0a8a9d 959static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
1da177e4 960{
8d0a8a9d 961 struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
1da177e4 962
8d0a8a9d 963 dst->fl_u.nfs4_fl.owner = lsp;
194bc1f4 964 refcount_inc(&lsp->ls_count);
8d0a8a9d 965}
1da177e4 966
8d0a8a9d 967static void nfs4_fl_release_lock(struct file_lock *fl)
1da177e4 968{
8d0a8a9d 969 nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
1da177e4
LT
970}
971
6aed6285 972static const struct file_lock_operations nfs4_fl_lock_ops = {
8d0a8a9d
TM
973 .fl_copy_lock = nfs4_fl_copy_lock,
974 .fl_release_private = nfs4_fl_release_lock,
975};
976
977int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
1da177e4 978{
8d0a8a9d
TM
979 struct nfs4_lock_state *lsp;
980
981 if (fl->fl_ops != NULL)
982 return 0;
8003d3c4 983 lsp = nfs4_get_lock_state(state, fl->fl_owner);
8d0a8a9d
TM
984 if (lsp == NULL)
985 return -ENOMEM;
986 fl->fl_u.nfs4_fl.owner = lsp;
987 fl->fl_ops = &nfs4_fl_lock_ops;
988 return 0;
1da177e4
LT
989}
990
5521abfd
TM
991static int nfs4_copy_lock_stateid(nfs4_stateid *dst,
992 struct nfs4_state *state,
17393475 993 const struct nfs_lock_context *l_ctx)
1da177e4 994{
8d0a8a9d 995 struct nfs4_lock_state *lsp;
8d424431 996 fl_owner_t fl_owner, fl_flock_owner;
5521abfd 997 int ret = -ENOENT;
1da177e4 998
17393475 999 if (l_ctx == NULL)
2a369153
TM
1000 goto out;
1001
8d0a8a9d 1002 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
4fc8796d 1003 goto out;
1da177e4 1004
d51fdb87 1005 fl_owner = l_ctx->lockowner;
8d424431
N
1006 fl_flock_owner = l_ctx->open_context->flock_owner;
1007
8d0a8a9d 1008 spin_lock(&state->state_lock);
8d424431 1009 lsp = __nfs4_find_lock_state(state, fl_owner, fl_flock_owner);
ef1820f9
N
1010 if (lsp && test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
1011 ret = -EIO;
1012 else if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
f597c537 1013 nfs4_stateid_copy(dst, &lsp->ls_stateid);
5521abfd 1014 ret = 0;
4fc8796d 1015 }
8d0a8a9d 1016 spin_unlock(&state->state_lock);
1da177e4 1017 nfs4_put_lock_state(lsp);
4fc8796d
TM
1018out:
1019 return ret;
1020}
1021
c82bac6f 1022bool nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
4fc8796d 1023{
c82bac6f 1024 bool ret;
92b40e93 1025 const nfs4_stateid *src;
4fc8796d
TM
1026 int seq;
1027
1028 do {
c82bac6f 1029 ret = false;
92b40e93 1030 src = &zero_stateid;
4fc8796d 1031 seq = read_seqbegin(&state->seqlock);
c82bac6f 1032 if (test_bit(NFS_OPEN_STATE, &state->flags)) {
92b40e93 1033 src = &state->open_stateid;
c82bac6f
TM
1034 ret = true;
1035 }
92b40e93 1036 nfs4_stateid_copy(dst, src);
4fc8796d 1037 } while (read_seqretry(&state->seqlock, seq));
c82bac6f 1038 return ret;
4fc8796d
TM
1039}
1040
1041/*
1042 * Byte-range lock aware utility to initialize the stateid of read/write
1043 * requests.
1044 */
abf4e13c 1045int nfs4_select_rw_stateid(struct nfs4_state *state,
17393475 1046 fmode_t fmode, const struct nfs_lock_context *l_ctx,
a52458b4 1047 nfs4_stateid *dst, const struct cred **cred)
4fc8796d 1048{
abf4e13c
TM
1049 int ret;
1050
7ebeb7fe
TM
1051 if (!nfs4_valid_open_stateid(state))
1052 return -EIO;
abf4e13c
TM
1053 if (cred != NULL)
1054 *cred = NULL;
17393475 1055 ret = nfs4_copy_lock_stateid(dst, state, l_ctx);
ef1820f9
N
1056 if (ret == -EIO)
1057 /* A lost lock - don't even consider delegations */
1058 goto out;
146d70ca 1059 /* returns true if delegation stateid found and copied */
abf4e13c 1060 if (nfs4_copy_delegation_stateid(state->inode, fmode, dst, cred)) {
146d70ca 1061 ret = 0;
5521abfd 1062 goto out;
146d70ca 1063 }
5521abfd 1064 if (ret != -ENOENT)
ef1820f9
N
1065 /* nfs4_copy_delegation_stateid() didn't over-write
1066 * dst, so it still has the lock stateid which we now
1067 * choose to use.
1068 */
5521abfd 1069 goto out;
d9aba2b4 1070 ret = nfs4_copy_open_stateid(dst, state) ? 0 : -EAGAIN;
5521abfd 1071out:
3b66486c
TM
1072 if (nfs_server_capable(state->inode, NFS_CAP_STATEID_NFSV41))
1073 dst->seqid = 0;
5521abfd 1074 return ret;
1da177e4
LT
1075}
1076
8535b2be 1077struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
cee54fc9 1078{
cee54fc9
TM
1079 struct nfs_seqid *new;
1080
8535b2be 1081 new = kmalloc(sizeof(*new), gfp_mask);
badc76dd
TM
1082 if (new == NULL)
1083 return ERR_PTR(-ENOMEM);
1084 new->sequence = counter;
1085 INIT_LIST_HEAD(&new->list);
1086 new->task = NULL;
cee54fc9
TM
1087 return new;
1088}
1089
72211dbe 1090void nfs_release_seqid(struct nfs_seqid *seqid)
1da177e4 1091{
4601df20 1092 struct nfs_seqid_counter *sequence;
cee54fc9 1093
a6796419 1094 if (seqid == NULL || list_empty(&seqid->list))
4601df20
TM
1095 return;
1096 sequence = seqid->sequence;
1097 spin_lock(&sequence->lock);
1098 list_del_init(&seqid->list);
1099 if (!list_empty(&sequence->list)) {
1100 struct nfs_seqid *next;
1101
1102 next = list_first_entry(&sequence->list,
1103 struct nfs_seqid, list);
1104 rpc_wake_up_queued_task(&sequence->wait, next->task);
2f74c0a0 1105 }
4601df20 1106 spin_unlock(&sequence->lock);
72211dbe
TM
1107}
1108
1109void nfs_free_seqid(struct nfs_seqid *seqid)
1110{
1111 nfs_release_seqid(seqid);
cee54fc9 1112 kfree(seqid);
1da177e4
LT
1113}
1114
1115/*
cee54fc9
TM
1116 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
1117 * failed with a seqid incrementing error -
16a6ddc7 1118 * see comments nfs4.h:seqid_mutating_error()
cee54fc9 1119 */
88d90939 1120static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
cee54fc9
TM
1121{
1122 switch (status) {
1123 case 0:
1124 break;
1125 case -NFS4ERR_BAD_SEQID:
6f43ddcc
TM
1126 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
1127 return;
9a3ba432 1128 pr_warn_ratelimited("NFS: v4 server returned a bad"
497799e7
DM
1129 " sequence-id error on an"
1130 " unconfirmed sequence %p!\n",
6f43ddcc 1131 seqid->sequence);
ffb81717 1132 return;
cee54fc9
TM
1133 case -NFS4ERR_STALE_CLIENTID:
1134 case -NFS4ERR_STALE_STATEID:
1135 case -NFS4ERR_BAD_STATEID:
1136 case -NFS4ERR_BADXDR:
1137 case -NFS4ERR_RESOURCE:
1138 case -NFS4ERR_NOFILEHANDLE:
406dab84 1139 case -NFS4ERR_MOVED:
cee54fc9
TM
1140 /* Non-seqid mutating errors */
1141 return;
8b98a532 1142 }
cee54fc9
TM
1143 /*
1144 * Note: no locking needed as we are guaranteed to be first
1145 * on the sequence list
1146 */
1147 seqid->sequence->counter++;
1148}
1149
1150void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
1151{
a6796419
TM
1152 struct nfs4_state_owner *sp;
1153
1154 if (seqid == NULL)
1155 return;
34dc1ad7 1156
a6796419 1157 sp = container_of(seqid->sequence, struct nfs4_state_owner, so_seqid);
34dc1ad7 1158 if (status == -NFS4ERR_BAD_SEQID)
86cfb041 1159 nfs4_reset_state_owner(sp);
a6796419 1160 if (!nfs4_has_session(sp->so_server->nfs_client))
34dc1ad7 1161 nfs_increment_seqid(status, seqid);
cee54fc9
TM
1162}
1163
1164/*
cee54fc9
TM
1165 * Increment the seqid if the LOCK/LOCKU succeeded, or
1166 * failed with a seqid incrementing error -
16a6ddc7 1167 * see comments nfs4.h:seqid_mutating_error()
cee54fc9
TM
1168 */
1169void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
1170{
a6796419
TM
1171 if (seqid != NULL)
1172 nfs_increment_seqid(status, seqid);
cee54fc9
TM
1173}
1174
1175int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
1176{
a6796419 1177 struct nfs_seqid_counter *sequence;
cee54fc9
TM
1178 int status = 0;
1179
a6796419
TM
1180 if (seqid == NULL)
1181 goto out;
1182 sequence = seqid->sequence;
cee54fc9 1183 spin_lock(&sequence->lock);
4601df20 1184 seqid->task = task;
2f74c0a0
TM
1185 if (list_empty(&seqid->list))
1186 list_add_tail(&seqid->list, &sequence->list);
1187 if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
1188 goto unlock;
5d00837b 1189 rpc_sleep_on(&sequence->wait, task, NULL);
2f74c0a0
TM
1190 status = -EAGAIN;
1191unlock:
cee54fc9 1192 spin_unlock(&sequence->lock);
a6796419 1193out:
cee54fc9 1194 return status;
1da177e4
LT
1195}
1196
e005e804 1197static int nfs4_run_state_manager(void *);
1da177e4 1198
e005e804 1199static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
433fbe4c 1200{
43d20e80 1201 clear_and_wake_up_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
433fbe4c
TM
1202 rpc_wake_up(&clp->cl_rpcwaitq);
1203}
1204
1da177e4 1205/*
e005e804 1206 * Schedule the nfs_client asynchronous state management routine
1da177e4 1207 */
b0d3ded1 1208void nfs4_schedule_state_manager(struct nfs_client *clp)
1da177e4 1209{
5043e900 1210 struct task_struct *task;
2446ab60 1211 char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
956fd46f
TM
1212 struct rpc_clnt *clnt = clp->cl_rpcclient;
1213 bool swapon = false;
1da177e4 1214
956fd46f 1215 if (clnt->cl_shutdown)
6ad477a6
BC
1216 return;
1217
aeabb3c9 1218 set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
956fd46f
TM
1219
1220 if (atomic_read(&clnt->cl_swapper)) {
1221 swapon = !test_and_set_bit(NFS4CLNT_MANAGER_AVAILABLE,
1222 &clp->cl_state);
1223 if (!swapon) {
1224 wake_up_var(&clp->cl_state);
1225 return;
1226 }
4dc73c67 1227 }
956fd46f
TM
1228
1229 if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
1230 return;
1231
5043e900 1232 __module_get(THIS_MODULE);
212bf41d 1233 refcount_inc(&clp->cl_count);
2446ab60
TM
1234
1235 /* The rcu_read_lock() is not strictly necessary, as the state
1236 * manager is the only thread that ever changes the rpc_xprt
1237 * after it's initialized. At this point, we're single threaded. */
1238 rcu_read_lock();
1239 snprintf(buf, sizeof(buf), "%s-manager",
1240 rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR));
1241 rcu_read_unlock();
f170168b 1242 task = kthread_run(nfs4_run_state_manager, clp, "%s", buf);
2446ab60
TM
1243 if (IS_ERR(task)) {
1244 printk(KERN_ERR "%s: kthread_run: %ld\n",
1245 __func__, PTR_ERR(task));
b4e4f669
TM
1246 if (!nfs_client_init_is_complete(clp))
1247 nfs_mark_client_ready(clp, PTR_ERR(task));
956fd46f
TM
1248 if (swapon)
1249 clear_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state);
2446ab60
TM
1250 nfs4_clear_state_manager_bit(clp);
1251 nfs_put_client(clp);
1252 module_put(THIS_MODULE);
1253 }
1da177e4
LT
1254}
1255
1256/*
0400a6b0 1257 * Schedule a lease recovery attempt
1da177e4 1258 */
0400a6b0 1259void nfs4_schedule_lease_recovery(struct nfs_client *clp)
1da177e4
LT
1260{
1261 if (!clp)
1262 return;
e598d843
TM
1263 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1264 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
cc0a9843
TM
1265 dprintk("%s: scheduling lease recovery for server %s\n", __func__,
1266 clp->cl_hostname);
e005e804 1267 nfs4_schedule_state_manager(clp);
1da177e4 1268}
9cb81968 1269EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
1da177e4 1270
c9fdeb28
CL
1271/**
1272 * nfs4_schedule_migration_recovery - trigger migration recovery
1273 *
1274 * @server: FSID that is migrating
1275 *
1276 * Returns zero if recovery has started, otherwise a negative NFS4ERR
1277 * value is returned.
1278 */
1279int nfs4_schedule_migration_recovery(const struct nfs_server *server)
1280{
1281 struct nfs_client *clp = server->nfs_client;
1282
1283 if (server->fh_expire_type != NFS4_FH_PERSISTENT) {
1284 pr_err("NFS: volatile file handles not supported (server %s)\n",
1285 clp->cl_hostname);
1286 return -NFS4ERR_IO;
1287 }
1288
1289 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
1290 return -NFS4ERR_IO;
1291
1292 dprintk("%s: scheduling migration recovery for (%llx:%llx) on %s\n",
1293 __func__,
1294 (unsigned long long)server->fsid.major,
1295 (unsigned long long)server->fsid.minor,
1296 clp->cl_hostname);
1297
1298 set_bit(NFS_MIG_IN_TRANSITION,
1299 &((struct nfs_server *)server)->mig_status);
1300 set_bit(NFS4CLNT_MOVED, &clp->cl_state);
1301
1302 nfs4_schedule_state_manager(clp);
1303 return 0;
1304}
1305EXPORT_SYMBOL_GPL(nfs4_schedule_migration_recovery);
1306
b7f7a66e
CL
1307/**
1308 * nfs4_schedule_lease_moved_recovery - start lease-moved recovery
1309 *
1310 * @clp: server to check for moved leases
1311 *
1312 */
1313void nfs4_schedule_lease_moved_recovery(struct nfs_client *clp)
1314{
1315 dprintk("%s: scheduling lease-moved recovery for client ID %llx on %s\n",
1316 __func__, clp->cl_clientid, clp->cl_hostname);
1317
1318 set_bit(NFS4CLNT_LEASE_MOVED, &clp->cl_state);
1319 nfs4_schedule_state_manager(clp);
1320}
1321EXPORT_SYMBOL_GPL(nfs4_schedule_lease_moved_recovery);
1322
33021279
TM
1323int nfs4_wait_clnt_recover(struct nfs_client *clp)
1324{
1325 int res;
1326
1327 might_sleep();
1328
212bf41d 1329 refcount_inc(&clp->cl_count);
74316201 1330 res = wait_on_bit_action(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
f5d39b02
PZ
1331 nfs_wait_bit_killable,
1332 TASK_KILLABLE|TASK_FREEZABLE_UNSAFE);
33021279 1333 if (res)
0625c2dd 1334 goto out;
33021279 1335 if (clp->cl_cons_state < 0)
0625c2dd
CL
1336 res = clp->cl_cons_state;
1337out:
1338 nfs_put_client(clp);
1339 return res;
33021279
TM
1340}
1341
1342int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1343{
1344 unsigned int loop;
1345 int ret;
1346
1347 for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1348 ret = nfs4_wait_clnt_recover(clp);
1349 if (ret != 0)
1350 break;
1351 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1352 !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1353 break;
1354 nfs4_schedule_state_manager(clp);
1355 ret = -EIO;
1356 }
1357 return ret;
1358}
1359
ad1e3968
TM
1360/*
1361 * nfs40_handle_cb_pathdown - return all delegations after NFS4ERR_CB_PATH_DOWN
1362 * @clp: client to process
1363 *
1364 * Set the NFS4CLNT_LEASE_EXPIRED state in order to force a
1365 * resend of the SETCLIENTID and hence re-establish the
1366 * callback channel. Then return all existing delegations.
1367 */
1368static void nfs40_handle_cb_pathdown(struct nfs_client *clp)
1369{
1370 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1371 nfs_expire_all_delegations(clp);
cc0a9843
TM
1372 dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
1373 clp->cl_hostname);
ad1e3968
TM
1374}
1375
042b60be
TM
1376void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
1377{
ad1e3968 1378 nfs40_handle_cb_pathdown(clp);
042b60be
TM
1379 nfs4_schedule_state_manager(clp);
1380}
1381
f9feab1e 1382static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
b79a4a1b
TM
1383{
1384
7ebeb7fe
TM
1385 if (!nfs4_valid_open_stateid(state))
1386 return 0;
b79a4a1b
TM
1387 set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1388 /* Don't recover state that expired before the reboot */
1389 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
1390 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1391 return 0;
1392 }
7eff03ae 1393 set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
b79a4a1b
TM
1394 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1395 return 1;
1396}
1397
4f14c194 1398int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
b79a4a1b 1399{
7ebeb7fe
TM
1400 if (!nfs4_valid_open_stateid(state))
1401 return 0;
b79a4a1b
TM
1402 set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
1403 clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
7eff03ae 1404 set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
b79a4a1b
TM
1405 set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
1406 return 1;
1407}
1408
5d422301 1409int nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
0400a6b0
TM
1410{
1411 struct nfs_client *clp = server->nfs_client;
1412
7ebeb7fe 1413 if (!nfs4_state_mark_reclaim_nograce(clp, state))
5d422301 1414 return -EBADF;
994b15b9
TM
1415 nfs_inode_find_delegation_state_and_recover(state->inode,
1416 &state->stateid);
cc0a9843
TM
1417 dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
1418 clp->cl_hostname);
0400a6b0 1419 nfs4_schedule_state_manager(clp);
5d422301 1420 return 0;
0400a6b0 1421}
9cb81968 1422EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
0400a6b0 1423
6c2d8f8d
TM
1424static struct nfs4_lock_state *
1425nfs_state_find_lock_state_by_stateid(struct nfs4_state *state,
1426 const nfs4_stateid *stateid)
1427{
1428 struct nfs4_lock_state *pos;
1429
1430 list_for_each_entry(pos, &state->lock_states, ls_locks) {
1431 if (!test_bit(NFS_LOCK_INITIALIZED, &pos->ls_flags))
1432 continue;
42c304c3 1433 if (nfs4_stateid_match_or_older(&pos->ls_stateid, stateid))
6c2d8f8d
TM
1434 return pos;
1435 }
1436 return NULL;
1437}
1438
1439static bool nfs_state_lock_state_matches_stateid(struct nfs4_state *state,
1440 const nfs4_stateid *stateid)
1441{
1442 bool found = false;
1443
1444 if (test_bit(LK_STATE_IN_USE, &state->flags)) {
1445 spin_lock(&state->state_lock);
1446 if (nfs_state_find_lock_state_by_stateid(state, stateid))
1447 found = true;
1448 spin_unlock(&state->state_lock);
1449 }
1450 return found;
1451}
1452
a1d0b5ee
TM
1453void nfs_inode_find_state_and_recover(struct inode *inode,
1454 const nfs4_stateid *stateid)
1455{
1456 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1457 struct nfs_inode *nfsi = NFS_I(inode);
1458 struct nfs_open_context *ctx;
1459 struct nfs4_state *state;
1460 bool found = false;
1461
0de43976
TM
1462 rcu_read_lock();
1463 list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
a1d0b5ee
TM
1464 state = ctx->state;
1465 if (state == NULL)
1466 continue;
42c304c3 1467 if (nfs4_stateid_match_or_older(&state->stateid, stateid) &&
7ebeb7fe 1468 nfs4_state_mark_reclaim_nograce(clp, state)) {
6c2d8f8d 1469 found = true;
a1d0b5ee 1470 continue;
6c2d8f8d 1471 }
42c304c3
TM
1472 if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1473 nfs4_stateid_match_or_older(&state->open_stateid, stateid) &&
46280d9d
TM
1474 nfs4_state_mark_reclaim_nograce(clp, state)) {
1475 found = true;
1476 continue;
1477 }
7ebeb7fe
TM
1478 if (nfs_state_lock_state_matches_stateid(state, stateid) &&
1479 nfs4_state_mark_reclaim_nograce(clp, state))
6c2d8f8d 1480 found = true;
a1d0b5ee 1481 }
0de43976 1482 rcu_read_unlock();
6c2d8f8d
TM
1483
1484 nfs_inode_find_delegation_state_and_recover(inode, stateid);
a1d0b5ee
TM
1485 if (found)
1486 nfs4_schedule_state_manager(clp);
1487}
1488
86dbd08b 1489static void nfs4_state_mark_open_context_bad(struct nfs4_state *state, int err)
c58c8441
TM
1490{
1491 struct inode *inode = state->inode;
1492 struct nfs_inode *nfsi = NFS_I(inode);
1493 struct nfs_open_context *ctx;
1494
0de43976
TM
1495 rcu_read_lock();
1496 list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
c58c8441
TM
1497 if (ctx->state != state)
1498 continue;
1499 set_bit(NFS_CONTEXT_BAD, &ctx->flags);
86dbd08b
TM
1500 pr_warn("NFSv4: state recovery failed for open file %pd2, "
1501 "error = %d\n", ctx->dentry, err);
c58c8441 1502 }
0de43976 1503 rcu_read_unlock();
c58c8441
TM
1504}
1505
5d422301
TM
1506static void nfs4_state_mark_recovery_failed(struct nfs4_state *state, int error)
1507{
1508 set_bit(NFS_STATE_RECOVERY_FAILED, &state->flags);
86dbd08b 1509 nfs4_state_mark_open_context_bad(state, error);
5d422301
TM
1510}
1511
a1d0b5ee 1512
02860014 1513static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
1da177e4
LT
1514{
1515 struct inode *inode = state->inode;
19e03c57 1516 struct nfs_inode *nfsi = NFS_I(inode);
1da177e4 1517 struct file_lock *fl;
dce2630c 1518 struct nfs4_lock_state *lsp;
1da177e4 1519 int status = 0;
17b985de 1520 struct file_lock_context *flctx = locks_inode_context(inode);
bd61e0a9 1521 struct list_head *list;
1da177e4 1522
bd61e0a9 1523 if (flctx == NULL)
3f09df70
TM
1524 return 0;
1525
bd61e0a9
JL
1526 list = &flctx->flc_posix;
1527
3f09df70 1528 /* Guard against delegation returns and new lock/unlock calls */
19e03c57 1529 down_write(&nfsi->rwsem);
6109c850 1530 spin_lock(&flctx->flc_lock);
bd61e0a9
JL
1531restart:
1532 list_for_each_entry(fl, list, fl_list) {
5263e31e
JL
1533 if (nfs_file_open_context(fl->fl_file)->state != state)
1534 continue;
6109c850 1535 spin_unlock(&flctx->flc_lock);
5263e31e
JL
1536 status = ops->recover_lock(state, fl);
1537 switch (status) {
1538 case 0:
1539 break;
67e7b52d 1540 case -ETIMEDOUT:
5263e31e
JL
1541 case -ESTALE:
1542 case -NFS4ERR_ADMIN_REVOKED:
1543 case -NFS4ERR_STALE_STATEID:
1544 case -NFS4ERR_BAD_STATEID:
1545 case -NFS4ERR_EXPIRED:
1546 case -NFS4ERR_NO_GRACE:
1547 case -NFS4ERR_STALE_CLIENTID:
1548 case -NFS4ERR_BADSESSION:
1549 case -NFS4ERR_BADSLOT:
1550 case -NFS4ERR_BAD_HIGH_SLOT:
1551 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1552 goto out;
1553 default:
1554 pr_err("NFS: %s: unhandled error %d\n",
1555 __func__, status);
df561f66 1556 fallthrough;
5263e31e
JL
1557 case -ENOMEM:
1558 case -NFS4ERR_DENIED:
1559 case -NFS4ERR_RECLAIM_BAD:
1560 case -NFS4ERR_RECLAIM_CONFLICT:
dce2630c
N
1561 lsp = fl->fl_u.nfs4_fl.owner;
1562 if (lsp)
1563 set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
5263e31e
JL
1564 status = 0;
1565 }
6109c850 1566 spin_lock(&flctx->flc_lock);
5263e31e 1567 }
bd61e0a9
JL
1568 if (list == &flctx->flc_posix) {
1569 list = &flctx->flc_flock;
1570 goto restart;
1571 }
6109c850 1572 spin_unlock(&flctx->flc_lock);
965b5d67 1573out:
19e03c57 1574 up_write(&nfsi->rwsem);
1da177e4
LT
1575 return status;
1576}
1577
80f42368
AS
1578#ifdef CONFIG_NFS_V4_2
1579static void nfs42_complete_copies(struct nfs4_state_owner *sp, struct nfs4_state *state)
1580{
1581 struct nfs4_copy_state *copy;
1582
0e65a32c
OK
1583 if (!test_bit(NFS_CLNT_DST_SSC_COPY_STATE, &state->flags) &&
1584 !test_bit(NFS_CLNT_SRC_SSC_COPY_STATE, &state->flags))
80f42368
AS
1585 return;
1586
1587 spin_lock(&sp->so_server->nfs_client->cl_lock);
1588 list_for_each_entry(copy, &sp->so_server->ss_copies, copies) {
0e65a32c
OK
1589 if ((test_bit(NFS_CLNT_DST_SSC_COPY_STATE, &state->flags) &&
1590 !nfs4_stateid_match_other(&state->stateid,
1591 &copy->parent_dst_state->stateid)))
1592 continue;
80f42368 1593 copy->flags = 1;
0e65a32c
OK
1594 if (test_and_clear_bit(NFS_CLNT_DST_SSC_COPY_STATE,
1595 &state->flags)) {
1596 clear_bit(NFS_CLNT_SRC_SSC_COPY_STATE, &state->flags);
1597 complete(&copy->completion);
1598 }
1599 }
1600 list_for_each_entry(copy, &sp->so_server->ss_copies, src_copies) {
1601 if ((test_bit(NFS_CLNT_SRC_SSC_COPY_STATE, &state->flags) &&
1602 !nfs4_stateid_match_other(&state->stateid,
1603 &copy->parent_src_state->stateid)))
1604 continue;
1605 copy->flags = 1;
1606 if (test_and_clear_bit(NFS_CLNT_DST_SSC_COPY_STATE,
1607 &state->flags))
1608 complete(&copy->completion);
80f42368
AS
1609 }
1610 spin_unlock(&sp->so_server->nfs_client->cl_lock);
1611}
1612#else /* !CONFIG_NFS_V4_2 */
1613static inline void nfs42_complete_copies(struct nfs4_state_owner *sp,
1614 struct nfs4_state *state)
1615{
1616}
1617#endif /* CONFIG_NFS_V4_2 */
1618
cb7a8384 1619static int __nfs4_reclaim_open_state(struct nfs4_state_owner *sp, struct nfs4_state *state,
3e2910c7
N
1620 const struct nfs4_state_recovery_ops *ops,
1621 int *lost_locks)
cb7a8384
AS
1622{
1623 struct nfs4_lock_state *lock;
1624 int status;
1625
1626 status = ops->recover_open(sp, state);
1627 if (status < 0)
1628 return status;
1629
1630 status = nfs4_reclaim_locks(state, ops);
1631 if (status < 0)
1632 return status;
1633
1634 if (!test_bit(NFS_DELEGATED_STATE, &state->flags)) {
1635 spin_lock(&state->state_lock);
1636 list_for_each_entry(lock, &state->lock_states, ls_locks) {
21f86d2d 1637 trace_nfs4_state_lock_reclaim(state, lock);
ef8d98f2
N
1638 if (!test_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags) &&
1639 !test_bit(NFS_LOCK_UNLOCKING, &lock->ls_flags))
3e2910c7 1640 *lost_locks += 1;
cb7a8384
AS
1641 }
1642 spin_unlock(&state->state_lock);
1643 }
1644
80f42368 1645 nfs42_complete_copies(sp, state);
cb7a8384
AS
1646 clear_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
1647 return status;
1648}
1649
3e2910c7
N
1650static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp,
1651 const struct nfs4_state_recovery_ops *ops,
1652 int *lost_locks)
1da177e4
LT
1653{
1654 struct nfs4_state *state;
c34fae00 1655 unsigned int loop = 0;
1da177e4 1656 int status = 0;
0b9018b9
OK
1657#ifdef CONFIG_NFS_V4_2
1658 bool found_ssc_copy_state = false;
1659#endif /* CONFIG_NFS_V4_2 */
1da177e4
LT
1660
1661 /* Note: we rely on the sp->so_states list being ordered
1662 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
1663 * states first.
1664 * This is needed to ensure that the server won't give us any
1665 * read delegations that we have to return if, say, we are
1666 * recovering after a network partition or a reboot from a
1667 * server that doesn't support a grace period.
1668 */
fe1d8195 1669 spin_lock(&sp->so_lock);
abbec2da 1670 raw_write_seqcount_begin(&sp->so_reclaim_seqcount);
c137afab 1671restart:
1da177e4 1672 list_for_each_entry(state, &sp->so_states, open_states) {
b79a4a1b
TM
1673 if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
1674 continue;
5d422301
TM
1675 if (!nfs4_valid_open_stateid(state))
1676 continue;
1da177e4
LT
1677 if (state->state == 0)
1678 continue;
0b9018b9
OK
1679#ifdef CONFIG_NFS_V4_2
1680 if (test_bit(NFS_SRV_SSC_COPY_STATE, &state->flags)) {
1681 nfs4_state_mark_recovery_failed(state, -EIO);
1682 found_ssc_copy_state = true;
1683 continue;
1684 }
1685#endif /* CONFIG_NFS_V4_2 */
ace9fad4 1686 refcount_inc(&state->count);
fe1d8195 1687 spin_unlock(&sp->so_lock);
3e2910c7 1688 status = __nfs4_reclaim_open_state(sp, state, ops, lost_locks);
cb7a8384 1689
1da177e4 1690 switch (status) {
35a61606 1691 default:
c34fae00
TM
1692 if (status >= 0) {
1693 loop = 0;
b79a4a1b 1694 break;
c34fae00 1695 }
35a61606 1696 printk(KERN_ERR "NFS: %s: unhandled error %d\n", __func__, status);
df561f66 1697 fallthrough;
35a61606
AS
1698 case -ENOENT:
1699 case -ENOMEM:
1700 case -EACCES:
1701 case -EROFS:
1702 case -EIO:
1703 case -ESTALE:
1704 /* Open state on this file cannot be recovered */
1705 nfs4_state_mark_recovery_failed(state, status);
1706 break;
1707 case -EAGAIN:
1708 ssleep(1);
c34fae00
TM
1709 if (loop++ < 10) {
1710 set_bit(ops->state_flag_bit, &state->flags);
1711 break;
1712 }
df561f66 1713 fallthrough;
35a61606
AS
1714 case -NFS4ERR_ADMIN_REVOKED:
1715 case -NFS4ERR_STALE_STATEID:
1716 case -NFS4ERR_OLD_STATEID:
1717 case -NFS4ERR_BAD_STATEID:
1718 case -NFS4ERR_RECLAIM_BAD:
1719 case -NFS4ERR_RECLAIM_CONFLICT:
1720 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1721 break;
1722 case -NFS4ERR_EXPIRED:
1723 case -NFS4ERR_NO_GRACE:
1724 nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
df561f66 1725 fallthrough;
35a61606
AS
1726 case -NFS4ERR_STALE_CLIENTID:
1727 case -NFS4ERR_BADSESSION:
1728 case -NFS4ERR_BADSLOT:
1729 case -NFS4ERR_BAD_HIGH_SLOT:
1730 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
67e7b52d 1731 case -ETIMEDOUT:
35a61606 1732 goto out_err;
1da177e4 1733 }
fe1d8195 1734 nfs4_put_open_state(state);
c137afab 1735 spin_lock(&sp->so_lock);
fe1d8195 1736 goto restart;
1da177e4 1737 }
abbec2da 1738 raw_write_seqcount_end(&sp->so_reclaim_seqcount);
fe1d8195 1739 spin_unlock(&sp->so_lock);
0b9018b9
OK
1740#ifdef CONFIG_NFS_V4_2
1741 if (found_ssc_copy_state)
1742 return -EIO;
1743#endif /* CONFIG_NFS_V4_2 */
1da177e4
LT
1744 return 0;
1745out_err:
fe1d8195 1746 nfs4_put_open_state(state);
c137afab 1747 spin_lock(&sp->so_lock);
abbec2da 1748 raw_write_seqcount_end(&sp->so_reclaim_seqcount);
c137afab 1749 spin_unlock(&sp->so_lock);
1da177e4
LT
1750 return status;
1751}
1752
b79a4a1b
TM
1753static void nfs4_clear_open_state(struct nfs4_state *state)
1754{
1755 struct nfs4_lock_state *lock;
1756
1757 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1758 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1759 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1760 clear_bit(NFS_O_RDWR_STATE, &state->flags);
4b44b40e 1761 spin_lock(&state->state_lock);
b79a4a1b 1762 list_for_each_entry(lock, &state->lock_states, ls_locks) {
b79a4a1b 1763 lock->ls_seqid.flags = 0;
795a88c9 1764 clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
b79a4a1b 1765 }
4b44b40e 1766 spin_unlock(&state->state_lock);
b79a4a1b
TM
1767}
1768
24d292b8
CL
1769static void nfs4_reset_seqids(struct nfs_server *server,
1770 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
cee54fc9 1771{
24d292b8 1772 struct nfs_client *clp = server->nfs_client;
cee54fc9 1773 struct nfs4_state_owner *sp;
9f958ab8 1774 struct rb_node *pos;
cee54fc9 1775 struct nfs4_state *state;
cee54fc9 1776
24d292b8
CL
1777 spin_lock(&clp->cl_lock);
1778 for (pos = rb_first(&server->state_owners);
1779 pos != NULL;
1780 pos = rb_next(pos)) {
1781 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
cee54fc9 1782 sp->so_seqid.flags = 0;
ec073428 1783 spin_lock(&sp->so_lock);
cee54fc9 1784 list_for_each_entry(state, &sp->so_states, open_states) {
b79a4a1b
TM
1785 if (mark_reclaim(clp, state))
1786 nfs4_clear_open_state(state);
cee54fc9 1787 }
ec073428 1788 spin_unlock(&sp->so_lock);
cee54fc9 1789 }
24d292b8
CL
1790 spin_unlock(&clp->cl_lock);
1791}
1792
1793static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp,
1794 int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1795{
1796 struct nfs_server *server;
1797
1798 rcu_read_lock();
1799 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1800 nfs4_reset_seqids(server, mark_reclaim);
1801 rcu_read_unlock();
cee54fc9
TM
1802}
1803
b79a4a1b
TM
1804static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
1805{
e59679f2 1806 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
b79a4a1b
TM
1807 /* Mark all delegations for reclaim */
1808 nfs_delegation_mark_reclaim(clp);
1809 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
1810}
1811
0048fdd0 1812static int nfs4_reclaim_complete(struct nfs_client *clp,
965e9c23 1813 const struct nfs4_state_recovery_ops *ops,
a52458b4 1814 const struct cred *cred)
fce5c838
RL
1815{
1816 /* Notify the server we're done reclaiming our state */
1817 if (ops->reclaim_complete)
0048fdd0
TM
1818 return ops->reclaim_complete(clp, cred);
1819 return 0;
fce5c838
RL
1820}
1821
24d292b8 1822static void nfs4_clear_reclaim_server(struct nfs_server *server)
b79a4a1b 1823{
24d292b8 1824 struct nfs_client *clp = server->nfs_client;
b79a4a1b
TM
1825 struct nfs4_state_owner *sp;
1826 struct rb_node *pos;
1827 struct nfs4_state *state;
1828
24d292b8
CL
1829 spin_lock(&clp->cl_lock);
1830 for (pos = rb_first(&server->state_owners);
1831 pos != NULL;
1832 pos = rb_next(pos)) {
1833 sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
b79a4a1b
TM
1834 spin_lock(&sp->so_lock);
1835 list_for_each_entry(state, &sp->so_states, open_states) {
24d292b8
CL
1836 if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
1837 &state->flags))
b79a4a1b
TM
1838 continue;
1839 nfs4_state_mark_reclaim_nograce(clp, state);
1840 }
1841 spin_unlock(&sp->so_lock);
1842 }
24d292b8
CL
1843 spin_unlock(&clp->cl_lock);
1844}
1845
1846static int nfs4_state_clear_reclaim_reboot(struct nfs_client *clp)
1847{
1848 struct nfs_server *server;
1849
1850 if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
1851 return 0;
1852
1853 rcu_read_lock();
1854 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
1855 nfs4_clear_reclaim_server(server);
1856 rcu_read_unlock();
b79a4a1b
TM
1857
1858 nfs_delegation_reap_unclaimed(clp);
6eaa6149
TM
1859 return 1;
1860}
1861
1862static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
1863{
965e9c23 1864 const struct nfs4_state_recovery_ops *ops;
a52458b4 1865 const struct cred *cred;
0048fdd0 1866 int err;
965e9c23 1867
6eaa6149
TM
1868 if (!nfs4_state_clear_reclaim_reboot(clp))
1869 return;
965e9c23 1870 ops = clp->cl_mvops->reboot_recovery_ops;
73d8bde5 1871 cred = nfs4_get_clid_cred(clp);
0048fdd0 1872 err = nfs4_reclaim_complete(clp, ops, cred);
a52458b4 1873 put_cred(cred);
0048fdd0
TM
1874 if (err == -NFS4ERR_CONN_NOT_BOUND_TO_SESSION)
1875 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
b79a4a1b
TM
1876}
1877
b79a4a1b
TM
1878static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
1879{
45870d69 1880 nfs_mark_test_expired_all_delegations(clp);
b79a4a1b
TM
1881 nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
1882}
1883
4f7cdf18 1884static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
e598d843
TM
1885{
1886 switch (error) {
000d3f95
AS
1887 case 0:
1888 break;
1889 case -NFS4ERR_CB_PATH_DOWN:
1890 nfs40_handle_cb_pathdown(clp);
1891 break;
1892 case -NFS4ERR_NO_GRACE:
1893 nfs4_state_end_reclaim_reboot(clp);
1894 break;
1895 case -NFS4ERR_STALE_CLIENTID:
1896 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1897 nfs4_state_start_reclaim_reboot(clp);
1898 break;
1899 case -NFS4ERR_EXPIRED:
1900 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1901 nfs4_state_start_reclaim_nograce(clp);
1902 break;
1903 case -NFS4ERR_BADSESSION:
1904 case -NFS4ERR_BADSLOT:
1905 case -NFS4ERR_BAD_HIGH_SLOT:
1906 case -NFS4ERR_DEADSESSION:
1907 case -NFS4ERR_SEQ_FALSE_RETRY:
1908 case -NFS4ERR_SEQ_MISORDERED:
1909 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1910 /* Zero session reset errors */
1911 break;
1912 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1913 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1914 break;
1915 default:
1916 dprintk("%s: failed to handle error %d for server %s\n",
1917 __func__, error, clp->cl_hostname);
1918 return error;
e598d843 1919 }
cc0a9843
TM
1920 dprintk("%s: handled error %d for server %s\n", __func__, error,
1921 clp->cl_hostname);
4f38e4aa 1922 return 0;
e598d843
TM
1923}
1924
02860014 1925static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
1da177e4 1926{
24d292b8
CL
1927 struct nfs4_state_owner *sp;
1928 struct nfs_server *server;
9f958ab8 1929 struct rb_node *pos;
c77e2283 1930 LIST_HEAD(freeme);
1da177e4 1931 int status = 0;
3e2910c7 1932 int lost_locks = 0;
1da177e4 1933
7eff03ae 1934restart:
24d292b8
CL
1935 rcu_read_lock();
1936 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
c77e2283 1937 nfs4_purge_state_owners(server, &freeme);
24d292b8
CL
1938 spin_lock(&clp->cl_lock);
1939 for (pos = rb_first(&server->state_owners);
1940 pos != NULL;
1941 pos = rb_next(pos)) {
1942 sp = rb_entry(pos,
1943 struct nfs4_state_owner, so_server_node);
1944 if (!test_and_clear_bit(ops->owner_flag_bit,
1945 &sp->so_flags))
1946 continue;
4a0954ef
TM
1947 if (!atomic_inc_not_zero(&sp->so_count))
1948 continue;
24d292b8
CL
1949 spin_unlock(&clp->cl_lock);
1950 rcu_read_unlock();
1951
3e2910c7 1952 status = nfs4_reclaim_open_state(sp, ops, &lost_locks);
24d292b8 1953 if (status < 0) {
3e2910c7
N
1954 if (lost_locks)
1955 pr_warn("NFS: %s: lost %d locks\n",
1956 clp->cl_hostname, lost_locks);
24d292b8
CL
1957 set_bit(ops->owner_flag_bit, &sp->so_flags);
1958 nfs4_put_state_owner(sp);
df817ba3
TM
1959 status = nfs4_recovery_handle_error(clp, status);
1960 return (status != 0) ? status : -EAGAIN;
24d292b8
CL
1961 }
1962
7eff03ae 1963 nfs4_put_state_owner(sp);
24d292b8 1964 goto restart;
7eff03ae 1965 }
24d292b8 1966 spin_unlock(&clp->cl_lock);
02860014 1967 }
24d292b8 1968 rcu_read_unlock();
c77e2283 1969 nfs4_free_state_owners(&freeme);
3e2910c7
N
1970 if (lost_locks)
1971 pr_warn("NFS: %s: lost %d locks\n",
1972 clp->cl_hostname, lost_locks);
df817ba3 1973 return 0;
02860014
TM
1974}
1975
1976static int nfs4_check_lease(struct nfs_client *clp)
1977{
a52458b4 1978 const struct cred *cred;
c48f4f35
TM
1979 const struct nfs4_state_maintenance_ops *ops =
1980 clp->cl_mvops->state_renewal_ops;
4f38e4aa 1981 int status;
1da177e4 1982
0f605b56
TM
1983 /* Is the client already known to have an expired lease? */
1984 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1985 return 0;
f15e1e8b 1986 cred = ops->get_state_renewal_cred(clp);
0f605b56 1987 if (cred == NULL) {
73d8bde5 1988 cred = nfs4_get_clid_cred(clp);
4f38e4aa 1989 status = -ENOKEY;
0f605b56
TM
1990 if (cred == NULL)
1991 goto out;
286d7d6a 1992 }
8e69514f 1993 status = ops->renew_lease(clp, cred);
a52458b4 1994 put_cred(cred);
bc7a05ca
TM
1995 if (status == -ETIMEDOUT) {
1996 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1997 return 0;
1998 }
0f605b56 1999out:
4f7cdf18 2000 return nfs4_recovery_handle_error(clp, status);
02860014
TM
2001}
2002
47b803c8
AA
2003/* Set NFS4CLNT_LEASE_EXPIRED and reclaim reboot state for all v4.0 errors
2004 * and for recoverable errors on EXCHANGE_ID for v4.1
2a6ee6aa
TM
2005 */
2006static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
2007{
2008 switch (status) {
89a21736
TM
2009 case -NFS4ERR_SEQ_MISORDERED:
2010 if (test_and_set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state))
2011 return -ESERVERFAULT;
2012 /* Lease confirmation error: retry after purging the lease */
2013 ssleep(1);
47b803c8
AA
2014 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
2015 break;
2a6ee6aa
TM
2016 case -NFS4ERR_STALE_CLIENTID:
2017 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
47b803c8 2018 nfs4_state_start_reclaim_reboot(clp);
2a6ee6aa 2019 break;
de734831
CL
2020 case -NFS4ERR_CLID_INUSE:
2021 pr_err("NFS: Server %s reports our clientid is in use\n",
2022 clp->cl_hostname);
2023 nfs_mark_client_ready(clp, -EPERM);
2024 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
2025 return -EPERM;
2a6ee6aa 2026 case -EACCES:
2a6ee6aa 2027 case -NFS4ERR_DELAY:
2a6ee6aa
TM
2028 case -EAGAIN:
2029 ssleep(1);
2030 break;
2031
2032 case -NFS4ERR_MINOR_VERS_MISMATCH:
2033 if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
2034 nfs_mark_client_ready(clp, -EPROTONOSUPPORT);
cc0a9843
TM
2035 dprintk("%s: exit with error %d for server %s\n",
2036 __func__, -EPROTONOSUPPORT, clp->cl_hostname);
2a6ee6aa 2037 return -EPROTONOSUPPORT;
ea027cb2
OK
2038 case -ENOSPC:
2039 if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
2040 nfs_mark_client_ready(clp, -EIO);
2041 return -EIO;
2a6ee6aa
TM
2042 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
2043 * in nfs4_exchange_id */
2044 default:
cc0a9843
TM
2045 dprintk("%s: exit with error %d for server %s\n", __func__,
2046 status, clp->cl_hostname);
2a6ee6aa
TM
2047 return status;
2048 }
2049 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
cc0a9843
TM
2050 dprintk("%s: handled error %d for server %s\n", __func__, status,
2051 clp->cl_hostname);
2a6ee6aa
TM
2052 return 0;
2053}
2054
b42353ff 2055static int nfs4_establish_lease(struct nfs_client *clp)
02860014 2056{
a52458b4 2057 const struct cred *cred;
c48f4f35
TM
2058 const struct nfs4_state_recovery_ops *ops =
2059 clp->cl_mvops->reboot_recovery_ops;
2a6ee6aa 2060 int status;
02860014 2061
8aafd2fd
TM
2062 status = nfs4_begin_drain_session(clp);
2063 if (status != 0)
2064 return status;
73d8bde5 2065 cred = nfs4_get_clid_cred(clp);
2a6ee6aa
TM
2066 if (cred == NULL)
2067 return -ENOENT;
2068 status = ops->establish_clid(clp, cred);
a52458b4 2069 put_cred(cred);
2a6ee6aa 2070 if (status != 0)
b42353ff
TM
2071 return status;
2072 pnfs_destroy_all_layouts(clp);
2073 return 0;
2074}
2075
6bbb4ae8
CL
2076/*
2077 * Returns zero or a negative errno. NFS4ERR values are converted
2078 * to local errno values.
2079 */
b42353ff
TM
2080static int nfs4_reclaim_lease(struct nfs_client *clp)
2081{
2082 int status;
2083
2084 status = nfs4_establish_lease(clp);
2085 if (status < 0)
2086 return nfs4_handle_reclaim_lease_error(clp, status);
2087 if (test_and_clear_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state))
2088 nfs4_state_start_reclaim_nograce(clp);
2089 if (!test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
2090 set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
2091 clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
2092 clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
2093 return 0;
2094}
2095
2096static int nfs4_purge_lease(struct nfs_client *clp)
2097{
2098 int status;
2099
2100 status = nfs4_establish_lease(clp);
2101 if (status < 0)
2a6ee6aa 2102 return nfs4_handle_reclaim_lease_error(clp, status);
b42353ff
TM
2103 clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
2104 set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
2105 nfs4_state_start_reclaim_nograce(clp);
2a6ee6aa 2106 return 0;
02860014
TM
2107}
2108
c9fdeb28
CL
2109/*
2110 * Try remote migration of one FSID from a source server to a
2111 * destination server. The source server provides a list of
2112 * potential destinations.
2113 *
2114 * Returns zero or a negative NFS4ERR status code.
2115 */
a52458b4 2116static int nfs4_try_migration(struct nfs_server *server, const struct cred *cred)
c9fdeb28
CL
2117{
2118 struct nfs_client *clp = server->nfs_client;
2119 struct nfs4_fs_locations *locations = NULL;
2120 struct inode *inode;
2121 struct page *page;
2122 int status, result;
2123
2124 dprintk("--> %s: FSID %llx:%llx on \"%s\"\n", __func__,
2125 (unsigned long long)server->fsid.major,
2126 (unsigned long long)server->fsid.minor,
2127 clp->cl_hostname);
2128
2129 result = 0;
2130 page = alloc_page(GFP_KERNEL);
2131 locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2132 if (page == NULL || locations == NULL) {
2133 dprintk("<-- %s: no memory\n", __func__);
2134 goto out;
2135 }
c3ed2227
BC
2136 locations->fattr = nfs_alloc_fattr();
2137 if (locations->fattr == NULL) {
2138 dprintk("<-- %s: no memory\n", __func__);
2139 goto out;
2140 }
c9fdeb28 2141
2b0143b5 2142 inode = d_inode(server->super->s_root);
1976b2b3
OK
2143 result = nfs4_proc_get_locations(server, NFS_FH(inode), locations,
2144 page, cred);
c9fdeb28
CL
2145 if (result) {
2146 dprintk("<-- %s: failed to retrieve fs_locations: %d\n",
2147 __func__, result);
2148 goto out;
2149 }
2150
2151 result = -NFS4ERR_NXIO;
90e12a31
OK
2152 if (!locations->nlocations)
2153 goto out;
2154
c3ed2227 2155 if (!(locations->fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)) {
c9fdeb28
CL
2156 dprintk("<-- %s: No fs_locations data, migration skipped\n",
2157 __func__);
2158 goto out;
2159 }
2160
8aafd2fd 2161 status = nfs4_begin_drain_session(clp);
1e672e36
WW
2162 if (status != 0) {
2163 result = status;
2164 goto out;
2165 }
c9fdeb28
CL
2166
2167 status = nfs4_replace_transport(server, locations);
2168 if (status != 0) {
2169 dprintk("<-- %s: failed to replace transport: %d\n",
2170 __func__, status);
2171 goto out;
2172 }
2173
2174 result = 0;
2175 dprintk("<-- %s: migration succeeded\n", __func__);
2176
2177out:
2178 if (page != NULL)
2179 __free_page(page);
c3ed2227
BC
2180 if (locations != NULL)
2181 kfree(locations->fattr);
c9fdeb28
CL
2182 kfree(locations);
2183 if (result) {
2184 pr_err("NFS: migration recovery failed (server %s)\n",
2185 clp->cl_hostname);
2186 set_bit(NFS_MIG_FAILED, &server->mig_status);
2187 }
2188 return result;
2189}
2190
2191/*
2192 * Returns zero or a negative NFS4ERR status code.
2193 */
2194static int nfs4_handle_migration(struct nfs_client *clp)
2195{
2196 const struct nfs4_state_maintenance_ops *ops =
2197 clp->cl_mvops->state_renewal_ops;
2198 struct nfs_server *server;
a52458b4 2199 const struct cred *cred;
c9fdeb28
CL
2200
2201 dprintk("%s: migration reported on \"%s\"\n", __func__,
2202 clp->cl_hostname);
2203
f15e1e8b 2204 cred = ops->get_state_renewal_cred(clp);
c9fdeb28
CL
2205 if (cred == NULL)
2206 return -NFS4ERR_NOENT;
2207
2208 clp->cl_mig_gen++;
2209restart:
2210 rcu_read_lock();
2211 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
2212 int status;
2213
2214 if (server->mig_gen == clp->cl_mig_gen)
2215 continue;
2216 server->mig_gen = clp->cl_mig_gen;
2217
2218 if (!test_and_clear_bit(NFS_MIG_IN_TRANSITION,
2219 &server->mig_status))
2220 continue;
2221
2222 rcu_read_unlock();
2223 status = nfs4_try_migration(server, cred);
2224 if (status < 0) {
a52458b4 2225 put_cred(cred);
c9fdeb28
CL
2226 return status;
2227 }
2228 goto restart;
2229 }
2230 rcu_read_unlock();
a52458b4 2231 put_cred(cred);
c9fdeb28
CL
2232 return 0;
2233}
2234
b7f7a66e
CL
2235/*
2236 * Test each nfs_server on the clp's cl_superblocks list to see
2237 * if it's moved to another server. Stop when the server no longer
2238 * returns NFS4ERR_LEASE_MOVED.
2239 */
2240static int nfs4_handle_lease_moved(struct nfs_client *clp)
2241{
2242 const struct nfs4_state_maintenance_ops *ops =
2243 clp->cl_mvops->state_renewal_ops;
2244 struct nfs_server *server;
a52458b4 2245 const struct cred *cred;
b7f7a66e
CL
2246
2247 dprintk("%s: lease moved reported on \"%s\"\n", __func__,
2248 clp->cl_hostname);
2249
f15e1e8b 2250 cred = ops->get_state_renewal_cred(clp);
b7f7a66e
CL
2251 if (cred == NULL)
2252 return -NFS4ERR_NOENT;
2253
2254 clp->cl_mig_gen++;
2255restart:
2256 rcu_read_lock();
2257 list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
2258 struct inode *inode;
2259 int status;
2260
2261 if (server->mig_gen == clp->cl_mig_gen)
2262 continue;
2263 server->mig_gen = clp->cl_mig_gen;
2264
2265 rcu_read_unlock();
2266
2b0143b5 2267 inode = d_inode(server->super->s_root);
b7f7a66e
CL
2268 status = nfs4_proc_fsid_present(inode, cred);
2269 if (status != -NFS4ERR_MOVED)
2270 goto restart; /* wasn't this one */
2271 if (nfs4_try_migration(server, cred) == -NFS4ERR_LEASE_MOVED)
2272 goto restart; /* there are more */
2273 goto out;
2274 }
2275 rcu_read_unlock();
2276
2277out:
a52458b4 2278 put_cred(cred);
b7f7a66e
CL
2279 return 0;
2280}
2281
05f4c350
CL
2282/**
2283 * nfs4_discover_server_trunking - Detect server IP address trunking
2284 *
2285 * @clp: nfs_client under test
2286 * @result: OUT: found nfs_client, or clp
2287 *
2288 * Returns zero or a negative errno. If zero is returned,
2289 * an nfs_client pointer is planted in "result".
2290 *
2291 * Note: since we are invoked in process context, and
2292 * not from inside the state manager, we cannot use
2293 * nfs4_handle_reclaim_lease_error().
2294 */
2295int nfs4_discover_server_trunking(struct nfs_client *clp,
2296 struct nfs_client **result)
2297{
2298 const struct nfs4_state_recovery_ops *ops =
2299 clp->cl_mvops->reboot_recovery_ops;
05f4c350 2300 struct rpc_clnt *clnt;
a52458b4 2301 const struct cred *cred;
4edaa308 2302 int i, status;
05f4c350
CL
2303
2304 dprintk("NFS: %s: testing '%s'\n", __func__, clp->cl_hostname);
2305
05f4c350 2306 clnt = clp->cl_rpcclient;
05f4c350
CL
2307 i = 0;
2308
2309 mutex_lock(&nfs_clid_init_mutex);
05f4c350 2310again:
ea33e6c3 2311 status = -ENOENT;
73d8bde5 2312 cred = nfs4_get_clid_cred(clp);
05f4c350
CL
2313 if (cred == NULL)
2314 goto out_unlock;
2315
2316 status = ops->detect_trunking(clp, result, cred);
a52458b4 2317 put_cred(cred);
05f4c350
CL
2318 switch (status) {
2319 case 0:
898fc11b
TM
2320 case -EINTR:
2321 case -ERESTARTSYS:
05f4c350 2322 break;
05f4c350 2323 case -ETIMEDOUT:
150e7260
TM
2324 if (clnt->cl_softrtry)
2325 break;
df561f66 2326 fallthrough;
150e7260 2327 case -NFS4ERR_DELAY:
05f4c350
CL
2328 case -EAGAIN:
2329 ssleep(1);
df561f66 2330 fallthrough;
202c312d 2331 case -NFS4ERR_STALE_CLIENTID:
05f4c350
CL
2332 dprintk("NFS: %s after status %d, retrying\n",
2333 __func__, status);
2334 goto again;
4edaa308 2335 case -EACCES:
d688f7b8
CL
2336 if (i++ == 0) {
2337 nfs4_root_machine_cred(clp);
2338 goto again;
2339 }
6d769f1e 2340 if (clnt->cl_auth->au_flavor == RPC_AUTH_UNIX)
4edaa308 2341 break;
df561f66 2342 fallthrough;
05f4c350
CL
2343 case -NFS4ERR_CLID_INUSE:
2344 case -NFS4ERR_WRONGSEC:
6d769f1e
JL
2345 /* No point in retrying if we already used RPC_AUTH_UNIX */
2346 if (clnt->cl_auth->au_flavor == RPC_AUTH_UNIX) {
2347 status = -EPERM;
2348 break;
2349 }
79d852bf 2350 clnt = rpc_clone_client_set_auth(clnt, RPC_AUTH_UNIX);
05f4c350
CL
2351 if (IS_ERR(clnt)) {
2352 status = PTR_ERR(clnt);
2353 break;
2354 }
b193d59a
TM
2355 /* Note: this is safe because we haven't yet marked the
2356 * client as ready, so we are the only user of
2357 * clp->cl_rpcclient
2358 */
2359 clnt = xchg(&clp->cl_rpcclient, clnt);
2360 rpc_shutdown_client(clnt);
2361 clnt = clp->cl_rpcclient;
05f4c350
CL
2362 goto again;
2363
2364 case -NFS4ERR_MINOR_VERS_MISMATCH:
2365 status = -EPROTONOSUPPORT;
2366 break;
2367
2368 case -EKEYEXPIRED:
05f4c350
CL
2369 case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
2370 * in nfs4_exchange_id */
2371 status = -EKEYEXPIRED;
ea33e6c3
TM
2372 break;
2373 default:
2374 pr_warn("NFS: %s unhandled error %d. Exiting with error EIO\n",
2375 __func__, status);
2376 status = -EIO;
05f4c350
CL
2377 }
2378
2379out_unlock:
2380 mutex_unlock(&nfs_clid_init_mutex);
05f4c350
CL
2381 dprintk("NFS: %s: status = %d\n", __func__, status);
2382 return status;
2383}
2384
76db6d95 2385#ifdef CONFIG_NFS_V4_1
9f594791 2386void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
0400a6b0 2387{
444f72fe
TM
2388 struct nfs_client *clp = session->clp;
2389
9f594791
TM
2390 switch (err) {
2391 default:
2392 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
2393 break;
2394 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2395 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2396 }
d94cbf6c 2397 nfs4_schedule_state_manager(clp);
0400a6b0 2398}
cbdabc7f 2399EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
0400a6b0 2400
3f10a6af 2401void nfs41_notify_server(struct nfs_client *clp)
ac074835
TM
2402{
2403 /* Use CHECK_LEASE to ping the server with a SEQUENCE */
2404 set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
2405 nfs4_schedule_state_manager(clp);
2406}
2407
0f79fd6f
TM
2408static void nfs4_reset_all_state(struct nfs_client *clp)
2409{
2410 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
2c820d9a 2411 set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
be0bfed0 2412 clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
0f79fd6f 2413 nfs4_state_start_reclaim_nograce(clp);
cc0a9843
TM
2414 dprintk("%s: scheduling reset of all state for server %s!\n",
2415 __func__, clp->cl_hostname);
0400a6b0 2416 nfs4_schedule_state_manager(clp);
0f79fd6f
TM
2417 }
2418}
2419
2420static void nfs41_handle_server_reboot(struct nfs_client *clp)
2421{
2422 if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
2423 nfs4_state_start_reclaim_reboot(clp);
cc0a9843
TM
2424 dprintk("%s: server %s rebooted!\n", __func__,
2425 clp->cl_hostname);
0400a6b0 2426 nfs4_schedule_state_manager(clp);
0f79fd6f
TM
2427 }
2428}
2429
8b895ce6 2430static void nfs41_handle_all_state_revoked(struct nfs_client *clp)
0f79fd6f 2431{
0f79fd6f 2432 nfs4_reset_all_state(clp);
cc0a9843 2433 dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
0f79fd6f
TM
2434}
2435
8b895ce6
TM
2436static void nfs41_handle_some_state_revoked(struct nfs_client *clp)
2437{
45870d69 2438 nfs4_state_start_reclaim_nograce(clp);
8b895ce6
TM
2439 nfs4_schedule_state_manager(clp);
2440
2441 dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
2442}
2443
0f79fd6f
TM
2444static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
2445{
4099287f
TM
2446 /* FIXME: For now, we destroy all layouts. */
2447 pnfs_destroy_all_layouts(clp);
8ca017c8 2448 nfs_test_expired_all_delegations(clp);
cc0a9843
TM
2449 dprintk("%s: Recallable state revoked on server %s!\n", __func__,
2450 clp->cl_hostname);
0f79fd6f
TM
2451}
2452
a9e64442 2453static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
0f79fd6f 2454{
b1352905
TM
2455 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
2456 nfs4_schedule_state_manager(clp);
2457
cc0a9843
TM
2458 dprintk("%s: server %s declared a backchannel fault\n", __func__,
2459 clp->cl_hostname);
0f79fd6f
TM
2460}
2461
a9e64442
WAA
2462static void nfs41_handle_cb_path_down(struct nfs_client *clp)
2463{
2464 if (test_and_set_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2465 &clp->cl_state) == 0)
2466 nfs4_schedule_state_manager(clp);
2467}
2468
0a014a44
TM
2469void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags,
2470 bool recovery)
0629e370
AB
2471{
2472 if (!flags)
2473 return;
2c820d9a
CL
2474
2475 dprintk("%s: \"%s\" (client ID %llx) flags=0x%08x\n",
2476 __func__, clp->cl_hostname, clp->cl_clientid, flags);
0a014a44
TM
2477 /*
2478 * If we're called from the state manager thread, then assume we're
2479 * already handling the RECLAIM_NEEDED and/or STATE_REVOKED.
2480 * Those flags are expected to remain set until we're done
2481 * recovering (see RFC5661, section 18.46.3).
2482 */
2483 if (recovery)
2484 goto out_recovery;
2c820d9a 2485
111d489f 2486 if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
0f79fd6f 2487 nfs41_handle_server_reboot(clp);
8b895ce6
TM
2488 if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED))
2489 nfs41_handle_all_state_revoked(clp);
2490 if (flags & (SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
d1c2331e 2491 SEQ4_STATUS_ADMIN_STATE_REVOKED))
8b895ce6 2492 nfs41_handle_some_state_revoked(clp);
d1c2331e
CL
2493 if (flags & SEQ4_STATUS_LEASE_MOVED)
2494 nfs4_schedule_lease_moved_recovery(clp);
111d489f 2495 if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
0f79fd6f 2496 nfs41_handle_recallable_state_revoked(clp);
0a014a44 2497out_recovery:
a9e64442
WAA
2498 if (flags & SEQ4_STATUS_BACKCHANNEL_FAULT)
2499 nfs41_handle_backchannel_fault(clp);
2500 else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
2501 SEQ4_STATUS_CB_PATH_DOWN_SESSION))
0f79fd6f 2502 nfs41_handle_cb_path_down(clp);
0629e370
AB
2503}
2504
c3fad1b1
AA
2505static int nfs4_reset_session(struct nfs_client *clp)
2506{
a52458b4 2507 const struct cred *cred;
c3fad1b1
AA
2508 int status;
2509
1a47e7a6
TM
2510 if (!nfs4_has_session(clp))
2511 return 0;
8aafd2fd
TM
2512 status = nfs4_begin_drain_session(clp);
2513 if (status != 0)
2514 return status;
73d8bde5 2515 cred = nfs4_get_clid_cred(clp);
848f5bda 2516 status = nfs4_proc_destroy_session(clp->cl_session, cred);
c489ee29
TM
2517 switch (status) {
2518 case 0:
2519 case -NFS4ERR_BADSESSION:
2520 case -NFS4ERR_DEADSESSION:
2521 break;
2522 case -NFS4ERR_BACK_CHAN_BUSY:
2523 case -NFS4ERR_DELAY:
2524 set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
2525 status = 0;
2526 ssleep(1);
2527 goto out;
2528 default:
f455848a 2529 status = nfs4_recovery_handle_error(clp, status);
c3fad1b1
AA
2530 goto out;
2531 }
2532
2533 memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
848f5bda 2534 status = nfs4_proc_create_session(clp, cred);
41f54a55 2535 if (status) {
cc0a9843
TM
2536 dprintk("%s: session reset failed with status %d for server %s!\n",
2537 __func__, status, clp->cl_hostname);
f2c1b510 2538 status = nfs4_handle_reclaim_lease_error(clp, status);
41f54a55
AA
2539 goto out;
2540 }
bda197f5 2541 nfs41_finish_session_reset(clp);
cc0a9843
TM
2542 dprintk("%s: session reset was successful for server %s!\n",
2543 __func__, clp->cl_hostname);
41f54a55 2544out:
a52458b4 2545 put_cred(cred);
c3fad1b1
AA
2546 return status;
2547}
76db6d95 2548
a9e64442
WAA
2549static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2550{
a52458b4 2551 const struct cred *cred;
2cf047c9
TM
2552 int ret;
2553
1a47e7a6
TM
2554 if (!nfs4_has_session(clp))
2555 return 0;
8aafd2fd
TM
2556 ret = nfs4_begin_drain_session(clp);
2557 if (ret != 0)
2558 return ret;
73d8bde5 2559 cred = nfs4_get_clid_cred(clp);
2cf047c9 2560 ret = nfs4_proc_bind_conn_to_session(clp, cred);
a52458b4 2561 put_cred(cred);
43ac544c 2562 clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
bf674c82
TM
2563 switch (ret) {
2564 case 0:
cc0a9843
TM
2565 dprintk("%s: bind_conn_to_session was successful for server %s!\n",
2566 __func__, clp->cl_hostname);
bf674c82
TM
2567 break;
2568 case -NFS4ERR_DELAY:
2569 ssleep(1);
2570 set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2571 break;
2572 default:
2573 return nfs4_recovery_handle_error(clp, ret);
2574 }
2575 return 0;
a9e64442 2576}
b5fdf841
TM
2577
2578static void nfs4_layoutreturn_any_run(struct nfs_client *clp)
2579{
2580 int iomode = 0;
2581
2582 if (test_and_clear_bit(NFS4CLNT_RECALL_ANY_LAYOUT_READ, &clp->cl_state))
2583 iomode += IOMODE_READ;
2584 if (test_and_clear_bit(NFS4CLNT_RECALL_ANY_LAYOUT_RW, &clp->cl_state))
2585 iomode += IOMODE_RW;
2586 /* Note: IOMODE_READ + IOMODE_RW == IOMODE_ANY */
2587 if (iomode) {
2588 pnfs_layout_return_unused_byclid(clp, iomode);
2589 set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
2590 }
2591}
76db6d95 2592#else /* CONFIG_NFS_V4_1 */
c3fad1b1 2593static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
a9e64442
WAA
2594
2595static int nfs4_bind_conn_to_session(struct nfs_client *clp)
2596{
2597 return 0;
2598}
b5fdf841
TM
2599
2600static void nfs4_layoutreturn_any_run(struct nfs_client *clp)
2601{
2602}
76db6d95
AA
2603#endif /* CONFIG_NFS_V4_1 */
2604
e005e804 2605static void nfs4_state_manager(struct nfs_client *clp)
02860014 2606{
3e17898a 2607 unsigned int memflags;
02860014 2608 int status = 0;
8ed27d4f 2609 const char *section = "", *section_sep = "";
02860014 2610
3e17898a
TM
2611 /*
2612 * State recovery can deadlock if the direct reclaim code tries
2613 * start NFS writeback. So ensure memory allocations are all
2614 * GFP_NOFS.
2615 */
2616 memflags = memalloc_nofs_save();
2617
02860014 2618 /* Ensure exclusive access to NFSv4 state */
47c2199b 2619 do {
511ba52e 2620 trace_nfs4_state_mgr(clp);
aeabb3c9 2621 clear_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
2c820d9a 2622 if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
8ed27d4f 2623 section = "purge state";
b42353ff 2624 status = nfs4_purge_lease(clp);
2a6ee6aa
TM
2625 if (status < 0)
2626 goto out_error;
b42353ff 2627 continue;
2c820d9a
CL
2628 }
2629
b42353ff 2630 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
8ed27d4f 2631 section = "lease expired";
b79a4a1b
TM
2632 /* We're going to have to re-establish a clientid */
2633 status = nfs4_reclaim_lease(clp);
2a6ee6aa 2634 if (status < 0)
b79a4a1b 2635 goto out_error;
b42353ff 2636 continue;
e598d843
TM
2637 }
2638
c3fad1b1 2639 /* Initialize or reset the session */
1a47e7a6 2640 if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
8ed27d4f 2641 section = "reset session";
4d643d1d 2642 status = nfs4_reset_session(clp);
b6d408ba
TM
2643 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
2644 continue;
2645 if (status < 0)
76db6d95 2646 goto out_error;
76db6d95 2647 }
b6d408ba 2648
a9e64442
WAA
2649 /* Send BIND_CONN_TO_SESSION */
2650 if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
1a47e7a6 2651 &clp->cl_state)) {
8ed27d4f 2652 section = "bind conn to session";
a9e64442
WAA
2653 status = nfs4_bind_conn_to_session(clp);
2654 if (status < 0)
2655 goto out_error;
bf674c82 2656 continue;
a9e64442
WAA
2657 }
2658
5df904ae
TM
2659 if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
2660 section = "check lease";
2661 status = nfs4_check_lease(clp);
2662 if (status < 0)
2663 goto out_error;
8faaa6d5 2664 continue;
c9fdeb28
CL
2665 }
2666
2667 if (test_and_clear_bit(NFS4CLNT_MOVED, &clp->cl_state)) {
2668 section = "migration";
2669 status = nfs4_handle_migration(clp);
2670 if (status < 0)
2671 goto out_error;
5df904ae
TM
2672 }
2673
b7f7a66e
CL
2674 if (test_and_clear_bit(NFS4CLNT_LEASE_MOVED, &clp->cl_state)) {
2675 section = "lease moved";
2676 status = nfs4_handle_lease_moved(clp);
2677 if (status < 0)
2678 goto out_error;
2679 }
2680
b79a4a1b 2681 /* First recover reboot state... */
e345e88a 2682 if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
8ed27d4f 2683 section = "reclaim reboot";
591d71cb 2684 status = nfs4_do_reclaim(clp,
c48f4f35 2685 clp->cl_mvops->reboot_recovery_ops);
df817ba3 2686 if (status == -EAGAIN)
b6d408ba
TM
2687 continue;
2688 if (status < 0)
2689 goto out_error;
df817ba3 2690 nfs4_state_end_reclaim_reboot(clp);
5d917cba 2691 continue;
02860014
TM
2692 }
2693
45870d69
TM
2694 /* Detect expired delegations... */
2695 if (test_and_clear_bit(NFS4CLNT_DELEGATION_EXPIRED, &clp->cl_state)) {
2696 section = "detect expired delegations";
2697 nfs_reap_expired_delegations(clp);
2698 continue;
2699 }
2700
b79a4a1b 2701 /* Now recover expired state... */
67e7b52d 2702 if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
8ed27d4f 2703 section = "reclaim nograce";
591d71cb 2704 status = nfs4_do_reclaim(clp,
c48f4f35 2705 clp->cl_mvops->nograce_recovery_ops);
df817ba3 2706 if (status == -EAGAIN)
b6d408ba
TM
2707 continue;
2708 if (status < 0)
b79a4a1b 2709 goto out_error;
67e7b52d 2710 clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
1da177e4 2711 }
707fb4b3 2712
3e17898a 2713 memalloc_nofs_restore(memflags);
5601a00d 2714 nfs4_end_drain_session(clp);
aeabb3c9
TM
2715 nfs4_clear_state_manager_bit(clp);
2716
ed1cc05a
TM
2717 if (test_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state) &&
2718 !test_and_set_bit(NFS4CLNT_MANAGER_RUNNING,
2719 &clp->cl_state)) {
2720 memflags = memalloc_nofs_save();
2721 continue;
2722 }
2723
b5fdf841 2724 if (!test_and_set_bit(NFS4CLNT_RECALL_RUNNING, &clp->cl_state)) {
aeabb3c9
TM
2725 if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
2726 nfs_client_return_marked_delegations(clp);
2727 set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
2728 }
b5fdf841
TM
2729 nfs4_layoutreturn_any_run(clp);
2730 clear_bit(NFS4CLNT_RECALL_RUNNING, &clp->cl_state);
707fb4b3 2731 }
e005e804 2732
4dc73c67
N
2733 return;
2734
a1aa09be 2735 } while (refcount_read(&clp->cl_count) > 1 && !signalled());
21a446cf
TM
2736 goto out_drain;
2737
1da177e4 2738out_error:
8ed27d4f
WAA
2739 if (strlen(section))
2740 section_sep = ": ";
511ba52e 2741 trace_nfs4_state_mgr_failed(clp, section, status);
8ed27d4f
WAA
2742 pr_warn_ratelimited("NFS: state manager%s%s failed on NFSv4 server %s"
2743 " with error %d\n", section_sep, section,
2744 clp->cl_hostname, -status);
ffe5a830 2745 ssleep(1);
21a446cf 2746out_drain:
3e17898a 2747 memalloc_nofs_restore(memflags);
5601a00d 2748 nfs4_end_drain_session(clp);
e005e804
TM
2749 nfs4_clear_state_manager_bit(clp);
2750}
2751
2752static int nfs4_run_state_manager(void *ptr)
2753{
2754 struct nfs_client *clp = ptr;
4dc73c67
N
2755 struct rpc_clnt *cl = clp->cl_rpcclient;
2756
2757 while (cl != cl->cl_parent)
2758 cl = cl->cl_parent;
e005e804
TM
2759
2760 allow_signal(SIGKILL);
4dc73c67 2761again:
e005e804 2762 nfs4_state_manager(clp);
956fd46f
TM
2763
2764 if (test_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state) &&
2765 !test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state)) {
4dc73c67
N
2766 wait_var_event_interruptible(&clp->cl_state,
2767 test_bit(NFS4CLNT_RUN_MANAGER,
2768 &clp->cl_state));
956fd46f
TM
2769 if (!atomic_read(&cl->cl_swapper))
2770 clear_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state);
2771 if (refcount_read(&clp->cl_count) > 1 && !signalled() &&
2772 !test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state))
4dc73c67
N
2773 goto again;
2774 /* Either no longer a swapper, or were signalled */
956fd46f 2775 clear_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state);
4dc73c67 2776 }
4dc73c67
N
2777
2778 if (refcount_read(&clp->cl_count) > 1 && !signalled() &&
2779 test_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state) &&
956fd46f 2780 !test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state))
4dc73c67
N
2781 goto again;
2782
e005e804 2783 nfs_put_client(clp);
080abad7 2784 module_put_and_kthread_exit(0);
e005e804 2785 return 0;
1da177e4 2786}