nfs4state.c 47.5 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
/*
 *  fs/nfs/nfs4state.c
 *
 *  Client-side XDR for NFSv4.
 *
 *  Copyright (c) 2002 The Regents of the University of Michigan.
 *  All rights reserved.
 *
 *  Kendrick Smith <kmsmith@umich.edu>
 *
 *  Redistribution and use in source and binary forms, with or without
 *  modification, are permitted provided that the following conditions
 *  are met:
 *
 *  1. Redistributions of source code must retain the above copyright
 *     notice, this list of conditions and the following disclaimer.
 *  2. Redistributions in binary form must reproduce the above copyright
 *     notice, this list of conditions and the following disclaimer in the
 *     documentation and/or other materials provided with the distribution.
 *  3. Neither the name of the University nor the names of its
 *     contributors may be used to endorse or promote products derived
 *     from this software without specific prior written permission.
 *
 *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
 *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
 *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
 *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
 *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
 *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
 *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
 *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * Implementation of the NFSv4 state model.  For the time being,
 * this is minimal, but will be made much more complex in a
 * subsequent patch.
 */

41
#include <linux/kernel.h>
L
Linus Torvalds 已提交
42
#include <linux/slab.h>
43
#include <linux/fs.h>
L
Linus Torvalds 已提交
44 45
#include <linux/nfs_fs.h>
#include <linux/nfs_idmap.h>
46 47
#include <linux/kthread.h>
#include <linux/module.h>
48
#include <linux/random.h>
49
#include <linux/ratelimit.h>
L
Linus Torvalds 已提交
50 51
#include <linux/workqueue.h>
#include <linux/bitops.h>
52
#include <linux/jiffies.h>
L
Linus Torvalds 已提交
53

54
#include "nfs4_fs.h"
L
Linus Torvalds 已提交
55 56
#include "callback.h"
#include "delegation.h"
57
#include "internal.h"
58
#include "pnfs.h"
L
Linus Torvalds 已提交
59 60 61

#define OPENOWNER_POOL_SIZE	8

62
const nfs4_stateid zero_stateid;
L
Linus Torvalds 已提交
63 64 65

static LIST_HEAD(nfs4_clientid_list);

66
int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
L
Linus Torvalds 已提交
67
{
68 69 70 71
	struct nfs4_setclientid_res clid = {
		.clientid = clp->cl_clientid,
		.confirm = clp->cl_confirm,
	};
72 73 74
	unsigned short port;
	int status;

75 76
	if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
		goto do_confirm;
77 78 79 80
	port = nfs_callback_tcpport;
	if (clp->cl_addr.ss_family == AF_INET6)
		port = nfs_callback_tcpport6;

81 82 83
	status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
	if (status != 0)
		goto out;
84 85 86 87
	clp->cl_clientid = clid.clientid;
	clp->cl_confirm = clid.confirm;
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
do_confirm:
88 89 90
	status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
	if (status != 0)
		goto out;
91
	clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
92 93
	nfs4_schedule_state_renewal(clp);
out:
L
Linus Torvalds 已提交
94 95 96
	return status;
}

97
struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117
{
	struct rpc_cred *cred = NULL;

	if (clp->cl_machine_cred != NULL)
		cred = get_rpccred(clp->cl_machine_cred);
	return cred;
}

static void nfs4_clear_machine_cred(struct nfs_client *clp)
{
	struct rpc_cred *cred;

	spin_lock(&clp->cl_lock);
	cred = clp->cl_machine_cred;
	clp->cl_machine_cred = NULL;
	spin_unlock(&clp->cl_lock);
	if (cred != NULL)
		put_rpccred(cred);
}

118 119
static struct rpc_cred *
nfs4_get_renew_cred_server_locked(struct nfs_server *server)
120
{
121
	struct rpc_cred *cred = NULL;
122
	struct nfs4_state_owner *sp;
123
	struct rb_node *pos;
124

125 126 127 128
	for (pos = rb_first(&server->state_owners);
	     pos != NULL;
	     pos = rb_next(pos)) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
129 130 131 132 133 134 135 136
		if (list_empty(&sp->so_states))
			continue;
		cred = get_rpccred(sp->so_cred);
		break;
	}
	return cred;
}

137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158
/**
 * nfs4_get_renew_cred_locked - Acquire credential for a renew operation
 * @clp: client state handle
 *
 * Returns an rpc_cred with reference count bumped, or NULL.
 * Caller must hold clp->cl_lock.
 */
struct rpc_cred *nfs4_get_renew_cred_locked(struct nfs_client *clp)
{
	struct rpc_cred *cred = NULL;
	struct nfs_server *server;

	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
		cred = nfs4_get_renew_cred_server_locked(server);
		if (cred != NULL)
			break;
	}
	rcu_read_unlock();
	return cred;
}

159 160
#if defined(CONFIG_NFS_V4_1)

R
Ricardo Labiaga 已提交
161 162 163 164 165
static int nfs41_setup_state_renewal(struct nfs_client *clp)
{
	int status;
	struct nfs_fsinfo fsinfo;

166 167 168 169 170
	if (!test_bit(NFS_CS_CHECK_LEASE_TIME, &clp->cl_res_state)) {
		nfs4_schedule_state_renewal(clp);
		return 0;
	}

R
Ricardo Labiaga 已提交
171 172 173 174 175 176 177 178 179 180 181 182 183 184
	status = nfs4_proc_get_lease_time(clp, &fsinfo);
	if (status == 0) {
		/* Update lease time and schedule renewal */
		spin_lock(&clp->cl_lock);
		clp->cl_lease_time = fsinfo.lease_time * HZ;
		clp->cl_last_renewal = jiffies;
		spin_unlock(&clp->cl_lock);

		nfs4_schedule_state_renewal(clp);
	}

	return status;
}

185 186 187 188 189
/*
 * Back channel returns NFS4ERR_DELAY for new requests when
 * NFS4_SESSION_DRAINING is set so there is no work to be done when draining
 * is ended.
 */
190
static void nfs4_end_drain_session(struct nfs_client *clp)
191
{
192
	struct nfs4_session *ses = clp->cl_session;
193
	struct nfs4_slot_table *tbl;
194 195
	int max_slots;

196 197
	if (ses == NULL)
		return;
198
	tbl = &ses->fc_slot_table;
199
	if (test_and_clear_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
200 201
		spin_lock(&tbl->slot_tbl_lock);
		max_slots = tbl->max_slots;
202
		while (max_slots--) {
203 204 205
			if (rpc_wake_up_first(&tbl->slot_tbl_waitq,
						nfs4_set_task_privileged,
						NULL) == NULL)
206 207
				break;
		}
208
		spin_unlock(&tbl->slot_tbl_lock);
209
	}
210 211
}

212
static int nfs4_wait_on_slot_tbl(struct nfs4_slot_table *tbl)
213 214
{
	spin_lock(&tbl->slot_tbl_lock);
215
	if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
216
		INIT_COMPLETION(tbl->complete);
217
		spin_unlock(&tbl->slot_tbl_lock);
218
		return wait_for_completion_interruptible(&tbl->complete);
219 220 221 222 223
	}
	spin_unlock(&tbl->slot_tbl_lock);
	return 0;
}

224 225 226 227 228 229 230 231 232 233 234 235 236 237
static int nfs4_begin_drain_session(struct nfs_client *clp)
{
	struct nfs4_session *ses = clp->cl_session;
	int ret = 0;

	set_bit(NFS4_SESSION_DRAINING, &ses->session_state);
	/* back channel */
	ret = nfs4_wait_on_slot_tbl(&ses->bc_slot_table);
	if (ret)
		return ret;
	/* fore channel */
	return nfs4_wait_on_slot_tbl(&ses->fc_slot_table);
}

238 239 240 241
int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
{
	int status;

242 243
	if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
		goto do_confirm;
244
	nfs4_begin_drain_session(clp);
245
	status = nfs4_proc_exchange_id(clp, cred);
R
Ricardo Labiaga 已提交
246 247
	if (status != 0)
		goto out;
248 249
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
do_confirm:
R
Ricardo Labiaga 已提交
250 251 252
	status = nfs4_proc_create_session(clp);
	if (status != 0)
		goto out;
253
	clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
R
Ricardo Labiaga 已提交
254 255 256
	nfs41_setup_state_renewal(clp);
	nfs_mark_client_ready(clp, NFS_CS_READY);
out:
257 258 259
	return status;
}

260 261 262 263 264 265 266 267 268 269 270 271
struct rpc_cred *nfs4_get_exchange_id_cred(struct nfs_client *clp)
{
	struct rpc_cred *cred;

	spin_lock(&clp->cl_lock);
	cred = nfs4_get_machine_cred_locked(clp);
	spin_unlock(&clp->cl_lock);
	return cred;
}

#endif /* CONFIG_NFS_V4_1 */

272 273
static struct rpc_cred *
nfs4_get_setclientid_cred_server(struct nfs_server *server)
274
{
275 276
	struct nfs_client *clp = server->nfs_client;
	struct rpc_cred *cred = NULL;
277
	struct nfs4_state_owner *sp;
278
	struct rb_node *pos;
279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298

	spin_lock(&clp->cl_lock);
	pos = rb_first(&server->state_owners);
	if (pos != NULL) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
		cred = get_rpccred(sp->so_cred);
	}
	spin_unlock(&clp->cl_lock);
	return cred;
}

/**
 * nfs4_get_setclientid_cred - Acquire credential for a setclientid operation
 * @clp: client state handle
 *
 * Returns an rpc_cred with reference count bumped, or NULL.
 */
struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
{
	struct nfs_server *server;
299
	struct rpc_cred *cred;
300

301 302
	spin_lock(&clp->cl_lock);
	cred = nfs4_get_machine_cred_locked(clp);
303
	spin_unlock(&clp->cl_lock);
304 305
	if (cred != NULL)
		goto out;
306 307 308 309 310 311

	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
		cred = nfs4_get_setclientid_cred_server(server);
		if (cred != NULL)
			break;
312
	}
313 314
	rcu_read_unlock();

315 316
out:
	return cred;
317 318
}

L
Linus Torvalds 已提交
319
static struct nfs4_state_owner *
320
nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
L
Linus Torvalds 已提交
321
{
322
	struct rb_node **p = &server->state_owners.rb_node,
323
		       *parent = NULL;
324
	struct nfs4_state_owner *sp;
L
Linus Torvalds 已提交
325

326 327
	while (*p != NULL) {
		parent = *p;
328
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
329 330 331 332 333 334

		if (cred < sp->so_cred)
			p = &parent->rb_left;
		else if (cred > sp->so_cred)
			p = &parent->rb_right;
		else {
335 336
			if (!list_empty(&sp->so_lru))
				list_del_init(&sp->so_lru);
337
			atomic_inc(&sp->so_count);
338
			return sp;
339
		}
L
Linus Torvalds 已提交
340
	}
341
	return NULL;
L
Linus Torvalds 已提交
342 343
}

344
static struct nfs4_state_owner *
345
nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
346
{
347 348
	struct nfs_server *server = new->so_server;
	struct rb_node **p = &server->state_owners.rb_node,
349 350
		       *parent = NULL;
	struct nfs4_state_owner *sp;
351
	int err;
352 353 354

	while (*p != NULL) {
		parent = *p;
355
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
356 357 358 359 360 361

		if (new->so_cred < sp->so_cred)
			p = &parent->rb_left;
		else if (new->so_cred > sp->so_cred)
			p = &parent->rb_right;
		else {
362 363
			if (!list_empty(&sp->so_lru))
				list_del_init(&sp->so_lru);
364 365 366 367
			atomic_inc(&sp->so_count);
			return sp;
		}
	}
368
	err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
369 370
	if (err)
		return ERR_PTR(err);
371 372
	rb_link_node(&new->so_server_node, parent, p);
	rb_insert_color(&new->so_server_node, &server->state_owners);
373 374 375 376
	return new;
}

static void
377
nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
378
{
379 380 381 382
	struct nfs_server *server = sp->so_server;

	if (!RB_EMPTY_NODE(&sp->so_server_node))
		rb_erase(&sp->so_server_node, &server->state_owners);
383
	ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
384 385
}

386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401
static void
nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
{
	sc->flags = 0;
	sc->counter = 0;
	spin_lock_init(&sc->lock);
	INIT_LIST_HEAD(&sc->list);
	rpc_init_wait_queue(&sc->wait, "Seqid_waitqueue");
}

static void
nfs4_destroy_seqid_counter(struct nfs_seqid_counter *sc)
{
	rpc_destroy_wait_queue(&sc->wait);
}

L
Linus Torvalds 已提交
402 403 404 405 406 407
/*
 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
 * create a new state_owner.
 *
 */
static struct nfs4_state_owner *
408 409 410
nfs4_alloc_state_owner(struct nfs_server *server,
		struct rpc_cred *cred,
		gfp_t gfp_flags)
L
Linus Torvalds 已提交
411 412 413
{
	struct nfs4_state_owner *sp;

414
	sp = kzalloc(sizeof(*sp), gfp_flags);
L
Linus Torvalds 已提交
415 416
	if (!sp)
		return NULL;
417 418
	sp->so_server = server;
	sp->so_cred = get_rpccred(cred);
419
	spin_lock_init(&sp->so_lock);
L
Linus Torvalds 已提交
420
	INIT_LIST_HEAD(&sp->so_states);
421
	nfs4_init_seqid_counter(&sp->so_seqid);
L
Linus Torvalds 已提交
422
	atomic_set(&sp->so_count, 1);
423
	INIT_LIST_HEAD(&sp->so_lru);
L
Linus Torvalds 已提交
424 425 426
	return sp;
}

427
static void
L
Linus Torvalds 已提交
428 429
nfs4_drop_state_owner(struct nfs4_state_owner *sp)
{
430 431 432
	if (!RB_EMPTY_NODE(&sp->so_server_node)) {
		struct nfs_server *server = sp->so_server;
		struct nfs_client *clp = server->nfs_client;
433 434

		spin_lock(&clp->cl_lock);
435 436
		rb_erase(&sp->so_server_node, &server->state_owners);
		RB_CLEAR_NODE(&sp->so_server_node);
437 438
		spin_unlock(&clp->cl_lock);
	}
L
Linus Torvalds 已提交
439 440
}

441 442
static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
{
443
	nfs4_destroy_seqid_counter(&sp->so_seqid);
444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472
	put_rpccred(sp->so_cred);
	kfree(sp);
}

static void nfs4_gc_state_owners(struct nfs_server *server)
{
	struct nfs_client *clp = server->nfs_client;
	struct nfs4_state_owner *sp, *tmp;
	unsigned long time_min, time_max;
	LIST_HEAD(doomed);

	spin_lock(&clp->cl_lock);
	time_max = jiffies;
	time_min = (long)time_max - (long)clp->cl_lease_time;
	list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
		/* NB: LRU is sorted so that oldest is at the head */
		if (time_in_range(sp->so_expires, time_min, time_max))
			break;
		list_move(&sp->so_lru, &doomed);
		nfs4_remove_state_owner_locked(sp);
	}
	spin_unlock(&clp->cl_lock);

	list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
		list_del(&sp->so_lru);
		nfs4_free_state_owner(sp);
	}
}

473 474 475 476 477 478 479 480
/**
 * nfs4_get_state_owner - Look up a state owner given a credential
 * @server: nfs_server to search
 * @cred: RPC credential to match
 *
 * Returns a pointer to an instantiated nfs4_state_owner struct, or NULL.
 */
struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server,
481 482
					      struct rpc_cred *cred,
					      gfp_t gfp_flags)
L
Linus Torvalds 已提交
483
{
484
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
485 486 487
	struct nfs4_state_owner *sp, *new;

	spin_lock(&clp->cl_lock);
488
	sp = nfs4_find_state_owner_locked(server, cred);
L
Linus Torvalds 已提交
489 490
	spin_unlock(&clp->cl_lock);
	if (sp != NULL)
491
		goto out;
492
	new = nfs4_alloc_state_owner(server, cred, gfp_flags);
493
	if (new == NULL)
494
		goto out;
495 496 497 498 499 500 501
	do {
		if (ida_pre_get(&server->openowner_id, gfp_flags) == 0)
			break;
		spin_lock(&clp->cl_lock);
		sp = nfs4_insert_state_owner_locked(new);
		spin_unlock(&clp->cl_lock);
	} while (sp == ERR_PTR(-EAGAIN));
502 503
	if (sp != new)
		nfs4_free_state_owner(new);
504 505
out:
	nfs4_gc_state_owners(server);
506
	return sp;
L
Linus Torvalds 已提交
507 508
}

509 510 511 512
/**
 * nfs4_put_state_owner - Release a nfs4_state_owner
 * @sp: state owner data to release
 */
L
Linus Torvalds 已提交
513 514
void nfs4_put_state_owner(struct nfs4_state_owner *sp)
{
515 516
	struct nfs_server *server = sp->so_server;
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
517 518 519

	if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
		return;
520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549

	if (!RB_EMPTY_NODE(&sp->so_server_node)) {
		sp->so_expires = jiffies;
		list_add_tail(&sp->so_lru, &server->state_owners_lru);
		spin_unlock(&clp->cl_lock);
	} else {
		nfs4_remove_state_owner_locked(sp);
		spin_unlock(&clp->cl_lock);
		nfs4_free_state_owner(sp);
	}
}

/**
 * nfs4_purge_state_owners - Release all cached state owners
 * @server: nfs_server with cached state owners to release
 *
 * Called at umount time.  Remaining state owners will be on
 * the LRU with ref count of zero.
 */
void nfs4_purge_state_owners(struct nfs_server *server)
{
	struct nfs_client *clp = server->nfs_client;
	struct nfs4_state_owner *sp, *tmp;
	LIST_HEAD(doomed);

	spin_lock(&clp->cl_lock);
	list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
		list_move(&sp->so_lru, &doomed);
		nfs4_remove_state_owner_locked(sp);
	}
L
Linus Torvalds 已提交
550
	spin_unlock(&clp->cl_lock);
551 552 553 554 555

	list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
		list_del(&sp->so_lru);
		nfs4_free_state_owner(sp);
	}
L
Linus Torvalds 已提交
556 557 558 559 560 561 562
}

static struct nfs4_state *
nfs4_alloc_open_state(void)
{
	struct nfs4_state *state;

563
	state = kzalloc(sizeof(*state), GFP_NOFS);
L
Linus Torvalds 已提交
564 565 566 567
	if (!state)
		return NULL;
	atomic_set(&state->count, 1);
	INIT_LIST_HEAD(&state->lock_states);
568
	spin_lock_init(&state->state_lock);
569
	seqlock_init(&state->seqlock);
L
Linus Torvalds 已提交
570 571 572
	return state;
}

573
void
574
nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
575
{
576
	if (state->state == fmode)
577 578
		return;
	/* NB! List reordering - see the reclaim code for why.  */
579 580
	if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
		if (fmode & FMODE_WRITE)
581 582 583 584
			list_move(&state->open_states, &state->owner->so_states);
		else
			list_move_tail(&state->open_states, &state->owner->so_states);
	}
585
	state->state = fmode;
586 587
}

L
Linus Torvalds 已提交
588 589 590 591 592 593 594
static struct nfs4_state *
__nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs4_state *state;

	list_for_each_entry(state, &nfsi->open_states, inode_states) {
595
		if (state->owner != owner)
L
Linus Torvalds 已提交
596
			continue;
597
		if (atomic_inc_not_zero(&state->count))
L
Linus Torvalds 已提交
598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620
			return state;
	}
	return NULL;
}

static void
nfs4_free_open_state(struct nfs4_state *state)
{
	kfree(state);
}

struct nfs4_state *
nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
{
	struct nfs4_state *state, *new;
	struct nfs_inode *nfsi = NFS_I(inode);

	spin_lock(&inode->i_lock);
	state = __nfs4_find_state_byowner(inode, owner);
	spin_unlock(&inode->i_lock);
	if (state)
		goto out;
	new = nfs4_alloc_open_state();
621
	spin_lock(&owner->so_lock);
L
Linus Torvalds 已提交
622 623 624 625 626 627 628
	spin_lock(&inode->i_lock);
	state = __nfs4_find_state_byowner(inode, owner);
	if (state == NULL && new != NULL) {
		state = new;
		state->owner = owner;
		atomic_inc(&owner->so_count);
		list_add(&state->inode_states, &nfsi->open_states);
629 630
		ihold(inode);
		state->inode = inode;
L
Linus Torvalds 已提交
631
		spin_unlock(&inode->i_lock);
632 633 634 635
		/* Note: The reclaim code dictates that we add stateless
		 * and read-only stateids to the end of the list */
		list_add_tail(&state->open_states, &owner->so_states);
		spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
636 637
	} else {
		spin_unlock(&inode->i_lock);
638
		spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
639 640 641 642 643 644 645 646 647 648 649 650
		if (new)
			nfs4_free_open_state(new);
	}
out:
	return state;
}

void nfs4_put_open_state(struct nfs4_state *state)
{
	struct inode *inode = state->inode;
	struct nfs4_state_owner *owner = state->owner;

651
	if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
L
Linus Torvalds 已提交
652
		return;
653
	spin_lock(&inode->i_lock);
654
	list_del(&state->inode_states);
L
Linus Torvalds 已提交
655
	list_del(&state->open_states);
656 657
	spin_unlock(&inode->i_lock);
	spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
658 659 660 661 662 663
	iput(inode);
	nfs4_free_open_state(state);
	nfs4_put_state_owner(owner);
}

/*
664
 * Close the current file.
L
Linus Torvalds 已提交
665
 */
666
static void __nfs4_close(struct nfs4_state *state,
667
		fmode_t fmode, gfp_t gfp_mask, int wait)
L
Linus Torvalds 已提交
668 669
{
	struct nfs4_state_owner *owner = state->owner;
670
	int call_close = 0;
671
	fmode_t newstate;
L
Linus Torvalds 已提交
672 673 674

	atomic_inc(&owner->so_count);
	/* Protect against nfs4_find_state() */
675
	spin_lock(&owner->so_lock);
676
	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
677 678 679 680 681 682 683 684 685
		case FMODE_READ:
			state->n_rdonly--;
			break;
		case FMODE_WRITE:
			state->n_wronly--;
			break;
		case FMODE_READ|FMODE_WRITE:
			state->n_rdwr--;
	}
686
	newstate = FMODE_READ|FMODE_WRITE;
687
	if (state->n_rdwr == 0) {
688
		if (state->n_rdonly == 0) {
689
			newstate &= ~FMODE_READ;
690 691 692 693
			call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
			call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
		}
		if (state->n_wronly == 0) {
694
			newstate &= ~FMODE_WRITE;
695 696 697 698 699
			call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
			call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
		}
		if (newstate == 0)
			clear_bit(NFS_DELEGATED_STATE, &state->flags);
700
	}
701
	nfs4_state_set_mode_locked(state, newstate);
702
	spin_unlock(&owner->so_lock);
703

704
	if (!call_close) {
705 706
		nfs4_put_open_state(state);
		nfs4_put_state_owner(owner);
F
Fred Isaman 已提交
707 708 709
	} else {
		bool roc = pnfs_roc(state->inode);

710
		nfs4_do_close(state, gfp_mask, wait, roc);
F
Fred Isaman 已提交
711
	}
712 713
}

714
void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
715
{
716
	__nfs4_close(state, fmode, GFP_NOFS, 0);
717 718
}

719
void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
720
{
721
	__nfs4_close(state, fmode, GFP_KERNEL, 1);
L
Linus Torvalds 已提交
722 723 724 725 726 727 728
}

/*
 * Search the state->lock_states for an existing lock_owner
 * that is compatible with current->files
 */
static struct nfs4_lock_state *
729
__nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type)
L
Linus Torvalds 已提交
730 731 732
{
	struct nfs4_lock_state *pos;
	list_for_each_entry(pos, &state->lock_states, ls_locks) {
733
		if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
L
Linus Torvalds 已提交
734
			continue;
735 736 737 738 739 740 741 742 743
		switch (pos->ls_owner.lo_type) {
		case NFS4_POSIX_LOCK_TYPE:
			if (pos->ls_owner.lo_u.posix_owner != fl_owner)
				continue;
			break;
		case NFS4_FLOCK_LOCK_TYPE:
			if (pos->ls_owner.lo_u.flock_owner != fl_pid)
				continue;
		}
L
Linus Torvalds 已提交
744 745 746 747 748 749 750 751 752 753 754
		atomic_inc(&pos->ls_count);
		return pos;
	}
	return NULL;
}

/*
 * Return a compatible lock_state. If no initialized lock_state structure
 * exists, return an uninitialized one.
 *
 */
755
static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type)
L
Linus Torvalds 已提交
756 757
{
	struct nfs4_lock_state *lsp;
758
	struct nfs_server *server = state->owner->so_server;
L
Linus Torvalds 已提交
759

760
	lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
L
Linus Torvalds 已提交
761 762
	if (lsp == NULL)
		return NULL;
763
	nfs4_init_seqid_counter(&lsp->ls_seqid);
L
Linus Torvalds 已提交
764
	atomic_set(&lsp->ls_count, 1);
765
	lsp->ls_state = state;
766 767 768 769 770 771 772 773 774
	lsp->ls_owner.lo_type = type;
	switch (lsp->ls_owner.lo_type) {
	case NFS4_FLOCK_LOCK_TYPE:
		lsp->ls_owner.lo_u.flock_owner = fl_pid;
		break;
	case NFS4_POSIX_LOCK_TYPE:
		lsp->ls_owner.lo_u.posix_owner = fl_owner;
		break;
	default:
775
		goto out_free;
776
	}
777 778
	lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
	if (lsp->ls_seqid.owner_id < 0)
779
		goto out_free;
780
	INIT_LIST_HEAD(&lsp->ls_locks);
L
Linus Torvalds 已提交
781
	return lsp;
782 783 784
out_free:
	kfree(lsp);
	return NULL;
L
Linus Torvalds 已提交
785 786
}

787 788
static void nfs4_free_lock_state(struct nfs4_lock_state *lsp)
{
789
	struct nfs_server *server = lsp->ls_state->owner->so_server;
790

791
	ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
792
	nfs4_destroy_seqid_counter(&lsp->ls_seqid);
793 794 795
	kfree(lsp);
}

L
Linus Torvalds 已提交
796 797 798 799 800
/*
 * Return a compatible lock_state. If no initialized lock_state structure
 * exists, return an uninitialized one.
 *
 */
801
static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner, pid_t pid, unsigned int type)
L
Linus Torvalds 已提交
802
{
803
	struct nfs4_lock_state *lsp, *new = NULL;
L
Linus Torvalds 已提交
804
	
805 806
	for(;;) {
		spin_lock(&state->state_lock);
807
		lsp = __nfs4_find_lock_state(state, owner, pid, type);
808 809 810 811 812 813 814 815 816 817
		if (lsp != NULL)
			break;
		if (new != NULL) {
			list_add(&new->ls_locks, &state->lock_states);
			set_bit(LK_STATE_IN_USE, &state->flags);
			lsp = new;
			new = NULL;
			break;
		}
		spin_unlock(&state->state_lock);
818
		new = nfs4_alloc_lock_state(state, owner, pid, type);
819 820 821 822
		if (new == NULL)
			return NULL;
	}
	spin_unlock(&state->state_lock);
823 824
	if (new != NULL)
		nfs4_free_lock_state(new);
L
Linus Torvalds 已提交
825 826 827 828
	return lsp;
}

/*
829 830
 * Release reference to lock_state, and free it if we see that
 * it is no longer in use
L
Linus Torvalds 已提交
831
 */
832
void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
L
Linus Torvalds 已提交
833
{
834
	struct nfs4_state *state;
L
Linus Torvalds 已提交
835

836 837 838 839 840 841 842 843 844
	if (lsp == NULL)
		return;
	state = lsp->ls_state;
	if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
		return;
	list_del(&lsp->ls_locks);
	if (list_empty(&state->lock_states))
		clear_bit(LK_STATE_IN_USE, &state->flags);
	spin_unlock(&state->state_lock);
845 846
	if (lsp->ls_flags & NFS_LOCK_INITIALIZED)
		nfs4_release_lockowner(lsp);
847
	nfs4_free_lock_state(lsp);
L
Linus Torvalds 已提交
848 849
}

850
static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
L
Linus Torvalds 已提交
851
{
852
	struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
L
Linus Torvalds 已提交
853

854 855 856
	dst->fl_u.nfs4_fl.owner = lsp;
	atomic_inc(&lsp->ls_count);
}
L
Linus Torvalds 已提交
857

858
static void nfs4_fl_release_lock(struct file_lock *fl)
L
Linus Torvalds 已提交
859
{
860
	nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
L
Linus Torvalds 已提交
861 862
}

863
static const struct file_lock_operations nfs4_fl_lock_ops = {
864 865 866 867 868
	.fl_copy_lock = nfs4_fl_copy_lock,
	.fl_release_private = nfs4_fl_release_lock,
};

int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
L
Linus Torvalds 已提交
869
{
870 871 872 873
	struct nfs4_lock_state *lsp;

	if (fl->fl_ops != NULL)
		return 0;
874 875 876 877 878 879
	if (fl->fl_flags & FL_POSIX)
		lsp = nfs4_get_lock_state(state, fl->fl_owner, 0, NFS4_POSIX_LOCK_TYPE);
	else if (fl->fl_flags & FL_FLOCK)
		lsp = nfs4_get_lock_state(state, 0, fl->fl_pid, NFS4_FLOCK_LOCK_TYPE);
	else
		return -EINVAL;
880 881 882 883 884
	if (lsp == NULL)
		return -ENOMEM;
	fl->fl_u.nfs4_fl.owner = lsp;
	fl->fl_ops = &nfs4_fl_lock_ops;
	return 0;
L
Linus Torvalds 已提交
885 886
}

887 888 889
/*
 * Byte-range lock aware utility to initialize the stateid of read/write
 * requests.
L
Linus Torvalds 已提交
890
 */
891
void nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid)
L
Linus Torvalds 已提交
892
{
893
	struct nfs4_lock_state *lsp;
894
	int seq;
L
Linus Torvalds 已提交
895

896 897 898 899
	do {
		seq = read_seqbegin(&state->seqlock);
		memcpy(dst, &state->stateid, sizeof(*dst));
	} while (read_seqretry(&state->seqlock, seq));
900 901
	if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
		return;
L
Linus Torvalds 已提交
902

903
	spin_lock(&state->state_lock);
904
	lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
905 906 907
	if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
		memcpy(dst, &lsp->ls_stateid, sizeof(*dst));
	spin_unlock(&state->state_lock);
L
Linus Torvalds 已提交
908 909 910
	nfs4_put_lock_state(lsp);
}

911
struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
912 913 914
{
	struct nfs_seqid *new;

915
	new = kmalloc(sizeof(*new), gfp_mask);
916 917
	if (new != NULL) {
		new->sequence = counter;
918
		INIT_LIST_HEAD(&new->list);
919
		new->task = NULL;
920 921 922 923
	}
	return new;
}

924
void nfs_release_seqid(struct nfs_seqid *seqid)
L
Linus Torvalds 已提交
925
{
926
	struct nfs_seqid_counter *sequence;
927

928 929 930 931 932 933 934 935 936 937 938
	if (list_empty(&seqid->list))
		return;
	sequence = seqid->sequence;
	spin_lock(&sequence->lock);
	list_del_init(&seqid->list);
	if (!list_empty(&sequence->list)) {
		struct nfs_seqid *next;

		next = list_first_entry(&sequence->list,
				struct nfs_seqid, list);
		rpc_wake_up_queued_task(&sequence->wait, next->task);
939
	}
940
	spin_unlock(&sequence->lock);
941 942 943 944 945
}

void nfs_free_seqid(struct nfs_seqid *seqid)
{
	nfs_release_seqid(seqid);
946
	kfree(seqid);
L
Linus Torvalds 已提交
947 948 949
}

/*
950 951 952 953
 * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
 * failed with a seqid incrementing error -
 * see comments nfs_fs.h:seqid_mutating_error()
 */
954
static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
955
{
956
	BUG_ON(list_first_entry(&seqid->sequence->list, struct nfs_seqid, list) != seqid);
957 958 959 960
	switch (status) {
		case 0:
			break;
		case -NFS4ERR_BAD_SEQID:
961 962 963
			if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
				return;
			printk(KERN_WARNING "NFS: v4 server returned a bad"
D
Dan Muntz 已提交
964 965
					" sequence-id error on an"
					" unconfirmed sequence %p!\n",
966
					seqid->sequence);
967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984
		case -NFS4ERR_STALE_CLIENTID:
		case -NFS4ERR_STALE_STATEID:
		case -NFS4ERR_BAD_STATEID:
		case -NFS4ERR_BADXDR:
		case -NFS4ERR_RESOURCE:
		case -NFS4ERR_NOFILEHANDLE:
			/* Non-seqid mutating errors */
			return;
	};
	/*
	 * Note: no locking needed as we are guaranteed to be first
	 * on the sequence list
	 */
	seqid->sequence->counter++;
}

void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
{
985 986 987 988 989
	struct nfs4_state_owner *sp = container_of(seqid->sequence,
					struct nfs4_state_owner, so_seqid);
	struct nfs_server *server = sp->so_server;

	if (status == -NFS4ERR_BAD_SEQID)
L
Linus Torvalds 已提交
990
		nfs4_drop_state_owner(sp);
991 992
	if (!nfs4_has_session(server->nfs_client))
		nfs_increment_seqid(status, seqid);
993 994 995 996 997 998 999 1000 1001
}

/*
 * Increment the seqid if the LOCK/LOCKU succeeded, or
 * failed with a seqid incrementing error -
 * see comments nfs_fs.h:seqid_mutating_error()
 */
void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
{
1002
	nfs_increment_seqid(status, seqid);
1003 1004 1005 1006
}

int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
{
1007
	struct nfs_seqid_counter *sequence = seqid->sequence;
1008 1009 1010
	int status = 0;

	spin_lock(&sequence->lock);
1011
	seqid->task = task;
1012 1013 1014 1015
	if (list_empty(&seqid->list))
		list_add_tail(&seqid->list, &sequence->list);
	if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
		goto unlock;
1016
	rpc_sleep_on(&sequence->wait, task, NULL);
1017 1018
	status = -EAGAIN;
unlock:
1019 1020
	spin_unlock(&sequence->lock);
	return status;
L
Linus Torvalds 已提交
1021 1022
}

1023
static int nfs4_run_state_manager(void *);
L
Linus Torvalds 已提交
1024

1025
static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
T
Trond Myklebust 已提交
1026 1027
{
	smp_mb__before_clear_bit();
1028
	clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
T
Trond Myklebust 已提交
1029
	smp_mb__after_clear_bit();
1030
	wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
T
Trond Myklebust 已提交
1031 1032 1033
	rpc_wake_up(&clp->cl_rpcwaitq);
}

L
Linus Torvalds 已提交
1034
/*
1035
 * Schedule the nfs_client asynchronous state management routine
L
Linus Torvalds 已提交
1036
 */
1037
void nfs4_schedule_state_manager(struct nfs_client *clp)
L
Linus Torvalds 已提交
1038
{
1039
	struct task_struct *task;
1040
	char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
L
Linus Torvalds 已提交
1041

1042 1043
	if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
		return;
1044 1045
	__module_get(THIS_MODULE);
	atomic_inc(&clp->cl_count);
1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061

	/* The rcu_read_lock() is not strictly necessary, as the state
	 * manager is the only thread that ever changes the rpc_xprt
	 * after it's initialized.  At this point, we're single threaded. */
	rcu_read_lock();
	snprintf(buf, sizeof(buf), "%s-manager",
			rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR));
	rcu_read_unlock();
	task = kthread_run(nfs4_run_state_manager, clp, buf);
	if (IS_ERR(task)) {
		printk(KERN_ERR "%s: kthread_run: %ld\n",
			__func__, PTR_ERR(task));
		nfs4_clear_state_manager_bit(clp);
		nfs_put_client(clp);
		module_put(THIS_MODULE);
	}
L
Linus Torvalds 已提交
1062 1063 1064
}

/*
1065
 * Schedule a lease recovery attempt
L
Linus Torvalds 已提交
1066
 */
1067
void nfs4_schedule_lease_recovery(struct nfs_client *clp)
L
Linus Torvalds 已提交
1068 1069 1070
{
	if (!clp)
		return;
1071 1072
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1073
	nfs4_schedule_state_manager(clp);
L
Linus Torvalds 已提交
1074 1075
}

1076 1077 1078 1079 1080 1081
void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
{
	nfs_handle_cb_pathdown(clp);
	nfs4_schedule_state_manager(clp);
}

1082
static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1083 1084 1085 1086 1087 1088 1089 1090
{

	set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
	/* Don't recover state that expired before the reboot */
	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
		clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
		return 0;
	}
1091
	set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1092 1093 1094 1095
	set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
	return 1;
}

1096
static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1097 1098 1099
{
	set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
	clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1100
	set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1101 1102 1103 1104
	set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
	return 1;
}

1105 1106 1107 1108 1109 1110 1111 1112
void nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
{
	struct nfs_client *clp = server->nfs_client;

	nfs4_state_mark_reclaim_nograce(clp, state);
	nfs4_schedule_state_manager(clp);
}

1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139
void nfs_inode_find_state_and_recover(struct inode *inode,
		const nfs4_stateid *stateid)
{
	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_open_context *ctx;
	struct nfs4_state *state;
	bool found = false;

	spin_lock(&inode->i_lock);
	list_for_each_entry(ctx, &nfsi->open_files, list) {
		state = ctx->state;
		if (state == NULL)
			continue;
		if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
			continue;
		if (memcmp(state->stateid.data, stateid->data, sizeof(state->stateid.data)) != 0)
			continue;
		nfs4_state_mark_reclaim_nograce(clp, state);
		found = true;
	}
	spin_unlock(&inode->i_lock);
	if (found)
		nfs4_schedule_state_manager(clp);
}


1140
static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1141 1142
{
	struct inode *inode = state->inode;
1143
	struct nfs_inode *nfsi = NFS_I(inode);
L
Linus Torvalds 已提交
1144 1145 1146
	struct file_lock *fl;
	int status = 0;

1147 1148 1149 1150
	if (inode->i_flock == NULL)
		return 0;

	/* Guard against delegation returns and new lock/unlock calls */
1151
	down_write(&nfsi->rwsem);
1152
	/* Protect inode->i_flock using the BKL */
1153
	lock_flocks();
H
Harvey Harrison 已提交
1154
	for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1155
		if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
L
Linus Torvalds 已提交
1156
			continue;
1157
		if (nfs_file_open_context(fl->fl_file)->state != state)
L
Linus Torvalds 已提交
1158
			continue;
1159
		unlock_flocks();
L
Linus Torvalds 已提交
1160 1161
		status = ops->recover_lock(state, fl);
		switch (status) {
1162 1163 1164 1165 1166 1167 1168 1169 1170
			case 0:
				break;
			case -ESTALE:
			case -NFS4ERR_ADMIN_REVOKED:
			case -NFS4ERR_STALE_STATEID:
			case -NFS4ERR_BAD_STATEID:
			case -NFS4ERR_EXPIRED:
			case -NFS4ERR_NO_GRACE:
			case -NFS4ERR_STALE_CLIENTID:
1171 1172 1173 1174
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1175
				goto out;
L
Linus Torvalds 已提交
1176
			default:
1177 1178
				printk(KERN_ERR "NFS: %s: unhandled error %d. "
					"Zeroing state\n", __func__, status);
1179 1180
			case -ENOMEM:
			case -NFS4ERR_DENIED:
L
Linus Torvalds 已提交
1181 1182
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1183
				/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1184
				status = 0;
L
Linus Torvalds 已提交
1185
		}
1186
		lock_flocks();
L
Linus Torvalds 已提交
1187
	}
1188
	unlock_flocks();
1189
out:
1190
	up_write(&nfsi->rwsem);
L
Linus Torvalds 已提交
1191 1192 1193
	return status;
}

1194
static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207
{
	struct nfs4_state *state;
	struct nfs4_lock_state *lock;
	int status = 0;

	/* Note: we rely on the sp->so_states list being ordered 
	 * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
	 * states first.
	 * This is needed to ensure that the server won't give us any
	 * read delegations that we have to return if, say, we are
	 * recovering after a network partition or a reboot from a
	 * server that doesn't support a grace period.
	 */
1208 1209
restart:
	spin_lock(&sp->so_lock);
L
Linus Torvalds 已提交
1210
	list_for_each_entry(state, &sp->so_states, open_states) {
1211 1212
		if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
			continue;
L
Linus Torvalds 已提交
1213 1214
		if (state->state == 0)
			continue;
1215 1216
		atomic_inc(&state->count);
		spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1217 1218
		status = ops->recover_open(sp, state);
		if (status >= 0) {
1219 1220
			status = nfs4_reclaim_locks(state, ops);
			if (status >= 0) {
1221
				spin_lock(&state->state_lock);
1222 1223
				list_for_each_entry(lock, &state->lock_states, ls_locks) {
					if (!(lock->ls_flags & NFS_LOCK_INITIALIZED))
1224 1225
						printk("NFS: %s: Lock reclaim "
							"failed!\n", __func__);
1226
				}
1227
				spin_unlock(&state->state_lock);
1228 1229
				nfs4_put_open_state(state);
				goto restart;
L
Linus Torvalds 已提交
1230 1231 1232 1233
			}
		}
		switch (status) {
			default:
1234 1235
				printk(KERN_ERR "NFS: %s: unhandled error %d. "
					"Zeroing state\n", __func__, status);
L
Linus Torvalds 已提交
1236
			case -ENOENT:
1237
			case -ENOMEM:
1238
			case -ESTALE:
L
Linus Torvalds 已提交
1239 1240 1241 1242 1243 1244 1245 1246 1247
				/*
				 * Open state on this file cannot be recovered
				 * All we can do is revert to using the zero stateid.
				 */
				memset(state->stateid.data, 0,
					sizeof(state->stateid.data));
				/* Mark the file as being 'closed' */
				state->state = 0;
				break;
1248 1249 1250 1251 1252 1253 1254 1255
			case -EKEYEXPIRED:
				/*
				 * User RPCSEC_GSS context has expired.
				 * We cannot recover this stateid now, so
				 * skip it and allow recovery thread to
				 * proceed.
				 */
				break;
1256 1257 1258
			case -NFS4ERR_ADMIN_REVOKED:
			case -NFS4ERR_STALE_STATEID:
			case -NFS4ERR_BAD_STATEID:
1259 1260
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1261
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1262
				break;
L
Linus Torvalds 已提交
1263 1264
			case -NFS4ERR_EXPIRED:
			case -NFS4ERR_NO_GRACE:
1265
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
L
Linus Torvalds 已提交
1266
			case -NFS4ERR_STALE_CLIENTID:
1267 1268 1269 1270
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
L
Linus Torvalds 已提交
1271 1272
				goto out_err;
		}
1273 1274
		nfs4_put_open_state(state);
		goto restart;
L
Linus Torvalds 已提交
1275
	}
1276
	spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1277 1278
	return 0;
out_err:
1279
	nfs4_put_open_state(state);
L
Linus Torvalds 已提交
1280 1281 1282
	return status;
}

1283 1284 1285 1286 1287 1288 1289 1290
static void nfs4_clear_open_state(struct nfs4_state *state)
{
	struct nfs4_lock_state *lock;

	clear_bit(NFS_DELEGATED_STATE, &state->flags);
	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1291
	spin_lock(&state->state_lock);
1292 1293 1294 1295
	list_for_each_entry(lock, &state->lock_states, ls_locks) {
		lock->ls_seqid.flags = 0;
		lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
	}
1296
	spin_unlock(&state->state_lock);
1297 1298
}

1299 1300
static void nfs4_reset_seqids(struct nfs_server *server,
	int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1301
{
1302
	struct nfs_client *clp = server->nfs_client;
1303
	struct nfs4_state_owner *sp;
1304
	struct rb_node *pos;
1305 1306
	struct nfs4_state *state;

1307 1308 1309 1310 1311
	spin_lock(&clp->cl_lock);
	for (pos = rb_first(&server->state_owners);
	     pos != NULL;
	     pos = rb_next(pos)) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1312
		sp->so_seqid.flags = 0;
1313
		spin_lock(&sp->so_lock);
1314
		list_for_each_entry(state, &sp->so_states, open_states) {
1315 1316
			if (mark_reclaim(clp, state))
				nfs4_clear_open_state(state);
1317
		}
1318
		spin_unlock(&sp->so_lock);
1319
	}
1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331
	spin_unlock(&clp->cl_lock);
}

static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp,
	int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
{
	struct nfs_server *server;

	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
		nfs4_reset_seqids(server, mark_reclaim);
	rcu_read_unlock();
1332 1333
}

1334 1335 1336 1337 1338 1339 1340
static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
{
	/* Mark all delegations for reclaim */
	nfs_delegation_mark_reclaim(clp);
	nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
}

1341 1342 1343 1344 1345 1346 1347 1348
static void nfs4_reclaim_complete(struct nfs_client *clp,
				 const struct nfs4_state_recovery_ops *ops)
{
	/* Notify the server we're done reclaiming our state */
	if (ops->reclaim_complete)
		(void)ops->reclaim_complete(clp);
}

1349
static void nfs4_clear_reclaim_server(struct nfs_server *server)
1350
{
1351
	struct nfs_client *clp = server->nfs_client;
1352 1353 1354 1355
	struct nfs4_state_owner *sp;
	struct rb_node *pos;
	struct nfs4_state *state;

1356 1357 1358 1359 1360
	spin_lock(&clp->cl_lock);
	for (pos = rb_first(&server->state_owners);
	     pos != NULL;
	     pos = rb_next(pos)) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1361 1362
		spin_lock(&sp->so_lock);
		list_for_each_entry(state, &sp->so_states, open_states) {
1363 1364
			if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
						&state->flags))
1365 1366 1367 1368 1369
				continue;
			nfs4_state_mark_reclaim_nograce(clp, state);
		}
		spin_unlock(&sp->so_lock);
	}
1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383
	spin_unlock(&clp->cl_lock);
}

static int nfs4_state_clear_reclaim_reboot(struct nfs_client *clp)
{
	struct nfs_server *server;

	if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
		return 0;

	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
		nfs4_clear_reclaim_server(server);
	rcu_read_unlock();
1384 1385

	nfs_delegation_reap_unclaimed(clp);
1386 1387 1388 1389 1390 1391 1392 1393
	return 1;
}

static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
{
	if (!nfs4_state_clear_reclaim_reboot(clp))
		return;
	nfs4_reclaim_complete(clp, clp->cl_mvops->reboot_recovery_ops);
1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407
}

static void nfs_delegation_clear_all(struct nfs_client *clp)
{
	nfs_delegation_mark_reclaim(clp);
	nfs_delegation_reap_unclaimed(clp);
}

static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
{
	nfs_delegation_clear_all(clp);
	nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
}

1408 1409 1410 1411 1412 1413 1414 1415
static void nfs4_warn_keyexpired(const char *s)
{
	printk_ratelimited(KERN_WARNING "Error: state manager"
			" encountered RPCSEC_GSS session"
			" expired against NFSv4 server %s.\n",
			s);
}

1416
static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1417 1418
{
	switch (error) {
1419 1420
		case 0:
			break;
1421
		case -NFS4ERR_CB_PATH_DOWN:
1422
			nfs_handle_cb_pathdown(clp);
1423
			break;
1424 1425
		case -NFS4ERR_NO_GRACE:
			nfs4_state_end_reclaim_reboot(clp);
1426
			break;
1427 1428 1429
		case -NFS4ERR_STALE_CLIENTID:
		case -NFS4ERR_LEASE_MOVED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1430
			nfs4_state_clear_reclaim_reboot(clp);
1431 1432 1433 1434 1435
			nfs4_state_start_reclaim_reboot(clp);
			break;
		case -NFS4ERR_EXPIRED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
			nfs4_state_start_reclaim_nograce(clp);
1436
			break;
1437 1438 1439 1440 1441 1442 1443
		case -NFS4ERR_BADSESSION:
		case -NFS4ERR_BADSLOT:
		case -NFS4ERR_BAD_HIGH_SLOT:
		case -NFS4ERR_DEADSESSION:
		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
		case -NFS4ERR_SEQ_FALSE_RETRY:
		case -NFS4ERR_SEQ_MISORDERED:
1444
			set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1445
			/* Zero session reset errors */
1446
			break;
1447 1448 1449
		case -EKEYEXPIRED:
			/* Nothing we can do */
			nfs4_warn_keyexpired(clp->cl_hostname);
1450 1451 1452
			break;
		default:
			return error;
1453
	}
1454
	return 0;
1455 1456
}

1457
static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1458
{
1459 1460
	struct nfs4_state_owner *sp;
	struct nfs_server *server;
1461
	struct rb_node *pos;
L
Linus Torvalds 已提交
1462 1463
	int status = 0;

1464
restart:
1465 1466
	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1467
		nfs4_purge_state_owners(server);
1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487
		spin_lock(&clp->cl_lock);
		for (pos = rb_first(&server->state_owners);
		     pos != NULL;
		     pos = rb_next(pos)) {
			sp = rb_entry(pos,
				struct nfs4_state_owner, so_server_node);
			if (!test_and_clear_bit(ops->owner_flag_bit,
							&sp->so_flags))
				continue;
			atomic_inc(&sp->so_count);
			spin_unlock(&clp->cl_lock);
			rcu_read_unlock();

			status = nfs4_reclaim_open_state(sp, ops);
			if (status < 0) {
				set_bit(ops->owner_flag_bit, &sp->so_flags);
				nfs4_put_state_owner(sp);
				return nfs4_recovery_handle_error(clp, status);
			}

1488
			nfs4_put_state_owner(sp);
1489
			goto restart;
1490
		}
1491
		spin_unlock(&clp->cl_lock);
1492
	}
1493
	rcu_read_unlock();
1494 1495 1496 1497 1498 1499
	return status;
}

static int nfs4_check_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1500 1501
	const struct nfs4_state_maintenance_ops *ops =
		clp->cl_mvops->state_renewal_ops;
1502
	int status;
L
Linus Torvalds 已提交
1503

1504 1505 1506
	/* Is the client already known to have an expired lease? */
	if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		return 0;
1507 1508 1509
	spin_lock(&clp->cl_lock);
	cred = ops->get_state_renewal_cred_locked(clp);
	spin_unlock(&clp->cl_lock);
1510 1511
	if (cred == NULL) {
		cred = nfs4_get_setclientid_cred(clp);
1512
		status = -ENOKEY;
1513 1514
		if (cred == NULL)
			goto out;
1515
	}
1516
	status = ops->renew_lease(clp, cred);
1517 1518
	put_rpccred(cred);
out:
1519
	return nfs4_recovery_handle_error(clp, status);
1520 1521 1522 1523 1524
}

static int nfs4_reclaim_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1525 1526
	const struct nfs4_state_recovery_ops *ops =
		clp->cl_mvops->reboot_recovery_ops;
1527 1528
	int status = -ENOENT;

1529
	cred = ops->get_clid_cred(clp);
1530
	if (cred != NULL) {
1531
		status = ops->establish_clid(clp, cred);
1532
		put_rpccred(cred);
1533 1534 1535
		/* Handle case where the user hasn't set up machine creds */
		if (status == -EACCES && cred == clp->cl_machine_cred) {
			nfs4_clear_machine_cred(clp);
1536
			status = -EAGAIN;
1537
		}
1538 1539
		if (status == -NFS4ERR_MINOR_VERS_MISMATCH)
			status = -EPROTONOSUPPORT;
1540
	}
1541 1542 1543
	return status;
}

1544
#ifdef CONFIG_NFS_V4_1
1545 1546
void nfs4_schedule_session_recovery(struct nfs4_session *session)
{
1547 1548 1549 1550
	struct nfs_client *clp = session->clp;

	set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
	nfs4_schedule_lease_recovery(clp);
1551
}
1552
EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
1553

A
Andy Adamson 已提交
1554 1555 1556
void nfs41_handle_recall_slot(struct nfs_client *clp)
{
	set_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
1557
	nfs4_schedule_state_manager(clp);
A
Andy Adamson 已提交
1558 1559
}

1560 1561 1562 1563 1564
static void nfs4_reset_all_state(struct nfs_client *clp)
{
	if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
		clp->cl_boot_time = CURRENT_TIME;
		nfs4_state_start_reclaim_nograce(clp);
1565
		nfs4_schedule_state_manager(clp);
1566 1567 1568 1569 1570 1571 1572
	}
}

static void nfs41_handle_server_reboot(struct nfs_client *clp)
{
	if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
		nfs4_state_start_reclaim_reboot(clp);
1573
		nfs4_schedule_state_manager(clp);
1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592
	}
}

static void nfs41_handle_state_revoked(struct nfs_client *clp)
{
	/* Temporary */
	nfs4_reset_all_state(clp);
}

static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
{
	/* This will need to handle layouts too */
	nfs_expire_all_delegations(clp);
}

static void nfs41_handle_cb_path_down(struct nfs_client *clp)
{
	nfs_expire_all_delegations(clp);
	if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
1593
		nfs4_schedule_state_manager(clp);
1594 1595
}

1596 1597 1598 1599
void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
{
	if (!flags)
		return;
1600
	if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
1601
		nfs41_handle_server_reboot(clp);
1602
	if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
1603 1604
			    SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
			    SEQ4_STATUS_ADMIN_STATE_REVOKED |
1605 1606
			    SEQ4_STATUS_LEASE_MOVED))
		nfs41_handle_state_revoked(clp);
1607
	if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
1608
		nfs41_handle_recallable_state_revoked(clp);
1609
	if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
1610 1611
			    SEQ4_STATUS_BACKCHANNEL_FAULT |
			    SEQ4_STATUS_CB_PATH_DOWN_SESSION))
1612
		nfs41_handle_cb_path_down(clp);
1613 1614
}

1615 1616 1617 1618
static int nfs4_reset_session(struct nfs_client *clp)
{
	int status;

1619
	nfs4_begin_drain_session(clp);
1620 1621 1622
	status = nfs4_proc_destroy_session(clp->cl_session);
	if (status && status != -NFS4ERR_BADSESSION &&
	    status != -NFS4ERR_DEADSESSION) {
1623
		status = nfs4_recovery_handle_error(clp, status);
1624 1625 1626 1627
		goto out;
	}

	memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
1628
	status = nfs4_proc_create_session(clp);
1629
	if (status) {
1630
		status = nfs4_recovery_handle_error(clp, status);
1631 1632
		goto out;
	}
1633
	clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1634 1635
	/* create_session negotiated new slot table */
	clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
1636

1637 1638
	 /* Let the state manager reestablish state */
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1639
		nfs41_setup_state_renewal(clp);
1640
out:
1641 1642
	return status;
}
1643

A
Andy Adamson 已提交
1644 1645 1646 1647 1648 1649 1650 1651 1652
static int nfs4_recall_slot(struct nfs_client *clp)
{
	struct nfs4_slot_table *fc_tbl = &clp->cl_session->fc_slot_table;
	struct nfs4_channel_attrs *fc_attrs = &clp->cl_session->fc_attrs;
	struct nfs4_slot *new, *old;
	int i;

	nfs4_begin_drain_session(clp);
	new = kmalloc(fc_tbl->target_max_slots * sizeof(struct nfs4_slot),
1653
		      GFP_NOFS);
A
Andy Adamson 已提交
1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671
        if (!new)
		return -ENOMEM;

	spin_lock(&fc_tbl->slot_tbl_lock);
	for (i = 0; i < fc_tbl->target_max_slots; i++)
		new[i].seq_nr = fc_tbl->slots[i].seq_nr;
	old = fc_tbl->slots;
	fc_tbl->slots = new;
	fc_tbl->max_slots = fc_tbl->target_max_slots;
	fc_tbl->target_max_slots = 0;
	fc_attrs->max_reqs = fc_tbl->max_slots;
	spin_unlock(&fc_tbl->slot_tbl_lock);

	kfree(old);
	nfs4_end_drain_session(clp);
	return 0;
}

1672
#else /* CONFIG_NFS_V4_1 */
1673
static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
1674
static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
A
Andy Adamson 已提交
1675
static int nfs4_recall_slot(struct nfs_client *clp) { return 0; }
1676 1677
#endif /* CONFIG_NFS_V4_1 */

1678 1679 1680 1681 1682
/* Set NFS4CLNT_LEASE_EXPIRED for all v4.0 errors and for recoverable errors
 * on EXCHANGE_ID for v4.1
 */
static void nfs4_set_lease_expired(struct nfs_client *clp, int status)
{
1683 1684 1685 1686 1687 1688
	switch (status) {
	case -NFS4ERR_CLID_INUSE:
	case -NFS4ERR_STALE_CLIENTID:
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
		break;
	case -NFS4ERR_DELAY:
1689
	case -ETIMEDOUT:
1690 1691 1692
	case -EAGAIN:
		ssleep(1);
		break;
1693

1694 1695 1696 1697 1698 1699
	case -EKEYEXPIRED:
		nfs4_warn_keyexpired(clp->cl_hostname);
	case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
				 * in nfs4_exchange_id */
	default:
		return;
1700 1701 1702 1703
	}
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
}

1704
static void nfs4_state_manager(struct nfs_client *clp)
1705 1706 1707 1708
{
	int status = 0;

	/* Ensure exclusive access to NFSv4 state */
1709
	do {
1710 1711 1712 1713
		if (test_and_clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
			/* We're going to have to re-establish a clientid */
			status = nfs4_reclaim_lease(clp);
			if (status) {
1714
				nfs4_set_lease_expired(clp, status);
1715 1716
				if (test_bit(NFS4CLNT_LEASE_EXPIRED,
							&clp->cl_state))
1717
					continue;
1718 1719 1720
				if (clp->cl_cons_state ==
							NFS_CS_SESSION_INITING)
					nfs_mark_client_ready(clp, status);
1721 1722
				goto out_error;
			}
1723
			clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1724 1725 1726 1727 1728 1729 1730 1731

			if (test_and_clear_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH,
					       &clp->cl_state))
				nfs4_state_start_reclaim_nograce(clp);
			else
				set_bit(NFS4CLNT_RECLAIM_REBOOT,
					&clp->cl_state);

1732
			pnfs_destroy_all_layouts(clp);
1733 1734 1735 1736
		}

		if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
			status = nfs4_check_lease(clp);
1737 1738
			if (status < 0)
				goto out_error;
1739
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1740
				continue;
1741
		}
1742

1743
		/* Initialize or reset the session */
1744
		if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)
1745
		   && nfs4_has_session(clp)) {
1746
			status = nfs4_reset_session(clp);
1747 1748 1749
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
				continue;
			if (status < 0)
1750 1751
				goto out_error;
		}
1752

1753
		/* First recover reboot state... */
1754
		if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
1755
			status = nfs4_do_reclaim(clp,
1756
				clp->cl_mvops->reboot_recovery_ops);
1757
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1758
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
1759
				continue;
1760
			nfs4_state_end_reclaim_reboot(clp);
1761 1762 1763 1764
			if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
				continue;
			if (status < 0)
				goto out_error;
1765 1766
		}

1767 1768
		/* Now recover expired state... */
		if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
1769
			status = nfs4_do_reclaim(clp,
1770
				clp->cl_mvops->nograce_recovery_ops);
1771
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1772
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
1773 1774 1775
			    test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
				continue;
			if (status < 0)
1776
				goto out_error;
L
Linus Torvalds 已提交
1777
		}
1778

1779
		nfs4_end_drain_session(clp);
1780 1781 1782 1783
		if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
			nfs_client_return_marked_delegations(clp);
			continue;
		}
A
Andy Adamson 已提交
1784 1785 1786 1787 1788 1789 1790 1791 1792
		/* Recall session slots */
		if (test_and_clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state)
		   && nfs4_has_session(clp)) {
			status = nfs4_recall_slot(clp);
			if (status < 0)
				goto out_error;
			continue;
		}

1793 1794

		nfs4_clear_state_manager_bit(clp);
1795 1796 1797 1798 1799
		/* Did we race with an attempt to give us more work? */
		if (clp->cl_state == 0)
			break;
		if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
			break;
1800
	} while (atomic_read(&clp->cl_count) > 1);
1801
	return;
L
Linus Torvalds 已提交
1802
out_error:
1803
	printk(KERN_WARNING "NFS: state manager failed on NFSv4 server %s"
1804
			" with error %d\n", clp->cl_hostname, -status);
1805
	nfs4_end_drain_session(clp);
1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817
	nfs4_clear_state_manager_bit(clp);
}

static int nfs4_run_state_manager(void *ptr)
{
	struct nfs_client *clp = ptr;

	allow_signal(SIGKILL);
	nfs4_state_manager(clp);
	nfs_put_client(clp);
	module_put_and_exit(0);
	return 0;
L
Linus Torvalds 已提交
1818 1819 1820 1821 1822 1823 1824
}

/*
 * Local variables:
 *  c-basic-offset: 8
 * End:
 */