nfs4state.c 52.5 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

C
Chuck Lever 已提交
60 61
#define NFSDBG_FACILITY		NFSDBG_STATE

L
Linus Torvalds 已提交
62 63
#define OPENOWNER_POOL_SIZE	8

64
const nfs4_stateid zero_stateid;
L
Linus Torvalds 已提交
65 66 67

static LIST_HEAD(nfs4_clientid_list);

68
int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
L
Linus Torvalds 已提交
69
{
70 71 72 73
	struct nfs4_setclientid_res clid = {
		.clientid = clp->cl_clientid,
		.confirm = clp->cl_confirm,
	};
74 75 76
	unsigned short port;
	int status;

77 78
	if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
		goto do_confirm;
79 80 81 82
	port = nfs_callback_tcpport;
	if (clp->cl_addr.ss_family == AF_INET6)
		port = nfs_callback_tcpport6;

83 84 85
	status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
	if (status != 0)
		goto out;
86 87 88 89
	clp->cl_clientid = clid.clientid;
	clp->cl_confirm = clid.confirm;
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
do_confirm:
90 91 92
	status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
	if (status != 0)
		goto out;
93
	clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
94 95
	nfs4_schedule_state_renewal(clp);
out:
L
Linus Torvalds 已提交
96 97 98
	return status;
}

99
struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119
{
	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);
}

120 121
static struct rpc_cred *
nfs4_get_renew_cred_server_locked(struct nfs_server *server)
122
{
123
	struct rpc_cred *cred = NULL;
124
	struct nfs4_state_owner *sp;
125
	struct rb_node *pos;
126

127 128 129 130
	for (pos = rb_first(&server->state_owners);
	     pos != NULL;
	     pos = rb_next(pos)) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
131 132 133 134 135 136 137 138
		if (list_empty(&sp->so_states))
			continue;
		cred = get_rpccred(sp->so_cred);
		break;
	}
	return cred;
}

139 140 141 142 143 144 145 146 147 148 149 150
/**
 * 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;

151 152 153 154 155
	/* Use machine credentials if available */
	cred = nfs4_get_machine_cred_locked(clp);
	if (cred != NULL)
		goto out;

156 157 158 159 160 161 162
	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();
163 164

out:
165 166 167
	return cred;
}

168 169
#if defined(CONFIG_NFS_V4_1)

R
Ricardo Labiaga 已提交
170 171 172 173 174
static int nfs41_setup_state_renewal(struct nfs_client *clp)
{
	int status;
	struct nfs_fsinfo fsinfo;

175 176 177 178 179
	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
180 181 182 183 184 185 186 187 188 189 190 191 192 193
	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;
}

194 195 196 197 198
/*
 * 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.
 */
199
static void nfs4_end_drain_session(struct nfs_client *clp)
200
{
201
	struct nfs4_session *ses = clp->cl_session;
202
	struct nfs4_slot_table *tbl;
203 204
	int max_slots;

205 206
	if (ses == NULL)
		return;
207
	tbl = &ses->fc_slot_table;
208
	if (test_and_clear_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
209 210
		spin_lock(&tbl->slot_tbl_lock);
		max_slots = tbl->max_slots;
211
		while (max_slots--) {
212 213 214
			if (rpc_wake_up_first(&tbl->slot_tbl_waitq,
						nfs4_set_task_privileged,
						NULL) == NULL)
215 216
				break;
		}
217
		spin_unlock(&tbl->slot_tbl_lock);
218
	}
219 220
}

221
static int nfs4_wait_on_slot_tbl(struct nfs4_slot_table *tbl)
222 223
{
	spin_lock(&tbl->slot_tbl_lock);
224
	if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
225
		INIT_COMPLETION(tbl->complete);
226
		spin_unlock(&tbl->slot_tbl_lock);
227
		return wait_for_completion_interruptible(&tbl->complete);
228 229 230 231 232
	}
	spin_unlock(&tbl->slot_tbl_lock);
	return 0;
}

233 234 235 236 237 238 239 240 241 242 243 244 245 246
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);
}

247 248 249 250 251 252 253 254 255 256
static void nfs41_finish_session_reset(struct nfs_client *clp)
{
	clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
	clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
	/* create_session negotiated new slot table */
	clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
	clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
	nfs41_setup_state_renewal(clp);
}

257 258 259 260
int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
{
	int status;

261 262
	if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
		goto do_confirm;
263
	nfs4_begin_drain_session(clp);
264
	status = nfs4_proc_exchange_id(clp, cred);
R
Ricardo Labiaga 已提交
265 266
	if (status != 0)
		goto out;
267 268
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
do_confirm:
269
	status = nfs4_proc_create_session(clp, cred);
R
Ricardo Labiaga 已提交
270 271
	if (status != 0)
		goto out;
272
	nfs41_finish_session_reset(clp);
R
Ricardo Labiaga 已提交
273 274
	nfs_mark_client_ready(clp, NFS_CS_READY);
out:
275 276 277
	return status;
}

278 279 280 281 282 283 284 285 286 287 288 289
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 */

290 291
static struct rpc_cred *
nfs4_get_setclientid_cred_server(struct nfs_server *server)
292
{
293 294
	struct nfs_client *clp = server->nfs_client;
	struct rpc_cred *cred = NULL;
295
	struct nfs4_state_owner *sp;
296
	struct rb_node *pos;
297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316

	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;
317
	struct rpc_cred *cred;
318

319 320
	spin_lock(&clp->cl_lock);
	cred = nfs4_get_machine_cred_locked(clp);
321
	spin_unlock(&clp->cl_lock);
322 323
	if (cred != NULL)
		goto out;
324 325 326 327 328 329

	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;
330
	}
331 332
	rcu_read_unlock();

333 334
out:
	return cred;
335 336
}

L
Linus Torvalds 已提交
337
static struct nfs4_state_owner *
338
nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
L
Linus Torvalds 已提交
339
{
340
	struct rb_node **p = &server->state_owners.rb_node,
341
		       *parent = NULL;
342
	struct nfs4_state_owner *sp;
L
Linus Torvalds 已提交
343

344 345
	while (*p != NULL) {
		parent = *p;
346
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
347 348 349 350 351 352

		if (cred < sp->so_cred)
			p = &parent->rb_left;
		else if (cred > sp->so_cred)
			p = &parent->rb_right;
		else {
353 354
			if (!list_empty(&sp->so_lru))
				list_del_init(&sp->so_lru);
355
			atomic_inc(&sp->so_count);
356
			return sp;
357
		}
L
Linus Torvalds 已提交
358
	}
359
	return NULL;
L
Linus Torvalds 已提交
360 361
}

362
static struct nfs4_state_owner *
363
nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
364
{
365 366
	struct nfs_server *server = new->so_server;
	struct rb_node **p = &server->state_owners.rb_node,
367 368
		       *parent = NULL;
	struct nfs4_state_owner *sp;
369
	int err;
370 371 372

	while (*p != NULL) {
		parent = *p;
373
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
374 375 376 377 378 379

		if (new->so_cred < sp->so_cred)
			p = &parent->rb_left;
		else if (new->so_cred > sp->so_cred)
			p = &parent->rb_right;
		else {
380 381
			if (!list_empty(&sp->so_lru))
				list_del_init(&sp->so_lru);
382 383 384 385
			atomic_inc(&sp->so_count);
			return sp;
		}
	}
386
	err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
387 388
	if (err)
		return ERR_PTR(err);
389 390
	rb_link_node(&new->so_server_node, parent, p);
	rb_insert_color(&new->so_server_node, &server->state_owners);
391 392 393 394
	return new;
}

static void
395
nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
396
{
397 398 399 400
	struct nfs_server *server = sp->so_server;

	if (!RB_EMPTY_NODE(&sp->so_server_node))
		rb_erase(&sp->so_server_node, &server->state_owners);
401
	ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
402 403
}

404 405 406
static void
nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
{
407
	sc->create_time = ktime_get();
408 409 410 411 412 413 414 415 416 417 418 419 420
	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
421 422 423 424 425 426
/*
 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
 * create a new state_owner.
 *
 */
static struct nfs4_state_owner *
427 428 429
nfs4_alloc_state_owner(struct nfs_server *server,
		struct rpc_cred *cred,
		gfp_t gfp_flags)
L
Linus Torvalds 已提交
430 431 432
{
	struct nfs4_state_owner *sp;

433
	sp = kzalloc(sizeof(*sp), gfp_flags);
L
Linus Torvalds 已提交
434 435
	if (!sp)
		return NULL;
436 437
	sp->so_server = server;
	sp->so_cred = get_rpccred(cred);
438
	spin_lock_init(&sp->so_lock);
L
Linus Torvalds 已提交
439
	INIT_LIST_HEAD(&sp->so_states);
440
	nfs4_init_seqid_counter(&sp->so_seqid);
L
Linus Torvalds 已提交
441
	atomic_set(&sp->so_count, 1);
442
	INIT_LIST_HEAD(&sp->so_lru);
L
Linus Torvalds 已提交
443 444 445
	return sp;
}

446
static void
L
Linus Torvalds 已提交
447 448
nfs4_drop_state_owner(struct nfs4_state_owner *sp)
{
449 450 451
	struct rb_node *rb_node = &sp->so_server_node;

	if (!RB_EMPTY_NODE(rb_node)) {
452 453
		struct nfs_server *server = sp->so_server;
		struct nfs_client *clp = server->nfs_client;
454 455

		spin_lock(&clp->cl_lock);
456 457 458 459
		if (!RB_EMPTY_NODE(rb_node)) {
			rb_erase(rb_node, &server->state_owners);
			RB_CLEAR_NODE(rb_node);
		}
460 461
		spin_unlock(&clp->cl_lock);
	}
L
Linus Torvalds 已提交
462 463
}

464 465
static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
{
466
	nfs4_destroy_seqid_counter(&sp->so_seqid);
467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495
	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);
	}
}

496 497 498 499 500 501 502 503
/**
 * 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,
504 505
					      struct rpc_cred *cred,
					      gfp_t gfp_flags)
L
Linus Torvalds 已提交
506
{
507
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
508 509 510
	struct nfs4_state_owner *sp, *new;

	spin_lock(&clp->cl_lock);
511
	sp = nfs4_find_state_owner_locked(server, cred);
L
Linus Torvalds 已提交
512 513
	spin_unlock(&clp->cl_lock);
	if (sp != NULL)
514
		goto out;
515
	new = nfs4_alloc_state_owner(server, cred, gfp_flags);
516
	if (new == NULL)
517
		goto out;
518 519 520 521 522 523 524
	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));
525 526
	if (sp != new)
		nfs4_free_state_owner(new);
527 528
out:
	nfs4_gc_state_owners(server);
529
	return sp;
L
Linus Torvalds 已提交
530 531
}

532 533 534
/**
 * nfs4_put_state_owner - Release a nfs4_state_owner
 * @sp: state owner data to release
535 536 537 538 539 540 541 542
 *
 * Note that we keep released state owners on an LRU
 * list.
 * This caches valid state owners so that they can be
 * reused, to avoid the OPEN_CONFIRM on minor version 0.
 * It also pins the uniquifier of dropped state owners for
 * a while, to ensure that those state owner names are
 * never reused.
543
 */
L
Linus Torvalds 已提交
544 545
void nfs4_put_state_owner(struct nfs4_state_owner *sp)
{
546 547
	struct nfs_server *server = sp->so_server;
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
548 549 550

	if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
		return;
551

552 553 554
	sp->so_expires = jiffies;
	list_add_tail(&sp->so_lru, &server->state_owners_lru);
	spin_unlock(&clp->cl_lock);
555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574
}

/**
 * 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
575
	spin_unlock(&clp->cl_lock);
576 577 578 579 580

	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
581 582 583 584 585 586 587
}

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

588
	state = kzalloc(sizeof(*state), GFP_NOFS);
L
Linus Torvalds 已提交
589 590 591 592
	if (!state)
		return NULL;
	atomic_set(&state->count, 1);
	INIT_LIST_HEAD(&state->lock_states);
593
	spin_lock_init(&state->state_lock);
594
	seqlock_init(&state->seqlock);
L
Linus Torvalds 已提交
595 596 597
	return state;
}

598
void
599
nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
600
{
601
	if (state->state == fmode)
602 603
		return;
	/* NB! List reordering - see the reclaim code for why.  */
604 605
	if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
		if (fmode & FMODE_WRITE)
606 607 608 609
			list_move(&state->open_states, &state->owner->so_states);
		else
			list_move_tail(&state->open_states, &state->owner->so_states);
	}
610
	state->state = fmode;
611 612
}

L
Linus Torvalds 已提交
613 614 615 616 617 618 619
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) {
620
		if (state->owner != owner)
L
Linus Torvalds 已提交
621
			continue;
622
		if (atomic_inc_not_zero(&state->count))
L
Linus Torvalds 已提交
623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645
			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();
646
	spin_lock(&owner->so_lock);
L
Linus Torvalds 已提交
647 648 649 650 651 652 653
	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);
654 655
		ihold(inode);
		state->inode = inode;
L
Linus Torvalds 已提交
656
		spin_unlock(&inode->i_lock);
657 658 659 660
		/* 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
661 662
	} else {
		spin_unlock(&inode->i_lock);
663
		spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
664 665 666 667 668 669 670 671 672 673 674 675
		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;

676
	if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
L
Linus Torvalds 已提交
677
		return;
678
	spin_lock(&inode->i_lock);
679
	list_del(&state->inode_states);
L
Linus Torvalds 已提交
680
	list_del(&state->open_states);
681 682
	spin_unlock(&inode->i_lock);
	spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
683 684 685 686 687 688
	iput(inode);
	nfs4_free_open_state(state);
	nfs4_put_state_owner(owner);
}

/*
689
 * Close the current file.
L
Linus Torvalds 已提交
690
 */
691
static void __nfs4_close(struct nfs4_state *state,
692
		fmode_t fmode, gfp_t gfp_mask, int wait)
L
Linus Torvalds 已提交
693 694
{
	struct nfs4_state_owner *owner = state->owner;
695
	int call_close = 0;
696
	fmode_t newstate;
L
Linus Torvalds 已提交
697 698 699

	atomic_inc(&owner->so_count);
	/* Protect against nfs4_find_state() */
700
	spin_lock(&owner->so_lock);
701
	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
702 703 704 705 706 707 708 709 710
		case FMODE_READ:
			state->n_rdonly--;
			break;
		case FMODE_WRITE:
			state->n_wronly--;
			break;
		case FMODE_READ|FMODE_WRITE:
			state->n_rdwr--;
	}
711
	newstate = FMODE_READ|FMODE_WRITE;
712
	if (state->n_rdwr == 0) {
713
		if (state->n_rdonly == 0) {
714
			newstate &= ~FMODE_READ;
715 716 717 718
			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) {
719
			newstate &= ~FMODE_WRITE;
720 721 722 723 724
			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);
725
	}
726
	nfs4_state_set_mode_locked(state, newstate);
727
	spin_unlock(&owner->so_lock);
728

729
	if (!call_close) {
730 731
		nfs4_put_open_state(state);
		nfs4_put_state_owner(owner);
F
Fred Isaman 已提交
732 733 734
	} else {
		bool roc = pnfs_roc(state->inode);

735
		nfs4_do_close(state, gfp_mask, wait, roc);
F
Fred Isaman 已提交
736
	}
737 738
}

739
void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
740
{
741
	__nfs4_close(state, fmode, GFP_NOFS, 0);
742 743
}

744
void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
745
{
746
	__nfs4_close(state, fmode, GFP_KERNEL, 1);
L
Linus Torvalds 已提交
747 748 749 750 751 752 753
}

/*
 * Search the state->lock_states for an existing lock_owner
 * that is compatible with current->files
 */
static struct nfs4_lock_state *
754
__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
755 756 757
{
	struct nfs4_lock_state *pos;
	list_for_each_entry(pos, &state->lock_states, ls_locks) {
758
		if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
L
Linus Torvalds 已提交
759
			continue;
760 761 762 763 764 765 766 767 768
		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
769 770 771 772 773 774 775 776 777 778 779
		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.
 *
 */
780
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
781 782
{
	struct nfs4_lock_state *lsp;
783
	struct nfs_server *server = state->owner->so_server;
L
Linus Torvalds 已提交
784

785
	lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
L
Linus Torvalds 已提交
786 787
	if (lsp == NULL)
		return NULL;
788
	nfs4_init_seqid_counter(&lsp->ls_seqid);
L
Linus Torvalds 已提交
789
	atomic_set(&lsp->ls_count, 1);
790
	lsp->ls_state = state;
791 792 793 794 795 796 797 798 799
	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:
800
		goto out_free;
801
	}
802 803
	lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
	if (lsp->ls_seqid.owner_id < 0)
804
		goto out_free;
805
	INIT_LIST_HEAD(&lsp->ls_locks);
L
Linus Torvalds 已提交
806
	return lsp;
807 808 809
out_free:
	kfree(lsp);
	return NULL;
L
Linus Torvalds 已提交
810 811
}

812
void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
813
{
814
	ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
815
	nfs4_destroy_seqid_counter(&lsp->ls_seqid);
816 817 818
	kfree(lsp);
}

L
Linus Torvalds 已提交
819 820 821 822 823
/*
 * Return a compatible lock_state. If no initialized lock_state structure
 * exists, return an uninitialized one.
 *
 */
824
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
825
{
826
	struct nfs4_lock_state *lsp, *new = NULL;
L
Linus Torvalds 已提交
827
	
828 829
	for(;;) {
		spin_lock(&state->state_lock);
830
		lsp = __nfs4_find_lock_state(state, owner, pid, type);
831 832 833 834 835 836 837 838 839 840
		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);
841
		new = nfs4_alloc_lock_state(state, owner, pid, type);
842 843 844 845
		if (new == NULL)
			return NULL;
	}
	spin_unlock(&state->state_lock);
846
	if (new != NULL)
847
		nfs4_free_lock_state(state->owner->so_server, new);
L
Linus Torvalds 已提交
848 849 850 851
	return lsp;
}

/*
852 853
 * Release reference to lock_state, and free it if we see that
 * it is no longer in use
L
Linus Torvalds 已提交
854
 */
855
void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
L
Linus Torvalds 已提交
856
{
857
	struct nfs4_state *state;
L
Linus Torvalds 已提交
858

859 860 861 862 863 864 865 866 867
	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);
868 869 870 871
	if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
		if (nfs4_release_lockowner(lsp) == 0)
			return;
	}
872
	nfs4_free_lock_state(lsp->ls_state->owner->so_server, lsp);
L
Linus Torvalds 已提交
873 874
}

875
static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
L
Linus Torvalds 已提交
876
{
877
	struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
L
Linus Torvalds 已提交
878

879 880 881
	dst->fl_u.nfs4_fl.owner = lsp;
	atomic_inc(&lsp->ls_count);
}
L
Linus Torvalds 已提交
882

883
static void nfs4_fl_release_lock(struct file_lock *fl)
L
Linus Torvalds 已提交
884
{
885
	nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
L
Linus Torvalds 已提交
886 887
}

888
static const struct file_lock_operations nfs4_fl_lock_ops = {
889 890 891 892 893
	.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
894
{
895 896 897 898
	struct nfs4_lock_state *lsp;

	if (fl->fl_ops != NULL)
		return 0;
899 900 901
	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)
902 903
		lsp = nfs4_get_lock_state(state, NULL, fl->fl_pid,
				NFS4_FLOCK_LOCK_TYPE);
904 905
	else
		return -EINVAL;
906 907 908 909 910
	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
911 912
}

913 914
static bool nfs4_copy_lock_stateid(nfs4_stateid *dst, struct nfs4_state *state,
		fl_owner_t fl_owner, pid_t fl_pid)
L
Linus Torvalds 已提交
915
{
916
	struct nfs4_lock_state *lsp;
917
	bool ret = false;
L
Linus Torvalds 已提交
918

919
	if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
920
		goto out;
L
Linus Torvalds 已提交
921

922
	spin_lock(&state->state_lock);
923
	lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
924
	if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0) {
925
		nfs4_stateid_copy(dst, &lsp->ls_stateid);
926 927
		ret = true;
	}
928
	spin_unlock(&state->state_lock);
L
Linus Torvalds 已提交
929
	nfs4_put_lock_state(lsp);
930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955
out:
	return ret;
}

static void nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
{
	int seq;

	do {
		seq = read_seqbegin(&state->seqlock);
		nfs4_stateid_copy(dst, &state->stateid);
	} while (read_seqretry(&state->seqlock, seq));
}

/*
 * Byte-range lock aware utility to initialize the stateid of read/write
 * requests.
 */
void nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state,
		fmode_t fmode, fl_owner_t fl_owner, pid_t fl_pid)
{
	if (nfs4_copy_delegation_stateid(dst, state->inode, fmode))
		return;
	if (nfs4_copy_lock_stateid(dst, state, fl_owner, fl_pid))
		return;
	nfs4_copy_open_stateid(dst, state);
L
Linus Torvalds 已提交
956 957
}

958
struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
959 960 961
{
	struct nfs_seqid *new;

962
	new = kmalloc(sizeof(*new), gfp_mask);
963 964
	if (new != NULL) {
		new->sequence = counter;
965
		INIT_LIST_HEAD(&new->list);
966
		new->task = NULL;
967 968 969 970
	}
	return new;
}

971
void nfs_release_seqid(struct nfs_seqid *seqid)
L
Linus Torvalds 已提交
972
{
973
	struct nfs_seqid_counter *sequence;
974

975 976 977 978 979 980 981 982 983 984 985
	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);
986
	}
987
	spin_unlock(&sequence->lock);
988 989 990 991 992
}

void nfs_free_seqid(struct nfs_seqid *seqid)
{
	nfs_release_seqid(seqid);
993
	kfree(seqid);
L
Linus Torvalds 已提交
994 995 996
}

/*
997 998 999 1000
 * 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()
 */
1001
static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
1002
{
1003
	BUG_ON(list_first_entry(&seqid->sequence->list, struct nfs_seqid, list) != seqid);
1004 1005 1006 1007
	switch (status) {
		case 0:
			break;
		case -NFS4ERR_BAD_SEQID:
1008 1009
			if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
				return;
1010
			pr_warn_ratelimited("NFS: v4 server returned a bad"
D
Dan Muntz 已提交
1011 1012
					" sequence-id error on an"
					" unconfirmed sequence %p!\n",
1013
					seqid->sequence);
1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031
		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)
{
1032 1033 1034 1035 1036
	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
1037
		nfs4_drop_state_owner(sp);
1038 1039
	if (!nfs4_has_session(server->nfs_client))
		nfs_increment_seqid(status, seqid);
1040 1041 1042 1043 1044 1045 1046 1047 1048
}

/*
 * 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)
{
1049
	nfs_increment_seqid(status, seqid);
1050 1051 1052 1053
}

int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
{
1054
	struct nfs_seqid_counter *sequence = seqid->sequence;
1055 1056 1057
	int status = 0;

	spin_lock(&sequence->lock);
1058
	seqid->task = task;
1059 1060 1061 1062
	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;
1063
	rpc_sleep_on(&sequence->wait, task, NULL);
1064 1065
	status = -EAGAIN;
unlock:
1066 1067
	spin_unlock(&sequence->lock);
	return status;
L
Linus Torvalds 已提交
1068 1069
}

1070
static int nfs4_run_state_manager(void *);
L
Linus Torvalds 已提交
1071

1072
static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
T
Trond Myklebust 已提交
1073 1074
{
	smp_mb__before_clear_bit();
1075
	clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
T
Trond Myklebust 已提交
1076
	smp_mb__after_clear_bit();
1077
	wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
T
Trond Myklebust 已提交
1078 1079 1080
	rpc_wake_up(&clp->cl_rpcwaitq);
}

L
Linus Torvalds 已提交
1081
/*
1082
 * Schedule the nfs_client asynchronous state management routine
L
Linus Torvalds 已提交
1083
 */
1084
void nfs4_schedule_state_manager(struct nfs_client *clp)
L
Linus Torvalds 已提交
1085
{
1086
	struct task_struct *task;
1087
	char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
L
Linus Torvalds 已提交
1088

1089 1090
	if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
		return;
1091 1092
	__module_get(THIS_MODULE);
	atomic_inc(&clp->cl_count);
1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108

	/* 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
1109 1110 1111
}

/*
1112
 * Schedule a lease recovery attempt
L
Linus Torvalds 已提交
1113
 */
1114
void nfs4_schedule_lease_recovery(struct nfs_client *clp)
L
Linus Torvalds 已提交
1115 1116 1117
{
	if (!clp)
		return;
1118 1119
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1120 1121
	dprintk("%s: scheduling lease recovery for server %s\n", __func__,
			clp->cl_hostname);
1122
	nfs4_schedule_state_manager(clp);
L
Linus Torvalds 已提交
1123
}
A
Andy Adamson 已提交
1124
EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
L
Linus Torvalds 已提交
1125

1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137
/*
 * nfs40_handle_cb_pathdown - return all delegations after NFS4ERR_CB_PATH_DOWN
 * @clp: client to process
 *
 * Set the NFS4CLNT_LEASE_EXPIRED state in order to force a
 * resend of the SETCLIENTID and hence re-establish the
 * callback channel. Then return all existing delegations.
 */
static void nfs40_handle_cb_pathdown(struct nfs_client *clp)
{
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
	nfs_expire_all_delegations(clp);
1138 1139
	dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
			clp->cl_hostname);
1140 1141
}

1142 1143
void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
{
1144
	nfs40_handle_cb_pathdown(clp);
1145 1146 1147
	nfs4_schedule_state_manager(clp);
}

1148
static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1149 1150 1151 1152 1153 1154 1155 1156
{

	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;
	}
1157
	set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1158 1159 1160 1161
	set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
	return 1;
}

1162
static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1163 1164 1165
{
	set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
	clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1166
	set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1167 1168 1169 1170
	set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
	return 1;
}

1171 1172 1173 1174 1175
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);
1176 1177
	dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
			clp->cl_hostname);
1178 1179
	nfs4_schedule_state_manager(clp);
}
A
Andy Adamson 已提交
1180
EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1181

1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197
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;
1198
		if (!nfs4_stateid_match(&state->stateid, stateid))
1199 1200 1201 1202 1203 1204 1205 1206 1207 1208
			continue;
		nfs4_state_mark_reclaim_nograce(clp, state);
		found = true;
	}
	spin_unlock(&inode->i_lock);
	if (found)
		nfs4_schedule_state_manager(clp);
}


1209
static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1210 1211
{
	struct inode *inode = state->inode;
1212
	struct nfs_inode *nfsi = NFS_I(inode);
L
Linus Torvalds 已提交
1213 1214 1215
	struct file_lock *fl;
	int status = 0;

1216 1217 1218 1219
	if (inode->i_flock == NULL)
		return 0;

	/* Guard against delegation returns and new lock/unlock calls */
1220
	down_write(&nfsi->rwsem);
1221
	/* Protect inode->i_flock using the BKL */
1222
	lock_flocks();
H
Harvey Harrison 已提交
1223
	for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1224
		if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
L
Linus Torvalds 已提交
1225
			continue;
1226
		if (nfs_file_open_context(fl->fl_file)->state != state)
L
Linus Torvalds 已提交
1227
			continue;
1228
		unlock_flocks();
L
Linus Torvalds 已提交
1229 1230
		status = ops->recover_lock(state, fl);
		switch (status) {
1231 1232 1233 1234 1235 1236 1237 1238 1239
			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:
1240 1241 1242 1243
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1244
				goto out;
L
Linus Torvalds 已提交
1245
			default:
1246 1247
				printk(KERN_ERR "NFS: %s: unhandled error %d. "
					"Zeroing state\n", __func__, status);
1248 1249
			case -ENOMEM:
			case -NFS4ERR_DENIED:
L
Linus Torvalds 已提交
1250 1251
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1252
				/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1253
				status = 0;
L
Linus Torvalds 已提交
1254
		}
1255
		lock_flocks();
L
Linus Torvalds 已提交
1256
	}
1257
	unlock_flocks();
1258
out:
1259
	up_write(&nfsi->rwsem);
L
Linus Torvalds 已提交
1260 1261 1262
	return status;
}

1263
static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276
{
	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.
	 */
1277 1278
restart:
	spin_lock(&sp->so_lock);
L
Linus Torvalds 已提交
1279
	list_for_each_entry(state, &sp->so_states, open_states) {
1280 1281
		if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
			continue;
L
Linus Torvalds 已提交
1282 1283
		if (state->state == 0)
			continue;
1284 1285
		atomic_inc(&state->count);
		spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1286 1287
		status = ops->recover_open(sp, state);
		if (status >= 0) {
1288 1289
			status = nfs4_reclaim_locks(state, ops);
			if (status >= 0) {
1290
				spin_lock(&state->state_lock);
1291 1292
				list_for_each_entry(lock, &state->lock_states, ls_locks) {
					if (!(lock->ls_flags & NFS_LOCK_INITIALIZED))
1293 1294
						pr_warn_ratelimited("NFS: "
							"%s: Lock reclaim "
1295
							"failed!\n", __func__);
1296
				}
1297
				spin_unlock(&state->state_lock);
1298 1299
				nfs4_put_open_state(state);
				goto restart;
L
Linus Torvalds 已提交
1300 1301 1302 1303
			}
		}
		switch (status) {
			default:
1304 1305
				printk(KERN_ERR "NFS: %s: unhandled error %d. "
					"Zeroing state\n", __func__, status);
L
Linus Torvalds 已提交
1306
			case -ENOENT:
1307
			case -ENOMEM:
1308
			case -ESTALE:
L
Linus Torvalds 已提交
1309 1310 1311 1312
				/*
				 * Open state on this file cannot be recovered
				 * All we can do is revert to using the zero stateid.
				 */
1313 1314
				memset(&state->stateid, 0,
					sizeof(state->stateid));
L
Linus Torvalds 已提交
1315 1316 1317
				/* Mark the file as being 'closed' */
				state->state = 0;
				break;
1318 1319 1320 1321 1322 1323 1324 1325
			case -EKEYEXPIRED:
				/*
				 * User RPCSEC_GSS context has expired.
				 * We cannot recover this stateid now, so
				 * skip it and allow recovery thread to
				 * proceed.
				 */
				break;
1326 1327 1328
			case -NFS4ERR_ADMIN_REVOKED:
			case -NFS4ERR_STALE_STATEID:
			case -NFS4ERR_BAD_STATEID:
1329 1330
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1331
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1332
				break;
L
Linus Torvalds 已提交
1333 1334
			case -NFS4ERR_EXPIRED:
			case -NFS4ERR_NO_GRACE:
1335
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
L
Linus Torvalds 已提交
1336
			case -NFS4ERR_STALE_CLIENTID:
1337 1338 1339 1340
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
L
Linus Torvalds 已提交
1341 1342
				goto out_err;
		}
1343 1344
		nfs4_put_open_state(state);
		goto restart;
L
Linus Torvalds 已提交
1345
	}
1346
	spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1347 1348
	return 0;
out_err:
1349
	nfs4_put_open_state(state);
L
Linus Torvalds 已提交
1350 1351 1352
	return status;
}

1353 1354 1355 1356 1357 1358 1359 1360
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);
1361
	spin_lock(&state->state_lock);
1362 1363 1364 1365
	list_for_each_entry(lock, &state->lock_states, ls_locks) {
		lock->ls_seqid.flags = 0;
		lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
	}
1366
	spin_unlock(&state->state_lock);
1367 1368
}

1369 1370
static void nfs4_reset_seqids(struct nfs_server *server,
	int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1371
{
1372
	struct nfs_client *clp = server->nfs_client;
1373
	struct nfs4_state_owner *sp;
1374
	struct rb_node *pos;
1375 1376
	struct nfs4_state *state;

1377 1378 1379 1380 1381
	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);
1382
		sp->so_seqid.flags = 0;
1383
		spin_lock(&sp->so_lock);
1384
		list_for_each_entry(state, &sp->so_states, open_states) {
1385 1386
			if (mark_reclaim(clp, state))
				nfs4_clear_open_state(state);
1387
		}
1388
		spin_unlock(&sp->so_lock);
1389
	}
1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401
	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();
1402 1403
}

1404 1405 1406 1407 1408 1409 1410
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);
}

1411 1412 1413 1414 1415 1416 1417 1418
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);
}

1419
static void nfs4_clear_reclaim_server(struct nfs_server *server)
1420
{
1421
	struct nfs_client *clp = server->nfs_client;
1422 1423 1424 1425
	struct nfs4_state_owner *sp;
	struct rb_node *pos;
	struct nfs4_state *state;

1426 1427 1428 1429 1430
	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);
1431 1432
		spin_lock(&sp->so_lock);
		list_for_each_entry(state, &sp->so_states, open_states) {
1433 1434
			if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
						&state->flags))
1435 1436 1437 1438 1439
				continue;
			nfs4_state_mark_reclaim_nograce(clp, state);
		}
		spin_unlock(&sp->so_lock);
	}
1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453
	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();
1454 1455

	nfs_delegation_reap_unclaimed(clp);
1456 1457 1458 1459 1460 1461 1462 1463
	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);
1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477
}

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);
}

1478 1479 1480 1481 1482 1483 1484 1485
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);
}

1486
static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1487 1488
{
	switch (error) {
1489 1490
		case 0:
			break;
1491
		case -NFS4ERR_CB_PATH_DOWN:
1492
			nfs40_handle_cb_pathdown(clp);
1493
			break;
1494 1495
		case -NFS4ERR_NO_GRACE:
			nfs4_state_end_reclaim_reboot(clp);
1496
			break;
1497 1498 1499
		case -NFS4ERR_STALE_CLIENTID:
		case -NFS4ERR_LEASE_MOVED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1500
			nfs4_state_clear_reclaim_reboot(clp);
1501 1502 1503 1504 1505
			nfs4_state_start_reclaim_reboot(clp);
			break;
		case -NFS4ERR_EXPIRED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
			nfs4_state_start_reclaim_nograce(clp);
1506
			break;
1507 1508 1509 1510 1511 1512
		case -NFS4ERR_BADSESSION:
		case -NFS4ERR_BADSLOT:
		case -NFS4ERR_BAD_HIGH_SLOT:
		case -NFS4ERR_DEADSESSION:
		case -NFS4ERR_SEQ_FALSE_RETRY:
		case -NFS4ERR_SEQ_MISORDERED:
1513
			set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1514
			/* Zero session reset errors */
1515
			break;
1516 1517 1518
		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
			set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
			break;
1519 1520 1521
		case -EKEYEXPIRED:
			/* Nothing we can do */
			nfs4_warn_keyexpired(clp->cl_hostname);
1522 1523
			break;
		default:
1524 1525
			dprintk("%s: failed to handle error %d for server %s\n",
					__func__, error, clp->cl_hostname);
1526
			return error;
1527
	}
1528 1529
	dprintk("%s: handled error %d for server %s\n", __func__, error,
			clp->cl_hostname);
1530
	return 0;
1531 1532
}

1533
static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1534
{
1535 1536
	struct nfs4_state_owner *sp;
	struct nfs_server *server;
1537
	struct rb_node *pos;
L
Linus Torvalds 已提交
1538 1539
	int status = 0;

1540
restart:
1541 1542
	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1543
		nfs4_purge_state_owners(server);
1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563
		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);
			}

1564
			nfs4_put_state_owner(sp);
1565
			goto restart;
1566
		}
1567
		spin_unlock(&clp->cl_lock);
1568
	}
1569
	rcu_read_unlock();
1570 1571 1572 1573 1574 1575
	return status;
}

static int nfs4_check_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1576 1577
	const struct nfs4_state_maintenance_ops *ops =
		clp->cl_mvops->state_renewal_ops;
1578
	int status;
L
Linus Torvalds 已提交
1579

1580 1581 1582
	/* Is the client already known to have an expired lease? */
	if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		return 0;
1583 1584 1585
	spin_lock(&clp->cl_lock);
	cred = ops->get_state_renewal_cred_locked(clp);
	spin_unlock(&clp->cl_lock);
1586 1587
	if (cred == NULL) {
		cred = nfs4_get_setclientid_cred(clp);
1588
		status = -ENOKEY;
1589 1590
		if (cred == NULL)
			goto out;
1591
	}
1592
	status = ops->renew_lease(clp, cred);
1593 1594
	put_rpccred(cred);
out:
1595
	return nfs4_recovery_handle_error(clp, status);
1596 1597
}

1598 1599 1600 1601 1602 1603
/* Set NFS4CLNT_LEASE_EXPIRED for all v4.0 errors and for recoverable errors
 * on EXCHANGE_ID for v4.1
 */
static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
{
	switch (status) {
1604 1605 1606 1607 1608
	case -NFS4ERR_SEQ_MISORDERED:
		if (test_and_set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state))
			return -ESERVERFAULT;
		/* Lease confirmation error: retry after purging the lease */
		ssleep(1);
1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626
	case -NFS4ERR_CLID_INUSE:
	case -NFS4ERR_STALE_CLIENTID:
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
		break;
	case -EACCES:
		if (clp->cl_machine_cred == NULL)
			return -EACCES;
		/* Handle case where the user hasn't set up machine creds */
		nfs4_clear_machine_cred(clp);
	case -NFS4ERR_DELAY:
	case -ETIMEDOUT:
	case -EAGAIN:
		ssleep(1);
		break;

	case -NFS4ERR_MINOR_VERS_MISMATCH:
		if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
			nfs_mark_client_ready(clp, -EPROTONOSUPPORT);
1627 1628
		dprintk("%s: exit with error %d for server %s\n",
				__func__, -EPROTONOSUPPORT, clp->cl_hostname);
1629 1630 1631 1632 1633 1634
		return -EPROTONOSUPPORT;
	case -EKEYEXPIRED:
		nfs4_warn_keyexpired(clp->cl_hostname);
	case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
				 * in nfs4_exchange_id */
	default:
1635 1636
		dprintk("%s: exit with error %d for server %s\n", __func__,
				status, clp->cl_hostname);
1637 1638 1639
		return status;
	}
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1640 1641
	dprintk("%s: handled error %d for server %s\n", __func__, status,
			clp->cl_hostname);
1642 1643 1644
	return 0;
}

1645 1646 1647
static int nfs4_reclaim_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1648 1649
	const struct nfs4_state_recovery_ops *ops =
		clp->cl_mvops->reboot_recovery_ops;
1650
	int status;
1651

1652
	cred = ops->get_clid_cred(clp);
1653 1654 1655 1656 1657 1658 1659
	if (cred == NULL)
		return -ENOENT;
	status = ops->establish_clid(clp, cred);
	put_rpccred(cred);
	if (status != 0)
		return nfs4_handle_reclaim_lease_error(clp, status);
	return 0;
1660 1661
}

1662
#ifdef CONFIG_NFS_V4_1
1663
void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
1664
{
1665 1666
	struct nfs_client *clp = session->clp;

1667 1668 1669 1670 1671 1672 1673
	switch (err) {
	default:
		set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
		break;
	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
		set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
	}
1674
	nfs4_schedule_lease_recovery(clp);
1675
}
1676
EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
1677

A
Andy Adamson 已提交
1678 1679 1680
void nfs41_handle_recall_slot(struct nfs_client *clp)
{
	set_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
1681 1682
	dprintk("%s: scheduling slot recall for server %s\n", __func__,
			clp->cl_hostname);
1683
	nfs4_schedule_state_manager(clp);
A
Andy Adamson 已提交
1684 1685
}

1686 1687 1688
static void nfs4_reset_all_state(struct nfs_client *clp)
{
	if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1689
		set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1690
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1691
		nfs4_state_start_reclaim_nograce(clp);
1692 1693
		dprintk("%s: scheduling reset of all state for server %s!\n",
				__func__, clp->cl_hostname);
1694
		nfs4_schedule_state_manager(clp);
1695 1696 1697 1698 1699 1700 1701
	}
}

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);
1702 1703
		dprintk("%s: server %s rebooted!\n", __func__,
				clp->cl_hostname);
1704
		nfs4_schedule_state_manager(clp);
1705 1706 1707 1708 1709 1710
	}
}

static void nfs41_handle_state_revoked(struct nfs_client *clp)
{
	nfs4_reset_all_state(clp);
1711
	dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
1712 1713 1714 1715 1716 1717
}

static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
{
	/* This will need to handle layouts too */
	nfs_expire_all_delegations(clp);
1718 1719
	dprintk("%s: Recallable state revoked on server %s!\n", __func__,
			clp->cl_hostname);
1720 1721
}

1722
static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
1723 1724 1725
{
	nfs_expire_all_delegations(clp);
	if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
1726
		nfs4_schedule_state_manager(clp);
1727 1728
	dprintk("%s: server %s declared a backchannel fault\n", __func__,
			clp->cl_hostname);
1729 1730
}

1731 1732 1733 1734 1735 1736 1737
static void nfs41_handle_cb_path_down(struct nfs_client *clp)
{
	if (test_and_set_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
		&clp->cl_state) == 0)
		nfs4_schedule_state_manager(clp);
}

1738 1739 1740 1741
void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
{
	if (!flags)
		return;
1742 1743 1744 1745

	dprintk("%s: \"%s\" (client ID %llx) flags=0x%08x\n",
		__func__, clp->cl_hostname, clp->cl_clientid, flags);

1746
	if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
1747
		nfs41_handle_server_reboot(clp);
1748
	if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
1749 1750
			    SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
			    SEQ4_STATUS_ADMIN_STATE_REVOKED |
1751 1752
			    SEQ4_STATUS_LEASE_MOVED))
		nfs41_handle_state_revoked(clp);
1753
	if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
1754
		nfs41_handle_recallable_state_revoked(clp);
1755 1756 1757 1758
	if (flags & SEQ4_STATUS_BACKCHANNEL_FAULT)
		nfs41_handle_backchannel_fault(clp);
	else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
				SEQ4_STATUS_CB_PATH_DOWN_SESSION))
1759
		nfs41_handle_cb_path_down(clp);
1760 1761
}

1762 1763
static int nfs4_reset_session(struct nfs_client *clp)
{
1764
	struct rpc_cred *cred;
1765 1766
	int status;

1767
	nfs4_begin_drain_session(clp);
1768 1769
	cred = nfs4_get_exchange_id_cred(clp);
	status = nfs4_proc_destroy_session(clp->cl_session, cred);
1770 1771
	if (status && status != -NFS4ERR_BADSESSION &&
	    status != -NFS4ERR_DEADSESSION) {
1772
		status = nfs4_recovery_handle_error(clp, status);
1773 1774 1775 1776
		goto out;
	}

	memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
1777
	status = nfs4_proc_create_session(clp, cred);
1778
	if (status) {
1779 1780
		dprintk("%s: session reset failed with status %d for server %s!\n",
			__func__, status, clp->cl_hostname);
1781
		status = nfs4_handle_reclaim_lease_error(clp, status);
1782 1783
		goto out;
	}
1784
	nfs41_finish_session_reset(clp);
1785 1786
	dprintk("%s: session reset was successful for server %s!\n",
			__func__, clp->cl_hostname);
1787
out:
1788 1789
	if (cred)
		put_rpccred(cred);
1790 1791
	return status;
}
1792

A
Andy Adamson 已提交
1793 1794 1795 1796 1797 1798 1799 1800 1801
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),
1802
		      GFP_NOFS);
A
Andy Adamson 已提交
1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820
        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;
}

1821 1822
static int nfs4_bind_conn_to_session(struct nfs_client *clp)
{
1823 1824 1825
	struct rpc_cred *cred;
	int ret;

1826
	nfs4_begin_drain_session(clp);
1827 1828 1829 1830
	cred = nfs4_get_exchange_id_cred(clp);
	ret = nfs4_proc_bind_conn_to_session(clp, cred);
	if (cred)
		put_rpccred(cred);
1831
	clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1832 1833
	switch (ret) {
	case 0:
1834 1835
		dprintk("%s: bind_conn_to_session was successful for server %s!\n",
			__func__, clp->cl_hostname);
1836 1837 1838 1839 1840 1841 1842 1843 1844
		break;
	case -NFS4ERR_DELAY:
		ssleep(1);
		set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
		break;
	default:
		return nfs4_recovery_handle_error(clp, ret);
	}
	return 0;
1845
}
1846
#else /* CONFIG_NFS_V4_1 */
1847
static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
1848
static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
A
Andy Adamson 已提交
1849
static int nfs4_recall_slot(struct nfs_client *clp) { return 0; }
1850 1851 1852 1853 1854

static int nfs4_bind_conn_to_session(struct nfs_client *clp)
{
	return 0;
}
1855 1856
#endif /* CONFIG_NFS_V4_1 */

1857
static void nfs4_state_manager(struct nfs_client *clp)
1858 1859 1860 1861
{
	int status = 0;

	/* Ensure exclusive access to NFSv4 state */
1862
	do {
1863
		if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
1864 1865 1866
			status = nfs4_reclaim_lease(clp);
			if (status < 0)
				goto out_error;
1867 1868 1869 1870
			clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
		}

1871 1872 1873
		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);
1874
			if (status < 0)
1875
				goto out_error;
1876 1877
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
				continue;
1878
			clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1879 1880 1881 1882 1883 1884 1885 1886

			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);

1887
			pnfs_destroy_all_layouts(clp);
1888 1889 1890 1891
		}

		if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
			status = nfs4_check_lease(clp);
1892 1893
			if (status < 0)
				goto out_error;
1894
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1895
				continue;
1896
		}
1897

1898
		/* Initialize or reset the session */
1899
		if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)
1900
		   && nfs4_has_session(clp)) {
1901
			status = nfs4_reset_session(clp);
1902 1903 1904
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
				continue;
			if (status < 0)
1905 1906
				goto out_error;
		}
1907

1908 1909 1910 1911 1912 1913
		/* Send BIND_CONN_TO_SESSION */
		if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
				&clp->cl_state) && nfs4_has_session(clp)) {
			status = nfs4_bind_conn_to_session(clp);
			if (status < 0)
				goto out_error;
1914
			continue;
1915 1916
		}

1917
		/* First recover reboot state... */
1918
		if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
1919
			status = nfs4_do_reclaim(clp,
1920
				clp->cl_mvops->reboot_recovery_ops);
1921
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1922
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
1923
				continue;
1924
			nfs4_state_end_reclaim_reboot(clp);
1925 1926 1927 1928
			if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
				continue;
			if (status < 0)
				goto out_error;
1929 1930
		}

1931 1932
		/* Now recover expired state... */
		if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
1933
			status = nfs4_do_reclaim(clp,
1934
				clp->cl_mvops->nograce_recovery_ops);
1935
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1936
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
1937 1938 1939
			    test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
				continue;
			if (status < 0)
1940
				goto out_error;
L
Linus Torvalds 已提交
1941
		}
1942

1943
		nfs4_end_drain_session(clp);
1944 1945 1946 1947
		if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
			nfs_client_return_marked_delegations(clp);
			continue;
		}
A
Andy Adamson 已提交
1948 1949 1950 1951 1952 1953 1954 1955 1956
		/* 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;
		}

1957 1958

		nfs4_clear_state_manager_bit(clp);
1959 1960 1961 1962 1963
		/* 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;
1964
	} while (atomic_read(&clp->cl_count) > 1);
1965
	return;
L
Linus Torvalds 已提交
1966
out_error:
1967
	pr_warn_ratelimited("NFS: state manager failed on NFSv4 server %s"
1968
			" with error %d\n", clp->cl_hostname, -status);
1969
	nfs4_end_drain_session(clp);
1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981
	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
1982 1983 1984 1985 1986 1987 1988
}

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