nfs4state.c 53.7 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"
59
#include "netns.h"
L
Linus Torvalds 已提交
60

C
Chuck Lever 已提交
61 62
#define NFSDBG_FACILITY		NFSDBG_STATE

L
Linus Torvalds 已提交
63 64
#define OPENOWNER_POOL_SIZE	8

65
const nfs4_stateid zero_stateid;
L
Linus Torvalds 已提交
66 67 68

static LIST_HEAD(nfs4_clientid_list);

69
int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
L
Linus Torvalds 已提交
70
{
71 72 73 74
	struct nfs4_setclientid_res clid = {
		.clientid = clp->cl_clientid,
		.confirm = clp->cl_confirm,
	};
75 76
	unsigned short port;
	int status;
77
	struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
78

79 80
	if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
		goto do_confirm;
81
	port = nn->nfs_callback_tcpport;
82
	if (clp->cl_addr.ss_family == AF_INET6)
83
		port = nn->nfs_callback_tcpport6;
84

85 86 87
	status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
	if (status != 0)
		goto out;
88 89 90 91
	clp->cl_clientid = clid.clientid;
	clp->cl_confirm = clid.confirm;
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
do_confirm:
92 93 94
	status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
	if (status != 0)
		goto out;
95
	clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
96 97
	nfs4_schedule_state_renewal(clp);
out:
L
Linus Torvalds 已提交
98 99 100
	return status;
}

101
struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121
{
	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);
}

122 123
static struct rpc_cred *
nfs4_get_renew_cred_server_locked(struct nfs_server *server)
124
{
125
	struct rpc_cred *cred = NULL;
126
	struct nfs4_state_owner *sp;
127
	struct rb_node *pos;
128

129 130 131 132
	for (pos = rb_first(&server->state_owners);
	     pos != NULL;
	     pos = rb_next(pos)) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
133 134 135 136 137 138 139 140
		if (list_empty(&sp->so_states))
			continue;
		cred = get_rpccred(sp->so_cred);
		break;
	}
	return cred;
}

141 142 143 144 145 146 147 148 149 150 151 152
/**
 * 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;

153 154 155 156 157
	/* Use machine credentials if available */
	cred = nfs4_get_machine_cred_locked(clp);
	if (cred != NULL)
		goto out;

158 159 160 161 162 163 164
	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();
165 166

out:
167 168 169
	return cred;
}

170 171
#if defined(CONFIG_NFS_V4_1)

R
Ricardo Labiaga 已提交
172 173 174 175 176
static int nfs41_setup_state_renewal(struct nfs_client *clp)
{
	int status;
	struct nfs_fsinfo fsinfo;

177 178 179 180 181
	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
182 183 184 185 186 187 188 189 190 191 192 193 194 195
	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;
}

196 197 198 199 200
/*
 * 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.
 */
201
static void nfs4_end_drain_session(struct nfs_client *clp)
202
{
203
	struct nfs4_session *ses = clp->cl_session;
204
	struct nfs4_slot_table *tbl;
205 206
	int max_slots;

207 208
	if (ses == NULL)
		return;
209
	tbl = &ses->fc_slot_table;
210
	if (test_and_clear_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
211 212
		spin_lock(&tbl->slot_tbl_lock);
		max_slots = tbl->max_slots;
213
		while (max_slots--) {
214 215 216
			if (rpc_wake_up_first(&tbl->slot_tbl_waitq,
						nfs4_set_task_privileged,
						NULL) == NULL)
217 218
				break;
		}
219
		spin_unlock(&tbl->slot_tbl_lock);
220
	}
221 222
}

223
static int nfs4_wait_on_slot_tbl(struct nfs4_slot_table *tbl)
224 225
{
	spin_lock(&tbl->slot_tbl_lock);
226
	if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
227
		INIT_COMPLETION(tbl->complete);
228
		spin_unlock(&tbl->slot_tbl_lock);
229
		return wait_for_completion_interruptible(&tbl->complete);
230 231 232 233 234
	}
	spin_unlock(&tbl->slot_tbl_lock);
	return 0;
}

235 236 237 238 239 240 241 242 243 244 245 246 247 248
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);
}

249 250 251 252 253 254 255 256 257 258
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);
}

259 260 261 262
int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
{
	int status;

263 264
	if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
		goto do_confirm;
265
	nfs4_begin_drain_session(clp);
266
	status = nfs4_proc_exchange_id(clp, cred);
R
Ricardo Labiaga 已提交
267 268
	if (status != 0)
		goto out;
269 270
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
do_confirm:
271
	status = nfs4_proc_create_session(clp, cred);
R
Ricardo Labiaga 已提交
272 273
	if (status != 0)
		goto out;
274
	nfs41_finish_session_reset(clp);
R
Ricardo Labiaga 已提交
275 276
	nfs_mark_client_ready(clp, NFS_CS_READY);
out:
277 278 279
	return status;
}

280 281 282 283 284 285 286 287 288 289 290 291
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 */

292 293
static struct rpc_cred *
nfs4_get_setclientid_cred_server(struct nfs_server *server)
294
{
295 296
	struct nfs_client *clp = server->nfs_client;
	struct rpc_cred *cred = NULL;
297
	struct nfs4_state_owner *sp;
298
	struct rb_node *pos;
299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318

	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;
319
	struct rpc_cred *cred;
320

321 322
	spin_lock(&clp->cl_lock);
	cred = nfs4_get_machine_cred_locked(clp);
323
	spin_unlock(&clp->cl_lock);
324 325
	if (cred != NULL)
		goto out;
326 327 328 329 330 331

	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;
332
	}
333 334
	rcu_read_unlock();

335 336
out:
	return cred;
337 338
}

L
Linus Torvalds 已提交
339
static struct nfs4_state_owner *
340
nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
L
Linus Torvalds 已提交
341
{
342
	struct rb_node **p = &server->state_owners.rb_node,
343
		       *parent = NULL;
344
	struct nfs4_state_owner *sp;
L
Linus Torvalds 已提交
345

346 347
	while (*p != NULL) {
		parent = *p;
348
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
349 350 351 352 353 354

		if (cred < sp->so_cred)
			p = &parent->rb_left;
		else if (cred > sp->so_cred)
			p = &parent->rb_right;
		else {
355 356
			if (!list_empty(&sp->so_lru))
				list_del_init(&sp->so_lru);
357
			atomic_inc(&sp->so_count);
358
			return sp;
359
		}
L
Linus Torvalds 已提交
360
	}
361
	return NULL;
L
Linus Torvalds 已提交
362 363
}

364
static struct nfs4_state_owner *
365
nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
366
{
367 368
	struct nfs_server *server = new->so_server;
	struct rb_node **p = &server->state_owners.rb_node,
369 370
		       *parent = NULL;
	struct nfs4_state_owner *sp;
371
	int err;
372 373 374

	while (*p != NULL) {
		parent = *p;
375
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
376 377 378 379 380 381

		if (new->so_cred < sp->so_cred)
			p = &parent->rb_left;
		else if (new->so_cred > sp->so_cred)
			p = &parent->rb_right;
		else {
382 383
			if (!list_empty(&sp->so_lru))
				list_del_init(&sp->so_lru);
384 385 386 387
			atomic_inc(&sp->so_count);
			return sp;
		}
	}
388
	err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
389 390
	if (err)
		return ERR_PTR(err);
391 392
	rb_link_node(&new->so_server_node, parent, p);
	rb_insert_color(&new->so_server_node, &server->state_owners);
393 394 395 396
	return new;
}

static void
397
nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
398
{
399 400 401 402
	struct nfs_server *server = sp->so_server;

	if (!RB_EMPTY_NODE(&sp->so_server_node))
		rb_erase(&sp->so_server_node, &server->state_owners);
403
	ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
404 405
}

406 407 408
static void
nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
{
409
	sc->create_time = ktime_get();
410 411 412 413 414 415 416 417 418 419 420 421 422
	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
423 424 425 426 427 428
/*
 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
 * create a new state_owner.
 *
 */
static struct nfs4_state_owner *
429 430 431
nfs4_alloc_state_owner(struct nfs_server *server,
		struct rpc_cred *cred,
		gfp_t gfp_flags)
L
Linus Torvalds 已提交
432 433 434
{
	struct nfs4_state_owner *sp;

435
	sp = kzalloc(sizeof(*sp), gfp_flags);
L
Linus Torvalds 已提交
436 437
	if (!sp)
		return NULL;
438 439
	sp->so_server = server;
	sp->so_cred = get_rpccred(cred);
440
	spin_lock_init(&sp->so_lock);
L
Linus Torvalds 已提交
441
	INIT_LIST_HEAD(&sp->so_states);
442
	nfs4_init_seqid_counter(&sp->so_seqid);
L
Linus Torvalds 已提交
443
	atomic_set(&sp->so_count, 1);
444
	INIT_LIST_HEAD(&sp->so_lru);
L
Linus Torvalds 已提交
445 446 447
	return sp;
}

448
static void
L
Linus Torvalds 已提交
449 450
nfs4_drop_state_owner(struct nfs4_state_owner *sp)
{
451 452 453
	struct rb_node *rb_node = &sp->so_server_node;

	if (!RB_EMPTY_NODE(rb_node)) {
454 455
		struct nfs_server *server = sp->so_server;
		struct nfs_client *clp = server->nfs_client;
456 457

		spin_lock(&clp->cl_lock);
458 459 460 461
		if (!RB_EMPTY_NODE(rb_node)) {
			rb_erase(rb_node, &server->state_owners);
			RB_CLEAR_NODE(rb_node);
		}
462 463
		spin_unlock(&clp->cl_lock);
	}
L
Linus Torvalds 已提交
464 465
}

466 467
static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
{
468
	nfs4_destroy_seqid_counter(&sp->so_seqid);
469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497
	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);
	}
}

498 499 500 501 502 503 504 505
/**
 * 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,
506 507
					      struct rpc_cred *cred,
					      gfp_t gfp_flags)
L
Linus Torvalds 已提交
508
{
509
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
510 511 512
	struct nfs4_state_owner *sp, *new;

	spin_lock(&clp->cl_lock);
513
	sp = nfs4_find_state_owner_locked(server, cred);
L
Linus Torvalds 已提交
514 515
	spin_unlock(&clp->cl_lock);
	if (sp != NULL)
516
		goto out;
517
	new = nfs4_alloc_state_owner(server, cred, gfp_flags);
518
	if (new == NULL)
519
		goto out;
520 521 522 523 524 525 526
	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));
527 528
	if (sp != new)
		nfs4_free_state_owner(new);
529 530
out:
	nfs4_gc_state_owners(server);
531
	return sp;
L
Linus Torvalds 已提交
532 533
}

534 535 536
/**
 * nfs4_put_state_owner - Release a nfs4_state_owner
 * @sp: state owner data to release
537 538 539 540 541 542 543 544
 *
 * 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.
545
 */
L
Linus Torvalds 已提交
546 547
void nfs4_put_state_owner(struct nfs4_state_owner *sp)
{
548 549
	struct nfs_server *server = sp->so_server;
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
550 551 552

	if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
		return;
553

554 555 556
	sp->so_expires = jiffies;
	list_add_tail(&sp->so_lru, &server->state_owners_lru);
	spin_unlock(&clp->cl_lock);
557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576
}

/**
 * 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
577
	spin_unlock(&clp->cl_lock);
578 579 580 581 582

	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
583 584 585 586 587 588 589
}

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

590
	state = kzalloc(sizeof(*state), GFP_NOFS);
L
Linus Torvalds 已提交
591 592 593 594
	if (!state)
		return NULL;
	atomic_set(&state->count, 1);
	INIT_LIST_HEAD(&state->lock_states);
595
	spin_lock_init(&state->state_lock);
596
	seqlock_init(&state->seqlock);
L
Linus Torvalds 已提交
597 598 599
	return state;
}

600
void
601
nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
602
{
603
	if (state->state == fmode)
604 605
		return;
	/* NB! List reordering - see the reclaim code for why.  */
606 607
	if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
		if (fmode & FMODE_WRITE)
608 609 610 611
			list_move(&state->open_states, &state->owner->so_states);
		else
			list_move_tail(&state->open_states, &state->owner->so_states);
	}
612
	state->state = fmode;
613 614
}

L
Linus Torvalds 已提交
615 616 617 618 619 620 621
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) {
622
		if (state->owner != owner)
L
Linus Torvalds 已提交
623
			continue;
624
		if (atomic_inc_not_zero(&state->count))
L
Linus Torvalds 已提交
625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647
			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();
648
	spin_lock(&owner->so_lock);
L
Linus Torvalds 已提交
649 650 651 652 653 654 655
	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);
656 657
		ihold(inode);
		state->inode = inode;
L
Linus Torvalds 已提交
658
		spin_unlock(&inode->i_lock);
659 660 661 662
		/* 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
663 664
	} else {
		spin_unlock(&inode->i_lock);
665
		spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
666 667 668 669 670 671 672 673 674 675 676 677
		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;

678
	if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
L
Linus Torvalds 已提交
679
		return;
680
	spin_lock(&inode->i_lock);
681
	list_del(&state->inode_states);
L
Linus Torvalds 已提交
682
	list_del(&state->open_states);
683 684
	spin_unlock(&inode->i_lock);
	spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
685 686 687 688 689 690
	iput(inode);
	nfs4_free_open_state(state);
	nfs4_put_state_owner(owner);
}

/*
691
 * Close the current file.
L
Linus Torvalds 已提交
692
 */
693
static void __nfs4_close(struct nfs4_state *state,
694
		fmode_t fmode, gfp_t gfp_mask, int wait)
L
Linus Torvalds 已提交
695 696
{
	struct nfs4_state_owner *owner = state->owner;
697
	int call_close = 0;
698
	fmode_t newstate;
L
Linus Torvalds 已提交
699 700 701

	atomic_inc(&owner->so_count);
	/* Protect against nfs4_find_state() */
702
	spin_lock(&owner->so_lock);
703
	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
704 705 706 707 708 709 710 711 712
		case FMODE_READ:
			state->n_rdonly--;
			break;
		case FMODE_WRITE:
			state->n_wronly--;
			break;
		case FMODE_READ|FMODE_WRITE:
			state->n_rdwr--;
	}
713
	newstate = FMODE_READ|FMODE_WRITE;
714
	if (state->n_rdwr == 0) {
715
		if (state->n_rdonly == 0) {
716
			newstate &= ~FMODE_READ;
717 718 719 720
			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) {
721
			newstate &= ~FMODE_WRITE;
722 723 724 725 726
			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);
727
	}
728
	nfs4_state_set_mode_locked(state, newstate);
729
	spin_unlock(&owner->so_lock);
730

731
	if (!call_close) {
732 733
		nfs4_put_open_state(state);
		nfs4_put_state_owner(owner);
734 735
	} else
		nfs4_do_close(state, gfp_mask, wait);
736 737
}

738
void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
739
{
740
	__nfs4_close(state, fmode, GFP_NOFS, 0);
741 742
}

743
void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
744
{
745
	__nfs4_close(state, fmode, GFP_KERNEL, 1);
L
Linus Torvalds 已提交
746 747 748 749 750 751 752
}

/*
 * Search the state->lock_states for an existing lock_owner
 * that is compatible with current->files
 */
static struct nfs4_lock_state *
753
__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
754 755 756
{
	struct nfs4_lock_state *pos;
	list_for_each_entry(pos, &state->lock_states, ls_locks) {
757
		if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
L
Linus Torvalds 已提交
758
			continue;
759 760 761 762 763 764 765 766 767
		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
768 769 770 771 772 773 774 775 776 777 778
		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.
 *
 */
779
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
780 781
{
	struct nfs4_lock_state *lsp;
782
	struct nfs_server *server = state->owner->so_server;
L
Linus Torvalds 已提交
783

784
	lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
L
Linus Torvalds 已提交
785 786
	if (lsp == NULL)
		return NULL;
787
	nfs4_init_seqid_counter(&lsp->ls_seqid);
L
Linus Torvalds 已提交
788
	atomic_set(&lsp->ls_count, 1);
789
	lsp->ls_state = state;
790 791 792 793 794 795 796 797 798
	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:
799
		goto out_free;
800
	}
801 802
	lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
	if (lsp->ls_seqid.owner_id < 0)
803
		goto out_free;
804
	INIT_LIST_HEAD(&lsp->ls_locks);
L
Linus Torvalds 已提交
805
	return lsp;
806 807 808
out_free:
	kfree(lsp);
	return NULL;
L
Linus Torvalds 已提交
809 810
}

811
void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
812
{
813
	ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
814
	nfs4_destroy_seqid_counter(&lsp->ls_seqid);
815 816 817
	kfree(lsp);
}

L
Linus Torvalds 已提交
818 819 820 821 822
/*
 * Return a compatible lock_state. If no initialized lock_state structure
 * exists, return an uninitialized one.
 *
 */
823
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
824
{
825
	struct nfs4_lock_state *lsp, *new = NULL;
L
Linus Torvalds 已提交
826
	
827 828
	for(;;) {
		spin_lock(&state->state_lock);
829
		lsp = __nfs4_find_lock_state(state, owner, pid, type);
830 831 832 833 834 835 836 837 838 839
		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);
840
		new = nfs4_alloc_lock_state(state, owner, pid, type);
841 842 843 844
		if (new == NULL)
			return NULL;
	}
	spin_unlock(&state->state_lock);
845
	if (new != NULL)
846
		nfs4_free_lock_state(state->owner->so_server, new);
L
Linus Torvalds 已提交
847 848 849 850
	return lsp;
}

/*
851 852
 * Release reference to lock_state, and free it if we see that
 * it is no longer in use
L
Linus Torvalds 已提交
853
 */
854
void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
L
Linus Torvalds 已提交
855
{
856
	struct nfs4_state *state;
L
Linus Torvalds 已提交
857

858 859 860 861 862 863 864 865 866
	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);
867
	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
868 869 870
		if (nfs4_release_lockowner(lsp) == 0)
			return;
	}
871
	nfs4_free_lock_state(lsp->ls_state->owner->so_server, lsp);
L
Linus Torvalds 已提交
872 873
}

874
static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
L
Linus Torvalds 已提交
875
{
876
	struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
L
Linus Torvalds 已提交
877

878 879 880
	dst->fl_u.nfs4_fl.owner = lsp;
	atomic_inc(&lsp->ls_count);
}
L
Linus Torvalds 已提交
881

882
static void nfs4_fl_release_lock(struct file_lock *fl)
L
Linus Torvalds 已提交
883
{
884
	nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
L
Linus Torvalds 已提交
885 886
}

887
static const struct file_lock_operations nfs4_fl_lock_ops = {
888 889 890 891 892
	.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
893
{
894 895 896 897
	struct nfs4_lock_state *lsp;

	if (fl->fl_ops != NULL)
		return 0;
898 899 900
	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)
901 902
		lsp = nfs4_get_lock_state(state, NULL, fl->fl_pid,
				NFS4_FLOCK_LOCK_TYPE);
903 904
	else
		return -EINVAL;
905 906 907 908 909
	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
910 911
}

912
static bool nfs4_copy_lock_stateid(nfs4_stateid *dst, struct nfs4_state *state,
913
		const struct nfs_lockowner *lockowner)
L
Linus Torvalds 已提交
914
{
915
	struct nfs4_lock_state *lsp;
916 917
	fl_owner_t fl_owner;
	pid_t fl_pid;
918
	bool ret = false;
L
Linus Torvalds 已提交
919

920 921 922 923

	if (lockowner == NULL)
		goto out;

924
	if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
925
		goto out;
L
Linus Torvalds 已提交
926

927 928
	fl_owner = lockowner->l_owner;
	fl_pid = lockowner->l_pid;
929
	spin_lock(&state->state_lock);
930
	lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
931
	if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
932
		nfs4_stateid_copy(dst, &lsp->ls_stateid);
933 934
		ret = true;
	}
935
	spin_unlock(&state->state_lock);
L
Linus Torvalds 已提交
936
	nfs4_put_lock_state(lsp);
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,
956
		fmode_t fmode, const struct nfs_lockowner *lockowner)
957 958 959
{
	if (nfs4_copy_delegation_stateid(dst, state->inode, fmode))
		return;
960
	if (nfs4_copy_lock_stateid(dst, state, lockowner))
961 962
		return;
	nfs4_copy_open_stateid(dst, state);
L
Linus Torvalds 已提交
963 964
}

965
struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
966 967 968
{
	struct nfs_seqid *new;

969
	new = kmalloc(sizeof(*new), gfp_mask);
970 971
	if (new != NULL) {
		new->sequence = counter;
972
		INIT_LIST_HEAD(&new->list);
973
		new->task = NULL;
974 975 976 977
	}
	return new;
}

978
void nfs_release_seqid(struct nfs_seqid *seqid)
L
Linus Torvalds 已提交
979
{
980
	struct nfs_seqid_counter *sequence;
981

982 983 984 985 986 987 988 989 990 991 992
	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);
993
	}
994
	spin_unlock(&sequence->lock);
995 996 997 998 999
}

void nfs_free_seqid(struct nfs_seqid *seqid)
{
	nfs_release_seqid(seqid);
1000
	kfree(seqid);
L
Linus Torvalds 已提交
1001 1002 1003
}

/*
1004 1005 1006 1007
 * 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()
 */
1008
static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
1009
{
1010
	BUG_ON(list_first_entry(&seqid->sequence->list, struct nfs_seqid, list) != seqid);
1011 1012 1013 1014
	switch (status) {
		case 0:
			break;
		case -NFS4ERR_BAD_SEQID:
1015 1016
			if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
				return;
1017
			pr_warn_ratelimited("NFS: v4 server returned a bad"
D
Dan Muntz 已提交
1018 1019
					" sequence-id error on an"
					" unconfirmed sequence %p!\n",
1020
					seqid->sequence);
1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038
		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)
{
1039 1040 1041 1042 1043
	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
1044
		nfs4_drop_state_owner(sp);
1045 1046
	if (!nfs4_has_session(server->nfs_client))
		nfs_increment_seqid(status, seqid);
1047 1048 1049 1050 1051 1052 1053 1054 1055
}

/*
 * 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)
{
1056
	nfs_increment_seqid(status, seqid);
1057 1058 1059 1060
}

int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
{
1061
	struct nfs_seqid_counter *sequence = seqid->sequence;
1062 1063 1064
	int status = 0;

	spin_lock(&sequence->lock);
1065
	seqid->task = task;
1066 1067 1068 1069
	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;
1070
	rpc_sleep_on(&sequence->wait, task, NULL);
1071 1072
	status = -EAGAIN;
unlock:
1073 1074
	spin_unlock(&sequence->lock);
	return status;
L
Linus Torvalds 已提交
1075 1076
}

1077
static int nfs4_run_state_manager(void *);
L
Linus Torvalds 已提交
1078

1079
static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
T
Trond Myklebust 已提交
1080 1081
{
	smp_mb__before_clear_bit();
1082
	clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
T
Trond Myklebust 已提交
1083
	smp_mb__after_clear_bit();
1084
	wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
T
Trond Myklebust 已提交
1085 1086 1087
	rpc_wake_up(&clp->cl_rpcwaitq);
}

L
Linus Torvalds 已提交
1088
/*
1089
 * Schedule the nfs_client asynchronous state management routine
L
Linus Torvalds 已提交
1090
 */
1091
void nfs4_schedule_state_manager(struct nfs_client *clp)
L
Linus Torvalds 已提交
1092
{
1093
	struct task_struct *task;
1094
	char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
L
Linus Torvalds 已提交
1095

1096 1097
	if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
		return;
1098 1099
	__module_get(THIS_MODULE);
	atomic_inc(&clp->cl_count);
1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115

	/* 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
1116 1117 1118
}

/*
1119
 * Schedule a lease recovery attempt
L
Linus Torvalds 已提交
1120
 */
1121
void nfs4_schedule_lease_recovery(struct nfs_client *clp)
L
Linus Torvalds 已提交
1122 1123 1124
{
	if (!clp)
		return;
1125 1126
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1127 1128
	dprintk("%s: scheduling lease recovery for server %s\n", __func__,
			clp->cl_hostname);
1129
	nfs4_schedule_state_manager(clp);
L
Linus Torvalds 已提交
1130
}
A
Andy Adamson 已提交
1131
EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
L
Linus Torvalds 已提交
1132

1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144
/*
 * 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);
1145 1146
	dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
			clp->cl_hostname);
1147 1148
}

1149 1150
void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
{
1151
	nfs40_handle_cb_pathdown(clp);
1152 1153 1154
	nfs4_schedule_state_manager(clp);
}

1155
static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1156 1157 1158 1159 1160 1161 1162 1163
{

	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;
	}
1164
	set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1165 1166 1167 1168
	set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
	return 1;
}

1169
static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1170 1171 1172
{
	set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
	clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1173
	set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1174 1175 1176 1177
	set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
	return 1;
}

1178 1179 1180 1181 1182
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);
1183 1184
	dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
			clp->cl_hostname);
1185 1186
	nfs4_schedule_state_manager(clp);
}
A
Andy Adamson 已提交
1187
EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1188

1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
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;
1205
		if (!nfs4_stateid_match(&state->stateid, stateid))
1206 1207 1208 1209 1210 1211 1212 1213 1214 1215
			continue;
		nfs4_state_mark_reclaim_nograce(clp, state);
		found = true;
	}
	spin_unlock(&inode->i_lock);
	if (found)
		nfs4_schedule_state_manager(clp);
}


1216
static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1217 1218
{
	struct inode *inode = state->inode;
1219
	struct nfs_inode *nfsi = NFS_I(inode);
L
Linus Torvalds 已提交
1220 1221 1222
	struct file_lock *fl;
	int status = 0;

1223 1224 1225 1226
	if (inode->i_flock == NULL)
		return 0;

	/* Guard against delegation returns and new lock/unlock calls */
1227
	down_write(&nfsi->rwsem);
1228
	/* Protect inode->i_flock using the BKL */
1229
	lock_flocks();
H
Harvey Harrison 已提交
1230
	for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1231
		if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
L
Linus Torvalds 已提交
1232
			continue;
1233
		if (nfs_file_open_context(fl->fl_file)->state != state)
L
Linus Torvalds 已提交
1234
			continue;
1235
		unlock_flocks();
L
Linus Torvalds 已提交
1236 1237
		status = ops->recover_lock(state, fl);
		switch (status) {
1238 1239 1240 1241 1242 1243 1244 1245 1246
			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:
1247 1248 1249 1250
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1251
				goto out;
L
Linus Torvalds 已提交
1252
			default:
1253 1254
				printk(KERN_ERR "NFS: %s: unhandled error %d. "
					"Zeroing state\n", __func__, status);
1255 1256
			case -ENOMEM:
			case -NFS4ERR_DENIED:
L
Linus Torvalds 已提交
1257 1258
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1259
				/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1260
				status = 0;
L
Linus Torvalds 已提交
1261
		}
1262
		lock_flocks();
L
Linus Torvalds 已提交
1263
	}
1264
	unlock_flocks();
1265
out:
1266
	up_write(&nfsi->rwsem);
L
Linus Torvalds 已提交
1267 1268 1269
	return status;
}

1270
static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283
{
	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.
	 */
1284 1285
restart:
	spin_lock(&sp->so_lock);
L
Linus Torvalds 已提交
1286
	list_for_each_entry(state, &sp->so_states, open_states) {
1287 1288
		if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
			continue;
L
Linus Torvalds 已提交
1289 1290
		if (state->state == 0)
			continue;
1291 1292
		atomic_inc(&state->count);
		spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1293 1294
		status = ops->recover_open(sp, state);
		if (status >= 0) {
1295 1296
			status = nfs4_reclaim_locks(state, ops);
			if (status >= 0) {
1297
				spin_lock(&state->state_lock);
1298
				list_for_each_entry(lock, &state->lock_states, ls_locks) {
1299
					if (!test_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags))
1300 1301
						pr_warn_ratelimited("NFS: "
							"%s: Lock reclaim "
1302
							"failed!\n", __func__);
1303
				}
1304
				spin_unlock(&state->state_lock);
1305 1306
				nfs4_put_open_state(state);
				goto restart;
L
Linus Torvalds 已提交
1307 1308 1309 1310
			}
		}
		switch (status) {
			default:
1311 1312
				printk(KERN_ERR "NFS: %s: unhandled error %d. "
					"Zeroing state\n", __func__, status);
L
Linus Torvalds 已提交
1313
			case -ENOENT:
1314
			case -ENOMEM:
1315
			case -ESTALE:
L
Linus Torvalds 已提交
1316 1317 1318 1319
				/*
				 * Open state on this file cannot be recovered
				 * All we can do is revert to using the zero stateid.
				 */
1320 1321
				memset(&state->stateid, 0,
					sizeof(state->stateid));
L
Linus Torvalds 已提交
1322 1323 1324
				/* Mark the file as being 'closed' */
				state->state = 0;
				break;
1325 1326 1327 1328 1329 1330 1331 1332
			case -EKEYEXPIRED:
				/*
				 * User RPCSEC_GSS context has expired.
				 * We cannot recover this stateid now, so
				 * skip it and allow recovery thread to
				 * proceed.
				 */
				break;
1333 1334 1335
			case -NFS4ERR_ADMIN_REVOKED:
			case -NFS4ERR_STALE_STATEID:
			case -NFS4ERR_BAD_STATEID:
1336 1337
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1338
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1339
				break;
L
Linus Torvalds 已提交
1340 1341
			case -NFS4ERR_EXPIRED:
			case -NFS4ERR_NO_GRACE:
1342
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
L
Linus Torvalds 已提交
1343
			case -NFS4ERR_STALE_CLIENTID:
1344 1345 1346 1347
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
L
Linus Torvalds 已提交
1348 1349
				goto out_err;
		}
1350 1351
		nfs4_put_open_state(state);
		goto restart;
L
Linus Torvalds 已提交
1352
	}
1353
	spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1354 1355
	return 0;
out_err:
1356
	nfs4_put_open_state(state);
L
Linus Torvalds 已提交
1357 1358 1359
	return status;
}

1360 1361 1362 1363 1364 1365 1366 1367
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);
1368
	spin_lock(&state->state_lock);
1369 1370
	list_for_each_entry(lock, &state->lock_states, ls_locks) {
		lock->ls_seqid.flags = 0;
1371
		clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
1372
	}
1373
	spin_unlock(&state->state_lock);
1374 1375
}

1376 1377
static void nfs4_reset_seqids(struct nfs_server *server,
	int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1378
{
1379
	struct nfs_client *clp = server->nfs_client;
1380
	struct nfs4_state_owner *sp;
1381
	struct rb_node *pos;
1382 1383
	struct nfs4_state *state;

1384 1385 1386 1387 1388
	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);
1389
		sp->so_seqid.flags = 0;
1390
		spin_lock(&sp->so_lock);
1391
		list_for_each_entry(state, &sp->so_states, open_states) {
1392 1393
			if (mark_reclaim(clp, state))
				nfs4_clear_open_state(state);
1394
		}
1395
		spin_unlock(&sp->so_lock);
1396
	}
1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408
	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();
1409 1410
}

1411 1412 1413 1414 1415 1416 1417
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);
}

1418 1419 1420 1421 1422 1423 1424 1425
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);
}

1426
static void nfs4_clear_reclaim_server(struct nfs_server *server)
1427
{
1428
	struct nfs_client *clp = server->nfs_client;
1429 1430 1431 1432
	struct nfs4_state_owner *sp;
	struct rb_node *pos;
	struct nfs4_state *state;

1433 1434 1435 1436 1437
	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);
1438 1439
		spin_lock(&sp->so_lock);
		list_for_each_entry(state, &sp->so_states, open_states) {
1440 1441
			if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
						&state->flags))
1442 1443 1444 1445 1446
				continue;
			nfs4_state_mark_reclaim_nograce(clp, state);
		}
		spin_unlock(&sp->so_lock);
	}
1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460
	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();
1461 1462

	nfs_delegation_reap_unclaimed(clp);
1463 1464 1465 1466 1467 1468 1469 1470
	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);
1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484
}

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);
}

1485 1486 1487 1488 1489 1490 1491 1492
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);
}

1493
static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1494 1495
{
	switch (error) {
1496 1497
		case 0:
			break;
1498
		case -NFS4ERR_CB_PATH_DOWN:
1499
			nfs40_handle_cb_pathdown(clp);
1500
			break;
1501 1502
		case -NFS4ERR_NO_GRACE:
			nfs4_state_end_reclaim_reboot(clp);
1503
			break;
1504 1505 1506
		case -NFS4ERR_STALE_CLIENTID:
		case -NFS4ERR_LEASE_MOVED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1507
			nfs4_state_clear_reclaim_reboot(clp);
1508 1509 1510 1511 1512
			nfs4_state_start_reclaim_reboot(clp);
			break;
		case -NFS4ERR_EXPIRED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
			nfs4_state_start_reclaim_nograce(clp);
1513
			break;
1514 1515 1516 1517 1518 1519
		case -NFS4ERR_BADSESSION:
		case -NFS4ERR_BADSLOT:
		case -NFS4ERR_BAD_HIGH_SLOT:
		case -NFS4ERR_DEADSESSION:
		case -NFS4ERR_SEQ_FALSE_RETRY:
		case -NFS4ERR_SEQ_MISORDERED:
1520
			set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1521
			/* Zero session reset errors */
1522
			break;
1523 1524 1525
		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
			set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
			break;
1526 1527 1528
		case -EKEYEXPIRED:
			/* Nothing we can do */
			nfs4_warn_keyexpired(clp->cl_hostname);
1529 1530
			break;
		default:
1531 1532
			dprintk("%s: failed to handle error %d for server %s\n",
					__func__, error, clp->cl_hostname);
1533
			return error;
1534
	}
1535 1536
	dprintk("%s: handled error %d for server %s\n", __func__, error,
			clp->cl_hostname);
1537
	return 0;
1538 1539
}

1540
static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1541
{
1542 1543
	struct nfs4_state_owner *sp;
	struct nfs_server *server;
1544
	struct rb_node *pos;
L
Linus Torvalds 已提交
1545 1546
	int status = 0;

1547
restart:
1548 1549
	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1550
		nfs4_purge_state_owners(server);
1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570
		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);
			}

1571
			nfs4_put_state_owner(sp);
1572
			goto restart;
1573
		}
1574
		spin_unlock(&clp->cl_lock);
1575
	}
1576
	rcu_read_unlock();
1577 1578 1579 1580 1581 1582
	return status;
}

static int nfs4_check_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1583 1584
	const struct nfs4_state_maintenance_ops *ops =
		clp->cl_mvops->state_renewal_ops;
1585
	int status;
L
Linus Torvalds 已提交
1586

1587 1588 1589
	/* Is the client already known to have an expired lease? */
	if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		return 0;
1590 1591 1592
	spin_lock(&clp->cl_lock);
	cred = ops->get_state_renewal_cred_locked(clp);
	spin_unlock(&clp->cl_lock);
1593 1594
	if (cred == NULL) {
		cred = nfs4_get_setclientid_cred(clp);
1595
		status = -ENOKEY;
1596 1597
		if (cred == NULL)
			goto out;
1598
	}
1599
	status = ops->renew_lease(clp, cred);
1600 1601
	put_rpccred(cred);
out:
1602
	return nfs4_recovery_handle_error(clp, status);
1603 1604
}

1605 1606 1607 1608 1609 1610
/* 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) {
1611 1612 1613 1614 1615
	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);
1616 1617 1618
	case -NFS4ERR_STALE_CLIENTID:
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
		break;
1619 1620 1621 1622 1623 1624
	case -NFS4ERR_CLID_INUSE:
		pr_err("NFS: Server %s reports our clientid is in use\n",
			clp->cl_hostname);
		nfs_mark_client_ready(clp, -EPERM);
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
		return -EPERM;
1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638
	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);
1639 1640
		dprintk("%s: exit with error %d for server %s\n",
				__func__, -EPROTONOSUPPORT, clp->cl_hostname);
1641 1642 1643 1644 1645 1646
		return -EPROTONOSUPPORT;
	case -EKEYEXPIRED:
		nfs4_warn_keyexpired(clp->cl_hostname);
	case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
				 * in nfs4_exchange_id */
	default:
1647 1648
		dprintk("%s: exit with error %d for server %s\n", __func__,
				status, clp->cl_hostname);
1649 1650 1651
		return status;
	}
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1652 1653
	dprintk("%s: handled error %d for server %s\n", __func__, status,
			clp->cl_hostname);
1654 1655 1656
	return 0;
}

1657
static int nfs4_establish_lease(struct nfs_client *clp)
1658 1659
{
	struct rpc_cred *cred;
1660 1661
	const struct nfs4_state_recovery_ops *ops =
		clp->cl_mvops->reboot_recovery_ops;
1662
	int status;
1663

1664
	cred = ops->get_clid_cred(clp);
1665 1666 1667 1668 1669
	if (cred == NULL)
		return -ENOENT;
	status = ops->establish_clid(clp, cred);
	put_rpccred(cred);
	if (status != 0)
1670 1671 1672 1673 1674
		return status;
	pnfs_destroy_all_layouts(clp);
	return 0;
}

1675 1676 1677 1678
/*
 * Returns zero or a negative errno.  NFS4ERR values are converted
 * to local errno values.
 */
1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700
static int nfs4_reclaim_lease(struct nfs_client *clp)
{
	int status;

	status = nfs4_establish_lease(clp);
	if (status < 0)
		return nfs4_handle_reclaim_lease_error(clp, status);
	if (test_and_clear_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state))
		nfs4_state_start_reclaim_nograce(clp);
	if (!test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
		set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
	clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
	clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
	return 0;
}

static int nfs4_purge_lease(struct nfs_client *clp)
{
	int status;

	status = nfs4_establish_lease(clp);
	if (status < 0)
1701
		return nfs4_handle_reclaim_lease_error(clp, status);
1702 1703 1704
	clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
	nfs4_state_start_reclaim_nograce(clp);
1705
	return 0;
1706 1707
}

1708
#ifdef CONFIG_NFS_V4_1
1709
void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
1710
{
1711 1712
	struct nfs_client *clp = session->clp;

1713 1714 1715 1716 1717 1718 1719
	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);
	}
1720
	nfs4_schedule_lease_recovery(clp);
1721
}
1722
EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
1723

A
Andy Adamson 已提交
1724 1725 1726
void nfs41_handle_recall_slot(struct nfs_client *clp)
{
	set_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
1727 1728
	dprintk("%s: scheduling slot recall for server %s\n", __func__,
			clp->cl_hostname);
1729
	nfs4_schedule_state_manager(clp);
A
Andy Adamson 已提交
1730 1731
}

1732 1733 1734
static void nfs4_reset_all_state(struct nfs_client *clp)
{
	if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1735
		set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1736
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1737
		nfs4_state_start_reclaim_nograce(clp);
1738 1739
		dprintk("%s: scheduling reset of all state for server %s!\n",
				__func__, clp->cl_hostname);
1740
		nfs4_schedule_state_manager(clp);
1741 1742 1743 1744 1745 1746 1747
	}
}

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);
1748 1749
		dprintk("%s: server %s rebooted!\n", __func__,
				clp->cl_hostname);
1750
		nfs4_schedule_state_manager(clp);
1751 1752 1753 1754 1755 1756
	}
}

static void nfs41_handle_state_revoked(struct nfs_client *clp)
{
	nfs4_reset_all_state(clp);
1757
	dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
1758 1759 1760 1761 1762 1763
}

static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
{
	/* This will need to handle layouts too */
	nfs_expire_all_delegations(clp);
1764 1765
	dprintk("%s: Recallable state revoked on server %s!\n", __func__,
			clp->cl_hostname);
1766 1767
}

1768
static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
1769 1770 1771
{
	nfs_expire_all_delegations(clp);
	if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
1772
		nfs4_schedule_state_manager(clp);
1773 1774
	dprintk("%s: server %s declared a backchannel fault\n", __func__,
			clp->cl_hostname);
1775 1776
}

1777 1778 1779 1780 1781 1782 1783
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);
}

1784 1785 1786 1787
void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
{
	if (!flags)
		return;
1788 1789 1790 1791

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

1792
	if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
1793
		nfs41_handle_server_reboot(clp);
1794
	if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
1795 1796
			    SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
			    SEQ4_STATUS_ADMIN_STATE_REVOKED |
1797 1798
			    SEQ4_STATUS_LEASE_MOVED))
		nfs41_handle_state_revoked(clp);
1799
	if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
1800
		nfs41_handle_recallable_state_revoked(clp);
1801 1802 1803 1804
	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))
1805
		nfs41_handle_cb_path_down(clp);
1806 1807
}

1808 1809
static int nfs4_reset_session(struct nfs_client *clp)
{
1810
	struct rpc_cred *cred;
1811 1812
	int status;

1813 1814
	if (!nfs4_has_session(clp))
		return 0;
1815
	nfs4_begin_drain_session(clp);
1816 1817
	cred = nfs4_get_exchange_id_cred(clp);
	status = nfs4_proc_destroy_session(clp->cl_session, cred);
1818 1819
	if (status && status != -NFS4ERR_BADSESSION &&
	    status != -NFS4ERR_DEADSESSION) {
1820
		status = nfs4_recovery_handle_error(clp, status);
1821 1822 1823 1824
		goto out;
	}

	memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
1825
	status = nfs4_proc_create_session(clp, cred);
1826
	if (status) {
1827 1828
		dprintk("%s: session reset failed with status %d for server %s!\n",
			__func__, status, clp->cl_hostname);
1829
		status = nfs4_handle_reclaim_lease_error(clp, status);
1830 1831
		goto out;
	}
1832
	nfs41_finish_session_reset(clp);
1833 1834
	dprintk("%s: session reset was successful for server %s!\n",
			__func__, clp->cl_hostname);
1835
out:
1836 1837
	if (cred)
		put_rpccred(cred);
1838 1839
	return status;
}
1840

A
Andy Adamson 已提交
1841 1842
static int nfs4_recall_slot(struct nfs_client *clp)
{
1843
	struct nfs4_slot_table *fc_tbl;
A
Andy Adamson 已提交
1844 1845 1846
	struct nfs4_slot *new, *old;
	int i;

1847 1848
	if (!nfs4_has_session(clp))
		return 0;
A
Andy Adamson 已提交
1849
	nfs4_begin_drain_session(clp);
1850
	fc_tbl = &clp->cl_session->fc_slot_table;
A
Andy Adamson 已提交
1851
	new = kmalloc(fc_tbl->target_max_slots * sizeof(struct nfs4_slot),
1852
		      GFP_NOFS);
A
Andy Adamson 已提交
1853 1854 1855 1856 1857 1858 1859 1860 1861 1862
        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;
1863
	clp->cl_session->fc_attrs.max_reqs = fc_tbl->max_slots;
A
Andy Adamson 已提交
1864 1865 1866 1867 1868 1869
	spin_unlock(&fc_tbl->slot_tbl_lock);

	kfree(old);
	return 0;
}

1870 1871
static int nfs4_bind_conn_to_session(struct nfs_client *clp)
{
1872 1873 1874
	struct rpc_cred *cred;
	int ret;

1875 1876
	if (!nfs4_has_session(clp))
		return 0;
1877
	nfs4_begin_drain_session(clp);
1878 1879 1880 1881
	cred = nfs4_get_exchange_id_cred(clp);
	ret = nfs4_proc_bind_conn_to_session(clp, cred);
	if (cred)
		put_rpccred(cred);
1882
	clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
1883 1884
	switch (ret) {
	case 0:
1885 1886
		dprintk("%s: bind_conn_to_session was successful for server %s!\n",
			__func__, clp->cl_hostname);
1887 1888 1889 1890 1891 1892 1893 1894 1895
		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;
1896
}
1897
#else /* CONFIG_NFS_V4_1 */
1898
static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
1899
static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
A
Andy Adamson 已提交
1900
static int nfs4_recall_slot(struct nfs_client *clp) { return 0; }
1901 1902 1903 1904 1905

static int nfs4_bind_conn_to_session(struct nfs_client *clp)
{
	return 0;
}
1906 1907
#endif /* CONFIG_NFS_V4_1 */

1908
static void nfs4_state_manager(struct nfs_client *clp)
1909 1910
{
	int status = 0;
1911
	const char *section = "", *section_sep = "";
1912 1913

	/* Ensure exclusive access to NFSv4 state */
1914
	do {
1915
		if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
1916
			section = "purge state";
1917
			status = nfs4_purge_lease(clp);
1918 1919
			if (status < 0)
				goto out_error;
1920
			continue;
1921 1922
		}

1923
		if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
1924
			section = "lease expired";
1925 1926
			/* We're going to have to re-establish a clientid */
			status = nfs4_reclaim_lease(clp);
1927
			if (status < 0)
1928
				goto out_error;
1929
			continue;
1930 1931 1932
		}

		if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
1933
			section = "check lease";
1934
			status = nfs4_check_lease(clp);
1935 1936
			if (status < 0)
				goto out_error;
1937
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1938
				continue;
1939
		}
1940

1941
		/* Initialize or reset the session */
1942
		if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
1943
			section = "reset session";
1944
			status = nfs4_reset_session(clp);
1945 1946 1947
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
				continue;
			if (status < 0)
1948 1949
				goto out_error;
		}
1950

1951 1952
		/* Send BIND_CONN_TO_SESSION */
		if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
1953
				&clp->cl_state)) {
1954
			section = "bind conn to session";
1955 1956 1957
			status = nfs4_bind_conn_to_session(clp);
			if (status < 0)
				goto out_error;
1958
			continue;
1959 1960
		}

1961
		/* Recall session slots */
1962
		if (test_and_clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state)) {
1963 1964 1965 1966 1967 1968 1969
			section = "recall slot";
			status = nfs4_recall_slot(clp);
			if (status < 0)
				goto out_error;
			continue;
		}

1970
		/* First recover reboot state... */
1971
		if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
1972
			section = "reclaim reboot";
1973
			status = nfs4_do_reclaim(clp,
1974
				clp->cl_mvops->reboot_recovery_ops);
1975
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1976
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
1977
				continue;
1978
			nfs4_state_end_reclaim_reboot(clp);
1979 1980 1981 1982
			if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
				continue;
			if (status < 0)
				goto out_error;
1983 1984
		}

1985 1986
		/* Now recover expired state... */
		if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
1987
			section = "reclaim nograce";
1988
			status = nfs4_do_reclaim(clp,
1989
				clp->cl_mvops->nograce_recovery_ops);
1990
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1991
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
1992 1993 1994
			    test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
				continue;
			if (status < 0)
1995
				goto out_error;
L
Linus Torvalds 已提交
1996
		}
1997

1998
		nfs4_end_drain_session(clp);
1999 2000 2001 2002
		if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
			nfs_client_return_marked_delegations(clp);
			continue;
		}
2003 2004

		nfs4_clear_state_manager_bit(clp);
2005 2006 2007 2008 2009
		/* 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;
2010
	} while (atomic_read(&clp->cl_count) > 1);
2011
	return;
L
Linus Torvalds 已提交
2012
out_error:
2013 2014 2015 2016 2017
	if (strlen(section))
		section_sep = ": ";
	pr_warn_ratelimited("NFS: state manager%s%s failed on NFSv4 server %s"
			" with error %d\n", section_sep, section,
			clp->cl_hostname, -status);
2018
	nfs4_end_drain_session(clp);
2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030
	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
2031 2032 2033 2034 2035 2036 2037
}

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