nfs4state.c 58.3 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 55
#include <linux/sunrpc/clnt.h>

56
#include "nfs4_fs.h"
L
Linus Torvalds 已提交
57 58
#include "callback.h"
#include "delegation.h"
59
#include "internal.h"
60
#include "nfs4session.h"
61
#include "pnfs.h"
62
#include "netns.h"
L
Linus Torvalds 已提交
63

C
Chuck Lever 已提交
64 65
#define NFSDBG_FACILITY		NFSDBG_STATE

L
Linus Torvalds 已提交
66 67
#define OPENOWNER_POOL_SIZE	8

68
const nfs4_stateid zero_stateid;
69
static DEFINE_MUTEX(nfs_clid_init_mutex);
L
Linus Torvalds 已提交
70

71
int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
L
Linus Torvalds 已提交
72
{
73 74 75 76
	struct nfs4_setclientid_res clid = {
		.clientid = clp->cl_clientid,
		.confirm = clp->cl_confirm,
	};
77 78
	unsigned short port;
	int status;
79
	struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
80

81 82
	if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
		goto do_confirm;
83
	port = nn->nfs_callback_tcpport;
84
	if (clp->cl_addr.ss_family == AF_INET6)
85
		port = nn->nfs_callback_tcpport6;
86

87 88 89
	status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
	if (status != 0)
		goto out;
90 91 92 93
	clp->cl_clientid = clid.clientid;
	clp->cl_confirm = clid.confirm;
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
do_confirm:
94 95 96
	status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
	if (status != 0)
		goto out;
97
	clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
98 99
	nfs4_schedule_state_renewal(clp);
out:
L
Linus Torvalds 已提交
100 101 102
	return status;
}

103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123
/**
 * nfs40_discover_server_trunking - Detect server IP address trunking (mv0)
 *
 * @clp: nfs_client under test
 * @result: OUT: found nfs_client, or clp
 * @cred: credential to use for trunking test
 *
 * Returns zero, a negative errno, or a negative NFS4ERR status.
 * If zero is returned, an nfs_client pointer is planted in
 * "result".
 *
 * Note: The returned client may not yet be marked ready.
 */
int nfs40_discover_server_trunking(struct nfs_client *clp,
				   struct nfs_client **result,
				   struct rpc_cred *cred)
{
	struct nfs4_setclientid_res clid = {
		.clientid = clp->cl_clientid,
		.confirm = clp->cl_confirm,
	};
124
	struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
125 126 127
	unsigned short port;
	int status;

128
	port = nn->nfs_callback_tcpport;
129
	if (clp->cl_addr.ss_family == AF_INET6)
130
		port = nn->nfs_callback_tcpport6;
131 132 133 134 135 136 137 138

	status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
	if (status != 0)
		goto out;
	clp->cl_clientid = clid.clientid;
	clp->cl_confirm = clid.confirm;

	status = nfs40_walk_client_list(clp, result, cred);
139
	if (status == 0) {
140 141 142 143 144 145 146 147
		/* Sustain the lease, even if it's empty.  If the clientid4
		 * goes stale it's of no use for trunking discovery. */
		nfs4_schedule_state_renewal(*result);
	}
out:
	return status;
}

148
struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
149 150 151 152 153 154 155 156
{
	struct rpc_cred *cred = NULL;

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

157 158 159 160 161 162 163 164 165 166 167 168 169
static void nfs4_root_machine_cred(struct nfs_client *clp)
{
	struct rpc_cred *cred, *new;

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

170 171
static struct rpc_cred *
nfs4_get_renew_cred_server_locked(struct nfs_server *server)
172
{
173
	struct rpc_cred *cred = NULL;
174
	struct nfs4_state_owner *sp;
175
	struct rb_node *pos;
176

177 178 179 180
	for (pos = rb_first(&server->state_owners);
	     pos != NULL;
	     pos = rb_next(pos)) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
181 182 183 184 185 186 187 188
		if (list_empty(&sp->so_states))
			continue;
		cred = get_rpccred(sp->so_cred);
		break;
	}
	return cred;
}

189 190 191 192 193 194 195 196 197 198 199 200
/**
 * 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;

201 202 203 204 205
	/* Use machine credentials if available */
	cred = nfs4_get_machine_cred_locked(clp);
	if (cred != NULL)
		goto out;

206 207 208 209 210 211 212
	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();
213 214

out:
215 216 217
	return cred;
}

218
static void nfs4_end_drain_slot_table(struct nfs4_slot_table *tbl)
219
{
220
	if (test_and_clear_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state)) {
221
		spin_lock(&tbl->slot_tbl_lock);
222
		nfs41_wake_slot_table(tbl);
223
		spin_unlock(&tbl->slot_tbl_lock);
224
	}
225 226
}

227 228 229 230
static void nfs4_end_drain_session(struct nfs_client *clp)
{
	struct nfs4_session *ses = clp->cl_session;

231 232 233 234 235
	if (clp->cl_slot_tbl) {
		nfs4_end_drain_slot_table(clp->cl_slot_tbl);
		return;
	}

236 237 238 239 240 241
	if (ses != NULL) {
		nfs4_end_drain_slot_table(&ses->bc_slot_table);
		nfs4_end_drain_slot_table(&ses->fc_slot_table);
	}
}

242 243
#if defined(CONFIG_NFS_V4_1)

244
static int nfs4_drain_slot_tbl(struct nfs4_slot_table *tbl)
245
{
246
	set_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state);
247
	spin_lock(&tbl->slot_tbl_lock);
248
	if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
249
		INIT_COMPLETION(tbl->complete);
250
		spin_unlock(&tbl->slot_tbl_lock);
251
		return wait_for_completion_interruptible(&tbl->complete);
252 253 254 255 256
	}
	spin_unlock(&tbl->slot_tbl_lock);
	return 0;
}

257 258 259 260 261
static int nfs4_begin_drain_session(struct nfs_client *clp)
{
	struct nfs4_session *ses = clp->cl_session;
	int ret = 0;

262 263 264
	if (clp->cl_slot_tbl)
		return nfs4_drain_slot_tbl(clp->cl_slot_tbl);

265
	/* back channel */
266
	ret = nfs4_drain_slot_tbl(&ses->bc_slot_table);
267 268 269
	if (ret)
		return ret;
	/* fore channel */
270
	return nfs4_drain_slot_tbl(&ses->fc_slot_table);
271 272
}

273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296
static int nfs41_setup_state_renewal(struct nfs_client *clp)
{
	int status;
	struct nfs_fsinfo fsinfo;

	if (!test_bit(NFS_CS_CHECK_LEASE_TIME, &clp->cl_res_state)) {
		nfs4_schedule_state_renewal(clp);
		return 0;
	}

	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;
}

297 298 299 300 301 302 303 304 305
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_BIND_CONN_TO_SESSION, &clp->cl_state);
	nfs41_setup_state_renewal(clp);
}

306 307 308 309
int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
{
	int status;

310 311
	if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
		goto do_confirm;
312
	nfs4_begin_drain_session(clp);
313
	status = nfs4_proc_exchange_id(clp, cred);
R
Ricardo Labiaga 已提交
314 315
	if (status != 0)
		goto out;
316 317
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
do_confirm:
318
	status = nfs4_proc_create_session(clp, cred);
R
Ricardo Labiaga 已提交
319 320
	if (status != 0)
		goto out;
321
	nfs41_finish_session_reset(clp);
R
Ricardo Labiaga 已提交
322 323
	nfs_mark_client_ready(clp, NFS_CS_READY);
out:
324 325 326
	return status;
}

327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348
/**
 * nfs41_discover_server_trunking - Detect server IP address trunking (mv1)
 *
 * @clp: nfs_client under test
 * @result: OUT: found nfs_client, or clp
 * @cred: credential to use for trunking test
 *
 * Returns NFS4_OK, a negative errno, or a negative NFS4ERR status.
 * If NFS4_OK is returned, an nfs_client pointer is planted in
 * "result".
 *
 * Note: The returned client may not yet be marked ready.
 */
int nfs41_discover_server_trunking(struct nfs_client *clp,
				   struct nfs_client **result,
				   struct rpc_cred *cred)
{
	int status;

	status = nfs4_proc_exchange_id(clp, cred);
	if (status != NFS4_OK)
		return status;
349
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
350 351 352 353

	return nfs41_walk_client_list(clp, result, cred);
}

354 355
#endif /* CONFIG_NFS_V4_1 */

356
/**
357
 * nfs4_get_clid_cred - Acquire credential for a setclientid operation
358 359 360 361
 * @clp: client state handle
 *
 * Returns an rpc_cred with reference count bumped, or NULL.
 */
362
struct rpc_cred *nfs4_get_clid_cred(struct nfs_client *clp)
363
{
364
	struct rpc_cred *cred;
365

366 367
	spin_lock(&clp->cl_lock);
	cred = nfs4_get_machine_cred_locked(clp);
368
	spin_unlock(&clp->cl_lock);
369
	return cred;
370 371
}

L
Linus Torvalds 已提交
372
static struct nfs4_state_owner *
373
nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
L
Linus Torvalds 已提交
374
{
375
	struct rb_node **p = &server->state_owners.rb_node,
376
		       *parent = NULL;
377
	struct nfs4_state_owner *sp;
L
Linus Torvalds 已提交
378

379 380
	while (*p != NULL) {
		parent = *p;
381
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
382 383 384 385 386 387

		if (cred < sp->so_cred)
			p = &parent->rb_left;
		else if (cred > sp->so_cred)
			p = &parent->rb_right;
		else {
388 389
			if (!list_empty(&sp->so_lru))
				list_del_init(&sp->so_lru);
390
			atomic_inc(&sp->so_count);
391
			return sp;
392
		}
L
Linus Torvalds 已提交
393
	}
394
	return NULL;
L
Linus Torvalds 已提交
395 396
}

397
static struct nfs4_state_owner *
398
nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
399
{
400 401
	struct nfs_server *server = new->so_server;
	struct rb_node **p = &server->state_owners.rb_node,
402 403
		       *parent = NULL;
	struct nfs4_state_owner *sp;
404
	int err;
405 406 407

	while (*p != NULL) {
		parent = *p;
408
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
409 410 411 412 413 414

		if (new->so_cred < sp->so_cred)
			p = &parent->rb_left;
		else if (new->so_cred > sp->so_cred)
			p = &parent->rb_right;
		else {
415 416
			if (!list_empty(&sp->so_lru))
				list_del_init(&sp->so_lru);
417 418 419 420
			atomic_inc(&sp->so_count);
			return sp;
		}
	}
421
	err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
422 423
	if (err)
		return ERR_PTR(err);
424 425
	rb_link_node(&new->so_server_node, parent, p);
	rb_insert_color(&new->so_server_node, &server->state_owners);
426 427 428 429
	return new;
}

static void
430
nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
431
{
432 433 434 435
	struct nfs_server *server = sp->so_server;

	if (!RB_EMPTY_NODE(&sp->so_server_node))
		rb_erase(&sp->so_server_node, &server->state_owners);
436
	ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
437 438
}

439 440 441
static void
nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
{
442
	sc->create_time = ktime_get();
443 444 445 446 447 448 449 450 451 452 453 454 455
	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
456 457 458 459 460 461
/*
 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
 * create a new state_owner.
 *
 */
static struct nfs4_state_owner *
462 463 464
nfs4_alloc_state_owner(struct nfs_server *server,
		struct rpc_cred *cred,
		gfp_t gfp_flags)
L
Linus Torvalds 已提交
465 466 467
{
	struct nfs4_state_owner *sp;

468
	sp = kzalloc(sizeof(*sp), gfp_flags);
L
Linus Torvalds 已提交
469 470
	if (!sp)
		return NULL;
471 472
	sp->so_server = server;
	sp->so_cred = get_rpccred(cred);
473
	spin_lock_init(&sp->so_lock);
L
Linus Torvalds 已提交
474
	INIT_LIST_HEAD(&sp->so_states);
475
	nfs4_init_seqid_counter(&sp->so_seqid);
L
Linus Torvalds 已提交
476
	atomic_set(&sp->so_count, 1);
477
	INIT_LIST_HEAD(&sp->so_lru);
478
	seqcount_init(&sp->so_reclaim_seqcount);
479
	mutex_init(&sp->so_delegreturn_mutex);
L
Linus Torvalds 已提交
480 481 482
	return sp;
}

483
static void
L
Linus Torvalds 已提交
484 485
nfs4_drop_state_owner(struct nfs4_state_owner *sp)
{
486 487 488
	struct rb_node *rb_node = &sp->so_server_node;

	if (!RB_EMPTY_NODE(rb_node)) {
489 490
		struct nfs_server *server = sp->so_server;
		struct nfs_client *clp = server->nfs_client;
491 492

		spin_lock(&clp->cl_lock);
493 494 495 496
		if (!RB_EMPTY_NODE(rb_node)) {
			rb_erase(rb_node, &server->state_owners);
			RB_CLEAR_NODE(rb_node);
		}
497 498
		spin_unlock(&clp->cl_lock);
	}
L
Linus Torvalds 已提交
499 500
}

501 502
static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
{
503
	nfs4_destroy_seqid_counter(&sp->so_seqid);
504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532
	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);
	}
}

533 534 535 536 537 538 539 540
/**
 * 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,
541 542
					      struct rpc_cred *cred,
					      gfp_t gfp_flags)
L
Linus Torvalds 已提交
543
{
544
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
545 546 547
	struct nfs4_state_owner *sp, *new;

	spin_lock(&clp->cl_lock);
548
	sp = nfs4_find_state_owner_locked(server, cred);
L
Linus Torvalds 已提交
549 550
	spin_unlock(&clp->cl_lock);
	if (sp != NULL)
551
		goto out;
552
	new = nfs4_alloc_state_owner(server, cred, gfp_flags);
553
	if (new == NULL)
554
		goto out;
555 556 557 558 559 560 561
	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));
562 563
	if (sp != new)
		nfs4_free_state_owner(new);
564 565
out:
	nfs4_gc_state_owners(server);
566
	return sp;
L
Linus Torvalds 已提交
567 568
}

569 570 571
/**
 * nfs4_put_state_owner - Release a nfs4_state_owner
 * @sp: state owner data to release
572 573 574 575 576 577 578 579
 *
 * 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.
580
 */
L
Linus Torvalds 已提交
581 582
void nfs4_put_state_owner(struct nfs4_state_owner *sp)
{
583 584
	struct nfs_server *server = sp->so_server;
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
585 586 587

	if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
		return;
588

589 590 591
	sp->so_expires = jiffies;
	list_add_tail(&sp->so_lru, &server->state_owners_lru);
	spin_unlock(&clp->cl_lock);
592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611
}

/**
 * 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
612
	spin_unlock(&clp->cl_lock);
613 614 615 616 617

	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
618 619 620 621 622 623 624
}

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

625
	state = kzalloc(sizeof(*state), GFP_NOFS);
L
Linus Torvalds 已提交
626 627 628 629
	if (!state)
		return NULL;
	atomic_set(&state->count, 1);
	INIT_LIST_HEAD(&state->lock_states);
630
	spin_lock_init(&state->state_lock);
631
	seqlock_init(&state->seqlock);
L
Linus Torvalds 已提交
632 633 634
	return state;
}

635
void
636
nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
637
{
638
	if (state->state == fmode)
639 640
		return;
	/* NB! List reordering - see the reclaim code for why.  */
641 642
	if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
		if (fmode & FMODE_WRITE)
643 644 645 646
			list_move(&state->open_states, &state->owner->so_states);
		else
			list_move_tail(&state->open_states, &state->owner->so_states);
	}
647
	state->state = fmode;
648 649
}

L
Linus Torvalds 已提交
650 651 652 653 654 655 656
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) {
657
		if (state->owner != owner)
L
Linus Torvalds 已提交
658
			continue;
659 660
		if (!nfs4_valid_open_stateid(state))
			continue;
661
		if (atomic_inc_not_zero(&state->count))
L
Linus Torvalds 已提交
662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684
			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();
685
	spin_lock(&owner->so_lock);
L
Linus Torvalds 已提交
686 687 688 689 690 691 692
	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);
693 694
		ihold(inode);
		state->inode = inode;
L
Linus Torvalds 已提交
695
		spin_unlock(&inode->i_lock);
696 697 698 699
		/* 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
700 701
	} else {
		spin_unlock(&inode->i_lock);
702
		spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
703 704 705 706 707 708 709 710 711 712 713 714
		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;

715
	if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
L
Linus Torvalds 已提交
716
		return;
717
	spin_lock(&inode->i_lock);
718
	list_del(&state->inode_states);
L
Linus Torvalds 已提交
719
	list_del(&state->open_states);
720 721
	spin_unlock(&inode->i_lock);
	spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
722 723 724 725 726 727
	iput(inode);
	nfs4_free_open_state(state);
	nfs4_put_state_owner(owner);
}

/*
728
 * Close the current file.
L
Linus Torvalds 已提交
729
 */
730
static void __nfs4_close(struct nfs4_state *state,
731
		fmode_t fmode, gfp_t gfp_mask, int wait)
L
Linus Torvalds 已提交
732 733
{
	struct nfs4_state_owner *owner = state->owner;
734
	int call_close = 0;
735
	fmode_t newstate;
L
Linus Torvalds 已提交
736 737 738

	atomic_inc(&owner->so_count);
	/* Protect against nfs4_find_state() */
739
	spin_lock(&owner->so_lock);
740
	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
741 742 743 744 745 746 747 748 749
		case FMODE_READ:
			state->n_rdonly--;
			break;
		case FMODE_WRITE:
			state->n_wronly--;
			break;
		case FMODE_READ|FMODE_WRITE:
			state->n_rdwr--;
	}
750
	newstate = FMODE_READ|FMODE_WRITE;
751
	if (state->n_rdwr == 0) {
752
		if (state->n_rdonly == 0) {
753
			newstate &= ~FMODE_READ;
754 755 756 757
			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) {
758
			newstate &= ~FMODE_WRITE;
759 760 761 762 763
			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);
764
	}
765
	nfs4_state_set_mode_locked(state, newstate);
766
	spin_unlock(&owner->so_lock);
767

768
	if (!call_close) {
769 770
		nfs4_put_open_state(state);
		nfs4_put_state_owner(owner);
771 772
	} else
		nfs4_do_close(state, gfp_mask, wait);
773 774
}

775
void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
776
{
777
	__nfs4_close(state, fmode, GFP_NOFS, 0);
778 779
}

780
void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
781
{
782
	__nfs4_close(state, fmode, GFP_KERNEL, 1);
L
Linus Torvalds 已提交
783 784 785 786 787 788 789
}

/*
 * Search the state->lock_states for an existing lock_owner
 * that is compatible with current->files
 */
static struct nfs4_lock_state *
790
__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
791 792 793
{
	struct nfs4_lock_state *pos;
	list_for_each_entry(pos, &state->lock_states, ls_locks) {
794
		if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
L
Linus Torvalds 已提交
795
			continue;
796 797 798 799 800 801 802 803 804
		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
805 806 807 808 809 810 811 812 813 814 815
		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.
 *
 */
816
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
817 818
{
	struct nfs4_lock_state *lsp;
819
	struct nfs_server *server = state->owner->so_server;
L
Linus Torvalds 已提交
820

821
	lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
L
Linus Torvalds 已提交
822 823
	if (lsp == NULL)
		return NULL;
824
	nfs4_init_seqid_counter(&lsp->ls_seqid);
L
Linus Torvalds 已提交
825
	atomic_set(&lsp->ls_count, 1);
826
	lsp->ls_state = state;
827 828 829 830 831 832 833 834 835
	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:
836
		goto out_free;
837
	}
838 839
	lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
	if (lsp->ls_seqid.owner_id < 0)
840
		goto out_free;
841
	INIT_LIST_HEAD(&lsp->ls_locks);
L
Linus Torvalds 已提交
842
	return lsp;
843 844 845
out_free:
	kfree(lsp);
	return NULL;
L
Linus Torvalds 已提交
846 847
}

848
void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
849
{
850
	ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
851
	nfs4_destroy_seqid_counter(&lsp->ls_seqid);
852 853 854
	kfree(lsp);
}

L
Linus Torvalds 已提交
855 856 857 858 859
/*
 * Return a compatible lock_state. If no initialized lock_state structure
 * exists, return an uninitialized one.
 *
 */
860
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
861
{
862
	struct nfs4_lock_state *lsp, *new = NULL;
L
Linus Torvalds 已提交
863
	
864 865
	for(;;) {
		spin_lock(&state->state_lock);
866
		lsp = __nfs4_find_lock_state(state, owner, pid, type);
867 868 869 870 871 872 873 874 875 876
		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);
877
		new = nfs4_alloc_lock_state(state, owner, pid, type);
878 879 880 881
		if (new == NULL)
			return NULL;
	}
	spin_unlock(&state->state_lock);
882
	if (new != NULL)
883
		nfs4_free_lock_state(state->owner->so_server, new);
L
Linus Torvalds 已提交
884 885 886 887
	return lsp;
}

/*
888 889
 * Release reference to lock_state, and free it if we see that
 * it is no longer in use
L
Linus Torvalds 已提交
890
 */
891
void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
L
Linus Torvalds 已提交
892
{
893
	struct nfs_server *server;
894
	struct nfs4_state *state;
L
Linus Torvalds 已提交
895

896 897 898 899 900 901 902 903 904
	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);
905
	server = state->owner->so_server;
906
	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
907 908 909 910 911
		struct nfs_client *clp = server->nfs_client;

		clp->cl_mvops->free_lock_state(server, lsp);
	} else
		nfs4_free_lock_state(server, lsp);
L
Linus Torvalds 已提交
912 913
}

914
static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
L
Linus Torvalds 已提交
915
{
916
	struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
L
Linus Torvalds 已提交
917

918 919 920
	dst->fl_u.nfs4_fl.owner = lsp;
	atomic_inc(&lsp->ls_count);
}
L
Linus Torvalds 已提交
921

922
static void nfs4_fl_release_lock(struct file_lock *fl)
L
Linus Torvalds 已提交
923
{
924
	nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
L
Linus Torvalds 已提交
925 926
}

927
static const struct file_lock_operations nfs4_fl_lock_ops = {
928 929 930 931 932
	.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
933
{
934 935 936 937
	struct nfs4_lock_state *lsp;

	if (fl->fl_ops != NULL)
		return 0;
938 939 940
	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)
941 942
		lsp = nfs4_get_lock_state(state, NULL, fl->fl_pid,
				NFS4_FLOCK_LOCK_TYPE);
943 944
	else
		return -EINVAL;
945 946 947 948 949
	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
950 951
}

952 953
static int nfs4_copy_lock_stateid(nfs4_stateid *dst,
		struct nfs4_state *state,
954
		const struct nfs_lockowner *lockowner)
L
Linus Torvalds 已提交
955
{
956
	struct nfs4_lock_state *lsp;
957 958
	fl_owner_t fl_owner;
	pid_t fl_pid;
959
	int ret = -ENOENT;
L
Linus Torvalds 已提交
960

961 962 963 964

	if (lockowner == NULL)
		goto out;

965
	if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
966
		goto out;
L
Linus Torvalds 已提交
967

968 969
	fl_owner = lockowner->l_owner;
	fl_pid = lockowner->l_pid;
970
	spin_lock(&state->state_lock);
971
	lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
972 973 974
	if (lsp && test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
		ret = -EIO;
	else if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
975
		nfs4_stateid_copy(dst, &lsp->ls_stateid);
976 977 978 979
		ret = 0;
		smp_rmb();
		if (!list_empty(&lsp->ls_seqid.list))
			ret = -EWOULDBLOCK;
980
	}
981
	spin_unlock(&state->state_lock);
L
Linus Torvalds 已提交
982
	nfs4_put_lock_state(lsp);
983 984 985 986
out:
	return ret;
}

987
static int nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
988
{
989
	const nfs4_stateid *src;
990
	int ret;
991 992 993
	int seq;

	do {
994
		src = &zero_stateid;
995
		seq = read_seqbegin(&state->seqlock);
996 997 998
		if (test_bit(NFS_OPEN_STATE, &state->flags))
			src = &state->open_stateid;
		nfs4_stateid_copy(dst, src);
999 1000 1001 1002
		ret = 0;
		smp_rmb();
		if (!list_empty(&state->owner->so_seqid.list))
			ret = -EWOULDBLOCK;
1003
	} while (read_seqretry(&state->seqlock, seq));
1004
	return ret;
1005 1006 1007 1008 1009 1010
}

/*
 * Byte-range lock aware utility to initialize the stateid of read/write
 * requests.
 */
1011
int nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state,
1012
		fmode_t fmode, const struct nfs_lockowner *lockowner)
1013
{
1014 1015 1016 1017
	int ret = nfs4_copy_lock_stateid(dst, state, lockowner);
	if (ret == -EIO)
		/* A lost lock - don't even consider delegations */
		goto out;
1018
	if (nfs4_copy_delegation_stateid(dst, state->inode, fmode))
1019 1020
		goto out;
	if (ret != -ENOENT)
1021 1022 1023 1024
		/* nfs4_copy_delegation_stateid() didn't over-write
		 * dst, so it still has the lock stateid which we now
		 * choose to use.
		 */
1025 1026 1027
		goto out;
	ret = nfs4_copy_open_stateid(dst, state);
out:
1028 1029
	if (nfs_server_capable(state->inode, NFS_CAP_STATEID_NFSV41))
		dst->seqid = 0;
1030
	return ret;
L
Linus Torvalds 已提交
1031 1032
}

1033
struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
1034 1035 1036
{
	struct nfs_seqid *new;

1037
	new = kmalloc(sizeof(*new), gfp_mask);
1038 1039
	if (new != NULL) {
		new->sequence = counter;
1040
		INIT_LIST_HEAD(&new->list);
1041
		new->task = NULL;
1042 1043 1044 1045
	}
	return new;
}

1046
void nfs_release_seqid(struct nfs_seqid *seqid)
L
Linus Torvalds 已提交
1047
{
1048
	struct nfs_seqid_counter *sequence;
1049

1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
	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);
1061
	}
1062
	spin_unlock(&sequence->lock);
1063 1064 1065 1066 1067
}

void nfs_free_seqid(struct nfs_seqid *seqid)
{
	nfs_release_seqid(seqid);
1068
	kfree(seqid);
L
Linus Torvalds 已提交
1069 1070 1071
}

/*
1072 1073 1074 1075
 * 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()
 */
1076
static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
1077 1078 1079 1080 1081
{
	switch (status) {
		case 0:
			break;
		case -NFS4ERR_BAD_SEQID:
1082 1083
			if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
				return;
1084
			pr_warn_ratelimited("NFS: v4 server returned a bad"
D
Dan Muntz 已提交
1085 1086
					" sequence-id error on an"
					" unconfirmed sequence %p!\n",
1087
					seqid->sequence);
1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105
		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)
{
1106 1107 1108 1109 1110
	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
1111
		nfs4_drop_state_owner(sp);
1112 1113
	if (!nfs4_has_session(server->nfs_client))
		nfs_increment_seqid(status, seqid);
1114 1115 1116 1117 1118 1119 1120 1121 1122
}

/*
 * 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)
{
1123
	nfs_increment_seqid(status, seqid);
1124 1125 1126 1127
}

int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
{
1128
	struct nfs_seqid_counter *sequence = seqid->sequence;
1129 1130 1131
	int status = 0;

	spin_lock(&sequence->lock);
1132
	seqid->task = task;
1133 1134 1135 1136
	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;
1137
	rpc_sleep_on(&sequence->wait, task, NULL);
1138 1139
	status = -EAGAIN;
unlock:
1140 1141
	spin_unlock(&sequence->lock);
	return status;
L
Linus Torvalds 已提交
1142 1143
}

1144
static int nfs4_run_state_manager(void *);
L
Linus Torvalds 已提交
1145

1146
static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
T
Trond Myklebust 已提交
1147 1148
{
	smp_mb__before_clear_bit();
1149
	clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
T
Trond Myklebust 已提交
1150
	smp_mb__after_clear_bit();
1151
	wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
T
Trond Myklebust 已提交
1152 1153 1154
	rpc_wake_up(&clp->cl_rpcwaitq);
}

L
Linus Torvalds 已提交
1155
/*
1156
 * Schedule the nfs_client asynchronous state management routine
L
Linus Torvalds 已提交
1157
 */
1158
void nfs4_schedule_state_manager(struct nfs_client *clp)
L
Linus Torvalds 已提交
1159
{
1160
	struct task_struct *task;
1161
	char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
L
Linus Torvalds 已提交
1162

1163 1164
	if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
		return;
1165 1166
	__module_get(THIS_MODULE);
	atomic_inc(&clp->cl_count);
1167 1168 1169 1170 1171 1172 1173 1174

	/* 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();
1175
	task = kthread_run(nfs4_run_state_manager, clp, "%s", buf);
1176 1177 1178 1179 1180 1181 1182
	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
1183 1184 1185
}

/*
1186
 * Schedule a lease recovery attempt
L
Linus Torvalds 已提交
1187
 */
1188
void nfs4_schedule_lease_recovery(struct nfs_client *clp)
L
Linus Torvalds 已提交
1189 1190 1191
{
	if (!clp)
		return;
1192 1193
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1194 1195
	dprintk("%s: scheduling lease recovery for server %s\n", __func__,
			clp->cl_hostname);
1196
	nfs4_schedule_state_manager(clp);
L
Linus Torvalds 已提交
1197
}
A
Andy Adamson 已提交
1198
EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
L
Linus Torvalds 已提交
1199

1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233
int nfs4_wait_clnt_recover(struct nfs_client *clp)
{
	int res;

	might_sleep();

	res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
			nfs_wait_bit_killable, TASK_KILLABLE);
	if (res)
		return res;

	if (clp->cl_cons_state < 0)
		return clp->cl_cons_state;
	return 0;
}

int nfs4_client_recover_expired_lease(struct nfs_client *clp)
{
	unsigned int loop;
	int ret;

	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
		ret = nfs4_wait_clnt_recover(clp);
		if (ret != 0)
			break;
		if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
		    !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
			break;
		nfs4_schedule_state_manager(clp);
		ret = -EIO;
	}
	return ret;
}

1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245
/*
 * 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);
1246 1247
	dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
			clp->cl_hostname);
1248 1249
}

1250 1251
void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
{
1252
	nfs40_handle_cb_pathdown(clp);
1253 1254 1255
	nfs4_schedule_state_manager(clp);
}

1256
static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1257 1258 1259 1260 1261 1262 1263 1264
{

	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;
	}
1265
	set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1266 1267 1268 1269
	set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
	return 1;
}

1270
static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1271 1272 1273
{
	set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
	clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1274
	set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1275 1276 1277 1278
	set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
	return 1;
}

1279
int nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
1280 1281 1282
{
	struct nfs_client *clp = server->nfs_client;

1283 1284
	if (!nfs4_valid_open_stateid(state))
		return -EBADF;
1285
	nfs4_state_mark_reclaim_nograce(clp, state);
1286 1287
	dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
			clp->cl_hostname);
1288
	nfs4_schedule_state_manager(clp);
1289
	return 0;
1290
}
A
Andy Adamson 已提交
1291
EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1292

1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308
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;
1309
		if (!nfs4_stateid_match(&state->stateid, stateid))
1310 1311 1312 1313 1314 1315 1316 1317 1318
			continue;
		nfs4_state_mark_reclaim_nograce(clp, state);
		found = true;
	}
	spin_unlock(&inode->i_lock);
	if (found)
		nfs4_schedule_state_manager(clp);
}

1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333
static void nfs4_state_mark_open_context_bad(struct nfs4_state *state)
{
	struct inode *inode = state->inode;
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_open_context *ctx;

	spin_lock(&inode->i_lock);
	list_for_each_entry(ctx, &nfsi->open_files, list) {
		if (ctx->state != state)
			continue;
		set_bit(NFS_CONTEXT_BAD, &ctx->flags);
	}
	spin_unlock(&inode->i_lock);
}

1334 1335 1336
static void nfs4_state_mark_recovery_failed(struct nfs4_state *state, int error)
{
	set_bit(NFS_STATE_RECOVERY_FAILED, &state->flags);
1337
	nfs4_state_mark_open_context_bad(state);
1338 1339
}

1340

1341
static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1342 1343
{
	struct inode *inode = state->inode;
1344
	struct nfs_inode *nfsi = NFS_I(inode);
L
Linus Torvalds 已提交
1345 1346 1347
	struct file_lock *fl;
	int status = 0;

1348 1349 1350 1351
	if (inode->i_flock == NULL)
		return 0;

	/* Guard against delegation returns and new lock/unlock calls */
1352
	down_write(&nfsi->rwsem);
1353
	/* Protect inode->i_flock using the BKL */
1354
	spin_lock(&inode->i_lock);
H
Harvey Harrison 已提交
1355
	for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1356
		if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
L
Linus Torvalds 已提交
1357
			continue;
1358
		if (nfs_file_open_context(fl->fl_file)->state != state)
L
Linus Torvalds 已提交
1359
			continue;
1360
		spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1361 1362
		status = ops->recover_lock(state, fl);
		switch (status) {
1363 1364 1365 1366 1367 1368 1369 1370 1371
			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:
1372 1373 1374 1375
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1376
				goto out;
L
Linus Torvalds 已提交
1377
			default:
1378 1379
				printk(KERN_ERR "NFS: %s: unhandled error %d. "
					"Zeroing state\n", __func__, status);
1380 1381
			case -ENOMEM:
			case -NFS4ERR_DENIED:
L
Linus Torvalds 已提交
1382 1383
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1384
				/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1385
				status = 0;
L
Linus Torvalds 已提交
1386
		}
1387
		spin_lock(&inode->i_lock);
L
Linus Torvalds 已提交
1388
	}
1389
	spin_unlock(&inode->i_lock);
1390
out:
1391
	up_write(&nfsi->rwsem);
L
Linus Torvalds 已提交
1392 1393 1394
	return status;
}

1395
static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408
{
	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.
	 */
1409
	spin_lock(&sp->so_lock);
1410 1411
	write_seqcount_begin(&sp->so_reclaim_seqcount);
restart:
L
Linus Torvalds 已提交
1412
	list_for_each_entry(state, &sp->so_states, open_states) {
1413 1414
		if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
			continue;
1415 1416
		if (!nfs4_valid_open_stateid(state))
			continue;
L
Linus Torvalds 已提交
1417 1418
		if (state->state == 0)
			continue;
1419 1420
		atomic_inc(&state->count);
		spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1421 1422
		status = ops->recover_open(sp, state);
		if (status >= 0) {
1423 1424
			status = nfs4_reclaim_locks(state, ops);
			if (status >= 0) {
1425 1426 1427 1428 1429 1430 1431 1432 1433
				if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0) {
					spin_lock(&state->state_lock);
					list_for_each_entry(lock, &state->lock_states, ls_locks) {
						if (!test_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags))
							pr_warn_ratelimited("NFS: "
									    "%s: Lock reclaim "
									    "failed!\n", __func__);
					}
					spin_unlock(&state->state_lock);
1434
				}
1435
				nfs4_put_open_state(state);
1436
				spin_lock(&sp->so_lock);
1437
				goto restart;
L
Linus Torvalds 已提交
1438 1439 1440 1441
			}
		}
		switch (status) {
			default:
1442 1443
				printk(KERN_ERR "NFS: %s: unhandled error %d. "
					"Zeroing state\n", __func__, status);
L
Linus Torvalds 已提交
1444
			case -ENOENT:
1445
			case -ENOMEM:
1446
			case -ESTALE:
L
Linus Torvalds 已提交
1447 1448 1449 1450
				/*
				 * Open state on this file cannot be recovered
				 * All we can do is revert to using the zero stateid.
				 */
1451
				nfs4_state_mark_recovery_failed(state, status);
L
Linus Torvalds 已提交
1452
				break;
1453 1454
			case -EAGAIN:
				ssleep(1);
1455 1456 1457
			case -NFS4ERR_ADMIN_REVOKED:
			case -NFS4ERR_STALE_STATEID:
			case -NFS4ERR_BAD_STATEID:
1458 1459
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1460
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1461
				break;
L
Linus Torvalds 已提交
1462 1463
			case -NFS4ERR_EXPIRED:
			case -NFS4ERR_NO_GRACE:
1464
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
L
Linus Torvalds 已提交
1465
			case -NFS4ERR_STALE_CLIENTID:
1466 1467 1468 1469
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
L
Linus Torvalds 已提交
1470 1471
				goto out_err;
		}
1472
		nfs4_put_open_state(state);
1473
		spin_lock(&sp->so_lock);
1474
		goto restart;
L
Linus Torvalds 已提交
1475
	}
1476
	write_seqcount_end(&sp->so_reclaim_seqcount);
1477
	spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1478 1479
	return 0;
out_err:
1480
	nfs4_put_open_state(state);
1481 1482 1483
	spin_lock(&sp->so_lock);
	write_seqcount_end(&sp->so_reclaim_seqcount);
	spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1484 1485 1486
	return status;
}

1487 1488 1489 1490 1491 1492 1493 1494
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);
1495
	spin_lock(&state->state_lock);
1496 1497
	list_for_each_entry(lock, &state->lock_states, ls_locks) {
		lock->ls_seqid.flags = 0;
1498
		clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
1499
	}
1500
	spin_unlock(&state->state_lock);
1501 1502
}

1503 1504
static void nfs4_reset_seqids(struct nfs_server *server,
	int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1505
{
1506
	struct nfs_client *clp = server->nfs_client;
1507
	struct nfs4_state_owner *sp;
1508
	struct rb_node *pos;
1509 1510
	struct nfs4_state *state;

1511 1512 1513 1514 1515
	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);
1516
		sp->so_seqid.flags = 0;
1517
		spin_lock(&sp->so_lock);
1518
		list_for_each_entry(state, &sp->so_states, open_states) {
1519 1520
			if (mark_reclaim(clp, state))
				nfs4_clear_open_state(state);
1521
		}
1522
		spin_unlock(&sp->so_lock);
1523
	}
1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535
	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();
1536 1537
}

1538 1539 1540 1541 1542 1543 1544
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);
}

1545
static void nfs4_reclaim_complete(struct nfs_client *clp,
1546 1547
				 const struct nfs4_state_recovery_ops *ops,
				 struct rpc_cred *cred)
1548 1549 1550
{
	/* Notify the server we're done reclaiming our state */
	if (ops->reclaim_complete)
1551
		(void)ops->reclaim_complete(clp, cred);
1552 1553
}

1554
static void nfs4_clear_reclaim_server(struct nfs_server *server)
1555
{
1556
	struct nfs_client *clp = server->nfs_client;
1557 1558 1559 1560
	struct nfs4_state_owner *sp;
	struct rb_node *pos;
	struct nfs4_state *state;

1561 1562 1563 1564 1565
	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);
1566 1567
		spin_lock(&sp->so_lock);
		list_for_each_entry(state, &sp->so_states, open_states) {
1568 1569
			if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
						&state->flags))
1570 1571 1572 1573 1574
				continue;
			nfs4_state_mark_reclaim_nograce(clp, state);
		}
		spin_unlock(&sp->so_lock);
	}
1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588
	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();
1589 1590

	nfs_delegation_reap_unclaimed(clp);
1591 1592 1593 1594 1595
	return 1;
}

static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
{
1596 1597 1598
	const struct nfs4_state_recovery_ops *ops;
	struct rpc_cred *cred;

1599 1600
	if (!nfs4_state_clear_reclaim_reboot(clp))
		return;
1601
	ops = clp->cl_mvops->reboot_recovery_ops;
1602
	cred = nfs4_get_clid_cred(clp);
1603 1604
	nfs4_reclaim_complete(clp, ops, cred);
	put_rpccred(cred);
1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618
}

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);
}

1619
static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1620 1621
{
	switch (error) {
1622 1623
		case 0:
			break;
1624
		case -NFS4ERR_CB_PATH_DOWN:
1625
			nfs40_handle_cb_pathdown(clp);
1626
			break;
1627 1628
		case -NFS4ERR_NO_GRACE:
			nfs4_state_end_reclaim_reboot(clp);
1629
			break;
1630 1631 1632
		case -NFS4ERR_STALE_CLIENTID:
		case -NFS4ERR_LEASE_MOVED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1633
			nfs4_state_clear_reclaim_reboot(clp);
1634 1635 1636 1637 1638
			nfs4_state_start_reclaim_reboot(clp);
			break;
		case -NFS4ERR_EXPIRED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
			nfs4_state_start_reclaim_nograce(clp);
1639
			break;
1640 1641 1642 1643 1644 1645
		case -NFS4ERR_BADSESSION:
		case -NFS4ERR_BADSLOT:
		case -NFS4ERR_BAD_HIGH_SLOT:
		case -NFS4ERR_DEADSESSION:
		case -NFS4ERR_SEQ_FALSE_RETRY:
		case -NFS4ERR_SEQ_MISORDERED:
1646
			set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1647
			/* Zero session reset errors */
1648
			break;
1649 1650 1651
		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
			set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
			break;
1652
		default:
1653 1654
			dprintk("%s: failed to handle error %d for server %s\n",
					__func__, error, clp->cl_hostname);
1655
			return error;
1656
	}
1657 1658
	dprintk("%s: handled error %d for server %s\n", __func__, error,
			clp->cl_hostname);
1659
	return 0;
1660 1661
}

1662
static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1663
{
1664 1665
	struct nfs4_state_owner *sp;
	struct nfs_server *server;
1666
	struct rb_node *pos;
L
Linus Torvalds 已提交
1667 1668
	int status = 0;

1669
restart:
1670 1671
	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1672
		nfs4_purge_state_owners(server);
1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692
		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);
			}

1693
			nfs4_put_state_owner(sp);
1694
			goto restart;
1695
		}
1696
		spin_unlock(&clp->cl_lock);
1697
	}
1698
	rcu_read_unlock();
1699 1700 1701 1702 1703 1704
	return status;
}

static int nfs4_check_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1705 1706
	const struct nfs4_state_maintenance_ops *ops =
		clp->cl_mvops->state_renewal_ops;
1707
	int status;
L
Linus Torvalds 已提交
1708

1709 1710 1711
	/* Is the client already known to have an expired lease? */
	if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		return 0;
1712 1713 1714
	spin_lock(&clp->cl_lock);
	cred = ops->get_state_renewal_cred_locked(clp);
	spin_unlock(&clp->cl_lock);
1715
	if (cred == NULL) {
1716
		cred = nfs4_get_clid_cred(clp);
1717
		status = -ENOKEY;
1718 1719
		if (cred == NULL)
			goto out;
1720
	}
1721
	status = ops->renew_lease(clp, cred);
1722
	put_rpccred(cred);
1723 1724 1725 1726
	if (status == -ETIMEDOUT) {
		set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
		return 0;
	}
1727
out:
1728
	return nfs4_recovery_handle_error(clp, status);
1729 1730
}

1731 1732
/* Set NFS4CLNT_LEASE_EXPIRED and reclaim reboot state for all v4.0 errors
 * and for recoverable errors on EXCHANGE_ID for v4.1
1733 1734 1735 1736
 */
static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
{
	switch (status) {
1737 1738 1739 1740 1741
	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);
1742 1743
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
		break;
1744 1745
	case -NFS4ERR_STALE_CLIENTID:
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1746 1747
		nfs4_state_clear_reclaim_reboot(clp);
		nfs4_state_start_reclaim_reboot(clp);
1748
		break;
1749 1750 1751 1752 1753 1754
	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;
1755 1756 1757 1758 1759 1760 1761 1762 1763 1764
	case -EACCES:
	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);
1765 1766
		dprintk("%s: exit with error %d for server %s\n",
				__func__, -EPROTONOSUPPORT, clp->cl_hostname);
1767 1768 1769 1770
		return -EPROTONOSUPPORT;
	case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
				 * in nfs4_exchange_id */
	default:
1771 1772
		dprintk("%s: exit with error %d for server %s\n", __func__,
				status, clp->cl_hostname);
1773 1774 1775
		return status;
	}
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1776 1777
	dprintk("%s: handled error %d for server %s\n", __func__, status,
			clp->cl_hostname);
1778 1779 1780
	return 0;
}

1781
static int nfs4_establish_lease(struct nfs_client *clp)
1782 1783
{
	struct rpc_cred *cred;
1784 1785
	const struct nfs4_state_recovery_ops *ops =
		clp->cl_mvops->reboot_recovery_ops;
1786
	int status;
1787

1788
	cred = nfs4_get_clid_cred(clp);
1789 1790 1791 1792 1793
	if (cred == NULL)
		return -ENOENT;
	status = ops->establish_clid(clp, cred);
	put_rpccred(cred);
	if (status != 0)
1794 1795 1796 1797 1798
		return status;
	pnfs_destroy_all_layouts(clp);
	return 0;
}

1799 1800 1801 1802
/*
 * Returns zero or a negative errno.  NFS4ERR values are converted
 * to local errno values.
 */
1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824
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)
1825
		return nfs4_handle_reclaim_lease_error(clp, status);
1826 1827 1828
	clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
	nfs4_state_start_reclaim_nograce(clp);
1829
	return 0;
1830 1831
}

1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851
/**
 * nfs4_discover_server_trunking - Detect server IP address trunking
 *
 * @clp: nfs_client under test
 * @result: OUT: found nfs_client, or clp
 *
 * Returns zero or a negative errno.  If zero is returned,
 * an nfs_client pointer is planted in "result".
 *
 * Note: since we are invoked in process context, and
 * not from inside the state manager, we cannot use
 * nfs4_handle_reclaim_lease_error().
 */
int nfs4_discover_server_trunking(struct nfs_client *clp,
				  struct nfs_client **result)
{
	const struct nfs4_state_recovery_ops *ops =
				clp->cl_mvops->reboot_recovery_ops;
	struct rpc_clnt *clnt;
	struct rpc_cred *cred;
1852
	int i, status;
1853 1854 1855 1856 1857 1858 1859 1860

	dprintk("NFS: %s: testing '%s'\n", __func__, clp->cl_hostname);

	clnt = clp->cl_rpcclient;
	i = 0;

	mutex_lock(&nfs_clid_init_mutex);
again:
1861
	status  = -ENOENT;
1862
	cred = nfs4_get_clid_cred(clp);
1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874
	if (cred == NULL)
		goto out_unlock;

	status = ops->detect_trunking(clp, result, cred);
	put_rpccred(cred);
	switch (status) {
	case 0:
		break;
	case -NFS4ERR_DELAY:
	case -ETIMEDOUT:
	case -EAGAIN:
		ssleep(1);
1875
	case -NFS4ERR_STALE_CLIENTID:
1876 1877 1878
		dprintk("NFS: %s after status %d, retrying\n",
			__func__, status);
		goto again;
1879
	case -EACCES:
1880 1881 1882 1883 1884
		if (i++ == 0) {
			nfs4_root_machine_cred(clp);
			goto again;
		}
		if (i > 2)
1885
			break;
1886 1887
	case -NFS4ERR_CLID_INUSE:
	case -NFS4ERR_WRONGSEC:
1888
		clnt = rpc_clone_client_set_auth(clnt, RPC_AUTH_UNIX);
1889 1890 1891 1892
		if (IS_ERR(clnt)) {
			status = PTR_ERR(clnt);
			break;
		}
1893 1894 1895 1896 1897 1898 1899
		/* Note: this is safe because we haven't yet marked the
		 * client as ready, so we are the only user of
		 * clp->cl_rpcclient
		 */
		clnt = xchg(&clp->cl_rpcclient, clnt);
		rpc_shutdown_client(clnt);
		clnt = clp->cl_rpcclient;
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909
		goto again;

	case -NFS4ERR_MINOR_VERS_MISMATCH:
		status = -EPROTONOSUPPORT;
		break;

	case -EKEYEXPIRED:
	case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
				 * in nfs4_exchange_id */
		status = -EKEYEXPIRED;
1910 1911 1912 1913 1914
		break;
	default:
		pr_warn("NFS: %s unhandled error %d. Exiting with error EIO\n",
				__func__, status);
		status = -EIO;
1915 1916 1917 1918 1919 1920 1921 1922
	}

out_unlock:
	mutex_unlock(&nfs_clid_init_mutex);
	dprintk("NFS: %s: status = %d\n", __func__, status);
	return status;
}

1923
#ifdef CONFIG_NFS_V4_1
1924
void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
1925
{
1926 1927
	struct nfs_client *clp = session->clp;

1928 1929 1930 1931 1932 1933 1934
	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);
	}
1935
	nfs4_schedule_lease_recovery(clp);
1936
}
1937
EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
1938

1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950
static void nfs41_ping_server(struct nfs_client *clp)
{
	/* Use CHECK_LEASE to ping the server with a SEQUENCE */
	set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
	nfs4_schedule_state_manager(clp);
}

void nfs41_server_notify_target_slotid_update(struct nfs_client *clp)
{
	nfs41_ping_server(clp);
}

1951 1952 1953 1954 1955
void nfs41_server_notify_highest_slotid_update(struct nfs_client *clp)
{
	nfs41_ping_server(clp);
}

1956 1957 1958
static void nfs4_reset_all_state(struct nfs_client *clp)
{
	if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1959
		set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1960
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1961
		nfs4_state_start_reclaim_nograce(clp);
1962 1963
		dprintk("%s: scheduling reset of all state for server %s!\n",
				__func__, clp->cl_hostname);
1964
		nfs4_schedule_state_manager(clp);
1965 1966 1967 1968 1969 1970 1971
	}
}

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);
1972 1973
		dprintk("%s: server %s rebooted!\n", __func__,
				clp->cl_hostname);
1974
		nfs4_schedule_state_manager(clp);
1975 1976 1977 1978 1979 1980
	}
}

static void nfs41_handle_state_revoked(struct nfs_client *clp)
{
	nfs4_reset_all_state(clp);
1981
	dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
1982 1983 1984 1985 1986 1987
}

static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
{
	/* This will need to handle layouts too */
	nfs_expire_all_delegations(clp);
1988 1989
	dprintk("%s: Recallable state revoked on server %s!\n", __func__,
			clp->cl_hostname);
1990 1991
}

1992
static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
1993 1994 1995
{
	nfs_expire_all_delegations(clp);
	if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
1996
		nfs4_schedule_state_manager(clp);
1997 1998
	dprintk("%s: server %s declared a backchannel fault\n", __func__,
			clp->cl_hostname);
1999 2000
}

2001 2002 2003 2004 2005 2006 2007
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);
}

2008 2009 2010 2011
void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
{
	if (!flags)
		return;
2012 2013 2014 2015

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

2016
	if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
2017
		nfs41_handle_server_reboot(clp);
2018
	if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
2019 2020
			    SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
			    SEQ4_STATUS_ADMIN_STATE_REVOKED |
2021 2022
			    SEQ4_STATUS_LEASE_MOVED))
		nfs41_handle_state_revoked(clp);
2023
	if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
2024
		nfs41_handle_recallable_state_revoked(clp);
2025 2026 2027 2028
	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))
2029
		nfs41_handle_cb_path_down(clp);
2030 2031
}

2032 2033
static int nfs4_reset_session(struct nfs_client *clp)
{
2034
	struct rpc_cred *cred;
2035 2036
	int status;

2037 2038
	if (!nfs4_has_session(clp))
		return 0;
2039
	nfs4_begin_drain_session(clp);
2040
	cred = nfs4_get_clid_cred(clp);
2041
	status = nfs4_proc_destroy_session(clp->cl_session, cred);
2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053
	switch (status) {
	case 0:
	case -NFS4ERR_BADSESSION:
	case -NFS4ERR_DEADSESSION:
		break;
	case -NFS4ERR_BACK_CHAN_BUSY:
	case -NFS4ERR_DELAY:
		set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
		status = 0;
		ssleep(1);
		goto out;
	default:
2054
		status = nfs4_recovery_handle_error(clp, status);
2055 2056 2057 2058
		goto out;
	}

	memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
2059
	status = nfs4_proc_create_session(clp, cred);
2060
	if (status) {
2061 2062
		dprintk("%s: session reset failed with status %d for server %s!\n",
			__func__, status, clp->cl_hostname);
2063
		status = nfs4_handle_reclaim_lease_error(clp, status);
2064 2065
		goto out;
	}
2066
	nfs41_finish_session_reset(clp);
2067 2068
	dprintk("%s: session reset was successful for server %s!\n",
			__func__, clp->cl_hostname);
2069
out:
2070 2071
	if (cred)
		put_rpccred(cred);
2072 2073
	return status;
}
2074

2075 2076
static int nfs4_bind_conn_to_session(struct nfs_client *clp)
{
2077 2078 2079
	struct rpc_cred *cred;
	int ret;

2080 2081
	if (!nfs4_has_session(clp))
		return 0;
2082
	nfs4_begin_drain_session(clp);
2083
	cred = nfs4_get_clid_cred(clp);
2084 2085 2086
	ret = nfs4_proc_bind_conn_to_session(clp, cred);
	if (cred)
		put_rpccred(cred);
2087
	clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2088 2089
	switch (ret) {
	case 0:
2090 2091
		dprintk("%s: bind_conn_to_session was successful for server %s!\n",
			__func__, clp->cl_hostname);
2092 2093 2094 2095 2096 2097 2098 2099 2100
		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;
2101
}
2102
#else /* CONFIG_NFS_V4_1 */
2103
static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
2104 2105 2106 2107 2108

static int nfs4_bind_conn_to_session(struct nfs_client *clp)
{
	return 0;
}
2109 2110
#endif /* CONFIG_NFS_V4_1 */

2111
static void nfs4_state_manager(struct nfs_client *clp)
2112 2113
{
	int status = 0;
2114
	const char *section = "", *section_sep = "";
2115 2116

	/* Ensure exclusive access to NFSv4 state */
2117
	do {
2118
		if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
2119
			section = "purge state";
2120
			status = nfs4_purge_lease(clp);
2121 2122
			if (status < 0)
				goto out_error;
2123
			continue;
2124 2125
		}

2126
		if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
2127
			section = "lease expired";
2128 2129
			/* We're going to have to re-establish a clientid */
			status = nfs4_reclaim_lease(clp);
2130
			if (status < 0)
2131
				goto out_error;
2132
			continue;
2133 2134
		}

2135
		/* Initialize or reset the session */
2136
		if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
2137
			section = "reset session";
2138
			status = nfs4_reset_session(clp);
2139 2140 2141
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
				continue;
			if (status < 0)
2142 2143
				goto out_error;
		}
2144

2145 2146
		/* Send BIND_CONN_TO_SESSION */
		if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2147
				&clp->cl_state)) {
2148
			section = "bind conn to session";
2149 2150 2151
			status = nfs4_bind_conn_to_session(clp);
			if (status < 0)
				goto out_error;
2152
			continue;
2153 2154
		}

2155 2156 2157 2158 2159 2160 2161 2162
		if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
			section = "check lease";
			status = nfs4_check_lease(clp);
			if (status < 0)
				goto out_error;
			continue;
		}

2163
		/* First recover reboot state... */
2164
		if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
2165
			section = "reclaim reboot";
2166
			status = nfs4_do_reclaim(clp,
2167
				clp->cl_mvops->reboot_recovery_ops);
2168
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2169
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
2170
				continue;
2171
			nfs4_state_end_reclaim_reboot(clp);
2172 2173 2174 2175
			if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
				continue;
			if (status < 0)
				goto out_error;
2176 2177
		}

2178 2179
		/* Now recover expired state... */
		if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
2180
			section = "reclaim nograce";
2181
			status = nfs4_do_reclaim(clp,
2182
				clp->cl_mvops->nograce_recovery_ops);
2183
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2184
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
2185 2186 2187
			    test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
				continue;
			if (status < 0)
2188
				goto out_error;
L
Linus Torvalds 已提交
2189
		}
2190

2191
		nfs4_end_drain_session(clp);
2192 2193 2194 2195
		if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
			nfs_client_return_marked_delegations(clp);
			continue;
		}
2196 2197

		nfs4_clear_state_manager_bit(clp);
2198 2199 2200 2201 2202
		/* 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;
2203
	} while (atomic_read(&clp->cl_count) > 1);
2204
	return;
L
Linus Torvalds 已提交
2205
out_error:
2206 2207 2208 2209 2210
	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);
2211
	ssleep(1);
2212
	nfs4_end_drain_session(clp);
2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224
	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
2225 2226 2227 2228 2229 2230 2231
}

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