nfs4state.c 58.9 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)
149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168
{
	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);
}

169 170
static struct rpc_cred *
nfs4_get_renew_cred_server_locked(struct nfs_server *server)
171
{
172
	struct rpc_cred *cred = NULL;
173
	struct nfs4_state_owner *sp;
174
	struct rb_node *pos;
175

176 177 178 179
	for (pos = rb_first(&server->state_owners);
	     pos != NULL;
	     pos = rb_next(pos)) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
180 181 182 183 184 185 186 187
		if (list_empty(&sp->so_states))
			continue;
		cred = get_rpccred(sp->so_cred);
		break;
	}
	return cred;
}

188 189 190 191 192 193 194 195 196 197 198 199
/**
 * 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;

200 201 202 203 204
	/* Use machine credentials if available */
	cred = nfs4_get_machine_cred_locked(clp);
	if (cred != NULL)
		goto out;

205 206 207 208 209 210 211
	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();
212 213

out:
214 215 216
	return cred;
}

217 218
#if defined(CONFIG_NFS_V4_1)

R
Ricardo Labiaga 已提交
219 220 221 222 223
static int nfs41_setup_state_renewal(struct nfs_client *clp)
{
	int status;
	struct nfs_fsinfo fsinfo;

224 225 226 227 228
	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
229 230 231 232 233 234 235 236 237 238 239 240 241 242
	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;
}

243 244 245 246 247
/*
 * 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.
 */
248
static void nfs4_end_drain_session(struct nfs_client *clp)
249
{
250
	struct nfs4_session *ses = clp->cl_session;
251
	struct nfs4_slot_table *tbl;
252

253 254
	if (ses == NULL)
		return;
255
	tbl = &ses->fc_slot_table;
256
	if (test_and_clear_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
257
		spin_lock(&tbl->slot_tbl_lock);
258
		nfs41_wake_slot_table(tbl);
259
		spin_unlock(&tbl->slot_tbl_lock);
260
	}
261 262
}

263 264 265 266 267 268 269 270 271 272
/*
 * Signal state manager thread if session fore channel is drained
 */
void nfs4_session_drain_complete(struct nfs4_session *session,
		struct nfs4_slot_table *tbl)
{
	if (nfs4_session_draining(session))
		complete(&tbl->complete);
}

273
static int nfs4_wait_on_slot_tbl(struct nfs4_slot_table *tbl)
274 275
{
	spin_lock(&tbl->slot_tbl_lock);
276
	if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
277
		INIT_COMPLETION(tbl->complete);
278
		spin_unlock(&tbl->slot_tbl_lock);
279
		return wait_for_completion_interruptible(&tbl->complete);
280 281 282 283 284
	}
	spin_unlock(&tbl->slot_tbl_lock);
	return 0;
}

285 286 287 288 289 290 291 292 293 294 295 296 297 298
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);
}

299 300 301 302 303 304 305 306 307
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);
}

308 309 310 311
int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
{
	int status;

312 313
	if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
		goto do_confirm;
314
	nfs4_begin_drain_session(clp);
315
	status = nfs4_proc_exchange_id(clp, cred);
R
Ricardo Labiaga 已提交
316 317
	if (status != 0)
		goto out;
318 319
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
do_confirm:
320
	status = nfs4_proc_create_session(clp, cred);
R
Ricardo Labiaga 已提交
321 322
	if (status != 0)
		goto out;
323
	nfs41_finish_session_reset(clp);
R
Ricardo Labiaga 已提交
324 325
	nfs_mark_client_ready(clp, NFS_CS_READY);
out:
326 327 328
	return status;
}

329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350
/**
 * 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;
351
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
352 353 354 355

	return nfs41_walk_client_list(clp, result, cred);
}

356 357 358 359 360 361 362 363 364 365 366 367
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 */

368 369
static struct rpc_cred *
nfs4_get_setclientid_cred_server(struct nfs_server *server)
370
{
371 372
	struct nfs_client *clp = server->nfs_client;
	struct rpc_cred *cred = NULL;
373
	struct nfs4_state_owner *sp;
374
	struct rb_node *pos;
375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394

	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;
395
	struct rpc_cred *cred;
396

397 398
	spin_lock(&clp->cl_lock);
	cred = nfs4_get_machine_cred_locked(clp);
399
	spin_unlock(&clp->cl_lock);
400 401
	if (cred != NULL)
		goto out;
402 403 404 405 406 407

	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;
408
	}
409 410
	rcu_read_unlock();

411 412
out:
	return cred;
413 414
}

L
Linus Torvalds 已提交
415
static struct nfs4_state_owner *
416
nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
L
Linus Torvalds 已提交
417
{
418
	struct rb_node **p = &server->state_owners.rb_node,
419
		       *parent = NULL;
420
	struct nfs4_state_owner *sp;
L
Linus Torvalds 已提交
421

422 423
	while (*p != NULL) {
		parent = *p;
424
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
425 426 427 428 429 430

		if (cred < sp->so_cred)
			p = &parent->rb_left;
		else if (cred > sp->so_cred)
			p = &parent->rb_right;
		else {
431 432
			if (!list_empty(&sp->so_lru))
				list_del_init(&sp->so_lru);
433
			atomic_inc(&sp->so_count);
434
			return sp;
435
		}
L
Linus Torvalds 已提交
436
	}
437
	return NULL;
L
Linus Torvalds 已提交
438 439
}

440
static struct nfs4_state_owner *
441
nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
442
{
443 444
	struct nfs_server *server = new->so_server;
	struct rb_node **p = &server->state_owners.rb_node,
445 446
		       *parent = NULL;
	struct nfs4_state_owner *sp;
447
	int err;
448 449 450

	while (*p != NULL) {
		parent = *p;
451
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
452 453 454 455 456 457

		if (new->so_cred < sp->so_cred)
			p = &parent->rb_left;
		else if (new->so_cred > sp->so_cred)
			p = &parent->rb_right;
		else {
458 459
			if (!list_empty(&sp->so_lru))
				list_del_init(&sp->so_lru);
460 461 462 463
			atomic_inc(&sp->so_count);
			return sp;
		}
	}
464
	err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
465 466
	if (err)
		return ERR_PTR(err);
467 468
	rb_link_node(&new->so_server_node, parent, p);
	rb_insert_color(&new->so_server_node, &server->state_owners);
469 470 471 472
	return new;
}

static void
473
nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
474
{
475 476 477 478
	struct nfs_server *server = sp->so_server;

	if (!RB_EMPTY_NODE(&sp->so_server_node))
		rb_erase(&sp->so_server_node, &server->state_owners);
479
	ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
480 481
}

482 483 484
static void
nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
{
485
	sc->create_time = ktime_get();
486 487 488 489 490 491 492 493 494 495 496 497 498
	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
499 500 501 502 503 504
/*
 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
 * create a new state_owner.
 *
 */
static struct nfs4_state_owner *
505 506 507
nfs4_alloc_state_owner(struct nfs_server *server,
		struct rpc_cred *cred,
		gfp_t gfp_flags)
L
Linus Torvalds 已提交
508 509 510
{
	struct nfs4_state_owner *sp;

511
	sp = kzalloc(sizeof(*sp), gfp_flags);
L
Linus Torvalds 已提交
512 513
	if (!sp)
		return NULL;
514 515
	sp->so_server = server;
	sp->so_cred = get_rpccred(cred);
516
	spin_lock_init(&sp->so_lock);
L
Linus Torvalds 已提交
517
	INIT_LIST_HEAD(&sp->so_states);
518
	nfs4_init_seqid_counter(&sp->so_seqid);
L
Linus Torvalds 已提交
519
	atomic_set(&sp->so_count, 1);
520
	INIT_LIST_HEAD(&sp->so_lru);
521
	seqcount_init(&sp->so_reclaim_seqcount);
522
	mutex_init(&sp->so_delegreturn_mutex);
L
Linus Torvalds 已提交
523 524 525
	return sp;
}

526
static void
L
Linus Torvalds 已提交
527 528
nfs4_drop_state_owner(struct nfs4_state_owner *sp)
{
529 530 531
	struct rb_node *rb_node = &sp->so_server_node;

	if (!RB_EMPTY_NODE(rb_node)) {
532 533
		struct nfs_server *server = sp->so_server;
		struct nfs_client *clp = server->nfs_client;
534 535

		spin_lock(&clp->cl_lock);
536 537 538 539
		if (!RB_EMPTY_NODE(rb_node)) {
			rb_erase(rb_node, &server->state_owners);
			RB_CLEAR_NODE(rb_node);
		}
540 541
		spin_unlock(&clp->cl_lock);
	}
L
Linus Torvalds 已提交
542 543
}

544 545
static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
{
546
	nfs4_destroy_seqid_counter(&sp->so_seqid);
547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
	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);
	}
}

576 577 578 579 580 581 582 583
/**
 * 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,
584 585
					      struct rpc_cred *cred,
					      gfp_t gfp_flags)
L
Linus Torvalds 已提交
586
{
587
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
588 589 590
	struct nfs4_state_owner *sp, *new;

	spin_lock(&clp->cl_lock);
591
	sp = nfs4_find_state_owner_locked(server, cred);
L
Linus Torvalds 已提交
592 593
	spin_unlock(&clp->cl_lock);
	if (sp != NULL)
594
		goto out;
595
	new = nfs4_alloc_state_owner(server, cred, gfp_flags);
596
	if (new == NULL)
597
		goto out;
598 599 600 601 602 603 604
	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));
605 606
	if (sp != new)
		nfs4_free_state_owner(new);
607 608
out:
	nfs4_gc_state_owners(server);
609
	return sp;
L
Linus Torvalds 已提交
610 611
}

612 613 614
/**
 * nfs4_put_state_owner - Release a nfs4_state_owner
 * @sp: state owner data to release
615 616 617 618 619 620 621 622
 *
 * 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.
623
 */
L
Linus Torvalds 已提交
624 625
void nfs4_put_state_owner(struct nfs4_state_owner *sp)
{
626 627
	struct nfs_server *server = sp->so_server;
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
628 629 630

	if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
		return;
631

632 633 634
	sp->so_expires = jiffies;
	list_add_tail(&sp->so_lru, &server->state_owners_lru);
	spin_unlock(&clp->cl_lock);
635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654
}

/**
 * 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
655
	spin_unlock(&clp->cl_lock);
656 657 658 659 660

	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
661 662 663 664 665 666 667
}

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

668
	state = kzalloc(sizeof(*state), GFP_NOFS);
L
Linus Torvalds 已提交
669 670 671 672
	if (!state)
		return NULL;
	atomic_set(&state->count, 1);
	INIT_LIST_HEAD(&state->lock_states);
673
	spin_lock_init(&state->state_lock);
674
	seqlock_init(&state->seqlock);
L
Linus Torvalds 已提交
675 676 677
	return state;
}

678
void
679
nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
680
{
681
	if (state->state == fmode)
682 683
		return;
	/* NB! List reordering - see the reclaim code for why.  */
684 685
	if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
		if (fmode & FMODE_WRITE)
686 687 688 689
			list_move(&state->open_states, &state->owner->so_states);
		else
			list_move_tail(&state->open_states, &state->owner->so_states);
	}
690
	state->state = fmode;
691 692
}

L
Linus Torvalds 已提交
693 694 695 696 697 698 699
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) {
700
		if (state->owner != owner)
L
Linus Torvalds 已提交
701
			continue;
702 703
		if (!nfs4_valid_open_stateid(state))
			continue;
704
		if (atomic_inc_not_zero(&state->count))
L
Linus Torvalds 已提交
705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727
			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();
728
	spin_lock(&owner->so_lock);
L
Linus Torvalds 已提交
729 730 731 732 733 734 735
	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);
736 737
		ihold(inode);
		state->inode = inode;
L
Linus Torvalds 已提交
738
		spin_unlock(&inode->i_lock);
739 740 741 742
		/* 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
743 744
	} else {
		spin_unlock(&inode->i_lock);
745
		spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
746 747 748 749 750 751 752 753 754 755 756 757
		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;

758
	if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
L
Linus Torvalds 已提交
759
		return;
760
	spin_lock(&inode->i_lock);
761
	list_del(&state->inode_states);
L
Linus Torvalds 已提交
762
	list_del(&state->open_states);
763 764
	spin_unlock(&inode->i_lock);
	spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
765 766 767 768 769 770
	iput(inode);
	nfs4_free_open_state(state);
	nfs4_put_state_owner(owner);
}

/*
771
 * Close the current file.
L
Linus Torvalds 已提交
772
 */
773
static void __nfs4_close(struct nfs4_state *state,
774
		fmode_t fmode, gfp_t gfp_mask, int wait)
L
Linus Torvalds 已提交
775 776
{
	struct nfs4_state_owner *owner = state->owner;
777
	int call_close = 0;
778
	fmode_t newstate;
L
Linus Torvalds 已提交
779 780 781

	atomic_inc(&owner->so_count);
	/* Protect against nfs4_find_state() */
782
	spin_lock(&owner->so_lock);
783
	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
784 785 786 787 788 789 790 791 792
		case FMODE_READ:
			state->n_rdonly--;
			break;
		case FMODE_WRITE:
			state->n_wronly--;
			break;
		case FMODE_READ|FMODE_WRITE:
			state->n_rdwr--;
	}
793
	newstate = FMODE_READ|FMODE_WRITE;
794
	if (state->n_rdwr == 0) {
795
		if (state->n_rdonly == 0) {
796
			newstate &= ~FMODE_READ;
797 798 799 800
			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) {
801
			newstate &= ~FMODE_WRITE;
802 803 804 805 806
			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);
807
	}
808
	nfs4_state_set_mode_locked(state, newstate);
809
	spin_unlock(&owner->so_lock);
810

811
	if (!call_close) {
812 813
		nfs4_put_open_state(state);
		nfs4_put_state_owner(owner);
814 815
	} else
		nfs4_do_close(state, gfp_mask, wait);
816 817
}

818
void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
819
{
820
	__nfs4_close(state, fmode, GFP_NOFS, 0);
821 822
}

823
void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
824
{
825
	__nfs4_close(state, fmode, GFP_KERNEL, 1);
L
Linus Torvalds 已提交
826 827 828 829 830 831 832
}

/*
 * Search the state->lock_states for an existing lock_owner
 * that is compatible with current->files
 */
static struct nfs4_lock_state *
833
__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
834 835 836
{
	struct nfs4_lock_state *pos;
	list_for_each_entry(pos, &state->lock_states, ls_locks) {
837
		if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
L
Linus Torvalds 已提交
838
			continue;
839 840 841 842 843 844 845 846 847
		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
848 849 850 851 852 853 854 855 856 857 858
		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.
 *
 */
859
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
860 861
{
	struct nfs4_lock_state *lsp;
862
	struct nfs_server *server = state->owner->so_server;
L
Linus Torvalds 已提交
863

864
	lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
L
Linus Torvalds 已提交
865 866
	if (lsp == NULL)
		return NULL;
867
	nfs4_init_seqid_counter(&lsp->ls_seqid);
L
Linus Torvalds 已提交
868
	atomic_set(&lsp->ls_count, 1);
869
	lsp->ls_state = state;
870 871 872 873 874 875 876 877 878
	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:
879
		goto out_free;
880
	}
881 882
	lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
	if (lsp->ls_seqid.owner_id < 0)
883
		goto out_free;
884
	INIT_LIST_HEAD(&lsp->ls_locks);
L
Linus Torvalds 已提交
885
	return lsp;
886 887 888
out_free:
	kfree(lsp);
	return NULL;
L
Linus Torvalds 已提交
889 890
}

891
void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
892
{
893
	ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
894
	nfs4_destroy_seqid_counter(&lsp->ls_seqid);
895 896 897
	kfree(lsp);
}

L
Linus Torvalds 已提交
898 899 900 901 902
/*
 * Return a compatible lock_state. If no initialized lock_state structure
 * exists, return an uninitialized one.
 *
 */
903
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
904
{
905
	struct nfs4_lock_state *lsp, *new = NULL;
L
Linus Torvalds 已提交
906
	
907 908
	for(;;) {
		spin_lock(&state->state_lock);
909
		lsp = __nfs4_find_lock_state(state, owner, pid, type);
910 911 912 913 914 915 916 917 918 919
		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);
920
		new = nfs4_alloc_lock_state(state, owner, pid, type);
921 922 923 924
		if (new == NULL)
			return NULL;
	}
	spin_unlock(&state->state_lock);
925
	if (new != NULL)
926
		nfs4_free_lock_state(state->owner->so_server, new);
L
Linus Torvalds 已提交
927 928 929 930
	return lsp;
}

/*
931 932
 * Release reference to lock_state, and free it if we see that
 * it is no longer in use
L
Linus Torvalds 已提交
933
 */
934
void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
L
Linus Torvalds 已提交
935
{
936
	struct nfs4_state *state;
L
Linus Torvalds 已提交
937

938 939 940 941 942 943 944 945 946
	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);
947
	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
948 949 950
		if (nfs4_release_lockowner(lsp) == 0)
			return;
	}
951
	nfs4_free_lock_state(lsp->ls_state->owner->so_server, lsp);
L
Linus Torvalds 已提交
952 953
}

954
static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
L
Linus Torvalds 已提交
955
{
956
	struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
L
Linus Torvalds 已提交
957

958 959 960
	dst->fl_u.nfs4_fl.owner = lsp;
	atomic_inc(&lsp->ls_count);
}
L
Linus Torvalds 已提交
961

962
static void nfs4_fl_release_lock(struct file_lock *fl)
L
Linus Torvalds 已提交
963
{
964
	nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
L
Linus Torvalds 已提交
965 966
}

967
static const struct file_lock_operations nfs4_fl_lock_ops = {
968 969 970 971 972
	.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
973
{
974 975 976 977
	struct nfs4_lock_state *lsp;

	if (fl->fl_ops != NULL)
		return 0;
978 979 980
	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)
981 982
		lsp = nfs4_get_lock_state(state, NULL, fl->fl_pid,
				NFS4_FLOCK_LOCK_TYPE);
983 984
	else
		return -EINVAL;
985 986 987 988 989
	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
990 991
}

992 993
static int nfs4_copy_lock_stateid(nfs4_stateid *dst,
		struct nfs4_state *state,
994
		const struct nfs_lockowner *lockowner)
L
Linus Torvalds 已提交
995
{
996
	struct nfs4_lock_state *lsp;
997 998
	fl_owner_t fl_owner;
	pid_t fl_pid;
999
	int ret = -ENOENT;
L
Linus Torvalds 已提交
1000

1001 1002 1003 1004

	if (lockowner == NULL)
		goto out;

1005
	if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
1006
		goto out;
L
Linus Torvalds 已提交
1007

1008 1009
	fl_owner = lockowner->l_owner;
	fl_pid = lockowner->l_pid;
1010
	spin_lock(&state->state_lock);
1011
	lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
1012
	if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
1013
		nfs4_stateid_copy(dst, &lsp->ls_stateid);
1014 1015 1016 1017
		ret = 0;
		smp_rmb();
		if (!list_empty(&lsp->ls_seqid.list))
			ret = -EWOULDBLOCK;
1018
	}
1019
	spin_unlock(&state->state_lock);
L
Linus Torvalds 已提交
1020
	nfs4_put_lock_state(lsp);
1021 1022 1023 1024
out:
	return ret;
}

1025
static int nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
1026
{
1027
	int ret;
1028 1029 1030 1031 1032
	int seq;

	do {
		seq = read_seqbegin(&state->seqlock);
		nfs4_stateid_copy(dst, &state->stateid);
1033 1034 1035 1036
		ret = 0;
		smp_rmb();
		if (!list_empty(&state->owner->so_seqid.list))
			ret = -EWOULDBLOCK;
1037
	} while (read_seqretry(&state->seqlock, seq));
1038
	return ret;
1039 1040 1041 1042 1043 1044
}

/*
 * Byte-range lock aware utility to initialize the stateid of read/write
 * requests.
 */
1045
int nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state,
1046
		fmode_t fmode, const struct nfs_lockowner *lockowner)
1047
{
1048
	int ret = 0;
1049
	if (nfs4_copy_delegation_stateid(dst, state->inode, fmode))
1050 1051 1052 1053 1054 1055
		goto out;
	ret = nfs4_copy_lock_stateid(dst, state, lockowner);
	if (ret != -ENOENT)
		goto out;
	ret = nfs4_copy_open_stateid(dst, state);
out:
1056 1057
	if (nfs_server_capable(state->inode, NFS_CAP_STATEID_NFSV41))
		dst->seqid = 0;
1058
	return ret;
L
Linus Torvalds 已提交
1059 1060
}

1061
struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
1062 1063 1064
{
	struct nfs_seqid *new;

1065
	new = kmalloc(sizeof(*new), gfp_mask);
1066 1067
	if (new != NULL) {
		new->sequence = counter;
1068
		INIT_LIST_HEAD(&new->list);
1069
		new->task = NULL;
1070 1071 1072 1073
	}
	return new;
}

1074
void nfs_release_seqid(struct nfs_seqid *seqid)
L
Linus Torvalds 已提交
1075
{
1076
	struct nfs_seqid_counter *sequence;
1077

1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088
	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);
1089
	}
1090
	spin_unlock(&sequence->lock);
1091 1092 1093 1094 1095
}

void nfs_free_seqid(struct nfs_seqid *seqid)
{
	nfs_release_seqid(seqid);
1096
	kfree(seqid);
L
Linus Torvalds 已提交
1097 1098 1099
}

/*
1100 1101 1102 1103
 * 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()
 */
1104
static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
1105 1106 1107 1108 1109
{
	switch (status) {
		case 0:
			break;
		case -NFS4ERR_BAD_SEQID:
1110 1111
			if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
				return;
1112
			pr_warn_ratelimited("NFS: v4 server returned a bad"
D
Dan Muntz 已提交
1113 1114
					" sequence-id error on an"
					" unconfirmed sequence %p!\n",
1115
					seqid->sequence);
1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133
		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)
{
1134 1135 1136 1137 1138
	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
1139
		nfs4_drop_state_owner(sp);
1140 1141
	if (!nfs4_has_session(server->nfs_client))
		nfs_increment_seqid(status, seqid);
1142 1143 1144 1145 1146 1147 1148 1149 1150
}

/*
 * 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)
{
1151
	nfs_increment_seqid(status, seqid);
1152 1153 1154 1155
}

int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
{
1156
	struct nfs_seqid_counter *sequence = seqid->sequence;
1157 1158 1159
	int status = 0;

	spin_lock(&sequence->lock);
1160
	seqid->task = task;
1161 1162 1163 1164
	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;
1165
	rpc_sleep_on(&sequence->wait, task, NULL);
1166 1167
	status = -EAGAIN;
unlock:
1168 1169
	spin_unlock(&sequence->lock);
	return status;
L
Linus Torvalds 已提交
1170 1171
}

1172
static int nfs4_run_state_manager(void *);
L
Linus Torvalds 已提交
1173

1174
static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
T
Trond Myklebust 已提交
1175 1176
{
	smp_mb__before_clear_bit();
1177
	clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
T
Trond Myklebust 已提交
1178
	smp_mb__after_clear_bit();
1179
	wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
T
Trond Myklebust 已提交
1180 1181 1182
	rpc_wake_up(&clp->cl_rpcwaitq);
}

L
Linus Torvalds 已提交
1183
/*
1184
 * Schedule the nfs_client asynchronous state management routine
L
Linus Torvalds 已提交
1185
 */
1186
void nfs4_schedule_state_manager(struct nfs_client *clp)
L
Linus Torvalds 已提交
1187
{
1188
	struct task_struct *task;
1189
	char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
L
Linus Torvalds 已提交
1190

1191 1192
	if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
		return;
1193 1194
	__module_get(THIS_MODULE);
	atomic_inc(&clp->cl_count);
1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210

	/* 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
1211 1212 1213
}

/*
1214
 * Schedule a lease recovery attempt
L
Linus Torvalds 已提交
1215
 */
1216
void nfs4_schedule_lease_recovery(struct nfs_client *clp)
L
Linus Torvalds 已提交
1217 1218 1219
{
	if (!clp)
		return;
1220 1221
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1222 1223
	dprintk("%s: scheduling lease recovery for server %s\n", __func__,
			clp->cl_hostname);
1224
	nfs4_schedule_state_manager(clp);
L
Linus Torvalds 已提交
1225
}
A
Andy Adamson 已提交
1226
EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
L
Linus Torvalds 已提交
1227

1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261
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;
}

1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273
/*
 * 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);
1274 1275
	dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
			clp->cl_hostname);
1276 1277
}

1278 1279
void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
{
1280
	nfs40_handle_cb_pathdown(clp);
1281 1282 1283
	nfs4_schedule_state_manager(clp);
}

1284
static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1285 1286 1287 1288 1289 1290 1291 1292
{

	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;
	}
1293
	set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1294 1295 1296 1297
	set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
	return 1;
}

1298
static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1299 1300 1301
{
	set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
	clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1302
	set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1303 1304 1305 1306
	set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
	return 1;
}

1307
int nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
1308 1309 1310
{
	struct nfs_client *clp = server->nfs_client;

1311 1312
	if (!nfs4_valid_open_stateid(state))
		return -EBADF;
1313
	nfs4_state_mark_reclaim_nograce(clp, state);
1314 1315
	dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
			clp->cl_hostname);
1316
	nfs4_schedule_state_manager(clp);
1317
	return 0;
1318
}
A
Andy Adamson 已提交
1319
EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1320

1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336
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;
1337
		if (!nfs4_stateid_match(&state->stateid, stateid))
1338 1339 1340 1341 1342 1343 1344 1345 1346
			continue;
		nfs4_state_mark_reclaim_nograce(clp, state);
		found = true;
	}
	spin_unlock(&inode->i_lock);
	if (found)
		nfs4_schedule_state_manager(clp);
}

1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361
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);
}

1362 1363 1364
static void nfs4_state_mark_recovery_failed(struct nfs4_state *state, int error)
{
	set_bit(NFS_STATE_RECOVERY_FAILED, &state->flags);
1365
	nfs4_state_mark_open_context_bad(state);
1366 1367
}

1368

1369
static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1370 1371
{
	struct inode *inode = state->inode;
1372
	struct nfs_inode *nfsi = NFS_I(inode);
L
Linus Torvalds 已提交
1373 1374 1375
	struct file_lock *fl;
	int status = 0;

1376 1377 1378 1379
	if (inode->i_flock == NULL)
		return 0;

	/* Guard against delegation returns and new lock/unlock calls */
1380
	down_write(&nfsi->rwsem);
1381
	/* Protect inode->i_flock using the BKL */
1382
	lock_flocks();
H
Harvey Harrison 已提交
1383
	for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1384
		if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
L
Linus Torvalds 已提交
1385
			continue;
1386
		if (nfs_file_open_context(fl->fl_file)->state != state)
L
Linus Torvalds 已提交
1387
			continue;
1388
		unlock_flocks();
L
Linus Torvalds 已提交
1389 1390
		status = ops->recover_lock(state, fl);
		switch (status) {
1391 1392 1393 1394 1395 1396 1397 1398 1399
			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:
1400 1401 1402 1403
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1404
				goto out;
L
Linus Torvalds 已提交
1405
			default:
1406 1407
				printk(KERN_ERR "NFS: %s: unhandled error %d. "
					"Zeroing state\n", __func__, status);
1408 1409
			case -ENOMEM:
			case -NFS4ERR_DENIED:
L
Linus Torvalds 已提交
1410 1411
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1412
				/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1413
				status = 0;
L
Linus Torvalds 已提交
1414
		}
1415
		lock_flocks();
L
Linus Torvalds 已提交
1416
	}
1417
	unlock_flocks();
1418
out:
1419
	up_write(&nfsi->rwsem);
L
Linus Torvalds 已提交
1420 1421 1422
	return status;
}

1423
static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436
{
	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.
	 */
1437
	spin_lock(&sp->so_lock);
1438 1439
	write_seqcount_begin(&sp->so_reclaim_seqcount);
restart:
L
Linus Torvalds 已提交
1440
	list_for_each_entry(state, &sp->so_states, open_states) {
1441 1442
		if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
			continue;
1443 1444
		if (!nfs4_valid_open_stateid(state))
			continue;
L
Linus Torvalds 已提交
1445 1446
		if (state->state == 0)
			continue;
1447 1448
		atomic_inc(&state->count);
		spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1449 1450
		status = ops->recover_open(sp, state);
		if (status >= 0) {
1451 1452
			status = nfs4_reclaim_locks(state, ops);
			if (status >= 0) {
1453
				spin_lock(&state->state_lock);
1454
				list_for_each_entry(lock, &state->lock_states, ls_locks) {
1455
					if (!test_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags))
1456 1457
						pr_warn_ratelimited("NFS: "
							"%s: Lock reclaim "
1458
							"failed!\n", __func__);
1459
				}
1460
				spin_unlock(&state->state_lock);
1461
				nfs4_put_open_state(state);
1462
				spin_lock(&sp->so_lock);
1463
				goto restart;
L
Linus Torvalds 已提交
1464 1465 1466 1467
			}
		}
		switch (status) {
			default:
1468 1469
				printk(KERN_ERR "NFS: %s: unhandled error %d. "
					"Zeroing state\n", __func__, status);
L
Linus Torvalds 已提交
1470
			case -ENOENT:
1471
			case -ENOMEM:
1472
			case -ESTALE:
L
Linus Torvalds 已提交
1473 1474 1475 1476
				/*
				 * Open state on this file cannot be recovered
				 * All we can do is revert to using the zero stateid.
				 */
1477
				nfs4_state_mark_recovery_failed(state, status);
L
Linus Torvalds 已提交
1478
				break;
1479 1480
			case -EAGAIN:
				ssleep(1);
1481 1482 1483
			case -NFS4ERR_ADMIN_REVOKED:
			case -NFS4ERR_STALE_STATEID:
			case -NFS4ERR_BAD_STATEID:
1484 1485
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1486
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1487
				break;
L
Linus Torvalds 已提交
1488 1489
			case -NFS4ERR_EXPIRED:
			case -NFS4ERR_NO_GRACE:
1490
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
L
Linus Torvalds 已提交
1491
			case -NFS4ERR_STALE_CLIENTID:
1492 1493 1494 1495
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
L
Linus Torvalds 已提交
1496 1497
				goto out_err;
		}
1498
		nfs4_put_open_state(state);
1499
		spin_lock(&sp->so_lock);
1500
		goto restart;
L
Linus Torvalds 已提交
1501
	}
1502
	write_seqcount_end(&sp->so_reclaim_seqcount);
1503
	spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1504 1505
	return 0;
out_err:
1506
	nfs4_put_open_state(state);
1507 1508 1509
	spin_lock(&sp->so_lock);
	write_seqcount_end(&sp->so_reclaim_seqcount);
	spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1510 1511 1512
	return status;
}

1513 1514 1515 1516 1517 1518 1519 1520
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);
1521
	spin_lock(&state->state_lock);
1522 1523
	list_for_each_entry(lock, &state->lock_states, ls_locks) {
		lock->ls_seqid.flags = 0;
1524
		clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
1525
	}
1526
	spin_unlock(&state->state_lock);
1527 1528
}

1529 1530
static void nfs4_reset_seqids(struct nfs_server *server,
	int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1531
{
1532
	struct nfs_client *clp = server->nfs_client;
1533
	struct nfs4_state_owner *sp;
1534
	struct rb_node *pos;
1535 1536
	struct nfs4_state *state;

1537 1538 1539 1540 1541
	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);
1542
		sp->so_seqid.flags = 0;
1543
		spin_lock(&sp->so_lock);
1544
		list_for_each_entry(state, &sp->so_states, open_states) {
1545 1546
			if (mark_reclaim(clp, state))
				nfs4_clear_open_state(state);
1547
		}
1548
		spin_unlock(&sp->so_lock);
1549
	}
1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561
	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();
1562 1563
}

1564 1565 1566 1567 1568 1569 1570
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);
}

1571 1572 1573 1574 1575 1576 1577 1578
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);
}

1579
static void nfs4_clear_reclaim_server(struct nfs_server *server)
1580
{
1581
	struct nfs_client *clp = server->nfs_client;
1582 1583 1584 1585
	struct nfs4_state_owner *sp;
	struct rb_node *pos;
	struct nfs4_state *state;

1586 1587 1588 1589 1590
	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);
1591 1592
		spin_lock(&sp->so_lock);
		list_for_each_entry(state, &sp->so_states, open_states) {
1593 1594
			if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
						&state->flags))
1595 1596 1597 1598 1599
				continue;
			nfs4_state_mark_reclaim_nograce(clp, state);
		}
		spin_unlock(&sp->so_lock);
	}
1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613
	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();
1614 1615

	nfs_delegation_reap_unclaimed(clp);
1616 1617 1618 1619 1620 1621 1622 1623
	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);
1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637
}

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);
}

1638
static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1639 1640
{
	switch (error) {
1641 1642
		case 0:
			break;
1643
		case -NFS4ERR_CB_PATH_DOWN:
1644
			nfs40_handle_cb_pathdown(clp);
1645
			break;
1646 1647
		case -NFS4ERR_NO_GRACE:
			nfs4_state_end_reclaim_reboot(clp);
1648
			break;
1649 1650 1651
		case -NFS4ERR_STALE_CLIENTID:
		case -NFS4ERR_LEASE_MOVED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1652
			nfs4_state_clear_reclaim_reboot(clp);
1653 1654 1655 1656 1657
			nfs4_state_start_reclaim_reboot(clp);
			break;
		case -NFS4ERR_EXPIRED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
			nfs4_state_start_reclaim_nograce(clp);
1658
			break;
1659 1660 1661 1662 1663 1664
		case -NFS4ERR_BADSESSION:
		case -NFS4ERR_BADSLOT:
		case -NFS4ERR_BAD_HIGH_SLOT:
		case -NFS4ERR_DEADSESSION:
		case -NFS4ERR_SEQ_FALSE_RETRY:
		case -NFS4ERR_SEQ_MISORDERED:
1665
			set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1666
			/* Zero session reset errors */
1667
			break;
1668 1669 1670
		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
			set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
			break;
1671
		default:
1672 1673
			dprintk("%s: failed to handle error %d for server %s\n",
					__func__, error, clp->cl_hostname);
1674
			return error;
1675
	}
1676 1677
	dprintk("%s: handled error %d for server %s\n", __func__, error,
			clp->cl_hostname);
1678
	return 0;
1679 1680
}

1681
static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1682
{
1683 1684
	struct nfs4_state_owner *sp;
	struct nfs_server *server;
1685
	struct rb_node *pos;
L
Linus Torvalds 已提交
1686 1687
	int status = 0;

1688
restart:
1689 1690
	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1691
		nfs4_purge_state_owners(server);
1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711
		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);
			}

1712
			nfs4_put_state_owner(sp);
1713
			goto restart;
1714
		}
1715
		spin_unlock(&clp->cl_lock);
1716
	}
1717
	rcu_read_unlock();
1718 1719 1720 1721 1722 1723
	return status;
}

static int nfs4_check_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1724 1725
	const struct nfs4_state_maintenance_ops *ops =
		clp->cl_mvops->state_renewal_ops;
1726
	int status;
L
Linus Torvalds 已提交
1727

1728 1729 1730
	/* Is the client already known to have an expired lease? */
	if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		return 0;
1731 1732 1733
	spin_lock(&clp->cl_lock);
	cred = ops->get_state_renewal_cred_locked(clp);
	spin_unlock(&clp->cl_lock);
1734 1735
	if (cred == NULL) {
		cred = nfs4_get_setclientid_cred(clp);
1736
		status = -ENOKEY;
1737 1738
		if (cred == NULL)
			goto out;
1739
	}
1740
	status = ops->renew_lease(clp, cred);
1741 1742
	put_rpccred(cred);
out:
1743
	return nfs4_recovery_handle_error(clp, status);
1744 1745
}

1746 1747
/* Set NFS4CLNT_LEASE_EXPIRED and reclaim reboot state for all v4.0 errors
 * and for recoverable errors on EXCHANGE_ID for v4.1
1748 1749 1750 1751
 */
static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
{
	switch (status) {
1752 1753 1754 1755 1756
	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);
1757 1758
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
		break;
1759 1760
	case -NFS4ERR_STALE_CLIENTID:
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1761 1762
		nfs4_state_clear_reclaim_reboot(clp);
		nfs4_state_start_reclaim_reboot(clp);
1763
		break;
1764 1765 1766 1767 1768 1769
	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;
1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783
	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);
1784 1785
		dprintk("%s: exit with error %d for server %s\n",
				__func__, -EPROTONOSUPPORT, clp->cl_hostname);
1786 1787 1788 1789
		return -EPROTONOSUPPORT;
	case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
				 * in nfs4_exchange_id */
	default:
1790 1791
		dprintk("%s: exit with error %d for server %s\n", __func__,
				status, clp->cl_hostname);
1792 1793 1794
		return status;
	}
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1795 1796
	dprintk("%s: handled error %d for server %s\n", __func__, status,
			clp->cl_hostname);
1797 1798 1799
	return 0;
}

1800
static int nfs4_establish_lease(struct nfs_client *clp)
1801 1802
{
	struct rpc_cred *cred;
1803 1804
	const struct nfs4_state_recovery_ops *ops =
		clp->cl_mvops->reboot_recovery_ops;
1805
	int status;
1806

1807
	cred = ops->get_clid_cred(clp);
1808 1809 1810 1811 1812
	if (cred == NULL)
		return -ENOENT;
	status = ops->establish_clid(clp, cred);
	put_rpccred(cred);
	if (status != 0)
1813 1814 1815 1816 1817
		return status;
	pnfs_destroy_all_layouts(clp);
	return 0;
}

1818 1819 1820 1821
/*
 * Returns zero or a negative errno.  NFS4ERR values are converted
 * to local errno values.
 */
1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843
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)
1844
		return nfs4_handle_reclaim_lease_error(clp, status);
1845 1846 1847
	clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
	nfs4_state_start_reclaim_nograce(clp);
1848
	return 0;
1849 1850
}

1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912
/**
 * 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;
	rpc_authflavor_t *flavors, flav, save;
	struct rpc_clnt *clnt;
	struct rpc_cred *cred;
	int i, len, status;

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

	len = NFS_MAX_SECFLAVORS;
	flavors = kcalloc(len, sizeof(*flavors), GFP_KERNEL);
	if (flavors == NULL) {
		status = -ENOMEM;
		goto out;
	}
	len = rpcauth_list_flavors(flavors, len);
	if (len < 0) {
		status = len;
		goto out_free;
	}
	clnt = clp->cl_rpcclient;
	save = clnt->cl_auth->au_flavor;
	i = 0;

	mutex_lock(&nfs_clid_init_mutex);
	status  = -ENOENT;
again:
	cred = ops->get_clid_cred(clp);
	if (cred == NULL)
		goto out_unlock;

	status = ops->detect_trunking(clp, result, cred);
	put_rpccred(cred);
	switch (status) {
	case 0:
		break;

	case -EACCES:
		if (clp->cl_machine_cred == NULL)
			break;
		/* 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);
1913
	case -NFS4ERR_STALE_CLIENTID:
1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953
		dprintk("NFS: %s after status %d, retrying\n",
			__func__, status);
		goto again;

	case -NFS4ERR_CLID_INUSE:
	case -NFS4ERR_WRONGSEC:
		status = -EPERM;
		if (i >= len)
			break;

		flav = flavors[i++];
		if (flav == save)
			flav = flavors[i++];
		clnt = rpc_clone_client_set_auth(clnt, flav);
		if (IS_ERR(clnt)) {
			status = PTR_ERR(clnt);
			break;
		}
		clp->cl_rpcclient = clnt;
		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;
	}

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

1954
#ifdef CONFIG_NFS_V4_1
1955
void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
1956
{
1957 1958
	struct nfs_client *clp = session->clp;

1959 1960 1961 1962 1963 1964 1965
	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);
	}
1966
	nfs4_schedule_lease_recovery(clp);
1967
}
1968
EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
1969

1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981
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);
}

1982 1983 1984 1985 1986
void nfs41_server_notify_highest_slotid_update(struct nfs_client *clp)
{
	nfs41_ping_server(clp);
}

1987 1988 1989
static void nfs4_reset_all_state(struct nfs_client *clp)
{
	if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1990
		set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1991
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1992
		nfs4_state_start_reclaim_nograce(clp);
1993 1994
		dprintk("%s: scheduling reset of all state for server %s!\n",
				__func__, clp->cl_hostname);
1995
		nfs4_schedule_state_manager(clp);
1996 1997 1998 1999 2000 2001 2002
	}
}

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);
2003 2004
		dprintk("%s: server %s rebooted!\n", __func__,
				clp->cl_hostname);
2005
		nfs4_schedule_state_manager(clp);
2006 2007 2008 2009 2010 2011
	}
}

static void nfs41_handle_state_revoked(struct nfs_client *clp)
{
	nfs4_reset_all_state(clp);
2012
	dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
2013 2014 2015 2016 2017 2018
}

static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
{
	/* This will need to handle layouts too */
	nfs_expire_all_delegations(clp);
2019 2020
	dprintk("%s: Recallable state revoked on server %s!\n", __func__,
			clp->cl_hostname);
2021 2022
}

2023
static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
2024 2025 2026
{
	nfs_expire_all_delegations(clp);
	if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
2027
		nfs4_schedule_state_manager(clp);
2028 2029
	dprintk("%s: server %s declared a backchannel fault\n", __func__,
			clp->cl_hostname);
2030 2031
}

2032 2033 2034 2035 2036 2037 2038
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);
}

2039 2040 2041 2042
void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
{
	if (!flags)
		return;
2043 2044 2045 2046

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

2047
	if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
2048
		nfs41_handle_server_reboot(clp);
2049
	if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
2050 2051
			    SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
			    SEQ4_STATUS_ADMIN_STATE_REVOKED |
2052 2053
			    SEQ4_STATUS_LEASE_MOVED))
		nfs41_handle_state_revoked(clp);
2054
	if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
2055
		nfs41_handle_recallable_state_revoked(clp);
2056 2057 2058 2059
	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))
2060
		nfs41_handle_cb_path_down(clp);
2061 2062
}

2063 2064
static int nfs4_reset_session(struct nfs_client *clp)
{
2065
	struct rpc_cred *cred;
2066 2067
	int status;

2068 2069
	if (!nfs4_has_session(clp))
		return 0;
2070
	nfs4_begin_drain_session(clp);
2071 2072
	cred = nfs4_get_exchange_id_cred(clp);
	status = nfs4_proc_destroy_session(clp->cl_session, cred);
2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084
	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:
2085
		status = nfs4_recovery_handle_error(clp, status);
2086 2087 2088 2089
		goto out;
	}

	memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
2090
	status = nfs4_proc_create_session(clp, cred);
2091
	if (status) {
2092 2093
		dprintk("%s: session reset failed with status %d for server %s!\n",
			__func__, status, clp->cl_hostname);
2094
		status = nfs4_handle_reclaim_lease_error(clp, status);
2095 2096
		goto out;
	}
2097
	nfs41_finish_session_reset(clp);
2098 2099
	dprintk("%s: session reset was successful for server %s!\n",
			__func__, clp->cl_hostname);
2100
out:
2101 2102
	if (cred)
		put_rpccred(cred);
2103 2104
	return status;
}
2105

2106 2107
static int nfs4_bind_conn_to_session(struct nfs_client *clp)
{
2108 2109 2110
	struct rpc_cred *cred;
	int ret;

2111 2112
	if (!nfs4_has_session(clp))
		return 0;
2113
	nfs4_begin_drain_session(clp);
2114 2115 2116 2117
	cred = nfs4_get_exchange_id_cred(clp);
	ret = nfs4_proc_bind_conn_to_session(clp, cred);
	if (cred)
		put_rpccred(cred);
2118
	clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2119 2120
	switch (ret) {
	case 0:
2121 2122
		dprintk("%s: bind_conn_to_session was successful for server %s!\n",
			__func__, clp->cl_hostname);
2123 2124 2125 2126 2127 2128 2129 2130 2131
		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;
2132
}
2133
#else /* CONFIG_NFS_V4_1 */
2134
static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
2135
static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
2136 2137 2138 2139 2140

static int nfs4_bind_conn_to_session(struct nfs_client *clp)
{
	return 0;
}
2141 2142
#endif /* CONFIG_NFS_V4_1 */

2143
static void nfs4_state_manager(struct nfs_client *clp)
2144 2145
{
	int status = 0;
2146
	const char *section = "", *section_sep = "";
2147 2148

	/* Ensure exclusive access to NFSv4 state */
2149
	do {
2150
		if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
2151
			section = "purge state";
2152
			status = nfs4_purge_lease(clp);
2153 2154
			if (status < 0)
				goto out_error;
2155
			continue;
2156 2157
		}

2158
		if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
2159
			section = "lease expired";
2160 2161
			/* We're going to have to re-establish a clientid */
			status = nfs4_reclaim_lease(clp);
2162
			if (status < 0)
2163
				goto out_error;
2164
			continue;
2165 2166
		}

2167
		/* Initialize or reset the session */
2168
		if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
2169
			section = "reset session";
2170
			status = nfs4_reset_session(clp);
2171 2172 2173
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
				continue;
			if (status < 0)
2174 2175
				goto out_error;
		}
2176

2177 2178
		/* Send BIND_CONN_TO_SESSION */
		if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2179
				&clp->cl_state)) {
2180
			section = "bind conn to session";
2181 2182 2183
			status = nfs4_bind_conn_to_session(clp);
			if (status < 0)
				goto out_error;
2184
			continue;
2185 2186
		}

2187 2188 2189 2190 2191 2192 2193 2194
		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;
		}

2195
		/* First recover reboot state... */
2196
		if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
2197
			section = "reclaim reboot";
2198
			status = nfs4_do_reclaim(clp,
2199
				clp->cl_mvops->reboot_recovery_ops);
2200
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2201
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
2202
				continue;
2203
			nfs4_state_end_reclaim_reboot(clp);
2204 2205 2206 2207
			if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
				continue;
			if (status < 0)
				goto out_error;
2208 2209
		}

2210 2211
		/* Now recover expired state... */
		if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
2212
			section = "reclaim nograce";
2213
			status = nfs4_do_reclaim(clp,
2214
				clp->cl_mvops->nograce_recovery_ops);
2215
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2216
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
2217 2218 2219
			    test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
				continue;
			if (status < 0)
2220
				goto out_error;
L
Linus Torvalds 已提交
2221
		}
2222

2223
		nfs4_end_drain_session(clp);
2224 2225 2226 2227
		if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
			nfs_client_return_marked_delegations(clp);
			continue;
		}
2228 2229

		nfs4_clear_state_manager_bit(clp);
2230 2231 2232 2233 2234
		/* 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;
2235
	} while (atomic_read(&clp->cl_count) > 1);
2236
	return;
L
Linus Torvalds 已提交
2237
out_error:
2238 2239 2240 2241 2242
	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);
2243
	ssleep(1);
2244
	nfs4_end_drain_session(clp);
2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256
	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
2257 2258 2259 2260 2261 2262 2263
}

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