nfs4state.c 48.9 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
#include "nfs4_fs.h"
L
Linus Torvalds 已提交
55 56
#include "callback.h"
#include "delegation.h"
57
#include "internal.h"
58
#include "pnfs.h"
L
Linus Torvalds 已提交
59 60 61

#define OPENOWNER_POOL_SIZE	8

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

static LIST_HEAD(nfs4_clientid_list);

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

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

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

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

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

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

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

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

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

137 138 139 140 141 142 143 144 145 146 147 148
/**
 * 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;

149 150 151 152 153
	/* Use machine credentials if available */
	cred = nfs4_get_machine_cred_locked(clp);
	if (cred != NULL)
		goto out;

154 155 156 157 158 159 160
	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();
161 162

out:
163 164 165
	return cred;
}

166 167
#if defined(CONFIG_NFS_V4_1)

R
Ricardo Labiaga 已提交
168 169 170 171 172
static int nfs41_setup_state_renewal(struct nfs_client *clp)
{
	int status;
	struct nfs_fsinfo fsinfo;

173 174 175 176 177
	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
178 179 180 181 182 183 184 185 186 187 188 189 190 191
	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;
}

192 193 194 195 196
/*
 * 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.
 */
197
static void nfs4_end_drain_session(struct nfs_client *clp)
198
{
199
	struct nfs4_session *ses = clp->cl_session;
200
	struct nfs4_slot_table *tbl;
201 202
	int max_slots;

203 204
	if (ses == NULL)
		return;
205
	tbl = &ses->fc_slot_table;
206
	if (test_and_clear_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
207 208
		spin_lock(&tbl->slot_tbl_lock);
		max_slots = tbl->max_slots;
209
		while (max_slots--) {
210 211 212
			if (rpc_wake_up_first(&tbl->slot_tbl_waitq,
						nfs4_set_task_privileged,
						NULL) == NULL)
213 214
				break;
		}
215
		spin_unlock(&tbl->slot_tbl_lock);
216
	}
217 218
}

219
static int nfs4_wait_on_slot_tbl(struct nfs4_slot_table *tbl)
220 221
{
	spin_lock(&tbl->slot_tbl_lock);
222
	if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
223
		INIT_COMPLETION(tbl->complete);
224
		spin_unlock(&tbl->slot_tbl_lock);
225
		return wait_for_completion_interruptible(&tbl->complete);
226 227 228 229 230
	}
	spin_unlock(&tbl->slot_tbl_lock);
	return 0;
}

231 232 233 234 235 236 237 238 239 240 241 242 243 244
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);
}

245 246 247 248
int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
{
	int status;

249 250
	if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
		goto do_confirm;
251
	nfs4_begin_drain_session(clp);
252
	status = nfs4_proc_exchange_id(clp, cred);
R
Ricardo Labiaga 已提交
253 254
	if (status != 0)
		goto out;
255 256
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
do_confirm:
R
Ricardo Labiaga 已提交
257 258 259
	status = nfs4_proc_create_session(clp);
	if (status != 0)
		goto out;
260
	clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
R
Ricardo Labiaga 已提交
261 262 263
	nfs41_setup_state_renewal(clp);
	nfs_mark_client_ready(clp, NFS_CS_READY);
out:
264 265 266
	return status;
}

267 268 269 270 271 272 273 274 275 276 277 278
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 */

279 280
static struct rpc_cred *
nfs4_get_setclientid_cred_server(struct nfs_server *server)
281
{
282 283
	struct nfs_client *clp = server->nfs_client;
	struct rpc_cred *cred = NULL;
284
	struct nfs4_state_owner *sp;
285
	struct rb_node *pos;
286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305

	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;
306
	struct rpc_cred *cred;
307

308 309
	spin_lock(&clp->cl_lock);
	cred = nfs4_get_machine_cred_locked(clp);
310
	spin_unlock(&clp->cl_lock);
311 312
	if (cred != NULL)
		goto out;
313 314 315 316 317 318

	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;
319
	}
320 321
	rcu_read_unlock();

322 323
out:
	return cred;
324 325
}

L
Linus Torvalds 已提交
326
static struct nfs4_state_owner *
327
nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
L
Linus Torvalds 已提交
328
{
329
	struct rb_node **p = &server->state_owners.rb_node,
330
		       *parent = NULL;
331
	struct nfs4_state_owner *sp;
L
Linus Torvalds 已提交
332

333 334
	while (*p != NULL) {
		parent = *p;
335
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
336 337 338 339 340 341

		if (cred < sp->so_cred)
			p = &parent->rb_left;
		else if (cred > sp->so_cred)
			p = &parent->rb_right;
		else {
342 343
			if (!list_empty(&sp->so_lru))
				list_del_init(&sp->so_lru);
344
			atomic_inc(&sp->so_count);
345
			return sp;
346
		}
L
Linus Torvalds 已提交
347
	}
348
	return NULL;
L
Linus Torvalds 已提交
349 350
}

351
static struct nfs4_state_owner *
352
nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
353
{
354 355
	struct nfs_server *server = new->so_server;
	struct rb_node **p = &server->state_owners.rb_node,
356 357
		       *parent = NULL;
	struct nfs4_state_owner *sp;
358
	int err;
359 360 361

	while (*p != NULL) {
		parent = *p;
362
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
363 364 365 366 367 368

		if (new->so_cred < sp->so_cred)
			p = &parent->rb_left;
		else if (new->so_cred > sp->so_cred)
			p = &parent->rb_right;
		else {
369 370
			if (!list_empty(&sp->so_lru))
				list_del_init(&sp->so_lru);
371 372 373 374
			atomic_inc(&sp->so_count);
			return sp;
		}
	}
375
	err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
376 377
	if (err)
		return ERR_PTR(err);
378 379
	rb_link_node(&new->so_server_node, parent, p);
	rb_insert_color(&new->so_server_node, &server->state_owners);
380 381 382 383
	return new;
}

static void
384
nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
385
{
386 387 388 389
	struct nfs_server *server = sp->so_server;

	if (!RB_EMPTY_NODE(&sp->so_server_node))
		rb_erase(&sp->so_server_node, &server->state_owners);
390
	ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
391 392
}

393 394 395
static void
nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
{
396
	sc->create_time = ktime_get();
397 398 399 400 401 402 403 404 405 406 407 408 409
	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
410 411 412 413 414 415
/*
 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
 * create a new state_owner.
 *
 */
static struct nfs4_state_owner *
416 417 418
nfs4_alloc_state_owner(struct nfs_server *server,
		struct rpc_cred *cred,
		gfp_t gfp_flags)
L
Linus Torvalds 已提交
419 420 421
{
	struct nfs4_state_owner *sp;

422
	sp = kzalloc(sizeof(*sp), gfp_flags);
L
Linus Torvalds 已提交
423 424
	if (!sp)
		return NULL;
425 426
	sp->so_server = server;
	sp->so_cred = get_rpccred(cred);
427
	spin_lock_init(&sp->so_lock);
L
Linus Torvalds 已提交
428
	INIT_LIST_HEAD(&sp->so_states);
429
	nfs4_init_seqid_counter(&sp->so_seqid);
L
Linus Torvalds 已提交
430
	atomic_set(&sp->so_count, 1);
431
	INIT_LIST_HEAD(&sp->so_lru);
L
Linus Torvalds 已提交
432 433 434
	return sp;
}

435
static void
L
Linus Torvalds 已提交
436 437
nfs4_drop_state_owner(struct nfs4_state_owner *sp)
{
438 439 440
	struct rb_node *rb_node = &sp->so_server_node;

	if (!RB_EMPTY_NODE(rb_node)) {
441 442
		struct nfs_server *server = sp->so_server;
		struct nfs_client *clp = server->nfs_client;
443 444

		spin_lock(&clp->cl_lock);
445 446 447 448
		if (!RB_EMPTY_NODE(rb_node)) {
			rb_erase(rb_node, &server->state_owners);
			RB_CLEAR_NODE(rb_node);
		}
449 450
		spin_unlock(&clp->cl_lock);
	}
L
Linus Torvalds 已提交
451 452
}

453 454
static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
{
455
	nfs4_destroy_seqid_counter(&sp->so_seqid);
456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484
	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);
	}
}

485 486 487 488 489 490 491 492
/**
 * 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,
493 494
					      struct rpc_cred *cred,
					      gfp_t gfp_flags)
L
Linus Torvalds 已提交
495
{
496
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
497 498 499
	struct nfs4_state_owner *sp, *new;

	spin_lock(&clp->cl_lock);
500
	sp = nfs4_find_state_owner_locked(server, cred);
L
Linus Torvalds 已提交
501 502
	spin_unlock(&clp->cl_lock);
	if (sp != NULL)
503
		goto out;
504
	new = nfs4_alloc_state_owner(server, cred, gfp_flags);
505
	if (new == NULL)
506
		goto out;
507 508 509 510 511 512 513
	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));
514 515
	if (sp != new)
		nfs4_free_state_owner(new);
516 517
out:
	nfs4_gc_state_owners(server);
518
	return sp;
L
Linus Torvalds 已提交
519 520
}

521 522 523
/**
 * nfs4_put_state_owner - Release a nfs4_state_owner
 * @sp: state owner data to release
524 525 526 527 528 529 530 531
 *
 * 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.
532
 */
L
Linus Torvalds 已提交
533 534
void nfs4_put_state_owner(struct nfs4_state_owner *sp)
{
535 536
	struct nfs_server *server = sp->so_server;
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
537 538 539

	if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
		return;
540

541 542 543
	sp->so_expires = jiffies;
	list_add_tail(&sp->so_lru, &server->state_owners_lru);
	spin_unlock(&clp->cl_lock);
544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563
}

/**
 * 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
564
	spin_unlock(&clp->cl_lock);
565 566 567 568 569

	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
570 571 572 573 574 575 576
}

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

577
	state = kzalloc(sizeof(*state), GFP_NOFS);
L
Linus Torvalds 已提交
578 579 580 581
	if (!state)
		return NULL;
	atomic_set(&state->count, 1);
	INIT_LIST_HEAD(&state->lock_states);
582
	spin_lock_init(&state->state_lock);
583
	seqlock_init(&state->seqlock);
L
Linus Torvalds 已提交
584 585 586
	return state;
}

587
void
588
nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
589
{
590
	if (state->state == fmode)
591 592
		return;
	/* NB! List reordering - see the reclaim code for why.  */
593 594
	if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
		if (fmode & FMODE_WRITE)
595 596 597 598
			list_move(&state->open_states, &state->owner->so_states);
		else
			list_move_tail(&state->open_states, &state->owner->so_states);
	}
599
	state->state = fmode;
600 601
}

L
Linus Torvalds 已提交
602 603 604 605 606 607 608
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) {
609
		if (state->owner != owner)
L
Linus Torvalds 已提交
610
			continue;
611
		if (atomic_inc_not_zero(&state->count))
L
Linus Torvalds 已提交
612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634
			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();
635
	spin_lock(&owner->so_lock);
L
Linus Torvalds 已提交
636 637 638 639 640 641 642
	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);
643 644
		ihold(inode);
		state->inode = inode;
L
Linus Torvalds 已提交
645
		spin_unlock(&inode->i_lock);
646 647 648 649
		/* 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
650 651
	} else {
		spin_unlock(&inode->i_lock);
652
		spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
653 654 655 656 657 658 659 660 661 662 663 664
		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;

665
	if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
L
Linus Torvalds 已提交
666
		return;
667
	spin_lock(&inode->i_lock);
668
	list_del(&state->inode_states);
L
Linus Torvalds 已提交
669
	list_del(&state->open_states);
670 671
	spin_unlock(&inode->i_lock);
	spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
672 673 674 675 676 677
	iput(inode);
	nfs4_free_open_state(state);
	nfs4_put_state_owner(owner);
}

/*
678
 * Close the current file.
L
Linus Torvalds 已提交
679
 */
680
static void __nfs4_close(struct nfs4_state *state,
681
		fmode_t fmode, gfp_t gfp_mask, int wait)
L
Linus Torvalds 已提交
682 683
{
	struct nfs4_state_owner *owner = state->owner;
684
	int call_close = 0;
685
	fmode_t newstate;
L
Linus Torvalds 已提交
686 687 688

	atomic_inc(&owner->so_count);
	/* Protect against nfs4_find_state() */
689
	spin_lock(&owner->so_lock);
690
	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
691 692 693 694 695 696 697 698 699
		case FMODE_READ:
			state->n_rdonly--;
			break;
		case FMODE_WRITE:
			state->n_wronly--;
			break;
		case FMODE_READ|FMODE_WRITE:
			state->n_rdwr--;
	}
700
	newstate = FMODE_READ|FMODE_WRITE;
701
	if (state->n_rdwr == 0) {
702
		if (state->n_rdonly == 0) {
703
			newstate &= ~FMODE_READ;
704 705 706 707
			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) {
708
			newstate &= ~FMODE_WRITE;
709 710 711 712 713
			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);
714
	}
715
	nfs4_state_set_mode_locked(state, newstate);
716
	spin_unlock(&owner->so_lock);
717

718
	if (!call_close) {
719 720
		nfs4_put_open_state(state);
		nfs4_put_state_owner(owner);
F
Fred Isaman 已提交
721 722 723
	} else {
		bool roc = pnfs_roc(state->inode);

724
		nfs4_do_close(state, gfp_mask, wait, roc);
F
Fred Isaman 已提交
725
	}
726 727
}

728
void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
729
{
730
	__nfs4_close(state, fmode, GFP_NOFS, 0);
731 732
}

733
void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
734
{
735
	__nfs4_close(state, fmode, GFP_KERNEL, 1);
L
Linus Torvalds 已提交
736 737 738 739 740 741 742
}

/*
 * Search the state->lock_states for an existing lock_owner
 * that is compatible with current->files
 */
static struct nfs4_lock_state *
743
__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
744 745 746
{
	struct nfs4_lock_state *pos;
	list_for_each_entry(pos, &state->lock_states, ls_locks) {
747
		if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
L
Linus Torvalds 已提交
748
			continue;
749 750 751 752 753 754 755 756 757
		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
758 759 760 761 762 763 764 765 766 767 768
		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.
 *
 */
769
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
770 771
{
	struct nfs4_lock_state *lsp;
772
	struct nfs_server *server = state->owner->so_server;
L
Linus Torvalds 已提交
773

774
	lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
L
Linus Torvalds 已提交
775 776
	if (lsp == NULL)
		return NULL;
777
	nfs4_init_seqid_counter(&lsp->ls_seqid);
L
Linus Torvalds 已提交
778
	atomic_set(&lsp->ls_count, 1);
779
	lsp->ls_state = state;
780 781 782 783 784 785 786 787 788
	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:
789
		goto out_free;
790
	}
791 792
	lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
	if (lsp->ls_seqid.owner_id < 0)
793
		goto out_free;
794
	INIT_LIST_HEAD(&lsp->ls_locks);
L
Linus Torvalds 已提交
795
	return lsp;
796 797 798
out_free:
	kfree(lsp);
	return NULL;
L
Linus Torvalds 已提交
799 800
}

801
void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
802
{
803
	ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
804
	nfs4_destroy_seqid_counter(&lsp->ls_seqid);
805 806 807
	kfree(lsp);
}

L
Linus Torvalds 已提交
808 809 810 811 812
/*
 * Return a compatible lock_state. If no initialized lock_state structure
 * exists, return an uninitialized one.
 *
 */
813
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
814
{
815
	struct nfs4_lock_state *lsp, *new = NULL;
L
Linus Torvalds 已提交
816
	
817 818
	for(;;) {
		spin_lock(&state->state_lock);
819
		lsp = __nfs4_find_lock_state(state, owner, pid, type);
820 821 822 823 824 825 826 827 828 829
		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);
830
		new = nfs4_alloc_lock_state(state, owner, pid, type);
831 832 833 834
		if (new == NULL)
			return NULL;
	}
	spin_unlock(&state->state_lock);
835
	if (new != NULL)
836
		nfs4_free_lock_state(state->owner->so_server, new);
L
Linus Torvalds 已提交
837 838 839 840
	return lsp;
}

/*
841 842
 * Release reference to lock_state, and free it if we see that
 * it is no longer in use
L
Linus Torvalds 已提交
843
 */
844
void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
L
Linus Torvalds 已提交
845
{
846
	struct nfs4_state *state;
L
Linus Torvalds 已提交
847

848 849 850 851 852 853 854 855 856
	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);
857 858 859 860
	if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
		if (nfs4_release_lockowner(lsp) == 0)
			return;
	}
861
	nfs4_free_lock_state(lsp->ls_state->owner->so_server, lsp);
L
Linus Torvalds 已提交
862 863
}

864
static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
L
Linus Torvalds 已提交
865
{
866
	struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
L
Linus Torvalds 已提交
867

868 869 870
	dst->fl_u.nfs4_fl.owner = lsp;
	atomic_inc(&lsp->ls_count);
}
L
Linus Torvalds 已提交
871

872
static void nfs4_fl_release_lock(struct file_lock *fl)
L
Linus Torvalds 已提交
873
{
874
	nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
L
Linus Torvalds 已提交
875 876
}

877
static const struct file_lock_operations nfs4_fl_lock_ops = {
878 879 880 881 882
	.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
883
{
884 885 886 887
	struct nfs4_lock_state *lsp;

	if (fl->fl_ops != NULL)
		return 0;
888 889 890
	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)
891 892
		lsp = nfs4_get_lock_state(state, NULL, fl->fl_pid,
				NFS4_FLOCK_LOCK_TYPE);
893 894
	else
		return -EINVAL;
895 896 897 898 899
	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
900 901
}

902 903
static bool nfs4_copy_lock_stateid(nfs4_stateid *dst, struct nfs4_state *state,
		fl_owner_t fl_owner, pid_t fl_pid)
L
Linus Torvalds 已提交
904
{
905
	struct nfs4_lock_state *lsp;
906
	bool ret = false;
L
Linus Torvalds 已提交
907

908
	if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
909
		goto out;
L
Linus Torvalds 已提交
910

911
	spin_lock(&state->state_lock);
912
	lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
913
	if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0) {
914
		nfs4_stateid_copy(dst, &lsp->ls_stateid);
915 916
		ret = true;
	}
917
	spin_unlock(&state->state_lock);
L
Linus Torvalds 已提交
918
	nfs4_put_lock_state(lsp);
919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944
out:
	return ret;
}

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

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

/*
 * Byte-range lock aware utility to initialize the stateid of read/write
 * requests.
 */
void nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state,
		fmode_t fmode, fl_owner_t fl_owner, pid_t fl_pid)
{
	if (nfs4_copy_delegation_stateid(dst, state->inode, fmode))
		return;
	if (nfs4_copy_lock_stateid(dst, state, fl_owner, fl_pid))
		return;
	nfs4_copy_open_stateid(dst, state);
L
Linus Torvalds 已提交
945 946
}

947
struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
948 949 950
{
	struct nfs_seqid *new;

951
	new = kmalloc(sizeof(*new), gfp_mask);
952 953
	if (new != NULL) {
		new->sequence = counter;
954
		INIT_LIST_HEAD(&new->list);
955
		new->task = NULL;
956 957 958 959
	}
	return new;
}

960
void nfs_release_seqid(struct nfs_seqid *seqid)
L
Linus Torvalds 已提交
961
{
962
	struct nfs_seqid_counter *sequence;
963

964 965 966 967 968 969 970 971 972 973 974
	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);
975
	}
976
	spin_unlock(&sequence->lock);
977 978 979 980 981
}

void nfs_free_seqid(struct nfs_seqid *seqid)
{
	nfs_release_seqid(seqid);
982
	kfree(seqid);
L
Linus Torvalds 已提交
983 984 985
}

/*
986 987 988 989
 * 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()
 */
990
static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
991
{
992
	BUG_ON(list_first_entry(&seqid->sequence->list, struct nfs_seqid, list) != seqid);
993 994 995 996
	switch (status) {
		case 0:
			break;
		case -NFS4ERR_BAD_SEQID:
997 998
			if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
				return;
999
			pr_warn_ratelimited("NFS: v4 server returned a bad"
D
Dan Muntz 已提交
1000 1001
					" sequence-id error on an"
					" unconfirmed sequence %p!\n",
1002
					seqid->sequence);
1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020
		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)
{
1021 1022 1023 1024 1025
	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
1026
		nfs4_drop_state_owner(sp);
1027 1028
	if (!nfs4_has_session(server->nfs_client))
		nfs_increment_seqid(status, seqid);
1029 1030 1031 1032 1033 1034 1035 1036 1037
}

/*
 * 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)
{
1038
	nfs_increment_seqid(status, seqid);
1039 1040 1041 1042
}

int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
{
1043
	struct nfs_seqid_counter *sequence = seqid->sequence;
1044 1045 1046
	int status = 0;

	spin_lock(&sequence->lock);
1047
	seqid->task = task;
1048 1049 1050 1051
	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;
1052
	rpc_sleep_on(&sequence->wait, task, NULL);
1053 1054
	status = -EAGAIN;
unlock:
1055 1056
	spin_unlock(&sequence->lock);
	return status;
L
Linus Torvalds 已提交
1057 1058
}

1059
static int nfs4_run_state_manager(void *);
L
Linus Torvalds 已提交
1060

1061
static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
T
Trond Myklebust 已提交
1062 1063
{
	smp_mb__before_clear_bit();
1064
	clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
T
Trond Myklebust 已提交
1065
	smp_mb__after_clear_bit();
1066
	wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
T
Trond Myklebust 已提交
1067 1068 1069
	rpc_wake_up(&clp->cl_rpcwaitq);
}

L
Linus Torvalds 已提交
1070
/*
1071
 * Schedule the nfs_client asynchronous state management routine
L
Linus Torvalds 已提交
1072
 */
1073
void nfs4_schedule_state_manager(struct nfs_client *clp)
L
Linus Torvalds 已提交
1074
{
1075
	struct task_struct *task;
1076
	char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
L
Linus Torvalds 已提交
1077

1078 1079
	if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
		return;
1080 1081
	__module_get(THIS_MODULE);
	atomic_inc(&clp->cl_count);
1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097

	/* 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
1098 1099 1100
}

/*
1101
 * Schedule a lease recovery attempt
L
Linus Torvalds 已提交
1102
 */
1103
void nfs4_schedule_lease_recovery(struct nfs_client *clp)
L
Linus Torvalds 已提交
1104 1105 1106
{
	if (!clp)
		return;
1107 1108
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1109
	nfs4_schedule_state_manager(clp);
L
Linus Torvalds 已提交
1110
}
A
Andy Adamson 已提交
1111
EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
L
Linus Torvalds 已提交
1112

1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126
/*
 * 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);
}

1127 1128
void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
{
1129
	nfs40_handle_cb_pathdown(clp);
1130 1131 1132
	nfs4_schedule_state_manager(clp);
}

1133
static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1134 1135 1136 1137 1138 1139 1140 1141
{

	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;
	}
1142
	set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1143 1144 1145 1146
	set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
	return 1;
}

1147
static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1148 1149 1150
{
	set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
	clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1151
	set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1152 1153 1154 1155
	set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
	return 1;
}

1156 1157 1158 1159 1160 1161 1162
void nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
{
	struct nfs_client *clp = server->nfs_client;

	nfs4_state_mark_reclaim_nograce(clp, state);
	nfs4_schedule_state_manager(clp);
}
A
Andy Adamson 已提交
1163
EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1164

1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180
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;
1181
		if (!nfs4_stateid_match(&state->stateid, stateid))
1182 1183 1184 1185 1186 1187 1188 1189 1190 1191
			continue;
		nfs4_state_mark_reclaim_nograce(clp, state);
		found = true;
	}
	spin_unlock(&inode->i_lock);
	if (found)
		nfs4_schedule_state_manager(clp);
}


1192
static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1193 1194
{
	struct inode *inode = state->inode;
1195
	struct nfs_inode *nfsi = NFS_I(inode);
L
Linus Torvalds 已提交
1196 1197 1198
	struct file_lock *fl;
	int status = 0;

1199 1200 1201 1202
	if (inode->i_flock == NULL)
		return 0;

	/* Guard against delegation returns and new lock/unlock calls */
1203
	down_write(&nfsi->rwsem);
1204
	/* Protect inode->i_flock using the BKL */
1205
	lock_flocks();
H
Harvey Harrison 已提交
1206
	for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1207
		if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
L
Linus Torvalds 已提交
1208
			continue;
1209
		if (nfs_file_open_context(fl->fl_file)->state != state)
L
Linus Torvalds 已提交
1210
			continue;
1211
		unlock_flocks();
L
Linus Torvalds 已提交
1212 1213
		status = ops->recover_lock(state, fl);
		switch (status) {
1214 1215 1216 1217 1218 1219 1220 1221 1222
			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:
1223 1224 1225 1226
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1227
				goto out;
L
Linus Torvalds 已提交
1228
			default:
1229 1230
				printk(KERN_ERR "NFS: %s: unhandled error %d. "
					"Zeroing state\n", __func__, status);
1231 1232
			case -ENOMEM:
			case -NFS4ERR_DENIED:
L
Linus Torvalds 已提交
1233 1234
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1235
				/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1236
				status = 0;
L
Linus Torvalds 已提交
1237
		}
1238
		lock_flocks();
L
Linus Torvalds 已提交
1239
	}
1240
	unlock_flocks();
1241
out:
1242
	up_write(&nfsi->rwsem);
L
Linus Torvalds 已提交
1243 1244 1245
	return status;
}

1246
static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259
{
	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.
	 */
1260 1261
restart:
	spin_lock(&sp->so_lock);
L
Linus Torvalds 已提交
1262
	list_for_each_entry(state, &sp->so_states, open_states) {
1263 1264
		if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
			continue;
L
Linus Torvalds 已提交
1265 1266
		if (state->state == 0)
			continue;
1267 1268
		atomic_inc(&state->count);
		spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1269 1270
		status = ops->recover_open(sp, state);
		if (status >= 0) {
1271 1272
			status = nfs4_reclaim_locks(state, ops);
			if (status >= 0) {
1273
				spin_lock(&state->state_lock);
1274 1275
				list_for_each_entry(lock, &state->lock_states, ls_locks) {
					if (!(lock->ls_flags & NFS_LOCK_INITIALIZED))
1276 1277
						pr_warn_ratelimited("NFS: "
							"%s: Lock reclaim "
1278
							"failed!\n", __func__);
1279
				}
1280
				spin_unlock(&state->state_lock);
1281 1282
				nfs4_put_open_state(state);
				goto restart;
L
Linus Torvalds 已提交
1283 1284 1285 1286
			}
		}
		switch (status) {
			default:
1287 1288
				printk(KERN_ERR "NFS: %s: unhandled error %d. "
					"Zeroing state\n", __func__, status);
L
Linus Torvalds 已提交
1289
			case -ENOENT:
1290
			case -ENOMEM:
1291
			case -ESTALE:
L
Linus Torvalds 已提交
1292 1293 1294 1295
				/*
				 * Open state on this file cannot be recovered
				 * All we can do is revert to using the zero stateid.
				 */
1296 1297
				memset(&state->stateid, 0,
					sizeof(state->stateid));
L
Linus Torvalds 已提交
1298 1299 1300
				/* Mark the file as being 'closed' */
				state->state = 0;
				break;
1301 1302 1303 1304 1305 1306 1307 1308
			case -EKEYEXPIRED:
				/*
				 * User RPCSEC_GSS context has expired.
				 * We cannot recover this stateid now, so
				 * skip it and allow recovery thread to
				 * proceed.
				 */
				break;
1309 1310 1311
			case -NFS4ERR_ADMIN_REVOKED:
			case -NFS4ERR_STALE_STATEID:
			case -NFS4ERR_BAD_STATEID:
1312 1313
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1314
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1315
				break;
L
Linus Torvalds 已提交
1316 1317
			case -NFS4ERR_EXPIRED:
			case -NFS4ERR_NO_GRACE:
1318
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
L
Linus Torvalds 已提交
1319
			case -NFS4ERR_STALE_CLIENTID:
1320 1321 1322 1323
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
L
Linus Torvalds 已提交
1324 1325
				goto out_err;
		}
1326 1327
		nfs4_put_open_state(state);
		goto restart;
L
Linus Torvalds 已提交
1328
	}
1329
	spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1330 1331
	return 0;
out_err:
1332
	nfs4_put_open_state(state);
L
Linus Torvalds 已提交
1333 1334 1335
	return status;
}

1336 1337 1338 1339 1340 1341 1342 1343
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);
1344
	spin_lock(&state->state_lock);
1345 1346 1347 1348
	list_for_each_entry(lock, &state->lock_states, ls_locks) {
		lock->ls_seqid.flags = 0;
		lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
	}
1349
	spin_unlock(&state->state_lock);
1350 1351
}

1352 1353
static void nfs4_reset_seqids(struct nfs_server *server,
	int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1354
{
1355
	struct nfs_client *clp = server->nfs_client;
1356
	struct nfs4_state_owner *sp;
1357
	struct rb_node *pos;
1358 1359
	struct nfs4_state *state;

1360 1361 1362 1363 1364
	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);
1365
		sp->so_seqid.flags = 0;
1366
		spin_lock(&sp->so_lock);
1367
		list_for_each_entry(state, &sp->so_states, open_states) {
1368 1369
			if (mark_reclaim(clp, state))
				nfs4_clear_open_state(state);
1370
		}
1371
		spin_unlock(&sp->so_lock);
1372
	}
1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384
	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();
1385 1386
}

1387 1388 1389 1390 1391 1392 1393
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);
}

1394 1395 1396 1397 1398 1399 1400 1401
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);
}

1402
static void nfs4_clear_reclaim_server(struct nfs_server *server)
1403
{
1404
	struct nfs_client *clp = server->nfs_client;
1405 1406 1407 1408
	struct nfs4_state_owner *sp;
	struct rb_node *pos;
	struct nfs4_state *state;

1409 1410 1411 1412 1413
	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);
1414 1415
		spin_lock(&sp->so_lock);
		list_for_each_entry(state, &sp->so_states, open_states) {
1416 1417
			if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
						&state->flags))
1418 1419 1420 1421 1422
				continue;
			nfs4_state_mark_reclaim_nograce(clp, state);
		}
		spin_unlock(&sp->so_lock);
	}
1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436
	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();
1437 1438

	nfs_delegation_reap_unclaimed(clp);
1439 1440 1441 1442 1443 1444 1445 1446
	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);
1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460
}

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);
}

1461 1462 1463 1464 1465 1466 1467 1468
static void nfs4_warn_keyexpired(const char *s)
{
	printk_ratelimited(KERN_WARNING "Error: state manager"
			" encountered RPCSEC_GSS session"
			" expired against NFSv4 server %s.\n",
			s);
}

1469
static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1470 1471
{
	switch (error) {
1472 1473
		case 0:
			break;
1474
		case -NFS4ERR_CB_PATH_DOWN:
1475
			nfs40_handle_cb_pathdown(clp);
1476
			break;
1477 1478
		case -NFS4ERR_NO_GRACE:
			nfs4_state_end_reclaim_reboot(clp);
1479
			break;
1480 1481 1482
		case -NFS4ERR_STALE_CLIENTID:
		case -NFS4ERR_LEASE_MOVED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1483
			nfs4_state_clear_reclaim_reboot(clp);
1484 1485 1486 1487 1488
			nfs4_state_start_reclaim_reboot(clp);
			break;
		case -NFS4ERR_EXPIRED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
			nfs4_state_start_reclaim_nograce(clp);
1489
			break;
1490 1491 1492 1493 1494 1495 1496
		case -NFS4ERR_BADSESSION:
		case -NFS4ERR_BADSLOT:
		case -NFS4ERR_BAD_HIGH_SLOT:
		case -NFS4ERR_DEADSESSION:
		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
		case -NFS4ERR_SEQ_FALSE_RETRY:
		case -NFS4ERR_SEQ_MISORDERED:
1497
			set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1498
			/* Zero session reset errors */
1499
			break;
1500 1501 1502
		case -EKEYEXPIRED:
			/* Nothing we can do */
			nfs4_warn_keyexpired(clp->cl_hostname);
1503 1504 1505
			break;
		default:
			return error;
1506
	}
1507
	return 0;
1508 1509
}

1510
static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1511
{
1512 1513
	struct nfs4_state_owner *sp;
	struct nfs_server *server;
1514
	struct rb_node *pos;
L
Linus Torvalds 已提交
1515 1516
	int status = 0;

1517
restart:
1518 1519
	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1520
		nfs4_purge_state_owners(server);
1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540
		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);
			}

1541
			nfs4_put_state_owner(sp);
1542
			goto restart;
1543
		}
1544
		spin_unlock(&clp->cl_lock);
1545
	}
1546
	rcu_read_unlock();
1547 1548 1549 1550 1551 1552
	return status;
}

static int nfs4_check_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1553 1554
	const struct nfs4_state_maintenance_ops *ops =
		clp->cl_mvops->state_renewal_ops;
1555
	int status;
L
Linus Torvalds 已提交
1556

1557 1558 1559
	/* Is the client already known to have an expired lease? */
	if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		return 0;
1560 1561 1562
	spin_lock(&clp->cl_lock);
	cred = ops->get_state_renewal_cred_locked(clp);
	spin_unlock(&clp->cl_lock);
1563 1564
	if (cred == NULL) {
		cred = nfs4_get_setclientid_cred(clp);
1565
		status = -ENOKEY;
1566 1567
		if (cred == NULL)
			goto out;
1568
	}
1569
	status = ops->renew_lease(clp, cred);
1570 1571
	put_rpccred(cred);
out:
1572
	return nfs4_recovery_handle_error(clp, status);
1573 1574 1575 1576 1577
}

static int nfs4_reclaim_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1578 1579
	const struct nfs4_state_recovery_ops *ops =
		clp->cl_mvops->reboot_recovery_ops;
1580 1581
	int status = -ENOENT;

1582
	cred = ops->get_clid_cred(clp);
1583
	if (cred != NULL) {
1584
		status = ops->establish_clid(clp, cred);
1585
		put_rpccred(cred);
1586 1587 1588
		/* Handle case where the user hasn't set up machine creds */
		if (status == -EACCES && cred == clp->cl_machine_cred) {
			nfs4_clear_machine_cred(clp);
1589
			status = -EAGAIN;
1590
		}
1591 1592
		if (status == -NFS4ERR_MINOR_VERS_MISMATCH)
			status = -EPROTONOSUPPORT;
1593
	}
1594 1595 1596
	return status;
}

1597
#ifdef CONFIG_NFS_V4_1
1598 1599
void nfs4_schedule_session_recovery(struct nfs4_session *session)
{
1600 1601 1602 1603
	struct nfs_client *clp = session->clp;

	set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
	nfs4_schedule_lease_recovery(clp);
1604
}
1605
EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
1606

A
Andy Adamson 已提交
1607 1608 1609
void nfs41_handle_recall_slot(struct nfs_client *clp)
{
	set_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
1610
	nfs4_schedule_state_manager(clp);
A
Andy Adamson 已提交
1611 1612
}

1613 1614 1615 1616 1617
static void nfs4_reset_all_state(struct nfs_client *clp)
{
	if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
		clp->cl_boot_time = CURRENT_TIME;
		nfs4_state_start_reclaim_nograce(clp);
1618
		nfs4_schedule_state_manager(clp);
1619 1620 1621 1622 1623 1624 1625
	}
}

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);
1626
		nfs4_schedule_state_manager(clp);
1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645
	}
}

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

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

static void nfs41_handle_cb_path_down(struct nfs_client *clp)
{
	nfs_expire_all_delegations(clp);
	if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
1646
		nfs4_schedule_state_manager(clp);
1647 1648
}

1649 1650 1651 1652
void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
{
	if (!flags)
		return;
1653
	if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
1654
		nfs41_handle_server_reboot(clp);
1655
	if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
1656 1657
			    SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
			    SEQ4_STATUS_ADMIN_STATE_REVOKED |
1658 1659
			    SEQ4_STATUS_LEASE_MOVED))
		nfs41_handle_state_revoked(clp);
1660
	if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
1661
		nfs41_handle_recallable_state_revoked(clp);
1662
	if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
1663 1664
			    SEQ4_STATUS_BACKCHANNEL_FAULT |
			    SEQ4_STATUS_CB_PATH_DOWN_SESSION))
1665
		nfs41_handle_cb_path_down(clp);
1666 1667
}

1668 1669 1670 1671
static int nfs4_reset_session(struct nfs_client *clp)
{
	int status;

1672
	nfs4_begin_drain_session(clp);
1673 1674 1675
	status = nfs4_proc_destroy_session(clp->cl_session);
	if (status && status != -NFS4ERR_BADSESSION &&
	    status != -NFS4ERR_DEADSESSION) {
1676
		status = nfs4_recovery_handle_error(clp, status);
1677 1678 1679 1680
		goto out;
	}

	memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
1681
	status = nfs4_proc_create_session(clp);
1682
	if (status) {
1683
		status = nfs4_recovery_handle_error(clp, status);
1684 1685
		goto out;
	}
1686
	clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1687 1688
	/* create_session negotiated new slot table */
	clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
1689

1690 1691
	 /* Let the state manager reestablish state */
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1692
		nfs41_setup_state_renewal(clp);
1693
out:
1694 1695
	return status;
}
1696

A
Andy Adamson 已提交
1697 1698 1699 1700 1701 1702 1703 1704 1705
static int nfs4_recall_slot(struct nfs_client *clp)
{
	struct nfs4_slot_table *fc_tbl = &clp->cl_session->fc_slot_table;
	struct nfs4_channel_attrs *fc_attrs = &clp->cl_session->fc_attrs;
	struct nfs4_slot *new, *old;
	int i;

	nfs4_begin_drain_session(clp);
	new = kmalloc(fc_tbl->target_max_slots * sizeof(struct nfs4_slot),
1706
		      GFP_NOFS);
A
Andy Adamson 已提交
1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724
        if (!new)
		return -ENOMEM;

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

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

1725
#else /* CONFIG_NFS_V4_1 */
1726
static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
1727
static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
A
Andy Adamson 已提交
1728
static int nfs4_recall_slot(struct nfs_client *clp) { return 0; }
1729 1730
#endif /* CONFIG_NFS_V4_1 */

1731 1732 1733 1734 1735
/* Set NFS4CLNT_LEASE_EXPIRED for all v4.0 errors and for recoverable errors
 * on EXCHANGE_ID for v4.1
 */
static void nfs4_set_lease_expired(struct nfs_client *clp, int status)
{
1736 1737 1738 1739 1740 1741
	switch (status) {
	case -NFS4ERR_CLID_INUSE:
	case -NFS4ERR_STALE_CLIENTID:
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
		break;
	case -NFS4ERR_DELAY:
1742
	case -ETIMEDOUT:
1743 1744 1745
	case -EAGAIN:
		ssleep(1);
		break;
1746

1747 1748 1749 1750 1751 1752
	case -EKEYEXPIRED:
		nfs4_warn_keyexpired(clp->cl_hostname);
	case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
				 * in nfs4_exchange_id */
	default:
		return;
1753 1754 1755 1756
	}
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
}

1757
static void nfs4_state_manager(struct nfs_client *clp)
1758 1759 1760 1761
{
	int status = 0;

	/* Ensure exclusive access to NFSv4 state */
1762
	do {
1763 1764 1765 1766
		if (test_and_clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
			/* We're going to have to re-establish a clientid */
			status = nfs4_reclaim_lease(clp);
			if (status) {
1767
				nfs4_set_lease_expired(clp, status);
1768 1769
				if (test_bit(NFS4CLNT_LEASE_EXPIRED,
							&clp->cl_state))
1770
					continue;
1771 1772 1773
				if (clp->cl_cons_state ==
							NFS_CS_SESSION_INITING)
					nfs_mark_client_ready(clp, status);
1774 1775
				goto out_error;
			}
1776
			clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1777 1778 1779 1780 1781 1782 1783 1784

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

1785
			pnfs_destroy_all_layouts(clp);
1786 1787 1788 1789
		}

		if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
			status = nfs4_check_lease(clp);
1790 1791
			if (status < 0)
				goto out_error;
1792
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1793
				continue;
1794
		}
1795

1796
		/* Initialize or reset the session */
1797
		if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)
1798
		   && nfs4_has_session(clp)) {
1799
			status = nfs4_reset_session(clp);
1800 1801 1802
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
				continue;
			if (status < 0)
1803 1804
				goto out_error;
		}
1805

1806
		/* First recover reboot state... */
1807
		if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
1808
			status = nfs4_do_reclaim(clp,
1809
				clp->cl_mvops->reboot_recovery_ops);
1810
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1811
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
1812
				continue;
1813
			nfs4_state_end_reclaim_reboot(clp);
1814 1815 1816 1817
			if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
				continue;
			if (status < 0)
				goto out_error;
1818 1819
		}

1820 1821
		/* Now recover expired state... */
		if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
1822
			status = nfs4_do_reclaim(clp,
1823
				clp->cl_mvops->nograce_recovery_ops);
1824
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1825
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
1826 1827 1828
			    test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
				continue;
			if (status < 0)
1829
				goto out_error;
L
Linus Torvalds 已提交
1830
		}
1831

1832
		nfs4_end_drain_session(clp);
1833 1834 1835 1836
		if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
			nfs_client_return_marked_delegations(clp);
			continue;
		}
A
Andy Adamson 已提交
1837 1838 1839 1840 1841 1842 1843 1844 1845
		/* Recall session slots */
		if (test_and_clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state)
		   && nfs4_has_session(clp)) {
			status = nfs4_recall_slot(clp);
			if (status < 0)
				goto out_error;
			continue;
		}

1846 1847

		nfs4_clear_state_manager_bit(clp);
1848 1849 1850 1851 1852
		/* 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;
1853
	} while (atomic_read(&clp->cl_count) > 1);
1854
	return;
L
Linus Torvalds 已提交
1855
out_error:
1856
	pr_warn_ratelimited("NFS: state manager failed on NFSv4 server %s"
1857
			" with error %d\n", clp->cl_hostname, -status);
1858
	nfs4_end_drain_session(clp);
1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870
	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
1871 1872 1873 1874 1875 1876 1877
}

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