nfs4state.c 37.4 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 43 44 45
#include <linux/slab.h>
#include <linux/smp_lock.h>
#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>
L
Linus Torvalds 已提交
49 50 51
#include <linux/workqueue.h>
#include <linux/bitops.h>

52
#include "nfs4_fs.h"
L
Linus Torvalds 已提交
53 54
#include "callback.h"
#include "delegation.h"
55
#include "internal.h"
L
Linus Torvalds 已提交
56 57 58

#define OPENOWNER_POOL_SIZE	8

59
const nfs4_stateid zero_stateid;
L
Linus Torvalds 已提交
60 61 62

static LIST_HEAD(nfs4_clientid_list);

63
int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
L
Linus Torvalds 已提交
64
{
65 66 67 68 69 70 71 72
	unsigned short port;
	int status;

	port = nfs_callback_tcpport;
	if (clp->cl_addr.ss_family == AF_INET6)
		port = nfs_callback_tcpport6;

	status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred);
L
Linus Torvalds 已提交
73
	if (status == 0)
74
		status = nfs4_proc_setclientid_confirm(clp, cred);
L
Linus Torvalds 已提交
75 76 77 78 79
	if (status == 0)
		nfs4_schedule_state_renewal(clp);
	return status;
}

80
struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100
{
	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);
}

101
struct rpc_cred *nfs4_get_renew_cred_locked(struct nfs_client *clp)
102 103
{
	struct nfs4_state_owner *sp;
104
	struct rb_node *pos;
105 106
	struct rpc_cred *cred = NULL;

107 108
	for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
109 110 111 112 113 114 115 116
		if (list_empty(&sp->so_states))
			continue;
		cred = get_rpccred(sp->so_cred);
		break;
	}
	return cred;
}

117 118
#if defined(CONFIG_NFS_V4_1)

R
Ricardo Labiaga 已提交
119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137
static int nfs41_setup_state_renewal(struct nfs_client *clp)
{
	int status;
	struct nfs_fsinfo fsinfo;

	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;
}

138 139 140
static void nfs41_end_drain_session(struct nfs_client *clp,
		struct nfs4_session *ses)
{
141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156
	int max_slots;

	if (test_and_clear_bit(NFS4CLNT_SESSION_DRAINING, &clp->cl_state)) {
		spin_lock(&ses->fc_slot_table.slot_tbl_lock);
		max_slots = ses->fc_slot_table.max_slots;
		while (max_slots--) {
			struct rpc_task *task;

			task = rpc_wake_up_next(&ses->fc_slot_table.
						slot_tbl_waitq);
			if (!task)
				break;
			rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
		}
		spin_unlock(&ses->fc_slot_table.slot_tbl_lock);
	}
157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174
}

static int nfs41_begin_drain_session(struct nfs_client *clp,
		struct nfs4_session *ses)
{
	struct nfs4_slot_table *tbl = &ses->fc_slot_table;

	spin_lock(&tbl->slot_tbl_lock);
	set_bit(NFS4CLNT_SESSION_DRAINING, &clp->cl_state);
	if (tbl->highest_used_slotid != -1) {
		INIT_COMPLETION(ses->complete);
		spin_unlock(&tbl->slot_tbl_lock);
		return wait_for_completion_interruptible(&ses->complete);
	}
	spin_unlock(&tbl->slot_tbl_lock);
	return 0;
}

175 176 177 178
int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
{
	int status;

179 180 181
	status = nfs41_begin_drain_session(clp, clp->cl_session);
	if (status != 0)
		goto out;
182
	status = nfs4_proc_exchange_id(clp, cred);
R
Ricardo Labiaga 已提交
183 184 185 186 187
	if (status != 0)
		goto out;
	status = nfs4_proc_create_session(clp);
	if (status != 0)
		goto out;
188
	nfs41_end_drain_session(clp, clp->cl_session);
R
Ricardo Labiaga 已提交
189 190 191
	nfs41_setup_state_renewal(clp);
	nfs_mark_client_ready(clp, NFS_CS_READY);
out:
192 193 194
	return status;
}

195 196 197 198 199 200 201 202 203 204 205 206
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 */

207
struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
208 209
{
	struct nfs4_state_owner *sp;
210
	struct rb_node *pos;
211
	struct rpc_cred *cred;
212

213 214
	spin_lock(&clp->cl_lock);
	cred = nfs4_get_machine_cred_locked(clp);
215 216
	if (cred != NULL)
		goto out;
217 218 219
	pos = rb_first(&clp->cl_state_owners);
	if (pos != NULL) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
220
		cred = get_rpccred(sp->so_cred);
221
	}
222
out:
223
	spin_unlock(&clp->cl_lock);
224
	return cred;
225 226
}

227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281
static void nfs_alloc_unique_id(struct rb_root *root, struct nfs_unique_id *new,
		__u64 minval, int maxbits)
{
	struct rb_node **p, *parent;
	struct nfs_unique_id *pos;
	__u64 mask = ~0ULL;

	if (maxbits < 64)
		mask = (1ULL << maxbits) - 1ULL;

	/* Ensure distribution is more or less flat */
	get_random_bytes(&new->id, sizeof(new->id));
	new->id &= mask;
	if (new->id < minval)
		new->id += minval;
retry:
	p = &root->rb_node;
	parent = NULL;

	while (*p != NULL) {
		parent = *p;
		pos = rb_entry(parent, struct nfs_unique_id, rb_node);

		if (new->id < pos->id)
			p = &(*p)->rb_left;
		else if (new->id > pos->id)
			p = &(*p)->rb_right;
		else
			goto id_exists;
	}
	rb_link_node(&new->rb_node, parent, p);
	rb_insert_color(&new->rb_node, root);
	return;
id_exists:
	for (;;) {
		new->id++;
		if (new->id < minval || (new->id & mask) != new->id) {
			new->id = minval;
			break;
		}
		parent = rb_next(parent);
		if (parent == NULL)
			break;
		pos = rb_entry(parent, struct nfs_unique_id, rb_node);
		if (new->id < pos->id)
			break;
	}
	goto retry;
}

static void nfs_free_unique_id(struct rb_root *root, struct nfs_unique_id *id)
{
	rb_erase(&id->rb_node, root);
}

L
Linus Torvalds 已提交
282
static struct nfs4_state_owner *
283
nfs4_find_state_owner(struct nfs_server *server, struct rpc_cred *cred)
L
Linus Torvalds 已提交
284
{
285
	struct nfs_client *clp = server->nfs_client;
286 287
	struct rb_node **p = &clp->cl_state_owners.rb_node,
		       *parent = NULL;
L
Linus Torvalds 已提交
288 289
	struct nfs4_state_owner *sp, *res = NULL;

290 291 292 293
	while (*p != NULL) {
		parent = *p;
		sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);

294 295 296 297 298 299 300 301
		if (server < sp->so_server) {
			p = &parent->rb_left;
			continue;
		}
		if (server > sp->so_server) {
			p = &parent->rb_right;
			continue;
		}
302 303 304 305 306 307 308 309 310
		if (cred < sp->so_cred)
			p = &parent->rb_left;
		else if (cred > sp->so_cred)
			p = &parent->rb_right;
		else {
			atomic_inc(&sp->so_count);
			res = sp;
			break;
		}
L
Linus Torvalds 已提交
311 312 313 314
	}
	return res;
}

315 316 317 318 319 320 321 322 323 324 325
static struct nfs4_state_owner *
nfs4_insert_state_owner(struct nfs_client *clp, struct nfs4_state_owner *new)
{
	struct rb_node **p = &clp->cl_state_owners.rb_node,
		       *parent = NULL;
	struct nfs4_state_owner *sp;

	while (*p != NULL) {
		parent = *p;
		sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);

326 327 328 329 330 331 332 333
		if (new->so_server < sp->so_server) {
			p = &parent->rb_left;
			continue;
		}
		if (new->so_server > sp->so_server) {
			p = &parent->rb_right;
			continue;
		}
334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356
		if (new->so_cred < sp->so_cred)
			p = &parent->rb_left;
		else if (new->so_cred > sp->so_cred)
			p = &parent->rb_right;
		else {
			atomic_inc(&sp->so_count);
			return sp;
		}
	}
	nfs_alloc_unique_id(&clp->cl_openowner_id, &new->so_owner_id, 1, 64);
	rb_link_node(&new->so_client_node, parent, p);
	rb_insert_color(&new->so_client_node, &clp->cl_state_owners);
	return new;
}

static void
nfs4_remove_state_owner(struct nfs_client *clp, struct nfs4_state_owner *sp)
{
	if (!RB_EMPTY_NODE(&sp->so_client_node))
		rb_erase(&sp->so_client_node, &clp->cl_state_owners);
	nfs_free_unique_id(&clp->cl_openowner_id, &sp->so_owner_id);
}

L
Linus Torvalds 已提交
357 358 359 360 361 362 363 364 365 366
/*
 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
 * create a new state_owner.
 *
 */
static struct nfs4_state_owner *
nfs4_alloc_state_owner(void)
{
	struct nfs4_state_owner *sp;

367
	sp = kzalloc(sizeof(*sp),GFP_KERNEL);
L
Linus Torvalds 已提交
368 369
	if (!sp)
		return NULL;
370
	spin_lock_init(&sp->so_lock);
L
Linus Torvalds 已提交
371 372
	INIT_LIST_HEAD(&sp->so_states);
	INIT_LIST_HEAD(&sp->so_delegations);
373 374 375 376
	rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue");
	sp->so_seqid.sequence = &sp->so_sequence;
	spin_lock_init(&sp->so_sequence.lock);
	INIT_LIST_HEAD(&sp->so_sequence.list);
L
Linus Torvalds 已提交
377 378 379 380
	atomic_set(&sp->so_count, 1);
	return sp;
}

381
static void
L
Linus Torvalds 已提交
382 383
nfs4_drop_state_owner(struct nfs4_state_owner *sp)
{
384 385 386 387 388 389 390 391
	if (!RB_EMPTY_NODE(&sp->so_client_node)) {
		struct nfs_client *clp = sp->so_client;

		spin_lock(&clp->cl_lock);
		rb_erase(&sp->so_client_node, &clp->cl_state_owners);
		RB_CLEAR_NODE(&sp->so_client_node);
		spin_unlock(&clp->cl_lock);
	}
L
Linus Torvalds 已提交
392 393 394 395
}

struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct rpc_cred *cred)
{
396
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
397 398 399
	struct nfs4_state_owner *sp, *new;

	spin_lock(&clp->cl_lock);
400
	sp = nfs4_find_state_owner(server, cred);
L
Linus Torvalds 已提交
401 402 403
	spin_unlock(&clp->cl_lock);
	if (sp != NULL)
		return sp;
404 405 406 407
	new = nfs4_alloc_state_owner();
	if (new == NULL)
		return NULL;
	new->so_client = clp;
408
	new->so_server = server;
409 410 411 412 413 414
	new->so_cred = cred;
	spin_lock(&clp->cl_lock);
	sp = nfs4_insert_state_owner(clp, new);
	spin_unlock(&clp->cl_lock);
	if (sp == new)
		get_rpccred(cred);
415 416
	else {
		rpc_destroy_wait_queue(&new->so_sequence.wait);
417
		kfree(new);
418
	}
419
	return sp;
L
Linus Torvalds 已提交
420 421 422 423
}

void nfs4_put_state_owner(struct nfs4_state_owner *sp)
{
424
	struct nfs_client *clp = sp->so_client;
L
Linus Torvalds 已提交
425 426 427 428
	struct rpc_cred *cred = sp->so_cred;

	if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
		return;
429
	nfs4_remove_state_owner(clp, sp);
L
Linus Torvalds 已提交
430
	spin_unlock(&clp->cl_lock);
431
	rpc_destroy_wait_queue(&sp->so_sequence.wait);
L
Linus Torvalds 已提交
432 433 434 435 436 437 438 439 440
	put_rpccred(cred);
	kfree(sp);
}

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

441
	state = kzalloc(sizeof(*state), GFP_KERNEL);
L
Linus Torvalds 已提交
442 443 444 445
	if (!state)
		return NULL;
	atomic_set(&state->count, 1);
	INIT_LIST_HEAD(&state->lock_states);
446
	spin_lock_init(&state->state_lock);
447
	seqlock_init(&state->seqlock);
L
Linus Torvalds 已提交
448 449 450
	return state;
}

451
void
452
nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
453
{
454
	if (state->state == fmode)
455 456
		return;
	/* NB! List reordering - see the reclaim code for why.  */
457 458
	if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
		if (fmode & FMODE_WRITE)
459 460 461 462
			list_move(&state->open_states, &state->owner->so_states);
		else
			list_move_tail(&state->open_states, &state->owner->so_states);
	}
463
	state->state = fmode;
464 465
}

L
Linus Torvalds 已提交
466 467 468 469 470 471 472
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) {
473
		if (state->owner != owner)
L
Linus Torvalds 已提交
474
			continue;
475
		if (atomic_inc_not_zero(&state->count))
L
Linus Torvalds 已提交
476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498
			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();
499
	spin_lock(&owner->so_lock);
L
Linus Torvalds 已提交
500 501 502 503 504 505 506 507 508
	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);
		state->inode = igrab(inode);
		spin_unlock(&inode->i_lock);
509 510 511 512
		/* 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
513 514
	} else {
		spin_unlock(&inode->i_lock);
515
		spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
516 517 518 519 520 521 522 523 524 525 526 527
		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;

528
	if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
L
Linus Torvalds 已提交
529
		return;
530
	spin_lock(&inode->i_lock);
531
	list_del(&state->inode_states);
L
Linus Torvalds 已提交
532
	list_del(&state->open_states);
533 534
	spin_unlock(&inode->i_lock);
	spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
535 536 537 538 539 540
	iput(inode);
	nfs4_free_open_state(state);
	nfs4_put_state_owner(owner);
}

/*
541
 * Close the current file.
L
Linus Torvalds 已提交
542
 */
543
static void __nfs4_close(struct path *path, struct nfs4_state *state, fmode_t fmode, int wait)
L
Linus Torvalds 已提交
544 545
{
	struct nfs4_state_owner *owner = state->owner;
546
	int call_close = 0;
547
	fmode_t newstate;
L
Linus Torvalds 已提交
548 549 550

	atomic_inc(&owner->so_count);
	/* Protect against nfs4_find_state() */
551
	spin_lock(&owner->so_lock);
552
	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
553 554 555 556 557 558 559 560 561
		case FMODE_READ:
			state->n_rdonly--;
			break;
		case FMODE_WRITE:
			state->n_wronly--;
			break;
		case FMODE_READ|FMODE_WRITE:
			state->n_rdwr--;
	}
562
	newstate = FMODE_READ|FMODE_WRITE;
563
	if (state->n_rdwr == 0) {
564
		if (state->n_rdonly == 0) {
565
			newstate &= ~FMODE_READ;
566 567 568 569
			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) {
570
			newstate &= ~FMODE_WRITE;
571 572 573 574 575
			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);
576
	}
577
	nfs4_state_set_mode_locked(state, newstate);
578
	spin_unlock(&owner->so_lock);
579

580
	if (!call_close) {
581 582 583
		nfs4_put_open_state(state);
		nfs4_put_state_owner(owner);
	} else
584 585 586
		nfs4_do_close(path, state, wait);
}

587
void nfs4_close_state(struct path *path, struct nfs4_state *state, fmode_t fmode)
588
{
589
	__nfs4_close(path, state, fmode, 0);
590 591
}

592
void nfs4_close_sync(struct path *path, struct nfs4_state *state, fmode_t fmode)
593
{
594
	__nfs4_close(path, state, fmode, 1);
L
Linus Torvalds 已提交
595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621
}

/*
 * Search the state->lock_states for an existing lock_owner
 * that is compatible with current->files
 */
static struct nfs4_lock_state *
__nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
{
	struct nfs4_lock_state *pos;
	list_for_each_entry(pos, &state->lock_states, ls_locks) {
		if (pos->ls_owner != fl_owner)
			continue;
		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.
 *
 */
static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
{
	struct nfs4_lock_state *lsp;
622
	struct nfs_client *clp = state->owner->so_client;
L
Linus Torvalds 已提交
623

624
	lsp = kzalloc(sizeof(*lsp), GFP_KERNEL);
L
Linus Torvalds 已提交
625 626
	if (lsp == NULL)
		return NULL;
627 628 629 630
	rpc_init_wait_queue(&lsp->ls_sequence.wait, "lock_seqid_waitqueue");
	spin_lock_init(&lsp->ls_sequence.lock);
	INIT_LIST_HEAD(&lsp->ls_sequence.list);
	lsp->ls_seqid.sequence = &lsp->ls_sequence;
L
Linus Torvalds 已提交
631
	atomic_set(&lsp->ls_count, 1);
632
	lsp->ls_state = state;
L
Linus Torvalds 已提交
633 634
	lsp->ls_owner = fl_owner;
	spin_lock(&clp->cl_lock);
635
	nfs_alloc_unique_id(&clp->cl_lockowner_id, &lsp->ls_id, 1, 64);
L
Linus Torvalds 已提交
636
	spin_unlock(&clp->cl_lock);
637
	INIT_LIST_HEAD(&lsp->ls_locks);
L
Linus Torvalds 已提交
638 639 640
	return lsp;
}

641 642 643 644 645 646 647
static void nfs4_free_lock_state(struct nfs4_lock_state *lsp)
{
	struct nfs_client *clp = lsp->ls_state->owner->so_client;

	spin_lock(&clp->cl_lock);
	nfs_free_unique_id(&clp->cl_lockowner_id, &lsp->ls_id);
	spin_unlock(&clp->cl_lock);
648
	rpc_destroy_wait_queue(&lsp->ls_sequence.wait);
649 650 651
	kfree(lsp);
}

L
Linus Torvalds 已提交
652 653 654 655 656
/*
 * Return a compatible lock_state. If no initialized lock_state structure
 * exists, return an uninitialized one.
 *
 */
657
static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
L
Linus Torvalds 已提交
658
{
659
	struct nfs4_lock_state *lsp, *new = NULL;
L
Linus Torvalds 已提交
660
	
661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678
	for(;;) {
		spin_lock(&state->state_lock);
		lsp = __nfs4_find_lock_state(state, owner);
		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);
		new = nfs4_alloc_lock_state(state, owner);
		if (new == NULL)
			return NULL;
	}
	spin_unlock(&state->state_lock);
679 680
	if (new != NULL)
		nfs4_free_lock_state(new);
L
Linus Torvalds 已提交
681 682 683 684
	return lsp;
}

/*
685 686
 * Release reference to lock_state, and free it if we see that
 * it is no longer in use
L
Linus Torvalds 已提交
687
 */
688
void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
L
Linus Torvalds 已提交
689
{
690
	struct nfs4_state *state;
L
Linus Torvalds 已提交
691

692 693 694 695 696 697 698 699 700
	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);
701
	nfs4_free_lock_state(lsp);
L
Linus Torvalds 已提交
702 703
}

704
static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
L
Linus Torvalds 已提交
705
{
706
	struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
L
Linus Torvalds 已提交
707

708 709 710
	dst->fl_u.nfs4_fl.owner = lsp;
	atomic_inc(&lsp->ls_count);
}
L
Linus Torvalds 已提交
711

712
static void nfs4_fl_release_lock(struct file_lock *fl)
L
Linus Torvalds 已提交
713
{
714
	nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
L
Linus Torvalds 已提交
715 716
}

717
static const struct file_lock_operations nfs4_fl_lock_ops = {
718 719 720 721 722
	.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
723
{
724 725 726 727 728 729 730 731 732 733
	struct nfs4_lock_state *lsp;

	if (fl->fl_ops != NULL)
		return 0;
	lsp = nfs4_get_lock_state(state, fl->fl_owner);
	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
734 735
}

736 737 738
/*
 * Byte-range lock aware utility to initialize the stateid of read/write
 * requests.
L
Linus Torvalds 已提交
739
 */
740
void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner)
L
Linus Torvalds 已提交
741
{
742
	struct nfs4_lock_state *lsp;
743
	int seq;
L
Linus Torvalds 已提交
744

745 746 747 748
	do {
		seq = read_seqbegin(&state->seqlock);
		memcpy(dst, &state->stateid, sizeof(*dst));
	} while (read_seqretry(&state->seqlock, seq));
749 750
	if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
		return;
L
Linus Torvalds 已提交
751

752 753 754 755 756
	spin_lock(&state->state_lock);
	lsp = __nfs4_find_lock_state(state, fl_owner);
	if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
		memcpy(dst, &lsp->ls_stateid, sizeof(*dst));
	spin_unlock(&state->state_lock);
L
Linus Torvalds 已提交
757 758 759
	nfs4_put_lock_state(lsp);
}

760 761 762 763 764 765 766
struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter)
{
	struct nfs_seqid *new;

	new = kmalloc(sizeof(*new), GFP_KERNEL);
	if (new != NULL) {
		new->sequence = counter;
767
		INIT_LIST_HEAD(&new->list);
768 769 770 771 772
	}
	return new;
}

void nfs_free_seqid(struct nfs_seqid *seqid)
L
Linus Torvalds 已提交
773
{
774 775
	if (!list_empty(&seqid->list)) {
		struct rpc_sequence *sequence = seqid->sequence->sequence;
776

777 778 779 780 781
		spin_lock(&sequence->lock);
		list_del(&seqid->list);
		spin_unlock(&sequence->lock);
		rpc_wake_up(&sequence->wait);
	}
782
	kfree(seqid);
L
Linus Torvalds 已提交
783 784 785
}

/*
786 787 788 789
 * 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()
 */
790
static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
791
{
792
	BUG_ON(list_first_entry(&seqid->sequence->sequence->list, struct nfs_seqid, list) != seqid);
793 794 795 796
	switch (status) {
		case 0:
			break;
		case -NFS4ERR_BAD_SEQID:
797 798 799
			if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
				return;
			printk(KERN_WARNING "NFS: v4 server returned a bad"
D
Dan Muntz 已提交
800 801
					" sequence-id error on an"
					" unconfirmed sequence %p!\n",
802
					seqid->sequence);
803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820
		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)
{
821 822 823 824 825
	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
826
		nfs4_drop_state_owner(sp);
827 828
	if (!nfs4_has_session(server->nfs_client))
		nfs_increment_seqid(status, seqid);
829 830 831 832 833 834 835 836 837
}

/*
 * 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)
{
838
	nfs_increment_seqid(status, seqid);
839 840 841 842 843 844 845 846
}

int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
{
	struct rpc_sequence *sequence = seqid->sequence->sequence;
	int status = 0;

	spin_lock(&sequence->lock);
847 848 849 850
	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;
851
	rpc_sleep_on(&sequence->wait, task, NULL);
852 853
	status = -EAGAIN;
unlock:
854 855
	spin_unlock(&sequence->lock);
	return status;
L
Linus Torvalds 已提交
856 857
}

858
static int nfs4_run_state_manager(void *);
L
Linus Torvalds 已提交
859

860
static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
T
Trond Myklebust 已提交
861 862
{
	smp_mb__before_clear_bit();
863
	clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
T
Trond Myklebust 已提交
864
	smp_mb__after_clear_bit();
865
	wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
T
Trond Myklebust 已提交
866 867 868
	rpc_wake_up(&clp->cl_rpcwaitq);
}

L
Linus Torvalds 已提交
869
/*
870
 * Schedule the nfs_client asynchronous state management routine
L
Linus Torvalds 已提交
871
 */
872
void nfs4_schedule_state_manager(struct nfs_client *clp)
L
Linus Torvalds 已提交
873
{
874
	struct task_struct *task;
L
Linus Torvalds 已提交
875

876 877
	if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
		return;
878 879
	__module_get(THIS_MODULE);
	atomic_inc(&clp->cl_count);
880
	task = kthread_run(nfs4_run_state_manager, clp, "%s-manager",
881 882
				rpc_peeraddr2str(clp->cl_rpcclient,
							RPC_DISPLAY_ADDR));
883 884
	if (!IS_ERR(task))
		return;
885
	nfs4_clear_state_manager_bit(clp);
886
	nfs_put_client(clp);
887
	module_put(THIS_MODULE);
L
Linus Torvalds 已提交
888 889 890 891 892
}

/*
 * Schedule a state recovery attempt
 */
893
void nfs4_schedule_state_recovery(struct nfs_client *clp)
L
Linus Torvalds 已提交
894 895 896
{
	if (!clp)
		return;
897 898
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
899
	nfs4_schedule_state_manager(clp);
L
Linus Torvalds 已提交
900 901
}

902 903 904 905 906 907 908 909 910
static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
{

	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;
	}
911
	set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
912 913 914 915
	set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
	return 1;
}

916
int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
917 918 919
{
	set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
	clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
920
	set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
921 922 923 924
	set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
	return 1;
}

925
static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
926 927
{
	struct inode *inode = state->inode;
928
	struct nfs_inode *nfsi = NFS_I(inode);
L
Linus Torvalds 已提交
929 930 931
	struct file_lock *fl;
	int status = 0;

932 933 934 935
	if (inode->i_flock == NULL)
		return 0;

	/* Guard against delegation returns and new lock/unlock calls */
936
	down_write(&nfsi->rwsem);
937 938
	/* Protect inode->i_flock using the BKL */
	lock_kernel();
H
Harvey Harrison 已提交
939
	for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
940
		if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
L
Linus Torvalds 已提交
941
			continue;
942
		if (nfs_file_open_context(fl->fl_file)->state != state)
L
Linus Torvalds 已提交
943
			continue;
944
		unlock_kernel();
L
Linus Torvalds 已提交
945 946
		status = ops->recover_lock(state, fl);
		switch (status) {
947 948 949 950 951 952 953 954 955
			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:
956 957 958 959
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
960
				goto out;
L
Linus Torvalds 已提交
961 962
			default:
				printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
963
						__func__, status);
964 965
			case -ENOMEM:
			case -NFS4ERR_DENIED:
L
Linus Torvalds 已提交
966 967
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
968
				/* kill_proc(fl->fl_pid, SIGLOST, 1); */
969
				status = 0;
L
Linus Torvalds 已提交
970
		}
971
		lock_kernel();
L
Linus Torvalds 已提交
972
	}
973
	unlock_kernel();
974
out:
975
	up_write(&nfsi->rwsem);
L
Linus Torvalds 已提交
976 977 978
	return status;
}

979
static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
980 981 982 983 984 985 986 987 988 989 990 991 992
{
	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.
	 */
993 994
restart:
	spin_lock(&sp->so_lock);
L
Linus Torvalds 已提交
995
	list_for_each_entry(state, &sp->so_states, open_states) {
996 997
		if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
			continue;
L
Linus Torvalds 已提交
998 999
		if (state->state == 0)
			continue;
1000 1001
		atomic_inc(&state->count);
		spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1002 1003
		status = ops->recover_open(sp, state);
		if (status >= 0) {
1004 1005 1006 1007 1008
			status = nfs4_reclaim_locks(state, ops);
			if (status >= 0) {
				list_for_each_entry(lock, &state->lock_states, ls_locks) {
					if (!(lock->ls_flags & NFS_LOCK_INITIALIZED))
						printk("%s: Lock reclaim failed!\n",
1009
							__func__);
1010
				}
1011 1012
				nfs4_put_open_state(state);
				goto restart;
L
Linus Torvalds 已提交
1013 1014 1015 1016 1017
			}
		}
		switch (status) {
			default:
				printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
1018
						__func__, status);
L
Linus Torvalds 已提交
1019
			case -ENOENT:
1020
			case -ENOMEM:
1021
			case -ESTALE:
L
Linus Torvalds 已提交
1022 1023 1024 1025 1026 1027 1028 1029 1030
				/*
				 * Open state on this file cannot be recovered
				 * All we can do is revert to using the zero stateid.
				 */
				memset(state->stateid.data, 0,
					sizeof(state->stateid.data));
				/* Mark the file as being 'closed' */
				state->state = 0;
				break;
1031 1032 1033
			case -NFS4ERR_ADMIN_REVOKED:
			case -NFS4ERR_STALE_STATEID:
			case -NFS4ERR_BAD_STATEID:
1034 1035 1036 1037
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
				nfs4_state_mark_reclaim_nograce(sp->so_client, state);
				break;
L
Linus Torvalds 已提交
1038 1039
			case -NFS4ERR_EXPIRED:
			case -NFS4ERR_NO_GRACE:
1040
				nfs4_state_mark_reclaim_nograce(sp->so_client, state);
L
Linus Torvalds 已提交
1041
			case -NFS4ERR_STALE_CLIENTID:
1042 1043 1044 1045
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
L
Linus Torvalds 已提交
1046 1047
				goto out_err;
		}
1048 1049
		nfs4_put_open_state(state);
		goto restart;
L
Linus Torvalds 已提交
1050
	}
1051
	spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1052 1053
	return 0;
out_err:
1054
	nfs4_put_open_state(state);
L
Linus Torvalds 已提交
1055 1056 1057
	return status;
}

1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072
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);
	list_for_each_entry(lock, &state->lock_states, ls_locks) {
		lock->ls_seqid.flags = 0;
		lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
	}
}

static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp, int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1073 1074
{
	struct nfs4_state_owner *sp;
1075
	struct rb_node *pos;
1076 1077 1078
	struct nfs4_state *state;

	/* Reset all sequence ids to zero */
1079 1080
	for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1081
		sp->so_seqid.flags = 0;
1082
		spin_lock(&sp->so_lock);
1083
		list_for_each_entry(state, &sp->so_states, open_states) {
1084 1085
			if (mark_reclaim(clp, state))
				nfs4_clear_open_state(state);
1086
		}
1087
		spin_unlock(&sp->so_lock);
1088 1089 1090
	}
}

1091 1092 1093 1094 1095 1096 1097
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);
}

1098 1099 1100 1101 1102 1103 1104 1105
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);
}

1106 1107 1108 1109 1110 1111 1112 1113 1114
static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
{
	struct nfs4_state_owner *sp;
	struct rb_node *pos;
	struct nfs4_state *state;

	if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
		return;

1115 1116 1117
	nfs4_reclaim_complete(clp,
		nfs4_reboot_recovery_ops[clp->cl_minorversion]);

1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143
	for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
		spin_lock(&sp->so_lock);
		list_for_each_entry(state, &sp->so_states, open_states) {
			if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags))
				continue;
			nfs4_state_mark_reclaim_nograce(clp, state);
		}
		spin_unlock(&sp->so_lock);
	}

	nfs_delegation_reap_unclaimed(clp);
}

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);
}

1144
static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1145 1146 1147
{
	switch (error) {
		case -NFS4ERR_CB_PATH_DOWN:
1148
			nfs_handle_cb_pathdown(clp);
1149
			return 0;
1150 1151 1152
		case -NFS4ERR_NO_GRACE:
			nfs4_state_end_reclaim_reboot(clp);
			return 0;
1153 1154 1155
		case -NFS4ERR_STALE_CLIENTID:
		case -NFS4ERR_LEASE_MOVED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1156
			nfs4_state_end_reclaim_reboot(clp);
1157 1158 1159 1160 1161
			nfs4_state_start_reclaim_reboot(clp);
			break;
		case -NFS4ERR_EXPIRED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
			nfs4_state_start_reclaim_nograce(clp);
1162
			break;
1163 1164 1165 1166 1167 1168 1169
		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:
1170
			set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1171 1172
			/* Zero session reset errors */
			return 0;
1173
	}
1174
	return error;
1175 1176
}

1177
static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1178
{
1179
	struct rb_node *pos;
L
Linus Torvalds 已提交
1180 1181
	int status = 0;

1182 1183
restart:
	spin_lock(&clp->cl_lock);
1184 1185
	for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
		struct nfs4_state_owner *sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
1186 1187 1188 1189
		if (!test_and_clear_bit(ops->owner_flag_bit, &sp->so_flags))
			continue;
		atomic_inc(&sp->so_count);
		spin_unlock(&clp->cl_lock);
1190
		status = nfs4_reclaim_open_state(sp, ops);
1191 1192 1193
		if (status < 0) {
			set_bit(ops->owner_flag_bit, &sp->so_flags);
			nfs4_put_state_owner(sp);
1194
			return nfs4_recovery_handle_error(clp, status);
1195 1196 1197
		}
		nfs4_put_state_owner(sp);
		goto restart;
1198
	}
1199
	spin_unlock(&clp->cl_lock);
1200 1201 1202 1203 1204 1205
	return status;
}

static int nfs4_check_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1206 1207
	struct nfs4_state_maintenance_ops *ops =
		nfs4_state_renewal_ops[clp->cl_minorversion];
1208
	int status = -NFS4ERR_EXPIRED;
L
Linus Torvalds 已提交
1209

1210 1211 1212
	/* Is the client already known to have an expired lease? */
	if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		return 0;
1213 1214 1215
	spin_lock(&clp->cl_lock);
	cred = ops->get_state_renewal_cred_locked(clp);
	spin_unlock(&clp->cl_lock);
1216 1217 1218 1219
	if (cred == NULL) {
		cred = nfs4_get_setclientid_cred(clp);
		if (cred == NULL)
			goto out;
1220
	}
1221
	status = ops->renew_lease(clp, cred);
1222 1223
	put_rpccred(cred);
out:
1224
	return nfs4_recovery_handle_error(clp, status);
1225 1226 1227 1228 1229
}

static int nfs4_reclaim_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1230 1231
	struct nfs4_state_recovery_ops *ops =
		nfs4_reboot_recovery_ops[clp->cl_minorversion];
1232 1233
	int status = -ENOENT;

1234
	cred = ops->get_clid_cred(clp);
1235
	if (cred != NULL) {
1236
		status = ops->establish_clid(clp, cred);
1237
		put_rpccred(cred);
1238 1239 1240
		/* Handle case where the user hasn't set up machine creds */
		if (status == -EACCES && cred == clp->cl_machine_cred) {
			nfs4_clear_machine_cred(clp);
1241
			status = -EAGAIN;
1242
		}
1243 1244
		if (status == -NFS4ERR_MINOR_VERS_MISMATCH)
			status = -EPROTONOSUPPORT;
1245
	}
1246 1247 1248
	return status;
}

1249
#ifdef CONFIG_NFS_V4_1
1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271
void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
{
	if (!flags)
		return;
	else if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED) {
		set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
		nfs4_state_start_reclaim_reboot(clp);
		nfs4_schedule_state_recovery(clp);
	} else if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
			    SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
			    SEQ4_STATUS_ADMIN_STATE_REVOKED |
			    SEQ4_STATUS_RECALLABLE_STATE_REVOKED |
			    SEQ4_STATUS_LEASE_MOVED)) {
		set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
		nfs4_state_start_reclaim_nograce(clp);
		nfs4_schedule_state_recovery(clp);
	} else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
			    SEQ4_STATUS_BACKCHANNEL_FAULT |
			    SEQ4_STATUS_CB_PATH_DOWN_SESSION))
		nfs_expire_all_delegations(clp);
}

1272 1273
static int nfs4_reset_session(struct nfs_client *clp)
{
1274
	struct nfs4_session *ses = clp->cl_session;
1275 1276
	int status;

1277 1278 1279
	status = nfs41_begin_drain_session(clp, ses);
	if (status != 0)
		return status;
1280

1281 1282 1283
	status = nfs4_proc_destroy_session(clp->cl_session);
	if (status && status != -NFS4ERR_BADSESSION &&
	    status != -NFS4ERR_DEADSESSION) {
1284
		status = nfs4_recovery_handle_error(clp, status);
1285 1286 1287 1288
		goto out;
	}

	memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
1289
	status = nfs4_proc_create_session(clp);
1290
	if (status)
1291
		status = nfs4_recovery_handle_error(clp, status);
1292

1293
out:
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
	/*
	 * Let the state manager reestablish state
	 * without waking other tasks yet.
	 */
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
		/* Wake up the next rpc task */
		nfs41_end_drain_session(clp, ses);
		if (status == 0)
			nfs41_setup_state_renewal(clp);
	}
1304 1305
	return status;
}
1306 1307

#else /* CONFIG_NFS_V4_1 */
1308
static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
1309 1310
#endif /* CONFIG_NFS_V4_1 */

1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331
/* 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)
{
	if (nfs4_has_session(clp)) {
		switch (status) {
		case -NFS4ERR_DELAY:
		case -NFS4ERR_CLID_INUSE:
		case -EAGAIN:
			break;

		case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
					 * in nfs4_exchange_id */
		default:
			return;
		}
	}
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
}

1332
static void nfs4_state_manager(struct nfs_client *clp)
1333 1334 1335 1336
{
	int status = 0;

	/* Ensure exclusive access to NFSv4 state */
1337
	for(;;) {
1338 1339 1340 1341
		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) {
1342
				nfs4_set_lease_expired(clp, status);
1343 1344
				if (test_bit(NFS4CLNT_LEASE_EXPIRED,
							&clp->cl_state))
1345
					continue;
1346 1347 1348
				if (clp->cl_cons_state ==
							NFS_CS_SESSION_INITING)
					nfs_mark_client_ready(clp, status);
1349 1350
				goto out_error;
			}
1351
			clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1352
			set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1353 1354 1355 1356
		}

		if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
			status = nfs4_check_lease(clp);
1357
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1358
				continue;
1359 1360
			if (status < 0 && status != -NFS4ERR_CB_PATH_DOWN)
				goto out_error;
1361
		}
1362

1363
		/* Initialize or reset the session */
1364
		if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)
1365
		   && nfs4_has_session(clp)) {
1366
			status = nfs4_reset_session(clp);
1367 1368 1369
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
				continue;
			if (status < 0)
1370 1371
				goto out_error;
		}
1372

1373
		/* First recover reboot state... */
1374
		if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
1375 1376
			status = nfs4_do_reclaim(clp,
				nfs4_reboot_recovery_ops[clp->cl_minorversion]);
1377
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1378
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
1379
				continue;
1380
			nfs4_state_end_reclaim_reboot(clp);
1381 1382 1383 1384
			if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
				continue;
			if (status < 0)
				goto out_error;
1385 1386
		}

1387 1388
		/* Now recover expired state... */
		if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
1389 1390
			status = nfs4_do_reclaim(clp,
				nfs4_nograce_recovery_ops[clp->cl_minorversion]);
1391
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1392
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
1393 1394 1395
			    test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
				continue;
			if (status < 0)
1396
				goto out_error;
L
Linus Torvalds 已提交
1397
		}
1398 1399 1400 1401 1402

		if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
			nfs_client_return_marked_delegations(clp);
			continue;
		}
1403 1404

		nfs4_clear_state_manager_bit(clp);
1405 1406 1407 1408 1409
		/* 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;
L
Linus Torvalds 已提交
1410
	}
1411
	return;
L
Linus Torvalds 已提交
1412
out_error:
1413
	printk(KERN_WARNING "Error: state manager failed on NFSv4 server %s"
1414
			" with error %d\n", clp->cl_hostname, -status);
1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426
	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
1427 1428 1429 1430 1431 1432 1433
}

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