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
static void nfs4_end_drain_session(struct nfs_client *clp)
139
{
140
	struct nfs4_session *ses = clp->cl_session;
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
static int nfs4_begin_drain_session(struct nfs_client *clp)
160
{
161
	struct nfs4_session *ses = clp->cl_session;
162 163 164 165 166 167 168 169 170 171 172 173 174
	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
	nfs4_begin_drain_session(clp);
180
	status = nfs4_proc_exchange_id(clp, cred);
R
Ricardo Labiaga 已提交
181 182 183 184 185 186 187 188
	if (status != 0)
		goto out;
	status = nfs4_proc_create_session(clp);
	if (status != 0)
		goto out;
	nfs41_setup_state_renewal(clp);
	nfs_mark_client_ready(clp, NFS_CS_READY);
out:
189 190 191
	return status;
}

192 193 194 195 196 197 198 199 200 201 202 203
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 */

204
struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
205 206
{
	struct nfs4_state_owner *sp;
207
	struct rb_node *pos;
208
	struct rpc_cred *cred;
209

210 211
	spin_lock(&clp->cl_lock);
	cred = nfs4_get_machine_cred_locked(clp);
212 213
	if (cred != NULL)
		goto out;
214 215 216
	pos = rb_first(&clp->cl_state_owners);
	if (pos != NULL) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
217
		cred = get_rpccred(sp->so_cred);
218
	}
219
out:
220
	spin_unlock(&clp->cl_lock);
221
	return cred;
222 223
}

224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278
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
279
static struct nfs4_state_owner *
280
nfs4_find_state_owner(struct nfs_server *server, struct rpc_cred *cred)
L
Linus Torvalds 已提交
281
{
282
	struct nfs_client *clp = server->nfs_client;
283 284
	struct rb_node **p = &clp->cl_state_owners.rb_node,
		       *parent = NULL;
L
Linus Torvalds 已提交
285 286
	struct nfs4_state_owner *sp, *res = NULL;

287 288 289 290
	while (*p != NULL) {
		parent = *p;
		sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);

291 292 293 294 295 296 297 298
		if (server < sp->so_server) {
			p = &parent->rb_left;
			continue;
		}
		if (server > sp->so_server) {
			p = &parent->rb_right;
			continue;
		}
299 300 301 302 303 304 305 306 307
		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
308 309 310 311
	}
	return res;
}

312 313 314 315 316 317 318 319 320 321 322
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);

323 324 325 326 327 328 329 330
		if (new->so_server < sp->so_server) {
			p = &parent->rb_left;
			continue;
		}
		if (new->so_server > sp->so_server) {
			p = &parent->rb_right;
			continue;
		}
331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353
		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
354 355 356 357 358 359 360 361 362 363
/*
 * 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;

364
	sp = kzalloc(sizeof(*sp),GFP_KERNEL);
L
Linus Torvalds 已提交
365 366
	if (!sp)
		return NULL;
367
	spin_lock_init(&sp->so_lock);
L
Linus Torvalds 已提交
368 369
	INIT_LIST_HEAD(&sp->so_states);
	INIT_LIST_HEAD(&sp->so_delegations);
370 371 372 373
	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
374 375 376 377
	atomic_set(&sp->so_count, 1);
	return sp;
}

378
static void
L
Linus Torvalds 已提交
379 380
nfs4_drop_state_owner(struct nfs4_state_owner *sp)
{
381 382 383 384 385 386 387 388
	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
389 390 391 392
}

struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct rpc_cred *cred)
{
393
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
394 395 396
	struct nfs4_state_owner *sp, *new;

	spin_lock(&clp->cl_lock);
397
	sp = nfs4_find_state_owner(server, cred);
L
Linus Torvalds 已提交
398 399 400
	spin_unlock(&clp->cl_lock);
	if (sp != NULL)
		return sp;
401 402 403 404
	new = nfs4_alloc_state_owner();
	if (new == NULL)
		return NULL;
	new->so_client = clp;
405
	new->so_server = server;
406 407 408 409 410 411
	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);
412 413
	else {
		rpc_destroy_wait_queue(&new->so_sequence.wait);
414
		kfree(new);
415
	}
416
	return sp;
L
Linus Torvalds 已提交
417 418 419 420
}

void nfs4_put_state_owner(struct nfs4_state_owner *sp)
{
421
	struct nfs_client *clp = sp->so_client;
L
Linus Torvalds 已提交
422 423 424 425
	struct rpc_cred *cred = sp->so_cred;

	if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
		return;
426
	nfs4_remove_state_owner(clp, sp);
L
Linus Torvalds 已提交
427
	spin_unlock(&clp->cl_lock);
428
	rpc_destroy_wait_queue(&sp->so_sequence.wait);
L
Linus Torvalds 已提交
429 430 431 432 433 434 435 436 437
	put_rpccred(cred);
	kfree(sp);
}

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

438
	state = kzalloc(sizeof(*state), GFP_KERNEL);
L
Linus Torvalds 已提交
439 440 441 442
	if (!state)
		return NULL;
	atomic_set(&state->count, 1);
	INIT_LIST_HEAD(&state->lock_states);
443
	spin_lock_init(&state->state_lock);
444
	seqlock_init(&state->seqlock);
L
Linus Torvalds 已提交
445 446 447
	return state;
}

448
void
449
nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
450
{
451
	if (state->state == fmode)
452 453
		return;
	/* NB! List reordering - see the reclaim code for why.  */
454 455
	if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
		if (fmode & FMODE_WRITE)
456 457 458 459
			list_move(&state->open_states, &state->owner->so_states);
		else
			list_move_tail(&state->open_states, &state->owner->so_states);
	}
460
	state->state = fmode;
461 462
}

L
Linus Torvalds 已提交
463 464 465 466 467 468 469
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) {
470
		if (state->owner != owner)
L
Linus Torvalds 已提交
471
			continue;
472
		if (atomic_inc_not_zero(&state->count))
L
Linus Torvalds 已提交
473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495
			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();
496
	spin_lock(&owner->so_lock);
L
Linus Torvalds 已提交
497 498 499 500 501 502 503 504 505
	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);
506 507 508 509
		/* 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
510 511
	} else {
		spin_unlock(&inode->i_lock);
512
		spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
513 514 515 516 517 518 519 520 521 522 523 524
		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;

525
	if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
L
Linus Torvalds 已提交
526
		return;
527
	spin_lock(&inode->i_lock);
528
	list_del(&state->inode_states);
L
Linus Torvalds 已提交
529
	list_del(&state->open_states);
530 531
	spin_unlock(&inode->i_lock);
	spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
532 533 534 535 536 537
	iput(inode);
	nfs4_free_open_state(state);
	nfs4_put_state_owner(owner);
}

/*
538
 * Close the current file.
L
Linus Torvalds 已提交
539
 */
540
static void __nfs4_close(struct path *path, struct nfs4_state *state, fmode_t fmode, int wait)
L
Linus Torvalds 已提交
541 542
{
	struct nfs4_state_owner *owner = state->owner;
543
	int call_close = 0;
544
	fmode_t newstate;
L
Linus Torvalds 已提交
545 546 547

	atomic_inc(&owner->so_count);
	/* Protect against nfs4_find_state() */
548
	spin_lock(&owner->so_lock);
549
	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
550 551 552 553 554 555 556 557 558
		case FMODE_READ:
			state->n_rdonly--;
			break;
		case FMODE_WRITE:
			state->n_wronly--;
			break;
		case FMODE_READ|FMODE_WRITE:
			state->n_rdwr--;
	}
559
	newstate = FMODE_READ|FMODE_WRITE;
560
	if (state->n_rdwr == 0) {
561
		if (state->n_rdonly == 0) {
562
			newstate &= ~FMODE_READ;
563 564 565 566
			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) {
567
			newstate &= ~FMODE_WRITE;
568 569 570 571 572
			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);
573
	}
574
	nfs4_state_set_mode_locked(state, newstate);
575
	spin_unlock(&owner->so_lock);
576

577
	if (!call_close) {
578 579 580
		nfs4_put_open_state(state);
		nfs4_put_state_owner(owner);
	} else
581 582 583
		nfs4_do_close(path, state, wait);
}

584
void nfs4_close_state(struct path *path, struct nfs4_state *state, fmode_t fmode)
585
{
586
	__nfs4_close(path, state, fmode, 0);
587 588
}

589
void nfs4_close_sync(struct path *path, struct nfs4_state *state, fmode_t fmode)
590
{
591
	__nfs4_close(path, state, fmode, 1);
L
Linus Torvalds 已提交
592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618
}

/*
 * 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;
619
	struct nfs_client *clp = state->owner->so_client;
L
Linus Torvalds 已提交
620

621
	lsp = kzalloc(sizeof(*lsp), GFP_KERNEL);
L
Linus Torvalds 已提交
622 623
	if (lsp == NULL)
		return NULL;
624 625 626 627
	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
628
	atomic_set(&lsp->ls_count, 1);
629
	lsp->ls_state = state;
L
Linus Torvalds 已提交
630 631
	lsp->ls_owner = fl_owner;
	spin_lock(&clp->cl_lock);
632
	nfs_alloc_unique_id(&clp->cl_lockowner_id, &lsp->ls_id, 1, 64);
L
Linus Torvalds 已提交
633
	spin_unlock(&clp->cl_lock);
634
	INIT_LIST_HEAD(&lsp->ls_locks);
L
Linus Torvalds 已提交
635 636 637
	return lsp;
}

638 639 640 641 642 643 644
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);
645
	rpc_destroy_wait_queue(&lsp->ls_sequence.wait);
646 647 648
	kfree(lsp);
}

L
Linus Torvalds 已提交
649 650 651 652 653
/*
 * Return a compatible lock_state. If no initialized lock_state structure
 * exists, return an uninitialized one.
 *
 */
654
static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
L
Linus Torvalds 已提交
655
{
656
	struct nfs4_lock_state *lsp, *new = NULL;
L
Linus Torvalds 已提交
657
	
658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675
	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);
676 677
	if (new != NULL)
		nfs4_free_lock_state(new);
L
Linus Torvalds 已提交
678 679 680 681
	return lsp;
}

/*
682 683
 * Release reference to lock_state, and free it if we see that
 * it is no longer in use
L
Linus Torvalds 已提交
684
 */
685
void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
L
Linus Torvalds 已提交
686
{
687
	struct nfs4_state *state;
L
Linus Torvalds 已提交
688

689 690 691 692 693 694 695 696 697
	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);
698
	nfs4_free_lock_state(lsp);
L
Linus Torvalds 已提交
699 700
}

701
static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
L
Linus Torvalds 已提交
702
{
703
	struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
L
Linus Torvalds 已提交
704

705 706 707
	dst->fl_u.nfs4_fl.owner = lsp;
	atomic_inc(&lsp->ls_count);
}
L
Linus Torvalds 已提交
708

709
static void nfs4_fl_release_lock(struct file_lock *fl)
L
Linus Torvalds 已提交
710
{
711
	nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
L
Linus Torvalds 已提交
712 713
}

714
static const struct file_lock_operations nfs4_fl_lock_ops = {
715 716 717 718 719
	.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
720
{
721 722 723 724 725 726 727 728 729 730
	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
731 732
}

733 734 735
/*
 * Byte-range lock aware utility to initialize the stateid of read/write
 * requests.
L
Linus Torvalds 已提交
736
 */
737
void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner)
L
Linus Torvalds 已提交
738
{
739
	struct nfs4_lock_state *lsp;
740
	int seq;
L
Linus Torvalds 已提交
741

742 743 744 745
	do {
		seq = read_seqbegin(&state->seqlock);
		memcpy(dst, &state->stateid, sizeof(*dst));
	} while (read_seqretry(&state->seqlock, seq));
746 747
	if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
		return;
L
Linus Torvalds 已提交
748

749 750 751 752 753
	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
754 755 756
	nfs4_put_lock_state(lsp);
}

757 758 759 760 761 762 763
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;
764
		INIT_LIST_HEAD(&new->list);
765 766 767 768
	}
	return new;
}

769
void nfs_release_seqid(struct nfs_seqid *seqid)
L
Linus Torvalds 已提交
770
{
771 772
	if (!list_empty(&seqid->list)) {
		struct rpc_sequence *sequence = seqid->sequence->sequence;
773

774
		spin_lock(&sequence->lock);
775
		list_del_init(&seqid->list);
776 777 778
		spin_unlock(&sequence->lock);
		rpc_wake_up(&sequence->wait);
	}
779 780 781 782 783
}

void nfs_free_seqid(struct nfs_seqid *seqid)
{
	nfs_release_seqid(seqid);
784
	kfree(seqid);
L
Linus Torvalds 已提交
785 786 787
}

/*
788 789 790 791
 * 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()
 */
792
static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
793
{
794
	BUG_ON(list_first_entry(&seqid->sequence->sequence->list, struct nfs_seqid, list) != seqid);
795 796 797 798
	switch (status) {
		case 0:
			break;
		case -NFS4ERR_BAD_SEQID:
799 800 801
			if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
				return;
			printk(KERN_WARNING "NFS: v4 server returned a bad"
D
Dan Muntz 已提交
802 803
					" sequence-id error on an"
					" unconfirmed sequence %p!\n",
804
					seqid->sequence);
805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822
		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)
{
823 824 825 826 827
	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
828
		nfs4_drop_state_owner(sp);
829 830
	if (!nfs4_has_session(server->nfs_client))
		nfs_increment_seqid(status, seqid);
831 832 833 834 835 836 837 838 839
}

/*
 * 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)
{
840
	nfs_increment_seqid(status, seqid);
841 842 843 844 845 846 847 848
}

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);
849 850 851 852
	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;
853
	rpc_sleep_on(&sequence->wait, task, NULL);
854 855
	status = -EAGAIN;
unlock:
856 857
	spin_unlock(&sequence->lock);
	return status;
L
Linus Torvalds 已提交
858 859
}

860
static int nfs4_run_state_manager(void *);
L
Linus Torvalds 已提交
861

862
static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
T
Trond Myklebust 已提交
863 864
{
	smp_mb__before_clear_bit();
865
	clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
T
Trond Myklebust 已提交
866
	smp_mb__after_clear_bit();
867
	wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
T
Trond Myklebust 已提交
868 869 870
	rpc_wake_up(&clp->cl_rpcwaitq);
}

L
Linus Torvalds 已提交
871
/*
872
 * Schedule the nfs_client asynchronous state management routine
L
Linus Torvalds 已提交
873
 */
874
void nfs4_schedule_state_manager(struct nfs_client *clp)
L
Linus Torvalds 已提交
875
{
876
	struct task_struct *task;
L
Linus Torvalds 已提交
877

878 879
	if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
		return;
880 881
	__module_get(THIS_MODULE);
	atomic_inc(&clp->cl_count);
882
	task = kthread_run(nfs4_run_state_manager, clp, "%s-manager",
883 884
				rpc_peeraddr2str(clp->cl_rpcclient,
							RPC_DISPLAY_ADDR));
885 886
	if (!IS_ERR(task))
		return;
887
	nfs4_clear_state_manager_bit(clp);
888
	nfs_put_client(clp);
889
	module_put(THIS_MODULE);
L
Linus Torvalds 已提交
890 891 892 893 894
}

/*
 * Schedule a state recovery attempt
 */
895
void nfs4_schedule_state_recovery(struct nfs_client *clp)
L
Linus Torvalds 已提交
896 897 898
{
	if (!clp)
		return;
899 900
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
901
	nfs4_schedule_state_manager(clp);
L
Linus Torvalds 已提交
902 903
}

904
int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
905 906 907 908 909 910 911 912
{

	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;
	}
913
	set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
914 915 916 917
	set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
	return 1;
}

918
int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
919 920 921
{
	set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
	clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
922
	set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
923 924 925 926
	set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
	return 1;
}

927
static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
928 929
{
	struct inode *inode = state->inode;
930
	struct nfs_inode *nfsi = NFS_I(inode);
L
Linus Torvalds 已提交
931 932 933
	struct file_lock *fl;
	int status = 0;

934 935 936 937
	if (inode->i_flock == NULL)
		return 0;

	/* Guard against delegation returns and new lock/unlock calls */
938
	down_write(&nfsi->rwsem);
939 940
	/* Protect inode->i_flock using the BKL */
	lock_kernel();
H
Harvey Harrison 已提交
941
	for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
942
		if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
L
Linus Torvalds 已提交
943
			continue;
944
		if (nfs_file_open_context(fl->fl_file)->state != state)
L
Linus Torvalds 已提交
945
			continue;
946
		unlock_kernel();
L
Linus Torvalds 已提交
947 948
		status = ops->recover_lock(state, fl);
		switch (status) {
949 950 951 952 953 954 955 956 957
			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:
958 959 960 961
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
962
				goto out;
L
Linus Torvalds 已提交
963 964
			default:
				printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
965
						__func__, status);
966 967
			case -ENOMEM:
			case -NFS4ERR_DENIED:
L
Linus Torvalds 已提交
968 969
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
970
				/* kill_proc(fl->fl_pid, SIGLOST, 1); */
971
				status = 0;
L
Linus Torvalds 已提交
972
		}
973
		lock_kernel();
L
Linus Torvalds 已提交
974
	}
975
	unlock_kernel();
976
out:
977
	up_write(&nfsi->rwsem);
L
Linus Torvalds 已提交
978 979 980
	return status;
}

981
static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
982 983 984 985 986 987 988 989 990 991 992 993 994
{
	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.
	 */
995 996
restart:
	spin_lock(&sp->so_lock);
L
Linus Torvalds 已提交
997
	list_for_each_entry(state, &sp->so_states, open_states) {
998 999
		if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
			continue;
L
Linus Torvalds 已提交
1000 1001
		if (state->state == 0)
			continue;
1002 1003
		atomic_inc(&state->count);
		spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1004 1005
		status = ops->recover_open(sp, state);
		if (status >= 0) {
1006 1007 1008 1009 1010
			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",
1011
							__func__);
1012
				}
1013 1014
				nfs4_put_open_state(state);
				goto restart;
L
Linus Torvalds 已提交
1015 1016 1017 1018 1019
			}
		}
		switch (status) {
			default:
				printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
1020
						__func__, status);
L
Linus Torvalds 已提交
1021
			case -ENOENT:
1022
			case -ENOMEM:
1023
			case -ESTALE:
L
Linus Torvalds 已提交
1024 1025 1026 1027 1028 1029 1030 1031 1032
				/*
				 * 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;
1033 1034 1035
			case -NFS4ERR_ADMIN_REVOKED:
			case -NFS4ERR_STALE_STATEID:
			case -NFS4ERR_BAD_STATEID:
1036 1037 1038 1039
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
				nfs4_state_mark_reclaim_nograce(sp->so_client, state);
				break;
L
Linus Torvalds 已提交
1040 1041
			case -NFS4ERR_EXPIRED:
			case -NFS4ERR_NO_GRACE:
1042
				nfs4_state_mark_reclaim_nograce(sp->so_client, state);
L
Linus Torvalds 已提交
1043
			case -NFS4ERR_STALE_CLIENTID:
1044 1045 1046 1047
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
L
Linus Torvalds 已提交
1048 1049
				goto out_err;
		}
1050 1051
		nfs4_put_open_state(state);
		goto restart;
L
Linus Torvalds 已提交
1052
	}
1053
	spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1054 1055
	return 0;
out_err:
1056
	nfs4_put_open_state(state);
L
Linus Torvalds 已提交
1057 1058 1059
	return status;
}

1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074
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))
1075 1076
{
	struct nfs4_state_owner *sp;
1077
	struct rb_node *pos;
1078 1079 1080
	struct nfs4_state *state;

	/* Reset all sequence ids to zero */
1081 1082
	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);
1083
		sp->so_seqid.flags = 0;
1084
		spin_lock(&sp->so_lock);
1085
		list_for_each_entry(state, &sp->so_states, open_states) {
1086 1087
			if (mark_reclaim(clp, state))
				nfs4_clear_open_state(state);
1088
		}
1089
		spin_unlock(&sp->so_lock);
1090 1091 1092
	}
}

1093 1094 1095 1096 1097 1098 1099
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);
}

1100 1101 1102 1103 1104 1105 1106 1107
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);
}

1108 1109 1110 1111 1112 1113 1114 1115 1116
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;

1117 1118 1119
	nfs4_reclaim_complete(clp,
		nfs4_reboot_recovery_ops[clp->cl_minorversion]);

1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145
	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);
}

1146
static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1147 1148 1149
{
	switch (error) {
		case -NFS4ERR_CB_PATH_DOWN:
1150
			nfs_handle_cb_pathdown(clp);
1151
			return 0;
1152 1153 1154
		case -NFS4ERR_NO_GRACE:
			nfs4_state_end_reclaim_reboot(clp);
			return 0;
1155 1156 1157
		case -NFS4ERR_STALE_CLIENTID:
		case -NFS4ERR_LEASE_MOVED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1158
			nfs4_state_end_reclaim_reboot(clp);
1159 1160 1161 1162 1163
			nfs4_state_start_reclaim_reboot(clp);
			break;
		case -NFS4ERR_EXPIRED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
			nfs4_state_start_reclaim_nograce(clp);
1164
			break;
1165 1166 1167 1168 1169 1170 1171
		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:
1172
			set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1173 1174
			/* Zero session reset errors */
			return 0;
1175
	}
1176
	return error;
1177 1178
}

1179
static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1180
{
1181
	struct rb_node *pos;
L
Linus Torvalds 已提交
1182 1183
	int status = 0;

1184 1185
restart:
	spin_lock(&clp->cl_lock);
1186 1187
	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);
1188 1189 1190 1191
		if (!test_and_clear_bit(ops->owner_flag_bit, &sp->so_flags))
			continue;
		atomic_inc(&sp->so_count);
		spin_unlock(&clp->cl_lock);
1192
		status = nfs4_reclaim_open_state(sp, ops);
1193 1194 1195
		if (status < 0) {
			set_bit(ops->owner_flag_bit, &sp->so_flags);
			nfs4_put_state_owner(sp);
1196
			return nfs4_recovery_handle_error(clp, status);
1197 1198 1199
		}
		nfs4_put_state_owner(sp);
		goto restart;
1200
	}
1201
	spin_unlock(&clp->cl_lock);
1202 1203 1204 1205 1206 1207
	return status;
}

static int nfs4_check_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1208 1209
	struct nfs4_state_maintenance_ops *ops =
		nfs4_state_renewal_ops[clp->cl_minorversion];
1210
	int status = -NFS4ERR_EXPIRED;
L
Linus Torvalds 已提交
1211

1212 1213 1214
	/* Is the client already known to have an expired lease? */
	if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		return 0;
1215 1216 1217
	spin_lock(&clp->cl_lock);
	cred = ops->get_state_renewal_cred_locked(clp);
	spin_unlock(&clp->cl_lock);
1218 1219 1220 1221
	if (cred == NULL) {
		cred = nfs4_get_setclientid_cred(clp);
		if (cred == NULL)
			goto out;
1222
	}
1223
	status = ops->renew_lease(clp, cred);
1224 1225
	put_rpccred(cred);
out:
1226
	return nfs4_recovery_handle_error(clp, status);
1227 1228 1229 1230 1231
}

static int nfs4_reclaim_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1232 1233
	struct nfs4_state_recovery_ops *ops =
		nfs4_reboot_recovery_ops[clp->cl_minorversion];
1234 1235
	int status = -ENOENT;

1236
	cred = ops->get_clid_cred(clp);
1237
	if (cred != NULL) {
1238
		status = ops->establish_clid(clp, cred);
1239
		put_rpccred(cred);
1240 1241 1242
		/* Handle case where the user hasn't set up machine creds */
		if (status == -EACCES && cred == clp->cl_machine_cred) {
			nfs4_clear_machine_cred(clp);
1243
			status = -EAGAIN;
1244
		}
1245 1246
		if (status == -NFS4ERR_MINOR_VERS_MISMATCH)
			status = -EPROTONOSUPPORT;
1247
	}
1248 1249 1250
	return status;
}

1251
#ifdef CONFIG_NFS_V4_1
1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273
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);
}

1274 1275 1276 1277
static int nfs4_reset_session(struct nfs_client *clp)
{
	int status;

1278
	nfs4_begin_drain_session(clp);
1279 1280 1281
	status = nfs4_proc_destroy_session(clp->cl_session);
	if (status && status != -NFS4ERR_BADSESSION &&
	    status != -NFS4ERR_DEADSESSION) {
1282
		status = nfs4_recovery_handle_error(clp, status);
1283 1284 1285 1286
		goto out;
	}

	memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
1287
	status = nfs4_proc_create_session(clp);
1288
	if (status)
1289
		status = nfs4_recovery_handle_error(clp, status);
1290

1291
out:
1292 1293 1294
	/*
	 * Let the state manager reestablish state
	 */
1295 1296 1297 1298
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
	    status == 0)
		nfs41_setup_state_renewal(clp);

1299 1300
	return status;
}
1301 1302

#else /* CONFIG_NFS_V4_1 */
1303
static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
1304
static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
1305 1306
#endif /* CONFIG_NFS_V4_1 */

1307 1308 1309 1310 1311 1312 1313 1314 1315 1316
/* 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:
1317
		case -EKEYEXPIRED:
1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328
			break;

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

1329
static void nfs4_state_manager(struct nfs_client *clp)
1330 1331 1332 1333
{
	int status = 0;

	/* Ensure exclusive access to NFSv4 state */
1334
	for(;;) {
1335 1336 1337 1338
		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) {
1339
				nfs4_set_lease_expired(clp, status);
1340 1341
				if (test_bit(NFS4CLNT_LEASE_EXPIRED,
							&clp->cl_state))
1342
					continue;
1343 1344 1345
				if (clp->cl_cons_state ==
							NFS_CS_SESSION_INITING)
					nfs_mark_client_ready(clp, status);
1346 1347
				goto out_error;
			}
1348
			clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1349
			set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1350 1351 1352 1353
		}

		if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
			status = nfs4_check_lease(clp);
1354
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1355
				continue;
1356 1357
			if (status < 0 && status != -NFS4ERR_CB_PATH_DOWN)
				goto out_error;
1358
		}
1359

1360
		/* Initialize or reset the session */
1361
		if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)
1362
		   && nfs4_has_session(clp)) {
1363
			status = nfs4_reset_session(clp);
1364 1365 1366
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
				continue;
			if (status < 0)
1367 1368
				goto out_error;
		}
1369

1370
		/* First recover reboot state... */
1371
		if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
1372 1373
			status = nfs4_do_reclaim(clp,
				nfs4_reboot_recovery_ops[clp->cl_minorversion]);
1374
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1375
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
1376
				continue;
1377
			nfs4_state_end_reclaim_reboot(clp);
1378 1379 1380 1381
			if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
				continue;
			if (status < 0)
				goto out_error;
1382 1383
		}

1384 1385
		/* Now recover expired state... */
		if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
1386 1387
			status = nfs4_do_reclaim(clp,
				nfs4_nograce_recovery_ops[clp->cl_minorversion]);
1388
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1389
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
1390 1391 1392
			    test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
				continue;
			if (status < 0)
1393
				goto out_error;
L
Linus Torvalds 已提交
1394
		}
1395

1396
		nfs4_end_drain_session(clp);
1397 1398 1399 1400
		if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
			nfs_client_return_marked_delegations(clp);
			continue;
		}
1401 1402

		nfs4_clear_state_manager_bit(clp);
1403 1404 1405 1406 1407
		/* 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
1408
	}
1409
	return;
L
Linus Torvalds 已提交
1410
out_error:
1411
	printk(KERN_WARNING "Error: state manager failed on NFSv4 server %s"
1412
			" with error %d\n", clp->cl_hostname, -status);
1413
	nfs4_end_drain_session(clp);
1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425
	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
1426 1427 1428 1429 1430 1431 1432
}

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