nfs4state.c 44.1 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
/*
 *  fs/nfs/nfs4state.c
 *
 *  Client-side XDR for NFSv4.
 *
 *  Copyright (c) 2002 The Regents of the University of Michigan.
 *  All rights reserved.
 *
 *  Kendrick Smith <kmsmith@umich.edu>
 *
 *  Redistribution and use in source and binary forms, with or without
 *  modification, are permitted provided that the following conditions
 *  are met:
 *
 *  1. Redistributions of source code must retain the above copyright
 *     notice, this list of conditions and the following disclaimer.
 *  2. Redistributions in binary form must reproduce the above copyright
 *     notice, this list of conditions and the following disclaimer in the
 *     documentation and/or other materials provided with the distribution.
 *  3. Neither the name of the University nor the names of its
 *     contributors may be used to endorse or promote products derived
 *     from this software without specific prior written permission.
 *
 *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
 *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
 *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
 *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
 *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
 *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
 *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
 *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * Implementation of the NFSv4 state model.  For the time being,
 * this is minimal, but will be made much more complex in a
 * subsequent patch.
 */

41
#include <linux/kernel.h>
L
Linus Torvalds 已提交
42
#include <linux/slab.h>
43
#include <linux/fs.h>
L
Linus Torvalds 已提交
44 45
#include <linux/nfs_fs.h>
#include <linux/nfs_idmap.h>
46 47
#include <linux/kthread.h>
#include <linux/module.h>
48
#include <linux/random.h>
49
#include <linux/ratelimit.h>
L
Linus Torvalds 已提交
50 51 52
#include <linux/workqueue.h>
#include <linux/bitops.h>

53
#include "nfs4_fs.h"
L
Linus Torvalds 已提交
54 55
#include "callback.h"
#include "delegation.h"
56
#include "internal.h"
57
#include "pnfs.h"
L
Linus Torvalds 已提交
58 59 60

#define OPENOWNER_POOL_SIZE	8

61
const nfs4_stateid zero_stateid;
L
Linus Torvalds 已提交
62 63 64

static LIST_HEAD(nfs4_clientid_list);

65
int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
L
Linus Torvalds 已提交
66
{
67
	struct nfs4_setclientid_res clid;
68 69 70 71 72 73 74
	unsigned short port;
	int status;

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

75 76 77 78 79 80 81 82 83
	status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
	if (status != 0)
		goto out;
	status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
	if (status != 0)
		goto out;
	clp->cl_clientid = clid.clientid;
	nfs4_schedule_state_renewal(clp);
out:
L
Linus Torvalds 已提交
84 85 86
	return status;
}

87
struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107
{
	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);
}

108 109
static struct rpc_cred *
nfs4_get_renew_cred_server_locked(struct nfs_server *server)
110
{
111
	struct rpc_cred *cred = NULL;
112
	struct nfs4_state_owner *sp;
113
	struct rb_node *pos;
114

115 116 117 118
	for (pos = rb_first(&server->state_owners);
	     pos != NULL;
	     pos = rb_next(pos)) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
119 120 121 122 123 124 125 126
		if (list_empty(&sp->so_states))
			continue;
		cred = get_rpccred(sp->so_cred);
		break;
	}
	return cred;
}

127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148
/**
 * nfs4_get_renew_cred_locked - Acquire credential for a renew operation
 * @clp: client state handle
 *
 * Returns an rpc_cred with reference count bumped, or NULL.
 * Caller must hold clp->cl_lock.
 */
struct rpc_cred *nfs4_get_renew_cred_locked(struct nfs_client *clp)
{
	struct rpc_cred *cred = NULL;
	struct nfs_server *server;

	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
		cred = nfs4_get_renew_cred_server_locked(server);
		if (cred != NULL)
			break;
	}
	rcu_read_unlock();
	return cred;
}

149 150
#if defined(CONFIG_NFS_V4_1)

R
Ricardo Labiaga 已提交
151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169
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;
}

170 171 172 173 174
/*
 * Back channel returns NFS4ERR_DELAY for new requests when
 * NFS4_SESSION_DRAINING is set so there is no work to be done when draining
 * is ended.
 */
175
static void nfs4_end_drain_session(struct nfs_client *clp)
176
{
177
	struct nfs4_session *ses = clp->cl_session;
178 179
	int max_slots;

180 181 182
	if (ses == NULL)
		return;
	if (test_and_clear_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
183 184 185 186 187 188 189 190 191 192 193 194 195
		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);
	}
196 197
}

198
static int nfs4_wait_on_slot_tbl(struct nfs4_slot_table *tbl)
199 200 201
{
	spin_lock(&tbl->slot_tbl_lock);
	if (tbl->highest_used_slotid != -1) {
202
		INIT_COMPLETION(tbl->complete);
203
		spin_unlock(&tbl->slot_tbl_lock);
204
		return wait_for_completion_interruptible(&tbl->complete);
205 206 207 208 209
	}
	spin_unlock(&tbl->slot_tbl_lock);
	return 0;
}

210 211 212 213 214 215 216 217 218 219 220 221 222 223
static int nfs4_begin_drain_session(struct nfs_client *clp)
{
	struct nfs4_session *ses = clp->cl_session;
	int ret = 0;

	set_bit(NFS4_SESSION_DRAINING, &ses->session_state);
	/* back channel */
	ret = nfs4_wait_on_slot_tbl(&ses->bc_slot_table);
	if (ret)
		return ret;
	/* fore channel */
	return nfs4_wait_on_slot_tbl(&ses->fc_slot_table);
}

224 225 226 227
int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
{
	int status;

228
	nfs4_begin_drain_session(clp);
229
	status = nfs4_proc_exchange_id(clp, cred);
R
Ricardo Labiaga 已提交
230 231 232 233 234 235 236 237
	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:
238 239 240
	return status;
}

241 242 243 244 245 246 247 248 249 250 251 252
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 */

253 254
static struct rpc_cred *
nfs4_get_setclientid_cred_server(struct nfs_server *server)
255
{
256 257
	struct nfs_client *clp = server->nfs_client;
	struct rpc_cred *cred = NULL;
258
	struct nfs4_state_owner *sp;
259
	struct rb_node *pos;
260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279

	spin_lock(&clp->cl_lock);
	pos = rb_first(&server->state_owners);
	if (pos != NULL) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
		cred = get_rpccred(sp->so_cred);
	}
	spin_unlock(&clp->cl_lock);
	return cred;
}

/**
 * nfs4_get_setclientid_cred - Acquire credential for a setclientid operation
 * @clp: client state handle
 *
 * Returns an rpc_cred with reference count bumped, or NULL.
 */
struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
{
	struct nfs_server *server;
280
	struct rpc_cred *cred;
281

282 283
	spin_lock(&clp->cl_lock);
	cred = nfs4_get_machine_cred_locked(clp);
284
	spin_unlock(&clp->cl_lock);
285 286
	if (cred != NULL)
		goto out;
287 288 289 290 291 292

	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
		cred = nfs4_get_setclientid_cred_server(server);
		if (cred != NULL)
			break;
293
	}
294 295
	rcu_read_unlock();

296 297
out:
	return cred;
298 299
}

300 301 302
static void nfs_alloc_unique_id_locked(struct rb_root *root,
				       struct nfs_unique_id *new,
				       __u64 minval, int maxbits)
303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
{
	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
356
static struct nfs4_state_owner *
357
nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
L
Linus Torvalds 已提交
358
{
359
	struct rb_node **p = &server->state_owners.rb_node,
360
		       *parent = NULL;
L
Linus Torvalds 已提交
361 362
	struct nfs4_state_owner *sp, *res = NULL;

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

367 368 369 370 371 372 373 374
		if (server < sp->so_server) {
			p = &parent->rb_left;
			continue;
		}
		if (server > sp->so_server) {
			p = &parent->rb_right;
			continue;
		}
375 376 377 378 379 380 381 382 383
		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
384 385 386 387
	}
	return res;
}

388
static struct nfs4_state_owner *
389
nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
390
{
391 392
	struct nfs_server *server = new->so_server;
	struct rb_node **p = &server->state_owners.rb_node,
393 394 395 396 397
		       *parent = NULL;
	struct nfs4_state_owner *sp;

	while (*p != NULL) {
		parent = *p;
398
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
399 400 401 402 403 404 405 406 407 408

		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;
		}
	}
409 410 411 412
	nfs_alloc_unique_id_locked(&server->openowner_id,
					&new->so_owner_id, 1, 64);
	rb_link_node(&new->so_server_node, parent, p);
	rb_insert_color(&new->so_server_node, &server->state_owners);
413 414 415 416
	return new;
}

static void
417
nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
418
{
419 420 421 422 423
	struct nfs_server *server = sp->so_server;

	if (!RB_EMPTY_NODE(&sp->so_server_node))
		rb_erase(&sp->so_server_node, &server->state_owners);
	nfs_free_unique_id(&server->openowner_id, &sp->so_owner_id);
424 425
}

L
Linus Torvalds 已提交
426 427 428 429 430 431 432 433 434 435
/*
 * 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;

436
	sp = kzalloc(sizeof(*sp),GFP_NOFS);
L
Linus Torvalds 已提交
437 438
	if (!sp)
		return NULL;
439
	spin_lock_init(&sp->so_lock);
L
Linus Torvalds 已提交
440
	INIT_LIST_HEAD(&sp->so_states);
441 442 443 444
	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
445 446 447 448
	atomic_set(&sp->so_count, 1);
	return sp;
}

449
static void
L
Linus Torvalds 已提交
450 451
nfs4_drop_state_owner(struct nfs4_state_owner *sp)
{
452 453 454
	if (!RB_EMPTY_NODE(&sp->so_server_node)) {
		struct nfs_server *server = sp->so_server;
		struct nfs_client *clp = server->nfs_client;
455 456

		spin_lock(&clp->cl_lock);
457 458
		rb_erase(&sp->so_server_node, &server->state_owners);
		RB_CLEAR_NODE(&sp->so_server_node);
459 460
		spin_unlock(&clp->cl_lock);
	}
L
Linus Torvalds 已提交
461 462
}

463 464 465 466 467 468 469 470 471
/**
 * nfs4_get_state_owner - Look up a state owner given a credential
 * @server: nfs_server to search
 * @cred: RPC credential to match
 *
 * Returns a pointer to an instantiated nfs4_state_owner struct, or NULL.
 */
struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server,
					      struct rpc_cred *cred)
L
Linus Torvalds 已提交
472
{
473
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
474 475 476
	struct nfs4_state_owner *sp, *new;

	spin_lock(&clp->cl_lock);
477
	sp = nfs4_find_state_owner_locked(server, cred);
L
Linus Torvalds 已提交
478 479 480
	spin_unlock(&clp->cl_lock);
	if (sp != NULL)
		return sp;
481 482 483
	new = nfs4_alloc_state_owner();
	if (new == NULL)
		return NULL;
484
	new->so_server = server;
485 486
	new->so_cred = cred;
	spin_lock(&clp->cl_lock);
487
	sp = nfs4_insert_state_owner_locked(new);
488 489 490
	spin_unlock(&clp->cl_lock);
	if (sp == new)
		get_rpccred(cred);
491 492
	else {
		rpc_destroy_wait_queue(&new->so_sequence.wait);
493
		kfree(new);
494
	}
495
	return sp;
L
Linus Torvalds 已提交
496 497
}

498 499 500 501 502
/**
 * nfs4_put_state_owner - Release a nfs4_state_owner
 * @sp: state owner data to release
 *
 */
L
Linus Torvalds 已提交
503 504
void nfs4_put_state_owner(struct nfs4_state_owner *sp)
{
505
	struct nfs_client *clp = sp->so_server->nfs_client;
L
Linus Torvalds 已提交
506 507 508 509
	struct rpc_cred *cred = sp->so_cred;

	if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
		return;
510
	nfs4_remove_state_owner_locked(sp);
L
Linus Torvalds 已提交
511
	spin_unlock(&clp->cl_lock);
512
	rpc_destroy_wait_queue(&sp->so_sequence.wait);
L
Linus Torvalds 已提交
513 514 515 516 517 518 519 520 521
	put_rpccred(cred);
	kfree(sp);
}

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

522
	state = kzalloc(sizeof(*state), GFP_NOFS);
L
Linus Torvalds 已提交
523 524 525 526
	if (!state)
		return NULL;
	atomic_set(&state->count, 1);
	INIT_LIST_HEAD(&state->lock_states);
527
	spin_lock_init(&state->state_lock);
528
	seqlock_init(&state->seqlock);
L
Linus Torvalds 已提交
529 530 531
	return state;
}

532
void
533
nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
534
{
535
	if (state->state == fmode)
536 537
		return;
	/* NB! List reordering - see the reclaim code for why.  */
538 539
	if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
		if (fmode & FMODE_WRITE)
540 541 542 543
			list_move(&state->open_states, &state->owner->so_states);
		else
			list_move_tail(&state->open_states, &state->owner->so_states);
	}
544
	state->state = fmode;
545 546
}

L
Linus Torvalds 已提交
547 548 549 550 551 552 553
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) {
554
		if (state->owner != owner)
L
Linus Torvalds 已提交
555
			continue;
556
		if (atomic_inc_not_zero(&state->count))
L
Linus Torvalds 已提交
557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579
			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();
580
	spin_lock(&owner->so_lock);
L
Linus Torvalds 已提交
581 582 583 584 585 586 587 588 589
	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);
590 591 592 593
		/* 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
594 595
	} else {
		spin_unlock(&inode->i_lock);
596
		spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
597 598 599 600 601 602 603 604 605 606 607 608
		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;

609
	if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
L
Linus Torvalds 已提交
610
		return;
611
	spin_lock(&inode->i_lock);
612
	list_del(&state->inode_states);
L
Linus Torvalds 已提交
613
	list_del(&state->open_states);
614 615
	spin_unlock(&inode->i_lock);
	spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
616 617 618 619 620 621
	iput(inode);
	nfs4_free_open_state(state);
	nfs4_put_state_owner(owner);
}

/*
622
 * Close the current file.
L
Linus Torvalds 已提交
623
 */
624 625
static void __nfs4_close(struct path *path, struct nfs4_state *state,
		fmode_t fmode, gfp_t gfp_mask, int wait)
L
Linus Torvalds 已提交
626 627
{
	struct nfs4_state_owner *owner = state->owner;
628
	int call_close = 0;
629
	fmode_t newstate;
L
Linus Torvalds 已提交
630 631 632

	atomic_inc(&owner->so_count);
	/* Protect against nfs4_find_state() */
633
	spin_lock(&owner->so_lock);
634
	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
635 636 637 638 639 640 641 642 643
		case FMODE_READ:
			state->n_rdonly--;
			break;
		case FMODE_WRITE:
			state->n_wronly--;
			break;
		case FMODE_READ|FMODE_WRITE:
			state->n_rdwr--;
	}
644
	newstate = FMODE_READ|FMODE_WRITE;
645
	if (state->n_rdwr == 0) {
646
		if (state->n_rdonly == 0) {
647
			newstate &= ~FMODE_READ;
648 649 650 651
			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) {
652
			newstate &= ~FMODE_WRITE;
653 654 655 656 657
			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);
658
	}
659
	nfs4_state_set_mode_locked(state, newstate);
660
	spin_unlock(&owner->so_lock);
661

662
	if (!call_close) {
663 664
		nfs4_put_open_state(state);
		nfs4_put_state_owner(owner);
F
Fred Isaman 已提交
665 666 667 668 669
	} else {
		bool roc = pnfs_roc(state->inode);

		nfs4_do_close(path, state, gfp_mask, wait, roc);
	}
670 671
}

672
void nfs4_close_state(struct path *path, struct nfs4_state *state, fmode_t fmode)
673
{
674
	__nfs4_close(path, state, fmode, GFP_NOFS, 0);
675 676
}

677
void nfs4_close_sync(struct path *path, struct nfs4_state *state, fmode_t fmode)
678
{
679
	__nfs4_close(path, state, fmode, GFP_KERNEL, 1);
L
Linus Torvalds 已提交
680 681 682 683 684 685 686
}

/*
 * Search the state->lock_states for an existing lock_owner
 * that is compatible with current->files
 */
static struct nfs4_lock_state *
687
__nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type)
L
Linus Torvalds 已提交
688 689 690
{
	struct nfs4_lock_state *pos;
	list_for_each_entry(pos, &state->lock_states, ls_locks) {
691
		if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
L
Linus Torvalds 已提交
692
			continue;
693 694 695 696 697 698 699 700 701
		switch (pos->ls_owner.lo_type) {
		case NFS4_POSIX_LOCK_TYPE:
			if (pos->ls_owner.lo_u.posix_owner != fl_owner)
				continue;
			break;
		case NFS4_FLOCK_LOCK_TYPE:
			if (pos->ls_owner.lo_u.flock_owner != fl_pid)
				continue;
		}
L
Linus Torvalds 已提交
702 703 704 705 706 707 708 709 710 711 712
		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.
 *
 */
713
static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type)
L
Linus Torvalds 已提交
714 715
{
	struct nfs4_lock_state *lsp;
716 717
	struct nfs_server *server = state->owner->so_server;
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
718

719
	lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
L
Linus Torvalds 已提交
720 721
	if (lsp == NULL)
		return NULL;
722 723 724 725
	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
726
	atomic_set(&lsp->ls_count, 1);
727
	lsp->ls_state = state;
728 729 730 731 732 733 734 735 736 737 738 739
	lsp->ls_owner.lo_type = type;
	switch (lsp->ls_owner.lo_type) {
	case NFS4_FLOCK_LOCK_TYPE:
		lsp->ls_owner.lo_u.flock_owner = fl_pid;
		break;
	case NFS4_POSIX_LOCK_TYPE:
		lsp->ls_owner.lo_u.posix_owner = fl_owner;
		break;
	default:
		kfree(lsp);
		return NULL;
	}
L
Linus Torvalds 已提交
740
	spin_lock(&clp->cl_lock);
741
	nfs_alloc_unique_id_locked(&server->lockowner_id, &lsp->ls_id, 1, 64);
L
Linus Torvalds 已提交
742
	spin_unlock(&clp->cl_lock);
743
	INIT_LIST_HEAD(&lsp->ls_locks);
L
Linus Torvalds 已提交
744 745 746
	return lsp;
}

747 748
static void nfs4_free_lock_state(struct nfs4_lock_state *lsp)
{
749 750
	struct nfs_server *server = lsp->ls_state->owner->so_server;
	struct nfs_client *clp = server->nfs_client;
751 752

	spin_lock(&clp->cl_lock);
753
	nfs_free_unique_id(&server->lockowner_id, &lsp->ls_id);
754
	spin_unlock(&clp->cl_lock);
755
	rpc_destroy_wait_queue(&lsp->ls_sequence.wait);
756 757 758
	kfree(lsp);
}

L
Linus Torvalds 已提交
759 760 761 762 763
/*
 * Return a compatible lock_state. If no initialized lock_state structure
 * exists, return an uninitialized one.
 *
 */
764
static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner, pid_t pid, unsigned int type)
L
Linus Torvalds 已提交
765
{
766
	struct nfs4_lock_state *lsp, *new = NULL;
L
Linus Torvalds 已提交
767
	
768 769
	for(;;) {
		spin_lock(&state->state_lock);
770
		lsp = __nfs4_find_lock_state(state, owner, pid, type);
771 772 773 774 775 776 777 778 779 780
		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);
781
		new = nfs4_alloc_lock_state(state, owner, pid, type);
782 783 784 785
		if (new == NULL)
			return NULL;
	}
	spin_unlock(&state->state_lock);
786 787
	if (new != NULL)
		nfs4_free_lock_state(new);
L
Linus Torvalds 已提交
788 789 790 791
	return lsp;
}

/*
792 793
 * Release reference to lock_state, and free it if we see that
 * it is no longer in use
L
Linus Torvalds 已提交
794
 */
795
void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
L
Linus Torvalds 已提交
796
{
797
	struct nfs4_state *state;
L
Linus Torvalds 已提交
798

799 800 801 802 803 804 805 806 807
	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);
808 809
	if (lsp->ls_flags & NFS_LOCK_INITIALIZED)
		nfs4_release_lockowner(lsp);
810
	nfs4_free_lock_state(lsp);
L
Linus Torvalds 已提交
811 812
}

813
static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
L
Linus Torvalds 已提交
814
{
815
	struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
L
Linus Torvalds 已提交
816

817 818 819
	dst->fl_u.nfs4_fl.owner = lsp;
	atomic_inc(&lsp->ls_count);
}
L
Linus Torvalds 已提交
820

821
static void nfs4_fl_release_lock(struct file_lock *fl)
L
Linus Torvalds 已提交
822
{
823
	nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
L
Linus Torvalds 已提交
824 825
}

826
static const struct file_lock_operations nfs4_fl_lock_ops = {
827 828 829 830 831
	.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
832
{
833 834 835 836
	struct nfs4_lock_state *lsp;

	if (fl->fl_ops != NULL)
		return 0;
837 838 839 840 841 842
	if (fl->fl_flags & FL_POSIX)
		lsp = nfs4_get_lock_state(state, fl->fl_owner, 0, NFS4_POSIX_LOCK_TYPE);
	else if (fl->fl_flags & FL_FLOCK)
		lsp = nfs4_get_lock_state(state, 0, fl->fl_pid, NFS4_FLOCK_LOCK_TYPE);
	else
		return -EINVAL;
843 844 845 846 847
	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
848 849
}

850 851 852
/*
 * Byte-range lock aware utility to initialize the stateid of read/write
 * requests.
L
Linus Torvalds 已提交
853
 */
854
void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid)
L
Linus Torvalds 已提交
855
{
856
	struct nfs4_lock_state *lsp;
857
	int seq;
L
Linus Torvalds 已提交
858

859 860 861 862
	do {
		seq = read_seqbegin(&state->seqlock);
		memcpy(dst, &state->stateid, sizeof(*dst));
	} while (read_seqretry(&state->seqlock, seq));
863 864
	if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
		return;
L
Linus Torvalds 已提交
865

866
	spin_lock(&state->state_lock);
867
	lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
868 869 870
	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
871 872 873
	nfs4_put_lock_state(lsp);
}

874
struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
875 876 877
{
	struct nfs_seqid *new;

878
	new = kmalloc(sizeof(*new), gfp_mask);
879 880
	if (new != NULL) {
		new->sequence = counter;
881
		INIT_LIST_HEAD(&new->list);
882 883 884 885
	}
	return new;
}

886
void nfs_release_seqid(struct nfs_seqid *seqid)
L
Linus Torvalds 已提交
887
{
888 889
	if (!list_empty(&seqid->list)) {
		struct rpc_sequence *sequence = seqid->sequence->sequence;
890

891
		spin_lock(&sequence->lock);
892
		list_del_init(&seqid->list);
893 894 895
		spin_unlock(&sequence->lock);
		rpc_wake_up(&sequence->wait);
	}
896 897 898 899 900
}

void nfs_free_seqid(struct nfs_seqid *seqid)
{
	nfs_release_seqid(seqid);
901
	kfree(seqid);
L
Linus Torvalds 已提交
902 903 904
}

/*
905 906 907 908
 * 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()
 */
909
static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
910
{
911
	BUG_ON(list_first_entry(&seqid->sequence->sequence->list, struct nfs_seqid, list) != seqid);
912 913 914 915
	switch (status) {
		case 0:
			break;
		case -NFS4ERR_BAD_SEQID:
916 917 918
			if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
				return;
			printk(KERN_WARNING "NFS: v4 server returned a bad"
D
Dan Muntz 已提交
919 920
					" sequence-id error on an"
					" unconfirmed sequence %p!\n",
921
					seqid->sequence);
922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939
		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)
{
940 941 942 943 944
	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
945
		nfs4_drop_state_owner(sp);
946 947
	if (!nfs4_has_session(server->nfs_client))
		nfs_increment_seqid(status, seqid);
948 949 950 951 952 953 954 955 956
}

/*
 * 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)
{
957
	nfs_increment_seqid(status, seqid);
958 959 960 961 962 963 964 965
}

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);
966 967 968 969
	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;
970
	rpc_sleep_on(&sequence->wait, task, NULL);
971 972
	status = -EAGAIN;
unlock:
973 974
	spin_unlock(&sequence->lock);
	return status;
L
Linus Torvalds 已提交
975 976
}

977
static int nfs4_run_state_manager(void *);
L
Linus Torvalds 已提交
978

979
static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
T
Trond Myklebust 已提交
980 981
{
	smp_mb__before_clear_bit();
982
	clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
T
Trond Myklebust 已提交
983
	smp_mb__after_clear_bit();
984
	wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
T
Trond Myklebust 已提交
985 986 987
	rpc_wake_up(&clp->cl_rpcwaitq);
}

L
Linus Torvalds 已提交
988
/*
989
 * Schedule the nfs_client asynchronous state management routine
L
Linus Torvalds 已提交
990
 */
991
void nfs4_schedule_state_manager(struct nfs_client *clp)
L
Linus Torvalds 已提交
992
{
993
	struct task_struct *task;
L
Linus Torvalds 已提交
994

995 996
	if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
		return;
997 998
	__module_get(THIS_MODULE);
	atomic_inc(&clp->cl_count);
999
	task = kthread_run(nfs4_run_state_manager, clp, "%s-manager",
1000 1001
				rpc_peeraddr2str(clp->cl_rpcclient,
							RPC_DISPLAY_ADDR));
1002 1003
	if (!IS_ERR(task))
		return;
1004
	nfs4_clear_state_manager_bit(clp);
1005
	nfs_put_client(clp);
1006
	module_put(THIS_MODULE);
L
Linus Torvalds 已提交
1007 1008 1009 1010 1011
}

/*
 * Schedule a state recovery attempt
 */
1012
void nfs4_schedule_state_recovery(struct nfs_client *clp)
L
Linus Torvalds 已提交
1013 1014 1015
{
	if (!clp)
		return;
1016 1017
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1018
	nfs4_schedule_state_manager(clp);
L
Linus Torvalds 已提交
1019 1020
}

1021
int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1022 1023 1024 1025 1026 1027 1028 1029
{

	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;
	}
1030
	set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1031 1032 1033 1034
	set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
	return 1;
}

1035
int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1036 1037 1038
{
	set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
	clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1039
	set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1040 1041 1042 1043
	set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
	return 1;
}

1044
static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1045 1046
{
	struct inode *inode = state->inode;
1047
	struct nfs_inode *nfsi = NFS_I(inode);
L
Linus Torvalds 已提交
1048 1049 1050
	struct file_lock *fl;
	int status = 0;

1051 1052 1053 1054
	if (inode->i_flock == NULL)
		return 0;

	/* Guard against delegation returns and new lock/unlock calls */
1055
	down_write(&nfsi->rwsem);
1056
	/* Protect inode->i_flock using the BKL */
1057
	lock_flocks();
H
Harvey Harrison 已提交
1058
	for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1059
		if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
L
Linus Torvalds 已提交
1060
			continue;
1061
		if (nfs_file_open_context(fl->fl_file)->state != state)
L
Linus Torvalds 已提交
1062
			continue;
1063
		unlock_flocks();
L
Linus Torvalds 已提交
1064 1065
		status = ops->recover_lock(state, fl);
		switch (status) {
1066 1067 1068 1069 1070 1071 1072 1073 1074
			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:
1075 1076 1077 1078
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1079
				goto out;
L
Linus Torvalds 已提交
1080 1081
			default:
				printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
1082
						__func__, status);
1083 1084
			case -ENOMEM:
			case -NFS4ERR_DENIED:
L
Linus Torvalds 已提交
1085 1086
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1087
				/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1088
				status = 0;
L
Linus Torvalds 已提交
1089
		}
1090
		lock_flocks();
L
Linus Torvalds 已提交
1091
	}
1092
	unlock_flocks();
1093
out:
1094
	up_write(&nfsi->rwsem);
L
Linus Torvalds 已提交
1095 1096 1097
	return status;
}

1098
static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111
{
	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.
	 */
1112 1113
restart:
	spin_lock(&sp->so_lock);
L
Linus Torvalds 已提交
1114
	list_for_each_entry(state, &sp->so_states, open_states) {
1115 1116
		if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
			continue;
L
Linus Torvalds 已提交
1117 1118
		if (state->state == 0)
			continue;
1119 1120
		atomic_inc(&state->count);
		spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1121 1122
		status = ops->recover_open(sp, state);
		if (status >= 0) {
1123 1124 1125 1126 1127
			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",
1128
							__func__);
1129
				}
1130 1131
				nfs4_put_open_state(state);
				goto restart;
L
Linus Torvalds 已提交
1132 1133 1134 1135 1136
			}
		}
		switch (status) {
			default:
				printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
1137
						__func__, status);
L
Linus Torvalds 已提交
1138
			case -ENOENT:
1139
			case -ENOMEM:
1140
			case -ESTALE:
L
Linus Torvalds 已提交
1141 1142 1143 1144 1145 1146 1147 1148 1149
				/*
				 * 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;
1150 1151 1152 1153 1154 1155 1156 1157
			case -EKEYEXPIRED:
				/*
				 * User RPCSEC_GSS context has expired.
				 * We cannot recover this stateid now, so
				 * skip it and allow recovery thread to
				 * proceed.
				 */
				break;
1158 1159 1160
			case -NFS4ERR_ADMIN_REVOKED:
			case -NFS4ERR_STALE_STATEID:
			case -NFS4ERR_BAD_STATEID:
1161 1162
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1163
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1164
				break;
L
Linus Torvalds 已提交
1165 1166
			case -NFS4ERR_EXPIRED:
			case -NFS4ERR_NO_GRACE:
1167
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
L
Linus Torvalds 已提交
1168
			case -NFS4ERR_STALE_CLIENTID:
1169 1170 1171 1172
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
L
Linus Torvalds 已提交
1173 1174
				goto out_err;
		}
1175 1176
		nfs4_put_open_state(state);
		goto restart;
L
Linus Torvalds 已提交
1177
	}
1178
	spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1179 1180
	return 0;
out_err:
1181
	nfs4_put_open_state(state);
L
Linus Torvalds 已提交
1182 1183 1184
	return status;
}

1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198
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;
	}
}

1199 1200
static void nfs4_reset_seqids(struct nfs_server *server,
	int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1201
{
1202
	struct nfs_client *clp = server->nfs_client;
1203
	struct nfs4_state_owner *sp;
1204
	struct rb_node *pos;
1205 1206
	struct nfs4_state *state;

1207 1208 1209 1210 1211
	spin_lock(&clp->cl_lock);
	for (pos = rb_first(&server->state_owners);
	     pos != NULL;
	     pos = rb_next(pos)) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1212
		sp->so_seqid.flags = 0;
1213
		spin_lock(&sp->so_lock);
1214
		list_for_each_entry(state, &sp->so_states, open_states) {
1215 1216
			if (mark_reclaim(clp, state))
				nfs4_clear_open_state(state);
1217
		}
1218
		spin_unlock(&sp->so_lock);
1219
	}
1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231
	spin_unlock(&clp->cl_lock);
}

static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp,
	int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
{
	struct nfs_server *server;

	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
		nfs4_reset_seqids(server, mark_reclaim);
	rcu_read_unlock();
1232 1233
}

1234 1235 1236 1237 1238 1239 1240
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);
}

1241 1242 1243 1244 1245 1246 1247 1248
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);
}

1249
static void nfs4_clear_reclaim_server(struct nfs_server *server)
1250
{
1251
	struct nfs_client *clp = server->nfs_client;
1252 1253 1254 1255
	struct nfs4_state_owner *sp;
	struct rb_node *pos;
	struct nfs4_state *state;

1256 1257 1258 1259 1260
	spin_lock(&clp->cl_lock);
	for (pos = rb_first(&server->state_owners);
	     pos != NULL;
	     pos = rb_next(pos)) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
1261 1262
		spin_lock(&sp->so_lock);
		list_for_each_entry(state, &sp->so_states, open_states) {
1263 1264
			if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
						&state->flags))
1265 1266 1267 1268 1269
				continue;
			nfs4_state_mark_reclaim_nograce(clp, state);
		}
		spin_unlock(&sp->so_lock);
	}
1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283
	spin_unlock(&clp->cl_lock);
}

static int nfs4_state_clear_reclaim_reboot(struct nfs_client *clp)
{
	struct nfs_server *server;

	if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
		return 0;

	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
		nfs4_clear_reclaim_server(server);
	rcu_read_unlock();
1284 1285

	nfs_delegation_reap_unclaimed(clp);
1286 1287 1288 1289 1290 1291 1292 1293
	return 1;
}

static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
{
	if (!nfs4_state_clear_reclaim_reboot(clp))
		return;
	nfs4_reclaim_complete(clp, clp->cl_mvops->reboot_recovery_ops);
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307
}

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);
}

1308 1309 1310 1311 1312 1313 1314 1315
static void nfs4_warn_keyexpired(const char *s)
{
	printk_ratelimited(KERN_WARNING "Error: state manager"
			" encountered RPCSEC_GSS session"
			" expired against NFSv4 server %s.\n",
			s);
}

1316
static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1317 1318 1319
{
	switch (error) {
		case -NFS4ERR_CB_PATH_DOWN:
1320
			nfs_handle_cb_pathdown(clp);
1321
			return 0;
1322 1323 1324
		case -NFS4ERR_NO_GRACE:
			nfs4_state_end_reclaim_reboot(clp);
			return 0;
1325 1326 1327
		case -NFS4ERR_STALE_CLIENTID:
		case -NFS4ERR_LEASE_MOVED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1328
			nfs4_state_clear_reclaim_reboot(clp);
1329 1330 1331 1332 1333
			nfs4_state_start_reclaim_reboot(clp);
			break;
		case -NFS4ERR_EXPIRED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
			nfs4_state_start_reclaim_nograce(clp);
1334
			break;
1335 1336 1337 1338 1339 1340 1341
		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:
1342
			set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1343 1344
			/* Zero session reset errors */
			return 0;
1345 1346 1347 1348
		case -EKEYEXPIRED:
			/* Nothing we can do */
			nfs4_warn_keyexpired(clp->cl_hostname);
			return 0;
1349
	}
1350
	return error;
1351 1352
}

1353
static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1354
{
1355 1356
	struct nfs4_state_owner *sp;
	struct nfs_server *server;
1357
	struct rb_node *pos;
L
Linus Torvalds 已提交
1358 1359
	int status = 0;

1360
restart:
1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382
	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
		spin_lock(&clp->cl_lock);
		for (pos = rb_first(&server->state_owners);
		     pos != NULL;
		     pos = rb_next(pos)) {
			sp = rb_entry(pos,
				struct nfs4_state_owner, so_server_node);
			if (!test_and_clear_bit(ops->owner_flag_bit,
							&sp->so_flags))
				continue;
			atomic_inc(&sp->so_count);
			spin_unlock(&clp->cl_lock);
			rcu_read_unlock();

			status = nfs4_reclaim_open_state(sp, ops);
			if (status < 0) {
				set_bit(ops->owner_flag_bit, &sp->so_flags);
				nfs4_put_state_owner(sp);
				return nfs4_recovery_handle_error(clp, status);
			}

1383
			nfs4_put_state_owner(sp);
1384
			goto restart;
1385
		}
1386
		spin_unlock(&clp->cl_lock);
1387
	}
1388
	rcu_read_unlock();
1389 1390 1391 1392 1393 1394
	return status;
}

static int nfs4_check_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1395 1396
	const struct nfs4_state_maintenance_ops *ops =
		clp->cl_mvops->state_renewal_ops;
1397
	int status = -NFS4ERR_EXPIRED;
L
Linus Torvalds 已提交
1398

1399 1400 1401
	/* Is the client already known to have an expired lease? */
	if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		return 0;
1402 1403 1404
	spin_lock(&clp->cl_lock);
	cred = ops->get_state_renewal_cred_locked(clp);
	spin_unlock(&clp->cl_lock);
1405 1406 1407 1408
	if (cred == NULL) {
		cred = nfs4_get_setclientid_cred(clp);
		if (cred == NULL)
			goto out;
1409
	}
1410
	status = ops->renew_lease(clp, cred);
1411 1412
	put_rpccred(cred);
out:
1413
	return nfs4_recovery_handle_error(clp, status);
1414 1415 1416 1417 1418
}

static int nfs4_reclaim_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1419 1420
	const struct nfs4_state_recovery_ops *ops =
		clp->cl_mvops->reboot_recovery_ops;
1421 1422
	int status = -ENOENT;

1423
	cred = ops->get_clid_cred(clp);
1424
	if (cred != NULL) {
1425
		status = ops->establish_clid(clp, cred);
1426
		put_rpccred(cred);
1427 1428 1429
		/* Handle case where the user hasn't set up machine creds */
		if (status == -EACCES && cred == clp->cl_machine_cred) {
			nfs4_clear_machine_cred(clp);
1430
			status = -EAGAIN;
1431
		}
1432 1433
		if (status == -NFS4ERR_MINOR_VERS_MISMATCH)
			status = -EPROTONOSUPPORT;
1434
	}
1435 1436 1437
	return status;
}

1438
#ifdef CONFIG_NFS_V4_1
A
Andy Adamson 已提交
1439 1440 1441 1442 1443 1444
void nfs41_handle_recall_slot(struct nfs_client *clp)
{
	set_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
	nfs4_schedule_state_recovery(clp);
}

1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480
static void nfs4_reset_all_state(struct nfs_client *clp)
{
	if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
		clp->cl_boot_time = CURRENT_TIME;
		nfs4_state_start_reclaim_nograce(clp);
		nfs4_schedule_state_recovery(clp);
	}
}

static void nfs41_handle_server_reboot(struct nfs_client *clp)
{
	if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
		nfs4_state_start_reclaim_reboot(clp);
		nfs4_schedule_state_recovery(clp);
	}
}

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

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

static void nfs41_handle_cb_path_down(struct nfs_client *clp)
{
	nfs_expire_all_delegations(clp);
	if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
		nfs4_schedule_state_recovery(clp);
}

1481 1482 1483 1484
void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
{
	if (!flags)
		return;
1485 1486 1487
	else if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
		nfs41_handle_server_reboot(clp);
	else if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
1488 1489
			    SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
			    SEQ4_STATUS_ADMIN_STATE_REVOKED |
1490 1491 1492 1493 1494
			    SEQ4_STATUS_LEASE_MOVED))
		nfs41_handle_state_revoked(clp);
	else if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
		nfs41_handle_recallable_state_revoked(clp);
	else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
1495 1496
			    SEQ4_STATUS_BACKCHANNEL_FAULT |
			    SEQ4_STATUS_CB_PATH_DOWN_SESSION))
1497
		nfs41_handle_cb_path_down(clp);
1498 1499
}

1500 1501 1502 1503
static int nfs4_reset_session(struct nfs_client *clp)
{
	int status;

1504
	nfs4_begin_drain_session(clp);
1505 1506 1507
	status = nfs4_proc_destroy_session(clp->cl_session);
	if (status && status != -NFS4ERR_BADSESSION &&
	    status != -NFS4ERR_DEADSESSION) {
1508
		status = nfs4_recovery_handle_error(clp, status);
1509 1510 1511 1512
		goto out;
	}

	memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
1513
	status = nfs4_proc_create_session(clp);
1514
	if (status) {
1515
		status = nfs4_recovery_handle_error(clp, status);
1516 1517 1518 1519
		goto out;
	}
	/* create_session negotiated new slot table */
	clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state);
1520

1521 1522
	 /* Let the state manager reestablish state */
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1523
		nfs41_setup_state_renewal(clp);
1524
out:
1525 1526
	return status;
}
1527

A
Andy Adamson 已提交
1528 1529 1530 1531 1532 1533 1534 1535 1536
static int nfs4_recall_slot(struct nfs_client *clp)
{
	struct nfs4_slot_table *fc_tbl = &clp->cl_session->fc_slot_table;
	struct nfs4_channel_attrs *fc_attrs = &clp->cl_session->fc_attrs;
	struct nfs4_slot *new, *old;
	int i;

	nfs4_begin_drain_session(clp);
	new = kmalloc(fc_tbl->target_max_slots * sizeof(struct nfs4_slot),
1537
		      GFP_NOFS);
A
Andy Adamson 已提交
1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555
        if (!new)
		return -ENOMEM;

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

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

1556
#else /* CONFIG_NFS_V4_1 */
1557
static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
1558
static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
A
Andy Adamson 已提交
1559
static int nfs4_recall_slot(struct nfs_client *clp) { return 0; }
1560 1561
#endif /* CONFIG_NFS_V4_1 */

1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573
/* 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;

1574 1575
		case -EKEYEXPIRED:
			nfs4_warn_keyexpired(clp->cl_hostname);
1576 1577 1578 1579 1580 1581 1582 1583 1584
		case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
					 * in nfs4_exchange_id */
		default:
			return;
		}
	}
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
}

1585
static void nfs4_state_manager(struct nfs_client *clp)
1586 1587 1588 1589
{
	int status = 0;

	/* Ensure exclusive access to NFSv4 state */
1590
	for(;;) {
1591 1592 1593 1594
		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) {
1595
				nfs4_set_lease_expired(clp, status);
1596 1597
				if (test_bit(NFS4CLNT_LEASE_EXPIRED,
							&clp->cl_state))
1598
					continue;
1599 1600 1601
				if (clp->cl_cons_state ==
							NFS_CS_SESSION_INITING)
					nfs_mark_client_ready(clp, status);
1602 1603
				goto out_error;
			}
1604
			clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1605
			set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
1606
			pnfs_destroy_all_layouts(clp);
1607 1608 1609 1610
		}

		if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
			status = nfs4_check_lease(clp);
1611
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
1612
				continue;
1613 1614
			if (status < 0 && status != -NFS4ERR_CB_PATH_DOWN)
				goto out_error;
1615
		}
1616

1617
		/* Initialize or reset the session */
1618
		if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)
1619
		   && nfs4_has_session(clp)) {
1620
			status = nfs4_reset_session(clp);
1621 1622 1623
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
				continue;
			if (status < 0)
1624 1625
				goto out_error;
		}
1626

1627
		/* First recover reboot state... */
1628
		if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
1629
			status = nfs4_do_reclaim(clp,
1630
				clp->cl_mvops->reboot_recovery_ops);
1631
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1632
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
1633
				continue;
1634
			nfs4_state_end_reclaim_reboot(clp);
1635 1636 1637 1638
			if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
				continue;
			if (status < 0)
				goto out_error;
1639 1640
		}

1641 1642
		/* Now recover expired state... */
		if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
1643
			status = nfs4_do_reclaim(clp,
1644
				clp->cl_mvops->nograce_recovery_ops);
1645
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
1646
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
1647 1648 1649
			    test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
				continue;
			if (status < 0)
1650
				goto out_error;
L
Linus Torvalds 已提交
1651
		}
1652

1653
		nfs4_end_drain_session(clp);
1654 1655 1656 1657
		if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
			nfs_client_return_marked_delegations(clp);
			continue;
		}
A
Andy Adamson 已提交
1658 1659 1660 1661 1662 1663 1664 1665 1666
		/* Recall session slots */
		if (test_and_clear_bit(NFS4CLNT_RECALL_SLOT, &clp->cl_state)
		   && nfs4_has_session(clp)) {
			status = nfs4_recall_slot(clp);
			if (status < 0)
				goto out_error;
			continue;
		}

1667 1668

		nfs4_clear_state_manager_bit(clp);
1669 1670 1671 1672 1673
		/* 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
1674
	}
1675
	return;
L
Linus Torvalds 已提交
1676
out_error:
1677
	printk(KERN_WARNING "Error: state manager failed on NFSv4 server %s"
1678
			" with error %d\n", clp->cl_hostname, -status);
1679
	nfs4_end_drain_session(clp);
1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691
	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
1692 1693 1694 1695 1696 1697 1698
}

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