nfs4state.c 57.4 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
#include <linux/slab.h>
43
#include <linux/fs.h>
L
Linus Torvalds 已提交
44 45
#include <linux/nfs_fs.h>
#include <linux/nfs_idmap.h>
46 47
#include <linux/kthread.h>
#include <linux/module.h>
48
#include <linux/random.h>
49
#include <linux/ratelimit.h>
L
Linus Torvalds 已提交
50 51
#include <linux/workqueue.h>
#include <linux/bitops.h>
52
#include <linux/jiffies.h>
L
Linus Torvalds 已提交
53

54 55
#include <linux/sunrpc/clnt.h>

56
#include "nfs4_fs.h"
L
Linus Torvalds 已提交
57 58
#include "callback.h"
#include "delegation.h"
59
#include "internal.h"
60
#include "pnfs.h"
61
#include "netns.h"
L
Linus Torvalds 已提交
62

C
Chuck Lever 已提交
63 64
#define NFSDBG_FACILITY		NFSDBG_STATE

L
Linus Torvalds 已提交
65 66
#define OPENOWNER_POOL_SIZE	8

67
const nfs4_stateid zero_stateid;
68
static DEFINE_MUTEX(nfs_clid_init_mutex);
L
Linus Torvalds 已提交
69 70
static LIST_HEAD(nfs4_clientid_list);

71
int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
L
Linus Torvalds 已提交
72
{
73 74 75 76
	struct nfs4_setclientid_res clid = {
		.clientid = clp->cl_clientid,
		.confirm = clp->cl_confirm,
	};
77 78
	unsigned short port;
	int status;
79
	struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
80

81 82
	if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
		goto do_confirm;
83
	port = nn->nfs_callback_tcpport;
84
	if (clp->cl_addr.ss_family == AF_INET6)
85
		port = nn->nfs_callback_tcpport6;
86

87 88 89
	status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
	if (status != 0)
		goto out;
90 91 92 93
	clp->cl_clientid = clid.clientid;
	clp->cl_confirm = clid.confirm;
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
do_confirm:
94 95 96
	status = nfs4_proc_setclientid_confirm(clp, &clid, cred);
	if (status != 0)
		goto out;
97
	clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
98 99
	nfs4_schedule_state_renewal(clp);
out:
L
Linus Torvalds 已提交
100 101 102
	return status;
}

103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123
/**
 * nfs40_discover_server_trunking - Detect server IP address trunking (mv0)
 *
 * @clp: nfs_client under test
 * @result: OUT: found nfs_client, or clp
 * @cred: credential to use for trunking test
 *
 * Returns zero, a negative errno, or a negative NFS4ERR status.
 * If zero is returned, an nfs_client pointer is planted in
 * "result".
 *
 * Note: The returned client may not yet be marked ready.
 */
int nfs40_discover_server_trunking(struct nfs_client *clp,
				   struct nfs_client **result,
				   struct rpc_cred *cred)
{
	struct nfs4_setclientid_res clid = {
		.clientid = clp->cl_clientid,
		.confirm = clp->cl_confirm,
	};
124
	struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
125 126 127
	unsigned short port;
	int status;

128
	port = nn->nfs_callback_tcpport;
129
	if (clp->cl_addr.ss_family == AF_INET6)
130
		port = nn->nfs_callback_tcpport6;
131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152

	status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred, &clid);
	if (status != 0)
		goto out;
	clp->cl_clientid = clid.clientid;
	clp->cl_confirm = clid.confirm;

	status = nfs40_walk_client_list(clp, result, cred);
	switch (status) {
	case -NFS4ERR_STALE_CLIENTID:
		set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
	case 0:
		/* Sustain the lease, even if it's empty.  If the clientid4
		 * goes stale it's of no use for trunking discovery. */
		nfs4_schedule_state_renewal(*result);
		break;
	}

out:
	return status;
}

153
struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173
{
	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);
}

174 175
static struct rpc_cred *
nfs4_get_renew_cred_server_locked(struct nfs_server *server)
176
{
177
	struct rpc_cred *cred = NULL;
178
	struct nfs4_state_owner *sp;
179
	struct rb_node *pos;
180

181 182 183 184
	for (pos = rb_first(&server->state_owners);
	     pos != NULL;
	     pos = rb_next(pos)) {
		sp = rb_entry(pos, struct nfs4_state_owner, so_server_node);
185 186 187 188 189 190 191 192
		if (list_empty(&sp->so_states))
			continue;
		cred = get_rpccred(sp->so_cred);
		break;
	}
	return cred;
}

193 194 195 196 197 198 199 200 201 202 203 204
/**
 * 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;

205 206 207 208 209
	/* Use machine credentials if available */
	cred = nfs4_get_machine_cred_locked(clp);
	if (cred != NULL)
		goto out;

210 211 212 213 214 215 216
	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();
217 218

out:
219 220 221
	return cred;
}

222 223
#if defined(CONFIG_NFS_V4_1)

R
Ricardo Labiaga 已提交
224 225 226 227 228
static int nfs41_setup_state_renewal(struct nfs_client *clp)
{
	int status;
	struct nfs_fsinfo fsinfo;

229 230 231 232 233
	if (!test_bit(NFS_CS_CHECK_LEASE_TIME, &clp->cl_res_state)) {
		nfs4_schedule_state_renewal(clp);
		return 0;
	}

R
Ricardo Labiaga 已提交
234 235 236 237 238 239 240 241 242 243 244 245 246 247
	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;
}

248 249 250 251 252
/*
 * 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.
 */
253
static void nfs4_end_drain_session(struct nfs_client *clp)
254
{
255
	struct nfs4_session *ses = clp->cl_session;
256
	struct nfs4_slot_table *tbl;
257
	unsigned int i;
258

259 260
	if (ses == NULL)
		return;
261
	tbl = &ses->fc_slot_table;
262
	if (test_and_clear_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
263
		spin_lock(&tbl->slot_tbl_lock);
264
		for (i = 0; i <= tbl->max_slotid; i++) {
265 266 267
			if (rpc_wake_up_first(&tbl->slot_tbl_waitq,
						nfs4_set_task_privileged,
						NULL) == NULL)
268 269
				break;
		}
270
		spin_unlock(&tbl->slot_tbl_lock);
271
	}
272 273
}

274 275 276 277 278 279 280 281 282 283
/*
 * Signal state manager thread if session fore channel is drained
 */
void nfs4_session_drain_complete(struct nfs4_session *session,
		struct nfs4_slot_table *tbl)
{
	if (nfs4_session_draining(session))
		complete(&tbl->complete);
}

284
static int nfs4_wait_on_slot_tbl(struct nfs4_slot_table *tbl)
285 286
{
	spin_lock(&tbl->slot_tbl_lock);
287
	if (tbl->highest_used_slotid != NFS4_NO_SLOT) {
288
		INIT_COMPLETION(tbl->complete);
289
		spin_unlock(&tbl->slot_tbl_lock);
290
		return wait_for_completion_interruptible(&tbl->complete);
291 292 293 294 295
	}
	spin_unlock(&tbl->slot_tbl_lock);
	return 0;
}

296 297 298 299 300 301 302 303 304 305 306 307 308 309
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);
}

310 311 312 313 314 315 316 317 318
static void nfs41_finish_session_reset(struct nfs_client *clp)
{
	clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
	clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
	/* create_session negotiated new slot table */
	clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
	nfs41_setup_state_renewal(clp);
}

319 320 321 322
int nfs41_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
{
	int status;

323 324
	if (test_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state))
		goto do_confirm;
325
	nfs4_begin_drain_session(clp);
326
	status = nfs4_proc_exchange_id(clp, cred);
R
Ricardo Labiaga 已提交
327 328
	if (status != 0)
		goto out;
329 330
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
do_confirm:
331
	status = nfs4_proc_create_session(clp, cred);
R
Ricardo Labiaga 已提交
332 333
	if (status != 0)
		goto out;
334
	nfs41_finish_session_reset(clp);
R
Ricardo Labiaga 已提交
335 336
	nfs_mark_client_ready(clp, NFS_CS_READY);
out:
337 338 339
	return status;
}

340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361
/**
 * nfs41_discover_server_trunking - Detect server IP address trunking (mv1)
 *
 * @clp: nfs_client under test
 * @result: OUT: found nfs_client, or clp
 * @cred: credential to use for trunking test
 *
 * Returns NFS4_OK, a negative errno, or a negative NFS4ERR status.
 * If NFS4_OK is returned, an nfs_client pointer is planted in
 * "result".
 *
 * Note: The returned client may not yet be marked ready.
 */
int nfs41_discover_server_trunking(struct nfs_client *clp,
				   struct nfs_client **result,
				   struct rpc_cred *cred)
{
	int status;

	status = nfs4_proc_exchange_id(clp, cred);
	if (status != NFS4_OK)
		return status;
362
	set_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
363 364 365 366

	return nfs41_walk_client_list(clp, result, cred);
}

367 368 369 370 371 372 373 374 375 376 377 378
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 */

379 380
static struct rpc_cred *
nfs4_get_setclientid_cred_server(struct nfs_server *server)
381
{
382 383
	struct nfs_client *clp = server->nfs_client;
	struct rpc_cred *cred = NULL;
384
	struct nfs4_state_owner *sp;
385
	struct rb_node *pos;
386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405

	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;
406
	struct rpc_cred *cred;
407

408 409
	spin_lock(&clp->cl_lock);
	cred = nfs4_get_machine_cred_locked(clp);
410
	spin_unlock(&clp->cl_lock);
411 412
	if (cred != NULL)
		goto out;
413 414 415 416 417 418

	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;
419
	}
420 421
	rcu_read_unlock();

422 423
out:
	return cred;
424 425
}

L
Linus Torvalds 已提交
426
static struct nfs4_state_owner *
427
nfs4_find_state_owner_locked(struct nfs_server *server, struct rpc_cred *cred)
L
Linus Torvalds 已提交
428
{
429
	struct rb_node **p = &server->state_owners.rb_node,
430
		       *parent = NULL;
431
	struct nfs4_state_owner *sp;
L
Linus Torvalds 已提交
432

433 434
	while (*p != NULL) {
		parent = *p;
435
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
436 437 438 439 440 441

		if (cred < sp->so_cred)
			p = &parent->rb_left;
		else if (cred > sp->so_cred)
			p = &parent->rb_right;
		else {
442 443
			if (!list_empty(&sp->so_lru))
				list_del_init(&sp->so_lru);
444
			atomic_inc(&sp->so_count);
445
			return sp;
446
		}
L
Linus Torvalds 已提交
447
	}
448
	return NULL;
L
Linus Torvalds 已提交
449 450
}

451
static struct nfs4_state_owner *
452
nfs4_insert_state_owner_locked(struct nfs4_state_owner *new)
453
{
454 455
	struct nfs_server *server = new->so_server;
	struct rb_node **p = &server->state_owners.rb_node,
456 457
		       *parent = NULL;
	struct nfs4_state_owner *sp;
458
	int err;
459 460 461

	while (*p != NULL) {
		parent = *p;
462
		sp = rb_entry(parent, struct nfs4_state_owner, so_server_node);
463 464 465 466 467 468

		if (new->so_cred < sp->so_cred)
			p = &parent->rb_left;
		else if (new->so_cred > sp->so_cred)
			p = &parent->rb_right;
		else {
469 470
			if (!list_empty(&sp->so_lru))
				list_del_init(&sp->so_lru);
471 472 473 474
			atomic_inc(&sp->so_count);
			return sp;
		}
	}
475
	err = ida_get_new(&server->openowner_id, &new->so_seqid.owner_id);
476 477
	if (err)
		return ERR_PTR(err);
478 479
	rb_link_node(&new->so_server_node, parent, p);
	rb_insert_color(&new->so_server_node, &server->state_owners);
480 481 482 483
	return new;
}

static void
484
nfs4_remove_state_owner_locked(struct nfs4_state_owner *sp)
485
{
486 487 488 489
	struct nfs_server *server = sp->so_server;

	if (!RB_EMPTY_NODE(&sp->so_server_node))
		rb_erase(&sp->so_server_node, &server->state_owners);
490
	ida_remove(&server->openowner_id, sp->so_seqid.owner_id);
491 492
}

493 494 495
static void
nfs4_init_seqid_counter(struct nfs_seqid_counter *sc)
{
496
	sc->create_time = ktime_get();
497 498 499 500 501 502 503 504 505 506 507 508 509
	sc->flags = 0;
	sc->counter = 0;
	spin_lock_init(&sc->lock);
	INIT_LIST_HEAD(&sc->list);
	rpc_init_wait_queue(&sc->wait, "Seqid_waitqueue");
}

static void
nfs4_destroy_seqid_counter(struct nfs_seqid_counter *sc)
{
	rpc_destroy_wait_queue(&sc->wait);
}

L
Linus Torvalds 已提交
510 511 512 513 514 515
/*
 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
 * create a new state_owner.
 *
 */
static struct nfs4_state_owner *
516 517 518
nfs4_alloc_state_owner(struct nfs_server *server,
		struct rpc_cred *cred,
		gfp_t gfp_flags)
L
Linus Torvalds 已提交
519 520 521
{
	struct nfs4_state_owner *sp;

522
	sp = kzalloc(sizeof(*sp), gfp_flags);
L
Linus Torvalds 已提交
523 524
	if (!sp)
		return NULL;
525 526
	sp->so_server = server;
	sp->so_cred = get_rpccred(cred);
527
	spin_lock_init(&sp->so_lock);
L
Linus Torvalds 已提交
528
	INIT_LIST_HEAD(&sp->so_states);
529
	nfs4_init_seqid_counter(&sp->so_seqid);
L
Linus Torvalds 已提交
530
	atomic_set(&sp->so_count, 1);
531
	INIT_LIST_HEAD(&sp->so_lru);
L
Linus Torvalds 已提交
532 533 534
	return sp;
}

535
static void
L
Linus Torvalds 已提交
536 537
nfs4_drop_state_owner(struct nfs4_state_owner *sp)
{
538 539 540
	struct rb_node *rb_node = &sp->so_server_node;

	if (!RB_EMPTY_NODE(rb_node)) {
541 542
		struct nfs_server *server = sp->so_server;
		struct nfs_client *clp = server->nfs_client;
543 544

		spin_lock(&clp->cl_lock);
545 546 547 548
		if (!RB_EMPTY_NODE(rb_node)) {
			rb_erase(rb_node, &server->state_owners);
			RB_CLEAR_NODE(rb_node);
		}
549 550
		spin_unlock(&clp->cl_lock);
	}
L
Linus Torvalds 已提交
551 552
}

553 554
static void nfs4_free_state_owner(struct nfs4_state_owner *sp)
{
555
	nfs4_destroy_seqid_counter(&sp->so_seqid);
556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584
	put_rpccred(sp->so_cred);
	kfree(sp);
}

static void nfs4_gc_state_owners(struct nfs_server *server)
{
	struct nfs_client *clp = server->nfs_client;
	struct nfs4_state_owner *sp, *tmp;
	unsigned long time_min, time_max;
	LIST_HEAD(doomed);

	spin_lock(&clp->cl_lock);
	time_max = jiffies;
	time_min = (long)time_max - (long)clp->cl_lease_time;
	list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
		/* NB: LRU is sorted so that oldest is at the head */
		if (time_in_range(sp->so_expires, time_min, time_max))
			break;
		list_move(&sp->so_lru, &doomed);
		nfs4_remove_state_owner_locked(sp);
	}
	spin_unlock(&clp->cl_lock);

	list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
		list_del(&sp->so_lru);
		nfs4_free_state_owner(sp);
	}
}

585 586 587 588 589 590 591 592
/**
 * 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,
593 594
					      struct rpc_cred *cred,
					      gfp_t gfp_flags)
L
Linus Torvalds 已提交
595
{
596
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
597 598 599
	struct nfs4_state_owner *sp, *new;

	spin_lock(&clp->cl_lock);
600
	sp = nfs4_find_state_owner_locked(server, cred);
L
Linus Torvalds 已提交
601 602
	spin_unlock(&clp->cl_lock);
	if (sp != NULL)
603
		goto out;
604
	new = nfs4_alloc_state_owner(server, cred, gfp_flags);
605
	if (new == NULL)
606
		goto out;
607 608 609 610 611 612 613
	do {
		if (ida_pre_get(&server->openowner_id, gfp_flags) == 0)
			break;
		spin_lock(&clp->cl_lock);
		sp = nfs4_insert_state_owner_locked(new);
		spin_unlock(&clp->cl_lock);
	} while (sp == ERR_PTR(-EAGAIN));
614 615
	if (sp != new)
		nfs4_free_state_owner(new);
616 617
out:
	nfs4_gc_state_owners(server);
618
	return sp;
L
Linus Torvalds 已提交
619 620
}

621 622 623
/**
 * nfs4_put_state_owner - Release a nfs4_state_owner
 * @sp: state owner data to release
624 625 626 627 628 629 630 631
 *
 * Note that we keep released state owners on an LRU
 * list.
 * This caches valid state owners so that they can be
 * reused, to avoid the OPEN_CONFIRM on minor version 0.
 * It also pins the uniquifier of dropped state owners for
 * a while, to ensure that those state owner names are
 * never reused.
632
 */
L
Linus Torvalds 已提交
633 634
void nfs4_put_state_owner(struct nfs4_state_owner *sp)
{
635 636
	struct nfs_server *server = sp->so_server;
	struct nfs_client *clp = server->nfs_client;
L
Linus Torvalds 已提交
637 638 639

	if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
		return;
640

641 642 643
	sp->so_expires = jiffies;
	list_add_tail(&sp->so_lru, &server->state_owners_lru);
	spin_unlock(&clp->cl_lock);
644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663
}

/**
 * nfs4_purge_state_owners - Release all cached state owners
 * @server: nfs_server with cached state owners to release
 *
 * Called at umount time.  Remaining state owners will be on
 * the LRU with ref count of zero.
 */
void nfs4_purge_state_owners(struct nfs_server *server)
{
	struct nfs_client *clp = server->nfs_client;
	struct nfs4_state_owner *sp, *tmp;
	LIST_HEAD(doomed);

	spin_lock(&clp->cl_lock);
	list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) {
		list_move(&sp->so_lru, &doomed);
		nfs4_remove_state_owner_locked(sp);
	}
L
Linus Torvalds 已提交
664
	spin_unlock(&clp->cl_lock);
665 666 667 668 669

	list_for_each_entry_safe(sp, tmp, &doomed, so_lru) {
		list_del(&sp->so_lru);
		nfs4_free_state_owner(sp);
	}
L
Linus Torvalds 已提交
670 671 672 673 674 675 676
}

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

677
	state = kzalloc(sizeof(*state), GFP_NOFS);
L
Linus Torvalds 已提交
678 679 680 681
	if (!state)
		return NULL;
	atomic_set(&state->count, 1);
	INIT_LIST_HEAD(&state->lock_states);
682
	spin_lock_init(&state->state_lock);
683
	seqlock_init(&state->seqlock);
L
Linus Torvalds 已提交
684 685 686
	return state;
}

687
void
688
nfs4_state_set_mode_locked(struct nfs4_state *state, fmode_t fmode)
689
{
690
	if (state->state == fmode)
691 692
		return;
	/* NB! List reordering - see the reclaim code for why.  */
693 694
	if ((fmode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
		if (fmode & FMODE_WRITE)
695 696 697 698
			list_move(&state->open_states, &state->owner->so_states);
		else
			list_move_tail(&state->open_states, &state->owner->so_states);
	}
699
	state->state = fmode;
700 701
}

L
Linus Torvalds 已提交
702 703 704 705 706 707 708
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) {
709
		if (state->owner != owner)
L
Linus Torvalds 已提交
710
			continue;
711
		if (atomic_inc_not_zero(&state->count))
L
Linus Torvalds 已提交
712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734
			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();
735
	spin_lock(&owner->so_lock);
L
Linus Torvalds 已提交
736 737 738 739 740 741 742
	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);
743 744
		ihold(inode);
		state->inode = inode;
L
Linus Torvalds 已提交
745
		spin_unlock(&inode->i_lock);
746 747 748 749
		/* 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
750 751
	} else {
		spin_unlock(&inode->i_lock);
752
		spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
753 754 755 756 757 758 759 760 761 762 763 764
		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;

765
	if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
L
Linus Torvalds 已提交
766
		return;
767
	spin_lock(&inode->i_lock);
768
	list_del(&state->inode_states);
L
Linus Torvalds 已提交
769
	list_del(&state->open_states);
770 771
	spin_unlock(&inode->i_lock);
	spin_unlock(&owner->so_lock);
L
Linus Torvalds 已提交
772 773 774 775 776 777
	iput(inode);
	nfs4_free_open_state(state);
	nfs4_put_state_owner(owner);
}

/*
778
 * Close the current file.
L
Linus Torvalds 已提交
779
 */
780
static void __nfs4_close(struct nfs4_state *state,
781
		fmode_t fmode, gfp_t gfp_mask, int wait)
L
Linus Torvalds 已提交
782 783
{
	struct nfs4_state_owner *owner = state->owner;
784
	int call_close = 0;
785
	fmode_t newstate;
L
Linus Torvalds 已提交
786 787 788

	atomic_inc(&owner->so_count);
	/* Protect against nfs4_find_state() */
789
	spin_lock(&owner->so_lock);
790
	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
791 792 793 794 795 796 797 798 799
		case FMODE_READ:
			state->n_rdonly--;
			break;
		case FMODE_WRITE:
			state->n_wronly--;
			break;
		case FMODE_READ|FMODE_WRITE:
			state->n_rdwr--;
	}
800
	newstate = FMODE_READ|FMODE_WRITE;
801
	if (state->n_rdwr == 0) {
802
		if (state->n_rdonly == 0) {
803
			newstate &= ~FMODE_READ;
804 805 806 807
			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) {
808
			newstate &= ~FMODE_WRITE;
809 810 811 812 813
			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);
814
	}
815
	nfs4_state_set_mode_locked(state, newstate);
816
	spin_unlock(&owner->so_lock);
817

818
	if (!call_close) {
819 820
		nfs4_put_open_state(state);
		nfs4_put_state_owner(owner);
821 822
	} else
		nfs4_do_close(state, gfp_mask, wait);
823 824
}

825
void nfs4_close_state(struct nfs4_state *state, fmode_t fmode)
826
{
827
	__nfs4_close(state, fmode, GFP_NOFS, 0);
828 829
}

830
void nfs4_close_sync(struct nfs4_state *state, fmode_t fmode)
831
{
832
	__nfs4_close(state, fmode, GFP_KERNEL, 1);
L
Linus Torvalds 已提交
833 834 835 836 837 838 839
}

/*
 * Search the state->lock_states for an existing lock_owner
 * that is compatible with current->files
 */
static struct nfs4_lock_state *
840
__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
841 842 843
{
	struct nfs4_lock_state *pos;
	list_for_each_entry(pos, &state->lock_states, ls_locks) {
844
		if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type)
L
Linus Torvalds 已提交
845
			continue;
846 847 848 849 850 851 852 853 854
		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
855 856 857 858 859 860 861 862 863 864 865
		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.
 *
 */
866
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
867 868
{
	struct nfs4_lock_state *lsp;
869
	struct nfs_server *server = state->owner->so_server;
L
Linus Torvalds 已提交
870

871
	lsp = kzalloc(sizeof(*lsp), GFP_NOFS);
L
Linus Torvalds 已提交
872 873
	if (lsp == NULL)
		return NULL;
874
	nfs4_init_seqid_counter(&lsp->ls_seqid);
L
Linus Torvalds 已提交
875
	atomic_set(&lsp->ls_count, 1);
876
	lsp->ls_state = state;
877 878 879 880 881 882 883 884 885
	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:
886
		goto out_free;
887
	}
888 889
	lsp->ls_seqid.owner_id = ida_simple_get(&server->lockowner_id, 0, 0, GFP_NOFS);
	if (lsp->ls_seqid.owner_id < 0)
890
		goto out_free;
891
	INIT_LIST_HEAD(&lsp->ls_locks);
L
Linus Torvalds 已提交
892
	return lsp;
893 894 895
out_free:
	kfree(lsp);
	return NULL;
L
Linus Torvalds 已提交
896 897
}

898
void nfs4_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
899
{
900
	ida_simple_remove(&server->lockowner_id, lsp->ls_seqid.owner_id);
901
	nfs4_destroy_seqid_counter(&lsp->ls_seqid);
902 903 904
	kfree(lsp);
}

L
Linus Torvalds 已提交
905 906 907 908 909
/*
 * Return a compatible lock_state. If no initialized lock_state structure
 * exists, return an uninitialized one.
 *
 */
910
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
911
{
912
	struct nfs4_lock_state *lsp, *new = NULL;
L
Linus Torvalds 已提交
913
	
914 915
	for(;;) {
		spin_lock(&state->state_lock);
916
		lsp = __nfs4_find_lock_state(state, owner, pid, type);
917 918 919 920 921 922 923 924 925 926
		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);
927
		new = nfs4_alloc_lock_state(state, owner, pid, type);
928 929 930 931
		if (new == NULL)
			return NULL;
	}
	spin_unlock(&state->state_lock);
932
	if (new != NULL)
933
		nfs4_free_lock_state(state->owner->so_server, new);
L
Linus Torvalds 已提交
934 935 936 937
	return lsp;
}

/*
938 939
 * Release reference to lock_state, and free it if we see that
 * it is no longer in use
L
Linus Torvalds 已提交
940
 */
941
void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
L
Linus Torvalds 已提交
942
{
943
	struct nfs4_state *state;
L
Linus Torvalds 已提交
944

945 946 947 948 949 950 951 952 953
	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);
954
	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
955 956 957
		if (nfs4_release_lockowner(lsp) == 0)
			return;
	}
958
	nfs4_free_lock_state(lsp->ls_state->owner->so_server, lsp);
L
Linus Torvalds 已提交
959 960
}

961
static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
L
Linus Torvalds 已提交
962
{
963
	struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
L
Linus Torvalds 已提交
964

965 966 967
	dst->fl_u.nfs4_fl.owner = lsp;
	atomic_inc(&lsp->ls_count);
}
L
Linus Torvalds 已提交
968

969
static void nfs4_fl_release_lock(struct file_lock *fl)
L
Linus Torvalds 已提交
970
{
971
	nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
L
Linus Torvalds 已提交
972 973
}

974
static const struct file_lock_operations nfs4_fl_lock_ops = {
975 976 977 978 979
	.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
980
{
981 982 983 984
	struct nfs4_lock_state *lsp;

	if (fl->fl_ops != NULL)
		return 0;
985 986 987
	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)
988 989
		lsp = nfs4_get_lock_state(state, NULL, fl->fl_pid,
				NFS4_FLOCK_LOCK_TYPE);
990 991
	else
		return -EINVAL;
992 993 994 995 996
	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
997 998
}

999
static bool nfs4_copy_lock_stateid(nfs4_stateid *dst, struct nfs4_state *state,
1000
		const struct nfs_lockowner *lockowner)
L
Linus Torvalds 已提交
1001
{
1002
	struct nfs4_lock_state *lsp;
1003 1004
	fl_owner_t fl_owner;
	pid_t fl_pid;
1005
	bool ret = false;
L
Linus Torvalds 已提交
1006

1007 1008 1009 1010

	if (lockowner == NULL)
		goto out;

1011
	if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
1012
		goto out;
L
Linus Torvalds 已提交
1013

1014 1015
	fl_owner = lockowner->l_owner;
	fl_pid = lockowner->l_pid;
1016
	spin_lock(&state->state_lock);
1017
	lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE);
1018
	if (lsp != NULL && test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0) {
1019
		nfs4_stateid_copy(dst, &lsp->ls_stateid);
1020 1021
		ret = true;
	}
1022
	spin_unlock(&state->state_lock);
L
Linus Torvalds 已提交
1023
	nfs4_put_lock_state(lsp);
1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042
out:
	return ret;
}

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

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

/*
 * Byte-range lock aware utility to initialize the stateid of read/write
 * requests.
 */
void nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state,
1043
		fmode_t fmode, const struct nfs_lockowner *lockowner)
1044 1045 1046
{
	if (nfs4_copy_delegation_stateid(dst, state->inode, fmode))
		return;
1047
	if (nfs4_copy_lock_stateid(dst, state, lockowner))
1048 1049
		return;
	nfs4_copy_open_stateid(dst, state);
L
Linus Torvalds 已提交
1050 1051
}

1052
struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask)
1053 1054 1055
{
	struct nfs_seqid *new;

1056
	new = kmalloc(sizeof(*new), gfp_mask);
1057 1058
	if (new != NULL) {
		new->sequence = counter;
1059
		INIT_LIST_HEAD(&new->list);
1060
		new->task = NULL;
1061 1062 1063 1064
	}
	return new;
}

1065
void nfs_release_seqid(struct nfs_seqid *seqid)
L
Linus Torvalds 已提交
1066
{
1067
	struct nfs_seqid_counter *sequence;
1068

1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079
	if (list_empty(&seqid->list))
		return;
	sequence = seqid->sequence;
	spin_lock(&sequence->lock);
	list_del_init(&seqid->list);
	if (!list_empty(&sequence->list)) {
		struct nfs_seqid *next;

		next = list_first_entry(&sequence->list,
				struct nfs_seqid, list);
		rpc_wake_up_queued_task(&sequence->wait, next->task);
1080
	}
1081
	spin_unlock(&sequence->lock);
1082 1083 1084 1085 1086
}

void nfs_free_seqid(struct nfs_seqid *seqid)
{
	nfs_release_seqid(seqid);
1087
	kfree(seqid);
L
Linus Torvalds 已提交
1088 1089 1090
}

/*
1091 1092 1093 1094
 * 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()
 */
1095
static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
1096 1097 1098 1099 1100
{
	switch (status) {
		case 0:
			break;
		case -NFS4ERR_BAD_SEQID:
1101 1102
			if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
				return;
1103
			pr_warn_ratelimited("NFS: v4 server returned a bad"
D
Dan Muntz 已提交
1104 1105
					" sequence-id error on an"
					" unconfirmed sequence %p!\n",
1106
					seqid->sequence);
1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124
		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)
{
1125 1126 1127 1128 1129
	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
1130
		nfs4_drop_state_owner(sp);
1131 1132
	if (!nfs4_has_session(server->nfs_client))
		nfs_increment_seqid(status, seqid);
1133 1134 1135 1136 1137 1138 1139 1140 1141
}

/*
 * 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)
{
1142
	nfs_increment_seqid(status, seqid);
1143 1144 1145 1146
}

int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
{
1147
	struct nfs_seqid_counter *sequence = seqid->sequence;
1148 1149 1150
	int status = 0;

	spin_lock(&sequence->lock);
1151
	seqid->task = task;
1152 1153 1154 1155
	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;
1156
	rpc_sleep_on(&sequence->wait, task, NULL);
1157 1158
	status = -EAGAIN;
unlock:
1159 1160
	spin_unlock(&sequence->lock);
	return status;
L
Linus Torvalds 已提交
1161 1162
}

1163
static int nfs4_run_state_manager(void *);
L
Linus Torvalds 已提交
1164

1165
static void nfs4_clear_state_manager_bit(struct nfs_client *clp)
T
Trond Myklebust 已提交
1166 1167
{
	smp_mb__before_clear_bit();
1168
	clear_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state);
T
Trond Myklebust 已提交
1169
	smp_mb__after_clear_bit();
1170
	wake_up_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING);
T
Trond Myklebust 已提交
1171 1172 1173
	rpc_wake_up(&clp->cl_rpcwaitq);
}

L
Linus Torvalds 已提交
1174
/*
1175
 * Schedule the nfs_client asynchronous state management routine
L
Linus Torvalds 已提交
1176
 */
1177
void nfs4_schedule_state_manager(struct nfs_client *clp)
L
Linus Torvalds 已提交
1178
{
1179
	struct task_struct *task;
1180
	char buf[INET6_ADDRSTRLEN + sizeof("-manager") + 1];
L
Linus Torvalds 已提交
1181

1182 1183
	if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
		return;
1184 1185
	__module_get(THIS_MODULE);
	atomic_inc(&clp->cl_count);
1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201

	/* The rcu_read_lock() is not strictly necessary, as the state
	 * manager is the only thread that ever changes the rpc_xprt
	 * after it's initialized.  At this point, we're single threaded. */
	rcu_read_lock();
	snprintf(buf, sizeof(buf), "%s-manager",
			rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR));
	rcu_read_unlock();
	task = kthread_run(nfs4_run_state_manager, clp, buf);
	if (IS_ERR(task)) {
		printk(KERN_ERR "%s: kthread_run: %ld\n",
			__func__, PTR_ERR(task));
		nfs4_clear_state_manager_bit(clp);
		nfs_put_client(clp);
		module_put(THIS_MODULE);
	}
L
Linus Torvalds 已提交
1202 1203 1204
}

/*
1205
 * Schedule a lease recovery attempt
L
Linus Torvalds 已提交
1206
 */
1207
void nfs4_schedule_lease_recovery(struct nfs_client *clp)
L
Linus Torvalds 已提交
1208 1209 1210
{
	if (!clp)
		return;
1211 1212
	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
1213 1214
	dprintk("%s: scheduling lease recovery for server %s\n", __func__,
			clp->cl_hostname);
1215
	nfs4_schedule_state_manager(clp);
L
Linus Torvalds 已提交
1216
}
A
Andy Adamson 已提交
1217
EXPORT_SYMBOL_GPL(nfs4_schedule_lease_recovery);
L
Linus Torvalds 已提交
1218

1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230
/*
 * nfs40_handle_cb_pathdown - return all delegations after NFS4ERR_CB_PATH_DOWN
 * @clp: client to process
 *
 * Set the NFS4CLNT_LEASE_EXPIRED state in order to force a
 * resend of the SETCLIENTID and hence re-establish the
 * callback channel. Then return all existing delegations.
 */
static void nfs40_handle_cb_pathdown(struct nfs_client *clp)
{
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
	nfs_expire_all_delegations(clp);
1231 1232
	dprintk("%s: handling CB_PATHDOWN recovery for server %s\n", __func__,
			clp->cl_hostname);
1233 1234
}

1235 1236
void nfs4_schedule_path_down_recovery(struct nfs_client *clp)
{
1237
	nfs40_handle_cb_pathdown(clp);
1238 1239 1240
	nfs4_schedule_state_manager(clp);
}

1241
static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
1242 1243 1244 1245 1246 1247 1248 1249
{

	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;
	}
1250
	set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
1251 1252 1253 1254
	set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
	return 1;
}

1255
static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
1256 1257 1258
{
	set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
	clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
1259
	set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
1260 1261 1262 1263
	set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
	return 1;
}

1264 1265 1266 1267 1268
void nfs4_schedule_stateid_recovery(const struct nfs_server *server, struct nfs4_state *state)
{
	struct nfs_client *clp = server->nfs_client;

	nfs4_state_mark_reclaim_nograce(clp, state);
1269 1270
	dprintk("%s: scheduling stateid recovery for server %s\n", __func__,
			clp->cl_hostname);
1271 1272
	nfs4_schedule_state_manager(clp);
}
A
Andy Adamson 已提交
1273
EXPORT_SYMBOL_GPL(nfs4_schedule_stateid_recovery);
1274

1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290
void nfs_inode_find_state_and_recover(struct inode *inode,
		const nfs4_stateid *stateid)
{
	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_open_context *ctx;
	struct nfs4_state *state;
	bool found = false;

	spin_lock(&inode->i_lock);
	list_for_each_entry(ctx, &nfsi->open_files, list) {
		state = ctx->state;
		if (state == NULL)
			continue;
		if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
			continue;
1291
		if (!nfs4_stateid_match(&state->stateid, stateid))
1292 1293 1294 1295 1296 1297 1298 1299 1300 1301
			continue;
		nfs4_state_mark_reclaim_nograce(clp, state);
		found = true;
	}
	spin_unlock(&inode->i_lock);
	if (found)
		nfs4_schedule_state_manager(clp);
}


1302
static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1303 1304
{
	struct inode *inode = state->inode;
1305
	struct nfs_inode *nfsi = NFS_I(inode);
L
Linus Torvalds 已提交
1306 1307 1308
	struct file_lock *fl;
	int status = 0;

1309 1310 1311 1312
	if (inode->i_flock == NULL)
		return 0;

	/* Guard against delegation returns and new lock/unlock calls */
1313
	down_write(&nfsi->rwsem);
1314
	/* Protect inode->i_flock using the BKL */
1315
	lock_flocks();
H
Harvey Harrison 已提交
1316
	for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1317
		if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
L
Linus Torvalds 已提交
1318
			continue;
1319
		if (nfs_file_open_context(fl->fl_file)->state != state)
L
Linus Torvalds 已提交
1320
			continue;
1321
		unlock_flocks();
L
Linus Torvalds 已提交
1322 1323
		status = ops->recover_lock(state, fl);
		switch (status) {
1324 1325 1326 1327 1328 1329 1330 1331 1332
			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:
1333 1334 1335 1336
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1337
				goto out;
L
Linus Torvalds 已提交
1338
			default:
1339 1340
				printk(KERN_ERR "NFS: %s: unhandled error %d. "
					"Zeroing state\n", __func__, status);
1341 1342
			case -ENOMEM:
			case -NFS4ERR_DENIED:
L
Linus Torvalds 已提交
1343 1344
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1345
				/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1346
				status = 0;
L
Linus Torvalds 已提交
1347
		}
1348
		lock_flocks();
L
Linus Torvalds 已提交
1349
	}
1350
	unlock_flocks();
1351
out:
1352
	up_write(&nfsi->rwsem);
L
Linus Torvalds 已提交
1353 1354 1355
	return status;
}

1356
static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369
{
	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.
	 */
1370 1371
restart:
	spin_lock(&sp->so_lock);
L
Linus Torvalds 已提交
1372
	list_for_each_entry(state, &sp->so_states, open_states) {
1373 1374
		if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
			continue;
L
Linus Torvalds 已提交
1375 1376
		if (state->state == 0)
			continue;
1377 1378
		atomic_inc(&state->count);
		spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1379 1380
		status = ops->recover_open(sp, state);
		if (status >= 0) {
1381 1382
			status = nfs4_reclaim_locks(state, ops);
			if (status >= 0) {
1383
				spin_lock(&state->state_lock);
1384
				list_for_each_entry(lock, &state->lock_states, ls_locks) {
1385
					if (!test_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags))
1386 1387
						pr_warn_ratelimited("NFS: "
							"%s: Lock reclaim "
1388
							"failed!\n", __func__);
1389
				}
1390
				spin_unlock(&state->state_lock);
1391 1392
				nfs4_put_open_state(state);
				goto restart;
L
Linus Torvalds 已提交
1393 1394 1395 1396
			}
		}
		switch (status) {
			default:
1397 1398
				printk(KERN_ERR "NFS: %s: unhandled error %d. "
					"Zeroing state\n", __func__, status);
L
Linus Torvalds 已提交
1399
			case -ENOENT:
1400
			case -ENOMEM:
1401
			case -ESTALE:
L
Linus Torvalds 已提交
1402 1403 1404 1405
				/*
				 * Open state on this file cannot be recovered
				 * All we can do is revert to using the zero stateid.
				 */
1406 1407
				memset(&state->stateid, 0,
					sizeof(state->stateid));
L
Linus Torvalds 已提交
1408 1409 1410
				/* Mark the file as being 'closed' */
				state->state = 0;
				break;
1411 1412 1413 1414 1415 1416 1417 1418
			case -EKEYEXPIRED:
				/*
				 * User RPCSEC_GSS context has expired.
				 * We cannot recover this stateid now, so
				 * skip it and allow recovery thread to
				 * proceed.
				 */
				break;
1419 1420 1421
			case -NFS4ERR_ADMIN_REVOKED:
			case -NFS4ERR_STALE_STATEID:
			case -NFS4ERR_BAD_STATEID:
1422 1423
			case -NFS4ERR_RECLAIM_BAD:
			case -NFS4ERR_RECLAIM_CONFLICT:
1424
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
1425
				break;
L
Linus Torvalds 已提交
1426 1427
			case -NFS4ERR_EXPIRED:
			case -NFS4ERR_NO_GRACE:
1428
				nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state);
L
Linus Torvalds 已提交
1429
			case -NFS4ERR_STALE_CLIENTID:
1430 1431 1432 1433
			case -NFS4ERR_BADSESSION:
			case -NFS4ERR_BADSLOT:
			case -NFS4ERR_BAD_HIGH_SLOT:
			case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
L
Linus Torvalds 已提交
1434 1435
				goto out_err;
		}
1436 1437
		nfs4_put_open_state(state);
		goto restart;
L
Linus Torvalds 已提交
1438
	}
1439
	spin_unlock(&sp->so_lock);
L
Linus Torvalds 已提交
1440 1441
	return 0;
out_err:
1442
	nfs4_put_open_state(state);
L
Linus Torvalds 已提交
1443 1444 1445
	return status;
}

1446 1447 1448 1449 1450 1451 1452 1453
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);
1454
	spin_lock(&state->state_lock);
1455 1456
	list_for_each_entry(lock, &state->lock_states, ls_locks) {
		lock->ls_seqid.flags = 0;
1457
		clear_bit(NFS_LOCK_INITIALIZED, &lock->ls_flags);
1458
	}
1459
	spin_unlock(&state->state_lock);
1460 1461
}

1462 1463
static void nfs4_reset_seqids(struct nfs_server *server,
	int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
1464
{
1465
	struct nfs_client *clp = server->nfs_client;
1466
	struct nfs4_state_owner *sp;
1467
	struct rb_node *pos;
1468 1469
	struct nfs4_state *state;

1470 1471 1472 1473 1474
	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);
1475
		sp->so_seqid.flags = 0;
1476
		spin_lock(&sp->so_lock);
1477
		list_for_each_entry(state, &sp->so_states, open_states) {
1478 1479
			if (mark_reclaim(clp, state))
				nfs4_clear_open_state(state);
1480
		}
1481
		spin_unlock(&sp->so_lock);
1482
	}
1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494
	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();
1495 1496
}

1497 1498 1499 1500 1501 1502 1503
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);
}

1504 1505 1506 1507 1508 1509 1510 1511
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);
}

1512
static void nfs4_clear_reclaim_server(struct nfs_server *server)
1513
{
1514
	struct nfs_client *clp = server->nfs_client;
1515 1516 1517 1518
	struct nfs4_state_owner *sp;
	struct rb_node *pos;
	struct nfs4_state *state;

1519 1520 1521 1522 1523
	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);
1524 1525
		spin_lock(&sp->so_lock);
		list_for_each_entry(state, &sp->so_states, open_states) {
1526 1527
			if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT,
						&state->flags))
1528 1529 1530 1531 1532
				continue;
			nfs4_state_mark_reclaim_nograce(clp, state);
		}
		spin_unlock(&sp->so_lock);
	}
1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546
	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();
1547 1548

	nfs_delegation_reap_unclaimed(clp);
1549 1550 1551 1552 1553 1554 1555 1556
	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);
1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570
}

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);
}

1571 1572 1573 1574 1575 1576 1577 1578
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);
}

1579
static int nfs4_recovery_handle_error(struct nfs_client *clp, int error)
1580 1581
{
	switch (error) {
1582 1583
		case 0:
			break;
1584
		case -NFS4ERR_CB_PATH_DOWN:
1585
			nfs40_handle_cb_pathdown(clp);
1586
			break;
1587 1588
		case -NFS4ERR_NO_GRACE:
			nfs4_state_end_reclaim_reboot(clp);
1589
			break;
1590 1591 1592
		case -NFS4ERR_STALE_CLIENTID:
		case -NFS4ERR_LEASE_MOVED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1593
			nfs4_state_clear_reclaim_reboot(clp);
1594 1595 1596 1597 1598
			nfs4_state_start_reclaim_reboot(clp);
			break;
		case -NFS4ERR_EXPIRED:
			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
			nfs4_state_start_reclaim_nograce(clp);
1599
			break;
1600 1601 1602 1603 1604 1605
		case -NFS4ERR_BADSESSION:
		case -NFS4ERR_BADSLOT:
		case -NFS4ERR_BAD_HIGH_SLOT:
		case -NFS4ERR_DEADSESSION:
		case -NFS4ERR_SEQ_FALSE_RETRY:
		case -NFS4ERR_SEQ_MISORDERED:
1606
			set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
1607
			/* Zero session reset errors */
1608
			break;
1609 1610 1611
		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
			set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
			break;
1612 1613 1614
		case -EKEYEXPIRED:
			/* Nothing we can do */
			nfs4_warn_keyexpired(clp->cl_hostname);
1615 1616
			break;
		default:
1617 1618
			dprintk("%s: failed to handle error %d for server %s\n",
					__func__, error, clp->cl_hostname);
1619
			return error;
1620
	}
1621 1622
	dprintk("%s: handled error %d for server %s\n", __func__, error,
			clp->cl_hostname);
1623
	return 0;
1624 1625
}

1626
static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
L
Linus Torvalds 已提交
1627
{
1628 1629
	struct nfs4_state_owner *sp;
	struct nfs_server *server;
1630
	struct rb_node *pos;
L
Linus Torvalds 已提交
1631 1632
	int status = 0;

1633
restart:
1634 1635
	rcu_read_lock();
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
1636
		nfs4_purge_state_owners(server);
1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656
		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);
			}

1657
			nfs4_put_state_owner(sp);
1658
			goto restart;
1659
		}
1660
		spin_unlock(&clp->cl_lock);
1661
	}
1662
	rcu_read_unlock();
1663 1664 1665 1666 1667 1668
	return status;
}

static int nfs4_check_lease(struct nfs_client *clp)
{
	struct rpc_cred *cred;
1669 1670
	const struct nfs4_state_maintenance_ops *ops =
		clp->cl_mvops->state_renewal_ops;
1671
	int status;
L
Linus Torvalds 已提交
1672

1673 1674 1675
	/* Is the client already known to have an expired lease? */
	if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
		return 0;
1676 1677 1678
	spin_lock(&clp->cl_lock);
	cred = ops->get_state_renewal_cred_locked(clp);
	spin_unlock(&clp->cl_lock);
1679 1680
	if (cred == NULL) {
		cred = nfs4_get_setclientid_cred(clp);
1681
		status = -ENOKEY;
1682 1683
		if (cred == NULL)
			goto out;
1684
	}
1685
	status = ops->renew_lease(clp, cred);
1686 1687
	put_rpccred(cred);
out:
1688
	return nfs4_recovery_handle_error(clp, status);
1689 1690
}

1691 1692
/* Set NFS4CLNT_LEASE_EXPIRED and reclaim reboot state for all v4.0 errors
 * and for recoverable errors on EXCHANGE_ID for v4.1
1693 1694 1695 1696
 */
static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status)
{
	switch (status) {
1697 1698 1699 1700 1701
	case -NFS4ERR_SEQ_MISORDERED:
		if (test_and_set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state))
			return -ESERVERFAULT;
		/* Lease confirmation error: retry after purging the lease */
		ssleep(1);
1702 1703
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
		break;
1704 1705
	case -NFS4ERR_STALE_CLIENTID:
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1706 1707
		nfs4_state_clear_reclaim_reboot(clp);
		nfs4_state_start_reclaim_reboot(clp);
1708
		break;
1709 1710 1711 1712 1713 1714
	case -NFS4ERR_CLID_INUSE:
		pr_err("NFS: Server %s reports our clientid is in use\n",
			clp->cl_hostname);
		nfs_mark_client_ready(clp, -EPERM);
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
		return -EPERM;
1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728
	case -EACCES:
		if (clp->cl_machine_cred == NULL)
			return -EACCES;
		/* Handle case where the user hasn't set up machine creds */
		nfs4_clear_machine_cred(clp);
	case -NFS4ERR_DELAY:
	case -ETIMEDOUT:
	case -EAGAIN:
		ssleep(1);
		break;

	case -NFS4ERR_MINOR_VERS_MISMATCH:
		if (clp->cl_cons_state == NFS_CS_SESSION_INITING)
			nfs_mark_client_ready(clp, -EPROTONOSUPPORT);
1729 1730
		dprintk("%s: exit with error %d for server %s\n",
				__func__, -EPROTONOSUPPORT, clp->cl_hostname);
1731 1732 1733 1734 1735 1736
		return -EPROTONOSUPPORT;
	case -EKEYEXPIRED:
		nfs4_warn_keyexpired(clp->cl_hostname);
	case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
				 * in nfs4_exchange_id */
	default:
1737 1738
		dprintk("%s: exit with error %d for server %s\n", __func__,
				status, clp->cl_hostname);
1739 1740 1741
		return status;
	}
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
1742 1743
	dprintk("%s: handled error %d for server %s\n", __func__, status,
			clp->cl_hostname);
1744 1745 1746
	return 0;
}

1747
static int nfs4_establish_lease(struct nfs_client *clp)
1748 1749
{
	struct rpc_cred *cred;
1750 1751
	const struct nfs4_state_recovery_ops *ops =
		clp->cl_mvops->reboot_recovery_ops;
1752
	int status;
1753

1754
	cred = ops->get_clid_cred(clp);
1755 1756 1757 1758 1759
	if (cred == NULL)
		return -ENOENT;
	status = ops->establish_clid(clp, cred);
	put_rpccred(cred);
	if (status != 0)
1760 1761 1762 1763 1764
		return status;
	pnfs_destroy_all_layouts(clp);
	return 0;
}

1765 1766 1767 1768
/*
 * Returns zero or a negative errno.  NFS4ERR values are converted
 * to local errno values.
 */
1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790
static int nfs4_reclaim_lease(struct nfs_client *clp)
{
	int status;

	status = nfs4_establish_lease(clp);
	if (status < 0)
		return nfs4_handle_reclaim_lease_error(clp, status);
	if (test_and_clear_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state))
		nfs4_state_start_reclaim_nograce(clp);
	if (!test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
		set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
	clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
	clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
	return 0;
}

static int nfs4_purge_lease(struct nfs_client *clp)
{
	int status;

	status = nfs4_establish_lease(clp);
	if (status < 0)
1791
		return nfs4_handle_reclaim_lease_error(clp, status);
1792 1793 1794
	clear_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
	nfs4_state_start_reclaim_nograce(clp);
1795
	return 0;
1796 1797
}

1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900
/**
 * nfs4_discover_server_trunking - Detect server IP address trunking
 *
 * @clp: nfs_client under test
 * @result: OUT: found nfs_client, or clp
 *
 * Returns zero or a negative errno.  If zero is returned,
 * an nfs_client pointer is planted in "result".
 *
 * Note: since we are invoked in process context, and
 * not from inside the state manager, we cannot use
 * nfs4_handle_reclaim_lease_error().
 */
int nfs4_discover_server_trunking(struct nfs_client *clp,
				  struct nfs_client **result)
{
	const struct nfs4_state_recovery_ops *ops =
				clp->cl_mvops->reboot_recovery_ops;
	rpc_authflavor_t *flavors, flav, save;
	struct rpc_clnt *clnt;
	struct rpc_cred *cred;
	int i, len, status;

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

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

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

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

	case -EACCES:
		if (clp->cl_machine_cred == NULL)
			break;
		/* Handle case where the user hasn't set up machine creds */
		nfs4_clear_machine_cred(clp);
	case -NFS4ERR_DELAY:
	case -ETIMEDOUT:
	case -EAGAIN:
		ssleep(1);
		dprintk("NFS: %s after status %d, retrying\n",
			__func__, status);
		goto again;

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

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

	case -NFS4ERR_MINOR_VERS_MISMATCH:
		status = -EPROTONOSUPPORT;
		break;

	case -EKEYEXPIRED:
		nfs4_warn_keyexpired(clp->cl_hostname);
	case -NFS4ERR_NOT_SAME: /* FixMe: implement recovery
				 * in nfs4_exchange_id */
		status = -EKEYEXPIRED;
	}

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

1901
#ifdef CONFIG_NFS_V4_1
1902
void nfs4_schedule_session_recovery(struct nfs4_session *session, int err)
1903
{
1904 1905
	struct nfs_client *clp = session->clp;

1906 1907 1908 1909 1910 1911 1912
	switch (err) {
	default:
		set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
		break;
	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
		set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
	}
1913
	nfs4_schedule_lease_recovery(clp);
1914
}
1915
EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery);
1916

1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928
static void nfs41_ping_server(struct nfs_client *clp)
{
	/* Use CHECK_LEASE to ping the server with a SEQUENCE */
	set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
	nfs4_schedule_state_manager(clp);
}

void nfs41_server_notify_target_slotid_update(struct nfs_client *clp)
{
	nfs41_ping_server(clp);
}

1929 1930 1931
static void nfs4_reset_all_state(struct nfs_client *clp)
{
	if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) {
1932
		set_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state);
1933
		clear_bit(NFS4CLNT_LEASE_CONFIRM, &clp->cl_state);
1934
		nfs4_state_start_reclaim_nograce(clp);
1935 1936
		dprintk("%s: scheduling reset of all state for server %s!\n",
				__func__, clp->cl_hostname);
1937
		nfs4_schedule_state_manager(clp);
1938 1939 1940 1941 1942 1943 1944
	}
}

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);
1945 1946
		dprintk("%s: server %s rebooted!\n", __func__,
				clp->cl_hostname);
1947
		nfs4_schedule_state_manager(clp);
1948 1949 1950 1951 1952 1953
	}
}

static void nfs41_handle_state_revoked(struct nfs_client *clp)
{
	nfs4_reset_all_state(clp);
1954
	dprintk("%s: state revoked on server %s\n", __func__, clp->cl_hostname);
1955 1956 1957 1958 1959 1960
}

static void nfs41_handle_recallable_state_revoked(struct nfs_client *clp)
{
	/* This will need to handle layouts too */
	nfs_expire_all_delegations(clp);
1961 1962
	dprintk("%s: Recallable state revoked on server %s!\n", __func__,
			clp->cl_hostname);
1963 1964
}

1965
static void nfs41_handle_backchannel_fault(struct nfs_client *clp)
1966 1967 1968
{
	nfs_expire_all_delegations(clp);
	if (test_and_set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) == 0)
1969
		nfs4_schedule_state_manager(clp);
1970 1971
	dprintk("%s: server %s declared a backchannel fault\n", __func__,
			clp->cl_hostname);
1972 1973
}

1974 1975 1976 1977 1978 1979 1980
static void nfs41_handle_cb_path_down(struct nfs_client *clp)
{
	if (test_and_set_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
		&clp->cl_state) == 0)
		nfs4_schedule_state_manager(clp);
}

1981 1982 1983 1984
void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags)
{
	if (!flags)
		return;
1985 1986 1987 1988

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

1989
	if (flags & SEQ4_STATUS_RESTART_RECLAIM_NEEDED)
1990
		nfs41_handle_server_reboot(clp);
1991
	if (flags & (SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED |
1992 1993
			    SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED |
			    SEQ4_STATUS_ADMIN_STATE_REVOKED |
1994 1995
			    SEQ4_STATUS_LEASE_MOVED))
		nfs41_handle_state_revoked(clp);
1996
	if (flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED)
1997
		nfs41_handle_recallable_state_revoked(clp);
1998 1999 2000 2001
	if (flags & SEQ4_STATUS_BACKCHANNEL_FAULT)
		nfs41_handle_backchannel_fault(clp);
	else if (flags & (SEQ4_STATUS_CB_PATH_DOWN |
				SEQ4_STATUS_CB_PATH_DOWN_SESSION))
2002
		nfs41_handle_cb_path_down(clp);
2003 2004
}

2005 2006
static int nfs4_reset_session(struct nfs_client *clp)
{
2007
	struct rpc_cred *cred;
2008 2009
	int status;

2010 2011
	if (!nfs4_has_session(clp))
		return 0;
2012
	nfs4_begin_drain_session(clp);
2013 2014
	cred = nfs4_get_exchange_id_cred(clp);
	status = nfs4_proc_destroy_session(clp->cl_session, cred);
2015 2016
	if (status && status != -NFS4ERR_BADSESSION &&
	    status != -NFS4ERR_DEADSESSION) {
2017
		status = nfs4_recovery_handle_error(clp, status);
2018 2019 2020 2021
		goto out;
	}

	memset(clp->cl_session->sess_id.data, 0, NFS4_MAX_SESSIONID_LEN);
2022
	status = nfs4_proc_create_session(clp, cred);
2023
	if (status) {
2024 2025
		dprintk("%s: session reset failed with status %d for server %s!\n",
			__func__, status, clp->cl_hostname);
2026
		status = nfs4_handle_reclaim_lease_error(clp, status);
2027 2028
		goto out;
	}
2029
	nfs41_finish_session_reset(clp);
2030 2031
	dprintk("%s: session reset was successful for server %s!\n",
			__func__, clp->cl_hostname);
2032
out:
2033 2034
	if (cred)
		put_rpccred(cred);
2035 2036
	return status;
}
2037

2038 2039
static int nfs4_bind_conn_to_session(struct nfs_client *clp)
{
2040 2041 2042
	struct rpc_cred *cred;
	int ret;

2043 2044
	if (!nfs4_has_session(clp))
		return 0;
2045
	nfs4_begin_drain_session(clp);
2046 2047 2048 2049
	cred = nfs4_get_exchange_id_cred(clp);
	ret = nfs4_proc_bind_conn_to_session(clp, cred);
	if (cred)
		put_rpccred(cred);
2050
	clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
2051 2052
	switch (ret) {
	case 0:
2053 2054
		dprintk("%s: bind_conn_to_session was successful for server %s!\n",
			__func__, clp->cl_hostname);
2055 2056 2057 2058 2059 2060 2061 2062 2063
		break;
	case -NFS4ERR_DELAY:
		ssleep(1);
		set_bit(NFS4CLNT_BIND_CONN_TO_SESSION, &clp->cl_state);
		break;
	default:
		return nfs4_recovery_handle_error(clp, ret);
	}
	return 0;
2064
}
2065
#else /* CONFIG_NFS_V4_1 */
2066
static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
2067
static int nfs4_end_drain_session(struct nfs_client *clp) { return 0; }
2068 2069 2070 2071 2072

static int nfs4_bind_conn_to_session(struct nfs_client *clp)
{
	return 0;
}
2073 2074
#endif /* CONFIG_NFS_V4_1 */

2075
static void nfs4_state_manager(struct nfs_client *clp)
2076 2077
{
	int status = 0;
2078
	const char *section = "", *section_sep = "";
2079 2080

	/* Ensure exclusive access to NFSv4 state */
2081
	do {
2082
		if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
2083
			section = "purge state";
2084
			status = nfs4_purge_lease(clp);
2085 2086
			if (status < 0)
				goto out_error;
2087
			continue;
2088 2089
		}

2090
		if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
2091
			section = "lease expired";
2092 2093
			/* We're going to have to re-establish a clientid */
			status = nfs4_reclaim_lease(clp);
2094
			if (status < 0)
2095
				goto out_error;
2096
			continue;
2097 2098
		}

2099
		/* Initialize or reset the session */
2100
		if (test_and_clear_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) {
2101
			section = "reset session";
2102
			status = nfs4_reset_session(clp);
2103 2104 2105
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
				continue;
			if (status < 0)
2106 2107
				goto out_error;
		}
2108

2109 2110
		/* Send BIND_CONN_TO_SESSION */
		if (test_and_clear_bit(NFS4CLNT_BIND_CONN_TO_SESSION,
2111
				&clp->cl_state)) {
2112
			section = "bind conn to session";
2113 2114 2115
			status = nfs4_bind_conn_to_session(clp);
			if (status < 0)
				goto out_error;
2116
			continue;
2117 2118
		}

2119 2120 2121 2122 2123 2124 2125 2126
		if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
			section = "check lease";
			status = nfs4_check_lease(clp);
			if (status < 0)
				goto out_error;
			continue;
		}

2127
		/* First recover reboot state... */
2128
		if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
2129
			section = "reclaim reboot";
2130
			status = nfs4_do_reclaim(clp,
2131
				clp->cl_mvops->reboot_recovery_ops);
2132
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2133
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
2134
				continue;
2135
			nfs4_state_end_reclaim_reboot(clp);
2136 2137 2138 2139
			if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state))
				continue;
			if (status < 0)
				goto out_error;
2140 2141
		}

2142 2143
		/* Now recover expired state... */
		if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
2144
			section = "reclaim nograce";
2145
			status = nfs4_do_reclaim(clp,
2146
				clp->cl_mvops->nograce_recovery_ops);
2147
			if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) ||
2148
			    test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) ||
2149 2150 2151
			    test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
				continue;
			if (status < 0)
2152
				goto out_error;
L
Linus Torvalds 已提交
2153
		}
2154

2155
		nfs4_end_drain_session(clp);
2156 2157 2158 2159
		if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
			nfs_client_return_marked_delegations(clp);
			continue;
		}
2160 2161

		nfs4_clear_state_manager_bit(clp);
2162 2163 2164 2165 2166
		/* 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;
2167
	} while (atomic_read(&clp->cl_count) > 1);
2168
	return;
L
Linus Torvalds 已提交
2169
out_error:
2170 2171 2172 2173 2174
	if (strlen(section))
		section_sep = ": ";
	pr_warn_ratelimited("NFS: state manager%s%s failed on NFSv4 server %s"
			" with error %d\n", section_sep, section,
			clp->cl_hostname, -status);
2175
	ssleep(1);
2176
	nfs4_end_drain_session(clp);
2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188
	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
2189 2190 2191 2192 2193 2194 2195
}

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