svclock.c 27 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
Linus Torvalds's avatar
Linus Torvalds committed
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
/*
 * linux/fs/lockd/svclock.c
 *
 * Handling of server-side locks, mostly of the blocked variety.
 * This is the ugliest part of lockd because we tread on very thin ice.
 * GRANT and CANCEL calls may get stuck, meet in mid-flight, etc.
 * IMNSHO introducing the grant callback into the NLM protocol was one
 * of the worst ideas Sun ever had. Except maybe for the idea of doing
 * NFS file locking at all.
 *
 * I'm trying hard to avoid race conditions by protecting most accesses
 * to a file's list of blocked locks through a semaphore. The global
 * list of blocked locks is not protected in this fashion however.
 * Therefore, some functions (such as the RPC callback for the async grant
 * call) move blocked locks towards the head of the list *while some other
 * process might be traversing it*. This should not be a problem in
 * practice, because this will only cause functions traversing the list
 * to visit some blocks twice.
 *
 * Copyright (C) 1996, Olaf Kirch <okir@monad.swb.de>
 */

#include <linux/types.h>
25
#include <linux/slab.h>
Linus Torvalds's avatar
Linus Torvalds committed
26 27 28 29
#include <linux/errno.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/sunrpc/clnt.h>
30
#include <linux/sunrpc/svc_xprt.h>
Linus Torvalds's avatar
Linus Torvalds committed
31 32
#include <linux/lockd/nlm.h>
#include <linux/lockd/lockd.h>
33
#include <linux/kthread.h>
Linus Torvalds's avatar
Linus Torvalds committed
34 35 36 37 38 39 40 41 42

#define NLMDBG_FACILITY		NLMDBG_SVCLOCK

#ifdef CONFIG_LOCKD_V4
#define nlm_deadlock	nlm4_deadlock
#else
#define nlm_deadlock	nlm_lck_denied
#endif

43
static void nlmsvc_release_block(struct nlm_block *block);
Linus Torvalds's avatar
Linus Torvalds committed
44
static void	nlmsvc_insert_block(struct nlm_block *block, unsigned long);
45
static void	nlmsvc_remove_block(struct nlm_block *block);
46

47 48
static int nlmsvc_setgrantargs(struct nlm_rqst *call, struct nlm_lock *lock);
static void nlmsvc_freegrantargs(struct nlm_rqst *call);
49
static const struct rpc_call_ops nlmsvc_grant_ops;
Linus Torvalds's avatar
Linus Torvalds committed
50 51 52 53

/*
 * The list of blocked locks to retry
 */
54
static LIST_HEAD(nlm_blocked);
55
static DEFINE_SPINLOCK(nlm_blocked_lock);
Linus Torvalds's avatar
Linus Torvalds committed
56

Jeff Layton's avatar
Jeff Layton committed
57
#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
58 59 60
static const char *nlmdbg_cookie2a(const struct nlm_cookie *cookie)
{
	/*
61 62
	 * We can get away with a static buffer because this is only called
	 * from lockd, which is single-threaded.
63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
	 */
	static char buf[2*NLM_MAXCOOKIELEN+1];
	unsigned int i, len = sizeof(buf);
	char *p = buf;

	len--;	/* allow for trailing \0 */
	if (len < 3)
		return "???";
	for (i = 0 ; i < cookie->len ; i++) {
		if (len < 2) {
			strcpy(p-3, "...");
			break;
		}
		sprintf(p, "%02x", cookie->data[i]);
		p += 2;
		len -= 2;
	}
	*p = '\0';

	return buf;
}
#endif

Linus Torvalds's avatar
Linus Torvalds committed
86 87 88 89
/*
 * Insert a blocked lock into the global list
 */
static void
90
nlmsvc_insert_block_locked(struct nlm_block *block, unsigned long when)
Linus Torvalds's avatar
Linus Torvalds committed
91
{
92 93
	struct nlm_block *b;
	struct list_head *pos;
Linus Torvalds's avatar
Linus Torvalds committed
94 95

	dprintk("lockd: nlmsvc_insert_block(%p, %ld)\n", block, when);
96 97 98 99 100 101 102
	if (list_empty(&block->b_list)) {
		kref_get(&block->b_count);
	} else {
		list_del_init(&block->b_list);
	}

	pos = &nlm_blocked;
Linus Torvalds's avatar
Linus Torvalds committed
103 104 105
	if (when != NLM_NEVER) {
		if ((when += jiffies) == NLM_NEVER)
			when ++;
106 107 108 109 110 111 112 113 114
		list_for_each(pos, &nlm_blocked) {
			b = list_entry(pos, struct nlm_block, b_list);
			if (time_after(b->b_when,when) || b->b_when == NLM_NEVER)
				break;
		}
		/* On normal exit from the loop, pos == &nlm_blocked,
		 * so we will be adding to the end of the list - good
		 */
	}
Linus Torvalds's avatar
Linus Torvalds committed
115

116
	list_add_tail(&block->b_list, pos);
Linus Torvalds's avatar
Linus Torvalds committed
117 118 119
	block->b_when = when;
}

120 121 122 123 124 125 126
static void nlmsvc_insert_block(struct nlm_block *block, unsigned long when)
{
	spin_lock(&nlm_blocked_lock);
	nlmsvc_insert_block_locked(block, when);
	spin_unlock(&nlm_blocked_lock);
}

Linus Torvalds's avatar
Linus Torvalds committed
127 128 129
/*
 * Remove a block from the global list
 */
130
static inline void
Linus Torvalds's avatar
Linus Torvalds committed
131 132
nlmsvc_remove_block(struct nlm_block *block)
{
133
	if (!list_empty(&block->b_list)) {
134
		spin_lock(&nlm_blocked_lock);
135
		list_del_init(&block->b_list);
136
		spin_unlock(&nlm_blocked_lock);
137
		nlmsvc_release_block(block);
Linus Torvalds's avatar
Linus Torvalds committed
138 139 140 141
	}
}

/*
142
 * Find a block for a given lock
Linus Torvalds's avatar
Linus Torvalds committed
143 144
 */
static struct nlm_block *
145
nlmsvc_lookup_block(struct nlm_file *file, struct nlm_lock *lock)
Linus Torvalds's avatar
Linus Torvalds committed
146
{
147
	struct nlm_block	*block;
Linus Torvalds's avatar
Linus Torvalds committed
148 149 150 151 152 153
	struct file_lock	*fl;

	dprintk("lockd: nlmsvc_lookup_block f=%p pd=%d %Ld-%Ld ty=%d\n",
				file, lock->fl.fl_pid,
				(long long)lock->fl.fl_start,
				(long long)lock->fl.fl_end, lock->fl.fl_type);
154
	list_for_each_entry(block, &nlm_blocked, b_list) {
155
		fl = &block->b_call->a_args.lock.fl;
Linus Torvalds's avatar
Linus Torvalds committed
156 157 158 159
		dprintk("lockd: check f=%p pd=%d %Ld-%Ld ty=%d cookie=%s\n",
				block->b_file, fl->fl_pid,
				(long long)fl->fl_start,
				(long long)fl->fl_end, fl->fl_type,
160
				nlmdbg_cookie2a(&block->b_call->a_args.cookie));
Linus Torvalds's avatar
Linus Torvalds committed
161
		if (block->b_file == file && nlm_compare_locks(fl, &lock->fl)) {
162
			kref_get(&block->b_count);
Linus Torvalds's avatar
Linus Torvalds committed
163 164 165 166 167 168 169 170 171
			return block;
		}
	}

	return NULL;
}

static inline int nlm_cookie_match(struct nlm_cookie *a, struct nlm_cookie *b)
{
172
	if (a->len != b->len)
Linus Torvalds's avatar
Linus Torvalds committed
173
		return 0;
174
	if (memcmp(a->data, b->data, a->len))
Linus Torvalds's avatar
Linus Torvalds committed
175 176 177 178 179 180 181 182
		return 0;
	return 1;
}

/*
 * Find a block with a given NLM cookie.
 */
static inline struct nlm_block *
183
nlmsvc_find_block(struct nlm_cookie *cookie)
Linus Torvalds's avatar
Linus Torvalds committed
184 185 186
{
	struct nlm_block *block;

187
	list_for_each_entry(block, &nlm_blocked, b_list) {
188
		if (nlm_cookie_match(&block->b_call->a_args.cookie,cookie))
189
			goto found;
Linus Torvalds's avatar
Linus Torvalds committed
190 191
	}

192 193 194
	return NULL;

found:
195
	dprintk("nlmsvc_find_block(%s): block=%p\n", nlmdbg_cookie2a(cookie), block);
196
	kref_get(&block->b_count);
Linus Torvalds's avatar
Linus Torvalds committed
197 198 199 200 201 202 203 204 205 206 207 208
	return block;
}

/*
 * Create a block and initialize it.
 *
 * Note: we explicitly set the cookie of the grant reply to that of
 * the blocked lock request. The spec explicitly mentions that the client
 * should _not_ rely on the callback containing the same cookie as the
 * request, but (as I found out later) that's because some implementations
 * do just this. Never mind the standards comittees, they support our
 * logging industries.
209 210 211 212 213
 *
 * 10 years later: I hope we can safely ignore these old and broken
 * clients by now. Let's fix this so we can uniquely identify an incoming
 * GRANTED_RES message by cookie, without having to rely on the client's IP
 * address. --okir
Linus Torvalds's avatar
Linus Torvalds committed
214
 */
215 216 217 218
static struct nlm_block *
nlmsvc_create_block(struct svc_rqst *rqstp, struct nlm_host *host,
		    struct nlm_file *file, struct nlm_lock *lock,
		    struct nlm_cookie *cookie)
Linus Torvalds's avatar
Linus Torvalds committed
219 220
{
	struct nlm_block	*block;
221
	struct nlm_rqst		*call = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
222

223 224 225 226
	call = nlm_alloc_call(host);
	if (call == NULL)
		return NULL;

Linus Torvalds's avatar
Linus Torvalds committed
227
	/* Allocate memory for block, and initialize arguments */
228 229
	block = kzalloc(sizeof(*block), GFP_KERNEL);
	if (block == NULL)
Linus Torvalds's avatar
Linus Torvalds committed
230
		goto failed;
231
	kref_init(&block->b_count);
232 233
	INIT_LIST_HEAD(&block->b_list);
	INIT_LIST_HEAD(&block->b_flist);
Linus Torvalds's avatar
Linus Torvalds committed
234

235
	if (!nlmsvc_setgrantargs(call, lock))
Linus Torvalds's avatar
Linus Torvalds committed
236 237 238
		goto failed_free;

	/* Set notifier function for VFS, and init args */
239 240
	call->a_args.lock.fl.fl_flags |= FL_SLEEP;
	call->a_args.lock.fl.fl_lmops = &nlmsvc_lock_operations;
241
	nlmclnt_next_cookie(&call->a_args.cookie);
Linus Torvalds's avatar
Linus Torvalds committed
242 243 244 245 246 247 248

	dprintk("lockd: created block %p...\n", block);

	/* Create and initialize the block */
	block->b_daemon = rqstp->rq_server;
	block->b_host   = host;
	block->b_file   = file;
249
	file->f_count++;
Linus Torvalds's avatar
Linus Torvalds committed
250 251

	/* Add to file's list of blocks */
252
	list_add(&block->b_flist, &file->f_blocks);
Linus Torvalds's avatar
Linus Torvalds committed
253 254

	/* Set up RPC arguments for callback */
255
	block->b_call = call;
Linus Torvalds's avatar
Linus Torvalds committed
256
	call->a_flags   = RPC_TASK_ASYNC;
257
	call->a_block = block;
Linus Torvalds's avatar
Linus Torvalds committed
258 259 260 261 262 263

	return block;

failed_free:
	kfree(block);
failed:
Chuck Lever's avatar
Chuck Lever committed
264
	nlmsvc_release_call(call);
Linus Torvalds's avatar
Linus Torvalds committed
265 266 267 268
	return NULL;
}

/*
269
 * Delete a block.
Linus Torvalds's avatar
Linus Torvalds committed
270 271 272
 * It is the caller's responsibility to check whether the file
 * can be closed hereafter.
 */
273
static int nlmsvc_unlink_block(struct nlm_block *block)
Linus Torvalds's avatar
Linus Torvalds committed
274
{
275
	int status;
276
	dprintk("lockd: unlinking block %p...\n", block);
Linus Torvalds's avatar
Linus Torvalds committed
277 278

	/* Remove block from list */
279
	status = locks_delete_block(&block->b_call->a_args.lock.fl);
Linus Torvalds's avatar
Linus Torvalds committed
280
	nlmsvc_remove_block(block);
281 282
	return status;
}
Linus Torvalds's avatar
Linus Torvalds committed
283

284 285 286 287 288 289
static void nlmsvc_free_block(struct kref *kref)
{
	struct nlm_block *block = container_of(kref, struct nlm_block, b_count);
	struct nlm_file		*file = block->b_file;

	dprintk("lockd: freeing block %p...\n", block);
Linus Torvalds's avatar
Linus Torvalds committed
290 291

	/* Remove block from file's list of blocks */
292
	list_del_init(&block->b_flist);
293
	mutex_unlock(&file->f_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
294

295
	nlmsvc_freegrantargs(block->b_call);
Chuck Lever's avatar
Chuck Lever committed
296
	nlmsvc_release_call(block->b_call);
297
	nlm_release_file(block->b_file);
Linus Torvalds's avatar
Linus Torvalds committed
298
	kfree(block);
299 300 301 302 303
}

static void nlmsvc_release_block(struct nlm_block *block)
{
	if (block != NULL)
304
		kref_put_mutex(&block->b_count, nlmsvc_free_block, &block->b_file->f_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
305 306
}

307 308 309 310 311 312 313
/*
 * Loop over all blocks and delete blocks held by
 * a matching host.
 */
void nlmsvc_traverse_blocks(struct nlm_host *host,
			struct nlm_file *file,
			nlm_host_match_fn_t match)
314
{
315
	struct nlm_block *block, *next;
316 317

restart:
318
	mutex_lock(&file->f_mutex);
319
	list_for_each_entry_safe(block, next, &file->f_blocks, b_flist) {
320
		if (!match(block->b_host, host))
321
			continue;
322 323 324
		/* Do not destroy blocks that are not on
		 * the global retry list - why? */
		if (list_empty(&block->b_list))
325 326
			continue;
		kref_get(&block->b_count);
327
		mutex_unlock(&file->f_mutex);
328 329 330 331
		nlmsvc_unlink_block(block);
		nlmsvc_release_block(block);
		goto restart;
	}
332
	mutex_unlock(&file->f_mutex);
333 334
}

335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408
static struct nlm_lockowner *
nlmsvc_get_lockowner(struct nlm_lockowner *lockowner)
{
	refcount_inc(&lockowner->count);
	return lockowner;
}

static void nlmsvc_put_lockowner(struct nlm_lockowner *lockowner)
{
	if (!refcount_dec_and_lock(&lockowner->count, &lockowner->host->h_lock))
		return;
	list_del(&lockowner->list);
	spin_unlock(&lockowner->host->h_lock);
	nlmsvc_release_host(lockowner->host);
	kfree(lockowner);
}

static struct nlm_lockowner *__nlmsvc_find_lockowner(struct nlm_host *host, pid_t pid)
{
	struct nlm_lockowner *lockowner;
	list_for_each_entry(lockowner, &host->h_lockowners, list) {
		if (lockowner->pid != pid)
			continue;
		return nlmsvc_get_lockowner(lockowner);
	}
	return NULL;
}

static struct nlm_lockowner *nlmsvc_find_lockowner(struct nlm_host *host, pid_t pid)
{
	struct nlm_lockowner *res, *new = NULL;

	spin_lock(&host->h_lock);
	res = __nlmsvc_find_lockowner(host, pid);

	if (res == NULL) {
		spin_unlock(&host->h_lock);
		new = kmalloc(sizeof(*res), GFP_KERNEL);
		spin_lock(&host->h_lock);
		res = __nlmsvc_find_lockowner(host, pid);
		if (res == NULL && new != NULL) {
			res = new;
			/* fs/locks.c will manage the refcount through lock_ops */
			refcount_set(&new->count, 1);
			new->pid = pid;
			new->host = nlm_get_host(host);
			list_add(&new->list, &host->h_lockowners);
			new = NULL;
		}
	}

	spin_unlock(&host->h_lock);
	kfree(new);
	return res;
}

void
nlmsvc_release_lockowner(struct nlm_lock *lock)
{
	if (lock->fl.fl_owner)
		nlmsvc_put_lockowner(lock->fl.fl_owner);
}

static void nlmsvc_locks_copy_lock(struct file_lock *new, struct file_lock *fl)
{
	struct nlm_lockowner *nlm_lo = (struct nlm_lockowner *)fl->fl_owner;
	new->fl_owner = nlmsvc_get_lockowner(nlm_lo);
}

static void nlmsvc_locks_release_private(struct file_lock *fl)
{
	nlmsvc_put_lockowner((struct nlm_lockowner *)fl->fl_owner);
}

YueHaibing's avatar
YueHaibing committed
409
static const struct file_lock_operations nlmsvc_lock_ops = {
410 411 412 413 414 415 416 417 418 419 420 421
	.fl_copy_lock = nlmsvc_locks_copy_lock,
	.fl_release_private = nlmsvc_locks_release_private,
};

void nlmsvc_locks_init_private(struct file_lock *fl, struct nlm_host *host,
						pid_t pid)
{
	fl->fl_owner = nlmsvc_find_lockowner(host, pid);
	if (fl->fl_owner != NULL)
		fl->fl_ops = &nlmsvc_lock_ops;
}

422 423 424 425 426 427 428 429
/*
 * Initialize arguments for GRANTED call. The nlm_rqst structure
 * has been cleared already.
 */
static int nlmsvc_setgrantargs(struct nlm_rqst *call, struct nlm_lock *lock)
{
	locks_copy_lock(&call->a_args.lock.fl, &lock->fl);
	memcpy(&call->a_args.lock.fh, &lock->fh, sizeof(call->a_args.lock.fh));
430
	call->a_args.lock.caller = utsname()->nodename;
431 432 433 434
	call->a_args.lock.oh.len = lock->oh.len;

	/* set default data area */
	call->a_args.lock.oh.data = call->a_owner;
435
	call->a_args.lock.svid = ((struct nlm_lockowner *)lock->fl.fl_owner)->pid;
436 437 438

	if (lock->oh.len > NLMCLNT_OHSIZE) {
		void *data = kmalloc(lock->oh.len, GFP_KERNEL);
439
		if (!data)
440 441 442 443 444 445 446 447 448 449
			return 0;
		call->a_args.lock.oh.data = (u8 *) data;
	}

	memcpy(call->a_args.lock.oh.data, lock->oh.data, lock->oh.len);
	return 1;
}

static void nlmsvc_freegrantargs(struct nlm_rqst *call)
{
450
	if (call->a_args.lock.oh.data != call->a_owner)
451
		kfree(call->a_args.lock.oh.data);
452 453

	locks_release_private(&call->a_args.lock.fl);
454 455
}

456 457 458
/*
 * Deferred lock request handling for non-blocking lock
 */
459
static __be32
460 461
nlmsvc_defer_lock_rqst(struct svc_rqst *rqstp, struct nlm_block *block)
{
462
	__be32 status = nlm_lck_denied_nolocks;
463 464 465 466 467 468 469 470 471 472 473 474 475

	block->b_flags |= B_QUEUED;

	nlmsvc_insert_block(block, NLM_TIMEOUT);

	block->b_cache_req = &rqstp->rq_chandle;
	if (rqstp->rq_chandle.defer) {
		block->b_deferred_req =
			rqstp->rq_chandle.defer(block->b_cache_req);
		if (block->b_deferred_req != NULL)
			status = nlm_drop_reply;
	}
	dprintk("lockd: nlmsvc_defer_lock_rqst block %p flags %d status %d\n",
476
		block, block->b_flags, ntohl(status));
477 478 479 480

	return status;
}

Linus Torvalds's avatar
Linus Torvalds committed
481 482 483 484
/*
 * Attempt to establish a lock, and if it can't be granted, block it
 * if required.
 */
Al Viro's avatar
Al Viro committed
485
__be32
Linus Torvalds's avatar
Linus Torvalds committed
486
nlmsvc_lock(struct svc_rqst *rqstp, struct nlm_file *file,
487
	    struct nlm_host *host, struct nlm_lock *lock, int wait,
488
	    struct nlm_cookie *cookie, int reclaim)
Linus Torvalds's avatar
Linus Torvalds committed
489
{
490
	struct nlm_block	*block = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
491
	int			error;
Al Viro's avatar
Al Viro committed
492
	__be32			ret;
Linus Torvalds's avatar
Linus Torvalds committed
493 494

	dprintk("lockd: nlmsvc_lock(%s/%ld, ty=%d, pi=%d, %Ld-%Ld, bl=%d)\n",
495 496
				locks_inode(file->f_file)->i_sb->s_id,
				locks_inode(file->f_file)->i_ino,
Linus Torvalds's avatar
Linus Torvalds committed
497 498 499 500 501 502
				lock->fl.fl_type, lock->fl.fl_pid,
				(long long)lock->fl.fl_start,
				(long long)lock->fl.fl_end,
				wait);

	/* Lock file against concurrent access */
503
	mutex_lock(&file->f_mutex);
504 505 506
	/* Get existing block (in case client is busy-waiting)
	 * or create new block
	 */
507
	block = nlmsvc_lookup_block(file, lock);
508
	if (block == NULL) {
509
		block = nlmsvc_create_block(rqstp, host, file, lock, cookie);
510 511 512
		ret = nlm_lck_denied_nolocks;
		if (block == NULL)
			goto out;
513
		lock = &block->b_call->a_args.lock;
514 515
	} else
		lock->fl.fl_flags &= ~FL_SLEEP;
Linus Torvalds's avatar
Linus Torvalds committed
516

517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532
	if (block->b_flags & B_QUEUED) {
		dprintk("lockd: nlmsvc_lock deferred block %p flags %d\n",
							block, block->b_flags);
		if (block->b_granted) {
			nlmsvc_unlink_block(block);
			ret = nlm_granted;
			goto out;
		}
		if (block->b_flags & B_TIMED_OUT) {
			nlmsvc_unlink_block(block);
			ret = nlm_lck_denied;
			goto out;
		}
		ret = nlm_drop_reply;
		goto out;
	}
Linus Torvalds's avatar
Linus Torvalds committed
533

534
	if (locks_in_grace(SVC_NET(rqstp)) && !reclaim) {
535 536 537
		ret = nlm_lck_denied_grace_period;
		goto out;
	}
538
	if (reclaim && !locks_in_grace(SVC_NET(rqstp))) {
539 540 541
		ret = nlm_lck_denied_grace_period;
		goto out;
	}
542

543 544 545 546
	if (!wait)
		lock->fl.fl_flags &= ~FL_SLEEP;
	error = vfs_lock_file(file->f_file, F_SETLK, &lock->fl, NULL);
	lock->fl.fl_flags &= ~FL_SLEEP;
547

548
	dprintk("lockd: vfs_lock_file returned %d\n", error);
549
	switch (error) {
Linus Torvalds's avatar
Linus Torvalds committed
550
		case 0:
Andy Adamson's avatar
Andy Adamson committed
551 552
			ret = nlm_granted;
			goto out;
553
		case -EAGAIN:
554 555 556 557 558 559 560
			/*
			 * If this is a blocking request for an
			 * already pending lock request then we need
			 * to put it back on lockd's block list
			 */
			if (wait)
				break;
561
			ret = nlm_lck_denied;
562
			goto out;
563
		case FILE_LOCK_DEFERRED:
564 565 566 567 568 569
			if (wait)
				break;
			/* Filesystem lock operation is in progress
			   Add it to the queue waiting for callback */
			ret = nlmsvc_defer_lock_rqst(rqstp, block);
			goto out;
570
		case -EDEADLK:
Andy Adamson's avatar
Andy Adamson committed
571 572
			ret = nlm_deadlock;
			goto out;
Linus Torvalds's avatar
Linus Torvalds committed
573
		default:			/* includes ENOLCK */
Andy Adamson's avatar
Andy Adamson committed
574 575
			ret = nlm_lck_denied_nolocks;
			goto out;
Linus Torvalds's avatar
Linus Torvalds committed
576 577
	}

578
	ret = nlm_lck_blocked;
Linus Torvalds's avatar
Linus Torvalds committed
579 580

	/* Append to list of blocked */
581
	nlmsvc_insert_block(block, NLM_NEVER);
Andy Adamson's avatar
Andy Adamson committed
582
out:
583
	mutex_unlock(&file->f_mutex);
584
	nlmsvc_release_block(block);
Andy Adamson's avatar
Andy Adamson committed
585 586
	dprintk("lockd: nlmsvc_lock returned %u\n", ret);
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
587 588 589 590 591
}

/*
 * Test for presence of a conflicting lock.
 */
Al Viro's avatar
Al Viro committed
592
__be32
593
nlmsvc_testlock(struct svc_rqst *rqstp, struct nlm_file *file,
594 595
		struct nlm_host *host, struct nlm_lock *lock,
		struct nlm_lock *conflock, struct nlm_cookie *cookie)
Linus Torvalds's avatar
Linus Torvalds committed
596
{
Marc Eshel's avatar
Marc Eshel committed
597 598
	int			error;
	__be32			ret;
599
	struct nlm_lockowner	*test_owner;
Marc Eshel's avatar
Marc Eshel committed
600

Linus Torvalds's avatar
Linus Torvalds committed
601
	dprintk("lockd: nlmsvc_testlock(%s/%ld, ty=%d, %Ld-%Ld)\n",
602 603
				locks_inode(file->f_file)->i_sb->s_id,
				locks_inode(file->f_file)->i_ino,
Linus Torvalds's avatar
Linus Torvalds committed
604 605 606 607
				lock->fl.fl_type,
				(long long)lock->fl.fl_start,
				(long long)lock->fl.fl_end);

608
	if (locks_in_grace(SVC_NET(rqstp))) {
609 610 611
		ret = nlm_lck_denied_grace_period;
		goto out;
	}
612

613 614 615
	/* If there's a conflicting lock, remember to clean up the test lock */
	test_owner = (struct nlm_lockowner *)lock->fl.fl_owner;

Marc Eshel's avatar
Marc Eshel committed
616 617
	error = vfs_test_lock(file->f_file, &lock->fl);
	if (error) {
618 619 620 621
		/* We can't currently deal with deferred test requests */
		if (error == FILE_LOCK_DEFERRED)
			WARN_ON_ONCE(1);

Marc Eshel's avatar
Marc Eshel committed
622 623 624
		ret = nlm_lck_denied_nolocks;
		goto out;
	}
625

Marc Eshel's avatar
Marc Eshel committed
626 627 628 629 630 631 632 633 634 635 636
	if (lock->fl.fl_type == F_UNLCK) {
		ret = nlm_granted;
		goto out;
	}

	dprintk("lockd: conflicting lock(ty=%d, %Ld-%Ld)\n",
		lock->fl.fl_type, (long long)lock->fl.fl_start,
		(long long)lock->fl.fl_end);
	conflock->caller = "somehost";	/* FIXME */
	conflock->len = strlen(conflock->caller);
	conflock->oh.len = 0;		/* don't return OH info */
637
	conflock->svid = ((struct nlm_lockowner *)lock->fl.fl_owner)->pid;
Marc Eshel's avatar
Marc Eshel committed
638 639 640
	conflock->fl.fl_type = lock->fl.fl_type;
	conflock->fl.fl_start = lock->fl.fl_start;
	conflock->fl.fl_end = lock->fl.fl_end;
641
	locks_release_private(&lock->fl);
642 643 644 645 646

	/* Clean up the test lock */
	lock->fl.fl_owner = NULL;
	nlmsvc_put_lockowner(test_owner);

Marc Eshel's avatar
Marc Eshel committed
647 648 649
	ret = nlm_lck_denied;
out:
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
650 651 652 653 654 655 656 657 658
}

/*
 * Remove a lock.
 * This implies a CANCEL call: We send a GRANT_MSG, the client replies
 * with a GRANT_RES call which gets lost, and calls UNLOCK immediately
 * afterwards. In this case the block will still be there, and hence
 * must be removed.
 */
Al Viro's avatar
Al Viro committed
659
__be32
660
nlmsvc_unlock(struct net *net, struct nlm_file *file, struct nlm_lock *lock)
Linus Torvalds's avatar
Linus Torvalds committed
661 662 663 664
{
	int	error;

	dprintk("lockd: nlmsvc_unlock(%s/%ld, pi=%d, %Ld-%Ld)\n",
665 666
				locks_inode(file->f_file)->i_sb->s_id,
				locks_inode(file->f_file)->i_ino,
Linus Torvalds's avatar
Linus Torvalds committed
667 668 669 670 671
				lock->fl.fl_pid,
				(long long)lock->fl.fl_start,
				(long long)lock->fl.fl_end);

	/* First, cancel any lock that might be there */
672
	nlmsvc_cancel_blocked(net, file, lock);
Linus Torvalds's avatar
Linus Torvalds committed
673 674

	lock->fl.fl_type = F_UNLCK;
675
	error = vfs_lock_file(file->f_file, F_SETLK, &lock->fl, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
676 677 678 679 680 681 682 683 684 685 686

	return (error < 0)? nlm_lck_denied_nolocks : nlm_granted;
}

/*
 * Cancel a previously blocked request.
 *
 * A cancel request always overrides any grant that may currently
 * be in progress.
 * The calling procedure must check whether the file can be closed.
 */
Al Viro's avatar
Al Viro committed
687
__be32
688
nlmsvc_cancel_blocked(struct net *net, struct nlm_file *file, struct nlm_lock *lock)
Linus Torvalds's avatar
Linus Torvalds committed
689 690
{
	struct nlm_block	*block;
J. Bruce Fields's avatar
J. Bruce Fields committed
691
	int status = 0;
Linus Torvalds's avatar
Linus Torvalds committed
692 693

	dprintk("lockd: nlmsvc_cancel(%s/%ld, pi=%d, %Ld-%Ld)\n",
694 695
				locks_inode(file->f_file)->i_sb->s_id,
				locks_inode(file->f_file)->i_ino,
Linus Torvalds's avatar
Linus Torvalds committed
696 697 698 699
				lock->fl.fl_pid,
				(long long)lock->fl.fl_start,
				(long long)lock->fl.fl_end);

700
	if (locks_in_grace(net))
701 702
		return nlm_lck_denied_grace_period;

703
	mutex_lock(&file->f_mutex);
704
	block = nlmsvc_lookup_block(file, lock);
705
	mutex_unlock(&file->f_mutex);
706
	if (block != NULL) {
707 708
		vfs_cancel_lock(block->b_file->f_file,
				&block->b_call->a_args.lock.fl);
709 710 711
		status = nlmsvc_unlink_block(block);
		nlmsvc_release_block(block);
	}
J. Bruce Fields's avatar
J. Bruce Fields committed
712
	return status ? nlm_lck_denied : nlm_granted;
Linus Torvalds's avatar
Linus Torvalds committed
713 714
}

Marc Eshel's avatar
Marc Eshel committed
715 716
/*
 * This is a callback from the filesystem for VFS file lock requests.
717
 * It will be used if lm_grant is defined and the filesystem can not
Marc Eshel's avatar
Marc Eshel committed
718 719 720 721 722 723 724
 * respond to the request immediately.
 * For SETLK or SETLKW request it will get the local posix lock.
 * In all cases it will move the block to the head of nlm_blocked q where
 * nlmsvc_retry_blocked() can send back a reply for SETLKW or revisit the
 * deferred rpc for GETLK and SETLK.
 */
static void
725
nlmsvc_update_deferred_block(struct nlm_block *block, int result)
Marc Eshel's avatar
Marc Eshel committed
726 727 728 729 730 731 732 733
{
	block->b_flags |= B_GOT_CALLBACK;
	if (result == 0)
		block->b_granted = 1;
	else
		block->b_flags |= B_TIMED_OUT;
}

734
static int nlmsvc_grant_deferred(struct file_lock *fl, int result)
Marc Eshel's avatar
Marc Eshel committed
735 736 737 738
{
	struct nlm_block *block;
	int rc = -ENOENT;

739
	spin_lock(&nlm_blocked_lock);
Marc Eshel's avatar
Marc Eshel committed
740 741 742 743 744 745 746 747 748
	list_for_each_entry(block, &nlm_blocked, b_list) {
		if (nlm_compare_locks(&block->b_call->a_args.lock.fl, fl)) {
			dprintk("lockd: nlmsvc_notify_blocked block %p flags %d\n",
							block, block->b_flags);
			if (block->b_flags & B_QUEUED) {
				if (block->b_flags & B_TIMED_OUT) {
					rc = -ENOLCK;
					break;
				}
749
				nlmsvc_update_deferred_block(block, result);
Marc Eshel's avatar
Marc Eshel committed
750 751 752
			} else if (result == 0)
				block->b_granted = 1;

753
			nlmsvc_insert_block_locked(block, 0);
Marc Eshel's avatar
Marc Eshel committed
754 755 756 757 758
			svc_wake_up(block->b_daemon);
			rc = 0;
			break;
		}
	}
759
	spin_unlock(&nlm_blocked_lock);
Marc Eshel's avatar
Marc Eshel committed
760 761 762 763 764
	if (rc == -ENOENT)
		printk(KERN_WARNING "lockd: grant for unknown block\n");
	return rc;
}

Linus Torvalds's avatar
Linus Torvalds committed
765 766 767 768 769 770 771 772 773 774
/*
 * Unblock a blocked lock request. This is a callback invoked from the
 * VFS layer when a lock on which we blocked is removed.
 *
 * This function doesn't grant the blocked lock instantly, but rather moves
 * the block to the head of nlm_blocked where it can be picked up by lockd.
 */
static void
nlmsvc_notify_blocked(struct file_lock *fl)
{
775
	struct nlm_block	*block;
Linus Torvalds's avatar
Linus Torvalds committed
776 777

	dprintk("lockd: VFS unblock notification for block %p\n", fl);
778
	spin_lock(&nlm_blocked_lock);
779
	list_for_each_entry(block, &nlm_blocked, b_list) {
780
		if (nlm_compare_locks(&block->b_call->a_args.lock.fl, fl)) {
781 782
			nlmsvc_insert_block_locked(block, 0);
			spin_unlock(&nlm_blocked_lock);
Linus Torvalds's avatar
Linus Torvalds committed
783 784 785 786
			svc_wake_up(block->b_daemon);
			return;
		}
	}
787
	spin_unlock(&nlm_blocked_lock);
Linus Torvalds's avatar
Linus Torvalds committed
788 789 790
	printk(KERN_WARNING "lockd: notification for unknown block!\n");
}

791
const struct lock_manager_operations nlmsvc_lock_operations = {
792 793
	.lm_notify = nlmsvc_notify_blocked,
	.lm_grant = nlmsvc_grant_deferred,
Linus Torvalds's avatar
Linus Torvalds committed
794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810
};

/*
 * Try to claim a lock that was previously blocked.
 *
 * Note that we use both the RPC_GRANTED_MSG call _and_ an async
 * RPC thread when notifying the client. This seems like overkill...
 * Here's why:
 *  -	we don't want to use a synchronous RPC thread, otherwise
 *	we might find ourselves hanging on a dead portmapper.
 *  -	Some lockd implementations (e.g. HP) don't react to
 *	RPC_GRANTED calls; they seem to insist on RPC_GRANTED_MSG calls.
 */
static void
nlmsvc_grant_blocked(struct nlm_block *block)
{
	struct nlm_file		*file = block->b_file;
811
	struct nlm_lock		*lock = &block->b_call->a_args.lock;
Linus Torvalds's avatar
Linus Torvalds committed
812
	int			error;
813
	loff_t			fl_start, fl_end;
Linus Torvalds's avatar
Linus Torvalds committed
814 815 816

	dprintk("lockd: grant blocked lock %p\n", block);

Marc Eshel's avatar
Marc Eshel committed
817 818
	kref_get(&block->b_count);

Linus Torvalds's avatar
Linus Torvalds committed
819
	/* Unlink block request from list */
820
	nlmsvc_unlink_block(block);
Linus Torvalds's avatar
Linus Torvalds committed
821 822 823 824 825 826 827 828 829 830

	/* If b_granted is true this means we've been here before.
	 * Just retry the grant callback, possibly refreshing the RPC
	 * binding */
	if (block->b_granted) {
		nlm_rebind_host(block->b_host);
		goto callback;
	}

	/* Try the lock operation again */
831 832 833
	/* vfs_lock_file() can mangle fl_start and fl_end, but we need
	 * them unchanged for the GRANT_MSG
	 */
834
	lock->fl.fl_flags |= FL_SLEEP;
835 836
	fl_start = lock->fl.fl_start;
	fl_end = lock->fl.fl_end;
837
	error = vfs_lock_file(file->f_file, F_SETLK, &lock->fl, NULL);
838
	lock->fl.fl_flags &= ~FL_SLEEP;
839 840
	lock->fl.fl_start = fl_start;
	lock->fl.fl_end = fl_end;
841

842 843 844
	switch (error) {
	case 0:
		break;
845
	case FILE_LOCK_DEFERRED:
846
		dprintk("lockd: lock still blocked error %d\n", error);
Linus Torvalds's avatar
Linus Torvalds committed
847
		nlmsvc_insert_block(block, NLM_NEVER);
Marc Eshel's avatar
Marc Eshel committed
848
		nlmsvc_release_block(block);
849
		return;
850
	default:
Linus Torvalds's avatar
Linus Torvalds committed
851
		printk(KERN_WARNING "lockd: unexpected error %d in %s!\n",
852
				-error, __func__);
Linus Torvalds's avatar
Linus Torvalds committed
853
		nlmsvc_insert_block(block, 10 * HZ);
Marc Eshel's avatar
Marc Eshel committed
854
		nlmsvc_release_block(block);
855
		return;
Linus Torvalds's avatar
Linus Torvalds committed
856 857 858 859 860 861 862
	}

callback:
	/* Lock was granted by VFS. */
	dprintk("lockd: GRANTing blocked lock.\n");
	block->b_granted = 1;

863 864 865 866
	/* keep block on the list, but don't reattempt until the RPC
	 * completes or the submission fails
	 */
	nlmsvc_insert_block(block, NLM_NEVER);
Linus Torvalds's avatar
Linus Torvalds committed
867

868 869 870 871 872 873 874 875 876
	/* Call the client -- use a soft RPC task since nlmsvc_retry_blocked
	 * will queue up a new one if this one times out
	 */
	error = nlm_async_call(block->b_call, NLMPROC_GRANTED_MSG,
				&nlmsvc_grant_ops);

	/* RPC submission failed, wait a bit and retry */
	if (error < 0)
		nlmsvc_insert_block(block, 10 * HZ);
Linus Torvalds's avatar
Linus Torvalds committed
877 878 879 880 881 882 883 884 885 886
}

/*
 * This is the callback from the RPC layer when the NLM_GRANTED_MSG
 * RPC call has succeeded or timed out.
 * Like all RPC callbacks, it is invoked by the rpciod process, so it
 * better not sleep. Therefore, we put the blocked lock on the nlm_blocked
 * chain once more in order to have it removed by lockd itself (which can
 * then sleep on the file semaphore without disrupting e.g. the nfs client).
 */
887
static void nlmsvc_grant_callback(struct rpc_task *task, void *data)
Linus Torvalds's avatar
Linus Torvalds committed
888
{
889
	struct nlm_rqst		*call = data;
890
	struct nlm_block	*block = call->a_block;
Linus Torvalds's avatar
Linus Torvalds committed
891 892 893 894
	unsigned long		timeout;

	dprintk("lockd: GRANT_MSG RPC callback\n");

895
	spin_lock(&nlm_blocked_lock);
896 897 898 899 900 901 902 903 904
	/* if the block is not on a list at this point then it has
	 * been invalidated. Don't try to requeue it.
	 *
	 * FIXME: it's possible that the block is removed from the list
	 * after this check but before the nlmsvc_insert_block. In that
	 * case it will be added back. Perhaps we need better locking
	 * for nlm_blocked?
	 */
	if (list_empty(&block->b_list))
905
		goto out;
906

Linus Torvalds's avatar
Linus Torvalds committed
907 908 909 910 911 912 913 914 915 916
	/* Technically, we should down the file semaphore here. Since we
	 * move the block towards the head of the queue only, no harm
	 * can be done, though. */
	if (task->tk_status < 0) {
		/* RPC error: Re-insert for retransmission */
		timeout = 10 * HZ;
	} else {
		/* Call was successful, now wait for client callback */
		timeout = 60 * HZ;
	}
917
	nlmsvc_insert_block_locked(block, timeout);
Linus Torvalds's avatar
Linus Torvalds committed
918
	svc_wake_up(block->b_daemon);
919
out:
920
	spin_unlock(&nlm_blocked_lock);
921 922
}

923 924 925 926
/*
 * FIXME: nlmsvc_release_block() grabs a mutex.  This is not allowed for an
 * .rpc_release rpc_call_op
 */
Adrian Bunk's avatar
Adrian Bunk committed
927
static void nlmsvc_grant_release(void *data)
928
{
929 930
	struct nlm_rqst		*call = data;
	nlmsvc_release_block(call->a_block);
Linus Torvalds's avatar
Linus Torvalds committed
931 932
}

933 934
static const struct rpc_call_ops nlmsvc_grant_ops = {
	.rpc_call_done = nlmsvc_grant_callback,
935
	.rpc_release = nlmsvc_grant_release,
936 937
};

Linus Torvalds's avatar
Linus Torvalds committed
938 939 940 941 942
/*
 * We received a GRANT_RES callback. Try to find the corresponding
 * block.
 */
void
Al Viro's avatar
Al Viro committed
943
nlmsvc_grant_reply(struct nlm_cookie *cookie, __be32 status)
Linus Torvalds's avatar
Linus Torvalds committed
944 945 946
{
	struct nlm_block	*block;

947 948 949
	dprintk("grant_reply: looking for cookie %x, s=%d \n",
		*(unsigned int *)(cookie->data), status);
	if (!(block = nlmsvc_find_block(cookie)))
Linus Torvalds's avatar
Linus Torvalds committed
950 951
		return;

952 953 954 955 956 957 958 959 960
	if (status == nlm_lck_denied_grace_period) {
		/* Try again in a couple of seconds */
		nlmsvc_insert_block(block, 10 * HZ);
	} else {
		/*
		 * Lock is now held by client, or has been rejected.
		 * In both cases, the block should be removed.
		 */
		nlmsvc_unlink_block(block);
Linus Torvalds's avatar
Linus Torvalds committed
961
	}
962
	nlmsvc_release_block(block);
Linus Torvalds's avatar
Linus Torvalds committed
963 964
}

Marc Eshel's avatar
Marc Eshel committed
965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981
/* Helper function to handle retry of a deferred block.
 * If it is a blocking lock, call grant_blocked.
 * For a non-blocking lock or test lock, revisit the request.
 */
static void
retry_deferred_block(struct nlm_block *block)
{
	if (!(block->b_flags & B_GOT_CALLBACK))
		block->b_flags |= B_TIMED_OUT;
	nlmsvc_insert_block(block, NLM_TIMEOUT);
	dprintk("revisit block %p flags %d\n",	block, block->b_flags);
	if (block->b_deferred_req) {
		block->b_deferred_req->revisit(block->b_deferred_req, 0);
		block->b_deferred_req = NULL;
	}
}

Linus Torvalds's avatar
Linus Torvalds committed
982 983 984 985 986 987 988 989
/*
 * Retry all blocked locks that have been notified. This is where lockd
 * picks up locks that can be granted, or grant notifications that must
 * be retransmitted.
 */
unsigned long
nlmsvc_retry_blocked(void)
{
990 991 992
	unsigned long	timeout = MAX_SCHEDULE_TIMEOUT;
	struct nlm_block *block;

993
	spin_lock(&nlm_blocked_lock);
994
	while (!list_empty(&nlm_blocked) && !kthread_should_stop()) {
995
		block = list_entry(nlm_blocked.next, struct nlm_block, b_list);
Linus Torvalds's avatar
Linus Torvalds committed
996 997 998

		if (block->b_when == NLM_NEVER)
			break;
999
		if (time_after(block->b_when, jiffies)) {
1000
			timeout = block->b_when - jiffies;
Linus Torvalds's avatar
Linus Torvalds committed
1001
			break;
1002
		}
1003
		spin_unlock(&nlm_blocked_lock);
1004

J. Bruce Fields's avatar
J. Bruce Fields committed
1005 1006
		dprintk("nlmsvc_retry_blocked(%p, when=%ld)\n",
			block, block->b_when);
Marc Eshel's avatar
Marc Eshel committed
1007 1008 1009 1010 1011 1012
		if (block->b_flags & B_QUEUED) {
			dprintk("nlmsvc_retry_blocked delete block (%p, granted=%d, flags=%d)\n",
				block, block->b_granted, block->b_flags);
			retry_deferred_block(block);
		} else
			nlmsvc_grant_blocked(block);
1013
		spin_lock(&nlm_blocked_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1014
	}
1015
	spin_unlock(&nlm_blocked_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1016

1017
	return timeout;
Linus Torvalds's avatar
Linus Torvalds committed
1018
}