dcache.c 80.2 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
/*
 * fs/dcache.c
 *
 * Complete reimplementation
 * (C) 1997 Thomas Schoebel-Theuer,
 * with heavy changes by Linus Torvalds
 */

/*
 * Notes on the allocation strategy:
 *
 * The dcache is a master of the icache - whenever a dcache entry
 * exists, the inode will always exist. "iput()" is done either when
 * the dcache entry is deleted or garbage collected.
 */

#include <linux/syscalls.h>
#include <linux/string.h>
#include <linux/mm.h>
#include <linux/fs.h>
21
#include <linux/fsnotify.h>
L
Linus Torvalds 已提交
22 23 24 25
#include <linux/slab.h>
#include <linux/init.h>
#include <linux/hash.h>
#include <linux/cache.h>
26
#include <linux/export.h>
L
Linus Torvalds 已提交
27 28 29 30 31 32 33
#include <linux/mount.h>
#include <linux/file.h>
#include <asm/uaccess.h>
#include <linux/security.h>
#include <linux/seqlock.h>
#include <linux/swap.h>
#include <linux/bootmem.h>
34
#include <linux/fs_struct.h>
35
#include <linux/hardirq.h>
36 37
#include <linux/bit_spinlock.h>
#include <linux/rculist_bl.h>
38
#include <linux/prefetch.h>
39
#include <linux/ratelimit.h>
40
#include "internal.h"
41
#include "mount.h"
L
Linus Torvalds 已提交
42

N
Nick Piggin 已提交
43 44
/*
 * Usage:
45 46
 * dcache->d_inode->i_lock protects:
 *   - i_dentry, d_alias, d_inode of aliases
47 48 49 50
 * dcache_hash_bucket lock protects:
 *   - the dcache hash table
 * s_anon bl list spinlock protects:
 *   - the s_anon list (see __d_drop)
N
Nick Piggin 已提交
51 52 53 54 55 56
 * dcache_lru_lock protects:
 *   - the dcache lru lists and counters
 * d_lock protects:
 *   - d_flags
 *   - d_name
 *   - d_lru
N
Nick Piggin 已提交
57
 *   - d_count
N
Nick Piggin 已提交
58
 *   - d_unhashed()
N
Nick Piggin 已提交
59 60
 *   - d_parent and d_subdirs
 *   - childrens' d_child and d_parent
N
Nick Piggin 已提交
61
 *   - d_alias, d_inode
N
Nick Piggin 已提交
62 63
 *
 * Ordering:
64
 * dentry->d_inode->i_lock
N
Nick Piggin 已提交
65 66
 *   dentry->d_lock
 *     dcache_lru_lock
67 68
 *     dcache_hash_bucket lock
 *     s_anon lock
N
Nick Piggin 已提交
69
 *
N
Nick Piggin 已提交
70 71 72 73 74 75 76
 * If there is an ancestor relationship:
 * dentry->d_parent->...->d_parent->d_lock
 *   ...
 *     dentry->d_parent->d_lock
 *       dentry->d_lock
 *
 * If no ancestor relationship:
N
Nick Piggin 已提交
77 78 79 80
 * if (dentry1 < dentry2)
 *   dentry1->d_lock
 *     dentry2->d_lock
 */
81
int sysctl_vfs_cache_pressure __read_mostly = 100;
L
Linus Torvalds 已提交
82 83
EXPORT_SYMBOL_GPL(sysctl_vfs_cache_pressure);

N
Nick Piggin 已提交
84
static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lru_lock);
A
Al Viro 已提交
85
__cacheline_aligned_in_smp DEFINE_SEQLOCK(rename_lock);
L
Linus Torvalds 已提交
86

87
EXPORT_SYMBOL(rename_lock);
L
Linus Torvalds 已提交
88

89
static struct kmem_cache *dentry_cache __read_mostly;
L
Linus Torvalds 已提交
90 91 92 93 94 95 96 97 98 99 100 101

/*
 * This is the single most critical data structure when it comes
 * to the dcache: the hashtable for lookups. Somebody should try
 * to make this good - I've just made it work.
 *
 * This hash-function tries to avoid losing too many bits of hash
 * information, yet avoid using a prime hash-size or similar.
 */
#define D_HASHBITS     d_hash_shift
#define D_HASHMASK     d_hash_mask

102 103
static unsigned int d_hash_mask __read_mostly;
static unsigned int d_hash_shift __read_mostly;
104

105
static struct hlist_bl_head *dentry_hashtable __read_mostly;
106

107
static inline struct hlist_bl_head *d_hash(const struct dentry *parent,
108
					unsigned int hash)
109
{
110 111
	hash += (unsigned long) parent / L1_CACHE_BYTES;
	hash = hash + (hash >> D_HASHBITS);
112 113 114
	return dentry_hashtable + (hash & D_HASHMASK);
}

L
Linus Torvalds 已提交
115 116 117 118 119
/* Statistics gathering. */
struct dentry_stat_t dentry_stat = {
	.age_limit = 45,
};

120
static DEFINE_PER_CPU(unsigned int, nr_dentry);
121 122

#if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
123 124 125 126 127 128 129 130 131
static int get_nr_dentry(void)
{
	int i;
	int sum = 0;
	for_each_possible_cpu(i)
		sum += per_cpu(nr_dentry, i);
	return sum < 0 ? 0 : sum;
}

132 133 134
int proc_nr_dentry(ctl_table *table, int write, void __user *buffer,
		   size_t *lenp, loff_t *ppos)
{
135
	dentry_stat.nr_dentry = get_nr_dentry();
136 137 138 139
	return proc_dointvec(table, write, buffer, lenp, ppos);
}
#endif

140 141 142 143
/*
 * Compare 2 name strings, return 0 if they match, otherwise non-zero.
 * The strings are both count bytes long, and count is non-zero.
 */
144 145 146 147 148 149 150 151 152 153 154 155
#ifdef CONFIG_DCACHE_WORD_ACCESS

#include <asm/word-at-a-time.h>
/*
 * NOTE! 'cs' and 'scount' come from a dentry, so it has a
 * aligned allocation for this particular component. We don't
 * strictly need the load_unaligned_zeropad() safety, but it
 * doesn't hurt either.
 *
 * In contrast, 'ct' and 'tcount' can be from a pathname, and do
 * need the careful unaligned handling.
 */
156
static inline int dentry_string_cmp(const unsigned char *cs, const unsigned char *ct, unsigned tcount)
157
{
158 159 160
	unsigned long a,b,mask;

	for (;;) {
161
		a = *(unsigned long *)cs;
162
		b = load_unaligned_zeropad(ct);
163 164 165 166 167 168 169 170 171 172 173 174
		if (tcount < sizeof(unsigned long))
			break;
		if (unlikely(a != b))
			return 1;
		cs += sizeof(unsigned long);
		ct += sizeof(unsigned long);
		tcount -= sizeof(unsigned long);
		if (!tcount)
			return 0;
	}
	mask = ~(~0ul << tcount*8);
	return unlikely(!!((a ^ b) & mask));
175 176
}

177
#else
178

179
static inline int dentry_string_cmp(const unsigned char *cs, const unsigned char *ct, unsigned tcount)
180
{
181 182 183 184 185 186 187 188 189 190
	do {
		if (*cs != *ct)
			return 1;
		cs++;
		ct++;
		tcount--;
	} while (tcount);
	return 0;
}

191 192
#endif

193 194
static inline int dentry_cmp(const struct dentry *dentry, const unsigned char *ct, unsigned tcount)
{
195
	const unsigned char *cs;
196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211
	/*
	 * Be careful about RCU walk racing with rename:
	 * use ACCESS_ONCE to fetch the name pointer.
	 *
	 * NOTE! Even if a rename will mean that the length
	 * was not loaded atomically, we don't care. The
	 * RCU walk will check the sequence count eventually,
	 * and catch it. And we won't overrun the buffer,
	 * because we're reading the name pointer atomically,
	 * and a dentry name is guaranteed to be properly
	 * terminated with a NUL byte.
	 *
	 * End result: even if 'len' is wrong, we'll exit
	 * early because the data cannot match (there can
	 * be no NUL in the ct/tcount data)
	 */
212 213 214
	cs = ACCESS_ONCE(dentry->d_name.name);
	smp_read_barrier_depends();
	return dentry_string_cmp(cs, ct, tcount);
215 216
}

C
Christoph Hellwig 已提交
217
static void __d_free(struct rcu_head *head)
L
Linus Torvalds 已提交
218
{
C
Christoph Hellwig 已提交
219 220
	struct dentry *dentry = container_of(head, struct dentry, d_u.d_rcu);

221
	WARN_ON(!list_empty(&dentry->d_alias));
L
Linus Torvalds 已提交
222 223 224 225 226 227
	if (dname_external(dentry))
		kfree(dentry->d_name.name);
	kmem_cache_free(dentry_cache, dentry); 
}

/*
N
Nick Piggin 已提交
228
 * no locks, please.
L
Linus Torvalds 已提交
229 230 231
 */
static void d_free(struct dentry *dentry)
{
N
Nick Piggin 已提交
232
	BUG_ON(dentry->d_count);
233
	this_cpu_dec(nr_dentry);
L
Linus Torvalds 已提交
234 235
	if (dentry->d_op && dentry->d_op->d_release)
		dentry->d_op->d_release(dentry);
236

237 238
	/* if dentry was never visible to RCU, immediate free is OK */
	if (!(dentry->d_flags & DCACHE_RCUACCESS))
C
Christoph Hellwig 已提交
239
		__d_free(&dentry->d_u.d_rcu);
240
	else
C
Christoph Hellwig 已提交
241
		call_rcu(&dentry->d_u.d_rcu, __d_free);
L
Linus Torvalds 已提交
242 243
}

N
Nick Piggin 已提交
244 245
/**
 * dentry_rcuwalk_barrier - invalidate in-progress rcu-walk lookups
246
 * @dentry: the target dentry
N
Nick Piggin 已提交
247 248 249 250 251 252 253 254 255 256 257
 * After this call, in-progress rcu-walk path lookup will fail. This
 * should be called after unhashing, and after changing d_inode (if
 * the dentry has not already been unhashed).
 */
static inline void dentry_rcuwalk_barrier(struct dentry *dentry)
{
	assert_spin_locked(&dentry->d_lock);
	/* Go through a barrier */
	write_seqcount_barrier(&dentry->d_seq);
}

L
Linus Torvalds 已提交
258 259
/*
 * Release the dentry's inode, using the filesystem
N
Nick Piggin 已提交
260 261
 * d_iput() operation if defined. Dentry has no refcount
 * and is unhashed.
L
Linus Torvalds 已提交
262
 */
263
static void dentry_iput(struct dentry * dentry)
264
	__releases(dentry->d_lock)
265
	__releases(dentry->d_inode->i_lock)
L
Linus Torvalds 已提交
266 267 268 269 270 271
{
	struct inode *inode = dentry->d_inode;
	if (inode) {
		dentry->d_inode = NULL;
		list_del_init(&dentry->d_alias);
		spin_unlock(&dentry->d_lock);
272
		spin_unlock(&inode->i_lock);
273 274
		if (!inode->i_nlink)
			fsnotify_inoderemove(inode);
L
Linus Torvalds 已提交
275 276 277 278 279 280 281 282 283
		if (dentry->d_op && dentry->d_op->d_iput)
			dentry->d_op->d_iput(dentry, inode);
		else
			iput(inode);
	} else {
		spin_unlock(&dentry->d_lock);
	}
}

N
Nick Piggin 已提交
284 285 286 287 288 289
/*
 * Release the dentry's inode, using the filesystem
 * d_iput() operation if defined. dentry remains in-use.
 */
static void dentry_unlink_inode(struct dentry * dentry)
	__releases(dentry->d_lock)
290
	__releases(dentry->d_inode->i_lock)
N
Nick Piggin 已提交
291 292 293 294 295 296
{
	struct inode *inode = dentry->d_inode;
	dentry->d_inode = NULL;
	list_del_init(&dentry->d_alias);
	dentry_rcuwalk_barrier(dentry);
	spin_unlock(&dentry->d_lock);
297
	spin_unlock(&inode->i_lock);
N
Nick Piggin 已提交
298 299 300 301 302 303 304 305
	if (!inode->i_nlink)
		fsnotify_inoderemove(inode);
	if (dentry->d_op && dentry->d_op->d_iput)
		dentry->d_op->d_iput(dentry, inode);
	else
		iput(inode);
}

306
/*
S
Sage Weil 已提交
307
 * dentry_lru_(add|del|prune|move_tail) must be called with d_lock held.
308 309 310
 */
static void dentry_lru_add(struct dentry *dentry)
{
311
	if (list_empty(&dentry->d_lru)) {
N
Nick Piggin 已提交
312
		spin_lock(&dcache_lru_lock);
313 314
		list_add(&dentry->d_lru, &dentry->d_sb->s_dentry_lru);
		dentry->d_sb->s_nr_dentry_unused++;
315
		dentry_stat.nr_unused++;
N
Nick Piggin 已提交
316
		spin_unlock(&dcache_lru_lock);
317
	}
318 319
}

N
Nick Piggin 已提交
320 321 322
static void __dentry_lru_del(struct dentry *dentry)
{
	list_del_init(&dentry->d_lru);
323
	dentry->d_flags &= ~DCACHE_SHRINK_LIST;
N
Nick Piggin 已提交
324 325 326 327
	dentry->d_sb->s_nr_dentry_unused--;
	dentry_stat.nr_unused--;
}

S
Sage Weil 已提交
328 329 330
/*
 * Remove a dentry with references from the LRU.
 */
331 332 333
static void dentry_lru_del(struct dentry *dentry)
{
	if (!list_empty(&dentry->d_lru)) {
N
Nick Piggin 已提交
334 335 336
		spin_lock(&dcache_lru_lock);
		__dentry_lru_del(dentry);
		spin_unlock(&dcache_lru_lock);
337 338 339
	}
}

S
Sage Weil 已提交
340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356
/*
 * Remove a dentry that is unreferenced and about to be pruned
 * (unhashed and destroyed) from the LRU, and inform the file system.
 * This wrapper should be called _prior_ to unhashing a victim dentry.
 */
static void dentry_lru_prune(struct dentry *dentry)
{
	if (!list_empty(&dentry->d_lru)) {
		if (dentry->d_flags & DCACHE_OP_PRUNE)
			dentry->d_op->d_prune(dentry);

		spin_lock(&dcache_lru_lock);
		__dentry_lru_del(dentry);
		spin_unlock(&dcache_lru_lock);
	}
}

357
static void dentry_lru_move_list(struct dentry *dentry, struct list_head *list)
358
{
N
Nick Piggin 已提交
359
	spin_lock(&dcache_lru_lock);
360
	if (list_empty(&dentry->d_lru)) {
361
		list_add_tail(&dentry->d_lru, list);
362
		dentry->d_sb->s_nr_dentry_unused++;
363
		dentry_stat.nr_unused++;
364
	} else {
365
		list_move_tail(&dentry->d_lru, list);
366
	}
N
Nick Piggin 已提交
367
	spin_unlock(&dcache_lru_lock);
368 369
}

370 371 372
/**
 * d_kill - kill dentry and return parent
 * @dentry: dentry to kill
373
 * @parent: parent dentry
374
 *
375
 * The dentry must already be unhashed and removed from the LRU.
376 377
 *
 * If this is the root of the dentry tree, return NULL.
N
Nick Piggin 已提交
378
 *
N
Nick Piggin 已提交
379 380
 * dentry->d_lock and parent->d_lock must be held by caller, and are dropped by
 * d_kill.
381
 */
N
Nick Piggin 已提交
382
static struct dentry *d_kill(struct dentry *dentry, struct dentry *parent)
383
	__releases(dentry->d_lock)
N
Nick Piggin 已提交
384
	__releases(parent->d_lock)
385
	__releases(dentry->d_inode->i_lock)
386 387
{
	list_del(&dentry->d_u.d_child);
388 389 390 391 392
	/*
	 * Inform try_to_ascend() that we are no longer attached to the
	 * dentry tree
	 */
	dentry->d_flags |= DCACHE_DISCONNECTED;
N
Nick Piggin 已提交
393 394
	if (parent)
		spin_unlock(&parent->d_lock);
395
	dentry_iput(dentry);
N
Nick Piggin 已提交
396 397 398 399
	/*
	 * dentry_iput drops the locks, at which point nobody (except
	 * transient RCU lookups) can reach this dentry.
	 */
400
	d_free(dentry);
401
	return parent;
402 403
}

404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424
/*
 * Unhash a dentry without inserting an RCU walk barrier or checking that
 * dentry->d_lock is locked.  The caller must take care of that, if
 * appropriate.
 */
static void __d_shrink(struct dentry *dentry)
{
	if (!d_unhashed(dentry)) {
		struct hlist_bl_head *b;
		if (unlikely(dentry->d_flags & DCACHE_DISCONNECTED))
			b = &dentry->d_sb->s_anon;
		else
			b = d_hash(dentry->d_parent, dentry->d_name.hash);

		hlist_bl_lock(b);
		__hlist_bl_del(&dentry->d_hash);
		dentry->d_hash.pprev = NULL;
		hlist_bl_unlock(b);
	}
}

N
Nick Piggin 已提交
425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441
/**
 * d_drop - drop a dentry
 * @dentry: dentry to drop
 *
 * d_drop() unhashes the entry from the parent dentry hashes, so that it won't
 * be found through a VFS lookup any more. Note that this is different from
 * deleting the dentry - d_delete will try to mark the dentry negative if
 * possible, giving a successful _negative_ lookup, while d_drop will
 * just make the cache lookup fail.
 *
 * d_drop() is used mainly for stuff that wants to invalidate a dentry for some
 * reason (NFS timeouts or autofs deletes).
 *
 * __d_drop requires dentry->d_lock.
 */
void __d_drop(struct dentry *dentry)
{
442
	if (!d_unhashed(dentry)) {
443
		__d_shrink(dentry);
444
		dentry_rcuwalk_barrier(dentry);
N
Nick Piggin 已提交
445 446 447 448 449 450 451 452 453 454 455 456
	}
}
EXPORT_SYMBOL(__d_drop);

void d_drop(struct dentry *dentry)
{
	spin_lock(&dentry->d_lock);
	__d_drop(dentry);
	spin_unlock(&dentry->d_lock);
}
EXPORT_SYMBOL(d_drop);

457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474
/*
 * d_clear_need_lookup - drop a dentry from cache and clear the need lookup flag
 * @dentry: dentry to drop
 *
 * This is called when we do a lookup on a placeholder dentry that needed to be
 * looked up.  The dentry should have been hashed in order for it to be found by
 * the lookup code, but now needs to be unhashed while we do the actual lookup
 * and clear the DCACHE_NEED_LOOKUP flag.
 */
void d_clear_need_lookup(struct dentry *dentry)
{
	spin_lock(&dentry->d_lock);
	__d_drop(dentry);
	dentry->d_flags &= ~DCACHE_NEED_LOOKUP;
	spin_unlock(&dentry->d_lock);
}
EXPORT_SYMBOL(d_clear_need_lookup);

475 476 477 478 479 480 481 482 483
/*
 * Finish off a dentry we've decided to kill.
 * dentry->d_lock must be held, returns with it unlocked.
 * If ref is non-zero, then decrement the refcount too.
 * Returns dentry requiring refcount drop, or NULL if we're done.
 */
static inline struct dentry *dentry_kill(struct dentry *dentry, int ref)
	__releases(dentry->d_lock)
{
484
	struct inode *inode;
485 486
	struct dentry *parent;

487 488
	inode = dentry->d_inode;
	if (inode && !spin_trylock(&inode->i_lock)) {
489 490 491 492 493 494 495 496 497 498
relock:
		spin_unlock(&dentry->d_lock);
		cpu_relax();
		return dentry; /* try again with same dentry */
	}
	if (IS_ROOT(dentry))
		parent = NULL;
	else
		parent = dentry->d_parent;
	if (parent && !spin_trylock(&parent->d_lock)) {
499 500
		if (inode)
			spin_unlock(&inode->i_lock);
501 502
		goto relock;
	}
N
Nick Piggin 已提交
503

504 505
	if (ref)
		dentry->d_count--;
S
Sage Weil 已提交
506 507 508 509 510 511
	/*
	 * if dentry was on the d_lru list delete it from there.
	 * inform the fs via d_prune that this dentry is about to be
	 * unhashed and destroyed.
	 */
	dentry_lru_prune(dentry);
512 513 514 515 516
	/* if it was on the hash then remove it */
	__d_drop(dentry);
	return d_kill(dentry, parent);
}

L
Linus Torvalds 已提交
517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548
/* 
 * This is dput
 *
 * This is complicated by the fact that we do not want to put
 * dentries that are no longer on any hash chain on the unused
 * list: we'd much rather just get rid of them immediately.
 *
 * However, that implies that we have to traverse the dentry
 * tree upwards to the parents which might _also_ now be
 * scheduled for deletion (it may have been only waiting for
 * its last child to go away).
 *
 * This tail recursion is done by hand as we don't want to depend
 * on the compiler to always get this right (gcc generally doesn't).
 * Real recursion would eat up our stack space.
 */

/*
 * dput - release a dentry
 * @dentry: dentry to release 
 *
 * Release a dentry. This will drop the usage count and if appropriate
 * call the dentry unlink method as well as removing it from the queues and
 * releasing its resources. If the parent dentries were scheduled for release
 * they too may now get deleted.
 */
void dput(struct dentry *dentry)
{
	if (!dentry)
		return;

repeat:
N
Nick Piggin 已提交
549
	if (dentry->d_count == 1)
L
Linus Torvalds 已提交
550 551
		might_sleep();
	spin_lock(&dentry->d_lock);
N
Nick Piggin 已提交
552 553 554
	BUG_ON(!dentry->d_count);
	if (dentry->d_count > 1) {
		dentry->d_count--;
L
Linus Torvalds 已提交
555 556 557 558
		spin_unlock(&dentry->d_lock);
		return;
	}

559
	if (dentry->d_flags & DCACHE_OP_DELETE) {
L
Linus Torvalds 已提交
560
		if (dentry->d_op->d_delete(dentry))
N
Nick Piggin 已提交
561
			goto kill_it;
L
Linus Torvalds 已提交
562
	}
N
Nick Piggin 已提交
563

L
Linus Torvalds 已提交
564 565 566
	/* Unreachable? Get rid of it */
 	if (d_unhashed(dentry))
		goto kill_it;
N
Nick Piggin 已提交
567

568 569 570 571 572 573 574
	/*
	 * If this dentry needs lookup, don't set the referenced flag so that it
	 * is more likely to be cleaned up by the dcache shrinker in case of
	 * memory pressure.
	 */
	if (!d_need_lookup(dentry))
		dentry->d_flags |= DCACHE_REFERENCED;
575
	dentry_lru_add(dentry);
N
Nick Piggin 已提交
576

N
Nick Piggin 已提交
577 578
	dentry->d_count--;
	spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
579 580
	return;

581
kill_it:
582
	dentry = dentry_kill(dentry, 1);
583 584
	if (dentry)
		goto repeat;
L
Linus Torvalds 已提交
585
}
586
EXPORT_SYMBOL(dput);
L
Linus Torvalds 已提交
587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604

/**
 * d_invalidate - invalidate a dentry
 * @dentry: dentry to invalidate
 *
 * Try to invalidate the dentry if it turns out to be
 * possible. If there are other dentries that can be
 * reached through this one we can't delete it and we
 * return -EBUSY. On success we return 0.
 *
 * no dcache lock.
 */
 
int d_invalidate(struct dentry * dentry)
{
	/*
	 * If it's already been dropped, return OK.
	 */
N
Nick Piggin 已提交
605
	spin_lock(&dentry->d_lock);
L
Linus Torvalds 已提交
606
	if (d_unhashed(dentry)) {
N
Nick Piggin 已提交
607
		spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
608 609 610 611 612 613 614
		return 0;
	}
	/*
	 * Check whether to do a partial shrink_dcache
	 * to get rid of unused child entries.
	 */
	if (!list_empty(&dentry->d_subdirs)) {
N
Nick Piggin 已提交
615
		spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
616
		shrink_dcache_parent(dentry);
N
Nick Piggin 已提交
617
		spin_lock(&dentry->d_lock);
L
Linus Torvalds 已提交
618 619 620 621 622 623 624 625 626 627 628
	}

	/*
	 * Somebody else still using it?
	 *
	 * If it's a directory, we can't drop it
	 * for fear of somebody re-populating it
	 * with children (even though dropping it
	 * would make it unreachable from the root,
	 * we might still populate it if it was a
	 * working directory or similar).
629 630
	 * We also need to leave mountpoints alone,
	 * directory or not.
L
Linus Torvalds 已提交
631
	 */
632 633
	if (dentry->d_count > 1 && dentry->d_inode) {
		if (S_ISDIR(dentry->d_inode->i_mode) || d_mountpoint(dentry)) {
L
Linus Torvalds 已提交
634 635 636 637 638 639 640 641 642
			spin_unlock(&dentry->d_lock);
			return -EBUSY;
		}
	}

	__d_drop(dentry);
	spin_unlock(&dentry->d_lock);
	return 0;
}
643
EXPORT_SYMBOL(d_invalidate);
L
Linus Torvalds 已提交
644

N
Nick Piggin 已提交
645
/* This must be called with d_lock held */
646
static inline void __dget_dlock(struct dentry *dentry)
N
Nick Piggin 已提交
647
{
N
Nick Piggin 已提交
648
	dentry->d_count++;
N
Nick Piggin 已提交
649 650
}

651
static inline void __dget(struct dentry *dentry)
L
Linus Torvalds 已提交
652
{
N
Nick Piggin 已提交
653
	spin_lock(&dentry->d_lock);
654
	__dget_dlock(dentry);
N
Nick Piggin 已提交
655
	spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
656 657
}

N
Nick Piggin 已提交
658 659 660 661 662
struct dentry *dget_parent(struct dentry *dentry)
{
	struct dentry *ret;

repeat:
N
Nick Piggin 已提交
663 664 665 666 667
	/*
	 * Don't need rcu_dereference because we re-check it was correct under
	 * the lock.
	 */
	rcu_read_lock();
N
Nick Piggin 已提交
668
	ret = dentry->d_parent;
N
Nick Piggin 已提交
669 670 671 672
	spin_lock(&ret->d_lock);
	if (unlikely(ret != dentry->d_parent)) {
		spin_unlock(&ret->d_lock);
		rcu_read_unlock();
N
Nick Piggin 已提交
673 674
		goto repeat;
	}
N
Nick Piggin 已提交
675
	rcu_read_unlock();
N
Nick Piggin 已提交
676 677 678 679 680 681 682
	BUG_ON(!ret->d_count);
	ret->d_count++;
	spin_unlock(&ret->d_lock);
	return ret;
}
EXPORT_SYMBOL(dget_parent);

L
Linus Torvalds 已提交
683 684 685
/**
 * d_find_alias - grab a hashed alias of inode
 * @inode: inode in question
686 687
 * @want_discon:  flag, used by d_splice_alias, to request
 *          that only a DISCONNECTED alias be returned.
L
Linus Torvalds 已提交
688 689 690 691 692 693 694
 *
 * If inode has a hashed alias, or is a directory and has any alias,
 * acquire the reference to alias and return it. Otherwise return NULL.
 * Notice that if inode is a directory there can be only one alias and
 * it can be unhashed only if it has no children, or if it is the root
 * of a filesystem.
 *
695
 * If the inode has an IS_ROOT, DCACHE_DISCONNECTED alias, then prefer
696 697
 * any other hashed alias over that one unless @want_discon is set,
 * in which case only return an IS_ROOT, DCACHE_DISCONNECTED alias.
L
Linus Torvalds 已提交
698
 */
699
static struct dentry *__d_find_alias(struct inode *inode, int want_discon)
L
Linus Torvalds 已提交
700
{
N
Nick Piggin 已提交
701
	struct dentry *alias, *discon_alias;
L
Linus Torvalds 已提交
702

N
Nick Piggin 已提交
703 704 705 706
again:
	discon_alias = NULL;
	list_for_each_entry(alias, &inode->i_dentry, d_alias) {
		spin_lock(&alias->d_lock);
L
Linus Torvalds 已提交
707
 		if (S_ISDIR(inode->i_mode) || !d_unhashed(alias)) {
708
			if (IS_ROOT(alias) &&
N
Nick Piggin 已提交
709
			    (alias->d_flags & DCACHE_DISCONNECTED)) {
L
Linus Torvalds 已提交
710
				discon_alias = alias;
711
			} else if (!want_discon) {
712
				__dget_dlock(alias);
N
Nick Piggin 已提交
713 714 715 716 717 718 719 720 721 722 723 724
				spin_unlock(&alias->d_lock);
				return alias;
			}
		}
		spin_unlock(&alias->d_lock);
	}
	if (discon_alias) {
		alias = discon_alias;
		spin_lock(&alias->d_lock);
		if (S_ISDIR(inode->i_mode) || !d_unhashed(alias)) {
			if (IS_ROOT(alias) &&
			    (alias->d_flags & DCACHE_DISCONNECTED)) {
725
				__dget_dlock(alias);
N
Nick Piggin 已提交
726
				spin_unlock(&alias->d_lock);
L
Linus Torvalds 已提交
727 728 729
				return alias;
			}
		}
N
Nick Piggin 已提交
730 731
		spin_unlock(&alias->d_lock);
		goto again;
L
Linus Torvalds 已提交
732
	}
N
Nick Piggin 已提交
733
	return NULL;
L
Linus Torvalds 已提交
734 735
}

N
Nick Piggin 已提交
736
struct dentry *d_find_alias(struct inode *inode)
L
Linus Torvalds 已提交
737
{
D
David Howells 已提交
738 739 740
	struct dentry *de = NULL;

	if (!list_empty(&inode->i_dentry)) {
741
		spin_lock(&inode->i_lock);
742
		de = __d_find_alias(inode, 0);
743
		spin_unlock(&inode->i_lock);
D
David Howells 已提交
744
	}
L
Linus Torvalds 已提交
745 746
	return de;
}
747
EXPORT_SYMBOL(d_find_alias);
L
Linus Torvalds 已提交
748 749 750 751 752 753 754

/*
 *	Try to kill dentries associated with this inode.
 * WARNING: you must own a reference to inode.
 */
void d_prune_aliases(struct inode *inode)
{
755
	struct dentry *dentry;
L
Linus Torvalds 已提交
756
restart:
757
	spin_lock(&inode->i_lock);
758
	list_for_each_entry(dentry, &inode->i_dentry, d_alias) {
L
Linus Torvalds 已提交
759
		spin_lock(&dentry->d_lock);
N
Nick Piggin 已提交
760
		if (!dentry->d_count) {
761
			__dget_dlock(dentry);
L
Linus Torvalds 已提交
762 763
			__d_drop(dentry);
			spin_unlock(&dentry->d_lock);
764
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
765 766 767 768 769
			dput(dentry);
			goto restart;
		}
		spin_unlock(&dentry->d_lock);
	}
770
	spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
771
}
772
EXPORT_SYMBOL(d_prune_aliases);
L
Linus Torvalds 已提交
773 774

/*
775 776 777
 * Try to throw away a dentry - free the inode, dput the parent.
 * Requires dentry->d_lock is held, and dentry->d_count == 0.
 * Releases dentry->d_lock.
778
 *
779
 * This may fail if locks cannot be acquired no problem, just try again.
L
Linus Torvalds 已提交
780
 */
781
static void try_prune_one_dentry(struct dentry *dentry)
782
	__releases(dentry->d_lock)
L
Linus Torvalds 已提交
783
{
784
	struct dentry *parent;
785

786
	parent = dentry_kill(dentry, 0);
787
	/*
788 789 790 791 792 793 794 795
	 * If dentry_kill returns NULL, we have nothing more to do.
	 * if it returns the same dentry, trylocks failed. In either
	 * case, just loop again.
	 *
	 * Otherwise, we need to prune ancestors too. This is necessary
	 * to prevent quadratic behavior of shrink_dcache_parent(), but
	 * is also expected to be beneficial in reducing dentry cache
	 * fragmentation.
796
	 */
797 798 799 800 801 802 803
	if (!parent)
		return;
	if (parent == dentry)
		return;

	/* Prune ancestors. */
	dentry = parent;
804
	while (dentry) {
N
Nick Piggin 已提交
805
		spin_lock(&dentry->d_lock);
806 807 808 809 810
		if (dentry->d_count > 1) {
			dentry->d_count--;
			spin_unlock(&dentry->d_lock);
			return;
		}
811
		dentry = dentry_kill(dentry, 1);
812
	}
L
Linus Torvalds 已提交
813 814
}

C
Christoph Hellwig 已提交
815
static void shrink_dentry_list(struct list_head *list)
L
Linus Torvalds 已提交
816
{
817 818
	struct dentry *dentry;

819 820 821 822 823 824 825 826
	rcu_read_lock();
	for (;;) {
		dentry = list_entry_rcu(list->prev, struct dentry, d_lru);
		if (&dentry->d_lru == list)
			break; /* empty */
		spin_lock(&dentry->d_lock);
		if (dentry != list_entry(list->prev, struct dentry, d_lru)) {
			spin_unlock(&dentry->d_lock);
N
Nick Piggin 已提交
827 828 829
			continue;
		}

L
Linus Torvalds 已提交
830 831
		/*
		 * We found an inuse dentry which was not removed from
832 833
		 * the LRU because of laziness during lookup.  Do not free
		 * it - just keep it off the LRU list.
L
Linus Torvalds 已提交
834
		 */
N
Nick Piggin 已提交
835
		if (dentry->d_count) {
836
			dentry_lru_del(dentry);
837
			spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
838 839
			continue;
		}
840 841

		rcu_read_unlock();
842 843 844

		try_prune_one_dentry(dentry);

845
		rcu_read_lock();
846
	}
847
	rcu_read_unlock();
C
Christoph Hellwig 已提交
848 849 850
}

/**
851 852 853 854 855 856 857
 * prune_dcache_sb - shrink the dcache
 * @sb: superblock
 * @count: number of entries to try to free
 *
 * Attempt to shrink the superblock dcache LRU by @count entries. This is
 * done when we need more memory an called from the superblock shrinker
 * function.
C
Christoph Hellwig 已提交
858
 *
859 860
 * This function may fail to free any resources if all the dentries are in
 * use.
C
Christoph Hellwig 已提交
861
 */
862
void prune_dcache_sb(struct super_block *sb, int count)
C
Christoph Hellwig 已提交
863 864 865 866 867
{
	struct dentry *dentry;
	LIST_HEAD(referenced);
	LIST_HEAD(tmp);

N
Nick Piggin 已提交
868 869
relock:
	spin_lock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
870 871 872 873 874
	while (!list_empty(&sb->s_dentry_lru)) {
		dentry = list_entry(sb->s_dentry_lru.prev,
				struct dentry, d_lru);
		BUG_ON(dentry->d_sb != sb);

N
Nick Piggin 已提交
875 876 877 878 879 880
		if (!spin_trylock(&dentry->d_lock)) {
			spin_unlock(&dcache_lru_lock);
			cpu_relax();
			goto relock;
		}

881
		if (dentry->d_flags & DCACHE_REFERENCED) {
N
Nick Piggin 已提交
882 883
			dentry->d_flags &= ~DCACHE_REFERENCED;
			list_move(&dentry->d_lru, &referenced);
C
Christoph Hellwig 已提交
884
			spin_unlock(&dentry->d_lock);
N
Nick Piggin 已提交
885 886
		} else {
			list_move_tail(&dentry->d_lru, &tmp);
887
			dentry->d_flags |= DCACHE_SHRINK_LIST;
N
Nick Piggin 已提交
888
			spin_unlock(&dentry->d_lock);
889
			if (!--count)
N
Nick Piggin 已提交
890
				break;
C
Christoph Hellwig 已提交
891
		}
892
		cond_resched_lock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
893
	}
894 895
	if (!list_empty(&referenced))
		list_splice(&referenced, &sb->s_dentry_lru);
N
Nick Piggin 已提交
896
	spin_unlock(&dcache_lru_lock);
897 898

	shrink_dentry_list(&tmp);
899 900
}

L
Linus Torvalds 已提交
901 902 903 904
/**
 * shrink_dcache_sb - shrink dcache for a superblock
 * @sb: superblock
 *
C
Christoph Hellwig 已提交
905 906
 * Shrink the dcache for the specified super block. This is used to free
 * the dcache before unmounting a file system.
L
Linus Torvalds 已提交
907
 */
C
Christoph Hellwig 已提交
908
void shrink_dcache_sb(struct super_block *sb)
L
Linus Torvalds 已提交
909
{
C
Christoph Hellwig 已提交
910 911
	LIST_HEAD(tmp);

N
Nick Piggin 已提交
912
	spin_lock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
913 914
	while (!list_empty(&sb->s_dentry_lru)) {
		list_splice_init(&sb->s_dentry_lru, &tmp);
915
		spin_unlock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
916
		shrink_dentry_list(&tmp);
917
		spin_lock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
918
	}
N
Nick Piggin 已提交
919
	spin_unlock(&dcache_lru_lock);
L
Linus Torvalds 已提交
920
}
921
EXPORT_SYMBOL(shrink_dcache_sb);
L
Linus Torvalds 已提交
922

923 924 925 926 927 928 929 930 931 932 933 934 935
/*
 * destroy a single subtree of dentries for unmount
 * - see the comments on shrink_dcache_for_umount() for a description of the
 *   locking
 */
static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
{
	struct dentry *parent;

	BUG_ON(!IS_ROOT(dentry));

	for (;;) {
		/* descend to the first leaf in the current subtree */
936
		while (!list_empty(&dentry->d_subdirs))
937 938 939 940 941 942 943 944
			dentry = list_entry(dentry->d_subdirs.next,
					    struct dentry, d_u.d_child);

		/* consume the dentries from this leaf up through its parents
		 * until we find one with children or run out altogether */
		do {
			struct inode *inode;

S
Sage Weil 已提交
945 946 947 948 949 950
			/*
			 * remove the dentry from the lru, and inform
			 * the fs that this dentry is about to be
			 * unhashed and destroyed.
			 */
			dentry_lru_prune(dentry);
951 952
			__d_shrink(dentry);

N
Nick Piggin 已提交
953
			if (dentry->d_count != 0) {
954 955 956 957 958 959 960 961
				printk(KERN_ERR
				       "BUG: Dentry %p{i=%lx,n=%s}"
				       " still in use (%d)"
				       " [unmount of %s %s]\n",
				       dentry,
				       dentry->d_inode ?
				       dentry->d_inode->i_ino : 0UL,
				       dentry->d_name.name,
N
Nick Piggin 已提交
962
				       dentry->d_count,
963 964 965 966 967
				       dentry->d_sb->s_type->name,
				       dentry->d_sb->s_id);
				BUG();
			}

N
Nick Piggin 已提交
968
			if (IS_ROOT(dentry)) {
969
				parent = NULL;
N
Nick Piggin 已提交
970 971
				list_del(&dentry->d_u.d_child);
			} else {
972
				parent = dentry->d_parent;
N
Nick Piggin 已提交
973
				parent->d_count--;
N
Nick Piggin 已提交
974
				list_del(&dentry->d_u.d_child);
975
			}
976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992

			inode = dentry->d_inode;
			if (inode) {
				dentry->d_inode = NULL;
				list_del_init(&dentry->d_alias);
				if (dentry->d_op && dentry->d_op->d_iput)
					dentry->d_op->d_iput(dentry, inode);
				else
					iput(inode);
			}

			d_free(dentry);

			/* finished when we fall off the top of the tree,
			 * otherwise we ascend to the parent and move to the
			 * next sibling if there is one */
			if (!parent)
993
				return;
994 995 996 997 998 999 1000 1001 1002 1003
			dentry = parent;
		} while (list_empty(&dentry->d_subdirs));

		dentry = list_entry(dentry->d_subdirs.next,
				    struct dentry, d_u.d_child);
	}
}

/*
 * destroy the dentries attached to a superblock on unmounting
N
Nick Piggin 已提交
1004
 * - we don't need to use dentry->d_lock because:
1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020
 *   - the superblock is detached from all mountings and open files, so the
 *     dentry trees will not be rearranged by the VFS
 *   - s_umount is write-locked, so the memory pressure shrinker will ignore
 *     any dentries belonging to this superblock that it comes across
 *   - the filesystem itself is no longer permitted to rearrange the dentries
 *     in this superblock
 */
void shrink_dcache_for_umount(struct super_block *sb)
{
	struct dentry *dentry;

	if (down_read_trylock(&sb->s_umount))
		BUG();

	dentry = sb->s_root;
	sb->s_root = NULL;
N
Nick Piggin 已提交
1021
	dentry->d_count--;
1022 1023
	shrink_dcache_for_umount_subtree(dentry);

1024 1025
	while (!hlist_bl_empty(&sb->s_anon)) {
		dentry = hlist_bl_entry(hlist_bl_first(&sb->s_anon), struct dentry, d_hash);
1026 1027 1028 1029
		shrink_dcache_for_umount_subtree(dentry);
	}
}

1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
/*
 * This tries to ascend one level of parenthood, but
 * we can race with renaming, so we need to re-check
 * the parenthood after dropping the lock and check
 * that the sequence number still matches.
 */
static struct dentry *try_to_ascend(struct dentry *old, int locked, unsigned seq)
{
	struct dentry *new = old->d_parent;

	rcu_read_lock();
	spin_unlock(&old->d_lock);
	spin_lock(&new->d_lock);

	/*
	 * might go back up the wrong parent if we have had a rename
	 * or deletion
	 */
	if (new != old->d_parent ||
1049
		 (old->d_flags & DCACHE_DISCONNECTED) ||
1050 1051 1052 1053 1054 1055 1056 1057 1058
		 (!locked && read_seqretry(&rename_lock, seq))) {
		spin_unlock(&new->d_lock);
		new = NULL;
	}
	rcu_read_unlock();
	return new;
}


L
Linus Torvalds 已提交
1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073
/*
 * Search for at least 1 mount point in the dentry's subdirs.
 * We descend to the next level whenever the d_subdirs
 * list is non-empty and continue searching.
 */
 
/**
 * have_submounts - check for mounts over a dentry
 * @parent: dentry to check.
 *
 * Return true if the parent or its subdirectories contain
 * a mount point
 */
int have_submounts(struct dentry *parent)
{
1074
	struct dentry *this_parent;
L
Linus Torvalds 已提交
1075
	struct list_head *next;
1076
	unsigned seq;
1077
	int locked = 0;
1078 1079

	seq = read_seqbegin(&rename_lock);
1080 1081
again:
	this_parent = parent;
L
Linus Torvalds 已提交
1082 1083 1084

	if (d_mountpoint(parent))
		goto positive;
N
Nick Piggin 已提交
1085
	spin_lock(&this_parent->d_lock);
L
Linus Torvalds 已提交
1086 1087 1088 1089 1090
repeat:
	next = this_parent->d_subdirs.next;
resume:
	while (next != &this_parent->d_subdirs) {
		struct list_head *tmp = next;
E
Eric Dumazet 已提交
1091
		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
L
Linus Torvalds 已提交
1092
		next = tmp->next;
N
Nick Piggin 已提交
1093 1094

		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
L
Linus Torvalds 已提交
1095
		/* Have we found a mount point ? */
N
Nick Piggin 已提交
1096 1097 1098
		if (d_mountpoint(dentry)) {
			spin_unlock(&dentry->d_lock);
			spin_unlock(&this_parent->d_lock);
L
Linus Torvalds 已提交
1099
			goto positive;
N
Nick Piggin 已提交
1100
		}
L
Linus Torvalds 已提交
1101
		if (!list_empty(&dentry->d_subdirs)) {
N
Nick Piggin 已提交
1102 1103
			spin_unlock(&this_parent->d_lock);
			spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
L
Linus Torvalds 已提交
1104
			this_parent = dentry;
N
Nick Piggin 已提交
1105
			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
L
Linus Torvalds 已提交
1106 1107
			goto repeat;
		}
N
Nick Piggin 已提交
1108
		spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
1109 1110 1111 1112 1113
	}
	/*
	 * All done at this level ... ascend and resume the search.
	 */
	if (this_parent != parent) {
1114 1115 1116
		struct dentry *child = this_parent;
		this_parent = try_to_ascend(this_parent, locked, seq);
		if (!this_parent)
1117 1118
			goto rename_retry;
		next = child->d_u.d_child.next;
L
Linus Torvalds 已提交
1119 1120
		goto resume;
	}
N
Nick Piggin 已提交
1121
	spin_unlock(&this_parent->d_lock);
1122
	if (!locked && read_seqretry(&rename_lock, seq))
1123
		goto rename_retry;
1124 1125
	if (locked)
		write_sequnlock(&rename_lock);
L
Linus Torvalds 已提交
1126 1127
	return 0; /* No mount points found in tree */
positive:
1128
	if (!locked && read_seqretry(&rename_lock, seq))
1129
		goto rename_retry;
1130 1131
	if (locked)
		write_sequnlock(&rename_lock);
L
Linus Torvalds 已提交
1132
	return 1;
1133 1134 1135 1136 1137

rename_retry:
	locked = 1;
	write_seqlock(&rename_lock);
	goto again;
L
Linus Torvalds 已提交
1138
}
1139
EXPORT_SYMBOL(have_submounts);
L
Linus Torvalds 已提交
1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154

/*
 * Search the dentry child list for the specified parent,
 * and move any unused dentries to the end of the unused
 * list for prune_dcache(). We descend to the next level
 * whenever the d_subdirs list is non-empty and continue
 * searching.
 *
 * It returns zero iff there are no unused children,
 * otherwise  it returns the number of children moved to
 * the end of the unused list. This may not be the total
 * number of unused children, because select_parent can
 * drop the lock and return early due to latency
 * constraints.
 */
1155
static int select_parent(struct dentry *parent, struct list_head *dispose)
L
Linus Torvalds 已提交
1156
{
1157
	struct dentry *this_parent;
L
Linus Torvalds 已提交
1158
	struct list_head *next;
1159
	unsigned seq;
L
Linus Torvalds 已提交
1160
	int found = 0;
1161
	int locked = 0;
L
Linus Torvalds 已提交
1162

1163
	seq = read_seqbegin(&rename_lock);
1164 1165
again:
	this_parent = parent;
N
Nick Piggin 已提交
1166
	spin_lock(&this_parent->d_lock);
L
Linus Torvalds 已提交
1167 1168 1169 1170 1171
repeat:
	next = this_parent->d_subdirs.next;
resume:
	while (next != &this_parent->d_subdirs) {
		struct list_head *tmp = next;
E
Eric Dumazet 已提交
1172
		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
L
Linus Torvalds 已提交
1173 1174
		next = tmp->next;

N
Nick Piggin 已提交
1175
		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
N
Nick Piggin 已提交
1176

1177 1178
		/*
		 * move only zero ref count dentries to the dispose list.
1179 1180 1181 1182 1183
		 *
		 * Those which are presently on the shrink list, being processed
		 * by shrink_dentry_list(), shouldn't be moved.  Otherwise the
		 * loop in shrink_dcache_parent() might not make any progress
		 * and loop forever.
L
Linus Torvalds 已提交
1184
		 */
1185 1186 1187
		if (dentry->d_count) {
			dentry_lru_del(dentry);
		} else if (!(dentry->d_flags & DCACHE_SHRINK_LIST)) {
1188
			dentry_lru_move_list(dentry, dispose);
1189
			dentry->d_flags |= DCACHE_SHRINK_LIST;
L
Linus Torvalds 已提交
1190 1191 1192 1193 1194 1195 1196
			found++;
		}
		/*
		 * We can return to the caller if we have found some (this
		 * ensures forward progress). We'll be coming back to find
		 * the rest.
		 */
N
Nick Piggin 已提交
1197 1198
		if (found && need_resched()) {
			spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
1199
			goto out;
N
Nick Piggin 已提交
1200
		}
L
Linus Torvalds 已提交
1201 1202 1203 1204 1205

		/*
		 * Descend a level if the d_subdirs list is non-empty.
		 */
		if (!list_empty(&dentry->d_subdirs)) {
N
Nick Piggin 已提交
1206 1207
			spin_unlock(&this_parent->d_lock);
			spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
L
Linus Torvalds 已提交
1208
			this_parent = dentry;
N
Nick Piggin 已提交
1209
			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
L
Linus Torvalds 已提交
1210 1211
			goto repeat;
		}
N
Nick Piggin 已提交
1212 1213

		spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
1214 1215 1216 1217 1218
	}
	/*
	 * All done at this level ... ascend and resume the search.
	 */
	if (this_parent != parent) {
1219 1220 1221
		struct dentry *child = this_parent;
		this_parent = try_to_ascend(this_parent, locked, seq);
		if (!this_parent)
1222 1223
			goto rename_retry;
		next = child->d_u.d_child.next;
L
Linus Torvalds 已提交
1224 1225 1226
		goto resume;
	}
out:
N
Nick Piggin 已提交
1227
	spin_unlock(&this_parent->d_lock);
1228
	if (!locked && read_seqretry(&rename_lock, seq))
1229
		goto rename_retry;
1230 1231
	if (locked)
		write_sequnlock(&rename_lock);
L
Linus Torvalds 已提交
1232
	return found;
1233 1234 1235 1236 1237 1238 1239

rename_retry:
	if (found)
		return found;
	locked = 1;
	write_seqlock(&rename_lock);
	goto again;
L
Linus Torvalds 已提交
1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
}

/**
 * shrink_dcache_parent - prune dcache
 * @parent: parent of entries to prune
 *
 * Prune the dcache to remove unused children of the parent dentry.
 */
void shrink_dcache_parent(struct dentry * parent)
{
1250
	LIST_HEAD(dispose);
L
Linus Torvalds 已提交
1251 1252
	int found;

1253 1254
	while ((found = select_parent(parent, &dispose)) != 0)
		shrink_dentry_list(&dispose);
L
Linus Torvalds 已提交
1255
}
1256
EXPORT_SYMBOL(shrink_dcache_parent);
L
Linus Torvalds 已提交
1257 1258

/**
1259 1260
 * __d_alloc	-	allocate a dcache entry
 * @sb: filesystem it will belong to
L
Linus Torvalds 已提交
1261 1262 1263 1264 1265 1266 1267
 * @name: qstr of the name
 *
 * Allocates a dentry. It returns %NULL if there is insufficient memory
 * available. On a success the dentry is returned. The name passed in is
 * copied and the copy passed in may be reused after this call.
 */
 
1268
struct dentry *__d_alloc(struct super_block *sb, const struct qstr *name)
L
Linus Torvalds 已提交
1269 1270 1271 1272
{
	struct dentry *dentry;
	char *dname;

1273
	dentry = kmem_cache_alloc(dentry_cache, GFP_KERNEL);
L
Linus Torvalds 已提交
1274 1275 1276
	if (!dentry)
		return NULL;

1277 1278 1279 1280 1281 1282 1283
	/*
	 * We guarantee that the inline name is always NUL-terminated.
	 * This way the memcpy() done by the name switching in rename
	 * will still always have a NUL at the end, even if we might
	 * be overwriting an internal NUL character
	 */
	dentry->d_iname[DNAME_INLINE_LEN-1] = 0;
L
Linus Torvalds 已提交
1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298
	if (name->len > DNAME_INLINE_LEN-1) {
		dname = kmalloc(name->len + 1, GFP_KERNEL);
		if (!dname) {
			kmem_cache_free(dentry_cache, dentry); 
			return NULL;
		}
	} else  {
		dname = dentry->d_iname;
	}	

	dentry->d_name.len = name->len;
	dentry->d_name.hash = name->hash;
	memcpy(dname, name->name, name->len);
	dname[name->len] = 0;

1299 1300 1301 1302
	/* Make sure we always see the terminating NUL character */
	smp_wmb();
	dentry->d_name.name = dname;

N
Nick Piggin 已提交
1303
	dentry->d_count = 1;
1304
	dentry->d_flags = 0;
L
Linus Torvalds 已提交
1305
	spin_lock_init(&dentry->d_lock);
N
Nick Piggin 已提交
1306
	seqcount_init(&dentry->d_seq);
L
Linus Torvalds 已提交
1307
	dentry->d_inode = NULL;
1308 1309
	dentry->d_parent = dentry;
	dentry->d_sb = sb;
L
Linus Torvalds 已提交
1310 1311
	dentry->d_op = NULL;
	dentry->d_fsdata = NULL;
1312
	INIT_HLIST_BL_NODE(&dentry->d_hash);
L
Linus Torvalds 已提交
1313 1314 1315
	INIT_LIST_HEAD(&dentry->d_lru);
	INIT_LIST_HEAD(&dentry->d_subdirs);
	INIT_LIST_HEAD(&dentry->d_alias);
N
Nick Piggin 已提交
1316
	INIT_LIST_HEAD(&dentry->d_u.d_child);
1317
	d_set_d_op(dentry, dentry->d_sb->s_d_op);
L
Linus Torvalds 已提交
1318

1319
	this_cpu_inc(nr_dentry);
1320

L
Linus Torvalds 已提交
1321 1322
	return dentry;
}
1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350

/**
 * d_alloc	-	allocate a dcache entry
 * @parent: parent of entry to allocate
 * @name: qstr of the name
 *
 * Allocates a dentry. It returns %NULL if there is insufficient memory
 * available. On a success the dentry is returned. The name passed in is
 * copied and the copy passed in may be reused after this call.
 */
struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
{
	struct dentry *dentry = __d_alloc(parent->d_sb, name);
	if (!dentry)
		return NULL;

	spin_lock(&parent->d_lock);
	/*
	 * don't need child lock because it is not subject
	 * to concurrency here
	 */
	__dget_dlock(parent);
	dentry->d_parent = parent;
	list_add(&dentry->d_u.d_child, &parent->d_subdirs);
	spin_unlock(&parent->d_lock);

	return dentry;
}
1351
EXPORT_SYMBOL(d_alloc);
L
Linus Torvalds 已提交
1352

1353 1354
struct dentry *d_alloc_pseudo(struct super_block *sb, const struct qstr *name)
{
1355 1356
	struct dentry *dentry = __d_alloc(sb, name);
	if (dentry)
1357 1358 1359 1360 1361
		dentry->d_flags |= DCACHE_DISCONNECTED;
	return dentry;
}
EXPORT_SYMBOL(d_alloc_pseudo);

L
Linus Torvalds 已提交
1362 1363 1364 1365 1366 1367 1368 1369 1370
struct dentry *d_alloc_name(struct dentry *parent, const char *name)
{
	struct qstr q;

	q.name = name;
	q.len = strlen(name);
	q.hash = full_name_hash(q.name, q.len);
	return d_alloc(parent, &q);
}
1371
EXPORT_SYMBOL(d_alloc_name);
L
Linus Torvalds 已提交
1372

1373 1374
void d_set_d_op(struct dentry *dentry, const struct dentry_operations *op)
{
1375 1376
	WARN_ON_ONCE(dentry->d_op);
	WARN_ON_ONCE(dentry->d_flags & (DCACHE_OP_HASH	|
1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390
				DCACHE_OP_COMPARE	|
				DCACHE_OP_REVALIDATE	|
				DCACHE_OP_DELETE ));
	dentry->d_op = op;
	if (!op)
		return;
	if (op->d_hash)
		dentry->d_flags |= DCACHE_OP_HASH;
	if (op->d_compare)
		dentry->d_flags |= DCACHE_OP_COMPARE;
	if (op->d_revalidate)
		dentry->d_flags |= DCACHE_OP_REVALIDATE;
	if (op->d_delete)
		dentry->d_flags |= DCACHE_OP_DELETE;
S
Sage Weil 已提交
1391 1392
	if (op->d_prune)
		dentry->d_flags |= DCACHE_OP_PRUNE;
1393 1394 1395 1396

}
EXPORT_SYMBOL(d_set_d_op);

1397 1398
static void __d_instantiate(struct dentry *dentry, struct inode *inode)
{
N
Nick Piggin 已提交
1399
	spin_lock(&dentry->d_lock);
1400 1401 1402
	if (inode) {
		if (unlikely(IS_AUTOMOUNT(inode)))
			dentry->d_flags |= DCACHE_NEED_AUTOMOUNT;
1403
		list_add(&dentry->d_alias, &inode->i_dentry);
1404
	}
1405
	dentry->d_inode = inode;
N
Nick Piggin 已提交
1406
	dentry_rcuwalk_barrier(dentry);
N
Nick Piggin 已提交
1407
	spin_unlock(&dentry->d_lock);
1408 1409 1410
	fsnotify_d_instantiate(dentry, inode);
}

L
Linus Torvalds 已提交
1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427
/**
 * d_instantiate - fill in inode information for a dentry
 * @entry: dentry to complete
 * @inode: inode to attach to this dentry
 *
 * Fill in inode information in the entry.
 *
 * This turns negative dentries into productive full members
 * of society.
 *
 * NOTE! This assumes that the inode count has been incremented
 * (or otherwise set) by the caller to indicate that it is now
 * in use by the dcache.
 */
 
void d_instantiate(struct dentry *entry, struct inode * inode)
{
1428
	BUG_ON(!list_empty(&entry->d_alias));
1429 1430
	if (inode)
		spin_lock(&inode->i_lock);
1431
	__d_instantiate(entry, inode);
1432 1433
	if (inode)
		spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1434 1435
	security_d_instantiate(entry, inode);
}
1436
EXPORT_SYMBOL(d_instantiate);
L
Linus Torvalds 已提交
1437 1438 1439 1440 1441 1442 1443 1444

/**
 * d_instantiate_unique - instantiate a non-aliased dentry
 * @entry: dentry to instantiate
 * @inode: inode to attach to this dentry
 *
 * Fill in inode information in the entry. On success, it returns NULL.
 * If an unhashed alias of "entry" already exists, then we return the
1445
 * aliased dentry instead and drop one reference to inode.
L
Linus Torvalds 已提交
1446 1447 1448
 *
 * Note that in order to avoid conflicts with rename() etc, the caller
 * had better be holding the parent directory semaphore.
1449 1450 1451 1452
 *
 * This also assumes that the inode count has been incremented
 * (or otherwise set) by the caller to indicate that it is now
 * in use by the dcache.
L
Linus Torvalds 已提交
1453
 */
1454 1455
static struct dentry *__d_instantiate_unique(struct dentry *entry,
					     struct inode *inode)
L
Linus Torvalds 已提交
1456 1457 1458 1459 1460 1461
{
	struct dentry *alias;
	int len = entry->d_name.len;
	const char *name = entry->d_name.name;
	unsigned int hash = entry->d_name.hash;

1462
	if (!inode) {
1463
		__d_instantiate(entry, NULL);
1464 1465 1466
		return NULL;
	}

L
Linus Torvalds 已提交
1467
	list_for_each_entry(alias, &inode->i_dentry, d_alias) {
N
Nick Piggin 已提交
1468 1469 1470 1471 1472
		/*
		 * Don't need alias->d_lock here, because aliases with
		 * d_parent == entry->d_parent are not subject to name or
		 * parent changes, because the parent inode i_mutex is held.
		 */
1473
		if (alias->d_name.hash != hash)
L
Linus Torvalds 已提交
1474 1475 1476
			continue;
		if (alias->d_parent != entry->d_parent)
			continue;
1477 1478
		if (alias->d_name.len != len)
			continue;
1479
		if (dentry_cmp(alias, name, len))
L
Linus Torvalds 已提交
1480
			continue;
1481
		__dget(alias);
L
Linus Torvalds 已提交
1482 1483
		return alias;
	}
1484

1485
	__d_instantiate(entry, inode);
L
Linus Torvalds 已提交
1486 1487
	return NULL;
}
1488 1489 1490 1491 1492 1493 1494

struct dentry *d_instantiate_unique(struct dentry *entry, struct inode *inode)
{
	struct dentry *result;

	BUG_ON(!list_empty(&entry->d_alias));

1495 1496
	if (inode)
		spin_lock(&inode->i_lock);
1497
	result = __d_instantiate_unique(entry, inode);
1498 1499
	if (inode)
		spin_unlock(&inode->i_lock);
1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510

	if (!result) {
		security_d_instantiate(entry, inode);
		return NULL;
	}

	BUG_ON(!d_unhashed(result));
	iput(inode);
	return result;
}

L
Linus Torvalds 已提交
1511 1512
EXPORT_SYMBOL(d_instantiate_unique);

A
Al Viro 已提交
1513 1514 1515 1516 1517
struct dentry *d_make_root(struct inode *root_inode)
{
	struct dentry *res = NULL;

	if (root_inode) {
1518
		static const struct qstr name = QSTR_INIT("/", 1);
A
Al Viro 已提交
1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529

		res = __d_alloc(root_inode->i_sb, &name);
		if (res)
			d_instantiate(res, root_inode);
		else
			iput(root_inode);
	}
	return res;
}
EXPORT_SYMBOL(d_make_root);

1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540
static struct dentry * __d_find_any_alias(struct inode *inode)
{
	struct dentry *alias;

	if (list_empty(&inode->i_dentry))
		return NULL;
	alias = list_first_entry(&inode->i_dentry, struct dentry, d_alias);
	__dget(alias);
	return alias;
}

1541 1542 1543 1544 1545 1546 1547 1548
/**
 * d_find_any_alias - find any alias for a given inode
 * @inode: inode to find an alias for
 *
 * If any aliases exist for the given inode, take and return a
 * reference for one of them.  If no aliases exist, return %NULL.
 */
struct dentry *d_find_any_alias(struct inode *inode)
1549 1550 1551 1552 1553 1554 1555 1556
{
	struct dentry *de;

	spin_lock(&inode->i_lock);
	de = __d_find_any_alias(inode);
	spin_unlock(&inode->i_lock);
	return de;
}
1557
EXPORT_SYMBOL(d_find_any_alias);
1558

1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571
/**
 * d_obtain_alias - find or allocate a dentry for a given inode
 * @inode: inode to allocate the dentry for
 *
 * Obtain a dentry for an inode resulting from NFS filehandle conversion or
 * similar open by handle operations.  The returned dentry may be anonymous,
 * or may have a full name (if the inode was already in the cache).
 *
 * When called on a directory inode, we must ensure that the inode only ever
 * has one dentry.  If a dentry is found, that is returned instead of
 * allocating a new one.
 *
 * On successful return, the reference to the inode has been transferred
1572 1573 1574 1575
 * to the dentry.  In case of an error the reference on the inode is released.
 * To make it easier to use in export operations a %NULL or IS_ERR inode may
 * be passed in and will be the error will be propagate to the return value,
 * with a %NULL @inode replaced by ERR_PTR(-ESTALE).
1576 1577 1578
 */
struct dentry *d_obtain_alias(struct inode *inode)
{
C
Christoph Hellwig 已提交
1579 1580 1581
	static const struct qstr anonstring = { .name = "" };
	struct dentry *tmp;
	struct dentry *res;
1582 1583

	if (!inode)
1584
		return ERR_PTR(-ESTALE);
1585 1586 1587
	if (IS_ERR(inode))
		return ERR_CAST(inode);

1588
	res = d_find_any_alias(inode);
C
Christoph Hellwig 已提交
1589 1590 1591
	if (res)
		goto out_iput;

1592
	tmp = __d_alloc(inode->i_sb, &anonstring);
C
Christoph Hellwig 已提交
1593 1594 1595
	if (!tmp) {
		res = ERR_PTR(-ENOMEM);
		goto out_iput;
1596
	}
N
Nick Piggin 已提交
1597

1598
	spin_lock(&inode->i_lock);
1599
	res = __d_find_any_alias(inode);
C
Christoph Hellwig 已提交
1600
	if (res) {
1601
		spin_unlock(&inode->i_lock);
C
Christoph Hellwig 已提交
1602 1603 1604 1605 1606 1607 1608 1609 1610
		dput(tmp);
		goto out_iput;
	}

	/* attach a disconnected dentry */
	spin_lock(&tmp->d_lock);
	tmp->d_inode = inode;
	tmp->d_flags |= DCACHE_DISCONNECTED;
	list_add(&tmp->d_alias, &inode->i_dentry);
1611
	hlist_bl_lock(&tmp->d_sb->s_anon);
1612
	hlist_bl_add_head(&tmp->d_hash, &tmp->d_sb->s_anon);
1613
	hlist_bl_unlock(&tmp->d_sb->s_anon);
C
Christoph Hellwig 已提交
1614
	spin_unlock(&tmp->d_lock);
1615
	spin_unlock(&inode->i_lock);
1616
	security_d_instantiate(tmp, inode);
C
Christoph Hellwig 已提交
1617 1618 1619 1620

	return tmp;

 out_iput:
1621 1622
	if (res && !IS_ERR(res))
		security_d_instantiate(res, inode);
C
Christoph Hellwig 已提交
1623 1624
	iput(inode);
	return res;
1625
}
1626
EXPORT_SYMBOL(d_obtain_alias);
L
Linus Torvalds 已提交
1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647

/**
 * d_splice_alias - splice a disconnected dentry into the tree if one exists
 * @inode:  the inode which may have a disconnected dentry
 * @dentry: a negative dentry which we want to point to the inode.
 *
 * If inode is a directory and has a 'disconnected' dentry (i.e. IS_ROOT and
 * DCACHE_DISCONNECTED), then d_move that in place of the given dentry
 * and return it, else simply d_add the inode to the dentry and return NULL.
 *
 * This is needed in the lookup routine of any filesystem that is exportable
 * (via knfsd) so that we can build dcache paths to directories effectively.
 *
 * If a dentry was found and moved, then it is returned.  Otherwise NULL
 * is returned.  This matches the expected return value of ->lookup.
 *
 */
struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
{
	struct dentry *new = NULL;

1648 1649 1650
	if (IS_ERR(inode))
		return ERR_CAST(inode);

1651
	if (inode && S_ISDIR(inode->i_mode)) {
1652
		spin_lock(&inode->i_lock);
1653
		new = __d_find_alias(inode, 1);
L
Linus Torvalds 已提交
1654
		if (new) {
1655
			BUG_ON(!(new->d_flags & DCACHE_DISCONNECTED));
1656
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1657 1658 1659 1660
			security_d_instantiate(new, inode);
			d_move(new, dentry);
			iput(inode);
		} else {
1661
			/* already taking inode->i_lock, so d_add() by hand */
1662
			__d_instantiate(dentry, inode);
1663
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1664 1665 1666 1667 1668 1669 1670
			security_d_instantiate(dentry, inode);
			d_rehash(dentry);
		}
	} else
		d_add(dentry, inode);
	return new;
}
1671
EXPORT_SYMBOL(d_splice_alias);
L
Linus Torvalds 已提交
1672

1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688
/**
 * d_add_ci - lookup or allocate new dentry with case-exact name
 * @inode:  the inode case-insensitive lookup has found
 * @dentry: the negative dentry that was passed to the parent's lookup func
 * @name:   the case-exact name to be associated with the returned dentry
 *
 * This is to avoid filling the dcache with case-insensitive names to the
 * same inode, only the actual correct case is stored in the dcache for
 * case-insensitive filesystems.
 *
 * For a case-insensitive lookup match and if the the case-exact dentry
 * already exists in in the dcache, use it and return it.
 *
 * If no entry exists with the exact case name, allocate new dentry with
 * the exact case, and return the spliced entry.
 */
1689
struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
1690 1691 1692 1693 1694 1695
			struct qstr *name)
{
	int error;
	struct dentry *found;
	struct dentry *new;

C
Christoph Hellwig 已提交
1696 1697 1698 1699
	/*
	 * First check if a dentry matching the name already exists,
	 * if not go ahead and create it now.
	 */
1700 1701 1702 1703 1704 1705 1706
	found = d_hash_and_lookup(dentry->d_parent, name);
	if (!found) {
		new = d_alloc(dentry->d_parent, name);
		if (!new) {
			error = -ENOMEM;
			goto err_out;
		}
C
Christoph Hellwig 已提交
1707

1708 1709 1710 1711 1712 1713 1714
		found = d_splice_alias(inode, new);
		if (found) {
			dput(new);
			return found;
		}
		return new;
	}
C
Christoph Hellwig 已提交
1715 1716 1717 1718 1719 1720 1721

	/*
	 * If a matching dentry exists, and it's not negative use it.
	 *
	 * Decrement the reference count to balance the iget() done
	 * earlier on.
	 */
1722 1723 1724 1725 1726 1727 1728 1729 1730
	if (found->d_inode) {
		if (unlikely(found->d_inode != inode)) {
			/* This can't happen because bad inodes are unhashed. */
			BUG_ON(!is_bad_inode(inode));
			BUG_ON(!is_bad_inode(found->d_inode));
		}
		iput(inode);
		return found;
	}
C
Christoph Hellwig 已提交
1731

1732
	/*
1733 1734
	 * We are going to instantiate this dentry, unhash it and clear the
	 * lookup flag so we can do that.
1735
	 */
1736 1737
	if (unlikely(d_need_lookup(found)))
		d_clear_need_lookup(found);
C
Christoph Hellwig 已提交
1738

1739 1740
	/*
	 * Negative dentry: instantiate it unless the inode is a directory and
C
Christoph Hellwig 已提交
1741
	 * already has a dentry.
1742
	 */
1743 1744 1745 1746
	new = d_splice_alias(inode, found);
	if (new) {
		dput(found);
		found = new;
1747
	}
1748
	return found;
1749 1750 1751 1752 1753

err_out:
	iput(inode);
	return ERR_PTR(error);
}
1754
EXPORT_SYMBOL(d_add_ci);
L
Linus Torvalds 已提交
1755

1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797
/*
 * Do the slow-case of the dentry name compare.
 *
 * Unlike the dentry_cmp() function, we need to atomically
 * load the name, length and inode information, so that the
 * filesystem can rely on them, and can use the 'name' and
 * 'len' information without worrying about walking off the
 * end of memory etc.
 *
 * Thus the read_seqcount_retry() and the "duplicate" info
 * in arguments (the low-level filesystem should not look
 * at the dentry inode or name contents directly, since
 * rename can change them while we're in RCU mode).
 */
enum slow_d_compare {
	D_COMP_OK,
	D_COMP_NOMATCH,
	D_COMP_SEQRETRY,
};

static noinline enum slow_d_compare slow_dentry_cmp(
		const struct dentry *parent,
		struct inode *inode,
		struct dentry *dentry,
		unsigned int seq,
		const struct qstr *name)
{
	int tlen = dentry->d_name.len;
	const char *tname = dentry->d_name.name;
	struct inode *i = dentry->d_inode;

	if (read_seqcount_retry(&dentry->d_seq, seq)) {
		cpu_relax();
		return D_COMP_SEQRETRY;
	}
	if (parent->d_op->d_compare(parent, inode,
				dentry, i,
				tlen, tname, name))
		return D_COMP_NOMATCH;
	return D_COMP_OK;
}

N
Nick Piggin 已提交
1798 1799 1800 1801
/**
 * __d_lookup_rcu - search for a dentry (racy, store-free)
 * @parent: parent dentry
 * @name: qstr of name we wish to find
1802
 * @seqp: returns d_seq value at the point where the dentry was found
N
Nick Piggin 已提交
1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823
 * @inode: returns dentry->d_inode when the inode was found valid.
 * Returns: dentry, or NULL
 *
 * __d_lookup_rcu is the dcache lookup function for rcu-walk name
 * resolution (store-free path walking) design described in
 * Documentation/filesystems/path-lookup.txt.
 *
 * This is not to be used outside core vfs.
 *
 * __d_lookup_rcu must only be used in rcu-walk mode, ie. with vfsmount lock
 * held, and rcu_read_lock held. The returned dentry must not be stored into
 * without taking d_lock and checking d_seq sequence count against @seq
 * returned here.
 *
 * A refcount may be taken on the found dentry with the __d_rcu_to_refcount
 * function.
 *
 * Alternatively, __d_lookup_rcu may be called again to look up the child of
 * the returned dentry, so long as its parent's seqlock is checked after the
 * child is looked up. Thus, an interlocking stepping of sequence lock checks
 * is formed, giving integrity down the path walk.
1824 1825 1826
 *
 * NOTE! The caller *has* to check the resulting dentry against the sequence
 * number we've returned before using any of the resulting dentry state!
N
Nick Piggin 已提交
1827
 */
1828 1829
struct dentry *__d_lookup_rcu(const struct dentry *parent,
				const struct qstr *name,
1830
				unsigned *seqp, struct inode *inode)
N
Nick Piggin 已提交
1831
{
1832
	u64 hashlen = name->hash_len;
N
Nick Piggin 已提交
1833
	const unsigned char *str = name->name;
1834
	struct hlist_bl_head *b = d_hash(parent, hashlen_hash(hashlen));
1835
	struct hlist_bl_node *node;
N
Nick Piggin 已提交
1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855
	struct dentry *dentry;

	/*
	 * Note: There is significant duplication with __d_lookup_rcu which is
	 * required to prevent single threaded performance regressions
	 * especially on architectures where smp_rmb (in seqcounts) are costly.
	 * Keep the two functions in sync.
	 */

	/*
	 * The hash list is protected using RCU.
	 *
	 * Carefully use d_seq when comparing a candidate dentry, to avoid
	 * races with d_move().
	 *
	 * It is possible that concurrent renames can mess up our list
	 * walk here and result in missing our dentry, resulting in the
	 * false-negative result. d_lookup() protects against concurrent
	 * renames using rename_lock seqlock.
	 *
1856
	 * See Documentation/filesystems/path-lookup.txt for more details.
N
Nick Piggin 已提交
1857
	 */
1858
	hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
1859
		unsigned seq;
N
Nick Piggin 已提交
1860 1861

seqretry:
1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877
		/*
		 * The dentry sequence count protects us from concurrent
		 * renames, and thus protects inode, parent and name fields.
		 *
		 * The caller must perform a seqcount check in order
		 * to do anything useful with the returned dentry,
		 * including using the 'd_inode' pointer.
		 *
		 * NOTE! We do a "raw" seqcount_begin here. That means that
		 * we don't wait for the sequence count to stabilize if it
		 * is in the middle of a sequence change. If we do the slow
		 * dentry compare, we will do seqretries until it is stable,
		 * and if we end up with a successful lookup, we actually
		 * want to exit RCU lookup anyway.
		 */
		seq = raw_seqcount_begin(&dentry->d_seq);
N
Nick Piggin 已提交
1878 1879
		if (dentry->d_parent != parent)
			continue;
1880 1881
		if (d_unhashed(dentry))
			continue;
1882 1883
		*seqp = seq;

1884
		if (unlikely(parent->d_flags & DCACHE_OP_COMPARE)) {
1885 1886
			if (dentry->d_name.hash != hashlen_hash(hashlen))
				continue;
1887 1888 1889 1890
			switch (slow_dentry_cmp(parent, inode, dentry, seq, name)) {
			case D_COMP_OK:
				return dentry;
			case D_COMP_NOMATCH:
N
Nick Piggin 已提交
1891
				continue;
1892 1893 1894
			default:
				goto seqretry;
			}
N
Nick Piggin 已提交
1895
		}
1896

1897
		if (dentry->d_name.hash_len != hashlen)
1898
			continue;
1899
		if (!dentry_cmp(dentry, str, hashlen_len(hashlen)))
1900
			return dentry;
N
Nick Piggin 已提交
1901 1902 1903 1904
	}
	return NULL;
}

L
Linus Torvalds 已提交
1905 1906 1907 1908
/**
 * d_lookup - search for a dentry
 * @parent: parent dentry
 * @name: qstr of name we wish to find
1909
 * Returns: dentry, or NULL
L
Linus Torvalds 已提交
1910
 *
1911 1912 1913 1914
 * d_lookup searches the children of the parent dentry for the name in
 * question. If the dentry is found its reference count is incremented and the
 * dentry is returned. The caller must use dput to free the entry when it has
 * finished using it. %NULL is returned if the dentry does not exist.
L
Linus Torvalds 已提交
1915
 */
N
Nick Piggin 已提交
1916
struct dentry *d_lookup(struct dentry *parent, struct qstr *name)
L
Linus Torvalds 已提交
1917
{
N
Nick Piggin 已提交
1918
	struct dentry *dentry;
1919
	unsigned seq;
L
Linus Torvalds 已提交
1920 1921 1922 1923 1924 1925 1926 1927 1928

        do {
                seq = read_seqbegin(&rename_lock);
                dentry = __d_lookup(parent, name);
                if (dentry)
			break;
	} while (read_seqretry(&rename_lock, seq));
	return dentry;
}
1929
EXPORT_SYMBOL(d_lookup);
L
Linus Torvalds 已提交
1930

N
Nick Piggin 已提交
1931
/**
1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945
 * __d_lookup - search for a dentry (racy)
 * @parent: parent dentry
 * @name: qstr of name we wish to find
 * Returns: dentry, or NULL
 *
 * __d_lookup is like d_lookup, however it may (rarely) return a
 * false-negative result due to unrelated rename activity.
 *
 * __d_lookup is slightly faster by avoiding rename_lock read seqlock,
 * however it must be used carefully, eg. with a following d_lookup in
 * the case of failure.
 *
 * __d_lookup callers must be commented.
 */
N
Nick Piggin 已提交
1946
struct dentry *__d_lookup(struct dentry *parent, struct qstr *name)
L
Linus Torvalds 已提交
1947 1948 1949 1950
{
	unsigned int len = name->len;
	unsigned int hash = name->hash;
	const unsigned char *str = name->name;
1951
	struct hlist_bl_head *b = d_hash(parent, hash);
1952
	struct hlist_bl_node *node;
N
Nick Piggin 已提交
1953
	struct dentry *found = NULL;
1954
	struct dentry *dentry;
L
Linus Torvalds 已提交
1955

N
Nick Piggin 已提交
1956 1957 1958 1959 1960 1961 1962
	/*
	 * Note: There is significant duplication with __d_lookup_rcu which is
	 * required to prevent single threaded performance regressions
	 * especially on architectures where smp_rmb (in seqcounts) are costly.
	 * Keep the two functions in sync.
	 */

1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973
	/*
	 * The hash list is protected using RCU.
	 *
	 * Take d_lock when comparing a candidate dentry, to avoid races
	 * with d_move().
	 *
	 * It is possible that concurrent renames can mess up our list
	 * walk here and result in missing our dentry, resulting in the
	 * false-negative result. d_lookup() protects against concurrent
	 * renames using rename_lock seqlock.
	 *
1974
	 * See Documentation/filesystems/path-lookup.txt for more details.
1975
	 */
L
Linus Torvalds 已提交
1976 1977
	rcu_read_lock();
	
1978
	hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
L
Linus Torvalds 已提交
1979 1980 1981 1982 1983 1984 1985

		if (dentry->d_name.hash != hash)
			continue;

		spin_lock(&dentry->d_lock);
		if (dentry->d_parent != parent)
			goto next;
1986 1987 1988
		if (d_unhashed(dentry))
			goto next;

L
Linus Torvalds 已提交
1989 1990 1991 1992
		/*
		 * It is safe to compare names since d_move() cannot
		 * change the qstr (protected by d_lock).
		 */
1993
		if (parent->d_flags & DCACHE_OP_COMPARE) {
1994 1995
			int tlen = dentry->d_name.len;
			const char *tname = dentry->d_name.name;
N
Nick Piggin 已提交
1996 1997
			if (parent->d_op->d_compare(parent, parent->d_inode,
						dentry, dentry->d_inode,
N
Nick Piggin 已提交
1998
						tlen, tname, name))
L
Linus Torvalds 已提交
1999 2000
				goto next;
		} else {
2001 2002
			if (dentry->d_name.len != len)
				goto next;
2003
			if (dentry_cmp(dentry, str, len))
L
Linus Torvalds 已提交
2004 2005 2006
				goto next;
		}

N
Nick Piggin 已提交
2007
		dentry->d_count++;
2008
		found = dentry;
L
Linus Torvalds 已提交
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
		spin_unlock(&dentry->d_lock);
		break;
next:
		spin_unlock(&dentry->d_lock);
 	}
 	rcu_read_unlock();

 	return found;
}

2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035
/**
 * d_hash_and_lookup - hash the qstr then search for a dentry
 * @dir: Directory to search in
 * @name: qstr of name we wish to find
 *
 * On hash failure or on lookup failure NULL is returned.
 */
struct dentry *d_hash_and_lookup(struct dentry *dir, struct qstr *name)
{
	struct dentry *dentry = NULL;

	/*
	 * Check for a fs-specific hash function. Note that we must
	 * calculate the standard hash first, as the d_op->d_hash()
	 * routine may choose to leave the hash value unchanged.
	 */
	name->hash = full_name_hash(name->name, name->len);
2036
	if (dir->d_flags & DCACHE_OP_HASH) {
N
Nick Piggin 已提交
2037
		if (dir->d_op->d_hash(dir, dir->d_inode, name) < 0)
2038 2039 2040 2041 2042 2043 2044
			goto out;
	}
	dentry = d_lookup(dir, name);
out:
	return dentry;
}

L
Linus Torvalds 已提交
2045
/**
N
Nick Piggin 已提交
2046
 * d_validate - verify dentry provided from insecure source (deprecated)
L
Linus Torvalds 已提交
2047
 * @dentry: The dentry alleged to be valid child of @dparent
2048
 * @dparent: The parent dentry (known to be valid)
L
Linus Torvalds 已提交
2049 2050 2051 2052
 *
 * An insecure source has sent us a dentry, here we verify it and dget() it.
 * This is used by ncpfs in its readdir implementation.
 * Zero is returned in the dentry is invalid.
N
Nick Piggin 已提交
2053 2054
 *
 * This function is slow for big directories, and deprecated, do not use it.
L
Linus Torvalds 已提交
2055
 */
2056
int d_validate(struct dentry *dentry, struct dentry *dparent)
L
Linus Torvalds 已提交
2057
{
N
Nick Piggin 已提交
2058
	struct dentry *child;
2059

N
Nick Piggin 已提交
2060
	spin_lock(&dparent->d_lock);
N
Nick Piggin 已提交
2061 2062
	list_for_each_entry(child, &dparent->d_subdirs, d_u.d_child) {
		if (dentry == child) {
N
Nick Piggin 已提交
2063
			spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
2064
			__dget_dlock(dentry);
N
Nick Piggin 已提交
2065 2066
			spin_unlock(&dentry->d_lock);
			spin_unlock(&dparent->d_lock);
L
Linus Torvalds 已提交
2067 2068 2069
			return 1;
		}
	}
N
Nick Piggin 已提交
2070
	spin_unlock(&dparent->d_lock);
N
Nick Piggin 已提交
2071

L
Linus Torvalds 已提交
2072 2073
	return 0;
}
2074
EXPORT_SYMBOL(d_validate);
L
Linus Torvalds 已提交
2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098

/*
 * When a file is deleted, we have two options:
 * - turn this dentry into a negative dentry
 * - unhash this dentry and free it.
 *
 * Usually, we want to just turn this into
 * a negative dentry, but if anybody else is
 * currently using the dentry or the inode
 * we can't do that and we fall back on removing
 * it from the hash queues and waiting for
 * it to be deleted later when it has no users
 */
 
/**
 * d_delete - delete a dentry
 * @dentry: The dentry to delete
 *
 * Turn the dentry into a negative dentry if possible, otherwise
 * remove it from the hash queues so it can be deleted later
 */
 
void d_delete(struct dentry * dentry)
{
2099
	struct inode *inode;
2100
	int isdir = 0;
L
Linus Torvalds 已提交
2101 2102 2103
	/*
	 * Are we the only user?
	 */
2104
again:
L
Linus Torvalds 已提交
2105
	spin_lock(&dentry->d_lock);
2106 2107
	inode = dentry->d_inode;
	isdir = S_ISDIR(inode->i_mode);
N
Nick Piggin 已提交
2108
	if (dentry->d_count == 1) {
2109
		if (inode && !spin_trylock(&inode->i_lock)) {
2110 2111 2112 2113
			spin_unlock(&dentry->d_lock);
			cpu_relax();
			goto again;
		}
A
Al Viro 已提交
2114
		dentry->d_flags &= ~DCACHE_CANT_MOUNT;
N
Nick Piggin 已提交
2115
		dentry_unlink_inode(dentry);
2116
		fsnotify_nameremove(dentry, isdir);
L
Linus Torvalds 已提交
2117 2118 2119 2120 2121 2122 2123
		return;
	}

	if (!d_unhashed(dentry))
		__d_drop(dentry);

	spin_unlock(&dentry->d_lock);
2124 2125

	fsnotify_nameremove(dentry, isdir);
L
Linus Torvalds 已提交
2126
}
2127
EXPORT_SYMBOL(d_delete);
L
Linus Torvalds 已提交
2128

2129
static void __d_rehash(struct dentry * entry, struct hlist_bl_head *b)
L
Linus Torvalds 已提交
2130
{
2131
	BUG_ON(!d_unhashed(entry));
2132
	hlist_bl_lock(b);
2133
	entry->d_flags |= DCACHE_RCUACCESS;
2134
	hlist_bl_add_head_rcu(&entry->d_hash, b);
2135
	hlist_bl_unlock(b);
L
Linus Torvalds 已提交
2136 2137
}

2138 2139 2140 2141 2142
static void _d_rehash(struct dentry * entry)
{
	__d_rehash(entry, d_hash(entry->d_parent, entry->d_name.hash));
}

L
Linus Torvalds 已提交
2143 2144 2145 2146 2147 2148 2149 2150 2151 2152
/**
 * d_rehash	- add an entry back to the hash
 * @entry: dentry to add to the hash
 *
 * Adds a dentry to the hash according to its name.
 */
 
void d_rehash(struct dentry * entry)
{
	spin_lock(&entry->d_lock);
2153
	_d_rehash(entry);
L
Linus Torvalds 已提交
2154 2155
	spin_unlock(&entry->d_lock);
}
2156
EXPORT_SYMBOL(d_rehash);
L
Linus Torvalds 已提交
2157

N
Nick Piggin 已提交
2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173
/**
 * dentry_update_name_case - update case insensitive dentry with a new name
 * @dentry: dentry to be updated
 * @name: new name
 *
 * Update a case insensitive dentry with new case of name.
 *
 * dentry must have been returned by d_lookup with name @name. Old and new
 * name lengths must match (ie. no d_compare which allows mismatched name
 * lengths).
 *
 * Parent inode i_mutex must be held over d_lookup and into this call (to
 * keep renames and concurrent inserts, and readdir(2) away).
 */
void dentry_update_name_case(struct dentry *dentry, struct qstr *name)
{
2174
	BUG_ON(!mutex_is_locked(&dentry->d_parent->d_inode->i_mutex));
N
Nick Piggin 已提交
2175 2176 2177
	BUG_ON(dentry->d_name.len != name->len); /* d_lookup gives this */

	spin_lock(&dentry->d_lock);
N
Nick Piggin 已提交
2178
	write_seqcount_begin(&dentry->d_seq);
N
Nick Piggin 已提交
2179
	memcpy((unsigned char *)dentry->d_name.name, name->name, name->len);
N
Nick Piggin 已提交
2180
	write_seqcount_end(&dentry->d_seq);
N
Nick Piggin 已提交
2181 2182 2183 2184
	spin_unlock(&dentry->d_lock);
}
EXPORT_SYMBOL(dentry_update_name_case);

L
Linus Torvalds 已提交
2185 2186 2187 2188 2189 2190 2191
static void switch_names(struct dentry *dentry, struct dentry *target)
{
	if (dname_external(target)) {
		if (dname_external(dentry)) {
			/*
			 * Both external: swap the pointers
			 */
2192
			swap(target->d_name.name, dentry->d_name.name);
L
Linus Torvalds 已提交
2193 2194 2195 2196 2197
		} else {
			/*
			 * dentry:internal, target:external.  Steal target's
			 * storage and make target internal.
			 */
2198 2199
			memcpy(target->d_iname, dentry->d_name.name,
					dentry->d_name.len + 1);
L
Linus Torvalds 已提交
2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218
			dentry->d_name.name = target->d_name.name;
			target->d_name.name = target->d_iname;
		}
	} else {
		if (dname_external(dentry)) {
			/*
			 * dentry:external, target:internal.  Give dentry's
			 * storage to target and make dentry internal
			 */
			memcpy(dentry->d_iname, target->d_name.name,
					target->d_name.len + 1);
			target->d_name.name = dentry->d_name.name;
			dentry->d_name.name = dentry->d_iname;
		} else {
			/*
			 * Both are internal.  Just copy target to dentry
			 */
			memcpy(dentry->d_iname, target->d_name.name,
					target->d_name.len + 1);
2219 2220
			dentry->d_name.len = target->d_name.len;
			return;
L
Linus Torvalds 已提交
2221 2222
		}
	}
2223
	swap(dentry->d_name.len, target->d_name.len);
L
Linus Torvalds 已提交
2224 2225
}

N
Nick Piggin 已提交
2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261
static void dentry_lock_for_move(struct dentry *dentry, struct dentry *target)
{
	/*
	 * XXXX: do we really need to take target->d_lock?
	 */
	if (IS_ROOT(dentry) || dentry->d_parent == target->d_parent)
		spin_lock(&target->d_parent->d_lock);
	else {
		if (d_ancestor(dentry->d_parent, target->d_parent)) {
			spin_lock(&dentry->d_parent->d_lock);
			spin_lock_nested(&target->d_parent->d_lock,
						DENTRY_D_LOCK_NESTED);
		} else {
			spin_lock(&target->d_parent->d_lock);
			spin_lock_nested(&dentry->d_parent->d_lock,
						DENTRY_D_LOCK_NESTED);
		}
	}
	if (target < dentry) {
		spin_lock_nested(&target->d_lock, 2);
		spin_lock_nested(&dentry->d_lock, 3);
	} else {
		spin_lock_nested(&dentry->d_lock, 2);
		spin_lock_nested(&target->d_lock, 3);
	}
}

static void dentry_unlock_parents_for_move(struct dentry *dentry,
					struct dentry *target)
{
	if (target->d_parent != dentry->d_parent)
		spin_unlock(&dentry->d_parent->d_lock);
	if (target->d_parent != target)
		spin_unlock(&target->d_parent->d_lock);
}

L
Linus Torvalds 已提交
2262
/*
N
Nick Piggin 已提交
2263 2264 2265 2266 2267 2268 2269 2270 2271
 * When switching names, the actual string doesn't strictly have to
 * be preserved in the target - because we're dropping the target
 * anyway. As such, we can just do a simple memcpy() to copy over
 * the new name before we switch.
 *
 * Note that we have to be a lot more careful about getting the hash
 * switched - we have to switch the hash value properly even if it
 * then no longer matches the actual (corrupted) string of the target.
 * The hash value has to match the hash queue that the dentry is on..
L
Linus Torvalds 已提交
2272
 */
2273
/*
2274
 * __d_move - move a dentry
L
Linus Torvalds 已提交
2275 2276 2277 2278
 * @dentry: entry to move
 * @target: new dentry
 *
 * Update the dcache to reflect the move of a file name. Negative
2279 2280 2281
 * dcache entries should not be moved in this way. Caller must hold
 * rename_lock, the i_mutex of the source and target directories,
 * and the sb->s_vfs_rename_mutex if they differ. See lock_rename().
L
Linus Torvalds 已提交
2282
 */
2283
static void __d_move(struct dentry * dentry, struct dentry * target)
L
Linus Torvalds 已提交
2284 2285 2286 2287
{
	if (!dentry->d_inode)
		printk(KERN_WARNING "VFS: moving negative dcache entry\n");

N
Nick Piggin 已提交
2288 2289 2290 2291
	BUG_ON(d_ancestor(dentry, target));
	BUG_ON(d_ancestor(target, dentry));

	dentry_lock_for_move(dentry, target);
L
Linus Torvalds 已提交
2292

N
Nick Piggin 已提交
2293 2294 2295
	write_seqcount_begin(&dentry->d_seq);
	write_seqcount_begin(&target->d_seq);

2296 2297 2298 2299 2300 2301 2302
	/* __d_drop does write_seqcount_barrier, but they're OK to nest. */

	/*
	 * Move the dentry to the target hash queue. Don't bother checking
	 * for the same hash queue because of how unlikely it is.
	 */
	__d_drop(dentry);
N
Nick Piggin 已提交
2303
	__d_rehash(dentry, d_hash(target->d_parent, target->d_name.hash));
L
Linus Torvalds 已提交
2304 2305 2306 2307

	/* Unhash the target: dput() will then get rid of it */
	__d_drop(target);

E
Eric Dumazet 已提交
2308 2309
	list_del(&dentry->d_u.d_child);
	list_del(&target->d_u.d_child);
L
Linus Torvalds 已提交
2310 2311 2312

	/* Switch the names.. */
	switch_names(dentry, target);
2313
	swap(dentry->d_name.hash, target->d_name.hash);
L
Linus Torvalds 已提交
2314 2315 2316 2317 2318

	/* ... and switch the parents */
	if (IS_ROOT(dentry)) {
		dentry->d_parent = target->d_parent;
		target->d_parent = target;
E
Eric Dumazet 已提交
2319
		INIT_LIST_HEAD(&target->d_u.d_child);
L
Linus Torvalds 已提交
2320
	} else {
2321
		swap(dentry->d_parent, target->d_parent);
L
Linus Torvalds 已提交
2322 2323

		/* And add them back to the (new) parent lists */
E
Eric Dumazet 已提交
2324
		list_add(&target->d_u.d_child, &target->d_parent->d_subdirs);
L
Linus Torvalds 已提交
2325 2326
	}

E
Eric Dumazet 已提交
2327
	list_add(&dentry->d_u.d_child, &dentry->d_parent->d_subdirs);
N
Nick Piggin 已提交
2328

N
Nick Piggin 已提交
2329 2330 2331
	write_seqcount_end(&target->d_seq);
	write_seqcount_end(&dentry->d_seq);

N
Nick Piggin 已提交
2332
	dentry_unlock_parents_for_move(dentry, target);
L
Linus Torvalds 已提交
2333
	spin_unlock(&target->d_lock);
2334
	fsnotify_d_move(dentry);
L
Linus Torvalds 已提交
2335
	spin_unlock(&dentry->d_lock);
2336 2337 2338 2339 2340 2341 2342 2343
}

/*
 * d_move - move a dentry
 * @dentry: entry to move
 * @target: new dentry
 *
 * Update the dcache to reflect the move of a file name. Negative
2344 2345
 * dcache entries should not be moved in this way. See the locking
 * requirements for __d_move.
2346 2347 2348 2349 2350
 */
void d_move(struct dentry *dentry, struct dentry *target)
{
	write_seqlock(&rename_lock);
	__d_move(dentry, target);
L
Linus Torvalds 已提交
2351
	write_sequnlock(&rename_lock);
2352
}
2353
EXPORT_SYMBOL(d_move);
L
Linus Torvalds 已提交
2354

2355 2356 2357 2358 2359 2360 2361
/**
 * d_ancestor - search for an ancestor
 * @p1: ancestor dentry
 * @p2: child dentry
 *
 * Returns the ancestor dentry of p2 which is a child of p1, if p1 is
 * an ancestor of p2, else NULL.
2362
 */
2363
struct dentry *d_ancestor(struct dentry *p1, struct dentry *p2)
2364 2365 2366
{
	struct dentry *p;

2367
	for (p = p2; !IS_ROOT(p); p = p->d_parent) {
2368
		if (p->d_parent == p1)
2369
			return p;
2370
	}
2371
	return NULL;
2372 2373 2374 2375 2376 2377
}

/*
 * This helper attempts to cope with remotely renamed directories
 *
 * It assumes that the caller is already holding
2378
 * dentry->d_parent->d_inode->i_mutex, inode->i_lock and rename_lock
2379 2380 2381 2382
 *
 * Note: If ever the locking in lock_rename() changes, then please
 * remember to update this too...
 */
2383 2384
static struct dentry *__d_unalias(struct inode *inode,
		struct dentry *dentry, struct dentry *alias)
2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401
{
	struct mutex *m1 = NULL, *m2 = NULL;
	struct dentry *ret;

	/* If alias and dentry share a parent, then no extra locks required */
	if (alias->d_parent == dentry->d_parent)
		goto out_unalias;

	/* See lock_rename() */
	ret = ERR_PTR(-EBUSY);
	if (!mutex_trylock(&dentry->d_sb->s_vfs_rename_mutex))
		goto out_err;
	m1 = &dentry->d_sb->s_vfs_rename_mutex;
	if (!mutex_trylock(&alias->d_parent->d_inode->i_mutex))
		goto out_err;
	m2 = &alias->d_parent->d_inode->i_mutex;
out_unalias:
2402
	__d_move(alias, dentry);
2403 2404
	ret = alias;
out_err:
2405
	spin_unlock(&inode->i_lock);
2406 2407 2408 2409 2410 2411 2412
	if (m2)
		mutex_unlock(m2);
	if (m1)
		mutex_unlock(m1);
	return ret;
}

2413 2414 2415
/*
 * Prepare an anonymous dentry for life in the superblock's dentry tree as a
 * named dentry in place of the dentry to be replaced.
N
Nick Piggin 已提交
2416
 * returns with anon->d_lock held!
2417 2418 2419 2420 2421
 */
static void __d_materialise_dentry(struct dentry *dentry, struct dentry *anon)
{
	struct dentry *dparent, *aparent;

N
Nick Piggin 已提交
2422
	dentry_lock_for_move(anon, dentry);
2423

N
Nick Piggin 已提交
2424 2425 2426
	write_seqcount_begin(&dentry->d_seq);
	write_seqcount_begin(&anon->d_seq);

2427 2428 2429
	dparent = dentry->d_parent;
	aparent = anon->d_parent;

N
Nick Piggin 已提交
2430 2431 2432
	switch_names(dentry, anon);
	swap(dentry->d_name.hash, anon->d_name.hash);

2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446
	dentry->d_parent = (aparent == anon) ? dentry : aparent;
	list_del(&dentry->d_u.d_child);
	if (!IS_ROOT(dentry))
		list_add(&dentry->d_u.d_child, &dentry->d_parent->d_subdirs);
	else
		INIT_LIST_HEAD(&dentry->d_u.d_child);

	anon->d_parent = (dparent == dentry) ? anon : dparent;
	list_del(&anon->d_u.d_child);
	if (!IS_ROOT(anon))
		list_add(&anon->d_u.d_child, &anon->d_parent->d_subdirs);
	else
		INIT_LIST_HEAD(&anon->d_u.d_child);

N
Nick Piggin 已提交
2447 2448 2449
	write_seqcount_end(&dentry->d_seq);
	write_seqcount_end(&anon->d_seq);

N
Nick Piggin 已提交
2450 2451 2452 2453
	dentry_unlock_parents_for_move(anon, dentry);
	spin_unlock(&dentry->d_lock);

	/* anon->d_lock still locked, returns locked */
2454 2455 2456 2457 2458 2459 2460 2461 2462
	anon->d_flags &= ~DCACHE_DISCONNECTED;
}

/**
 * d_materialise_unique - introduce an inode into the tree
 * @dentry: candidate dentry
 * @inode: inode to bind to the dentry, to which aliases may be attached
 *
 * Introduces an dentry into the tree, substituting an extant disconnected
2463 2464
 * root directory alias in its place if there is one. Caller must hold the
 * i_mutex of the parent directory.
2465 2466 2467
 */
struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
{
2468
	struct dentry *actual;
2469 2470 2471 2472 2473

	BUG_ON(!d_unhashed(dentry));

	if (!inode) {
		actual = dentry;
2474
		__d_instantiate(dentry, NULL);
2475 2476
		d_rehash(actual);
		goto out_nolock;
2477 2478
	}

2479
	spin_lock(&inode->i_lock);
2480

2481 2482 2483 2484
	if (S_ISDIR(inode->i_mode)) {
		struct dentry *alias;

		/* Does an aliased dentry already exist? */
2485
		alias = __d_find_alias(inode, 0);
2486 2487
		if (alias) {
			actual = alias;
2488 2489 2490 2491 2492
			write_seqlock(&rename_lock);

			if (d_ancestor(alias, dentry)) {
				/* Check for loops */
				actual = ERR_PTR(-ELOOP);
2493
				spin_unlock(&inode->i_lock);
2494 2495 2496
			} else if (IS_ROOT(alias)) {
				/* Is this an anonymous mountpoint that we
				 * could splice into our tree? */
2497
				__d_materialise_dentry(dentry, alias);
2498
				write_sequnlock(&rename_lock);
2499 2500
				__d_drop(alias);
				goto found;
2501 2502
			} else {
				/* Nope, but we must(!) avoid directory
2503
				 * aliasing. This drops inode->i_lock */
2504
				actual = __d_unalias(inode, dentry, alias);
2505
			}
2506
			write_sequnlock(&rename_lock);
2507 2508 2509 2510 2511 2512 2513 2514
			if (IS_ERR(actual)) {
				if (PTR_ERR(actual) == -ELOOP)
					pr_warn_ratelimited(
						"VFS: Lookup of '%s' in %s %s"
						" would have caused loop\n",
						dentry->d_name.name,
						inode->i_sb->s_type->name,
						inode->i_sb->s_id);
2515
				dput(alias);
2516
			}
2517 2518
			goto out_nolock;
		}
2519 2520 2521 2522 2523 2524
	}

	/* Add a unique reference */
	actual = __d_instantiate_unique(dentry, inode);
	if (!actual)
		actual = dentry;
2525 2526
	else
		BUG_ON(!d_unhashed(actual));
2527 2528 2529 2530 2531

	spin_lock(&actual->d_lock);
found:
	_d_rehash(actual);
	spin_unlock(&actual->d_lock);
2532
	spin_unlock(&inode->i_lock);
2533
out_nolock:
2534 2535 2536 2537 2538 2539 2540 2541
	if (actual == dentry) {
		security_d_instantiate(dentry, inode);
		return NULL;
	}

	iput(inode);
	return actual;
}
2542
EXPORT_SYMBOL_GPL(d_materialise_unique);
2543

2544
static int prepend(char **buffer, int *buflen, const char *str, int namelen)
2545 2546 2547 2548 2549 2550 2551 2552 2553
{
	*buflen -= namelen;
	if (*buflen < 0)
		return -ENAMETOOLONG;
	*buffer -= namelen;
	memcpy(*buffer, str, namelen);
	return 0;
}

2554 2555 2556 2557 2558
static int prepend_name(char **buffer, int *buflen, struct qstr *name)
{
	return prepend(buffer, buflen, name->name, name->len);
}

L
Linus Torvalds 已提交
2559
/**
2560
 * prepend_path - Prepend path string to a buffer
2561
 * @path: the dentry/vfsmount to report
2562
 * @root: root vfsmnt/dentry
M
Miklos Szeredi 已提交
2563 2564
 * @buffer: pointer to the end of the buffer
 * @buflen: pointer to buffer length
2565
 *
2566
 * Caller holds the rename_lock.
L
Linus Torvalds 已提交
2567
 */
2568 2569
static int prepend_path(const struct path *path,
			const struct path *root,
M
Miklos Szeredi 已提交
2570
			char **buffer, int *buflen)
L
Linus Torvalds 已提交
2571
{
2572 2573
	struct dentry *dentry = path->dentry;
	struct vfsmount *vfsmnt = path->mnt;
2574
	struct mount *mnt = real_mount(vfsmnt);
M
Miklos Szeredi 已提交
2575 2576
	bool slash = false;
	int error = 0;
2577

A
Andi Kleen 已提交
2578
	br_read_lock(&vfsmount_lock);
M
Miklos Szeredi 已提交
2579
	while (dentry != root->dentry || vfsmnt != root->mnt) {
L
Linus Torvalds 已提交
2580 2581 2582
		struct dentry * parent;

		if (dentry == vfsmnt->mnt_root || IS_ROOT(dentry)) {
2583
			/* Global root? */
2584
			if (!mnt_has_parent(mnt))
L
Linus Torvalds 已提交
2585
				goto global_root;
2586
			dentry = mnt->mnt_mountpoint;
2587 2588
			mnt = mnt->mnt_parent;
			vfsmnt = &mnt->mnt;
L
Linus Torvalds 已提交
2589 2590 2591 2592
			continue;
		}
		parent = dentry->d_parent;
		prefetch(parent);
N
Nick Piggin 已提交
2593
		spin_lock(&dentry->d_lock);
M
Miklos Szeredi 已提交
2594
		error = prepend_name(buffer, buflen, &dentry->d_name);
N
Nick Piggin 已提交
2595
		spin_unlock(&dentry->d_lock);
M
Miklos Szeredi 已提交
2596 2597 2598 2599 2600 2601
		if (!error)
			error = prepend(buffer, buflen, "/", 1);
		if (error)
			break;

		slash = true;
L
Linus Torvalds 已提交
2602 2603 2604
		dentry = parent;
	}

M
Miklos Szeredi 已提交
2605 2606 2607
	if (!error && !slash)
		error = prepend(buffer, buflen, "/", 1);

2608
out:
A
Andi Kleen 已提交
2609
	br_read_unlock(&vfsmount_lock);
M
Miklos Szeredi 已提交
2610
	return error;
L
Linus Torvalds 已提交
2611 2612

global_root:
2613 2614 2615 2616 2617 2618 2619 2620 2621
	/*
	 * Filesystems needing to implement special "root names"
	 * should do so with ->d_dname()
	 */
	if (IS_ROOT(dentry) &&
	    (dentry->d_name.len != 1 || dentry->d_name.name[0] != '/')) {
		WARN(1, "Root dentry has weird name <%.*s>\n",
		     (int) dentry->d_name.len, dentry->d_name.name);
	}
2622 2623 2624
	if (!slash)
		error = prepend(buffer, buflen, "/", 1);
	if (!error)
A
Al Viro 已提交
2625
		error = is_mounted(vfsmnt) ? 1 : 2;
2626
	goto out;
M
Miklos Szeredi 已提交
2627
}
2628

M
Miklos Szeredi 已提交
2629 2630 2631
/**
 * __d_path - return the path of a dentry
 * @path: the dentry/vfsmount to report
2632
 * @root: root vfsmnt/dentry
2633
 * @buf: buffer to return value in
M
Miklos Szeredi 已提交
2634 2635
 * @buflen: buffer length
 *
2636
 * Convert a dentry into an ASCII path name.
M
Miklos Szeredi 已提交
2637 2638 2639 2640
 *
 * Returns a pointer into the buffer or an error code if the
 * path was too long.
 *
2641
 * "buflen" should be positive.
M
Miklos Szeredi 已提交
2642
 *
2643
 * If the path is not reachable from the supplied root, return %NULL.
M
Miklos Szeredi 已提交
2644
 */
2645 2646
char *__d_path(const struct path *path,
	       const struct path *root,
M
Miklos Szeredi 已提交
2647 2648 2649 2650 2651 2652
	       char *buf, int buflen)
{
	char *res = buf + buflen;
	int error;

	prepend(&res, &buflen, "\0", 1);
2653
	write_seqlock(&rename_lock);
M
Miklos Szeredi 已提交
2654
	error = prepend_path(path, root, &res, &buflen);
2655
	write_sequnlock(&rename_lock);
2656

2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678
	if (error < 0)
		return ERR_PTR(error);
	if (error > 0)
		return NULL;
	return res;
}

char *d_absolute_path(const struct path *path,
	       char *buf, int buflen)
{
	struct path root = {};
	char *res = buf + buflen;
	int error;

	prepend(&res, &buflen, "\0", 1);
	write_seqlock(&rename_lock);
	error = prepend_path(path, &root, &res, &buflen);
	write_sequnlock(&rename_lock);

	if (error > 1)
		error = -EINVAL;
	if (error < 0)
M
Miklos Szeredi 已提交
2679 2680
		return ERR_PTR(error);
	return res;
L
Linus Torvalds 已提交
2681 2682
}

2683 2684 2685
/*
 * same as __d_path but appends "(deleted)" for unlinked files.
 */
2686 2687 2688
static int path_with_deleted(const struct path *path,
			     const struct path *root,
			     char **buf, int *buflen)
2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699
{
	prepend(buf, buflen, "\0", 1);
	if (d_unlinked(path->dentry)) {
		int error = prepend(buf, buflen, " (deleted)", 10);
		if (error)
			return error;
	}

	return prepend_path(path, root, buf, buflen);
}

2700 2701 2702 2703 2704
static int prepend_unreachable(char **buffer, int *buflen)
{
	return prepend(buffer, buflen, "(unreachable)", 13);
}

J
Jan Blunck 已提交
2705 2706
/**
 * d_path - return the path of a dentry
2707
 * @path: path to report
J
Jan Blunck 已提交
2708 2709 2710 2711 2712 2713
 * @buf: buffer to return value in
 * @buflen: buffer length
 *
 * Convert a dentry into an ASCII path name. If the entry has been deleted
 * the string " (deleted)" is appended. Note that this is ambiguous.
 *
2714 2715 2716 2717
 * Returns a pointer into the buffer or an error code if the path was
 * too long. Note: Callers should use the returned pointer, not the passed
 * in buffer, to use the name! The implementation often starts at an offset
 * into the buffer, and may leave 0 bytes at the start.
J
Jan Blunck 已提交
2718
 *
2719
 * "buflen" should be positive.
J
Jan Blunck 已提交
2720
 */
2721
char *d_path(const struct path *path, char *buf, int buflen)
L
Linus Torvalds 已提交
2722
{
2723
	char *res = buf + buflen;
J
Jan Blunck 已提交
2724
	struct path root;
2725
	int error;
L
Linus Torvalds 已提交
2726

2727 2728 2729 2730 2731 2732 2733
	/*
	 * We have various synthetic filesystems that never get mounted.  On
	 * these filesystems dentries are never used for lookup purposes, and
	 * thus don't need to be hashed.  They also don't need a name until a
	 * user wants to identify the object in /proc/pid/fd/.  The little hack
	 * below allows us to generate a name for these objects on demand:
	 */
2734 2735
	if (path->dentry->d_op && path->dentry->d_op->d_dname)
		return path->dentry->d_op->d_dname(path->dentry, buf, buflen);
2736

2737
	get_fs_root(current->fs, &root);
2738
	write_seqlock(&rename_lock);
2739 2740
	error = path_with_deleted(path, &root, &res, &buflen);
	if (error < 0)
2741
		res = ERR_PTR(error);
2742
	write_sequnlock(&rename_lock);
J
Jan Blunck 已提交
2743
	path_put(&root);
L
Linus Torvalds 已提交
2744 2745
	return res;
}
2746
EXPORT_SYMBOL(d_path);
L
Linus Torvalds 已提交
2747

2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766
/**
 * d_path_with_unreachable - return the path of a dentry
 * @path: path to report
 * @buf: buffer to return value in
 * @buflen: buffer length
 *
 * The difference from d_path() is that this prepends "(unreachable)"
 * to paths which are unreachable from the current process' root.
 */
char *d_path_with_unreachable(const struct path *path, char *buf, int buflen)
{
	char *res = buf + buflen;
	struct path root;
	int error;

	if (path->dentry->d_op && path->dentry->d_op->d_dname)
		return path->dentry->d_op->d_dname(path->dentry, buf, buflen);

	get_fs_root(current->fs, &root);
2767
	write_seqlock(&rename_lock);
2768 2769
	error = path_with_deleted(path, &root, &res, &buflen);
	if (error > 0)
2770
		error = prepend_unreachable(&res, &buflen);
2771
	write_sequnlock(&rename_lock);
2772 2773 2774 2775 2776 2777 2778
	path_put(&root);
	if (error)
		res =  ERR_PTR(error);

	return res;
}

2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799
/*
 * Helper function for dentry_operations.d_dname() members
 */
char *dynamic_dname(struct dentry *dentry, char *buffer, int buflen,
			const char *fmt, ...)
{
	va_list args;
	char temp[64];
	int sz;

	va_start(args, fmt);
	sz = vsnprintf(temp, sizeof(temp), fmt, args) + 1;
	va_end(args);

	if (sz > sizeof(temp) || sz > buflen)
		return ERR_PTR(-ENAMETOOLONG);

	buffer += buflen - sz;
	return memcpy(buffer, temp, sz);
}

2800 2801 2802
/*
 * Write full pathname from the root of the filesystem into the buffer.
 */
N
Nick Piggin 已提交
2803
static char *__dentry_path(struct dentry *dentry, char *buf, int buflen)
2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814
{
	char *end = buf + buflen;
	char *retval;

	prepend(&end, &buflen, "\0", 1);
	if (buflen < 1)
		goto Elong;
	/* Get '/' right */
	retval = end-1;
	*retval = '/';

2815 2816
	while (!IS_ROOT(dentry)) {
		struct dentry *parent = dentry->d_parent;
N
Nick Piggin 已提交
2817
		int error;
2818 2819

		prefetch(parent);
N
Nick Piggin 已提交
2820 2821 2822 2823
		spin_lock(&dentry->d_lock);
		error = prepend_name(&end, &buflen, &dentry->d_name);
		spin_unlock(&dentry->d_lock);
		if (error != 0 || prepend(&end, &buflen, "/", 1) != 0)
2824 2825 2826 2827 2828
			goto Elong;

		retval = end;
		dentry = parent;
	}
A
Al Viro 已提交
2829 2830 2831 2832
	return retval;
Elong:
	return ERR_PTR(-ENAMETOOLONG);
}
N
Nick Piggin 已提交
2833 2834 2835 2836 2837

char *dentry_path_raw(struct dentry *dentry, char *buf, int buflen)
{
	char *retval;

2838
	write_seqlock(&rename_lock);
N
Nick Piggin 已提交
2839
	retval = __dentry_path(dentry, buf, buflen);
2840
	write_sequnlock(&rename_lock);
N
Nick Piggin 已提交
2841 2842 2843 2844

	return retval;
}
EXPORT_SYMBOL(dentry_path_raw);
A
Al Viro 已提交
2845 2846 2847 2848 2849 2850

char *dentry_path(struct dentry *dentry, char *buf, int buflen)
{
	char *p = NULL;
	char *retval;

2851
	write_seqlock(&rename_lock);
A
Al Viro 已提交
2852 2853 2854 2855 2856 2857 2858
	if (d_unlinked(dentry)) {
		p = buf + buflen;
		if (prepend(&p, &buflen, "//deleted", 10) != 0)
			goto Elong;
		buflen++;
	}
	retval = __dentry_path(dentry, buf, buflen);
2859
	write_sequnlock(&rename_lock);
A
Al Viro 已提交
2860 2861
	if (!IS_ERR(retval) && p)
		*p = '/';	/* restore '/' overriden with '\0' */
2862 2863 2864 2865 2866
	return retval;
Elong:
	return ERR_PTR(-ENAMETOOLONG);
}

L
Linus Torvalds 已提交
2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884
/*
 * NOTE! The user-level library version returns a
 * character pointer. The kernel system call just
 * returns the length of the buffer filled (which
 * includes the ending '\0' character), or a negative
 * error value. So libc would do something like
 *
 *	char *getcwd(char * buf, size_t size)
 *	{
 *		int retval;
 *
 *		retval = sys_getcwd(buf, size);
 *		if (retval >= 0)
 *			return buf;
 *		errno = -retval;
 *		return NULL;
 *	}
 */
2885
SYSCALL_DEFINE2(getcwd, char __user *, buf, unsigned long, size)
L
Linus Torvalds 已提交
2886
{
2887
	int error;
J
Jan Blunck 已提交
2888
	struct path pwd, root;
2889
	char *page = (char *) __get_free_page(GFP_USER);
L
Linus Torvalds 已提交
2890 2891 2892 2893

	if (!page)
		return -ENOMEM;

2894
	get_fs_root_and_pwd(current->fs, &root, &pwd);
L
Linus Torvalds 已提交
2895

2896
	error = -ENOENT;
2897
	write_seqlock(&rename_lock);
2898
	if (!d_unlinked(pwd.dentry)) {
2899
		unsigned long len;
2900 2901
		char *cwd = page + PAGE_SIZE;
		int buflen = PAGE_SIZE;
L
Linus Torvalds 已提交
2902

2903
		prepend(&cwd, &buflen, "\0", 1);
2904
		error = prepend_path(&pwd, &root, &cwd, &buflen);
2905
		write_sequnlock(&rename_lock);
2906

2907
		if (error < 0)
2908 2909
			goto out;

2910
		/* Unreachable from current root */
2911
		if (error > 0) {
2912 2913 2914 2915 2916
			error = prepend_unreachable(&cwd, &buflen);
			if (error)
				goto out;
		}

2917 2918 2919 2920 2921 2922 2923
		error = -ERANGE;
		len = PAGE_SIZE + page - cwd;
		if (len <= size) {
			error = len;
			if (copy_to_user(buf, cwd, len))
				error = -EFAULT;
		}
2924 2925 2926
	} else {
		write_sequnlock(&rename_lock);
	}
L
Linus Torvalds 已提交
2927 2928

out:
J
Jan Blunck 已提交
2929 2930
	path_put(&pwd);
	path_put(&root);
L
Linus Torvalds 已提交
2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950
	free_page((unsigned long) page);
	return error;
}

/*
 * Test whether new_dentry is a subdirectory of old_dentry.
 *
 * Trivially implemented using the dcache structure
 */

/**
 * is_subdir - is new dentry a subdirectory of old_dentry
 * @new_dentry: new dentry
 * @old_dentry: old dentry
 *
 * Returns 1 if new_dentry is a subdirectory of the parent (at any depth).
 * Returns 0 otherwise.
 * Caller must ensure that "new_dentry" is pinned before calling is_subdir()
 */
  
2951
int is_subdir(struct dentry *new_dentry, struct dentry *old_dentry)
L
Linus Torvalds 已提交
2952 2953
{
	int result;
2954
	unsigned seq;
L
Linus Torvalds 已提交
2955

2956 2957 2958 2959
	if (new_dentry == old_dentry)
		return 1;

	do {
L
Linus Torvalds 已提交
2960 2961
		/* for restarting inner loop in case of seq retry */
		seq = read_seqbegin(&rename_lock);
2962 2963 2964 2965 2966
		/*
		 * Need rcu_readlock to protect against the d_parent trashing
		 * due to d_move
		 */
		rcu_read_lock();
2967
		if (d_ancestor(old_dentry, new_dentry))
L
Linus Torvalds 已提交
2968
			result = 1;
2969 2970
		else
			result = 0;
2971
		rcu_read_unlock();
L
Linus Torvalds 已提交
2972 2973 2974 2975 2976 2977 2978
	} while (read_seqretry(&rename_lock, seq));

	return result;
}

void d_genocide(struct dentry *root)
{
2979
	struct dentry *this_parent;
L
Linus Torvalds 已提交
2980
	struct list_head *next;
2981
	unsigned seq;
2982
	int locked = 0;
L
Linus Torvalds 已提交
2983

2984
	seq = read_seqbegin(&rename_lock);
2985 2986
again:
	this_parent = root;
N
Nick Piggin 已提交
2987
	spin_lock(&this_parent->d_lock);
L
Linus Torvalds 已提交
2988 2989 2990 2991 2992
repeat:
	next = this_parent->d_subdirs.next;
resume:
	while (next != &this_parent->d_subdirs) {
		struct list_head *tmp = next;
E
Eric Dumazet 已提交
2993
		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
L
Linus Torvalds 已提交
2994
		next = tmp->next;
2995

N
Nick Piggin 已提交
2996 2997 2998
		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
		if (d_unhashed(dentry) || !dentry->d_inode) {
			spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
2999
			continue;
N
Nick Piggin 已提交
3000
		}
L
Linus Torvalds 已提交
3001
		if (!list_empty(&dentry->d_subdirs)) {
N
Nick Piggin 已提交
3002 3003
			spin_unlock(&this_parent->d_lock);
			spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
L
Linus Torvalds 已提交
3004
			this_parent = dentry;
N
Nick Piggin 已提交
3005
			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
L
Linus Torvalds 已提交
3006 3007
			goto repeat;
		}
3008 3009 3010 3011
		if (!(dentry->d_flags & DCACHE_GENOCIDE)) {
			dentry->d_flags |= DCACHE_GENOCIDE;
			dentry->d_count--;
		}
N
Nick Piggin 已提交
3012
		spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
3013 3014
	}
	if (this_parent != root) {
3015
		struct dentry *child = this_parent;
3016 3017 3018 3019
		if (!(this_parent->d_flags & DCACHE_GENOCIDE)) {
			this_parent->d_flags |= DCACHE_GENOCIDE;
			this_parent->d_count--;
		}
3020 3021
		this_parent = try_to_ascend(this_parent, locked, seq);
		if (!this_parent)
3022 3023
			goto rename_retry;
		next = child->d_u.d_child.next;
L
Linus Torvalds 已提交
3024 3025
		goto resume;
	}
N
Nick Piggin 已提交
3026
	spin_unlock(&this_parent->d_lock);
3027
	if (!locked && read_seqretry(&rename_lock, seq))
3028
		goto rename_retry;
3029 3030 3031 3032 3033 3034 3035 3036
	if (locked)
		write_sequnlock(&rename_lock);
	return;

rename_retry:
	locked = 1;
	write_seqlock(&rename_lock);
	goto again;
L
Linus Torvalds 已提交
3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057
}

/**
 * find_inode_number - check for dentry with name
 * @dir: directory to check
 * @name: Name to find.
 *
 * Check whether a dentry already exists for the given name,
 * and return the inode number if it has an inode. Otherwise
 * 0 is returned.
 *
 * This routine is used to post-process directory listings for
 * filesystems using synthetic inode numbers, and is necessary
 * to keep getcwd() working.
 */
 
ino_t find_inode_number(struct dentry *dir, struct qstr *name)
{
	struct dentry * dentry;
	ino_t ino = 0;

3058 3059
	dentry = d_hash_and_lookup(dir, name);
	if (dentry) {
L
Linus Torvalds 已提交
3060 3061 3062 3063 3064 3065
		if (dentry->d_inode)
			ino = dentry->d_inode->i_ino;
		dput(dentry);
	}
	return ino;
}
3066
EXPORT_SYMBOL(find_inode_number);
L
Linus Torvalds 已提交
3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079

static __initdata unsigned long dhash_entries;
static int __init set_dhash_entries(char *str)
{
	if (!str)
		return 0;
	dhash_entries = simple_strtoul(str, &str, 0);
	return 1;
}
__setup("dhash_entries=", set_dhash_entries);

static void __init dcache_init_early(void)
{
3080
	unsigned int loop;
L
Linus Torvalds 已提交
3081 3082 3083 3084 3085 3086 3087 3088 3089

	/* If hashes are distributed across NUMA nodes, defer
	 * hash allocation until vmalloc space is available.
	 */
	if (hashdist)
		return;

	dentry_hashtable =
		alloc_large_system_hash("Dentry cache",
3090
					sizeof(struct hlist_bl_head),
L
Linus Torvalds 已提交
3091 3092 3093 3094 3095
					dhash_entries,
					13,
					HASH_EARLY,
					&d_hash_shift,
					&d_hash_mask,
3096
					0,
L
Linus Torvalds 已提交
3097 3098
					0);

3099
	for (loop = 0; loop < (1U << d_hash_shift); loop++)
3100
		INIT_HLIST_BL_HEAD(dentry_hashtable + loop);
L
Linus Torvalds 已提交
3101 3102
}

3103
static void __init dcache_init(void)
L
Linus Torvalds 已提交
3104
{
3105
	unsigned int loop;
L
Linus Torvalds 已提交
3106 3107 3108 3109 3110 3111

	/* 
	 * A constructor could be added for stable state like the lists,
	 * but it is probably not worth it because of the cache nature
	 * of the dcache. 
	 */
3112 3113
	dentry_cache = KMEM_CACHE(dentry,
		SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|SLAB_MEM_SPREAD);
L
Linus Torvalds 已提交
3114 3115 3116 3117 3118 3119 3120

	/* Hash may have been set up in dcache_init_early */
	if (!hashdist)
		return;

	dentry_hashtable =
		alloc_large_system_hash("Dentry cache",
3121
					sizeof(struct hlist_bl_head),
L
Linus Torvalds 已提交
3122 3123 3124 3125 3126
					dhash_entries,
					13,
					0,
					&d_hash_shift,
					&d_hash_mask,
3127
					0,
L
Linus Torvalds 已提交
3128 3129
					0);

3130
	for (loop = 0; loop < (1U << d_hash_shift); loop++)
3131
		INIT_HLIST_BL_HEAD(dentry_hashtable + loop);
L
Linus Torvalds 已提交
3132 3133 3134
}

/* SLAB cache for __getname() consumers */
3135
struct kmem_cache *names_cachep __read_mostly;
3136
EXPORT_SYMBOL(names_cachep);
L
Linus Torvalds 已提交
3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156

EXPORT_SYMBOL(d_genocide);

void __init vfs_caches_init_early(void)
{
	dcache_init_early();
	inode_init_early();
}

void __init vfs_caches_init(unsigned long mempages)
{
	unsigned long reserve;

	/* Base hash sizes on available memory, with a reserve equal to
           150% of current kernel size */

	reserve = min((mempages - nr_free_pages()) * 3/2, mempages - 1);
	mempages -= reserve;

	names_cachep = kmem_cache_create("names_cache", PATH_MAX, 0,
3157
			SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
L
Linus Torvalds 已提交
3158

3159 3160
	dcache_init();
	inode_init();
L
Linus Torvalds 已提交
3161
	files_init(mempages);
3162
	mnt_init();
L
Linus Torvalds 已提交
3163 3164 3165
	bdev_cache_init();
	chrdev_init();
}