dcache.c 82.9 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
/**
 * read_seqbegin_or_lock - begin a sequence number check or locking block
93 94
 * @lock: sequence lock
 * @seq : sequence number to be checked
95 96 97 98 99 100 101 102
 *
 * First try it once optimistically without taking the lock. If that fails,
 * take the lock. The sequence number is also used as a marker for deciding
 * whether to be a reader (even) or writer (odd).
 * N.B. seq must be initialized to an even number to begin with.
 */
static inline void read_seqbegin_or_lock(seqlock_t *lock, int *seq)
{
103
	if (!(*seq & 1))	/* Even */
104
		*seq = read_seqbegin(lock);
105
	else			/* Odd */
106
		read_seqlock_excl(lock);
107 108
}

109
static inline int need_seqretry(seqlock_t *lock, int seq)
110
{
111 112 113 114 115 116
	return !(seq & 1) && read_seqretry(lock, seq);
}

static inline void done_seqretry(seqlock_t *lock, int seq)
{
	if (seq & 1)
117
		read_sequnlock_excl(lock);
118 119
}

L
Linus Torvalds 已提交
120 121 122 123 124 125 126 127 128 129 130
/*
 * 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

131 132
static unsigned int d_hash_mask __read_mostly;
static unsigned int d_hash_shift __read_mostly;
133

134
static struct hlist_bl_head *dentry_hashtable __read_mostly;
135

136
static inline struct hlist_bl_head *d_hash(const struct dentry *parent,
137
					unsigned int hash)
138
{
139 140
	hash += (unsigned long) parent / L1_CACHE_BYTES;
	hash = hash + (hash >> D_HASHBITS);
141 142 143
	return dentry_hashtable + (hash & D_HASHMASK);
}

L
Linus Torvalds 已提交
144 145 146 147 148
/* Statistics gathering. */
struct dentry_stat_t dentry_stat = {
	.age_limit = 45,
};

149
static DEFINE_PER_CPU(unsigned int, nr_dentry);
150 151

#if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
152 153 154 155 156 157 158 159 160
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;
}

161 162 163
int proc_nr_dentry(ctl_table *table, int write, void __user *buffer,
		   size_t *lenp, loff_t *ppos)
{
164
	dentry_stat.nr_dentry = get_nr_dentry();
165 166 167 168
	return proc_dointvec(table, write, buffer, lenp, ppos);
}
#endif

169 170 171 172
/*
 * Compare 2 name strings, return 0 if they match, otherwise non-zero.
 * The strings are both count bytes long, and count is non-zero.
 */
173 174 175 176 177 178 179 180 181 182 183 184
#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.
 */
185
static inline int dentry_string_cmp(const unsigned char *cs, const unsigned char *ct, unsigned tcount)
186
{
187 188 189
	unsigned long a,b,mask;

	for (;;) {
190
		a = *(unsigned long *)cs;
191
		b = load_unaligned_zeropad(ct);
192 193 194 195 196 197 198 199 200 201 202 203
		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));
204 205
}

206
#else
207

208
static inline int dentry_string_cmp(const unsigned char *cs, const unsigned char *ct, unsigned tcount)
209
{
210 211 212 213 214 215 216 217 218 219
	do {
		if (*cs != *ct)
			return 1;
		cs++;
		ct++;
		tcount--;
	} while (tcount);
	return 0;
}

220 221
#endif

222 223
static inline int dentry_cmp(const struct dentry *dentry, const unsigned char *ct, unsigned tcount)
{
224
	const unsigned char *cs;
225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240
	/*
	 * 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)
	 */
241 242 243
	cs = ACCESS_ONCE(dentry->d_name.name);
	smp_read_barrier_depends();
	return dentry_string_cmp(cs, ct, tcount);
244 245
}

C
Christoph Hellwig 已提交
246
static void __d_free(struct rcu_head *head)
L
Linus Torvalds 已提交
247
{
C
Christoph Hellwig 已提交
248 249
	struct dentry *dentry = container_of(head, struct dentry, d_u.d_rcu);

A
Al Viro 已提交
250
	WARN_ON(!hlist_unhashed(&dentry->d_alias));
L
Linus Torvalds 已提交
251 252 253 254 255 256
	if (dname_external(dentry))
		kfree(dentry->d_name.name);
	kmem_cache_free(dentry_cache, dentry); 
}

/*
N
Nick Piggin 已提交
257
 * no locks, please.
L
Linus Torvalds 已提交
258 259 260
 */
static void d_free(struct dentry *dentry)
{
261
	BUG_ON((int)dentry->d_lockref.count > 0);
262
	this_cpu_dec(nr_dentry);
L
Linus Torvalds 已提交
263 264
	if (dentry->d_op && dentry->d_op->d_release)
		dentry->d_op->d_release(dentry);
265

266 267
	/* if dentry was never visible to RCU, immediate free is OK */
	if (!(dentry->d_flags & DCACHE_RCUACCESS))
C
Christoph Hellwig 已提交
268
		__d_free(&dentry->d_u.d_rcu);
269
	else
C
Christoph Hellwig 已提交
270
		call_rcu(&dentry->d_u.d_rcu, __d_free);
L
Linus Torvalds 已提交
271 272
}

N
Nick Piggin 已提交
273 274
/**
 * dentry_rcuwalk_barrier - invalidate in-progress rcu-walk lookups
275
 * @dentry: the target dentry
N
Nick Piggin 已提交
276 277 278 279 280 281 282 283 284 285 286
 * 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
287 288
/*
 * Release the dentry's inode, using the filesystem
N
Nick Piggin 已提交
289 290
 * d_iput() operation if defined. Dentry has no refcount
 * and is unhashed.
L
Linus Torvalds 已提交
291
 */
292
static void dentry_iput(struct dentry * dentry)
293
	__releases(dentry->d_lock)
294
	__releases(dentry->d_inode->i_lock)
L
Linus Torvalds 已提交
295 296 297 298
{
	struct inode *inode = dentry->d_inode;
	if (inode) {
		dentry->d_inode = NULL;
A
Al Viro 已提交
299
		hlist_del_init(&dentry->d_alias);
L
Linus Torvalds 已提交
300
		spin_unlock(&dentry->d_lock);
301
		spin_unlock(&inode->i_lock);
302 303
		if (!inode->i_nlink)
			fsnotify_inoderemove(inode);
L
Linus Torvalds 已提交
304 305 306 307 308 309 310 311 312
		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
313 314 315 316 317 318
/*
 * 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)
319
	__releases(dentry->d_inode->i_lock)
N
Nick Piggin 已提交
320 321 322
{
	struct inode *inode = dentry->d_inode;
	dentry->d_inode = NULL;
A
Al Viro 已提交
323
	hlist_del_init(&dentry->d_alias);
N
Nick Piggin 已提交
324 325
	dentry_rcuwalk_barrier(dentry);
	spin_unlock(&dentry->d_lock);
326
	spin_unlock(&inode->i_lock);
N
Nick Piggin 已提交
327 328 329 330 331 332 333 334
	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);
}

335
/*
S
Sage Weil 已提交
336
 * dentry_lru_(add|del|prune|move_tail) must be called with d_lock held.
337 338 339
 */
static void dentry_lru_add(struct dentry *dentry)
{
340
	if (unlikely(!(dentry->d_flags & DCACHE_LRU_LIST))) {
N
Nick Piggin 已提交
341
		spin_lock(&dcache_lru_lock);
342
		dentry->d_flags |= DCACHE_LRU_LIST;
343 344
		list_add(&dentry->d_lru, &dentry->d_sb->s_dentry_lru);
		dentry->d_sb->s_nr_dentry_unused++;
345
		dentry_stat.nr_unused++;
N
Nick Piggin 已提交
346
		spin_unlock(&dcache_lru_lock);
347
	}
348 349
}

N
Nick Piggin 已提交
350 351 352
static void __dentry_lru_del(struct dentry *dentry)
{
	list_del_init(&dentry->d_lru);
353
	dentry->d_flags &= ~(DCACHE_SHRINK_LIST | DCACHE_LRU_LIST);
N
Nick Piggin 已提交
354 355 356 357
	dentry->d_sb->s_nr_dentry_unused--;
	dentry_stat.nr_unused--;
}

S
Sage Weil 已提交
358 359 360
/*
 * Remove a dentry with references from the LRU.
 */
361 362 363
static void dentry_lru_del(struct dentry *dentry)
{
	if (!list_empty(&dentry->d_lru)) {
N
Nick Piggin 已提交
364 365 366
		spin_lock(&dcache_lru_lock);
		__dentry_lru_del(dentry);
		spin_unlock(&dcache_lru_lock);
367 368 369
	}
}

370
static void dentry_lru_move_list(struct dentry *dentry, struct list_head *list)
371
{
N
Nick Piggin 已提交
372
	spin_lock(&dcache_lru_lock);
373
	if (list_empty(&dentry->d_lru)) {
374
		dentry->d_flags |= DCACHE_LRU_LIST;
375
		list_add_tail(&dentry->d_lru, list);
376
		dentry->d_sb->s_nr_dentry_unused++;
377
		dentry_stat.nr_unused++;
378
	} else {
379
		list_move_tail(&dentry->d_lru, list);
380
	}
N
Nick Piggin 已提交
381
	spin_unlock(&dcache_lru_lock);
382 383
}

384 385 386
/**
 * d_kill - kill dentry and return parent
 * @dentry: dentry to kill
387
 * @parent: parent dentry
388
 *
389
 * The dentry must already be unhashed and removed from the LRU.
390 391
 *
 * If this is the root of the dentry tree, return NULL.
N
Nick Piggin 已提交
392
 *
N
Nick Piggin 已提交
393 394
 * dentry->d_lock and parent->d_lock must be held by caller, and are dropped by
 * d_kill.
395
 */
N
Nick Piggin 已提交
396
static struct dentry *d_kill(struct dentry *dentry, struct dentry *parent)
397
	__releases(dentry->d_lock)
N
Nick Piggin 已提交
398
	__releases(parent->d_lock)
399
	__releases(dentry->d_inode->i_lock)
400 401
{
	list_del(&dentry->d_u.d_child);
402 403 404 405
	/*
	 * Inform try_to_ascend() that we are no longer attached to the
	 * dentry tree
	 */
406
	dentry->d_flags |= DCACHE_DENTRY_KILLED;
N
Nick Piggin 已提交
407 408
	if (parent)
		spin_unlock(&parent->d_lock);
409
	dentry_iput(dentry);
N
Nick Piggin 已提交
410 411 412 413
	/*
	 * dentry_iput drops the locks, at which point nobody (except
	 * transient RCU lookups) can reach this dentry.
	 */
414
	d_free(dentry);
415
	return parent;
416 417
}

418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438
/*
 * 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
439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455
/**
 * 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)
{
456
	if (!d_unhashed(dentry)) {
457
		__d_shrink(dentry);
458
		dentry_rcuwalk_barrier(dentry);
N
Nick Piggin 已提交
459 460 461 462 463 464 465 466 467 468 469 470
	}
}
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);

471 472 473 474 475 476
/*
 * 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.
 */
477
static inline struct dentry *dentry_kill(struct dentry *dentry)
478 479
	__releases(dentry->d_lock)
{
480
	struct inode *inode;
481 482
	struct dentry *parent;

483 484
	inode = dentry->d_inode;
	if (inode && !spin_trylock(&inode->i_lock)) {
485 486 487 488 489 490 491 492 493 494
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)) {
495 496
		if (inode)
			spin_unlock(&inode->i_lock);
497 498
		goto relock;
	}
N
Nick Piggin 已提交
499

500 501 502 503 504
	/*
	 * The dentry is now unrecoverably dead to the world.
	 */
	lockref_mark_dead(&dentry->d_lockref);

S
Sage Weil 已提交
505 506 507 508
	/*
	 * inform the fs via d_prune that this dentry is about to be
	 * unhashed and destroyed.
	 */
509
	if ((dentry->d_flags & DCACHE_OP_PRUNE) && !d_unhashed(dentry))
Y
Yan, Zheng 已提交
510 511 512
		dentry->d_op->d_prune(dentry);

	dentry_lru_del(dentry);
513 514 515 516 517
	/* if it was on the hash then remove it */
	__d_drop(dentry);
	return d_kill(dentry, parent);
}

L
Linus Torvalds 已提交
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
/* 
 * 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)
{
546
	if (unlikely(!dentry))
L
Linus Torvalds 已提交
547 548 549
		return;

repeat:
550
	if (lockref_put_or_lock(&dentry->d_lockref))
L
Linus Torvalds 已提交
551 552
		return;

553 554 555 556 557
	/* Unreachable? Get rid of it */
	if (unlikely(d_unhashed(dentry)))
		goto kill_it;

	if (unlikely(dentry->d_flags & DCACHE_OP_DELETE)) {
L
Linus Torvalds 已提交
558
		if (dentry->d_op->d_delete(dentry))
N
Nick Piggin 已提交
559
			goto kill_it;
L
Linus Torvalds 已提交
560
	}
N
Nick Piggin 已提交
561

J
Jeff Layton 已提交
562
	dentry->d_flags |= DCACHE_REFERENCED;
563
	dentry_lru_add(dentry);
N
Nick Piggin 已提交
564

565
	dentry->d_lockref.count--;
N
Nick Piggin 已提交
566
	spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
567 568
	return;

569
kill_it:
570
	dentry = dentry_kill(dentry);
571 572
	if (dentry)
		goto repeat;
L
Linus Torvalds 已提交
573
}
574
EXPORT_SYMBOL(dput);
L
Linus Torvalds 已提交
575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592

/**
 * 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
593
	spin_lock(&dentry->d_lock);
L
Linus Torvalds 已提交
594
	if (d_unhashed(dentry)) {
N
Nick Piggin 已提交
595
		spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
596 597 598 599 600 601 602
		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
603
		spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
604
		shrink_dcache_parent(dentry);
N
Nick Piggin 已提交
605
		spin_lock(&dentry->d_lock);
L
Linus Torvalds 已提交
606 607 608 609 610 611 612 613 614 615 616
	}

	/*
	 * 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).
617 618
	 * We also need to leave mountpoints alone,
	 * directory or not.
L
Linus Torvalds 已提交
619
	 */
620
	if (dentry->d_lockref.count > 1 && dentry->d_inode) {
621
		if (S_ISDIR(dentry->d_inode->i_mode) || d_mountpoint(dentry)) {
L
Linus Torvalds 已提交
622 623 624 625 626 627 628 629 630
			spin_unlock(&dentry->d_lock);
			return -EBUSY;
		}
	}

	__d_drop(dentry);
	spin_unlock(&dentry->d_lock);
	return 0;
}
631
EXPORT_SYMBOL(d_invalidate);
L
Linus Torvalds 已提交
632

N
Nick Piggin 已提交
633
/* This must be called with d_lock held */
634
static inline void __dget_dlock(struct dentry *dentry)
N
Nick Piggin 已提交
635
{
636
	dentry->d_lockref.count++;
N
Nick Piggin 已提交
637 638
}

639
static inline void __dget(struct dentry *dentry)
L
Linus Torvalds 已提交
640
{
641
	lockref_get(&dentry->d_lockref);
L
Linus Torvalds 已提交
642 643
}

N
Nick Piggin 已提交
644 645
struct dentry *dget_parent(struct dentry *dentry)
{
646
	int gotref;
N
Nick Piggin 已提交
647 648
	struct dentry *ret;

649 650 651 652 653 654 655 656 657 658 659 660 661 662
	/*
	 * Do optimistic parent lookup without any
	 * locking.
	 */
	rcu_read_lock();
	ret = ACCESS_ONCE(dentry->d_parent);
	gotref = lockref_get_not_zero(&ret->d_lockref);
	rcu_read_unlock();
	if (likely(gotref)) {
		if (likely(ret == ACCESS_ONCE(dentry->d_parent)))
			return ret;
		dput(ret);
	}

N
Nick Piggin 已提交
663
repeat:
N
Nick Piggin 已提交
664 665 666 667 668
	/*
	 * Don't need rcu_dereference because we re-check it was correct under
	 * the lock.
	 */
	rcu_read_lock();
N
Nick Piggin 已提交
669
	ret = dentry->d_parent;
N
Nick Piggin 已提交
670 671 672 673
	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
674 675
		goto repeat;
	}
N
Nick Piggin 已提交
676
	rcu_read_unlock();
677 678
	BUG_ON(!ret->d_lockref.count);
	ret->d_lockref.count++;
N
Nick Piggin 已提交
679 680 681 682 683
	spin_unlock(&ret->d_lock);
	return ret;
}
EXPORT_SYMBOL(dget_parent);

L
Linus Torvalds 已提交
684 685 686
/**
 * d_find_alias - grab a hashed alias of inode
 * @inode: inode in question
687 688
 * @want_discon:  flag, used by d_splice_alias, to request
 *          that only a DISCONNECTED alias be returned.
L
Linus Torvalds 已提交
689 690 691 692 693 694 695
 *
 * 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.
 *
696
 * If the inode has an IS_ROOT, DCACHE_DISCONNECTED alias, then prefer
697 698
 * 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
699
 */
700
static struct dentry *__d_find_alias(struct inode *inode, int want_discon)
L
Linus Torvalds 已提交
701
{
N
Nick Piggin 已提交
702
	struct dentry *alias, *discon_alias;
L
Linus Torvalds 已提交
703

N
Nick Piggin 已提交
704 705
again:
	discon_alias = NULL;
706
	hlist_for_each_entry(alias, &inode->i_dentry, d_alias) {
N
Nick Piggin 已提交
707
		spin_lock(&alias->d_lock);
L
Linus Torvalds 已提交
708
 		if (S_ISDIR(inode->i_mode) || !d_unhashed(alias)) {
709
			if (IS_ROOT(alias) &&
N
Nick Piggin 已提交
710
			    (alias->d_flags & DCACHE_DISCONNECTED)) {
L
Linus Torvalds 已提交
711
				discon_alias = alias;
712
			} else if (!want_discon) {
713
				__dget_dlock(alias);
N
Nick Piggin 已提交
714 715 716 717 718 719 720 721 722 723 724 725
				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)) {
726
				__dget_dlock(alias);
N
Nick Piggin 已提交
727
				spin_unlock(&alias->d_lock);
L
Linus Torvalds 已提交
728 729 730
				return alias;
			}
		}
N
Nick Piggin 已提交
731 732
		spin_unlock(&alias->d_lock);
		goto again;
L
Linus Torvalds 已提交
733
	}
N
Nick Piggin 已提交
734
	return NULL;
L
Linus Torvalds 已提交
735 736
}

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

A
Al Viro 已提交
741
	if (!hlist_empty(&inode->i_dentry)) {
742
		spin_lock(&inode->i_lock);
743
		de = __d_find_alias(inode, 0);
744
		spin_unlock(&inode->i_lock);
D
David Howells 已提交
745
	}
L
Linus Torvalds 已提交
746 747
	return de;
}
748
EXPORT_SYMBOL(d_find_alias);
L
Linus Torvalds 已提交
749 750 751 752 753 754 755

/*
 *	Try to kill dentries associated with this inode.
 * WARNING: you must own a reference to inode.
 */
void d_prune_aliases(struct inode *inode)
{
756
	struct dentry *dentry;
L
Linus Torvalds 已提交
757
restart:
758
	spin_lock(&inode->i_lock);
759
	hlist_for_each_entry(dentry, &inode->i_dentry, d_alias) {
L
Linus Torvalds 已提交
760
		spin_lock(&dentry->d_lock);
761
		if (!dentry->d_lockref.count) {
762 763 764 765 766 767 768 769
			/*
			 * inform the fs via d_prune that this dentry
			 * is about to be unhashed and destroyed.
			 */
			if ((dentry->d_flags & DCACHE_OP_PRUNE) &&
			    !d_unhashed(dentry))
				dentry->d_op->d_prune(dentry);

770
			__dget_dlock(dentry);
L
Linus Torvalds 已提交
771 772
			__d_drop(dentry);
			spin_unlock(&dentry->d_lock);
773
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
774 775 776 777 778
			dput(dentry);
			goto restart;
		}
		spin_unlock(&dentry->d_lock);
	}
779
	spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
780
}
781
EXPORT_SYMBOL(d_prune_aliases);
L
Linus Torvalds 已提交
782 783

/*
784 785 786
 * 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.
787
 *
788
 * This may fail if locks cannot be acquired no problem, just try again.
L
Linus Torvalds 已提交
789
 */
790
static void try_prune_one_dentry(struct dentry *dentry)
791
	__releases(dentry->d_lock)
L
Linus Torvalds 已提交
792
{
793
	struct dentry *parent;
794

795
	parent = dentry_kill(dentry);
796
	/*
797 798 799 800 801 802 803 804
	 * 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.
805
	 */
806 807 808 809 810 811 812
	if (!parent)
		return;
	if (parent == dentry)
		return;

	/* Prune ancestors. */
	dentry = parent;
813
	while (dentry) {
814
		if (lockref_put_or_lock(&dentry->d_lockref))
815
			return;
816
		dentry = dentry_kill(dentry);
817
	}
L
Linus Torvalds 已提交
818 819
}

C
Christoph Hellwig 已提交
820
static void shrink_dentry_list(struct list_head *list)
L
Linus Torvalds 已提交
821
{
822 823
	struct dentry *dentry;

824 825 826 827 828 829 830 831
	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
832 833 834
			continue;
		}

L
Linus Torvalds 已提交
835 836
		/*
		 * We found an inuse dentry which was not removed from
837 838
		 * the LRU because of laziness during lookup.  Do not free
		 * it - just keep it off the LRU list.
L
Linus Torvalds 已提交
839
		 */
840
		if (dentry->d_lockref.count) {
841
			dentry_lru_del(dentry);
842
			spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
843 844
			continue;
		}
845 846

		rcu_read_unlock();
847 848 849

		try_prune_one_dentry(dentry);

850
		rcu_read_lock();
851
	}
852
	rcu_read_unlock();
C
Christoph Hellwig 已提交
853 854 855
}

/**
856 857 858 859 860 861 862
 * 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
863
 *
864 865
 * This function may fail to free any resources if all the dentries are in
 * use.
C
Christoph Hellwig 已提交
866
 */
867
void prune_dcache_sb(struct super_block *sb, int count)
C
Christoph Hellwig 已提交
868 869 870 871 872
{
	struct dentry *dentry;
	LIST_HEAD(referenced);
	LIST_HEAD(tmp);

N
Nick Piggin 已提交
873 874
relock:
	spin_lock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
875 876 877 878 879
	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
880 881 882 883 884 885
		if (!spin_trylock(&dentry->d_lock)) {
			spin_unlock(&dcache_lru_lock);
			cpu_relax();
			goto relock;
		}

886
		if (dentry->d_flags & DCACHE_REFERENCED) {
N
Nick Piggin 已提交
887 888
			dentry->d_flags &= ~DCACHE_REFERENCED;
			list_move(&dentry->d_lru, &referenced);
C
Christoph Hellwig 已提交
889
			spin_unlock(&dentry->d_lock);
N
Nick Piggin 已提交
890 891
		} else {
			list_move_tail(&dentry->d_lru, &tmp);
892
			dentry->d_flags |= DCACHE_SHRINK_LIST;
N
Nick Piggin 已提交
893
			spin_unlock(&dentry->d_lock);
894
			if (!--count)
N
Nick Piggin 已提交
895
				break;
C
Christoph Hellwig 已提交
896
		}
897
		cond_resched_lock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
898
	}
899 900
	if (!list_empty(&referenced))
		list_splice(&referenced, &sb->s_dentry_lru);
N
Nick Piggin 已提交
901
	spin_unlock(&dcache_lru_lock);
902 903

	shrink_dentry_list(&tmp);
904 905
}

L
Linus Torvalds 已提交
906 907 908 909
/**
 * shrink_dcache_sb - shrink dcache for a superblock
 * @sb: superblock
 *
C
Christoph Hellwig 已提交
910 911
 * 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
912
 */
C
Christoph Hellwig 已提交
913
void shrink_dcache_sb(struct super_block *sb)
L
Linus Torvalds 已提交
914
{
C
Christoph Hellwig 已提交
915 916
	LIST_HEAD(tmp);

N
Nick Piggin 已提交
917
	spin_lock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
918 919
	while (!list_empty(&sb->s_dentry_lru)) {
		list_splice_init(&sb->s_dentry_lru, &tmp);
920
		spin_unlock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
921
		shrink_dentry_list(&tmp);
922
		spin_lock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
923
	}
N
Nick Piggin 已提交
924
	spin_unlock(&dcache_lru_lock);
L
Linus Torvalds 已提交
925
}
926
EXPORT_SYMBOL(shrink_dcache_sb);
L
Linus Torvalds 已提交
927

928 929 930 931 932 933 934 935 936 937 938 939 940
/*
 * 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 */
941
		while (!list_empty(&dentry->d_subdirs))
942 943 944 945 946 947 948 949
			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
950
			/*
Y
Yan, Zheng 已提交
951
			 * inform the fs that this dentry is about to be
S
Sage Weil 已提交
952 953
			 * unhashed and destroyed.
			 */
954 955
			if ((dentry->d_flags & DCACHE_OP_PRUNE) &&
			    !d_unhashed(dentry))
Y
Yan, Zheng 已提交
956 957 958
				dentry->d_op->d_prune(dentry);

			dentry_lru_del(dentry);
959 960
			__d_shrink(dentry);

961
			if (dentry->d_lockref.count != 0) {
962 963 964 965 966 967 968 969
				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,
970
				       dentry->d_lockref.count,
971 972 973 974 975
				       dentry->d_sb->s_type->name,
				       dentry->d_sb->s_id);
				BUG();
			}

N
Nick Piggin 已提交
976
			if (IS_ROOT(dentry)) {
977
				parent = NULL;
N
Nick Piggin 已提交
978 979
				list_del(&dentry->d_u.d_child);
			} else {
980
				parent = dentry->d_parent;
981
				parent->d_lockref.count--;
N
Nick Piggin 已提交
982
				list_del(&dentry->d_u.d_child);
983
			}
984 985 986 987

			inode = dentry->d_inode;
			if (inode) {
				dentry->d_inode = NULL;
A
Al Viro 已提交
988
				hlist_del_init(&dentry->d_alias);
989 990 991 992 993 994 995 996 997 998 999 1000
				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)
1001
				return;
1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
			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
1012
 * - we don't need to use dentry->d_lock because:
1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
 *   - 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;
1029
	dentry->d_lockref.count--;
1030 1031
	shrink_dcache_for_umount_subtree(dentry);

1032 1033
	while (!hlist_bl_empty(&sb->s_anon)) {
		dentry = hlist_bl_entry(hlist_bl_first(&sb->s_anon), struct dentry, d_hash);
1034 1035 1036 1037
		shrink_dcache_for_umount_subtree(dentry);
	}
}

1038 1039 1040 1041 1042 1043
/*
 * 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.
 */
1044
static struct dentry *try_to_ascend(struct dentry *old, unsigned seq)
1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056
{
	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 ||
1057
		 (old->d_flags & DCACHE_DENTRY_KILLED) ||
1058
		 need_seqretry(&rename_lock, seq)) {
1059 1060 1061 1062 1063 1064 1065
		spin_unlock(&new->d_lock);
		new = NULL;
	}
	rcu_read_unlock();
	return new;
}

M
Miklos Szeredi 已提交
1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
/**
 * enum d_walk_ret - action to talke during tree walk
 * @D_WALK_CONTINUE:	contrinue walk
 * @D_WALK_QUIT:	quit walk
 * @D_WALK_NORETRY:	quit when retry is needed
 * @D_WALK_SKIP:	skip this dentry and its children
 */
enum d_walk_ret {
	D_WALK_CONTINUE,
	D_WALK_QUIT,
	D_WALK_NORETRY,
	D_WALK_SKIP,
};
1079

L
Linus Torvalds 已提交
1080
/**
M
Miklos Szeredi 已提交
1081 1082 1083 1084 1085
 * d_walk - walk the dentry tree
 * @parent:	start of walk
 * @data:	data passed to @enter() and @finish()
 * @enter:	callback when first entering the dentry
 * @finish:	callback when successfully finished the walk
L
Linus Torvalds 已提交
1086
 *
M
Miklos Szeredi 已提交
1087
 * The @enter() and @finish() callbacks are called with d_lock held.
L
Linus Torvalds 已提交
1088
 */
M
Miklos Szeredi 已提交
1089 1090 1091
static void d_walk(struct dentry *parent, void *data,
		   enum d_walk_ret (*enter)(void *, struct dentry *),
		   void (*finish)(void *))
L
Linus Torvalds 已提交
1092
{
1093
	struct dentry *this_parent;
L
Linus Torvalds 已提交
1094
	struct list_head *next;
1095
	unsigned seq = 0;
M
Miklos Szeredi 已提交
1096 1097
	enum d_walk_ret ret;
	bool retry = true;
1098

1099
again:
1100
	read_seqbegin_or_lock(&rename_lock, &seq);
1101
	this_parent = parent;
N
Nick Piggin 已提交
1102
	spin_lock(&this_parent->d_lock);
M
Miklos Szeredi 已提交
1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114

	ret = enter(data, this_parent);
	switch (ret) {
	case D_WALK_CONTINUE:
		break;
	case D_WALK_QUIT:
	case D_WALK_SKIP:
		goto out_unlock;
	case D_WALK_NORETRY:
		retry = false;
		break;
	}
L
Linus Torvalds 已提交
1115 1116 1117 1118 1119
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
1120
		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
L
Linus Torvalds 已提交
1121
		next = tmp->next;
N
Nick Piggin 已提交
1122 1123

		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
M
Miklos Szeredi 已提交
1124 1125 1126 1127 1128 1129

		ret = enter(data, dentry);
		switch (ret) {
		case D_WALK_CONTINUE:
			break;
		case D_WALK_QUIT:
N
Nick Piggin 已提交
1130
			spin_unlock(&dentry->d_lock);
M
Miklos Szeredi 已提交
1131 1132 1133 1134 1135 1136 1137
			goto out_unlock;
		case D_WALK_NORETRY:
			retry = false;
			break;
		case D_WALK_SKIP:
			spin_unlock(&dentry->d_lock);
			continue;
N
Nick Piggin 已提交
1138
		}
M
Miklos Szeredi 已提交
1139

L
Linus Torvalds 已提交
1140
		if (!list_empty(&dentry->d_subdirs)) {
N
Nick Piggin 已提交
1141 1142
			spin_unlock(&this_parent->d_lock);
			spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
L
Linus Torvalds 已提交
1143
			this_parent = dentry;
N
Nick Piggin 已提交
1144
			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
L
Linus Torvalds 已提交
1145 1146
			goto repeat;
		}
N
Nick Piggin 已提交
1147
		spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
1148 1149 1150 1151 1152
	}
	/*
	 * All done at this level ... ascend and resume the search.
	 */
	if (this_parent != parent) {
1153
		struct dentry *child = this_parent;
1154
		this_parent = try_to_ascend(this_parent, seq);
1155
		if (!this_parent)
1156 1157
			goto rename_retry;
		next = child->d_u.d_child.next;
L
Linus Torvalds 已提交
1158 1159
		goto resume;
	}
1160
	if (need_seqretry(&rename_lock, seq)) {
M
Miklos Szeredi 已提交
1161
		spin_unlock(&this_parent->d_lock);
1162
		goto rename_retry;
M
Miklos Szeredi 已提交
1163 1164 1165 1166 1167 1168
	}
	if (finish)
		finish(data);

out_unlock:
	spin_unlock(&this_parent->d_lock);
1169
	done_seqretry(&rename_lock, seq);
M
Miklos Szeredi 已提交
1170
	return;
1171 1172

rename_retry:
M
Miklos Szeredi 已提交
1173 1174
	if (!retry)
		return;
1175
	seq = 1;
1176
	goto again;
L
Linus Torvalds 已提交
1177
}
M
Miklos Szeredi 已提交
1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210

/*
 * 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
 */

static enum d_walk_ret check_mount(void *data, struct dentry *dentry)
{
	int *ret = data;
	if (d_mountpoint(dentry)) {
		*ret = 1;
		return D_WALK_QUIT;
	}
	return D_WALK_CONTINUE;
}

int have_submounts(struct dentry *parent)
{
	int ret = 0;

	d_walk(parent, &ret, check_mount, NULL);

	return ret;
}
1211
EXPORT_SYMBOL(have_submounts);
L
Linus Torvalds 已提交
1212

1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245
/*
 * Called by mount code to set a mountpoint and check if the mountpoint is
 * reachable (e.g. NFS can unhash a directory dentry and then the complete
 * subtree can become unreachable).
 *
 * Only one of check_submounts_and_drop() and d_set_mounted() must succeed.  For
 * this reason take rename_lock and d_lock on dentry and ancestors.
 */
int d_set_mounted(struct dentry *dentry)
{
	struct dentry *p;
	int ret = -ENOENT;
	write_seqlock(&rename_lock);
	for (p = dentry->d_parent; !IS_ROOT(p); p = p->d_parent) {
		/* Need exclusion wrt. check_submounts_and_drop() */
		spin_lock(&p->d_lock);
		if (unlikely(d_unhashed(p))) {
			spin_unlock(&p->d_lock);
			goto out;
		}
		spin_unlock(&p->d_lock);
	}
	spin_lock(&dentry->d_lock);
	if (!d_unlinked(dentry)) {
		dentry->d_flags |= DCACHE_MOUNTED;
		ret = 0;
	}
 	spin_unlock(&dentry->d_lock);
out:
	write_sequnlock(&rename_lock);
	return ret;
}

L
Linus Torvalds 已提交
1246
/*
1247
 * Search the dentry child list of the specified parent,
L
Linus Torvalds 已提交
1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260
 * 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.
 */

M
Miklos Szeredi 已提交
1261 1262 1263 1264 1265
struct select_data {
	struct dentry *start;
	struct list_head dispose;
	int found;
};
N
Nick Piggin 已提交
1266

M
Miklos Szeredi 已提交
1267 1268 1269 1270
static enum d_walk_ret select_collect(void *_data, struct dentry *dentry)
{
	struct select_data *data = _data;
	enum d_walk_ret ret = D_WALK_CONTINUE;
L
Linus Torvalds 已提交
1271

M
Miklos Szeredi 已提交
1272 1273
	if (data->start == dentry)
		goto out;
N
Nick Piggin 已提交
1274

L
Linus Torvalds 已提交
1275
	/*
M
Miklos Szeredi 已提交
1276 1277 1278 1279 1280 1281
	 * move only zero ref count dentries to the dispose list.
	 *
	 * 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
1282
	 */
M
Miklos Szeredi 已提交
1283 1284 1285 1286 1287 1288 1289
	if (dentry->d_lockref.count) {
		dentry_lru_del(dentry);
	} else if (!(dentry->d_flags & DCACHE_SHRINK_LIST)) {
		dentry_lru_move_list(dentry, &data->dispose);
		dentry->d_flags |= DCACHE_SHRINK_LIST;
		data->found++;
		ret = D_WALK_NORETRY;
L
Linus Torvalds 已提交
1290
	}
M
Miklos Szeredi 已提交
1291 1292 1293 1294 1295 1296 1297
	/*
	 * We can return to the caller if we have found some (this
	 * ensures forward progress). We'll be coming back to find
	 * the rest.
	 */
	if (data->found && need_resched())
		ret = D_WALK_QUIT;
L
Linus Torvalds 已提交
1298
out:
M
Miklos Szeredi 已提交
1299
	return ret;
L
Linus Torvalds 已提交
1300 1301 1302 1303 1304 1305 1306 1307
}

/**
 * shrink_dcache_parent - prune dcache
 * @parent: parent of entries to prune
 *
 * Prune the dcache to remove unused children of the parent dentry.
 */
M
Miklos Szeredi 已提交
1308
void shrink_dcache_parent(struct dentry *parent)
L
Linus Torvalds 已提交
1309
{
M
Miklos Szeredi 已提交
1310 1311
	for (;;) {
		struct select_data data;
L
Linus Torvalds 已提交
1312

M
Miklos Szeredi 已提交
1313 1314 1315 1316 1317 1318 1319 1320 1321
		INIT_LIST_HEAD(&data.dispose);
		data.start = parent;
		data.found = 0;

		d_walk(parent, &data, select_collect, NULL);
		if (!data.found)
			break;

		shrink_dentry_list(&data.dispose);
1322 1323
		cond_resched();
	}
L
Linus Torvalds 已提交
1324
}
1325
EXPORT_SYMBOL(shrink_dcache_parent);
L
Linus Torvalds 已提交
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 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391
static enum d_walk_ret check_and_collect(void *_data, struct dentry *dentry)
{
	struct select_data *data = _data;

	if (d_mountpoint(dentry)) {
		data->found = -EBUSY;
		return D_WALK_QUIT;
	}

	return select_collect(_data, dentry);
}

static void check_and_drop(void *_data)
{
	struct select_data *data = _data;

	if (d_mountpoint(data->start))
		data->found = -EBUSY;
	if (!data->found)
		__d_drop(data->start);
}

/**
 * check_submounts_and_drop - prune dcache, check for submounts and drop
 *
 * All done as a single atomic operation relative to has_unlinked_ancestor().
 * Returns 0 if successfully unhashed @parent.  If there were submounts then
 * return -EBUSY.
 *
 * @dentry: dentry to prune and drop
 */
int check_submounts_and_drop(struct dentry *dentry)
{
	int ret = 0;

	/* Negative dentries can be dropped without further checks */
	if (!dentry->d_inode) {
		d_drop(dentry);
		goto out;
	}

	for (;;) {
		struct select_data data;

		INIT_LIST_HEAD(&data.dispose);
		data.start = dentry;
		data.found = 0;

		d_walk(dentry, &data, check_and_collect, check_and_drop);
		ret = data.found;

		if (!list_empty(&data.dispose))
			shrink_dentry_list(&data.dispose);

		if (ret <= 0)
			break;

		cond_resched();
	}

out:
	return ret;
}
EXPORT_SYMBOL(check_submounts_and_drop);

L
Linus Torvalds 已提交
1392
/**
1393 1394
 * __d_alloc	-	allocate a dcache entry
 * @sb: filesystem it will belong to
L
Linus Torvalds 已提交
1395 1396 1397 1398 1399 1400 1401
 * @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.
 */
 
1402
struct dentry *__d_alloc(struct super_block *sb, const struct qstr *name)
L
Linus Torvalds 已提交
1403 1404 1405 1406
{
	struct dentry *dentry;
	char *dname;

1407
	dentry = kmem_cache_alloc(dentry_cache, GFP_KERNEL);
L
Linus Torvalds 已提交
1408 1409 1410
	if (!dentry)
		return NULL;

1411 1412 1413 1414 1415 1416 1417
	/*
	 * 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
1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432
	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;

1433 1434 1435 1436
	/* Make sure we always see the terminating NUL character */
	smp_wmb();
	dentry->d_name.name = dname;

1437
	dentry->d_lockref.count = 1;
1438
	dentry->d_flags = 0;
L
Linus Torvalds 已提交
1439
	spin_lock_init(&dentry->d_lock);
N
Nick Piggin 已提交
1440
	seqcount_init(&dentry->d_seq);
L
Linus Torvalds 已提交
1441
	dentry->d_inode = NULL;
1442 1443
	dentry->d_parent = dentry;
	dentry->d_sb = sb;
L
Linus Torvalds 已提交
1444 1445
	dentry->d_op = NULL;
	dentry->d_fsdata = NULL;
1446
	INIT_HLIST_BL_NODE(&dentry->d_hash);
L
Linus Torvalds 已提交
1447 1448
	INIT_LIST_HEAD(&dentry->d_lru);
	INIT_LIST_HEAD(&dentry->d_subdirs);
A
Al Viro 已提交
1449
	INIT_HLIST_NODE(&dentry->d_alias);
N
Nick Piggin 已提交
1450
	INIT_LIST_HEAD(&dentry->d_u.d_child);
1451
	d_set_d_op(dentry, dentry->d_sb->s_d_op);
L
Linus Torvalds 已提交
1452

1453
	this_cpu_inc(nr_dentry);
1454

L
Linus Torvalds 已提交
1455 1456
	return dentry;
}
1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484

/**
 * 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;
}
1485
EXPORT_SYMBOL(d_alloc);
L
Linus Torvalds 已提交
1486

1487 1488
struct dentry *d_alloc_pseudo(struct super_block *sb, const struct qstr *name)
{
1489 1490
	struct dentry *dentry = __d_alloc(sb, name);
	if (dentry)
1491 1492 1493 1494 1495
		dentry->d_flags |= DCACHE_DISCONNECTED;
	return dentry;
}
EXPORT_SYMBOL(d_alloc_pseudo);

L
Linus Torvalds 已提交
1496 1497 1498 1499 1500 1501 1502 1503 1504
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);
}
1505
EXPORT_SYMBOL(d_alloc_name);
L
Linus Torvalds 已提交
1506

1507 1508
void d_set_d_op(struct dentry *dentry, const struct dentry_operations *op)
{
1509 1510
	WARN_ON_ONCE(dentry->d_op);
	WARN_ON_ONCE(dentry->d_flags & (DCACHE_OP_HASH	|
1511 1512
				DCACHE_OP_COMPARE	|
				DCACHE_OP_REVALIDATE	|
1513
				DCACHE_OP_WEAK_REVALIDATE	|
1514 1515 1516 1517 1518 1519 1520 1521 1522 1523
				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;
1524 1525
	if (op->d_weak_revalidate)
		dentry->d_flags |= DCACHE_OP_WEAK_REVALIDATE;
1526 1527
	if (op->d_delete)
		dentry->d_flags |= DCACHE_OP_DELETE;
S
Sage Weil 已提交
1528 1529
	if (op->d_prune)
		dentry->d_flags |= DCACHE_OP_PRUNE;
1530 1531 1532 1533

}
EXPORT_SYMBOL(d_set_d_op);

1534 1535
static void __d_instantiate(struct dentry *dentry, struct inode *inode)
{
N
Nick Piggin 已提交
1536
	spin_lock(&dentry->d_lock);
1537 1538 1539
	if (inode) {
		if (unlikely(IS_AUTOMOUNT(inode)))
			dentry->d_flags |= DCACHE_NEED_AUTOMOUNT;
A
Al Viro 已提交
1540
		hlist_add_head(&dentry->d_alias, &inode->i_dentry);
1541
	}
1542
	dentry->d_inode = inode;
N
Nick Piggin 已提交
1543
	dentry_rcuwalk_barrier(dentry);
N
Nick Piggin 已提交
1544
	spin_unlock(&dentry->d_lock);
1545 1546 1547
	fsnotify_d_instantiate(dentry, inode);
}

L
Linus Torvalds 已提交
1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564
/**
 * 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)
{
A
Al Viro 已提交
1565
	BUG_ON(!hlist_unhashed(&entry->d_alias));
1566 1567
	if (inode)
		spin_lock(&inode->i_lock);
1568
	__d_instantiate(entry, inode);
1569 1570
	if (inode)
		spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1571 1572
	security_d_instantiate(entry, inode);
}
1573
EXPORT_SYMBOL(d_instantiate);
L
Linus Torvalds 已提交
1574 1575 1576 1577 1578 1579 1580 1581

/**
 * 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
1582
 * aliased dentry instead and drop one reference to inode.
L
Linus Torvalds 已提交
1583 1584 1585
 *
 * Note that in order to avoid conflicts with rename() etc, the caller
 * had better be holding the parent directory semaphore.
1586 1587 1588 1589
 *
 * 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
1590
 */
1591 1592
static struct dentry *__d_instantiate_unique(struct dentry *entry,
					     struct inode *inode)
L
Linus Torvalds 已提交
1593 1594 1595 1596 1597 1598
{
	struct dentry *alias;
	int len = entry->d_name.len;
	const char *name = entry->d_name.name;
	unsigned int hash = entry->d_name.hash;

1599
	if (!inode) {
1600
		__d_instantiate(entry, NULL);
1601 1602 1603
		return NULL;
	}

1604
	hlist_for_each_entry(alias, &inode->i_dentry, d_alias) {
N
Nick Piggin 已提交
1605 1606 1607 1608 1609
		/*
		 * 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.
		 */
1610
		if (alias->d_name.hash != hash)
L
Linus Torvalds 已提交
1611 1612 1613
			continue;
		if (alias->d_parent != entry->d_parent)
			continue;
1614 1615
		if (alias->d_name.len != len)
			continue;
1616
		if (dentry_cmp(alias, name, len))
L
Linus Torvalds 已提交
1617
			continue;
1618
		__dget(alias);
L
Linus Torvalds 已提交
1619 1620
		return alias;
	}
1621

1622
	__d_instantiate(entry, inode);
L
Linus Torvalds 已提交
1623 1624
	return NULL;
}
1625 1626 1627 1628 1629

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

A
Al Viro 已提交
1630
	BUG_ON(!hlist_unhashed(&entry->d_alias));
1631

1632 1633
	if (inode)
		spin_lock(&inode->i_lock);
1634
	result = __d_instantiate_unique(entry, inode);
1635 1636
	if (inode)
		spin_unlock(&inode->i_lock);
1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647

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

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

L
Linus Torvalds 已提交
1648 1649
EXPORT_SYMBOL(d_instantiate_unique);

A
Al Viro 已提交
1650 1651 1652 1653 1654
struct dentry *d_make_root(struct inode *root_inode)
{
	struct dentry *res = NULL;

	if (root_inode) {
1655
		static const struct qstr name = QSTR_INIT("/", 1);
A
Al Viro 已提交
1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666

		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);

1667 1668 1669 1670
static struct dentry * __d_find_any_alias(struct inode *inode)
{
	struct dentry *alias;

A
Al Viro 已提交
1671
	if (hlist_empty(&inode->i_dentry))
1672
		return NULL;
A
Al Viro 已提交
1673
	alias = hlist_entry(inode->i_dentry.first, struct dentry, d_alias);
1674 1675 1676 1677
	__dget(alias);
	return alias;
}

1678 1679 1680 1681 1682 1683 1684 1685
/**
 * 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)
1686 1687 1688 1689 1690 1691 1692 1693
{
	struct dentry *de;

	spin_lock(&inode->i_lock);
	de = __d_find_any_alias(inode);
	spin_unlock(&inode->i_lock);
	return de;
}
1694
EXPORT_SYMBOL(d_find_any_alias);
1695

1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708
/**
 * 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
1709 1710 1711 1712
 * 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).
1713 1714 1715
 */
struct dentry *d_obtain_alias(struct inode *inode)
{
1716
	static const struct qstr anonstring = QSTR_INIT("/", 1);
C
Christoph Hellwig 已提交
1717 1718
	struct dentry *tmp;
	struct dentry *res;
1719 1720

	if (!inode)
1721
		return ERR_PTR(-ESTALE);
1722 1723 1724
	if (IS_ERR(inode))
		return ERR_CAST(inode);

1725
	res = d_find_any_alias(inode);
C
Christoph Hellwig 已提交
1726 1727 1728
	if (res)
		goto out_iput;

1729
	tmp = __d_alloc(inode->i_sb, &anonstring);
C
Christoph Hellwig 已提交
1730 1731 1732
	if (!tmp) {
		res = ERR_PTR(-ENOMEM);
		goto out_iput;
1733
	}
N
Nick Piggin 已提交
1734

1735
	spin_lock(&inode->i_lock);
1736
	res = __d_find_any_alias(inode);
C
Christoph Hellwig 已提交
1737
	if (res) {
1738
		spin_unlock(&inode->i_lock);
C
Christoph Hellwig 已提交
1739 1740 1741 1742 1743 1744 1745 1746
		dput(tmp);
		goto out_iput;
	}

	/* attach a disconnected dentry */
	spin_lock(&tmp->d_lock);
	tmp->d_inode = inode;
	tmp->d_flags |= DCACHE_DISCONNECTED;
A
Al Viro 已提交
1747
	hlist_add_head(&tmp->d_alias, &inode->i_dentry);
1748
	hlist_bl_lock(&tmp->d_sb->s_anon);
1749
	hlist_bl_add_head(&tmp->d_hash, &tmp->d_sb->s_anon);
1750
	hlist_bl_unlock(&tmp->d_sb->s_anon);
C
Christoph Hellwig 已提交
1751
	spin_unlock(&tmp->d_lock);
1752
	spin_unlock(&inode->i_lock);
1753
	security_d_instantiate(tmp, inode);
C
Christoph Hellwig 已提交
1754 1755 1756 1757

	return tmp;

 out_iput:
1758 1759
	if (res && !IS_ERR(res))
		security_d_instantiate(res, inode);
C
Christoph Hellwig 已提交
1760 1761
	iput(inode);
	return res;
1762
}
1763
EXPORT_SYMBOL(d_obtain_alias);
L
Linus Torvalds 已提交
1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779

/**
 * 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.
 *
1780 1781 1782 1783
 * Cluster filesystems may call this function with a negative, hashed dentry.
 * In that case, we know that the inode will be a regular file, and also this
 * will only occur during atomic_open. So we need to check for the dentry
 * being already hashed only in the final case.
L
Linus Torvalds 已提交
1784 1785 1786 1787 1788
 */
struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
{
	struct dentry *new = NULL;

1789 1790 1791
	if (IS_ERR(inode))
		return ERR_CAST(inode);

1792
	if (inode && S_ISDIR(inode->i_mode)) {
1793
		spin_lock(&inode->i_lock);
1794
		new = __d_find_alias(inode, 1);
L
Linus Torvalds 已提交
1795
		if (new) {
1796
			BUG_ON(!(new->d_flags & DCACHE_DISCONNECTED));
1797
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1798 1799 1800 1801
			security_d_instantiate(new, inode);
			d_move(new, dentry);
			iput(inode);
		} else {
1802
			/* already taking inode->i_lock, so d_add() by hand */
1803
			__d_instantiate(dentry, inode);
1804
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1805 1806 1807
			security_d_instantiate(dentry, inode);
			d_rehash(dentry);
		}
1808 1809 1810 1811 1812
	} else {
		d_instantiate(dentry, inode);
		if (d_unhashed(dentry))
			d_rehash(dentry);
	}
L
Linus Torvalds 已提交
1813 1814
	return new;
}
1815
EXPORT_SYMBOL(d_splice_alias);
L
Linus Torvalds 已提交
1816

1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832
/**
 * 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.
 */
1833
struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
1834 1835 1836 1837 1838
			struct qstr *name)
{
	struct dentry *found;
	struct dentry *new;

C
Christoph Hellwig 已提交
1839 1840 1841 1842
	/*
	 * First check if a dentry matching the name already exists,
	 * if not go ahead and create it now.
	 */
1843
	found = d_hash_and_lookup(dentry->d_parent, name);
1844 1845
	if (unlikely(IS_ERR(found)))
		goto err_out;
1846 1847 1848
	if (!found) {
		new = d_alloc(dentry->d_parent, name);
		if (!new) {
1849
			found = ERR_PTR(-ENOMEM);
1850 1851
			goto err_out;
		}
C
Christoph Hellwig 已提交
1852

1853 1854 1855 1856 1857 1858 1859
		found = d_splice_alias(inode, new);
		if (found) {
			dput(new);
			return found;
		}
		return new;
	}
C
Christoph Hellwig 已提交
1860 1861 1862 1863 1864 1865 1866

	/*
	 * If a matching dentry exists, and it's not negative use it.
	 *
	 * Decrement the reference count to balance the iget() done
	 * earlier on.
	 */
1867 1868 1869 1870 1871 1872 1873 1874 1875
	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
1876

1877 1878
	/*
	 * Negative dentry: instantiate it unless the inode is a directory and
C
Christoph Hellwig 已提交
1879
	 * already has a dentry.
1880
	 */
1881 1882 1883 1884
	new = d_splice_alias(inode, found);
	if (new) {
		dput(found);
		found = new;
1885
	}
1886
	return found;
1887 1888 1889

err_out:
	iput(inode);
1890
	return found;
1891
}
1892
EXPORT_SYMBOL(d_add_ci);
L
Linus Torvalds 已提交
1893

1894 1895 1896 1897
/*
 * Do the slow-case of the dentry name compare.
 *
 * Unlike the dentry_cmp() function, we need to atomically
1898
 * load the name and length information, so that the
1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926
 * 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 dentry *dentry,
		unsigned int seq,
		const struct qstr *name)
{
	int tlen = dentry->d_name.len;
	const char *tname = dentry->d_name.name;

	if (read_seqcount_retry(&dentry->d_seq, seq)) {
		cpu_relax();
		return D_COMP_SEQRETRY;
	}
1927
	if (parent->d_op->d_compare(parent, dentry, tlen, tname, name))
1928 1929 1930 1931
		return D_COMP_NOMATCH;
	return D_COMP_OK;
}

N
Nick Piggin 已提交
1932 1933 1934 1935
/**
 * __d_lookup_rcu - search for a dentry (racy, store-free)
 * @parent: parent dentry
 * @name: qstr of name we wish to find
1936
 * @seqp: returns d_seq value at the point where the dentry was found
N
Nick Piggin 已提交
1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949
 * 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.
 *
1950
 * A refcount may be taken on the found dentry with the d_rcu_to_refcount
N
Nick Piggin 已提交
1951 1952 1953 1954 1955 1956
 * 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.
1957 1958 1959
 *
 * 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
1960
 */
1961 1962
struct dentry *__d_lookup_rcu(const struct dentry *parent,
				const struct qstr *name,
1963
				unsigned *seqp)
N
Nick Piggin 已提交
1964
{
1965
	u64 hashlen = name->hash_len;
N
Nick Piggin 已提交
1966
	const unsigned char *str = name->name;
1967
	struct hlist_bl_head *b = d_hash(parent, hashlen_hash(hashlen));
1968
	struct hlist_bl_node *node;
N
Nick Piggin 已提交
1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988
	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.
	 *
1989
	 * See Documentation/filesystems/path-lookup.txt for more details.
N
Nick Piggin 已提交
1990
	 */
1991
	hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
1992
		unsigned seq;
N
Nick Piggin 已提交
1993 1994

seqretry:
1995 1996
		/*
		 * The dentry sequence count protects us from concurrent
1997
		 * renames, and thus protects parent and name fields.
1998 1999
		 *
		 * The caller must perform a seqcount check in order
2000
		 * to do anything useful with the returned dentry.
2001 2002 2003 2004 2005 2006 2007 2008 2009
		 *
		 * 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
2010 2011
		if (dentry->d_parent != parent)
			continue;
2012 2013
		if (d_unhashed(dentry))
			continue;
2014

2015
		if (unlikely(parent->d_flags & DCACHE_OP_COMPARE)) {
2016 2017
			if (dentry->d_name.hash != hashlen_hash(hashlen))
				continue;
2018 2019
			*seqp = seq;
			switch (slow_dentry_cmp(parent, dentry, seq, name)) {
2020 2021 2022
			case D_COMP_OK:
				return dentry;
			case D_COMP_NOMATCH:
N
Nick Piggin 已提交
2023
				continue;
2024 2025 2026
			default:
				goto seqretry;
			}
N
Nick Piggin 已提交
2027
		}
2028

2029
		if (dentry->d_name.hash_len != hashlen)
2030
			continue;
2031
		*seqp = seq;
2032
		if (!dentry_cmp(dentry, str, hashlen_len(hashlen)))
2033
			return dentry;
N
Nick Piggin 已提交
2034 2035 2036 2037
	}
	return NULL;
}

L
Linus Torvalds 已提交
2038 2039 2040 2041
/**
 * d_lookup - search for a dentry
 * @parent: parent dentry
 * @name: qstr of name we wish to find
2042
 * Returns: dentry, or NULL
L
Linus Torvalds 已提交
2043
 *
2044 2045 2046 2047
 * 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
2048
 */
A
Al Viro 已提交
2049
struct dentry *d_lookup(const struct dentry *parent, const struct qstr *name)
L
Linus Torvalds 已提交
2050
{
N
Nick Piggin 已提交
2051
	struct dentry *dentry;
2052
	unsigned seq;
L
Linus Torvalds 已提交
2053 2054 2055 2056 2057 2058 2059 2060 2061

        do {
                seq = read_seqbegin(&rename_lock);
                dentry = __d_lookup(parent, name);
                if (dentry)
			break;
	} while (read_seqretry(&rename_lock, seq));
	return dentry;
}
2062
EXPORT_SYMBOL(d_lookup);
L
Linus Torvalds 已提交
2063

N
Nick Piggin 已提交
2064
/**
2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078
 * __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.
 */
A
Al Viro 已提交
2079
struct dentry *__d_lookup(const struct dentry *parent, const struct qstr *name)
L
Linus Torvalds 已提交
2080 2081 2082 2083
{
	unsigned int len = name->len;
	unsigned int hash = name->hash;
	const unsigned char *str = name->name;
2084
	struct hlist_bl_head *b = d_hash(parent, hash);
2085
	struct hlist_bl_node *node;
N
Nick Piggin 已提交
2086
	struct dentry *found = NULL;
2087
	struct dentry *dentry;
L
Linus Torvalds 已提交
2088

N
Nick Piggin 已提交
2089 2090 2091 2092 2093 2094 2095
	/*
	 * 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.
	 */

2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106
	/*
	 * 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.
	 *
2107
	 * See Documentation/filesystems/path-lookup.txt for more details.
2108
	 */
L
Linus Torvalds 已提交
2109 2110
	rcu_read_lock();
	
2111
	hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
L
Linus Torvalds 已提交
2112 2113 2114 2115 2116 2117 2118

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

		spin_lock(&dentry->d_lock);
		if (dentry->d_parent != parent)
			goto next;
2119 2120 2121
		if (d_unhashed(dentry))
			goto next;

L
Linus Torvalds 已提交
2122 2123 2124 2125
		/*
		 * It is safe to compare names since d_move() cannot
		 * change the qstr (protected by d_lock).
		 */
2126
		if (parent->d_flags & DCACHE_OP_COMPARE) {
2127 2128
			int tlen = dentry->d_name.len;
			const char *tname = dentry->d_name.name;
2129
			if (parent->d_op->d_compare(parent, dentry, tlen, tname, name))
L
Linus Torvalds 已提交
2130 2131
				goto next;
		} else {
2132 2133
			if (dentry->d_name.len != len)
				goto next;
2134
			if (dentry_cmp(dentry, str, len))
L
Linus Torvalds 已提交
2135 2136 2137
				goto next;
		}

2138
		dentry->d_lockref.count++;
2139
		found = dentry;
L
Linus Torvalds 已提交
2140 2141 2142 2143 2144 2145 2146 2147 2148 2149
		spin_unlock(&dentry->d_lock);
		break;
next:
		spin_unlock(&dentry->d_lock);
 	}
 	rcu_read_unlock();

 	return found;
}

2150 2151 2152 2153 2154
/**
 * 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
 *
2155
 * On lookup failure NULL is returned; on bad name - ERR_PTR(-error)
2156 2157 2158 2159 2160 2161 2162 2163 2164
 */
struct dentry *d_hash_and_lookup(struct dentry *dir, struct qstr *name)
{
	/*
	 * 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);
2165
	if (dir->d_flags & DCACHE_OP_HASH) {
2166
		int err = dir->d_op->d_hash(dir, name);
2167 2168
		if (unlikely(err < 0))
			return ERR_PTR(err);
2169
	}
2170
	return d_lookup(dir, name);
2171
}
2172
EXPORT_SYMBOL(d_hash_and_lookup);
2173

L
Linus Torvalds 已提交
2174
/**
N
Nick Piggin 已提交
2175
 * d_validate - verify dentry provided from insecure source (deprecated)
L
Linus Torvalds 已提交
2176
 * @dentry: The dentry alleged to be valid child of @dparent
2177
 * @dparent: The parent dentry (known to be valid)
L
Linus Torvalds 已提交
2178 2179 2180 2181
 *
 * 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
2182 2183
 *
 * This function is slow for big directories, and deprecated, do not use it.
L
Linus Torvalds 已提交
2184
 */
2185
int d_validate(struct dentry *dentry, struct dentry *dparent)
L
Linus Torvalds 已提交
2186
{
N
Nick Piggin 已提交
2187
	struct dentry *child;
2188

N
Nick Piggin 已提交
2189
	spin_lock(&dparent->d_lock);
N
Nick Piggin 已提交
2190 2191
	list_for_each_entry(child, &dparent->d_subdirs, d_u.d_child) {
		if (dentry == child) {
N
Nick Piggin 已提交
2192
			spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
2193
			__dget_dlock(dentry);
N
Nick Piggin 已提交
2194 2195
			spin_unlock(&dentry->d_lock);
			spin_unlock(&dparent->d_lock);
L
Linus Torvalds 已提交
2196 2197 2198
			return 1;
		}
	}
N
Nick Piggin 已提交
2199
	spin_unlock(&dparent->d_lock);
N
Nick Piggin 已提交
2200

L
Linus Torvalds 已提交
2201 2202
	return 0;
}
2203
EXPORT_SYMBOL(d_validate);
L
Linus Torvalds 已提交
2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227

/*
 * 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)
{
2228
	struct inode *inode;
2229
	int isdir = 0;
L
Linus Torvalds 已提交
2230 2231 2232
	/*
	 * Are we the only user?
	 */
2233
again:
L
Linus Torvalds 已提交
2234
	spin_lock(&dentry->d_lock);
2235 2236
	inode = dentry->d_inode;
	isdir = S_ISDIR(inode->i_mode);
2237
	if (dentry->d_lockref.count == 1) {
A
Alan Cox 已提交
2238
		if (!spin_trylock(&inode->i_lock)) {
2239 2240 2241 2242
			spin_unlock(&dentry->d_lock);
			cpu_relax();
			goto again;
		}
A
Al Viro 已提交
2243
		dentry->d_flags &= ~DCACHE_CANT_MOUNT;
N
Nick Piggin 已提交
2244
		dentry_unlink_inode(dentry);
2245
		fsnotify_nameremove(dentry, isdir);
L
Linus Torvalds 已提交
2246 2247 2248 2249 2250 2251 2252
		return;
	}

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

	spin_unlock(&dentry->d_lock);
2253 2254

	fsnotify_nameremove(dentry, isdir);
L
Linus Torvalds 已提交
2255
}
2256
EXPORT_SYMBOL(d_delete);
L
Linus Torvalds 已提交
2257

2258
static void __d_rehash(struct dentry * entry, struct hlist_bl_head *b)
L
Linus Torvalds 已提交
2259
{
2260
	BUG_ON(!d_unhashed(entry));
2261
	hlist_bl_lock(b);
2262
	entry->d_flags |= DCACHE_RCUACCESS;
2263
	hlist_bl_add_head_rcu(&entry->d_hash, b);
2264
	hlist_bl_unlock(b);
L
Linus Torvalds 已提交
2265 2266
}

2267 2268 2269 2270 2271
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
2272 2273 2274 2275 2276 2277 2278 2279 2280 2281
/**
 * 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);
2282
	_d_rehash(entry);
L
Linus Torvalds 已提交
2283 2284
	spin_unlock(&entry->d_lock);
}
2285
EXPORT_SYMBOL(d_rehash);
L
Linus Torvalds 已提交
2286

N
Nick Piggin 已提交
2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302
/**
 * 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)
{
2303
	BUG_ON(!mutex_is_locked(&dentry->d_parent->d_inode->i_mutex));
N
Nick Piggin 已提交
2304 2305 2306
	BUG_ON(dentry->d_name.len != name->len); /* d_lookup gives this */

	spin_lock(&dentry->d_lock);
N
Nick Piggin 已提交
2307
	write_seqcount_begin(&dentry->d_seq);
N
Nick Piggin 已提交
2308
	memcpy((unsigned char *)dentry->d_name.name, name->name, name->len);
N
Nick Piggin 已提交
2309
	write_seqcount_end(&dentry->d_seq);
N
Nick Piggin 已提交
2310 2311 2312 2313
	spin_unlock(&dentry->d_lock);
}
EXPORT_SYMBOL(dentry_update_name_case);

L
Linus Torvalds 已提交
2314 2315 2316 2317 2318 2319 2320
static void switch_names(struct dentry *dentry, struct dentry *target)
{
	if (dname_external(target)) {
		if (dname_external(dentry)) {
			/*
			 * Both external: swap the pointers
			 */
2321
			swap(target->d_name.name, dentry->d_name.name);
L
Linus Torvalds 已提交
2322 2323 2324 2325 2326
		} else {
			/*
			 * dentry:internal, target:external.  Steal target's
			 * storage and make target internal.
			 */
2327 2328
			memcpy(target->d_iname, dentry->d_name.name,
					dentry->d_name.len + 1);
L
Linus Torvalds 已提交
2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347
			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);
2348 2349
			dentry->d_name.len = target->d_name.len;
			return;
L
Linus Torvalds 已提交
2350 2351
		}
	}
2352
	swap(dentry->d_name.len, target->d_name.len);
L
Linus Torvalds 已提交
2353 2354
}

N
Nick Piggin 已提交
2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390
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
2391
/*
N
Nick Piggin 已提交
2392 2393 2394 2395 2396 2397 2398 2399 2400
 * 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
2401
 */
2402
/*
2403
 * __d_move - move a dentry
L
Linus Torvalds 已提交
2404 2405 2406 2407
 * @dentry: entry to move
 * @target: new dentry
 *
 * Update the dcache to reflect the move of a file name. Negative
2408 2409 2410
 * 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
2411
 */
2412
static void __d_move(struct dentry * dentry, struct dentry * target)
L
Linus Torvalds 已提交
2413 2414 2415 2416
{
	if (!dentry->d_inode)
		printk(KERN_WARNING "VFS: moving negative dcache entry\n");

N
Nick Piggin 已提交
2417 2418 2419 2420
	BUG_ON(d_ancestor(dentry, target));
	BUG_ON(d_ancestor(target, dentry));

	dentry_lock_for_move(dentry, target);
L
Linus Torvalds 已提交
2421

N
Nick Piggin 已提交
2422 2423 2424
	write_seqcount_begin(&dentry->d_seq);
	write_seqcount_begin(&target->d_seq);

2425 2426 2427 2428 2429 2430 2431
	/* __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
2432
	__d_rehash(dentry, d_hash(target->d_parent, target->d_name.hash));
L
Linus Torvalds 已提交
2433 2434 2435 2436

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

E
Eric Dumazet 已提交
2437 2438
	list_del(&dentry->d_u.d_child);
	list_del(&target->d_u.d_child);
L
Linus Torvalds 已提交
2439 2440 2441

	/* Switch the names.. */
	switch_names(dentry, target);
2442
	swap(dentry->d_name.hash, target->d_name.hash);
L
Linus Torvalds 已提交
2443 2444 2445 2446 2447

	/* ... and switch the parents */
	if (IS_ROOT(dentry)) {
		dentry->d_parent = target->d_parent;
		target->d_parent = target;
E
Eric Dumazet 已提交
2448
		INIT_LIST_HEAD(&target->d_u.d_child);
L
Linus Torvalds 已提交
2449
	} else {
2450
		swap(dentry->d_parent, target->d_parent);
L
Linus Torvalds 已提交
2451 2452

		/* And add them back to the (new) parent lists */
E
Eric Dumazet 已提交
2453
		list_add(&target->d_u.d_child, &target->d_parent->d_subdirs);
L
Linus Torvalds 已提交
2454 2455
	}

E
Eric Dumazet 已提交
2456
	list_add(&dentry->d_u.d_child, &dentry->d_parent->d_subdirs);
N
Nick Piggin 已提交
2457

N
Nick Piggin 已提交
2458 2459 2460
	write_seqcount_end(&target->d_seq);
	write_seqcount_end(&dentry->d_seq);

N
Nick Piggin 已提交
2461
	dentry_unlock_parents_for_move(dentry, target);
L
Linus Torvalds 已提交
2462
	spin_unlock(&target->d_lock);
2463
	fsnotify_d_move(dentry);
L
Linus Torvalds 已提交
2464
	spin_unlock(&dentry->d_lock);
2465 2466 2467 2468 2469 2470 2471 2472
}

/*
 * d_move - move a dentry
 * @dentry: entry to move
 * @target: new dentry
 *
 * Update the dcache to reflect the move of a file name. Negative
2473 2474
 * dcache entries should not be moved in this way. See the locking
 * requirements for __d_move.
2475 2476 2477 2478 2479
 */
void d_move(struct dentry *dentry, struct dentry *target)
{
	write_seqlock(&rename_lock);
	__d_move(dentry, target);
L
Linus Torvalds 已提交
2480
	write_sequnlock(&rename_lock);
2481
}
2482
EXPORT_SYMBOL(d_move);
L
Linus Torvalds 已提交
2483

2484 2485 2486 2487 2488 2489 2490
/**
 * 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.
2491
 */
2492
struct dentry *d_ancestor(struct dentry *p1, struct dentry *p2)
2493 2494 2495
{
	struct dentry *p;

2496
	for (p = p2; !IS_ROOT(p); p = p->d_parent) {
2497
		if (p->d_parent == p1)
2498
			return p;
2499
	}
2500
	return NULL;
2501 2502 2503 2504 2505 2506
}

/*
 * This helper attempts to cope with remotely renamed directories
 *
 * It assumes that the caller is already holding
2507
 * dentry->d_parent->d_inode->i_mutex, inode->i_lock and rename_lock
2508 2509 2510 2511
 *
 * Note: If ever the locking in lock_rename() changes, then please
 * remember to update this too...
 */
2512 2513
static struct dentry *__d_unalias(struct inode *inode,
		struct dentry *dentry, struct dentry *alias)
2514 2515
{
	struct mutex *m1 = NULL, *m2 = NULL;
2516
	struct dentry *ret = ERR_PTR(-EBUSY);
2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529

	/* 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() */
	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:
2530 2531 2532 2533
	if (likely(!d_mountpoint(alias))) {
		__d_move(alias, dentry);
		ret = alias;
	}
2534
out_err:
2535
	spin_unlock(&inode->i_lock);
2536 2537 2538 2539 2540 2541 2542
	if (m2)
		mutex_unlock(m2);
	if (m1)
		mutex_unlock(m1);
	return ret;
}

2543 2544 2545
/*
 * 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
2546
 * returns with anon->d_lock held!
2547 2548 2549
 */
static void __d_materialise_dentry(struct dentry *dentry, struct dentry *anon)
{
2550
	struct dentry *dparent;
2551

N
Nick Piggin 已提交
2552
	dentry_lock_for_move(anon, dentry);
2553

N
Nick Piggin 已提交
2554 2555 2556
	write_seqcount_begin(&dentry->d_seq);
	write_seqcount_begin(&anon->d_seq);

2557 2558
	dparent = dentry->d_parent;

N
Nick Piggin 已提交
2559 2560 2561
	switch_names(dentry, anon);
	swap(dentry->d_name.hash, anon->d_name.hash);

2562 2563 2564
	dentry->d_parent = dentry;
	list_del_init(&dentry->d_u.d_child);
	anon->d_parent = dparent;
2565
	list_move(&anon->d_u.d_child, &dparent->d_subdirs);
2566

N
Nick Piggin 已提交
2567 2568 2569
	write_seqcount_end(&dentry->d_seq);
	write_seqcount_end(&anon->d_seq);

N
Nick Piggin 已提交
2570 2571 2572 2573
	dentry_unlock_parents_for_move(anon, dentry);
	spin_unlock(&dentry->d_lock);

	/* anon->d_lock still locked, returns locked */
2574 2575 2576 2577 2578 2579 2580 2581 2582
	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
2583 2584
 * root directory alias in its place if there is one. Caller must hold the
 * i_mutex of the parent directory.
2585 2586 2587
 */
struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
{
2588
	struct dentry *actual;
2589 2590 2591 2592 2593

	BUG_ON(!d_unhashed(dentry));

	if (!inode) {
		actual = dentry;
2594
		__d_instantiate(dentry, NULL);
2595 2596
		d_rehash(actual);
		goto out_nolock;
2597 2598
	}

2599
	spin_lock(&inode->i_lock);
2600

2601 2602 2603 2604
	if (S_ISDIR(inode->i_mode)) {
		struct dentry *alias;

		/* Does an aliased dentry already exist? */
2605
		alias = __d_find_alias(inode, 0);
2606 2607
		if (alias) {
			actual = alias;
2608 2609 2610 2611 2612
			write_seqlock(&rename_lock);

			if (d_ancestor(alias, dentry)) {
				/* Check for loops */
				actual = ERR_PTR(-ELOOP);
2613
				spin_unlock(&inode->i_lock);
2614 2615 2616
			} else if (IS_ROOT(alias)) {
				/* Is this an anonymous mountpoint that we
				 * could splice into our tree? */
2617
				__d_materialise_dentry(dentry, alias);
2618
				write_sequnlock(&rename_lock);
2619 2620
				__d_drop(alias);
				goto found;
2621 2622
			} else {
				/* Nope, but we must(!) avoid directory
2623
				 * aliasing. This drops inode->i_lock */
2624
				actual = __d_unalias(inode, dentry, alias);
2625
			}
2626
			write_sequnlock(&rename_lock);
2627 2628 2629 2630 2631 2632 2633 2634
			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);
2635
				dput(alias);
2636
			}
2637 2638
			goto out_nolock;
		}
2639 2640 2641 2642 2643 2644
	}

	/* Add a unique reference */
	actual = __d_instantiate_unique(dentry, inode);
	if (!actual)
		actual = dentry;
2645 2646
	else
		BUG_ON(!d_unhashed(actual));
2647 2648 2649 2650 2651

	spin_lock(&actual->d_lock);
found:
	_d_rehash(actual);
	spin_unlock(&actual->d_lock);
2652
	spin_unlock(&inode->i_lock);
2653
out_nolock:
2654 2655 2656 2657 2658 2659 2660 2661
	if (actual == dentry) {
		security_d_instantiate(dentry, inode);
		return NULL;
	}

	iput(inode);
	return actual;
}
2662
EXPORT_SYMBOL_GPL(d_materialise_unique);
2663

2664
static int prepend(char **buffer, int *buflen, const char *str, int namelen)
2665 2666 2667 2668 2669 2670 2671 2672 2673
{
	*buflen -= namelen;
	if (*buflen < 0)
		return -ENAMETOOLONG;
	*buffer -= namelen;
	memcpy(*buffer, str, namelen);
	return 0;
}

2674 2675
/**
 * prepend_name - prepend a pathname in front of current buffer pointer
2676 2677 2678
 * @buffer: buffer pointer
 * @buflen: allocated length of the buffer
 * @name:   name string and length qstr structure
2679 2680 2681 2682 2683 2684 2685 2686 2687 2688
 *
 * With RCU path tracing, it may race with d_move(). Use ACCESS_ONCE() to
 * make sure that either the old or the new name pointer and length are
 * fetched. However, there may be mismatch between length and pointer.
 * The length cannot be trusted, we need to copy it byte-by-byte until
 * the length is reached or a null byte is found. It also prepends "/" at
 * the beginning of the name. The sequence number check at the caller will
 * retry it again when a d_move() does happen. So any garbage in the buffer
 * due to mismatched pointer and length will be discarded.
 */
2689 2690
static int prepend_name(char **buffer, int *buflen, struct qstr *name)
{
2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706
	const char *dname = ACCESS_ONCE(name->name);
	u32 dlen = ACCESS_ONCE(name->len);
	char *p;

	if (*buflen < dlen + 1)
		return -ENAMETOOLONG;
	*buflen -= dlen + 1;
	p = *buffer -= dlen + 1;
	*p++ = '/';
	while (dlen--) {
		char c = *dname++;
		if (!c)
			break;
		*p++ = c;
	}
	return 0;
2707 2708
}

L
Linus Torvalds 已提交
2709
/**
2710
 * prepend_path - Prepend path string to a buffer
2711
 * @path: the dentry/vfsmount to report
2712
 * @root: root vfsmnt/dentry
M
Miklos Szeredi 已提交
2713 2714
 * @buffer: pointer to the end of the buffer
 * @buflen: pointer to buffer length
2715
 *
2716 2717 2718 2719 2720 2721 2722 2723 2724
 * The function will first try to write out the pathname without taking any
 * lock other than the RCU read lock to make sure that dentries won't go away.
 * It only checks the sequence number of the global rename_lock as any change
 * in the dentry's d_seq will be preceded by changes in the rename_lock
 * sequence number. If the sequence number had been changed, it will restart
 * the whole pathname back-tracing sequence again by taking the rename_lock.
 * In this case, there is no need to take the RCU read lock as the recursive
 * parent pointer references will keep the dentry chain alive as long as no
 * rename operation is performed.
L
Linus Torvalds 已提交
2725
 */
2726 2727
static int prepend_path(const struct path *path,
			const struct path *root,
M
Miklos Szeredi 已提交
2728
			char **buffer, int *buflen)
L
Linus Torvalds 已提交
2729
{
2730 2731
	struct dentry *dentry = path->dentry;
	struct vfsmount *vfsmnt = path->mnt;
2732
	struct mount *mnt = real_mount(vfsmnt);
M
Miklos Szeredi 已提交
2733
	int error = 0;
2734 2735 2736
	unsigned seq = 0;
	char *bptr;
	int blen;
2737

2738
	rcu_read_lock();
2739 2740 2741 2742
restart:
	bptr = *buffer;
	blen = *buflen;
	read_seqbegin_or_lock(&rename_lock, &seq);
M
Miklos Szeredi 已提交
2743
	while (dentry != root->dentry || vfsmnt != root->mnt) {
L
Linus Torvalds 已提交
2744 2745 2746
		struct dentry * parent;

		if (dentry == vfsmnt->mnt_root || IS_ROOT(dentry)) {
2747
			/* Global root? */
2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767
			if (mnt_has_parent(mnt)) {
				dentry = mnt->mnt_mountpoint;
				mnt = mnt->mnt_parent;
				vfsmnt = &mnt->mnt;
				continue;
			}
			/*
			 * 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);
			}
			if (!error)
				error = is_mounted(vfsmnt) ? 1 : 2;
			break;
L
Linus Torvalds 已提交
2768 2769 2770
		}
		parent = dentry->d_parent;
		prefetch(parent);
2771
		error = prepend_name(&bptr, &blen, &dentry->d_name);
M
Miklos Szeredi 已提交
2772 2773 2774
		if (error)
			break;

L
Linus Torvalds 已提交
2775 2776
		dentry = parent;
	}
2777 2778 2779 2780
	if (!(seq & 1))
		rcu_read_unlock();
	if (need_seqretry(&rename_lock, seq)) {
		seq = 1;
2781
		goto restart;
2782 2783
	}
	done_seqretry(&rename_lock, seq);
L
Linus Torvalds 已提交
2784

2785 2786 2787 2788 2789 2790 2791 2792
	if (error >= 0 && bptr == *buffer) {
		if (--blen < 0)
			error = -ENAMETOOLONG;
		else
			*--bptr = '/';
	}
	*buffer = bptr;
	*buflen = blen;
A
Al Viro 已提交
2793
	return error;
M
Miklos Szeredi 已提交
2794
}
2795

M
Miklos Szeredi 已提交
2796 2797 2798
/**
 * __d_path - return the path of a dentry
 * @path: the dentry/vfsmount to report
2799
 * @root: root vfsmnt/dentry
2800
 * @buf: buffer to return value in
M
Miklos Szeredi 已提交
2801 2802
 * @buflen: buffer length
 *
2803
 * Convert a dentry into an ASCII path name.
M
Miklos Szeredi 已提交
2804 2805 2806 2807
 *
 * Returns a pointer into the buffer or an error code if the
 * path was too long.
 *
2808
 * "buflen" should be positive.
M
Miklos Szeredi 已提交
2809
 *
2810
 * If the path is not reachable from the supplied root, return %NULL.
M
Miklos Szeredi 已提交
2811
 */
2812 2813
char *__d_path(const struct path *path,
	       const struct path *root,
M
Miklos Szeredi 已提交
2814 2815 2816 2817 2818 2819
	       char *buf, int buflen)
{
	char *res = buf + buflen;
	int error;

	prepend(&res, &buflen, "\0", 1);
A
Al Viro 已提交
2820
	br_read_lock(&vfsmount_lock);
M
Miklos Szeredi 已提交
2821
	error = prepend_path(path, root, &res, &buflen);
A
Al Viro 已提交
2822
	br_read_unlock(&vfsmount_lock);
2823

2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838
	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);
A
Al Viro 已提交
2839
	br_read_lock(&vfsmount_lock);
2840
	error = prepend_path(path, &root, &res, &buflen);
A
Al Viro 已提交
2841
	br_read_unlock(&vfsmount_lock);
2842 2843 2844 2845

	if (error > 1)
		error = -EINVAL;
	if (error < 0)
M
Miklos Szeredi 已提交
2846 2847
		return ERR_PTR(error);
	return res;
L
Linus Torvalds 已提交
2848 2849
}

2850 2851 2852
/*
 * same as __d_path but appends "(deleted)" for unlinked files.
 */
2853 2854 2855
static int path_with_deleted(const struct path *path,
			     const struct path *root,
			     char **buf, int *buflen)
2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866
{
	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);
}

2867 2868 2869 2870 2871
static int prepend_unreachable(char **buffer, int *buflen)
{
	return prepend(buffer, buflen, "(unreachable)", 13);
}

2872 2873 2874 2875 2876 2877 2878 2879 2880 2881
static void get_fs_root_rcu(struct fs_struct *fs, struct path *root)
{
	unsigned seq;

	do {
		seq = read_seqcount_begin(&fs->seq);
		*root = fs->root;
	} while (read_seqcount_retry(&fs->seq, seq));
}

J
Jan Blunck 已提交
2882 2883
/**
 * d_path - return the path of a dentry
2884
 * @path: path to report
J
Jan Blunck 已提交
2885 2886 2887 2888 2889 2890
 * @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.
 *
2891 2892 2893 2894
 * 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
2895
 *
2896
 * "buflen" should be positive.
J
Jan Blunck 已提交
2897
 */
2898
char *d_path(const struct path *path, char *buf, int buflen)
L
Linus Torvalds 已提交
2899
{
2900
	char *res = buf + buflen;
J
Jan Blunck 已提交
2901
	struct path root;
2902
	int error;
L
Linus Torvalds 已提交
2903

2904 2905 2906 2907 2908 2909 2910
	/*
	 * 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:
	 */
2911 2912
	if (path->dentry->d_op && path->dentry->d_op->d_dname)
		return path->dentry->d_op->d_dname(path->dentry, buf, buflen);
2913

2914 2915
	rcu_read_lock();
	get_fs_root_rcu(current->fs, &root);
A
Al Viro 已提交
2916
	br_read_lock(&vfsmount_lock);
2917
	error = path_with_deleted(path, &root, &res, &buflen);
A
Al Viro 已提交
2918
	br_read_unlock(&vfsmount_lock);
2919 2920
	rcu_read_unlock();

2921
	if (error < 0)
2922
		res = ERR_PTR(error);
L
Linus Torvalds 已提交
2923 2924
	return res;
}
2925
EXPORT_SYMBOL(d_path);
L
Linus Torvalds 已提交
2926

2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947
/*
 * 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);
}

2948 2949 2950 2951 2952
char *simple_dname(struct dentry *dentry, char *buffer, int buflen)
{
	char *end = buffer + buflen;
	/* these dentries are never renamed, so d_lock is not needed */
	if (prepend(&end, &buflen, " (deleted)", 11) ||
2953
	    prepend(&end, &buflen, dentry->d_name.name, dentry->d_name.len) ||
2954 2955
	    prepend(&end, &buflen, "/", 1))  
		end = ERR_PTR(-ENAMETOOLONG);
2956
	return end;
2957 2958
}

2959 2960 2961
/*
 * Write full pathname from the root of the filesystem into the buffer.
 */
N
Nick Piggin 已提交
2962
static char *__dentry_path(struct dentry *dentry, char *buf, int buflen)
2963
{
2964 2965 2966
	char *end, *retval;
	int len, seq = 0;
	int error = 0;
2967

2968
	rcu_read_lock();
2969 2970 2971 2972
restart:
	end = buf + buflen;
	len = buflen;
	prepend(&end, &len, "\0", 1);
2973 2974 2975 2976 2977
	if (buflen < 1)
		goto Elong;
	/* Get '/' right */
	retval = end-1;
	*retval = '/';
2978
	read_seqbegin_or_lock(&rename_lock, &seq);
2979 2980
	while (!IS_ROOT(dentry)) {
		struct dentry *parent = dentry->d_parent;
N
Nick Piggin 已提交
2981
		int error;
2982 2983

		prefetch(parent);
2984 2985 2986
		error = prepend_name(&end, &len, &dentry->d_name);
		if (error)
			break;
2987 2988 2989 2990

		retval = end;
		dentry = parent;
	}
2991 2992 2993 2994
	if (!(seq & 1))
		rcu_read_unlock();
	if (need_seqretry(&rename_lock, seq)) {
		seq = 1;
2995
		goto restart;
2996 2997
	}
	done_seqretry(&rename_lock, seq);
2998 2999
	if (error)
		goto Elong;
A
Al Viro 已提交
3000 3001 3002 3003
	return retval;
Elong:
	return ERR_PTR(-ENAMETOOLONG);
}
N
Nick Piggin 已提交
3004 3005 3006

char *dentry_path_raw(struct dentry *dentry, char *buf, int buflen)
{
3007
	return __dentry_path(dentry, buf, buflen);
N
Nick Piggin 已提交
3008 3009
}
EXPORT_SYMBOL(dentry_path_raw);
A
Al Viro 已提交
3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024

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

	if (d_unlinked(dentry)) {
		p = buf + buflen;
		if (prepend(&p, &buflen, "//deleted", 10) != 0)
			goto Elong;
		buflen++;
	}
	retval = __dentry_path(dentry, buf, buflen);
	if (!IS_ERR(retval) && p)
		*p = '/';	/* restore '/' overriden with '\0' */
3025 3026 3027 3028 3029
	return retval;
Elong:
	return ERR_PTR(-ENAMETOOLONG);
}

3030 3031
static void get_fs_root_and_pwd_rcu(struct fs_struct *fs, struct path *root,
				    struct path *pwd)
3032
{
3033 3034 3035 3036 3037 3038 3039
	unsigned seq;

	do {
		seq = read_seqcount_begin(&fs->seq);
		*root = fs->root;
		*pwd = fs->pwd;
	} while (read_seqcount_retry(&fs->seq, seq));
3040 3041
}

L
Linus Torvalds 已提交
3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059
/*
 * 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;
 *	}
 */
3060
SYSCALL_DEFINE2(getcwd, char __user *, buf, unsigned long, size)
L
Linus Torvalds 已提交
3061
{
3062
	int error;
J
Jan Blunck 已提交
3063
	struct path pwd, root;
3064
	char *page = __getname();
L
Linus Torvalds 已提交
3065 3066 3067 3068

	if (!page)
		return -ENOMEM;

3069 3070
	rcu_read_lock();
	get_fs_root_and_pwd_rcu(current->fs, &root, &pwd);
L
Linus Torvalds 已提交
3071

3072
	error = -ENOENT;
A
Al Viro 已提交
3073
	br_read_lock(&vfsmount_lock);
3074
	if (!d_unlinked(pwd.dentry)) {
3075
		unsigned long len;
3076 3077
		char *cwd = page + PATH_MAX;
		int buflen = PATH_MAX;
L
Linus Torvalds 已提交
3078

3079
		prepend(&cwd, &buflen, "\0", 1);
3080
		error = prepend_path(&pwd, &root, &cwd, &buflen);
A
Al Viro 已提交
3081
		br_read_unlock(&vfsmount_lock);
3082
		rcu_read_unlock();
3083

3084
		if (error < 0)
3085 3086
			goto out;

3087
		/* Unreachable from current root */
3088
		if (error > 0) {
3089 3090 3091 3092 3093
			error = prepend_unreachable(&cwd, &buflen);
			if (error)
				goto out;
		}

3094
		error = -ERANGE;
3095
		len = PATH_MAX + page - cwd;
3096 3097 3098 3099 3100
		if (len <= size) {
			error = len;
			if (copy_to_user(buf, cwd, len))
				error = -EFAULT;
		}
3101
	} else {
A
Al Viro 已提交
3102
		br_read_unlock(&vfsmount_lock);
3103
		rcu_read_unlock();
3104
	}
L
Linus Torvalds 已提交
3105 3106

out:
3107
	__putname(page);
L
Linus Torvalds 已提交
3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126
	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()
 */
  
3127
int is_subdir(struct dentry *new_dentry, struct dentry *old_dentry)
L
Linus Torvalds 已提交
3128 3129
{
	int result;
3130
	unsigned seq;
L
Linus Torvalds 已提交
3131

3132 3133 3134 3135
	if (new_dentry == old_dentry)
		return 1;

	do {
L
Linus Torvalds 已提交
3136 3137
		/* for restarting inner loop in case of seq retry */
		seq = read_seqbegin(&rename_lock);
3138 3139 3140 3141 3142
		/*
		 * Need rcu_readlock to protect against the d_parent trashing
		 * due to d_move
		 */
		rcu_read_lock();
3143
		if (d_ancestor(old_dentry, new_dentry))
L
Linus Torvalds 已提交
3144
			result = 1;
3145 3146
		else
			result = 0;
3147
		rcu_read_unlock();
L
Linus Torvalds 已提交
3148 3149 3150 3151 3152
	} while (read_seqretry(&rename_lock, seq));

	return result;
}

M
Miklos Szeredi 已提交
3153
static enum d_walk_ret d_genocide_kill(void *data, struct dentry *dentry)
L
Linus Torvalds 已提交
3154
{
M
Miklos Szeredi 已提交
3155 3156 3157 3158
	struct dentry *root = data;
	if (dentry != root) {
		if (d_unhashed(dentry) || !dentry->d_inode)
			return D_WALK_SKIP;
L
Linus Torvalds 已提交
3159

M
Miklos Szeredi 已提交
3160 3161 3162 3163
		if (!(dentry->d_flags & DCACHE_GENOCIDE)) {
			dentry->d_flags |= DCACHE_GENOCIDE;
			dentry->d_lockref.count--;
		}
L
Linus Torvalds 已提交
3164
	}
M
Miklos Szeredi 已提交
3165 3166
	return D_WALK_CONTINUE;
}
3167

M
Miklos Szeredi 已提交
3168 3169 3170
void d_genocide(struct dentry *parent)
{
	d_walk(parent, parent, d_genocide_kill, NULL);
L
Linus Torvalds 已提交
3171 3172
}

3173
void d_tmpfile(struct dentry *dentry, struct inode *inode)
L
Linus Torvalds 已提交
3174
{
3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185
	inode_dec_link_count(inode);
	BUG_ON(dentry->d_name.name != dentry->d_iname ||
		!hlist_unhashed(&dentry->d_alias) ||
		!d_unlinked(dentry));
	spin_lock(&dentry->d_parent->d_lock);
	spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
	dentry->d_name.len = sprintf(dentry->d_iname, "#%llu",
				(unsigned long long)inode->i_ino);
	spin_unlock(&dentry->d_lock);
	spin_unlock(&dentry->d_parent->d_lock);
	d_instantiate(dentry, inode);
L
Linus Torvalds 已提交
3186
}
3187
EXPORT_SYMBOL(d_tmpfile);
L
Linus Torvalds 已提交
3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200

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)
{
3201
	unsigned int loop;
L
Linus Torvalds 已提交
3202 3203 3204 3205 3206 3207 3208 3209 3210

	/* 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",
3211
					sizeof(struct hlist_bl_head),
L
Linus Torvalds 已提交
3212 3213 3214 3215 3216
					dhash_entries,
					13,
					HASH_EARLY,
					&d_hash_shift,
					&d_hash_mask,
3217
					0,
L
Linus Torvalds 已提交
3218 3219
					0);

3220
	for (loop = 0; loop < (1U << d_hash_shift); loop++)
3221
		INIT_HLIST_BL_HEAD(dentry_hashtable + loop);
L
Linus Torvalds 已提交
3222 3223
}

3224
static void __init dcache_init(void)
L
Linus Torvalds 已提交
3225
{
3226
	unsigned int loop;
L
Linus Torvalds 已提交
3227 3228 3229 3230 3231 3232

	/* 
	 * 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. 
	 */
3233 3234
	dentry_cache = KMEM_CACHE(dentry,
		SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|SLAB_MEM_SPREAD);
L
Linus Torvalds 已提交
3235 3236 3237 3238 3239 3240 3241

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

	dentry_hashtable =
		alloc_large_system_hash("Dentry cache",
3242
					sizeof(struct hlist_bl_head),
L
Linus Torvalds 已提交
3243 3244 3245 3246 3247
					dhash_entries,
					13,
					0,
					&d_hash_shift,
					&d_hash_mask,
3248
					0,
L
Linus Torvalds 已提交
3249 3250
					0);

3251
	for (loop = 0; loop < (1U << d_hash_shift); loop++)
3252
		INIT_HLIST_BL_HEAD(dentry_hashtable + loop);
L
Linus Torvalds 已提交
3253 3254 3255
}

/* SLAB cache for __getname() consumers */
3256
struct kmem_cache *names_cachep __read_mostly;
3257
EXPORT_SYMBOL(names_cachep);
L
Linus Torvalds 已提交
3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277

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,
3278
			SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
L
Linus Torvalds 已提交
3279

3280 3281
	dcache_init();
	inode_init();
L
Linus Torvalds 已提交
3282
	files_init(mempages);
3283
	mnt_init();
L
Linus Torvalds 已提交
3284 3285 3286
	bdev_cache_init();
	chrdev_init();
}