dcache.c 78.2 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
/*
 * fs/dcache.c
 *
 * Complete reimplementation
 * (C) 1997 Thomas Schoebel-Theuer,
 * with heavy changes by Linus Torvalds
 */

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

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

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

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

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

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

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

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

105
static struct hlist_bl_head *dentry_hashtable __read_mostly;
106

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

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

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

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

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

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

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

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

177
#else
178

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

191 192
#endif

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

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

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

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

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

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

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

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

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

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

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

340
static void dentry_lru_move_list(struct dentry *dentry, struct list_head *list)
341
{
N
Nick Piggin 已提交
342
	spin_lock(&dcache_lru_lock);
343
	if (list_empty(&dentry->d_lru)) {
344
		list_add_tail(&dentry->d_lru, list);
345
		dentry->d_sb->s_nr_dentry_unused++;
346
		dentry_stat.nr_unused++;
347
	} else {
348
		list_move_tail(&dentry->d_lru, list);
349
	}
N
Nick Piggin 已提交
350
	spin_unlock(&dcache_lru_lock);
351 352
}

353 354 355
/**
 * d_kill - kill dentry and return parent
 * @dentry: dentry to kill
356
 * @parent: parent dentry
357
 *
358
 * The dentry must already be unhashed and removed from the LRU.
359 360
 *
 * If this is the root of the dentry tree, return NULL.
N
Nick Piggin 已提交
361
 *
N
Nick Piggin 已提交
362 363
 * dentry->d_lock and parent->d_lock must be held by caller, and are dropped by
 * d_kill.
364
 */
N
Nick Piggin 已提交
365
static struct dentry *d_kill(struct dentry *dentry, struct dentry *parent)
366
	__releases(dentry->d_lock)
N
Nick Piggin 已提交
367
	__releases(parent->d_lock)
368
	__releases(dentry->d_inode->i_lock)
369 370
{
	list_del(&dentry->d_u.d_child);
371 372 373 374
	/*
	 * Inform try_to_ascend() that we are no longer attached to the
	 * dentry tree
	 */
375
	dentry->d_flags |= DCACHE_DENTRY_KILLED;
N
Nick Piggin 已提交
376 377
	if (parent)
		spin_unlock(&parent->d_lock);
378
	dentry_iput(dentry);
N
Nick Piggin 已提交
379 380 381 382
	/*
	 * dentry_iput drops the locks, at which point nobody (except
	 * transient RCU lookups) can reach this dentry.
	 */
383
	d_free(dentry);
384
	return parent;
385 386
}

387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407
/*
 * 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
408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424
/**
 * 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)
{
425
	if (!d_unhashed(dentry)) {
426
		__d_shrink(dentry);
427
		dentry_rcuwalk_barrier(dentry);
N
Nick Piggin 已提交
428 429 430 431 432 433 434 435 436 437 438 439
	}
}
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);

440 441 442 443 444 445 446 447 448
/*
 * Finish off a dentry we've decided to kill.
 * dentry->d_lock must be held, returns with it unlocked.
 * If ref is non-zero, then decrement the refcount too.
 * Returns dentry requiring refcount drop, or NULL if we're done.
 */
static inline struct dentry *dentry_kill(struct dentry *dentry, int ref)
	__releases(dentry->d_lock)
{
449
	struct inode *inode;
450 451
	struct dentry *parent;

452 453
	inode = dentry->d_inode;
	if (inode && !spin_trylock(&inode->i_lock)) {
454 455 456 457 458 459 460 461 462 463
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)) {
464 465
		if (inode)
			spin_unlock(&inode->i_lock);
466 467
		goto relock;
	}
N
Nick Piggin 已提交
468

469 470
	if (ref)
		dentry->d_count--;
S
Sage Weil 已提交
471 472 473 474
	/*
	 * inform the fs via d_prune that this dentry is about to be
	 * unhashed and destroyed.
	 */
Y
Yan, Zheng 已提交
475 476 477 478
	if (dentry->d_flags & DCACHE_OP_PRUNE)
		dentry->d_op->d_prune(dentry);

	dentry_lru_del(dentry);
479 480 481 482 483
	/* if it was on the hash then remove it */
	__d_drop(dentry);
	return d_kill(dentry, parent);
}

L
Linus Torvalds 已提交
484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515
/* 
 * This is dput
 *
 * This is complicated by the fact that we do not want to put
 * dentries that are no longer on any hash chain on the unused
 * list: we'd much rather just get rid of them immediately.
 *
 * However, that implies that we have to traverse the dentry
 * tree upwards to the parents which might _also_ now be
 * scheduled for deletion (it may have been only waiting for
 * its last child to go away).
 *
 * This tail recursion is done by hand as we don't want to depend
 * on the compiler to always get this right (gcc generally doesn't).
 * Real recursion would eat up our stack space.
 */

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

repeat:
N
Nick Piggin 已提交
516
	if (dentry->d_count == 1)
L
Linus Torvalds 已提交
517 518
		might_sleep();
	spin_lock(&dentry->d_lock);
N
Nick Piggin 已提交
519 520 521
	BUG_ON(!dentry->d_count);
	if (dentry->d_count > 1) {
		dentry->d_count--;
L
Linus Torvalds 已提交
522 523 524 525
		spin_unlock(&dentry->d_lock);
		return;
	}

526
	if (dentry->d_flags & DCACHE_OP_DELETE) {
L
Linus Torvalds 已提交
527
		if (dentry->d_op->d_delete(dentry))
N
Nick Piggin 已提交
528
			goto kill_it;
L
Linus Torvalds 已提交
529
	}
N
Nick Piggin 已提交
530

L
Linus Torvalds 已提交
531 532 533
	/* Unreachable? Get rid of it */
 	if (d_unhashed(dentry))
		goto kill_it;
N
Nick Piggin 已提交
534

J
Jeff Layton 已提交
535
	dentry->d_flags |= DCACHE_REFERENCED;
536
	dentry_lru_add(dentry);
N
Nick Piggin 已提交
537

N
Nick Piggin 已提交
538 539
	dentry->d_count--;
	spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
540 541
	return;

542
kill_it:
543
	dentry = dentry_kill(dentry, 1);
544 545
	if (dentry)
		goto repeat;
L
Linus Torvalds 已提交
546
}
547
EXPORT_SYMBOL(dput);
L
Linus Torvalds 已提交
548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565

/**
 * 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
566
	spin_lock(&dentry->d_lock);
L
Linus Torvalds 已提交
567
	if (d_unhashed(dentry)) {
N
Nick Piggin 已提交
568
		spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
569 570 571 572 573 574 575
		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
576
		spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
577
		shrink_dcache_parent(dentry);
N
Nick Piggin 已提交
578
		spin_lock(&dentry->d_lock);
L
Linus Torvalds 已提交
579 580 581 582 583 584 585 586 587 588 589
	}

	/*
	 * 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).
590 591
	 * We also need to leave mountpoints alone,
	 * directory or not.
L
Linus Torvalds 已提交
592
	 */
593 594
	if (dentry->d_count > 1 && dentry->d_inode) {
		if (S_ISDIR(dentry->d_inode->i_mode) || d_mountpoint(dentry)) {
L
Linus Torvalds 已提交
595 596 597 598 599 600 601 602 603
			spin_unlock(&dentry->d_lock);
			return -EBUSY;
		}
	}

	__d_drop(dentry);
	spin_unlock(&dentry->d_lock);
	return 0;
}
604
EXPORT_SYMBOL(d_invalidate);
L
Linus Torvalds 已提交
605

N
Nick Piggin 已提交
606
/* This must be called with d_lock held */
607
static inline void __dget_dlock(struct dentry *dentry)
N
Nick Piggin 已提交
608
{
N
Nick Piggin 已提交
609
	dentry->d_count++;
N
Nick Piggin 已提交
610 611
}

612
static inline void __dget(struct dentry *dentry)
L
Linus Torvalds 已提交
613
{
N
Nick Piggin 已提交
614
	spin_lock(&dentry->d_lock);
615
	__dget_dlock(dentry);
N
Nick Piggin 已提交
616
	spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
617 618
}

N
Nick Piggin 已提交
619 620 621 622 623
struct dentry *dget_parent(struct dentry *dentry)
{
	struct dentry *ret;

repeat:
N
Nick Piggin 已提交
624 625 626 627 628
	/*
	 * Don't need rcu_dereference because we re-check it was correct under
	 * the lock.
	 */
	rcu_read_lock();
N
Nick Piggin 已提交
629
	ret = dentry->d_parent;
N
Nick Piggin 已提交
630 631 632 633
	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
634 635
		goto repeat;
	}
N
Nick Piggin 已提交
636
	rcu_read_unlock();
N
Nick Piggin 已提交
637 638 639 640 641 642 643
	BUG_ON(!ret->d_count);
	ret->d_count++;
	spin_unlock(&ret->d_lock);
	return ret;
}
EXPORT_SYMBOL(dget_parent);

L
Linus Torvalds 已提交
644 645 646
/**
 * d_find_alias - grab a hashed alias of inode
 * @inode: inode in question
647 648
 * @want_discon:  flag, used by d_splice_alias, to request
 *          that only a DISCONNECTED alias be returned.
L
Linus Torvalds 已提交
649 650 651 652 653 654 655
 *
 * 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.
 *
656
 * If the inode has an IS_ROOT, DCACHE_DISCONNECTED alias, then prefer
657 658
 * 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
659
 */
660
static struct dentry *__d_find_alias(struct inode *inode, int want_discon)
L
Linus Torvalds 已提交
661
{
N
Nick Piggin 已提交
662
	struct dentry *alias, *discon_alias;
L
Linus Torvalds 已提交
663

N
Nick Piggin 已提交
664 665
again:
	discon_alias = NULL;
666
	hlist_for_each_entry(alias, &inode->i_dentry, d_alias) {
N
Nick Piggin 已提交
667
		spin_lock(&alias->d_lock);
L
Linus Torvalds 已提交
668
 		if (S_ISDIR(inode->i_mode) || !d_unhashed(alias)) {
669
			if (IS_ROOT(alias) &&
N
Nick Piggin 已提交
670
			    (alias->d_flags & DCACHE_DISCONNECTED)) {
L
Linus Torvalds 已提交
671
				discon_alias = alias;
672
			} else if (!want_discon) {
673
				__dget_dlock(alias);
N
Nick Piggin 已提交
674 675 676 677 678 679 680 681 682 683 684 685
				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)) {
686
				__dget_dlock(alias);
N
Nick Piggin 已提交
687
				spin_unlock(&alias->d_lock);
L
Linus Torvalds 已提交
688 689 690
				return alias;
			}
		}
N
Nick Piggin 已提交
691 692
		spin_unlock(&alias->d_lock);
		goto again;
L
Linus Torvalds 已提交
693
	}
N
Nick Piggin 已提交
694
	return NULL;
L
Linus Torvalds 已提交
695 696
}

N
Nick Piggin 已提交
697
struct dentry *d_find_alias(struct inode *inode)
L
Linus Torvalds 已提交
698
{
D
David Howells 已提交
699 700
	struct dentry *de = NULL;

A
Al Viro 已提交
701
	if (!hlist_empty(&inode->i_dentry)) {
702
		spin_lock(&inode->i_lock);
703
		de = __d_find_alias(inode, 0);
704
		spin_unlock(&inode->i_lock);
D
David Howells 已提交
705
	}
L
Linus Torvalds 已提交
706 707
	return de;
}
708
EXPORT_SYMBOL(d_find_alias);
L
Linus Torvalds 已提交
709 710 711 712 713 714 715

/*
 *	Try to kill dentries associated with this inode.
 * WARNING: you must own a reference to inode.
 */
void d_prune_aliases(struct inode *inode)
{
716
	struct dentry *dentry;
L
Linus Torvalds 已提交
717
restart:
718
	spin_lock(&inode->i_lock);
719
	hlist_for_each_entry(dentry, &inode->i_dentry, d_alias) {
L
Linus Torvalds 已提交
720
		spin_lock(&dentry->d_lock);
N
Nick Piggin 已提交
721
		if (!dentry->d_count) {
722
			__dget_dlock(dentry);
L
Linus Torvalds 已提交
723 724
			__d_drop(dentry);
			spin_unlock(&dentry->d_lock);
725
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
726 727 728 729 730
			dput(dentry);
			goto restart;
		}
		spin_unlock(&dentry->d_lock);
	}
731
	spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
732
}
733
EXPORT_SYMBOL(d_prune_aliases);
L
Linus Torvalds 已提交
734 735

/*
736 737 738
 * 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.
739
 *
740
 * This may fail if locks cannot be acquired no problem, just try again.
L
Linus Torvalds 已提交
741
 */
742
static void try_prune_one_dentry(struct dentry *dentry)
743
	__releases(dentry->d_lock)
L
Linus Torvalds 已提交
744
{
745
	struct dentry *parent;
746

747
	parent = dentry_kill(dentry, 0);
748
	/*
749 750 751 752 753 754 755 756
	 * 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.
757
	 */
758 759 760 761 762 763 764
	if (!parent)
		return;
	if (parent == dentry)
		return;

	/* Prune ancestors. */
	dentry = parent;
765
	while (dentry) {
N
Nick Piggin 已提交
766
		spin_lock(&dentry->d_lock);
767 768 769 770 771
		if (dentry->d_count > 1) {
			dentry->d_count--;
			spin_unlock(&dentry->d_lock);
			return;
		}
772
		dentry = dentry_kill(dentry, 1);
773
	}
L
Linus Torvalds 已提交
774 775
}

C
Christoph Hellwig 已提交
776
static void shrink_dentry_list(struct list_head *list)
L
Linus Torvalds 已提交
777
{
778 779
	struct dentry *dentry;

780 781 782 783 784 785 786 787
	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
788 789 790
			continue;
		}

L
Linus Torvalds 已提交
791 792
		/*
		 * We found an inuse dentry which was not removed from
793 794
		 * the LRU because of laziness during lookup.  Do not free
		 * it - just keep it off the LRU list.
L
Linus Torvalds 已提交
795
		 */
N
Nick Piggin 已提交
796
		if (dentry->d_count) {
797
			dentry_lru_del(dentry);
798
			spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
799 800
			continue;
		}
801 802

		rcu_read_unlock();
803 804 805

		try_prune_one_dentry(dentry);

806
		rcu_read_lock();
807
	}
808
	rcu_read_unlock();
C
Christoph Hellwig 已提交
809 810 811
}

/**
812 813 814 815 816 817 818
 * 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
819
 *
820 821
 * This function may fail to free any resources if all the dentries are in
 * use.
C
Christoph Hellwig 已提交
822
 */
823
void prune_dcache_sb(struct super_block *sb, int count)
C
Christoph Hellwig 已提交
824 825 826 827 828
{
	struct dentry *dentry;
	LIST_HEAD(referenced);
	LIST_HEAD(tmp);

N
Nick Piggin 已提交
829 830
relock:
	spin_lock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
831 832 833 834 835
	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
836 837 838 839 840 841
		if (!spin_trylock(&dentry->d_lock)) {
			spin_unlock(&dcache_lru_lock);
			cpu_relax();
			goto relock;
		}

842
		if (dentry->d_flags & DCACHE_REFERENCED) {
N
Nick Piggin 已提交
843 844
			dentry->d_flags &= ~DCACHE_REFERENCED;
			list_move(&dentry->d_lru, &referenced);
C
Christoph Hellwig 已提交
845
			spin_unlock(&dentry->d_lock);
N
Nick Piggin 已提交
846 847
		} else {
			list_move_tail(&dentry->d_lru, &tmp);
848
			dentry->d_flags |= DCACHE_SHRINK_LIST;
N
Nick Piggin 已提交
849
			spin_unlock(&dentry->d_lock);
850
			if (!--count)
N
Nick Piggin 已提交
851
				break;
C
Christoph Hellwig 已提交
852
		}
853
		cond_resched_lock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
854
	}
855 856
	if (!list_empty(&referenced))
		list_splice(&referenced, &sb->s_dentry_lru);
N
Nick Piggin 已提交
857
	spin_unlock(&dcache_lru_lock);
858 859

	shrink_dentry_list(&tmp);
860 861
}

L
Linus Torvalds 已提交
862 863 864 865
/**
 * shrink_dcache_sb - shrink dcache for a superblock
 * @sb: superblock
 *
C
Christoph Hellwig 已提交
866 867
 * 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
868
 */
C
Christoph Hellwig 已提交
869
void shrink_dcache_sb(struct super_block *sb)
L
Linus Torvalds 已提交
870
{
C
Christoph Hellwig 已提交
871 872
	LIST_HEAD(tmp);

N
Nick Piggin 已提交
873
	spin_lock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
874 875
	while (!list_empty(&sb->s_dentry_lru)) {
		list_splice_init(&sb->s_dentry_lru, &tmp);
876
		spin_unlock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
877
		shrink_dentry_list(&tmp);
878
		spin_lock(&dcache_lru_lock);
C
Christoph Hellwig 已提交
879
	}
N
Nick Piggin 已提交
880
	spin_unlock(&dcache_lru_lock);
L
Linus Torvalds 已提交
881
}
882
EXPORT_SYMBOL(shrink_dcache_sb);
L
Linus Torvalds 已提交
883

884 885 886 887 888 889 890 891 892 893 894 895 896
/*
 * 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 */
897
		while (!list_empty(&dentry->d_subdirs))
898 899 900 901 902 903 904 905
			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
906
			/*
Y
Yan, Zheng 已提交
907
			 * inform the fs that this dentry is about to be
S
Sage Weil 已提交
908 909
			 * unhashed and destroyed.
			 */
Y
Yan, Zheng 已提交
910 911 912 913
			if (dentry->d_flags & DCACHE_OP_PRUNE)
				dentry->d_op->d_prune(dentry);

			dentry_lru_del(dentry);
914 915
			__d_shrink(dentry);

N
Nick Piggin 已提交
916
			if (dentry->d_count != 0) {
917 918 919 920 921 922 923 924
				printk(KERN_ERR
				       "BUG: Dentry %p{i=%lx,n=%s}"
				       " still in use (%d)"
				       " [unmount of %s %s]\n",
				       dentry,
				       dentry->d_inode ?
				       dentry->d_inode->i_ino : 0UL,
				       dentry->d_name.name,
N
Nick Piggin 已提交
925
				       dentry->d_count,
926 927 928 929 930
				       dentry->d_sb->s_type->name,
				       dentry->d_sb->s_id);
				BUG();
			}

N
Nick Piggin 已提交
931
			if (IS_ROOT(dentry)) {
932
				parent = NULL;
N
Nick Piggin 已提交
933 934
				list_del(&dentry->d_u.d_child);
			} else {
935
				parent = dentry->d_parent;
N
Nick Piggin 已提交
936
				parent->d_count--;
N
Nick Piggin 已提交
937
				list_del(&dentry->d_u.d_child);
938
			}
939 940 941 942

			inode = dentry->d_inode;
			if (inode) {
				dentry->d_inode = NULL;
A
Al Viro 已提交
943
				hlist_del_init(&dentry->d_alias);
944 945 946 947 948 949 950 951 952 953 954 955
				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)
956
				return;
957 958 959 960 961 962 963 964 965 966
			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
967
 * - we don't need to use dentry->d_lock because:
968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983
 *   - the superblock is detached from all mountings and open files, so the
 *     dentry trees will not be rearranged by the VFS
 *   - s_umount is write-locked, so the memory pressure shrinker will ignore
 *     any dentries belonging to this superblock that it comes across
 *   - the filesystem itself is no longer permitted to rearrange the dentries
 *     in this superblock
 */
void shrink_dcache_for_umount(struct super_block *sb)
{
	struct dentry *dentry;

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

	dentry = sb->s_root;
	sb->s_root = NULL;
N
Nick Piggin 已提交
984
	dentry->d_count--;
985 986
	shrink_dcache_for_umount_subtree(dentry);

987 988
	while (!hlist_bl_empty(&sb->s_anon)) {
		dentry = hlist_bl_entry(hlist_bl_first(&sb->s_anon), struct dentry, d_hash);
989 990 991 992
		shrink_dcache_for_umount_subtree(dentry);
	}
}

993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
/*
 * This tries to ascend one level of parenthood, but
 * we can race with renaming, so we need to re-check
 * the parenthood after dropping the lock and check
 * that the sequence number still matches.
 */
static struct dentry *try_to_ascend(struct dentry *old, int locked, unsigned seq)
{
	struct dentry *new = old->d_parent;

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

	/*
	 * might go back up the wrong parent if we have had a rename
	 * or deletion
	 */
	if (new != old->d_parent ||
1012
		 (old->d_flags & DCACHE_DENTRY_KILLED) ||
1013 1014 1015 1016 1017 1018 1019 1020 1021
		 (!locked && read_seqretry(&rename_lock, seq))) {
		spin_unlock(&new->d_lock);
		new = NULL;
	}
	rcu_read_unlock();
	return new;
}


L
Linus Torvalds 已提交
1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036
/*
 * Search for at least 1 mount point in the dentry's subdirs.
 * We descend to the next level whenever the d_subdirs
 * list is non-empty and continue searching.
 */
 
/**
 * have_submounts - check for mounts over a dentry
 * @parent: dentry to check.
 *
 * Return true if the parent or its subdirectories contain
 * a mount point
 */
int have_submounts(struct dentry *parent)
{
1037
	struct dentry *this_parent;
L
Linus Torvalds 已提交
1038
	struct list_head *next;
1039
	unsigned seq;
1040
	int locked = 0;
1041 1042

	seq = read_seqbegin(&rename_lock);
1043 1044
again:
	this_parent = parent;
L
Linus Torvalds 已提交
1045 1046 1047

	if (d_mountpoint(parent))
		goto positive;
N
Nick Piggin 已提交
1048
	spin_lock(&this_parent->d_lock);
L
Linus Torvalds 已提交
1049 1050 1051 1052 1053
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
1054
		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
L
Linus Torvalds 已提交
1055
		next = tmp->next;
N
Nick Piggin 已提交
1056 1057

		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
L
Linus Torvalds 已提交
1058
		/* Have we found a mount point ? */
N
Nick Piggin 已提交
1059 1060 1061
		if (d_mountpoint(dentry)) {
			spin_unlock(&dentry->d_lock);
			spin_unlock(&this_parent->d_lock);
L
Linus Torvalds 已提交
1062
			goto positive;
N
Nick Piggin 已提交
1063
		}
L
Linus Torvalds 已提交
1064
		if (!list_empty(&dentry->d_subdirs)) {
N
Nick Piggin 已提交
1065 1066
			spin_unlock(&this_parent->d_lock);
			spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
L
Linus Torvalds 已提交
1067
			this_parent = dentry;
N
Nick Piggin 已提交
1068
			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
L
Linus Torvalds 已提交
1069 1070
			goto repeat;
		}
N
Nick Piggin 已提交
1071
		spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
1072 1073 1074 1075 1076
	}
	/*
	 * All done at this level ... ascend and resume the search.
	 */
	if (this_parent != parent) {
1077 1078 1079
		struct dentry *child = this_parent;
		this_parent = try_to_ascend(this_parent, locked, seq);
		if (!this_parent)
1080 1081
			goto rename_retry;
		next = child->d_u.d_child.next;
L
Linus Torvalds 已提交
1082 1083
		goto resume;
	}
N
Nick Piggin 已提交
1084
	spin_unlock(&this_parent->d_lock);
1085
	if (!locked && read_seqretry(&rename_lock, seq))
1086
		goto rename_retry;
1087 1088
	if (locked)
		write_sequnlock(&rename_lock);
L
Linus Torvalds 已提交
1089 1090
	return 0; /* No mount points found in tree */
positive:
1091
	if (!locked && read_seqretry(&rename_lock, seq))
1092
		goto rename_retry;
1093 1094
	if (locked)
		write_sequnlock(&rename_lock);
L
Linus Torvalds 已提交
1095
	return 1;
1096 1097

rename_retry:
1098 1099
	if (locked)
		goto again;
1100 1101 1102
	locked = 1;
	write_seqlock(&rename_lock);
	goto again;
L
Linus Torvalds 已提交
1103
}
1104
EXPORT_SYMBOL(have_submounts);
L
Linus Torvalds 已提交
1105 1106

/*
1107
 * Search the dentry child list of the specified parent,
L
Linus Torvalds 已提交
1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119
 * 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.
 */
1120
static int select_parent(struct dentry *parent, struct list_head *dispose)
L
Linus Torvalds 已提交
1121
{
1122
	struct dentry *this_parent;
L
Linus Torvalds 已提交
1123
	struct list_head *next;
1124
	unsigned seq;
L
Linus Torvalds 已提交
1125
	int found = 0;
1126
	int locked = 0;
L
Linus Torvalds 已提交
1127

1128
	seq = read_seqbegin(&rename_lock);
1129 1130
again:
	this_parent = parent;
N
Nick Piggin 已提交
1131
	spin_lock(&this_parent->d_lock);
L
Linus Torvalds 已提交
1132 1133 1134 1135 1136
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
1137
		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
L
Linus Torvalds 已提交
1138 1139
		next = tmp->next;

N
Nick Piggin 已提交
1140
		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
N
Nick Piggin 已提交
1141

1142 1143
		/*
		 * move only zero ref count dentries to the dispose list.
1144 1145 1146 1147 1148
		 *
		 * 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
1149
		 */
1150 1151 1152
		if (dentry->d_count) {
			dentry_lru_del(dentry);
		} else if (!(dentry->d_flags & DCACHE_SHRINK_LIST)) {
1153
			dentry_lru_move_list(dentry, dispose);
1154
			dentry->d_flags |= DCACHE_SHRINK_LIST;
L
Linus Torvalds 已提交
1155 1156 1157 1158 1159 1160 1161
			found++;
		}
		/*
		 * We can return to the caller if we have found some (this
		 * ensures forward progress). We'll be coming back to find
		 * the rest.
		 */
N
Nick Piggin 已提交
1162 1163
		if (found && need_resched()) {
			spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
1164
			goto out;
N
Nick Piggin 已提交
1165
		}
L
Linus Torvalds 已提交
1166 1167 1168 1169 1170

		/*
		 * Descend a level if the d_subdirs list is non-empty.
		 */
		if (!list_empty(&dentry->d_subdirs)) {
N
Nick Piggin 已提交
1171 1172
			spin_unlock(&this_parent->d_lock);
			spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
L
Linus Torvalds 已提交
1173
			this_parent = dentry;
N
Nick Piggin 已提交
1174
			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
L
Linus Torvalds 已提交
1175 1176
			goto repeat;
		}
N
Nick Piggin 已提交
1177 1178

		spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
1179 1180 1181 1182 1183
	}
	/*
	 * All done at this level ... ascend and resume the search.
	 */
	if (this_parent != parent) {
1184 1185 1186
		struct dentry *child = this_parent;
		this_parent = try_to_ascend(this_parent, locked, seq);
		if (!this_parent)
1187 1188
			goto rename_retry;
		next = child->d_u.d_child.next;
L
Linus Torvalds 已提交
1189 1190 1191
		goto resume;
	}
out:
N
Nick Piggin 已提交
1192
	spin_unlock(&this_parent->d_lock);
1193
	if (!locked && read_seqretry(&rename_lock, seq))
1194
		goto rename_retry;
1195 1196
	if (locked)
		write_sequnlock(&rename_lock);
L
Linus Torvalds 已提交
1197
	return found;
1198 1199 1200 1201

rename_retry:
	if (found)
		return found;
1202 1203
	if (locked)
		goto again;
1204 1205 1206
	locked = 1;
	write_seqlock(&rename_lock);
	goto again;
L
Linus Torvalds 已提交
1207 1208 1209 1210 1211 1212 1213 1214 1215 1216
}

/**
 * shrink_dcache_parent - prune dcache
 * @parent: parent of entries to prune
 *
 * Prune the dcache to remove unused children of the parent dentry.
 */
void shrink_dcache_parent(struct dentry * parent)
{
1217
	LIST_HEAD(dispose);
L
Linus Torvalds 已提交
1218 1219
	int found;

1220
	while ((found = select_parent(parent, &dispose)) != 0) {
1221
		shrink_dentry_list(&dispose);
1222 1223
		cond_resched();
	}
L
Linus Torvalds 已提交
1224
}
1225
EXPORT_SYMBOL(shrink_dcache_parent);
L
Linus Torvalds 已提交
1226 1227

/**
1228 1229
 * __d_alloc	-	allocate a dcache entry
 * @sb: filesystem it will belong to
L
Linus Torvalds 已提交
1230 1231 1232 1233 1234 1235 1236
 * @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.
 */
 
1237
struct dentry *__d_alloc(struct super_block *sb, const struct qstr *name)
L
Linus Torvalds 已提交
1238 1239 1240 1241
{
	struct dentry *dentry;
	char *dname;

1242
	dentry = kmem_cache_alloc(dentry_cache, GFP_KERNEL);
L
Linus Torvalds 已提交
1243 1244 1245
	if (!dentry)
		return NULL;

1246 1247 1248 1249 1250 1251 1252
	/*
	 * 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
1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267
	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;

1268 1269 1270 1271
	/* Make sure we always see the terminating NUL character */
	smp_wmb();
	dentry->d_name.name = dname;

N
Nick Piggin 已提交
1272
	dentry->d_count = 1;
1273
	dentry->d_flags = 0;
L
Linus Torvalds 已提交
1274
	spin_lock_init(&dentry->d_lock);
N
Nick Piggin 已提交
1275
	seqcount_init(&dentry->d_seq);
L
Linus Torvalds 已提交
1276
	dentry->d_inode = NULL;
1277 1278
	dentry->d_parent = dentry;
	dentry->d_sb = sb;
L
Linus Torvalds 已提交
1279 1280
	dentry->d_op = NULL;
	dentry->d_fsdata = NULL;
1281
	INIT_HLIST_BL_NODE(&dentry->d_hash);
L
Linus Torvalds 已提交
1282 1283
	INIT_LIST_HEAD(&dentry->d_lru);
	INIT_LIST_HEAD(&dentry->d_subdirs);
A
Al Viro 已提交
1284
	INIT_HLIST_NODE(&dentry->d_alias);
N
Nick Piggin 已提交
1285
	INIT_LIST_HEAD(&dentry->d_u.d_child);
1286
	d_set_d_op(dentry, dentry->d_sb->s_d_op);
L
Linus Torvalds 已提交
1287

1288
	this_cpu_inc(nr_dentry);
1289

L
Linus Torvalds 已提交
1290 1291
	return dentry;
}
1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319

/**
 * 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;
}
1320
EXPORT_SYMBOL(d_alloc);
L
Linus Torvalds 已提交
1321

1322 1323
struct dentry *d_alloc_pseudo(struct super_block *sb, const struct qstr *name)
{
1324 1325
	struct dentry *dentry = __d_alloc(sb, name);
	if (dentry)
1326 1327 1328 1329 1330
		dentry->d_flags |= DCACHE_DISCONNECTED;
	return dentry;
}
EXPORT_SYMBOL(d_alloc_pseudo);

L
Linus Torvalds 已提交
1331 1332 1333 1334 1335 1336 1337 1338 1339
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);
}
1340
EXPORT_SYMBOL(d_alloc_name);
L
Linus Torvalds 已提交
1341

1342 1343
void d_set_d_op(struct dentry *dentry, const struct dentry_operations *op)
{
1344 1345
	WARN_ON_ONCE(dentry->d_op);
	WARN_ON_ONCE(dentry->d_flags & (DCACHE_OP_HASH	|
1346 1347
				DCACHE_OP_COMPARE	|
				DCACHE_OP_REVALIDATE	|
1348
				DCACHE_OP_WEAK_REVALIDATE	|
1349 1350 1351 1352 1353 1354 1355 1356 1357 1358
				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;
1359 1360
	if (op->d_weak_revalidate)
		dentry->d_flags |= DCACHE_OP_WEAK_REVALIDATE;
1361 1362
	if (op->d_delete)
		dentry->d_flags |= DCACHE_OP_DELETE;
S
Sage Weil 已提交
1363 1364
	if (op->d_prune)
		dentry->d_flags |= DCACHE_OP_PRUNE;
1365 1366 1367 1368

}
EXPORT_SYMBOL(d_set_d_op);

1369 1370
static void __d_instantiate(struct dentry *dentry, struct inode *inode)
{
N
Nick Piggin 已提交
1371
	spin_lock(&dentry->d_lock);
1372 1373 1374
	if (inode) {
		if (unlikely(IS_AUTOMOUNT(inode)))
			dentry->d_flags |= DCACHE_NEED_AUTOMOUNT;
A
Al Viro 已提交
1375
		hlist_add_head(&dentry->d_alias, &inode->i_dentry);
1376
	}
1377
	dentry->d_inode = inode;
N
Nick Piggin 已提交
1378
	dentry_rcuwalk_barrier(dentry);
N
Nick Piggin 已提交
1379
	spin_unlock(&dentry->d_lock);
1380 1381 1382
	fsnotify_d_instantiate(dentry, inode);
}

L
Linus Torvalds 已提交
1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399
/**
 * 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
1400
	BUG_ON(!hlist_unhashed(&entry->d_alias));
1401 1402
	if (inode)
		spin_lock(&inode->i_lock);
1403
	__d_instantiate(entry, inode);
1404 1405
	if (inode)
		spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1406 1407
	security_d_instantiate(entry, inode);
}
1408
EXPORT_SYMBOL(d_instantiate);
L
Linus Torvalds 已提交
1409 1410 1411 1412 1413 1414 1415 1416

/**
 * 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
1417
 * aliased dentry instead and drop one reference to inode.
L
Linus Torvalds 已提交
1418 1419 1420
 *
 * Note that in order to avoid conflicts with rename() etc, the caller
 * had better be holding the parent directory semaphore.
1421 1422 1423 1424
 *
 * 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
1425
 */
1426 1427
static struct dentry *__d_instantiate_unique(struct dentry *entry,
					     struct inode *inode)
L
Linus Torvalds 已提交
1428 1429 1430 1431 1432 1433
{
	struct dentry *alias;
	int len = entry->d_name.len;
	const char *name = entry->d_name.name;
	unsigned int hash = entry->d_name.hash;

1434
	if (!inode) {
1435
		__d_instantiate(entry, NULL);
1436 1437 1438
		return NULL;
	}

1439
	hlist_for_each_entry(alias, &inode->i_dentry, d_alias) {
N
Nick Piggin 已提交
1440 1441 1442 1443 1444
		/*
		 * 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.
		 */
1445
		if (alias->d_name.hash != hash)
L
Linus Torvalds 已提交
1446 1447 1448
			continue;
		if (alias->d_parent != entry->d_parent)
			continue;
1449 1450
		if (alias->d_name.len != len)
			continue;
1451
		if (dentry_cmp(alias, name, len))
L
Linus Torvalds 已提交
1452
			continue;
1453
		__dget(alias);
L
Linus Torvalds 已提交
1454 1455
		return alias;
	}
1456

1457
	__d_instantiate(entry, inode);
L
Linus Torvalds 已提交
1458 1459
	return NULL;
}
1460 1461 1462 1463 1464

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

A
Al Viro 已提交
1465
	BUG_ON(!hlist_unhashed(&entry->d_alias));
1466

1467 1468
	if (inode)
		spin_lock(&inode->i_lock);
1469
	result = __d_instantiate_unique(entry, inode);
1470 1471
	if (inode)
		spin_unlock(&inode->i_lock);
1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482

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

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

L
Linus Torvalds 已提交
1483 1484
EXPORT_SYMBOL(d_instantiate_unique);

A
Al Viro 已提交
1485 1486 1487 1488 1489
struct dentry *d_make_root(struct inode *root_inode)
{
	struct dentry *res = NULL;

	if (root_inode) {
1490
		static const struct qstr name = QSTR_INIT("/", 1);
A
Al Viro 已提交
1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501

		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);

1502 1503 1504 1505
static struct dentry * __d_find_any_alias(struct inode *inode)
{
	struct dentry *alias;

A
Al Viro 已提交
1506
	if (hlist_empty(&inode->i_dentry))
1507
		return NULL;
A
Al Viro 已提交
1508
	alias = hlist_entry(inode->i_dentry.first, struct dentry, d_alias);
1509 1510 1511 1512
	__dget(alias);
	return alias;
}

1513 1514 1515 1516 1517 1518 1519 1520
/**
 * 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)
1521 1522 1523 1524 1525 1526 1527 1528
{
	struct dentry *de;

	spin_lock(&inode->i_lock);
	de = __d_find_any_alias(inode);
	spin_unlock(&inode->i_lock);
	return de;
}
1529
EXPORT_SYMBOL(d_find_any_alias);
1530

1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543
/**
 * 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
1544 1545 1546 1547
 * 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).
1548 1549 1550
 */
struct dentry *d_obtain_alias(struct inode *inode)
{
1551
	static const struct qstr anonstring = QSTR_INIT("/", 1);
C
Christoph Hellwig 已提交
1552 1553
	struct dentry *tmp;
	struct dentry *res;
1554 1555

	if (!inode)
1556
		return ERR_PTR(-ESTALE);
1557 1558 1559
	if (IS_ERR(inode))
		return ERR_CAST(inode);

1560
	res = d_find_any_alias(inode);
C
Christoph Hellwig 已提交
1561 1562 1563
	if (res)
		goto out_iput;

1564
	tmp = __d_alloc(inode->i_sb, &anonstring);
C
Christoph Hellwig 已提交
1565 1566 1567
	if (!tmp) {
		res = ERR_PTR(-ENOMEM);
		goto out_iput;
1568
	}
N
Nick Piggin 已提交
1569

1570
	spin_lock(&inode->i_lock);
1571
	res = __d_find_any_alias(inode);
C
Christoph Hellwig 已提交
1572
	if (res) {
1573
		spin_unlock(&inode->i_lock);
C
Christoph Hellwig 已提交
1574 1575 1576 1577 1578 1579 1580 1581
		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
1582
	hlist_add_head(&tmp->d_alias, &inode->i_dentry);
1583
	hlist_bl_lock(&tmp->d_sb->s_anon);
1584
	hlist_bl_add_head(&tmp->d_hash, &tmp->d_sb->s_anon);
1585
	hlist_bl_unlock(&tmp->d_sb->s_anon);
C
Christoph Hellwig 已提交
1586
	spin_unlock(&tmp->d_lock);
1587
	spin_unlock(&inode->i_lock);
1588
	security_d_instantiate(tmp, inode);
C
Christoph Hellwig 已提交
1589 1590 1591 1592

	return tmp;

 out_iput:
1593 1594
	if (res && !IS_ERR(res))
		security_d_instantiate(res, inode);
C
Christoph Hellwig 已提交
1595 1596
	iput(inode);
	return res;
1597
}
1598
EXPORT_SYMBOL(d_obtain_alias);
L
Linus Torvalds 已提交
1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614

/**
 * 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.
 *
1615 1616 1617 1618
 * 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
1619 1620 1621 1622 1623
 */
struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
{
	struct dentry *new = NULL;

1624 1625 1626
	if (IS_ERR(inode))
		return ERR_CAST(inode);

1627
	if (inode && S_ISDIR(inode->i_mode)) {
1628
		spin_lock(&inode->i_lock);
1629
		new = __d_find_alias(inode, 1);
L
Linus Torvalds 已提交
1630
		if (new) {
1631
			BUG_ON(!(new->d_flags & DCACHE_DISCONNECTED));
1632
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1633 1634 1635 1636
			security_d_instantiate(new, inode);
			d_move(new, dentry);
			iput(inode);
		} else {
1637
			/* already taking inode->i_lock, so d_add() by hand */
1638
			__d_instantiate(dentry, inode);
1639
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1640 1641 1642
			security_d_instantiate(dentry, inode);
			d_rehash(dentry);
		}
1643 1644 1645 1646 1647
	} else {
		d_instantiate(dentry, inode);
		if (d_unhashed(dentry))
			d_rehash(dentry);
	}
L
Linus Torvalds 已提交
1648 1649
	return new;
}
1650
EXPORT_SYMBOL(d_splice_alias);
L
Linus Torvalds 已提交
1651

1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667
/**
 * 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.
 */
1668
struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
1669 1670 1671 1672 1673
			struct qstr *name)
{
	struct dentry *found;
	struct dentry *new;

C
Christoph Hellwig 已提交
1674 1675 1676 1677
	/*
	 * First check if a dentry matching the name already exists,
	 * if not go ahead and create it now.
	 */
1678
	found = d_hash_and_lookup(dentry->d_parent, name);
1679 1680
	if (unlikely(IS_ERR(found)))
		goto err_out;
1681 1682 1683
	if (!found) {
		new = d_alloc(dentry->d_parent, name);
		if (!new) {
1684
			found = ERR_PTR(-ENOMEM);
1685 1686
			goto err_out;
		}
C
Christoph Hellwig 已提交
1687

1688 1689 1690 1691 1692 1693 1694
		found = d_splice_alias(inode, new);
		if (found) {
			dput(new);
			return found;
		}
		return new;
	}
C
Christoph Hellwig 已提交
1695 1696 1697 1698 1699 1700 1701

	/*
	 * If a matching dentry exists, and it's not negative use it.
	 *
	 * Decrement the reference count to balance the iget() done
	 * earlier on.
	 */
1702 1703 1704 1705 1706 1707 1708 1709 1710
	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
1711

1712 1713
	/*
	 * Negative dentry: instantiate it unless the inode is a directory and
C
Christoph Hellwig 已提交
1714
	 * already has a dentry.
1715
	 */
1716 1717 1718 1719
	new = d_splice_alias(inode, found);
	if (new) {
		dput(found);
		found = new;
1720
	}
1721
	return found;
1722 1723 1724

err_out:
	iput(inode);
1725
	return found;
1726
}
1727
EXPORT_SYMBOL(d_add_ci);
L
Linus Torvalds 已提交
1728

1729 1730 1731 1732
/*
 * Do the slow-case of the dentry name compare.
 *
 * Unlike the dentry_cmp() function, we need to atomically
1733
 * load the name and length information, so that the
1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761
 * 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;
	}
1762
	if (parent->d_op->d_compare(parent, dentry, tlen, tname, name))
1763 1764 1765 1766
		return D_COMP_NOMATCH;
	return D_COMP_OK;
}

N
Nick Piggin 已提交
1767 1768 1769 1770
/**
 * __d_lookup_rcu - search for a dentry (racy, store-free)
 * @parent: parent dentry
 * @name: qstr of name we wish to find
1771
 * @seqp: returns d_seq value at the point where the dentry was found
N
Nick Piggin 已提交
1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791
 * Returns: dentry, or NULL
 *
 * __d_lookup_rcu is the dcache lookup function for rcu-walk name
 * resolution (store-free path walking) design described in
 * Documentation/filesystems/path-lookup.txt.
 *
 * This is not to be used outside core vfs.
 *
 * __d_lookup_rcu must only be used in rcu-walk mode, ie. with vfsmount lock
 * held, and rcu_read_lock held. The returned dentry must not be stored into
 * without taking d_lock and checking d_seq sequence count against @seq
 * returned here.
 *
 * A refcount may be taken on the found dentry with the __d_rcu_to_refcount
 * function.
 *
 * Alternatively, __d_lookup_rcu may be called again to look up the child of
 * the returned dentry, so long as its parent's seqlock is checked after the
 * child is looked up. Thus, an interlocking stepping of sequence lock checks
 * is formed, giving integrity down the path walk.
1792 1793 1794
 *
 * 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
1795
 */
1796 1797
struct dentry *__d_lookup_rcu(const struct dentry *parent,
				const struct qstr *name,
1798
				unsigned *seqp)
N
Nick Piggin 已提交
1799
{
1800
	u64 hashlen = name->hash_len;
N
Nick Piggin 已提交
1801
	const unsigned char *str = name->name;
1802
	struct hlist_bl_head *b = d_hash(parent, hashlen_hash(hashlen));
1803
	struct hlist_bl_node *node;
N
Nick Piggin 已提交
1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823
	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.
	 *
1824
	 * See Documentation/filesystems/path-lookup.txt for more details.
N
Nick Piggin 已提交
1825
	 */
1826
	hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
1827
		unsigned seq;
N
Nick Piggin 已提交
1828 1829

seqretry:
1830 1831
		/*
		 * The dentry sequence count protects us from concurrent
1832
		 * renames, and thus protects parent and name fields.
1833 1834
		 *
		 * The caller must perform a seqcount check in order
1835
		 * to do anything useful with the returned dentry.
1836 1837 1838 1839 1840 1841 1842 1843 1844
		 *
		 * 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
1845 1846
		if (dentry->d_parent != parent)
			continue;
1847 1848
		if (d_unhashed(dentry))
			continue;
1849

1850
		if (unlikely(parent->d_flags & DCACHE_OP_COMPARE)) {
1851 1852
			if (dentry->d_name.hash != hashlen_hash(hashlen))
				continue;
1853 1854
			*seqp = seq;
			switch (slow_dentry_cmp(parent, dentry, seq, name)) {
1855 1856 1857
			case D_COMP_OK:
				return dentry;
			case D_COMP_NOMATCH:
N
Nick Piggin 已提交
1858
				continue;
1859 1860 1861
			default:
				goto seqretry;
			}
N
Nick Piggin 已提交
1862
		}
1863

1864
		if (dentry->d_name.hash_len != hashlen)
1865
			continue;
1866
		*seqp = seq;
1867
		if (!dentry_cmp(dentry, str, hashlen_len(hashlen)))
1868
			return dentry;
N
Nick Piggin 已提交
1869 1870 1871 1872
	}
	return NULL;
}

L
Linus Torvalds 已提交
1873 1874 1875 1876
/**
 * d_lookup - search for a dentry
 * @parent: parent dentry
 * @name: qstr of name we wish to find
1877
 * Returns: dentry, or NULL
L
Linus Torvalds 已提交
1878
 *
1879 1880 1881 1882
 * 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
1883
 */
A
Al Viro 已提交
1884
struct dentry *d_lookup(const struct dentry *parent, const struct qstr *name)
L
Linus Torvalds 已提交
1885
{
N
Nick Piggin 已提交
1886
	struct dentry *dentry;
1887
	unsigned seq;
L
Linus Torvalds 已提交
1888 1889 1890 1891 1892 1893 1894 1895 1896

        do {
                seq = read_seqbegin(&rename_lock);
                dentry = __d_lookup(parent, name);
                if (dentry)
			break;
	} while (read_seqretry(&rename_lock, seq));
	return dentry;
}
1897
EXPORT_SYMBOL(d_lookup);
L
Linus Torvalds 已提交
1898

N
Nick Piggin 已提交
1899
/**
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913
 * __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
1914
struct dentry *__d_lookup(const struct dentry *parent, const struct qstr *name)
L
Linus Torvalds 已提交
1915 1916 1917 1918
{
	unsigned int len = name->len;
	unsigned int hash = name->hash;
	const unsigned char *str = name->name;
1919
	struct hlist_bl_head *b = d_hash(parent, hash);
1920
	struct hlist_bl_node *node;
N
Nick Piggin 已提交
1921
	struct dentry *found = NULL;
1922
	struct dentry *dentry;
L
Linus Torvalds 已提交
1923

N
Nick Piggin 已提交
1924 1925 1926 1927 1928 1929 1930
	/*
	 * 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.
	 */

1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941
	/*
	 * 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.
	 *
1942
	 * See Documentation/filesystems/path-lookup.txt for more details.
1943
	 */
L
Linus Torvalds 已提交
1944 1945
	rcu_read_lock();
	
1946
	hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
L
Linus Torvalds 已提交
1947 1948 1949 1950 1951 1952 1953

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

		spin_lock(&dentry->d_lock);
		if (dentry->d_parent != parent)
			goto next;
1954 1955 1956
		if (d_unhashed(dentry))
			goto next;

L
Linus Torvalds 已提交
1957 1958 1959 1960
		/*
		 * It is safe to compare names since d_move() cannot
		 * change the qstr (protected by d_lock).
		 */
1961
		if (parent->d_flags & DCACHE_OP_COMPARE) {
1962 1963
			int tlen = dentry->d_name.len;
			const char *tname = dentry->d_name.name;
1964
			if (parent->d_op->d_compare(parent, dentry, tlen, tname, name))
L
Linus Torvalds 已提交
1965 1966
				goto next;
		} else {
1967 1968
			if (dentry->d_name.len != len)
				goto next;
1969
			if (dentry_cmp(dentry, str, len))
L
Linus Torvalds 已提交
1970 1971 1972
				goto next;
		}

N
Nick Piggin 已提交
1973
		dentry->d_count++;
1974
		found = dentry;
L
Linus Torvalds 已提交
1975 1976 1977 1978 1979 1980 1981 1982 1983 1984
		spin_unlock(&dentry->d_lock);
		break;
next:
		spin_unlock(&dentry->d_lock);
 	}
 	rcu_read_unlock();

 	return found;
}

1985 1986 1987 1988 1989
/**
 * 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
 *
1990
 * On lookup failure NULL is returned; on bad name - ERR_PTR(-error)
1991 1992 1993 1994 1995 1996 1997 1998 1999
 */
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);
2000
	if (dir->d_flags & DCACHE_OP_HASH) {
2001
		int err = dir->d_op->d_hash(dir, name);
2002 2003
		if (unlikely(err < 0))
			return ERR_PTR(err);
2004
	}
2005
	return d_lookup(dir, name);
2006
}
2007
EXPORT_SYMBOL(d_hash_and_lookup);
2008

L
Linus Torvalds 已提交
2009
/**
N
Nick Piggin 已提交
2010
 * d_validate - verify dentry provided from insecure source (deprecated)
L
Linus Torvalds 已提交
2011
 * @dentry: The dentry alleged to be valid child of @dparent
2012
 * @dparent: The parent dentry (known to be valid)
L
Linus Torvalds 已提交
2013 2014 2015 2016
 *
 * 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
2017 2018
 *
 * This function is slow for big directories, and deprecated, do not use it.
L
Linus Torvalds 已提交
2019
 */
2020
int d_validate(struct dentry *dentry, struct dentry *dparent)
L
Linus Torvalds 已提交
2021
{
N
Nick Piggin 已提交
2022
	struct dentry *child;
2023

N
Nick Piggin 已提交
2024
	spin_lock(&dparent->d_lock);
N
Nick Piggin 已提交
2025 2026
	list_for_each_entry(child, &dparent->d_subdirs, d_u.d_child) {
		if (dentry == child) {
N
Nick Piggin 已提交
2027
			spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
2028
			__dget_dlock(dentry);
N
Nick Piggin 已提交
2029 2030
			spin_unlock(&dentry->d_lock);
			spin_unlock(&dparent->d_lock);
L
Linus Torvalds 已提交
2031 2032 2033
			return 1;
		}
	}
N
Nick Piggin 已提交
2034
	spin_unlock(&dparent->d_lock);
N
Nick Piggin 已提交
2035

L
Linus Torvalds 已提交
2036 2037
	return 0;
}
2038
EXPORT_SYMBOL(d_validate);
L
Linus Torvalds 已提交
2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062

/*
 * 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)
{
2063
	struct inode *inode;
2064
	int isdir = 0;
L
Linus Torvalds 已提交
2065 2066 2067
	/*
	 * Are we the only user?
	 */
2068
again:
L
Linus Torvalds 已提交
2069
	spin_lock(&dentry->d_lock);
2070 2071
	inode = dentry->d_inode;
	isdir = S_ISDIR(inode->i_mode);
N
Nick Piggin 已提交
2072
	if (dentry->d_count == 1) {
A
Alan Cox 已提交
2073
		if (!spin_trylock(&inode->i_lock)) {
2074 2075 2076 2077
			spin_unlock(&dentry->d_lock);
			cpu_relax();
			goto again;
		}
A
Al Viro 已提交
2078
		dentry->d_flags &= ~DCACHE_CANT_MOUNT;
N
Nick Piggin 已提交
2079
		dentry_unlink_inode(dentry);
2080
		fsnotify_nameremove(dentry, isdir);
L
Linus Torvalds 已提交
2081 2082 2083 2084 2085 2086 2087
		return;
	}

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

	spin_unlock(&dentry->d_lock);
2088 2089

	fsnotify_nameremove(dentry, isdir);
L
Linus Torvalds 已提交
2090
}
2091
EXPORT_SYMBOL(d_delete);
L
Linus Torvalds 已提交
2092

2093
static void __d_rehash(struct dentry * entry, struct hlist_bl_head *b)
L
Linus Torvalds 已提交
2094
{
2095
	BUG_ON(!d_unhashed(entry));
2096
	hlist_bl_lock(b);
2097
	entry->d_flags |= DCACHE_RCUACCESS;
2098
	hlist_bl_add_head_rcu(&entry->d_hash, b);
2099
	hlist_bl_unlock(b);
L
Linus Torvalds 已提交
2100 2101
}

2102 2103 2104 2105 2106
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
2107 2108 2109 2110 2111 2112 2113 2114 2115 2116
/**
 * 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);
2117
	_d_rehash(entry);
L
Linus Torvalds 已提交
2118 2119
	spin_unlock(&entry->d_lock);
}
2120
EXPORT_SYMBOL(d_rehash);
L
Linus Torvalds 已提交
2121

N
Nick Piggin 已提交
2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137
/**
 * 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)
{
2138
	BUG_ON(!mutex_is_locked(&dentry->d_parent->d_inode->i_mutex));
N
Nick Piggin 已提交
2139 2140 2141
	BUG_ON(dentry->d_name.len != name->len); /* d_lookup gives this */

	spin_lock(&dentry->d_lock);
N
Nick Piggin 已提交
2142
	write_seqcount_begin(&dentry->d_seq);
N
Nick Piggin 已提交
2143
	memcpy((unsigned char *)dentry->d_name.name, name->name, name->len);
N
Nick Piggin 已提交
2144
	write_seqcount_end(&dentry->d_seq);
N
Nick Piggin 已提交
2145 2146 2147 2148
	spin_unlock(&dentry->d_lock);
}
EXPORT_SYMBOL(dentry_update_name_case);

L
Linus Torvalds 已提交
2149 2150 2151 2152 2153 2154 2155
static void switch_names(struct dentry *dentry, struct dentry *target)
{
	if (dname_external(target)) {
		if (dname_external(dentry)) {
			/*
			 * Both external: swap the pointers
			 */
2156
			swap(target->d_name.name, dentry->d_name.name);
L
Linus Torvalds 已提交
2157 2158 2159 2160 2161
		} else {
			/*
			 * dentry:internal, target:external.  Steal target's
			 * storage and make target internal.
			 */
2162 2163
			memcpy(target->d_iname, dentry->d_name.name,
					dentry->d_name.len + 1);
L
Linus Torvalds 已提交
2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182
			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);
2183 2184
			dentry->d_name.len = target->d_name.len;
			return;
L
Linus Torvalds 已提交
2185 2186
		}
	}
2187
	swap(dentry->d_name.len, target->d_name.len);
L
Linus Torvalds 已提交
2188 2189
}

N
Nick Piggin 已提交
2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225
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
2226
/*
N
Nick Piggin 已提交
2227 2228 2229 2230 2231 2232 2233 2234 2235
 * 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
2236
 */
2237
/*
2238
 * __d_move - move a dentry
L
Linus Torvalds 已提交
2239 2240 2241 2242
 * @dentry: entry to move
 * @target: new dentry
 *
 * Update the dcache to reflect the move of a file name. Negative
2243 2244 2245
 * 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
2246
 */
2247
static void __d_move(struct dentry * dentry, struct dentry * target)
L
Linus Torvalds 已提交
2248 2249 2250 2251
{
	if (!dentry->d_inode)
		printk(KERN_WARNING "VFS: moving negative dcache entry\n");

N
Nick Piggin 已提交
2252 2253 2254 2255
	BUG_ON(d_ancestor(dentry, target));
	BUG_ON(d_ancestor(target, dentry));

	dentry_lock_for_move(dentry, target);
L
Linus Torvalds 已提交
2256

N
Nick Piggin 已提交
2257 2258 2259
	write_seqcount_begin(&dentry->d_seq);
	write_seqcount_begin(&target->d_seq);

2260 2261 2262 2263 2264 2265 2266
	/* __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
2267
	__d_rehash(dentry, d_hash(target->d_parent, target->d_name.hash));
L
Linus Torvalds 已提交
2268 2269 2270 2271

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

E
Eric Dumazet 已提交
2272 2273
	list_del(&dentry->d_u.d_child);
	list_del(&target->d_u.d_child);
L
Linus Torvalds 已提交
2274 2275 2276

	/* Switch the names.. */
	switch_names(dentry, target);
2277
	swap(dentry->d_name.hash, target->d_name.hash);
L
Linus Torvalds 已提交
2278 2279 2280 2281 2282

	/* ... and switch the parents */
	if (IS_ROOT(dentry)) {
		dentry->d_parent = target->d_parent;
		target->d_parent = target;
E
Eric Dumazet 已提交
2283
		INIT_LIST_HEAD(&target->d_u.d_child);
L
Linus Torvalds 已提交
2284
	} else {
2285
		swap(dentry->d_parent, target->d_parent);
L
Linus Torvalds 已提交
2286 2287

		/* And add them back to the (new) parent lists */
E
Eric Dumazet 已提交
2288
		list_add(&target->d_u.d_child, &target->d_parent->d_subdirs);
L
Linus Torvalds 已提交
2289 2290
	}

E
Eric Dumazet 已提交
2291
	list_add(&dentry->d_u.d_child, &dentry->d_parent->d_subdirs);
N
Nick Piggin 已提交
2292

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

N
Nick Piggin 已提交
2296
	dentry_unlock_parents_for_move(dentry, target);
L
Linus Torvalds 已提交
2297
	spin_unlock(&target->d_lock);
2298
	fsnotify_d_move(dentry);
L
Linus Torvalds 已提交
2299
	spin_unlock(&dentry->d_lock);
2300 2301 2302 2303 2304 2305 2306 2307
}

/*
 * d_move - move a dentry
 * @dentry: entry to move
 * @target: new dentry
 *
 * Update the dcache to reflect the move of a file name. Negative
2308 2309
 * dcache entries should not be moved in this way. See the locking
 * requirements for __d_move.
2310 2311 2312 2313 2314
 */
void d_move(struct dentry *dentry, struct dentry *target)
{
	write_seqlock(&rename_lock);
	__d_move(dentry, target);
L
Linus Torvalds 已提交
2315
	write_sequnlock(&rename_lock);
2316
}
2317
EXPORT_SYMBOL(d_move);
L
Linus Torvalds 已提交
2318

2319 2320 2321 2322 2323 2324 2325
/**
 * 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.
2326
 */
2327
struct dentry *d_ancestor(struct dentry *p1, struct dentry *p2)
2328 2329 2330
{
	struct dentry *p;

2331
	for (p = p2; !IS_ROOT(p); p = p->d_parent) {
2332
		if (p->d_parent == p1)
2333
			return p;
2334
	}
2335
	return NULL;
2336 2337 2338 2339 2340 2341
}

/*
 * This helper attempts to cope with remotely renamed directories
 *
 * It assumes that the caller is already holding
2342
 * dentry->d_parent->d_inode->i_mutex, inode->i_lock and rename_lock
2343 2344 2345 2346
 *
 * Note: If ever the locking in lock_rename() changes, then please
 * remember to update this too...
 */
2347 2348
static struct dentry *__d_unalias(struct inode *inode,
		struct dentry *dentry, struct dentry *alias)
2349 2350
{
	struct mutex *m1 = NULL, *m2 = NULL;
2351
	struct dentry *ret = ERR_PTR(-EBUSY);
2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364

	/* 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:
2365 2366 2367 2368
	if (likely(!d_mountpoint(alias))) {
		__d_move(alias, dentry);
		ret = alias;
	}
2369
out_err:
2370
	spin_unlock(&inode->i_lock);
2371 2372 2373 2374 2375 2376 2377
	if (m2)
		mutex_unlock(m2);
	if (m1)
		mutex_unlock(m1);
	return ret;
}

2378 2379 2380
/*
 * 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
2381
 * returns with anon->d_lock held!
2382 2383 2384
 */
static void __d_materialise_dentry(struct dentry *dentry, struct dentry *anon)
{
2385
	struct dentry *dparent;
2386

N
Nick Piggin 已提交
2387
	dentry_lock_for_move(anon, dentry);
2388

N
Nick Piggin 已提交
2389 2390 2391
	write_seqcount_begin(&dentry->d_seq);
	write_seqcount_begin(&anon->d_seq);

2392 2393
	dparent = dentry->d_parent;

N
Nick Piggin 已提交
2394 2395 2396
	switch_names(dentry, anon);
	swap(dentry->d_name.hash, anon->d_name.hash);

2397 2398 2399
	dentry->d_parent = dentry;
	list_del_init(&dentry->d_u.d_child);
	anon->d_parent = dparent;
2400
	list_move(&anon->d_u.d_child, &dparent->d_subdirs);
2401

N
Nick Piggin 已提交
2402 2403 2404
	write_seqcount_end(&dentry->d_seq);
	write_seqcount_end(&anon->d_seq);

N
Nick Piggin 已提交
2405 2406 2407 2408
	dentry_unlock_parents_for_move(anon, dentry);
	spin_unlock(&dentry->d_lock);

	/* anon->d_lock still locked, returns locked */
2409 2410 2411 2412 2413 2414 2415 2416 2417
	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
2418 2419
 * root directory alias in its place if there is one. Caller must hold the
 * i_mutex of the parent directory.
2420 2421 2422
 */
struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
{
2423
	struct dentry *actual;
2424 2425 2426 2427 2428

	BUG_ON(!d_unhashed(dentry));

	if (!inode) {
		actual = dentry;
2429
		__d_instantiate(dentry, NULL);
2430 2431
		d_rehash(actual);
		goto out_nolock;
2432 2433
	}

2434
	spin_lock(&inode->i_lock);
2435

2436 2437 2438 2439
	if (S_ISDIR(inode->i_mode)) {
		struct dentry *alias;

		/* Does an aliased dentry already exist? */
2440
		alias = __d_find_alias(inode, 0);
2441 2442
		if (alias) {
			actual = alias;
2443 2444 2445 2446 2447
			write_seqlock(&rename_lock);

			if (d_ancestor(alias, dentry)) {
				/* Check for loops */
				actual = ERR_PTR(-ELOOP);
2448
				spin_unlock(&inode->i_lock);
2449 2450 2451
			} else if (IS_ROOT(alias)) {
				/* Is this an anonymous mountpoint that we
				 * could splice into our tree? */
2452
				__d_materialise_dentry(dentry, alias);
2453
				write_sequnlock(&rename_lock);
2454 2455
				__d_drop(alias);
				goto found;
2456 2457
			} else {
				/* Nope, but we must(!) avoid directory
2458
				 * aliasing. This drops inode->i_lock */
2459
				actual = __d_unalias(inode, dentry, alias);
2460
			}
2461
			write_sequnlock(&rename_lock);
2462 2463 2464 2465 2466 2467 2468 2469
			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);
2470
				dput(alias);
2471
			}
2472 2473
			goto out_nolock;
		}
2474 2475 2476 2477 2478 2479
	}

	/* Add a unique reference */
	actual = __d_instantiate_unique(dentry, inode);
	if (!actual)
		actual = dentry;
2480 2481
	else
		BUG_ON(!d_unhashed(actual));
2482 2483 2484 2485 2486

	spin_lock(&actual->d_lock);
found:
	_d_rehash(actual);
	spin_unlock(&actual->d_lock);
2487
	spin_unlock(&inode->i_lock);
2488
out_nolock:
2489 2490 2491 2492 2493 2494 2495 2496
	if (actual == dentry) {
		security_d_instantiate(dentry, inode);
		return NULL;
	}

	iput(inode);
	return actual;
}
2497
EXPORT_SYMBOL_GPL(d_materialise_unique);
2498

2499
static int prepend(char **buffer, int *buflen, const char *str, int namelen)
2500 2501 2502 2503 2504 2505 2506 2507 2508
{
	*buflen -= namelen;
	if (*buflen < 0)
		return -ENAMETOOLONG;
	*buffer -= namelen;
	memcpy(*buffer, str, namelen);
	return 0;
}

2509 2510 2511 2512 2513
static int prepend_name(char **buffer, int *buflen, struct qstr *name)
{
	return prepend(buffer, buflen, name->name, name->len);
}

L
Linus Torvalds 已提交
2514
/**
2515
 * prepend_path - Prepend path string to a buffer
2516
 * @path: the dentry/vfsmount to report
2517
 * @root: root vfsmnt/dentry
M
Miklos Szeredi 已提交
2518 2519
 * @buffer: pointer to the end of the buffer
 * @buflen: pointer to buffer length
2520
 *
2521
 * Caller holds the rename_lock.
L
Linus Torvalds 已提交
2522
 */
2523 2524
static int prepend_path(const struct path *path,
			const struct path *root,
M
Miklos Szeredi 已提交
2525
			char **buffer, int *buflen)
L
Linus Torvalds 已提交
2526
{
2527 2528
	struct dentry *dentry = path->dentry;
	struct vfsmount *vfsmnt = path->mnt;
2529
	struct mount *mnt = real_mount(vfsmnt);
M
Miklos Szeredi 已提交
2530 2531
	bool slash = false;
	int error = 0;
2532

M
Miklos Szeredi 已提交
2533
	while (dentry != root->dentry || vfsmnt != root->mnt) {
L
Linus Torvalds 已提交
2534 2535 2536
		struct dentry * parent;

		if (dentry == vfsmnt->mnt_root || IS_ROOT(dentry)) {
2537
			/* Global root? */
2538
			if (!mnt_has_parent(mnt))
L
Linus Torvalds 已提交
2539
				goto global_root;
2540
			dentry = mnt->mnt_mountpoint;
2541 2542
			mnt = mnt->mnt_parent;
			vfsmnt = &mnt->mnt;
L
Linus Torvalds 已提交
2543 2544 2545 2546
			continue;
		}
		parent = dentry->d_parent;
		prefetch(parent);
N
Nick Piggin 已提交
2547
		spin_lock(&dentry->d_lock);
M
Miklos Szeredi 已提交
2548
		error = prepend_name(buffer, buflen, &dentry->d_name);
N
Nick Piggin 已提交
2549
		spin_unlock(&dentry->d_lock);
M
Miklos Szeredi 已提交
2550 2551 2552 2553 2554 2555
		if (!error)
			error = prepend(buffer, buflen, "/", 1);
		if (error)
			break;

		slash = true;
L
Linus Torvalds 已提交
2556 2557 2558
		dentry = parent;
	}

M
Miklos Szeredi 已提交
2559 2560 2561 2562
	if (!error && !slash)
		error = prepend(buffer, buflen, "/", 1);

	return error;
L
Linus Torvalds 已提交
2563 2564

global_root:
2565 2566 2567 2568 2569 2570 2571 2572 2573
	/*
	 * 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);
	}
2574 2575 2576
	if (!slash)
		error = prepend(buffer, buflen, "/", 1);
	if (!error)
A
Al Viro 已提交
2577
		error = is_mounted(vfsmnt) ? 1 : 2;
A
Al Viro 已提交
2578
	return error;
M
Miklos Szeredi 已提交
2579
}
2580

M
Miklos Szeredi 已提交
2581 2582 2583
/**
 * __d_path - return the path of a dentry
 * @path: the dentry/vfsmount to report
2584
 * @root: root vfsmnt/dentry
2585
 * @buf: buffer to return value in
M
Miklos Szeredi 已提交
2586 2587
 * @buflen: buffer length
 *
2588
 * Convert a dentry into an ASCII path name.
M
Miklos Szeredi 已提交
2589 2590 2591 2592
 *
 * Returns a pointer into the buffer or an error code if the
 * path was too long.
 *
2593
 * "buflen" should be positive.
M
Miklos Szeredi 已提交
2594
 *
2595
 * If the path is not reachable from the supplied root, return %NULL.
M
Miklos Szeredi 已提交
2596
 */
2597 2598
char *__d_path(const struct path *path,
	       const struct path *root,
M
Miklos Szeredi 已提交
2599 2600 2601 2602 2603 2604
	       char *buf, int buflen)
{
	char *res = buf + buflen;
	int error;

	prepend(&res, &buflen, "\0", 1);
A
Al Viro 已提交
2605
	br_read_lock(&vfsmount_lock);
2606
	write_seqlock(&rename_lock);
M
Miklos Szeredi 已提交
2607
	error = prepend_path(path, root, &res, &buflen);
2608
	write_sequnlock(&rename_lock);
A
Al Viro 已提交
2609
	br_read_unlock(&vfsmount_lock);
2610

2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625
	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
2626
	br_read_lock(&vfsmount_lock);
2627 2628 2629
	write_seqlock(&rename_lock);
	error = prepend_path(path, &root, &res, &buflen);
	write_sequnlock(&rename_lock);
A
Al Viro 已提交
2630
	br_read_unlock(&vfsmount_lock);
2631 2632 2633 2634

	if (error > 1)
		error = -EINVAL;
	if (error < 0)
M
Miklos Szeredi 已提交
2635 2636
		return ERR_PTR(error);
	return res;
L
Linus Torvalds 已提交
2637 2638
}

2639 2640 2641
/*
 * same as __d_path but appends "(deleted)" for unlinked files.
 */
2642 2643 2644
static int path_with_deleted(const struct path *path,
			     const struct path *root,
			     char **buf, int *buflen)
2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655
{
	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);
}

2656 2657 2658 2659 2660
static int prepend_unreachable(char **buffer, int *buflen)
{
	return prepend(buffer, buflen, "(unreachable)", 13);
}

J
Jan Blunck 已提交
2661 2662
/**
 * d_path - return the path of a dentry
2663
 * @path: path to report
J
Jan Blunck 已提交
2664 2665 2666 2667 2668 2669
 * @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.
 *
2670 2671 2672 2673
 * 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
2674
 *
2675
 * "buflen" should be positive.
J
Jan Blunck 已提交
2676
 */
2677
char *d_path(const struct path *path, char *buf, int buflen)
L
Linus Torvalds 已提交
2678
{
2679
	char *res = buf + buflen;
J
Jan Blunck 已提交
2680
	struct path root;
2681
	int error;
L
Linus Torvalds 已提交
2682

2683 2684 2685 2686 2687 2688 2689
	/*
	 * 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:
	 */
2690 2691
	if (path->dentry->d_op && path->dentry->d_op->d_dname)
		return path->dentry->d_op->d_dname(path->dentry, buf, buflen);
2692

2693
	get_fs_root(current->fs, &root);
A
Al Viro 已提交
2694
	br_read_lock(&vfsmount_lock);
2695
	write_seqlock(&rename_lock);
2696
	error = path_with_deleted(path, &root, &res, &buflen);
A
Al Viro 已提交
2697 2698
	write_sequnlock(&rename_lock);
	br_read_unlock(&vfsmount_lock);
2699
	if (error < 0)
2700
		res = ERR_PTR(error);
J
Jan Blunck 已提交
2701
	path_put(&root);
L
Linus Torvalds 已提交
2702 2703
	return res;
}
2704
EXPORT_SYMBOL(d_path);
L
Linus Torvalds 已提交
2705

2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726
/*
 * 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);
}

2727 2728 2729
/*
 * Write full pathname from the root of the filesystem into the buffer.
 */
N
Nick Piggin 已提交
2730
static char *__dentry_path(struct dentry *dentry, char *buf, int buflen)
2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741
{
	char *end = buf + buflen;
	char *retval;

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

2742 2743
	while (!IS_ROOT(dentry)) {
		struct dentry *parent = dentry->d_parent;
N
Nick Piggin 已提交
2744
		int error;
2745 2746

		prefetch(parent);
N
Nick Piggin 已提交
2747 2748 2749 2750
		spin_lock(&dentry->d_lock);
		error = prepend_name(&end, &buflen, &dentry->d_name);
		spin_unlock(&dentry->d_lock);
		if (error != 0 || prepend(&end, &buflen, "/", 1) != 0)
2751 2752 2753 2754 2755
			goto Elong;

		retval = end;
		dentry = parent;
	}
A
Al Viro 已提交
2756 2757 2758 2759
	return retval;
Elong:
	return ERR_PTR(-ENAMETOOLONG);
}
N
Nick Piggin 已提交
2760 2761 2762 2763 2764

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

2765
	write_seqlock(&rename_lock);
N
Nick Piggin 已提交
2766
	retval = __dentry_path(dentry, buf, buflen);
2767
	write_sequnlock(&rename_lock);
N
Nick Piggin 已提交
2768 2769 2770 2771

	return retval;
}
EXPORT_SYMBOL(dentry_path_raw);
A
Al Viro 已提交
2772 2773 2774 2775 2776 2777

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

2778
	write_seqlock(&rename_lock);
A
Al Viro 已提交
2779 2780 2781 2782 2783 2784 2785
	if (d_unlinked(dentry)) {
		p = buf + buflen;
		if (prepend(&p, &buflen, "//deleted", 10) != 0)
			goto Elong;
		buflen++;
	}
	retval = __dentry_path(dentry, buf, buflen);
2786
	write_sequnlock(&rename_lock);
A
Al Viro 已提交
2787 2788
	if (!IS_ERR(retval) && p)
		*p = '/';	/* restore '/' overriden with '\0' */
2789 2790 2791 2792 2793
	return retval;
Elong:
	return ERR_PTR(-ENAMETOOLONG);
}

L
Linus Torvalds 已提交
2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811
/*
 * 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;
 *	}
 */
2812
SYSCALL_DEFINE2(getcwd, char __user *, buf, unsigned long, size)
L
Linus Torvalds 已提交
2813
{
2814
	int error;
J
Jan Blunck 已提交
2815
	struct path pwd, root;
2816
	char *page = (char *) __get_free_page(GFP_USER);
L
Linus Torvalds 已提交
2817 2818 2819 2820

	if (!page)
		return -ENOMEM;

2821
	get_fs_root_and_pwd(current->fs, &root, &pwd);
L
Linus Torvalds 已提交
2822

2823
	error = -ENOENT;
A
Al Viro 已提交
2824
	br_read_lock(&vfsmount_lock);
2825
	write_seqlock(&rename_lock);
2826
	if (!d_unlinked(pwd.dentry)) {
2827
		unsigned long len;
2828 2829
		char *cwd = page + PAGE_SIZE;
		int buflen = PAGE_SIZE;
L
Linus Torvalds 已提交
2830

2831
		prepend(&cwd, &buflen, "\0", 1);
2832
		error = prepend_path(&pwd, &root, &cwd, &buflen);
2833
		write_sequnlock(&rename_lock);
A
Al Viro 已提交
2834
		br_read_unlock(&vfsmount_lock);
2835

2836
		if (error < 0)
2837 2838
			goto out;

2839
		/* Unreachable from current root */
2840
		if (error > 0) {
2841 2842 2843 2844 2845
			error = prepend_unreachable(&cwd, &buflen);
			if (error)
				goto out;
		}

2846 2847 2848 2849 2850 2851 2852
		error = -ERANGE;
		len = PAGE_SIZE + page - cwd;
		if (len <= size) {
			error = len;
			if (copy_to_user(buf, cwd, len))
				error = -EFAULT;
		}
2853 2854
	} else {
		write_sequnlock(&rename_lock);
A
Al Viro 已提交
2855
		br_read_unlock(&vfsmount_lock);
2856
	}
L
Linus Torvalds 已提交
2857 2858

out:
J
Jan Blunck 已提交
2859 2860
	path_put(&pwd);
	path_put(&root);
L
Linus Torvalds 已提交
2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880
	free_page((unsigned long) page);
	return error;
}

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

/**
 * is_subdir - is new dentry a subdirectory of old_dentry
 * @new_dentry: new dentry
 * @old_dentry: old dentry
 *
 * Returns 1 if new_dentry is a subdirectory of the parent (at any depth).
 * Returns 0 otherwise.
 * Caller must ensure that "new_dentry" is pinned before calling is_subdir()
 */
  
2881
int is_subdir(struct dentry *new_dentry, struct dentry *old_dentry)
L
Linus Torvalds 已提交
2882 2883
{
	int result;
2884
	unsigned seq;
L
Linus Torvalds 已提交
2885

2886 2887 2888 2889
	if (new_dentry == old_dentry)
		return 1;

	do {
L
Linus Torvalds 已提交
2890 2891
		/* for restarting inner loop in case of seq retry */
		seq = read_seqbegin(&rename_lock);
2892 2893 2894 2895 2896
		/*
		 * Need rcu_readlock to protect against the d_parent trashing
		 * due to d_move
		 */
		rcu_read_lock();
2897
		if (d_ancestor(old_dentry, new_dentry))
L
Linus Torvalds 已提交
2898
			result = 1;
2899 2900
		else
			result = 0;
2901
		rcu_read_unlock();
L
Linus Torvalds 已提交
2902 2903 2904 2905 2906 2907 2908
	} while (read_seqretry(&rename_lock, seq));

	return result;
}

void d_genocide(struct dentry *root)
{
2909
	struct dentry *this_parent;
L
Linus Torvalds 已提交
2910
	struct list_head *next;
2911
	unsigned seq;
2912
	int locked = 0;
L
Linus Torvalds 已提交
2913

2914
	seq = read_seqbegin(&rename_lock);
2915 2916
again:
	this_parent = root;
N
Nick Piggin 已提交
2917
	spin_lock(&this_parent->d_lock);
L
Linus Torvalds 已提交
2918 2919 2920 2921 2922
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
2923
		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
L
Linus Torvalds 已提交
2924
		next = tmp->next;
2925

N
Nick Piggin 已提交
2926 2927 2928
		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
		if (d_unhashed(dentry) || !dentry->d_inode) {
			spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
2929
			continue;
N
Nick Piggin 已提交
2930
		}
L
Linus Torvalds 已提交
2931
		if (!list_empty(&dentry->d_subdirs)) {
N
Nick Piggin 已提交
2932 2933
			spin_unlock(&this_parent->d_lock);
			spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
L
Linus Torvalds 已提交
2934
			this_parent = dentry;
N
Nick Piggin 已提交
2935
			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
L
Linus Torvalds 已提交
2936 2937
			goto repeat;
		}
2938 2939 2940 2941
		if (!(dentry->d_flags & DCACHE_GENOCIDE)) {
			dentry->d_flags |= DCACHE_GENOCIDE;
			dentry->d_count--;
		}
N
Nick Piggin 已提交
2942
		spin_unlock(&dentry->d_lock);
L
Linus Torvalds 已提交
2943 2944
	}
	if (this_parent != root) {
2945
		struct dentry *child = this_parent;
2946 2947 2948 2949
		if (!(this_parent->d_flags & DCACHE_GENOCIDE)) {
			this_parent->d_flags |= DCACHE_GENOCIDE;
			this_parent->d_count--;
		}
2950 2951
		this_parent = try_to_ascend(this_parent, locked, seq);
		if (!this_parent)
2952 2953
			goto rename_retry;
		next = child->d_u.d_child.next;
L
Linus Torvalds 已提交
2954 2955
		goto resume;
	}
N
Nick Piggin 已提交
2956
	spin_unlock(&this_parent->d_lock);
2957
	if (!locked && read_seqretry(&rename_lock, seq))
2958
		goto rename_retry;
2959 2960 2961 2962 2963
	if (locked)
		write_sequnlock(&rename_lock);
	return;

rename_retry:
2964 2965
	if (locked)
		goto again;
2966 2967 2968
	locked = 1;
	write_seqlock(&rename_lock);
	goto again;
L
Linus Torvalds 已提交
2969 2970
}

2971
void d_tmpfile(struct dentry *dentry, struct inode *inode)
L
Linus Torvalds 已提交
2972
{
2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983
	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
2984
}
2985
EXPORT_SYMBOL(d_tmpfile);
L
Linus Torvalds 已提交
2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998

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)
{
2999
	unsigned int loop;
L
Linus Torvalds 已提交
3000 3001 3002 3003 3004 3005 3006 3007 3008

	/* 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",
3009
					sizeof(struct hlist_bl_head),
L
Linus Torvalds 已提交
3010 3011 3012 3013 3014
					dhash_entries,
					13,
					HASH_EARLY,
					&d_hash_shift,
					&d_hash_mask,
3015
					0,
L
Linus Torvalds 已提交
3016 3017
					0);

3018
	for (loop = 0; loop < (1U << d_hash_shift); loop++)
3019
		INIT_HLIST_BL_HEAD(dentry_hashtable + loop);
L
Linus Torvalds 已提交
3020 3021
}

3022
static void __init dcache_init(void)
L
Linus Torvalds 已提交
3023
{
3024
	unsigned int loop;
L
Linus Torvalds 已提交
3025 3026 3027 3028 3029 3030

	/* 
	 * 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. 
	 */
3031 3032
	dentry_cache = KMEM_CACHE(dentry,
		SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|SLAB_MEM_SPREAD);
L
Linus Torvalds 已提交
3033 3034 3035 3036 3037 3038 3039

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

	dentry_hashtable =
		alloc_large_system_hash("Dentry cache",
3040
					sizeof(struct hlist_bl_head),
L
Linus Torvalds 已提交
3041 3042 3043 3044 3045
					dhash_entries,
					13,
					0,
					&d_hash_shift,
					&d_hash_mask,
3046
					0,
L
Linus Torvalds 已提交
3047 3048
					0);

3049
	for (loop = 0; loop < (1U << d_hash_shift); loop++)
3050
		INIT_HLIST_BL_HEAD(dentry_hashtable + loop);
L
Linus Torvalds 已提交
3051 3052 3053
}

/* SLAB cache for __getname() consumers */
3054
struct kmem_cache *names_cachep __read_mostly;
3055
EXPORT_SYMBOL(names_cachep);
L
Linus Torvalds 已提交
3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075

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,
3076
			SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
L
Linus Torvalds 已提交
3077

3078 3079
	dcache_init();
	inode_init();
L
Linus Torvalds 已提交
3080
	files_init(mempages);
3081
	mnt_init();
L
Linus Torvalds 已提交
3082 3083 3084
	bdev_cache_init();
	chrdev_init();
}