blob: d68631f18df122824a0d36a307c647f15931c7bd [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * fs/dcache.c
3 *
4 * Complete reimplementation
5 * (C) 1997 Thomas Schoebel-Theuer,
6 * with heavy changes by Linus Torvalds
7 */
8
9/*
10 * Notes on the allocation strategy:
11 *
12 * The dcache is a master of the icache - whenever a dcache entry
13 * exists, the inode will always exist. "iput()" is done either when
14 * the dcache entry is deleted or garbage collected.
15 */
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/syscalls.h>
18#include <linux/string.h>
19#include <linux/mm.h>
20#include <linux/fs.h>
John McCutchan7a91bf72005-08-08 13:52:16 -040021#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/slab.h>
23#include <linux/init.h>
24#include <linux/smp_lock.h>
25#include <linux/hash.h>
26#include <linux/cache.h>
27#include <linux/module.h>
28#include <linux/mount.h>
29#include <linux/file.h>
30#include <asm/uaccess.h>
31#include <linux/security.h>
32#include <linux/seqlock.h>
33#include <linux/swap.h>
34#include <linux/bootmem.h>
David Howells07f3f052006-09-30 20:52:18 +020035#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070036
Linus Torvalds1da177e2005-04-16 15:20:36 -070037
Eric Dumazetfa3536c2006-03-26 01:37:24 -080038int sysctl_vfs_cache_pressure __read_mostly = 100;
Linus Torvalds1da177e2005-04-16 15:20:36 -070039EXPORT_SYMBOL_GPL(sysctl_vfs_cache_pressure);
40
41 __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lock);
Ingo Molnare4d91912006-07-03 00:24:34 -070042static __cacheline_aligned_in_smp DEFINE_SEQLOCK(rename_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
44EXPORT_SYMBOL(dcache_lock);
45
Christoph Lametere18b8902006-12-06 20:33:20 -080046static struct kmem_cache *dentry_cache __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -070047
48#define DNAME_INLINE_LEN (sizeof(struct dentry)-offsetof(struct dentry,d_iname))
49
50/*
51 * This is the single most critical data structure when it comes
52 * to the dcache: the hashtable for lookups. Somebody should try
53 * to make this good - I've just made it work.
54 *
55 * This hash-function tries to avoid losing too many bits of hash
56 * information, yet avoid using a prime hash-size or similar.
57 */
58#define D_HASHBITS d_hash_shift
59#define D_HASHMASK d_hash_mask
60
Eric Dumazetfa3536c2006-03-26 01:37:24 -080061static unsigned int d_hash_mask __read_mostly;
62static unsigned int d_hash_shift __read_mostly;
63static struct hlist_head *dentry_hashtable __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -070064static LIST_HEAD(dentry_unused);
65
66/* Statistics gathering. */
67struct dentry_stat_t dentry_stat = {
68 .age_limit = 45,
69};
70
Eric Dumazetb3423412006-12-06 20:38:48 -080071static void __d_free(struct dentry *dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -070072{
Linus Torvalds1da177e2005-04-16 15:20:36 -070073 if (dname_external(dentry))
74 kfree(dentry->d_name.name);
75 kmem_cache_free(dentry_cache, dentry);
76}
77
Eric Dumazetb3423412006-12-06 20:38:48 -080078static void d_callback(struct rcu_head *head)
79{
80 struct dentry * dentry = container_of(head, struct dentry, d_u.d_rcu);
81 __d_free(dentry);
82}
83
Linus Torvalds1da177e2005-04-16 15:20:36 -070084/*
85 * no dcache_lock, please. The caller must decrement dentry_stat.nr_dentry
86 * inside dcache_lock.
87 */
88static void d_free(struct dentry *dentry)
89{
90 if (dentry->d_op && dentry->d_op->d_release)
91 dentry->d_op->d_release(dentry);
Eric Dumazetb3423412006-12-06 20:38:48 -080092 /* if dentry was never inserted into hash, immediate free is OK */
93 if (dentry->d_hash.pprev == NULL)
94 __d_free(dentry);
95 else
96 call_rcu(&dentry->d_u.d_rcu, d_callback);
Linus Torvalds1da177e2005-04-16 15:20:36 -070097}
98
99/*
100 * Release the dentry's inode, using the filesystem
101 * d_iput() operation if defined.
102 * Called with dcache_lock and per dentry lock held, drops both.
103 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800104static void dentry_iput(struct dentry * dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105{
106 struct inode *inode = dentry->d_inode;
107 if (inode) {
108 dentry->d_inode = NULL;
109 list_del_init(&dentry->d_alias);
110 spin_unlock(&dentry->d_lock);
111 spin_unlock(&dcache_lock);
Linus Torvaldsf805fbd2005-09-19 19:54:29 -0700112 if (!inode->i_nlink)
113 fsnotify_inoderemove(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114 if (dentry->d_op && dentry->d_op->d_iput)
115 dentry->d_op->d_iput(dentry, inode);
116 else
117 iput(inode);
118 } else {
119 spin_unlock(&dentry->d_lock);
120 spin_unlock(&dcache_lock);
121 }
122}
123
124/*
125 * This is dput
126 *
127 * This is complicated by the fact that we do not want to put
128 * dentries that are no longer on any hash chain on the unused
129 * list: we'd much rather just get rid of them immediately.
130 *
131 * However, that implies that we have to traverse the dentry
132 * tree upwards to the parents which might _also_ now be
133 * scheduled for deletion (it may have been only waiting for
134 * its last child to go away).
135 *
136 * This tail recursion is done by hand as we don't want to depend
137 * on the compiler to always get this right (gcc generally doesn't).
138 * Real recursion would eat up our stack space.
139 */
140
141/*
142 * dput - release a dentry
143 * @dentry: dentry to release
144 *
145 * Release a dentry. This will drop the usage count and if appropriate
146 * call the dentry unlink method as well as removing it from the queues and
147 * releasing its resources. If the parent dentries were scheduled for release
148 * they too may now get deleted.
149 *
150 * no dcache lock, please.
151 */
152
153void dput(struct dentry *dentry)
154{
155 if (!dentry)
156 return;
157
158repeat:
159 if (atomic_read(&dentry->d_count) == 1)
160 might_sleep();
161 if (!atomic_dec_and_lock(&dentry->d_count, &dcache_lock))
162 return;
163
164 spin_lock(&dentry->d_lock);
165 if (atomic_read(&dentry->d_count)) {
166 spin_unlock(&dentry->d_lock);
167 spin_unlock(&dcache_lock);
168 return;
169 }
170
171 /*
172 * AV: ->d_delete() is _NOT_ allowed to block now.
173 */
174 if (dentry->d_op && dentry->d_op->d_delete) {
175 if (dentry->d_op->d_delete(dentry))
176 goto unhash_it;
177 }
178 /* Unreachable? Get rid of it */
179 if (d_unhashed(dentry))
180 goto kill_it;
181 if (list_empty(&dentry->d_lru)) {
182 dentry->d_flags |= DCACHE_REFERENCED;
183 list_add(&dentry->d_lru, &dentry_unused);
184 dentry_stat.nr_unused++;
185 }
186 spin_unlock(&dentry->d_lock);
187 spin_unlock(&dcache_lock);
188 return;
189
190unhash_it:
191 __d_drop(dentry);
192
193kill_it: {
194 struct dentry *parent;
195
196 /* If dentry was on d_lru list
197 * delete it from there
198 */
199 if (!list_empty(&dentry->d_lru)) {
200 list_del(&dentry->d_lru);
201 dentry_stat.nr_unused--;
202 }
Eric Dumazet5160ee62006-01-08 01:03:32 -0800203 list_del(&dentry->d_u.d_child);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 dentry_stat.nr_dentry--; /* For d_free, below */
205 /*drops the locks, at that point nobody can reach this dentry */
206 dentry_iput(dentry);
207 parent = dentry->d_parent;
208 d_free(dentry);
209 if (dentry == parent)
210 return;
211 dentry = parent;
212 goto repeat;
213 }
214}
215
216/**
217 * d_invalidate - invalidate a dentry
218 * @dentry: dentry to invalidate
219 *
220 * Try to invalidate the dentry if it turns out to be
221 * possible. If there are other dentries that can be
222 * reached through this one we can't delete it and we
223 * return -EBUSY. On success we return 0.
224 *
225 * no dcache lock.
226 */
227
228int d_invalidate(struct dentry * dentry)
229{
230 /*
231 * If it's already been dropped, return OK.
232 */
233 spin_lock(&dcache_lock);
234 if (d_unhashed(dentry)) {
235 spin_unlock(&dcache_lock);
236 return 0;
237 }
238 /*
239 * Check whether to do a partial shrink_dcache
240 * to get rid of unused child entries.
241 */
242 if (!list_empty(&dentry->d_subdirs)) {
243 spin_unlock(&dcache_lock);
244 shrink_dcache_parent(dentry);
245 spin_lock(&dcache_lock);
246 }
247
248 /*
249 * Somebody else still using it?
250 *
251 * If it's a directory, we can't drop it
252 * for fear of somebody re-populating it
253 * with children (even though dropping it
254 * would make it unreachable from the root,
255 * we might still populate it if it was a
256 * working directory or similar).
257 */
258 spin_lock(&dentry->d_lock);
259 if (atomic_read(&dentry->d_count) > 1) {
260 if (dentry->d_inode && S_ISDIR(dentry->d_inode->i_mode)) {
261 spin_unlock(&dentry->d_lock);
262 spin_unlock(&dcache_lock);
263 return -EBUSY;
264 }
265 }
266
267 __d_drop(dentry);
268 spin_unlock(&dentry->d_lock);
269 spin_unlock(&dcache_lock);
270 return 0;
271}
272
273/* This should be called _only_ with dcache_lock held */
274
275static inline struct dentry * __dget_locked(struct dentry *dentry)
276{
277 atomic_inc(&dentry->d_count);
278 if (!list_empty(&dentry->d_lru)) {
279 dentry_stat.nr_unused--;
280 list_del_init(&dentry->d_lru);
281 }
282 return dentry;
283}
284
285struct dentry * dget_locked(struct dentry *dentry)
286{
287 return __dget_locked(dentry);
288}
289
290/**
291 * d_find_alias - grab a hashed alias of inode
292 * @inode: inode in question
293 * @want_discon: flag, used by d_splice_alias, to request
294 * that only a DISCONNECTED alias be returned.
295 *
296 * If inode has a hashed alias, or is a directory and has any alias,
297 * acquire the reference to alias and return it. Otherwise return NULL.
298 * Notice that if inode is a directory there can be only one alias and
299 * it can be unhashed only if it has no children, or if it is the root
300 * of a filesystem.
301 *
NeilBrown21c0d8f2006-10-04 02:16:16 -0700302 * If the inode has an IS_ROOT, DCACHE_DISCONNECTED alias, then prefer
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 * any other hashed alias over that one unless @want_discon is set,
NeilBrown21c0d8f2006-10-04 02:16:16 -0700304 * in which case only return an IS_ROOT, DCACHE_DISCONNECTED alias.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 */
306
307static struct dentry * __d_find_alias(struct inode *inode, int want_discon)
308{
309 struct list_head *head, *next, *tmp;
310 struct dentry *alias, *discon_alias=NULL;
311
312 head = &inode->i_dentry;
313 next = inode->i_dentry.next;
314 while (next != head) {
315 tmp = next;
316 next = tmp->next;
317 prefetch(next);
318 alias = list_entry(tmp, struct dentry, d_alias);
319 if (S_ISDIR(inode->i_mode) || !d_unhashed(alias)) {
NeilBrown21c0d8f2006-10-04 02:16:16 -0700320 if (IS_ROOT(alias) &&
321 (alias->d_flags & DCACHE_DISCONNECTED))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 discon_alias = alias;
323 else if (!want_discon) {
324 __dget_locked(alias);
325 return alias;
326 }
327 }
328 }
329 if (discon_alias)
330 __dget_locked(discon_alias);
331 return discon_alias;
332}
333
334struct dentry * d_find_alias(struct inode *inode)
335{
David Howells214fda12006-03-25 03:06:36 -0800336 struct dentry *de = NULL;
337
338 if (!list_empty(&inode->i_dentry)) {
339 spin_lock(&dcache_lock);
340 de = __d_find_alias(inode, 0);
341 spin_unlock(&dcache_lock);
342 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 return de;
344}
345
346/*
347 * Try to kill dentries associated with this inode.
348 * WARNING: you must own a reference to inode.
349 */
350void d_prune_aliases(struct inode *inode)
351{
Domen Puncer0cdca3f2005-09-10 00:27:07 -0700352 struct dentry *dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353restart:
354 spin_lock(&dcache_lock);
Domen Puncer0cdca3f2005-09-10 00:27:07 -0700355 list_for_each_entry(dentry, &inode->i_dentry, d_alias) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 spin_lock(&dentry->d_lock);
357 if (!atomic_read(&dentry->d_count)) {
358 __dget_locked(dentry);
359 __d_drop(dentry);
360 spin_unlock(&dentry->d_lock);
361 spin_unlock(&dcache_lock);
362 dput(dentry);
363 goto restart;
364 }
365 spin_unlock(&dentry->d_lock);
366 }
367 spin_unlock(&dcache_lock);
368}
369
370/*
Andrew Mortond702ccb2006-06-22 14:47:31 -0700371 * Throw away a dentry - free the inode, dput the parent. This requires that
372 * the LRU list has already been removed.
373 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 * Called with dcache_lock, drops it and then regains.
Andrew Mortond702ccb2006-06-22 14:47:31 -0700375 * Called with dentry->d_lock held, drops it.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 */
Andrew Mortond702ccb2006-06-22 14:47:31 -0700377static void prune_one_dentry(struct dentry * dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378{
379 struct dentry * parent;
380
381 __d_drop(dentry);
Eric Dumazet5160ee62006-01-08 01:03:32 -0800382 list_del(&dentry->d_u.d_child);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 dentry_stat.nr_dentry--; /* For d_free, below */
384 dentry_iput(dentry);
385 parent = dentry->d_parent;
386 d_free(dentry);
387 if (parent != dentry)
388 dput(parent);
389 spin_lock(&dcache_lock);
390}
391
392/**
393 * prune_dcache - shrink the dcache
394 * @count: number of entries to try and free
NeilBrown0feae5c2006-06-22 14:47:28 -0700395 * @sb: if given, ignore dentries for other superblocks
396 * which are being unmounted.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 *
398 * Shrink the dcache. This is done when we need
399 * more memory, or simply when we need to unmount
400 * something (at which point we need to unuse
401 * all dentries).
402 *
403 * This function may fail to free any resources if
404 * all the dentries are in use.
405 */
406
NeilBrown0feae5c2006-06-22 14:47:28 -0700407static void prune_dcache(int count, struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408{
409 spin_lock(&dcache_lock);
410 for (; count ; count--) {
411 struct dentry *dentry;
412 struct list_head *tmp;
NeilBrown0feae5c2006-06-22 14:47:28 -0700413 struct rw_semaphore *s_umount;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414
415 cond_resched_lock(&dcache_lock);
416
417 tmp = dentry_unused.prev;
Hua Zhongf58a1eb2006-06-25 05:49:32 -0700418 if (sb) {
NeilBrown0feae5c2006-06-22 14:47:28 -0700419 /* Try to find a dentry for this sb, but don't try
420 * too hard, if they aren't near the tail they will
421 * be moved down again soon
422 */
423 int skip = count;
424 while (skip && tmp != &dentry_unused &&
425 list_entry(tmp, struct dentry, d_lru)->d_sb != sb) {
426 skip--;
427 tmp = tmp->prev;
428 }
429 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 if (tmp == &dentry_unused)
431 break;
432 list_del_init(tmp);
433 prefetch(dentry_unused.prev);
434 dentry_stat.nr_unused--;
435 dentry = list_entry(tmp, struct dentry, d_lru);
436
437 spin_lock(&dentry->d_lock);
438 /*
439 * We found an inuse dentry which was not removed from
440 * dentry_unused because of laziness during lookup. Do not free
441 * it - just keep it off the dentry_unused list.
442 */
443 if (atomic_read(&dentry->d_count)) {
444 spin_unlock(&dentry->d_lock);
445 continue;
446 }
447 /* If the dentry was recently referenced, don't free it. */
448 if (dentry->d_flags & DCACHE_REFERENCED) {
449 dentry->d_flags &= ~DCACHE_REFERENCED;
450 list_add(&dentry->d_lru, &dentry_unused);
451 dentry_stat.nr_unused++;
452 spin_unlock(&dentry->d_lock);
453 continue;
454 }
NeilBrown0feae5c2006-06-22 14:47:28 -0700455 /*
456 * If the dentry is not DCACHED_REFERENCED, it is time
457 * to remove it from the dcache, provided the super block is
458 * NULL (which means we are trying to reclaim memory)
459 * or this dentry belongs to the same super block that
460 * we want to shrink.
461 */
462 /*
463 * If this dentry is for "my" filesystem, then I can prune it
464 * without taking the s_umount lock (I already hold it).
465 */
466 if (sb && dentry->d_sb == sb) {
467 prune_one_dentry(dentry);
468 continue;
469 }
470 /*
471 * ...otherwise we need to be sure this filesystem isn't being
472 * unmounted, otherwise we could race with
473 * generic_shutdown_super(), and end up holding a reference to
474 * an inode while the filesystem is unmounted.
475 * So we try to get s_umount, and make sure s_root isn't NULL.
476 * (Take a local copy of s_umount to avoid a use-after-free of
477 * `dentry').
478 */
479 s_umount = &dentry->d_sb->s_umount;
480 if (down_read_trylock(s_umount)) {
481 if (dentry->d_sb->s_root != NULL) {
482 prune_one_dentry(dentry);
483 up_read(s_umount);
484 continue;
485 }
486 up_read(s_umount);
487 }
488 spin_unlock(&dentry->d_lock);
Vasily Averin6eac3f92006-10-28 10:38:44 -0700489 /*
490 * Insert dentry at the head of the list as inserting at the
491 * tail leads to a cycle.
NeilBrown0feae5c2006-06-22 14:47:28 -0700492 */
Vasily Averin6eac3f92006-10-28 10:38:44 -0700493 list_add(&dentry->d_lru, &dentry_unused);
494 dentry_stat.nr_unused++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 }
496 spin_unlock(&dcache_lock);
497}
498
499/*
500 * Shrink the dcache for the specified super block.
501 * This allows us to unmount a device without disturbing
502 * the dcache for the other devices.
503 *
504 * This implementation makes just two traversals of the
505 * unused list. On the first pass we move the selected
506 * dentries to the most recent end, and on the second
507 * pass we free them. The second pass must restart after
508 * each dput(), but since the target dentries are all at
509 * the end, it's really just a single traversal.
510 */
511
512/**
513 * shrink_dcache_sb - shrink dcache for a superblock
514 * @sb: superblock
515 *
516 * Shrink the dcache for the specified super block. This
517 * is used to free the dcache before unmounting a file
518 * system
519 */
520
521void shrink_dcache_sb(struct super_block * sb)
522{
523 struct list_head *tmp, *next;
524 struct dentry *dentry;
525
526 /*
527 * Pass one ... move the dentries for the specified
528 * superblock to the most recent end of the unused list.
529 */
530 spin_lock(&dcache_lock);
Domen Puncer0cdca3f2005-09-10 00:27:07 -0700531 list_for_each_safe(tmp, next, &dentry_unused) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 dentry = list_entry(tmp, struct dentry, d_lru);
533 if (dentry->d_sb != sb)
534 continue;
Akinobu Mita1bfba4e2006-06-26 00:24:40 -0700535 list_move(tmp, &dentry_unused);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 }
537
538 /*
539 * Pass two ... free the dentries for this superblock.
540 */
541repeat:
Domen Puncer0cdca3f2005-09-10 00:27:07 -0700542 list_for_each_safe(tmp, next, &dentry_unused) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 dentry = list_entry(tmp, struct dentry, d_lru);
544 if (dentry->d_sb != sb)
545 continue;
546 dentry_stat.nr_unused--;
547 list_del_init(tmp);
548 spin_lock(&dentry->d_lock);
549 if (atomic_read(&dentry->d_count)) {
550 spin_unlock(&dentry->d_lock);
551 continue;
552 }
553 prune_one_dentry(dentry);
Kirill Korotaev2ab13462006-03-25 03:07:45 -0800554 cond_resched_lock(&dcache_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 goto repeat;
556 }
557 spin_unlock(&dcache_lock);
558}
559
560/*
David Howellsc636ebd2006-10-11 01:22:19 -0700561 * destroy a single subtree of dentries for unmount
562 * - see the comments on shrink_dcache_for_umount() for a description of the
563 * locking
564 */
565static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
566{
567 struct dentry *parent;
David Howellsf8713572006-10-28 10:38:46 -0700568 unsigned detached = 0;
David Howellsc636ebd2006-10-11 01:22:19 -0700569
570 BUG_ON(!IS_ROOT(dentry));
571
572 /* detach this root from the system */
573 spin_lock(&dcache_lock);
574 if (!list_empty(&dentry->d_lru)) {
575 dentry_stat.nr_unused--;
576 list_del_init(&dentry->d_lru);
577 }
578 __d_drop(dentry);
579 spin_unlock(&dcache_lock);
580
581 for (;;) {
582 /* descend to the first leaf in the current subtree */
583 while (!list_empty(&dentry->d_subdirs)) {
584 struct dentry *loop;
585
586 /* this is a branch with children - detach all of them
587 * from the system in one go */
588 spin_lock(&dcache_lock);
589 list_for_each_entry(loop, &dentry->d_subdirs,
590 d_u.d_child) {
591 if (!list_empty(&loop->d_lru)) {
592 dentry_stat.nr_unused--;
593 list_del_init(&loop->d_lru);
594 }
595
596 __d_drop(loop);
597 cond_resched_lock(&dcache_lock);
598 }
599 spin_unlock(&dcache_lock);
600
601 /* move to the first child */
602 dentry = list_entry(dentry->d_subdirs.next,
603 struct dentry, d_u.d_child);
604 }
605
606 /* consume the dentries from this leaf up through its parents
607 * until we find one with children or run out altogether */
608 do {
609 struct inode *inode;
610
611 if (atomic_read(&dentry->d_count) != 0) {
612 printk(KERN_ERR
613 "BUG: Dentry %p{i=%lx,n=%s}"
614 " still in use (%d)"
615 " [unmount of %s %s]\n",
616 dentry,
617 dentry->d_inode ?
618 dentry->d_inode->i_ino : 0UL,
619 dentry->d_name.name,
620 atomic_read(&dentry->d_count),
621 dentry->d_sb->s_type->name,
622 dentry->d_sb->s_id);
623 BUG();
624 }
625
626 parent = dentry->d_parent;
627 if (parent == dentry)
628 parent = NULL;
629 else
630 atomic_dec(&parent->d_count);
631
632 list_del(&dentry->d_u.d_child);
David Howellsf8713572006-10-28 10:38:46 -0700633 detached++;
David Howellsc636ebd2006-10-11 01:22:19 -0700634
635 inode = dentry->d_inode;
636 if (inode) {
637 dentry->d_inode = NULL;
638 list_del_init(&dentry->d_alias);
639 if (dentry->d_op && dentry->d_op->d_iput)
640 dentry->d_op->d_iput(dentry, inode);
641 else
642 iput(inode);
643 }
644
645 d_free(dentry);
646
647 /* finished when we fall off the top of the tree,
648 * otherwise we ascend to the parent and move to the
649 * next sibling if there is one */
650 if (!parent)
David Howellsf8713572006-10-28 10:38:46 -0700651 goto out;
David Howellsc636ebd2006-10-11 01:22:19 -0700652
653 dentry = parent;
654
655 } while (list_empty(&dentry->d_subdirs));
656
657 dentry = list_entry(dentry->d_subdirs.next,
658 struct dentry, d_u.d_child);
659 }
David Howellsf8713572006-10-28 10:38:46 -0700660out:
661 /* several dentries were freed, need to correct nr_dentry */
662 spin_lock(&dcache_lock);
663 dentry_stat.nr_dentry -= detached;
664 spin_unlock(&dcache_lock);
David Howellsc636ebd2006-10-11 01:22:19 -0700665}
666
667/*
668 * destroy the dentries attached to a superblock on unmounting
669 * - we don't need to use dentry->d_lock, and only need dcache_lock when
670 * removing the dentry from the system lists and hashes because:
671 * - the superblock is detached from all mountings and open files, so the
672 * dentry trees will not be rearranged by the VFS
673 * - s_umount is write-locked, so the memory pressure shrinker will ignore
674 * any dentries belonging to this superblock that it comes across
675 * - the filesystem itself is no longer permitted to rearrange the dentries
676 * in this superblock
677 */
678void shrink_dcache_for_umount(struct super_block *sb)
679{
680 struct dentry *dentry;
681
682 if (down_read_trylock(&sb->s_umount))
683 BUG();
684
685 dentry = sb->s_root;
686 sb->s_root = NULL;
687 atomic_dec(&dentry->d_count);
688 shrink_dcache_for_umount_subtree(dentry);
689
690 while (!hlist_empty(&sb->s_anon)) {
691 dentry = hlist_entry(sb->s_anon.first, struct dentry, d_hash);
692 shrink_dcache_for_umount_subtree(dentry);
693 }
694}
695
696/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697 * Search for at least 1 mount point in the dentry's subdirs.
698 * We descend to the next level whenever the d_subdirs
699 * list is non-empty and continue searching.
700 */
701
702/**
703 * have_submounts - check for mounts over a dentry
704 * @parent: dentry to check.
705 *
706 * Return true if the parent or its subdirectories contain
707 * a mount point
708 */
709
710int have_submounts(struct dentry *parent)
711{
712 struct dentry *this_parent = parent;
713 struct list_head *next;
714
715 spin_lock(&dcache_lock);
716 if (d_mountpoint(parent))
717 goto positive;
718repeat:
719 next = this_parent->d_subdirs.next;
720resume:
721 while (next != &this_parent->d_subdirs) {
722 struct list_head *tmp = next;
Eric Dumazet5160ee62006-01-08 01:03:32 -0800723 struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 next = tmp->next;
725 /* Have we found a mount point ? */
726 if (d_mountpoint(dentry))
727 goto positive;
728 if (!list_empty(&dentry->d_subdirs)) {
729 this_parent = dentry;
730 goto repeat;
731 }
732 }
733 /*
734 * All done at this level ... ascend and resume the search.
735 */
736 if (this_parent != parent) {
Eric Dumazet5160ee62006-01-08 01:03:32 -0800737 next = this_parent->d_u.d_child.next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 this_parent = this_parent->d_parent;
739 goto resume;
740 }
741 spin_unlock(&dcache_lock);
742 return 0; /* No mount points found in tree */
743positive:
744 spin_unlock(&dcache_lock);
745 return 1;
746}
747
748/*
749 * Search the dentry child list for the specified parent,
750 * and move any unused dentries to the end of the unused
751 * list for prune_dcache(). We descend to the next level
752 * whenever the d_subdirs list is non-empty and continue
753 * searching.
754 *
755 * It returns zero iff there are no unused children,
756 * otherwise it returns the number of children moved to
757 * the end of the unused list. This may not be the total
758 * number of unused children, because select_parent can
759 * drop the lock and return early due to latency
760 * constraints.
761 */
762static int select_parent(struct dentry * parent)
763{
764 struct dentry *this_parent = parent;
765 struct list_head *next;
766 int found = 0;
767
768 spin_lock(&dcache_lock);
769repeat:
770 next = this_parent->d_subdirs.next;
771resume:
772 while (next != &this_parent->d_subdirs) {
773 struct list_head *tmp = next;
Eric Dumazet5160ee62006-01-08 01:03:32 -0800774 struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 next = tmp->next;
776
777 if (!list_empty(&dentry->d_lru)) {
778 dentry_stat.nr_unused--;
779 list_del_init(&dentry->d_lru);
780 }
781 /*
782 * move only zero ref count dentries to the end
783 * of the unused list for prune_dcache
784 */
785 if (!atomic_read(&dentry->d_count)) {
Akinobu Mita8e130592006-06-26 00:24:37 -0700786 list_add_tail(&dentry->d_lru, &dentry_unused);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 dentry_stat.nr_unused++;
788 found++;
789 }
790
791 /*
792 * We can return to the caller if we have found some (this
793 * ensures forward progress). We'll be coming back to find
794 * the rest.
795 */
796 if (found && need_resched())
797 goto out;
798
799 /*
800 * Descend a level if the d_subdirs list is non-empty.
801 */
802 if (!list_empty(&dentry->d_subdirs)) {
803 this_parent = dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 goto repeat;
805 }
806 }
807 /*
808 * All done at this level ... ascend and resume the search.
809 */
810 if (this_parent != parent) {
Eric Dumazet5160ee62006-01-08 01:03:32 -0800811 next = this_parent->d_u.d_child.next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 this_parent = this_parent->d_parent;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 goto resume;
814 }
815out:
816 spin_unlock(&dcache_lock);
817 return found;
818}
819
820/**
821 * shrink_dcache_parent - prune dcache
822 * @parent: parent of entries to prune
823 *
824 * Prune the dcache to remove unused children of the parent dentry.
825 */
826
827void shrink_dcache_parent(struct dentry * parent)
828{
829 int found;
830
831 while ((found = select_parent(parent)) != 0)
NeilBrown0feae5c2006-06-22 14:47:28 -0700832 prune_dcache(found, parent->d_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833}
834
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835/*
836 * Scan `nr' dentries and return the number which remain.
837 *
838 * We need to avoid reentering the filesystem if the caller is performing a
839 * GFP_NOFS allocation attempt. One example deadlock is:
840 *
841 * ext2_new_block->getblk->GFP->shrink_dcache_memory->prune_dcache->
842 * prune_one_dentry->dput->dentry_iput->iput->inode->i_sb->s_op->put_inode->
843 * ext2_discard_prealloc->ext2_free_blocks->lock_super->DEADLOCK.
844 *
845 * In this case we return -1 to tell the caller that we baled.
846 */
Al Viro27496a82005-10-21 03:20:48 -0400847static int shrink_dcache_memory(int nr, gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848{
849 if (nr) {
850 if (!(gfp_mask & __GFP_FS))
851 return -1;
NeilBrown0feae5c2006-06-22 14:47:28 -0700852 prune_dcache(nr, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 }
854 return (dentry_stat.nr_unused / 100) * sysctl_vfs_cache_pressure;
855}
856
857/**
858 * d_alloc - allocate a dcache entry
859 * @parent: parent of entry to allocate
860 * @name: qstr of the name
861 *
862 * Allocates a dentry. It returns %NULL if there is insufficient memory
863 * available. On a success the dentry is returned. The name passed in is
864 * copied and the copy passed in may be reused after this call.
865 */
866
867struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
868{
869 struct dentry *dentry;
870 char *dname;
871
872 dentry = kmem_cache_alloc(dentry_cache, GFP_KERNEL);
873 if (!dentry)
874 return NULL;
875
876 if (name->len > DNAME_INLINE_LEN-1) {
877 dname = kmalloc(name->len + 1, GFP_KERNEL);
878 if (!dname) {
879 kmem_cache_free(dentry_cache, dentry);
880 return NULL;
881 }
882 } else {
883 dname = dentry->d_iname;
884 }
885 dentry->d_name.name = dname;
886
887 dentry->d_name.len = name->len;
888 dentry->d_name.hash = name->hash;
889 memcpy(dname, name->name, name->len);
890 dname[name->len] = 0;
891
892 atomic_set(&dentry->d_count, 1);
893 dentry->d_flags = DCACHE_UNHASHED;
894 spin_lock_init(&dentry->d_lock);
895 dentry->d_inode = NULL;
896 dentry->d_parent = NULL;
897 dentry->d_sb = NULL;
898 dentry->d_op = NULL;
899 dentry->d_fsdata = NULL;
900 dentry->d_mounted = 0;
Marcelo Tosatti47ba87e2006-02-03 03:04:06 -0800901#ifdef CONFIG_PROFILING
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 dentry->d_cookie = NULL;
Marcelo Tosatti47ba87e2006-02-03 03:04:06 -0800903#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 INIT_HLIST_NODE(&dentry->d_hash);
905 INIT_LIST_HEAD(&dentry->d_lru);
906 INIT_LIST_HEAD(&dentry->d_subdirs);
907 INIT_LIST_HEAD(&dentry->d_alias);
908
909 if (parent) {
910 dentry->d_parent = dget(parent);
911 dentry->d_sb = parent->d_sb;
912 } else {
Eric Dumazet5160ee62006-01-08 01:03:32 -0800913 INIT_LIST_HEAD(&dentry->d_u.d_child);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 }
915
916 spin_lock(&dcache_lock);
917 if (parent)
Eric Dumazet5160ee62006-01-08 01:03:32 -0800918 list_add(&dentry->d_u.d_child, &parent->d_subdirs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 dentry_stat.nr_dentry++;
920 spin_unlock(&dcache_lock);
921
922 return dentry;
923}
924
925struct dentry *d_alloc_name(struct dentry *parent, const char *name)
926{
927 struct qstr q;
928
929 q.name = name;
930 q.len = strlen(name);
931 q.hash = full_name_hash(q.name, q.len);
932 return d_alloc(parent, &q);
933}
934
935/**
936 * d_instantiate - fill in inode information for a dentry
937 * @entry: dentry to complete
938 * @inode: inode to attach to this dentry
939 *
940 * Fill in inode information in the entry.
941 *
942 * This turns negative dentries into productive full members
943 * of society.
944 *
945 * NOTE! This assumes that the inode count has been incremented
946 * (or otherwise set) by the caller to indicate that it is now
947 * in use by the dcache.
948 */
949
950void d_instantiate(struct dentry *entry, struct inode * inode)
951{
Eric Sesterhenn28133c72006-03-26 18:25:39 +0200952 BUG_ON(!list_empty(&entry->d_alias));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953 spin_lock(&dcache_lock);
954 if (inode)
955 list_add(&entry->d_alias, &inode->i_dentry);
956 entry->d_inode = inode;
Nick Pigginc32ccd82006-03-25 03:07:09 -0800957 fsnotify_d_instantiate(entry, inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 spin_unlock(&dcache_lock);
959 security_d_instantiate(entry, inode);
960}
961
962/**
963 * d_instantiate_unique - instantiate a non-aliased dentry
964 * @entry: dentry to instantiate
965 * @inode: inode to attach to this dentry
966 *
967 * Fill in inode information in the entry. On success, it returns NULL.
968 * If an unhashed alias of "entry" already exists, then we return the
Oleg Drokine866cfa2006-01-09 20:52:51 -0800969 * aliased dentry instead and drop one reference to inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 *
971 * Note that in order to avoid conflicts with rename() etc, the caller
972 * had better be holding the parent directory semaphore.
Oleg Drokine866cfa2006-01-09 20:52:51 -0800973 *
974 * This also assumes that the inode count has been incremented
975 * (or otherwise set) by the caller to indicate that it is now
976 * in use by the dcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 */
David Howells770bfad2006-08-22 20:06:07 -0400978static struct dentry *__d_instantiate_unique(struct dentry *entry,
979 struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980{
981 struct dentry *alias;
982 int len = entry->d_name.len;
983 const char *name = entry->d_name.name;
984 unsigned int hash = entry->d_name.hash;
985
David Howells770bfad2006-08-22 20:06:07 -0400986 if (!inode) {
987 entry->d_inode = NULL;
988 return NULL;
989 }
990
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 list_for_each_entry(alias, &inode->i_dentry, d_alias) {
992 struct qstr *qstr = &alias->d_name;
993
994 if (qstr->hash != hash)
995 continue;
996 if (alias->d_parent != entry->d_parent)
997 continue;
998 if (qstr->len != len)
999 continue;
1000 if (memcmp(qstr->name, name, len))
1001 continue;
1002 dget_locked(alias);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 return alias;
1004 }
David Howells770bfad2006-08-22 20:06:07 -04001005
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006 list_add(&entry->d_alias, &inode->i_dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 entry->d_inode = inode;
Nick Pigginc32ccd82006-03-25 03:07:09 -08001008 fsnotify_d_instantiate(entry, inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 return NULL;
1010}
David Howells770bfad2006-08-22 20:06:07 -04001011
1012struct dentry *d_instantiate_unique(struct dentry *entry, struct inode *inode)
1013{
1014 struct dentry *result;
1015
1016 BUG_ON(!list_empty(&entry->d_alias));
1017
1018 spin_lock(&dcache_lock);
1019 result = __d_instantiate_unique(entry, inode);
1020 spin_unlock(&dcache_lock);
1021
1022 if (!result) {
1023 security_d_instantiate(entry, inode);
1024 return NULL;
1025 }
1026
1027 BUG_ON(!d_unhashed(result));
1028 iput(inode);
1029 return result;
1030}
1031
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032EXPORT_SYMBOL(d_instantiate_unique);
1033
1034/**
1035 * d_alloc_root - allocate root dentry
1036 * @root_inode: inode to allocate the root for
1037 *
1038 * Allocate a root ("/") dentry for the inode given. The inode is
1039 * instantiated and returned. %NULL is returned if there is insufficient
1040 * memory or the inode passed is %NULL.
1041 */
1042
1043struct dentry * d_alloc_root(struct inode * root_inode)
1044{
1045 struct dentry *res = NULL;
1046
1047 if (root_inode) {
1048 static const struct qstr name = { .name = "/", .len = 1 };
1049
1050 res = d_alloc(NULL, &name);
1051 if (res) {
1052 res->d_sb = root_inode->i_sb;
1053 res->d_parent = res;
1054 d_instantiate(res, root_inode);
1055 }
1056 }
1057 return res;
1058}
1059
1060static inline struct hlist_head *d_hash(struct dentry *parent,
1061 unsigned long hash)
1062{
1063 hash += ((unsigned long) parent ^ GOLDEN_RATIO_PRIME) / L1_CACHE_BYTES;
1064 hash = hash ^ ((hash ^ GOLDEN_RATIO_PRIME) >> D_HASHBITS);
1065 return dentry_hashtable + (hash & D_HASHMASK);
1066}
1067
1068/**
1069 * d_alloc_anon - allocate an anonymous dentry
1070 * @inode: inode to allocate the dentry for
1071 *
1072 * This is similar to d_alloc_root. It is used by filesystems when
1073 * creating a dentry for a given inode, often in the process of
1074 * mapping a filehandle to a dentry. The returned dentry may be
1075 * anonymous, or may have a full name (if the inode was already
1076 * in the cache). The file system may need to make further
1077 * efforts to connect this dentry into the dcache properly.
1078 *
1079 * When called on a directory inode, we must ensure that
1080 * the inode only ever has one dentry. If a dentry is
1081 * found, that is returned instead of allocating a new one.
1082 *
1083 * On successful return, the reference to the inode has been transferred
1084 * to the dentry. If %NULL is returned (indicating kmalloc failure),
1085 * the reference on the inode has not been released.
1086 */
1087
1088struct dentry * d_alloc_anon(struct inode *inode)
1089{
1090 static const struct qstr anonstring = { .name = "" };
1091 struct dentry *tmp;
1092 struct dentry *res;
1093
1094 if ((res = d_find_alias(inode))) {
1095 iput(inode);
1096 return res;
1097 }
1098
1099 tmp = d_alloc(NULL, &anonstring);
1100 if (!tmp)
1101 return NULL;
1102
1103 tmp->d_parent = tmp; /* make sure dput doesn't croak */
1104
1105 spin_lock(&dcache_lock);
1106 res = __d_find_alias(inode, 0);
1107 if (!res) {
1108 /* attach a disconnected dentry */
1109 res = tmp;
1110 tmp = NULL;
1111 spin_lock(&res->d_lock);
1112 res->d_sb = inode->i_sb;
1113 res->d_parent = res;
1114 res->d_inode = inode;
1115 res->d_flags |= DCACHE_DISCONNECTED;
1116 res->d_flags &= ~DCACHE_UNHASHED;
1117 list_add(&res->d_alias, &inode->i_dentry);
1118 hlist_add_head(&res->d_hash, &inode->i_sb->s_anon);
1119 spin_unlock(&res->d_lock);
1120
1121 inode = NULL; /* don't drop reference */
1122 }
1123 spin_unlock(&dcache_lock);
1124
1125 if (inode)
1126 iput(inode);
1127 if (tmp)
1128 dput(tmp);
1129 return res;
1130}
1131
1132
1133/**
1134 * d_splice_alias - splice a disconnected dentry into the tree if one exists
1135 * @inode: the inode which may have a disconnected dentry
1136 * @dentry: a negative dentry which we want to point to the inode.
1137 *
1138 * If inode is a directory and has a 'disconnected' dentry (i.e. IS_ROOT and
1139 * DCACHE_DISCONNECTED), then d_move that in place of the given dentry
1140 * and return it, else simply d_add the inode to the dentry and return NULL.
1141 *
1142 * This is needed in the lookup routine of any filesystem that is exportable
1143 * (via knfsd) so that we can build dcache paths to directories effectively.
1144 *
1145 * If a dentry was found and moved, then it is returned. Otherwise NULL
1146 * is returned. This matches the expected return value of ->lookup.
1147 *
1148 */
1149struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
1150{
1151 struct dentry *new = NULL;
1152
NeilBrown21c0d8f2006-10-04 02:16:16 -07001153 if (inode && S_ISDIR(inode->i_mode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154 spin_lock(&dcache_lock);
1155 new = __d_find_alias(inode, 1);
1156 if (new) {
1157 BUG_ON(!(new->d_flags & DCACHE_DISCONNECTED));
Nick Pigginc32ccd82006-03-25 03:07:09 -08001158 fsnotify_d_instantiate(new, inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 spin_unlock(&dcache_lock);
1160 security_d_instantiate(new, inode);
1161 d_rehash(dentry);
1162 d_move(new, dentry);
1163 iput(inode);
1164 } else {
1165 /* d_instantiate takes dcache_lock, so we do it by hand */
1166 list_add(&dentry->d_alias, &inode->i_dentry);
1167 dentry->d_inode = inode;
Nick Pigginc32ccd82006-03-25 03:07:09 -08001168 fsnotify_d_instantiate(dentry, inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169 spin_unlock(&dcache_lock);
1170 security_d_instantiate(dentry, inode);
1171 d_rehash(dentry);
1172 }
1173 } else
1174 d_add(dentry, inode);
1175 return new;
1176}
1177
1178
1179/**
1180 * d_lookup - search for a dentry
1181 * @parent: parent dentry
1182 * @name: qstr of name we wish to find
1183 *
1184 * Searches the children of the parent dentry for the name in question. If
1185 * the dentry is found its reference count is incremented and the dentry
1186 * is returned. The caller must use d_put to free the entry when it has
1187 * finished using it. %NULL is returned on failure.
1188 *
1189 * __d_lookup is dcache_lock free. The hash list is protected using RCU.
1190 * Memory barriers are used while updating and doing lockless traversal.
1191 * To avoid races with d_move while rename is happening, d_lock is used.
1192 *
1193 * Overflows in memcmp(), while d_move, are avoided by keeping the length
1194 * and name pointer in one structure pointed by d_qstr.
1195 *
1196 * rcu_read_lock() and rcu_read_unlock() are used to disable preemption while
1197 * lookup is going on.
1198 *
1199 * dentry_unused list is not updated even if lookup finds the required dentry
1200 * in there. It is updated in places such as prune_dcache, shrink_dcache_sb,
1201 * select_parent and __dget_locked. This laziness saves lookup from dcache_lock
1202 * acquisition.
1203 *
1204 * d_lookup() is protected against the concurrent renames in some unrelated
1205 * directory using the seqlockt_t rename_lock.
1206 */
1207
1208struct dentry * d_lookup(struct dentry * parent, struct qstr * name)
1209{
1210 struct dentry * dentry = NULL;
1211 unsigned long seq;
1212
1213 do {
1214 seq = read_seqbegin(&rename_lock);
1215 dentry = __d_lookup(parent, name);
1216 if (dentry)
1217 break;
1218 } while (read_seqretry(&rename_lock, seq));
1219 return dentry;
1220}
1221
1222struct dentry * __d_lookup(struct dentry * parent, struct qstr * name)
1223{
1224 unsigned int len = name->len;
1225 unsigned int hash = name->hash;
1226 const unsigned char *str = name->name;
1227 struct hlist_head *head = d_hash(parent,hash);
1228 struct dentry *found = NULL;
1229 struct hlist_node *node;
Paul E. McKenney665a7582005-11-07 00:59:17 -08001230 struct dentry *dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231
1232 rcu_read_lock();
1233
Paul E. McKenney665a7582005-11-07 00:59:17 -08001234 hlist_for_each_entry_rcu(dentry, node, head, d_hash) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 struct qstr *qstr;
1236
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 if (dentry->d_name.hash != hash)
1238 continue;
1239 if (dentry->d_parent != parent)
1240 continue;
1241
1242 spin_lock(&dentry->d_lock);
1243
1244 /*
1245 * Recheck the dentry after taking the lock - d_move may have
1246 * changed things. Don't bother checking the hash because we're
1247 * about to compare the whole name anyway.
1248 */
1249 if (dentry->d_parent != parent)
1250 goto next;
1251
1252 /*
1253 * It is safe to compare names since d_move() cannot
1254 * change the qstr (protected by d_lock).
1255 */
1256 qstr = &dentry->d_name;
1257 if (parent->d_op && parent->d_op->d_compare) {
1258 if (parent->d_op->d_compare(parent, qstr, name))
1259 goto next;
1260 } else {
1261 if (qstr->len != len)
1262 goto next;
1263 if (memcmp(qstr->name, str, len))
1264 goto next;
1265 }
1266
1267 if (!d_unhashed(dentry)) {
1268 atomic_inc(&dentry->d_count);
1269 found = dentry;
1270 }
1271 spin_unlock(&dentry->d_lock);
1272 break;
1273next:
1274 spin_unlock(&dentry->d_lock);
1275 }
1276 rcu_read_unlock();
1277
1278 return found;
1279}
1280
1281/**
Eric W. Biederman3e7e2412006-03-31 02:31:43 -08001282 * d_hash_and_lookup - hash the qstr then search for a dentry
1283 * @dir: Directory to search in
1284 * @name: qstr of name we wish to find
1285 *
1286 * On hash failure or on lookup failure NULL is returned.
1287 */
1288struct dentry *d_hash_and_lookup(struct dentry *dir, struct qstr *name)
1289{
1290 struct dentry *dentry = NULL;
1291
1292 /*
1293 * Check for a fs-specific hash function. Note that we must
1294 * calculate the standard hash first, as the d_op->d_hash()
1295 * routine may choose to leave the hash value unchanged.
1296 */
1297 name->hash = full_name_hash(name->name, name->len);
1298 if (dir->d_op && dir->d_op->d_hash) {
1299 if (dir->d_op->d_hash(dir, name) < 0)
1300 goto out;
1301 }
1302 dentry = d_lookup(dir, name);
1303out:
1304 return dentry;
1305}
1306
1307/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308 * d_validate - verify dentry provided from insecure source
1309 * @dentry: The dentry alleged to be valid child of @dparent
1310 * @dparent: The parent dentry (known to be valid)
1311 * @hash: Hash of the dentry
1312 * @len: Length of the name
1313 *
1314 * An insecure source has sent us a dentry, here we verify it and dget() it.
1315 * This is used by ncpfs in its readdir implementation.
1316 * Zero is returned in the dentry is invalid.
1317 */
1318
1319int d_validate(struct dentry *dentry, struct dentry *dparent)
1320{
1321 struct hlist_head *base;
1322 struct hlist_node *lhp;
1323
1324 /* Check whether the ptr might be valid at all.. */
1325 if (!kmem_ptr_validate(dentry_cache, dentry))
1326 goto out;
1327
1328 if (dentry->d_parent != dparent)
1329 goto out;
1330
1331 spin_lock(&dcache_lock);
1332 base = d_hash(dparent, dentry->d_name.hash);
1333 hlist_for_each(lhp,base) {
Paul E. McKenney665a7582005-11-07 00:59:17 -08001334 /* hlist_for_each_entry_rcu() not required for d_hash list
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 * as it is parsed under dcache_lock
1336 */
1337 if (dentry == hlist_entry(lhp, struct dentry, d_hash)) {
1338 __dget_locked(dentry);
1339 spin_unlock(&dcache_lock);
1340 return 1;
1341 }
1342 }
1343 spin_unlock(&dcache_lock);
1344out:
1345 return 0;
1346}
1347
1348/*
1349 * When a file is deleted, we have two options:
1350 * - turn this dentry into a negative dentry
1351 * - unhash this dentry and free it.
1352 *
1353 * Usually, we want to just turn this into
1354 * a negative dentry, but if anybody else is
1355 * currently using the dentry or the inode
1356 * we can't do that and we fall back on removing
1357 * it from the hash queues and waiting for
1358 * it to be deleted later when it has no users
1359 */
1360
1361/**
1362 * d_delete - delete a dentry
1363 * @dentry: The dentry to delete
1364 *
1365 * Turn the dentry into a negative dentry if possible, otherwise
1366 * remove it from the hash queues so it can be deleted later
1367 */
1368
1369void d_delete(struct dentry * dentry)
1370{
John McCutchan7a91bf72005-08-08 13:52:16 -04001371 int isdir = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 /*
1373 * Are we the only user?
1374 */
1375 spin_lock(&dcache_lock);
1376 spin_lock(&dentry->d_lock);
John McCutchan7a91bf72005-08-08 13:52:16 -04001377 isdir = S_ISDIR(dentry->d_inode->i_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378 if (atomic_read(&dentry->d_count) == 1) {
1379 dentry_iput(dentry);
John McCutchan7a91bf72005-08-08 13:52:16 -04001380 fsnotify_nameremove(dentry, isdir);
Amy Griffis7a2bd3f2006-03-31 02:30:54 -08001381
1382 /* remove this and other inotify debug checks after 2.6.18 */
1383 dentry->d_flags &= ~DCACHE_INOTIFY_PARENT_WATCHED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384 return;
1385 }
1386
1387 if (!d_unhashed(dentry))
1388 __d_drop(dentry);
1389
1390 spin_unlock(&dentry->d_lock);
1391 spin_unlock(&dcache_lock);
John McCutchan7a91bf72005-08-08 13:52:16 -04001392
1393 fsnotify_nameremove(dentry, isdir);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394}
1395
1396static void __d_rehash(struct dentry * entry, struct hlist_head *list)
1397{
1398
1399 entry->d_flags &= ~DCACHE_UNHASHED;
1400 hlist_add_head_rcu(&entry->d_hash, list);
1401}
1402
David Howells770bfad2006-08-22 20:06:07 -04001403static void _d_rehash(struct dentry * entry)
1404{
1405 __d_rehash(entry, d_hash(entry->d_parent, entry->d_name.hash));
1406}
1407
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408/**
1409 * d_rehash - add an entry back to the hash
1410 * @entry: dentry to add to the hash
1411 *
1412 * Adds a dentry to the hash according to its name.
1413 */
1414
1415void d_rehash(struct dentry * entry)
1416{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417 spin_lock(&dcache_lock);
1418 spin_lock(&entry->d_lock);
David Howells770bfad2006-08-22 20:06:07 -04001419 _d_rehash(entry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 spin_unlock(&entry->d_lock);
1421 spin_unlock(&dcache_lock);
1422}
1423
1424#define do_switch(x,y) do { \
1425 __typeof__ (x) __tmp = x; \
1426 x = y; y = __tmp; } while (0)
1427
1428/*
1429 * When switching names, the actual string doesn't strictly have to
1430 * be preserved in the target - because we're dropping the target
1431 * anyway. As such, we can just do a simple memcpy() to copy over
1432 * the new name before we switch.
1433 *
1434 * Note that we have to be a lot more careful about getting the hash
1435 * switched - we have to switch the hash value properly even if it
1436 * then no longer matches the actual (corrupted) string of the target.
1437 * The hash value has to match the hash queue that the dentry is on..
1438 */
1439static void switch_names(struct dentry *dentry, struct dentry *target)
1440{
1441 if (dname_external(target)) {
1442 if (dname_external(dentry)) {
1443 /*
1444 * Both external: swap the pointers
1445 */
1446 do_switch(target->d_name.name, dentry->d_name.name);
1447 } else {
1448 /*
1449 * dentry:internal, target:external. Steal target's
1450 * storage and make target internal.
1451 */
1452 dentry->d_name.name = target->d_name.name;
1453 target->d_name.name = target->d_iname;
1454 }
1455 } else {
1456 if (dname_external(dentry)) {
1457 /*
1458 * dentry:external, target:internal. Give dentry's
1459 * storage to target and make dentry internal
1460 */
1461 memcpy(dentry->d_iname, target->d_name.name,
1462 target->d_name.len + 1);
1463 target->d_name.name = dentry->d_name.name;
1464 dentry->d_name.name = dentry->d_iname;
1465 } else {
1466 /*
1467 * Both are internal. Just copy target to dentry
1468 */
1469 memcpy(dentry->d_iname, target->d_name.name,
1470 target->d_name.len + 1);
1471 }
1472 }
1473}
1474
1475/*
1476 * We cannibalize "target" when moving dentry on top of it,
1477 * because it's going to be thrown away anyway. We could be more
1478 * polite about it, though.
1479 *
1480 * This forceful removal will result in ugly /proc output if
1481 * somebody holds a file open that got deleted due to a rename.
1482 * We could be nicer about the deleted file, and let it show
1483 * up under the name it got deleted rather than the name that
1484 * deleted it.
1485 */
1486
Trond Myklebust9eaef272006-10-21 10:24:20 -07001487/*
1488 * d_move_locked - move a dentry
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489 * @dentry: entry to move
1490 * @target: new dentry
1491 *
1492 * Update the dcache to reflect the move of a file name. Negative
1493 * dcache entries should not be moved in this way.
1494 */
Trond Myklebust9eaef272006-10-21 10:24:20 -07001495static void d_move_locked(struct dentry * dentry, struct dentry * target)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496{
1497 struct hlist_head *list;
1498
1499 if (!dentry->d_inode)
1500 printk(KERN_WARNING "VFS: moving negative dcache entry\n");
1501
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502 write_seqlock(&rename_lock);
1503 /*
1504 * XXXX: do we really need to take target->d_lock?
1505 */
1506 if (target < dentry) {
1507 spin_lock(&target->d_lock);
Ingo Molnara90b9c02006-07-03 00:25:04 -07001508 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509 } else {
1510 spin_lock(&dentry->d_lock);
Ingo Molnara90b9c02006-07-03 00:25:04 -07001511 spin_lock_nested(&target->d_lock, DENTRY_D_LOCK_NESTED);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 }
1513
1514 /* Move the dentry to the target hash queue, if on different bucket */
1515 if (dentry->d_flags & DCACHE_UNHASHED)
1516 goto already_unhashed;
1517
1518 hlist_del_rcu(&dentry->d_hash);
1519
1520already_unhashed:
1521 list = d_hash(target->d_parent, target->d_name.hash);
1522 __d_rehash(dentry, list);
1523
1524 /* Unhash the target: dput() will then get rid of it */
1525 __d_drop(target);
1526
Eric Dumazet5160ee62006-01-08 01:03:32 -08001527 list_del(&dentry->d_u.d_child);
1528 list_del(&target->d_u.d_child);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529
1530 /* Switch the names.. */
1531 switch_names(dentry, target);
1532 do_switch(dentry->d_name.len, target->d_name.len);
1533 do_switch(dentry->d_name.hash, target->d_name.hash);
1534
1535 /* ... and switch the parents */
1536 if (IS_ROOT(dentry)) {
1537 dentry->d_parent = target->d_parent;
1538 target->d_parent = target;
Eric Dumazet5160ee62006-01-08 01:03:32 -08001539 INIT_LIST_HEAD(&target->d_u.d_child);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 } else {
1541 do_switch(dentry->d_parent, target->d_parent);
1542
1543 /* And add them back to the (new) parent lists */
Eric Dumazet5160ee62006-01-08 01:03:32 -08001544 list_add(&target->d_u.d_child, &target->d_parent->d_subdirs);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545 }
1546
Eric Dumazet5160ee62006-01-08 01:03:32 -08001547 list_add(&dentry->d_u.d_child, &dentry->d_parent->d_subdirs);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 spin_unlock(&target->d_lock);
Nick Pigginc32ccd82006-03-25 03:07:09 -08001549 fsnotify_d_move(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 spin_unlock(&dentry->d_lock);
1551 write_sequnlock(&rename_lock);
Trond Myklebust9eaef272006-10-21 10:24:20 -07001552}
1553
1554/**
1555 * d_move - move a dentry
1556 * @dentry: entry to move
1557 * @target: new dentry
1558 *
1559 * Update the dcache to reflect the move of a file name. Negative
1560 * dcache entries should not be moved in this way.
1561 */
1562
1563void d_move(struct dentry * dentry, struct dentry * target)
1564{
1565 spin_lock(&dcache_lock);
1566 d_move_locked(dentry, target);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 spin_unlock(&dcache_lock);
1568}
1569
David Howells770bfad2006-08-22 20:06:07 -04001570/*
Trond Myklebust9eaef272006-10-21 10:24:20 -07001571 * Helper that returns 1 if p1 is a parent of p2, else 0
1572 */
1573static int d_isparent(struct dentry *p1, struct dentry *p2)
1574{
1575 struct dentry *p;
1576
1577 for (p = p2; p->d_parent != p; p = p->d_parent) {
1578 if (p->d_parent == p1)
1579 return 1;
1580 }
1581 return 0;
1582}
1583
1584/*
1585 * This helper attempts to cope with remotely renamed directories
1586 *
1587 * It assumes that the caller is already holding
1588 * dentry->d_parent->d_inode->i_mutex and the dcache_lock
1589 *
1590 * Note: If ever the locking in lock_rename() changes, then please
1591 * remember to update this too...
1592 *
1593 * On return, dcache_lock will have been unlocked.
1594 */
1595static struct dentry *__d_unalias(struct dentry *dentry, struct dentry *alias)
1596{
1597 struct mutex *m1 = NULL, *m2 = NULL;
1598 struct dentry *ret;
1599
1600 /* If alias and dentry share a parent, then no extra locks required */
1601 if (alias->d_parent == dentry->d_parent)
1602 goto out_unalias;
1603
1604 /* Check for loops */
1605 ret = ERR_PTR(-ELOOP);
1606 if (d_isparent(alias, dentry))
1607 goto out_err;
1608
1609 /* See lock_rename() */
1610 ret = ERR_PTR(-EBUSY);
1611 if (!mutex_trylock(&dentry->d_sb->s_vfs_rename_mutex))
1612 goto out_err;
1613 m1 = &dentry->d_sb->s_vfs_rename_mutex;
1614 if (!mutex_trylock(&alias->d_parent->d_inode->i_mutex))
1615 goto out_err;
1616 m2 = &alias->d_parent->d_inode->i_mutex;
1617out_unalias:
1618 d_move_locked(alias, dentry);
1619 ret = alias;
1620out_err:
1621 spin_unlock(&dcache_lock);
1622 if (m2)
1623 mutex_unlock(m2);
1624 if (m1)
1625 mutex_unlock(m1);
1626 return ret;
1627}
1628
1629/*
David Howells770bfad2006-08-22 20:06:07 -04001630 * Prepare an anonymous dentry for life in the superblock's dentry tree as a
1631 * named dentry in place of the dentry to be replaced.
1632 */
1633static void __d_materialise_dentry(struct dentry *dentry, struct dentry *anon)
1634{
1635 struct dentry *dparent, *aparent;
1636
1637 switch_names(dentry, anon);
1638 do_switch(dentry->d_name.len, anon->d_name.len);
1639 do_switch(dentry->d_name.hash, anon->d_name.hash);
1640
1641 dparent = dentry->d_parent;
1642 aparent = anon->d_parent;
1643
1644 dentry->d_parent = (aparent == anon) ? dentry : aparent;
1645 list_del(&dentry->d_u.d_child);
1646 if (!IS_ROOT(dentry))
1647 list_add(&dentry->d_u.d_child, &dentry->d_parent->d_subdirs);
1648 else
1649 INIT_LIST_HEAD(&dentry->d_u.d_child);
1650
1651 anon->d_parent = (dparent == dentry) ? anon : dparent;
1652 list_del(&anon->d_u.d_child);
1653 if (!IS_ROOT(anon))
1654 list_add(&anon->d_u.d_child, &anon->d_parent->d_subdirs);
1655 else
1656 INIT_LIST_HEAD(&anon->d_u.d_child);
1657
1658 anon->d_flags &= ~DCACHE_DISCONNECTED;
1659}
1660
1661/**
1662 * d_materialise_unique - introduce an inode into the tree
1663 * @dentry: candidate dentry
1664 * @inode: inode to bind to the dentry, to which aliases may be attached
1665 *
1666 * Introduces an dentry into the tree, substituting an extant disconnected
1667 * root directory alias in its place if there is one
1668 */
1669struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
1670{
Trond Myklebust9eaef272006-10-21 10:24:20 -07001671 struct dentry *actual;
David Howells770bfad2006-08-22 20:06:07 -04001672
1673 BUG_ON(!d_unhashed(dentry));
1674
1675 spin_lock(&dcache_lock);
1676
1677 if (!inode) {
1678 actual = dentry;
1679 dentry->d_inode = NULL;
1680 goto found_lock;
1681 }
1682
Trond Myklebust9eaef272006-10-21 10:24:20 -07001683 if (S_ISDIR(inode->i_mode)) {
1684 struct dentry *alias;
David Howells770bfad2006-08-22 20:06:07 -04001685
Trond Myklebust9eaef272006-10-21 10:24:20 -07001686 /* Does an aliased dentry already exist? */
1687 alias = __d_find_alias(inode, 0);
1688 if (alias) {
1689 actual = alias;
1690 /* Is this an anonymous mountpoint that we could splice
1691 * into our tree? */
1692 if (IS_ROOT(alias)) {
1693 spin_lock(&alias->d_lock);
1694 __d_materialise_dentry(dentry, alias);
1695 __d_drop(alias);
1696 goto found;
1697 }
1698 /* Nope, but we must(!) avoid directory aliasing */
1699 actual = __d_unalias(dentry, alias);
1700 if (IS_ERR(actual))
1701 dput(alias);
1702 goto out_nolock;
1703 }
David Howells770bfad2006-08-22 20:06:07 -04001704 }
1705
1706 /* Add a unique reference */
1707 actual = __d_instantiate_unique(dentry, inode);
1708 if (!actual)
1709 actual = dentry;
1710 else if (unlikely(!d_unhashed(actual)))
1711 goto shouldnt_be_hashed;
1712
1713found_lock:
1714 spin_lock(&actual->d_lock);
1715found:
1716 _d_rehash(actual);
1717 spin_unlock(&actual->d_lock);
1718 spin_unlock(&dcache_lock);
Trond Myklebust9eaef272006-10-21 10:24:20 -07001719out_nolock:
David Howells770bfad2006-08-22 20:06:07 -04001720 if (actual == dentry) {
1721 security_d_instantiate(dentry, inode);
1722 return NULL;
1723 }
1724
1725 iput(inode);
1726 return actual;
1727
David Howells770bfad2006-08-22 20:06:07 -04001728shouldnt_be_hashed:
1729 spin_unlock(&dcache_lock);
1730 BUG();
1731 goto shouldnt_be_hashed;
1732}
1733
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734/**
1735 * d_path - return the path of a dentry
1736 * @dentry: dentry to report
1737 * @vfsmnt: vfsmnt to which the dentry belongs
1738 * @root: root dentry
1739 * @rootmnt: vfsmnt to which the root dentry belongs
1740 * @buffer: buffer to return value in
1741 * @buflen: buffer length
1742 *
1743 * Convert a dentry into an ASCII path name. If the entry has been deleted
1744 * the string " (deleted)" is appended. Note that this is ambiguous.
1745 *
1746 * Returns the buffer or an error code if the path was too long.
1747 *
1748 * "buflen" should be positive. Caller holds the dcache_lock.
1749 */
1750static char * __d_path( struct dentry *dentry, struct vfsmount *vfsmnt,
1751 struct dentry *root, struct vfsmount *rootmnt,
1752 char *buffer, int buflen)
1753{
1754 char * end = buffer+buflen;
1755 char * retval;
1756 int namelen;
1757
1758 *--end = '\0';
1759 buflen--;
1760 if (!IS_ROOT(dentry) && d_unhashed(dentry)) {
1761 buflen -= 10;
1762 end -= 10;
1763 if (buflen < 0)
1764 goto Elong;
1765 memcpy(end, " (deleted)", 10);
1766 }
1767
1768 if (buflen < 1)
1769 goto Elong;
1770 /* Get '/' right */
1771 retval = end-1;
1772 *retval = '/';
1773
1774 for (;;) {
1775 struct dentry * parent;
1776
1777 if (dentry == root && vfsmnt == rootmnt)
1778 break;
1779 if (dentry == vfsmnt->mnt_root || IS_ROOT(dentry)) {
1780 /* Global root? */
1781 spin_lock(&vfsmount_lock);
1782 if (vfsmnt->mnt_parent == vfsmnt) {
1783 spin_unlock(&vfsmount_lock);
1784 goto global_root;
1785 }
1786 dentry = vfsmnt->mnt_mountpoint;
1787 vfsmnt = vfsmnt->mnt_parent;
1788 spin_unlock(&vfsmount_lock);
1789 continue;
1790 }
1791 parent = dentry->d_parent;
1792 prefetch(parent);
1793 namelen = dentry->d_name.len;
1794 buflen -= namelen + 1;
1795 if (buflen < 0)
1796 goto Elong;
1797 end -= namelen;
1798 memcpy(end, dentry->d_name.name, namelen);
1799 *--end = '/';
1800 retval = end;
1801 dentry = parent;
1802 }
1803
1804 return retval;
1805
1806global_root:
1807 namelen = dentry->d_name.len;
1808 buflen -= namelen;
1809 if (buflen < 0)
1810 goto Elong;
1811 retval -= namelen-1; /* hit the slash */
1812 memcpy(retval, dentry->d_name.name, namelen);
1813 return retval;
1814Elong:
1815 return ERR_PTR(-ENAMETOOLONG);
1816}
1817
1818/* write full pathname into buffer and return start of pathname */
1819char * d_path(struct dentry *dentry, struct vfsmount *vfsmnt,
1820 char *buf, int buflen)
1821{
1822 char *res;
1823 struct vfsmount *rootmnt;
1824 struct dentry *root;
1825
1826 read_lock(&current->fs->lock);
1827 rootmnt = mntget(current->fs->rootmnt);
1828 root = dget(current->fs->root);
1829 read_unlock(&current->fs->lock);
1830 spin_lock(&dcache_lock);
1831 res = __d_path(dentry, vfsmnt, root, rootmnt, buf, buflen);
1832 spin_unlock(&dcache_lock);
1833 dput(root);
1834 mntput(rootmnt);
1835 return res;
1836}
1837
1838/*
1839 * NOTE! The user-level library version returns a
1840 * character pointer. The kernel system call just
1841 * returns the length of the buffer filled (which
1842 * includes the ending '\0' character), or a negative
1843 * error value. So libc would do something like
1844 *
1845 * char *getcwd(char * buf, size_t size)
1846 * {
1847 * int retval;
1848 *
1849 * retval = sys_getcwd(buf, size);
1850 * if (retval >= 0)
1851 * return buf;
1852 * errno = -retval;
1853 * return NULL;
1854 * }
1855 */
1856asmlinkage long sys_getcwd(char __user *buf, unsigned long size)
1857{
1858 int error;
1859 struct vfsmount *pwdmnt, *rootmnt;
1860 struct dentry *pwd, *root;
1861 char *page = (char *) __get_free_page(GFP_USER);
1862
1863 if (!page)
1864 return -ENOMEM;
1865
1866 read_lock(&current->fs->lock);
1867 pwdmnt = mntget(current->fs->pwdmnt);
1868 pwd = dget(current->fs->pwd);
1869 rootmnt = mntget(current->fs->rootmnt);
1870 root = dget(current->fs->root);
1871 read_unlock(&current->fs->lock);
1872
1873 error = -ENOENT;
1874 /* Has the current directory has been unlinked? */
1875 spin_lock(&dcache_lock);
1876 if (pwd->d_parent == pwd || !d_unhashed(pwd)) {
1877 unsigned long len;
1878 char * cwd;
1879
1880 cwd = __d_path(pwd, pwdmnt, root, rootmnt, page, PAGE_SIZE);
1881 spin_unlock(&dcache_lock);
1882
1883 error = PTR_ERR(cwd);
1884 if (IS_ERR(cwd))
1885 goto out;
1886
1887 error = -ERANGE;
1888 len = PAGE_SIZE + page - cwd;
1889 if (len <= size) {
1890 error = len;
1891 if (copy_to_user(buf, cwd, len))
1892 error = -EFAULT;
1893 }
1894 } else
1895 spin_unlock(&dcache_lock);
1896
1897out:
1898 dput(pwd);
1899 mntput(pwdmnt);
1900 dput(root);
1901 mntput(rootmnt);
1902 free_page((unsigned long) page);
1903 return error;
1904}
1905
1906/*
1907 * Test whether new_dentry is a subdirectory of old_dentry.
1908 *
1909 * Trivially implemented using the dcache structure
1910 */
1911
1912/**
1913 * is_subdir - is new dentry a subdirectory of old_dentry
1914 * @new_dentry: new dentry
1915 * @old_dentry: old dentry
1916 *
1917 * Returns 1 if new_dentry is a subdirectory of the parent (at any depth).
1918 * Returns 0 otherwise.
1919 * Caller must ensure that "new_dentry" is pinned before calling is_subdir()
1920 */
1921
1922int is_subdir(struct dentry * new_dentry, struct dentry * old_dentry)
1923{
1924 int result;
1925 struct dentry * saved = new_dentry;
1926 unsigned long seq;
1927
1928 /* need rcu_readlock to protect against the d_parent trashing due to
1929 * d_move
1930 */
1931 rcu_read_lock();
1932 do {
1933 /* for restarting inner loop in case of seq retry */
1934 new_dentry = saved;
1935 result = 0;
1936 seq = read_seqbegin(&rename_lock);
1937 for (;;) {
1938 if (new_dentry != old_dentry) {
1939 struct dentry * parent = new_dentry->d_parent;
1940 if (parent == new_dentry)
1941 break;
1942 new_dentry = parent;
1943 continue;
1944 }
1945 result = 1;
1946 break;
1947 }
1948 } while (read_seqretry(&rename_lock, seq));
1949 rcu_read_unlock();
1950
1951 return result;
1952}
1953
1954void d_genocide(struct dentry *root)
1955{
1956 struct dentry *this_parent = root;
1957 struct list_head *next;
1958
1959 spin_lock(&dcache_lock);
1960repeat:
1961 next = this_parent->d_subdirs.next;
1962resume:
1963 while (next != &this_parent->d_subdirs) {
1964 struct list_head *tmp = next;
Eric Dumazet5160ee62006-01-08 01:03:32 -08001965 struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966 next = tmp->next;
1967 if (d_unhashed(dentry)||!dentry->d_inode)
1968 continue;
1969 if (!list_empty(&dentry->d_subdirs)) {
1970 this_parent = dentry;
1971 goto repeat;
1972 }
1973 atomic_dec(&dentry->d_count);
1974 }
1975 if (this_parent != root) {
Eric Dumazet5160ee62006-01-08 01:03:32 -08001976 next = this_parent->d_u.d_child.next;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977 atomic_dec(&this_parent->d_count);
1978 this_parent = this_parent->d_parent;
1979 goto resume;
1980 }
1981 spin_unlock(&dcache_lock);
1982}
1983
1984/**
1985 * find_inode_number - check for dentry with name
1986 * @dir: directory to check
1987 * @name: Name to find.
1988 *
1989 * Check whether a dentry already exists for the given name,
1990 * and return the inode number if it has an inode. Otherwise
1991 * 0 is returned.
1992 *
1993 * This routine is used to post-process directory listings for
1994 * filesystems using synthetic inode numbers, and is necessary
1995 * to keep getcwd() working.
1996 */
1997
1998ino_t find_inode_number(struct dentry *dir, struct qstr *name)
1999{
2000 struct dentry * dentry;
2001 ino_t ino = 0;
2002
Eric W. Biederman3e7e2412006-03-31 02:31:43 -08002003 dentry = d_hash_and_lookup(dir, name);
2004 if (dentry) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005 if (dentry->d_inode)
2006 ino = dentry->d_inode->i_ino;
2007 dput(dentry);
2008 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 return ino;
2010}
2011
2012static __initdata unsigned long dhash_entries;
2013static int __init set_dhash_entries(char *str)
2014{
2015 if (!str)
2016 return 0;
2017 dhash_entries = simple_strtoul(str, &str, 0);
2018 return 1;
2019}
2020__setup("dhash_entries=", set_dhash_entries);
2021
2022static void __init dcache_init_early(void)
2023{
2024 int loop;
2025
2026 /* If hashes are distributed across NUMA nodes, defer
2027 * hash allocation until vmalloc space is available.
2028 */
2029 if (hashdist)
2030 return;
2031
2032 dentry_hashtable =
2033 alloc_large_system_hash("Dentry cache",
2034 sizeof(struct hlist_head),
2035 dhash_entries,
2036 13,
2037 HASH_EARLY,
2038 &d_hash_shift,
2039 &d_hash_mask,
2040 0);
2041
2042 for (loop = 0; loop < (1 << d_hash_shift); loop++)
2043 INIT_HLIST_HEAD(&dentry_hashtable[loop]);
2044}
2045
2046static void __init dcache_init(unsigned long mempages)
2047{
2048 int loop;
2049
2050 /*
2051 * A constructor could be added for stable state like the lists,
2052 * but it is probably not worth it because of the cache nature
2053 * of the dcache.
2054 */
2055 dentry_cache = kmem_cache_create("dentry_cache",
2056 sizeof(struct dentry),
2057 0,
Paul Jacksonb0196002006-03-24 03:16:09 -08002058 (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|
2059 SLAB_MEM_SPREAD),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060 NULL, NULL);
2061
2062 set_shrinker(DEFAULT_SEEKS, shrink_dcache_memory);
2063
2064 /* Hash may have been set up in dcache_init_early */
2065 if (!hashdist)
2066 return;
2067
2068 dentry_hashtable =
2069 alloc_large_system_hash("Dentry cache",
2070 sizeof(struct hlist_head),
2071 dhash_entries,
2072 13,
2073 0,
2074 &d_hash_shift,
2075 &d_hash_mask,
2076 0);
2077
2078 for (loop = 0; loop < (1 << d_hash_shift); loop++)
2079 INIT_HLIST_HEAD(&dentry_hashtable[loop]);
2080}
2081
2082/* SLAB cache for __getname() consumers */
Christoph Lametere18b8902006-12-06 20:33:20 -08002083struct kmem_cache *names_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084
2085/* SLAB cache for file structures */
Christoph Lametere18b8902006-12-06 20:33:20 -08002086struct kmem_cache *filp_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087
2088EXPORT_SYMBOL(d_genocide);
2089
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090void __init vfs_caches_init_early(void)
2091{
2092 dcache_init_early();
2093 inode_init_early();
2094}
2095
2096void __init vfs_caches_init(unsigned long mempages)
2097{
2098 unsigned long reserve;
2099
2100 /* Base hash sizes on available memory, with a reserve equal to
2101 150% of current kernel size */
2102
2103 reserve = min((mempages - nr_free_pages()) * 3/2, mempages - 1);
2104 mempages -= reserve;
2105
2106 names_cachep = kmem_cache_create("names_cache", PATH_MAX, 0,
2107 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
2108
2109 filp_cachep = kmem_cache_create("filp", sizeof(struct file), 0,
Dipankar Sarma529bf6b2006-03-07 21:55:35 -08002110 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111
2112 dcache_init(mempages);
2113 inode_init(mempages);
2114 files_init(mempages);
2115 mnt_init(mempages);
2116 bdev_cache_init();
2117 chrdev_init();
2118}
2119
2120EXPORT_SYMBOL(d_alloc);
2121EXPORT_SYMBOL(d_alloc_anon);
2122EXPORT_SYMBOL(d_alloc_root);
2123EXPORT_SYMBOL(d_delete);
2124EXPORT_SYMBOL(d_find_alias);
2125EXPORT_SYMBOL(d_instantiate);
2126EXPORT_SYMBOL(d_invalidate);
2127EXPORT_SYMBOL(d_lookup);
2128EXPORT_SYMBOL(d_move);
David Howells770bfad2006-08-22 20:06:07 -04002129EXPORT_SYMBOL_GPL(d_materialise_unique);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130EXPORT_SYMBOL(d_path);
2131EXPORT_SYMBOL(d_prune_aliases);
2132EXPORT_SYMBOL(d_rehash);
2133EXPORT_SYMBOL(d_splice_alias);
2134EXPORT_SYMBOL(d_validate);
2135EXPORT_SYMBOL(dget_locked);
2136EXPORT_SYMBOL(dput);
2137EXPORT_SYMBOL(find_inode_number);
2138EXPORT_SYMBOL(have_submounts);
2139EXPORT_SYMBOL(names_cachep);
2140EXPORT_SYMBOL(shrink_dcache_parent);
2141EXPORT_SYMBOL(shrink_dcache_sb);