4 * Complete reimplementation
5 * (C) 1997 Thomas Schoebel-Theuer,
6 * with heavy changes by Linus Torvalds
10 * Notes on the allocation strategy:
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.
17 #include <linux/syscalls.h>
18 #include <linux/string.h>
21 #include <linux/fsnotify.h>
22 #include <linux/slab.h>
23 #include <linux/init.h>
24 #include <linux/hash.h>
25 #include <linux/cache.h>
26 #include <linux/module.h>
27 #include <linux/mount.h>
28 #include <linux/file.h>
29 #include <asm/uaccess.h>
30 #include <linux/security.h>
31 #include <linux/seqlock.h>
32 #include <linux/swap.h>
33 #include <linux/bootmem.h>
34 #include <linux/fs_struct.h>
35 #include <linux/hardirq.h>
36 #include <linux/bit_spinlock.h>
37 #include <linux/rculist_bl.h>
42 * dcache->d_inode->i_lock protects:
43 * - i_dentry, d_alias, d_inode of aliases
44 * dcache_hash_bucket lock protects:
45 * - the dcache hash table
46 * s_anon bl list spinlock protects:
47 * - the s_anon list (see __d_drop)
48 * dcache_lru_lock protects:
49 * - the dcache lru lists and counters
56 * - d_parent and d_subdirs
57 * - childrens' d_child and d_parent
61 * dentry->d_inode->i_lock
64 * dcache_hash_bucket lock
67 * If there is an ancestor relationship:
68 * dentry->d_parent->...->d_parent->d_lock
70 * dentry->d_parent->d_lock
73 * If no ancestor relationship:
74 * if (dentry1 < dentry2)
78 int sysctl_vfs_cache_pressure __read_mostly = 100;
79 EXPORT_SYMBOL_GPL(sysctl_vfs_cache_pressure);
81 static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lru_lock);
82 __cacheline_aligned_in_smp DEFINE_SEQLOCK(rename_lock);
84 EXPORT_SYMBOL(rename_lock);
86 static struct kmem_cache *dentry_cache __read_mostly;
89 * This is the single most critical data structure when it comes
90 * to the dcache: the hashtable for lookups. Somebody should try
91 * to make this good - I've just made it work.
93 * This hash-function tries to avoid losing too many bits of hash
94 * information, yet avoid using a prime hash-size or similar.
96 #define D_HASHBITS d_hash_shift
97 #define D_HASHMASK d_hash_mask
99 static unsigned int d_hash_mask __read_mostly;
100 static unsigned int d_hash_shift __read_mostly;
102 static struct hlist_bl_head *dentry_hashtable __read_mostly;
104 static inline struct hlist_bl_head *d_hash(struct dentry *parent,
107 hash += ((unsigned long) parent ^ GOLDEN_RATIO_PRIME) / L1_CACHE_BYTES;
108 hash = hash ^ ((hash ^ GOLDEN_RATIO_PRIME) >> D_HASHBITS);
109 return dentry_hashtable + (hash & D_HASHMASK);
112 /* Statistics gathering. */
113 struct dentry_stat_t dentry_stat = {
117 static DEFINE_PER_CPU(unsigned int, nr_dentry);
119 #if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
120 static int get_nr_dentry(void)
124 for_each_possible_cpu(i)
125 sum += per_cpu(nr_dentry, i);
126 return sum < 0 ? 0 : sum;
129 int proc_nr_dentry(ctl_table *table, int write, void __user *buffer,
130 size_t *lenp, loff_t *ppos)
132 dentry_stat.nr_dentry = get_nr_dentry();
133 return proc_dointvec(table, write, buffer, lenp, ppos);
137 static void __d_free(struct rcu_head *head)
139 struct dentry *dentry = container_of(head, struct dentry, d_u.d_rcu);
141 WARN_ON(!list_empty(&dentry->d_alias));
142 if (dname_external(dentry))
143 kfree(dentry->d_name.name);
144 kmem_cache_free(dentry_cache, dentry);
150 static void d_free(struct dentry *dentry)
152 BUG_ON(dentry->d_count);
153 this_cpu_dec(nr_dentry);
154 if (dentry->d_op && dentry->d_op->d_release)
155 dentry->d_op->d_release(dentry);
157 /* if dentry was never visible to RCU, immediate free is OK */
158 if (!(dentry->d_flags & DCACHE_RCUACCESS))
159 __d_free(&dentry->d_u.d_rcu);
161 call_rcu(&dentry->d_u.d_rcu, __d_free);
165 * dentry_rcuwalk_barrier - invalidate in-progress rcu-walk lookups
166 * @dentry: the target dentry
167 * After this call, in-progress rcu-walk path lookup will fail. This
168 * should be called after unhashing, and after changing d_inode (if
169 * the dentry has not already been unhashed).
171 static inline void dentry_rcuwalk_barrier(struct dentry *dentry)
173 assert_spin_locked(&dentry->d_lock);
174 /* Go through a barrier */
175 write_seqcount_barrier(&dentry->d_seq);
179 * Release the dentry's inode, using the filesystem
180 * d_iput() operation if defined. Dentry has no refcount
183 static void dentry_iput(struct dentry * dentry)
184 __releases(dentry->d_lock)
185 __releases(dentry->d_inode->i_lock)
187 struct inode *inode = dentry->d_inode;
189 dentry->d_inode = NULL;
190 list_del_init(&dentry->d_alias);
191 spin_unlock(&dentry->d_lock);
192 spin_unlock(&inode->i_lock);
194 fsnotify_inoderemove(inode);
195 if (dentry->d_op && dentry->d_op->d_iput)
196 dentry->d_op->d_iput(dentry, inode);
200 spin_unlock(&dentry->d_lock);
205 * Release the dentry's inode, using the filesystem
206 * d_iput() operation if defined. dentry remains in-use.
208 static void dentry_unlink_inode(struct dentry * dentry)
209 __releases(dentry->d_lock)
210 __releases(dentry->d_inode->i_lock)
212 struct inode *inode = dentry->d_inode;
213 dentry->d_inode = NULL;
214 list_del_init(&dentry->d_alias);
215 dentry_rcuwalk_barrier(dentry);
216 spin_unlock(&dentry->d_lock);
217 spin_unlock(&inode->i_lock);
219 fsnotify_inoderemove(inode);
220 if (dentry->d_op && dentry->d_op->d_iput)
221 dentry->d_op->d_iput(dentry, inode);
227 * dentry_lru_(add|del|move_tail) must be called with d_lock held.
229 static void dentry_lru_add(struct dentry *dentry)
231 if (list_empty(&dentry->d_lru)) {
232 spin_lock(&dcache_lru_lock);
233 list_add(&dentry->d_lru, &dentry->d_sb->s_dentry_lru);
234 dentry->d_sb->s_nr_dentry_unused++;
235 dentry_stat.nr_unused++;
236 spin_unlock(&dcache_lru_lock);
240 static void __dentry_lru_del(struct dentry *dentry)
242 list_del_init(&dentry->d_lru);
243 dentry->d_sb->s_nr_dentry_unused--;
244 dentry_stat.nr_unused--;
247 static void dentry_lru_del(struct dentry *dentry)
249 if (!list_empty(&dentry->d_lru)) {
250 spin_lock(&dcache_lru_lock);
251 __dentry_lru_del(dentry);
252 spin_unlock(&dcache_lru_lock);
256 static void dentry_lru_move_tail(struct dentry *dentry)
258 spin_lock(&dcache_lru_lock);
259 if (list_empty(&dentry->d_lru)) {
260 list_add_tail(&dentry->d_lru, &dentry->d_sb->s_dentry_lru);
261 dentry->d_sb->s_nr_dentry_unused++;
262 dentry_stat.nr_unused++;
264 list_move_tail(&dentry->d_lru, &dentry->d_sb->s_dentry_lru);
266 spin_unlock(&dcache_lru_lock);
270 * d_kill - kill dentry and return parent
271 * @dentry: dentry to kill
272 * @parent: parent dentry
274 * The dentry must already be unhashed and removed from the LRU.
276 * If this is the root of the dentry tree, return NULL.
278 * dentry->d_lock and parent->d_lock must be held by caller, and are dropped by
281 static struct dentry *d_kill(struct dentry *dentry, struct dentry *parent)
282 __releases(dentry->d_lock)
283 __releases(parent->d_lock)
284 __releases(dentry->d_inode->i_lock)
286 list_del(&dentry->d_u.d_child);
288 * Inform try_to_ascend() that we are no longer attached to the
291 dentry->d_flags |= DCACHE_DISCONNECTED;
293 spin_unlock(&parent->d_lock);
296 * dentry_iput drops the locks, at which point nobody (except
297 * transient RCU lookups) can reach this dentry.
304 * d_drop - drop a dentry
305 * @dentry: dentry to drop
307 * d_drop() unhashes the entry from the parent dentry hashes, so that it won't
308 * be found through a VFS lookup any more. Note that this is different from
309 * deleting the dentry - d_delete will try to mark the dentry negative if
310 * possible, giving a successful _negative_ lookup, while d_drop will
311 * just make the cache lookup fail.
313 * d_drop() is used mainly for stuff that wants to invalidate a dentry for some
314 * reason (NFS timeouts or autofs deletes).
316 * __d_drop requires dentry->d_lock.
318 void __d_drop(struct dentry *dentry)
320 if (!d_unhashed(dentry)) {
321 struct hlist_bl_head *b;
322 if (unlikely(dentry->d_flags & DCACHE_DISCONNECTED))
323 b = &dentry->d_sb->s_anon;
325 b = d_hash(dentry->d_parent, dentry->d_name.hash);
328 __hlist_bl_del(&dentry->d_hash);
329 dentry->d_hash.pprev = NULL;
332 dentry_rcuwalk_barrier(dentry);
335 EXPORT_SYMBOL(__d_drop);
337 void d_drop(struct dentry *dentry)
339 spin_lock(&dentry->d_lock);
341 spin_unlock(&dentry->d_lock);
343 EXPORT_SYMBOL(d_drop);
346 * Finish off a dentry we've decided to kill.
347 * dentry->d_lock must be held, returns with it unlocked.
348 * If ref is non-zero, then decrement the refcount too.
349 * Returns dentry requiring refcount drop, or NULL if we're done.
351 static inline struct dentry *dentry_kill(struct dentry *dentry, int ref)
352 __releases(dentry->d_lock)
355 struct dentry *parent;
357 inode = dentry->d_inode;
358 if (inode && !spin_trylock(&inode->i_lock)) {
360 spin_unlock(&dentry->d_lock);
362 return dentry; /* try again with same dentry */
367 parent = dentry->d_parent;
368 if (parent && !spin_trylock(&parent->d_lock)) {
370 spin_unlock(&inode->i_lock);
376 /* if dentry was on the d_lru list delete it from there */
377 dentry_lru_del(dentry);
378 /* if it was on the hash then remove it */
380 return d_kill(dentry, parent);
386 * This is complicated by the fact that we do not want to put
387 * dentries that are no longer on any hash chain on the unused
388 * list: we'd much rather just get rid of them immediately.
390 * However, that implies that we have to traverse the dentry
391 * tree upwards to the parents which might _also_ now be
392 * scheduled for deletion (it may have been only waiting for
393 * its last child to go away).
395 * This tail recursion is done by hand as we don't want to depend
396 * on the compiler to always get this right (gcc generally doesn't).
397 * Real recursion would eat up our stack space.
401 * dput - release a dentry
402 * @dentry: dentry to release
404 * Release a dentry. This will drop the usage count and if appropriate
405 * call the dentry unlink method as well as removing it from the queues and
406 * releasing its resources. If the parent dentries were scheduled for release
407 * they too may now get deleted.
409 void dput(struct dentry *dentry)
415 if (dentry->d_count == 1)
417 spin_lock(&dentry->d_lock);
418 BUG_ON(!dentry->d_count);
419 if (dentry->d_count > 1) {
421 spin_unlock(&dentry->d_lock);
425 if (dentry->d_flags & DCACHE_OP_DELETE) {
426 if (dentry->d_op->d_delete(dentry))
430 /* Unreachable? Get rid of it */
431 if (d_unhashed(dentry))
434 /* Otherwise leave it cached and ensure it's on the LRU */
435 dentry->d_flags |= DCACHE_REFERENCED;
436 dentry_lru_add(dentry);
439 spin_unlock(&dentry->d_lock);
443 dentry = dentry_kill(dentry, 1);
450 * d_invalidate - invalidate a dentry
451 * @dentry: dentry to invalidate
453 * Try to invalidate the dentry if it turns out to be
454 * possible. If there are other dentries that can be
455 * reached through this one we can't delete it and we
456 * return -EBUSY. On success we return 0.
461 int d_invalidate(struct dentry * dentry)
464 * If it's already been dropped, return OK.
466 spin_lock(&dentry->d_lock);
467 if (d_unhashed(dentry)) {
468 spin_unlock(&dentry->d_lock);
472 * Check whether to do a partial shrink_dcache
473 * to get rid of unused child entries.
475 if (!list_empty(&dentry->d_subdirs)) {
476 spin_unlock(&dentry->d_lock);
477 shrink_dcache_parent(dentry);
478 spin_lock(&dentry->d_lock);
482 * Somebody else still using it?
484 * If it's a directory, we can't drop it
485 * for fear of somebody re-populating it
486 * with children (even though dropping it
487 * would make it unreachable from the root,
488 * we might still populate it if it was a
489 * working directory or similar).
491 if (dentry->d_count > 1) {
492 if (dentry->d_inode && S_ISDIR(dentry->d_inode->i_mode)) {
493 spin_unlock(&dentry->d_lock);
499 spin_unlock(&dentry->d_lock);
502 EXPORT_SYMBOL(d_invalidate);
504 /* This must be called with d_lock held */
505 static inline void __dget_dlock(struct dentry *dentry)
510 static inline void __dget(struct dentry *dentry)
512 spin_lock(&dentry->d_lock);
513 __dget_dlock(dentry);
514 spin_unlock(&dentry->d_lock);
517 struct dentry *dget_parent(struct dentry *dentry)
523 * Don't need rcu_dereference because we re-check it was correct under
527 ret = dentry->d_parent;
532 spin_lock(&ret->d_lock);
533 if (unlikely(ret != dentry->d_parent)) {
534 spin_unlock(&ret->d_lock);
539 BUG_ON(!ret->d_count);
541 spin_unlock(&ret->d_lock);
545 EXPORT_SYMBOL(dget_parent);
548 * d_find_alias - grab a hashed alias of inode
549 * @inode: inode in question
550 * @want_discon: flag, used by d_splice_alias, to request
551 * that only a DISCONNECTED alias be returned.
553 * If inode has a hashed alias, or is a directory and has any alias,
554 * acquire the reference to alias and return it. Otherwise return NULL.
555 * Notice that if inode is a directory there can be only one alias and
556 * it can be unhashed only if it has no children, or if it is the root
559 * If the inode has an IS_ROOT, DCACHE_DISCONNECTED alias, then prefer
560 * any other hashed alias over that one unless @want_discon is set,
561 * in which case only return an IS_ROOT, DCACHE_DISCONNECTED alias.
563 static struct dentry *__d_find_alias(struct inode *inode, int want_discon)
565 struct dentry *alias, *discon_alias;
569 list_for_each_entry(alias, &inode->i_dentry, d_alias) {
570 spin_lock(&alias->d_lock);
571 if (S_ISDIR(inode->i_mode) || !d_unhashed(alias)) {
572 if (IS_ROOT(alias) &&
573 (alias->d_flags & DCACHE_DISCONNECTED)) {
574 discon_alias = alias;
575 } else if (!want_discon) {
577 spin_unlock(&alias->d_lock);
581 spin_unlock(&alias->d_lock);
584 alias = discon_alias;
585 spin_lock(&alias->d_lock);
586 if (S_ISDIR(inode->i_mode) || !d_unhashed(alias)) {
587 if (IS_ROOT(alias) &&
588 (alias->d_flags & DCACHE_DISCONNECTED)) {
590 spin_unlock(&alias->d_lock);
594 spin_unlock(&alias->d_lock);
600 struct dentry *d_find_alias(struct inode *inode)
602 struct dentry *de = NULL;
604 if (!list_empty(&inode->i_dentry)) {
605 spin_lock(&inode->i_lock);
606 de = __d_find_alias(inode, 0);
607 spin_unlock(&inode->i_lock);
611 EXPORT_SYMBOL(d_find_alias);
614 * Try to kill dentries associated with this inode.
615 * WARNING: you must own a reference to inode.
617 void d_prune_aliases(struct inode *inode)
619 struct dentry *dentry;
621 spin_lock(&inode->i_lock);
622 list_for_each_entry(dentry, &inode->i_dentry, d_alias) {
623 spin_lock(&dentry->d_lock);
624 if (!dentry->d_count) {
625 __dget_dlock(dentry);
627 spin_unlock(&dentry->d_lock);
628 spin_unlock(&inode->i_lock);
632 spin_unlock(&dentry->d_lock);
634 spin_unlock(&inode->i_lock);
636 EXPORT_SYMBOL(d_prune_aliases);
639 * Try to throw away a dentry - free the inode, dput the parent.
640 * Requires dentry->d_lock is held, and dentry->d_count == 0.
641 * Releases dentry->d_lock.
643 * This may fail if locks cannot be acquired no problem, just try again.
645 static void try_prune_one_dentry(struct dentry *dentry)
646 __releases(dentry->d_lock)
648 struct dentry *parent;
650 parent = dentry_kill(dentry, 0);
652 * If dentry_kill returns NULL, we have nothing more to do.
653 * if it returns the same dentry, trylocks failed. In either
654 * case, just loop again.
656 * Otherwise, we need to prune ancestors too. This is necessary
657 * to prevent quadratic behavior of shrink_dcache_parent(), but
658 * is also expected to be beneficial in reducing dentry cache
663 if (parent == dentry)
666 /* Prune ancestors. */
669 spin_lock(&dentry->d_lock);
670 if (dentry->d_count > 1) {
672 spin_unlock(&dentry->d_lock);
675 dentry = dentry_kill(dentry, 1);
679 static void shrink_dentry_list(struct list_head *list)
681 struct dentry *dentry;
685 dentry = list_entry_rcu(list->prev, struct dentry, d_lru);
686 if (&dentry->d_lru == list)
688 spin_lock(&dentry->d_lock);
689 if (dentry != list_entry(list->prev, struct dentry, d_lru)) {
690 spin_unlock(&dentry->d_lock);
695 * We found an inuse dentry which was not removed from
696 * the LRU because of laziness during lookup. Do not free
697 * it - just keep it off the LRU list.
699 if (dentry->d_count) {
700 dentry_lru_del(dentry);
701 spin_unlock(&dentry->d_lock);
707 try_prune_one_dentry(dentry);
715 * __shrink_dcache_sb - shrink the dentry LRU on a given superblock
716 * @sb: superblock to shrink dentry LRU.
717 * @count: number of entries to prune
718 * @flags: flags to control the dentry processing
720 * If flags contains DCACHE_REFERENCED reference dentries will not be pruned.
722 static void __shrink_dcache_sb(struct super_block *sb, int *count, int flags)
724 /* called from prune_dcache() and shrink_dcache_parent() */
725 struct dentry *dentry;
726 LIST_HEAD(referenced);
731 spin_lock(&dcache_lru_lock);
732 while (!list_empty(&sb->s_dentry_lru)) {
733 dentry = list_entry(sb->s_dentry_lru.prev,
734 struct dentry, d_lru);
735 BUG_ON(dentry->d_sb != sb);
737 if (!spin_trylock(&dentry->d_lock)) {
738 spin_unlock(&dcache_lru_lock);
744 * If we are honouring the DCACHE_REFERENCED flag and the
745 * dentry has this flag set, don't free it. Clear the flag
746 * and put it back on the LRU.
748 if (flags & DCACHE_REFERENCED &&
749 dentry->d_flags & DCACHE_REFERENCED) {
750 dentry->d_flags &= ~DCACHE_REFERENCED;
751 list_move(&dentry->d_lru, &referenced);
752 spin_unlock(&dentry->d_lock);
754 list_move_tail(&dentry->d_lru, &tmp);
755 spin_unlock(&dentry->d_lock);
759 cond_resched_lock(&dcache_lru_lock);
761 if (!list_empty(&referenced))
762 list_splice(&referenced, &sb->s_dentry_lru);
763 spin_unlock(&dcache_lru_lock);
765 shrink_dentry_list(&tmp);
771 * prune_dcache - shrink the dcache
772 * @count: number of entries to try to free
774 * Shrink the dcache. This is done when we need more memory, or simply when we
775 * need to unmount something (at which point we need to unuse all dentries).
777 * This function may fail to free any resources if all the dentries are in use.
779 static void prune_dcache(int count)
781 struct super_block *sb, *p = NULL;
783 int unused = dentry_stat.nr_unused;
787 if (unused == 0 || count == 0)
792 prune_ratio = unused / count;
794 list_for_each_entry(sb, &super_blocks, s_list) {
795 if (list_empty(&sb->s_instances))
797 if (sb->s_nr_dentry_unused == 0)
800 /* Now, we reclaim unused dentrins with fairness.
801 * We reclaim them same percentage from each superblock.
802 * We calculate number of dentries to scan on this sb
803 * as follows, but the implementation is arranged to avoid
805 * number of dentries to scan on this sb =
806 * count * (number of dentries on this sb /
807 * number of dentries in the machine)
809 spin_unlock(&sb_lock);
810 if (prune_ratio != 1)
811 w_count = (sb->s_nr_dentry_unused / prune_ratio) + 1;
813 w_count = sb->s_nr_dentry_unused;
816 * We need to be sure this filesystem isn't being unmounted,
817 * otherwise we could race with generic_shutdown_super(), and
818 * end up holding a reference to an inode while the filesystem
819 * is unmounted. So we try to get s_umount, and make sure
822 if (down_read_trylock(&sb->s_umount)) {
823 if ((sb->s_root != NULL) &&
824 (!list_empty(&sb->s_dentry_lru))) {
825 __shrink_dcache_sb(sb, &w_count,
829 up_read(&sb->s_umount);
836 /* more work left to do? */
842 spin_unlock(&sb_lock);
846 * shrink_dcache_sb - shrink dcache for a superblock
849 * Shrink the dcache for the specified super block. This is used to free
850 * the dcache before unmounting a file system.
852 void shrink_dcache_sb(struct super_block *sb)
856 spin_lock(&dcache_lru_lock);
857 while (!list_empty(&sb->s_dentry_lru)) {
858 list_splice_init(&sb->s_dentry_lru, &tmp);
859 spin_unlock(&dcache_lru_lock);
860 shrink_dentry_list(&tmp);
861 spin_lock(&dcache_lru_lock);
863 spin_unlock(&dcache_lru_lock);
865 EXPORT_SYMBOL(shrink_dcache_sb);
868 * destroy a single subtree of dentries for unmount
869 * - see the comments on shrink_dcache_for_umount() for a description of the
872 static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
874 struct dentry *parent;
875 unsigned detached = 0;
877 BUG_ON(!IS_ROOT(dentry));
879 /* detach this root from the system */
880 spin_lock(&dentry->d_lock);
881 dentry_lru_del(dentry);
883 spin_unlock(&dentry->d_lock);
886 /* descend to the first leaf in the current subtree */
887 while (!list_empty(&dentry->d_subdirs)) {
890 /* this is a branch with children - detach all of them
891 * from the system in one go */
892 spin_lock(&dentry->d_lock);
893 list_for_each_entry(loop, &dentry->d_subdirs,
895 spin_lock_nested(&loop->d_lock,
896 DENTRY_D_LOCK_NESTED);
897 dentry_lru_del(loop);
899 spin_unlock(&loop->d_lock);
901 spin_unlock(&dentry->d_lock);
903 /* move to the first child */
904 dentry = list_entry(dentry->d_subdirs.next,
905 struct dentry, d_u.d_child);
908 /* consume the dentries from this leaf up through its parents
909 * until we find one with children or run out altogether */
913 if (dentry->d_count != 0) {
915 "BUG: Dentry %p{i=%lx,n=%s}"
917 " [unmount of %s %s]\n",
920 dentry->d_inode->i_ino : 0UL,
923 dentry->d_sb->s_type->name,
928 if (IS_ROOT(dentry)) {
930 list_del(&dentry->d_u.d_child);
932 parent = dentry->d_parent;
933 spin_lock(&parent->d_lock);
935 list_del(&dentry->d_u.d_child);
936 spin_unlock(&parent->d_lock);
941 inode = dentry->d_inode;
943 dentry->d_inode = NULL;
944 list_del_init(&dentry->d_alias);
945 if (dentry->d_op && dentry->d_op->d_iput)
946 dentry->d_op->d_iput(dentry, inode);
953 /* finished when we fall off the top of the tree,
954 * otherwise we ascend to the parent and move to the
955 * next sibling if there is one */
959 } while (list_empty(&dentry->d_subdirs));
961 dentry = list_entry(dentry->d_subdirs.next,
962 struct dentry, d_u.d_child);
967 * destroy the dentries attached to a superblock on unmounting
968 * - we don't need to use dentry->d_lock because:
969 * - the superblock is detached from all mountings and open files, so the
970 * dentry trees will not be rearranged by the VFS
971 * - s_umount is write-locked, so the memory pressure shrinker will ignore
972 * any dentries belonging to this superblock that it comes across
973 * - the filesystem itself is no longer permitted to rearrange the dentries
976 void shrink_dcache_for_umount(struct super_block *sb)
978 struct dentry *dentry;
980 if (down_read_trylock(&sb->s_umount))
985 spin_lock(&dentry->d_lock);
987 spin_unlock(&dentry->d_lock);
988 shrink_dcache_for_umount_subtree(dentry);
990 while (!hlist_bl_empty(&sb->s_anon)) {
991 dentry = hlist_bl_entry(hlist_bl_first(&sb->s_anon), struct dentry, d_hash);
992 shrink_dcache_for_umount_subtree(dentry);
997 * This tries to ascend one level of parenthood, but
998 * we can race with renaming, so we need to re-check
999 * the parenthood after dropping the lock and check
1000 * that the sequence number still matches.
1002 static struct dentry *try_to_ascend(struct dentry *old, int locked, unsigned seq)
1004 struct dentry *new = old->d_parent;
1007 spin_unlock(&old->d_lock);
1008 spin_lock(&new->d_lock);
1011 * might go back up the wrong parent if we have had a rename
1014 if (new != old->d_parent ||
1015 (old->d_flags & DCACHE_DISCONNECTED) ||
1016 (!locked && read_seqretry(&rename_lock, seq))) {
1017 spin_unlock(&new->d_lock);
1026 * Search for at least 1 mount point in the dentry's subdirs.
1027 * We descend to the next level whenever the d_subdirs
1028 * list is non-empty and continue searching.
1032 * have_submounts - check for mounts over a dentry
1033 * @parent: dentry to check.
1035 * Return true if the parent or its subdirectories contain
1038 int have_submounts(struct dentry *parent)
1040 struct dentry *this_parent;
1041 struct list_head *next;
1045 seq = read_seqbegin(&rename_lock);
1047 this_parent = parent;
1049 if (d_mountpoint(parent))
1051 spin_lock(&this_parent->d_lock);
1053 next = this_parent->d_subdirs.next;
1055 while (next != &this_parent->d_subdirs) {
1056 struct list_head *tmp = next;
1057 struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
1060 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
1061 /* Have we found a mount point ? */
1062 if (d_mountpoint(dentry)) {
1063 spin_unlock(&dentry->d_lock);
1064 spin_unlock(&this_parent->d_lock);
1067 if (!list_empty(&dentry->d_subdirs)) {
1068 spin_unlock(&this_parent->d_lock);
1069 spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
1070 this_parent = dentry;
1071 spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
1074 spin_unlock(&dentry->d_lock);
1077 * All done at this level ... ascend and resume the search.
1079 if (this_parent != parent) {
1080 struct dentry *child = this_parent;
1081 this_parent = try_to_ascend(this_parent, locked, seq);
1084 next = child->d_u.d_child.next;
1087 spin_unlock(&this_parent->d_lock);
1088 if (!locked && read_seqretry(&rename_lock, seq))
1091 write_sequnlock(&rename_lock);
1092 return 0; /* No mount points found in tree */
1094 if (!locked && read_seqretry(&rename_lock, seq))
1097 write_sequnlock(&rename_lock);
1102 write_seqlock(&rename_lock);
1105 EXPORT_SYMBOL(have_submounts);
1108 * Search the dentry child list for the specified parent,
1109 * and move any unused dentries to the end of the unused
1110 * list for prune_dcache(). We descend to the next level
1111 * whenever the d_subdirs list is non-empty and continue
1114 * It returns zero iff there are no unused children,
1115 * otherwise it returns the number of children moved to
1116 * the end of the unused list. This may not be the total
1117 * number of unused children, because select_parent can
1118 * drop the lock and return early due to latency
1121 static int select_parent(struct dentry * parent)
1123 struct dentry *this_parent;
1124 struct list_head *next;
1129 seq = read_seqbegin(&rename_lock);
1131 this_parent = parent;
1132 spin_lock(&this_parent->d_lock);
1134 next = this_parent->d_subdirs.next;
1136 while (next != &this_parent->d_subdirs) {
1137 struct list_head *tmp = next;
1138 struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
1141 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
1144 * move only zero ref count dentries to the end
1145 * of the unused list for prune_dcache
1147 if (!dentry->d_count) {
1148 dentry_lru_move_tail(dentry);
1151 dentry_lru_del(dentry);
1155 * We can return to the caller if we have found some (this
1156 * ensures forward progress). We'll be coming back to find
1159 if (found && need_resched()) {
1160 spin_unlock(&dentry->d_lock);
1165 * Descend a level if the d_subdirs list is non-empty.
1167 if (!list_empty(&dentry->d_subdirs)) {
1168 spin_unlock(&this_parent->d_lock);
1169 spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
1170 this_parent = dentry;
1171 spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
1175 spin_unlock(&dentry->d_lock);
1178 * All done at this level ... ascend and resume the search.
1180 if (this_parent != parent) {
1181 struct dentry *child = this_parent;
1182 this_parent = try_to_ascend(this_parent, locked, seq);
1185 next = child->d_u.d_child.next;
1189 spin_unlock(&this_parent->d_lock);
1190 if (!locked && read_seqretry(&rename_lock, seq))
1193 write_sequnlock(&rename_lock);
1200 write_seqlock(&rename_lock);
1205 * shrink_dcache_parent - prune dcache
1206 * @parent: parent of entries to prune
1208 * Prune the dcache to remove unused children of the parent dentry.
1211 void shrink_dcache_parent(struct dentry * parent)
1213 struct super_block *sb = parent->d_sb;
1216 while ((found = select_parent(parent)) != 0)
1217 __shrink_dcache_sb(sb, &found, 0);
1219 EXPORT_SYMBOL(shrink_dcache_parent);
1222 * Scan `nr' dentries and return the number which remain.
1224 * We need to avoid reentering the filesystem if the caller is performing a
1225 * GFP_NOFS allocation attempt. One example deadlock is:
1227 * ext2_new_block->getblk->GFP->shrink_dcache_memory->prune_dcache->
1228 * prune_one_dentry->dput->dentry_iput->iput->inode->i_sb->s_op->put_inode->
1229 * ext2_discard_prealloc->ext2_free_blocks->lock_super->DEADLOCK.
1231 * In this case we return -1 to tell the caller that we baled.
1233 static int shrink_dcache_memory(struct shrinker *shrink, int nr, gfp_t gfp_mask)
1236 if (!(gfp_mask & __GFP_FS))
1241 return (dentry_stat.nr_unused / 100) * sysctl_vfs_cache_pressure;
1244 static struct shrinker dcache_shrinker = {
1245 .shrink = shrink_dcache_memory,
1246 .seeks = DEFAULT_SEEKS,
1250 * d_alloc - allocate a dcache entry
1251 * @parent: parent of entry to allocate
1252 * @name: qstr of the name
1254 * Allocates a dentry. It returns %NULL if there is insufficient memory
1255 * available. On a success the dentry is returned. The name passed in is
1256 * copied and the copy passed in may be reused after this call.
1259 struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
1261 struct dentry *dentry;
1264 dentry = kmem_cache_alloc(dentry_cache, GFP_KERNEL);
1268 if (name->len > DNAME_INLINE_LEN-1) {
1269 dname = kmalloc(name->len + 1, GFP_KERNEL);
1271 kmem_cache_free(dentry_cache, dentry);
1275 dname = dentry->d_iname;
1277 dentry->d_name.name = dname;
1279 dentry->d_name.len = name->len;
1280 dentry->d_name.hash = name->hash;
1281 memcpy(dname, name->name, name->len);
1282 dname[name->len] = 0;
1284 dentry->d_count = 1;
1285 dentry->d_flags = 0;
1286 spin_lock_init(&dentry->d_lock);
1287 seqcount_init(&dentry->d_seq);
1288 dentry->d_inode = NULL;
1289 dentry->d_parent = NULL;
1290 dentry->d_sb = NULL;
1291 dentry->d_op = NULL;
1292 dentry->d_fsdata = NULL;
1293 INIT_HLIST_BL_NODE(&dentry->d_hash);
1294 INIT_LIST_HEAD(&dentry->d_lru);
1295 INIT_LIST_HEAD(&dentry->d_subdirs);
1296 INIT_LIST_HEAD(&dentry->d_alias);
1297 INIT_LIST_HEAD(&dentry->d_u.d_child);
1300 spin_lock(&parent->d_lock);
1302 * don't need child lock because it is not subject
1303 * to concurrency here
1305 __dget_dlock(parent);
1306 dentry->d_parent = parent;
1307 dentry->d_sb = parent->d_sb;
1308 d_set_d_op(dentry, dentry->d_sb->s_d_op);
1309 list_add(&dentry->d_u.d_child, &parent->d_subdirs);
1310 spin_unlock(&parent->d_lock);
1313 this_cpu_inc(nr_dentry);
1317 EXPORT_SYMBOL(d_alloc);
1319 struct dentry *d_alloc_pseudo(struct super_block *sb, const struct qstr *name)
1321 struct dentry *dentry = d_alloc(NULL, name);
1324 d_set_d_op(dentry, dentry->d_sb->s_d_op);
1325 dentry->d_parent = dentry;
1326 dentry->d_flags |= DCACHE_DISCONNECTED;
1330 EXPORT_SYMBOL(d_alloc_pseudo);
1332 struct dentry *d_alloc_name(struct dentry *parent, const char *name)
1337 q.len = strlen(name);
1338 q.hash = full_name_hash(q.name, q.len);
1339 return d_alloc(parent, &q);
1341 EXPORT_SYMBOL(d_alloc_name);
1343 void d_set_d_op(struct dentry *dentry, const struct dentry_operations *op)
1345 WARN_ON_ONCE(dentry->d_op);
1346 WARN_ON_ONCE(dentry->d_flags & (DCACHE_OP_HASH |
1348 DCACHE_OP_REVALIDATE |
1349 DCACHE_OP_DELETE ));
1354 dentry->d_flags |= DCACHE_OP_HASH;
1356 dentry->d_flags |= DCACHE_OP_COMPARE;
1357 if (op->d_revalidate)
1358 dentry->d_flags |= DCACHE_OP_REVALIDATE;
1360 dentry->d_flags |= DCACHE_OP_DELETE;
1363 EXPORT_SYMBOL(d_set_d_op);
1365 static void __d_instantiate(struct dentry *dentry, struct inode *inode)
1367 spin_lock(&dentry->d_lock);
1369 if (unlikely(IS_AUTOMOUNT(inode)))
1370 dentry->d_flags |= DCACHE_NEED_AUTOMOUNT;
1371 list_add(&dentry->d_alias, &inode->i_dentry);
1373 dentry->d_inode = inode;
1374 dentry_rcuwalk_barrier(dentry);
1375 spin_unlock(&dentry->d_lock);
1376 fsnotify_d_instantiate(dentry, inode);
1380 * d_instantiate - fill in inode information for a dentry
1381 * @entry: dentry to complete
1382 * @inode: inode to attach to this dentry
1384 * Fill in inode information in the entry.
1386 * This turns negative dentries into productive full members
1389 * NOTE! This assumes that the inode count has been incremented
1390 * (or otherwise set) by the caller to indicate that it is now
1391 * in use by the dcache.
1394 void d_instantiate(struct dentry *entry, struct inode * inode)
1396 BUG_ON(!list_empty(&entry->d_alias));
1398 spin_lock(&inode->i_lock);
1399 __d_instantiate(entry, inode);
1401 spin_unlock(&inode->i_lock);
1402 security_d_instantiate(entry, inode);
1404 EXPORT_SYMBOL(d_instantiate);
1407 * d_instantiate_unique - instantiate a non-aliased dentry
1408 * @entry: dentry to instantiate
1409 * @inode: inode to attach to this dentry
1411 * Fill in inode information in the entry. On success, it returns NULL.
1412 * If an unhashed alias of "entry" already exists, then we return the
1413 * aliased dentry instead and drop one reference to inode.
1415 * Note that in order to avoid conflicts with rename() etc, the caller
1416 * had better be holding the parent directory semaphore.
1418 * This also assumes that the inode count has been incremented
1419 * (or otherwise set) by the caller to indicate that it is now
1420 * in use by the dcache.
1422 static struct dentry *__d_instantiate_unique(struct dentry *entry,
1423 struct inode *inode)
1425 struct dentry *alias;
1426 int len = entry->d_name.len;
1427 const char *name = entry->d_name.name;
1428 unsigned int hash = entry->d_name.hash;
1431 __d_instantiate(entry, NULL);
1435 list_for_each_entry(alias, &inode->i_dentry, d_alias) {
1436 struct qstr *qstr = &alias->d_name;
1439 * Don't need alias->d_lock here, because aliases with
1440 * d_parent == entry->d_parent are not subject to name or
1441 * parent changes, because the parent inode i_mutex is held.
1443 if (qstr->hash != hash)
1445 if (alias->d_parent != entry->d_parent)
1447 if (dentry_cmp(qstr->name, qstr->len, name, len))
1453 __d_instantiate(entry, inode);
1457 struct dentry *d_instantiate_unique(struct dentry *entry, struct inode *inode)
1459 struct dentry *result;
1461 BUG_ON(!list_empty(&entry->d_alias));
1464 spin_lock(&inode->i_lock);
1465 result = __d_instantiate_unique(entry, inode);
1467 spin_unlock(&inode->i_lock);
1470 security_d_instantiate(entry, inode);
1474 BUG_ON(!d_unhashed(result));
1479 EXPORT_SYMBOL(d_instantiate_unique);
1482 * d_alloc_root - allocate root dentry
1483 * @root_inode: inode to allocate the root for
1485 * Allocate a root ("/") dentry for the inode given. The inode is
1486 * instantiated and returned. %NULL is returned if there is insufficient
1487 * memory or the inode passed is %NULL.
1490 struct dentry * d_alloc_root(struct inode * root_inode)
1492 struct dentry *res = NULL;
1495 static const struct qstr name = { .name = "/", .len = 1 };
1497 res = d_alloc(NULL, &name);
1499 res->d_sb = root_inode->i_sb;
1500 d_set_d_op(res, res->d_sb->s_d_op);
1501 res->d_parent = res;
1502 d_instantiate(res, root_inode);
1507 EXPORT_SYMBOL(d_alloc_root);
1509 static struct dentry * __d_find_any_alias(struct inode *inode)
1511 struct dentry *alias;
1513 if (list_empty(&inode->i_dentry))
1515 alias = list_first_entry(&inode->i_dentry, struct dentry, d_alias);
1520 static struct dentry * d_find_any_alias(struct inode *inode)
1524 spin_lock(&inode->i_lock);
1525 de = __d_find_any_alias(inode);
1526 spin_unlock(&inode->i_lock);
1532 * d_obtain_alias - find or allocate a dentry for a given inode
1533 * @inode: inode to allocate the dentry for
1535 * Obtain a dentry for an inode resulting from NFS filehandle conversion or
1536 * similar open by handle operations. The returned dentry may be anonymous,
1537 * or may have a full name (if the inode was already in the cache).
1539 * When called on a directory inode, we must ensure that the inode only ever
1540 * has one dentry. If a dentry is found, that is returned instead of
1541 * allocating a new one.
1543 * On successful return, the reference to the inode has been transferred
1544 * to the dentry. In case of an error the reference on the inode is released.
1545 * To make it easier to use in export operations a %NULL or IS_ERR inode may
1546 * be passed in and will be the error will be propagate to the return value,
1547 * with a %NULL @inode replaced by ERR_PTR(-ESTALE).
1549 struct dentry *d_obtain_alias(struct inode *inode)
1551 static const struct qstr anonstring = { .name = "" };
1556 return ERR_PTR(-ESTALE);
1558 return ERR_CAST(inode);
1560 res = d_find_any_alias(inode);
1564 tmp = d_alloc(NULL, &anonstring);
1566 res = ERR_PTR(-ENOMEM);
1569 tmp->d_parent = tmp; /* make sure dput doesn't croak */
1572 spin_lock(&inode->i_lock);
1573 res = __d_find_any_alias(inode);
1575 spin_unlock(&inode->i_lock);
1580 /* attach a disconnected dentry */
1581 spin_lock(&tmp->d_lock);
1582 tmp->d_sb = inode->i_sb;
1583 d_set_d_op(tmp, tmp->d_sb->s_d_op);
1584 tmp->d_inode = inode;
1585 tmp->d_flags |= DCACHE_DISCONNECTED;
1586 list_add(&tmp->d_alias, &inode->i_dentry);
1587 hlist_bl_lock(&tmp->d_sb->s_anon);
1588 hlist_bl_add_head(&tmp->d_hash, &tmp->d_sb->s_anon);
1589 hlist_bl_unlock(&tmp->d_sb->s_anon);
1590 spin_unlock(&tmp->d_lock);
1591 spin_unlock(&inode->i_lock);
1592 security_d_instantiate(tmp, inode);
1597 if (res && !IS_ERR(res))
1598 security_d_instantiate(res, inode);
1602 EXPORT_SYMBOL(d_obtain_alias);
1605 * d_splice_alias - splice a disconnected dentry into the tree if one exists
1606 * @inode: the inode which may have a disconnected dentry
1607 * @dentry: a negative dentry which we want to point to the inode.
1609 * If inode is a directory and has a 'disconnected' dentry (i.e. IS_ROOT and
1610 * DCACHE_DISCONNECTED), then d_move that in place of the given dentry
1611 * and return it, else simply d_add the inode to the dentry and return NULL.
1613 * This is needed in the lookup routine of any filesystem that is exportable
1614 * (via knfsd) so that we can build dcache paths to directories effectively.
1616 * If a dentry was found and moved, then it is returned. Otherwise NULL
1617 * is returned. This matches the expected return value of ->lookup.
1620 struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
1622 struct dentry *new = NULL;
1624 if (inode && S_ISDIR(inode->i_mode)) {
1625 spin_lock(&inode->i_lock);
1626 new = __d_find_alias(inode, 1);
1628 BUG_ON(!(new->d_flags & DCACHE_DISCONNECTED));
1629 spin_unlock(&inode->i_lock);
1630 security_d_instantiate(new, inode);
1631 d_move(new, dentry);
1634 /* already taking inode->i_lock, so d_add() by hand */
1635 __d_instantiate(dentry, inode);
1636 spin_unlock(&inode->i_lock);
1637 security_d_instantiate(dentry, inode);
1641 d_add(dentry, inode);
1644 EXPORT_SYMBOL(d_splice_alias);
1647 * d_add_ci - lookup or allocate new dentry with case-exact name
1648 * @inode: the inode case-insensitive lookup has found
1649 * @dentry: the negative dentry that was passed to the parent's lookup func
1650 * @name: the case-exact name to be associated with the returned dentry
1652 * This is to avoid filling the dcache with case-insensitive names to the
1653 * same inode, only the actual correct case is stored in the dcache for
1654 * case-insensitive filesystems.
1656 * For a case-insensitive lookup match and if the the case-exact dentry
1657 * already exists in in the dcache, use it and return it.
1659 * If no entry exists with the exact case name, allocate new dentry with
1660 * the exact case, and return the spliced entry.
1662 struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
1666 struct dentry *found;
1670 * First check if a dentry matching the name already exists,
1671 * if not go ahead and create it now.
1673 found = d_hash_and_lookup(dentry->d_parent, name);
1675 new = d_alloc(dentry->d_parent, name);
1681 found = d_splice_alias(inode, new);
1690 * If a matching dentry exists, and it's not negative use it.
1692 * Decrement the reference count to balance the iget() done
1695 if (found->d_inode) {
1696 if (unlikely(found->d_inode != inode)) {
1697 /* This can't happen because bad inodes are unhashed. */
1698 BUG_ON(!is_bad_inode(inode));
1699 BUG_ON(!is_bad_inode(found->d_inode));
1706 * Negative dentry: instantiate it unless the inode is a directory and
1707 * already has a dentry.
1709 spin_lock(&inode->i_lock);
1710 if (!S_ISDIR(inode->i_mode) || list_empty(&inode->i_dentry)) {
1711 __d_instantiate(found, inode);
1712 spin_unlock(&inode->i_lock);
1713 security_d_instantiate(found, inode);
1718 * In case a directory already has a (disconnected) entry grab a
1719 * reference to it, move it in place and use it.
1721 new = list_entry(inode->i_dentry.next, struct dentry, d_alias);
1723 spin_unlock(&inode->i_lock);
1724 security_d_instantiate(found, inode);
1732 return ERR_PTR(error);
1734 EXPORT_SYMBOL(d_add_ci);
1737 * __d_lookup_rcu - search for a dentry (racy, store-free)
1738 * @parent: parent dentry
1739 * @name: qstr of name we wish to find
1740 * @seq: returns d_seq value at the point where the dentry was found
1741 * @inode: returns dentry->d_inode when the inode was found valid.
1742 * Returns: dentry, or NULL
1744 * __d_lookup_rcu is the dcache lookup function for rcu-walk name
1745 * resolution (store-free path walking) design described in
1746 * Documentation/filesystems/path-lookup.txt.
1748 * This is not to be used outside core vfs.
1750 * __d_lookup_rcu must only be used in rcu-walk mode, ie. with vfsmount lock
1751 * held, and rcu_read_lock held. The returned dentry must not be stored into
1752 * without taking d_lock and checking d_seq sequence count against @seq
1755 * A refcount may be taken on the found dentry with the __d_rcu_to_refcount
1758 * Alternatively, __d_lookup_rcu may be called again to look up the child of
1759 * the returned dentry, so long as its parent's seqlock is checked after the
1760 * child is looked up. Thus, an interlocking stepping of sequence lock checks
1761 * is formed, giving integrity down the path walk.
1763 struct dentry *__d_lookup_rcu(struct dentry *parent, struct qstr *name,
1764 unsigned *seq, struct inode **inode)
1766 unsigned int len = name->len;
1767 unsigned int hash = name->hash;
1768 const unsigned char *str = name->name;
1769 struct hlist_bl_head *b = d_hash(parent, hash);
1770 struct hlist_bl_node *node;
1771 struct dentry *dentry;
1774 * Note: There is significant duplication with __d_lookup_rcu which is
1775 * required to prevent single threaded performance regressions
1776 * especially on architectures where smp_rmb (in seqcounts) are costly.
1777 * Keep the two functions in sync.
1781 * The hash list is protected using RCU.
1783 * Carefully use d_seq when comparing a candidate dentry, to avoid
1784 * races with d_move().
1786 * It is possible that concurrent renames can mess up our list
1787 * walk here and result in missing our dentry, resulting in the
1788 * false-negative result. d_lookup() protects against concurrent
1789 * renames using rename_lock seqlock.
1791 * See Documentation/filesystems/path-lookup.txt for more details.
1793 hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
1798 if (dentry->d_name.hash != hash)
1802 *seq = read_seqcount_begin(&dentry->d_seq);
1803 if (dentry->d_parent != parent)
1805 if (d_unhashed(dentry))
1807 tlen = dentry->d_name.len;
1808 tname = dentry->d_name.name;
1809 i = dentry->d_inode;
1814 * This seqcount check is required to ensure name and
1815 * len are loaded atomically, so as not to walk off the
1816 * edge of memory when walking. If we could load this
1817 * atomically some other way, we could drop this check.
1819 if (read_seqcount_retry(&dentry->d_seq, *seq))
1821 if (parent->d_flags & DCACHE_OP_COMPARE) {
1822 if (parent->d_op->d_compare(parent, *inode,
1827 if (dentry_cmp(tname, tlen, str, len))
1831 * No extra seqcount check is required after the name
1832 * compare. The caller must perform a seqcount check in
1833 * order to do anything useful with the returned dentry
1843 * d_lookup - search for a dentry
1844 * @parent: parent dentry
1845 * @name: qstr of name we wish to find
1846 * Returns: dentry, or NULL
1848 * d_lookup searches the children of the parent dentry for the name in
1849 * question. If the dentry is found its reference count is incremented and the
1850 * dentry is returned. The caller must use dput to free the entry when it has
1851 * finished using it. %NULL is returned if the dentry does not exist.
1853 struct dentry *d_lookup(struct dentry *parent, struct qstr *name)
1855 struct dentry *dentry;
1859 seq = read_seqbegin(&rename_lock);
1860 dentry = __d_lookup(parent, name);
1863 } while (read_seqretry(&rename_lock, seq));
1866 EXPORT_SYMBOL(d_lookup);
1869 * __d_lookup - search for a dentry (racy)
1870 * @parent: parent dentry
1871 * @name: qstr of name we wish to find
1872 * Returns: dentry, or NULL
1874 * __d_lookup is like d_lookup, however it may (rarely) return a
1875 * false-negative result due to unrelated rename activity.
1877 * __d_lookup is slightly faster by avoiding rename_lock read seqlock,
1878 * however it must be used carefully, eg. with a following d_lookup in
1879 * the case of failure.
1881 * __d_lookup callers must be commented.
1883 struct dentry *__d_lookup(struct dentry *parent, struct qstr *name)
1885 unsigned int len = name->len;
1886 unsigned int hash = name->hash;
1887 const unsigned char *str = name->name;
1888 struct hlist_bl_head *b = d_hash(parent, hash);
1889 struct hlist_bl_node *node;
1890 struct dentry *found = NULL;
1891 struct dentry *dentry;
1894 * Note: There is significant duplication with __d_lookup_rcu which is
1895 * required to prevent single threaded performance regressions
1896 * especially on architectures where smp_rmb (in seqcounts) are costly.
1897 * Keep the two functions in sync.
1901 * The hash list is protected using RCU.
1903 * Take d_lock when comparing a candidate dentry, to avoid races
1906 * It is possible that concurrent renames can mess up our list
1907 * walk here and result in missing our dentry, resulting in the
1908 * false-negative result. d_lookup() protects against concurrent
1909 * renames using rename_lock seqlock.
1911 * See Documentation/filesystems/path-lookup.txt for more details.
1915 hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
1919 if (dentry->d_name.hash != hash)
1922 spin_lock(&dentry->d_lock);
1923 if (dentry->d_parent != parent)
1925 if (d_unhashed(dentry))
1929 * It is safe to compare names since d_move() cannot
1930 * change the qstr (protected by d_lock).
1932 tlen = dentry->d_name.len;
1933 tname = dentry->d_name.name;
1934 if (parent->d_flags & DCACHE_OP_COMPARE) {
1935 if (parent->d_op->d_compare(parent, parent->d_inode,
1936 dentry, dentry->d_inode,
1940 if (dentry_cmp(tname, tlen, str, len))
1946 spin_unlock(&dentry->d_lock);
1949 spin_unlock(&dentry->d_lock);
1957 * d_hash_and_lookup - hash the qstr then search for a dentry
1958 * @dir: Directory to search in
1959 * @name: qstr of name we wish to find
1961 * On hash failure or on lookup failure NULL is returned.
1963 struct dentry *d_hash_and_lookup(struct dentry *dir, struct qstr *name)
1965 struct dentry *dentry = NULL;
1968 * Check for a fs-specific hash function. Note that we must
1969 * calculate the standard hash first, as the d_op->d_hash()
1970 * routine may choose to leave the hash value unchanged.
1972 name->hash = full_name_hash(name->name, name->len);
1973 if (dir->d_flags & DCACHE_OP_HASH) {
1974 if (dir->d_op->d_hash(dir, dir->d_inode, name) < 0)
1977 dentry = d_lookup(dir, name);
1983 * d_validate - verify dentry provided from insecure source (deprecated)
1984 * @dentry: The dentry alleged to be valid child of @dparent
1985 * @dparent: The parent dentry (known to be valid)
1987 * An insecure source has sent us a dentry, here we verify it and dget() it.
1988 * This is used by ncpfs in its readdir implementation.
1989 * Zero is returned in the dentry is invalid.
1991 * This function is slow for big directories, and deprecated, do not use it.
1993 int d_validate(struct dentry *dentry, struct dentry *dparent)
1995 struct dentry *child;
1997 spin_lock(&dparent->d_lock);
1998 list_for_each_entry(child, &dparent->d_subdirs, d_u.d_child) {
1999 if (dentry == child) {
2000 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
2001 __dget_dlock(dentry);
2002 spin_unlock(&dentry->d_lock);
2003 spin_unlock(&dparent->d_lock);
2007 spin_unlock(&dparent->d_lock);
2011 EXPORT_SYMBOL(d_validate);
2014 * When a file is deleted, we have two options:
2015 * - turn this dentry into a negative dentry
2016 * - unhash this dentry and free it.
2018 * Usually, we want to just turn this into
2019 * a negative dentry, but if anybody else is
2020 * currently using the dentry or the inode
2021 * we can't do that and we fall back on removing
2022 * it from the hash queues and waiting for
2023 * it to be deleted later when it has no users
2027 * d_delete - delete a dentry
2028 * @dentry: The dentry to delete
2030 * Turn the dentry into a negative dentry if possible, otherwise
2031 * remove it from the hash queues so it can be deleted later
2034 void d_delete(struct dentry * dentry)
2036 struct inode *inode;
2039 * Are we the only user?
2042 spin_lock(&dentry->d_lock);
2043 inode = dentry->d_inode;
2044 isdir = S_ISDIR(inode->i_mode);
2045 if (dentry->d_count == 1) {
2046 if (inode && !spin_trylock(&inode->i_lock)) {
2047 spin_unlock(&dentry->d_lock);
2051 dentry->d_flags &= ~DCACHE_CANT_MOUNT;
2052 dentry_unlink_inode(dentry);
2053 fsnotify_nameremove(dentry, isdir);
2057 if (!d_unhashed(dentry))
2060 spin_unlock(&dentry->d_lock);
2062 fsnotify_nameremove(dentry, isdir);
2064 EXPORT_SYMBOL(d_delete);
2066 static void __d_rehash(struct dentry * entry, struct hlist_bl_head *b)
2068 BUG_ON(!d_unhashed(entry));
2070 entry->d_flags |= DCACHE_RCUACCESS;
2071 hlist_bl_add_head_rcu(&entry->d_hash, b);
2075 static void _d_rehash(struct dentry * entry)
2077 __d_rehash(entry, d_hash(entry->d_parent, entry->d_name.hash));
2081 * d_rehash - add an entry back to the hash
2082 * @entry: dentry to add to the hash
2084 * Adds a dentry to the hash according to its name.
2087 void d_rehash(struct dentry * entry)
2089 spin_lock(&entry->d_lock);
2091 spin_unlock(&entry->d_lock);
2093 EXPORT_SYMBOL(d_rehash);
2096 * dentry_update_name_case - update case insensitive dentry with a new name
2097 * @dentry: dentry to be updated
2100 * Update a case insensitive dentry with new case of name.
2102 * dentry must have been returned by d_lookup with name @name. Old and new
2103 * name lengths must match (ie. no d_compare which allows mismatched name
2106 * Parent inode i_mutex must be held over d_lookup and into this call (to
2107 * keep renames and concurrent inserts, and readdir(2) away).
2109 void dentry_update_name_case(struct dentry *dentry, struct qstr *name)
2111 BUG_ON(!mutex_is_locked(&dentry->d_parent->d_inode->i_mutex));
2112 BUG_ON(dentry->d_name.len != name->len); /* d_lookup gives this */
2114 spin_lock(&dentry->d_lock);
2115 write_seqcount_begin(&dentry->d_seq);
2116 memcpy((unsigned char *)dentry->d_name.name, name->name, name->len);
2117 write_seqcount_end(&dentry->d_seq);
2118 spin_unlock(&dentry->d_lock);
2120 EXPORT_SYMBOL(dentry_update_name_case);
2122 static void switch_names(struct dentry *dentry, struct dentry *target)
2124 if (dname_external(target)) {
2125 if (dname_external(dentry)) {
2127 * Both external: swap the pointers
2129 swap(target->d_name.name, dentry->d_name.name);
2132 * dentry:internal, target:external. Steal target's
2133 * storage and make target internal.
2135 memcpy(target->d_iname, dentry->d_name.name,
2136 dentry->d_name.len + 1);
2137 dentry->d_name.name = target->d_name.name;
2138 target->d_name.name = target->d_iname;
2141 if (dname_external(dentry)) {
2143 * dentry:external, target:internal. Give dentry's
2144 * storage to target and make dentry internal
2146 memcpy(dentry->d_iname, target->d_name.name,
2147 target->d_name.len + 1);
2148 target->d_name.name = dentry->d_name.name;
2149 dentry->d_name.name = dentry->d_iname;
2152 * Both are internal. Just copy target to dentry
2154 memcpy(dentry->d_iname, target->d_name.name,
2155 target->d_name.len + 1);
2156 dentry->d_name.len = target->d_name.len;
2160 swap(dentry->d_name.len, target->d_name.len);
2163 static void dentry_lock_for_move(struct dentry *dentry, struct dentry *target)
2166 * XXXX: do we really need to take target->d_lock?
2168 if (IS_ROOT(dentry) || dentry->d_parent == target->d_parent)
2169 spin_lock(&target->d_parent->d_lock);
2171 if (d_ancestor(dentry->d_parent, target->d_parent)) {
2172 spin_lock(&dentry->d_parent->d_lock);
2173 spin_lock_nested(&target->d_parent->d_lock,
2174 DENTRY_D_LOCK_NESTED);
2176 spin_lock(&target->d_parent->d_lock);
2177 spin_lock_nested(&dentry->d_parent->d_lock,
2178 DENTRY_D_LOCK_NESTED);
2181 if (target < dentry) {
2182 spin_lock_nested(&target->d_lock, 2);
2183 spin_lock_nested(&dentry->d_lock, 3);
2185 spin_lock_nested(&dentry->d_lock, 2);
2186 spin_lock_nested(&target->d_lock, 3);
2190 static void dentry_unlock_parents_for_move(struct dentry *dentry,
2191 struct dentry *target)
2193 if (target->d_parent != dentry->d_parent)
2194 spin_unlock(&dentry->d_parent->d_lock);
2195 if (target->d_parent != target)
2196 spin_unlock(&target->d_parent->d_lock);
2200 * When switching names, the actual string doesn't strictly have to
2201 * be preserved in the target - because we're dropping the target
2202 * anyway. As such, we can just do a simple memcpy() to copy over
2203 * the new name before we switch.
2205 * Note that we have to be a lot more careful about getting the hash
2206 * switched - we have to switch the hash value properly even if it
2207 * then no longer matches the actual (corrupted) string of the target.
2208 * The hash value has to match the hash queue that the dentry is on..
2211 * d_move - move a dentry
2212 * @dentry: entry to move
2213 * @target: new dentry
2215 * Update the dcache to reflect the move of a file name. Negative
2216 * dcache entries should not be moved in this way.
2218 void d_move(struct dentry * dentry, struct dentry * target)
2220 if (!dentry->d_inode)
2221 printk(KERN_WARNING "VFS: moving negative dcache entry\n");
2223 BUG_ON(d_ancestor(dentry, target));
2224 BUG_ON(d_ancestor(target, dentry));
2226 write_seqlock(&rename_lock);
2228 dentry_lock_for_move(dentry, target);
2230 write_seqcount_begin(&dentry->d_seq);
2231 write_seqcount_begin(&target->d_seq);
2233 /* __d_drop does write_seqcount_barrier, but they're OK to nest. */
2236 * Move the dentry to the target hash queue. Don't bother checking
2237 * for the same hash queue because of how unlikely it is.
2240 __d_rehash(dentry, d_hash(target->d_parent, target->d_name.hash));
2242 /* Unhash the target: dput() will then get rid of it */
2245 list_del(&dentry->d_u.d_child);
2246 list_del(&target->d_u.d_child);
2248 /* Switch the names.. */
2249 switch_names(dentry, target);
2250 swap(dentry->d_name.hash, target->d_name.hash);
2252 /* ... and switch the parents */
2253 if (IS_ROOT(dentry)) {
2254 dentry->d_parent = target->d_parent;
2255 target->d_parent = target;
2256 INIT_LIST_HEAD(&target->d_u.d_child);
2258 swap(dentry->d_parent, target->d_parent);
2260 /* And add them back to the (new) parent lists */
2261 list_add(&target->d_u.d_child, &target->d_parent->d_subdirs);
2264 list_add(&dentry->d_u.d_child, &dentry->d_parent->d_subdirs);
2266 write_seqcount_end(&target->d_seq);
2267 write_seqcount_end(&dentry->d_seq);
2269 dentry_unlock_parents_for_move(dentry, target);
2270 spin_unlock(&target->d_lock);
2271 fsnotify_d_move(dentry);
2272 spin_unlock(&dentry->d_lock);
2273 write_sequnlock(&rename_lock);
2275 EXPORT_SYMBOL(d_move);
2278 * d_ancestor - search for an ancestor
2279 * @p1: ancestor dentry
2282 * Returns the ancestor dentry of p2 which is a child of p1, if p1 is
2283 * an ancestor of p2, else NULL.
2285 struct dentry *d_ancestor(struct dentry *p1, struct dentry *p2)
2289 for (p = p2; !IS_ROOT(p); p = p->d_parent) {
2290 if (p->d_parent == p1)
2297 * This helper attempts to cope with remotely renamed directories
2299 * It assumes that the caller is already holding
2300 * dentry->d_parent->d_inode->i_mutex and the inode->i_lock
2302 * Note: If ever the locking in lock_rename() changes, then please
2303 * remember to update this too...
2305 static struct dentry *__d_unalias(struct inode *inode,
2306 struct dentry *dentry, struct dentry *alias)
2308 struct mutex *m1 = NULL, *m2 = NULL;
2311 /* If alias and dentry share a parent, then no extra locks required */
2312 if (alias->d_parent == dentry->d_parent)
2315 /* Check for loops */
2316 ret = ERR_PTR(-ELOOP);
2317 if (d_ancestor(alias, dentry))
2320 /* See lock_rename() */
2321 ret = ERR_PTR(-EBUSY);
2322 if (!mutex_trylock(&dentry->d_sb->s_vfs_rename_mutex))
2324 m1 = &dentry->d_sb->s_vfs_rename_mutex;
2325 if (!mutex_trylock(&alias->d_parent->d_inode->i_mutex))
2327 m2 = &alias->d_parent->d_inode->i_mutex;
2329 d_move(alias, dentry);
2332 spin_unlock(&inode->i_lock);
2341 * Prepare an anonymous dentry for life in the superblock's dentry tree as a
2342 * named dentry in place of the dentry to be replaced.
2343 * returns with anon->d_lock held!
2345 static void __d_materialise_dentry(struct dentry *dentry, struct dentry *anon)
2347 struct dentry *dparent, *aparent;
2349 dentry_lock_for_move(anon, dentry);
2351 write_seqcount_begin(&dentry->d_seq);
2352 write_seqcount_begin(&anon->d_seq);
2354 dparent = dentry->d_parent;
2355 aparent = anon->d_parent;
2357 switch_names(dentry, anon);
2358 swap(dentry->d_name.hash, anon->d_name.hash);
2360 dentry->d_parent = (aparent == anon) ? dentry : aparent;
2361 list_del(&dentry->d_u.d_child);
2362 if (!IS_ROOT(dentry))
2363 list_add(&dentry->d_u.d_child, &dentry->d_parent->d_subdirs);
2365 INIT_LIST_HEAD(&dentry->d_u.d_child);
2367 anon->d_parent = (dparent == dentry) ? anon : dparent;
2368 list_del(&anon->d_u.d_child);
2370 list_add(&anon->d_u.d_child, &anon->d_parent->d_subdirs);
2372 INIT_LIST_HEAD(&anon->d_u.d_child);
2374 write_seqcount_end(&dentry->d_seq);
2375 write_seqcount_end(&anon->d_seq);
2377 dentry_unlock_parents_for_move(anon, dentry);
2378 spin_unlock(&dentry->d_lock);
2380 /* anon->d_lock still locked, returns locked */
2381 anon->d_flags &= ~DCACHE_DISCONNECTED;
2385 * d_materialise_unique - introduce an inode into the tree
2386 * @dentry: candidate dentry
2387 * @inode: inode to bind to the dentry, to which aliases may be attached
2389 * Introduces an dentry into the tree, substituting an extant disconnected
2390 * root directory alias in its place if there is one
2392 struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
2394 struct dentry *actual;
2396 BUG_ON(!d_unhashed(dentry));
2400 __d_instantiate(dentry, NULL);
2405 spin_lock(&inode->i_lock);
2407 if (S_ISDIR(inode->i_mode)) {
2408 struct dentry *alias;
2410 /* Does an aliased dentry already exist? */
2411 alias = __d_find_alias(inode, 0);
2414 /* Is this an anonymous mountpoint that we could splice
2416 if (IS_ROOT(alias)) {
2417 __d_materialise_dentry(dentry, alias);
2421 /* Nope, but we must(!) avoid directory aliasing */
2422 actual = __d_unalias(inode, dentry, alias);
2429 /* Add a unique reference */
2430 actual = __d_instantiate_unique(dentry, inode);
2434 BUG_ON(!d_unhashed(actual));
2436 spin_lock(&actual->d_lock);
2439 spin_unlock(&actual->d_lock);
2440 spin_unlock(&inode->i_lock);
2442 if (actual == dentry) {
2443 security_d_instantiate(dentry, inode);
2450 EXPORT_SYMBOL_GPL(d_materialise_unique);
2452 static int prepend(char **buffer, int *buflen, const char *str, int namelen)
2456 return -ENAMETOOLONG;
2458 memcpy(*buffer, str, namelen);
2462 static int prepend_name(char **buffer, int *buflen, struct qstr *name)
2464 return prepend(buffer, buflen, name->name, name->len);
2468 * prepend_path - Prepend path string to a buffer
2469 * @path: the dentry/vfsmount to report
2470 * @root: root vfsmnt/dentry (may be modified by this function)
2471 * @buffer: pointer to the end of the buffer
2472 * @buflen: pointer to buffer length
2474 * Caller holds the rename_lock.
2476 * If path is not reachable from the supplied root, then the value of
2477 * root is changed (without modifying refcounts).
2479 static int prepend_path(const struct path *path, struct path *root,
2480 char **buffer, int *buflen)
2482 struct dentry *dentry = path->dentry;
2483 struct vfsmount *vfsmnt = path->mnt;
2487 br_read_lock(vfsmount_lock);
2488 while (dentry != root->dentry || vfsmnt != root->mnt) {
2489 struct dentry * parent;
2491 if (dentry == vfsmnt->mnt_root || IS_ROOT(dentry)) {
2493 if (vfsmnt->mnt_parent == vfsmnt) {
2496 dentry = vfsmnt->mnt_mountpoint;
2497 vfsmnt = vfsmnt->mnt_parent;
2500 parent = dentry->d_parent;
2502 spin_lock(&dentry->d_lock);
2503 error = prepend_name(buffer, buflen, &dentry->d_name);
2504 spin_unlock(&dentry->d_lock);
2506 error = prepend(buffer, buflen, "/", 1);
2515 if (!error && !slash)
2516 error = prepend(buffer, buflen, "/", 1);
2518 br_read_unlock(vfsmount_lock);
2523 * Filesystems needing to implement special "root names"
2524 * should do so with ->d_dname()
2526 if (IS_ROOT(dentry) &&
2527 (dentry->d_name.len != 1 || dentry->d_name.name[0] != '/')) {
2528 WARN(1, "Root dentry has weird name <%.*s>\n",
2529 (int) dentry->d_name.len, dentry->d_name.name);
2532 root->dentry = dentry;
2537 * __d_path - return the path of a dentry
2538 * @path: the dentry/vfsmount to report
2539 * @root: root vfsmnt/dentry (may be modified by this function)
2540 * @buf: buffer to return value in
2541 * @buflen: buffer length
2543 * Convert a dentry into an ASCII path name.
2545 * Returns a pointer into the buffer or an error code if the
2546 * path was too long.
2548 * "buflen" should be positive.
2550 * If path is not reachable from the supplied root, then the value of
2551 * root is changed (without modifying refcounts).
2553 char *__d_path(const struct path *path, struct path *root,
2554 char *buf, int buflen)
2556 char *res = buf + buflen;
2559 prepend(&res, &buflen, "\0", 1);
2560 write_seqlock(&rename_lock);
2561 error = prepend_path(path, root, &res, &buflen);
2562 write_sequnlock(&rename_lock);
2565 return ERR_PTR(error);
2570 * same as __d_path but appends "(deleted)" for unlinked files.
2572 static int path_with_deleted(const struct path *path, struct path *root,
2573 char **buf, int *buflen)
2575 prepend(buf, buflen, "\0", 1);
2576 if (d_unlinked(path->dentry)) {
2577 int error = prepend(buf, buflen, " (deleted)", 10);
2582 return prepend_path(path, root, buf, buflen);
2585 static int prepend_unreachable(char **buffer, int *buflen)
2587 return prepend(buffer, buflen, "(unreachable)", 13);
2591 * d_path - return the path of a dentry
2592 * @path: path to report
2593 * @buf: buffer to return value in
2594 * @buflen: buffer length
2596 * Convert a dentry into an ASCII path name. If the entry has been deleted
2597 * the string " (deleted)" is appended. Note that this is ambiguous.
2599 * Returns a pointer into the buffer or an error code if the path was
2600 * too long. Note: Callers should use the returned pointer, not the passed
2601 * in buffer, to use the name! The implementation often starts at an offset
2602 * into the buffer, and may leave 0 bytes at the start.
2604 * "buflen" should be positive.
2606 char *d_path(const struct path *path, char *buf, int buflen)
2608 char *res = buf + buflen;
2614 * We have various synthetic filesystems that never get mounted. On
2615 * these filesystems dentries are never used for lookup purposes, and
2616 * thus don't need to be hashed. They also don't need a name until a
2617 * user wants to identify the object in /proc/pid/fd/. The little hack
2618 * below allows us to generate a name for these objects on demand:
2620 if (path->dentry->d_op && path->dentry->d_op->d_dname)
2621 return path->dentry->d_op->d_dname(path->dentry, buf, buflen);
2623 get_fs_root(current->fs, &root);
2624 write_seqlock(&rename_lock);
2626 error = path_with_deleted(path, &tmp, &res, &buflen);
2628 res = ERR_PTR(error);
2629 write_sequnlock(&rename_lock);
2633 EXPORT_SYMBOL(d_path);
2636 * d_path_with_unreachable - return the path of a dentry
2637 * @path: path to report
2638 * @buf: buffer to return value in
2639 * @buflen: buffer length
2641 * The difference from d_path() is that this prepends "(unreachable)"
2642 * to paths which are unreachable from the current process' root.
2644 char *d_path_with_unreachable(const struct path *path, char *buf, int buflen)
2646 char *res = buf + buflen;
2651 if (path->dentry->d_op && path->dentry->d_op->d_dname)
2652 return path->dentry->d_op->d_dname(path->dentry, buf, buflen);
2654 get_fs_root(current->fs, &root);
2655 write_seqlock(&rename_lock);
2657 error = path_with_deleted(path, &tmp, &res, &buflen);
2658 if (!error && !path_equal(&tmp, &root))
2659 error = prepend_unreachable(&res, &buflen);
2660 write_sequnlock(&rename_lock);
2663 res = ERR_PTR(error);
2669 * Helper function for dentry_operations.d_dname() members
2671 char *dynamic_dname(struct dentry *dentry, char *buffer, int buflen,
2672 const char *fmt, ...)
2678 va_start(args, fmt);
2679 sz = vsnprintf(temp, sizeof(temp), fmt, args) + 1;
2682 if (sz > sizeof(temp) || sz > buflen)
2683 return ERR_PTR(-ENAMETOOLONG);
2685 buffer += buflen - sz;
2686 return memcpy(buffer, temp, sz);
2690 * Write full pathname from the root of the filesystem into the buffer.
2692 static char *__dentry_path(struct dentry *dentry, char *buf, int buflen)
2694 char *end = buf + buflen;
2697 prepend(&end, &buflen, "\0", 1);
2704 while (!IS_ROOT(dentry)) {
2705 struct dentry *parent = dentry->d_parent;
2709 spin_lock(&dentry->d_lock);
2710 error = prepend_name(&end, &buflen, &dentry->d_name);
2711 spin_unlock(&dentry->d_lock);
2712 if (error != 0 || prepend(&end, &buflen, "/", 1) != 0)
2720 return ERR_PTR(-ENAMETOOLONG);
2723 char *dentry_path_raw(struct dentry *dentry, char *buf, int buflen)
2727 write_seqlock(&rename_lock);
2728 retval = __dentry_path(dentry, buf, buflen);
2729 write_sequnlock(&rename_lock);
2733 EXPORT_SYMBOL(dentry_path_raw);
2735 char *dentry_path(struct dentry *dentry, char *buf, int buflen)
2740 write_seqlock(&rename_lock);
2741 if (d_unlinked(dentry)) {
2743 if (prepend(&p, &buflen, "//deleted", 10) != 0)
2747 retval = __dentry_path(dentry, buf, buflen);
2748 write_sequnlock(&rename_lock);
2749 if (!IS_ERR(retval) && p)
2750 *p = '/'; /* restore '/' overriden with '\0' */
2753 return ERR_PTR(-ENAMETOOLONG);
2757 * NOTE! The user-level library version returns a
2758 * character pointer. The kernel system call just
2759 * returns the length of the buffer filled (which
2760 * includes the ending '\0' character), or a negative
2761 * error value. So libc would do something like
2763 * char *getcwd(char * buf, size_t size)
2767 * retval = sys_getcwd(buf, size);
2774 SYSCALL_DEFINE2(getcwd, char __user *, buf, unsigned long, size)
2777 struct path pwd, root;
2778 char *page = (char *) __get_free_page(GFP_USER);
2783 get_fs_root_and_pwd(current->fs, &root, &pwd);
2786 write_seqlock(&rename_lock);
2787 if (!d_unlinked(pwd.dentry)) {
2789 struct path tmp = root;
2790 char *cwd = page + PAGE_SIZE;
2791 int buflen = PAGE_SIZE;
2793 prepend(&cwd, &buflen, "\0", 1);
2794 error = prepend_path(&pwd, &tmp, &cwd, &buflen);
2795 write_sequnlock(&rename_lock);
2800 /* Unreachable from current root */
2801 if (!path_equal(&tmp, &root)) {
2802 error = prepend_unreachable(&cwd, &buflen);
2808 len = PAGE_SIZE + page - cwd;
2811 if (copy_to_user(buf, cwd, len))
2815 write_sequnlock(&rename_lock);
2821 free_page((unsigned long) page);
2826 * Test whether new_dentry is a subdirectory of old_dentry.
2828 * Trivially implemented using the dcache structure
2832 * is_subdir - is new dentry a subdirectory of old_dentry
2833 * @new_dentry: new dentry
2834 * @old_dentry: old dentry
2836 * Returns 1 if new_dentry is a subdirectory of the parent (at any depth).
2837 * Returns 0 otherwise.
2838 * Caller must ensure that "new_dentry" is pinned before calling is_subdir()
2841 int is_subdir(struct dentry *new_dentry, struct dentry *old_dentry)
2846 if (new_dentry == old_dentry)
2850 /* for restarting inner loop in case of seq retry */
2851 seq = read_seqbegin(&rename_lock);
2853 * Need rcu_readlock to protect against the d_parent trashing
2857 if (d_ancestor(old_dentry, new_dentry))
2862 } while (read_seqretry(&rename_lock, seq));
2867 int path_is_under(struct path *path1, struct path *path2)
2869 struct vfsmount *mnt = path1->mnt;
2870 struct dentry *dentry = path1->dentry;
2873 br_read_lock(vfsmount_lock);
2874 if (mnt != path2->mnt) {
2876 if (mnt->mnt_parent == mnt) {
2877 br_read_unlock(vfsmount_lock);
2880 if (mnt->mnt_parent == path2->mnt)
2882 mnt = mnt->mnt_parent;
2884 dentry = mnt->mnt_mountpoint;
2886 res = is_subdir(dentry, path2->dentry);
2887 br_read_unlock(vfsmount_lock);
2890 EXPORT_SYMBOL(path_is_under);
2892 void d_genocide(struct dentry *root)
2894 struct dentry *this_parent;
2895 struct list_head *next;
2899 seq = read_seqbegin(&rename_lock);
2902 spin_lock(&this_parent->d_lock);
2904 next = this_parent->d_subdirs.next;
2906 while (next != &this_parent->d_subdirs) {
2907 struct list_head *tmp = next;
2908 struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
2911 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
2912 if (d_unhashed(dentry) || !dentry->d_inode) {
2913 spin_unlock(&dentry->d_lock);
2916 if (!list_empty(&dentry->d_subdirs)) {
2917 spin_unlock(&this_parent->d_lock);
2918 spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
2919 this_parent = dentry;
2920 spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
2923 if (!(dentry->d_flags & DCACHE_GENOCIDE)) {
2924 dentry->d_flags |= DCACHE_GENOCIDE;
2927 spin_unlock(&dentry->d_lock);
2929 if (this_parent != root) {
2930 struct dentry *child = this_parent;
2931 if (!(this_parent->d_flags & DCACHE_GENOCIDE)) {
2932 this_parent->d_flags |= DCACHE_GENOCIDE;
2933 this_parent->d_count--;
2935 this_parent = try_to_ascend(this_parent, locked, seq);
2938 next = child->d_u.d_child.next;
2941 spin_unlock(&this_parent->d_lock);
2942 if (!locked && read_seqretry(&rename_lock, seq))
2945 write_sequnlock(&rename_lock);
2950 write_seqlock(&rename_lock);
2955 * find_inode_number - check for dentry with name
2956 * @dir: directory to check
2957 * @name: Name to find.
2959 * Check whether a dentry already exists for the given name,
2960 * and return the inode number if it has an inode. Otherwise
2963 * This routine is used to post-process directory listings for
2964 * filesystems using synthetic inode numbers, and is necessary
2965 * to keep getcwd() working.
2968 ino_t find_inode_number(struct dentry *dir, struct qstr *name)
2970 struct dentry * dentry;
2973 dentry = d_hash_and_lookup(dir, name);
2975 if (dentry->d_inode)
2976 ino = dentry->d_inode->i_ino;
2981 EXPORT_SYMBOL(find_inode_number);
2983 static __initdata unsigned long dhash_entries;
2984 static int __init set_dhash_entries(char *str)
2988 dhash_entries = simple_strtoul(str, &str, 0);
2991 __setup("dhash_entries=", set_dhash_entries);
2993 static void __init dcache_init_early(void)
2997 /* If hashes are distributed across NUMA nodes, defer
2998 * hash allocation until vmalloc space is available.
3004 alloc_large_system_hash("Dentry cache",
3005 sizeof(struct hlist_bl_head),
3013 for (loop = 0; loop < (1 << d_hash_shift); loop++)
3014 INIT_HLIST_BL_HEAD(dentry_hashtable + loop);
3017 static void __init dcache_init(void)
3022 * A constructor could be added for stable state like the lists,
3023 * but it is probably not worth it because of the cache nature
3026 dentry_cache = KMEM_CACHE(dentry,
3027 SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|SLAB_MEM_SPREAD);
3029 register_shrinker(&dcache_shrinker);
3031 /* Hash may have been set up in dcache_init_early */
3036 alloc_large_system_hash("Dentry cache",
3037 sizeof(struct hlist_bl_head),
3045 for (loop = 0; loop < (1 << d_hash_shift); loop++)
3046 INIT_HLIST_BL_HEAD(dentry_hashtable + loop);
3049 /* SLAB cache for __getname() consumers */
3050 struct kmem_cache *names_cachep __read_mostly;
3051 EXPORT_SYMBOL(names_cachep);
3053 EXPORT_SYMBOL(d_genocide);
3055 void __init vfs_caches_init_early(void)
3057 dcache_init_early();
3061 void __init vfs_caches_init(unsigned long mempages)
3063 unsigned long reserve;
3065 /* Base hash sizes on available memory, with a reserve equal to
3066 150% of current kernel size */
3068 reserve = min((mempages - nr_free_pages()) * 3/2, mempages - 1);
3069 mempages -= reserve;
3071 names_cachep = kmem_cache_create("names_cache", PATH_MAX, 0,
3072 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
3076 files_init(mempages);