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/export.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>
38 #include <linux/prefetch.h>
39 #include <linux/ratelimit.h>
45 * dcache->d_inode->i_lock protects:
46 * - i_dentry, d_alias, d_inode of aliases
47 * dcache_hash_bucket lock protects:
48 * - the dcache hash table
49 * s_anon bl list spinlock protects:
50 * - the s_anon list (see __d_drop)
51 * dcache_lru_lock protects:
52 * - the dcache lru lists and counters
59 * - d_parent and d_subdirs
60 * - childrens' d_child and d_parent
64 * dentry->d_inode->i_lock
67 * dcache_hash_bucket lock
70 * If there is an ancestor relationship:
71 * dentry->d_parent->...->d_parent->d_lock
73 * dentry->d_parent->d_lock
76 * If no ancestor relationship:
77 * if (dentry1 < dentry2)
81 int sysctl_vfs_cache_pressure __read_mostly = 100;
82 EXPORT_SYMBOL_GPL(sysctl_vfs_cache_pressure);
84 static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dcache_lru_lock);
85 __cacheline_aligned_in_smp DEFINE_SEQLOCK(rename_lock);
87 EXPORT_SYMBOL(rename_lock);
89 static struct kmem_cache *dentry_cache __read_mostly;
92 * read_seqbegin_or_lock - begin a sequence number check or locking block
94 * seq : sequence number to be checked
96 * First try it once optimistically without taking the lock. If that fails,
97 * take the lock. The sequence number is also used as a marker for deciding
98 * whether to be a reader (even) or writer (odd).
99 * N.B. seq must be initialized to an even number to begin with.
101 static inline void read_seqbegin_or_lock(seqlock_t *lock, int *seq)
103 if (!(*seq & 1)) /* Even */
104 *seq = read_seqbegin(lock);
109 static inline int need_seqretry(seqlock_t *lock, int seq)
111 return !(seq & 1) && read_seqretry(lock, seq);
114 static inline void done_seqretry(seqlock_t *lock, int seq)
117 write_sequnlock(lock);
121 * This is the single most critical data structure when it comes
122 * to the dcache: the hashtable for lookups. Somebody should try
123 * to make this good - I've just made it work.
125 * This hash-function tries to avoid losing too many bits of hash
126 * information, yet avoid using a prime hash-size or similar.
128 #define D_HASHBITS d_hash_shift
129 #define D_HASHMASK d_hash_mask
131 static unsigned int d_hash_mask __read_mostly;
132 static unsigned int d_hash_shift __read_mostly;
134 static struct hlist_bl_head *dentry_hashtable __read_mostly;
136 static inline struct hlist_bl_head *d_hash(const struct dentry *parent,
139 hash += (unsigned long) parent / L1_CACHE_BYTES;
140 hash = hash + (hash >> D_HASHBITS);
141 return dentry_hashtable + (hash & D_HASHMASK);
144 /* Statistics gathering. */
145 struct dentry_stat_t dentry_stat = {
149 static DEFINE_PER_CPU(long, nr_dentry);
151 #if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
152 static long get_nr_dentry(void)
156 for_each_possible_cpu(i)
157 sum += per_cpu(nr_dentry, i);
158 return sum < 0 ? 0 : sum;
161 int proc_nr_dentry(ctl_table *table, int write, void __user *buffer,
162 size_t *lenp, loff_t *ppos)
164 dentry_stat.nr_dentry = get_nr_dentry();
165 return proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
170 * Compare 2 name strings, return 0 if they match, otherwise non-zero.
171 * The strings are both count bytes long, and count is non-zero.
173 #ifdef CONFIG_DCACHE_WORD_ACCESS
175 #include <asm/word-at-a-time.h>
177 * NOTE! 'cs' and 'scount' come from a dentry, so it has a
178 * aligned allocation for this particular component. We don't
179 * strictly need the load_unaligned_zeropad() safety, but it
180 * doesn't hurt either.
182 * In contrast, 'ct' and 'tcount' can be from a pathname, and do
183 * need the careful unaligned handling.
185 static inline int dentry_string_cmp(const unsigned char *cs, const unsigned char *ct, unsigned tcount)
187 unsigned long a,b,mask;
190 a = *(unsigned long *)cs;
191 b = load_unaligned_zeropad(ct);
192 if (tcount < sizeof(unsigned long))
194 if (unlikely(a != b))
196 cs += sizeof(unsigned long);
197 ct += sizeof(unsigned long);
198 tcount -= sizeof(unsigned long);
202 mask = ~(~0ul << tcount*8);
203 return unlikely(!!((a ^ b) & mask));
208 static inline int dentry_string_cmp(const unsigned char *cs, const unsigned char *ct, unsigned tcount)
222 static inline int dentry_cmp(const struct dentry *dentry, const unsigned char *ct, unsigned tcount)
224 const unsigned char *cs;
226 * Be careful about RCU walk racing with rename:
227 * use ACCESS_ONCE to fetch the name pointer.
229 * NOTE! Even if a rename will mean that the length
230 * was not loaded atomically, we don't care. The
231 * RCU walk will check the sequence count eventually,
232 * and catch it. And we won't overrun the buffer,
233 * because we're reading the name pointer atomically,
234 * and a dentry name is guaranteed to be properly
235 * terminated with a NUL byte.
237 * End result: even if 'len' is wrong, we'll exit
238 * early because the data cannot match (there can
239 * be no NUL in the ct/tcount data)
241 cs = ACCESS_ONCE(dentry->d_name.name);
242 smp_read_barrier_depends();
243 return dentry_string_cmp(cs, ct, tcount);
246 static void __d_free(struct rcu_head *head)
248 struct dentry *dentry = container_of(head, struct dentry, d_u.d_rcu);
250 WARN_ON(!hlist_unhashed(&dentry->d_alias));
251 if (dname_external(dentry))
252 kfree(dentry->d_name.name);
253 kmem_cache_free(dentry_cache, dentry);
259 static void d_free(struct dentry *dentry)
261 BUG_ON((int)dentry->d_lockref.count > 0);
262 this_cpu_dec(nr_dentry);
263 if (dentry->d_op && dentry->d_op->d_release)
264 dentry->d_op->d_release(dentry);
266 /* if dentry was never visible to RCU, immediate free is OK */
267 if (!(dentry->d_flags & DCACHE_RCUACCESS))
268 __d_free(&dentry->d_u.d_rcu);
270 call_rcu(&dentry->d_u.d_rcu, __d_free);
274 * dentry_rcuwalk_barrier - invalidate in-progress rcu-walk lookups
275 * @dentry: the target dentry
276 * After this call, in-progress rcu-walk path lookup will fail. This
277 * should be called after unhashing, and after changing d_inode (if
278 * the dentry has not already been unhashed).
280 static inline void dentry_rcuwalk_barrier(struct dentry *dentry)
282 assert_spin_locked(&dentry->d_lock);
283 /* Go through a barrier */
284 write_seqcount_barrier(&dentry->d_seq);
288 * Release the dentry's inode, using the filesystem
289 * d_iput() operation if defined. Dentry has no refcount
292 static void dentry_iput(struct dentry * dentry)
293 __releases(dentry->d_lock)
294 __releases(dentry->d_inode->i_lock)
296 struct inode *inode = dentry->d_inode;
298 dentry->d_inode = NULL;
299 hlist_del_init(&dentry->d_alias);
300 spin_unlock(&dentry->d_lock);
301 spin_unlock(&inode->i_lock);
303 fsnotify_inoderemove(inode);
304 if (dentry->d_op && dentry->d_op->d_iput)
305 dentry->d_op->d_iput(dentry, inode);
309 spin_unlock(&dentry->d_lock);
314 * Release the dentry's inode, using the filesystem
315 * d_iput() operation if defined. dentry remains in-use.
317 static void dentry_unlink_inode(struct dentry * dentry)
318 __releases(dentry->d_lock)
319 __releases(dentry->d_inode->i_lock)
321 struct inode *inode = dentry->d_inode;
322 dentry->d_inode = NULL;
323 hlist_del_init(&dentry->d_alias);
324 dentry_rcuwalk_barrier(dentry);
325 spin_unlock(&dentry->d_lock);
326 spin_unlock(&inode->i_lock);
328 fsnotify_inoderemove(inode);
329 if (dentry->d_op && dentry->d_op->d_iput)
330 dentry->d_op->d_iput(dentry, inode);
336 * dentry_lru_(add|del|prune|move_tail) must be called with d_lock held.
338 static void dentry_lru_add(struct dentry *dentry)
340 if (unlikely(!(dentry->d_flags & DCACHE_LRU_LIST))) {
341 spin_lock(&dcache_lru_lock);
342 dentry->d_flags |= DCACHE_LRU_LIST;
343 list_add(&dentry->d_lru, &dentry->d_sb->s_dentry_lru);
344 dentry->d_sb->s_nr_dentry_unused++;
345 dentry_stat.nr_unused++;
346 spin_unlock(&dcache_lru_lock);
350 static void __dentry_lru_del(struct dentry *dentry)
352 list_del_init(&dentry->d_lru);
353 dentry->d_flags &= ~(DCACHE_SHRINK_LIST | DCACHE_LRU_LIST);
354 dentry->d_sb->s_nr_dentry_unused--;
355 dentry_stat.nr_unused--;
359 * Remove a dentry with references from the LRU.
361 static void dentry_lru_del(struct dentry *dentry)
363 if (!list_empty(&dentry->d_lru)) {
364 spin_lock(&dcache_lru_lock);
365 __dentry_lru_del(dentry);
366 spin_unlock(&dcache_lru_lock);
370 static void dentry_lru_move_list(struct dentry *dentry, struct list_head *list)
372 spin_lock(&dcache_lru_lock);
373 if (list_empty(&dentry->d_lru)) {
374 dentry->d_flags |= DCACHE_LRU_LIST;
375 list_add_tail(&dentry->d_lru, list);
376 dentry->d_sb->s_nr_dentry_unused++;
377 dentry_stat.nr_unused++;
379 list_move_tail(&dentry->d_lru, list);
381 spin_unlock(&dcache_lru_lock);
385 * d_kill - kill dentry and return parent
386 * @dentry: dentry to kill
387 * @parent: parent dentry
389 * The dentry must already be unhashed and removed from the LRU.
391 * If this is the root of the dentry tree, return NULL.
393 * dentry->d_lock and parent->d_lock must be held by caller, and are dropped by
396 static struct dentry *d_kill(struct dentry *dentry, struct dentry *parent)
397 __releases(dentry->d_lock)
398 __releases(parent->d_lock)
399 __releases(dentry->d_inode->i_lock)
401 list_del(&dentry->d_u.d_child);
403 * Inform try_to_ascend() that we are no longer attached to the
406 dentry->d_flags |= DCACHE_DENTRY_KILLED;
408 spin_unlock(&parent->d_lock);
411 * dentry_iput drops the locks, at which point nobody (except
412 * transient RCU lookups) can reach this dentry.
419 * Unhash a dentry without inserting an RCU walk barrier or checking that
420 * dentry->d_lock is locked. The caller must take care of that, if
423 static void __d_shrink(struct dentry *dentry)
425 if (!d_unhashed(dentry)) {
426 struct hlist_bl_head *b;
427 if (unlikely(dentry->d_flags & DCACHE_DISCONNECTED))
428 b = &dentry->d_sb->s_anon;
430 b = d_hash(dentry->d_parent, dentry->d_name.hash);
433 __hlist_bl_del(&dentry->d_hash);
434 dentry->d_hash.pprev = NULL;
440 * d_drop - drop a dentry
441 * @dentry: dentry to drop
443 * d_drop() unhashes the entry from the parent dentry hashes, so that it won't
444 * be found through a VFS lookup any more. Note that this is different from
445 * deleting the dentry - d_delete will try to mark the dentry negative if
446 * possible, giving a successful _negative_ lookup, while d_drop will
447 * just make the cache lookup fail.
449 * d_drop() is used mainly for stuff that wants to invalidate a dentry for some
450 * reason (NFS timeouts or autofs deletes).
452 * __d_drop requires dentry->d_lock.
454 void __d_drop(struct dentry *dentry)
456 if (!d_unhashed(dentry)) {
458 dentry_rcuwalk_barrier(dentry);
461 EXPORT_SYMBOL(__d_drop);
463 void d_drop(struct dentry *dentry)
465 spin_lock(&dentry->d_lock);
467 spin_unlock(&dentry->d_lock);
469 EXPORT_SYMBOL(d_drop);
472 * Finish off a dentry we've decided to kill.
473 * dentry->d_lock must be held, returns with it unlocked.
474 * If ref is non-zero, then decrement the refcount too.
475 * Returns dentry requiring refcount drop, or NULL if we're done.
477 static inline struct dentry *dentry_kill(struct dentry *dentry)
478 __releases(dentry->d_lock)
481 struct dentry *parent;
483 inode = dentry->d_inode;
484 if (inode && !spin_trylock(&inode->i_lock)) {
486 spin_unlock(&dentry->d_lock);
488 return dentry; /* try again with same dentry */
493 parent = dentry->d_parent;
494 if (parent && !spin_trylock(&parent->d_lock)) {
496 spin_unlock(&inode->i_lock);
501 * The dentry is now unrecoverably dead to the world.
503 lockref_mark_dead(&dentry->d_lockref);
506 * inform the fs via d_prune that this dentry is about to be
507 * unhashed and destroyed.
509 if ((dentry->d_flags & DCACHE_OP_PRUNE) && !d_unhashed(dentry))
510 dentry->d_op->d_prune(dentry);
512 dentry_lru_del(dentry);
513 /* if it was on the hash then remove it */
515 return d_kill(dentry, parent);
521 * This is complicated by the fact that we do not want to put
522 * dentries that are no longer on any hash chain on the unused
523 * list: we'd much rather just get rid of them immediately.
525 * However, that implies that we have to traverse the dentry
526 * tree upwards to the parents which might _also_ now be
527 * scheduled for deletion (it may have been only waiting for
528 * its last child to go away).
530 * This tail recursion is done by hand as we don't want to depend
531 * on the compiler to always get this right (gcc generally doesn't).
532 * Real recursion would eat up our stack space.
536 * dput - release a dentry
537 * @dentry: dentry to release
539 * Release a dentry. This will drop the usage count and if appropriate
540 * call the dentry unlink method as well as removing it from the queues and
541 * releasing its resources. If the parent dentries were scheduled for release
542 * they too may now get deleted.
544 void dput(struct dentry *dentry)
546 if (unlikely(!dentry))
550 if (lockref_put_or_lock(&dentry->d_lockref))
553 /* Unreachable? Get rid of it */
554 if (unlikely(d_unhashed(dentry)))
557 if (unlikely(dentry->d_flags & DCACHE_OP_DELETE)) {
558 if (dentry->d_op->d_delete(dentry))
562 dentry->d_flags |= DCACHE_REFERENCED;
563 dentry_lru_add(dentry);
565 dentry->d_lockref.count--;
566 spin_unlock(&dentry->d_lock);
570 dentry = dentry_kill(dentry);
577 * d_invalidate - invalidate a dentry
578 * @dentry: dentry to invalidate
580 * Try to invalidate the dentry if it turns out to be
581 * possible. If there are other dentries that can be
582 * reached through this one we can't delete it and we
583 * return -EBUSY. On success we return 0.
588 int d_invalidate(struct dentry * dentry)
591 * If it's already been dropped, return OK.
593 spin_lock(&dentry->d_lock);
594 if (d_unhashed(dentry)) {
595 spin_unlock(&dentry->d_lock);
599 * Check whether to do a partial shrink_dcache
600 * to get rid of unused child entries.
602 if (!list_empty(&dentry->d_subdirs)) {
603 spin_unlock(&dentry->d_lock);
604 shrink_dcache_parent(dentry);
605 spin_lock(&dentry->d_lock);
609 * Somebody else still using it?
611 * If it's a directory, we can't drop it
612 * for fear of somebody re-populating it
613 * with children (even though dropping it
614 * would make it unreachable from the root,
615 * we might still populate it if it was a
616 * working directory or similar).
617 * We also need to leave mountpoints alone,
620 if (dentry->d_lockref.count > 1 && dentry->d_inode) {
621 if (S_ISDIR(dentry->d_inode->i_mode) || d_mountpoint(dentry)) {
622 spin_unlock(&dentry->d_lock);
628 spin_unlock(&dentry->d_lock);
631 EXPORT_SYMBOL(d_invalidate);
633 /* This must be called with d_lock held */
634 static inline void __dget_dlock(struct dentry *dentry)
636 dentry->d_lockref.count++;
639 static inline void __dget(struct dentry *dentry)
641 lockref_get(&dentry->d_lockref);
644 struct dentry *dget_parent(struct dentry *dentry)
650 * Do optimistic parent lookup without any
654 ret = ACCESS_ONCE(dentry->d_parent);
655 gotref = lockref_get_not_zero(&ret->d_lockref);
657 if (likely(gotref)) {
658 if (likely(ret == ACCESS_ONCE(dentry->d_parent)))
665 * Don't need rcu_dereference because we re-check it was correct under
669 ret = dentry->d_parent;
670 spin_lock(&ret->d_lock);
671 if (unlikely(ret != dentry->d_parent)) {
672 spin_unlock(&ret->d_lock);
677 BUG_ON(!ret->d_lockref.count);
678 ret->d_lockref.count++;
679 spin_unlock(&ret->d_lock);
682 EXPORT_SYMBOL(dget_parent);
685 * d_find_alias - grab a hashed alias of inode
686 * @inode: inode in question
687 * @want_discon: flag, used by d_splice_alias, to request
688 * that only a DISCONNECTED alias be returned.
690 * If inode has a hashed alias, or is a directory and has any alias,
691 * acquire the reference to alias and return it. Otherwise return NULL.
692 * Notice that if inode is a directory there can be only one alias and
693 * it can be unhashed only if it has no children, or if it is the root
696 * If the inode has an IS_ROOT, DCACHE_DISCONNECTED alias, then prefer
697 * any other hashed alias over that one unless @want_discon is set,
698 * in which case only return an IS_ROOT, DCACHE_DISCONNECTED alias.
700 static struct dentry *__d_find_alias(struct inode *inode, int want_discon)
702 struct dentry *alias, *discon_alias;
706 hlist_for_each_entry(alias, &inode->i_dentry, d_alias) {
707 spin_lock(&alias->d_lock);
708 if (S_ISDIR(inode->i_mode) || !d_unhashed(alias)) {
709 if (IS_ROOT(alias) &&
710 (alias->d_flags & DCACHE_DISCONNECTED)) {
711 discon_alias = alias;
712 } else if (!want_discon) {
714 spin_unlock(&alias->d_lock);
718 spin_unlock(&alias->d_lock);
721 alias = discon_alias;
722 spin_lock(&alias->d_lock);
723 if (S_ISDIR(inode->i_mode) || !d_unhashed(alias)) {
724 if (IS_ROOT(alias) &&
725 (alias->d_flags & DCACHE_DISCONNECTED)) {
727 spin_unlock(&alias->d_lock);
731 spin_unlock(&alias->d_lock);
737 struct dentry *d_find_alias(struct inode *inode)
739 struct dentry *de = NULL;
741 if (!hlist_empty(&inode->i_dentry)) {
742 spin_lock(&inode->i_lock);
743 de = __d_find_alias(inode, 0);
744 spin_unlock(&inode->i_lock);
748 EXPORT_SYMBOL(d_find_alias);
751 * Try to kill dentries associated with this inode.
752 * WARNING: you must own a reference to inode.
754 void d_prune_aliases(struct inode *inode)
756 struct dentry *dentry;
758 spin_lock(&inode->i_lock);
759 hlist_for_each_entry(dentry, &inode->i_dentry, d_alias) {
760 spin_lock(&dentry->d_lock);
761 if (!dentry->d_lockref.count) {
763 * inform the fs via d_prune that this dentry
764 * is about to be unhashed and destroyed.
766 if ((dentry->d_flags & DCACHE_OP_PRUNE) &&
768 dentry->d_op->d_prune(dentry);
770 __dget_dlock(dentry);
772 spin_unlock(&dentry->d_lock);
773 spin_unlock(&inode->i_lock);
777 spin_unlock(&dentry->d_lock);
779 spin_unlock(&inode->i_lock);
781 EXPORT_SYMBOL(d_prune_aliases);
784 * Try to throw away a dentry - free the inode, dput the parent.
785 * Requires dentry->d_lock is held, and dentry->d_count == 0.
786 * Releases dentry->d_lock.
788 * This may fail if locks cannot be acquired no problem, just try again.
790 static void try_prune_one_dentry(struct dentry *dentry)
791 __releases(dentry->d_lock)
793 struct dentry *parent;
795 parent = dentry_kill(dentry);
797 * If dentry_kill returns NULL, we have nothing more to do.
798 * if it returns the same dentry, trylocks failed. In either
799 * case, just loop again.
801 * Otherwise, we need to prune ancestors too. This is necessary
802 * to prevent quadratic behavior of shrink_dcache_parent(), but
803 * is also expected to be beneficial in reducing dentry cache
808 if (parent == dentry)
811 /* Prune ancestors. */
814 if (lockref_put_or_lock(&dentry->d_lockref))
816 dentry = dentry_kill(dentry);
820 static void shrink_dentry_list(struct list_head *list)
822 struct dentry *dentry;
826 dentry = list_entry_rcu(list->prev, struct dentry, d_lru);
827 if (&dentry->d_lru == list)
829 spin_lock(&dentry->d_lock);
830 if (dentry != list_entry(list->prev, struct dentry, d_lru)) {
831 spin_unlock(&dentry->d_lock);
836 * We found an inuse dentry which was not removed from
837 * the LRU because of laziness during lookup. Do not free
838 * it - just keep it off the LRU list.
840 if (dentry->d_lockref.count) {
841 dentry_lru_del(dentry);
842 spin_unlock(&dentry->d_lock);
848 try_prune_one_dentry(dentry);
856 * prune_dcache_sb - shrink the dcache
858 * @count: number of entries to try to free
860 * Attempt to shrink the superblock dcache LRU by @count entries. This is
861 * done when we need more memory an called from the superblock shrinker
864 * This function may fail to free any resources if all the dentries are in
867 void prune_dcache_sb(struct super_block *sb, int count)
869 struct dentry *dentry;
870 LIST_HEAD(referenced);
874 spin_lock(&dcache_lru_lock);
875 while (!list_empty(&sb->s_dentry_lru)) {
876 dentry = list_entry(sb->s_dentry_lru.prev,
877 struct dentry, d_lru);
878 BUG_ON(dentry->d_sb != sb);
880 if (!spin_trylock(&dentry->d_lock)) {
881 spin_unlock(&dcache_lru_lock);
886 if (dentry->d_flags & DCACHE_REFERENCED) {
887 dentry->d_flags &= ~DCACHE_REFERENCED;
888 list_move(&dentry->d_lru, &referenced);
889 spin_unlock(&dentry->d_lock);
891 list_move_tail(&dentry->d_lru, &tmp);
892 dentry->d_flags |= DCACHE_SHRINK_LIST;
893 spin_unlock(&dentry->d_lock);
897 cond_resched_lock(&dcache_lru_lock);
899 if (!list_empty(&referenced))
900 list_splice(&referenced, &sb->s_dentry_lru);
901 spin_unlock(&dcache_lru_lock);
903 shrink_dentry_list(&tmp);
907 * shrink_dcache_sb - shrink dcache for a superblock
910 * Shrink the dcache for the specified super block. This is used to free
911 * the dcache before unmounting a file system.
913 void shrink_dcache_sb(struct super_block *sb)
917 spin_lock(&dcache_lru_lock);
918 while (!list_empty(&sb->s_dentry_lru)) {
919 list_splice_init(&sb->s_dentry_lru, &tmp);
920 spin_unlock(&dcache_lru_lock);
921 shrink_dentry_list(&tmp);
922 spin_lock(&dcache_lru_lock);
924 spin_unlock(&dcache_lru_lock);
926 EXPORT_SYMBOL(shrink_dcache_sb);
929 * destroy a single subtree of dentries for unmount
930 * - see the comments on shrink_dcache_for_umount() for a description of the
933 static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
935 struct dentry *parent;
937 BUG_ON(!IS_ROOT(dentry));
940 /* descend to the first leaf in the current subtree */
941 while (!list_empty(&dentry->d_subdirs))
942 dentry = list_entry(dentry->d_subdirs.next,
943 struct dentry, d_u.d_child);
945 /* consume the dentries from this leaf up through its parents
946 * until we find one with children or run out altogether */
951 * inform the fs that this dentry is about to be
952 * unhashed and destroyed.
954 if ((dentry->d_flags & DCACHE_OP_PRUNE) &&
956 dentry->d_op->d_prune(dentry);
958 dentry_lru_del(dentry);
961 if (dentry->d_lockref.count != 0) {
963 "BUG: Dentry %p{i=%lx,n=%s}"
965 " [unmount of %s %s]\n",
968 dentry->d_inode->i_ino : 0UL,
970 dentry->d_lockref.count,
971 dentry->d_sb->s_type->name,
976 if (IS_ROOT(dentry)) {
978 list_del(&dentry->d_u.d_child);
980 parent = dentry->d_parent;
981 parent->d_lockref.count--;
982 list_del(&dentry->d_u.d_child);
985 inode = dentry->d_inode;
987 dentry->d_inode = NULL;
988 hlist_del_init(&dentry->d_alias);
989 if (dentry->d_op && dentry->d_op->d_iput)
990 dentry->d_op->d_iput(dentry, inode);
997 /* finished when we fall off the top of the tree,
998 * otherwise we ascend to the parent and move to the
999 * next sibling if there is one */
1003 } while (list_empty(&dentry->d_subdirs));
1005 dentry = list_entry(dentry->d_subdirs.next,
1006 struct dentry, d_u.d_child);
1011 * destroy the dentries attached to a superblock on unmounting
1012 * - we don't need to use dentry->d_lock because:
1013 * - the superblock is detached from all mountings and open files, so the
1014 * dentry trees will not be rearranged by the VFS
1015 * - s_umount is write-locked, so the memory pressure shrinker will ignore
1016 * any dentries belonging to this superblock that it comes across
1017 * - the filesystem itself is no longer permitted to rearrange the dentries
1018 * in this superblock
1020 void shrink_dcache_for_umount(struct super_block *sb)
1022 struct dentry *dentry;
1024 if (down_read_trylock(&sb->s_umount))
1027 dentry = sb->s_root;
1029 dentry->d_lockref.count--;
1030 shrink_dcache_for_umount_subtree(dentry);
1032 while (!hlist_bl_empty(&sb->s_anon)) {
1033 dentry = hlist_bl_entry(hlist_bl_first(&sb->s_anon), struct dentry, d_hash);
1034 shrink_dcache_for_umount_subtree(dentry);
1039 * This tries to ascend one level of parenthood, but
1040 * we can race with renaming, so we need to re-check
1041 * the parenthood after dropping the lock and check
1042 * that the sequence number still matches.
1044 static struct dentry *try_to_ascend(struct dentry *old, unsigned seq)
1046 struct dentry *new = old->d_parent;
1049 spin_unlock(&old->d_lock);
1050 spin_lock(&new->d_lock);
1053 * might go back up the wrong parent if we have had a rename
1056 if (new != old->d_parent ||
1057 (old->d_flags & DCACHE_DENTRY_KILLED) ||
1058 need_seqretry(&rename_lock, seq)) {
1059 spin_unlock(&new->d_lock);
1067 * enum d_walk_ret - action to talke during tree walk
1068 * @D_WALK_CONTINUE: contrinue walk
1069 * @D_WALK_QUIT: quit walk
1070 * @D_WALK_NORETRY: quit when retry is needed
1071 * @D_WALK_SKIP: skip this dentry and its children
1081 * d_walk - walk the dentry tree
1082 * @parent: start of walk
1083 * @data: data passed to @enter() and @finish()
1084 * @enter: callback when first entering the dentry
1085 * @finish: callback when successfully finished the walk
1087 * The @enter() and @finish() callbacks are called with d_lock held.
1089 static void d_walk(struct dentry *parent, void *data,
1090 enum d_walk_ret (*enter)(void *, struct dentry *),
1091 void (*finish)(void *))
1093 struct dentry *this_parent;
1094 struct list_head *next;
1096 enum d_walk_ret ret;
1100 read_seqbegin_or_lock(&rename_lock, &seq);
1101 this_parent = parent;
1102 spin_lock(&this_parent->d_lock);
1104 ret = enter(data, this_parent);
1106 case D_WALK_CONTINUE:
1111 case D_WALK_NORETRY:
1116 next = this_parent->d_subdirs.next;
1118 while (next != &this_parent->d_subdirs) {
1119 struct list_head *tmp = next;
1120 struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
1123 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
1125 ret = enter(data, dentry);
1127 case D_WALK_CONTINUE:
1130 spin_unlock(&dentry->d_lock);
1132 case D_WALK_NORETRY:
1136 spin_unlock(&dentry->d_lock);
1140 if (!list_empty(&dentry->d_subdirs)) {
1141 spin_unlock(&this_parent->d_lock);
1142 spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
1143 this_parent = dentry;
1144 spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
1147 spin_unlock(&dentry->d_lock);
1150 * All done at this level ... ascend and resume the search.
1152 if (this_parent != parent) {
1153 struct dentry *child = this_parent;
1154 this_parent = try_to_ascend(this_parent, seq);
1157 next = child->d_u.d_child.next;
1160 if (need_seqretry(&rename_lock, seq)) {
1161 spin_unlock(&this_parent->d_lock);
1168 spin_unlock(&this_parent->d_lock);
1169 done_seqretry(&rename_lock, seq);
1180 * Search for at least 1 mount point in the dentry's subdirs.
1181 * We descend to the next level whenever the d_subdirs
1182 * list is non-empty and continue searching.
1186 * have_submounts - check for mounts over a dentry
1187 * @parent: dentry to check.
1189 * Return true if the parent or its subdirectories contain
1193 static enum d_walk_ret check_mount(void *data, struct dentry *dentry)
1196 if (d_mountpoint(dentry)) {
1200 return D_WALK_CONTINUE;
1203 int have_submounts(struct dentry *parent)
1207 d_walk(parent, &ret, check_mount, NULL);
1211 EXPORT_SYMBOL(have_submounts);
1214 * Called by mount code to set a mountpoint and check if the mountpoint is
1215 * reachable (e.g. NFS can unhash a directory dentry and then the complete
1216 * subtree can become unreachable).
1218 * Only one of check_submounts_and_drop() and d_set_mounted() must succeed. For
1219 * this reason take rename_lock and d_lock on dentry and ancestors.
1221 int d_set_mounted(struct dentry *dentry)
1225 write_seqlock(&rename_lock);
1226 for (p = dentry->d_parent; !IS_ROOT(p); p = p->d_parent) {
1227 /* Need exclusion wrt. check_submounts_and_drop() */
1228 spin_lock(&p->d_lock);
1229 if (unlikely(d_unhashed(p))) {
1230 spin_unlock(&p->d_lock);
1233 spin_unlock(&p->d_lock);
1235 spin_lock(&dentry->d_lock);
1236 if (!d_unlinked(dentry)) {
1237 dentry->d_flags |= DCACHE_MOUNTED;
1240 spin_unlock(&dentry->d_lock);
1242 write_sequnlock(&rename_lock);
1247 * Search the dentry child list of the specified parent,
1248 * and move any unused dentries to the end of the unused
1249 * list for prune_dcache(). We descend to the next level
1250 * whenever the d_subdirs list is non-empty and continue
1253 * It returns zero iff there are no unused children,
1254 * otherwise it returns the number of children moved to
1255 * the end of the unused list. This may not be the total
1256 * number of unused children, because select_parent can
1257 * drop the lock and return early due to latency
1261 struct select_data {
1262 struct dentry *start;
1263 struct list_head dispose;
1267 static enum d_walk_ret select_collect(void *_data, struct dentry *dentry)
1269 struct select_data *data = _data;
1270 enum d_walk_ret ret = D_WALK_CONTINUE;
1272 if (data->start == dentry)
1276 * move only zero ref count dentries to the dispose list.
1278 * Those which are presently on the shrink list, being processed
1279 * by shrink_dentry_list(), shouldn't be moved. Otherwise the
1280 * loop in shrink_dcache_parent() might not make any progress
1283 if (dentry->d_lockref.count) {
1284 dentry_lru_del(dentry);
1285 } else if (!(dentry->d_flags & DCACHE_SHRINK_LIST)) {
1286 dentry_lru_move_list(dentry, &data->dispose);
1287 dentry->d_flags |= DCACHE_SHRINK_LIST;
1289 ret = D_WALK_NORETRY;
1292 * We can return to the caller if we have found some (this
1293 * ensures forward progress). We'll be coming back to find
1296 if (data->found && need_resched())
1303 * shrink_dcache_parent - prune dcache
1304 * @parent: parent of entries to prune
1306 * Prune the dcache to remove unused children of the parent dentry.
1308 void shrink_dcache_parent(struct dentry *parent)
1311 struct select_data data;
1313 INIT_LIST_HEAD(&data.dispose);
1314 data.start = parent;
1317 d_walk(parent, &data, select_collect, NULL);
1321 shrink_dentry_list(&data.dispose);
1325 EXPORT_SYMBOL(shrink_dcache_parent);
1327 static enum d_walk_ret check_and_collect(void *_data, struct dentry *dentry)
1329 struct select_data *data = _data;
1331 if (d_mountpoint(dentry)) {
1332 data->found = -EBUSY;
1336 return select_collect(_data, dentry);
1339 static void check_and_drop(void *_data)
1341 struct select_data *data = _data;
1343 if (d_mountpoint(data->start))
1344 data->found = -EBUSY;
1346 __d_drop(data->start);
1350 * check_submounts_and_drop - prune dcache, check for submounts and drop
1352 * All done as a single atomic operation relative to has_unlinked_ancestor().
1353 * Returns 0 if successfully unhashed @parent. If there were submounts then
1356 * @dentry: dentry to prune and drop
1358 int check_submounts_and_drop(struct dentry *dentry)
1362 /* Negative dentries can be dropped without further checks */
1363 if (!dentry->d_inode) {
1369 struct select_data data;
1371 INIT_LIST_HEAD(&data.dispose);
1372 data.start = dentry;
1375 d_walk(dentry, &data, check_and_collect, check_and_drop);
1378 if (!list_empty(&data.dispose))
1379 shrink_dentry_list(&data.dispose);
1390 EXPORT_SYMBOL(check_submounts_and_drop);
1393 * __d_alloc - allocate a dcache entry
1394 * @sb: filesystem it will belong to
1395 * @name: qstr of the name
1397 * Allocates a dentry. It returns %NULL if there is insufficient memory
1398 * available. On a success the dentry is returned. The name passed in is
1399 * copied and the copy passed in may be reused after this call.
1402 struct dentry *__d_alloc(struct super_block *sb, const struct qstr *name)
1404 struct dentry *dentry;
1407 dentry = kmem_cache_alloc(dentry_cache, GFP_KERNEL);
1412 * We guarantee that the inline name is always NUL-terminated.
1413 * This way the memcpy() done by the name switching in rename
1414 * will still always have a NUL at the end, even if we might
1415 * be overwriting an internal NUL character
1417 dentry->d_iname[DNAME_INLINE_LEN-1] = 0;
1418 if (name->len > DNAME_INLINE_LEN-1) {
1419 dname = kmalloc(name->len + 1, GFP_KERNEL);
1421 kmem_cache_free(dentry_cache, dentry);
1425 dname = dentry->d_iname;
1428 dentry->d_name.len = name->len;
1429 dentry->d_name.hash = name->hash;
1430 memcpy(dname, name->name, name->len);
1431 dname[name->len] = 0;
1433 /* Make sure we always see the terminating NUL character */
1435 dentry->d_name.name = dname;
1437 dentry->d_lockref.count = 1;
1438 dentry->d_flags = 0;
1439 spin_lock_init(&dentry->d_lock);
1440 seqcount_init(&dentry->d_seq);
1441 dentry->d_inode = NULL;
1442 dentry->d_parent = dentry;
1444 dentry->d_op = NULL;
1445 dentry->d_fsdata = NULL;
1446 INIT_HLIST_BL_NODE(&dentry->d_hash);
1447 INIT_LIST_HEAD(&dentry->d_lru);
1448 INIT_LIST_HEAD(&dentry->d_subdirs);
1449 INIT_HLIST_NODE(&dentry->d_alias);
1450 INIT_LIST_HEAD(&dentry->d_u.d_child);
1451 d_set_d_op(dentry, dentry->d_sb->s_d_op);
1453 this_cpu_inc(nr_dentry);
1459 * d_alloc - allocate a dcache entry
1460 * @parent: parent of entry to allocate
1461 * @name: qstr of the name
1463 * Allocates a dentry. It returns %NULL if there is insufficient memory
1464 * available. On a success the dentry is returned. The name passed in is
1465 * copied and the copy passed in may be reused after this call.
1467 struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
1469 struct dentry *dentry = __d_alloc(parent->d_sb, name);
1473 spin_lock(&parent->d_lock);
1475 * don't need child lock because it is not subject
1476 * to concurrency here
1478 __dget_dlock(parent);
1479 dentry->d_parent = parent;
1480 list_add(&dentry->d_u.d_child, &parent->d_subdirs);
1481 spin_unlock(&parent->d_lock);
1485 EXPORT_SYMBOL(d_alloc);
1487 struct dentry *d_alloc_pseudo(struct super_block *sb, const struct qstr *name)
1489 struct dentry *dentry = __d_alloc(sb, name);
1491 dentry->d_flags |= DCACHE_DISCONNECTED;
1494 EXPORT_SYMBOL(d_alloc_pseudo);
1496 struct dentry *d_alloc_name(struct dentry *parent, const char *name)
1501 q.len = strlen(name);
1502 q.hash = full_name_hash(q.name, q.len);
1503 return d_alloc(parent, &q);
1505 EXPORT_SYMBOL(d_alloc_name);
1507 void d_set_d_op(struct dentry *dentry, const struct dentry_operations *op)
1509 WARN_ON_ONCE(dentry->d_op);
1510 WARN_ON_ONCE(dentry->d_flags & (DCACHE_OP_HASH |
1512 DCACHE_OP_REVALIDATE |
1513 DCACHE_OP_WEAK_REVALIDATE |
1514 DCACHE_OP_DELETE ));
1519 dentry->d_flags |= DCACHE_OP_HASH;
1521 dentry->d_flags |= DCACHE_OP_COMPARE;
1522 if (op->d_revalidate)
1523 dentry->d_flags |= DCACHE_OP_REVALIDATE;
1524 if (op->d_weak_revalidate)
1525 dentry->d_flags |= DCACHE_OP_WEAK_REVALIDATE;
1527 dentry->d_flags |= DCACHE_OP_DELETE;
1529 dentry->d_flags |= DCACHE_OP_PRUNE;
1532 EXPORT_SYMBOL(d_set_d_op);
1534 static void __d_instantiate(struct dentry *dentry, struct inode *inode)
1536 spin_lock(&dentry->d_lock);
1538 if (unlikely(IS_AUTOMOUNT(inode)))
1539 dentry->d_flags |= DCACHE_NEED_AUTOMOUNT;
1540 hlist_add_head(&dentry->d_alias, &inode->i_dentry);
1542 dentry->d_inode = inode;
1543 dentry_rcuwalk_barrier(dentry);
1544 spin_unlock(&dentry->d_lock);
1545 fsnotify_d_instantiate(dentry, inode);
1549 * d_instantiate - fill in inode information for a dentry
1550 * @entry: dentry to complete
1551 * @inode: inode to attach to this dentry
1553 * Fill in inode information in the entry.
1555 * This turns negative dentries into productive full members
1558 * NOTE! This assumes that the inode count has been incremented
1559 * (or otherwise set) by the caller to indicate that it is now
1560 * in use by the dcache.
1563 void d_instantiate(struct dentry *entry, struct inode * inode)
1565 BUG_ON(!hlist_unhashed(&entry->d_alias));
1567 spin_lock(&inode->i_lock);
1568 __d_instantiate(entry, inode);
1570 spin_unlock(&inode->i_lock);
1571 security_d_instantiate(entry, inode);
1573 EXPORT_SYMBOL(d_instantiate);
1576 * d_instantiate_unique - instantiate a non-aliased dentry
1577 * @entry: dentry to instantiate
1578 * @inode: inode to attach to this dentry
1580 * Fill in inode information in the entry. On success, it returns NULL.
1581 * If an unhashed alias of "entry" already exists, then we return the
1582 * aliased dentry instead and drop one reference to inode.
1584 * Note that in order to avoid conflicts with rename() etc, the caller
1585 * had better be holding the parent directory semaphore.
1587 * This also assumes that the inode count has been incremented
1588 * (or otherwise set) by the caller to indicate that it is now
1589 * in use by the dcache.
1591 static struct dentry *__d_instantiate_unique(struct dentry *entry,
1592 struct inode *inode)
1594 struct dentry *alias;
1595 int len = entry->d_name.len;
1596 const char *name = entry->d_name.name;
1597 unsigned int hash = entry->d_name.hash;
1600 __d_instantiate(entry, NULL);
1604 hlist_for_each_entry(alias, &inode->i_dentry, d_alias) {
1606 * Don't need alias->d_lock here, because aliases with
1607 * d_parent == entry->d_parent are not subject to name or
1608 * parent changes, because the parent inode i_mutex is held.
1610 if (alias->d_name.hash != hash)
1612 if (alias->d_parent != entry->d_parent)
1614 if (alias->d_name.len != len)
1616 if (dentry_cmp(alias, name, len))
1622 __d_instantiate(entry, inode);
1626 struct dentry *d_instantiate_unique(struct dentry *entry, struct inode *inode)
1628 struct dentry *result;
1630 BUG_ON(!hlist_unhashed(&entry->d_alias));
1633 spin_lock(&inode->i_lock);
1634 result = __d_instantiate_unique(entry, inode);
1636 spin_unlock(&inode->i_lock);
1639 security_d_instantiate(entry, inode);
1643 BUG_ON(!d_unhashed(result));
1648 EXPORT_SYMBOL(d_instantiate_unique);
1650 struct dentry *d_make_root(struct inode *root_inode)
1652 struct dentry *res = NULL;
1655 static const struct qstr name = QSTR_INIT("/", 1);
1657 res = __d_alloc(root_inode->i_sb, &name);
1659 d_instantiate(res, root_inode);
1665 EXPORT_SYMBOL(d_make_root);
1667 static struct dentry * __d_find_any_alias(struct inode *inode)
1669 struct dentry *alias;
1671 if (hlist_empty(&inode->i_dentry))
1673 alias = hlist_entry(inode->i_dentry.first, struct dentry, d_alias);
1679 * d_find_any_alias - find any alias for a given inode
1680 * @inode: inode to find an alias for
1682 * If any aliases exist for the given inode, take and return a
1683 * reference for one of them. If no aliases exist, return %NULL.
1685 struct dentry *d_find_any_alias(struct inode *inode)
1689 spin_lock(&inode->i_lock);
1690 de = __d_find_any_alias(inode);
1691 spin_unlock(&inode->i_lock);
1694 EXPORT_SYMBOL(d_find_any_alias);
1697 * d_obtain_alias - find or allocate a dentry for a given inode
1698 * @inode: inode to allocate the dentry for
1700 * Obtain a dentry for an inode resulting from NFS filehandle conversion or
1701 * similar open by handle operations. The returned dentry may be anonymous,
1702 * or may have a full name (if the inode was already in the cache).
1704 * When called on a directory inode, we must ensure that the inode only ever
1705 * has one dentry. If a dentry is found, that is returned instead of
1706 * allocating a new one.
1708 * On successful return, the reference to the inode has been transferred
1709 * to the dentry. In case of an error the reference on the inode is released.
1710 * To make it easier to use in export operations a %NULL or IS_ERR inode may
1711 * be passed in and will be the error will be propagate to the return value,
1712 * with a %NULL @inode replaced by ERR_PTR(-ESTALE).
1714 struct dentry *d_obtain_alias(struct inode *inode)
1716 static const struct qstr anonstring = QSTR_INIT("/", 1);
1721 return ERR_PTR(-ESTALE);
1723 return ERR_CAST(inode);
1725 res = d_find_any_alias(inode);
1729 tmp = __d_alloc(inode->i_sb, &anonstring);
1731 res = ERR_PTR(-ENOMEM);
1735 spin_lock(&inode->i_lock);
1736 res = __d_find_any_alias(inode);
1738 spin_unlock(&inode->i_lock);
1743 /* attach a disconnected dentry */
1744 spin_lock(&tmp->d_lock);
1745 tmp->d_inode = inode;
1746 tmp->d_flags |= DCACHE_DISCONNECTED;
1747 hlist_add_head(&tmp->d_alias, &inode->i_dentry);
1748 hlist_bl_lock(&tmp->d_sb->s_anon);
1749 hlist_bl_add_head(&tmp->d_hash, &tmp->d_sb->s_anon);
1750 hlist_bl_unlock(&tmp->d_sb->s_anon);
1751 spin_unlock(&tmp->d_lock);
1752 spin_unlock(&inode->i_lock);
1753 security_d_instantiate(tmp, inode);
1758 if (res && !IS_ERR(res))
1759 security_d_instantiate(res, inode);
1763 EXPORT_SYMBOL(d_obtain_alias);
1766 * d_splice_alias - splice a disconnected dentry into the tree if one exists
1767 * @inode: the inode which may have a disconnected dentry
1768 * @dentry: a negative dentry which we want to point to the inode.
1770 * If inode is a directory and has a 'disconnected' dentry (i.e. IS_ROOT and
1771 * DCACHE_DISCONNECTED), then d_move that in place of the given dentry
1772 * and return it, else simply d_add the inode to the dentry and return NULL.
1774 * This is needed in the lookup routine of any filesystem that is exportable
1775 * (via knfsd) so that we can build dcache paths to directories effectively.
1777 * If a dentry was found and moved, then it is returned. Otherwise NULL
1778 * is returned. This matches the expected return value of ->lookup.
1780 * Cluster filesystems may call this function with a negative, hashed dentry.
1781 * In that case, we know that the inode will be a regular file, and also this
1782 * will only occur during atomic_open. So we need to check for the dentry
1783 * being already hashed only in the final case.
1785 struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
1787 struct dentry *new = NULL;
1790 return ERR_CAST(inode);
1792 if (inode && S_ISDIR(inode->i_mode)) {
1793 spin_lock(&inode->i_lock);
1794 new = __d_find_alias(inode, 1);
1796 BUG_ON(!(new->d_flags & DCACHE_DISCONNECTED));
1797 spin_unlock(&inode->i_lock);
1798 security_d_instantiate(new, inode);
1799 d_move(new, dentry);
1802 /* already taking inode->i_lock, so d_add() by hand */
1803 __d_instantiate(dentry, inode);
1804 spin_unlock(&inode->i_lock);
1805 security_d_instantiate(dentry, inode);
1809 d_instantiate(dentry, inode);
1810 if (d_unhashed(dentry))
1815 EXPORT_SYMBOL(d_splice_alias);
1818 * d_add_ci - lookup or allocate new dentry with case-exact name
1819 * @inode: the inode case-insensitive lookup has found
1820 * @dentry: the negative dentry that was passed to the parent's lookup func
1821 * @name: the case-exact name to be associated with the returned dentry
1823 * This is to avoid filling the dcache with case-insensitive names to the
1824 * same inode, only the actual correct case is stored in the dcache for
1825 * case-insensitive filesystems.
1827 * For a case-insensitive lookup match and if the the case-exact dentry
1828 * already exists in in the dcache, use it and return it.
1830 * If no entry exists with the exact case name, allocate new dentry with
1831 * the exact case, and return the spliced entry.
1833 struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
1836 struct dentry *found;
1840 * First check if a dentry matching the name already exists,
1841 * if not go ahead and create it now.
1843 found = d_hash_and_lookup(dentry->d_parent, name);
1844 if (unlikely(IS_ERR(found)))
1847 new = d_alloc(dentry->d_parent, name);
1849 found = ERR_PTR(-ENOMEM);
1853 found = d_splice_alias(inode, new);
1862 * If a matching dentry exists, and it's not negative use it.
1864 * Decrement the reference count to balance the iget() done
1867 if (found->d_inode) {
1868 if (unlikely(found->d_inode != inode)) {
1869 /* This can't happen because bad inodes are unhashed. */
1870 BUG_ON(!is_bad_inode(inode));
1871 BUG_ON(!is_bad_inode(found->d_inode));
1878 * Negative dentry: instantiate it unless the inode is a directory and
1879 * already has a dentry.
1881 new = d_splice_alias(inode, found);
1892 EXPORT_SYMBOL(d_add_ci);
1895 * Do the slow-case of the dentry name compare.
1897 * Unlike the dentry_cmp() function, we need to atomically
1898 * load the name and length information, so that the
1899 * filesystem can rely on them, and can use the 'name' and
1900 * 'len' information without worrying about walking off the
1901 * end of memory etc.
1903 * Thus the read_seqcount_retry() and the "duplicate" info
1904 * in arguments (the low-level filesystem should not look
1905 * at the dentry inode or name contents directly, since
1906 * rename can change them while we're in RCU mode).
1908 enum slow_d_compare {
1914 static noinline enum slow_d_compare slow_dentry_cmp(
1915 const struct dentry *parent,
1916 struct dentry *dentry,
1918 const struct qstr *name)
1920 int tlen = dentry->d_name.len;
1921 const char *tname = dentry->d_name.name;
1923 if (read_seqcount_retry(&dentry->d_seq, seq)) {
1925 return D_COMP_SEQRETRY;
1927 if (parent->d_op->d_compare(parent, dentry, tlen, tname, name))
1928 return D_COMP_NOMATCH;
1933 * __d_lookup_rcu - search for a dentry (racy, store-free)
1934 * @parent: parent dentry
1935 * @name: qstr of name we wish to find
1936 * @seqp: returns d_seq value at the point where the dentry was found
1937 * Returns: dentry, or NULL
1939 * __d_lookup_rcu is the dcache lookup function for rcu-walk name
1940 * resolution (store-free path walking) design described in
1941 * Documentation/filesystems/path-lookup.txt.
1943 * This is not to be used outside core vfs.
1945 * __d_lookup_rcu must only be used in rcu-walk mode, ie. with vfsmount lock
1946 * held, and rcu_read_lock held. The returned dentry must not be stored into
1947 * without taking d_lock and checking d_seq sequence count against @seq
1950 * A refcount may be taken on the found dentry with the d_rcu_to_refcount
1953 * Alternatively, __d_lookup_rcu may be called again to look up the child of
1954 * the returned dentry, so long as its parent's seqlock is checked after the
1955 * child is looked up. Thus, an interlocking stepping of sequence lock checks
1956 * is formed, giving integrity down the path walk.
1958 * NOTE! The caller *has* to check the resulting dentry against the sequence
1959 * number we've returned before using any of the resulting dentry state!
1961 struct dentry *__d_lookup_rcu(const struct dentry *parent,
1962 const struct qstr *name,
1965 u64 hashlen = name->hash_len;
1966 const unsigned char *str = name->name;
1967 struct hlist_bl_head *b = d_hash(parent, hashlen_hash(hashlen));
1968 struct hlist_bl_node *node;
1969 struct dentry *dentry;
1972 * Note: There is significant duplication with __d_lookup_rcu which is
1973 * required to prevent single threaded performance regressions
1974 * especially on architectures where smp_rmb (in seqcounts) are costly.
1975 * Keep the two functions in sync.
1979 * The hash list is protected using RCU.
1981 * Carefully use d_seq when comparing a candidate dentry, to avoid
1982 * races with d_move().
1984 * It is possible that concurrent renames can mess up our list
1985 * walk here and result in missing our dentry, resulting in the
1986 * false-negative result. d_lookup() protects against concurrent
1987 * renames using rename_lock seqlock.
1989 * See Documentation/filesystems/path-lookup.txt for more details.
1991 hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
1996 * The dentry sequence count protects us from concurrent
1997 * renames, and thus protects parent and name fields.
1999 * The caller must perform a seqcount check in order
2000 * to do anything useful with the returned dentry.
2002 * NOTE! We do a "raw" seqcount_begin here. That means that
2003 * we don't wait for the sequence count to stabilize if it
2004 * is in the middle of a sequence change. If we do the slow
2005 * dentry compare, we will do seqretries until it is stable,
2006 * and if we end up with a successful lookup, we actually
2007 * want to exit RCU lookup anyway.
2009 seq = raw_seqcount_begin(&dentry->d_seq);
2010 if (dentry->d_parent != parent)
2012 if (d_unhashed(dentry))
2015 if (unlikely(parent->d_flags & DCACHE_OP_COMPARE)) {
2016 if (dentry->d_name.hash != hashlen_hash(hashlen))
2019 switch (slow_dentry_cmp(parent, dentry, seq, name)) {
2022 case D_COMP_NOMATCH:
2029 if (dentry->d_name.hash_len != hashlen)
2032 if (!dentry_cmp(dentry, str, hashlen_len(hashlen)))
2039 * d_lookup - search for a dentry
2040 * @parent: parent dentry
2041 * @name: qstr of name we wish to find
2042 * Returns: dentry, or NULL
2044 * d_lookup searches the children of the parent dentry for the name in
2045 * question. If the dentry is found its reference count is incremented and the
2046 * dentry is returned. The caller must use dput to free the entry when it has
2047 * finished using it. %NULL is returned if the dentry does not exist.
2049 struct dentry *d_lookup(const struct dentry *parent, const struct qstr *name)
2051 struct dentry *dentry;
2055 seq = read_seqbegin(&rename_lock);
2056 dentry = __d_lookup(parent, name);
2059 } while (read_seqretry(&rename_lock, seq));
2062 EXPORT_SYMBOL(d_lookup);
2065 * __d_lookup - search for a dentry (racy)
2066 * @parent: parent dentry
2067 * @name: qstr of name we wish to find
2068 * Returns: dentry, or NULL
2070 * __d_lookup is like d_lookup, however it may (rarely) return a
2071 * false-negative result due to unrelated rename activity.
2073 * __d_lookup is slightly faster by avoiding rename_lock read seqlock,
2074 * however it must be used carefully, eg. with a following d_lookup in
2075 * the case of failure.
2077 * __d_lookup callers must be commented.
2079 struct dentry *__d_lookup(const struct dentry *parent, const struct qstr *name)
2081 unsigned int len = name->len;
2082 unsigned int hash = name->hash;
2083 const unsigned char *str = name->name;
2084 struct hlist_bl_head *b = d_hash(parent, hash);
2085 struct hlist_bl_node *node;
2086 struct dentry *found = NULL;
2087 struct dentry *dentry;
2090 * Note: There is significant duplication with __d_lookup_rcu which is
2091 * required to prevent single threaded performance regressions
2092 * especially on architectures where smp_rmb (in seqcounts) are costly.
2093 * Keep the two functions in sync.
2097 * The hash list is protected using RCU.
2099 * Take d_lock when comparing a candidate dentry, to avoid races
2102 * It is possible that concurrent renames can mess up our list
2103 * walk here and result in missing our dentry, resulting in the
2104 * false-negative result. d_lookup() protects against concurrent
2105 * renames using rename_lock seqlock.
2107 * See Documentation/filesystems/path-lookup.txt for more details.
2111 hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
2113 if (dentry->d_name.hash != hash)
2116 spin_lock(&dentry->d_lock);
2117 if (dentry->d_parent != parent)
2119 if (d_unhashed(dentry))
2123 * It is safe to compare names since d_move() cannot
2124 * change the qstr (protected by d_lock).
2126 if (parent->d_flags & DCACHE_OP_COMPARE) {
2127 int tlen = dentry->d_name.len;
2128 const char *tname = dentry->d_name.name;
2129 if (parent->d_op->d_compare(parent, dentry, tlen, tname, name))
2132 if (dentry->d_name.len != len)
2134 if (dentry_cmp(dentry, str, len))
2138 dentry->d_lockref.count++;
2140 spin_unlock(&dentry->d_lock);
2143 spin_unlock(&dentry->d_lock);
2151 * d_hash_and_lookup - hash the qstr then search for a dentry
2152 * @dir: Directory to search in
2153 * @name: qstr of name we wish to find
2155 * On lookup failure NULL is returned; on bad name - ERR_PTR(-error)
2157 struct dentry *d_hash_and_lookup(struct dentry *dir, struct qstr *name)
2160 * Check for a fs-specific hash function. Note that we must
2161 * calculate the standard hash first, as the d_op->d_hash()
2162 * routine may choose to leave the hash value unchanged.
2164 name->hash = full_name_hash(name->name, name->len);
2165 if (dir->d_flags & DCACHE_OP_HASH) {
2166 int err = dir->d_op->d_hash(dir, name);
2167 if (unlikely(err < 0))
2168 return ERR_PTR(err);
2170 return d_lookup(dir, name);
2172 EXPORT_SYMBOL(d_hash_and_lookup);
2175 * d_validate - verify dentry provided from insecure source (deprecated)
2176 * @dentry: The dentry alleged to be valid child of @dparent
2177 * @dparent: The parent dentry (known to be valid)
2179 * An insecure source has sent us a dentry, here we verify it and dget() it.
2180 * This is used by ncpfs in its readdir implementation.
2181 * Zero is returned in the dentry is invalid.
2183 * This function is slow for big directories, and deprecated, do not use it.
2185 int d_validate(struct dentry *dentry, struct dentry *dparent)
2187 struct dentry *child;
2189 spin_lock(&dparent->d_lock);
2190 list_for_each_entry(child, &dparent->d_subdirs, d_u.d_child) {
2191 if (dentry == child) {
2192 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
2193 __dget_dlock(dentry);
2194 spin_unlock(&dentry->d_lock);
2195 spin_unlock(&dparent->d_lock);
2199 spin_unlock(&dparent->d_lock);
2203 EXPORT_SYMBOL(d_validate);
2206 * When a file is deleted, we have two options:
2207 * - turn this dentry into a negative dentry
2208 * - unhash this dentry and free it.
2210 * Usually, we want to just turn this into
2211 * a negative dentry, but if anybody else is
2212 * currently using the dentry or the inode
2213 * we can't do that and we fall back on removing
2214 * it from the hash queues and waiting for
2215 * it to be deleted later when it has no users
2219 * d_delete - delete a dentry
2220 * @dentry: The dentry to delete
2222 * Turn the dentry into a negative dentry if possible, otherwise
2223 * remove it from the hash queues so it can be deleted later
2226 void d_delete(struct dentry * dentry)
2228 struct inode *inode;
2231 * Are we the only user?
2234 spin_lock(&dentry->d_lock);
2235 inode = dentry->d_inode;
2236 isdir = S_ISDIR(inode->i_mode);
2237 if (dentry->d_lockref.count == 1) {
2238 if (!spin_trylock(&inode->i_lock)) {
2239 spin_unlock(&dentry->d_lock);
2243 dentry->d_flags &= ~DCACHE_CANT_MOUNT;
2244 dentry_unlink_inode(dentry);
2245 fsnotify_nameremove(dentry, isdir);
2249 if (!d_unhashed(dentry))
2252 spin_unlock(&dentry->d_lock);
2254 fsnotify_nameremove(dentry, isdir);
2256 EXPORT_SYMBOL(d_delete);
2258 static void __d_rehash(struct dentry * entry, struct hlist_bl_head *b)
2260 BUG_ON(!d_unhashed(entry));
2262 entry->d_flags |= DCACHE_RCUACCESS;
2263 hlist_bl_add_head_rcu(&entry->d_hash, b);
2267 static void _d_rehash(struct dentry * entry)
2269 __d_rehash(entry, d_hash(entry->d_parent, entry->d_name.hash));
2273 * d_rehash - add an entry back to the hash
2274 * @entry: dentry to add to the hash
2276 * Adds a dentry to the hash according to its name.
2279 void d_rehash(struct dentry * entry)
2281 spin_lock(&entry->d_lock);
2283 spin_unlock(&entry->d_lock);
2285 EXPORT_SYMBOL(d_rehash);
2288 * dentry_update_name_case - update case insensitive dentry with a new name
2289 * @dentry: dentry to be updated
2292 * Update a case insensitive dentry with new case of name.
2294 * dentry must have been returned by d_lookup with name @name. Old and new
2295 * name lengths must match (ie. no d_compare which allows mismatched name
2298 * Parent inode i_mutex must be held over d_lookup and into this call (to
2299 * keep renames and concurrent inserts, and readdir(2) away).
2301 void dentry_update_name_case(struct dentry *dentry, struct qstr *name)
2303 BUG_ON(!mutex_is_locked(&dentry->d_parent->d_inode->i_mutex));
2304 BUG_ON(dentry->d_name.len != name->len); /* d_lookup gives this */
2306 spin_lock(&dentry->d_lock);
2307 write_seqcount_begin(&dentry->d_seq);
2308 memcpy((unsigned char *)dentry->d_name.name, name->name, name->len);
2309 write_seqcount_end(&dentry->d_seq);
2310 spin_unlock(&dentry->d_lock);
2312 EXPORT_SYMBOL(dentry_update_name_case);
2314 static void switch_names(struct dentry *dentry, struct dentry *target)
2316 if (dname_external(target)) {
2317 if (dname_external(dentry)) {
2319 * Both external: swap the pointers
2321 swap(target->d_name.name, dentry->d_name.name);
2324 * dentry:internal, target:external. Steal target's
2325 * storage and make target internal.
2327 memcpy(target->d_iname, dentry->d_name.name,
2328 dentry->d_name.len + 1);
2329 dentry->d_name.name = target->d_name.name;
2330 target->d_name.name = target->d_iname;
2333 if (dname_external(dentry)) {
2335 * dentry:external, target:internal. Give dentry's
2336 * storage to target and make dentry internal
2338 memcpy(dentry->d_iname, target->d_name.name,
2339 target->d_name.len + 1);
2340 target->d_name.name = dentry->d_name.name;
2341 dentry->d_name.name = dentry->d_iname;
2344 * Both are internal. Just copy target to dentry
2346 memcpy(dentry->d_iname, target->d_name.name,
2347 target->d_name.len + 1);
2348 dentry->d_name.len = target->d_name.len;
2352 swap(dentry->d_name.len, target->d_name.len);
2355 static void dentry_lock_for_move(struct dentry *dentry, struct dentry *target)
2358 * XXXX: do we really need to take target->d_lock?
2360 if (IS_ROOT(dentry) || dentry->d_parent == target->d_parent)
2361 spin_lock(&target->d_parent->d_lock);
2363 if (d_ancestor(dentry->d_parent, target->d_parent)) {
2364 spin_lock(&dentry->d_parent->d_lock);
2365 spin_lock_nested(&target->d_parent->d_lock,
2366 DENTRY_D_LOCK_NESTED);
2368 spin_lock(&target->d_parent->d_lock);
2369 spin_lock_nested(&dentry->d_parent->d_lock,
2370 DENTRY_D_LOCK_NESTED);
2373 if (target < dentry) {
2374 spin_lock_nested(&target->d_lock, 2);
2375 spin_lock_nested(&dentry->d_lock, 3);
2377 spin_lock_nested(&dentry->d_lock, 2);
2378 spin_lock_nested(&target->d_lock, 3);
2382 static void dentry_unlock_parents_for_move(struct dentry *dentry,
2383 struct dentry *target)
2385 if (target->d_parent != dentry->d_parent)
2386 spin_unlock(&dentry->d_parent->d_lock);
2387 if (target->d_parent != target)
2388 spin_unlock(&target->d_parent->d_lock);
2392 * When switching names, the actual string doesn't strictly have to
2393 * be preserved in the target - because we're dropping the target
2394 * anyway. As such, we can just do a simple memcpy() to copy over
2395 * the new name before we switch.
2397 * Note that we have to be a lot more careful about getting the hash
2398 * switched - we have to switch the hash value properly even if it
2399 * then no longer matches the actual (corrupted) string of the target.
2400 * The hash value has to match the hash queue that the dentry is on..
2403 * __d_move - move a dentry
2404 * @dentry: entry to move
2405 * @target: new dentry
2407 * Update the dcache to reflect the move of a file name. Negative
2408 * dcache entries should not be moved in this way. Caller must hold
2409 * rename_lock, the i_mutex of the source and target directories,
2410 * and the sb->s_vfs_rename_mutex if they differ. See lock_rename().
2412 static void __d_move(struct dentry * dentry, struct dentry * target)
2414 if (!dentry->d_inode)
2415 printk(KERN_WARNING "VFS: moving negative dcache entry\n");
2417 BUG_ON(d_ancestor(dentry, target));
2418 BUG_ON(d_ancestor(target, dentry));
2420 dentry_lock_for_move(dentry, target);
2422 write_seqcount_begin(&dentry->d_seq);
2423 write_seqcount_begin(&target->d_seq);
2425 /* __d_drop does write_seqcount_barrier, but they're OK to nest. */
2428 * Move the dentry to the target hash queue. Don't bother checking
2429 * for the same hash queue because of how unlikely it is.
2432 __d_rehash(dentry, d_hash(target->d_parent, target->d_name.hash));
2434 /* Unhash the target: dput() will then get rid of it */
2437 list_del(&dentry->d_u.d_child);
2438 list_del(&target->d_u.d_child);
2440 /* Switch the names.. */
2441 switch_names(dentry, target);
2442 swap(dentry->d_name.hash, target->d_name.hash);
2444 /* ... and switch the parents */
2445 if (IS_ROOT(dentry)) {
2446 dentry->d_parent = target->d_parent;
2447 target->d_parent = target;
2448 INIT_LIST_HEAD(&target->d_u.d_child);
2450 swap(dentry->d_parent, target->d_parent);
2452 /* And add them back to the (new) parent lists */
2453 list_add(&target->d_u.d_child, &target->d_parent->d_subdirs);
2456 list_add(&dentry->d_u.d_child, &dentry->d_parent->d_subdirs);
2458 write_seqcount_end(&target->d_seq);
2459 write_seqcount_end(&dentry->d_seq);
2461 dentry_unlock_parents_for_move(dentry, target);
2462 spin_unlock(&target->d_lock);
2463 fsnotify_d_move(dentry);
2464 spin_unlock(&dentry->d_lock);
2468 * d_move - move a dentry
2469 * @dentry: entry to move
2470 * @target: new dentry
2472 * Update the dcache to reflect the move of a file name. Negative
2473 * dcache entries should not be moved in this way. See the locking
2474 * requirements for __d_move.
2476 void d_move(struct dentry *dentry, struct dentry *target)
2478 write_seqlock(&rename_lock);
2479 __d_move(dentry, target);
2480 write_sequnlock(&rename_lock);
2482 EXPORT_SYMBOL(d_move);
2485 * d_ancestor - search for an ancestor
2486 * @p1: ancestor dentry
2489 * Returns the ancestor dentry of p2 which is a child of p1, if p1 is
2490 * an ancestor of p2, else NULL.
2492 struct dentry *d_ancestor(struct dentry *p1, struct dentry *p2)
2496 for (p = p2; !IS_ROOT(p); p = p->d_parent) {
2497 if (p->d_parent == p1)
2504 * This helper attempts to cope with remotely renamed directories
2506 * It assumes that the caller is already holding
2507 * dentry->d_parent->d_inode->i_mutex, inode->i_lock and rename_lock
2509 * Note: If ever the locking in lock_rename() changes, then please
2510 * remember to update this too...
2512 static struct dentry *__d_unalias(struct inode *inode,
2513 struct dentry *dentry, struct dentry *alias)
2515 struct mutex *m1 = NULL, *m2 = NULL;
2516 struct dentry *ret = ERR_PTR(-EBUSY);
2518 /* If alias and dentry share a parent, then no extra locks required */
2519 if (alias->d_parent == dentry->d_parent)
2522 /* See lock_rename() */
2523 if (!mutex_trylock(&dentry->d_sb->s_vfs_rename_mutex))
2525 m1 = &dentry->d_sb->s_vfs_rename_mutex;
2526 if (!mutex_trylock(&alias->d_parent->d_inode->i_mutex))
2528 m2 = &alias->d_parent->d_inode->i_mutex;
2530 if (likely(!d_mountpoint(alias))) {
2531 __d_move(alias, dentry);
2535 spin_unlock(&inode->i_lock);
2544 * Prepare an anonymous dentry for life in the superblock's dentry tree as a
2545 * named dentry in place of the dentry to be replaced.
2546 * returns with anon->d_lock held!
2548 static void __d_materialise_dentry(struct dentry *dentry, struct dentry *anon)
2550 struct dentry *dparent;
2552 dentry_lock_for_move(anon, dentry);
2554 write_seqcount_begin(&dentry->d_seq);
2555 write_seqcount_begin(&anon->d_seq);
2557 dparent = dentry->d_parent;
2559 switch_names(dentry, anon);
2560 swap(dentry->d_name.hash, anon->d_name.hash);
2562 dentry->d_parent = dentry;
2563 list_del_init(&dentry->d_u.d_child);
2564 anon->d_parent = dparent;
2565 list_move(&anon->d_u.d_child, &dparent->d_subdirs);
2567 write_seqcount_end(&dentry->d_seq);
2568 write_seqcount_end(&anon->d_seq);
2570 dentry_unlock_parents_for_move(anon, dentry);
2571 spin_unlock(&dentry->d_lock);
2573 /* anon->d_lock still locked, returns locked */
2574 anon->d_flags &= ~DCACHE_DISCONNECTED;
2578 * d_materialise_unique - introduce an inode into the tree
2579 * @dentry: candidate dentry
2580 * @inode: inode to bind to the dentry, to which aliases may be attached
2582 * Introduces an dentry into the tree, substituting an extant disconnected
2583 * root directory alias in its place if there is one. Caller must hold the
2584 * i_mutex of the parent directory.
2586 struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
2588 struct dentry *actual;
2590 BUG_ON(!d_unhashed(dentry));
2594 __d_instantiate(dentry, NULL);
2599 spin_lock(&inode->i_lock);
2601 if (S_ISDIR(inode->i_mode)) {
2602 struct dentry *alias;
2604 /* Does an aliased dentry already exist? */
2605 alias = __d_find_alias(inode, 0);
2608 write_seqlock(&rename_lock);
2610 if (d_ancestor(alias, dentry)) {
2611 /* Check for loops */
2612 actual = ERR_PTR(-ELOOP);
2613 spin_unlock(&inode->i_lock);
2614 } else if (IS_ROOT(alias)) {
2615 /* Is this an anonymous mountpoint that we
2616 * could splice into our tree? */
2617 __d_materialise_dentry(dentry, alias);
2618 write_sequnlock(&rename_lock);
2622 /* Nope, but we must(!) avoid directory
2623 * aliasing. This drops inode->i_lock */
2624 actual = __d_unalias(inode, dentry, alias);
2626 write_sequnlock(&rename_lock);
2627 if (IS_ERR(actual)) {
2628 if (PTR_ERR(actual) == -ELOOP)
2629 pr_warn_ratelimited(
2630 "VFS: Lookup of '%s' in %s %s"
2631 " would have caused loop\n",
2632 dentry->d_name.name,
2633 inode->i_sb->s_type->name,
2641 /* Add a unique reference */
2642 actual = __d_instantiate_unique(dentry, inode);
2646 BUG_ON(!d_unhashed(actual));
2648 spin_lock(&actual->d_lock);
2651 spin_unlock(&actual->d_lock);
2652 spin_unlock(&inode->i_lock);
2654 if (actual == dentry) {
2655 security_d_instantiate(dentry, inode);
2662 EXPORT_SYMBOL_GPL(d_materialise_unique);
2664 static int prepend(char **buffer, int *buflen, const char *str, int namelen)
2668 return -ENAMETOOLONG;
2670 memcpy(*buffer, str, namelen);
2675 * prepend_name - prepend a pathname in front of current buffer pointer
2676 * buffer: buffer pointer
2677 * buflen: allocated length of the buffer
2678 * name: name string and length qstr structure
2680 * With RCU path tracing, it may race with d_move(). Use ACCESS_ONCE() to
2681 * make sure that either the old or the new name pointer and length are
2682 * fetched. However, there may be mismatch between length and pointer.
2683 * The length cannot be trusted, we need to copy it byte-by-byte until
2684 * the length is reached or a null byte is found. It also prepends "/" at
2685 * the beginning of the name. The sequence number check at the caller will
2686 * retry it again when a d_move() does happen. So any garbage in the buffer
2687 * due to mismatched pointer and length will be discarded.
2689 static int prepend_name(char **buffer, int *buflen, struct qstr *name)
2691 const char *dname = ACCESS_ONCE(name->name);
2692 u32 dlen = ACCESS_ONCE(name->len);
2695 if (*buflen < dlen + 1)
2696 return -ENAMETOOLONG;
2697 *buflen -= dlen + 1;
2698 p = *buffer -= dlen + 1;
2710 * prepend_path - Prepend path string to a buffer
2711 * @path: the dentry/vfsmount to report
2712 * @root: root vfsmnt/dentry
2713 * @buffer: pointer to the end of the buffer
2714 * @buflen: pointer to buffer length
2716 * The function tries to write out the pathname without taking any lock other
2717 * than the RCU read lock to make sure that dentries won't go away. It only
2718 * checks the sequence number of the global rename_lock as any change in the
2719 * dentry's d_seq will be preceded by changes in the rename_lock sequence
2720 * number. If the sequence number had been change, it will restart the whole
2721 * pathname back-tracing sequence again. It performs a total of 3 trials of
2722 * lockless back-tracing sequences before falling back to take the
2725 static int prepend_path(const struct path *path,
2726 const struct path *root,
2727 char **buffer, int *buflen)
2729 struct dentry *dentry = path->dentry;
2730 struct vfsmount *vfsmnt = path->mnt;
2731 struct mount *mnt = real_mount(vfsmnt);
2741 read_seqbegin_or_lock(&rename_lock, &seq);
2742 while (dentry != root->dentry || vfsmnt != root->mnt) {
2743 struct dentry * parent;
2745 if (dentry == vfsmnt->mnt_root || IS_ROOT(dentry)) {
2747 if (mnt_has_parent(mnt)) {
2748 dentry = mnt->mnt_mountpoint;
2749 mnt = mnt->mnt_parent;
2754 * Filesystems needing to implement special "root names"
2755 * should do so with ->d_dname()
2757 if (IS_ROOT(dentry) &&
2758 (dentry->d_name.len != 1 ||
2759 dentry->d_name.name[0] != '/')) {
2760 WARN(1, "Root dentry has weird name <%.*s>\n",
2761 (int) dentry->d_name.len,
2762 dentry->d_name.name);
2765 error = is_mounted(vfsmnt) ? 1 : 2;
2768 parent = dentry->d_parent;
2770 error = prepend_name(&bptr, &blen, &dentry->d_name);
2778 if (need_seqretry(&rename_lock, seq)) {
2782 done_seqretry(&rename_lock, seq);
2784 if (error >= 0 && bptr == *buffer) {
2786 error = -ENAMETOOLONG;
2796 * __d_path - return the path of a dentry
2797 * @path: the dentry/vfsmount to report
2798 * @root: root vfsmnt/dentry
2799 * @buf: buffer to return value in
2800 * @buflen: buffer length
2802 * Convert a dentry into an ASCII path name.
2804 * Returns a pointer into the buffer or an error code if the
2805 * path was too long.
2807 * "buflen" should be positive.
2809 * If the path is not reachable from the supplied root, return %NULL.
2811 char *__d_path(const struct path *path,
2812 const struct path *root,
2813 char *buf, int buflen)
2815 char *res = buf + buflen;
2818 prepend(&res, &buflen, "\0", 1);
2819 br_read_lock(&vfsmount_lock);
2820 error = prepend_path(path, root, &res, &buflen);
2821 br_read_unlock(&vfsmount_lock);
2824 return ERR_PTR(error);
2830 char *d_absolute_path(const struct path *path,
2831 char *buf, int buflen)
2833 struct path root = {};
2834 char *res = buf + buflen;
2837 prepend(&res, &buflen, "\0", 1);
2838 br_read_lock(&vfsmount_lock);
2839 error = prepend_path(path, &root, &res, &buflen);
2840 br_read_unlock(&vfsmount_lock);
2845 return ERR_PTR(error);
2850 * same as __d_path but appends "(deleted)" for unlinked files.
2852 static int path_with_deleted(const struct path *path,
2853 const struct path *root,
2854 char **buf, int *buflen)
2856 prepend(buf, buflen, "\0", 1);
2857 if (d_unlinked(path->dentry)) {
2858 int error = prepend(buf, buflen, " (deleted)", 10);
2863 return prepend_path(path, root, buf, buflen);
2866 static int prepend_unreachable(char **buffer, int *buflen)
2868 return prepend(buffer, buflen, "(unreachable)", 13);
2872 * d_path - return the path of a dentry
2873 * @path: path to report
2874 * @buf: buffer to return value in
2875 * @buflen: buffer length
2877 * Convert a dentry into an ASCII path name. If the entry has been deleted
2878 * the string " (deleted)" is appended. Note that this is ambiguous.
2880 * Returns a pointer into the buffer or an error code if the path was
2881 * too long. Note: Callers should use the returned pointer, not the passed
2882 * in buffer, to use the name! The implementation often starts at an offset
2883 * into the buffer, and may leave 0 bytes at the start.
2885 * "buflen" should be positive.
2887 char *d_path(const struct path *path, char *buf, int buflen)
2889 char *res = buf + buflen;
2894 * We have various synthetic filesystems that never get mounted. On
2895 * these filesystems dentries are never used for lookup purposes, and
2896 * thus don't need to be hashed. They also don't need a name until a
2897 * user wants to identify the object in /proc/pid/fd/. The little hack
2898 * below allows us to generate a name for these objects on demand:
2900 if (path->dentry->d_op && path->dentry->d_op->d_dname)
2901 return path->dentry->d_op->d_dname(path->dentry, buf, buflen);
2903 get_fs_root(current->fs, &root);
2904 br_read_lock(&vfsmount_lock);
2905 error = path_with_deleted(path, &root, &res, &buflen);
2906 br_read_unlock(&vfsmount_lock);
2908 res = ERR_PTR(error);
2912 EXPORT_SYMBOL(d_path);
2915 * Helper function for dentry_operations.d_dname() members
2917 char *dynamic_dname(struct dentry *dentry, char *buffer, int buflen,
2918 const char *fmt, ...)
2924 va_start(args, fmt);
2925 sz = vsnprintf(temp, sizeof(temp), fmt, args) + 1;
2928 if (sz > sizeof(temp) || sz > buflen)
2929 return ERR_PTR(-ENAMETOOLONG);
2931 buffer += buflen - sz;
2932 return memcpy(buffer, temp, sz);
2935 char *simple_dname(struct dentry *dentry, char *buffer, int buflen)
2937 char *end = buffer + buflen;
2938 /* these dentries are never renamed, so d_lock is not needed */
2939 if (prepend(&end, &buflen, " (deleted)", 11) ||
2940 prepend(&end, &buflen, dentry->d_name.name, dentry->d_name.len) ||
2941 prepend(&end, &buflen, "/", 1))
2942 end = ERR_PTR(-ENAMETOOLONG);
2947 * Write full pathname from the root of the filesystem into the buffer.
2949 static char *__dentry_path(struct dentry *dentry, char *buf, int buflen)
2959 prepend(&end, &len, "\0", 1);
2965 read_seqbegin_or_lock(&rename_lock, &seq);
2966 while (!IS_ROOT(dentry)) {
2967 struct dentry *parent = dentry->d_parent;
2971 error = prepend_name(&end, &len, &dentry->d_name);
2980 if (need_seqretry(&rename_lock, seq)) {
2984 done_seqretry(&rename_lock, seq);
2989 return ERR_PTR(-ENAMETOOLONG);
2992 char *dentry_path_raw(struct dentry *dentry, char *buf, int buflen)
2994 return __dentry_path(dentry, buf, buflen);
2996 EXPORT_SYMBOL(dentry_path_raw);
2998 char *dentry_path(struct dentry *dentry, char *buf, int buflen)
3003 if (d_unlinked(dentry)) {
3005 if (prepend(&p, &buflen, "//deleted", 10) != 0)
3009 retval = __dentry_path(dentry, buf, buflen);
3010 if (!IS_ERR(retval) && p)
3011 *p = '/'; /* restore '/' overriden with '\0' */
3014 return ERR_PTR(-ENAMETOOLONG);
3018 * NOTE! The user-level library version returns a
3019 * character pointer. The kernel system call just
3020 * returns the length of the buffer filled (which
3021 * includes the ending '\0' character), or a negative
3022 * error value. So libc would do something like
3024 * char *getcwd(char * buf, size_t size)
3028 * retval = sys_getcwd(buf, size);
3035 SYSCALL_DEFINE2(getcwd, char __user *, buf, unsigned long, size)
3038 struct path pwd, root;
3039 char *page = (char *) __get_free_page(GFP_USER);
3044 get_fs_root_and_pwd(current->fs, &root, &pwd);
3047 br_read_lock(&vfsmount_lock);
3048 if (!d_unlinked(pwd.dentry)) {
3050 char *cwd = page + PAGE_SIZE;
3051 int buflen = PAGE_SIZE;
3053 prepend(&cwd, &buflen, "\0", 1);
3054 error = prepend_path(&pwd, &root, &cwd, &buflen);
3055 br_read_unlock(&vfsmount_lock);
3060 /* Unreachable from current root */
3062 error = prepend_unreachable(&cwd, &buflen);
3068 len = PAGE_SIZE + page - cwd;
3071 if (copy_to_user(buf, cwd, len))
3075 br_read_unlock(&vfsmount_lock);
3081 free_page((unsigned long) page);
3086 * Test whether new_dentry is a subdirectory of old_dentry.
3088 * Trivially implemented using the dcache structure
3092 * is_subdir - is new dentry a subdirectory of old_dentry
3093 * @new_dentry: new dentry
3094 * @old_dentry: old dentry
3096 * Returns 1 if new_dentry is a subdirectory of the parent (at any depth).
3097 * Returns 0 otherwise.
3098 * Caller must ensure that "new_dentry" is pinned before calling is_subdir()
3101 int is_subdir(struct dentry *new_dentry, struct dentry *old_dentry)
3106 if (new_dentry == old_dentry)
3110 /* for restarting inner loop in case of seq retry */
3111 seq = read_seqbegin(&rename_lock);
3113 * Need rcu_readlock to protect against the d_parent trashing
3117 if (d_ancestor(old_dentry, new_dentry))
3122 } while (read_seqretry(&rename_lock, seq));
3127 static enum d_walk_ret d_genocide_kill(void *data, struct dentry *dentry)
3129 struct dentry *root = data;
3130 if (dentry != root) {
3131 if (d_unhashed(dentry) || !dentry->d_inode)
3134 if (!(dentry->d_flags & DCACHE_GENOCIDE)) {
3135 dentry->d_flags |= DCACHE_GENOCIDE;
3136 dentry->d_lockref.count--;
3139 return D_WALK_CONTINUE;
3142 void d_genocide(struct dentry *parent)
3144 d_walk(parent, parent, d_genocide_kill, NULL);
3147 void d_tmpfile(struct dentry *dentry, struct inode *inode)
3149 inode_dec_link_count(inode);
3150 BUG_ON(dentry->d_name.name != dentry->d_iname ||
3151 !hlist_unhashed(&dentry->d_alias) ||
3152 !d_unlinked(dentry));
3153 spin_lock(&dentry->d_parent->d_lock);
3154 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
3155 dentry->d_name.len = sprintf(dentry->d_iname, "#%llu",
3156 (unsigned long long)inode->i_ino);
3157 spin_unlock(&dentry->d_lock);
3158 spin_unlock(&dentry->d_parent->d_lock);
3159 d_instantiate(dentry, inode);
3161 EXPORT_SYMBOL(d_tmpfile);
3163 static __initdata unsigned long dhash_entries;
3164 static int __init set_dhash_entries(char *str)
3168 dhash_entries = simple_strtoul(str, &str, 0);
3171 __setup("dhash_entries=", set_dhash_entries);
3173 static void __init dcache_init_early(void)
3177 /* If hashes are distributed across NUMA nodes, defer
3178 * hash allocation until vmalloc space is available.
3184 alloc_large_system_hash("Dentry cache",
3185 sizeof(struct hlist_bl_head),
3194 for (loop = 0; loop < (1U << d_hash_shift); loop++)
3195 INIT_HLIST_BL_HEAD(dentry_hashtable + loop);
3198 static void __init dcache_init(void)
3203 * A constructor could be added for stable state like the lists,
3204 * but it is probably not worth it because of the cache nature
3207 dentry_cache = KMEM_CACHE(dentry,
3208 SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|SLAB_MEM_SPREAD);
3210 /* Hash may have been set up in dcache_init_early */
3215 alloc_large_system_hash("Dentry cache",
3216 sizeof(struct hlist_bl_head),
3225 for (loop = 0; loop < (1U << d_hash_shift); loop++)
3226 INIT_HLIST_BL_HEAD(dentry_hashtable + loop);
3229 /* SLAB cache for __getname() consumers */
3230 struct kmem_cache *names_cachep __read_mostly;
3231 EXPORT_SYMBOL(names_cachep);
3233 EXPORT_SYMBOL(d_genocide);
3235 void __init vfs_caches_init_early(void)
3237 dcache_init_early();
3241 void __init vfs_caches_init(unsigned long mempages)
3243 unsigned long reserve;
3245 /* Base hash sizes on available memory, with a reserve equal to
3246 150% of current kernel size */
3248 reserve = min((mempages - nr_free_pages()) * 3/2, mempages - 1);
3249 mempages -= reserve;
3251 names_cachep = kmem_cache_create("names_cache", PATH_MAX, 0,
3252 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
3256 files_init(mempages);