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/ratelimit.h>
18 #include <linux/string.h>
21 #include <linux/fscrypt.h>
22 #include <linux/fsnotify.h>
23 #include <linux/slab.h>
24 #include <linux/init.h>
25 #include <linux/hash.h>
26 #include <linux/cache.h>
27 #include <linux/export.h>
28 #include <linux/security.h>
29 #include <linux/seqlock.h>
30 #include <linux/memblock.h>
31 #include <linux/bit_spinlock.h>
32 #include <linux/rculist_bl.h>
33 #include <linux/list_lru.h>
39 * dcache->d_inode->i_lock protects:
40 * - i_dentry, d_u.d_alias, d_inode of aliases
41 * dcache_hash_bucket lock protects:
42 * - the dcache hash table
43 * s_roots bl list spinlock protects:
44 * - the s_roots list (see __d_drop)
45 * dentry->d_sb->s_dentry_lru_lock protects:
46 * - the dcache lru lists and counters
53 * - d_parent and d_subdirs
54 * - childrens' d_child and d_parent
55 * - d_u.d_alias, d_inode
58 * dentry->d_inode->i_lock
60 * dentry->d_sb->s_dentry_lru_lock
61 * dcache_hash_bucket lock
64 * If there is an ancestor relationship:
65 * dentry->d_parent->...->d_parent->d_lock
67 * dentry->d_parent->d_lock
70 * If no ancestor relationship:
71 * arbitrary, since it's serialized on rename_lock
73 int sysctl_vfs_cache_pressure __read_mostly = 100;
74 EXPORT_SYMBOL_GPL(sysctl_vfs_cache_pressure);
76 __cacheline_aligned_in_smp DEFINE_SEQLOCK(rename_lock);
78 EXPORT_SYMBOL(rename_lock);
80 static struct kmem_cache *dentry_cache __read_mostly;
82 const struct qstr empty_name = QSTR_INIT("", 0);
83 EXPORT_SYMBOL(empty_name);
84 const struct qstr slash_name = QSTR_INIT("/", 1);
85 EXPORT_SYMBOL(slash_name);
88 * This is the single most critical data structure when it comes
89 * to the dcache: the hashtable for lookups. Somebody should try
90 * to make this good - I've just made it work.
92 * This hash-function tries to avoid losing too many bits of hash
93 * information, yet avoid using a prime hash-size or similar.
96 static unsigned int d_hash_shift __read_mostly;
98 static struct hlist_bl_head *dentry_hashtable __read_mostly;
100 static inline struct hlist_bl_head *d_hash(unsigned int hash)
102 return dentry_hashtable + (hash >> d_hash_shift);
105 #define IN_LOOKUP_SHIFT 10
106 static struct hlist_bl_head in_lookup_hashtable[1 << IN_LOOKUP_SHIFT];
108 static inline struct hlist_bl_head *in_lookup_hash(const struct dentry *parent,
111 hash += (unsigned long) parent / L1_CACHE_BYTES;
112 return in_lookup_hashtable + hash_32(hash, IN_LOOKUP_SHIFT);
116 /* Statistics gathering. */
117 struct dentry_stat_t dentry_stat = {
121 static DEFINE_PER_CPU(long, nr_dentry);
122 static DEFINE_PER_CPU(long, nr_dentry_unused);
123 static DEFINE_PER_CPU(long, nr_dentry_negative);
125 #if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
128 * Here we resort to our own counters instead of using generic per-cpu counters
129 * for consistency with what the vfs inode code does. We are expected to harvest
130 * better code and performance by having our own specialized counters.
132 * Please note that the loop is done over all possible CPUs, not over all online
133 * CPUs. The reason for this is that we don't want to play games with CPUs going
134 * on and off. If one of them goes off, we will just keep their counters.
136 * glommer: See cffbc8a for details, and if you ever intend to change this,
137 * please update all vfs counters to match.
139 static long get_nr_dentry(void)
143 for_each_possible_cpu(i)
144 sum += per_cpu(nr_dentry, i);
145 return sum < 0 ? 0 : sum;
148 static long get_nr_dentry_unused(void)
152 for_each_possible_cpu(i)
153 sum += per_cpu(nr_dentry_unused, i);
154 return sum < 0 ? 0 : sum;
157 static long get_nr_dentry_negative(void)
162 for_each_possible_cpu(i)
163 sum += per_cpu(nr_dentry_negative, i);
164 return sum < 0 ? 0 : sum;
167 int proc_nr_dentry(struct ctl_table *table, int write, void __user *buffer,
168 size_t *lenp, loff_t *ppos)
170 dentry_stat.nr_dentry = get_nr_dentry();
171 dentry_stat.nr_unused = get_nr_dentry_unused();
172 dentry_stat.nr_negative = get_nr_dentry_negative();
173 return proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
178 * Compare 2 name strings, return 0 if they match, otherwise non-zero.
179 * The strings are both count bytes long, and count is non-zero.
181 #ifdef CONFIG_DCACHE_WORD_ACCESS
183 #include <asm/word-at-a-time.h>
185 * NOTE! 'cs' and 'scount' come from a dentry, so it has a
186 * aligned allocation for this particular component. We don't
187 * strictly need the load_unaligned_zeropad() safety, but it
188 * doesn't hurt either.
190 * In contrast, 'ct' and 'tcount' can be from a pathname, and do
191 * need the careful unaligned handling.
193 static inline int dentry_string_cmp(const unsigned char *cs, const unsigned char *ct, unsigned tcount)
195 unsigned long a,b,mask;
198 a = read_word_at_a_time(cs);
199 b = load_unaligned_zeropad(ct);
200 if (tcount < sizeof(unsigned long))
202 if (unlikely(a != b))
204 cs += sizeof(unsigned long);
205 ct += sizeof(unsigned long);
206 tcount -= sizeof(unsigned long);
210 mask = bytemask_from_count(tcount);
211 return unlikely(!!((a ^ b) & mask));
216 static inline int dentry_string_cmp(const unsigned char *cs, const unsigned char *ct, unsigned tcount)
230 static inline int dentry_cmp(const struct dentry *dentry, const unsigned char *ct, unsigned tcount)
233 * Be careful about RCU walk racing with rename:
234 * use 'READ_ONCE' to fetch the name pointer.
236 * NOTE! Even if a rename will mean that the length
237 * was not loaded atomically, we don't care. The
238 * RCU walk will check the sequence count eventually,
239 * and catch it. And we won't overrun the buffer,
240 * because we're reading the name pointer atomically,
241 * and a dentry name is guaranteed to be properly
242 * terminated with a NUL byte.
244 * End result: even if 'len' is wrong, we'll exit
245 * early because the data cannot match (there can
246 * be no NUL in the ct/tcount data)
248 const unsigned char *cs = READ_ONCE(dentry->d_name.name);
250 return dentry_string_cmp(cs, ct, tcount);
253 struct external_name {
256 struct rcu_head head;
258 unsigned char name[];
261 static inline struct external_name *external_name(struct dentry *dentry)
263 return container_of(dentry->d_name.name, struct external_name, name[0]);
266 static void __d_free(struct rcu_head *head)
268 struct dentry *dentry = container_of(head, struct dentry, d_u.d_rcu);
270 kmem_cache_free(dentry_cache, dentry);
273 static void __d_free_external(struct rcu_head *head)
275 struct dentry *dentry = container_of(head, struct dentry, d_u.d_rcu);
276 kfree(external_name(dentry));
277 kmem_cache_free(dentry_cache, dentry);
280 static inline int dname_external(const struct dentry *dentry)
282 return dentry->d_name.name != dentry->d_iname;
285 void take_dentry_name_snapshot(struct name_snapshot *name, struct dentry *dentry)
287 spin_lock(&dentry->d_lock);
288 if (unlikely(dname_external(dentry))) {
289 struct external_name *p = external_name(dentry);
290 atomic_inc(&p->u.count);
291 spin_unlock(&dentry->d_lock);
292 name->name = p->name;
294 memcpy(name->inline_name, dentry->d_iname,
295 dentry->d_name.len + 1);
296 spin_unlock(&dentry->d_lock);
297 name->name = name->inline_name;
300 EXPORT_SYMBOL(take_dentry_name_snapshot);
302 void release_dentry_name_snapshot(struct name_snapshot *name)
304 if (unlikely(name->name != name->inline_name)) {
305 struct external_name *p;
306 p = container_of(name->name, struct external_name, name[0]);
307 if (unlikely(atomic_dec_and_test(&p->u.count)))
308 kfree_rcu(p, u.head);
311 EXPORT_SYMBOL(release_dentry_name_snapshot);
313 static inline void __d_set_inode_and_type(struct dentry *dentry,
319 dentry->d_inode = inode;
320 flags = READ_ONCE(dentry->d_flags);
321 flags &= ~(DCACHE_ENTRY_TYPE | DCACHE_FALLTHRU);
323 WRITE_ONCE(dentry->d_flags, flags);
326 static inline void __d_clear_type_and_inode(struct dentry *dentry)
328 unsigned flags = READ_ONCE(dentry->d_flags);
330 flags &= ~(DCACHE_ENTRY_TYPE | DCACHE_FALLTHRU);
331 WRITE_ONCE(dentry->d_flags, flags);
332 dentry->d_inode = NULL;
333 if (dentry->d_flags & DCACHE_LRU_LIST)
334 this_cpu_inc(nr_dentry_negative);
337 static void dentry_free(struct dentry *dentry)
339 WARN_ON(!hlist_unhashed(&dentry->d_u.d_alias));
340 if (unlikely(dname_external(dentry))) {
341 struct external_name *p = external_name(dentry);
342 if (likely(atomic_dec_and_test(&p->u.count))) {
343 call_rcu(&dentry->d_u.d_rcu, __d_free_external);
347 /* if dentry was never visible to RCU, immediate free is OK */
348 if (!(dentry->d_flags & DCACHE_RCUACCESS))
349 __d_free(&dentry->d_u.d_rcu);
351 call_rcu(&dentry->d_u.d_rcu, __d_free);
355 * Release the dentry's inode, using the filesystem
356 * d_iput() operation if defined.
358 static void dentry_unlink_inode(struct dentry * dentry)
359 __releases(dentry->d_lock)
360 __releases(dentry->d_inode->i_lock)
362 struct inode *inode = dentry->d_inode;
364 raw_write_seqcount_begin(&dentry->d_seq);
365 __d_clear_type_and_inode(dentry);
366 hlist_del_init(&dentry->d_u.d_alias);
367 raw_write_seqcount_end(&dentry->d_seq);
368 spin_unlock(&dentry->d_lock);
369 spin_unlock(&inode->i_lock);
371 fsnotify_inoderemove(inode);
372 if (dentry->d_op && dentry->d_op->d_iput)
373 dentry->d_op->d_iput(dentry, inode);
379 * The DCACHE_LRU_LIST bit is set whenever the 'd_lru' entry
380 * is in use - which includes both the "real" per-superblock
381 * LRU list _and_ the DCACHE_SHRINK_LIST use.
383 * The DCACHE_SHRINK_LIST bit is set whenever the dentry is
384 * on the shrink list (ie not on the superblock LRU list).
386 * The per-cpu "nr_dentry_unused" counters are updated with
387 * the DCACHE_LRU_LIST bit.
389 * The per-cpu "nr_dentry_negative" counters are only updated
390 * when deleted from or added to the per-superblock LRU list, not
391 * from/to the shrink list. That is to avoid an unneeded dec/inc
392 * pair when moving from LRU to shrink list in select_collect().
394 * These helper functions make sure we always follow the
395 * rules. d_lock must be held by the caller.
397 #define D_FLAG_VERIFY(dentry,x) WARN_ON_ONCE(((dentry)->d_flags & (DCACHE_LRU_LIST | DCACHE_SHRINK_LIST)) != (x))
398 static void d_lru_add(struct dentry *dentry)
400 D_FLAG_VERIFY(dentry, 0);
401 dentry->d_flags |= DCACHE_LRU_LIST;
402 this_cpu_inc(nr_dentry_unused);
403 if (d_is_negative(dentry))
404 this_cpu_inc(nr_dentry_negative);
405 WARN_ON_ONCE(!list_lru_add(&dentry->d_sb->s_dentry_lru, &dentry->d_lru));
408 static void d_lru_del(struct dentry *dentry)
410 D_FLAG_VERIFY(dentry, DCACHE_LRU_LIST);
411 dentry->d_flags &= ~DCACHE_LRU_LIST;
412 this_cpu_dec(nr_dentry_unused);
413 if (d_is_negative(dentry))
414 this_cpu_dec(nr_dentry_negative);
415 WARN_ON_ONCE(!list_lru_del(&dentry->d_sb->s_dentry_lru, &dentry->d_lru));
418 static void d_shrink_del(struct dentry *dentry)
420 D_FLAG_VERIFY(dentry, DCACHE_SHRINK_LIST | DCACHE_LRU_LIST);
421 list_del_init(&dentry->d_lru);
422 dentry->d_flags &= ~(DCACHE_SHRINK_LIST | DCACHE_LRU_LIST);
423 this_cpu_dec(nr_dentry_unused);
426 static void d_shrink_add(struct dentry *dentry, struct list_head *list)
428 D_FLAG_VERIFY(dentry, 0);
429 list_add(&dentry->d_lru, list);
430 dentry->d_flags |= DCACHE_SHRINK_LIST | DCACHE_LRU_LIST;
431 this_cpu_inc(nr_dentry_unused);
435 * These can only be called under the global LRU lock, ie during the
436 * callback for freeing the LRU list. "isolate" removes it from the
437 * LRU lists entirely, while shrink_move moves it to the indicated
440 static void d_lru_isolate(struct list_lru_one *lru, struct dentry *dentry)
442 D_FLAG_VERIFY(dentry, DCACHE_LRU_LIST);
443 dentry->d_flags &= ~DCACHE_LRU_LIST;
444 this_cpu_dec(nr_dentry_unused);
445 if (d_is_negative(dentry))
446 this_cpu_dec(nr_dentry_negative);
447 list_lru_isolate(lru, &dentry->d_lru);
450 static void d_lru_shrink_move(struct list_lru_one *lru, struct dentry *dentry,
451 struct list_head *list)
453 D_FLAG_VERIFY(dentry, DCACHE_LRU_LIST);
454 dentry->d_flags |= DCACHE_SHRINK_LIST;
455 if (d_is_negative(dentry))
456 this_cpu_dec(nr_dentry_negative);
457 list_lru_isolate_move(lru, &dentry->d_lru, list);
461 * d_drop - drop a dentry
462 * @dentry: dentry to drop
464 * d_drop() unhashes the entry from the parent dentry hashes, so that it won't
465 * be found through a VFS lookup any more. Note that this is different from
466 * deleting the dentry - d_delete will try to mark the dentry negative if
467 * possible, giving a successful _negative_ lookup, while d_drop will
468 * just make the cache lookup fail.
470 * d_drop() is used mainly for stuff that wants to invalidate a dentry for some
471 * reason (NFS timeouts or autofs deletes).
473 * __d_drop requires dentry->d_lock
474 * ___d_drop doesn't mark dentry as "unhashed"
475 * (dentry->d_hash.pprev will be LIST_POISON2, not NULL).
477 static void ___d_drop(struct dentry *dentry)
479 struct hlist_bl_head *b;
481 * Hashed dentries are normally on the dentry hashtable,
482 * with the exception of those newly allocated by
483 * d_obtain_root, which are always IS_ROOT:
485 if (unlikely(IS_ROOT(dentry)))
486 b = &dentry->d_sb->s_roots;
488 b = d_hash(dentry->d_name.hash);
491 __hlist_bl_del(&dentry->d_hash);
495 void __d_drop(struct dentry *dentry)
497 if (!d_unhashed(dentry)) {
499 dentry->d_hash.pprev = NULL;
500 write_seqcount_invalidate(&dentry->d_seq);
503 EXPORT_SYMBOL(__d_drop);
505 void d_drop(struct dentry *dentry)
507 spin_lock(&dentry->d_lock);
509 spin_unlock(&dentry->d_lock);
511 EXPORT_SYMBOL(d_drop);
513 static inline void dentry_unlist(struct dentry *dentry, struct dentry *parent)
517 * Inform d_walk() and shrink_dentry_list() that we are no longer
518 * attached to the dentry tree
520 dentry->d_flags |= DCACHE_DENTRY_KILLED;
521 if (unlikely(list_empty(&dentry->d_child)))
523 __list_del_entry(&dentry->d_child);
525 * Cursors can move around the list of children. While we'd been
526 * a normal list member, it didn't matter - ->d_child.next would've
527 * been updated. However, from now on it won't be and for the
528 * things like d_walk() it might end up with a nasty surprise.
529 * Normally d_walk() doesn't care about cursors moving around -
530 * ->d_lock on parent prevents that and since a cursor has no children
531 * of its own, we get through it without ever unlocking the parent.
532 * There is one exception, though - if we ascend from a child that
533 * gets killed as soon as we unlock it, the next sibling is found
534 * using the value left in its ->d_child.next. And if _that_
535 * pointed to a cursor, and cursor got moved (e.g. by lseek())
536 * before d_walk() regains parent->d_lock, we'll end up skipping
537 * everything the cursor had been moved past.
539 * Solution: make sure that the pointer left behind in ->d_child.next
540 * points to something that won't be moving around. I.e. skip the
543 while (dentry->d_child.next != &parent->d_subdirs) {
544 next = list_entry(dentry->d_child.next, struct dentry, d_child);
545 if (likely(!(next->d_flags & DCACHE_DENTRY_CURSOR)))
547 dentry->d_child.next = next->d_child.next;
551 static void __dentry_kill(struct dentry *dentry)
553 struct dentry *parent = NULL;
554 bool can_free = true;
555 if (!IS_ROOT(dentry))
556 parent = dentry->d_parent;
559 * The dentry is now unrecoverably dead to the world.
561 lockref_mark_dead(&dentry->d_lockref);
564 * inform the fs via d_prune that this dentry is about to be
565 * unhashed and destroyed.
567 if (dentry->d_flags & DCACHE_OP_PRUNE)
568 dentry->d_op->d_prune(dentry);
570 if (dentry->d_flags & DCACHE_LRU_LIST) {
571 if (!(dentry->d_flags & DCACHE_SHRINK_LIST))
574 /* if it was on the hash then remove it */
576 dentry_unlist(dentry, parent);
578 spin_unlock(&parent->d_lock);
580 dentry_unlink_inode(dentry);
582 spin_unlock(&dentry->d_lock);
583 this_cpu_dec(nr_dentry);
584 if (dentry->d_op && dentry->d_op->d_release)
585 dentry->d_op->d_release(dentry);
587 spin_lock(&dentry->d_lock);
588 if (dentry->d_flags & DCACHE_SHRINK_LIST) {
589 dentry->d_flags |= DCACHE_MAY_FREE;
592 spin_unlock(&dentry->d_lock);
593 if (likely(can_free))
598 static struct dentry *__lock_parent(struct dentry *dentry)
600 struct dentry *parent;
602 spin_unlock(&dentry->d_lock);
604 parent = READ_ONCE(dentry->d_parent);
605 spin_lock(&parent->d_lock);
607 * We can't blindly lock dentry until we are sure
608 * that we won't violate the locking order.
609 * Any changes of dentry->d_parent must have
610 * been done with parent->d_lock held, so
611 * spin_lock() above is enough of a barrier
612 * for checking if it's still our child.
614 if (unlikely(parent != dentry->d_parent)) {
615 spin_unlock(&parent->d_lock);
619 if (parent != dentry)
620 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
626 static inline struct dentry *lock_parent(struct dentry *dentry)
628 struct dentry *parent = dentry->d_parent;
631 if (likely(spin_trylock(&parent->d_lock)))
633 return __lock_parent(dentry);
636 static inline bool retain_dentry(struct dentry *dentry)
638 WARN_ON(d_in_lookup(dentry));
640 /* Unreachable? Get rid of it */
641 if (unlikely(d_unhashed(dentry)))
644 if (unlikely(dentry->d_flags & DCACHE_DISCONNECTED))
647 if (unlikely(dentry->d_flags & DCACHE_OP_DELETE)) {
648 if (dentry->d_op->d_delete(dentry))
651 /* retain; LRU fodder */
652 dentry->d_lockref.count--;
653 if (unlikely(!(dentry->d_flags & DCACHE_LRU_LIST)))
655 else if (unlikely(!(dentry->d_flags & DCACHE_REFERENCED)))
656 dentry->d_flags |= DCACHE_REFERENCED;
661 * Finish off a dentry we've decided to kill.
662 * dentry->d_lock must be held, returns with it unlocked.
663 * Returns dentry requiring refcount drop, or NULL if we're done.
665 static struct dentry *dentry_kill(struct dentry *dentry)
666 __releases(dentry->d_lock)
668 struct inode *inode = dentry->d_inode;
669 struct dentry *parent = NULL;
671 if (inode && unlikely(!spin_trylock(&inode->i_lock)))
674 if (!IS_ROOT(dentry)) {
675 parent = dentry->d_parent;
676 if (unlikely(!spin_trylock(&parent->d_lock))) {
677 parent = __lock_parent(dentry);
678 if (likely(inode || !dentry->d_inode))
680 /* negative that became positive */
682 spin_unlock(&parent->d_lock);
683 inode = dentry->d_inode;
687 __dentry_kill(dentry);
691 spin_unlock(&dentry->d_lock);
692 spin_lock(&inode->i_lock);
693 spin_lock(&dentry->d_lock);
694 parent = lock_parent(dentry);
696 if (unlikely(dentry->d_lockref.count != 1)) {
697 dentry->d_lockref.count--;
698 } else if (likely(!retain_dentry(dentry))) {
699 __dentry_kill(dentry);
702 /* we are keeping it, after all */
704 spin_unlock(&inode->i_lock);
706 spin_unlock(&parent->d_lock);
707 spin_unlock(&dentry->d_lock);
712 * Try to do a lockless dput(), and return whether that was successful.
714 * If unsuccessful, we return false, having already taken the dentry lock.
716 * The caller needs to hold the RCU read lock, so that the dentry is
717 * guaranteed to stay around even if the refcount goes down to zero!
719 static inline bool fast_dput(struct dentry *dentry)
722 unsigned int d_flags;
725 * If we have a d_op->d_delete() operation, we sould not
726 * let the dentry count go to zero, so use "put_or_lock".
728 if (unlikely(dentry->d_flags & DCACHE_OP_DELETE))
729 return lockref_put_or_lock(&dentry->d_lockref);
732 * .. otherwise, we can try to just decrement the
733 * lockref optimistically.
735 ret = lockref_put_return(&dentry->d_lockref);
738 * If the lockref_put_return() failed due to the lock being held
739 * by somebody else, the fast path has failed. We will need to
740 * get the lock, and then check the count again.
742 if (unlikely(ret < 0)) {
743 spin_lock(&dentry->d_lock);
744 if (dentry->d_lockref.count > 1) {
745 dentry->d_lockref.count--;
746 spin_unlock(&dentry->d_lock);
753 * If we weren't the last ref, we're done.
759 * Careful, careful. The reference count went down
760 * to zero, but we don't hold the dentry lock, so
761 * somebody else could get it again, and do another
762 * dput(), and we need to not race with that.
764 * However, there is a very special and common case
765 * where we don't care, because there is nothing to
766 * do: the dentry is still hashed, it does not have
767 * a 'delete' op, and it's referenced and already on
770 * NOTE! Since we aren't locked, these values are
771 * not "stable". However, it is sufficient that at
772 * some point after we dropped the reference the
773 * dentry was hashed and the flags had the proper
774 * value. Other dentry users may have re-gotten
775 * a reference to the dentry and change that, but
776 * our work is done - we can leave the dentry
777 * around with a zero refcount.
780 d_flags = READ_ONCE(dentry->d_flags);
781 d_flags &= DCACHE_REFERENCED | DCACHE_LRU_LIST | DCACHE_DISCONNECTED;
783 /* Nothing to do? Dropping the reference was all we needed? */
784 if (d_flags == (DCACHE_REFERENCED | DCACHE_LRU_LIST) && !d_unhashed(dentry))
788 * Not the fast normal case? Get the lock. We've already decremented
789 * the refcount, but we'll need to re-check the situation after
792 spin_lock(&dentry->d_lock);
795 * Did somebody else grab a reference to it in the meantime, and
796 * we're no longer the last user after all? Alternatively, somebody
797 * else could have killed it and marked it dead. Either way, we
798 * don't need to do anything else.
800 if (dentry->d_lockref.count) {
801 spin_unlock(&dentry->d_lock);
806 * Re-get the reference we optimistically dropped. We hold the
807 * lock, and we just tested that it was zero, so we can just
810 dentry->d_lockref.count = 1;
818 * This is complicated by the fact that we do not want to put
819 * dentries that are no longer on any hash chain on the unused
820 * list: we'd much rather just get rid of them immediately.
822 * However, that implies that we have to traverse the dentry
823 * tree upwards to the parents which might _also_ now be
824 * scheduled for deletion (it may have been only waiting for
825 * its last child to go away).
827 * This tail recursion is done by hand as we don't want to depend
828 * on the compiler to always get this right (gcc generally doesn't).
829 * Real recursion would eat up our stack space.
833 * dput - release a dentry
834 * @dentry: dentry to release
836 * Release a dentry. This will drop the usage count and if appropriate
837 * call the dentry unlink method as well as removing it from the queues and
838 * releasing its resources. If the parent dentries were scheduled for release
839 * they too may now get deleted.
841 void dput(struct dentry *dentry)
847 if (likely(fast_dput(dentry))) {
852 /* Slow case: now with the dentry lock held */
855 if (likely(retain_dentry(dentry))) {
856 spin_unlock(&dentry->d_lock);
860 dentry = dentry_kill(dentry);
866 /* This must be called with d_lock held */
867 static inline void __dget_dlock(struct dentry *dentry)
869 dentry->d_lockref.count++;
872 static inline void __dget(struct dentry *dentry)
874 lockref_get(&dentry->d_lockref);
877 struct dentry *dget_parent(struct dentry *dentry)
883 * Do optimistic parent lookup without any
887 ret = READ_ONCE(dentry->d_parent);
888 gotref = lockref_get_not_zero(&ret->d_lockref);
890 if (likely(gotref)) {
891 if (likely(ret == READ_ONCE(dentry->d_parent)))
898 * Don't need rcu_dereference because we re-check it was correct under
902 ret = dentry->d_parent;
903 spin_lock(&ret->d_lock);
904 if (unlikely(ret != dentry->d_parent)) {
905 spin_unlock(&ret->d_lock);
910 BUG_ON(!ret->d_lockref.count);
911 ret->d_lockref.count++;
912 spin_unlock(&ret->d_lock);
915 EXPORT_SYMBOL(dget_parent);
917 static struct dentry * __d_find_any_alias(struct inode *inode)
919 struct dentry *alias;
921 if (hlist_empty(&inode->i_dentry))
923 alias = hlist_entry(inode->i_dentry.first, struct dentry, d_u.d_alias);
929 * d_find_any_alias - find any alias for a given inode
930 * @inode: inode to find an alias for
932 * If any aliases exist for the given inode, take and return a
933 * reference for one of them. If no aliases exist, return %NULL.
935 struct dentry *d_find_any_alias(struct inode *inode)
939 spin_lock(&inode->i_lock);
940 de = __d_find_any_alias(inode);
941 spin_unlock(&inode->i_lock);
944 EXPORT_SYMBOL(d_find_any_alias);
947 * d_find_alias - grab a hashed alias of inode
948 * @inode: inode in question
950 * If inode has a hashed alias, or is a directory and has any alias,
951 * acquire the reference to alias and return it. Otherwise return NULL.
952 * Notice that if inode is a directory there can be only one alias and
953 * it can be unhashed only if it has no children, or if it is the root
954 * of a filesystem, or if the directory was renamed and d_revalidate
955 * was the first vfs operation to notice.
957 * If the inode has an IS_ROOT, DCACHE_DISCONNECTED alias, then prefer
958 * any other hashed alias over that one.
960 static struct dentry *__d_find_alias(struct inode *inode)
962 struct dentry *alias;
964 if (S_ISDIR(inode->i_mode))
965 return __d_find_any_alias(inode);
967 hlist_for_each_entry(alias, &inode->i_dentry, d_u.d_alias) {
968 spin_lock(&alias->d_lock);
969 if (!d_unhashed(alias)) {
971 spin_unlock(&alias->d_lock);
974 spin_unlock(&alias->d_lock);
979 struct dentry *d_find_alias(struct inode *inode)
981 struct dentry *de = NULL;
983 if (!hlist_empty(&inode->i_dentry)) {
984 spin_lock(&inode->i_lock);
985 de = __d_find_alias(inode);
986 spin_unlock(&inode->i_lock);
990 EXPORT_SYMBOL(d_find_alias);
993 * Try to kill dentries associated with this inode.
994 * WARNING: you must own a reference to inode.
996 void d_prune_aliases(struct inode *inode)
998 struct dentry *dentry;
1000 spin_lock(&inode->i_lock);
1001 hlist_for_each_entry(dentry, &inode->i_dentry, d_u.d_alias) {
1002 spin_lock(&dentry->d_lock);
1003 if (!dentry->d_lockref.count) {
1004 struct dentry *parent = lock_parent(dentry);
1005 if (likely(!dentry->d_lockref.count)) {
1006 __dentry_kill(dentry);
1011 spin_unlock(&parent->d_lock);
1013 spin_unlock(&dentry->d_lock);
1015 spin_unlock(&inode->i_lock);
1017 EXPORT_SYMBOL(d_prune_aliases);
1020 * Lock a dentry from shrink list.
1021 * Called under rcu_read_lock() and dentry->d_lock; the former
1022 * guarantees that nothing we access will be freed under us.
1023 * Note that dentry is *not* protected from concurrent dentry_kill(),
1026 * Return false if dentry has been disrupted or grabbed, leaving
1027 * the caller to kick it off-list. Otherwise, return true and have
1028 * that dentry's inode and parent both locked.
1030 static bool shrink_lock_dentry(struct dentry *dentry)
1032 struct inode *inode;
1033 struct dentry *parent;
1035 if (dentry->d_lockref.count)
1038 inode = dentry->d_inode;
1039 if (inode && unlikely(!spin_trylock(&inode->i_lock))) {
1040 spin_unlock(&dentry->d_lock);
1041 spin_lock(&inode->i_lock);
1042 spin_lock(&dentry->d_lock);
1043 if (unlikely(dentry->d_lockref.count))
1045 /* changed inode means that somebody had grabbed it */
1046 if (unlikely(inode != dentry->d_inode))
1050 parent = dentry->d_parent;
1051 if (IS_ROOT(dentry) || likely(spin_trylock(&parent->d_lock)))
1054 spin_unlock(&dentry->d_lock);
1055 spin_lock(&parent->d_lock);
1056 if (unlikely(parent != dentry->d_parent)) {
1057 spin_unlock(&parent->d_lock);
1058 spin_lock(&dentry->d_lock);
1061 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
1062 if (likely(!dentry->d_lockref.count))
1064 spin_unlock(&parent->d_lock);
1067 spin_unlock(&inode->i_lock);
1071 static void shrink_dentry_list(struct list_head *list)
1073 while (!list_empty(list)) {
1074 struct dentry *dentry, *parent;
1076 dentry = list_entry(list->prev, struct dentry, d_lru);
1077 spin_lock(&dentry->d_lock);
1079 if (!shrink_lock_dentry(dentry)) {
1080 bool can_free = false;
1082 d_shrink_del(dentry);
1083 if (dentry->d_lockref.count < 0)
1084 can_free = dentry->d_flags & DCACHE_MAY_FREE;
1085 spin_unlock(&dentry->d_lock);
1087 dentry_free(dentry);
1091 d_shrink_del(dentry);
1092 parent = dentry->d_parent;
1093 __dentry_kill(dentry);
1094 if (parent == dentry)
1097 * We need to prune ancestors too. This is necessary to prevent
1098 * quadratic behavior of shrink_dcache_parent(), but is also
1099 * expected to be beneficial in reducing dentry cache
1103 while (dentry && !lockref_put_or_lock(&dentry->d_lockref))
1104 dentry = dentry_kill(dentry);
1108 static enum lru_status dentry_lru_isolate(struct list_head *item,
1109 struct list_lru_one *lru, spinlock_t *lru_lock, void *arg)
1111 struct list_head *freeable = arg;
1112 struct dentry *dentry = container_of(item, struct dentry, d_lru);
1116 * we are inverting the lru lock/dentry->d_lock here,
1117 * so use a trylock. If we fail to get the lock, just skip
1120 if (!spin_trylock(&dentry->d_lock))
1124 * Referenced dentries are still in use. If they have active
1125 * counts, just remove them from the LRU. Otherwise give them
1126 * another pass through the LRU.
1128 if (dentry->d_lockref.count) {
1129 d_lru_isolate(lru, dentry);
1130 spin_unlock(&dentry->d_lock);
1134 if (dentry->d_flags & DCACHE_REFERENCED) {
1135 dentry->d_flags &= ~DCACHE_REFERENCED;
1136 spin_unlock(&dentry->d_lock);
1139 * The list move itself will be made by the common LRU code. At
1140 * this point, we've dropped the dentry->d_lock but keep the
1141 * lru lock. This is safe to do, since every list movement is
1142 * protected by the lru lock even if both locks are held.
1144 * This is guaranteed by the fact that all LRU management
1145 * functions are intermediated by the LRU API calls like
1146 * list_lru_add and list_lru_del. List movement in this file
1147 * only ever occur through this functions or through callbacks
1148 * like this one, that are called from the LRU API.
1150 * The only exceptions to this are functions like
1151 * shrink_dentry_list, and code that first checks for the
1152 * DCACHE_SHRINK_LIST flag. Those are guaranteed to be
1153 * operating only with stack provided lists after they are
1154 * properly isolated from the main list. It is thus, always a
1160 d_lru_shrink_move(lru, dentry, freeable);
1161 spin_unlock(&dentry->d_lock);
1167 * prune_dcache_sb - shrink the dcache
1169 * @sc: shrink control, passed to list_lru_shrink_walk()
1171 * Attempt to shrink the superblock dcache LRU by @sc->nr_to_scan entries. This
1172 * is done when we need more memory and called from the superblock shrinker
1175 * This function may fail to free any resources if all the dentries are in
1178 long prune_dcache_sb(struct super_block *sb, struct shrink_control *sc)
1183 freed = list_lru_shrink_walk(&sb->s_dentry_lru, sc,
1184 dentry_lru_isolate, &dispose);
1185 shrink_dentry_list(&dispose);
1189 static enum lru_status dentry_lru_isolate_shrink(struct list_head *item,
1190 struct list_lru_one *lru, spinlock_t *lru_lock, void *arg)
1192 struct list_head *freeable = arg;
1193 struct dentry *dentry = container_of(item, struct dentry, d_lru);
1196 * we are inverting the lru lock/dentry->d_lock here,
1197 * so use a trylock. If we fail to get the lock, just skip
1200 if (!spin_trylock(&dentry->d_lock))
1203 d_lru_shrink_move(lru, dentry, freeable);
1204 spin_unlock(&dentry->d_lock);
1211 * shrink_dcache_sb - shrink dcache for a superblock
1214 * Shrink the dcache for the specified super block. This is used to free
1215 * the dcache before unmounting a file system.
1217 void shrink_dcache_sb(struct super_block *sb)
1222 list_lru_walk(&sb->s_dentry_lru,
1223 dentry_lru_isolate_shrink, &dispose, 1024);
1224 shrink_dentry_list(&dispose);
1225 } while (list_lru_count(&sb->s_dentry_lru) > 0);
1227 EXPORT_SYMBOL(shrink_dcache_sb);
1230 * enum d_walk_ret - action to talke during tree walk
1231 * @D_WALK_CONTINUE: contrinue walk
1232 * @D_WALK_QUIT: quit walk
1233 * @D_WALK_NORETRY: quit when retry is needed
1234 * @D_WALK_SKIP: skip this dentry and its children
1244 * d_walk - walk the dentry tree
1245 * @parent: start of walk
1246 * @data: data passed to @enter() and @finish()
1247 * @enter: callback when first entering the dentry
1249 * The @enter() callbacks are called with d_lock held.
1251 static void d_walk(struct dentry *parent, void *data,
1252 enum d_walk_ret (*enter)(void *, struct dentry *))
1254 struct dentry *this_parent;
1255 struct list_head *next;
1257 enum d_walk_ret ret;
1261 read_seqbegin_or_lock(&rename_lock, &seq);
1262 this_parent = parent;
1263 spin_lock(&this_parent->d_lock);
1265 ret = enter(data, this_parent);
1267 case D_WALK_CONTINUE:
1272 case D_WALK_NORETRY:
1277 next = this_parent->d_subdirs.next;
1279 while (next != &this_parent->d_subdirs) {
1280 struct list_head *tmp = next;
1281 struct dentry *dentry = list_entry(tmp, struct dentry, d_child);
1284 if (unlikely(dentry->d_flags & DCACHE_DENTRY_CURSOR))
1287 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
1289 ret = enter(data, dentry);
1291 case D_WALK_CONTINUE:
1294 spin_unlock(&dentry->d_lock);
1296 case D_WALK_NORETRY:
1300 spin_unlock(&dentry->d_lock);
1304 if (!list_empty(&dentry->d_subdirs)) {
1305 spin_unlock(&this_parent->d_lock);
1306 spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
1307 this_parent = dentry;
1308 spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
1311 spin_unlock(&dentry->d_lock);
1314 * All done at this level ... ascend and resume the search.
1318 if (this_parent != parent) {
1319 struct dentry *child = this_parent;
1320 this_parent = child->d_parent;
1322 spin_unlock(&child->d_lock);
1323 spin_lock(&this_parent->d_lock);
1325 /* might go back up the wrong parent if we have had a rename. */
1326 if (need_seqretry(&rename_lock, seq))
1328 /* go into the first sibling still alive */
1330 next = child->d_child.next;
1331 if (next == &this_parent->d_subdirs)
1333 child = list_entry(next, struct dentry, d_child);
1334 } while (unlikely(child->d_flags & DCACHE_DENTRY_KILLED));
1338 if (need_seqretry(&rename_lock, seq))
1343 spin_unlock(&this_parent->d_lock);
1344 done_seqretry(&rename_lock, seq);
1348 spin_unlock(&this_parent->d_lock);
1357 struct check_mount {
1358 struct vfsmount *mnt;
1359 unsigned int mounted;
1362 static enum d_walk_ret path_check_mount(void *data, struct dentry *dentry)
1364 struct check_mount *info = data;
1365 struct path path = { .mnt = info->mnt, .dentry = dentry };
1367 if (likely(!d_mountpoint(dentry)))
1368 return D_WALK_CONTINUE;
1369 if (__path_is_mountpoint(&path)) {
1373 return D_WALK_CONTINUE;
1377 * path_has_submounts - check for mounts over a dentry in the
1378 * current namespace.
1379 * @parent: path to check.
1381 * Return true if the parent or its subdirectories contain
1382 * a mount point in the current namespace.
1384 int path_has_submounts(const struct path *parent)
1386 struct check_mount data = { .mnt = parent->mnt, .mounted = 0 };
1388 read_seqlock_excl(&mount_lock);
1389 d_walk(parent->dentry, &data, path_check_mount);
1390 read_sequnlock_excl(&mount_lock);
1392 return data.mounted;
1394 EXPORT_SYMBOL(path_has_submounts);
1397 * Called by mount code to set a mountpoint and check if the mountpoint is
1398 * reachable (e.g. NFS can unhash a directory dentry and then the complete
1399 * subtree can become unreachable).
1401 * Only one of d_invalidate() and d_set_mounted() must succeed. For
1402 * this reason take rename_lock and d_lock on dentry and ancestors.
1404 int d_set_mounted(struct dentry *dentry)
1408 write_seqlock(&rename_lock);
1409 for (p = dentry->d_parent; !IS_ROOT(p); p = p->d_parent) {
1410 /* Need exclusion wrt. d_invalidate() */
1411 spin_lock(&p->d_lock);
1412 if (unlikely(d_unhashed(p))) {
1413 spin_unlock(&p->d_lock);
1416 spin_unlock(&p->d_lock);
1418 spin_lock(&dentry->d_lock);
1419 if (!d_unlinked(dentry)) {
1421 if (!d_mountpoint(dentry)) {
1422 dentry->d_flags |= DCACHE_MOUNTED;
1426 spin_unlock(&dentry->d_lock);
1428 write_sequnlock(&rename_lock);
1433 * Search the dentry child list of the specified parent,
1434 * and move any unused dentries to the end of the unused
1435 * list for prune_dcache(). We descend to the next level
1436 * whenever the d_subdirs list is non-empty and continue
1439 * It returns zero iff there are no unused children,
1440 * otherwise it returns the number of children moved to
1441 * the end of the unused list. This may not be the total
1442 * number of unused children, because select_parent can
1443 * drop the lock and return early due to latency
1447 struct select_data {
1448 struct dentry *start;
1449 struct list_head dispose;
1453 static enum d_walk_ret select_collect(void *_data, struct dentry *dentry)
1455 struct select_data *data = _data;
1456 enum d_walk_ret ret = D_WALK_CONTINUE;
1458 if (data->start == dentry)
1461 if (dentry->d_flags & DCACHE_SHRINK_LIST) {
1464 if (dentry->d_flags & DCACHE_LRU_LIST)
1466 if (!dentry->d_lockref.count) {
1467 d_shrink_add(dentry, &data->dispose);
1472 * We can return to the caller if we have found some (this
1473 * ensures forward progress). We'll be coming back to find
1476 if (!list_empty(&data->dispose))
1477 ret = need_resched() ? D_WALK_QUIT : D_WALK_NORETRY;
1483 * shrink_dcache_parent - prune dcache
1484 * @parent: parent of entries to prune
1486 * Prune the dcache to remove unused children of the parent dentry.
1488 void shrink_dcache_parent(struct dentry *parent)
1491 struct select_data data;
1493 INIT_LIST_HEAD(&data.dispose);
1494 data.start = parent;
1497 d_walk(parent, &data, select_collect);
1499 if (!list_empty(&data.dispose)) {
1500 shrink_dentry_list(&data.dispose);
1509 EXPORT_SYMBOL(shrink_dcache_parent);
1511 static enum d_walk_ret umount_check(void *_data, struct dentry *dentry)
1513 /* it has busy descendents; complain about those instead */
1514 if (!list_empty(&dentry->d_subdirs))
1515 return D_WALK_CONTINUE;
1517 /* root with refcount 1 is fine */
1518 if (dentry == _data && dentry->d_lockref.count == 1)
1519 return D_WALK_CONTINUE;
1521 printk(KERN_ERR "BUG: Dentry %p{i=%lx,n=%pd} "
1522 " still in use (%d) [unmount of %s %s]\n",
1525 dentry->d_inode->i_ino : 0UL,
1527 dentry->d_lockref.count,
1528 dentry->d_sb->s_type->name,
1529 dentry->d_sb->s_id);
1531 return D_WALK_CONTINUE;
1534 static void do_one_tree(struct dentry *dentry)
1536 shrink_dcache_parent(dentry);
1537 d_walk(dentry, dentry, umount_check);
1543 * destroy the dentries attached to a superblock on unmounting
1545 void shrink_dcache_for_umount(struct super_block *sb)
1547 struct dentry *dentry;
1549 WARN(down_read_trylock(&sb->s_umount), "s_umount should've been locked");
1551 dentry = sb->s_root;
1553 do_one_tree(dentry);
1555 while (!hlist_bl_empty(&sb->s_roots)) {
1556 dentry = dget(hlist_bl_entry(hlist_bl_first(&sb->s_roots), struct dentry, d_hash));
1557 do_one_tree(dentry);
1561 static enum d_walk_ret find_submount(void *_data, struct dentry *dentry)
1563 struct dentry **victim = _data;
1564 if (d_mountpoint(dentry)) {
1565 __dget_dlock(dentry);
1569 return D_WALK_CONTINUE;
1573 * d_invalidate - detach submounts, prune dcache, and drop
1574 * @dentry: dentry to invalidate (aka detach, prune and drop)
1576 void d_invalidate(struct dentry *dentry)
1578 bool had_submounts = false;
1579 spin_lock(&dentry->d_lock);
1580 if (d_unhashed(dentry)) {
1581 spin_unlock(&dentry->d_lock);
1585 spin_unlock(&dentry->d_lock);
1587 /* Negative dentries can be dropped without further checks */
1588 if (!dentry->d_inode)
1591 shrink_dcache_parent(dentry);
1593 struct dentry *victim = NULL;
1594 d_walk(dentry, &victim, find_submount);
1597 shrink_dcache_parent(dentry);
1600 had_submounts = true;
1601 detach_mounts(victim);
1605 EXPORT_SYMBOL(d_invalidate);
1608 * __d_alloc - allocate a dcache entry
1609 * @sb: filesystem it will belong to
1610 * @name: qstr of the name
1612 * Allocates a dentry. It returns %NULL if there is insufficient memory
1613 * available. On a success the dentry is returned. The name passed in is
1614 * copied and the copy passed in may be reused after this call.
1617 struct dentry *__d_alloc(struct super_block *sb, const struct qstr *name)
1619 struct dentry *dentry;
1623 dentry = kmem_cache_alloc(dentry_cache, GFP_KERNEL);
1628 * We guarantee that the inline name is always NUL-terminated.
1629 * This way the memcpy() done by the name switching in rename
1630 * will still always have a NUL at the end, even if we might
1631 * be overwriting an internal NUL character
1633 dentry->d_iname[DNAME_INLINE_LEN-1] = 0;
1634 if (unlikely(!name)) {
1636 dname = dentry->d_iname;
1637 } else if (name->len > DNAME_INLINE_LEN-1) {
1638 size_t size = offsetof(struct external_name, name[1]);
1639 struct external_name *p = kmalloc(size + name->len,
1640 GFP_KERNEL_ACCOUNT |
1643 kmem_cache_free(dentry_cache, dentry);
1646 atomic_set(&p->u.count, 1);
1649 dname = dentry->d_iname;
1652 dentry->d_name.len = name->len;
1653 dentry->d_name.hash = name->hash;
1654 memcpy(dname, name->name, name->len);
1655 dname[name->len] = 0;
1657 /* Make sure we always see the terminating NUL character */
1658 smp_store_release(&dentry->d_name.name, dname); /* ^^^ */
1660 dentry->d_lockref.count = 1;
1661 dentry->d_flags = 0;
1662 spin_lock_init(&dentry->d_lock);
1663 seqcount_init(&dentry->d_seq);
1664 dentry->d_inode = NULL;
1665 dentry->d_parent = dentry;
1667 dentry->d_op = NULL;
1668 dentry->d_fsdata = NULL;
1669 INIT_HLIST_BL_NODE(&dentry->d_hash);
1670 INIT_LIST_HEAD(&dentry->d_lru);
1671 INIT_LIST_HEAD(&dentry->d_subdirs);
1672 INIT_HLIST_NODE(&dentry->d_u.d_alias);
1673 INIT_LIST_HEAD(&dentry->d_child);
1674 d_set_d_op(dentry, dentry->d_sb->s_d_op);
1676 if (dentry->d_op && dentry->d_op->d_init) {
1677 err = dentry->d_op->d_init(dentry);
1679 if (dname_external(dentry))
1680 kfree(external_name(dentry));
1681 kmem_cache_free(dentry_cache, dentry);
1686 this_cpu_inc(nr_dentry);
1692 * d_alloc - allocate a dcache entry
1693 * @parent: parent of entry to allocate
1694 * @name: qstr of the name
1696 * Allocates a dentry. It returns %NULL if there is insufficient memory
1697 * available. On a success the dentry is returned. The name passed in is
1698 * copied and the copy passed in may be reused after this call.
1700 struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
1702 struct dentry *dentry = __d_alloc(parent->d_sb, name);
1705 dentry->d_flags |= DCACHE_RCUACCESS;
1706 spin_lock(&parent->d_lock);
1708 * don't need child lock because it is not subject
1709 * to concurrency here
1711 __dget_dlock(parent);
1712 dentry->d_parent = parent;
1713 list_add(&dentry->d_child, &parent->d_subdirs);
1714 spin_unlock(&parent->d_lock);
1718 EXPORT_SYMBOL(d_alloc);
1720 struct dentry *d_alloc_anon(struct super_block *sb)
1722 return __d_alloc(sb, NULL);
1724 EXPORT_SYMBOL(d_alloc_anon);
1726 struct dentry *d_alloc_cursor(struct dentry * parent)
1728 struct dentry *dentry = d_alloc_anon(parent->d_sb);
1730 dentry->d_flags |= DCACHE_RCUACCESS | DCACHE_DENTRY_CURSOR;
1731 dentry->d_parent = dget(parent);
1737 * d_alloc_pseudo - allocate a dentry (for lookup-less filesystems)
1738 * @sb: the superblock
1739 * @name: qstr of the name
1741 * For a filesystem that just pins its dentries in memory and never
1742 * performs lookups at all, return an unhashed IS_ROOT dentry.
1744 struct dentry *d_alloc_pseudo(struct super_block *sb, const struct qstr *name)
1746 return __d_alloc(sb, name);
1748 EXPORT_SYMBOL(d_alloc_pseudo);
1750 struct dentry *d_alloc_name(struct dentry *parent, const char *name)
1755 q.hash_len = hashlen_string(parent, name);
1756 return d_alloc(parent, &q);
1758 EXPORT_SYMBOL(d_alloc_name);
1760 void d_set_d_op(struct dentry *dentry, const struct dentry_operations *op)
1762 WARN_ON_ONCE(dentry->d_op);
1763 WARN_ON_ONCE(dentry->d_flags & (DCACHE_OP_HASH |
1765 DCACHE_OP_REVALIDATE |
1766 DCACHE_OP_WEAK_REVALIDATE |
1773 dentry->d_flags |= DCACHE_OP_HASH;
1775 dentry->d_flags |= DCACHE_OP_COMPARE;
1776 if (op->d_revalidate)
1777 dentry->d_flags |= DCACHE_OP_REVALIDATE;
1778 if (op->d_weak_revalidate)
1779 dentry->d_flags |= DCACHE_OP_WEAK_REVALIDATE;
1781 dentry->d_flags |= DCACHE_OP_DELETE;
1783 dentry->d_flags |= DCACHE_OP_PRUNE;
1785 dentry->d_flags |= DCACHE_OP_REAL;
1788 EXPORT_SYMBOL(d_set_d_op);
1792 * d_set_fallthru - Mark a dentry as falling through to a lower layer
1793 * @dentry - The dentry to mark
1795 * Mark a dentry as falling through to the lower layer (as set with
1796 * d_pin_lower()). This flag may be recorded on the medium.
1798 void d_set_fallthru(struct dentry *dentry)
1800 spin_lock(&dentry->d_lock);
1801 dentry->d_flags |= DCACHE_FALLTHRU;
1802 spin_unlock(&dentry->d_lock);
1804 EXPORT_SYMBOL(d_set_fallthru);
1806 static unsigned d_flags_for_inode(struct inode *inode)
1808 unsigned add_flags = DCACHE_REGULAR_TYPE;
1811 return DCACHE_MISS_TYPE;
1813 if (S_ISDIR(inode->i_mode)) {
1814 add_flags = DCACHE_DIRECTORY_TYPE;
1815 if (unlikely(!(inode->i_opflags & IOP_LOOKUP))) {
1816 if (unlikely(!inode->i_op->lookup))
1817 add_flags = DCACHE_AUTODIR_TYPE;
1819 inode->i_opflags |= IOP_LOOKUP;
1821 goto type_determined;
1824 if (unlikely(!(inode->i_opflags & IOP_NOFOLLOW))) {
1825 if (unlikely(inode->i_op->get_link)) {
1826 add_flags = DCACHE_SYMLINK_TYPE;
1827 goto type_determined;
1829 inode->i_opflags |= IOP_NOFOLLOW;
1832 if (unlikely(!S_ISREG(inode->i_mode)))
1833 add_flags = DCACHE_SPECIAL_TYPE;
1836 if (unlikely(IS_AUTOMOUNT(inode)))
1837 add_flags |= DCACHE_NEED_AUTOMOUNT;
1841 static void __d_instantiate(struct dentry *dentry, struct inode *inode)
1843 unsigned add_flags = d_flags_for_inode(inode);
1844 WARN_ON(d_in_lookup(dentry));
1846 spin_lock(&dentry->d_lock);
1848 * Decrement negative dentry count if it was in the LRU list.
1850 if (dentry->d_flags & DCACHE_LRU_LIST)
1851 this_cpu_dec(nr_dentry_negative);
1852 hlist_add_head(&dentry->d_u.d_alias, &inode->i_dentry);
1853 raw_write_seqcount_begin(&dentry->d_seq);
1854 __d_set_inode_and_type(dentry, inode, add_flags);
1855 raw_write_seqcount_end(&dentry->d_seq);
1856 fsnotify_update_flags(dentry);
1857 spin_unlock(&dentry->d_lock);
1861 * d_instantiate - fill in inode information for a dentry
1862 * @entry: dentry to complete
1863 * @inode: inode to attach to this dentry
1865 * Fill in inode information in the entry.
1867 * This turns negative dentries into productive full members
1870 * NOTE! This assumes that the inode count has been incremented
1871 * (or otherwise set) by the caller to indicate that it is now
1872 * in use by the dcache.
1875 void d_instantiate(struct dentry *entry, struct inode * inode)
1877 BUG_ON(!hlist_unhashed(&entry->d_u.d_alias));
1879 security_d_instantiate(entry, inode);
1880 spin_lock(&inode->i_lock);
1881 __d_instantiate(entry, inode);
1882 spin_unlock(&inode->i_lock);
1885 EXPORT_SYMBOL(d_instantiate);
1888 * This should be equivalent to d_instantiate() + unlock_new_inode(),
1889 * with lockdep-related part of unlock_new_inode() done before
1890 * anything else. Use that instead of open-coding d_instantiate()/
1891 * unlock_new_inode() combinations.
1893 void d_instantiate_new(struct dentry *entry, struct inode *inode)
1895 BUG_ON(!hlist_unhashed(&entry->d_u.d_alias));
1897 lockdep_annotate_inode_mutex_key(inode);
1898 security_d_instantiate(entry, inode);
1899 spin_lock(&inode->i_lock);
1900 __d_instantiate(entry, inode);
1901 WARN_ON(!(inode->i_state & I_NEW));
1902 inode->i_state &= ~I_NEW & ~I_CREATING;
1904 wake_up_bit(&inode->i_state, __I_NEW);
1905 spin_unlock(&inode->i_lock);
1907 EXPORT_SYMBOL(d_instantiate_new);
1909 struct dentry *d_make_root(struct inode *root_inode)
1911 struct dentry *res = NULL;
1914 res = d_alloc_anon(root_inode->i_sb);
1916 res->d_flags |= DCACHE_RCUACCESS;
1917 d_instantiate(res, root_inode);
1924 EXPORT_SYMBOL(d_make_root);
1926 static struct dentry *__d_instantiate_anon(struct dentry *dentry,
1927 struct inode *inode,
1933 security_d_instantiate(dentry, inode);
1934 spin_lock(&inode->i_lock);
1935 res = __d_find_any_alias(inode);
1937 spin_unlock(&inode->i_lock);
1942 /* attach a disconnected dentry */
1943 add_flags = d_flags_for_inode(inode);
1946 add_flags |= DCACHE_DISCONNECTED;
1948 spin_lock(&dentry->d_lock);
1949 __d_set_inode_and_type(dentry, inode, add_flags);
1950 hlist_add_head(&dentry->d_u.d_alias, &inode->i_dentry);
1951 if (!disconnected) {
1952 hlist_bl_lock(&dentry->d_sb->s_roots);
1953 hlist_bl_add_head(&dentry->d_hash, &dentry->d_sb->s_roots);
1954 hlist_bl_unlock(&dentry->d_sb->s_roots);
1956 spin_unlock(&dentry->d_lock);
1957 spin_unlock(&inode->i_lock);
1966 struct dentry *d_instantiate_anon(struct dentry *dentry, struct inode *inode)
1968 return __d_instantiate_anon(dentry, inode, true);
1970 EXPORT_SYMBOL(d_instantiate_anon);
1972 static struct dentry *__d_obtain_alias(struct inode *inode, bool disconnected)
1978 return ERR_PTR(-ESTALE);
1980 return ERR_CAST(inode);
1982 res = d_find_any_alias(inode);
1986 tmp = d_alloc_anon(inode->i_sb);
1988 res = ERR_PTR(-ENOMEM);
1992 return __d_instantiate_anon(tmp, inode, disconnected);
2000 * d_obtain_alias - find or allocate a DISCONNECTED dentry for a given inode
2001 * @inode: inode to allocate the dentry for
2003 * Obtain a dentry for an inode resulting from NFS filehandle conversion or
2004 * similar open by handle operations. The returned dentry may be anonymous,
2005 * or may have a full name (if the inode was already in the cache).
2007 * When called on a directory inode, we must ensure that the inode only ever
2008 * has one dentry. If a dentry is found, that is returned instead of
2009 * allocating a new one.
2011 * On successful return, the reference to the inode has been transferred
2012 * to the dentry. In case of an error the reference on the inode is released.
2013 * To make it easier to use in export operations a %NULL or IS_ERR inode may
2014 * be passed in and the error will be propagated to the return value,
2015 * with a %NULL @inode replaced by ERR_PTR(-ESTALE).
2017 struct dentry *d_obtain_alias(struct inode *inode)
2019 return __d_obtain_alias(inode, true);
2021 EXPORT_SYMBOL(d_obtain_alias);
2024 * d_obtain_root - find or allocate a dentry for a given inode
2025 * @inode: inode to allocate the dentry for
2027 * Obtain an IS_ROOT dentry for the root of a filesystem.
2029 * We must ensure that directory inodes only ever have one dentry. If a
2030 * dentry is found, that is returned instead of allocating a new one.
2032 * On successful return, the reference to the inode has been transferred
2033 * to the dentry. In case of an error the reference on the inode is
2034 * released. A %NULL or IS_ERR inode may be passed in and will be the
2035 * error will be propagate to the return value, with a %NULL @inode
2036 * replaced by ERR_PTR(-ESTALE).
2038 struct dentry *d_obtain_root(struct inode *inode)
2040 return __d_obtain_alias(inode, false);
2042 EXPORT_SYMBOL(d_obtain_root);
2045 * d_add_ci - lookup or allocate new dentry with case-exact name
2046 * @inode: the inode case-insensitive lookup has found
2047 * @dentry: the negative dentry that was passed to the parent's lookup func
2048 * @name: the case-exact name to be associated with the returned dentry
2050 * This is to avoid filling the dcache with case-insensitive names to the
2051 * same inode, only the actual correct case is stored in the dcache for
2052 * case-insensitive filesystems.
2054 * For a case-insensitive lookup match and if the the case-exact dentry
2055 * already exists in in the dcache, use it and return it.
2057 * If no entry exists with the exact case name, allocate new dentry with
2058 * the exact case, and return the spliced entry.
2060 struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
2063 struct dentry *found, *res;
2066 * First check if a dentry matching the name already exists,
2067 * if not go ahead and create it now.
2069 found = d_hash_and_lookup(dentry->d_parent, name);
2074 if (d_in_lookup(dentry)) {
2075 found = d_alloc_parallel(dentry->d_parent, name,
2077 if (IS_ERR(found) || !d_in_lookup(found)) {
2082 found = d_alloc(dentry->d_parent, name);
2085 return ERR_PTR(-ENOMEM);
2088 res = d_splice_alias(inode, found);
2095 EXPORT_SYMBOL(d_add_ci);
2098 static inline bool d_same_name(const struct dentry *dentry,
2099 const struct dentry *parent,
2100 const struct qstr *name)
2102 if (likely(!(parent->d_flags & DCACHE_OP_COMPARE))) {
2103 if (dentry->d_name.len != name->len)
2105 return dentry_cmp(dentry, name->name, name->len) == 0;
2107 return parent->d_op->d_compare(dentry,
2108 dentry->d_name.len, dentry->d_name.name,
2113 * __d_lookup_rcu - search for a dentry (racy, store-free)
2114 * @parent: parent dentry
2115 * @name: qstr of name we wish to find
2116 * @seqp: returns d_seq value at the point where the dentry was found
2117 * Returns: dentry, or NULL
2119 * __d_lookup_rcu is the dcache lookup function for rcu-walk name
2120 * resolution (store-free path walking) design described in
2121 * Documentation/filesystems/path-lookup.txt.
2123 * This is not to be used outside core vfs.
2125 * __d_lookup_rcu must only be used in rcu-walk mode, ie. with vfsmount lock
2126 * held, and rcu_read_lock held. The returned dentry must not be stored into
2127 * without taking d_lock and checking d_seq sequence count against @seq
2130 * A refcount may be taken on the found dentry with the d_rcu_to_refcount
2133 * Alternatively, __d_lookup_rcu may be called again to look up the child of
2134 * the returned dentry, so long as its parent's seqlock is checked after the
2135 * child is looked up. Thus, an interlocking stepping of sequence lock checks
2136 * is formed, giving integrity down the path walk.
2138 * NOTE! The caller *has* to check the resulting dentry against the sequence
2139 * number we've returned before using any of the resulting dentry state!
2141 struct dentry *__d_lookup_rcu(const struct dentry *parent,
2142 const struct qstr *name,
2145 u64 hashlen = name->hash_len;
2146 const unsigned char *str = name->name;
2147 struct hlist_bl_head *b = d_hash(hashlen_hash(hashlen));
2148 struct hlist_bl_node *node;
2149 struct dentry *dentry;
2152 * Note: There is significant duplication with __d_lookup_rcu which is
2153 * required to prevent single threaded performance regressions
2154 * especially on architectures where smp_rmb (in seqcounts) are costly.
2155 * Keep the two functions in sync.
2159 * The hash list is protected using RCU.
2161 * Carefully use d_seq when comparing a candidate dentry, to avoid
2162 * races with d_move().
2164 * It is possible that concurrent renames can mess up our list
2165 * walk here and result in missing our dentry, resulting in the
2166 * false-negative result. d_lookup() protects against concurrent
2167 * renames using rename_lock seqlock.
2169 * See Documentation/filesystems/path-lookup.txt for more details.
2171 hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
2176 * The dentry sequence count protects us from concurrent
2177 * renames, and thus protects parent and name fields.
2179 * The caller must perform a seqcount check in order
2180 * to do anything useful with the returned dentry.
2182 * NOTE! We do a "raw" seqcount_begin here. That means that
2183 * we don't wait for the sequence count to stabilize if it
2184 * is in the middle of a sequence change. If we do the slow
2185 * dentry compare, we will do seqretries until it is stable,
2186 * and if we end up with a successful lookup, we actually
2187 * want to exit RCU lookup anyway.
2189 * Note that raw_seqcount_begin still *does* smp_rmb(), so
2190 * we are still guaranteed NUL-termination of ->d_name.name.
2192 seq = raw_seqcount_begin(&dentry->d_seq);
2193 if (dentry->d_parent != parent)
2195 if (d_unhashed(dentry))
2198 if (unlikely(parent->d_flags & DCACHE_OP_COMPARE)) {
2201 if (dentry->d_name.hash != hashlen_hash(hashlen))
2203 tlen = dentry->d_name.len;
2204 tname = dentry->d_name.name;
2205 /* we want a consistent (name,len) pair */
2206 if (read_seqcount_retry(&dentry->d_seq, seq)) {
2210 if (parent->d_op->d_compare(dentry,
2211 tlen, tname, name) != 0)
2214 if (dentry->d_name.hash_len != hashlen)
2216 if (dentry_cmp(dentry, str, hashlen_len(hashlen)) != 0)
2226 * d_lookup - search for a dentry
2227 * @parent: parent dentry
2228 * @name: qstr of name we wish to find
2229 * Returns: dentry, or NULL
2231 * d_lookup searches the children of the parent dentry for the name in
2232 * question. If the dentry is found its reference count is incremented and the
2233 * dentry is returned. The caller must use dput to free the entry when it has
2234 * finished using it. %NULL is returned if the dentry does not exist.
2236 struct dentry *d_lookup(const struct dentry *parent, const struct qstr *name)
2238 struct dentry *dentry;
2242 seq = read_seqbegin(&rename_lock);
2243 dentry = __d_lookup(parent, name);
2246 } while (read_seqretry(&rename_lock, seq));
2249 EXPORT_SYMBOL(d_lookup);
2252 * __d_lookup - search for a dentry (racy)
2253 * @parent: parent dentry
2254 * @name: qstr of name we wish to find
2255 * Returns: dentry, or NULL
2257 * __d_lookup is like d_lookup, however it may (rarely) return a
2258 * false-negative result due to unrelated rename activity.
2260 * __d_lookup is slightly faster by avoiding rename_lock read seqlock,
2261 * however it must be used carefully, eg. with a following d_lookup in
2262 * the case of failure.
2264 * __d_lookup callers must be commented.
2266 struct dentry *__d_lookup(const struct dentry *parent, const struct qstr *name)
2268 unsigned int hash = name->hash;
2269 struct hlist_bl_head *b = d_hash(hash);
2270 struct hlist_bl_node *node;
2271 struct dentry *found = NULL;
2272 struct dentry *dentry;
2275 * Note: There is significant duplication with __d_lookup_rcu which is
2276 * required to prevent single threaded performance regressions
2277 * especially on architectures where smp_rmb (in seqcounts) are costly.
2278 * Keep the two functions in sync.
2282 * The hash list is protected using RCU.
2284 * Take d_lock when comparing a candidate dentry, to avoid races
2287 * It is possible that concurrent renames can mess up our list
2288 * walk here and result in missing our dentry, resulting in the
2289 * false-negative result. d_lookup() protects against concurrent
2290 * renames using rename_lock seqlock.
2292 * See Documentation/filesystems/path-lookup.txt for more details.
2296 hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
2298 if (dentry->d_name.hash != hash)
2301 spin_lock(&dentry->d_lock);
2302 if (dentry->d_parent != parent)
2304 if (d_unhashed(dentry))
2307 if (!d_same_name(dentry, parent, name))
2310 dentry->d_lockref.count++;
2312 spin_unlock(&dentry->d_lock);
2315 spin_unlock(&dentry->d_lock);
2323 * d_hash_and_lookup - hash the qstr then search for a dentry
2324 * @dir: Directory to search in
2325 * @name: qstr of name we wish to find
2327 * On lookup failure NULL is returned; on bad name - ERR_PTR(-error)
2329 struct dentry *d_hash_and_lookup(struct dentry *dir, struct qstr *name)
2332 * Check for a fs-specific hash function. Note that we must
2333 * calculate the standard hash first, as the d_op->d_hash()
2334 * routine may choose to leave the hash value unchanged.
2336 name->hash = full_name_hash(dir, name->name, name->len);
2337 if (dir->d_flags & DCACHE_OP_HASH) {
2338 int err = dir->d_op->d_hash(dir, name);
2339 if (unlikely(err < 0))
2340 return ERR_PTR(err);
2342 return d_lookup(dir, name);
2344 EXPORT_SYMBOL(d_hash_and_lookup);
2347 * When a file is deleted, we have two options:
2348 * - turn this dentry into a negative dentry
2349 * - unhash this dentry and free it.
2351 * Usually, we want to just turn this into
2352 * a negative dentry, but if anybody else is
2353 * currently using the dentry or the inode
2354 * we can't do that and we fall back on removing
2355 * it from the hash queues and waiting for
2356 * it to be deleted later when it has no users
2360 * d_delete - delete a dentry
2361 * @dentry: The dentry to delete
2363 * Turn the dentry into a negative dentry if possible, otherwise
2364 * remove it from the hash queues so it can be deleted later
2367 void d_delete(struct dentry * dentry)
2369 struct inode *inode = dentry->d_inode;
2370 int isdir = d_is_dir(dentry);
2372 spin_lock(&inode->i_lock);
2373 spin_lock(&dentry->d_lock);
2375 * Are we the only user?
2377 if (dentry->d_lockref.count == 1) {
2378 dentry->d_flags &= ~DCACHE_CANT_MOUNT;
2379 dentry_unlink_inode(dentry);
2382 spin_unlock(&dentry->d_lock);
2383 spin_unlock(&inode->i_lock);
2385 fsnotify_nameremove(dentry, isdir);
2387 EXPORT_SYMBOL(d_delete);
2389 static void __d_rehash(struct dentry *entry)
2391 struct hlist_bl_head *b = d_hash(entry->d_name.hash);
2394 hlist_bl_add_head_rcu(&entry->d_hash, b);
2399 * d_rehash - add an entry back to the hash
2400 * @entry: dentry to add to the hash
2402 * Adds a dentry to the hash according to its name.
2405 void d_rehash(struct dentry * entry)
2407 spin_lock(&entry->d_lock);
2409 spin_unlock(&entry->d_lock);
2411 EXPORT_SYMBOL(d_rehash);
2413 static inline unsigned start_dir_add(struct inode *dir)
2417 unsigned n = dir->i_dir_seq;
2418 if (!(n & 1) && cmpxchg(&dir->i_dir_seq, n, n + 1) == n)
2424 static inline void end_dir_add(struct inode *dir, unsigned n)
2426 smp_store_release(&dir->i_dir_seq, n + 2);
2429 static void d_wait_lookup(struct dentry *dentry)
2431 if (d_in_lookup(dentry)) {
2432 DECLARE_WAITQUEUE(wait, current);
2433 add_wait_queue(dentry->d_wait, &wait);
2435 set_current_state(TASK_UNINTERRUPTIBLE);
2436 spin_unlock(&dentry->d_lock);
2438 spin_lock(&dentry->d_lock);
2439 } while (d_in_lookup(dentry));
2443 struct dentry *d_alloc_parallel(struct dentry *parent,
2444 const struct qstr *name,
2445 wait_queue_head_t *wq)
2447 unsigned int hash = name->hash;
2448 struct hlist_bl_head *b = in_lookup_hash(parent, hash);
2449 struct hlist_bl_node *node;
2450 struct dentry *new = d_alloc(parent, name);
2451 struct dentry *dentry;
2452 unsigned seq, r_seq, d_seq;
2455 return ERR_PTR(-ENOMEM);
2459 seq = smp_load_acquire(&parent->d_inode->i_dir_seq);
2460 r_seq = read_seqbegin(&rename_lock);
2461 dentry = __d_lookup_rcu(parent, name, &d_seq);
2462 if (unlikely(dentry)) {
2463 if (!lockref_get_not_dead(&dentry->d_lockref)) {
2467 if (read_seqcount_retry(&dentry->d_seq, d_seq)) {
2476 if (unlikely(read_seqretry(&rename_lock, r_seq))) {
2481 if (unlikely(seq & 1)) {
2487 if (unlikely(READ_ONCE(parent->d_inode->i_dir_seq) != seq)) {
2493 * No changes for the parent since the beginning of d_lookup().
2494 * Since all removals from the chain happen with hlist_bl_lock(),
2495 * any potential in-lookup matches are going to stay here until
2496 * we unlock the chain. All fields are stable in everything
2499 hlist_bl_for_each_entry(dentry, node, b, d_u.d_in_lookup_hash) {
2500 if (dentry->d_name.hash != hash)
2502 if (dentry->d_parent != parent)
2504 if (!d_same_name(dentry, parent, name))
2507 /* now we can try to grab a reference */
2508 if (!lockref_get_not_dead(&dentry->d_lockref)) {
2515 * somebody is likely to be still doing lookup for it;
2516 * wait for them to finish
2518 spin_lock(&dentry->d_lock);
2519 d_wait_lookup(dentry);
2521 * it's not in-lookup anymore; in principle we should repeat
2522 * everything from dcache lookup, but it's likely to be what
2523 * d_lookup() would've found anyway. If it is, just return it;
2524 * otherwise we really have to repeat the whole thing.
2526 if (unlikely(dentry->d_name.hash != hash))
2528 if (unlikely(dentry->d_parent != parent))
2530 if (unlikely(d_unhashed(dentry)))
2532 if (unlikely(!d_same_name(dentry, parent, name)))
2534 /* OK, it *is* a hashed match; return it */
2535 spin_unlock(&dentry->d_lock);
2540 /* we can't take ->d_lock here; it's OK, though. */
2541 new->d_flags |= DCACHE_PAR_LOOKUP;
2543 hlist_bl_add_head_rcu(&new->d_u.d_in_lookup_hash, b);
2547 spin_unlock(&dentry->d_lock);
2551 EXPORT_SYMBOL(d_alloc_parallel);
2553 void __d_lookup_done(struct dentry *dentry)
2555 struct hlist_bl_head *b = in_lookup_hash(dentry->d_parent,
2556 dentry->d_name.hash);
2558 dentry->d_flags &= ~DCACHE_PAR_LOOKUP;
2559 __hlist_bl_del(&dentry->d_u.d_in_lookup_hash);
2560 wake_up_all(dentry->d_wait);
2561 dentry->d_wait = NULL;
2563 INIT_HLIST_NODE(&dentry->d_u.d_alias);
2564 INIT_LIST_HEAD(&dentry->d_lru);
2566 EXPORT_SYMBOL(__d_lookup_done);
2568 /* inode->i_lock held if inode is non-NULL */
2570 static inline void __d_add(struct dentry *dentry, struct inode *inode)
2572 struct inode *dir = NULL;
2574 spin_lock(&dentry->d_lock);
2575 if (unlikely(d_in_lookup(dentry))) {
2576 dir = dentry->d_parent->d_inode;
2577 n = start_dir_add(dir);
2578 __d_lookup_done(dentry);
2581 unsigned add_flags = d_flags_for_inode(inode);
2582 hlist_add_head(&dentry->d_u.d_alias, &inode->i_dentry);
2583 raw_write_seqcount_begin(&dentry->d_seq);
2584 __d_set_inode_and_type(dentry, inode, add_flags);
2585 raw_write_seqcount_end(&dentry->d_seq);
2586 fsnotify_update_flags(dentry);
2590 end_dir_add(dir, n);
2591 spin_unlock(&dentry->d_lock);
2593 spin_unlock(&inode->i_lock);
2597 * d_add - add dentry to hash queues
2598 * @entry: dentry to add
2599 * @inode: The inode to attach to this dentry
2601 * This adds the entry to the hash queues and initializes @inode.
2602 * The entry was actually filled in earlier during d_alloc().
2605 void d_add(struct dentry *entry, struct inode *inode)
2608 security_d_instantiate(entry, inode);
2609 spin_lock(&inode->i_lock);
2611 __d_add(entry, inode);
2613 EXPORT_SYMBOL(d_add);
2616 * d_exact_alias - find and hash an exact unhashed alias
2617 * @entry: dentry to add
2618 * @inode: The inode to go with this dentry
2620 * If an unhashed dentry with the same name/parent and desired
2621 * inode already exists, hash and return it. Otherwise, return
2624 * Parent directory should be locked.
2626 struct dentry *d_exact_alias(struct dentry *entry, struct inode *inode)
2628 struct dentry *alias;
2629 unsigned int hash = entry->d_name.hash;
2631 spin_lock(&inode->i_lock);
2632 hlist_for_each_entry(alias, &inode->i_dentry, d_u.d_alias) {
2634 * Don't need alias->d_lock here, because aliases with
2635 * d_parent == entry->d_parent are not subject to name or
2636 * parent changes, because the parent inode i_mutex is held.
2638 if (alias->d_name.hash != hash)
2640 if (alias->d_parent != entry->d_parent)
2642 if (!d_same_name(alias, entry->d_parent, &entry->d_name))
2644 spin_lock(&alias->d_lock);
2645 if (!d_unhashed(alias)) {
2646 spin_unlock(&alias->d_lock);
2649 __dget_dlock(alias);
2651 spin_unlock(&alias->d_lock);
2653 spin_unlock(&inode->i_lock);
2656 spin_unlock(&inode->i_lock);
2659 EXPORT_SYMBOL(d_exact_alias);
2661 static void swap_names(struct dentry *dentry, struct dentry *target)
2663 if (unlikely(dname_external(target))) {
2664 if (unlikely(dname_external(dentry))) {
2666 * Both external: swap the pointers
2668 swap(target->d_name.name, dentry->d_name.name);
2671 * dentry:internal, target:external. Steal target's
2672 * storage and make target internal.
2674 memcpy(target->d_iname, dentry->d_name.name,
2675 dentry->d_name.len + 1);
2676 dentry->d_name.name = target->d_name.name;
2677 target->d_name.name = target->d_iname;
2680 if (unlikely(dname_external(dentry))) {
2682 * dentry:external, target:internal. Give dentry's
2683 * storage to target and make dentry internal
2685 memcpy(dentry->d_iname, target->d_name.name,
2686 target->d_name.len + 1);
2687 target->d_name.name = dentry->d_name.name;
2688 dentry->d_name.name = dentry->d_iname;
2691 * Both are internal.
2694 BUILD_BUG_ON(!IS_ALIGNED(DNAME_INLINE_LEN, sizeof(long)));
2695 for (i = 0; i < DNAME_INLINE_LEN / sizeof(long); i++) {
2696 swap(((long *) &dentry->d_iname)[i],
2697 ((long *) &target->d_iname)[i]);
2701 swap(dentry->d_name.hash_len, target->d_name.hash_len);
2704 static void copy_name(struct dentry *dentry, struct dentry *target)
2706 struct external_name *old_name = NULL;
2707 if (unlikely(dname_external(dentry)))
2708 old_name = external_name(dentry);
2709 if (unlikely(dname_external(target))) {
2710 atomic_inc(&external_name(target)->u.count);
2711 dentry->d_name = target->d_name;
2713 memcpy(dentry->d_iname, target->d_name.name,
2714 target->d_name.len + 1);
2715 dentry->d_name.name = dentry->d_iname;
2716 dentry->d_name.hash_len = target->d_name.hash_len;
2718 if (old_name && likely(atomic_dec_and_test(&old_name->u.count)))
2719 kfree_rcu(old_name, u.head);
2723 * __d_move - move a dentry
2724 * @dentry: entry to move
2725 * @target: new dentry
2726 * @exchange: exchange the two dentries
2728 * Update the dcache to reflect the move of a file name. Negative
2729 * dcache entries should not be moved in this way. Caller must hold
2730 * rename_lock, the i_mutex of the source and target directories,
2731 * and the sb->s_vfs_rename_mutex if they differ. See lock_rename().
2733 static void __d_move(struct dentry *dentry, struct dentry *target,
2736 struct dentry *old_parent, *p;
2737 struct inode *dir = NULL;
2740 WARN_ON(!dentry->d_inode);
2741 if (WARN_ON(dentry == target))
2744 BUG_ON(d_ancestor(target, dentry));
2745 old_parent = dentry->d_parent;
2746 p = d_ancestor(old_parent, target);
2747 if (IS_ROOT(dentry)) {
2749 spin_lock(&target->d_parent->d_lock);
2751 /* target is not a descendent of dentry->d_parent */
2752 spin_lock(&target->d_parent->d_lock);
2753 spin_lock_nested(&old_parent->d_lock, DENTRY_D_LOCK_NESTED);
2755 BUG_ON(p == dentry);
2756 spin_lock(&old_parent->d_lock);
2758 spin_lock_nested(&target->d_parent->d_lock,
2759 DENTRY_D_LOCK_NESTED);
2761 spin_lock_nested(&dentry->d_lock, 2);
2762 spin_lock_nested(&target->d_lock, 3);
2764 if (unlikely(d_in_lookup(target))) {
2765 dir = target->d_parent->d_inode;
2766 n = start_dir_add(dir);
2767 __d_lookup_done(target);
2770 write_seqcount_begin(&dentry->d_seq);
2771 write_seqcount_begin_nested(&target->d_seq, DENTRY_D_LOCK_NESTED);
2774 if (!d_unhashed(dentry))
2776 if (!d_unhashed(target))
2779 /* ... and switch them in the tree */
2780 dentry->d_parent = target->d_parent;
2782 copy_name(dentry, target);
2783 target->d_hash.pprev = NULL;
2784 dentry->d_parent->d_lockref.count++;
2785 if (dentry == old_parent)
2786 dentry->d_flags |= DCACHE_RCUACCESS;
2788 WARN_ON(!--old_parent->d_lockref.count);
2790 target->d_parent = old_parent;
2791 swap_names(dentry, target);
2792 list_move(&target->d_child, &target->d_parent->d_subdirs);
2794 fsnotify_update_flags(target);
2796 list_move(&dentry->d_child, &dentry->d_parent->d_subdirs);
2798 fsnotify_update_flags(dentry);
2799 fscrypt_handle_d_move(dentry);
2801 write_seqcount_end(&target->d_seq);
2802 write_seqcount_end(&dentry->d_seq);
2805 end_dir_add(dir, n);
2807 if (dentry->d_parent != old_parent)
2808 spin_unlock(&dentry->d_parent->d_lock);
2809 if (dentry != old_parent)
2810 spin_unlock(&old_parent->d_lock);
2811 spin_unlock(&target->d_lock);
2812 spin_unlock(&dentry->d_lock);
2816 * d_move - move a dentry
2817 * @dentry: entry to move
2818 * @target: new dentry
2820 * Update the dcache to reflect the move of a file name. Negative
2821 * dcache entries should not be moved in this way. See the locking
2822 * requirements for __d_move.
2824 void d_move(struct dentry *dentry, struct dentry *target)
2826 write_seqlock(&rename_lock);
2827 __d_move(dentry, target, false);
2828 write_sequnlock(&rename_lock);
2830 EXPORT_SYMBOL(d_move);
2833 * d_exchange - exchange two dentries
2834 * @dentry1: first dentry
2835 * @dentry2: second dentry
2837 void d_exchange(struct dentry *dentry1, struct dentry *dentry2)
2839 write_seqlock(&rename_lock);
2841 WARN_ON(!dentry1->d_inode);
2842 WARN_ON(!dentry2->d_inode);
2843 WARN_ON(IS_ROOT(dentry1));
2844 WARN_ON(IS_ROOT(dentry2));
2846 __d_move(dentry1, dentry2, true);
2848 write_sequnlock(&rename_lock);
2852 * d_ancestor - search for an ancestor
2853 * @p1: ancestor dentry
2856 * Returns the ancestor dentry of p2 which is a child of p1, if p1 is
2857 * an ancestor of p2, else NULL.
2859 struct dentry *d_ancestor(struct dentry *p1, struct dentry *p2)
2863 for (p = p2; !IS_ROOT(p); p = p->d_parent) {
2864 if (p->d_parent == p1)
2871 * This helper attempts to cope with remotely renamed directories
2873 * It assumes that the caller is already holding
2874 * dentry->d_parent->d_inode->i_mutex, and rename_lock
2876 * Note: If ever the locking in lock_rename() changes, then please
2877 * remember to update this too...
2879 static int __d_unalias(struct inode *inode,
2880 struct dentry *dentry, struct dentry *alias)
2882 struct mutex *m1 = NULL;
2883 struct rw_semaphore *m2 = NULL;
2886 /* If alias and dentry share a parent, then no extra locks required */
2887 if (alias->d_parent == dentry->d_parent)
2890 /* See lock_rename() */
2891 if (!mutex_trylock(&dentry->d_sb->s_vfs_rename_mutex))
2893 m1 = &dentry->d_sb->s_vfs_rename_mutex;
2894 if (!inode_trylock_shared(alias->d_parent->d_inode))
2896 m2 = &alias->d_parent->d_inode->i_rwsem;
2898 __d_move(alias, dentry, false);
2909 * d_splice_alias - splice a disconnected dentry into the tree if one exists
2910 * @inode: the inode which may have a disconnected dentry
2911 * @dentry: a negative dentry which we want to point to the inode.
2913 * If inode is a directory and has an IS_ROOT alias, then d_move that in
2914 * place of the given dentry and return it, else simply d_add the inode
2915 * to the dentry and return NULL.
2917 * If a non-IS_ROOT directory is found, the filesystem is corrupt, and
2918 * we should error out: directories can't have multiple aliases.
2920 * This is needed in the lookup routine of any filesystem that is exportable
2921 * (via knfsd) so that we can build dcache paths to directories effectively.
2923 * If a dentry was found and moved, then it is returned. Otherwise NULL
2924 * is returned. This matches the expected return value of ->lookup.
2926 * Cluster filesystems may call this function with a negative, hashed dentry.
2927 * In that case, we know that the inode will be a regular file, and also this
2928 * will only occur during atomic_open. So we need to check for the dentry
2929 * being already hashed only in the final case.
2931 struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
2934 return ERR_CAST(inode);
2936 BUG_ON(!d_unhashed(dentry));
2941 security_d_instantiate(dentry, inode);
2942 spin_lock(&inode->i_lock);
2943 if (S_ISDIR(inode->i_mode)) {
2944 struct dentry *new = __d_find_any_alias(inode);
2945 if (unlikely(new)) {
2946 /* The reference to new ensures it remains an alias */
2947 spin_unlock(&inode->i_lock);
2948 write_seqlock(&rename_lock);
2949 if (unlikely(d_ancestor(new, dentry))) {
2950 write_sequnlock(&rename_lock);
2952 new = ERR_PTR(-ELOOP);
2953 pr_warn_ratelimited(
2954 "VFS: Lookup of '%s' in %s %s"
2955 " would have caused loop\n",
2956 dentry->d_name.name,
2957 inode->i_sb->s_type->name,
2959 } else if (!IS_ROOT(new)) {
2960 struct dentry *old_parent = dget(new->d_parent);
2961 int err = __d_unalias(inode, dentry, new);
2962 write_sequnlock(&rename_lock);
2969 __d_move(new, dentry, false);
2970 write_sequnlock(&rename_lock);
2977 __d_add(dentry, inode);
2980 EXPORT_SYMBOL(d_splice_alias);
2983 * Test whether new_dentry is a subdirectory of old_dentry.
2985 * Trivially implemented using the dcache structure
2989 * is_subdir - is new dentry a subdirectory of old_dentry
2990 * @new_dentry: new dentry
2991 * @old_dentry: old dentry
2993 * Returns true if new_dentry is a subdirectory of the parent (at any depth).
2994 * Returns false otherwise.
2995 * Caller must ensure that "new_dentry" is pinned before calling is_subdir()
2998 bool is_subdir(struct dentry *new_dentry, struct dentry *old_dentry)
3003 if (new_dentry == old_dentry)
3007 /* for restarting inner loop in case of seq retry */
3008 seq = read_seqbegin(&rename_lock);
3010 * Need rcu_readlock to protect against the d_parent trashing
3014 if (d_ancestor(old_dentry, new_dentry))
3019 } while (read_seqretry(&rename_lock, seq));
3023 EXPORT_SYMBOL(is_subdir);
3025 static enum d_walk_ret d_genocide_kill(void *data, struct dentry *dentry)
3027 struct dentry *root = data;
3028 if (dentry != root) {
3029 if (d_unhashed(dentry) || !dentry->d_inode)
3032 if (!(dentry->d_flags & DCACHE_GENOCIDE)) {
3033 dentry->d_flags |= DCACHE_GENOCIDE;
3034 dentry->d_lockref.count--;
3037 return D_WALK_CONTINUE;
3040 void d_genocide(struct dentry *parent)
3042 d_walk(parent, parent, d_genocide_kill);
3045 EXPORT_SYMBOL(d_genocide);
3047 void d_tmpfile(struct dentry *dentry, struct inode *inode)
3049 inode_dec_link_count(inode);
3050 BUG_ON(dentry->d_name.name != dentry->d_iname ||
3051 !hlist_unhashed(&dentry->d_u.d_alias) ||
3052 !d_unlinked(dentry));
3053 spin_lock(&dentry->d_parent->d_lock);
3054 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
3055 dentry->d_name.len = sprintf(dentry->d_iname, "#%llu",
3056 (unsigned long long)inode->i_ino);
3057 spin_unlock(&dentry->d_lock);
3058 spin_unlock(&dentry->d_parent->d_lock);
3059 d_instantiate(dentry, inode);
3061 EXPORT_SYMBOL(d_tmpfile);
3063 static __initdata unsigned long dhash_entries;
3064 static int __init set_dhash_entries(char *str)
3068 dhash_entries = simple_strtoul(str, &str, 0);
3071 __setup("dhash_entries=", set_dhash_entries);
3073 static void __init dcache_init_early(void)
3075 /* If hashes are distributed across NUMA nodes, defer
3076 * hash allocation until vmalloc space is available.
3082 alloc_large_system_hash("Dentry cache",
3083 sizeof(struct hlist_bl_head),
3086 HASH_EARLY | HASH_ZERO,
3091 d_hash_shift = 32 - d_hash_shift;
3094 static void __init dcache_init(void)
3097 * A constructor could be added for stable state like the lists,
3098 * but it is probably not worth it because of the cache nature
3101 dentry_cache = KMEM_CACHE_USERCOPY(dentry,
3102 SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|SLAB_MEM_SPREAD|SLAB_ACCOUNT,
3105 /* Hash may have been set up in dcache_init_early */
3110 alloc_large_system_hash("Dentry cache",
3111 sizeof(struct hlist_bl_head),
3119 d_hash_shift = 32 - d_hash_shift;
3122 /* SLAB cache for __getname() consumers */
3123 struct kmem_cache *names_cachep __read_mostly;
3124 EXPORT_SYMBOL(names_cachep);
3126 void __init vfs_caches_init_early(void)
3130 for (i = 0; i < ARRAY_SIZE(in_lookup_hashtable); i++)
3131 INIT_HLIST_BL_HEAD(&in_lookup_hashtable[i]);
3133 dcache_init_early();
3137 void __init vfs_caches_init(void)
3139 names_cachep = kmem_cache_create_usercopy("names_cache", PATH_MAX, 0,
3140 SLAB_HWCACHE_ALIGN|SLAB_PANIC, 0, PATH_MAX, NULL);
3145 files_maxfiles_init();