1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * fs/dcache.c
4 *
5 * Complete reimplementation
6 * (C) 1997 Thomas Schoebel-Theuer,
7 * with heavy changes by Linus Torvalds
8 */
9
10 /*
11 * Notes on the allocation strategy:
12 *
13 * The dcache is a master of the icache - whenever a dcache entry
14 * exists, the inode will always exist. "iput()" is done either when
15 * the dcache entry is deleted or garbage collected.
16 */
17
18 #include <linux/ratelimit.h>
19 #include <linux/string.h>
20 #include <linux/mm.h>
21 #include <linux/fs.h>
22 #include <linux/fscrypt.h>
23 #include <linux/fsnotify.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/hash.h>
27 #include <linux/cache.h>
28 #include <linux/export.h>
29 #include <linux/security.h>
30 #include <linux/seqlock.h>
31 #include <linux/memblock.h>
32 #include <linux/bit_spinlock.h>
33 #include <linux/rculist_bl.h>
34 #include <linux/list_lru.h>
35 #include "internal.h"
36 #include "mount.h"
37
38 /*
39 * Usage:
40 * dcache->d_inode->i_lock protects:
41 * - i_dentry, d_u.d_alias, d_inode of aliases
42 * dcache_hash_bucket lock protects:
43 * - the dcache hash table
44 * s_roots bl list spinlock protects:
45 * - the s_roots list (see __d_drop)
46 * dentry->d_sb->s_dentry_lru_lock protects:
47 * - the dcache lru lists and counters
48 * d_lock protects:
49 * - d_flags
50 * - d_name
51 * - d_lru
52 * - d_count
53 * - d_unhashed()
54 * - d_parent and d_subdirs
55 * - childrens' d_child and d_parent
56 * - d_u.d_alias, d_inode
57 *
58 * Ordering:
59 * dentry->d_inode->i_lock
60 * dentry->d_lock
61 * dentry->d_sb->s_dentry_lru_lock
62 * dcache_hash_bucket lock
63 * s_roots lock
64 *
65 * If there is an ancestor relationship:
66 * dentry->d_parent->...->d_parent->d_lock
67 * ...
68 * dentry->d_parent->d_lock
69 * dentry->d_lock
70 *
71 * If no ancestor relationship:
72 * arbitrary, since it's serialized on rename_lock
73 */
74 int sysctl_vfs_cache_pressure __read_mostly = 100;
75 EXPORT_SYMBOL_GPL(sysctl_vfs_cache_pressure);
76
77 __cacheline_aligned_in_smp DEFINE_SEQLOCK(rename_lock);
78
79 EXPORT_SYMBOL(rename_lock);
80
81 static struct kmem_cache *dentry_cache __read_mostly;
82
83 const struct qstr empty_name = QSTR_INIT("", 0);
84 EXPORT_SYMBOL(empty_name);
85 const struct qstr slash_name = QSTR_INIT("/", 1);
86 EXPORT_SYMBOL(slash_name);
87 const struct qstr dotdot_name = QSTR_INIT("..", 2);
88 EXPORT_SYMBOL(dotdot_name);
89
90 /*
91 * This is the single most critical data structure when it comes
92 * to the dcache: the hashtable for lookups. Somebody should try
93 * to make this good - I've just made it work.
94 *
95 * This hash-function tries to avoid losing too many bits of hash
96 * information, yet avoid using a prime hash-size or similar.
97 */
98
99 static unsigned int d_hash_shift __read_mostly;
100
101 static struct hlist_bl_head *dentry_hashtable __read_mostly;
102
d_hash(unsigned int hash)103 static inline struct hlist_bl_head *d_hash(unsigned int hash)
104 {
105 return dentry_hashtable + (hash >> d_hash_shift);
106 }
107
108 #define IN_LOOKUP_SHIFT 10
109 static struct hlist_bl_head in_lookup_hashtable[1 << IN_LOOKUP_SHIFT];
110
in_lookup_hash(const struct dentry * parent,unsigned int hash)111 static inline struct hlist_bl_head *in_lookup_hash(const struct dentry *parent,
112 unsigned int hash)
113 {
114 hash += (unsigned long) parent / L1_CACHE_BYTES;
115 return in_lookup_hashtable + hash_32(hash, IN_LOOKUP_SHIFT);
116 }
117
118 struct dentry_stat_t {
119 long nr_dentry;
120 long nr_unused;
121 long age_limit; /* age in seconds */
122 long want_pages; /* pages requested by system */
123 long nr_negative; /* # of unused negative dentries */
124 long dummy; /* Reserved for future use */
125 };
126
127 static DEFINE_PER_CPU(long, nr_dentry);
128 static DEFINE_PER_CPU(long, nr_dentry_unused);
129 static DEFINE_PER_CPU(long, nr_dentry_negative);
130
131 #if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
132 /* Statistics gathering. */
133 static struct dentry_stat_t dentry_stat = {
134 .age_limit = 45,
135 };
136
137 /*
138 * Here we resort to our own counters instead of using generic per-cpu counters
139 * for consistency with what the vfs inode code does. We are expected to harvest
140 * better code and performance by having our own specialized counters.
141 *
142 * Please note that the loop is done over all possible CPUs, not over all online
143 * CPUs. The reason for this is that we don't want to play games with CPUs going
144 * on and off. If one of them goes off, we will just keep their counters.
145 *
146 * glommer: See cffbc8a for details, and if you ever intend to change this,
147 * please update all vfs counters to match.
148 */
get_nr_dentry(void)149 static long get_nr_dentry(void)
150 {
151 int i;
152 long sum = 0;
153 for_each_possible_cpu(i)
154 sum += per_cpu(nr_dentry, i);
155 return sum < 0 ? 0 : sum;
156 }
157
get_nr_dentry_unused(void)158 static long get_nr_dentry_unused(void)
159 {
160 int i;
161 long sum = 0;
162 for_each_possible_cpu(i)
163 sum += per_cpu(nr_dentry_unused, i);
164 return sum < 0 ? 0 : sum;
165 }
166
get_nr_dentry_negative(void)167 static long get_nr_dentry_negative(void)
168 {
169 int i;
170 long sum = 0;
171
172 for_each_possible_cpu(i)
173 sum += per_cpu(nr_dentry_negative, i);
174 return sum < 0 ? 0 : sum;
175 }
176
proc_nr_dentry(struct ctl_table * table,int write,void * buffer,size_t * lenp,loff_t * ppos)177 static int proc_nr_dentry(struct ctl_table *table, int write, void *buffer,
178 size_t *lenp, loff_t *ppos)
179 {
180 dentry_stat.nr_dentry = get_nr_dentry();
181 dentry_stat.nr_unused = get_nr_dentry_unused();
182 dentry_stat.nr_negative = get_nr_dentry_negative();
183 return proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
184 }
185
186 static struct ctl_table fs_dcache_sysctls[] = {
187 {
188 .procname = "dentry-state",
189 .data = &dentry_stat,
190 .maxlen = 6*sizeof(long),
191 .mode = 0444,
192 .proc_handler = proc_nr_dentry,
193 },
194 { }
195 };
196
init_fs_dcache_sysctls(void)197 static int __init init_fs_dcache_sysctls(void)
198 {
199 register_sysctl_init("fs", fs_dcache_sysctls);
200 return 0;
201 }
202 fs_initcall(init_fs_dcache_sysctls);
203 #endif
204
205 /*
206 * Compare 2 name strings, return 0 if they match, otherwise non-zero.
207 * The strings are both count bytes long, and count is non-zero.
208 */
209 #ifdef CONFIG_DCACHE_WORD_ACCESS
210
211 #include <asm/word-at-a-time.h>
212 /*
213 * NOTE! 'cs' and 'scount' come from a dentry, so it has a
214 * aligned allocation for this particular component. We don't
215 * strictly need the load_unaligned_zeropad() safety, but it
216 * doesn't hurt either.
217 *
218 * In contrast, 'ct' and 'tcount' can be from a pathname, and do
219 * need the careful unaligned handling.
220 */
dentry_string_cmp(const unsigned char * cs,const unsigned char * ct,unsigned tcount)221 static inline int dentry_string_cmp(const unsigned char *cs, const unsigned char *ct, unsigned tcount)
222 {
223 unsigned long a,b,mask;
224
225 for (;;) {
226 a = read_word_at_a_time(cs);
227 b = load_unaligned_zeropad(ct);
228 if (tcount < sizeof(unsigned long))
229 break;
230 if (unlikely(a != b))
231 return 1;
232 cs += sizeof(unsigned long);
233 ct += sizeof(unsigned long);
234 tcount -= sizeof(unsigned long);
235 if (!tcount)
236 return 0;
237 }
238 mask = bytemask_from_count(tcount);
239 return unlikely(!!((a ^ b) & mask));
240 }
241
242 #else
243
dentry_string_cmp(const unsigned char * cs,const unsigned char * ct,unsigned tcount)244 static inline int dentry_string_cmp(const unsigned char *cs, const unsigned char *ct, unsigned tcount)
245 {
246 do {
247 if (*cs != *ct)
248 return 1;
249 cs++;
250 ct++;
251 tcount--;
252 } while (tcount);
253 return 0;
254 }
255
256 #endif
257
dentry_cmp(const struct dentry * dentry,const unsigned char * ct,unsigned tcount)258 static inline int dentry_cmp(const struct dentry *dentry, const unsigned char *ct, unsigned tcount)
259 {
260 /*
261 * Be careful about RCU walk racing with rename:
262 * use 'READ_ONCE' to fetch the name pointer.
263 *
264 * NOTE! Even if a rename will mean that the length
265 * was not loaded atomically, we don't care. The
266 * RCU walk will check the sequence count eventually,
267 * and catch it. And we won't overrun the buffer,
268 * because we're reading the name pointer atomically,
269 * and a dentry name is guaranteed to be properly
270 * terminated with a NUL byte.
271 *
272 * End result: even if 'len' is wrong, we'll exit
273 * early because the data cannot match (there can
274 * be no NUL in the ct/tcount data)
275 */
276 const unsigned char *cs = READ_ONCE(dentry->d_name.name);
277
278 return dentry_string_cmp(cs, ct, tcount);
279 }
280
281 struct external_name {
282 union {
283 atomic_t count;
284 struct rcu_head head;
285 } u;
286 unsigned char name[];
287 };
288
external_name(struct dentry * dentry)289 static inline struct external_name *external_name(struct dentry *dentry)
290 {
291 return container_of(dentry->d_name.name, struct external_name, name[0]);
292 }
293
__d_free(struct rcu_head * head)294 static void __d_free(struct rcu_head *head)
295 {
296 struct dentry *dentry = container_of(head, struct dentry, d_u.d_rcu);
297
298 kmem_cache_free(dentry_cache, dentry);
299 }
300
__d_free_external(struct rcu_head * head)301 static void __d_free_external(struct rcu_head *head)
302 {
303 struct dentry *dentry = container_of(head, struct dentry, d_u.d_rcu);
304 kfree(external_name(dentry));
305 kmem_cache_free(dentry_cache, dentry);
306 }
307
dname_external(const struct dentry * dentry)308 static inline int dname_external(const struct dentry *dentry)
309 {
310 return dentry->d_name.name != dentry->d_iname;
311 }
312
take_dentry_name_snapshot(struct name_snapshot * name,struct dentry * dentry)313 void take_dentry_name_snapshot(struct name_snapshot *name, struct dentry *dentry)
314 {
315 spin_lock(&dentry->d_lock);
316 name->name = dentry->d_name;
317 if (unlikely(dname_external(dentry))) {
318 atomic_inc(&external_name(dentry)->u.count);
319 } else {
320 memcpy(name->inline_name, dentry->d_iname,
321 dentry->d_name.len + 1);
322 name->name.name = name->inline_name;
323 }
324 spin_unlock(&dentry->d_lock);
325 }
326 EXPORT_SYMBOL(take_dentry_name_snapshot);
327
release_dentry_name_snapshot(struct name_snapshot * name)328 void release_dentry_name_snapshot(struct name_snapshot *name)
329 {
330 if (unlikely(name->name.name != name->inline_name)) {
331 struct external_name *p;
332 p = container_of(name->name.name, struct external_name, name[0]);
333 if (unlikely(atomic_dec_and_test(&p->u.count)))
334 kfree_rcu(p, u.head);
335 }
336 }
337 EXPORT_SYMBOL(release_dentry_name_snapshot);
338
__d_set_inode_and_type(struct dentry * dentry,struct inode * inode,unsigned type_flags)339 static inline void __d_set_inode_and_type(struct dentry *dentry,
340 struct inode *inode,
341 unsigned type_flags)
342 {
343 unsigned flags;
344
345 dentry->d_inode = inode;
346 flags = READ_ONCE(dentry->d_flags);
347 flags &= ~(DCACHE_ENTRY_TYPE | DCACHE_FALLTHRU);
348 flags |= type_flags;
349 smp_store_release(&dentry->d_flags, flags);
350 }
351
__d_clear_type_and_inode(struct dentry * dentry)352 static inline void __d_clear_type_and_inode(struct dentry *dentry)
353 {
354 unsigned flags = READ_ONCE(dentry->d_flags);
355
356 flags &= ~(DCACHE_ENTRY_TYPE | DCACHE_FALLTHRU);
357 WRITE_ONCE(dentry->d_flags, flags);
358 dentry->d_inode = NULL;
359 /*
360 * The negative counter only tracks dentries on the LRU. Don't inc if
361 * d_lru is on another list.
362 */
363 if ((flags & (DCACHE_LRU_LIST|DCACHE_SHRINK_LIST)) == DCACHE_LRU_LIST)
364 this_cpu_inc(nr_dentry_negative);
365 }
366
dentry_free(struct dentry * dentry)367 static void dentry_free(struct dentry *dentry)
368 {
369 WARN_ON(!hlist_unhashed(&dentry->d_u.d_alias));
370 if (unlikely(dname_external(dentry))) {
371 struct external_name *p = external_name(dentry);
372 if (likely(atomic_dec_and_test(&p->u.count))) {
373 call_rcu(&dentry->d_u.d_rcu, __d_free_external);
374 return;
375 }
376 }
377 /* if dentry was never visible to RCU, immediate free is OK */
378 if (dentry->d_flags & DCACHE_NORCU)
379 __d_free(&dentry->d_u.d_rcu);
380 else
381 call_rcu(&dentry->d_u.d_rcu, __d_free);
382 }
383
384 /*
385 * Release the dentry's inode, using the filesystem
386 * d_iput() operation if defined.
387 */
dentry_unlink_inode(struct dentry * dentry)388 static void dentry_unlink_inode(struct dentry * dentry)
389 __releases(dentry->d_lock)
390 __releases(dentry->d_inode->i_lock)
391 {
392 struct inode *inode = dentry->d_inode;
393
394 raw_write_seqcount_begin(&dentry->d_seq);
395 __d_clear_type_and_inode(dentry);
396 hlist_del_init(&dentry->d_u.d_alias);
397 raw_write_seqcount_end(&dentry->d_seq);
398 spin_unlock(&dentry->d_lock);
399 spin_unlock(&inode->i_lock);
400 if (!inode->i_nlink)
401 fsnotify_inoderemove(inode);
402 if (dentry->d_op && dentry->d_op->d_iput)
403 dentry->d_op->d_iput(dentry, inode);
404 else
405 iput(inode);
406 }
407
408 /*
409 * The DCACHE_LRU_LIST bit is set whenever the 'd_lru' entry
410 * is in use - which includes both the "real" per-superblock
411 * LRU list _and_ the DCACHE_SHRINK_LIST use.
412 *
413 * The DCACHE_SHRINK_LIST bit is set whenever the dentry is
414 * on the shrink list (ie not on the superblock LRU list).
415 *
416 * The per-cpu "nr_dentry_unused" counters are updated with
417 * the DCACHE_LRU_LIST bit.
418 *
419 * The per-cpu "nr_dentry_negative" counters are only updated
420 * when deleted from or added to the per-superblock LRU list, not
421 * from/to the shrink list. That is to avoid an unneeded dec/inc
422 * pair when moving from LRU to shrink list in select_collect().
423 *
424 * These helper functions make sure we always follow the
425 * rules. d_lock must be held by the caller.
426 */
427 #define D_FLAG_VERIFY(dentry,x) WARN_ON_ONCE(((dentry)->d_flags & (DCACHE_LRU_LIST | DCACHE_SHRINK_LIST)) != (x))
d_lru_add(struct dentry * dentry)428 static void d_lru_add(struct dentry *dentry)
429 {
430 D_FLAG_VERIFY(dentry, 0);
431 dentry->d_flags |= DCACHE_LRU_LIST;
432 this_cpu_inc(nr_dentry_unused);
433 if (d_is_negative(dentry))
434 this_cpu_inc(nr_dentry_negative);
435 WARN_ON_ONCE(!list_lru_add(&dentry->d_sb->s_dentry_lru, &dentry->d_lru));
436 }
437
d_lru_del(struct dentry * dentry)438 static void d_lru_del(struct dentry *dentry)
439 {
440 D_FLAG_VERIFY(dentry, DCACHE_LRU_LIST);
441 dentry->d_flags &= ~DCACHE_LRU_LIST;
442 this_cpu_dec(nr_dentry_unused);
443 if (d_is_negative(dentry))
444 this_cpu_dec(nr_dentry_negative);
445 WARN_ON_ONCE(!list_lru_del(&dentry->d_sb->s_dentry_lru, &dentry->d_lru));
446 }
447
d_shrink_del(struct dentry * dentry)448 static void d_shrink_del(struct dentry *dentry)
449 {
450 D_FLAG_VERIFY(dentry, DCACHE_SHRINK_LIST | DCACHE_LRU_LIST);
451 list_del_init(&dentry->d_lru);
452 dentry->d_flags &= ~(DCACHE_SHRINK_LIST | DCACHE_LRU_LIST);
453 this_cpu_dec(nr_dentry_unused);
454 }
455
d_shrink_add(struct dentry * dentry,struct list_head * list)456 static void d_shrink_add(struct dentry *dentry, struct list_head *list)
457 {
458 D_FLAG_VERIFY(dentry, 0);
459 list_add(&dentry->d_lru, list);
460 dentry->d_flags |= DCACHE_SHRINK_LIST | DCACHE_LRU_LIST;
461 this_cpu_inc(nr_dentry_unused);
462 }
463
464 /*
465 * These can only be called under the global LRU lock, ie during the
466 * callback for freeing the LRU list. "isolate" removes it from the
467 * LRU lists entirely, while shrink_move moves it to the indicated
468 * private list.
469 */
d_lru_isolate(struct list_lru_one * lru,struct dentry * dentry)470 static void d_lru_isolate(struct list_lru_one *lru, struct dentry *dentry)
471 {
472 D_FLAG_VERIFY(dentry, DCACHE_LRU_LIST);
473 dentry->d_flags &= ~DCACHE_LRU_LIST;
474 this_cpu_dec(nr_dentry_unused);
475 if (d_is_negative(dentry))
476 this_cpu_dec(nr_dentry_negative);
477 list_lru_isolate(lru, &dentry->d_lru);
478 }
479
d_lru_shrink_move(struct list_lru_one * lru,struct dentry * dentry,struct list_head * list)480 static void d_lru_shrink_move(struct list_lru_one *lru, struct dentry *dentry,
481 struct list_head *list)
482 {
483 D_FLAG_VERIFY(dentry, DCACHE_LRU_LIST);
484 dentry->d_flags |= DCACHE_SHRINK_LIST;
485 if (d_is_negative(dentry))
486 this_cpu_dec(nr_dentry_negative);
487 list_lru_isolate_move(lru, &dentry->d_lru, list);
488 }
489
___d_drop(struct dentry * dentry)490 static void ___d_drop(struct dentry *dentry)
491 {
492 struct hlist_bl_head *b;
493 /*
494 * Hashed dentries are normally on the dentry hashtable,
495 * with the exception of those newly allocated by
496 * d_obtain_root, which are always IS_ROOT:
497 */
498 if (unlikely(IS_ROOT(dentry)))
499 b = &dentry->d_sb->s_roots;
500 else
501 b = d_hash(dentry->d_name.hash);
502
503 hlist_bl_lock(b);
504 __hlist_bl_del(&dentry->d_hash);
505 hlist_bl_unlock(b);
506 }
507
__d_drop(struct dentry * dentry)508 void __d_drop(struct dentry *dentry)
509 {
510 if (!d_unhashed(dentry)) {
511 ___d_drop(dentry);
512 dentry->d_hash.pprev = NULL;
513 write_seqcount_invalidate(&dentry->d_seq);
514 }
515 }
516 EXPORT_SYMBOL(__d_drop);
517
518 /**
519 * d_drop - drop a dentry
520 * @dentry: dentry to drop
521 *
522 * d_drop() unhashes the entry from the parent dentry hashes, so that it won't
523 * be found through a VFS lookup any more. Note that this is different from
524 * deleting the dentry - d_delete will try to mark the dentry negative if
525 * possible, giving a successful _negative_ lookup, while d_drop will
526 * just make the cache lookup fail.
527 *
528 * d_drop() is used mainly for stuff that wants to invalidate a dentry for some
529 * reason (NFS timeouts or autofs deletes).
530 *
531 * __d_drop requires dentry->d_lock
532 *
533 * ___d_drop doesn't mark dentry as "unhashed"
534 * (dentry->d_hash.pprev will be LIST_POISON2, not NULL).
535 */
d_drop(struct dentry * dentry)536 void d_drop(struct dentry *dentry)
537 {
538 spin_lock(&dentry->d_lock);
539 __d_drop(dentry);
540 spin_unlock(&dentry->d_lock);
541 }
542 EXPORT_SYMBOL(d_drop);
543
dentry_unlist(struct dentry * dentry,struct dentry * parent)544 static inline void dentry_unlist(struct dentry *dentry, struct dentry *parent)
545 {
546 struct dentry *next;
547 /*
548 * Inform d_walk() and shrink_dentry_list() that we are no longer
549 * attached to the dentry tree
550 */
551 dentry->d_flags |= DCACHE_DENTRY_KILLED;
552 if (unlikely(list_empty(&dentry->d_child)))
553 return;
554 __list_del_entry(&dentry->d_child);
555 /*
556 * Cursors can move around the list of children. While we'd been
557 * a normal list member, it didn't matter - ->d_child.next would've
558 * been updated. However, from now on it won't be and for the
559 * things like d_walk() it might end up with a nasty surprise.
560 * Normally d_walk() doesn't care about cursors moving around -
561 * ->d_lock on parent prevents that and since a cursor has no children
562 * of its own, we get through it without ever unlocking the parent.
563 * There is one exception, though - if we ascend from a child that
564 * gets killed as soon as we unlock it, the next sibling is found
565 * using the value left in its ->d_child.next. And if _that_
566 * pointed to a cursor, and cursor got moved (e.g. by lseek())
567 * before d_walk() regains parent->d_lock, we'll end up skipping
568 * everything the cursor had been moved past.
569 *
570 * Solution: make sure that the pointer left behind in ->d_child.next
571 * points to something that won't be moving around. I.e. skip the
572 * cursors.
573 */
574 while (dentry->d_child.next != &parent->d_subdirs) {
575 next = list_entry(dentry->d_child.next, struct dentry, d_child);
576 if (likely(!(next->d_flags & DCACHE_DENTRY_CURSOR)))
577 break;
578 dentry->d_child.next = next->d_child.next;
579 }
580 }
581
__dentry_kill(struct dentry * dentry)582 static void __dentry_kill(struct dentry *dentry)
583 {
584 struct dentry *parent = NULL;
585 bool can_free = true;
586 if (!IS_ROOT(dentry))
587 parent = dentry->d_parent;
588
589 /*
590 * The dentry is now unrecoverably dead to the world.
591 */
592 lockref_mark_dead(&dentry->d_lockref);
593
594 /*
595 * inform the fs via d_prune that this dentry is about to be
596 * unhashed and destroyed.
597 */
598 if (dentry->d_flags & DCACHE_OP_PRUNE)
599 dentry->d_op->d_prune(dentry);
600
601 if (dentry->d_flags & DCACHE_LRU_LIST) {
602 if (!(dentry->d_flags & DCACHE_SHRINK_LIST))
603 d_lru_del(dentry);
604 }
605 /* if it was on the hash then remove it */
606 __d_drop(dentry);
607 dentry_unlist(dentry, parent);
608 if (parent)
609 spin_unlock(&parent->d_lock);
610 if (dentry->d_inode)
611 dentry_unlink_inode(dentry);
612 else
613 spin_unlock(&dentry->d_lock);
614 this_cpu_dec(nr_dentry);
615 if (dentry->d_op && dentry->d_op->d_release)
616 dentry->d_op->d_release(dentry);
617
618 spin_lock(&dentry->d_lock);
619 if (dentry->d_flags & DCACHE_SHRINK_LIST) {
620 dentry->d_flags |= DCACHE_MAY_FREE;
621 can_free = false;
622 }
623 spin_unlock(&dentry->d_lock);
624 if (likely(can_free))
625 dentry_free(dentry);
626 cond_resched();
627 }
628
__lock_parent(struct dentry * dentry)629 static struct dentry *__lock_parent(struct dentry *dentry)
630 {
631 struct dentry *parent;
632 rcu_read_lock();
633 spin_unlock(&dentry->d_lock);
634 again:
635 parent = READ_ONCE(dentry->d_parent);
636 spin_lock(&parent->d_lock);
637 /*
638 * We can't blindly lock dentry until we are sure
639 * that we won't violate the locking order.
640 * Any changes of dentry->d_parent must have
641 * been done with parent->d_lock held, so
642 * spin_lock() above is enough of a barrier
643 * for checking if it's still our child.
644 */
645 if (unlikely(parent != dentry->d_parent)) {
646 spin_unlock(&parent->d_lock);
647 goto again;
648 }
649 rcu_read_unlock();
650 if (parent != dentry)
651 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
652 else
653 parent = NULL;
654 return parent;
655 }
656
lock_parent(struct dentry * dentry)657 static inline struct dentry *lock_parent(struct dentry *dentry)
658 {
659 struct dentry *parent = dentry->d_parent;
660 if (IS_ROOT(dentry))
661 return NULL;
662 if (likely(spin_trylock(&parent->d_lock)))
663 return parent;
664 return __lock_parent(dentry);
665 }
666
retain_dentry(struct dentry * dentry)667 static inline bool retain_dentry(struct dentry *dentry)
668 {
669 WARN_ON(d_in_lookup(dentry));
670
671 /* Unreachable? Get rid of it */
672 if (unlikely(d_unhashed(dentry)))
673 return false;
674
675 if (unlikely(dentry->d_flags & DCACHE_DISCONNECTED))
676 return false;
677
678 if (unlikely(dentry->d_flags & DCACHE_OP_DELETE)) {
679 if (dentry->d_op->d_delete(dentry))
680 return false;
681 }
682
683 if (unlikely(dentry->d_flags & DCACHE_DONTCACHE))
684 return false;
685
686 /* retain; LRU fodder */
687 dentry->d_lockref.count--;
688 if (unlikely(!(dentry->d_flags & DCACHE_LRU_LIST)))
689 d_lru_add(dentry);
690 else if (unlikely(!(dentry->d_flags & DCACHE_REFERENCED)))
691 dentry->d_flags |= DCACHE_REFERENCED;
692 return true;
693 }
694
d_mark_dontcache(struct inode * inode)695 void d_mark_dontcache(struct inode *inode)
696 {
697 struct dentry *de;
698
699 spin_lock(&inode->i_lock);
700 hlist_for_each_entry(de, &inode->i_dentry, d_u.d_alias) {
701 spin_lock(&de->d_lock);
702 de->d_flags |= DCACHE_DONTCACHE;
703 spin_unlock(&de->d_lock);
704 }
705 inode->i_state |= I_DONTCACHE;
706 spin_unlock(&inode->i_lock);
707 }
708 EXPORT_SYMBOL(d_mark_dontcache);
709
710 /*
711 * Finish off a dentry we've decided to kill.
712 * dentry->d_lock must be held, returns with it unlocked.
713 * Returns dentry requiring refcount drop, or NULL if we're done.
714 */
dentry_kill(struct dentry * dentry)715 static struct dentry *dentry_kill(struct dentry *dentry)
716 __releases(dentry->d_lock)
717 {
718 struct inode *inode = dentry->d_inode;
719 struct dentry *parent = NULL;
720
721 if (inode && unlikely(!spin_trylock(&inode->i_lock)))
722 goto slow_positive;
723
724 if (!IS_ROOT(dentry)) {
725 parent = dentry->d_parent;
726 if (unlikely(!spin_trylock(&parent->d_lock))) {
727 parent = __lock_parent(dentry);
728 if (likely(inode || !dentry->d_inode))
729 goto got_locks;
730 /* negative that became positive */
731 if (parent)
732 spin_unlock(&parent->d_lock);
733 inode = dentry->d_inode;
734 goto slow_positive;
735 }
736 }
737 __dentry_kill(dentry);
738 return parent;
739
740 slow_positive:
741 spin_unlock(&dentry->d_lock);
742 spin_lock(&inode->i_lock);
743 spin_lock(&dentry->d_lock);
744 parent = lock_parent(dentry);
745 got_locks:
746 if (unlikely(dentry->d_lockref.count != 1)) {
747 dentry->d_lockref.count--;
748 } else if (likely(!retain_dentry(dentry))) {
749 __dentry_kill(dentry);
750 return parent;
751 }
752 /* we are keeping it, after all */
753 if (inode)
754 spin_unlock(&inode->i_lock);
755 if (parent)
756 spin_unlock(&parent->d_lock);
757 spin_unlock(&dentry->d_lock);
758 return NULL;
759 }
760
761 /*
762 * Try to do a lockless dput(), and return whether that was successful.
763 *
764 * If unsuccessful, we return false, having already taken the dentry lock.
765 *
766 * The caller needs to hold the RCU read lock, so that the dentry is
767 * guaranteed to stay around even if the refcount goes down to zero!
768 */
fast_dput(struct dentry * dentry)769 static inline bool fast_dput(struct dentry *dentry)
770 {
771 int ret;
772 unsigned int d_flags;
773
774 /*
775 * If we have a d_op->d_delete() operation, we sould not
776 * let the dentry count go to zero, so use "put_or_lock".
777 */
778 if (unlikely(dentry->d_flags & DCACHE_OP_DELETE))
779 return lockref_put_or_lock(&dentry->d_lockref);
780
781 /*
782 * .. otherwise, we can try to just decrement the
783 * lockref optimistically.
784 */
785 ret = lockref_put_return(&dentry->d_lockref);
786
787 /*
788 * If the lockref_put_return() failed due to the lock being held
789 * by somebody else, the fast path has failed. We will need to
790 * get the lock, and then check the count again.
791 */
792 if (unlikely(ret < 0)) {
793 spin_lock(&dentry->d_lock);
794 if (WARN_ON_ONCE(dentry->d_lockref.count <= 0)) {
795 spin_unlock(&dentry->d_lock);
796 return true;
797 }
798 dentry->d_lockref.count--;
799 goto locked;
800 }
801
802 /*
803 * If we weren't the last ref, we're done.
804 */
805 if (ret)
806 return true;
807
808 /*
809 * Careful, careful. The reference count went down
810 * to zero, but we don't hold the dentry lock, so
811 * somebody else could get it again, and do another
812 * dput(), and we need to not race with that.
813 *
814 * However, there is a very special and common case
815 * where we don't care, because there is nothing to
816 * do: the dentry is still hashed, it does not have
817 * a 'delete' op, and it's referenced and already on
818 * the LRU list.
819 *
820 * NOTE! Since we aren't locked, these values are
821 * not "stable". However, it is sufficient that at
822 * some point after we dropped the reference the
823 * dentry was hashed and the flags had the proper
824 * value. Other dentry users may have re-gotten
825 * a reference to the dentry and change that, but
826 * our work is done - we can leave the dentry
827 * around with a zero refcount.
828 *
829 * Nevertheless, there are two cases that we should kill
830 * the dentry anyway.
831 * 1. free disconnected dentries as soon as their refcount
832 * reached zero.
833 * 2. free dentries if they should not be cached.
834 */
835 smp_rmb();
836 d_flags = READ_ONCE(dentry->d_flags);
837 d_flags &= DCACHE_REFERENCED | DCACHE_LRU_LIST |
838 DCACHE_DISCONNECTED | DCACHE_DONTCACHE;
839
840 /* Nothing to do? Dropping the reference was all we needed? */
841 if (d_flags == (DCACHE_REFERENCED | DCACHE_LRU_LIST) && !d_unhashed(dentry))
842 return true;
843
844 /*
845 * Not the fast normal case? Get the lock. We've already decremented
846 * the refcount, but we'll need to re-check the situation after
847 * getting the lock.
848 */
849 spin_lock(&dentry->d_lock);
850
851 /*
852 * Did somebody else grab a reference to it in the meantime, and
853 * we're no longer the last user after all? Alternatively, somebody
854 * else could have killed it and marked it dead. Either way, we
855 * don't need to do anything else.
856 */
857 locked:
858 if (dentry->d_lockref.count) {
859 spin_unlock(&dentry->d_lock);
860 return true;
861 }
862
863 /*
864 * Re-get the reference we optimistically dropped. We hold the
865 * lock, and we just tested that it was zero, so we can just
866 * set it to 1.
867 */
868 dentry->d_lockref.count = 1;
869 return false;
870 }
871
872
873 /*
874 * This is dput
875 *
876 * This is complicated by the fact that we do not want to put
877 * dentries that are no longer on any hash chain on the unused
878 * list: we'd much rather just get rid of them immediately.
879 *
880 * However, that implies that we have to traverse the dentry
881 * tree upwards to the parents which might _also_ now be
882 * scheduled for deletion (it may have been only waiting for
883 * its last child to go away).
884 *
885 * This tail recursion is done by hand as we don't want to depend
886 * on the compiler to always get this right (gcc generally doesn't).
887 * Real recursion would eat up our stack space.
888 */
889
890 /*
891 * dput - release a dentry
892 * @dentry: dentry to release
893 *
894 * Release a dentry. This will drop the usage count and if appropriate
895 * call the dentry unlink method as well as removing it from the queues and
896 * releasing its resources. If the parent dentries were scheduled for release
897 * they too may now get deleted.
898 */
dput(struct dentry * dentry)899 void dput(struct dentry *dentry)
900 {
901 while (dentry) {
902 might_sleep();
903
904 rcu_read_lock();
905 if (likely(fast_dput(dentry))) {
906 rcu_read_unlock();
907 return;
908 }
909
910 /* Slow case: now with the dentry lock held */
911 rcu_read_unlock();
912
913 if (likely(retain_dentry(dentry))) {
914 spin_unlock(&dentry->d_lock);
915 return;
916 }
917
918 dentry = dentry_kill(dentry);
919 }
920 }
921 EXPORT_SYMBOL(dput);
922
__dput_to_list(struct dentry * dentry,struct list_head * list)923 static void __dput_to_list(struct dentry *dentry, struct list_head *list)
924 __must_hold(&dentry->d_lock)
925 {
926 if (dentry->d_flags & DCACHE_SHRINK_LIST) {
927 /* let the owner of the list it's on deal with it */
928 --dentry->d_lockref.count;
929 } else {
930 if (dentry->d_flags & DCACHE_LRU_LIST)
931 d_lru_del(dentry);
932 if (!--dentry->d_lockref.count)
933 d_shrink_add(dentry, list);
934 }
935 }
936
dput_to_list(struct dentry * dentry,struct list_head * list)937 void dput_to_list(struct dentry *dentry, struct list_head *list)
938 {
939 rcu_read_lock();
940 if (likely(fast_dput(dentry))) {
941 rcu_read_unlock();
942 return;
943 }
944 rcu_read_unlock();
945 if (!retain_dentry(dentry))
946 __dput_to_list(dentry, list);
947 spin_unlock(&dentry->d_lock);
948 }
949
950 /* This must be called with d_lock held */
__dget_dlock(struct dentry * dentry)951 static inline void __dget_dlock(struct dentry *dentry)
952 {
953 dentry->d_lockref.count++;
954 }
955
__dget(struct dentry * dentry)956 static inline void __dget(struct dentry *dentry)
957 {
958 lockref_get(&dentry->d_lockref);
959 }
960
dget_parent(struct dentry * dentry)961 struct dentry *dget_parent(struct dentry *dentry)
962 {
963 int gotref;
964 struct dentry *ret;
965 unsigned seq;
966
967 /*
968 * Do optimistic parent lookup without any
969 * locking.
970 */
971 rcu_read_lock();
972 seq = raw_seqcount_begin(&dentry->d_seq);
973 ret = READ_ONCE(dentry->d_parent);
974 gotref = lockref_get_not_zero(&ret->d_lockref);
975 rcu_read_unlock();
976 if (likely(gotref)) {
977 if (!read_seqcount_retry(&dentry->d_seq, seq))
978 return ret;
979 dput(ret);
980 }
981
982 repeat:
983 /*
984 * Don't need rcu_dereference because we re-check it was correct under
985 * the lock.
986 */
987 rcu_read_lock();
988 ret = dentry->d_parent;
989 spin_lock(&ret->d_lock);
990 if (unlikely(ret != dentry->d_parent)) {
991 spin_unlock(&ret->d_lock);
992 rcu_read_unlock();
993 goto repeat;
994 }
995 rcu_read_unlock();
996 BUG_ON(!ret->d_lockref.count);
997 ret->d_lockref.count++;
998 spin_unlock(&ret->d_lock);
999 return ret;
1000 }
1001 EXPORT_SYMBOL(dget_parent);
1002
__d_find_any_alias(struct inode * inode)1003 static struct dentry * __d_find_any_alias(struct inode *inode)
1004 {
1005 struct dentry *alias;
1006
1007 if (hlist_empty(&inode->i_dentry))
1008 return NULL;
1009 alias = hlist_entry(inode->i_dentry.first, struct dentry, d_u.d_alias);
1010 __dget(alias);
1011 return alias;
1012 }
1013
1014 /**
1015 * d_find_any_alias - find any alias for a given inode
1016 * @inode: inode to find an alias for
1017 *
1018 * If any aliases exist for the given inode, take and return a
1019 * reference for one of them. If no aliases exist, return %NULL.
1020 */
d_find_any_alias(struct inode * inode)1021 struct dentry *d_find_any_alias(struct inode *inode)
1022 {
1023 struct dentry *de;
1024
1025 spin_lock(&inode->i_lock);
1026 de = __d_find_any_alias(inode);
1027 spin_unlock(&inode->i_lock);
1028 return de;
1029 }
1030 EXPORT_SYMBOL(d_find_any_alias);
1031
__d_find_alias(struct inode * inode)1032 static struct dentry *__d_find_alias(struct inode *inode)
1033 {
1034 struct dentry *alias;
1035
1036 if (S_ISDIR(inode->i_mode))
1037 return __d_find_any_alias(inode);
1038
1039 hlist_for_each_entry(alias, &inode->i_dentry, d_u.d_alias) {
1040 spin_lock(&alias->d_lock);
1041 if (!d_unhashed(alias)) {
1042 __dget_dlock(alias);
1043 spin_unlock(&alias->d_lock);
1044 return alias;
1045 }
1046 spin_unlock(&alias->d_lock);
1047 }
1048 return NULL;
1049 }
1050
1051 /**
1052 * d_find_alias - grab a hashed alias of inode
1053 * @inode: inode in question
1054 *
1055 * If inode has a hashed alias, or is a directory and has any alias,
1056 * acquire the reference to alias and return it. Otherwise return NULL.
1057 * Notice that if inode is a directory there can be only one alias and
1058 * it can be unhashed only if it has no children, or if it is the root
1059 * of a filesystem, or if the directory was renamed and d_revalidate
1060 * was the first vfs operation to notice.
1061 *
1062 * If the inode has an IS_ROOT, DCACHE_DISCONNECTED alias, then prefer
1063 * any other hashed alias over that one.
1064 */
d_find_alias(struct inode * inode)1065 struct dentry *d_find_alias(struct inode *inode)
1066 {
1067 struct dentry *de = NULL;
1068
1069 if (!hlist_empty(&inode->i_dentry)) {
1070 spin_lock(&inode->i_lock);
1071 de = __d_find_alias(inode);
1072 spin_unlock(&inode->i_lock);
1073 }
1074 return de;
1075 }
1076 EXPORT_SYMBOL(d_find_alias);
1077
1078 /*
1079 * Caller MUST be holding rcu_read_lock() and be guaranteed
1080 * that inode won't get freed until rcu_read_unlock().
1081 */
d_find_alias_rcu(struct inode * inode)1082 struct dentry *d_find_alias_rcu(struct inode *inode)
1083 {
1084 struct hlist_head *l = &inode->i_dentry;
1085 struct dentry *de = NULL;
1086
1087 spin_lock(&inode->i_lock);
1088 // ->i_dentry and ->i_rcu are colocated, but the latter won't be
1089 // used without having I_FREEING set, which means no aliases left
1090 if (likely(!(inode->i_state & I_FREEING) && !hlist_empty(l))) {
1091 if (S_ISDIR(inode->i_mode)) {
1092 de = hlist_entry(l->first, struct dentry, d_u.d_alias);
1093 } else {
1094 hlist_for_each_entry(de, l, d_u.d_alias)
1095 if (!d_unhashed(de))
1096 break;
1097 }
1098 }
1099 spin_unlock(&inode->i_lock);
1100 return de;
1101 }
1102
1103 /*
1104 * Try to kill dentries associated with this inode.
1105 * WARNING: you must own a reference to inode.
1106 */
d_prune_aliases(struct inode * inode)1107 void d_prune_aliases(struct inode *inode)
1108 {
1109 struct dentry *dentry;
1110 restart:
1111 spin_lock(&inode->i_lock);
1112 hlist_for_each_entry(dentry, &inode->i_dentry, d_u.d_alias) {
1113 spin_lock(&dentry->d_lock);
1114 if (!dentry->d_lockref.count) {
1115 struct dentry *parent = lock_parent(dentry);
1116 if (likely(!dentry->d_lockref.count)) {
1117 __dentry_kill(dentry);
1118 dput(parent);
1119 goto restart;
1120 }
1121 if (parent)
1122 spin_unlock(&parent->d_lock);
1123 }
1124 spin_unlock(&dentry->d_lock);
1125 }
1126 spin_unlock(&inode->i_lock);
1127 }
1128 EXPORT_SYMBOL(d_prune_aliases);
1129
1130 /*
1131 * Lock a dentry from shrink list.
1132 * Called under rcu_read_lock() and dentry->d_lock; the former
1133 * guarantees that nothing we access will be freed under us.
1134 * Note that dentry is *not* protected from concurrent dentry_kill(),
1135 * d_delete(), etc.
1136 *
1137 * Return false if dentry has been disrupted or grabbed, leaving
1138 * the caller to kick it off-list. Otherwise, return true and have
1139 * that dentry's inode and parent both locked.
1140 */
shrink_lock_dentry(struct dentry * dentry)1141 static bool shrink_lock_dentry(struct dentry *dentry)
1142 {
1143 struct inode *inode;
1144 struct dentry *parent;
1145
1146 if (dentry->d_lockref.count)
1147 return false;
1148
1149 inode = dentry->d_inode;
1150 if (inode && unlikely(!spin_trylock(&inode->i_lock))) {
1151 spin_unlock(&dentry->d_lock);
1152 spin_lock(&inode->i_lock);
1153 spin_lock(&dentry->d_lock);
1154 if (unlikely(dentry->d_lockref.count))
1155 goto out;
1156 /* changed inode means that somebody had grabbed it */
1157 if (unlikely(inode != dentry->d_inode))
1158 goto out;
1159 }
1160
1161 parent = dentry->d_parent;
1162 if (IS_ROOT(dentry) || likely(spin_trylock(&parent->d_lock)))
1163 return true;
1164
1165 spin_unlock(&dentry->d_lock);
1166 spin_lock(&parent->d_lock);
1167 if (unlikely(parent != dentry->d_parent)) {
1168 spin_unlock(&parent->d_lock);
1169 spin_lock(&dentry->d_lock);
1170 goto out;
1171 }
1172 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
1173 if (likely(!dentry->d_lockref.count))
1174 return true;
1175 spin_unlock(&parent->d_lock);
1176 out:
1177 if (inode)
1178 spin_unlock(&inode->i_lock);
1179 return false;
1180 }
1181
shrink_dentry_list(struct list_head * list)1182 void shrink_dentry_list(struct list_head *list)
1183 {
1184 while (!list_empty(list)) {
1185 struct dentry *dentry, *parent;
1186
1187 dentry = list_entry(list->prev, struct dentry, d_lru);
1188 spin_lock(&dentry->d_lock);
1189 rcu_read_lock();
1190 if (!shrink_lock_dentry(dentry)) {
1191 bool can_free = false;
1192 rcu_read_unlock();
1193 d_shrink_del(dentry);
1194 if (dentry->d_lockref.count < 0)
1195 can_free = dentry->d_flags & DCACHE_MAY_FREE;
1196 spin_unlock(&dentry->d_lock);
1197 if (can_free)
1198 dentry_free(dentry);
1199 continue;
1200 }
1201 rcu_read_unlock();
1202 d_shrink_del(dentry);
1203 parent = dentry->d_parent;
1204 if (parent != dentry)
1205 __dput_to_list(parent, list);
1206 __dentry_kill(dentry);
1207 }
1208 }
1209
dentry_lru_isolate(struct list_head * item,struct list_lru_one * lru,spinlock_t * lru_lock,void * arg)1210 static enum lru_status dentry_lru_isolate(struct list_head *item,
1211 struct list_lru_one *lru, spinlock_t *lru_lock, void *arg)
1212 {
1213 struct list_head *freeable = arg;
1214 struct dentry *dentry = container_of(item, struct dentry, d_lru);
1215
1216
1217 /*
1218 * we are inverting the lru lock/dentry->d_lock here,
1219 * so use a trylock. If we fail to get the lock, just skip
1220 * it
1221 */
1222 if (!spin_trylock(&dentry->d_lock))
1223 return LRU_SKIP;
1224
1225 /*
1226 * Referenced dentries are still in use. If they have active
1227 * counts, just remove them from the LRU. Otherwise give them
1228 * another pass through the LRU.
1229 */
1230 if (dentry->d_lockref.count) {
1231 d_lru_isolate(lru, dentry);
1232 spin_unlock(&dentry->d_lock);
1233 return LRU_REMOVED;
1234 }
1235
1236 if (dentry->d_flags & DCACHE_REFERENCED) {
1237 dentry->d_flags &= ~DCACHE_REFERENCED;
1238 spin_unlock(&dentry->d_lock);
1239
1240 /*
1241 * The list move itself will be made by the common LRU code. At
1242 * this point, we've dropped the dentry->d_lock but keep the
1243 * lru lock. This is safe to do, since every list movement is
1244 * protected by the lru lock even if both locks are held.
1245 *
1246 * This is guaranteed by the fact that all LRU management
1247 * functions are intermediated by the LRU API calls like
1248 * list_lru_add and list_lru_del. List movement in this file
1249 * only ever occur through this functions or through callbacks
1250 * like this one, that are called from the LRU API.
1251 *
1252 * The only exceptions to this are functions like
1253 * shrink_dentry_list, and code that first checks for the
1254 * DCACHE_SHRINK_LIST flag. Those are guaranteed to be
1255 * operating only with stack provided lists after they are
1256 * properly isolated from the main list. It is thus, always a
1257 * local access.
1258 */
1259 return LRU_ROTATE;
1260 }
1261
1262 d_lru_shrink_move(lru, dentry, freeable);
1263 spin_unlock(&dentry->d_lock);
1264
1265 return LRU_REMOVED;
1266 }
1267
1268 /**
1269 * prune_dcache_sb - shrink the dcache
1270 * @sb: superblock
1271 * @sc: shrink control, passed to list_lru_shrink_walk()
1272 *
1273 * Attempt to shrink the superblock dcache LRU by @sc->nr_to_scan entries. This
1274 * is done when we need more memory and called from the superblock shrinker
1275 * function.
1276 *
1277 * This function may fail to free any resources if all the dentries are in
1278 * use.
1279 */
prune_dcache_sb(struct super_block * sb,struct shrink_control * sc)1280 long prune_dcache_sb(struct super_block *sb, struct shrink_control *sc)
1281 {
1282 LIST_HEAD(dispose);
1283 long freed;
1284
1285 freed = list_lru_shrink_walk(&sb->s_dentry_lru, sc,
1286 dentry_lru_isolate, &dispose);
1287 shrink_dentry_list(&dispose);
1288 return freed;
1289 }
1290
dentry_lru_isolate_shrink(struct list_head * item,struct list_lru_one * lru,spinlock_t * lru_lock,void * arg)1291 static enum lru_status dentry_lru_isolate_shrink(struct list_head *item,
1292 struct list_lru_one *lru, spinlock_t *lru_lock, void *arg)
1293 {
1294 struct list_head *freeable = arg;
1295 struct dentry *dentry = container_of(item, struct dentry, d_lru);
1296
1297 /*
1298 * we are inverting the lru lock/dentry->d_lock here,
1299 * so use a trylock. If we fail to get the lock, just skip
1300 * it
1301 */
1302 if (!spin_trylock(&dentry->d_lock))
1303 return LRU_SKIP;
1304
1305 d_lru_shrink_move(lru, dentry, freeable);
1306 spin_unlock(&dentry->d_lock);
1307
1308 return LRU_REMOVED;
1309 }
1310
1311
1312 /**
1313 * shrink_dcache_sb - shrink dcache for a superblock
1314 * @sb: superblock
1315 *
1316 * Shrink the dcache for the specified super block. This is used to free
1317 * the dcache before unmounting a file system.
1318 */
shrink_dcache_sb(struct super_block * sb)1319 void shrink_dcache_sb(struct super_block *sb)
1320 {
1321 do {
1322 LIST_HEAD(dispose);
1323
1324 list_lru_walk(&sb->s_dentry_lru,
1325 dentry_lru_isolate_shrink, &dispose, 1024);
1326 shrink_dentry_list(&dispose);
1327 } while (list_lru_count(&sb->s_dentry_lru) > 0);
1328 }
1329 EXPORT_SYMBOL(shrink_dcache_sb);
1330
1331 /**
1332 * enum d_walk_ret - action to talke during tree walk
1333 * @D_WALK_CONTINUE: contrinue walk
1334 * @D_WALK_QUIT: quit walk
1335 * @D_WALK_NORETRY: quit when retry is needed
1336 * @D_WALK_SKIP: skip this dentry and its children
1337 */
1338 enum d_walk_ret {
1339 D_WALK_CONTINUE,
1340 D_WALK_QUIT,
1341 D_WALK_NORETRY,
1342 D_WALK_SKIP,
1343 };
1344
1345 /**
1346 * d_walk - walk the dentry tree
1347 * @parent: start of walk
1348 * @data: data passed to @enter() and @finish()
1349 * @enter: callback when first entering the dentry
1350 *
1351 * The @enter() callbacks are called with d_lock held.
1352 */
d_walk(struct dentry * parent,void * data,enum d_walk_ret (* enter)(void *,struct dentry *))1353 static void d_walk(struct dentry *parent, void *data,
1354 enum d_walk_ret (*enter)(void *, struct dentry *))
1355 {
1356 struct dentry *this_parent;
1357 struct list_head *next;
1358 unsigned seq = 0;
1359 enum d_walk_ret ret;
1360 bool retry = true;
1361
1362 again:
1363 read_seqbegin_or_lock(&rename_lock, &seq);
1364 this_parent = parent;
1365 spin_lock(&this_parent->d_lock);
1366
1367 ret = enter(data, this_parent);
1368 switch (ret) {
1369 case D_WALK_CONTINUE:
1370 break;
1371 case D_WALK_QUIT:
1372 case D_WALK_SKIP:
1373 goto out_unlock;
1374 case D_WALK_NORETRY:
1375 retry = false;
1376 break;
1377 }
1378 repeat:
1379 next = this_parent->d_subdirs.next;
1380 resume:
1381 while (next != &this_parent->d_subdirs) {
1382 struct list_head *tmp = next;
1383 struct dentry *dentry = list_entry(tmp, struct dentry, d_child);
1384 next = tmp->next;
1385
1386 if (unlikely(dentry->d_flags & DCACHE_DENTRY_CURSOR))
1387 continue;
1388
1389 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
1390
1391 ret = enter(data, dentry);
1392 switch (ret) {
1393 case D_WALK_CONTINUE:
1394 break;
1395 case D_WALK_QUIT:
1396 spin_unlock(&dentry->d_lock);
1397 goto out_unlock;
1398 case D_WALK_NORETRY:
1399 retry = false;
1400 break;
1401 case D_WALK_SKIP:
1402 spin_unlock(&dentry->d_lock);
1403 continue;
1404 }
1405
1406 if (!list_empty(&dentry->d_subdirs)) {
1407 spin_unlock(&this_parent->d_lock);
1408 spin_release(&dentry->d_lock.dep_map, _RET_IP_);
1409 this_parent = dentry;
1410 spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
1411 goto repeat;
1412 }
1413 spin_unlock(&dentry->d_lock);
1414 }
1415 /*
1416 * All done at this level ... ascend and resume the search.
1417 */
1418 rcu_read_lock();
1419 ascend:
1420 if (this_parent != parent) {
1421 struct dentry *child = this_parent;
1422 this_parent = child->d_parent;
1423
1424 spin_unlock(&child->d_lock);
1425 spin_lock(&this_parent->d_lock);
1426
1427 /* might go back up the wrong parent if we have had a rename. */
1428 if (need_seqretry(&rename_lock, seq))
1429 goto rename_retry;
1430 /* go into the first sibling still alive */
1431 do {
1432 next = child->d_child.next;
1433 if (next == &this_parent->d_subdirs)
1434 goto ascend;
1435 child = list_entry(next, struct dentry, d_child);
1436 } while (unlikely(child->d_flags & DCACHE_DENTRY_KILLED));
1437 rcu_read_unlock();
1438 goto resume;
1439 }
1440 if (need_seqretry(&rename_lock, seq))
1441 goto rename_retry;
1442 rcu_read_unlock();
1443
1444 out_unlock:
1445 spin_unlock(&this_parent->d_lock);
1446 done_seqretry(&rename_lock, seq);
1447 return;
1448
1449 rename_retry:
1450 spin_unlock(&this_parent->d_lock);
1451 rcu_read_unlock();
1452 BUG_ON(seq & 1);
1453 if (!retry)
1454 return;
1455 seq = 1;
1456 goto again;
1457 }
1458
1459 struct check_mount {
1460 struct vfsmount *mnt;
1461 unsigned int mounted;
1462 };
1463
path_check_mount(void * data,struct dentry * dentry)1464 static enum d_walk_ret path_check_mount(void *data, struct dentry *dentry)
1465 {
1466 struct check_mount *info = data;
1467 struct path path = { .mnt = info->mnt, .dentry = dentry };
1468
1469 if (likely(!d_mountpoint(dentry)))
1470 return D_WALK_CONTINUE;
1471 if (__path_is_mountpoint(&path)) {
1472 info->mounted = 1;
1473 return D_WALK_QUIT;
1474 }
1475 return D_WALK_CONTINUE;
1476 }
1477
1478 /**
1479 * path_has_submounts - check for mounts over a dentry in the
1480 * current namespace.
1481 * @parent: path to check.
1482 *
1483 * Return true if the parent or its subdirectories contain
1484 * a mount point in the current namespace.
1485 */
path_has_submounts(const struct path * parent)1486 int path_has_submounts(const struct path *parent)
1487 {
1488 struct check_mount data = { .mnt = parent->mnt, .mounted = 0 };
1489
1490 read_seqlock_excl(&mount_lock);
1491 d_walk(parent->dentry, &data, path_check_mount);
1492 read_sequnlock_excl(&mount_lock);
1493
1494 return data.mounted;
1495 }
1496 EXPORT_SYMBOL(path_has_submounts);
1497
1498 /*
1499 * Called by mount code to set a mountpoint and check if the mountpoint is
1500 * reachable (e.g. NFS can unhash a directory dentry and then the complete
1501 * subtree can become unreachable).
1502 *
1503 * Only one of d_invalidate() and d_set_mounted() must succeed. For
1504 * this reason take rename_lock and d_lock on dentry and ancestors.
1505 */
d_set_mounted(struct dentry * dentry)1506 int d_set_mounted(struct dentry *dentry)
1507 {
1508 struct dentry *p;
1509 int ret = -ENOENT;
1510 write_seqlock(&rename_lock);
1511 for (p = dentry->d_parent; !IS_ROOT(p); p = p->d_parent) {
1512 /* Need exclusion wrt. d_invalidate() */
1513 spin_lock(&p->d_lock);
1514 if (unlikely(d_unhashed(p))) {
1515 spin_unlock(&p->d_lock);
1516 goto out;
1517 }
1518 spin_unlock(&p->d_lock);
1519 }
1520 spin_lock(&dentry->d_lock);
1521 if (!d_unlinked(dentry)) {
1522 ret = -EBUSY;
1523 if (!d_mountpoint(dentry)) {
1524 dentry->d_flags |= DCACHE_MOUNTED;
1525 ret = 0;
1526 }
1527 }
1528 spin_unlock(&dentry->d_lock);
1529 out:
1530 write_sequnlock(&rename_lock);
1531 return ret;
1532 }
1533
1534 /*
1535 * Search the dentry child list of the specified parent,
1536 * and move any unused dentries to the end of the unused
1537 * list for prune_dcache(). We descend to the next level
1538 * whenever the d_subdirs list is non-empty and continue
1539 * searching.
1540 *
1541 * It returns zero iff there are no unused children,
1542 * otherwise it returns the number of children moved to
1543 * the end of the unused list. This may not be the total
1544 * number of unused children, because select_parent can
1545 * drop the lock and return early due to latency
1546 * constraints.
1547 */
1548
1549 struct select_data {
1550 struct dentry *start;
1551 union {
1552 long found;
1553 struct dentry *victim;
1554 };
1555 struct list_head dispose;
1556 };
1557
select_collect(void * _data,struct dentry * dentry)1558 static enum d_walk_ret select_collect(void *_data, struct dentry *dentry)
1559 {
1560 struct select_data *data = _data;
1561 enum d_walk_ret ret = D_WALK_CONTINUE;
1562
1563 if (data->start == dentry)
1564 goto out;
1565
1566 if (dentry->d_flags & DCACHE_SHRINK_LIST) {
1567 data->found++;
1568 } else {
1569 if (dentry->d_flags & DCACHE_LRU_LIST)
1570 d_lru_del(dentry);
1571 if (!dentry->d_lockref.count) {
1572 d_shrink_add(dentry, &data->dispose);
1573 data->found++;
1574 }
1575 }
1576 /*
1577 * We can return to the caller if we have found some (this
1578 * ensures forward progress). We'll be coming back to find
1579 * the rest.
1580 */
1581 if (!list_empty(&data->dispose))
1582 ret = need_resched() ? D_WALK_QUIT : D_WALK_NORETRY;
1583 out:
1584 return ret;
1585 }
1586
select_collect2(void * _data,struct dentry * dentry)1587 static enum d_walk_ret select_collect2(void *_data, struct dentry *dentry)
1588 {
1589 struct select_data *data = _data;
1590 enum d_walk_ret ret = D_WALK_CONTINUE;
1591
1592 if (data->start == dentry)
1593 goto out;
1594
1595 if (dentry->d_flags & DCACHE_SHRINK_LIST) {
1596 if (!dentry->d_lockref.count) {
1597 rcu_read_lock();
1598 data->victim = dentry;
1599 return D_WALK_QUIT;
1600 }
1601 } else {
1602 if (dentry->d_flags & DCACHE_LRU_LIST)
1603 d_lru_del(dentry);
1604 if (!dentry->d_lockref.count)
1605 d_shrink_add(dentry, &data->dispose);
1606 }
1607 /*
1608 * We can return to the caller if we have found some (this
1609 * ensures forward progress). We'll be coming back to find
1610 * the rest.
1611 */
1612 if (!list_empty(&data->dispose))
1613 ret = need_resched() ? D_WALK_QUIT : D_WALK_NORETRY;
1614 out:
1615 return ret;
1616 }
1617
1618 /**
1619 * shrink_dcache_parent - prune dcache
1620 * @parent: parent of entries to prune
1621 *
1622 * Prune the dcache to remove unused children of the parent dentry.
1623 */
shrink_dcache_parent(struct dentry * parent)1624 void shrink_dcache_parent(struct dentry *parent)
1625 {
1626 for (;;) {
1627 struct select_data data = {.start = parent};
1628
1629 INIT_LIST_HEAD(&data.dispose);
1630 d_walk(parent, &data, select_collect);
1631
1632 if (!list_empty(&data.dispose)) {
1633 shrink_dentry_list(&data.dispose);
1634 continue;
1635 }
1636
1637 cond_resched();
1638 if (!data.found)
1639 break;
1640 data.victim = NULL;
1641 d_walk(parent, &data, select_collect2);
1642 if (data.victim) {
1643 struct dentry *parent;
1644 spin_lock(&data.victim->d_lock);
1645 if (!shrink_lock_dentry(data.victim)) {
1646 spin_unlock(&data.victim->d_lock);
1647 rcu_read_unlock();
1648 } else {
1649 rcu_read_unlock();
1650 parent = data.victim->d_parent;
1651 if (parent != data.victim)
1652 __dput_to_list(parent, &data.dispose);
1653 __dentry_kill(data.victim);
1654 }
1655 }
1656 if (!list_empty(&data.dispose))
1657 shrink_dentry_list(&data.dispose);
1658 }
1659 }
1660 EXPORT_SYMBOL(shrink_dcache_parent);
1661
umount_check(void * _data,struct dentry * dentry)1662 static enum d_walk_ret umount_check(void *_data, struct dentry *dentry)
1663 {
1664 /* it has busy descendents; complain about those instead */
1665 if (!list_empty(&dentry->d_subdirs))
1666 return D_WALK_CONTINUE;
1667
1668 /* root with refcount 1 is fine */
1669 if (dentry == _data && dentry->d_lockref.count == 1)
1670 return D_WALK_CONTINUE;
1671
1672 WARN(1, "BUG: Dentry %p{i=%lx,n=%pd} "
1673 " still in use (%d) [unmount of %s %s]\n",
1674 dentry,
1675 dentry->d_inode ?
1676 dentry->d_inode->i_ino : 0UL,
1677 dentry,
1678 dentry->d_lockref.count,
1679 dentry->d_sb->s_type->name,
1680 dentry->d_sb->s_id);
1681 return D_WALK_CONTINUE;
1682 }
1683
do_one_tree(struct dentry * dentry)1684 static void do_one_tree(struct dentry *dentry)
1685 {
1686 shrink_dcache_parent(dentry);
1687 d_walk(dentry, dentry, umount_check);
1688 d_drop(dentry);
1689 dput(dentry);
1690 }
1691
1692 /*
1693 * destroy the dentries attached to a superblock on unmounting
1694 */
shrink_dcache_for_umount(struct super_block * sb)1695 void shrink_dcache_for_umount(struct super_block *sb)
1696 {
1697 struct dentry *dentry;
1698
1699 WARN(down_read_trylock(&sb->s_umount), "s_umount should've been locked");
1700
1701 dentry = sb->s_root;
1702 sb->s_root = NULL;
1703 do_one_tree(dentry);
1704
1705 while (!hlist_bl_empty(&sb->s_roots)) {
1706 dentry = dget(hlist_bl_entry(hlist_bl_first(&sb->s_roots), struct dentry, d_hash));
1707 do_one_tree(dentry);
1708 }
1709 }
1710
find_submount(void * _data,struct dentry * dentry)1711 static enum d_walk_ret find_submount(void *_data, struct dentry *dentry)
1712 {
1713 struct dentry **victim = _data;
1714 if (d_mountpoint(dentry)) {
1715 __dget_dlock(dentry);
1716 *victim = dentry;
1717 return D_WALK_QUIT;
1718 }
1719 return D_WALK_CONTINUE;
1720 }
1721
1722 /**
1723 * d_invalidate - detach submounts, prune dcache, and drop
1724 * @dentry: dentry to invalidate (aka detach, prune and drop)
1725 */
d_invalidate(struct dentry * dentry)1726 void d_invalidate(struct dentry *dentry)
1727 {
1728 bool had_submounts = false;
1729 spin_lock(&dentry->d_lock);
1730 if (d_unhashed(dentry)) {
1731 spin_unlock(&dentry->d_lock);
1732 return;
1733 }
1734 __d_drop(dentry);
1735 spin_unlock(&dentry->d_lock);
1736
1737 /* Negative dentries can be dropped without further checks */
1738 if (!dentry->d_inode)
1739 return;
1740
1741 shrink_dcache_parent(dentry);
1742 for (;;) {
1743 struct dentry *victim = NULL;
1744 d_walk(dentry, &victim, find_submount);
1745 if (!victim) {
1746 if (had_submounts)
1747 shrink_dcache_parent(dentry);
1748 return;
1749 }
1750 had_submounts = true;
1751 detach_mounts(victim);
1752 dput(victim);
1753 }
1754 }
1755 EXPORT_SYMBOL(d_invalidate);
1756
1757 /**
1758 * __d_alloc - allocate a dcache entry
1759 * @sb: filesystem it will belong to
1760 * @name: qstr of the name
1761 *
1762 * Allocates a dentry. It returns %NULL if there is insufficient memory
1763 * available. On a success the dentry is returned. The name passed in is
1764 * copied and the copy passed in may be reused after this call.
1765 */
1766
__d_alloc(struct super_block * sb,const struct qstr * name)1767 static struct dentry *__d_alloc(struct super_block *sb, const struct qstr *name)
1768 {
1769 struct dentry *dentry;
1770 char *dname;
1771 int err;
1772
1773 dentry = kmem_cache_alloc_lru(dentry_cache, &sb->s_dentry_lru,
1774 GFP_KERNEL);
1775 if (!dentry)
1776 return NULL;
1777
1778 /*
1779 * We guarantee that the inline name is always NUL-terminated.
1780 * This way the memcpy() done by the name switching in rename
1781 * will still always have a NUL at the end, even if we might
1782 * be overwriting an internal NUL character
1783 */
1784 dentry->d_iname[DNAME_INLINE_LEN-1] = 0;
1785 if (unlikely(!name)) {
1786 name = &slash_name;
1787 dname = dentry->d_iname;
1788 } else if (name->len > DNAME_INLINE_LEN-1) {
1789 size_t size = offsetof(struct external_name, name[1]);
1790 struct external_name *p = kmalloc(size + name->len,
1791 GFP_KERNEL_ACCOUNT |
1792 __GFP_RECLAIMABLE);
1793 if (!p) {
1794 kmem_cache_free(dentry_cache, dentry);
1795 return NULL;
1796 }
1797 atomic_set(&p->u.count, 1);
1798 dname = p->name;
1799 } else {
1800 dname = dentry->d_iname;
1801 }
1802
1803 dentry->d_name.len = name->len;
1804 dentry->d_name.hash = name->hash;
1805 memcpy(dname, name->name, name->len);
1806 dname[name->len] = 0;
1807
1808 /* Make sure we always see the terminating NUL character */
1809 smp_store_release(&dentry->d_name.name, dname); /* ^^^ */
1810
1811 dentry->d_lockref.count = 1;
1812 dentry->d_flags = 0;
1813 spin_lock_init(&dentry->d_lock);
1814 seqcount_spinlock_init(&dentry->d_seq, &dentry->d_lock);
1815 dentry->d_inode = NULL;
1816 dentry->d_parent = dentry;
1817 dentry->d_sb = sb;
1818 dentry->d_op = NULL;
1819 dentry->d_fsdata = NULL;
1820 INIT_HLIST_BL_NODE(&dentry->d_hash);
1821 INIT_LIST_HEAD(&dentry->d_lru);
1822 INIT_LIST_HEAD(&dentry->d_subdirs);
1823 INIT_HLIST_NODE(&dentry->d_u.d_alias);
1824 INIT_LIST_HEAD(&dentry->d_child);
1825 d_set_d_op(dentry, dentry->d_sb->s_d_op);
1826
1827 if (dentry->d_op && dentry->d_op->d_init) {
1828 err = dentry->d_op->d_init(dentry);
1829 if (err) {
1830 if (dname_external(dentry))
1831 kfree(external_name(dentry));
1832 kmem_cache_free(dentry_cache, dentry);
1833 return NULL;
1834 }
1835 }
1836
1837 this_cpu_inc(nr_dentry);
1838
1839 return dentry;
1840 }
1841
1842 /**
1843 * d_alloc - allocate a dcache entry
1844 * @parent: parent of entry to allocate
1845 * @name: qstr of the name
1846 *
1847 * Allocates a dentry. It returns %NULL if there is insufficient memory
1848 * available. On a success the dentry is returned. The name passed in is
1849 * copied and the copy passed in may be reused after this call.
1850 */
d_alloc(struct dentry * parent,const struct qstr * name)1851 struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
1852 {
1853 struct dentry *dentry = __d_alloc(parent->d_sb, name);
1854 if (!dentry)
1855 return NULL;
1856 spin_lock(&parent->d_lock);
1857 /*
1858 * don't need child lock because it is not subject
1859 * to concurrency here
1860 */
1861 __dget_dlock(parent);
1862 dentry->d_parent = parent;
1863 list_add(&dentry->d_child, &parent->d_subdirs);
1864 spin_unlock(&parent->d_lock);
1865
1866 return dentry;
1867 }
1868 EXPORT_SYMBOL(d_alloc);
1869
d_alloc_anon(struct super_block * sb)1870 struct dentry *d_alloc_anon(struct super_block *sb)
1871 {
1872 return __d_alloc(sb, NULL);
1873 }
1874 EXPORT_SYMBOL(d_alloc_anon);
1875
d_alloc_cursor(struct dentry * parent)1876 struct dentry *d_alloc_cursor(struct dentry * parent)
1877 {
1878 struct dentry *dentry = d_alloc_anon(parent->d_sb);
1879 if (dentry) {
1880 dentry->d_flags |= DCACHE_DENTRY_CURSOR;
1881 dentry->d_parent = dget(parent);
1882 }
1883 return dentry;
1884 }
1885
1886 /**
1887 * d_alloc_pseudo - allocate a dentry (for lookup-less filesystems)
1888 * @sb: the superblock
1889 * @name: qstr of the name
1890 *
1891 * For a filesystem that just pins its dentries in memory and never
1892 * performs lookups at all, return an unhashed IS_ROOT dentry.
1893 * This is used for pipes, sockets et.al. - the stuff that should
1894 * never be anyone's children or parents. Unlike all other
1895 * dentries, these will not have RCU delay between dropping the
1896 * last reference and freeing them.
1897 *
1898 * The only user is alloc_file_pseudo() and that's what should
1899 * be considered a public interface. Don't use directly.
1900 */
d_alloc_pseudo(struct super_block * sb,const struct qstr * name)1901 struct dentry *d_alloc_pseudo(struct super_block *sb, const struct qstr *name)
1902 {
1903 struct dentry *dentry = __d_alloc(sb, name);
1904 if (likely(dentry))
1905 dentry->d_flags |= DCACHE_NORCU;
1906 return dentry;
1907 }
1908
d_alloc_name(struct dentry * parent,const char * name)1909 struct dentry *d_alloc_name(struct dentry *parent, const char *name)
1910 {
1911 struct qstr q;
1912
1913 q.name = name;
1914 q.hash_len = hashlen_string(parent, name);
1915 return d_alloc(parent, &q);
1916 }
1917 EXPORT_SYMBOL(d_alloc_name);
1918
d_set_d_op(struct dentry * dentry,const struct dentry_operations * op)1919 void d_set_d_op(struct dentry *dentry, const struct dentry_operations *op)
1920 {
1921 WARN_ON_ONCE(dentry->d_op);
1922 WARN_ON_ONCE(dentry->d_flags & (DCACHE_OP_HASH |
1923 DCACHE_OP_COMPARE |
1924 DCACHE_OP_REVALIDATE |
1925 DCACHE_OP_WEAK_REVALIDATE |
1926 DCACHE_OP_DELETE |
1927 DCACHE_OP_REAL));
1928 dentry->d_op = op;
1929 if (!op)
1930 return;
1931 if (op->d_hash)
1932 dentry->d_flags |= DCACHE_OP_HASH;
1933 if (op->d_compare)
1934 dentry->d_flags |= DCACHE_OP_COMPARE;
1935 if (op->d_revalidate)
1936 dentry->d_flags |= DCACHE_OP_REVALIDATE;
1937 if (op->d_weak_revalidate)
1938 dentry->d_flags |= DCACHE_OP_WEAK_REVALIDATE;
1939 if (op->d_delete)
1940 dentry->d_flags |= DCACHE_OP_DELETE;
1941 if (op->d_prune)
1942 dentry->d_flags |= DCACHE_OP_PRUNE;
1943 if (op->d_real)
1944 dentry->d_flags |= DCACHE_OP_REAL;
1945
1946 }
1947 EXPORT_SYMBOL(d_set_d_op);
1948
1949
1950 /*
1951 * d_set_fallthru - Mark a dentry as falling through to a lower layer
1952 * @dentry - The dentry to mark
1953 *
1954 * Mark a dentry as falling through to the lower layer (as set with
1955 * d_pin_lower()). This flag may be recorded on the medium.
1956 */
d_set_fallthru(struct dentry * dentry)1957 void d_set_fallthru(struct dentry *dentry)
1958 {
1959 spin_lock(&dentry->d_lock);
1960 dentry->d_flags |= DCACHE_FALLTHRU;
1961 spin_unlock(&dentry->d_lock);
1962 }
1963 EXPORT_SYMBOL(d_set_fallthru);
1964
d_flags_for_inode(struct inode * inode)1965 static unsigned d_flags_for_inode(struct inode *inode)
1966 {
1967 unsigned add_flags = DCACHE_REGULAR_TYPE;
1968
1969 if (!inode)
1970 return DCACHE_MISS_TYPE;
1971
1972 if (S_ISDIR(inode->i_mode)) {
1973 add_flags = DCACHE_DIRECTORY_TYPE;
1974 if (unlikely(!(inode->i_opflags & IOP_LOOKUP))) {
1975 if (unlikely(!inode->i_op->lookup))
1976 add_flags = DCACHE_AUTODIR_TYPE;
1977 else
1978 inode->i_opflags |= IOP_LOOKUP;
1979 }
1980 goto type_determined;
1981 }
1982
1983 if (unlikely(!(inode->i_opflags & IOP_NOFOLLOW))) {
1984 if (unlikely(inode->i_op->get_link)) {
1985 add_flags = DCACHE_SYMLINK_TYPE;
1986 goto type_determined;
1987 }
1988 inode->i_opflags |= IOP_NOFOLLOW;
1989 }
1990
1991 if (unlikely(!S_ISREG(inode->i_mode)))
1992 add_flags = DCACHE_SPECIAL_TYPE;
1993
1994 type_determined:
1995 if (unlikely(IS_AUTOMOUNT(inode)))
1996 add_flags |= DCACHE_NEED_AUTOMOUNT;
1997 return add_flags;
1998 }
1999
__d_instantiate(struct dentry * dentry,struct inode * inode)2000 static void __d_instantiate(struct dentry *dentry, struct inode *inode)
2001 {
2002 unsigned add_flags = d_flags_for_inode(inode);
2003 WARN_ON(d_in_lookup(dentry));
2004
2005 spin_lock(&dentry->d_lock);
2006 /*
2007 * The negative counter only tracks dentries on the LRU. Don't dec if
2008 * d_lru is on another list.
2009 */
2010 if ((dentry->d_flags &
2011 (DCACHE_LRU_LIST|DCACHE_SHRINK_LIST)) == DCACHE_LRU_LIST)
2012 this_cpu_dec(nr_dentry_negative);
2013 hlist_add_head(&dentry->d_u.d_alias, &inode->i_dentry);
2014 raw_write_seqcount_begin(&dentry->d_seq);
2015 __d_set_inode_and_type(dentry, inode, add_flags);
2016 raw_write_seqcount_end(&dentry->d_seq);
2017 fsnotify_update_flags(dentry);
2018 spin_unlock(&dentry->d_lock);
2019 }
2020
2021 /**
2022 * d_instantiate - fill in inode information for a dentry
2023 * @entry: dentry to complete
2024 * @inode: inode to attach to this dentry
2025 *
2026 * Fill in inode information in the entry.
2027 *
2028 * This turns negative dentries into productive full members
2029 * of society.
2030 *
2031 * NOTE! This assumes that the inode count has been incremented
2032 * (or otherwise set) by the caller to indicate that it is now
2033 * in use by the dcache.
2034 */
2035
d_instantiate(struct dentry * entry,struct inode * inode)2036 void d_instantiate(struct dentry *entry, struct inode * inode)
2037 {
2038 BUG_ON(!hlist_unhashed(&entry->d_u.d_alias));
2039 if (inode) {
2040 security_d_instantiate(entry, inode);
2041 spin_lock(&inode->i_lock);
2042 __d_instantiate(entry, inode);
2043 spin_unlock(&inode->i_lock);
2044 }
2045 }
2046 EXPORT_SYMBOL(d_instantiate);
2047
2048 /*
2049 * This should be equivalent to d_instantiate() + unlock_new_inode(),
2050 * with lockdep-related part of unlock_new_inode() done before
2051 * anything else. Use that instead of open-coding d_instantiate()/
2052 * unlock_new_inode() combinations.
2053 */
d_instantiate_new(struct dentry * entry,struct inode * inode)2054 void d_instantiate_new(struct dentry *entry, struct inode *inode)
2055 {
2056 BUG_ON(!hlist_unhashed(&entry->d_u.d_alias));
2057 BUG_ON(!inode);
2058 lockdep_annotate_inode_mutex_key(inode);
2059 security_d_instantiate(entry, inode);
2060 spin_lock(&inode->i_lock);
2061 __d_instantiate(entry, inode);
2062 WARN_ON(!(inode->i_state & I_NEW));
2063 inode->i_state &= ~I_NEW & ~I_CREATING;
2064 smp_mb();
2065 wake_up_bit(&inode->i_state, __I_NEW);
2066 spin_unlock(&inode->i_lock);
2067 }
2068 EXPORT_SYMBOL(d_instantiate_new);
2069
d_make_root(struct inode * root_inode)2070 struct dentry *d_make_root(struct inode *root_inode)
2071 {
2072 struct dentry *res = NULL;
2073
2074 if (root_inode) {
2075 res = d_alloc_anon(root_inode->i_sb);
2076 if (res)
2077 d_instantiate(res, root_inode);
2078 else
2079 iput(root_inode);
2080 }
2081 return res;
2082 }
2083 EXPORT_SYMBOL(d_make_root);
2084
__d_instantiate_anon(struct dentry * dentry,struct inode * inode,bool disconnected)2085 static struct dentry *__d_instantiate_anon(struct dentry *dentry,
2086 struct inode *inode,
2087 bool disconnected)
2088 {
2089 struct dentry *res;
2090 unsigned add_flags;
2091
2092 security_d_instantiate(dentry, inode);
2093 spin_lock(&inode->i_lock);
2094 res = __d_find_any_alias(inode);
2095 if (res) {
2096 spin_unlock(&inode->i_lock);
2097 dput(dentry);
2098 goto out_iput;
2099 }
2100
2101 /* attach a disconnected dentry */
2102 add_flags = d_flags_for_inode(inode);
2103
2104 if (disconnected)
2105 add_flags |= DCACHE_DISCONNECTED;
2106
2107 spin_lock(&dentry->d_lock);
2108 __d_set_inode_and_type(dentry, inode, add_flags);
2109 hlist_add_head(&dentry->d_u.d_alias, &inode->i_dentry);
2110 if (!disconnected) {
2111 hlist_bl_lock(&dentry->d_sb->s_roots);
2112 hlist_bl_add_head(&dentry->d_hash, &dentry->d_sb->s_roots);
2113 hlist_bl_unlock(&dentry->d_sb->s_roots);
2114 }
2115 spin_unlock(&dentry->d_lock);
2116 spin_unlock(&inode->i_lock);
2117
2118 return dentry;
2119
2120 out_iput:
2121 iput(inode);
2122 return res;
2123 }
2124
d_instantiate_anon(struct dentry * dentry,struct inode * inode)2125 struct dentry *d_instantiate_anon(struct dentry *dentry, struct inode *inode)
2126 {
2127 return __d_instantiate_anon(dentry, inode, true);
2128 }
2129 EXPORT_SYMBOL(d_instantiate_anon);
2130
__d_obtain_alias(struct inode * inode,bool disconnected)2131 static struct dentry *__d_obtain_alias(struct inode *inode, bool disconnected)
2132 {
2133 struct dentry *tmp;
2134 struct dentry *res;
2135
2136 if (!inode)
2137 return ERR_PTR(-ESTALE);
2138 if (IS_ERR(inode))
2139 return ERR_CAST(inode);
2140
2141 res = d_find_any_alias(inode);
2142 if (res)
2143 goto out_iput;
2144
2145 tmp = d_alloc_anon(inode->i_sb);
2146 if (!tmp) {
2147 res = ERR_PTR(-ENOMEM);
2148 goto out_iput;
2149 }
2150
2151 return __d_instantiate_anon(tmp, inode, disconnected);
2152
2153 out_iput:
2154 iput(inode);
2155 return res;
2156 }
2157
2158 /**
2159 * d_obtain_alias - find or allocate a DISCONNECTED dentry for a given inode
2160 * @inode: inode to allocate the dentry for
2161 *
2162 * Obtain a dentry for an inode resulting from NFS filehandle conversion or
2163 * similar open by handle operations. The returned dentry may be anonymous,
2164 * or may have a full name (if the inode was already in the cache).
2165 *
2166 * When called on a directory inode, we must ensure that the inode only ever
2167 * has one dentry. If a dentry is found, that is returned instead of
2168 * allocating a new one.
2169 *
2170 * On successful return, the reference to the inode has been transferred
2171 * to the dentry. In case of an error the reference on the inode is released.
2172 * To make it easier to use in export operations a %NULL or IS_ERR inode may
2173 * be passed in and the error will be propagated to the return value,
2174 * with a %NULL @inode replaced by ERR_PTR(-ESTALE).
2175 */
d_obtain_alias(struct inode * inode)2176 struct dentry *d_obtain_alias(struct inode *inode)
2177 {
2178 return __d_obtain_alias(inode, true);
2179 }
2180 EXPORT_SYMBOL(d_obtain_alias);
2181
2182 /**
2183 * d_obtain_root - find or allocate a dentry for a given inode
2184 * @inode: inode to allocate the dentry for
2185 *
2186 * Obtain an IS_ROOT dentry for the root of a filesystem.
2187 *
2188 * We must ensure that directory inodes only ever have one dentry. If a
2189 * dentry is found, that is returned instead of allocating a new one.
2190 *
2191 * On successful return, the reference to the inode has been transferred
2192 * to the dentry. In case of an error the reference on the inode is
2193 * released. A %NULL or IS_ERR inode may be passed in and will be the
2194 * error will be propagate to the return value, with a %NULL @inode
2195 * replaced by ERR_PTR(-ESTALE).
2196 */
d_obtain_root(struct inode * inode)2197 struct dentry *d_obtain_root(struct inode *inode)
2198 {
2199 return __d_obtain_alias(inode, false);
2200 }
2201 EXPORT_SYMBOL(d_obtain_root);
2202
2203 /**
2204 * d_add_ci - lookup or allocate new dentry with case-exact name
2205 * @inode: the inode case-insensitive lookup has found
2206 * @dentry: the negative dentry that was passed to the parent's lookup func
2207 * @name: the case-exact name to be associated with the returned dentry
2208 *
2209 * This is to avoid filling the dcache with case-insensitive names to the
2210 * same inode, only the actual correct case is stored in the dcache for
2211 * case-insensitive filesystems.
2212 *
2213 * For a case-insensitive lookup match and if the case-exact dentry
2214 * already exists in the dcache, use it and return it.
2215 *
2216 * If no entry exists with the exact case name, allocate new dentry with
2217 * the exact case, and return the spliced entry.
2218 */
d_add_ci(struct dentry * dentry,struct inode * inode,struct qstr * name)2219 struct dentry *d_add_ci(struct dentry *dentry, struct inode *inode,
2220 struct qstr *name)
2221 {
2222 struct dentry *found, *res;
2223
2224 /*
2225 * First check if a dentry matching the name already exists,
2226 * if not go ahead and create it now.
2227 */
2228 found = d_hash_and_lookup(dentry->d_parent, name);
2229 if (found) {
2230 iput(inode);
2231 return found;
2232 }
2233 if (d_in_lookup(dentry)) {
2234 found = d_alloc_parallel(dentry->d_parent, name,
2235 dentry->d_wait);
2236 if (IS_ERR(found) || !d_in_lookup(found)) {
2237 iput(inode);
2238 return found;
2239 }
2240 } else {
2241 found = d_alloc(dentry->d_parent, name);
2242 if (!found) {
2243 iput(inode);
2244 return ERR_PTR(-ENOMEM);
2245 }
2246 }
2247 res = d_splice_alias(inode, found);
2248 if (res) {
2249 d_lookup_done(found);
2250 dput(found);
2251 return res;
2252 }
2253 return found;
2254 }
2255 EXPORT_SYMBOL(d_add_ci);
2256
2257 /**
2258 * d_same_name - compare dentry name with case-exact name
2259 * @parent: parent dentry
2260 * @dentry: the negative dentry that was passed to the parent's lookup func
2261 * @name: the case-exact name to be associated with the returned dentry
2262 *
2263 * Return: true if names are same, or false
2264 */
d_same_name(const struct dentry * dentry,const struct dentry * parent,const struct qstr * name)2265 bool d_same_name(const struct dentry *dentry, const struct dentry *parent,
2266 const struct qstr *name)
2267 {
2268 if (likely(!(parent->d_flags & DCACHE_OP_COMPARE))) {
2269 if (dentry->d_name.len != name->len)
2270 return false;
2271 return dentry_cmp(dentry, name->name, name->len) == 0;
2272 }
2273 return parent->d_op->d_compare(dentry,
2274 dentry->d_name.len, dentry->d_name.name,
2275 name) == 0;
2276 }
2277 EXPORT_SYMBOL_GPL(d_same_name);
2278
2279 /*
2280 * This is __d_lookup_rcu() when the parent dentry has
2281 * DCACHE_OP_COMPARE, which makes things much nastier.
2282 */
__d_lookup_rcu_op_compare(const struct dentry * parent,const struct qstr * name,unsigned * seqp)2283 static noinline struct dentry *__d_lookup_rcu_op_compare(
2284 const struct dentry *parent,
2285 const struct qstr *name,
2286 unsigned *seqp)
2287 {
2288 u64 hashlen = name->hash_len;
2289 struct hlist_bl_head *b = d_hash(hashlen_hash(hashlen));
2290 struct hlist_bl_node *node;
2291 struct dentry *dentry;
2292
2293 hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
2294 int tlen;
2295 const char *tname;
2296 unsigned seq;
2297
2298 seqretry:
2299 seq = raw_seqcount_begin(&dentry->d_seq);
2300 if (dentry->d_parent != parent)
2301 continue;
2302 if (d_unhashed(dentry))
2303 continue;
2304 if (dentry->d_name.hash != hashlen_hash(hashlen))
2305 continue;
2306 tlen = dentry->d_name.len;
2307 tname = dentry->d_name.name;
2308 /* we want a consistent (name,len) pair */
2309 if (read_seqcount_retry(&dentry->d_seq, seq)) {
2310 cpu_relax();
2311 goto seqretry;
2312 }
2313 if (parent->d_op->d_compare(dentry, tlen, tname, name) != 0)
2314 continue;
2315 *seqp = seq;
2316 return dentry;
2317 }
2318 return NULL;
2319 }
2320
2321 /**
2322 * __d_lookup_rcu - search for a dentry (racy, store-free)
2323 * @parent: parent dentry
2324 * @name: qstr of name we wish to find
2325 * @seqp: returns d_seq value at the point where the dentry was found
2326 * Returns: dentry, or NULL
2327 *
2328 * __d_lookup_rcu is the dcache lookup function for rcu-walk name
2329 * resolution (store-free path walking) design described in
2330 * Documentation/filesystems/path-lookup.txt.
2331 *
2332 * This is not to be used outside core vfs.
2333 *
2334 * __d_lookup_rcu must only be used in rcu-walk mode, ie. with vfsmount lock
2335 * held, and rcu_read_lock held. The returned dentry must not be stored into
2336 * without taking d_lock and checking d_seq sequence count against @seq
2337 * returned here.
2338 *
2339 * A refcount may be taken on the found dentry with the d_rcu_to_refcount
2340 * function.
2341 *
2342 * Alternatively, __d_lookup_rcu may be called again to look up the child of
2343 * the returned dentry, so long as its parent's seqlock is checked after the
2344 * child is looked up. Thus, an interlocking stepping of sequence lock checks
2345 * is formed, giving integrity down the path walk.
2346 *
2347 * NOTE! The caller *has* to check the resulting dentry against the sequence
2348 * number we've returned before using any of the resulting dentry state!
2349 */
__d_lookup_rcu(const struct dentry * parent,const struct qstr * name,unsigned * seqp)2350 struct dentry *__d_lookup_rcu(const struct dentry *parent,
2351 const struct qstr *name,
2352 unsigned *seqp)
2353 {
2354 u64 hashlen = name->hash_len;
2355 const unsigned char *str = name->name;
2356 struct hlist_bl_head *b = d_hash(hashlen_hash(hashlen));
2357 struct hlist_bl_node *node;
2358 struct dentry *dentry;
2359
2360 /*
2361 * Note: There is significant duplication with __d_lookup_rcu which is
2362 * required to prevent single threaded performance regressions
2363 * especially on architectures where smp_rmb (in seqcounts) are costly.
2364 * Keep the two functions in sync.
2365 */
2366
2367 if (unlikely(parent->d_flags & DCACHE_OP_COMPARE))
2368 return __d_lookup_rcu_op_compare(parent, name, seqp);
2369
2370 /*
2371 * The hash list is protected using RCU.
2372 *
2373 * Carefully use d_seq when comparing a candidate dentry, to avoid
2374 * races with d_move().
2375 *
2376 * It is possible that concurrent renames can mess up our list
2377 * walk here and result in missing our dentry, resulting in the
2378 * false-negative result. d_lookup() protects against concurrent
2379 * renames using rename_lock seqlock.
2380 *
2381 * See Documentation/filesystems/path-lookup.txt for more details.
2382 */
2383 hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
2384 unsigned seq;
2385
2386 /*
2387 * The dentry sequence count protects us from concurrent
2388 * renames, and thus protects parent and name fields.
2389 *
2390 * The caller must perform a seqcount check in order
2391 * to do anything useful with the returned dentry.
2392 *
2393 * NOTE! We do a "raw" seqcount_begin here. That means that
2394 * we don't wait for the sequence count to stabilize if it
2395 * is in the middle of a sequence change. If we do the slow
2396 * dentry compare, we will do seqretries until it is stable,
2397 * and if we end up with a successful lookup, we actually
2398 * want to exit RCU lookup anyway.
2399 *
2400 * Note that raw_seqcount_begin still *does* smp_rmb(), so
2401 * we are still guaranteed NUL-termination of ->d_name.name.
2402 */
2403 seq = raw_seqcount_begin(&dentry->d_seq);
2404 if (dentry->d_parent != parent)
2405 continue;
2406 if (d_unhashed(dentry))
2407 continue;
2408 if (dentry->d_name.hash_len != hashlen)
2409 continue;
2410 if (dentry_cmp(dentry, str, hashlen_len(hashlen)) != 0)
2411 continue;
2412 *seqp = seq;
2413 return dentry;
2414 }
2415 return NULL;
2416 }
2417
2418 /**
2419 * d_lookup - search for a dentry
2420 * @parent: parent dentry
2421 * @name: qstr of name we wish to find
2422 * Returns: dentry, or NULL
2423 *
2424 * d_lookup searches the children of the parent dentry for the name in
2425 * question. If the dentry is found its reference count is incremented and the
2426 * dentry is returned. The caller must use dput to free the entry when it has
2427 * finished using it. %NULL is returned if the dentry does not exist.
2428 */
d_lookup(const struct dentry * parent,const struct qstr * name)2429 struct dentry *d_lookup(const struct dentry *parent, const struct qstr *name)
2430 {
2431 struct dentry *dentry;
2432 unsigned seq;
2433
2434 do {
2435 seq = read_seqbegin(&rename_lock);
2436 dentry = __d_lookup(parent, name);
2437 if (dentry)
2438 break;
2439 } while (read_seqretry(&rename_lock, seq));
2440 return dentry;
2441 }
2442 EXPORT_SYMBOL(d_lookup);
2443
2444 /**
2445 * __d_lookup - search for a dentry (racy)
2446 * @parent: parent dentry
2447 * @name: qstr of name we wish to find
2448 * Returns: dentry, or NULL
2449 *
2450 * __d_lookup is like d_lookup, however it may (rarely) return a
2451 * false-negative result due to unrelated rename activity.
2452 *
2453 * __d_lookup is slightly faster by avoiding rename_lock read seqlock,
2454 * however it must be used carefully, eg. with a following d_lookup in
2455 * the case of failure.
2456 *
2457 * __d_lookup callers must be commented.
2458 */
__d_lookup(const struct dentry * parent,const struct qstr * name)2459 struct dentry *__d_lookup(const struct dentry *parent, const struct qstr *name)
2460 {
2461 unsigned int hash = name->hash;
2462 struct hlist_bl_head *b = d_hash(hash);
2463 struct hlist_bl_node *node;
2464 struct dentry *found = NULL;
2465 struct dentry *dentry;
2466
2467 /*
2468 * Note: There is significant duplication with __d_lookup_rcu which is
2469 * required to prevent single threaded performance regressions
2470 * especially on architectures where smp_rmb (in seqcounts) are costly.
2471 * Keep the two functions in sync.
2472 */
2473
2474 /*
2475 * The hash list is protected using RCU.
2476 *
2477 * Take d_lock when comparing a candidate dentry, to avoid races
2478 * with d_move().
2479 *
2480 * It is possible that concurrent renames can mess up our list
2481 * walk here and result in missing our dentry, resulting in the
2482 * false-negative result. d_lookup() protects against concurrent
2483 * renames using rename_lock seqlock.
2484 *
2485 * See Documentation/filesystems/path-lookup.txt for more details.
2486 */
2487 rcu_read_lock();
2488
2489 hlist_bl_for_each_entry_rcu(dentry, node, b, d_hash) {
2490
2491 if (dentry->d_name.hash != hash)
2492 continue;
2493
2494 spin_lock(&dentry->d_lock);
2495 if (dentry->d_parent != parent)
2496 goto next;
2497 if (d_unhashed(dentry))
2498 goto next;
2499
2500 if (!d_same_name(dentry, parent, name))
2501 goto next;
2502
2503 dentry->d_lockref.count++;
2504 found = dentry;
2505 spin_unlock(&dentry->d_lock);
2506 break;
2507 next:
2508 spin_unlock(&dentry->d_lock);
2509 }
2510 rcu_read_unlock();
2511
2512 return found;
2513 }
2514
2515 /**
2516 * d_hash_and_lookup - hash the qstr then search for a dentry
2517 * @dir: Directory to search in
2518 * @name: qstr of name we wish to find
2519 *
2520 * On lookup failure NULL is returned; on bad name - ERR_PTR(-error)
2521 */
d_hash_and_lookup(struct dentry * dir,struct qstr * name)2522 struct dentry *d_hash_and_lookup(struct dentry *dir, struct qstr *name)
2523 {
2524 /*
2525 * Check for a fs-specific hash function. Note that we must
2526 * calculate the standard hash first, as the d_op->d_hash()
2527 * routine may choose to leave the hash value unchanged.
2528 */
2529 name->hash = full_name_hash(dir, name->name, name->len);
2530 if (dir->d_flags & DCACHE_OP_HASH) {
2531 int err = dir->d_op->d_hash(dir, name);
2532 if (unlikely(err < 0))
2533 return ERR_PTR(err);
2534 }
2535 return d_lookup(dir, name);
2536 }
2537 EXPORT_SYMBOL(d_hash_and_lookup);
2538
2539 /*
2540 * When a file is deleted, we have two options:
2541 * - turn this dentry into a negative dentry
2542 * - unhash this dentry and free it.
2543 *
2544 * Usually, we want to just turn this into
2545 * a negative dentry, but if anybody else is
2546 * currently using the dentry or the inode
2547 * we can't do that and we fall back on removing
2548 * it from the hash queues and waiting for
2549 * it to be deleted later when it has no users
2550 */
2551
2552 /**
2553 * d_delete - delete a dentry
2554 * @dentry: The dentry to delete
2555 *
2556 * Turn the dentry into a negative dentry if possible, otherwise
2557 * remove it from the hash queues so it can be deleted later
2558 */
2559
d_delete(struct dentry * dentry)2560 void d_delete(struct dentry * dentry)
2561 {
2562 struct inode *inode = dentry->d_inode;
2563
2564 spin_lock(&inode->i_lock);
2565 spin_lock(&dentry->d_lock);
2566 /*
2567 * Are we the only user?
2568 */
2569 if (dentry->d_lockref.count == 1) {
2570 dentry->d_flags &= ~DCACHE_CANT_MOUNT;
2571 dentry_unlink_inode(dentry);
2572 } else {
2573 __d_drop(dentry);
2574 spin_unlock(&dentry->d_lock);
2575 spin_unlock(&inode->i_lock);
2576 }
2577 }
2578 EXPORT_SYMBOL(d_delete);
2579
__d_rehash(struct dentry * entry)2580 static void __d_rehash(struct dentry *entry)
2581 {
2582 struct hlist_bl_head *b = d_hash(entry->d_name.hash);
2583
2584 hlist_bl_lock(b);
2585 hlist_bl_add_head_rcu(&entry->d_hash, b);
2586 hlist_bl_unlock(b);
2587 }
2588
2589 /**
2590 * d_rehash - add an entry back to the hash
2591 * @entry: dentry to add to the hash
2592 *
2593 * Adds a dentry to the hash according to its name.
2594 */
2595
d_rehash(struct dentry * entry)2596 void d_rehash(struct dentry * entry)
2597 {
2598 spin_lock(&entry->d_lock);
2599 __d_rehash(entry);
2600 spin_unlock(&entry->d_lock);
2601 }
2602 EXPORT_SYMBOL(d_rehash);
2603
start_dir_add(struct inode * dir)2604 static inline unsigned start_dir_add(struct inode *dir)
2605 {
2606 preempt_disable_nested();
2607 for (;;) {
2608 unsigned n = dir->i_dir_seq;
2609 if (!(n & 1) && cmpxchg(&dir->i_dir_seq, n, n + 1) == n)
2610 return n;
2611 cpu_relax();
2612 }
2613 }
2614
end_dir_add(struct inode * dir,unsigned int n,wait_queue_head_t * d_wait)2615 static inline void end_dir_add(struct inode *dir, unsigned int n,
2616 wait_queue_head_t *d_wait)
2617 {
2618 smp_store_release(&dir->i_dir_seq, n + 2);
2619 preempt_enable_nested();
2620 wake_up_all(d_wait);
2621 }
2622
d_wait_lookup(struct dentry * dentry)2623 static void d_wait_lookup(struct dentry *dentry)
2624 {
2625 if (d_in_lookup(dentry)) {
2626 DECLARE_WAITQUEUE(wait, current);
2627 add_wait_queue(dentry->d_wait, &wait);
2628 do {
2629 set_current_state(TASK_UNINTERRUPTIBLE);
2630 spin_unlock(&dentry->d_lock);
2631 schedule();
2632 spin_lock(&dentry->d_lock);
2633 } while (d_in_lookup(dentry));
2634 }
2635 }
2636
d_alloc_parallel(struct dentry * parent,const struct qstr * name,wait_queue_head_t * wq)2637 struct dentry *d_alloc_parallel(struct dentry *parent,
2638 const struct qstr *name,
2639 wait_queue_head_t *wq)
2640 {
2641 unsigned int hash = name->hash;
2642 struct hlist_bl_head *b = in_lookup_hash(parent, hash);
2643 struct hlist_bl_node *node;
2644 struct dentry *new = d_alloc(parent, name);
2645 struct dentry *dentry;
2646 unsigned seq, r_seq, d_seq;
2647
2648 if (unlikely(!new))
2649 return ERR_PTR(-ENOMEM);
2650
2651 retry:
2652 rcu_read_lock();
2653 seq = smp_load_acquire(&parent->d_inode->i_dir_seq);
2654 r_seq = read_seqbegin(&rename_lock);
2655 dentry = __d_lookup_rcu(parent, name, &d_seq);
2656 if (unlikely(dentry)) {
2657 if (!lockref_get_not_dead(&dentry->d_lockref)) {
2658 rcu_read_unlock();
2659 goto retry;
2660 }
2661 if (read_seqcount_retry(&dentry->d_seq, d_seq)) {
2662 rcu_read_unlock();
2663 dput(dentry);
2664 goto retry;
2665 }
2666 rcu_read_unlock();
2667 dput(new);
2668 return dentry;
2669 }
2670 if (unlikely(read_seqretry(&rename_lock, r_seq))) {
2671 rcu_read_unlock();
2672 goto retry;
2673 }
2674
2675 if (unlikely(seq & 1)) {
2676 rcu_read_unlock();
2677 goto retry;
2678 }
2679
2680 hlist_bl_lock(b);
2681 if (unlikely(READ_ONCE(parent->d_inode->i_dir_seq) != seq)) {
2682 hlist_bl_unlock(b);
2683 rcu_read_unlock();
2684 goto retry;
2685 }
2686 /*
2687 * No changes for the parent since the beginning of d_lookup().
2688 * Since all removals from the chain happen with hlist_bl_lock(),
2689 * any potential in-lookup matches are going to stay here until
2690 * we unlock the chain. All fields are stable in everything
2691 * we encounter.
2692 */
2693 hlist_bl_for_each_entry(dentry, node, b, d_u.d_in_lookup_hash) {
2694 if (dentry->d_name.hash != hash)
2695 continue;
2696 if (dentry->d_parent != parent)
2697 continue;
2698 if (!d_same_name(dentry, parent, name))
2699 continue;
2700 hlist_bl_unlock(b);
2701 /* now we can try to grab a reference */
2702 if (!lockref_get_not_dead(&dentry->d_lockref)) {
2703 rcu_read_unlock();
2704 goto retry;
2705 }
2706
2707 rcu_read_unlock();
2708 /*
2709 * somebody is likely to be still doing lookup for it;
2710 * wait for them to finish
2711 */
2712 spin_lock(&dentry->d_lock);
2713 d_wait_lookup(dentry);
2714 /*
2715 * it's not in-lookup anymore; in principle we should repeat
2716 * everything from dcache lookup, but it's likely to be what
2717 * d_lookup() would've found anyway. If it is, just return it;
2718 * otherwise we really have to repeat the whole thing.
2719 */
2720 if (unlikely(dentry->d_name.hash != hash))
2721 goto mismatch;
2722 if (unlikely(dentry->d_parent != parent))
2723 goto mismatch;
2724 if (unlikely(d_unhashed(dentry)))
2725 goto mismatch;
2726 if (unlikely(!d_same_name(dentry, parent, name)))
2727 goto mismatch;
2728 /* OK, it *is* a hashed match; return it */
2729 spin_unlock(&dentry->d_lock);
2730 dput(new);
2731 return dentry;
2732 }
2733 rcu_read_unlock();
2734 /* we can't take ->d_lock here; it's OK, though. */
2735 new->d_flags |= DCACHE_PAR_LOOKUP;
2736 new->d_wait = wq;
2737 hlist_bl_add_head_rcu(&new->d_u.d_in_lookup_hash, b);
2738 hlist_bl_unlock(b);
2739 return new;
2740 mismatch:
2741 spin_unlock(&dentry->d_lock);
2742 dput(dentry);
2743 goto retry;
2744 }
2745 EXPORT_SYMBOL(d_alloc_parallel);
2746
2747 /*
2748 * - Unhash the dentry
2749 * - Retrieve and clear the waitqueue head in dentry
2750 * - Return the waitqueue head
2751 */
__d_lookup_unhash(struct dentry * dentry)2752 static wait_queue_head_t *__d_lookup_unhash(struct dentry *dentry)
2753 {
2754 wait_queue_head_t *d_wait;
2755 struct hlist_bl_head *b;
2756
2757 lockdep_assert_held(&dentry->d_lock);
2758
2759 b = in_lookup_hash(dentry->d_parent, dentry->d_name.hash);
2760 hlist_bl_lock(b);
2761 dentry->d_flags &= ~DCACHE_PAR_LOOKUP;
2762 __hlist_bl_del(&dentry->d_u.d_in_lookup_hash);
2763 d_wait = dentry->d_wait;
2764 dentry->d_wait = NULL;
2765 hlist_bl_unlock(b);
2766 INIT_HLIST_NODE(&dentry->d_u.d_alias);
2767 INIT_LIST_HEAD(&dentry->d_lru);
2768 return d_wait;
2769 }
2770
__d_lookup_unhash_wake(struct dentry * dentry)2771 void __d_lookup_unhash_wake(struct dentry *dentry)
2772 {
2773 spin_lock(&dentry->d_lock);
2774 wake_up_all(__d_lookup_unhash(dentry));
2775 spin_unlock(&dentry->d_lock);
2776 }
2777 EXPORT_SYMBOL(__d_lookup_unhash_wake);
2778
2779 /* inode->i_lock held if inode is non-NULL */
2780
__d_add(struct dentry * dentry,struct inode * inode)2781 static inline void __d_add(struct dentry *dentry, struct inode *inode)
2782 {
2783 wait_queue_head_t *d_wait;
2784 struct inode *dir = NULL;
2785 unsigned n;
2786 spin_lock(&dentry->d_lock);
2787 if (unlikely(d_in_lookup(dentry))) {
2788 dir = dentry->d_parent->d_inode;
2789 n = start_dir_add(dir);
2790 d_wait = __d_lookup_unhash(dentry);
2791 }
2792 if (inode) {
2793 unsigned add_flags = d_flags_for_inode(inode);
2794 hlist_add_head(&dentry->d_u.d_alias, &inode->i_dentry);
2795 raw_write_seqcount_begin(&dentry->d_seq);
2796 __d_set_inode_and_type(dentry, inode, add_flags);
2797 raw_write_seqcount_end(&dentry->d_seq);
2798 fsnotify_update_flags(dentry);
2799 }
2800 __d_rehash(dentry);
2801 if (dir)
2802 end_dir_add(dir, n, d_wait);
2803 spin_unlock(&dentry->d_lock);
2804 if (inode)
2805 spin_unlock(&inode->i_lock);
2806 }
2807
2808 /**
2809 * d_add - add dentry to hash queues
2810 * @entry: dentry to add
2811 * @inode: The inode to attach to this dentry
2812 *
2813 * This adds the entry to the hash queues and initializes @inode.
2814 * The entry was actually filled in earlier during d_alloc().
2815 */
2816
d_add(struct dentry * entry,struct inode * inode)2817 void d_add(struct dentry *entry, struct inode *inode)
2818 {
2819 if (inode) {
2820 security_d_instantiate(entry, inode);
2821 spin_lock(&inode->i_lock);
2822 }
2823 __d_add(entry, inode);
2824 }
2825 EXPORT_SYMBOL(d_add);
2826
2827 /**
2828 * d_exact_alias - find and hash an exact unhashed alias
2829 * @entry: dentry to add
2830 * @inode: The inode to go with this dentry
2831 *
2832 * If an unhashed dentry with the same name/parent and desired
2833 * inode already exists, hash and return it. Otherwise, return
2834 * NULL.
2835 *
2836 * Parent directory should be locked.
2837 */
d_exact_alias(struct dentry * entry,struct inode * inode)2838 struct dentry *d_exact_alias(struct dentry *entry, struct inode *inode)
2839 {
2840 struct dentry *alias;
2841 unsigned int hash = entry->d_name.hash;
2842
2843 spin_lock(&inode->i_lock);
2844 hlist_for_each_entry(alias, &inode->i_dentry, d_u.d_alias) {
2845 /*
2846 * Don't need alias->d_lock here, because aliases with
2847 * d_parent == entry->d_parent are not subject to name or
2848 * parent changes, because the parent inode i_mutex is held.
2849 */
2850 if (alias->d_name.hash != hash)
2851 continue;
2852 if (alias->d_parent != entry->d_parent)
2853 continue;
2854 if (!d_same_name(alias, entry->d_parent, &entry->d_name))
2855 continue;
2856 spin_lock(&alias->d_lock);
2857 if (!d_unhashed(alias)) {
2858 spin_unlock(&alias->d_lock);
2859 alias = NULL;
2860 } else {
2861 __dget_dlock(alias);
2862 __d_rehash(alias);
2863 spin_unlock(&alias->d_lock);
2864 }
2865 spin_unlock(&inode->i_lock);
2866 return alias;
2867 }
2868 spin_unlock(&inode->i_lock);
2869 return NULL;
2870 }
2871 EXPORT_SYMBOL(d_exact_alias);
2872
swap_names(struct dentry * dentry,struct dentry * target)2873 static void swap_names(struct dentry *dentry, struct dentry *target)
2874 {
2875 if (unlikely(dname_external(target))) {
2876 if (unlikely(dname_external(dentry))) {
2877 /*
2878 * Both external: swap the pointers
2879 */
2880 swap(target->d_name.name, dentry->d_name.name);
2881 } else {
2882 /*
2883 * dentry:internal, target:external. Steal target's
2884 * storage and make target internal.
2885 */
2886 memcpy(target->d_iname, dentry->d_name.name,
2887 dentry->d_name.len + 1);
2888 dentry->d_name.name = target->d_name.name;
2889 target->d_name.name = target->d_iname;
2890 }
2891 } else {
2892 if (unlikely(dname_external(dentry))) {
2893 /*
2894 * dentry:external, target:internal. Give dentry's
2895 * storage to target and make dentry internal
2896 */
2897 memcpy(dentry->d_iname, target->d_name.name,
2898 target->d_name.len + 1);
2899 target->d_name.name = dentry->d_name.name;
2900 dentry->d_name.name = dentry->d_iname;
2901 } else {
2902 /*
2903 * Both are internal.
2904 */
2905 unsigned int i;
2906 BUILD_BUG_ON(!IS_ALIGNED(DNAME_INLINE_LEN, sizeof(long)));
2907 for (i = 0; i < DNAME_INLINE_LEN / sizeof(long); i++) {
2908 swap(((long *) &dentry->d_iname)[i],
2909 ((long *) &target->d_iname)[i]);
2910 }
2911 }
2912 }
2913 swap(dentry->d_name.hash_len, target->d_name.hash_len);
2914 }
2915
copy_name(struct dentry * dentry,struct dentry * target)2916 static void copy_name(struct dentry *dentry, struct dentry *target)
2917 {
2918 struct external_name *old_name = NULL;
2919 if (unlikely(dname_external(dentry)))
2920 old_name = external_name(dentry);
2921 if (unlikely(dname_external(target))) {
2922 atomic_inc(&external_name(target)->u.count);
2923 dentry->d_name = target->d_name;
2924 } else {
2925 memcpy(dentry->d_iname, target->d_name.name,
2926 target->d_name.len + 1);
2927 dentry->d_name.name = dentry->d_iname;
2928 dentry->d_name.hash_len = target->d_name.hash_len;
2929 }
2930 if (old_name && likely(atomic_dec_and_test(&old_name->u.count)))
2931 kfree_rcu(old_name, u.head);
2932 }
2933
2934 /*
2935 * __d_move - move a dentry
2936 * @dentry: entry to move
2937 * @target: new dentry
2938 * @exchange: exchange the two dentries
2939 *
2940 * Update the dcache to reflect the move of a file name. Negative
2941 * dcache entries should not be moved in this way. Caller must hold
2942 * rename_lock, the i_mutex of the source and target directories,
2943 * and the sb->s_vfs_rename_mutex if they differ. See lock_rename().
2944 */
__d_move(struct dentry * dentry,struct dentry * target,bool exchange)2945 static void __d_move(struct dentry *dentry, struct dentry *target,
2946 bool exchange)
2947 {
2948 struct dentry *old_parent, *p;
2949 wait_queue_head_t *d_wait;
2950 struct inode *dir = NULL;
2951 unsigned n;
2952
2953 WARN_ON(!dentry->d_inode);
2954 if (WARN_ON(dentry == target))
2955 return;
2956
2957 BUG_ON(d_ancestor(target, dentry));
2958 old_parent = dentry->d_parent;
2959 p = d_ancestor(old_parent, target);
2960 if (IS_ROOT(dentry)) {
2961 BUG_ON(p);
2962 spin_lock(&target->d_parent->d_lock);
2963 } else if (!p) {
2964 /* target is not a descendent of dentry->d_parent */
2965 spin_lock(&target->d_parent->d_lock);
2966 spin_lock_nested(&old_parent->d_lock, DENTRY_D_LOCK_NESTED);
2967 } else {
2968 BUG_ON(p == dentry);
2969 spin_lock(&old_parent->d_lock);
2970 if (p != target)
2971 spin_lock_nested(&target->d_parent->d_lock,
2972 DENTRY_D_LOCK_NESTED);
2973 }
2974 spin_lock_nested(&dentry->d_lock, 2);
2975 spin_lock_nested(&target->d_lock, 3);
2976
2977 if (unlikely(d_in_lookup(target))) {
2978 dir = target->d_parent->d_inode;
2979 n = start_dir_add(dir);
2980 d_wait = __d_lookup_unhash(target);
2981 }
2982
2983 write_seqcount_begin(&dentry->d_seq);
2984 write_seqcount_begin_nested(&target->d_seq, DENTRY_D_LOCK_NESTED);
2985
2986 /* unhash both */
2987 if (!d_unhashed(dentry))
2988 ___d_drop(dentry);
2989 if (!d_unhashed(target))
2990 ___d_drop(target);
2991
2992 /* ... and switch them in the tree */
2993 dentry->d_parent = target->d_parent;
2994 if (!exchange) {
2995 copy_name(dentry, target);
2996 target->d_hash.pprev = NULL;
2997 dentry->d_parent->d_lockref.count++;
2998 if (dentry != old_parent) /* wasn't IS_ROOT */
2999 WARN_ON(!--old_parent->d_lockref.count);
3000 } else {
3001 target->d_parent = old_parent;
3002 swap_names(dentry, target);
3003 list_move(&target->d_child, &target->d_parent->d_subdirs);
3004 __d_rehash(target);
3005 fsnotify_update_flags(target);
3006 }
3007 list_move(&dentry->d_child, &dentry->d_parent->d_subdirs);
3008 __d_rehash(dentry);
3009 fsnotify_update_flags(dentry);
3010 fscrypt_handle_d_move(dentry);
3011
3012 write_seqcount_end(&target->d_seq);
3013 write_seqcount_end(&dentry->d_seq);
3014
3015 if (dir)
3016 end_dir_add(dir, n, d_wait);
3017
3018 if (dentry->d_parent != old_parent)
3019 spin_unlock(&dentry->d_parent->d_lock);
3020 if (dentry != old_parent)
3021 spin_unlock(&old_parent->d_lock);
3022 spin_unlock(&target->d_lock);
3023 spin_unlock(&dentry->d_lock);
3024 }
3025
3026 /*
3027 * d_move - move a dentry
3028 * @dentry: entry to move
3029 * @target: new dentry
3030 *
3031 * Update the dcache to reflect the move of a file name. Negative
3032 * dcache entries should not be moved in this way. See the locking
3033 * requirements for __d_move.
3034 */
d_move(struct dentry * dentry,struct dentry * target)3035 void d_move(struct dentry *dentry, struct dentry *target)
3036 {
3037 write_seqlock(&rename_lock);
3038 __d_move(dentry, target, false);
3039 write_sequnlock(&rename_lock);
3040 }
3041 EXPORT_SYMBOL(d_move);
3042
3043 /*
3044 * d_exchange - exchange two dentries
3045 * @dentry1: first dentry
3046 * @dentry2: second dentry
3047 */
d_exchange(struct dentry * dentry1,struct dentry * dentry2)3048 void d_exchange(struct dentry *dentry1, struct dentry *dentry2)
3049 {
3050 write_seqlock(&rename_lock);
3051
3052 WARN_ON(!dentry1->d_inode);
3053 WARN_ON(!dentry2->d_inode);
3054 WARN_ON(IS_ROOT(dentry1));
3055 WARN_ON(IS_ROOT(dentry2));
3056
3057 __d_move(dentry1, dentry2, true);
3058
3059 write_sequnlock(&rename_lock);
3060 }
3061
3062 /**
3063 * d_ancestor - search for an ancestor
3064 * @p1: ancestor dentry
3065 * @p2: child dentry
3066 *
3067 * Returns the ancestor dentry of p2 which is a child of p1, if p1 is
3068 * an ancestor of p2, else NULL.
3069 */
d_ancestor(struct dentry * p1,struct dentry * p2)3070 struct dentry *d_ancestor(struct dentry *p1, struct dentry *p2)
3071 {
3072 struct dentry *p;
3073
3074 for (p = p2; !IS_ROOT(p); p = p->d_parent) {
3075 if (p->d_parent == p1)
3076 return p;
3077 }
3078 return NULL;
3079 }
3080
3081 /*
3082 * This helper attempts to cope with remotely renamed directories
3083 *
3084 * It assumes that the caller is already holding
3085 * dentry->d_parent->d_inode->i_mutex, and rename_lock
3086 *
3087 * Note: If ever the locking in lock_rename() changes, then please
3088 * remember to update this too...
3089 */
__d_unalias(struct inode * inode,struct dentry * dentry,struct dentry * alias)3090 static int __d_unalias(struct inode *inode,
3091 struct dentry *dentry, struct dentry *alias)
3092 {
3093 struct mutex *m1 = NULL;
3094 struct rw_semaphore *m2 = NULL;
3095 int ret = -ESTALE;
3096
3097 /* If alias and dentry share a parent, then no extra locks required */
3098 if (alias->d_parent == dentry->d_parent)
3099 goto out_unalias;
3100
3101 /* See lock_rename() */
3102 if (!mutex_trylock(&dentry->d_sb->s_vfs_rename_mutex))
3103 goto out_err;
3104 m1 = &dentry->d_sb->s_vfs_rename_mutex;
3105 if (!inode_trylock_shared(alias->d_parent->d_inode))
3106 goto out_err;
3107 m2 = &alias->d_parent->d_inode->i_rwsem;
3108 out_unalias:
3109 __d_move(alias, dentry, false);
3110 ret = 0;
3111 out_err:
3112 if (m2)
3113 up_read(m2);
3114 if (m1)
3115 mutex_unlock(m1);
3116 return ret;
3117 }
3118
3119 /**
3120 * d_splice_alias - splice a disconnected dentry into the tree if one exists
3121 * @inode: the inode which may have a disconnected dentry
3122 * @dentry: a negative dentry which we want to point to the inode.
3123 *
3124 * If inode is a directory and has an IS_ROOT alias, then d_move that in
3125 * place of the given dentry and return it, else simply d_add the inode
3126 * to the dentry and return NULL.
3127 *
3128 * If a non-IS_ROOT directory is found, the filesystem is corrupt, and
3129 * we should error out: directories can't have multiple aliases.
3130 *
3131 * This is needed in the lookup routine of any filesystem that is exportable
3132 * (via knfsd) so that we can build dcache paths to directories effectively.
3133 *
3134 * If a dentry was found and moved, then it is returned. Otherwise NULL
3135 * is returned. This matches the expected return value of ->lookup.
3136 *
3137 * Cluster filesystems may call this function with a negative, hashed dentry.
3138 * In that case, we know that the inode will be a regular file, and also this
3139 * will only occur during atomic_open. So we need to check for the dentry
3140 * being already hashed only in the final case.
3141 */
d_splice_alias(struct inode * inode,struct dentry * dentry)3142 struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
3143 {
3144 if (IS_ERR(inode))
3145 return ERR_CAST(inode);
3146
3147 BUG_ON(!d_unhashed(dentry));
3148
3149 if (!inode)
3150 goto out;
3151
3152 security_d_instantiate(dentry, inode);
3153 spin_lock(&inode->i_lock);
3154 if (S_ISDIR(inode->i_mode)) {
3155 struct dentry *new = __d_find_any_alias(inode);
3156 if (unlikely(new)) {
3157 /* The reference to new ensures it remains an alias */
3158 spin_unlock(&inode->i_lock);
3159 write_seqlock(&rename_lock);
3160 if (unlikely(d_ancestor(new, dentry))) {
3161 write_sequnlock(&rename_lock);
3162 dput(new);
3163 new = ERR_PTR(-ELOOP);
3164 pr_warn_ratelimited(
3165 "VFS: Lookup of '%s' in %s %s"
3166 " would have caused loop\n",
3167 dentry->d_name.name,
3168 inode->i_sb->s_type->name,
3169 inode->i_sb->s_id);
3170 } else if (!IS_ROOT(new)) {
3171 struct dentry *old_parent = dget(new->d_parent);
3172 int err = __d_unalias(inode, dentry, new);
3173 write_sequnlock(&rename_lock);
3174 if (err) {
3175 dput(new);
3176 new = ERR_PTR(err);
3177 }
3178 dput(old_parent);
3179 } else {
3180 __d_move(new, dentry, false);
3181 write_sequnlock(&rename_lock);
3182 }
3183 iput(inode);
3184 return new;
3185 }
3186 }
3187 out:
3188 __d_add(dentry, inode);
3189 return NULL;
3190 }
3191 EXPORT_SYMBOL(d_splice_alias);
3192
3193 /*
3194 * Test whether new_dentry is a subdirectory of old_dentry.
3195 *
3196 * Trivially implemented using the dcache structure
3197 */
3198
3199 /**
3200 * is_subdir - is new dentry a subdirectory of old_dentry
3201 * @new_dentry: new dentry
3202 * @old_dentry: old dentry
3203 *
3204 * Returns true if new_dentry is a subdirectory of the parent (at any depth).
3205 * Returns false otherwise.
3206 * Caller must ensure that "new_dentry" is pinned before calling is_subdir()
3207 */
3208
is_subdir(struct dentry * new_dentry,struct dentry * old_dentry)3209 bool is_subdir(struct dentry *new_dentry, struct dentry *old_dentry)
3210 {
3211 bool subdir;
3212 unsigned seq;
3213
3214 if (new_dentry == old_dentry)
3215 return true;
3216
3217 /* Access d_parent under rcu as d_move() may change it. */
3218 rcu_read_lock();
3219 seq = read_seqbegin(&rename_lock);
3220 subdir = d_ancestor(old_dentry, new_dentry);
3221 /* Try lockless once... */
3222 if (read_seqretry(&rename_lock, seq)) {
3223 /* ...else acquire lock for progress even on deep chains. */
3224 read_seqlock_excl(&rename_lock);
3225 subdir = d_ancestor(old_dentry, new_dentry);
3226 read_sequnlock_excl(&rename_lock);
3227 }
3228 rcu_read_unlock();
3229 return subdir;
3230 }
3231 EXPORT_SYMBOL(is_subdir);
3232
d_genocide_kill(void * data,struct dentry * dentry)3233 static enum d_walk_ret d_genocide_kill(void *data, struct dentry *dentry)
3234 {
3235 struct dentry *root = data;
3236 if (dentry != root) {
3237 if (d_unhashed(dentry) || !dentry->d_inode)
3238 return D_WALK_SKIP;
3239
3240 if (!(dentry->d_flags & DCACHE_GENOCIDE)) {
3241 dentry->d_flags |= DCACHE_GENOCIDE;
3242 dentry->d_lockref.count--;
3243 }
3244 }
3245 return D_WALK_CONTINUE;
3246 }
3247
d_genocide(struct dentry * parent)3248 void d_genocide(struct dentry *parent)
3249 {
3250 d_walk(parent, parent, d_genocide_kill);
3251 }
3252
d_tmpfile(struct file * file,struct inode * inode)3253 void d_tmpfile(struct file *file, struct inode *inode)
3254 {
3255 struct dentry *dentry = file->f_path.dentry;
3256
3257 inode_dec_link_count(inode);
3258 BUG_ON(dentry->d_name.name != dentry->d_iname ||
3259 !hlist_unhashed(&dentry->d_u.d_alias) ||
3260 !d_unlinked(dentry));
3261 spin_lock(&dentry->d_parent->d_lock);
3262 spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
3263 dentry->d_name.len = sprintf(dentry->d_iname, "#%llu",
3264 (unsigned long long)inode->i_ino);
3265 spin_unlock(&dentry->d_lock);
3266 spin_unlock(&dentry->d_parent->d_lock);
3267 d_instantiate(dentry, inode);
3268 }
3269 EXPORT_SYMBOL(d_tmpfile);
3270
3271 static __initdata unsigned long dhash_entries;
set_dhash_entries(char * str)3272 static int __init set_dhash_entries(char *str)
3273 {
3274 if (!str)
3275 return 0;
3276 dhash_entries = simple_strtoul(str, &str, 0);
3277 return 1;
3278 }
3279 __setup("dhash_entries=", set_dhash_entries);
3280
dcache_init_early(void)3281 static void __init dcache_init_early(void)
3282 {
3283 /* If hashes are distributed across NUMA nodes, defer
3284 * hash allocation until vmalloc space is available.
3285 */
3286 if (hashdist)
3287 return;
3288
3289 dentry_hashtable =
3290 alloc_large_system_hash("Dentry cache",
3291 sizeof(struct hlist_bl_head),
3292 dhash_entries,
3293 13,
3294 HASH_EARLY | HASH_ZERO,
3295 &d_hash_shift,
3296 NULL,
3297 0,
3298 0);
3299 d_hash_shift = 32 - d_hash_shift;
3300 }
3301
dcache_init(void)3302 static void __init dcache_init(void)
3303 {
3304 /*
3305 * A constructor could be added for stable state like the lists,
3306 * but it is probably not worth it because of the cache nature
3307 * of the dcache.
3308 */
3309 dentry_cache = KMEM_CACHE_USERCOPY(dentry,
3310 SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|SLAB_MEM_SPREAD|SLAB_ACCOUNT,
3311 d_iname);
3312
3313 /* Hash may have been set up in dcache_init_early */
3314 if (!hashdist)
3315 return;
3316
3317 dentry_hashtable =
3318 alloc_large_system_hash("Dentry cache",
3319 sizeof(struct hlist_bl_head),
3320 dhash_entries,
3321 13,
3322 HASH_ZERO,
3323 &d_hash_shift,
3324 NULL,
3325 0,
3326 0);
3327 d_hash_shift = 32 - d_hash_shift;
3328 }
3329
3330 /* SLAB cache for __getname() consumers */
3331 struct kmem_cache *names_cachep __read_mostly;
3332 EXPORT_SYMBOL(names_cachep);
3333
vfs_caches_init_early(void)3334 void __init vfs_caches_init_early(void)
3335 {
3336 int i;
3337
3338 for (i = 0; i < ARRAY_SIZE(in_lookup_hashtable); i++)
3339 INIT_HLIST_BL_HEAD(&in_lookup_hashtable[i]);
3340
3341 dcache_init_early();
3342 inode_init_early();
3343 }
3344
vfs_caches_init(void)3345 void __init vfs_caches_init(void)
3346 {
3347 names_cachep = kmem_cache_create_usercopy("names_cache", PATH_MAX, 0,
3348 SLAB_HWCACHE_ALIGN|SLAB_PANIC, 0, PATH_MAX, NULL);
3349
3350 dcache_init();
3351 inode_init();
3352 files_init();
3353 files_maxfiles_init();
3354 mnt_init();
3355 bdev_cache_init();
3356 chrdev_init();
3357 }
3358