2 * (C) 1997 Linus Torvalds
3 * (C) 1999 Andrea Arcangeli <andrea@suse.de> (dynamic inode allocation)
7 #include <linux/dcache.h>
8 #include <linux/init.h>
9 #include <linux/slab.h>
10 #include <linux/writeback.h>
11 #include <linux/module.h>
12 #include <linux/backing-dev.h>
13 #include <linux/wait.h>
14 #include <linux/rwsem.h>
15 #include <linux/hash.h>
16 #include <linux/swap.h>
17 #include <linux/security.h>
18 #include <linux/pagemap.h>
19 #include <linux/cdev.h>
20 #include <linux/bootmem.h>
21 #include <linux/fsnotify.h>
22 #include <linux/mount.h>
23 #include <linux/async.h>
24 #include <linux/posix_acl.h>
25 #include <linux/prefetch.h>
26 #include <linux/ima.h>
27 #include <linux/cred.h>
28 #include <linux/buffer_head.h> /* for inode_has_buffers */
29 #include <linux/ratelimit.h>
33 * Inode locking rules:
35 * inode->i_lock protects:
36 * inode->i_state, inode->i_hash, __iget()
37 * inode->i_sb->s_inode_lru_lock protects:
38 * inode->i_sb->s_inode_lru, inode->i_lru
39 * inode_sb_list_lock protects:
40 * sb->s_inodes, inode->i_sb_list
41 * bdi->wb.list_lock protects:
42 * bdi->wb.b_{dirty,io,more_io}, inode->i_wb_list
43 * inode_hash_lock protects:
44 * inode_hashtable, inode->i_hash
50 * inode->i_sb->s_inode_lru_lock
63 static unsigned int i_hash_mask __read_mostly
;
64 static unsigned int i_hash_shift __read_mostly
;
65 static struct hlist_head
*inode_hashtable __read_mostly
;
66 static __cacheline_aligned_in_smp
DEFINE_SPINLOCK(inode_hash_lock
);
68 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(inode_sb_list_lock
);
71 * Empty aops. Can be used for the cases where the user does not
72 * define any of the address_space operations.
74 const struct address_space_operations empty_aops
= {
76 EXPORT_SYMBOL(empty_aops
);
79 * Statistics gathering..
81 struct inodes_stat_t inodes_stat
;
83 static DEFINE_PER_CPU(unsigned int, nr_inodes
);
84 static DEFINE_PER_CPU(unsigned int, nr_unused
);
86 static struct kmem_cache
*inode_cachep __read_mostly
;
88 static int get_nr_inodes(void)
92 for_each_possible_cpu(i
)
93 sum
+= per_cpu(nr_inodes
, i
);
94 return sum
< 0 ? 0 : sum
;
97 static inline int get_nr_inodes_unused(void)
101 for_each_possible_cpu(i
)
102 sum
+= per_cpu(nr_unused
, i
);
103 return sum
< 0 ? 0 : sum
;
106 int get_nr_dirty_inodes(void)
108 /* not actually dirty inodes, but a wild approximation */
109 int nr_dirty
= get_nr_inodes() - get_nr_inodes_unused();
110 return nr_dirty
> 0 ? nr_dirty
: 0;
114 * Handle nr_inode sysctl
117 int proc_nr_inodes(ctl_table
*table
, int write
,
118 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
120 inodes_stat
.nr_inodes
= get_nr_inodes();
121 inodes_stat
.nr_unused
= get_nr_inodes_unused();
122 return proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
127 * inode_init_always - perform inode structure intialisation
128 * @sb: superblock inode belongs to
129 * @inode: inode to initialise
131 * These are initializations that need to be done on every inode
132 * allocation as the fields are not initialised by slab allocation.
134 int inode_init_always(struct super_block
*sb
, struct inode
*inode
)
136 static const struct inode_operations empty_iops
;
137 static const struct file_operations empty_fops
;
138 struct address_space
*const mapping
= &inode
->i_data
;
141 inode
->i_blkbits
= sb
->s_blocksize_bits
;
143 atomic_set(&inode
->i_count
, 1);
144 inode
->i_op
= &empty_iops
;
145 inode
->i_fop
= &empty_fops
;
146 inode
->__i_nlink
= 1;
147 inode
->i_opflags
= 0;
150 atomic_set(&inode
->i_writecount
, 0);
154 inode
->i_generation
= 0;
156 memset(&inode
->i_dquot
, 0, sizeof(inode
->i_dquot
));
158 inode
->i_pipe
= NULL
;
159 inode
->i_bdev
= NULL
;
160 inode
->i_cdev
= NULL
;
162 inode
->dirtied_when
= 0;
164 if (security_inode_alloc(inode
))
166 spin_lock_init(&inode
->i_lock
);
167 lockdep_set_class(&inode
->i_lock
, &sb
->s_type
->i_lock_key
);
169 mutex_init(&inode
->i_mutex
);
170 lockdep_set_class(&inode
->i_mutex
, &sb
->s_type
->i_mutex_key
);
172 atomic_set(&inode
->i_dio_count
, 0);
174 mapping
->a_ops
= &empty_aops
;
175 mapping
->host
= inode
;
177 mapping_set_gfp_mask(mapping
, GFP_HIGHUSER_MOVABLE
);
178 mapping
->assoc_mapping
= NULL
;
179 mapping
->backing_dev_info
= &default_backing_dev_info
;
180 mapping
->writeback_index
= 0;
183 * If the block_device provides a backing_dev_info for client
184 * inodes then use that. Otherwise the inode share the bdev's
188 struct backing_dev_info
*bdi
;
190 bdi
= sb
->s_bdev
->bd_inode
->i_mapping
->backing_dev_info
;
191 mapping
->backing_dev_info
= bdi
;
193 inode
->i_private
= NULL
;
194 inode
->i_mapping
= mapping
;
195 INIT_LIST_HEAD(&inode
->i_dentry
); /* buggered by rcu freeing */
196 #ifdef CONFIG_FS_POSIX_ACL
197 inode
->i_acl
= inode
->i_default_acl
= ACL_NOT_CACHED
;
200 #ifdef CONFIG_FSNOTIFY
201 inode
->i_fsnotify_mask
= 0;
204 this_cpu_inc(nr_inodes
);
210 EXPORT_SYMBOL(inode_init_always
);
212 static struct inode
*alloc_inode(struct super_block
*sb
)
216 if (sb
->s_op
->alloc_inode
)
217 inode
= sb
->s_op
->alloc_inode(sb
);
219 inode
= kmem_cache_alloc(inode_cachep
, GFP_KERNEL
);
224 if (unlikely(inode_init_always(sb
, inode
))) {
225 if (inode
->i_sb
->s_op
->destroy_inode
)
226 inode
->i_sb
->s_op
->destroy_inode(inode
);
228 kmem_cache_free(inode_cachep
, inode
);
235 void free_inode_nonrcu(struct inode
*inode
)
237 kmem_cache_free(inode_cachep
, inode
);
239 EXPORT_SYMBOL(free_inode_nonrcu
);
241 void __destroy_inode(struct inode
*inode
)
243 BUG_ON(inode_has_buffers(inode
));
244 security_inode_free(inode
);
245 fsnotify_inode_delete(inode
);
246 if (!inode
->i_nlink
) {
247 WARN_ON(atomic_long_read(&inode
->i_sb
->s_remove_count
) == 0);
248 atomic_long_dec(&inode
->i_sb
->s_remove_count
);
251 #ifdef CONFIG_FS_POSIX_ACL
252 if (inode
->i_acl
&& inode
->i_acl
!= ACL_NOT_CACHED
)
253 posix_acl_release(inode
->i_acl
);
254 if (inode
->i_default_acl
&& inode
->i_default_acl
!= ACL_NOT_CACHED
)
255 posix_acl_release(inode
->i_default_acl
);
257 this_cpu_dec(nr_inodes
);
259 EXPORT_SYMBOL(__destroy_inode
);
261 static void i_callback(struct rcu_head
*head
)
263 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
264 kmem_cache_free(inode_cachep
, inode
);
267 static void destroy_inode(struct inode
*inode
)
269 BUG_ON(!list_empty(&inode
->i_lru
));
270 __destroy_inode(inode
);
271 if (inode
->i_sb
->s_op
->destroy_inode
)
272 inode
->i_sb
->s_op
->destroy_inode(inode
);
274 call_rcu(&inode
->i_rcu
, i_callback
);
278 * drop_nlink - directly drop an inode's link count
281 * This is a low-level filesystem helper to replace any
282 * direct filesystem manipulation of i_nlink. In cases
283 * where we are attempting to track writes to the
284 * filesystem, a decrement to zero means an imminent
285 * write when the file is truncated and actually unlinked
288 void drop_nlink(struct inode
*inode
)
290 WARN_ON(inode
->i_nlink
== 0);
293 atomic_long_inc(&inode
->i_sb
->s_remove_count
);
295 EXPORT_SYMBOL(drop_nlink
);
298 * clear_nlink - directly zero an inode's link count
301 * This is a low-level filesystem helper to replace any
302 * direct filesystem manipulation of i_nlink. See
303 * drop_nlink() for why we care about i_nlink hitting zero.
305 void clear_nlink(struct inode
*inode
)
307 if (inode
->i_nlink
) {
308 inode
->__i_nlink
= 0;
309 atomic_long_inc(&inode
->i_sb
->s_remove_count
);
312 EXPORT_SYMBOL(clear_nlink
);
315 * set_nlink - directly set an inode's link count
317 * @nlink: new nlink (should be non-zero)
319 * This is a low-level filesystem helper to replace any
320 * direct filesystem manipulation of i_nlink.
322 void set_nlink(struct inode
*inode
, unsigned int nlink
)
325 printk_ratelimited(KERN_INFO
326 "set_nlink() clearing i_nlink on %s inode %li\n",
327 inode
->i_sb
->s_type
->name
, inode
->i_ino
);
330 /* Yes, some filesystems do change nlink from zero to one */
331 if (inode
->i_nlink
== 0)
332 atomic_long_dec(&inode
->i_sb
->s_remove_count
);
334 inode
->__i_nlink
= nlink
;
337 EXPORT_SYMBOL(set_nlink
);
340 * inc_nlink - directly increment an inode's link count
343 * This is a low-level filesystem helper to replace any
344 * direct filesystem manipulation of i_nlink. Currently,
345 * it is only here for parity with dec_nlink().
347 void inc_nlink(struct inode
*inode
)
349 if (WARN_ON(inode
->i_nlink
== 0))
350 atomic_long_dec(&inode
->i_sb
->s_remove_count
);
354 EXPORT_SYMBOL(inc_nlink
);
356 void address_space_init_once(struct address_space
*mapping
)
358 memset(mapping
, 0, sizeof(*mapping
));
359 INIT_RADIX_TREE(&mapping
->page_tree
, GFP_ATOMIC
);
360 spin_lock_init(&mapping
->tree_lock
);
361 mutex_init(&mapping
->i_mmap_mutex
);
362 INIT_LIST_HEAD(&mapping
->private_list
);
363 spin_lock_init(&mapping
->private_lock
);
364 INIT_RAW_PRIO_TREE_ROOT(&mapping
->i_mmap
);
365 INIT_LIST_HEAD(&mapping
->i_mmap_nonlinear
);
367 EXPORT_SYMBOL(address_space_init_once
);
370 * These are initializations that only need to be done
371 * once, because the fields are idempotent across use
372 * of the inode, so let the slab aware of that.
374 void inode_init_once(struct inode
*inode
)
376 memset(inode
, 0, sizeof(*inode
));
377 INIT_HLIST_NODE(&inode
->i_hash
);
378 INIT_LIST_HEAD(&inode
->i_devices
);
379 INIT_LIST_HEAD(&inode
->i_wb_list
);
380 INIT_LIST_HEAD(&inode
->i_lru
);
381 address_space_init_once(&inode
->i_data
);
382 i_size_ordered_init(inode
);
383 #ifdef CONFIG_FSNOTIFY
384 INIT_HLIST_HEAD(&inode
->i_fsnotify_marks
);
387 EXPORT_SYMBOL(inode_init_once
);
389 static void init_once(void *foo
)
391 struct inode
*inode
= (struct inode
*) foo
;
393 inode_init_once(inode
);
397 * inode->i_lock must be held
399 void __iget(struct inode
*inode
)
401 atomic_inc(&inode
->i_count
);
405 * get additional reference to inode; caller must already hold one.
407 void ihold(struct inode
*inode
)
409 WARN_ON(atomic_inc_return(&inode
->i_count
) < 2);
411 EXPORT_SYMBOL(ihold
);
413 static void inode_lru_list_add(struct inode
*inode
)
415 spin_lock(&inode
->i_sb
->s_inode_lru_lock
);
416 if (list_empty(&inode
->i_lru
)) {
417 list_add(&inode
->i_lru
, &inode
->i_sb
->s_inode_lru
);
418 inode
->i_sb
->s_nr_inodes_unused
++;
419 this_cpu_inc(nr_unused
);
421 spin_unlock(&inode
->i_sb
->s_inode_lru_lock
);
424 static void inode_lru_list_del(struct inode
*inode
)
426 spin_lock(&inode
->i_sb
->s_inode_lru_lock
);
427 if (!list_empty(&inode
->i_lru
)) {
428 list_del_init(&inode
->i_lru
);
429 inode
->i_sb
->s_nr_inodes_unused
--;
430 this_cpu_dec(nr_unused
);
432 spin_unlock(&inode
->i_sb
->s_inode_lru_lock
);
436 * inode_sb_list_add - add inode to the superblock list of inodes
437 * @inode: inode to add
439 void inode_sb_list_add(struct inode
*inode
)
441 spin_lock(&inode_sb_list_lock
);
442 list_add(&inode
->i_sb_list
, &inode
->i_sb
->s_inodes
);
443 spin_unlock(&inode_sb_list_lock
);
445 EXPORT_SYMBOL_GPL(inode_sb_list_add
);
447 static inline void inode_sb_list_del(struct inode
*inode
)
449 if (!list_empty(&inode
->i_sb_list
)) {
450 spin_lock(&inode_sb_list_lock
);
451 list_del_init(&inode
->i_sb_list
);
452 spin_unlock(&inode_sb_list_lock
);
456 static unsigned long hash(struct super_block
*sb
, unsigned long hashval
)
460 tmp
= (hashval
* (unsigned long)sb
) ^ (GOLDEN_RATIO_PRIME
+ hashval
) /
462 tmp
= tmp
^ ((tmp
^ GOLDEN_RATIO_PRIME
) >> i_hash_shift
);
463 return tmp
& i_hash_mask
;
467 * __insert_inode_hash - hash an inode
468 * @inode: unhashed inode
469 * @hashval: unsigned long value used to locate this object in the
472 * Add an inode to the inode hash for this superblock.
474 void __insert_inode_hash(struct inode
*inode
, unsigned long hashval
)
476 struct hlist_head
*b
= inode_hashtable
+ hash(inode
->i_sb
, hashval
);
478 spin_lock(&inode_hash_lock
);
479 spin_lock(&inode
->i_lock
);
480 hlist_add_head(&inode
->i_hash
, b
);
481 spin_unlock(&inode
->i_lock
);
482 spin_unlock(&inode_hash_lock
);
484 EXPORT_SYMBOL(__insert_inode_hash
);
487 * __remove_inode_hash - remove an inode from the hash
488 * @inode: inode to unhash
490 * Remove an inode from the superblock.
492 void __remove_inode_hash(struct inode
*inode
)
494 spin_lock(&inode_hash_lock
);
495 spin_lock(&inode
->i_lock
);
496 hlist_del_init(&inode
->i_hash
);
497 spin_unlock(&inode
->i_lock
);
498 spin_unlock(&inode_hash_lock
);
500 EXPORT_SYMBOL(__remove_inode_hash
);
502 void end_writeback(struct inode
*inode
)
506 * We have to cycle tree_lock here because reclaim can be still in the
507 * process of removing the last page (in __delete_from_page_cache())
508 * and we must not free mapping under it.
510 spin_lock_irq(&inode
->i_data
.tree_lock
);
511 BUG_ON(inode
->i_data
.nrpages
);
512 spin_unlock_irq(&inode
->i_data
.tree_lock
);
513 BUG_ON(!list_empty(&inode
->i_data
.private_list
));
514 BUG_ON(!(inode
->i_state
& I_FREEING
));
515 BUG_ON(inode
->i_state
& I_CLEAR
);
516 inode_sync_wait(inode
);
517 /* don't need i_lock here, no concurrent mods to i_state */
518 inode
->i_state
= I_FREEING
| I_CLEAR
;
520 EXPORT_SYMBOL(end_writeback
);
523 * Free the inode passed in, removing it from the lists it is still connected
524 * to. We remove any pages still attached to the inode and wait for any IO that
525 * is still in progress before finally destroying the inode.
527 * An inode must already be marked I_FREEING so that we avoid the inode being
528 * moved back onto lists if we race with other code that manipulates the lists
529 * (e.g. writeback_single_inode). The caller is responsible for setting this.
531 * An inode must already be removed from the LRU list before being evicted from
532 * the cache. This should occur atomically with setting the I_FREEING state
533 * flag, so no inodes here should ever be on the LRU when being evicted.
535 static void evict(struct inode
*inode
)
537 const struct super_operations
*op
= inode
->i_sb
->s_op
;
539 BUG_ON(!(inode
->i_state
& I_FREEING
));
540 BUG_ON(!list_empty(&inode
->i_lru
));
542 if (!list_empty(&inode
->i_wb_list
))
543 inode_wb_list_del(inode
);
545 inode_sb_list_del(inode
);
547 if (op
->evict_inode
) {
548 op
->evict_inode(inode
);
550 if (inode
->i_data
.nrpages
)
551 truncate_inode_pages(&inode
->i_data
, 0);
552 end_writeback(inode
);
554 if (S_ISBLK(inode
->i_mode
) && inode
->i_bdev
)
556 if (S_ISCHR(inode
->i_mode
) && inode
->i_cdev
)
559 remove_inode_hash(inode
);
561 spin_lock(&inode
->i_lock
);
562 wake_up_bit(&inode
->i_state
, __I_NEW
);
563 BUG_ON(inode
->i_state
!= (I_FREEING
| I_CLEAR
));
564 spin_unlock(&inode
->i_lock
);
566 destroy_inode(inode
);
570 * dispose_list - dispose of the contents of a local list
571 * @head: the head of the list to free
573 * Dispose-list gets a local list with local inodes in it, so it doesn't
574 * need to worry about list corruption and SMP locks.
576 static void dispose_list(struct list_head
*head
)
578 while (!list_empty(head
)) {
581 inode
= list_first_entry(head
, struct inode
, i_lru
);
582 list_del_init(&inode
->i_lru
);
589 * evict_inodes - evict all evictable inodes for a superblock
590 * @sb: superblock to operate on
592 * Make sure that no inodes with zero refcount are retained. This is
593 * called by superblock shutdown after having MS_ACTIVE flag removed,
594 * so any inode reaching zero refcount during or after that call will
595 * be immediately evicted.
597 void evict_inodes(struct super_block
*sb
)
599 struct inode
*inode
, *next
;
602 spin_lock(&inode_sb_list_lock
);
603 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
604 if (atomic_read(&inode
->i_count
))
607 spin_lock(&inode
->i_lock
);
608 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
609 spin_unlock(&inode
->i_lock
);
613 inode
->i_state
|= I_FREEING
;
614 inode_lru_list_del(inode
);
615 spin_unlock(&inode
->i_lock
);
616 list_add(&inode
->i_lru
, &dispose
);
618 spin_unlock(&inode_sb_list_lock
);
620 dispose_list(&dispose
);
624 * invalidate_inodes - attempt to free all inodes on a superblock
625 * @sb: superblock to operate on
626 * @kill_dirty: flag to guide handling of dirty inodes
628 * Attempts to free all inodes for a given superblock. If there were any
629 * busy inodes return a non-zero value, else zero.
630 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
633 int invalidate_inodes(struct super_block
*sb
, bool kill_dirty
)
636 struct inode
*inode
, *next
;
639 spin_lock(&inode_sb_list_lock
);
640 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
641 spin_lock(&inode
->i_lock
);
642 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
643 spin_unlock(&inode
->i_lock
);
646 if (inode
->i_state
& I_DIRTY
&& !kill_dirty
) {
647 spin_unlock(&inode
->i_lock
);
651 if (atomic_read(&inode
->i_count
)) {
652 spin_unlock(&inode
->i_lock
);
657 inode
->i_state
|= I_FREEING
;
658 inode_lru_list_del(inode
);
659 spin_unlock(&inode
->i_lock
);
660 list_add(&inode
->i_lru
, &dispose
);
662 spin_unlock(&inode_sb_list_lock
);
664 dispose_list(&dispose
);
669 static int can_unuse(struct inode
*inode
)
671 if (inode
->i_state
& ~I_REFERENCED
)
673 if (inode_has_buffers(inode
))
675 if (atomic_read(&inode
->i_count
))
677 if (inode
->i_data
.nrpages
)
683 * Walk the superblock inode LRU for freeable inodes and attempt to free them.
684 * This is called from the superblock shrinker function with a number of inodes
685 * to trim from the LRU. Inodes to be freed are moved to a temporary list and
686 * then are freed outside inode_lock by dispose_list().
688 * Any inodes which are pinned purely because of attached pagecache have their
689 * pagecache removed. If the inode has metadata buffers attached to
690 * mapping->private_list then try to remove them.
692 * If the inode has the I_REFERENCED flag set, then it means that it has been
693 * used recently - the flag is set in iput_final(). When we encounter such an
694 * inode, clear the flag and move it to the back of the LRU so it gets another
695 * pass through the LRU before it gets reclaimed. This is necessary because of
696 * the fact we are doing lazy LRU updates to minimise lock contention so the
697 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
698 * with this flag set because they are the inodes that are out of order.
700 void prune_icache_sb(struct super_block
*sb
, int nr_to_scan
)
704 unsigned long reap
= 0;
706 spin_lock(&sb
->s_inode_lru_lock
);
707 for (nr_scanned
= nr_to_scan
; nr_scanned
>= 0; nr_scanned
--) {
710 if (list_empty(&sb
->s_inode_lru
))
713 inode
= list_entry(sb
->s_inode_lru
.prev
, struct inode
, i_lru
);
716 * we are inverting the sb->s_inode_lru_lock/inode->i_lock here,
717 * so use a trylock. If we fail to get the lock, just move the
718 * inode to the back of the list so we don't spin on it.
720 if (!spin_trylock(&inode
->i_lock
)) {
721 list_move_tail(&inode
->i_lru
, &sb
->s_inode_lru
);
726 * Referenced or dirty inodes are still in use. Give them
727 * another pass through the LRU as we canot reclaim them now.
729 if (atomic_read(&inode
->i_count
) ||
730 (inode
->i_state
& ~I_REFERENCED
)) {
731 list_del_init(&inode
->i_lru
);
732 spin_unlock(&inode
->i_lock
);
733 sb
->s_nr_inodes_unused
--;
734 this_cpu_dec(nr_unused
);
738 /* recently referenced inodes get one more pass */
739 if (inode
->i_state
& I_REFERENCED
) {
740 inode
->i_state
&= ~I_REFERENCED
;
741 list_move(&inode
->i_lru
, &sb
->s_inode_lru
);
742 spin_unlock(&inode
->i_lock
);
745 if (inode_has_buffers(inode
) || inode
->i_data
.nrpages
) {
747 spin_unlock(&inode
->i_lock
);
748 spin_unlock(&sb
->s_inode_lru_lock
);
749 if (remove_inode_buffers(inode
))
750 reap
+= invalidate_mapping_pages(&inode
->i_data
,
753 spin_lock(&sb
->s_inode_lru_lock
);
755 if (inode
!= list_entry(sb
->s_inode_lru
.next
,
756 struct inode
, i_lru
))
757 continue; /* wrong inode or list_empty */
758 /* avoid lock inversions with trylock */
759 if (!spin_trylock(&inode
->i_lock
))
761 if (!can_unuse(inode
)) {
762 spin_unlock(&inode
->i_lock
);
766 WARN_ON(inode
->i_state
& I_NEW
);
767 inode
->i_state
|= I_FREEING
;
768 spin_unlock(&inode
->i_lock
);
770 list_move(&inode
->i_lru
, &freeable
);
771 sb
->s_nr_inodes_unused
--;
772 this_cpu_dec(nr_unused
);
774 if (current_is_kswapd())
775 __count_vm_events(KSWAPD_INODESTEAL
, reap
);
777 __count_vm_events(PGINODESTEAL
, reap
);
778 spin_unlock(&sb
->s_inode_lru_lock
);
780 dispose_list(&freeable
);
783 static void __wait_on_freeing_inode(struct inode
*inode
);
785 * Called with the inode lock held.
787 static struct inode
*find_inode(struct super_block
*sb
,
788 struct hlist_head
*head
,
789 int (*test
)(struct inode
*, void *),
792 struct hlist_node
*node
;
793 struct inode
*inode
= NULL
;
796 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
797 spin_lock(&inode
->i_lock
);
798 if (inode
->i_sb
!= sb
) {
799 spin_unlock(&inode
->i_lock
);
802 if (!test(inode
, data
)) {
803 spin_unlock(&inode
->i_lock
);
806 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
807 __wait_on_freeing_inode(inode
);
811 spin_unlock(&inode
->i_lock
);
818 * find_inode_fast is the fast path version of find_inode, see the comment at
819 * iget_locked for details.
821 static struct inode
*find_inode_fast(struct super_block
*sb
,
822 struct hlist_head
*head
, unsigned long ino
)
824 struct hlist_node
*node
;
825 struct inode
*inode
= NULL
;
828 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
829 spin_lock(&inode
->i_lock
);
830 if (inode
->i_ino
!= ino
) {
831 spin_unlock(&inode
->i_lock
);
834 if (inode
->i_sb
!= sb
) {
835 spin_unlock(&inode
->i_lock
);
838 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
839 __wait_on_freeing_inode(inode
);
843 spin_unlock(&inode
->i_lock
);
850 * Each cpu owns a range of LAST_INO_BATCH numbers.
851 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
852 * to renew the exhausted range.
854 * This does not significantly increase overflow rate because every CPU can
855 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
856 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
857 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
858 * overflow rate by 2x, which does not seem too significant.
860 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
861 * error if st_ino won't fit in target struct field. Use 32bit counter
862 * here to attempt to avoid that.
864 #define LAST_INO_BATCH 1024
865 static DEFINE_PER_CPU(unsigned int, last_ino
);
867 unsigned int get_next_ino(void)
869 unsigned int *p
= &get_cpu_var(last_ino
);
870 unsigned int res
= *p
;
873 if (unlikely((res
& (LAST_INO_BATCH
-1)) == 0)) {
874 static atomic_t shared_last_ino
;
875 int next
= atomic_add_return(LAST_INO_BATCH
, &shared_last_ino
);
877 res
= next
- LAST_INO_BATCH
;
882 put_cpu_var(last_ino
);
885 EXPORT_SYMBOL(get_next_ino
);
888 * new_inode_pseudo - obtain an inode
891 * Allocates a new inode for given superblock.
892 * Inode wont be chained in superblock s_inodes list
894 * - fs can't be unmount
895 * - quotas, fsnotify, writeback can't work
897 struct inode
*new_inode_pseudo(struct super_block
*sb
)
899 struct inode
*inode
= alloc_inode(sb
);
902 spin_lock(&inode
->i_lock
);
904 spin_unlock(&inode
->i_lock
);
905 INIT_LIST_HEAD(&inode
->i_sb_list
);
911 * new_inode - obtain an inode
914 * Allocates a new inode for given superblock. The default gfp_mask
915 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
916 * If HIGHMEM pages are unsuitable or it is known that pages allocated
917 * for the page cache are not reclaimable or migratable,
918 * mapping_set_gfp_mask() must be called with suitable flags on the
919 * newly created inode's mapping
922 struct inode
*new_inode(struct super_block
*sb
)
926 spin_lock_prefetch(&inode_sb_list_lock
);
928 inode
= new_inode_pseudo(sb
);
930 inode_sb_list_add(inode
);
933 EXPORT_SYMBOL(new_inode
);
935 #ifdef CONFIG_DEBUG_LOCK_ALLOC
936 void lockdep_annotate_inode_mutex_key(struct inode
*inode
)
938 if (S_ISDIR(inode
->i_mode
)) {
939 struct file_system_type
*type
= inode
->i_sb
->s_type
;
941 /* Set new key only if filesystem hasn't already changed it */
942 if (!lockdep_match_class(&inode
->i_mutex
,
943 &type
->i_mutex_key
)) {
945 * ensure nobody is actually holding i_mutex
947 mutex_destroy(&inode
->i_mutex
);
948 mutex_init(&inode
->i_mutex
);
949 lockdep_set_class(&inode
->i_mutex
,
950 &type
->i_mutex_dir_key
);
954 EXPORT_SYMBOL(lockdep_annotate_inode_mutex_key
);
958 * unlock_new_inode - clear the I_NEW state and wake up any waiters
959 * @inode: new inode to unlock
961 * Called when the inode is fully initialised to clear the new state of the
962 * inode and wake up anyone waiting for the inode to finish initialisation.
964 void unlock_new_inode(struct inode
*inode
)
966 lockdep_annotate_inode_mutex_key(inode
);
967 spin_lock(&inode
->i_lock
);
968 WARN_ON(!(inode
->i_state
& I_NEW
));
969 inode
->i_state
&= ~I_NEW
;
970 wake_up_bit(&inode
->i_state
, __I_NEW
);
971 spin_unlock(&inode
->i_lock
);
973 EXPORT_SYMBOL(unlock_new_inode
);
976 * iget5_locked - obtain an inode from a mounted file system
977 * @sb: super block of file system
978 * @hashval: hash value (usually inode number) to get
979 * @test: callback used for comparisons between inodes
980 * @set: callback used to initialize a new struct inode
981 * @data: opaque data pointer to pass to @test and @set
983 * Search for the inode specified by @hashval and @data in the inode cache,
984 * and if present it is return it with an increased reference count. This is
985 * a generalized version of iget_locked() for file systems where the inode
986 * number is not sufficient for unique identification of an inode.
988 * If the inode is not in cache, allocate a new inode and return it locked,
989 * hashed, and with the I_NEW flag set. The file system gets to fill it in
990 * before unlocking it via unlock_new_inode().
992 * Note both @test and @set are called with the inode_hash_lock held, so can't
995 struct inode
*iget5_locked(struct super_block
*sb
, unsigned long hashval
,
996 int (*test
)(struct inode
*, void *),
997 int (*set
)(struct inode
*, void *), void *data
)
999 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1000 struct inode
*inode
;
1002 spin_lock(&inode_hash_lock
);
1003 inode
= find_inode(sb
, head
, test
, data
);
1004 spin_unlock(&inode_hash_lock
);
1007 wait_on_inode(inode
);
1011 inode
= alloc_inode(sb
);
1015 spin_lock(&inode_hash_lock
);
1016 /* We released the lock, so.. */
1017 old
= find_inode(sb
, head
, test
, data
);
1019 if (set(inode
, data
))
1022 spin_lock(&inode
->i_lock
);
1023 inode
->i_state
= I_NEW
;
1024 hlist_add_head(&inode
->i_hash
, head
);
1025 spin_unlock(&inode
->i_lock
);
1026 inode_sb_list_add(inode
);
1027 spin_unlock(&inode_hash_lock
);
1029 /* Return the locked inode with I_NEW set, the
1030 * caller is responsible for filling in the contents
1036 * Uhhuh, somebody else created the same inode under
1037 * us. Use the old inode instead of the one we just
1040 spin_unlock(&inode_hash_lock
);
1041 destroy_inode(inode
);
1043 wait_on_inode(inode
);
1048 spin_unlock(&inode_hash_lock
);
1049 destroy_inode(inode
);
1052 EXPORT_SYMBOL(iget5_locked
);
1055 * iget_locked - obtain an inode from a mounted file system
1056 * @sb: super block of file system
1057 * @ino: inode number to get
1059 * Search for the inode specified by @ino in the inode cache and if present
1060 * return it with an increased reference count. This is for file systems
1061 * where the inode number is sufficient for unique identification of an inode.
1063 * If the inode is not in cache, allocate a new inode and return it locked,
1064 * hashed, and with the I_NEW flag set. The file system gets to fill it in
1065 * before unlocking it via unlock_new_inode().
1067 struct inode
*iget_locked(struct super_block
*sb
, unsigned long ino
)
1069 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1070 struct inode
*inode
;
1072 spin_lock(&inode_hash_lock
);
1073 inode
= find_inode_fast(sb
, head
, ino
);
1074 spin_unlock(&inode_hash_lock
);
1076 wait_on_inode(inode
);
1080 inode
= alloc_inode(sb
);
1084 spin_lock(&inode_hash_lock
);
1085 /* We released the lock, so.. */
1086 old
= find_inode_fast(sb
, head
, ino
);
1089 spin_lock(&inode
->i_lock
);
1090 inode
->i_state
= I_NEW
;
1091 hlist_add_head(&inode
->i_hash
, head
);
1092 spin_unlock(&inode
->i_lock
);
1093 inode_sb_list_add(inode
);
1094 spin_unlock(&inode_hash_lock
);
1096 /* Return the locked inode with I_NEW set, the
1097 * caller is responsible for filling in the contents
1103 * Uhhuh, somebody else created the same inode under
1104 * us. Use the old inode instead of the one we just
1107 spin_unlock(&inode_hash_lock
);
1108 destroy_inode(inode
);
1110 wait_on_inode(inode
);
1114 EXPORT_SYMBOL(iget_locked
);
1117 * search the inode cache for a matching inode number.
1118 * If we find one, then the inode number we are trying to
1119 * allocate is not unique and so we should not use it.
1121 * Returns 1 if the inode number is unique, 0 if it is not.
1123 static int test_inode_iunique(struct super_block
*sb
, unsigned long ino
)
1125 struct hlist_head
*b
= inode_hashtable
+ hash(sb
, ino
);
1126 struct hlist_node
*node
;
1127 struct inode
*inode
;
1129 spin_lock(&inode_hash_lock
);
1130 hlist_for_each_entry(inode
, node
, b
, i_hash
) {
1131 if (inode
->i_ino
== ino
&& inode
->i_sb
== sb
) {
1132 spin_unlock(&inode_hash_lock
);
1136 spin_unlock(&inode_hash_lock
);
1142 * iunique - get a unique inode number
1144 * @max_reserved: highest reserved inode number
1146 * Obtain an inode number that is unique on the system for a given
1147 * superblock. This is used by file systems that have no natural
1148 * permanent inode numbering system. An inode number is returned that
1149 * is higher than the reserved limit but unique.
1152 * With a large number of inodes live on the file system this function
1153 * currently becomes quite slow.
1155 ino_t
iunique(struct super_block
*sb
, ino_t max_reserved
)
1158 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1159 * error if st_ino won't fit in target struct field. Use 32bit counter
1160 * here to attempt to avoid that.
1162 static DEFINE_SPINLOCK(iunique_lock
);
1163 static unsigned int counter
;
1166 spin_lock(&iunique_lock
);
1168 if (counter
<= max_reserved
)
1169 counter
= max_reserved
+ 1;
1171 } while (!test_inode_iunique(sb
, res
));
1172 spin_unlock(&iunique_lock
);
1176 EXPORT_SYMBOL(iunique
);
1178 struct inode
*igrab(struct inode
*inode
)
1180 spin_lock(&inode
->i_lock
);
1181 if (!(inode
->i_state
& (I_FREEING
|I_WILL_FREE
))) {
1183 spin_unlock(&inode
->i_lock
);
1185 spin_unlock(&inode
->i_lock
);
1187 * Handle the case where s_op->clear_inode is not been
1188 * called yet, and somebody is calling igrab
1189 * while the inode is getting freed.
1195 EXPORT_SYMBOL(igrab
);
1198 * ilookup5_nowait - search for an inode in the inode cache
1199 * @sb: super block of file system to search
1200 * @hashval: hash value (usually inode number) to search for
1201 * @test: callback used for comparisons between inodes
1202 * @data: opaque data pointer to pass to @test
1204 * Search for the inode specified by @hashval and @data in the inode cache.
1205 * If the inode is in the cache, the inode is returned with an incremented
1208 * Note: I_NEW is not waited upon so you have to be very careful what you do
1209 * with the returned inode. You probably should be using ilookup5() instead.
1211 * Note2: @test is called with the inode_hash_lock held, so can't sleep.
1213 struct inode
*ilookup5_nowait(struct super_block
*sb
, unsigned long hashval
,
1214 int (*test
)(struct inode
*, void *), void *data
)
1216 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1217 struct inode
*inode
;
1219 spin_lock(&inode_hash_lock
);
1220 inode
= find_inode(sb
, head
, test
, data
);
1221 spin_unlock(&inode_hash_lock
);
1225 EXPORT_SYMBOL(ilookup5_nowait
);
1228 * ilookup5 - search for an inode in the inode cache
1229 * @sb: super block of file system to search
1230 * @hashval: hash value (usually inode number) to search for
1231 * @test: callback used for comparisons between inodes
1232 * @data: opaque data pointer to pass to @test
1234 * Search for the inode specified by @hashval and @data in the inode cache,
1235 * and if the inode is in the cache, return the inode with an incremented
1236 * reference count. Waits on I_NEW before returning the inode.
1237 * returned with an incremented reference count.
1239 * This is a generalized version of ilookup() for file systems where the
1240 * inode number is not sufficient for unique identification of an inode.
1242 * Note: @test is called with the inode_hash_lock held, so can't sleep.
1244 struct inode
*ilookup5(struct super_block
*sb
, unsigned long hashval
,
1245 int (*test
)(struct inode
*, void *), void *data
)
1247 struct inode
*inode
= ilookup5_nowait(sb
, hashval
, test
, data
);
1250 wait_on_inode(inode
);
1253 EXPORT_SYMBOL(ilookup5
);
1256 * ilookup - search for an inode in the inode cache
1257 * @sb: super block of file system to search
1258 * @ino: inode number to search for
1260 * Search for the inode @ino in the inode cache, and if the inode is in the
1261 * cache, the inode is returned with an incremented reference count.
1263 struct inode
*ilookup(struct super_block
*sb
, unsigned long ino
)
1265 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1266 struct inode
*inode
;
1268 spin_lock(&inode_hash_lock
);
1269 inode
= find_inode_fast(sb
, head
, ino
);
1270 spin_unlock(&inode_hash_lock
);
1273 wait_on_inode(inode
);
1276 EXPORT_SYMBOL(ilookup
);
1278 int insert_inode_locked(struct inode
*inode
)
1280 struct super_block
*sb
= inode
->i_sb
;
1281 ino_t ino
= inode
->i_ino
;
1282 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1285 struct hlist_node
*node
;
1286 struct inode
*old
= NULL
;
1287 spin_lock(&inode_hash_lock
);
1288 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1289 if (old
->i_ino
!= ino
)
1291 if (old
->i_sb
!= sb
)
1293 spin_lock(&old
->i_lock
);
1294 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
1295 spin_unlock(&old
->i_lock
);
1300 if (likely(!node
)) {
1301 spin_lock(&inode
->i_lock
);
1302 inode
->i_state
|= I_NEW
;
1303 hlist_add_head(&inode
->i_hash
, head
);
1304 spin_unlock(&inode
->i_lock
);
1305 spin_unlock(&inode_hash_lock
);
1309 spin_unlock(&old
->i_lock
);
1310 spin_unlock(&inode_hash_lock
);
1312 if (unlikely(!inode_unhashed(old
))) {
1319 EXPORT_SYMBOL(insert_inode_locked
);
1321 int insert_inode_locked4(struct inode
*inode
, unsigned long hashval
,
1322 int (*test
)(struct inode
*, void *), void *data
)
1324 struct super_block
*sb
= inode
->i_sb
;
1325 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1328 struct hlist_node
*node
;
1329 struct inode
*old
= NULL
;
1331 spin_lock(&inode_hash_lock
);
1332 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1333 if (old
->i_sb
!= sb
)
1335 if (!test(old
, data
))
1337 spin_lock(&old
->i_lock
);
1338 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
1339 spin_unlock(&old
->i_lock
);
1344 if (likely(!node
)) {
1345 spin_lock(&inode
->i_lock
);
1346 inode
->i_state
|= I_NEW
;
1347 hlist_add_head(&inode
->i_hash
, head
);
1348 spin_unlock(&inode
->i_lock
);
1349 spin_unlock(&inode_hash_lock
);
1353 spin_unlock(&old
->i_lock
);
1354 spin_unlock(&inode_hash_lock
);
1356 if (unlikely(!inode_unhashed(old
))) {
1363 EXPORT_SYMBOL(insert_inode_locked4
);
1366 int generic_delete_inode(struct inode
*inode
)
1370 EXPORT_SYMBOL(generic_delete_inode
);
1373 * Normal UNIX filesystem behaviour: delete the
1374 * inode when the usage count drops to zero, and
1377 int generic_drop_inode(struct inode
*inode
)
1379 return !inode
->i_nlink
|| inode_unhashed(inode
);
1381 EXPORT_SYMBOL_GPL(generic_drop_inode
);
1384 * Called when we're dropping the last reference
1387 * Call the FS "drop_inode()" function, defaulting to
1388 * the legacy UNIX filesystem behaviour. If it tells
1389 * us to evict inode, do so. Otherwise, retain inode
1390 * in cache if fs is alive, sync and evict if fs is
1393 static void iput_final(struct inode
*inode
)
1395 struct super_block
*sb
= inode
->i_sb
;
1396 const struct super_operations
*op
= inode
->i_sb
->s_op
;
1399 WARN_ON(inode
->i_state
& I_NEW
);
1402 drop
= op
->drop_inode(inode
);
1404 drop
= generic_drop_inode(inode
);
1406 if (!drop
&& (sb
->s_flags
& MS_ACTIVE
)) {
1407 inode
->i_state
|= I_REFERENCED
;
1408 if (!(inode
->i_state
& (I_DIRTY
|I_SYNC
)))
1409 inode_lru_list_add(inode
);
1410 spin_unlock(&inode
->i_lock
);
1415 inode
->i_state
|= I_WILL_FREE
;
1416 spin_unlock(&inode
->i_lock
);
1417 write_inode_now(inode
, 1);
1418 spin_lock(&inode
->i_lock
);
1419 WARN_ON(inode
->i_state
& I_NEW
);
1420 inode
->i_state
&= ~I_WILL_FREE
;
1423 inode
->i_state
|= I_FREEING
;
1424 if (!list_empty(&inode
->i_lru
))
1425 inode_lru_list_del(inode
);
1426 spin_unlock(&inode
->i_lock
);
1432 * iput - put an inode
1433 * @inode: inode to put
1435 * Puts an inode, dropping its usage count. If the inode use count hits
1436 * zero, the inode is then freed and may also be destroyed.
1438 * Consequently, iput() can sleep.
1440 void iput(struct inode
*inode
)
1443 BUG_ON(inode
->i_state
& I_CLEAR
);
1445 if (atomic_dec_and_lock(&inode
->i_count
, &inode
->i_lock
))
1449 EXPORT_SYMBOL(iput
);
1452 * bmap - find a block number in a file
1453 * @inode: inode of file
1454 * @block: block to find
1456 * Returns the block number on the device holding the inode that
1457 * is the disk block number for the block of the file requested.
1458 * That is, asked for block 4 of inode 1 the function will return the
1459 * disk block relative to the disk start that holds that block of the
1462 sector_t
bmap(struct inode
*inode
, sector_t block
)
1465 if (inode
->i_mapping
->a_ops
->bmap
)
1466 res
= inode
->i_mapping
->a_ops
->bmap(inode
->i_mapping
, block
);
1469 EXPORT_SYMBOL(bmap
);
1472 * With relative atime, only update atime if the previous atime is
1473 * earlier than either the ctime or mtime or if at least a day has
1474 * passed since the last atime update.
1476 static int relatime_need_update(struct vfsmount
*mnt
, struct inode
*inode
,
1477 struct timespec now
)
1480 if (!(mnt
->mnt_flags
& MNT_RELATIME
))
1483 * Is mtime younger than atime? If yes, update atime:
1485 if (timespec_compare(&inode
->i_mtime
, &inode
->i_atime
) >= 0)
1488 * Is ctime younger than atime? If yes, update atime:
1490 if (timespec_compare(&inode
->i_ctime
, &inode
->i_atime
) >= 0)
1494 * Is the previous atime value older than a day? If yes,
1497 if ((long)(now
.tv_sec
- inode
->i_atime
.tv_sec
) >= 24*60*60)
1500 * Good, we can skip the atime update:
1506 * touch_atime - update the access time
1507 * @mnt: mount the inode is accessed on
1508 * @dentry: dentry accessed
1510 * Update the accessed time on an inode and mark it for writeback.
1511 * This function automatically handles read only file systems and media,
1512 * as well as the "noatime" flag and inode specific "noatime" markers.
1514 void touch_atime(struct vfsmount
*mnt
, struct dentry
*dentry
)
1516 struct inode
*inode
= dentry
->d_inode
;
1517 struct timespec now
;
1519 if (inode
->i_flags
& S_NOATIME
)
1521 if (IS_NOATIME(inode
))
1523 if ((inode
->i_sb
->s_flags
& MS_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1526 if (mnt
->mnt_flags
& MNT_NOATIME
)
1528 if ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1531 now
= current_fs_time(inode
->i_sb
);
1533 if (!relatime_need_update(mnt
, inode
, now
))
1536 if (timespec_equal(&inode
->i_atime
, &now
))
1539 if (mnt_want_write(mnt
))
1542 inode
->i_atime
= now
;
1543 mark_inode_dirty_sync(inode
);
1544 mnt_drop_write(mnt
);
1546 EXPORT_SYMBOL(touch_atime
);
1549 * file_update_time - update mtime and ctime time
1550 * @file: file accessed
1552 * Update the mtime and ctime members of an inode and mark the inode
1553 * for writeback. Note that this function is meant exclusively for
1554 * usage in the file write path of filesystems, and filesystems may
1555 * choose to explicitly ignore update via this function with the
1556 * S_NOCMTIME inode flag, e.g. for network filesystem where these
1557 * timestamps are handled by the server.
1560 void file_update_time(struct file
*file
)
1562 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1563 struct timespec now
;
1564 enum { S_MTIME
= 1, S_CTIME
= 2, S_VERSION
= 4 } sync_it
= 0;
1566 /* First try to exhaust all avenues to not sync */
1567 if (IS_NOCMTIME(inode
))
1570 now
= current_fs_time(inode
->i_sb
);
1571 if (!timespec_equal(&inode
->i_mtime
, &now
))
1574 if (!timespec_equal(&inode
->i_ctime
, &now
))
1577 if (IS_I_VERSION(inode
))
1578 sync_it
|= S_VERSION
;
1583 /* Finally allowed to write? Takes lock. */
1584 if (mnt_want_write_file(file
))
1587 /* Only change inode inside the lock region */
1588 if (sync_it
& S_VERSION
)
1589 inode_inc_iversion(inode
);
1590 if (sync_it
& S_CTIME
)
1591 inode
->i_ctime
= now
;
1592 if (sync_it
& S_MTIME
)
1593 inode
->i_mtime
= now
;
1594 mark_inode_dirty_sync(inode
);
1595 mnt_drop_write_file(file
);
1597 EXPORT_SYMBOL(file_update_time
);
1599 int inode_needs_sync(struct inode
*inode
)
1603 if (S_ISDIR(inode
->i_mode
) && IS_DIRSYNC(inode
))
1607 EXPORT_SYMBOL(inode_needs_sync
);
1609 int inode_wait(void *word
)
1614 EXPORT_SYMBOL(inode_wait
);
1617 * If we try to find an inode in the inode hash while it is being
1618 * deleted, we have to wait until the filesystem completes its
1619 * deletion before reporting that it isn't found. This function waits
1620 * until the deletion _might_ have completed. Callers are responsible
1621 * to recheck inode state.
1623 * It doesn't matter if I_NEW is not set initially, a call to
1624 * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
1627 static void __wait_on_freeing_inode(struct inode
*inode
)
1629 wait_queue_head_t
*wq
;
1630 DEFINE_WAIT_BIT(wait
, &inode
->i_state
, __I_NEW
);
1631 wq
= bit_waitqueue(&inode
->i_state
, __I_NEW
);
1632 prepare_to_wait(wq
, &wait
.wait
, TASK_UNINTERRUPTIBLE
);
1633 spin_unlock(&inode
->i_lock
);
1634 spin_unlock(&inode_hash_lock
);
1636 finish_wait(wq
, &wait
.wait
);
1637 spin_lock(&inode_hash_lock
);
1640 static __initdata
unsigned long ihash_entries
;
1641 static int __init
set_ihash_entries(char *str
)
1645 ihash_entries
= simple_strtoul(str
, &str
, 0);
1648 __setup("ihash_entries=", set_ihash_entries
);
1651 * Initialize the waitqueues and inode hash table.
1653 void __init
inode_init_early(void)
1657 /* If hashes are distributed across NUMA nodes, defer
1658 * hash allocation until vmalloc space is available.
1664 alloc_large_system_hash("Inode-cache",
1665 sizeof(struct hlist_head
),
1673 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1674 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1677 void __init
inode_init(void)
1681 /* inode slab cache */
1682 inode_cachep
= kmem_cache_create("inode_cache",
1683 sizeof(struct inode
),
1685 (SLAB_RECLAIM_ACCOUNT
|SLAB_PANIC
|
1689 /* Hash may have been set up in inode_init_early */
1694 alloc_large_system_hash("Inode-cache",
1695 sizeof(struct hlist_head
),
1703 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1704 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1707 void init_special_inode(struct inode
*inode
, umode_t mode
, dev_t rdev
)
1709 inode
->i_mode
= mode
;
1710 if (S_ISCHR(mode
)) {
1711 inode
->i_fop
= &def_chr_fops
;
1712 inode
->i_rdev
= rdev
;
1713 } else if (S_ISBLK(mode
)) {
1714 inode
->i_fop
= &def_blk_fops
;
1715 inode
->i_rdev
= rdev
;
1716 } else if (S_ISFIFO(mode
))
1717 inode
->i_fop
= &def_fifo_fops
;
1718 else if (S_ISSOCK(mode
))
1719 inode
->i_fop
= &bad_sock_fops
;
1721 printk(KERN_DEBUG
"init_special_inode: bogus i_mode (%o) for"
1722 " inode %s:%lu\n", mode
, inode
->i_sb
->s_id
,
1725 EXPORT_SYMBOL(init_special_inode
);
1728 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
1730 * @dir: Directory inode
1731 * @mode: mode of the new inode
1733 void inode_init_owner(struct inode
*inode
, const struct inode
*dir
,
1736 inode
->i_uid
= current_fsuid();
1737 if (dir
&& dir
->i_mode
& S_ISGID
) {
1738 inode
->i_gid
= dir
->i_gid
;
1742 inode
->i_gid
= current_fsgid();
1743 inode
->i_mode
= mode
;
1745 EXPORT_SYMBOL(inode_init_owner
);
1748 * inode_owner_or_capable - check current task permissions to inode
1749 * @inode: inode being checked
1751 * Return true if current either has CAP_FOWNER to the inode, or
1754 bool inode_owner_or_capable(const struct inode
*inode
)
1756 struct user_namespace
*ns
= inode_userns(inode
);
1758 if (current_user_ns() == ns
&& current_fsuid() == inode
->i_uid
)
1760 if (ns_capable(ns
, CAP_FOWNER
))
1764 EXPORT_SYMBOL(inode_owner_or_capable
);