4 * (C) 1997 Linus Torvalds
9 #include <linux/dcache.h>
10 #include <linux/init.h>
11 #include <linux/quotaops.h>
12 #include <linux/slab.h>
13 #include <linux/writeback.h>
14 #include <linux/module.h>
15 #include <linux/backing-dev.h>
16 #include <linux/wait.h>
17 #include <linux/hash.h>
18 #include <linux/swap.h>
19 #include <linux/security.h>
20 #include <linux/pagemap.h>
21 #include <linux/cdev.h>
22 #include <linux/bootmem.h>
23 #include <linux/inotify.h>
24 #include <linux/mount.h>
27 * This is needed for the following functions:
29 * - invalidate_inode_buffers
32 * FIXME: remove all knowledge of the buffer layer from this file
34 #include <linux/buffer_head.h>
37 * New inode.c implementation.
39 * This implementation has the basic premise of trying
40 * to be extremely low-overhead and SMP-safe, yet be
41 * simple enough to be "obviously correct".
46 /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */
48 /* #define INODE_PARANOIA 1 */
49 /* #define INODE_DEBUG 1 */
52 * Inode lookup is no longer as critical as it used to be:
53 * most of the lookups are going to be through the dcache.
55 #define I_HASHBITS i_hash_shift
56 #define I_HASHMASK i_hash_mask
58 static unsigned int i_hash_mask __read_mostly
;
59 static unsigned int i_hash_shift __read_mostly
;
62 * Each inode can be on two separate lists. One is
63 * the hash list of the inode, used for lookups. The
64 * other linked list is the "type" list:
65 * "in_use" - valid inode, i_count > 0, i_nlink > 0
66 * "dirty" - as "in_use" but also dirty
67 * "unused" - valid inode, i_count = 0
69 * A "dirty" list is maintained for each super block,
70 * allowing for low-overhead inode sync() operations.
73 LIST_HEAD(inode_in_use
);
74 LIST_HEAD(inode_unused
);
75 static struct hlist_head
*inode_hashtable __read_mostly
;
78 * A simple spinlock to protect the list manipulations.
80 * NOTE! You also have to own the lock if you change
81 * the i_state of an inode while it is in use..
83 DEFINE_SPINLOCK(inode_lock
);
86 * iprune_mutex provides exclusion between the kswapd or try_to_free_pages
87 * icache shrinking path, and the umount path. Without this exclusion,
88 * by the time prune_icache calls iput for the inode whose pages it has
89 * been invalidating, or by the time it calls clear_inode & destroy_inode
90 * from its final dispose_list, the struct super_block they refer to
91 * (for inode->i_sb->s_op) may already have been freed and reused.
93 static DEFINE_MUTEX(iprune_mutex
);
96 * Statistics gathering..
98 struct inodes_stat_t inodes_stat
;
100 static kmem_cache_t
* inode_cachep __read_mostly
;
102 static struct inode
*alloc_inode(struct super_block
*sb
)
104 static const struct address_space_operations empty_aops
;
105 static struct inode_operations empty_iops
;
106 static const struct file_operations empty_fops
;
109 if (sb
->s_op
->alloc_inode
)
110 inode
= sb
->s_op
->alloc_inode(sb
);
112 inode
= (struct inode
*) kmem_cache_alloc(inode_cachep
, SLAB_KERNEL
);
115 struct address_space
* const mapping
= &inode
->i_data
;
118 inode
->i_blkbits
= sb
->s_blocksize_bits
;
120 atomic_set(&inode
->i_count
, 1);
121 inode
->i_op
= &empty_iops
;
122 inode
->i_fop
= &empty_fops
;
124 atomic_set(&inode
->i_writecount
, 0);
128 inode
->i_generation
= 0;
130 memset(&inode
->i_dquot
, 0, sizeof(inode
->i_dquot
));
132 inode
->i_pipe
= NULL
;
133 inode
->i_bdev
= NULL
;
134 inode
->i_cdev
= NULL
;
136 inode
->i_security
= NULL
;
137 inode
->dirtied_when
= 0;
138 if (security_inode_alloc(inode
)) {
139 if (inode
->i_sb
->s_op
->destroy_inode
)
140 inode
->i_sb
->s_op
->destroy_inode(inode
);
142 kmem_cache_free(inode_cachep
, (inode
));
146 mapping
->a_ops
= &empty_aops
;
147 mapping
->host
= inode
;
149 mapping_set_gfp_mask(mapping
, GFP_HIGHUSER
);
150 mapping
->assoc_mapping
= NULL
;
151 mapping
->backing_dev_info
= &default_backing_dev_info
;
154 * If the block_device provides a backing_dev_info for client
155 * inodes then use that. Otherwise the inode share the bdev's
159 struct backing_dev_info
*bdi
;
161 bdi
= sb
->s_bdev
->bd_inode_backing_dev_info
;
163 bdi
= sb
->s_bdev
->bd_inode
->i_mapping
->backing_dev_info
;
164 mapping
->backing_dev_info
= bdi
;
166 memset(&inode
->u
, 0, sizeof(inode
->u
));
167 inode
->i_mapping
= mapping
;
172 void destroy_inode(struct inode
*inode
)
174 BUG_ON(inode_has_buffers(inode
));
175 security_inode_free(inode
);
176 if (inode
->i_sb
->s_op
->destroy_inode
)
177 inode
->i_sb
->s_op
->destroy_inode(inode
);
179 kmem_cache_free(inode_cachep
, (inode
));
184 * These are initializations that only need to be done
185 * once, because the fields are idempotent across use
186 * of the inode, so let the slab aware of that.
188 void inode_init_once(struct inode
*inode
)
190 memset(inode
, 0, sizeof(*inode
));
191 INIT_HLIST_NODE(&inode
->i_hash
);
192 INIT_LIST_HEAD(&inode
->i_dentry
);
193 INIT_LIST_HEAD(&inode
->i_devices
);
194 mutex_init(&inode
->i_mutex
);
195 init_rwsem(&inode
->i_alloc_sem
);
196 INIT_RADIX_TREE(&inode
->i_data
.page_tree
, GFP_ATOMIC
);
197 rwlock_init(&inode
->i_data
.tree_lock
);
198 spin_lock_init(&inode
->i_data
.i_mmap_lock
);
199 INIT_LIST_HEAD(&inode
->i_data
.private_list
);
200 spin_lock_init(&inode
->i_data
.private_lock
);
201 INIT_RAW_PRIO_TREE_ROOT(&inode
->i_data
.i_mmap
);
202 INIT_LIST_HEAD(&inode
->i_data
.i_mmap_nonlinear
);
203 spin_lock_init(&inode
->i_lock
);
204 i_size_ordered_init(inode
);
205 #ifdef CONFIG_INOTIFY
206 INIT_LIST_HEAD(&inode
->inotify_watches
);
207 mutex_init(&inode
->inotify_mutex
);
211 EXPORT_SYMBOL(inode_init_once
);
213 static void init_once(void * foo
, kmem_cache_t
* cachep
, unsigned long flags
)
215 struct inode
* inode
= (struct inode
*) foo
;
217 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
)) ==
218 SLAB_CTOR_CONSTRUCTOR
)
219 inode_init_once(inode
);
223 * inode_lock must be held
225 void __iget(struct inode
* inode
)
227 if (atomic_read(&inode
->i_count
)) {
228 atomic_inc(&inode
->i_count
);
231 atomic_inc(&inode
->i_count
);
232 if (!(inode
->i_state
& (I_DIRTY
|I_LOCK
)))
233 list_move(&inode
->i_list
, &inode_in_use
);
234 inodes_stat
.nr_unused
--;
238 * clear_inode - clear an inode
239 * @inode: inode to clear
241 * This is called by the filesystem to tell us
242 * that the inode is no longer useful. We just
243 * terminate it with extreme prejudice.
245 void clear_inode(struct inode
*inode
)
248 invalidate_inode_buffers(inode
);
250 BUG_ON(inode
->i_data
.nrpages
);
251 BUG_ON(!(inode
->i_state
& I_FREEING
));
252 BUG_ON(inode
->i_state
& I_CLEAR
);
253 wait_on_inode(inode
);
255 if (inode
->i_sb
&& inode
->i_sb
->s_op
->clear_inode
)
256 inode
->i_sb
->s_op
->clear_inode(inode
);
261 inode
->i_state
= I_CLEAR
;
264 EXPORT_SYMBOL(clear_inode
);
267 * dispose_list - dispose of the contents of a local list
268 * @head: the head of the list to free
270 * Dispose-list gets a local list with local inodes in it, so it doesn't
271 * need to worry about list corruption and SMP locks.
273 static void dispose_list(struct list_head
*head
)
277 while (!list_empty(head
)) {
280 inode
= list_entry(head
->next
, struct inode
, i_list
);
281 list_del(&inode
->i_list
);
283 if (inode
->i_data
.nrpages
)
284 truncate_inode_pages(&inode
->i_data
, 0);
287 spin_lock(&inode_lock
);
288 hlist_del_init(&inode
->i_hash
);
289 list_del_init(&inode
->i_sb_list
);
290 spin_unlock(&inode_lock
);
292 wake_up_inode(inode
);
293 destroy_inode(inode
);
296 spin_lock(&inode_lock
);
297 inodes_stat
.nr_inodes
-= nr_disposed
;
298 spin_unlock(&inode_lock
);
302 * Invalidate all inodes for a device.
304 static int invalidate_list(struct list_head
*head
, struct list_head
*dispose
)
306 struct list_head
*next
;
307 int busy
= 0, count
= 0;
311 struct list_head
* tmp
= next
;
312 struct inode
* inode
;
315 * We can reschedule here without worrying about the list's
316 * consistency because the per-sb list of inodes must not
317 * change during umount anymore, and because iprune_mutex keeps
318 * shrink_icache_memory() away.
320 cond_resched_lock(&inode_lock
);
325 inode
= list_entry(tmp
, struct inode
, i_sb_list
);
326 invalidate_inode_buffers(inode
);
327 if (!atomic_read(&inode
->i_count
)) {
328 list_move(&inode
->i_list
, dispose
);
329 inode
->i_state
|= I_FREEING
;
335 /* only unused inodes may be cached with i_count zero */
336 inodes_stat
.nr_unused
-= count
;
341 * invalidate_inodes - discard the inodes on a device
344 * Discard all of the inodes for a given superblock. If the discard
345 * fails because there are busy inodes then a non zero value is returned.
346 * If the discard is successful all the inodes have been discarded.
348 int invalidate_inodes(struct super_block
* sb
)
351 LIST_HEAD(throw_away
);
353 mutex_lock(&iprune_mutex
);
354 spin_lock(&inode_lock
);
355 inotify_unmount_inodes(&sb
->s_inodes
);
356 busy
= invalidate_list(&sb
->s_inodes
, &throw_away
);
357 spin_unlock(&inode_lock
);
359 dispose_list(&throw_away
);
360 mutex_unlock(&iprune_mutex
);
365 EXPORT_SYMBOL(invalidate_inodes
);
367 int __invalidate_device(struct block_device
*bdev
)
369 struct super_block
*sb
= get_super(bdev
);
374 * no need to lock the super, get_super holds the
375 * read mutex so the filesystem cannot go away
376 * under us (->put_super runs with the write lock
379 shrink_dcache_sb(sb
);
380 res
= invalidate_inodes(sb
);
383 invalidate_bdev(bdev
, 0);
386 EXPORT_SYMBOL(__invalidate_device
);
388 static int can_unuse(struct inode
*inode
)
392 if (inode_has_buffers(inode
))
394 if (atomic_read(&inode
->i_count
))
396 if (inode
->i_data
.nrpages
)
402 * Scan `goal' inodes on the unused list for freeable ones. They are moved to
403 * a temporary list and then are freed outside inode_lock by dispose_list().
405 * Any inodes which are pinned purely because of attached pagecache have their
406 * pagecache removed. We expect the final iput() on that inode to add it to
407 * the front of the inode_unused list. So look for it there and if the
408 * inode is still freeable, proceed. The right inode is found 99.9% of the
409 * time in testing on a 4-way.
411 * If the inode has metadata buffers attached to mapping->private_list then
412 * try to remove them.
414 static void prune_icache(int nr_to_scan
)
419 unsigned long reap
= 0;
421 mutex_lock(&iprune_mutex
);
422 spin_lock(&inode_lock
);
423 for (nr_scanned
= 0; nr_scanned
< nr_to_scan
; nr_scanned
++) {
426 if (list_empty(&inode_unused
))
429 inode
= list_entry(inode_unused
.prev
, struct inode
, i_list
);
431 if (inode
->i_state
|| atomic_read(&inode
->i_count
)) {
432 list_move(&inode
->i_list
, &inode_unused
);
435 if (inode_has_buffers(inode
) || inode
->i_data
.nrpages
) {
437 spin_unlock(&inode_lock
);
438 if (remove_inode_buffers(inode
))
439 reap
+= invalidate_inode_pages(&inode
->i_data
);
441 spin_lock(&inode_lock
);
443 if (inode
!= list_entry(inode_unused
.next
,
444 struct inode
, i_list
))
445 continue; /* wrong inode or list_empty */
446 if (!can_unuse(inode
))
449 list_move(&inode
->i_list
, &freeable
);
450 inode
->i_state
|= I_FREEING
;
453 inodes_stat
.nr_unused
-= nr_pruned
;
454 spin_unlock(&inode_lock
);
456 dispose_list(&freeable
);
457 mutex_unlock(&iprune_mutex
);
459 if (current_is_kswapd())
460 mod_page_state(kswapd_inodesteal
, reap
);
462 mod_page_state(pginodesteal
, reap
);
466 * shrink_icache_memory() will attempt to reclaim some unused inodes. Here,
467 * "unused" means that no dentries are referring to the inodes: the files are
468 * not open and the dcache references to those inodes have already been
471 * This function is passed the number of inodes to scan, and it returns the
472 * total number of remaining possibly-reclaimable inodes.
474 static int shrink_icache_memory(int nr
, gfp_t gfp_mask
)
478 * Nasty deadlock avoidance. We may hold various FS locks,
479 * and we don't want to recurse into the FS that called us
480 * in clear_inode() and friends..
482 if (!(gfp_mask
& __GFP_FS
))
486 return (inodes_stat
.nr_unused
/ 100) * sysctl_vfs_cache_pressure
;
489 static void __wait_on_freeing_inode(struct inode
*inode
);
491 * Called with the inode lock held.
492 * NOTE: we are not increasing the inode-refcount, you must call __iget()
493 * by hand after calling find_inode now! This simplifies iunique and won't
494 * add any additional branch in the common code.
496 static struct inode
* find_inode(struct super_block
* sb
, struct hlist_head
*head
, int (*test
)(struct inode
*, void *), void *data
)
498 struct hlist_node
*node
;
499 struct inode
* inode
= NULL
;
502 hlist_for_each (node
, head
) {
503 inode
= hlist_entry(node
, struct inode
, i_hash
);
504 if (inode
->i_sb
!= sb
)
506 if (!test(inode
, data
))
508 if (inode
->i_state
& (I_FREEING
|I_CLEAR
|I_WILL_FREE
)) {
509 __wait_on_freeing_inode(inode
);
514 return node
? inode
: NULL
;
518 * find_inode_fast is the fast path version of find_inode, see the comment at
519 * iget_locked for details.
521 static struct inode
* find_inode_fast(struct super_block
* sb
, struct hlist_head
*head
, unsigned long ino
)
523 struct hlist_node
*node
;
524 struct inode
* inode
= NULL
;
527 hlist_for_each (node
, head
) {
528 inode
= hlist_entry(node
, struct inode
, i_hash
);
529 if (inode
->i_ino
!= ino
)
531 if (inode
->i_sb
!= sb
)
533 if (inode
->i_state
& (I_FREEING
|I_CLEAR
|I_WILL_FREE
)) {
534 __wait_on_freeing_inode(inode
);
539 return node
? inode
: NULL
;
543 * new_inode - obtain an inode
546 * Allocates a new inode for given superblock.
548 struct inode
*new_inode(struct super_block
*sb
)
550 static unsigned long last_ino
;
551 struct inode
* inode
;
553 spin_lock_prefetch(&inode_lock
);
555 inode
= alloc_inode(sb
);
557 spin_lock(&inode_lock
);
558 inodes_stat
.nr_inodes
++;
559 list_add(&inode
->i_list
, &inode_in_use
);
560 list_add(&inode
->i_sb_list
, &sb
->s_inodes
);
561 inode
->i_ino
= ++last_ino
;
563 spin_unlock(&inode_lock
);
568 EXPORT_SYMBOL(new_inode
);
570 void unlock_new_inode(struct inode
*inode
)
573 * This is special! We do not need the spinlock
574 * when clearing I_LOCK, because we're guaranteed
575 * that nobody else tries to do anything about the
576 * state of the inode when it is locked, as we
577 * just created it (so there can be no old holders
578 * that haven't tested I_LOCK).
580 inode
->i_state
&= ~(I_LOCK
|I_NEW
);
581 wake_up_inode(inode
);
584 EXPORT_SYMBOL(unlock_new_inode
);
587 * This is called without the inode lock held.. Be careful.
589 * We no longer cache the sb_flags in i_flags - see fs.h
590 * -- rmk@arm.uk.linux.org
592 static struct inode
* get_new_inode(struct super_block
*sb
, struct hlist_head
*head
, int (*test
)(struct inode
*, void *), int (*set
)(struct inode
*, void *), void *data
)
594 struct inode
* inode
;
596 inode
= alloc_inode(sb
);
600 spin_lock(&inode_lock
);
601 /* We released the lock, so.. */
602 old
= find_inode(sb
, head
, test
, data
);
604 if (set(inode
, data
))
607 inodes_stat
.nr_inodes
++;
608 list_add(&inode
->i_list
, &inode_in_use
);
609 list_add(&inode
->i_sb_list
, &sb
->s_inodes
);
610 hlist_add_head(&inode
->i_hash
, head
);
611 inode
->i_state
= I_LOCK
|I_NEW
;
612 spin_unlock(&inode_lock
);
614 /* Return the locked inode with I_NEW set, the
615 * caller is responsible for filling in the contents
621 * Uhhuh, somebody else created the same inode under
622 * us. Use the old inode instead of the one we just
626 spin_unlock(&inode_lock
);
627 destroy_inode(inode
);
629 wait_on_inode(inode
);
634 spin_unlock(&inode_lock
);
635 destroy_inode(inode
);
640 * get_new_inode_fast is the fast path version of get_new_inode, see the
641 * comment at iget_locked for details.
643 static struct inode
* get_new_inode_fast(struct super_block
*sb
, struct hlist_head
*head
, unsigned long ino
)
645 struct inode
* inode
;
647 inode
= alloc_inode(sb
);
651 spin_lock(&inode_lock
);
652 /* We released the lock, so.. */
653 old
= find_inode_fast(sb
, head
, ino
);
656 inodes_stat
.nr_inodes
++;
657 list_add(&inode
->i_list
, &inode_in_use
);
658 list_add(&inode
->i_sb_list
, &sb
->s_inodes
);
659 hlist_add_head(&inode
->i_hash
, head
);
660 inode
->i_state
= I_LOCK
|I_NEW
;
661 spin_unlock(&inode_lock
);
663 /* Return the locked inode with I_NEW set, the
664 * caller is responsible for filling in the contents
670 * Uhhuh, somebody else created the same inode under
671 * us. Use the old inode instead of the one we just
675 spin_unlock(&inode_lock
);
676 destroy_inode(inode
);
678 wait_on_inode(inode
);
683 static inline unsigned long hash(struct super_block
*sb
, unsigned long hashval
)
687 tmp
= (hashval
* (unsigned long)sb
) ^ (GOLDEN_RATIO_PRIME
+ hashval
) /
689 tmp
= tmp
^ ((tmp
^ GOLDEN_RATIO_PRIME
) >> I_HASHBITS
);
690 return tmp
& I_HASHMASK
;
694 * iunique - get a unique inode number
696 * @max_reserved: highest reserved inode number
698 * Obtain an inode number that is unique on the system for a given
699 * superblock. This is used by file systems that have no natural
700 * permanent inode numbering system. An inode number is returned that
701 * is higher than the reserved limit but unique.
704 * With a large number of inodes live on the file system this function
705 * currently becomes quite slow.
707 ino_t
iunique(struct super_block
*sb
, ino_t max_reserved
)
709 static ino_t counter
;
711 struct hlist_head
* head
;
713 spin_lock(&inode_lock
);
715 if (counter
> max_reserved
) {
716 head
= inode_hashtable
+ hash(sb
,counter
);
718 inode
= find_inode_fast(sb
, head
, res
);
720 spin_unlock(&inode_lock
);
724 counter
= max_reserved
+ 1;
730 EXPORT_SYMBOL(iunique
);
732 struct inode
*igrab(struct inode
*inode
)
734 spin_lock(&inode_lock
);
735 if (!(inode
->i_state
& (I_FREEING
|I_WILL_FREE
)))
739 * Handle the case where s_op->clear_inode is not been
740 * called yet, and somebody is calling igrab
741 * while the inode is getting freed.
744 spin_unlock(&inode_lock
);
748 EXPORT_SYMBOL(igrab
);
751 * ifind - internal function, you want ilookup5() or iget5().
752 * @sb: super block of file system to search
753 * @head: the head of the list to search
754 * @test: callback used for comparisons between inodes
755 * @data: opaque data pointer to pass to @test
756 * @wait: if true wait for the inode to be unlocked, if false do not
758 * ifind() searches for the inode specified by @data in the inode
759 * cache. This is a generalized version of ifind_fast() for file systems where
760 * the inode number is not sufficient for unique identification of an inode.
762 * If the inode is in the cache, the inode is returned with an incremented
765 * Otherwise NULL is returned.
767 * Note, @test is called with the inode_lock held, so can't sleep.
769 static struct inode
*ifind(struct super_block
*sb
,
770 struct hlist_head
*head
, int (*test
)(struct inode
*, void *),
771 void *data
, const int wait
)
775 spin_lock(&inode_lock
);
776 inode
= find_inode(sb
, head
, test
, data
);
779 spin_unlock(&inode_lock
);
781 wait_on_inode(inode
);
784 spin_unlock(&inode_lock
);
789 * ifind_fast - internal function, you want ilookup() or iget().
790 * @sb: super block of file system to search
791 * @head: head of the list to search
792 * @ino: inode number to search for
794 * ifind_fast() searches for the inode @ino in the inode cache. This is for
795 * file systems where the inode number is sufficient for unique identification
798 * If the inode is in the cache, the inode is returned with an incremented
801 * Otherwise NULL is returned.
803 static struct inode
*ifind_fast(struct super_block
*sb
,
804 struct hlist_head
*head
, unsigned long ino
)
808 spin_lock(&inode_lock
);
809 inode
= find_inode_fast(sb
, head
, ino
);
812 spin_unlock(&inode_lock
);
813 wait_on_inode(inode
);
816 spin_unlock(&inode_lock
);
821 * ilookup5_nowait - search for an inode in the inode cache
822 * @sb: super block of file system to search
823 * @hashval: hash value (usually inode number) to search for
824 * @test: callback used for comparisons between inodes
825 * @data: opaque data pointer to pass to @test
827 * ilookup5() uses ifind() to search for the inode specified by @hashval and
828 * @data in the inode cache. This is a generalized version of ilookup() for
829 * file systems where the inode number is not sufficient for unique
830 * identification of an inode.
832 * If the inode is in the cache, the inode is returned with an incremented
833 * reference count. Note, the inode lock is not waited upon so you have to be
834 * very careful what you do with the returned inode. You probably should be
835 * using ilookup5() instead.
837 * Otherwise NULL is returned.
839 * Note, @test is called with the inode_lock held, so can't sleep.
841 struct inode
*ilookup5_nowait(struct super_block
*sb
, unsigned long hashval
,
842 int (*test
)(struct inode
*, void *), void *data
)
844 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
846 return ifind(sb
, head
, test
, data
, 0);
849 EXPORT_SYMBOL(ilookup5_nowait
);
852 * ilookup5 - search for an inode in the inode cache
853 * @sb: super block of file system to search
854 * @hashval: hash value (usually inode number) to search for
855 * @test: callback used for comparisons between inodes
856 * @data: opaque data pointer to pass to @test
858 * ilookup5() uses ifind() to search for the inode specified by @hashval and
859 * @data in the inode cache. This is a generalized version of ilookup() for
860 * file systems where the inode number is not sufficient for unique
861 * identification of an inode.
863 * If the inode is in the cache, the inode lock is waited upon and the inode is
864 * returned with an incremented reference count.
866 * Otherwise NULL is returned.
868 * Note, @test is called with the inode_lock held, so can't sleep.
870 struct inode
*ilookup5(struct super_block
*sb
, unsigned long hashval
,
871 int (*test
)(struct inode
*, void *), void *data
)
873 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
875 return ifind(sb
, head
, test
, data
, 1);
878 EXPORT_SYMBOL(ilookup5
);
881 * ilookup - search for an inode in the inode cache
882 * @sb: super block of file system to search
883 * @ino: inode number to search for
885 * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache.
886 * This is for file systems where the inode number is sufficient for unique
887 * identification of an inode.
889 * If the inode is in the cache, the inode is returned with an incremented
892 * Otherwise NULL is returned.
894 struct inode
*ilookup(struct super_block
*sb
, unsigned long ino
)
896 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
898 return ifind_fast(sb
, head
, ino
);
901 EXPORT_SYMBOL(ilookup
);
904 * iget5_locked - obtain an inode from a mounted file system
905 * @sb: super block of file system
906 * @hashval: hash value (usually inode number) to get
907 * @test: callback used for comparisons between inodes
908 * @set: callback used to initialize a new struct inode
909 * @data: opaque data pointer to pass to @test and @set
911 * This is iget() without the read_inode() portion of get_new_inode().
913 * iget5_locked() uses ifind() to search for the inode specified by @hashval
914 * and @data in the inode cache and if present it is returned with an increased
915 * reference count. This is a generalized version of iget_locked() for file
916 * systems where the inode number is not sufficient for unique identification
919 * If the inode is not in cache, get_new_inode() is called to allocate a new
920 * inode and this is returned locked, hashed, and with the I_NEW flag set. The
921 * file system gets to fill it in before unlocking it via unlock_new_inode().
923 * Note both @test and @set are called with the inode_lock held, so can't sleep.
925 struct inode
*iget5_locked(struct super_block
*sb
, unsigned long hashval
,
926 int (*test
)(struct inode
*, void *),
927 int (*set
)(struct inode
*, void *), void *data
)
929 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
932 inode
= ifind(sb
, head
, test
, data
, 1);
936 * get_new_inode() will do the right thing, re-trying the search
937 * in case it had to block at any point.
939 return get_new_inode(sb
, head
, test
, set
, data
);
942 EXPORT_SYMBOL(iget5_locked
);
945 * iget_locked - obtain an inode from a mounted file system
946 * @sb: super block of file system
947 * @ino: inode number to get
949 * This is iget() without the read_inode() portion of get_new_inode_fast().
951 * iget_locked() uses ifind_fast() to search for the inode specified by @ino in
952 * the inode cache and if present it is returned with an increased reference
953 * count. This is for file systems where the inode number is sufficient for
954 * unique identification of an inode.
956 * If the inode is not in cache, get_new_inode_fast() is called to allocate a
957 * new inode and this is returned locked, hashed, and with the I_NEW flag set.
958 * The file system gets to fill it in before unlocking it via
959 * unlock_new_inode().
961 struct inode
*iget_locked(struct super_block
*sb
, unsigned long ino
)
963 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
966 inode
= ifind_fast(sb
, head
, ino
);
970 * get_new_inode_fast() will do the right thing, re-trying the search
971 * in case it had to block at any point.
973 return get_new_inode_fast(sb
, head
, ino
);
976 EXPORT_SYMBOL(iget_locked
);
979 * __insert_inode_hash - hash an inode
980 * @inode: unhashed inode
981 * @hashval: unsigned long value used to locate this object in the
984 * Add an inode to the inode hash for this superblock.
986 void __insert_inode_hash(struct inode
*inode
, unsigned long hashval
)
988 struct hlist_head
*head
= inode_hashtable
+ hash(inode
->i_sb
, hashval
);
989 spin_lock(&inode_lock
);
990 hlist_add_head(&inode
->i_hash
, head
);
991 spin_unlock(&inode_lock
);
994 EXPORT_SYMBOL(__insert_inode_hash
);
997 * remove_inode_hash - remove an inode from the hash
998 * @inode: inode to unhash
1000 * Remove an inode from the superblock.
1002 void remove_inode_hash(struct inode
*inode
)
1004 spin_lock(&inode_lock
);
1005 hlist_del_init(&inode
->i_hash
);
1006 spin_unlock(&inode_lock
);
1009 EXPORT_SYMBOL(remove_inode_hash
);
1012 * Tell the filesystem that this inode is no longer of any interest and should
1013 * be completely destroyed.
1015 * We leave the inode in the inode hash table until *after* the filesystem's
1016 * ->delete_inode completes. This ensures that an iget (such as nfsd might
1017 * instigate) will always find up-to-date information either in the hash or on
1020 * I_FREEING is set so that no-one will take a new reference to the inode while
1021 * it is being deleted.
1023 void generic_delete_inode(struct inode
*inode
)
1025 struct super_operations
*op
= inode
->i_sb
->s_op
;
1027 list_del_init(&inode
->i_list
);
1028 list_del_init(&inode
->i_sb_list
);
1029 inode
->i_state
|=I_FREEING
;
1030 inodes_stat
.nr_inodes
--;
1031 spin_unlock(&inode_lock
);
1033 security_inode_delete(inode
);
1035 if (op
->delete_inode
) {
1036 void (*delete)(struct inode
*) = op
->delete_inode
;
1037 if (!is_bad_inode(inode
))
1039 /* Filesystems implementing their own
1040 * s_op->delete_inode are required to call
1041 * truncate_inode_pages and clear_inode()
1045 truncate_inode_pages(&inode
->i_data
, 0);
1048 spin_lock(&inode_lock
);
1049 hlist_del_init(&inode
->i_hash
);
1050 spin_unlock(&inode_lock
);
1051 wake_up_inode(inode
);
1052 BUG_ON(inode
->i_state
!= I_CLEAR
);
1053 destroy_inode(inode
);
1056 EXPORT_SYMBOL(generic_delete_inode
);
1058 static void generic_forget_inode(struct inode
*inode
)
1060 struct super_block
*sb
= inode
->i_sb
;
1062 if (!hlist_unhashed(&inode
->i_hash
)) {
1063 if (!(inode
->i_state
& (I_DIRTY
|I_LOCK
)))
1064 list_move(&inode
->i_list
, &inode_unused
);
1065 inodes_stat
.nr_unused
++;
1066 if (!sb
|| (sb
->s_flags
& MS_ACTIVE
)) {
1067 spin_unlock(&inode_lock
);
1070 inode
->i_state
|= I_WILL_FREE
;
1071 spin_unlock(&inode_lock
);
1072 write_inode_now(inode
, 1);
1073 spin_lock(&inode_lock
);
1074 inode
->i_state
&= ~I_WILL_FREE
;
1075 inodes_stat
.nr_unused
--;
1076 hlist_del_init(&inode
->i_hash
);
1078 list_del_init(&inode
->i_list
);
1079 list_del_init(&inode
->i_sb_list
);
1080 inode
->i_state
|= I_FREEING
;
1081 inodes_stat
.nr_inodes
--;
1082 spin_unlock(&inode_lock
);
1083 if (inode
->i_data
.nrpages
)
1084 truncate_inode_pages(&inode
->i_data
, 0);
1086 wake_up_inode(inode
);
1087 destroy_inode(inode
);
1091 * Normal UNIX filesystem behaviour: delete the
1092 * inode when the usage count drops to zero, and
1095 void generic_drop_inode(struct inode
*inode
)
1097 if (!inode
->i_nlink
)
1098 generic_delete_inode(inode
);
1100 generic_forget_inode(inode
);
1103 EXPORT_SYMBOL_GPL(generic_drop_inode
);
1106 * Called when we're dropping the last reference
1109 * Call the FS "drop()" function, defaulting to
1110 * the legacy UNIX filesystem behaviour..
1112 * NOTE! NOTE! NOTE! We're called with the inode lock
1113 * held, and the drop function is supposed to release
1116 static inline void iput_final(struct inode
*inode
)
1118 struct super_operations
*op
= inode
->i_sb
->s_op
;
1119 void (*drop
)(struct inode
*) = generic_drop_inode
;
1121 if (op
&& op
->drop_inode
)
1122 drop
= op
->drop_inode
;
1127 * iput - put an inode
1128 * @inode: inode to put
1130 * Puts an inode, dropping its usage count. If the inode use count hits
1131 * zero, the inode is then freed and may also be destroyed.
1133 * Consequently, iput() can sleep.
1135 void iput(struct inode
*inode
)
1138 struct super_operations
*op
= inode
->i_sb
->s_op
;
1140 BUG_ON(inode
->i_state
== I_CLEAR
);
1142 if (op
&& op
->put_inode
)
1143 op
->put_inode(inode
);
1145 if (atomic_dec_and_lock(&inode
->i_count
, &inode_lock
))
1150 EXPORT_SYMBOL(iput
);
1153 * bmap - find a block number in a file
1154 * @inode: inode of file
1155 * @block: block to find
1157 * Returns the block number on the device holding the inode that
1158 * is the disk block number for the block of the file requested.
1159 * That is, asked for block 4 of inode 1 the function will return the
1160 * disk block relative to the disk start that holds that block of the
1163 sector_t
bmap(struct inode
* inode
, sector_t block
)
1166 if (inode
->i_mapping
->a_ops
->bmap
)
1167 res
= inode
->i_mapping
->a_ops
->bmap(inode
->i_mapping
, block
);
1171 EXPORT_SYMBOL(bmap
);
1174 * touch_atime - update the access time
1175 * @mnt: mount the inode is accessed on
1176 * @dentry: dentry accessed
1178 * Update the accessed time on an inode and mark it for writeback.
1179 * This function automatically handles read only file systems and media,
1180 * as well as the "noatime" flag and inode specific "noatime" markers.
1182 void touch_atime(struct vfsmount
*mnt
, struct dentry
*dentry
)
1184 struct inode
*inode
= dentry
->d_inode
;
1185 struct timespec now
;
1187 if (IS_RDONLY(inode
))
1190 if ((inode
->i_flags
& S_NOATIME
) ||
1191 (inode
->i_sb
->s_flags
& MS_NOATIME
) ||
1192 ((inode
->i_sb
->s_flags
& MS_NODIRATIME
) && S_ISDIR(inode
->i_mode
)))
1196 * We may have a NULL vfsmount when coming from NFSD
1199 ((mnt
->mnt_flags
& MNT_NOATIME
) ||
1200 ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
))))
1203 now
= current_fs_time(inode
->i_sb
);
1204 if (!timespec_equal(&inode
->i_atime
, &now
)) {
1205 inode
->i_atime
= now
;
1206 mark_inode_dirty_sync(inode
);
1210 EXPORT_SYMBOL(touch_atime
);
1213 * file_update_time - update mtime and ctime time
1214 * @file: file accessed
1216 * Update the mtime and ctime members of an inode and mark the inode
1217 * for writeback. Note that this function is meant exclusively for
1218 * usage in the file write path of filesystems, and filesystems may
1219 * choose to explicitly ignore update via this function with the
1220 * S_NOCTIME inode flag, e.g. for network filesystem where these
1221 * timestamps are handled by the server.
1224 void file_update_time(struct file
*file
)
1226 struct inode
*inode
= file
->f_dentry
->d_inode
;
1227 struct timespec now
;
1230 if (IS_NOCMTIME(inode
))
1232 if (IS_RDONLY(inode
))
1235 now
= current_fs_time(inode
->i_sb
);
1236 if (!timespec_equal(&inode
->i_mtime
, &now
))
1238 inode
->i_mtime
= now
;
1240 if (!timespec_equal(&inode
->i_ctime
, &now
))
1242 inode
->i_ctime
= now
;
1245 mark_inode_dirty_sync(inode
);
1248 EXPORT_SYMBOL(file_update_time
);
1250 int inode_needs_sync(struct inode
*inode
)
1254 if (S_ISDIR(inode
->i_mode
) && IS_DIRSYNC(inode
))
1259 EXPORT_SYMBOL(inode_needs_sync
);
1262 * Quota functions that want to walk the inode lists..
1266 /* Function back in dquot.c */
1267 int remove_inode_dquot_ref(struct inode
*, int, struct list_head
*);
1269 void remove_dquot_ref(struct super_block
*sb
, int type
,
1270 struct list_head
*tofree_head
)
1272 struct inode
*inode
;
1275 return; /* nothing to do */
1276 spin_lock(&inode_lock
); /* This lock is for inodes code */
1279 * We don't have to lock against quota code - test IS_QUOTAINIT is
1280 * just for speedup...
1282 list_for_each_entry(inode
, &sb
->s_inodes
, i_sb_list
)
1283 if (!IS_NOQUOTA(inode
))
1284 remove_inode_dquot_ref(inode
, type
, tofree_head
);
1286 spin_unlock(&inode_lock
);
1291 int inode_wait(void *word
)
1298 * If we try to find an inode in the inode hash while it is being
1299 * deleted, we have to wait until the filesystem completes its
1300 * deletion before reporting that it isn't found. This function waits
1301 * until the deletion _might_ have completed. Callers are responsible
1302 * to recheck inode state.
1304 * It doesn't matter if I_LOCK is not set initially, a call to
1305 * wake_up_inode() after removing from the hash list will DTRT.
1307 * This is called with inode_lock held.
1309 static void __wait_on_freeing_inode(struct inode
*inode
)
1311 wait_queue_head_t
*wq
;
1312 DEFINE_WAIT_BIT(wait
, &inode
->i_state
, __I_LOCK
);
1313 wq
= bit_waitqueue(&inode
->i_state
, __I_LOCK
);
1314 prepare_to_wait(wq
, &wait
.wait
, TASK_UNINTERRUPTIBLE
);
1315 spin_unlock(&inode_lock
);
1317 finish_wait(wq
, &wait
.wait
);
1318 spin_lock(&inode_lock
);
1321 void wake_up_inode(struct inode
*inode
)
1324 * Prevent speculative execution through spin_unlock(&inode_lock);
1327 wake_up_bit(&inode
->i_state
, __I_LOCK
);
1330 static __initdata
unsigned long ihash_entries
;
1331 static int __init
set_ihash_entries(char *str
)
1335 ihash_entries
= simple_strtoul(str
, &str
, 0);
1338 __setup("ihash_entries=", set_ihash_entries
);
1341 * Initialize the waitqueues and inode hash table.
1343 void __init
inode_init_early(void)
1347 /* If hashes are distributed across NUMA nodes, defer
1348 * hash allocation until vmalloc space is available.
1354 alloc_large_system_hash("Inode-cache",
1355 sizeof(struct hlist_head
),
1363 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1364 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1367 void __init
inode_init(unsigned long mempages
)
1371 /* inode slab cache */
1372 inode_cachep
= kmem_cache_create("inode_cache",
1373 sizeof(struct inode
),
1375 (SLAB_RECLAIM_ACCOUNT
|SLAB_PANIC
|
1379 set_shrinker(DEFAULT_SEEKS
, shrink_icache_memory
);
1381 /* Hash may have been set up in inode_init_early */
1386 alloc_large_system_hash("Inode-cache",
1387 sizeof(struct hlist_head
),
1395 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1396 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1399 void init_special_inode(struct inode
*inode
, umode_t mode
, dev_t rdev
)
1401 inode
->i_mode
= mode
;
1402 if (S_ISCHR(mode
)) {
1403 inode
->i_fop
= &def_chr_fops
;
1404 inode
->i_rdev
= rdev
;
1405 } else if (S_ISBLK(mode
)) {
1406 inode
->i_fop
= &def_blk_fops
;
1407 inode
->i_rdev
= rdev
;
1408 } else if (S_ISFIFO(mode
))
1409 inode
->i_fop
= &def_fifo_fops
;
1410 else if (S_ISSOCK(mode
))
1411 inode
->i_fop
= &bad_sock_fops
;
1413 printk(KERN_DEBUG
"init_special_inode: bogus i_mode (%o)\n",
1416 EXPORT_SYMBOL(init_special_inode
);