4 * (C) 1997 Linus Torvalds
9 #include <linux/dcache.h>
10 #include <linux/init.h>
11 #include <linux/slab.h>
12 #include <linux/writeback.h>
13 #include <linux/module.h>
14 #include <linux/backing-dev.h>
15 #include <linux/wait.h>
16 #include <linux/rwsem.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/fsnotify.h>
24 #include <linux/mount.h>
25 #include <linux/async.h>
26 #include <linux/posix_acl.h>
27 #include <linux/ima.h>
28 #include <linux/cred.h>
31 * inode locking rules.
33 * inode->i_lock protects:
34 * inode->i_state, inode->i_hash, __iget()
35 * inode_lru_lock protects:
36 * inode_lru, inode->i_lru
45 * This is needed for the following functions:
49 * FIXME: remove all knowledge of the buffer layer from this file
51 #include <linux/buffer_head.h>
54 * New inode.c implementation.
56 * This implementation has the basic premise of trying
57 * to be extremely low-overhead and SMP-safe, yet be
58 * simple enough to be "obviously correct".
63 /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */
65 /* #define INODE_PARANOIA 1 */
66 /* #define INODE_DEBUG 1 */
69 * Inode lookup is no longer as critical as it used to be:
70 * most of the lookups are going to be through the dcache.
72 #define I_HASHBITS i_hash_shift
73 #define I_HASHMASK i_hash_mask
75 static unsigned int i_hash_mask __read_mostly
;
76 static unsigned int i_hash_shift __read_mostly
;
79 * Each inode can be on two separate lists. One is
80 * the hash list of the inode, used for lookups. The
81 * other linked list is the "type" list:
82 * "in_use" - valid inode, i_count > 0, i_nlink > 0
83 * "dirty" - as "in_use" but also dirty
84 * "unused" - valid inode, i_count = 0
86 * A "dirty" list is maintained for each super block,
87 * allowing for low-overhead inode sync() operations.
90 static LIST_HEAD(inode_lru
);
91 static DEFINE_SPINLOCK(inode_lru_lock
);
92 static struct hlist_head
*inode_hashtable __read_mostly
;
95 * A simple spinlock to protect the list manipulations.
97 * NOTE! You also have to own the lock if you change
98 * the i_state of an inode while it is in use..
100 DEFINE_SPINLOCK(inode_lock
);
103 * iprune_sem provides exclusion between the icache shrinking and the
106 * We don't actually need it to protect anything in the umount path,
107 * but only need to cycle through it to make sure any inode that
108 * prune_icache took off the LRU list has been fully torn down by the
109 * time we are past evict_inodes.
111 static DECLARE_RWSEM(iprune_sem
);
114 * Statistics gathering..
116 struct inodes_stat_t inodes_stat
;
118 static DEFINE_PER_CPU(unsigned int, nr_inodes
);
120 static struct kmem_cache
*inode_cachep __read_mostly
;
122 static int get_nr_inodes(void)
126 for_each_possible_cpu(i
)
127 sum
+= per_cpu(nr_inodes
, i
);
128 return sum
< 0 ? 0 : sum
;
131 static inline int get_nr_inodes_unused(void)
133 return inodes_stat
.nr_unused
;
136 int get_nr_dirty_inodes(void)
138 /* not actually dirty inodes, but a wild approximation */
139 int nr_dirty
= get_nr_inodes() - get_nr_inodes_unused();
140 return nr_dirty
> 0 ? nr_dirty
: 0;
144 * Handle nr_inode sysctl
147 int proc_nr_inodes(ctl_table
*table
, int write
,
148 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
150 inodes_stat
.nr_inodes
= get_nr_inodes();
151 return proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
156 * inode_init_always - perform inode structure intialisation
157 * @sb: superblock inode belongs to
158 * @inode: inode to initialise
160 * These are initializations that need to be done on every inode
161 * allocation as the fields are not initialised by slab allocation.
163 int inode_init_always(struct super_block
*sb
, struct inode
*inode
)
165 static const struct address_space_operations empty_aops
;
166 static const struct inode_operations empty_iops
;
167 static const struct file_operations empty_fops
;
168 struct address_space
*const mapping
= &inode
->i_data
;
171 inode
->i_blkbits
= sb
->s_blocksize_bits
;
173 atomic_set(&inode
->i_count
, 1);
174 inode
->i_op
= &empty_iops
;
175 inode
->i_fop
= &empty_fops
;
179 atomic_set(&inode
->i_writecount
, 0);
183 inode
->i_generation
= 0;
185 memset(&inode
->i_dquot
, 0, sizeof(inode
->i_dquot
));
187 inode
->i_pipe
= NULL
;
188 inode
->i_bdev
= NULL
;
189 inode
->i_cdev
= NULL
;
191 inode
->dirtied_when
= 0;
193 if (security_inode_alloc(inode
))
195 spin_lock_init(&inode
->i_lock
);
196 lockdep_set_class(&inode
->i_lock
, &sb
->s_type
->i_lock_key
);
198 mutex_init(&inode
->i_mutex
);
199 lockdep_set_class(&inode
->i_mutex
, &sb
->s_type
->i_mutex_key
);
201 init_rwsem(&inode
->i_alloc_sem
);
202 lockdep_set_class(&inode
->i_alloc_sem
, &sb
->s_type
->i_alloc_sem_key
);
204 mapping
->a_ops
= &empty_aops
;
205 mapping
->host
= inode
;
207 mapping_set_gfp_mask(mapping
, GFP_HIGHUSER_MOVABLE
);
208 mapping
->assoc_mapping
= NULL
;
209 mapping
->backing_dev_info
= &default_backing_dev_info
;
210 mapping
->writeback_index
= 0;
213 * If the block_device provides a backing_dev_info for client
214 * inodes then use that. Otherwise the inode share the bdev's
218 struct backing_dev_info
*bdi
;
220 bdi
= sb
->s_bdev
->bd_inode
->i_mapping
->backing_dev_info
;
221 mapping
->backing_dev_info
= bdi
;
223 inode
->i_private
= NULL
;
224 inode
->i_mapping
= mapping
;
225 #ifdef CONFIG_FS_POSIX_ACL
226 inode
->i_acl
= inode
->i_default_acl
= ACL_NOT_CACHED
;
229 #ifdef CONFIG_FSNOTIFY
230 inode
->i_fsnotify_mask
= 0;
233 this_cpu_inc(nr_inodes
);
239 EXPORT_SYMBOL(inode_init_always
);
241 static struct inode
*alloc_inode(struct super_block
*sb
)
245 if (sb
->s_op
->alloc_inode
)
246 inode
= sb
->s_op
->alloc_inode(sb
);
248 inode
= kmem_cache_alloc(inode_cachep
, GFP_KERNEL
);
253 if (unlikely(inode_init_always(sb
, inode
))) {
254 if (inode
->i_sb
->s_op
->destroy_inode
)
255 inode
->i_sb
->s_op
->destroy_inode(inode
);
257 kmem_cache_free(inode_cachep
, inode
);
264 void free_inode_nonrcu(struct inode
*inode
)
266 kmem_cache_free(inode_cachep
, inode
);
268 EXPORT_SYMBOL(free_inode_nonrcu
);
270 void __destroy_inode(struct inode
*inode
)
272 BUG_ON(inode_has_buffers(inode
));
273 security_inode_free(inode
);
274 fsnotify_inode_delete(inode
);
275 #ifdef CONFIG_FS_POSIX_ACL
276 if (inode
->i_acl
&& inode
->i_acl
!= ACL_NOT_CACHED
)
277 posix_acl_release(inode
->i_acl
);
278 if (inode
->i_default_acl
&& inode
->i_default_acl
!= ACL_NOT_CACHED
)
279 posix_acl_release(inode
->i_default_acl
);
281 this_cpu_dec(nr_inodes
);
283 EXPORT_SYMBOL(__destroy_inode
);
285 static void i_callback(struct rcu_head
*head
)
287 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
288 INIT_LIST_HEAD(&inode
->i_dentry
);
289 kmem_cache_free(inode_cachep
, inode
);
292 static void destroy_inode(struct inode
*inode
)
294 BUG_ON(!list_empty(&inode
->i_lru
));
295 __destroy_inode(inode
);
296 if (inode
->i_sb
->s_op
->destroy_inode
)
297 inode
->i_sb
->s_op
->destroy_inode(inode
);
299 call_rcu(&inode
->i_rcu
, i_callback
);
302 void address_space_init_once(struct address_space
*mapping
)
304 memset(mapping
, 0, sizeof(*mapping
));
305 INIT_RADIX_TREE(&mapping
->page_tree
, GFP_ATOMIC
);
306 spin_lock_init(&mapping
->tree_lock
);
307 spin_lock_init(&mapping
->i_mmap_lock
);
308 INIT_LIST_HEAD(&mapping
->private_list
);
309 spin_lock_init(&mapping
->private_lock
);
310 INIT_RAW_PRIO_TREE_ROOT(&mapping
->i_mmap
);
311 INIT_LIST_HEAD(&mapping
->i_mmap_nonlinear
);
312 mutex_init(&mapping
->unmap_mutex
);
314 EXPORT_SYMBOL(address_space_init_once
);
317 * These are initializations that only need to be done
318 * once, because the fields are idempotent across use
319 * of the inode, so let the slab aware of that.
321 void inode_init_once(struct inode
*inode
)
323 memset(inode
, 0, sizeof(*inode
));
324 INIT_HLIST_NODE(&inode
->i_hash
);
325 INIT_LIST_HEAD(&inode
->i_dentry
);
326 INIT_LIST_HEAD(&inode
->i_devices
);
327 INIT_LIST_HEAD(&inode
->i_wb_list
);
328 INIT_LIST_HEAD(&inode
->i_lru
);
329 address_space_init_once(&inode
->i_data
);
330 i_size_ordered_init(inode
);
331 #ifdef CONFIG_FSNOTIFY
332 INIT_HLIST_HEAD(&inode
->i_fsnotify_marks
);
335 EXPORT_SYMBOL(inode_init_once
);
337 static void init_once(void *foo
)
339 struct inode
*inode
= (struct inode
*) foo
;
341 inode_init_once(inode
);
345 * inode->i_lock must be held
347 void __iget(struct inode
*inode
)
349 atomic_inc(&inode
->i_count
);
353 * get additional reference to inode; caller must already hold one.
355 void ihold(struct inode
*inode
)
357 WARN_ON(atomic_inc_return(&inode
->i_count
) < 2);
359 EXPORT_SYMBOL(ihold
);
361 static void inode_lru_list_add(struct inode
*inode
)
363 spin_lock(&inode_lru_lock
);
364 if (list_empty(&inode
->i_lru
)) {
365 list_add(&inode
->i_lru
, &inode_lru
);
366 inodes_stat
.nr_unused
++;
368 spin_unlock(&inode_lru_lock
);
371 static void inode_lru_list_del(struct inode
*inode
)
373 spin_lock(&inode_lru_lock
);
374 if (!list_empty(&inode
->i_lru
)) {
375 list_del_init(&inode
->i_lru
);
376 inodes_stat
.nr_unused
--;
378 spin_unlock(&inode_lru_lock
);
381 static inline void __inode_sb_list_add(struct inode
*inode
)
383 list_add(&inode
->i_sb_list
, &inode
->i_sb
->s_inodes
);
387 * inode_sb_list_add - add inode to the superblock list of inodes
388 * @inode: inode to add
390 void inode_sb_list_add(struct inode
*inode
)
392 spin_lock(&inode_lock
);
393 __inode_sb_list_add(inode
);
394 spin_unlock(&inode_lock
);
396 EXPORT_SYMBOL_GPL(inode_sb_list_add
);
398 static inline void __inode_sb_list_del(struct inode
*inode
)
400 list_del_init(&inode
->i_sb_list
);
403 static unsigned long hash(struct super_block
*sb
, unsigned long hashval
)
407 tmp
= (hashval
* (unsigned long)sb
) ^ (GOLDEN_RATIO_PRIME
+ hashval
) /
409 tmp
= tmp
^ ((tmp
^ GOLDEN_RATIO_PRIME
) >> I_HASHBITS
);
410 return tmp
& I_HASHMASK
;
414 * __insert_inode_hash - hash an inode
415 * @inode: unhashed inode
416 * @hashval: unsigned long value used to locate this object in the
419 * Add an inode to the inode hash for this superblock.
421 void __insert_inode_hash(struct inode
*inode
, unsigned long hashval
)
423 struct hlist_head
*b
= inode_hashtable
+ hash(inode
->i_sb
, hashval
);
425 spin_lock(&inode_lock
);
426 spin_lock(&inode
->i_lock
);
427 hlist_add_head(&inode
->i_hash
, b
);
428 spin_unlock(&inode
->i_lock
);
429 spin_unlock(&inode_lock
);
431 EXPORT_SYMBOL(__insert_inode_hash
);
434 * remove_inode_hash - remove an inode from the hash
435 * @inode: inode to unhash
437 * Remove an inode from the superblock.
439 void remove_inode_hash(struct inode
*inode
)
441 spin_lock(&inode_lock
);
442 spin_lock(&inode
->i_lock
);
443 hlist_del_init(&inode
->i_hash
);
444 spin_unlock(&inode
->i_lock
);
445 spin_unlock(&inode_lock
);
447 EXPORT_SYMBOL(remove_inode_hash
);
449 void end_writeback(struct inode
*inode
)
452 BUG_ON(inode
->i_data
.nrpages
);
453 BUG_ON(!list_empty(&inode
->i_data
.private_list
));
454 BUG_ON(!(inode
->i_state
& I_FREEING
));
455 BUG_ON(inode
->i_state
& I_CLEAR
);
456 inode_sync_wait(inode
);
457 /* don't need i_lock here, no concurrent mods to i_state */
458 inode
->i_state
= I_FREEING
| I_CLEAR
;
460 EXPORT_SYMBOL(end_writeback
);
463 * Free the inode passed in, removing it from the lists it is still connected
464 * to. We remove any pages still attached to the inode and wait for any IO that
465 * is still in progress before finally destroying the inode.
467 * An inode must already be marked I_FREEING so that we avoid the inode being
468 * moved back onto lists if we race with other code that manipulates the lists
469 * (e.g. writeback_single_inode). The caller is responsible for setting this.
471 * An inode must already be removed from the LRU list before being evicted from
472 * the cache. This should occur atomically with setting the I_FREEING state
473 * flag, so no inodes here should ever be on the LRU when being evicted.
475 static void evict(struct inode
*inode
)
477 const struct super_operations
*op
= inode
->i_sb
->s_op
;
479 BUG_ON(!(inode
->i_state
& I_FREEING
));
480 BUG_ON(!list_empty(&inode
->i_lru
));
482 spin_lock(&inode_lock
);
483 list_del_init(&inode
->i_wb_list
);
484 __inode_sb_list_del(inode
);
485 spin_unlock(&inode_lock
);
487 if (op
->evict_inode
) {
488 op
->evict_inode(inode
);
490 if (inode
->i_data
.nrpages
)
491 truncate_inode_pages(&inode
->i_data
, 0);
492 end_writeback(inode
);
494 if (S_ISBLK(inode
->i_mode
) && inode
->i_bdev
)
496 if (S_ISCHR(inode
->i_mode
) && inode
->i_cdev
)
499 remove_inode_hash(inode
);
501 spin_lock(&inode
->i_lock
);
502 wake_up_bit(&inode
->i_state
, __I_NEW
);
503 BUG_ON(inode
->i_state
!= (I_FREEING
| I_CLEAR
));
504 spin_unlock(&inode
->i_lock
);
506 destroy_inode(inode
);
510 * dispose_list - dispose of the contents of a local list
511 * @head: the head of the list to free
513 * Dispose-list gets a local list with local inodes in it, so it doesn't
514 * need to worry about list corruption and SMP locks.
516 static void dispose_list(struct list_head
*head
)
518 while (!list_empty(head
)) {
521 inode
= list_first_entry(head
, struct inode
, i_lru
);
522 list_del_init(&inode
->i_lru
);
529 * evict_inodes - evict all evictable inodes for a superblock
530 * @sb: superblock to operate on
532 * Make sure that no inodes with zero refcount are retained. This is
533 * called by superblock shutdown after having MS_ACTIVE flag removed,
534 * so any inode reaching zero refcount during or after that call will
535 * be immediately evicted.
537 void evict_inodes(struct super_block
*sb
)
539 struct inode
*inode
, *next
;
542 spin_lock(&inode_lock
);
543 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
544 if (atomic_read(&inode
->i_count
))
547 spin_lock(&inode
->i_lock
);
548 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
549 spin_unlock(&inode
->i_lock
);
553 inode
->i_state
|= I_FREEING
;
554 inode_lru_list_del(inode
);
555 spin_unlock(&inode
->i_lock
);
556 list_add(&inode
->i_lru
, &dispose
);
558 spin_unlock(&inode_lock
);
560 dispose_list(&dispose
);
563 * Cycle through iprune_sem to make sure any inode that prune_icache
564 * moved off the list before we took the lock has been fully torn
567 down_write(&iprune_sem
);
568 up_write(&iprune_sem
);
572 * invalidate_inodes - attempt to free all inodes on a superblock
573 * @sb: superblock to operate on
574 * @kill_dirty: flag to guide handling of dirty inodes
576 * Attempts to free all inodes for a given superblock. If there were any
577 * busy inodes return a non-zero value, else zero.
578 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
581 int invalidate_inodes(struct super_block
*sb
, bool kill_dirty
)
584 struct inode
*inode
, *next
;
587 spin_lock(&inode_lock
);
588 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
589 spin_lock(&inode
->i_lock
);
590 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
591 spin_unlock(&inode
->i_lock
);
594 if (inode
->i_state
& I_DIRTY
&& !kill_dirty
) {
595 spin_unlock(&inode
->i_lock
);
599 if (atomic_read(&inode
->i_count
)) {
600 spin_unlock(&inode
->i_lock
);
605 inode
->i_state
|= I_FREEING
;
606 inode_lru_list_del(inode
);
607 spin_unlock(&inode
->i_lock
);
608 list_add(&inode
->i_lru
, &dispose
);
610 spin_unlock(&inode_lock
);
612 dispose_list(&dispose
);
617 static int can_unuse(struct inode
*inode
)
619 if (inode
->i_state
& ~I_REFERENCED
)
621 if (inode_has_buffers(inode
))
623 if (atomic_read(&inode
->i_count
))
625 if (inode
->i_data
.nrpages
)
631 * Scan `goal' inodes on the unused list for freeable ones. They are moved to a
632 * temporary list and then are freed outside inode_lru_lock by dispose_list().
634 * Any inodes which are pinned purely because of attached pagecache have their
635 * pagecache removed. If the inode has metadata buffers attached to
636 * mapping->private_list then try to remove them.
638 * If the inode has the I_REFERENCED flag set, then it means that it has been
639 * used recently - the flag is set in iput_final(). When we encounter such an
640 * inode, clear the flag and move it to the back of the LRU so it gets another
641 * pass through the LRU before it gets reclaimed. This is necessary because of
642 * the fact we are doing lazy LRU updates to minimise lock contention so the
643 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
644 * with this flag set because they are the inodes that are out of order.
646 static void prune_icache(int nr_to_scan
)
650 unsigned long reap
= 0;
652 down_read(&iprune_sem
);
653 spin_lock(&inode_lru_lock
);
654 for (nr_scanned
= 0; nr_scanned
< nr_to_scan
; nr_scanned
++) {
657 if (list_empty(&inode_lru
))
660 inode
= list_entry(inode_lru
.prev
, struct inode
, i_lru
);
663 * we are inverting the inode_lru_lock/inode->i_lock here,
664 * so use a trylock. If we fail to get the lock, just move the
665 * inode to the back of the list so we don't spin on it.
667 if (!spin_trylock(&inode
->i_lock
)) {
668 list_move(&inode
->i_lru
, &inode_lru
);
673 * Referenced or dirty inodes are still in use. Give them
674 * another pass through the LRU as we canot reclaim them now.
676 if (atomic_read(&inode
->i_count
) ||
677 (inode
->i_state
& ~I_REFERENCED
)) {
678 list_del_init(&inode
->i_lru
);
679 spin_unlock(&inode
->i_lock
);
680 inodes_stat
.nr_unused
--;
684 /* recently referenced inodes get one more pass */
685 if (inode
->i_state
& I_REFERENCED
) {
686 inode
->i_state
&= ~I_REFERENCED
;
687 list_move(&inode
->i_lru
, &inode_lru
);
688 spin_unlock(&inode
->i_lock
);
691 if (inode_has_buffers(inode
) || inode
->i_data
.nrpages
) {
693 spin_unlock(&inode
->i_lock
);
694 spin_unlock(&inode_lru_lock
);
695 if (remove_inode_buffers(inode
))
696 reap
+= invalidate_mapping_pages(&inode
->i_data
,
699 spin_lock(&inode_lru_lock
);
701 if (inode
!= list_entry(inode_lru
.next
,
702 struct inode
, i_lru
))
703 continue; /* wrong inode or list_empty */
704 /* avoid lock inversions with trylock */
705 if (!spin_trylock(&inode
->i_lock
))
707 if (!can_unuse(inode
)) {
708 spin_unlock(&inode
->i_lock
);
712 WARN_ON(inode
->i_state
& I_NEW
);
713 inode
->i_state
|= I_FREEING
;
714 spin_unlock(&inode
->i_lock
);
716 list_move(&inode
->i_lru
, &freeable
);
717 inodes_stat
.nr_unused
--;
719 if (current_is_kswapd())
720 __count_vm_events(KSWAPD_INODESTEAL
, reap
);
722 __count_vm_events(PGINODESTEAL
, reap
);
723 spin_unlock(&inode_lru_lock
);
725 dispose_list(&freeable
);
726 up_read(&iprune_sem
);
730 * shrink_icache_memory() will attempt to reclaim some unused inodes. Here,
731 * "unused" means that no dentries are referring to the inodes: the files are
732 * not open and the dcache references to those inodes have already been
735 * This function is passed the number of inodes to scan, and it returns the
736 * total number of remaining possibly-reclaimable inodes.
738 static int shrink_icache_memory(struct shrinker
*shrink
, int nr
, gfp_t gfp_mask
)
742 * Nasty deadlock avoidance. We may hold various FS locks,
743 * and we don't want to recurse into the FS that called us
744 * in clear_inode() and friends..
746 if (!(gfp_mask
& __GFP_FS
))
750 return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure
;
753 static struct shrinker icache_shrinker
= {
754 .shrink
= shrink_icache_memory
,
755 .seeks
= DEFAULT_SEEKS
,
758 static void __wait_on_freeing_inode(struct inode
*inode
);
760 * Called with the inode lock held.
762 static struct inode
*find_inode(struct super_block
*sb
,
763 struct hlist_head
*head
,
764 int (*test
)(struct inode
*, void *),
767 struct hlist_node
*node
;
768 struct inode
*inode
= NULL
;
771 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
772 if (inode
->i_sb
!= sb
)
774 if (!test(inode
, data
))
776 spin_lock(&inode
->i_lock
);
777 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
778 __wait_on_freeing_inode(inode
);
782 spin_unlock(&inode
->i_lock
);
789 * find_inode_fast is the fast path version of find_inode, see the comment at
790 * iget_locked for details.
792 static struct inode
*find_inode_fast(struct super_block
*sb
,
793 struct hlist_head
*head
, unsigned long ino
)
795 struct hlist_node
*node
;
796 struct inode
*inode
= NULL
;
799 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
800 if (inode
->i_ino
!= ino
)
802 if (inode
->i_sb
!= sb
)
804 spin_lock(&inode
->i_lock
);
805 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
806 __wait_on_freeing_inode(inode
);
810 spin_unlock(&inode
->i_lock
);
817 * Each cpu owns a range of LAST_INO_BATCH numbers.
818 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
819 * to renew the exhausted range.
821 * This does not significantly increase overflow rate because every CPU can
822 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
823 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
824 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
825 * overflow rate by 2x, which does not seem too significant.
827 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
828 * error if st_ino won't fit in target struct field. Use 32bit counter
829 * here to attempt to avoid that.
831 #define LAST_INO_BATCH 1024
832 static DEFINE_PER_CPU(unsigned int, last_ino
);
834 unsigned int get_next_ino(void)
836 unsigned int *p
= &get_cpu_var(last_ino
);
837 unsigned int res
= *p
;
840 if (unlikely((res
& (LAST_INO_BATCH
-1)) == 0)) {
841 static atomic_t shared_last_ino
;
842 int next
= atomic_add_return(LAST_INO_BATCH
, &shared_last_ino
);
844 res
= next
- LAST_INO_BATCH
;
849 put_cpu_var(last_ino
);
852 EXPORT_SYMBOL(get_next_ino
);
855 * new_inode - obtain an inode
858 * Allocates a new inode for given superblock. The default gfp_mask
859 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
860 * If HIGHMEM pages are unsuitable or it is known that pages allocated
861 * for the page cache are not reclaimable or migratable,
862 * mapping_set_gfp_mask() must be called with suitable flags on the
863 * newly created inode's mapping
866 struct inode
*new_inode(struct super_block
*sb
)
870 spin_lock_prefetch(&inode_lock
);
872 inode
= alloc_inode(sb
);
874 spin_lock(&inode_lock
);
875 spin_lock(&inode
->i_lock
);
877 spin_unlock(&inode
->i_lock
);
878 __inode_sb_list_add(inode
);
879 spin_unlock(&inode_lock
);
883 EXPORT_SYMBOL(new_inode
);
886 * unlock_new_inode - clear the I_NEW state and wake up any waiters
887 * @inode: new inode to unlock
889 * Called when the inode is fully initialised to clear the new state of the
890 * inode and wake up anyone waiting for the inode to finish initialisation.
892 void unlock_new_inode(struct inode
*inode
)
894 #ifdef CONFIG_DEBUG_LOCK_ALLOC
895 if (S_ISDIR(inode
->i_mode
)) {
896 struct file_system_type
*type
= inode
->i_sb
->s_type
;
898 /* Set new key only if filesystem hasn't already changed it */
899 if (!lockdep_match_class(&inode
->i_mutex
,
900 &type
->i_mutex_key
)) {
902 * ensure nobody is actually holding i_mutex
904 mutex_destroy(&inode
->i_mutex
);
905 mutex_init(&inode
->i_mutex
);
906 lockdep_set_class(&inode
->i_mutex
,
907 &type
->i_mutex_dir_key
);
911 spin_lock(&inode
->i_lock
);
912 WARN_ON(!(inode
->i_state
& I_NEW
));
913 inode
->i_state
&= ~I_NEW
;
914 wake_up_bit(&inode
->i_state
, __I_NEW
);
915 spin_unlock(&inode
->i_lock
);
917 EXPORT_SYMBOL(unlock_new_inode
);
920 * This is called without the inode lock held.. Be careful.
922 * We no longer cache the sb_flags in i_flags - see fs.h
923 * -- rmk@arm.uk.linux.org
925 static struct inode
*get_new_inode(struct super_block
*sb
,
926 struct hlist_head
*head
,
927 int (*test
)(struct inode
*, void *),
928 int (*set
)(struct inode
*, void *),
933 inode
= alloc_inode(sb
);
937 spin_lock(&inode_lock
);
938 /* We released the lock, so.. */
939 old
= find_inode(sb
, head
, test
, data
);
941 if (set(inode
, data
))
944 spin_lock(&inode
->i_lock
);
945 inode
->i_state
= I_NEW
;
946 hlist_add_head(&inode
->i_hash
, head
);
947 spin_unlock(&inode
->i_lock
);
948 __inode_sb_list_add(inode
);
949 spin_unlock(&inode_lock
);
951 /* Return the locked inode with I_NEW set, the
952 * caller is responsible for filling in the contents
958 * Uhhuh, somebody else created the same inode under
959 * us. Use the old inode instead of the one we just
962 spin_unlock(&inode_lock
);
963 destroy_inode(inode
);
965 wait_on_inode(inode
);
970 spin_unlock(&inode_lock
);
971 destroy_inode(inode
);
976 * get_new_inode_fast is the fast path version of get_new_inode, see the
977 * comment at iget_locked for details.
979 static struct inode
*get_new_inode_fast(struct super_block
*sb
,
980 struct hlist_head
*head
, unsigned long ino
)
984 inode
= alloc_inode(sb
);
988 spin_lock(&inode_lock
);
989 /* We released the lock, so.. */
990 old
= find_inode_fast(sb
, head
, ino
);
993 spin_lock(&inode
->i_lock
);
994 inode
->i_state
= I_NEW
;
995 hlist_add_head(&inode
->i_hash
, head
);
996 spin_unlock(&inode
->i_lock
);
997 __inode_sb_list_add(inode
);
998 spin_unlock(&inode_lock
);
1000 /* Return the locked inode with I_NEW set, the
1001 * caller is responsible for filling in the contents
1007 * Uhhuh, somebody else created the same inode under
1008 * us. Use the old inode instead of the one we just
1011 spin_unlock(&inode_lock
);
1012 destroy_inode(inode
);
1014 wait_on_inode(inode
);
1020 * search the inode cache for a matching inode number.
1021 * If we find one, then the inode number we are trying to
1022 * allocate is not unique and so we should not use it.
1024 * Returns 1 if the inode number is unique, 0 if it is not.
1026 static int test_inode_iunique(struct super_block
*sb
, unsigned long ino
)
1028 struct hlist_head
*b
= inode_hashtable
+ hash(sb
, ino
);
1029 struct hlist_node
*node
;
1030 struct inode
*inode
;
1032 hlist_for_each_entry(inode
, node
, b
, i_hash
) {
1033 if (inode
->i_ino
== ino
&& inode
->i_sb
== sb
)
1041 * iunique - get a unique inode number
1043 * @max_reserved: highest reserved inode number
1045 * Obtain an inode number that is unique on the system for a given
1046 * superblock. This is used by file systems that have no natural
1047 * permanent inode numbering system. An inode number is returned that
1048 * is higher than the reserved limit but unique.
1051 * With a large number of inodes live on the file system this function
1052 * currently becomes quite slow.
1054 ino_t
iunique(struct super_block
*sb
, ino_t max_reserved
)
1057 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1058 * error if st_ino won't fit in target struct field. Use 32bit counter
1059 * here to attempt to avoid that.
1061 static DEFINE_SPINLOCK(iunique_lock
);
1062 static unsigned int counter
;
1065 spin_lock(&inode_lock
);
1066 spin_lock(&iunique_lock
);
1068 if (counter
<= max_reserved
)
1069 counter
= max_reserved
+ 1;
1071 } while (!test_inode_iunique(sb
, res
));
1072 spin_unlock(&iunique_lock
);
1073 spin_unlock(&inode_lock
);
1077 EXPORT_SYMBOL(iunique
);
1079 struct inode
*igrab(struct inode
*inode
)
1081 spin_lock(&inode
->i_lock
);
1082 if (!(inode
->i_state
& (I_FREEING
|I_WILL_FREE
))) {
1084 spin_unlock(&inode
->i_lock
);
1086 spin_unlock(&inode
->i_lock
);
1088 * Handle the case where s_op->clear_inode is not been
1089 * called yet, and somebody is calling igrab
1090 * while the inode is getting freed.
1096 EXPORT_SYMBOL(igrab
);
1099 * ifind - internal function, you want ilookup5() or iget5().
1100 * @sb: super block of file system to search
1101 * @head: the head of the list to search
1102 * @test: callback used for comparisons between inodes
1103 * @data: opaque data pointer to pass to @test
1104 * @wait: if true wait for the inode to be unlocked, if false do not
1106 * ifind() searches for the inode specified by @data in the inode
1107 * cache. This is a generalized version of ifind_fast() for file systems where
1108 * the inode number is not sufficient for unique identification of an inode.
1110 * If the inode is in the cache, the inode is returned with an incremented
1113 * Otherwise NULL is returned.
1115 * Note, @test is called with the inode_lock held, so can't sleep.
1117 static struct inode
*ifind(struct super_block
*sb
,
1118 struct hlist_head
*head
, int (*test
)(struct inode
*, void *),
1119 void *data
, const int wait
)
1121 struct inode
*inode
;
1123 spin_lock(&inode_lock
);
1124 inode
= find_inode(sb
, head
, test
, data
);
1126 spin_unlock(&inode_lock
);
1128 wait_on_inode(inode
);
1131 spin_unlock(&inode_lock
);
1136 * ifind_fast - internal function, you want ilookup() or iget().
1137 * @sb: super block of file system to search
1138 * @head: head of the list to search
1139 * @ino: inode number to search for
1141 * ifind_fast() searches for the inode @ino in the inode cache. This is for
1142 * file systems where the inode number is sufficient for unique identification
1145 * If the inode is in the cache, the inode is returned with an incremented
1148 * Otherwise NULL is returned.
1150 static struct inode
*ifind_fast(struct super_block
*sb
,
1151 struct hlist_head
*head
, unsigned long ino
)
1153 struct inode
*inode
;
1155 spin_lock(&inode_lock
);
1156 inode
= find_inode_fast(sb
, head
, ino
);
1158 spin_unlock(&inode_lock
);
1159 wait_on_inode(inode
);
1162 spin_unlock(&inode_lock
);
1167 * ilookup5_nowait - search for an inode in the inode cache
1168 * @sb: super block of file system to search
1169 * @hashval: hash value (usually inode number) to search for
1170 * @test: callback used for comparisons between inodes
1171 * @data: opaque data pointer to pass to @test
1173 * ilookup5() uses ifind() to search for the inode specified by @hashval and
1174 * @data in the inode cache. This is a generalized version of ilookup() for
1175 * file systems where the inode number is not sufficient for unique
1176 * identification of an inode.
1178 * If the inode is in the cache, the inode is returned with an incremented
1179 * reference count. Note, the inode lock is not waited upon so you have to be
1180 * very careful what you do with the returned inode. You probably should be
1181 * using ilookup5() instead.
1183 * Otherwise NULL is returned.
1185 * Note, @test is called with the inode_lock held, so can't sleep.
1187 struct inode
*ilookup5_nowait(struct super_block
*sb
, unsigned long hashval
,
1188 int (*test
)(struct inode
*, void *), void *data
)
1190 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1192 return ifind(sb
, head
, test
, data
, 0);
1194 EXPORT_SYMBOL(ilookup5_nowait
);
1197 * ilookup5 - search for an inode in the inode cache
1198 * @sb: super block of file system to search
1199 * @hashval: hash value (usually inode number) to search for
1200 * @test: callback used for comparisons between inodes
1201 * @data: opaque data pointer to pass to @test
1203 * ilookup5() uses ifind() to search for the inode specified by @hashval and
1204 * @data in the inode cache. This is a generalized version of ilookup() for
1205 * file systems where the inode number is not sufficient for unique
1206 * identification of an inode.
1208 * If the inode is in the cache, the inode lock is waited upon and the inode is
1209 * returned with an incremented reference count.
1211 * Otherwise NULL is returned.
1213 * Note, @test is called with the inode_lock held, so can't sleep.
1215 struct inode
*ilookup5(struct super_block
*sb
, unsigned long hashval
,
1216 int (*test
)(struct inode
*, void *), void *data
)
1218 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1220 return ifind(sb
, head
, test
, data
, 1);
1222 EXPORT_SYMBOL(ilookup5
);
1225 * ilookup - search for an inode in the inode cache
1226 * @sb: super block of file system to search
1227 * @ino: inode number to search for
1229 * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache.
1230 * This is for file systems where the inode number is sufficient for unique
1231 * identification of an inode.
1233 * If the inode is in the cache, the inode is returned with an incremented
1236 * Otherwise NULL is returned.
1238 struct inode
*ilookup(struct super_block
*sb
, unsigned long ino
)
1240 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1242 return ifind_fast(sb
, head
, ino
);
1244 EXPORT_SYMBOL(ilookup
);
1247 * iget5_locked - obtain an inode from a mounted file system
1248 * @sb: super block of file system
1249 * @hashval: hash value (usually inode number) to get
1250 * @test: callback used for comparisons between inodes
1251 * @set: callback used to initialize a new struct inode
1252 * @data: opaque data pointer to pass to @test and @set
1254 * iget5_locked() uses ifind() to search for the inode specified by @hashval
1255 * and @data in the inode cache and if present it is returned with an increased
1256 * reference count. This is a generalized version of iget_locked() for file
1257 * systems where the inode number is not sufficient for unique identification
1260 * If the inode is not in cache, get_new_inode() is called to allocate a new
1261 * inode and this is returned locked, hashed, and with the I_NEW flag set. The
1262 * file system gets to fill it in before unlocking it via unlock_new_inode().
1264 * Note both @test and @set are called with the inode_lock held, so can't sleep.
1266 struct inode
*iget5_locked(struct super_block
*sb
, unsigned long hashval
,
1267 int (*test
)(struct inode
*, void *),
1268 int (*set
)(struct inode
*, void *), void *data
)
1270 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1271 struct inode
*inode
;
1273 inode
= ifind(sb
, head
, test
, data
, 1);
1277 * get_new_inode() will do the right thing, re-trying the search
1278 * in case it had to block at any point.
1280 return get_new_inode(sb
, head
, test
, set
, data
);
1282 EXPORT_SYMBOL(iget5_locked
);
1285 * iget_locked - obtain an inode from a mounted file system
1286 * @sb: super block of file system
1287 * @ino: inode number to get
1289 * iget_locked() uses ifind_fast() to search for the inode specified by @ino in
1290 * the inode cache and if present it is returned with an increased reference
1291 * count. This is for file systems where the inode number is sufficient for
1292 * unique identification of an inode.
1294 * If the inode is not in cache, get_new_inode_fast() is called to allocate a
1295 * new inode and this is returned locked, hashed, and with the I_NEW flag set.
1296 * The file system gets to fill it in before unlocking it via
1297 * unlock_new_inode().
1299 struct inode
*iget_locked(struct super_block
*sb
, unsigned long ino
)
1301 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1302 struct inode
*inode
;
1304 inode
= ifind_fast(sb
, head
, ino
);
1308 * get_new_inode_fast() will do the right thing, re-trying the search
1309 * in case it had to block at any point.
1311 return get_new_inode_fast(sb
, head
, ino
);
1313 EXPORT_SYMBOL(iget_locked
);
1315 int insert_inode_locked(struct inode
*inode
)
1317 struct super_block
*sb
= inode
->i_sb
;
1318 ino_t ino
= inode
->i_ino
;
1319 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1322 struct hlist_node
*node
;
1323 struct inode
*old
= NULL
;
1324 spin_lock(&inode_lock
);
1325 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1326 if (old
->i_ino
!= ino
)
1328 if (old
->i_sb
!= sb
)
1330 spin_lock(&old
->i_lock
);
1331 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
1332 spin_unlock(&old
->i_lock
);
1337 if (likely(!node
)) {
1338 spin_lock(&inode
->i_lock
);
1339 inode
->i_state
|= I_NEW
;
1340 hlist_add_head(&inode
->i_hash
, head
);
1341 spin_unlock(&inode
->i_lock
);
1342 spin_unlock(&inode_lock
);
1346 spin_unlock(&old
->i_lock
);
1347 spin_unlock(&inode_lock
);
1349 if (unlikely(!inode_unhashed(old
))) {
1356 EXPORT_SYMBOL(insert_inode_locked
);
1358 int insert_inode_locked4(struct inode
*inode
, unsigned long hashval
,
1359 int (*test
)(struct inode
*, void *), void *data
)
1361 struct super_block
*sb
= inode
->i_sb
;
1362 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1365 struct hlist_node
*node
;
1366 struct inode
*old
= NULL
;
1368 spin_lock(&inode_lock
);
1369 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1370 if (old
->i_sb
!= sb
)
1372 if (!test(old
, data
))
1374 spin_lock(&old
->i_lock
);
1375 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
1376 spin_unlock(&old
->i_lock
);
1381 if (likely(!node
)) {
1382 spin_lock(&inode
->i_lock
);
1383 inode
->i_state
|= I_NEW
;
1384 hlist_add_head(&inode
->i_hash
, head
);
1385 spin_unlock(&inode
->i_lock
);
1386 spin_unlock(&inode_lock
);
1390 spin_unlock(&old
->i_lock
);
1391 spin_unlock(&inode_lock
);
1393 if (unlikely(!inode_unhashed(old
))) {
1400 EXPORT_SYMBOL(insert_inode_locked4
);
1403 int generic_delete_inode(struct inode
*inode
)
1407 EXPORT_SYMBOL(generic_delete_inode
);
1410 * Normal UNIX filesystem behaviour: delete the
1411 * inode when the usage count drops to zero, and
1414 int generic_drop_inode(struct inode
*inode
)
1416 return !inode
->i_nlink
|| inode_unhashed(inode
);
1418 EXPORT_SYMBOL_GPL(generic_drop_inode
);
1421 * Called when we're dropping the last reference
1424 * Call the FS "drop_inode()" function, defaulting to
1425 * the legacy UNIX filesystem behaviour. If it tells
1426 * us to evict inode, do so. Otherwise, retain inode
1427 * in cache if fs is alive, sync and evict if fs is
1430 static void iput_final(struct inode
*inode
)
1432 struct super_block
*sb
= inode
->i_sb
;
1433 const struct super_operations
*op
= inode
->i_sb
->s_op
;
1436 WARN_ON(inode
->i_state
& I_NEW
);
1438 if (op
&& op
->drop_inode
)
1439 drop
= op
->drop_inode(inode
);
1441 drop
= generic_drop_inode(inode
);
1443 if (!drop
&& (sb
->s_flags
& MS_ACTIVE
)) {
1444 inode
->i_state
|= I_REFERENCED
;
1445 if (!(inode
->i_state
& (I_DIRTY
|I_SYNC
)))
1446 inode_lru_list_add(inode
);
1447 spin_unlock(&inode
->i_lock
);
1452 inode
->i_state
|= I_WILL_FREE
;
1453 spin_unlock(&inode
->i_lock
);
1454 write_inode_now(inode
, 1);
1455 spin_lock(&inode
->i_lock
);
1456 WARN_ON(inode
->i_state
& I_NEW
);
1457 inode
->i_state
&= ~I_WILL_FREE
;
1460 inode
->i_state
|= I_FREEING
;
1461 inode_lru_list_del(inode
);
1462 spin_unlock(&inode
->i_lock
);
1468 * iput - put an inode
1469 * @inode: inode to put
1471 * Puts an inode, dropping its usage count. If the inode use count hits
1472 * zero, the inode is then freed and may also be destroyed.
1474 * Consequently, iput() can sleep.
1476 void iput(struct inode
*inode
)
1479 BUG_ON(inode
->i_state
& I_CLEAR
);
1481 if (atomic_dec_and_lock(&inode
->i_count
, &inode
->i_lock
))
1485 EXPORT_SYMBOL(iput
);
1488 * bmap - find a block number in a file
1489 * @inode: inode of file
1490 * @block: block to find
1492 * Returns the block number on the device holding the inode that
1493 * is the disk block number for the block of the file requested.
1494 * That is, asked for block 4 of inode 1 the function will return the
1495 * disk block relative to the disk start that holds that block of the
1498 sector_t
bmap(struct inode
*inode
, sector_t block
)
1501 if (inode
->i_mapping
->a_ops
->bmap
)
1502 res
= inode
->i_mapping
->a_ops
->bmap(inode
->i_mapping
, block
);
1505 EXPORT_SYMBOL(bmap
);
1508 * With relative atime, only update atime if the previous atime is
1509 * earlier than either the ctime or mtime or if at least a day has
1510 * passed since the last atime update.
1512 static int relatime_need_update(struct vfsmount
*mnt
, struct inode
*inode
,
1513 struct timespec now
)
1516 if (!(mnt
->mnt_flags
& MNT_RELATIME
))
1519 * Is mtime younger than atime? If yes, update atime:
1521 if (timespec_compare(&inode
->i_mtime
, &inode
->i_atime
) >= 0)
1524 * Is ctime younger than atime? If yes, update atime:
1526 if (timespec_compare(&inode
->i_ctime
, &inode
->i_atime
) >= 0)
1530 * Is the previous atime value older than a day? If yes,
1533 if ((long)(now
.tv_sec
- inode
->i_atime
.tv_sec
) >= 24*60*60)
1536 * Good, we can skip the atime update:
1542 * touch_atime - update the access time
1543 * @mnt: mount the inode is accessed on
1544 * @dentry: dentry accessed
1546 * Update the accessed time on an inode and mark it for writeback.
1547 * This function automatically handles read only file systems and media,
1548 * as well as the "noatime" flag and inode specific "noatime" markers.
1550 void touch_atime(struct vfsmount
*mnt
, struct dentry
*dentry
)
1552 struct inode
*inode
= dentry
->d_inode
;
1553 struct timespec now
;
1555 if (inode
->i_flags
& S_NOATIME
)
1557 if (IS_NOATIME(inode
))
1559 if ((inode
->i_sb
->s_flags
& MS_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1562 if (mnt
->mnt_flags
& MNT_NOATIME
)
1564 if ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1567 now
= current_fs_time(inode
->i_sb
);
1569 if (!relatime_need_update(mnt
, inode
, now
))
1572 if (timespec_equal(&inode
->i_atime
, &now
))
1575 if (mnt_want_write(mnt
))
1578 inode
->i_atime
= now
;
1579 mark_inode_dirty_sync(inode
);
1580 mnt_drop_write(mnt
);
1582 EXPORT_SYMBOL(touch_atime
);
1585 * file_update_time - update mtime and ctime time
1586 * @file: file accessed
1588 * Update the mtime and ctime members of an inode and mark the inode
1589 * for writeback. Note that this function is meant exclusively for
1590 * usage in the file write path of filesystems, and filesystems may
1591 * choose to explicitly ignore update via this function with the
1592 * S_NOCMTIME inode flag, e.g. for network filesystem where these
1593 * timestamps are handled by the server.
1596 void file_update_time(struct file
*file
)
1598 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1599 struct timespec now
;
1600 enum { S_MTIME
= 1, S_CTIME
= 2, S_VERSION
= 4 } sync_it
= 0;
1602 /* First try to exhaust all avenues to not sync */
1603 if (IS_NOCMTIME(inode
))
1606 now
= current_fs_time(inode
->i_sb
);
1607 if (!timespec_equal(&inode
->i_mtime
, &now
))
1610 if (!timespec_equal(&inode
->i_ctime
, &now
))
1613 if (IS_I_VERSION(inode
))
1614 sync_it
|= S_VERSION
;
1619 /* Finally allowed to write? Takes lock. */
1620 if (mnt_want_write_file(file
))
1623 /* Only change inode inside the lock region */
1624 if (sync_it
& S_VERSION
)
1625 inode_inc_iversion(inode
);
1626 if (sync_it
& S_CTIME
)
1627 inode
->i_ctime
= now
;
1628 if (sync_it
& S_MTIME
)
1629 inode
->i_mtime
= now
;
1630 mark_inode_dirty_sync(inode
);
1631 mnt_drop_write(file
->f_path
.mnt
);
1633 EXPORT_SYMBOL(file_update_time
);
1635 int inode_needs_sync(struct inode
*inode
)
1639 if (S_ISDIR(inode
->i_mode
) && IS_DIRSYNC(inode
))
1643 EXPORT_SYMBOL(inode_needs_sync
);
1645 int inode_wait(void *word
)
1650 EXPORT_SYMBOL(inode_wait
);
1653 * If we try to find an inode in the inode hash while it is being
1654 * deleted, we have to wait until the filesystem completes its
1655 * deletion before reporting that it isn't found. This function waits
1656 * until the deletion _might_ have completed. Callers are responsible
1657 * to recheck inode state.
1659 * It doesn't matter if I_NEW is not set initially, a call to
1660 * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
1663 static void __wait_on_freeing_inode(struct inode
*inode
)
1665 wait_queue_head_t
*wq
;
1666 DEFINE_WAIT_BIT(wait
, &inode
->i_state
, __I_NEW
);
1667 wq
= bit_waitqueue(&inode
->i_state
, __I_NEW
);
1668 prepare_to_wait(wq
, &wait
.wait
, TASK_UNINTERRUPTIBLE
);
1669 spin_unlock(&inode
->i_lock
);
1670 spin_unlock(&inode_lock
);
1672 finish_wait(wq
, &wait
.wait
);
1673 spin_lock(&inode_lock
);
1676 static __initdata
unsigned long ihash_entries
;
1677 static int __init
set_ihash_entries(char *str
)
1681 ihash_entries
= simple_strtoul(str
, &str
, 0);
1684 __setup("ihash_entries=", set_ihash_entries
);
1687 * Initialize the waitqueues and inode hash table.
1689 void __init
inode_init_early(void)
1693 /* If hashes are distributed across NUMA nodes, defer
1694 * hash allocation until vmalloc space is available.
1700 alloc_large_system_hash("Inode-cache",
1701 sizeof(struct hlist_head
),
1709 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1710 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1713 void __init
inode_init(void)
1717 /* inode slab cache */
1718 inode_cachep
= kmem_cache_create("inode_cache",
1719 sizeof(struct inode
),
1721 (SLAB_RECLAIM_ACCOUNT
|SLAB_PANIC
|
1724 register_shrinker(&icache_shrinker
);
1726 /* Hash may have been set up in inode_init_early */
1731 alloc_large_system_hash("Inode-cache",
1732 sizeof(struct hlist_head
),
1740 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1741 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1744 void init_special_inode(struct inode
*inode
, umode_t mode
, dev_t rdev
)
1746 inode
->i_mode
= mode
;
1747 if (S_ISCHR(mode
)) {
1748 inode
->i_fop
= &def_chr_fops
;
1749 inode
->i_rdev
= rdev
;
1750 } else if (S_ISBLK(mode
)) {
1751 inode
->i_fop
= &def_blk_fops
;
1752 inode
->i_rdev
= rdev
;
1753 } else if (S_ISFIFO(mode
))
1754 inode
->i_fop
= &def_fifo_fops
;
1755 else if (S_ISSOCK(mode
))
1756 inode
->i_fop
= &bad_sock_fops
;
1758 printk(KERN_DEBUG
"init_special_inode: bogus i_mode (%o) for"
1759 " inode %s:%lu\n", mode
, inode
->i_sb
->s_id
,
1762 EXPORT_SYMBOL(init_special_inode
);
1765 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
1767 * @dir: Directory inode
1768 * @mode: mode of the new inode
1770 void inode_init_owner(struct inode
*inode
, const struct inode
*dir
,
1773 inode
->i_uid
= current_fsuid();
1774 if (dir
&& dir
->i_mode
& S_ISGID
) {
1775 inode
->i_gid
= dir
->i_gid
;
1779 inode
->i_gid
= current_fsgid();
1780 inode
->i_mode
= mode
;
1782 EXPORT_SYMBOL(inode_init_owner
);
1785 * inode_owner_or_capable - check current task permissions to inode
1786 * @inode: inode being checked
1788 * Return true if current either has CAP_FOWNER to the inode, or
1791 bool inode_owner_or_capable(const struct inode
*inode
)
1793 struct user_namespace
*ns
= inode_userns(inode
);
1795 if (current_user_ns() == ns
&& current_fsuid() == inode
->i_uid
)
1797 if (ns_capable(ns
, CAP_FOWNER
))
1801 EXPORT_SYMBOL(inode_owner_or_capable
);