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/prefetch.h>
28 #include <linux/ima.h>
29 #include <linux/cred.h>
33 * inode locking rules.
35 * inode->i_lock protects:
36 * inode->i_state, inode->i_hash, __iget()
37 * inode_lru_lock protects:
38 * inode_lru, inode->i_lru
39 * inode_sb_list_lock protects:
40 * sb->s_inodes, inode->i_sb_list
41 * inode_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
64 * This is needed for the following functions:
68 * FIXME: remove all knowledge of the buffer layer from this file
70 #include <linux/buffer_head.h>
73 * New inode.c implementation.
75 * This implementation has the basic premise of trying
76 * to be extremely low-overhead and SMP-safe, yet be
77 * simple enough to be "obviously correct".
82 /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */
84 /* #define INODE_PARANOIA 1 */
85 /* #define INODE_DEBUG 1 */
88 * Inode lookup is no longer as critical as it used to be:
89 * most of the lookups are going to be through the dcache.
91 #define I_HASHBITS i_hash_shift
92 #define I_HASHMASK i_hash_mask
94 static unsigned int i_hash_mask __read_mostly
;
95 static unsigned int i_hash_shift __read_mostly
;
96 static struct hlist_head
*inode_hashtable __read_mostly
;
97 static __cacheline_aligned_in_smp
DEFINE_SPINLOCK(inode_hash_lock
);
100 * Each inode can be on two separate lists. One is
101 * the hash list of the inode, used for lookups. The
102 * other linked list is the "type" list:
103 * "in_use" - valid inode, i_count > 0, i_nlink > 0
104 * "dirty" - as "in_use" but also dirty
105 * "unused" - valid inode, i_count = 0
107 * A "dirty" list is maintained for each super block,
108 * allowing for low-overhead inode sync() operations.
111 static LIST_HEAD(inode_lru
);
112 static DEFINE_SPINLOCK(inode_lru_lock
);
114 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(inode_sb_list_lock
);
115 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(inode_wb_list_lock
);
118 * iprune_sem provides exclusion between the icache shrinking and the
121 * We don't actually need it to protect anything in the umount path,
122 * but only need to cycle through it to make sure any inode that
123 * prune_icache took off the LRU list has been fully torn down by the
124 * time we are past evict_inodes.
126 static DECLARE_RWSEM(iprune_sem
);
129 * Empty aops. Can be used for the cases where the user does not
130 * define any of the address_space operations.
132 const struct address_space_operations empty_aops
= {
134 EXPORT_SYMBOL(empty_aops
);
137 * Statistics gathering..
139 struct inodes_stat_t inodes_stat
;
141 static DEFINE_PER_CPU(unsigned int, nr_inodes
);
143 static struct kmem_cache
*inode_cachep __read_mostly
;
145 static int get_nr_inodes(void)
149 for_each_possible_cpu(i
)
150 sum
+= per_cpu(nr_inodes
, i
);
151 return sum
< 0 ? 0 : sum
;
154 static inline int get_nr_inodes_unused(void)
156 return inodes_stat
.nr_unused
;
159 int get_nr_dirty_inodes(void)
161 /* not actually dirty inodes, but a wild approximation */
162 int nr_dirty
= get_nr_inodes() - get_nr_inodes_unused();
163 return nr_dirty
> 0 ? nr_dirty
: 0;
167 * Handle nr_inode sysctl
170 int proc_nr_inodes(ctl_table
*table
, int write
,
171 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
173 inodes_stat
.nr_inodes
= get_nr_inodes();
174 return proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
179 * inode_init_always - perform inode structure intialisation
180 * @sb: superblock inode belongs to
181 * @inode: inode to initialise
183 * These are initializations that need to be done on every inode
184 * allocation as the fields are not initialised by slab allocation.
186 int inode_init_always(struct super_block
*sb
, struct inode
*inode
)
188 static const struct inode_operations empty_iops
;
189 static const struct file_operations empty_fops
;
190 struct address_space
*const mapping
= &inode
->i_data
;
193 inode
->i_blkbits
= sb
->s_blocksize_bits
;
195 atomic_set(&inode
->i_count
, 1);
196 inode
->i_op
= &empty_iops
;
197 inode
->i_fop
= &empty_fops
;
201 atomic_set(&inode
->i_writecount
, 0);
205 inode
->i_generation
= 0;
207 memset(&inode
->i_dquot
, 0, sizeof(inode
->i_dquot
));
209 inode
->i_pipe
= NULL
;
210 inode
->i_bdev
= NULL
;
211 inode
->i_cdev
= NULL
;
213 inode
->dirtied_when
= 0;
215 if (security_inode_alloc(inode
))
217 spin_lock_init(&inode
->i_lock
);
218 lockdep_set_class(&inode
->i_lock
, &sb
->s_type
->i_lock_key
);
220 mutex_init(&inode
->i_mutex
);
221 lockdep_set_class(&inode
->i_mutex
, &sb
->s_type
->i_mutex_key
);
223 init_rwsem(&inode
->i_alloc_sem
);
224 lockdep_set_class(&inode
->i_alloc_sem
, &sb
->s_type
->i_alloc_sem_key
);
226 mapping
->a_ops
= &empty_aops
;
227 mapping
->host
= inode
;
229 mapping_set_gfp_mask(mapping
, GFP_HIGHUSER_MOVABLE
);
230 mapping
->assoc_mapping
= NULL
;
231 mapping
->backing_dev_info
= &default_backing_dev_info
;
232 mapping
->writeback_index
= 0;
235 * If the block_device provides a backing_dev_info for client
236 * inodes then use that. Otherwise the inode share the bdev's
240 struct backing_dev_info
*bdi
;
242 bdi
= sb
->s_bdev
->bd_inode
->i_mapping
->backing_dev_info
;
243 mapping
->backing_dev_info
= bdi
;
245 inode
->i_private
= NULL
;
246 inode
->i_mapping
= mapping
;
247 #ifdef CONFIG_FS_POSIX_ACL
248 inode
->i_acl
= inode
->i_default_acl
= ACL_NOT_CACHED
;
251 #ifdef CONFIG_FSNOTIFY
252 inode
->i_fsnotify_mask
= 0;
255 this_cpu_inc(nr_inodes
);
261 EXPORT_SYMBOL(inode_init_always
);
263 static struct inode
*alloc_inode(struct super_block
*sb
)
267 if (sb
->s_op
->alloc_inode
)
268 inode
= sb
->s_op
->alloc_inode(sb
);
270 inode
= kmem_cache_alloc(inode_cachep
, GFP_KERNEL
);
275 if (unlikely(inode_init_always(sb
, inode
))) {
276 if (inode
->i_sb
->s_op
->destroy_inode
)
277 inode
->i_sb
->s_op
->destroy_inode(inode
);
279 kmem_cache_free(inode_cachep
, inode
);
286 void free_inode_nonrcu(struct inode
*inode
)
288 kmem_cache_free(inode_cachep
, inode
);
290 EXPORT_SYMBOL(free_inode_nonrcu
);
292 void __destroy_inode(struct inode
*inode
)
294 BUG_ON(inode_has_buffers(inode
));
295 security_inode_free(inode
);
296 fsnotify_inode_delete(inode
);
297 #ifdef CONFIG_FS_POSIX_ACL
298 if (inode
->i_acl
&& inode
->i_acl
!= ACL_NOT_CACHED
)
299 posix_acl_release(inode
->i_acl
);
300 if (inode
->i_default_acl
&& inode
->i_default_acl
!= ACL_NOT_CACHED
)
301 posix_acl_release(inode
->i_default_acl
);
303 this_cpu_dec(nr_inodes
);
305 EXPORT_SYMBOL(__destroy_inode
);
307 static void i_callback(struct rcu_head
*head
)
309 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
310 INIT_LIST_HEAD(&inode
->i_dentry
);
311 kmem_cache_free(inode_cachep
, inode
);
314 static void destroy_inode(struct inode
*inode
)
316 BUG_ON(!list_empty(&inode
->i_lru
));
317 __destroy_inode(inode
);
318 if (inode
->i_sb
->s_op
->destroy_inode
)
319 inode
->i_sb
->s_op
->destroy_inode(inode
);
321 call_rcu(&inode
->i_rcu
, i_callback
);
324 void address_space_init_once(struct address_space
*mapping
)
326 memset(mapping
, 0, sizeof(*mapping
));
327 INIT_RADIX_TREE(&mapping
->page_tree
, GFP_ATOMIC
);
328 spin_lock_init(&mapping
->tree_lock
);
329 spin_lock_init(&mapping
->i_mmap_lock
);
330 INIT_LIST_HEAD(&mapping
->private_list
);
331 spin_lock_init(&mapping
->private_lock
);
332 INIT_RAW_PRIO_TREE_ROOT(&mapping
->i_mmap
);
333 INIT_LIST_HEAD(&mapping
->i_mmap_nonlinear
);
334 mutex_init(&mapping
->unmap_mutex
);
336 EXPORT_SYMBOL(address_space_init_once
);
339 * These are initializations that only need to be done
340 * once, because the fields are idempotent across use
341 * of the inode, so let the slab aware of that.
343 void inode_init_once(struct inode
*inode
)
345 memset(inode
, 0, sizeof(*inode
));
346 INIT_HLIST_NODE(&inode
->i_hash
);
347 INIT_LIST_HEAD(&inode
->i_dentry
);
348 INIT_LIST_HEAD(&inode
->i_devices
);
349 INIT_LIST_HEAD(&inode
->i_wb_list
);
350 INIT_LIST_HEAD(&inode
->i_lru
);
351 address_space_init_once(&inode
->i_data
);
352 i_size_ordered_init(inode
);
353 #ifdef CONFIG_FSNOTIFY
354 INIT_HLIST_HEAD(&inode
->i_fsnotify_marks
);
357 EXPORT_SYMBOL(inode_init_once
);
359 static void init_once(void *foo
)
361 struct inode
*inode
= (struct inode
*) foo
;
363 inode_init_once(inode
);
367 * inode->i_lock must be held
369 void __iget(struct inode
*inode
)
371 atomic_inc(&inode
->i_count
);
375 * get additional reference to inode; caller must already hold one.
377 void ihold(struct inode
*inode
)
379 WARN_ON(atomic_inc_return(&inode
->i_count
) < 2);
381 EXPORT_SYMBOL(ihold
);
383 static void inode_lru_list_add(struct inode
*inode
)
385 spin_lock(&inode_lru_lock
);
386 if (list_empty(&inode
->i_lru
)) {
387 list_add(&inode
->i_lru
, &inode_lru
);
388 inodes_stat
.nr_unused
++;
390 spin_unlock(&inode_lru_lock
);
393 static void inode_lru_list_del(struct inode
*inode
)
395 spin_lock(&inode_lru_lock
);
396 if (!list_empty(&inode
->i_lru
)) {
397 list_del_init(&inode
->i_lru
);
398 inodes_stat
.nr_unused
--;
400 spin_unlock(&inode_lru_lock
);
404 * inode_sb_list_add - add inode to the superblock list of inodes
405 * @inode: inode to add
407 void inode_sb_list_add(struct inode
*inode
)
409 spin_lock(&inode_sb_list_lock
);
410 list_add(&inode
->i_sb_list
, &inode
->i_sb
->s_inodes
);
411 spin_unlock(&inode_sb_list_lock
);
413 EXPORT_SYMBOL_GPL(inode_sb_list_add
);
415 static inline void inode_sb_list_del(struct inode
*inode
)
417 spin_lock(&inode_sb_list_lock
);
418 list_del_init(&inode
->i_sb_list
);
419 spin_unlock(&inode_sb_list_lock
);
422 static unsigned long hash(struct super_block
*sb
, unsigned long hashval
)
426 tmp
= (hashval
* (unsigned long)sb
) ^ (GOLDEN_RATIO_PRIME
+ hashval
) /
428 tmp
= tmp
^ ((tmp
^ GOLDEN_RATIO_PRIME
) >> I_HASHBITS
);
429 return tmp
& I_HASHMASK
;
433 * __insert_inode_hash - hash an inode
434 * @inode: unhashed inode
435 * @hashval: unsigned long value used to locate this object in the
438 * Add an inode to the inode hash for this superblock.
440 void __insert_inode_hash(struct inode
*inode
, unsigned long hashval
)
442 struct hlist_head
*b
= inode_hashtable
+ hash(inode
->i_sb
, hashval
);
444 spin_lock(&inode_hash_lock
);
445 spin_lock(&inode
->i_lock
);
446 hlist_add_head(&inode
->i_hash
, b
);
447 spin_unlock(&inode
->i_lock
);
448 spin_unlock(&inode_hash_lock
);
450 EXPORT_SYMBOL(__insert_inode_hash
);
453 * remove_inode_hash - remove an inode from the hash
454 * @inode: inode to unhash
456 * Remove an inode from the superblock.
458 void remove_inode_hash(struct inode
*inode
)
460 spin_lock(&inode_hash_lock
);
461 spin_lock(&inode
->i_lock
);
462 hlist_del_init(&inode
->i_hash
);
463 spin_unlock(&inode
->i_lock
);
464 spin_unlock(&inode_hash_lock
);
466 EXPORT_SYMBOL(remove_inode_hash
);
468 void end_writeback(struct inode
*inode
)
471 BUG_ON(inode
->i_data
.nrpages
);
472 BUG_ON(!list_empty(&inode
->i_data
.private_list
));
473 BUG_ON(!(inode
->i_state
& I_FREEING
));
474 BUG_ON(inode
->i_state
& I_CLEAR
);
475 inode_sync_wait(inode
);
476 /* don't need i_lock here, no concurrent mods to i_state */
477 inode
->i_state
= I_FREEING
| I_CLEAR
;
479 EXPORT_SYMBOL(end_writeback
);
482 * Free the inode passed in, removing it from the lists it is still connected
483 * to. We remove any pages still attached to the inode and wait for any IO that
484 * is still in progress before finally destroying the inode.
486 * An inode must already be marked I_FREEING so that we avoid the inode being
487 * moved back onto lists if we race with other code that manipulates the lists
488 * (e.g. writeback_single_inode). The caller is responsible for setting this.
490 * An inode must already be removed from the LRU list before being evicted from
491 * the cache. This should occur atomically with setting the I_FREEING state
492 * flag, so no inodes here should ever be on the LRU when being evicted.
494 static void evict(struct inode
*inode
)
496 const struct super_operations
*op
= inode
->i_sb
->s_op
;
498 BUG_ON(!(inode
->i_state
& I_FREEING
));
499 BUG_ON(!list_empty(&inode
->i_lru
));
501 inode_wb_list_del(inode
);
502 inode_sb_list_del(inode
);
504 if (op
->evict_inode
) {
505 op
->evict_inode(inode
);
507 if (inode
->i_data
.nrpages
)
508 truncate_inode_pages(&inode
->i_data
, 0);
509 end_writeback(inode
);
511 if (S_ISBLK(inode
->i_mode
) && inode
->i_bdev
)
513 if (S_ISCHR(inode
->i_mode
) && inode
->i_cdev
)
516 remove_inode_hash(inode
);
518 spin_lock(&inode
->i_lock
);
519 wake_up_bit(&inode
->i_state
, __I_NEW
);
520 BUG_ON(inode
->i_state
!= (I_FREEING
| I_CLEAR
));
521 spin_unlock(&inode
->i_lock
);
523 destroy_inode(inode
);
527 * dispose_list - dispose of the contents of a local list
528 * @head: the head of the list to free
530 * Dispose-list gets a local list with local inodes in it, so it doesn't
531 * need to worry about list corruption and SMP locks.
533 static void dispose_list(struct list_head
*head
)
535 while (!list_empty(head
)) {
538 inode
= list_first_entry(head
, struct inode
, i_lru
);
539 list_del_init(&inode
->i_lru
);
546 * evict_inodes - evict all evictable inodes for a superblock
547 * @sb: superblock to operate on
549 * Make sure that no inodes with zero refcount are retained. This is
550 * called by superblock shutdown after having MS_ACTIVE flag removed,
551 * so any inode reaching zero refcount during or after that call will
552 * be immediately evicted.
554 void evict_inodes(struct super_block
*sb
)
556 struct inode
*inode
, *next
;
559 spin_lock(&inode_sb_list_lock
);
560 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
561 if (atomic_read(&inode
->i_count
))
564 spin_lock(&inode
->i_lock
);
565 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
566 spin_unlock(&inode
->i_lock
);
570 inode
->i_state
|= I_FREEING
;
571 inode_lru_list_del(inode
);
572 spin_unlock(&inode
->i_lock
);
573 list_add(&inode
->i_lru
, &dispose
);
575 spin_unlock(&inode_sb_list_lock
);
577 dispose_list(&dispose
);
580 * Cycle through iprune_sem to make sure any inode that prune_icache
581 * moved off the list before we took the lock has been fully torn
584 down_write(&iprune_sem
);
585 up_write(&iprune_sem
);
589 * invalidate_inodes - attempt to free all inodes on a superblock
590 * @sb: superblock to operate on
591 * @kill_dirty: flag to guide handling of dirty inodes
593 * Attempts to free all inodes for a given superblock. If there were any
594 * busy inodes return a non-zero value, else zero.
595 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
598 int invalidate_inodes(struct super_block
*sb
, bool kill_dirty
)
601 struct inode
*inode
, *next
;
604 spin_lock(&inode_sb_list_lock
);
605 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
606 spin_lock(&inode
->i_lock
);
607 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
608 spin_unlock(&inode
->i_lock
);
611 if (inode
->i_state
& I_DIRTY
&& !kill_dirty
) {
612 spin_unlock(&inode
->i_lock
);
616 if (atomic_read(&inode
->i_count
)) {
617 spin_unlock(&inode
->i_lock
);
622 inode
->i_state
|= I_FREEING
;
623 inode_lru_list_del(inode
);
624 spin_unlock(&inode
->i_lock
);
625 list_add(&inode
->i_lru
, &dispose
);
627 spin_unlock(&inode_sb_list_lock
);
629 dispose_list(&dispose
);
634 static int can_unuse(struct inode
*inode
)
636 if (inode
->i_state
& ~I_REFERENCED
)
638 if (inode_has_buffers(inode
))
640 if (atomic_read(&inode
->i_count
))
642 if (inode
->i_data
.nrpages
)
648 * Scan `goal' inodes on the unused list for freeable ones. They are moved to a
649 * temporary list and then are freed outside inode_lru_lock by dispose_list().
651 * Any inodes which are pinned purely because of attached pagecache have their
652 * pagecache removed. If the inode has metadata buffers attached to
653 * mapping->private_list then try to remove them.
655 * If the inode has the I_REFERENCED flag set, then it means that it has been
656 * used recently - the flag is set in iput_final(). When we encounter such an
657 * inode, clear the flag and move it to the back of the LRU so it gets another
658 * pass through the LRU before it gets reclaimed. This is necessary because of
659 * the fact we are doing lazy LRU updates to minimise lock contention so the
660 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
661 * with this flag set because they are the inodes that are out of order.
663 static void prune_icache(int nr_to_scan
)
667 unsigned long reap
= 0;
669 down_read(&iprune_sem
);
670 spin_lock(&inode_lru_lock
);
671 for (nr_scanned
= 0; nr_scanned
< nr_to_scan
; nr_scanned
++) {
674 if (list_empty(&inode_lru
))
677 inode
= list_entry(inode_lru
.prev
, struct inode
, i_lru
);
680 * we are inverting the inode_lru_lock/inode->i_lock here,
681 * so use a trylock. If we fail to get the lock, just move the
682 * inode to the back of the list so we don't spin on it.
684 if (!spin_trylock(&inode
->i_lock
)) {
685 list_move(&inode
->i_lru
, &inode_lru
);
690 * Referenced or dirty inodes are still in use. Give them
691 * another pass through the LRU as we canot reclaim them now.
693 if (atomic_read(&inode
->i_count
) ||
694 (inode
->i_state
& ~I_REFERENCED
)) {
695 list_del_init(&inode
->i_lru
);
696 spin_unlock(&inode
->i_lock
);
697 inodes_stat
.nr_unused
--;
701 /* recently referenced inodes get one more pass */
702 if (inode
->i_state
& I_REFERENCED
) {
703 inode
->i_state
&= ~I_REFERENCED
;
704 list_move(&inode
->i_lru
, &inode_lru
);
705 spin_unlock(&inode
->i_lock
);
708 if (inode_has_buffers(inode
) || inode
->i_data
.nrpages
) {
710 spin_unlock(&inode
->i_lock
);
711 spin_unlock(&inode_lru_lock
);
712 if (remove_inode_buffers(inode
))
713 reap
+= invalidate_mapping_pages(&inode
->i_data
,
716 spin_lock(&inode_lru_lock
);
718 if (inode
!= list_entry(inode_lru
.next
,
719 struct inode
, i_lru
))
720 continue; /* wrong inode or list_empty */
721 /* avoid lock inversions with trylock */
722 if (!spin_trylock(&inode
->i_lock
))
724 if (!can_unuse(inode
)) {
725 spin_unlock(&inode
->i_lock
);
729 WARN_ON(inode
->i_state
& I_NEW
);
730 inode
->i_state
|= I_FREEING
;
731 spin_unlock(&inode
->i_lock
);
733 list_move(&inode
->i_lru
, &freeable
);
734 inodes_stat
.nr_unused
--;
736 if (current_is_kswapd())
737 __count_vm_events(KSWAPD_INODESTEAL
, reap
);
739 __count_vm_events(PGINODESTEAL
, reap
);
740 spin_unlock(&inode_lru_lock
);
742 dispose_list(&freeable
);
743 up_read(&iprune_sem
);
747 * shrink_icache_memory() will attempt to reclaim some unused inodes. Here,
748 * "unused" means that no dentries are referring to the inodes: the files are
749 * not open and the dcache references to those inodes have already been
752 * This function is passed the number of inodes to scan, and it returns the
753 * total number of remaining possibly-reclaimable inodes.
755 static int shrink_icache_memory(struct shrinker
*shrink
, int nr
, gfp_t gfp_mask
)
759 * Nasty deadlock avoidance. We may hold various FS locks,
760 * and we don't want to recurse into the FS that called us
761 * in clear_inode() and friends..
763 if (!(gfp_mask
& __GFP_FS
))
767 return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure
;
770 static struct shrinker icache_shrinker
= {
771 .shrink
= shrink_icache_memory
,
772 .seeks
= DEFAULT_SEEKS
,
775 static void __wait_on_freeing_inode(struct inode
*inode
);
777 * Called with the inode lock held.
779 static struct inode
*find_inode(struct super_block
*sb
,
780 struct hlist_head
*head
,
781 int (*test
)(struct inode
*, void *),
784 struct hlist_node
*node
;
785 struct inode
*inode
= NULL
;
788 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
789 spin_lock(&inode
->i_lock
);
790 if (inode
->i_sb
!= sb
) {
791 spin_unlock(&inode
->i_lock
);
794 if (!test(inode
, data
)) {
795 spin_unlock(&inode
->i_lock
);
798 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
799 __wait_on_freeing_inode(inode
);
803 spin_unlock(&inode
->i_lock
);
810 * find_inode_fast is the fast path version of find_inode, see the comment at
811 * iget_locked for details.
813 static struct inode
*find_inode_fast(struct super_block
*sb
,
814 struct hlist_head
*head
, unsigned long ino
)
816 struct hlist_node
*node
;
817 struct inode
*inode
= NULL
;
820 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
821 spin_lock(&inode
->i_lock
);
822 if (inode
->i_ino
!= ino
) {
823 spin_unlock(&inode
->i_lock
);
826 if (inode
->i_sb
!= sb
) {
827 spin_unlock(&inode
->i_lock
);
830 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
831 __wait_on_freeing_inode(inode
);
835 spin_unlock(&inode
->i_lock
);
842 * Each cpu owns a range of LAST_INO_BATCH numbers.
843 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
844 * to renew the exhausted range.
846 * This does not significantly increase overflow rate because every CPU can
847 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
848 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
849 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
850 * overflow rate by 2x, which does not seem too significant.
852 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
853 * error if st_ino won't fit in target struct field. Use 32bit counter
854 * here to attempt to avoid that.
856 #define LAST_INO_BATCH 1024
857 static DEFINE_PER_CPU(unsigned int, last_ino
);
859 unsigned int get_next_ino(void)
861 unsigned int *p
= &get_cpu_var(last_ino
);
862 unsigned int res
= *p
;
865 if (unlikely((res
& (LAST_INO_BATCH
-1)) == 0)) {
866 static atomic_t shared_last_ino
;
867 int next
= atomic_add_return(LAST_INO_BATCH
, &shared_last_ino
);
869 res
= next
- LAST_INO_BATCH
;
874 put_cpu_var(last_ino
);
877 EXPORT_SYMBOL(get_next_ino
);
880 * new_inode - obtain an inode
883 * Allocates a new inode for given superblock. The default gfp_mask
884 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
885 * If HIGHMEM pages are unsuitable or it is known that pages allocated
886 * for the page cache are not reclaimable or migratable,
887 * mapping_set_gfp_mask() must be called with suitable flags on the
888 * newly created inode's mapping
891 struct inode
*new_inode(struct super_block
*sb
)
895 spin_lock_prefetch(&inode_sb_list_lock
);
897 inode
= alloc_inode(sb
);
899 spin_lock(&inode
->i_lock
);
901 spin_unlock(&inode
->i_lock
);
902 inode_sb_list_add(inode
);
906 EXPORT_SYMBOL(new_inode
);
909 * unlock_new_inode - clear the I_NEW state and wake up any waiters
910 * @inode: new inode to unlock
912 * Called when the inode is fully initialised to clear the new state of the
913 * inode and wake up anyone waiting for the inode to finish initialisation.
915 void unlock_new_inode(struct inode
*inode
)
917 #ifdef CONFIG_DEBUG_LOCK_ALLOC
918 if (S_ISDIR(inode
->i_mode
)) {
919 struct file_system_type
*type
= inode
->i_sb
->s_type
;
921 /* Set new key only if filesystem hasn't already changed it */
922 if (!lockdep_match_class(&inode
->i_mutex
,
923 &type
->i_mutex_key
)) {
925 * ensure nobody is actually holding i_mutex
927 mutex_destroy(&inode
->i_mutex
);
928 mutex_init(&inode
->i_mutex
);
929 lockdep_set_class(&inode
->i_mutex
,
930 &type
->i_mutex_dir_key
);
934 spin_lock(&inode
->i_lock
);
935 WARN_ON(!(inode
->i_state
& I_NEW
));
936 inode
->i_state
&= ~I_NEW
;
937 wake_up_bit(&inode
->i_state
, __I_NEW
);
938 spin_unlock(&inode
->i_lock
);
940 EXPORT_SYMBOL(unlock_new_inode
);
943 * iget5_locked - obtain an inode from a mounted file system
944 * @sb: super block of file system
945 * @hashval: hash value (usually inode number) to get
946 * @test: callback used for comparisons between inodes
947 * @set: callback used to initialize a new struct inode
948 * @data: opaque data pointer to pass to @test and @set
950 * Search for the inode specified by @hashval and @data in the inode cache,
951 * and if present it is return it with an increased reference count. This is
952 * a generalized version of iget_locked() for file systems where the inode
953 * number is not sufficient for unique identification of an inode.
955 * If the inode is not in cache, allocate a new inode and return it locked,
956 * hashed, and with the I_NEW flag set. The file system gets to fill it in
957 * before unlocking it via unlock_new_inode().
959 * Note both @test and @set are called with the inode_hash_lock held, so can't
962 struct inode
*iget5_locked(struct super_block
*sb
, unsigned long hashval
,
963 int (*test
)(struct inode
*, void *),
964 int (*set
)(struct inode
*, void *), void *data
)
966 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
969 spin_lock(&inode_hash_lock
);
970 inode
= find_inode(sb
, head
, test
, data
);
971 spin_unlock(&inode_hash_lock
);
974 wait_on_inode(inode
);
978 inode
= alloc_inode(sb
);
982 spin_lock(&inode_hash_lock
);
983 /* We released the lock, so.. */
984 old
= find_inode(sb
, head
, test
, data
);
986 if (set(inode
, data
))
989 spin_lock(&inode
->i_lock
);
990 inode
->i_state
= I_NEW
;
991 hlist_add_head(&inode
->i_hash
, head
);
992 spin_unlock(&inode
->i_lock
);
993 inode_sb_list_add(inode
);
994 spin_unlock(&inode_hash_lock
);
996 /* Return the locked inode with I_NEW set, the
997 * caller is responsible for filling in the contents
1003 * Uhhuh, somebody else created the same inode under
1004 * us. Use the old inode instead of the one we just
1007 spin_unlock(&inode_hash_lock
);
1008 destroy_inode(inode
);
1010 wait_on_inode(inode
);
1015 spin_unlock(&inode_hash_lock
);
1016 destroy_inode(inode
);
1019 EXPORT_SYMBOL(iget5_locked
);
1022 * iget_locked - obtain an inode from a mounted file system
1023 * @sb: super block of file system
1024 * @ino: inode number to get
1026 * Search for the inode specified by @ino in the inode cache and if present
1027 * return it with an increased reference count. This is for file systems
1028 * where the inode number is sufficient for unique identification of an inode.
1030 * If the inode is not in cache, allocate a new inode and return it locked,
1031 * hashed, and with the I_NEW flag set. The file system gets to fill it in
1032 * before unlocking it via unlock_new_inode().
1034 struct inode
*iget_locked(struct super_block
*sb
, unsigned long ino
)
1036 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1037 struct inode
*inode
;
1039 spin_lock(&inode_hash_lock
);
1040 inode
= find_inode_fast(sb
, head
, ino
);
1041 spin_unlock(&inode_hash_lock
);
1043 wait_on_inode(inode
);
1047 inode
= alloc_inode(sb
);
1051 spin_lock(&inode_hash_lock
);
1052 /* We released the lock, so.. */
1053 old
= find_inode_fast(sb
, head
, ino
);
1056 spin_lock(&inode
->i_lock
);
1057 inode
->i_state
= I_NEW
;
1058 hlist_add_head(&inode
->i_hash
, head
);
1059 spin_unlock(&inode
->i_lock
);
1060 inode_sb_list_add(inode
);
1061 spin_unlock(&inode_hash_lock
);
1063 /* Return the locked inode with I_NEW set, the
1064 * caller is responsible for filling in the contents
1070 * Uhhuh, somebody else created the same inode under
1071 * us. Use the old inode instead of the one we just
1074 spin_unlock(&inode_hash_lock
);
1075 destroy_inode(inode
);
1077 wait_on_inode(inode
);
1081 EXPORT_SYMBOL(iget_locked
);
1084 * search the inode cache for a matching inode number.
1085 * If we find one, then the inode number we are trying to
1086 * allocate is not unique and so we should not use it.
1088 * Returns 1 if the inode number is unique, 0 if it is not.
1090 static int test_inode_iunique(struct super_block
*sb
, unsigned long ino
)
1092 struct hlist_head
*b
= inode_hashtable
+ hash(sb
, ino
);
1093 struct hlist_node
*node
;
1094 struct inode
*inode
;
1096 spin_lock(&inode_hash_lock
);
1097 hlist_for_each_entry(inode
, node
, b
, i_hash
) {
1098 if (inode
->i_ino
== ino
&& inode
->i_sb
== sb
) {
1099 spin_unlock(&inode_hash_lock
);
1103 spin_unlock(&inode_hash_lock
);
1109 * iunique - get a unique inode number
1111 * @max_reserved: highest reserved inode number
1113 * Obtain an inode number that is unique on the system for a given
1114 * superblock. This is used by file systems that have no natural
1115 * permanent inode numbering system. An inode number is returned that
1116 * is higher than the reserved limit but unique.
1119 * With a large number of inodes live on the file system this function
1120 * currently becomes quite slow.
1122 ino_t
iunique(struct super_block
*sb
, ino_t max_reserved
)
1125 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1126 * error if st_ino won't fit in target struct field. Use 32bit counter
1127 * here to attempt to avoid that.
1129 static DEFINE_SPINLOCK(iunique_lock
);
1130 static unsigned int counter
;
1133 spin_lock(&iunique_lock
);
1135 if (counter
<= max_reserved
)
1136 counter
= max_reserved
+ 1;
1138 } while (!test_inode_iunique(sb
, res
));
1139 spin_unlock(&iunique_lock
);
1143 EXPORT_SYMBOL(iunique
);
1145 struct inode
*igrab(struct inode
*inode
)
1147 spin_lock(&inode
->i_lock
);
1148 if (!(inode
->i_state
& (I_FREEING
|I_WILL_FREE
))) {
1150 spin_unlock(&inode
->i_lock
);
1152 spin_unlock(&inode
->i_lock
);
1154 * Handle the case where s_op->clear_inode is not been
1155 * called yet, and somebody is calling igrab
1156 * while the inode is getting freed.
1162 EXPORT_SYMBOL(igrab
);
1165 * ilookup5_nowait - search for an inode in the inode cache
1166 * @sb: super block of file system to search
1167 * @hashval: hash value (usually inode number) to search for
1168 * @test: callback used for comparisons between inodes
1169 * @data: opaque data pointer to pass to @test
1171 * Search for the inode specified by @hashval and @data in the inode cache.
1172 * If the inode is in the cache, the inode is returned with an incremented
1175 * Note: I_NEW is not waited upon so you have to be very careful what you do
1176 * with the returned inode. You probably should be using ilookup5() instead.
1178 * Note2: @test is called with the inode_hash_lock held, so can't sleep.
1180 struct inode
*ilookup5_nowait(struct super_block
*sb
, unsigned long hashval
,
1181 int (*test
)(struct inode
*, void *), void *data
)
1183 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1184 struct inode
*inode
;
1186 spin_lock(&inode_hash_lock
);
1187 inode
= find_inode(sb
, head
, test
, data
);
1188 spin_unlock(&inode_hash_lock
);
1192 EXPORT_SYMBOL(ilookup5_nowait
);
1195 * ilookup5 - search for an inode in the inode cache
1196 * @sb: super block of file system to search
1197 * @hashval: hash value (usually inode number) to search for
1198 * @test: callback used for comparisons between inodes
1199 * @data: opaque data pointer to pass to @test
1201 * Search for the inode specified by @hashval and @data in the inode cache,
1202 * and if the inode is in the cache, return the inode with an incremented
1203 * reference count. Waits on I_NEW before returning the inode.
1204 * returned with an incremented reference count.
1206 * This is a generalized version of ilookup() for file systems where the
1207 * inode number is not sufficient for unique identification of an inode.
1209 * Note: @test is called with the inode_hash_lock held, so can't sleep.
1211 struct inode
*ilookup5(struct super_block
*sb
, unsigned long hashval
,
1212 int (*test
)(struct inode
*, void *), void *data
)
1214 struct inode
*inode
= ilookup5_nowait(sb
, hashval
, test
, data
);
1217 wait_on_inode(inode
);
1220 EXPORT_SYMBOL(ilookup5
);
1223 * ilookup - search for an inode in the inode cache
1224 * @sb: super block of file system to search
1225 * @ino: inode number to search for
1227 * Search for the inode @ino in the inode cache, and if the inode is in the
1228 * cache, the inode is returned with an incremented reference count.
1230 struct inode
*ilookup(struct super_block
*sb
, unsigned long ino
)
1232 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1233 struct inode
*inode
;
1235 spin_lock(&inode_hash_lock
);
1236 inode
= find_inode_fast(sb
, head
, ino
);
1237 spin_unlock(&inode_hash_lock
);
1240 wait_on_inode(inode
);
1243 EXPORT_SYMBOL(ilookup
);
1245 int insert_inode_locked(struct inode
*inode
)
1247 struct super_block
*sb
= inode
->i_sb
;
1248 ino_t ino
= inode
->i_ino
;
1249 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1252 struct hlist_node
*node
;
1253 struct inode
*old
= NULL
;
1254 spin_lock(&inode_hash_lock
);
1255 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1256 if (old
->i_ino
!= ino
)
1258 if (old
->i_sb
!= sb
)
1260 spin_lock(&old
->i_lock
);
1261 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
1262 spin_unlock(&old
->i_lock
);
1267 if (likely(!node
)) {
1268 spin_lock(&inode
->i_lock
);
1269 inode
->i_state
|= I_NEW
;
1270 hlist_add_head(&inode
->i_hash
, head
);
1271 spin_unlock(&inode
->i_lock
);
1272 spin_unlock(&inode_hash_lock
);
1276 spin_unlock(&old
->i_lock
);
1277 spin_unlock(&inode_hash_lock
);
1279 if (unlikely(!inode_unhashed(old
))) {
1286 EXPORT_SYMBOL(insert_inode_locked
);
1288 int insert_inode_locked4(struct inode
*inode
, unsigned long hashval
,
1289 int (*test
)(struct inode
*, void *), void *data
)
1291 struct super_block
*sb
= inode
->i_sb
;
1292 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1295 struct hlist_node
*node
;
1296 struct inode
*old
= NULL
;
1298 spin_lock(&inode_hash_lock
);
1299 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1300 if (old
->i_sb
!= sb
)
1302 if (!test(old
, data
))
1304 spin_lock(&old
->i_lock
);
1305 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
1306 spin_unlock(&old
->i_lock
);
1311 if (likely(!node
)) {
1312 spin_lock(&inode
->i_lock
);
1313 inode
->i_state
|= I_NEW
;
1314 hlist_add_head(&inode
->i_hash
, head
);
1315 spin_unlock(&inode
->i_lock
);
1316 spin_unlock(&inode_hash_lock
);
1320 spin_unlock(&old
->i_lock
);
1321 spin_unlock(&inode_hash_lock
);
1323 if (unlikely(!inode_unhashed(old
))) {
1330 EXPORT_SYMBOL(insert_inode_locked4
);
1333 int generic_delete_inode(struct inode
*inode
)
1337 EXPORT_SYMBOL(generic_delete_inode
);
1340 * Normal UNIX filesystem behaviour: delete the
1341 * inode when the usage count drops to zero, and
1344 int generic_drop_inode(struct inode
*inode
)
1346 return !inode
->i_nlink
|| inode_unhashed(inode
);
1348 EXPORT_SYMBOL_GPL(generic_drop_inode
);
1351 * Called when we're dropping the last reference
1354 * Call the FS "drop_inode()" function, defaulting to
1355 * the legacy UNIX filesystem behaviour. If it tells
1356 * us to evict inode, do so. Otherwise, retain inode
1357 * in cache if fs is alive, sync and evict if fs is
1360 static void iput_final(struct inode
*inode
)
1362 struct super_block
*sb
= inode
->i_sb
;
1363 const struct super_operations
*op
= inode
->i_sb
->s_op
;
1366 WARN_ON(inode
->i_state
& I_NEW
);
1368 if (op
&& op
->drop_inode
)
1369 drop
= op
->drop_inode(inode
);
1371 drop
= generic_drop_inode(inode
);
1373 if (!drop
&& (sb
->s_flags
& MS_ACTIVE
)) {
1374 inode
->i_state
|= I_REFERENCED
;
1375 if (!(inode
->i_state
& (I_DIRTY
|I_SYNC
)))
1376 inode_lru_list_add(inode
);
1377 spin_unlock(&inode
->i_lock
);
1382 inode
->i_state
|= I_WILL_FREE
;
1383 spin_unlock(&inode
->i_lock
);
1384 write_inode_now(inode
, 1);
1385 spin_lock(&inode
->i_lock
);
1386 WARN_ON(inode
->i_state
& I_NEW
);
1387 inode
->i_state
&= ~I_WILL_FREE
;
1390 inode
->i_state
|= I_FREEING
;
1391 inode_lru_list_del(inode
);
1392 spin_unlock(&inode
->i_lock
);
1398 * iput - put an inode
1399 * @inode: inode to put
1401 * Puts an inode, dropping its usage count. If the inode use count hits
1402 * zero, the inode is then freed and may also be destroyed.
1404 * Consequently, iput() can sleep.
1406 void iput(struct inode
*inode
)
1409 BUG_ON(inode
->i_state
& I_CLEAR
);
1411 if (atomic_dec_and_lock(&inode
->i_count
, &inode
->i_lock
))
1415 EXPORT_SYMBOL(iput
);
1418 * bmap - find a block number in a file
1419 * @inode: inode of file
1420 * @block: block to find
1422 * Returns the block number on the device holding the inode that
1423 * is the disk block number for the block of the file requested.
1424 * That is, asked for block 4 of inode 1 the function will return the
1425 * disk block relative to the disk start that holds that block of the
1428 sector_t
bmap(struct inode
*inode
, sector_t block
)
1431 if (inode
->i_mapping
->a_ops
->bmap
)
1432 res
= inode
->i_mapping
->a_ops
->bmap(inode
->i_mapping
, block
);
1435 EXPORT_SYMBOL(bmap
);
1438 * With relative atime, only update atime if the previous atime is
1439 * earlier than either the ctime or mtime or if at least a day has
1440 * passed since the last atime update.
1442 static int relatime_need_update(struct vfsmount
*mnt
, struct inode
*inode
,
1443 struct timespec now
)
1446 if (!(mnt
->mnt_flags
& MNT_RELATIME
))
1449 * Is mtime younger than atime? If yes, update atime:
1451 if (timespec_compare(&inode
->i_mtime
, &inode
->i_atime
) >= 0)
1454 * Is ctime younger than atime? If yes, update atime:
1456 if (timespec_compare(&inode
->i_ctime
, &inode
->i_atime
) >= 0)
1460 * Is the previous atime value older than a day? If yes,
1463 if ((long)(now
.tv_sec
- inode
->i_atime
.tv_sec
) >= 24*60*60)
1466 * Good, we can skip the atime update:
1472 * touch_atime - update the access time
1473 * @mnt: mount the inode is accessed on
1474 * @dentry: dentry accessed
1476 * Update the accessed time on an inode and mark it for writeback.
1477 * This function automatically handles read only file systems and media,
1478 * as well as the "noatime" flag and inode specific "noatime" markers.
1480 void touch_atime(struct vfsmount
*mnt
, struct dentry
*dentry
)
1482 struct inode
*inode
= dentry
->d_inode
;
1483 struct timespec now
;
1485 if (inode
->i_flags
& S_NOATIME
)
1487 if (IS_NOATIME(inode
))
1489 if ((inode
->i_sb
->s_flags
& MS_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1492 if (mnt
->mnt_flags
& MNT_NOATIME
)
1494 if ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1497 now
= current_fs_time(inode
->i_sb
);
1499 if (!relatime_need_update(mnt
, inode
, now
))
1502 if (timespec_equal(&inode
->i_atime
, &now
))
1505 if (mnt_want_write(mnt
))
1508 inode
->i_atime
= now
;
1509 mark_inode_dirty_sync(inode
);
1510 mnt_drop_write(mnt
);
1512 EXPORT_SYMBOL(touch_atime
);
1515 * file_update_time - update mtime and ctime time
1516 * @file: file accessed
1518 * Update the mtime and ctime members of an inode and mark the inode
1519 * for writeback. Note that this function is meant exclusively for
1520 * usage in the file write path of filesystems, and filesystems may
1521 * choose to explicitly ignore update via this function with the
1522 * S_NOCMTIME inode flag, e.g. for network filesystem where these
1523 * timestamps are handled by the server.
1526 void file_update_time(struct file
*file
)
1528 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1529 struct timespec now
;
1530 enum { S_MTIME
= 1, S_CTIME
= 2, S_VERSION
= 4 } sync_it
= 0;
1532 /* First try to exhaust all avenues to not sync */
1533 if (IS_NOCMTIME(inode
))
1536 now
= current_fs_time(inode
->i_sb
);
1537 if (!timespec_equal(&inode
->i_mtime
, &now
))
1540 if (!timespec_equal(&inode
->i_ctime
, &now
))
1543 if (IS_I_VERSION(inode
))
1544 sync_it
|= S_VERSION
;
1549 /* Finally allowed to write? Takes lock. */
1550 if (mnt_want_write_file(file
))
1553 /* Only change inode inside the lock region */
1554 if (sync_it
& S_VERSION
)
1555 inode_inc_iversion(inode
);
1556 if (sync_it
& S_CTIME
)
1557 inode
->i_ctime
= now
;
1558 if (sync_it
& S_MTIME
)
1559 inode
->i_mtime
= now
;
1560 mark_inode_dirty_sync(inode
);
1561 mnt_drop_write(file
->f_path
.mnt
);
1563 EXPORT_SYMBOL(file_update_time
);
1565 int inode_needs_sync(struct inode
*inode
)
1569 if (S_ISDIR(inode
->i_mode
) && IS_DIRSYNC(inode
))
1573 EXPORT_SYMBOL(inode_needs_sync
);
1575 int inode_wait(void *word
)
1580 EXPORT_SYMBOL(inode_wait
);
1583 * If we try to find an inode in the inode hash while it is being
1584 * deleted, we have to wait until the filesystem completes its
1585 * deletion before reporting that it isn't found. This function waits
1586 * until the deletion _might_ have completed. Callers are responsible
1587 * to recheck inode state.
1589 * It doesn't matter if I_NEW is not set initially, a call to
1590 * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
1593 static void __wait_on_freeing_inode(struct inode
*inode
)
1595 wait_queue_head_t
*wq
;
1596 DEFINE_WAIT_BIT(wait
, &inode
->i_state
, __I_NEW
);
1597 wq
= bit_waitqueue(&inode
->i_state
, __I_NEW
);
1598 prepare_to_wait(wq
, &wait
.wait
, TASK_UNINTERRUPTIBLE
);
1599 spin_unlock(&inode
->i_lock
);
1600 spin_unlock(&inode_hash_lock
);
1602 finish_wait(wq
, &wait
.wait
);
1603 spin_lock(&inode_hash_lock
);
1606 static __initdata
unsigned long ihash_entries
;
1607 static int __init
set_ihash_entries(char *str
)
1611 ihash_entries
= simple_strtoul(str
, &str
, 0);
1614 __setup("ihash_entries=", set_ihash_entries
);
1617 * Initialize the waitqueues and inode hash table.
1619 void __init
inode_init_early(void)
1623 /* If hashes are distributed across NUMA nodes, defer
1624 * hash allocation until vmalloc space is available.
1630 alloc_large_system_hash("Inode-cache",
1631 sizeof(struct hlist_head
),
1639 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1640 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1643 void __init
inode_init(void)
1647 /* inode slab cache */
1648 inode_cachep
= kmem_cache_create("inode_cache",
1649 sizeof(struct inode
),
1651 (SLAB_RECLAIM_ACCOUNT
|SLAB_PANIC
|
1654 register_shrinker(&icache_shrinker
);
1656 /* Hash may have been set up in inode_init_early */
1661 alloc_large_system_hash("Inode-cache",
1662 sizeof(struct hlist_head
),
1670 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1671 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1674 void init_special_inode(struct inode
*inode
, umode_t mode
, dev_t rdev
)
1676 inode
->i_mode
= mode
;
1677 if (S_ISCHR(mode
)) {
1678 inode
->i_fop
= &def_chr_fops
;
1679 inode
->i_rdev
= rdev
;
1680 } else if (S_ISBLK(mode
)) {
1681 inode
->i_fop
= &def_blk_fops
;
1682 inode
->i_rdev
= rdev
;
1683 } else if (S_ISFIFO(mode
))
1684 inode
->i_fop
= &def_fifo_fops
;
1685 else if (S_ISSOCK(mode
))
1686 inode
->i_fop
= &bad_sock_fops
;
1688 printk(KERN_DEBUG
"init_special_inode: bogus i_mode (%o) for"
1689 " inode %s:%lu\n", mode
, inode
->i_sb
->s_id
,
1692 EXPORT_SYMBOL(init_special_inode
);
1695 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
1697 * @dir: Directory inode
1698 * @mode: mode of the new inode
1700 void inode_init_owner(struct inode
*inode
, const struct inode
*dir
,
1703 inode
->i_uid
= current_fsuid();
1704 if (dir
&& dir
->i_mode
& S_ISGID
) {
1705 inode
->i_gid
= dir
->i_gid
;
1709 inode
->i_gid
= current_fsgid();
1710 inode
->i_mode
= mode
;
1712 EXPORT_SYMBOL(inode_init_owner
);
1715 * inode_owner_or_capable - check current task permissions to inode
1716 * @inode: inode being checked
1718 * Return true if current either has CAP_FOWNER to the inode, or
1721 bool inode_owner_or_capable(const struct inode
*inode
)
1723 struct user_namespace
*ns
= inode_userns(inode
);
1725 if (current_user_ns() == ns
&& current_fsuid() == inode
->i_uid
)
1727 if (ns_capable(ns
, CAP_FOWNER
))
1731 EXPORT_SYMBOL(inode_owner_or_capable
);