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 mutex_init(&mapping
->i_mmap_mutex
);
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
);
335 EXPORT_SYMBOL(address_space_init_once
);
338 * These are initializations that only need to be done
339 * once, because the fields are idempotent across use
340 * of the inode, so let the slab aware of that.
342 void inode_init_once(struct inode
*inode
)
344 memset(inode
, 0, sizeof(*inode
));
345 INIT_HLIST_NODE(&inode
->i_hash
);
346 INIT_LIST_HEAD(&inode
->i_dentry
);
347 INIT_LIST_HEAD(&inode
->i_devices
);
348 INIT_LIST_HEAD(&inode
->i_wb_list
);
349 INIT_LIST_HEAD(&inode
->i_lru
);
350 address_space_init_once(&inode
->i_data
);
351 i_size_ordered_init(inode
);
352 #ifdef CONFIG_FSNOTIFY
353 INIT_HLIST_HEAD(&inode
->i_fsnotify_marks
);
356 EXPORT_SYMBOL(inode_init_once
);
358 static void init_once(void *foo
)
360 struct inode
*inode
= (struct inode
*) foo
;
362 inode_init_once(inode
);
366 * inode->i_lock must be held
368 void __iget(struct inode
*inode
)
370 atomic_inc(&inode
->i_count
);
374 * get additional reference to inode; caller must already hold one.
376 void ihold(struct inode
*inode
)
378 WARN_ON(atomic_inc_return(&inode
->i_count
) < 2);
380 EXPORT_SYMBOL(ihold
);
382 static void inode_lru_list_add(struct inode
*inode
)
384 spin_lock(&inode_lru_lock
);
385 if (list_empty(&inode
->i_lru
)) {
386 list_add(&inode
->i_lru
, &inode_lru
);
387 inodes_stat
.nr_unused
++;
389 spin_unlock(&inode_lru_lock
);
392 static void inode_lru_list_del(struct inode
*inode
)
394 spin_lock(&inode_lru_lock
);
395 if (!list_empty(&inode
->i_lru
)) {
396 list_del_init(&inode
->i_lru
);
397 inodes_stat
.nr_unused
--;
399 spin_unlock(&inode_lru_lock
);
403 * inode_sb_list_add - add inode to the superblock list of inodes
404 * @inode: inode to add
406 void inode_sb_list_add(struct inode
*inode
)
408 spin_lock(&inode_sb_list_lock
);
409 list_add(&inode
->i_sb_list
, &inode
->i_sb
->s_inodes
);
410 spin_unlock(&inode_sb_list_lock
);
412 EXPORT_SYMBOL_GPL(inode_sb_list_add
);
414 static inline void inode_sb_list_del(struct inode
*inode
)
416 spin_lock(&inode_sb_list_lock
);
417 list_del_init(&inode
->i_sb_list
);
418 spin_unlock(&inode_sb_list_lock
);
421 static unsigned long hash(struct super_block
*sb
, unsigned long hashval
)
425 tmp
= (hashval
* (unsigned long)sb
) ^ (GOLDEN_RATIO_PRIME
+ hashval
) /
427 tmp
= tmp
^ ((tmp
^ GOLDEN_RATIO_PRIME
) >> I_HASHBITS
);
428 return tmp
& I_HASHMASK
;
432 * __insert_inode_hash - hash an inode
433 * @inode: unhashed inode
434 * @hashval: unsigned long value used to locate this object in the
437 * Add an inode to the inode hash for this superblock.
439 void __insert_inode_hash(struct inode
*inode
, unsigned long hashval
)
441 struct hlist_head
*b
= inode_hashtable
+ hash(inode
->i_sb
, hashval
);
443 spin_lock(&inode_hash_lock
);
444 spin_lock(&inode
->i_lock
);
445 hlist_add_head(&inode
->i_hash
, b
);
446 spin_unlock(&inode
->i_lock
);
447 spin_unlock(&inode_hash_lock
);
449 EXPORT_SYMBOL(__insert_inode_hash
);
452 * remove_inode_hash - remove an inode from the hash
453 * @inode: inode to unhash
455 * Remove an inode from the superblock.
457 void remove_inode_hash(struct inode
*inode
)
459 spin_lock(&inode_hash_lock
);
460 spin_lock(&inode
->i_lock
);
461 hlist_del_init(&inode
->i_hash
);
462 spin_unlock(&inode
->i_lock
);
463 spin_unlock(&inode_hash_lock
);
465 EXPORT_SYMBOL(remove_inode_hash
);
467 void end_writeback(struct inode
*inode
)
470 BUG_ON(inode
->i_data
.nrpages
);
471 BUG_ON(!list_empty(&inode
->i_data
.private_list
));
472 BUG_ON(!(inode
->i_state
& I_FREEING
));
473 BUG_ON(inode
->i_state
& I_CLEAR
);
474 inode_sync_wait(inode
);
475 /* don't need i_lock here, no concurrent mods to i_state */
476 inode
->i_state
= I_FREEING
| I_CLEAR
;
478 EXPORT_SYMBOL(end_writeback
);
481 * Free the inode passed in, removing it from the lists it is still connected
482 * to. We remove any pages still attached to the inode and wait for any IO that
483 * is still in progress before finally destroying the inode.
485 * An inode must already be marked I_FREEING so that we avoid the inode being
486 * moved back onto lists if we race with other code that manipulates the lists
487 * (e.g. writeback_single_inode). The caller is responsible for setting this.
489 * An inode must already be removed from the LRU list before being evicted from
490 * the cache. This should occur atomically with setting the I_FREEING state
491 * flag, so no inodes here should ever be on the LRU when being evicted.
493 static void evict(struct inode
*inode
)
495 const struct super_operations
*op
= inode
->i_sb
->s_op
;
497 BUG_ON(!(inode
->i_state
& I_FREEING
));
498 BUG_ON(!list_empty(&inode
->i_lru
));
500 inode_wb_list_del(inode
);
501 inode_sb_list_del(inode
);
503 if (op
->evict_inode
) {
504 op
->evict_inode(inode
);
506 if (inode
->i_data
.nrpages
)
507 truncate_inode_pages(&inode
->i_data
, 0);
508 end_writeback(inode
);
510 if (S_ISBLK(inode
->i_mode
) && inode
->i_bdev
)
512 if (S_ISCHR(inode
->i_mode
) && inode
->i_cdev
)
515 remove_inode_hash(inode
);
517 spin_lock(&inode
->i_lock
);
518 wake_up_bit(&inode
->i_state
, __I_NEW
);
519 BUG_ON(inode
->i_state
!= (I_FREEING
| I_CLEAR
));
520 spin_unlock(&inode
->i_lock
);
522 destroy_inode(inode
);
526 * dispose_list - dispose of the contents of a local list
527 * @head: the head of the list to free
529 * Dispose-list gets a local list with local inodes in it, so it doesn't
530 * need to worry about list corruption and SMP locks.
532 static void dispose_list(struct list_head
*head
)
534 while (!list_empty(head
)) {
537 inode
= list_first_entry(head
, struct inode
, i_lru
);
538 list_del_init(&inode
->i_lru
);
545 * evict_inodes - evict all evictable inodes for a superblock
546 * @sb: superblock to operate on
548 * Make sure that no inodes with zero refcount are retained. This is
549 * called by superblock shutdown after having MS_ACTIVE flag removed,
550 * so any inode reaching zero refcount during or after that call will
551 * be immediately evicted.
553 void evict_inodes(struct super_block
*sb
)
555 struct inode
*inode
, *next
;
558 spin_lock(&inode_sb_list_lock
);
559 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
560 if (atomic_read(&inode
->i_count
))
563 spin_lock(&inode
->i_lock
);
564 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
565 spin_unlock(&inode
->i_lock
);
569 inode
->i_state
|= I_FREEING
;
570 inode_lru_list_del(inode
);
571 spin_unlock(&inode
->i_lock
);
572 list_add(&inode
->i_lru
, &dispose
);
574 spin_unlock(&inode_sb_list_lock
);
576 dispose_list(&dispose
);
579 * Cycle through iprune_sem to make sure any inode that prune_icache
580 * moved off the list before we took the lock has been fully torn
583 down_write(&iprune_sem
);
584 up_write(&iprune_sem
);
588 * invalidate_inodes - attempt to free all inodes on a superblock
589 * @sb: superblock to operate on
590 * @kill_dirty: flag to guide handling of dirty inodes
592 * Attempts to free all inodes for a given superblock. If there were any
593 * busy inodes return a non-zero value, else zero.
594 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
597 int invalidate_inodes(struct super_block
*sb
, bool kill_dirty
)
600 struct inode
*inode
, *next
;
603 spin_lock(&inode_sb_list_lock
);
604 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
605 spin_lock(&inode
->i_lock
);
606 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
607 spin_unlock(&inode
->i_lock
);
610 if (inode
->i_state
& I_DIRTY
&& !kill_dirty
) {
611 spin_unlock(&inode
->i_lock
);
615 if (atomic_read(&inode
->i_count
)) {
616 spin_unlock(&inode
->i_lock
);
621 inode
->i_state
|= I_FREEING
;
622 inode_lru_list_del(inode
);
623 spin_unlock(&inode
->i_lock
);
624 list_add(&inode
->i_lru
, &dispose
);
626 spin_unlock(&inode_sb_list_lock
);
628 dispose_list(&dispose
);
633 static int can_unuse(struct inode
*inode
)
635 if (inode
->i_state
& ~I_REFERENCED
)
637 if (inode_has_buffers(inode
))
639 if (atomic_read(&inode
->i_count
))
641 if (inode
->i_data
.nrpages
)
647 * Scan `goal' inodes on the unused list for freeable ones. They are moved to a
648 * temporary list and then are freed outside inode_lru_lock by dispose_list().
650 * Any inodes which are pinned purely because of attached pagecache have their
651 * pagecache removed. If the inode has metadata buffers attached to
652 * mapping->private_list then try to remove them.
654 * If the inode has the I_REFERENCED flag set, then it means that it has been
655 * used recently - the flag is set in iput_final(). When we encounter such an
656 * inode, clear the flag and move it to the back of the LRU so it gets another
657 * pass through the LRU before it gets reclaimed. This is necessary because of
658 * the fact we are doing lazy LRU updates to minimise lock contention so the
659 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
660 * with this flag set because they are the inodes that are out of order.
662 static void prune_icache(int nr_to_scan
)
666 unsigned long reap
= 0;
668 down_read(&iprune_sem
);
669 spin_lock(&inode_lru_lock
);
670 for (nr_scanned
= 0; nr_scanned
< nr_to_scan
; nr_scanned
++) {
673 if (list_empty(&inode_lru
))
676 inode
= list_entry(inode_lru
.prev
, struct inode
, i_lru
);
679 * we are inverting the inode_lru_lock/inode->i_lock here,
680 * so use a trylock. If we fail to get the lock, just move the
681 * inode to the back of the list so we don't spin on it.
683 if (!spin_trylock(&inode
->i_lock
)) {
684 list_move(&inode
->i_lru
, &inode_lru
);
689 * Referenced or dirty inodes are still in use. Give them
690 * another pass through the LRU as we canot reclaim them now.
692 if (atomic_read(&inode
->i_count
) ||
693 (inode
->i_state
& ~I_REFERENCED
)) {
694 list_del_init(&inode
->i_lru
);
695 spin_unlock(&inode
->i_lock
);
696 inodes_stat
.nr_unused
--;
700 /* recently referenced inodes get one more pass */
701 if (inode
->i_state
& I_REFERENCED
) {
702 inode
->i_state
&= ~I_REFERENCED
;
703 list_move(&inode
->i_lru
, &inode_lru
);
704 spin_unlock(&inode
->i_lock
);
707 if (inode_has_buffers(inode
) || inode
->i_data
.nrpages
) {
709 spin_unlock(&inode
->i_lock
);
710 spin_unlock(&inode_lru_lock
);
711 if (remove_inode_buffers(inode
))
712 reap
+= invalidate_mapping_pages(&inode
->i_data
,
715 spin_lock(&inode_lru_lock
);
717 if (inode
!= list_entry(inode_lru
.next
,
718 struct inode
, i_lru
))
719 continue; /* wrong inode or list_empty */
720 /* avoid lock inversions with trylock */
721 if (!spin_trylock(&inode
->i_lock
))
723 if (!can_unuse(inode
)) {
724 spin_unlock(&inode
->i_lock
);
728 WARN_ON(inode
->i_state
& I_NEW
);
729 inode
->i_state
|= I_FREEING
;
730 spin_unlock(&inode
->i_lock
);
732 list_move(&inode
->i_lru
, &freeable
);
733 inodes_stat
.nr_unused
--;
735 if (current_is_kswapd())
736 __count_vm_events(KSWAPD_INODESTEAL
, reap
);
738 __count_vm_events(PGINODESTEAL
, reap
);
739 spin_unlock(&inode_lru_lock
);
741 dispose_list(&freeable
);
742 up_read(&iprune_sem
);
746 * shrink_icache_memory() will attempt to reclaim some unused inodes. Here,
747 * "unused" means that no dentries are referring to the inodes: the files are
748 * not open and the dcache references to those inodes have already been
751 * This function is passed the number of inodes to scan, and it returns the
752 * total number of remaining possibly-reclaimable inodes.
754 static int shrink_icache_memory(struct shrinker
*shrink
,
755 struct shrink_control
*sc
)
757 int nr
= sc
->nr_to_scan
;
758 gfp_t gfp_mask
= sc
->gfp_mask
;
762 * Nasty deadlock avoidance. We may hold various FS locks,
763 * and we don't want to recurse into the FS that called us
764 * in clear_inode() and friends..
766 if (!(gfp_mask
& __GFP_FS
))
770 return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure
;
773 static struct shrinker icache_shrinker
= {
774 .shrink
= shrink_icache_memory
,
775 .seeks
= DEFAULT_SEEKS
,
778 static void __wait_on_freeing_inode(struct inode
*inode
);
780 * Called with the inode lock held.
782 static struct inode
*find_inode(struct super_block
*sb
,
783 struct hlist_head
*head
,
784 int (*test
)(struct inode
*, void *),
787 struct hlist_node
*node
;
788 struct inode
*inode
= NULL
;
791 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
792 spin_lock(&inode
->i_lock
);
793 if (inode
->i_sb
!= sb
) {
794 spin_unlock(&inode
->i_lock
);
797 if (!test(inode
, data
)) {
798 spin_unlock(&inode
->i_lock
);
801 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
802 __wait_on_freeing_inode(inode
);
806 spin_unlock(&inode
->i_lock
);
813 * find_inode_fast is the fast path version of find_inode, see the comment at
814 * iget_locked for details.
816 static struct inode
*find_inode_fast(struct super_block
*sb
,
817 struct hlist_head
*head
, unsigned long ino
)
819 struct hlist_node
*node
;
820 struct inode
*inode
= NULL
;
823 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
824 spin_lock(&inode
->i_lock
);
825 if (inode
->i_ino
!= ino
) {
826 spin_unlock(&inode
->i_lock
);
829 if (inode
->i_sb
!= sb
) {
830 spin_unlock(&inode
->i_lock
);
833 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
834 __wait_on_freeing_inode(inode
);
838 spin_unlock(&inode
->i_lock
);
845 * Each cpu owns a range of LAST_INO_BATCH numbers.
846 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
847 * to renew the exhausted range.
849 * This does not significantly increase overflow rate because every CPU can
850 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
851 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
852 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
853 * overflow rate by 2x, which does not seem too significant.
855 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
856 * error if st_ino won't fit in target struct field. Use 32bit counter
857 * here to attempt to avoid that.
859 #define LAST_INO_BATCH 1024
860 static DEFINE_PER_CPU(unsigned int, last_ino
);
862 unsigned int get_next_ino(void)
864 unsigned int *p
= &get_cpu_var(last_ino
);
865 unsigned int res
= *p
;
868 if (unlikely((res
& (LAST_INO_BATCH
-1)) == 0)) {
869 static atomic_t shared_last_ino
;
870 int next
= atomic_add_return(LAST_INO_BATCH
, &shared_last_ino
);
872 res
= next
- LAST_INO_BATCH
;
877 put_cpu_var(last_ino
);
880 EXPORT_SYMBOL(get_next_ino
);
883 * new_inode - obtain an inode
886 * Allocates a new inode for given superblock. The default gfp_mask
887 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
888 * If HIGHMEM pages are unsuitable or it is known that pages allocated
889 * for the page cache are not reclaimable or migratable,
890 * mapping_set_gfp_mask() must be called with suitable flags on the
891 * newly created inode's mapping
894 struct inode
*new_inode(struct super_block
*sb
)
898 spin_lock_prefetch(&inode_sb_list_lock
);
900 inode
= alloc_inode(sb
);
902 spin_lock(&inode
->i_lock
);
904 spin_unlock(&inode
->i_lock
);
905 inode_sb_list_add(inode
);
909 EXPORT_SYMBOL(new_inode
);
912 * unlock_new_inode - clear the I_NEW state and wake up any waiters
913 * @inode: new inode to unlock
915 * Called when the inode is fully initialised to clear the new state of the
916 * inode and wake up anyone waiting for the inode to finish initialisation.
918 void unlock_new_inode(struct inode
*inode
)
920 #ifdef CONFIG_DEBUG_LOCK_ALLOC
921 if (S_ISDIR(inode
->i_mode
)) {
922 struct file_system_type
*type
= inode
->i_sb
->s_type
;
924 /* Set new key only if filesystem hasn't already changed it */
925 if (!lockdep_match_class(&inode
->i_mutex
,
926 &type
->i_mutex_key
)) {
928 * ensure nobody is actually holding i_mutex
930 mutex_destroy(&inode
->i_mutex
);
931 mutex_init(&inode
->i_mutex
);
932 lockdep_set_class(&inode
->i_mutex
,
933 &type
->i_mutex_dir_key
);
937 spin_lock(&inode
->i_lock
);
938 WARN_ON(!(inode
->i_state
& I_NEW
));
939 inode
->i_state
&= ~I_NEW
;
940 wake_up_bit(&inode
->i_state
, __I_NEW
);
941 spin_unlock(&inode
->i_lock
);
943 EXPORT_SYMBOL(unlock_new_inode
);
946 * iget5_locked - obtain an inode from a mounted file system
947 * @sb: super block of file system
948 * @hashval: hash value (usually inode number) to get
949 * @test: callback used for comparisons between inodes
950 * @set: callback used to initialize a new struct inode
951 * @data: opaque data pointer to pass to @test and @set
953 * Search for the inode specified by @hashval and @data in the inode cache,
954 * and if present it is return it with an increased reference count. This is
955 * a generalized version of iget_locked() for file systems where the inode
956 * number is not sufficient for unique identification of an inode.
958 * If the inode is not in cache, allocate a new inode and return it locked,
959 * hashed, and with the I_NEW flag set. The file system gets to fill it in
960 * before unlocking it via unlock_new_inode().
962 * Note both @test and @set are called with the inode_hash_lock held, so can't
965 struct inode
*iget5_locked(struct super_block
*sb
, unsigned long hashval
,
966 int (*test
)(struct inode
*, void *),
967 int (*set
)(struct inode
*, void *), void *data
)
969 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
972 spin_lock(&inode_hash_lock
);
973 inode
= find_inode(sb
, head
, test
, data
);
974 spin_unlock(&inode_hash_lock
);
977 wait_on_inode(inode
);
981 inode
= alloc_inode(sb
);
985 spin_lock(&inode_hash_lock
);
986 /* We released the lock, so.. */
987 old
= find_inode(sb
, head
, test
, data
);
989 if (set(inode
, data
))
992 spin_lock(&inode
->i_lock
);
993 inode
->i_state
= I_NEW
;
994 hlist_add_head(&inode
->i_hash
, head
);
995 spin_unlock(&inode
->i_lock
);
996 inode_sb_list_add(inode
);
997 spin_unlock(&inode_hash_lock
);
999 /* Return the locked inode with I_NEW set, the
1000 * caller is responsible for filling in the contents
1006 * Uhhuh, somebody else created the same inode under
1007 * us. Use the old inode instead of the one we just
1010 spin_unlock(&inode_hash_lock
);
1011 destroy_inode(inode
);
1013 wait_on_inode(inode
);
1018 spin_unlock(&inode_hash_lock
);
1019 destroy_inode(inode
);
1022 EXPORT_SYMBOL(iget5_locked
);
1025 * iget_locked - obtain an inode from a mounted file system
1026 * @sb: super block of file system
1027 * @ino: inode number to get
1029 * Search for the inode specified by @ino in the inode cache and if present
1030 * return it with an increased reference count. This is for file systems
1031 * where the inode number is sufficient for unique identification of an inode.
1033 * If the inode is not in cache, allocate a new inode and return it locked,
1034 * hashed, and with the I_NEW flag set. The file system gets to fill it in
1035 * before unlocking it via unlock_new_inode().
1037 struct inode
*iget_locked(struct super_block
*sb
, unsigned long ino
)
1039 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1040 struct inode
*inode
;
1042 spin_lock(&inode_hash_lock
);
1043 inode
= find_inode_fast(sb
, head
, ino
);
1044 spin_unlock(&inode_hash_lock
);
1046 wait_on_inode(inode
);
1050 inode
= alloc_inode(sb
);
1054 spin_lock(&inode_hash_lock
);
1055 /* We released the lock, so.. */
1056 old
= find_inode_fast(sb
, head
, ino
);
1059 spin_lock(&inode
->i_lock
);
1060 inode
->i_state
= I_NEW
;
1061 hlist_add_head(&inode
->i_hash
, head
);
1062 spin_unlock(&inode
->i_lock
);
1063 inode_sb_list_add(inode
);
1064 spin_unlock(&inode_hash_lock
);
1066 /* Return the locked inode with I_NEW set, the
1067 * caller is responsible for filling in the contents
1073 * Uhhuh, somebody else created the same inode under
1074 * us. Use the old inode instead of the one we just
1077 spin_unlock(&inode_hash_lock
);
1078 destroy_inode(inode
);
1080 wait_on_inode(inode
);
1084 EXPORT_SYMBOL(iget_locked
);
1087 * search the inode cache for a matching inode number.
1088 * If we find one, then the inode number we are trying to
1089 * allocate is not unique and so we should not use it.
1091 * Returns 1 if the inode number is unique, 0 if it is not.
1093 static int test_inode_iunique(struct super_block
*sb
, unsigned long ino
)
1095 struct hlist_head
*b
= inode_hashtable
+ hash(sb
, ino
);
1096 struct hlist_node
*node
;
1097 struct inode
*inode
;
1099 spin_lock(&inode_hash_lock
);
1100 hlist_for_each_entry(inode
, node
, b
, i_hash
) {
1101 if (inode
->i_ino
== ino
&& inode
->i_sb
== sb
) {
1102 spin_unlock(&inode_hash_lock
);
1106 spin_unlock(&inode_hash_lock
);
1112 * iunique - get a unique inode number
1114 * @max_reserved: highest reserved inode number
1116 * Obtain an inode number that is unique on the system for a given
1117 * superblock. This is used by file systems that have no natural
1118 * permanent inode numbering system. An inode number is returned that
1119 * is higher than the reserved limit but unique.
1122 * With a large number of inodes live on the file system this function
1123 * currently becomes quite slow.
1125 ino_t
iunique(struct super_block
*sb
, ino_t max_reserved
)
1128 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1129 * error if st_ino won't fit in target struct field. Use 32bit counter
1130 * here to attempt to avoid that.
1132 static DEFINE_SPINLOCK(iunique_lock
);
1133 static unsigned int counter
;
1136 spin_lock(&iunique_lock
);
1138 if (counter
<= max_reserved
)
1139 counter
= max_reserved
+ 1;
1141 } while (!test_inode_iunique(sb
, res
));
1142 spin_unlock(&iunique_lock
);
1146 EXPORT_SYMBOL(iunique
);
1148 struct inode
*igrab(struct inode
*inode
)
1150 spin_lock(&inode
->i_lock
);
1151 if (!(inode
->i_state
& (I_FREEING
|I_WILL_FREE
))) {
1153 spin_unlock(&inode
->i_lock
);
1155 spin_unlock(&inode
->i_lock
);
1157 * Handle the case where s_op->clear_inode is not been
1158 * called yet, and somebody is calling igrab
1159 * while the inode is getting freed.
1165 EXPORT_SYMBOL(igrab
);
1168 * ilookup5_nowait - search for an inode in the inode cache
1169 * @sb: super block of file system to search
1170 * @hashval: hash value (usually inode number) to search for
1171 * @test: callback used for comparisons between inodes
1172 * @data: opaque data pointer to pass to @test
1174 * Search for the inode specified by @hashval and @data in the inode cache.
1175 * If the inode is in the cache, the inode is returned with an incremented
1178 * Note: I_NEW is not waited upon so you have to be very careful what you do
1179 * with the returned inode. You probably should be using ilookup5() instead.
1181 * Note2: @test is called with the inode_hash_lock held, so can't sleep.
1183 struct inode
*ilookup5_nowait(struct super_block
*sb
, unsigned long hashval
,
1184 int (*test
)(struct inode
*, void *), void *data
)
1186 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1187 struct inode
*inode
;
1189 spin_lock(&inode_hash_lock
);
1190 inode
= find_inode(sb
, head
, test
, data
);
1191 spin_unlock(&inode_hash_lock
);
1195 EXPORT_SYMBOL(ilookup5_nowait
);
1198 * ilookup5 - search for an inode in the inode cache
1199 * @sb: super block of file system to search
1200 * @hashval: hash value (usually inode number) to search for
1201 * @test: callback used for comparisons between inodes
1202 * @data: opaque data pointer to pass to @test
1204 * Search for the inode specified by @hashval and @data in the inode cache,
1205 * and if the inode is in the cache, return the inode with an incremented
1206 * reference count. Waits on I_NEW before returning the inode.
1207 * returned with an incremented reference count.
1209 * This is a generalized version of ilookup() for file systems where the
1210 * inode number is not sufficient for unique identification of an inode.
1212 * Note: @test is called with the inode_hash_lock held, so can't sleep.
1214 struct inode
*ilookup5(struct super_block
*sb
, unsigned long hashval
,
1215 int (*test
)(struct inode
*, void *), void *data
)
1217 struct inode
*inode
= ilookup5_nowait(sb
, hashval
, test
, data
);
1220 wait_on_inode(inode
);
1223 EXPORT_SYMBOL(ilookup5
);
1226 * ilookup - search for an inode in the inode cache
1227 * @sb: super block of file system to search
1228 * @ino: inode number to search for
1230 * Search for the inode @ino in the inode cache, and if the inode is in the
1231 * cache, the inode is returned with an incremented reference count.
1233 struct inode
*ilookup(struct super_block
*sb
, unsigned long ino
)
1235 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1236 struct inode
*inode
;
1238 spin_lock(&inode_hash_lock
);
1239 inode
= find_inode_fast(sb
, head
, ino
);
1240 spin_unlock(&inode_hash_lock
);
1243 wait_on_inode(inode
);
1246 EXPORT_SYMBOL(ilookup
);
1248 int insert_inode_locked(struct inode
*inode
)
1250 struct super_block
*sb
= inode
->i_sb
;
1251 ino_t ino
= inode
->i_ino
;
1252 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1255 struct hlist_node
*node
;
1256 struct inode
*old
= NULL
;
1257 spin_lock(&inode_hash_lock
);
1258 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1259 if (old
->i_ino
!= ino
)
1261 if (old
->i_sb
!= sb
)
1263 spin_lock(&old
->i_lock
);
1264 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
1265 spin_unlock(&old
->i_lock
);
1270 if (likely(!node
)) {
1271 spin_lock(&inode
->i_lock
);
1272 inode
->i_state
|= I_NEW
;
1273 hlist_add_head(&inode
->i_hash
, head
);
1274 spin_unlock(&inode
->i_lock
);
1275 spin_unlock(&inode_hash_lock
);
1279 spin_unlock(&old
->i_lock
);
1280 spin_unlock(&inode_hash_lock
);
1282 if (unlikely(!inode_unhashed(old
))) {
1289 EXPORT_SYMBOL(insert_inode_locked
);
1291 int insert_inode_locked4(struct inode
*inode
, unsigned long hashval
,
1292 int (*test
)(struct inode
*, void *), void *data
)
1294 struct super_block
*sb
= inode
->i_sb
;
1295 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1298 struct hlist_node
*node
;
1299 struct inode
*old
= NULL
;
1301 spin_lock(&inode_hash_lock
);
1302 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1303 if (old
->i_sb
!= sb
)
1305 if (!test(old
, data
))
1307 spin_lock(&old
->i_lock
);
1308 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
1309 spin_unlock(&old
->i_lock
);
1314 if (likely(!node
)) {
1315 spin_lock(&inode
->i_lock
);
1316 inode
->i_state
|= I_NEW
;
1317 hlist_add_head(&inode
->i_hash
, head
);
1318 spin_unlock(&inode
->i_lock
);
1319 spin_unlock(&inode_hash_lock
);
1323 spin_unlock(&old
->i_lock
);
1324 spin_unlock(&inode_hash_lock
);
1326 if (unlikely(!inode_unhashed(old
))) {
1333 EXPORT_SYMBOL(insert_inode_locked4
);
1336 int generic_delete_inode(struct inode
*inode
)
1340 EXPORT_SYMBOL(generic_delete_inode
);
1343 * Normal UNIX filesystem behaviour: delete the
1344 * inode when the usage count drops to zero, and
1347 int generic_drop_inode(struct inode
*inode
)
1349 return !inode
->i_nlink
|| inode_unhashed(inode
);
1351 EXPORT_SYMBOL_GPL(generic_drop_inode
);
1354 * Called when we're dropping the last reference
1357 * Call the FS "drop_inode()" function, defaulting to
1358 * the legacy UNIX filesystem behaviour. If it tells
1359 * us to evict inode, do so. Otherwise, retain inode
1360 * in cache if fs is alive, sync and evict if fs is
1363 static void iput_final(struct inode
*inode
)
1365 struct super_block
*sb
= inode
->i_sb
;
1366 const struct super_operations
*op
= inode
->i_sb
->s_op
;
1369 WARN_ON(inode
->i_state
& I_NEW
);
1371 if (op
&& op
->drop_inode
)
1372 drop
= op
->drop_inode(inode
);
1374 drop
= generic_drop_inode(inode
);
1376 if (!drop
&& (sb
->s_flags
& MS_ACTIVE
)) {
1377 inode
->i_state
|= I_REFERENCED
;
1378 if (!(inode
->i_state
& (I_DIRTY
|I_SYNC
)))
1379 inode_lru_list_add(inode
);
1380 spin_unlock(&inode
->i_lock
);
1385 inode
->i_state
|= I_WILL_FREE
;
1386 spin_unlock(&inode
->i_lock
);
1387 write_inode_now(inode
, 1);
1388 spin_lock(&inode
->i_lock
);
1389 WARN_ON(inode
->i_state
& I_NEW
);
1390 inode
->i_state
&= ~I_WILL_FREE
;
1393 inode
->i_state
|= I_FREEING
;
1394 inode_lru_list_del(inode
);
1395 spin_unlock(&inode
->i_lock
);
1401 * iput - put an inode
1402 * @inode: inode to put
1404 * Puts an inode, dropping its usage count. If the inode use count hits
1405 * zero, the inode is then freed and may also be destroyed.
1407 * Consequently, iput() can sleep.
1409 void iput(struct inode
*inode
)
1412 BUG_ON(inode
->i_state
& I_CLEAR
);
1414 if (atomic_dec_and_lock(&inode
->i_count
, &inode
->i_lock
))
1418 EXPORT_SYMBOL(iput
);
1421 * bmap - find a block number in a file
1422 * @inode: inode of file
1423 * @block: block to find
1425 * Returns the block number on the device holding the inode that
1426 * is the disk block number for the block of the file requested.
1427 * That is, asked for block 4 of inode 1 the function will return the
1428 * disk block relative to the disk start that holds that block of the
1431 sector_t
bmap(struct inode
*inode
, sector_t block
)
1434 if (inode
->i_mapping
->a_ops
->bmap
)
1435 res
= inode
->i_mapping
->a_ops
->bmap(inode
->i_mapping
, block
);
1438 EXPORT_SYMBOL(bmap
);
1441 * With relative atime, only update atime if the previous atime is
1442 * earlier than either the ctime or mtime or if at least a day has
1443 * passed since the last atime update.
1445 static int relatime_need_update(struct vfsmount
*mnt
, struct inode
*inode
,
1446 struct timespec now
)
1449 if (!(mnt
->mnt_flags
& MNT_RELATIME
))
1452 * Is mtime younger than atime? If yes, update atime:
1454 if (timespec_compare(&inode
->i_mtime
, &inode
->i_atime
) >= 0)
1457 * Is ctime younger than atime? If yes, update atime:
1459 if (timespec_compare(&inode
->i_ctime
, &inode
->i_atime
) >= 0)
1463 * Is the previous atime value older than a day? If yes,
1466 if ((long)(now
.tv_sec
- inode
->i_atime
.tv_sec
) >= 24*60*60)
1469 * Good, we can skip the atime update:
1475 * touch_atime - update the access time
1476 * @mnt: mount the inode is accessed on
1477 * @dentry: dentry accessed
1479 * Update the accessed time on an inode and mark it for writeback.
1480 * This function automatically handles read only file systems and media,
1481 * as well as the "noatime" flag and inode specific "noatime" markers.
1483 void touch_atime(struct vfsmount
*mnt
, struct dentry
*dentry
)
1485 struct inode
*inode
= dentry
->d_inode
;
1486 struct timespec now
;
1488 if (inode
->i_flags
& S_NOATIME
)
1490 if (IS_NOATIME(inode
))
1492 if ((inode
->i_sb
->s_flags
& MS_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1495 if (mnt
->mnt_flags
& MNT_NOATIME
)
1497 if ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1500 now
= current_fs_time(inode
->i_sb
);
1502 if (!relatime_need_update(mnt
, inode
, now
))
1505 if (timespec_equal(&inode
->i_atime
, &now
))
1508 if (mnt_want_write(mnt
))
1511 inode
->i_atime
= now
;
1512 mark_inode_dirty_sync(inode
);
1513 mnt_drop_write(mnt
);
1515 EXPORT_SYMBOL(touch_atime
);
1518 * file_update_time - update mtime and ctime time
1519 * @file: file accessed
1521 * Update the mtime and ctime members of an inode and mark the inode
1522 * for writeback. Note that this function is meant exclusively for
1523 * usage in the file write path of filesystems, and filesystems may
1524 * choose to explicitly ignore update via this function with the
1525 * S_NOCMTIME inode flag, e.g. for network filesystem where these
1526 * timestamps are handled by the server.
1529 void file_update_time(struct file
*file
)
1531 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1532 struct timespec now
;
1533 enum { S_MTIME
= 1, S_CTIME
= 2, S_VERSION
= 4 } sync_it
= 0;
1535 /* First try to exhaust all avenues to not sync */
1536 if (IS_NOCMTIME(inode
))
1539 now
= current_fs_time(inode
->i_sb
);
1540 if (!timespec_equal(&inode
->i_mtime
, &now
))
1543 if (!timespec_equal(&inode
->i_ctime
, &now
))
1546 if (IS_I_VERSION(inode
))
1547 sync_it
|= S_VERSION
;
1552 /* Finally allowed to write? Takes lock. */
1553 if (mnt_want_write_file(file
))
1556 /* Only change inode inside the lock region */
1557 if (sync_it
& S_VERSION
)
1558 inode_inc_iversion(inode
);
1559 if (sync_it
& S_CTIME
)
1560 inode
->i_ctime
= now
;
1561 if (sync_it
& S_MTIME
)
1562 inode
->i_mtime
= now
;
1563 mark_inode_dirty_sync(inode
);
1564 mnt_drop_write(file
->f_path
.mnt
);
1566 EXPORT_SYMBOL(file_update_time
);
1568 int inode_needs_sync(struct inode
*inode
)
1572 if (S_ISDIR(inode
->i_mode
) && IS_DIRSYNC(inode
))
1576 EXPORT_SYMBOL(inode_needs_sync
);
1578 int inode_wait(void *word
)
1583 EXPORT_SYMBOL(inode_wait
);
1586 * If we try to find an inode in the inode hash while it is being
1587 * deleted, we have to wait until the filesystem completes its
1588 * deletion before reporting that it isn't found. This function waits
1589 * until the deletion _might_ have completed. Callers are responsible
1590 * to recheck inode state.
1592 * It doesn't matter if I_NEW is not set initially, a call to
1593 * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
1596 static void __wait_on_freeing_inode(struct inode
*inode
)
1598 wait_queue_head_t
*wq
;
1599 DEFINE_WAIT_BIT(wait
, &inode
->i_state
, __I_NEW
);
1600 wq
= bit_waitqueue(&inode
->i_state
, __I_NEW
);
1601 prepare_to_wait(wq
, &wait
.wait
, TASK_UNINTERRUPTIBLE
);
1602 spin_unlock(&inode
->i_lock
);
1603 spin_unlock(&inode_hash_lock
);
1605 finish_wait(wq
, &wait
.wait
);
1606 spin_lock(&inode_hash_lock
);
1609 static __initdata
unsigned long ihash_entries
;
1610 static int __init
set_ihash_entries(char *str
)
1614 ihash_entries
= simple_strtoul(str
, &str
, 0);
1617 __setup("ihash_entries=", set_ihash_entries
);
1620 * Initialize the waitqueues and inode hash table.
1622 void __init
inode_init_early(void)
1626 /* If hashes are distributed across NUMA nodes, defer
1627 * hash allocation until vmalloc space is available.
1633 alloc_large_system_hash("Inode-cache",
1634 sizeof(struct hlist_head
),
1642 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1643 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1646 void __init
inode_init(void)
1650 /* inode slab cache */
1651 inode_cachep
= kmem_cache_create("inode_cache",
1652 sizeof(struct inode
),
1654 (SLAB_RECLAIM_ACCOUNT
|SLAB_PANIC
|
1657 register_shrinker(&icache_shrinker
);
1659 /* Hash may have been set up in inode_init_early */
1664 alloc_large_system_hash("Inode-cache",
1665 sizeof(struct hlist_head
),
1673 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1674 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1677 void init_special_inode(struct inode
*inode
, umode_t mode
, dev_t rdev
)
1679 inode
->i_mode
= mode
;
1680 if (S_ISCHR(mode
)) {
1681 inode
->i_fop
= &def_chr_fops
;
1682 inode
->i_rdev
= rdev
;
1683 } else if (S_ISBLK(mode
)) {
1684 inode
->i_fop
= &def_blk_fops
;
1685 inode
->i_rdev
= rdev
;
1686 } else if (S_ISFIFO(mode
))
1687 inode
->i_fop
= &def_fifo_fops
;
1688 else if (S_ISSOCK(mode
))
1689 inode
->i_fop
= &bad_sock_fops
;
1691 printk(KERN_DEBUG
"init_special_inode: bogus i_mode (%o) for"
1692 " inode %s:%lu\n", mode
, inode
->i_sb
->s_id
,
1695 EXPORT_SYMBOL(init_special_inode
);
1698 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
1700 * @dir: Directory inode
1701 * @mode: mode of the new inode
1703 void inode_init_owner(struct inode
*inode
, const struct inode
*dir
,
1706 inode
->i_uid
= current_fsuid();
1707 if (dir
&& dir
->i_mode
& S_ISGID
) {
1708 inode
->i_gid
= dir
->i_gid
;
1712 inode
->i_gid
= current_fsgid();
1713 inode
->i_mode
= mode
;
1715 EXPORT_SYMBOL(inode_init_owner
);
1718 * inode_owner_or_capable - check current task permissions to inode
1719 * @inode: inode being checked
1721 * Return true if current either has CAP_FOWNER to the inode, or
1724 bool inode_owner_or_capable(const struct inode
*inode
)
1726 struct user_namespace
*ns
= inode_userns(inode
);
1728 if (current_user_ns() == ns
&& current_fsuid() == inode
->i_uid
)
1730 if (ns_capable(ns
, CAP_FOWNER
))
1734 EXPORT_SYMBOL(inode_owner_or_capable
);