2 * (C) 1997 Linus Torvalds
3 * (C) 1999 Andrea Arcangeli <andrea@suse.de> (dynamic inode allocation)
5 #include <linux/export.h>
8 #include <linux/backing-dev.h>
9 #include <linux/hash.h>
10 #include <linux/swap.h>
11 #include <linux/security.h>
12 #include <linux/cdev.h>
13 #include <linux/bootmem.h>
14 #include <linux/fsnotify.h>
15 #include <linux/mount.h>
16 #include <linux/posix_acl.h>
17 #include <linux/prefetch.h>
18 #include <linux/buffer_head.h> /* for inode_has_buffers */
19 #include <linux/ratelimit.h>
23 * Inode locking rules:
25 * inode->i_lock protects:
26 * inode->i_state, inode->i_hash, __iget()
27 * inode->i_sb->s_inode_lru_lock protects:
28 * inode->i_sb->s_inode_lru, inode->i_lru
29 * inode_sb_list_lock protects:
30 * sb->s_inodes, inode->i_sb_list
31 * bdi->wb.list_lock protects:
32 * bdi->wb.b_{dirty,io,more_io}, inode->i_wb_list
33 * inode_hash_lock protects:
34 * inode_hashtable, inode->i_hash
40 * inode->i_sb->s_inode_lru_lock
53 static unsigned int i_hash_mask __read_mostly
;
54 static unsigned int i_hash_shift __read_mostly
;
55 static struct hlist_head
*inode_hashtable __read_mostly
;
56 static __cacheline_aligned_in_smp
DEFINE_SPINLOCK(inode_hash_lock
);
58 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(inode_sb_list_lock
);
61 * Empty aops. Can be used for the cases where the user does not
62 * define any of the address_space operations.
64 const struct address_space_operations empty_aops
= {
66 EXPORT_SYMBOL(empty_aops
);
69 * Statistics gathering..
71 struct inodes_stat_t inodes_stat
;
73 static DEFINE_PER_CPU(unsigned int, nr_inodes
);
74 static DEFINE_PER_CPU(unsigned int, nr_unused
);
76 static struct kmem_cache
*inode_cachep __read_mostly
;
78 static int get_nr_inodes(void)
82 for_each_possible_cpu(i
)
83 sum
+= per_cpu(nr_inodes
, i
);
84 return sum
< 0 ? 0 : sum
;
87 static inline int get_nr_inodes_unused(void)
91 for_each_possible_cpu(i
)
92 sum
+= per_cpu(nr_unused
, i
);
93 return sum
< 0 ? 0 : sum
;
96 int get_nr_dirty_inodes(void)
98 /* not actually dirty inodes, but a wild approximation */
99 int nr_dirty
= get_nr_inodes() - get_nr_inodes_unused();
100 return nr_dirty
> 0 ? nr_dirty
: 0;
104 * Handle nr_inode sysctl
107 int proc_nr_inodes(ctl_table
*table
, int write
,
108 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
110 inodes_stat
.nr_inodes
= get_nr_inodes();
111 inodes_stat
.nr_unused
= get_nr_inodes_unused();
112 return proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
117 * inode_init_always - perform inode structure intialisation
118 * @sb: superblock inode belongs to
119 * @inode: inode to initialise
121 * These are initializations that need to be done on every inode
122 * allocation as the fields are not initialised by slab allocation.
124 int inode_init_always(struct super_block
*sb
, struct inode
*inode
)
126 static const struct inode_operations empty_iops
;
127 static const struct file_operations empty_fops
;
128 struct address_space
*const mapping
= &inode
->i_data
;
131 inode
->i_blkbits
= sb
->s_blocksize_bits
;
133 atomic_set(&inode
->i_count
, 1);
134 inode
->i_op
= &empty_iops
;
135 inode
->i_fop
= &empty_fops
;
136 inode
->__i_nlink
= 1;
137 inode
->i_opflags
= 0;
140 atomic_set(&inode
->i_writecount
, 0);
144 inode
->i_generation
= 0;
146 memset(&inode
->i_dquot
, 0, sizeof(inode
->i_dquot
));
148 inode
->i_pipe
= NULL
;
149 inode
->i_bdev
= NULL
;
150 inode
->i_cdev
= NULL
;
152 inode
->dirtied_when
= 0;
154 if (security_inode_alloc(inode
))
156 spin_lock_init(&inode
->i_lock
);
157 lockdep_set_class(&inode
->i_lock
, &sb
->s_type
->i_lock_key
);
159 mutex_init(&inode
->i_mutex
);
160 lockdep_set_class(&inode
->i_mutex
, &sb
->s_type
->i_mutex_key
);
162 atomic_set(&inode
->i_dio_count
, 0);
164 mapping
->a_ops
= &empty_aops
;
165 mapping
->host
= inode
;
167 mapping_set_gfp_mask(mapping
, GFP_HIGHUSER_MOVABLE
);
168 mapping
->assoc_mapping
= NULL
;
169 mapping
->backing_dev_info
= &default_backing_dev_info
;
170 mapping
->writeback_index
= 0;
173 * If the block_device provides a backing_dev_info for client
174 * inodes then use that. Otherwise the inode share the bdev's
178 struct backing_dev_info
*bdi
;
180 bdi
= sb
->s_bdev
->bd_inode
->i_mapping
->backing_dev_info
;
181 mapping
->backing_dev_info
= bdi
;
183 inode
->i_private
= NULL
;
184 inode
->i_mapping
= mapping
;
185 INIT_LIST_HEAD(&inode
->i_dentry
); /* buggered by rcu freeing */
186 #ifdef CONFIG_FS_POSIX_ACL
187 inode
->i_acl
= inode
->i_default_acl
= ACL_NOT_CACHED
;
190 #ifdef CONFIG_FSNOTIFY
191 inode
->i_fsnotify_mask
= 0;
194 this_cpu_inc(nr_inodes
);
200 EXPORT_SYMBOL(inode_init_always
);
202 static struct inode
*alloc_inode(struct super_block
*sb
)
206 if (sb
->s_op
->alloc_inode
)
207 inode
= sb
->s_op
->alloc_inode(sb
);
209 inode
= kmem_cache_alloc(inode_cachep
, GFP_KERNEL
);
214 if (unlikely(inode_init_always(sb
, inode
))) {
215 if (inode
->i_sb
->s_op
->destroy_inode
)
216 inode
->i_sb
->s_op
->destroy_inode(inode
);
218 kmem_cache_free(inode_cachep
, inode
);
225 void free_inode_nonrcu(struct inode
*inode
)
227 kmem_cache_free(inode_cachep
, inode
);
229 EXPORT_SYMBOL(free_inode_nonrcu
);
231 void __destroy_inode(struct inode
*inode
)
233 BUG_ON(inode_has_buffers(inode
));
234 security_inode_free(inode
);
235 fsnotify_inode_delete(inode
);
236 if (!inode
->i_nlink
) {
237 WARN_ON(atomic_long_read(&inode
->i_sb
->s_remove_count
) == 0);
238 atomic_long_dec(&inode
->i_sb
->s_remove_count
);
241 #ifdef CONFIG_FS_POSIX_ACL
242 if (inode
->i_acl
&& inode
->i_acl
!= ACL_NOT_CACHED
)
243 posix_acl_release(inode
->i_acl
);
244 if (inode
->i_default_acl
&& inode
->i_default_acl
!= ACL_NOT_CACHED
)
245 posix_acl_release(inode
->i_default_acl
);
247 this_cpu_dec(nr_inodes
);
249 EXPORT_SYMBOL(__destroy_inode
);
251 static void i_callback(struct rcu_head
*head
)
253 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
254 kmem_cache_free(inode_cachep
, inode
);
257 static void destroy_inode(struct inode
*inode
)
259 BUG_ON(!list_empty(&inode
->i_lru
));
260 __destroy_inode(inode
);
261 if (inode
->i_sb
->s_op
->destroy_inode
)
262 inode
->i_sb
->s_op
->destroy_inode(inode
);
264 call_rcu(&inode
->i_rcu
, i_callback
);
268 * drop_nlink - directly drop an inode's link count
271 * This is a low-level filesystem helper to replace any
272 * direct filesystem manipulation of i_nlink. In cases
273 * where we are attempting to track writes to the
274 * filesystem, a decrement to zero means an imminent
275 * write when the file is truncated and actually unlinked
278 void drop_nlink(struct inode
*inode
)
280 WARN_ON(inode
->i_nlink
== 0);
283 atomic_long_inc(&inode
->i_sb
->s_remove_count
);
285 EXPORT_SYMBOL(drop_nlink
);
288 * clear_nlink - directly zero an inode's link count
291 * This is a low-level filesystem helper to replace any
292 * direct filesystem manipulation of i_nlink. See
293 * drop_nlink() for why we care about i_nlink hitting zero.
295 void clear_nlink(struct inode
*inode
)
297 if (inode
->i_nlink
) {
298 inode
->__i_nlink
= 0;
299 atomic_long_inc(&inode
->i_sb
->s_remove_count
);
302 EXPORT_SYMBOL(clear_nlink
);
305 * set_nlink - directly set an inode's link count
307 * @nlink: new nlink (should be non-zero)
309 * This is a low-level filesystem helper to replace any
310 * direct filesystem manipulation of i_nlink.
312 void set_nlink(struct inode
*inode
, unsigned int nlink
)
317 /* Yes, some filesystems do change nlink from zero to one */
318 if (inode
->i_nlink
== 0)
319 atomic_long_dec(&inode
->i_sb
->s_remove_count
);
321 inode
->__i_nlink
= nlink
;
324 EXPORT_SYMBOL(set_nlink
);
327 * inc_nlink - directly increment an inode's link count
330 * This is a low-level filesystem helper to replace any
331 * direct filesystem manipulation of i_nlink. Currently,
332 * it is only here for parity with dec_nlink().
334 void inc_nlink(struct inode
*inode
)
336 if (WARN_ON(inode
->i_nlink
== 0))
337 atomic_long_dec(&inode
->i_sb
->s_remove_count
);
341 EXPORT_SYMBOL(inc_nlink
);
343 void address_space_init_once(struct address_space
*mapping
)
345 memset(mapping
, 0, sizeof(*mapping
));
346 INIT_RADIX_TREE(&mapping
->page_tree
, GFP_ATOMIC
);
347 spin_lock_init(&mapping
->tree_lock
);
348 mutex_init(&mapping
->i_mmap_mutex
);
349 INIT_LIST_HEAD(&mapping
->private_list
);
350 spin_lock_init(&mapping
->private_lock
);
351 INIT_RAW_PRIO_TREE_ROOT(&mapping
->i_mmap
);
352 INIT_LIST_HEAD(&mapping
->i_mmap_nonlinear
);
354 EXPORT_SYMBOL(address_space_init_once
);
357 * These are initializations that only need to be done
358 * once, because the fields are idempotent across use
359 * of the inode, so let the slab aware of that.
361 void inode_init_once(struct inode
*inode
)
363 memset(inode
, 0, sizeof(*inode
));
364 INIT_HLIST_NODE(&inode
->i_hash
);
365 INIT_LIST_HEAD(&inode
->i_devices
);
366 INIT_LIST_HEAD(&inode
->i_wb_list
);
367 INIT_LIST_HEAD(&inode
->i_lru
);
368 address_space_init_once(&inode
->i_data
);
369 i_size_ordered_init(inode
);
370 #ifdef CONFIG_FSNOTIFY
371 INIT_HLIST_HEAD(&inode
->i_fsnotify_marks
);
374 EXPORT_SYMBOL(inode_init_once
);
376 static void init_once(void *foo
)
378 struct inode
*inode
= (struct inode
*) foo
;
380 inode_init_once(inode
);
384 * inode->i_lock must be held
386 void __iget(struct inode
*inode
)
388 atomic_inc(&inode
->i_count
);
392 * get additional reference to inode; caller must already hold one.
394 void ihold(struct inode
*inode
)
396 WARN_ON(atomic_inc_return(&inode
->i_count
) < 2);
398 EXPORT_SYMBOL(ihold
);
400 static void inode_lru_list_add(struct inode
*inode
)
402 spin_lock(&inode
->i_sb
->s_inode_lru_lock
);
403 if (list_empty(&inode
->i_lru
)) {
404 list_add(&inode
->i_lru
, &inode
->i_sb
->s_inode_lru
);
405 inode
->i_sb
->s_nr_inodes_unused
++;
406 this_cpu_inc(nr_unused
);
408 spin_unlock(&inode
->i_sb
->s_inode_lru_lock
);
411 static void inode_lru_list_del(struct inode
*inode
)
413 spin_lock(&inode
->i_sb
->s_inode_lru_lock
);
414 if (!list_empty(&inode
->i_lru
)) {
415 list_del_init(&inode
->i_lru
);
416 inode
->i_sb
->s_nr_inodes_unused
--;
417 this_cpu_dec(nr_unused
);
419 spin_unlock(&inode
->i_sb
->s_inode_lru_lock
);
423 * inode_sb_list_add - add inode to the superblock list of inodes
424 * @inode: inode to add
426 void inode_sb_list_add(struct inode
*inode
)
428 spin_lock(&inode_sb_list_lock
);
429 list_add(&inode
->i_sb_list
, &inode
->i_sb
->s_inodes
);
430 spin_unlock(&inode_sb_list_lock
);
432 EXPORT_SYMBOL_GPL(inode_sb_list_add
);
434 static inline void inode_sb_list_del(struct inode
*inode
)
436 if (!list_empty(&inode
->i_sb_list
)) {
437 spin_lock(&inode_sb_list_lock
);
438 list_del_init(&inode
->i_sb_list
);
439 spin_unlock(&inode_sb_list_lock
);
443 static unsigned long hash(struct super_block
*sb
, unsigned long hashval
)
447 tmp
= (hashval
* (unsigned long)sb
) ^ (GOLDEN_RATIO_PRIME
+ hashval
) /
449 tmp
= tmp
^ ((tmp
^ GOLDEN_RATIO_PRIME
) >> i_hash_shift
);
450 return tmp
& i_hash_mask
;
454 * __insert_inode_hash - hash an inode
455 * @inode: unhashed inode
456 * @hashval: unsigned long value used to locate this object in the
459 * Add an inode to the inode hash for this superblock.
461 void __insert_inode_hash(struct inode
*inode
, unsigned long hashval
)
463 struct hlist_head
*b
= inode_hashtable
+ hash(inode
->i_sb
, hashval
);
465 spin_lock(&inode_hash_lock
);
466 spin_lock(&inode
->i_lock
);
467 hlist_add_head(&inode
->i_hash
, b
);
468 spin_unlock(&inode
->i_lock
);
469 spin_unlock(&inode_hash_lock
);
471 EXPORT_SYMBOL(__insert_inode_hash
);
474 * __remove_inode_hash - remove an inode from the hash
475 * @inode: inode to unhash
477 * Remove an inode from the superblock.
479 void __remove_inode_hash(struct inode
*inode
)
481 spin_lock(&inode_hash_lock
);
482 spin_lock(&inode
->i_lock
);
483 hlist_del_init(&inode
->i_hash
);
484 spin_unlock(&inode
->i_lock
);
485 spin_unlock(&inode_hash_lock
);
487 EXPORT_SYMBOL(__remove_inode_hash
);
489 void clear_inode(struct inode
*inode
)
493 * We have to cycle tree_lock here because reclaim can be still in the
494 * process of removing the last page (in __delete_from_page_cache())
495 * and we must not free mapping under it.
497 spin_lock_irq(&inode
->i_data
.tree_lock
);
498 BUG_ON(inode
->i_data
.nrpages
);
499 spin_unlock_irq(&inode
->i_data
.tree_lock
);
500 BUG_ON(!list_empty(&inode
->i_data
.private_list
));
501 BUG_ON(!(inode
->i_state
& I_FREEING
));
502 BUG_ON(inode
->i_state
& I_CLEAR
);
503 /* don't need i_lock here, no concurrent mods to i_state */
504 inode
->i_state
= I_FREEING
| I_CLEAR
;
506 EXPORT_SYMBOL(clear_inode
);
509 * Free the inode passed in, removing it from the lists it is still connected
510 * to. We remove any pages still attached to the inode and wait for any IO that
511 * is still in progress before finally destroying the inode.
513 * An inode must already be marked I_FREEING so that we avoid the inode being
514 * moved back onto lists if we race with other code that manipulates the lists
515 * (e.g. writeback_single_inode). The caller is responsible for setting this.
517 * An inode must already be removed from the LRU list before being evicted from
518 * the cache. This should occur atomically with setting the I_FREEING state
519 * flag, so no inodes here should ever be on the LRU when being evicted.
521 static void evict(struct inode
*inode
)
523 const struct super_operations
*op
= inode
->i_sb
->s_op
;
525 BUG_ON(!(inode
->i_state
& I_FREEING
));
526 BUG_ON(!list_empty(&inode
->i_lru
));
528 if (!list_empty(&inode
->i_wb_list
))
529 inode_wb_list_del(inode
);
531 inode_sb_list_del(inode
);
533 inode_sync_wait(inode
);
535 if (op
->evict_inode
) {
536 op
->evict_inode(inode
);
538 if (inode
->i_data
.nrpages
)
539 truncate_inode_pages(&inode
->i_data
, 0);
542 if (S_ISBLK(inode
->i_mode
) && inode
->i_bdev
)
544 if (S_ISCHR(inode
->i_mode
) && inode
->i_cdev
)
547 remove_inode_hash(inode
);
549 spin_lock(&inode
->i_lock
);
550 wake_up_bit(&inode
->i_state
, __I_NEW
);
551 BUG_ON(inode
->i_state
!= (I_FREEING
| I_CLEAR
));
552 spin_unlock(&inode
->i_lock
);
554 destroy_inode(inode
);
558 * dispose_list - dispose of the contents of a local list
559 * @head: the head of the list to free
561 * Dispose-list gets a local list with local inodes in it, so it doesn't
562 * need to worry about list corruption and SMP locks.
564 static void dispose_list(struct list_head
*head
)
566 while (!list_empty(head
)) {
569 inode
= list_first_entry(head
, struct inode
, i_lru
);
570 list_del_init(&inode
->i_lru
);
577 * evict_inodes - evict all evictable inodes for a superblock
578 * @sb: superblock to operate on
580 * Make sure that no inodes with zero refcount are retained. This is
581 * called by superblock shutdown after having MS_ACTIVE flag removed,
582 * so any inode reaching zero refcount during or after that call will
583 * be immediately evicted.
585 void evict_inodes(struct super_block
*sb
)
587 struct inode
*inode
, *next
;
590 spin_lock(&inode_sb_list_lock
);
591 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
592 if (atomic_read(&inode
->i_count
))
595 spin_lock(&inode
->i_lock
);
596 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
597 spin_unlock(&inode
->i_lock
);
601 inode
->i_state
|= I_FREEING
;
602 inode_lru_list_del(inode
);
603 spin_unlock(&inode
->i_lock
);
604 list_add(&inode
->i_lru
, &dispose
);
606 spin_unlock(&inode_sb_list_lock
);
608 dispose_list(&dispose
);
612 * invalidate_inodes - attempt to free all inodes on a superblock
613 * @sb: superblock to operate on
614 * @kill_dirty: flag to guide handling of dirty inodes
616 * Attempts to free all inodes for a given superblock. If there were any
617 * busy inodes return a non-zero value, else zero.
618 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
621 int invalidate_inodes(struct super_block
*sb
, bool kill_dirty
)
624 struct inode
*inode
, *next
;
627 spin_lock(&inode_sb_list_lock
);
628 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
629 spin_lock(&inode
->i_lock
);
630 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
631 spin_unlock(&inode
->i_lock
);
634 if (inode
->i_state
& I_DIRTY
&& !kill_dirty
) {
635 spin_unlock(&inode
->i_lock
);
639 if (atomic_read(&inode
->i_count
)) {
640 spin_unlock(&inode
->i_lock
);
645 inode
->i_state
|= I_FREEING
;
646 inode_lru_list_del(inode
);
647 spin_unlock(&inode
->i_lock
);
648 list_add(&inode
->i_lru
, &dispose
);
650 spin_unlock(&inode_sb_list_lock
);
652 dispose_list(&dispose
);
657 static int can_unuse(struct inode
*inode
)
659 if (inode
->i_state
& ~I_REFERENCED
)
661 if (inode_has_buffers(inode
))
663 if (atomic_read(&inode
->i_count
))
665 if (inode
->i_data
.nrpages
)
671 * Walk the superblock inode LRU for freeable inodes and attempt to free them.
672 * This is called from the superblock shrinker function with a number of inodes
673 * to trim from the LRU. Inodes to be freed are moved to a temporary list and
674 * then are freed outside inode_lock by dispose_list().
676 * Any inodes which are pinned purely because of attached pagecache have their
677 * pagecache removed. If the inode has metadata buffers attached to
678 * mapping->private_list then try to remove them.
680 * If the inode has the I_REFERENCED flag set, then it means that it has been
681 * used recently - the flag is set in iput_final(). When we encounter such an
682 * inode, clear the flag and move it to the back of the LRU so it gets another
683 * pass through the LRU before it gets reclaimed. This is necessary because of
684 * the fact we are doing lazy LRU updates to minimise lock contention so the
685 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
686 * with this flag set because they are the inodes that are out of order.
688 void prune_icache_sb(struct super_block
*sb
, int nr_to_scan
)
692 unsigned long reap
= 0;
694 spin_lock(&sb
->s_inode_lru_lock
);
695 for (nr_scanned
= nr_to_scan
; nr_scanned
>= 0; nr_scanned
--) {
698 if (list_empty(&sb
->s_inode_lru
))
701 inode
= list_entry(sb
->s_inode_lru
.prev
, struct inode
, i_lru
);
704 * we are inverting the sb->s_inode_lru_lock/inode->i_lock here,
705 * so use a trylock. If we fail to get the lock, just move the
706 * inode to the back of the list so we don't spin on it.
708 if (!spin_trylock(&inode
->i_lock
)) {
709 list_move_tail(&inode
->i_lru
, &sb
->s_inode_lru
);
714 * Referenced or dirty inodes are still in use. Give them
715 * another pass through the LRU as we canot reclaim them now.
717 if (atomic_read(&inode
->i_count
) ||
718 (inode
->i_state
& ~I_REFERENCED
)) {
719 list_del_init(&inode
->i_lru
);
720 spin_unlock(&inode
->i_lock
);
721 sb
->s_nr_inodes_unused
--;
722 this_cpu_dec(nr_unused
);
726 /* recently referenced inodes get one more pass */
727 if (inode
->i_state
& I_REFERENCED
) {
728 inode
->i_state
&= ~I_REFERENCED
;
729 list_move(&inode
->i_lru
, &sb
->s_inode_lru
);
730 spin_unlock(&inode
->i_lock
);
733 if (inode_has_buffers(inode
) || inode
->i_data
.nrpages
) {
735 spin_unlock(&inode
->i_lock
);
736 spin_unlock(&sb
->s_inode_lru_lock
);
737 if (remove_inode_buffers(inode
))
738 reap
+= invalidate_mapping_pages(&inode
->i_data
,
741 spin_lock(&sb
->s_inode_lru_lock
);
743 if (inode
!= list_entry(sb
->s_inode_lru
.next
,
744 struct inode
, i_lru
))
745 continue; /* wrong inode or list_empty */
746 /* avoid lock inversions with trylock */
747 if (!spin_trylock(&inode
->i_lock
))
749 if (!can_unuse(inode
)) {
750 spin_unlock(&inode
->i_lock
);
754 WARN_ON(inode
->i_state
& I_NEW
);
755 inode
->i_state
|= I_FREEING
;
756 spin_unlock(&inode
->i_lock
);
758 list_move(&inode
->i_lru
, &freeable
);
759 sb
->s_nr_inodes_unused
--;
760 this_cpu_dec(nr_unused
);
762 if (current_is_kswapd())
763 __count_vm_events(KSWAPD_INODESTEAL
, reap
);
765 __count_vm_events(PGINODESTEAL
, reap
);
766 spin_unlock(&sb
->s_inode_lru_lock
);
767 if (current
->reclaim_state
)
768 current
->reclaim_state
->reclaimed_slab
+= reap
;
770 dispose_list(&freeable
);
773 static void __wait_on_freeing_inode(struct inode
*inode
);
775 * Called with the inode lock held.
777 static struct inode
*find_inode(struct super_block
*sb
,
778 struct hlist_head
*head
,
779 int (*test
)(struct inode
*, void *),
782 struct hlist_node
*node
;
783 struct inode
*inode
= NULL
;
786 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
787 spin_lock(&inode
->i_lock
);
788 if (inode
->i_sb
!= sb
) {
789 spin_unlock(&inode
->i_lock
);
792 if (!test(inode
, data
)) {
793 spin_unlock(&inode
->i_lock
);
796 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
797 __wait_on_freeing_inode(inode
);
801 spin_unlock(&inode
->i_lock
);
808 * find_inode_fast is the fast path version of find_inode, see the comment at
809 * iget_locked for details.
811 static struct inode
*find_inode_fast(struct super_block
*sb
,
812 struct hlist_head
*head
, unsigned long ino
)
814 struct hlist_node
*node
;
815 struct inode
*inode
= NULL
;
818 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
819 spin_lock(&inode
->i_lock
);
820 if (inode
->i_ino
!= ino
) {
821 spin_unlock(&inode
->i_lock
);
824 if (inode
->i_sb
!= sb
) {
825 spin_unlock(&inode
->i_lock
);
828 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
829 __wait_on_freeing_inode(inode
);
833 spin_unlock(&inode
->i_lock
);
840 * Each cpu owns a range of LAST_INO_BATCH numbers.
841 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
842 * to renew the exhausted range.
844 * This does not significantly increase overflow rate because every CPU can
845 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
846 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
847 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
848 * overflow rate by 2x, which does not seem too significant.
850 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
851 * error if st_ino won't fit in target struct field. Use 32bit counter
852 * here to attempt to avoid that.
854 #define LAST_INO_BATCH 1024
855 static DEFINE_PER_CPU(unsigned int, last_ino
);
857 unsigned int get_next_ino(void)
859 unsigned int *p
= &get_cpu_var(last_ino
);
860 unsigned int res
= *p
;
863 if (unlikely((res
& (LAST_INO_BATCH
-1)) == 0)) {
864 static atomic_t shared_last_ino
;
865 int next
= atomic_add_return(LAST_INO_BATCH
, &shared_last_ino
);
867 res
= next
- LAST_INO_BATCH
;
872 put_cpu_var(last_ino
);
875 EXPORT_SYMBOL(get_next_ino
);
878 * new_inode_pseudo - obtain an inode
881 * Allocates a new inode for given superblock.
882 * Inode wont be chained in superblock s_inodes list
884 * - fs can't be unmount
885 * - quotas, fsnotify, writeback can't work
887 struct inode
*new_inode_pseudo(struct super_block
*sb
)
889 struct inode
*inode
= alloc_inode(sb
);
892 spin_lock(&inode
->i_lock
);
894 spin_unlock(&inode
->i_lock
);
895 INIT_LIST_HEAD(&inode
->i_sb_list
);
901 * new_inode - obtain an inode
904 * Allocates a new inode for given superblock. The default gfp_mask
905 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
906 * If HIGHMEM pages are unsuitable or it is known that pages allocated
907 * for the page cache are not reclaimable or migratable,
908 * mapping_set_gfp_mask() must be called with suitable flags on the
909 * newly created inode's mapping
912 struct inode
*new_inode(struct super_block
*sb
)
916 spin_lock_prefetch(&inode_sb_list_lock
);
918 inode
= new_inode_pseudo(sb
);
920 inode_sb_list_add(inode
);
923 EXPORT_SYMBOL(new_inode
);
925 #ifdef CONFIG_DEBUG_LOCK_ALLOC
926 void lockdep_annotate_inode_mutex_key(struct inode
*inode
)
928 if (S_ISDIR(inode
->i_mode
)) {
929 struct file_system_type
*type
= inode
->i_sb
->s_type
;
931 /* Set new key only if filesystem hasn't already changed it */
932 if (lockdep_match_class(&inode
->i_mutex
, &type
->i_mutex_key
)) {
934 * ensure nobody is actually holding i_mutex
936 mutex_destroy(&inode
->i_mutex
);
937 mutex_init(&inode
->i_mutex
);
938 lockdep_set_class(&inode
->i_mutex
,
939 &type
->i_mutex_dir_key
);
943 EXPORT_SYMBOL(lockdep_annotate_inode_mutex_key
);
947 * unlock_new_inode - clear the I_NEW state and wake up any waiters
948 * @inode: new inode to unlock
950 * Called when the inode is fully initialised to clear the new state of the
951 * inode and wake up anyone waiting for the inode to finish initialisation.
953 void unlock_new_inode(struct inode
*inode
)
955 lockdep_annotate_inode_mutex_key(inode
);
956 spin_lock(&inode
->i_lock
);
957 WARN_ON(!(inode
->i_state
& I_NEW
));
958 inode
->i_state
&= ~I_NEW
;
960 wake_up_bit(&inode
->i_state
, __I_NEW
);
961 spin_unlock(&inode
->i_lock
);
963 EXPORT_SYMBOL(unlock_new_inode
);
966 * iget5_locked - obtain an inode from a mounted file system
967 * @sb: super block of file system
968 * @hashval: hash value (usually inode number) to get
969 * @test: callback used for comparisons between inodes
970 * @set: callback used to initialize a new struct inode
971 * @data: opaque data pointer to pass to @test and @set
973 * Search for the inode specified by @hashval and @data in the inode cache,
974 * and if present it is return it with an increased reference count. This is
975 * a generalized version of iget_locked() for file systems where the inode
976 * number is not sufficient for unique identification of an inode.
978 * If the inode is not in cache, allocate a new inode and return it locked,
979 * hashed, and with the I_NEW flag set. The file system gets to fill it in
980 * before unlocking it via unlock_new_inode().
982 * Note both @test and @set are called with the inode_hash_lock held, so can't
985 struct inode
*iget5_locked(struct super_block
*sb
, unsigned long hashval
,
986 int (*test
)(struct inode
*, void *),
987 int (*set
)(struct inode
*, void *), void *data
)
989 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
992 spin_lock(&inode_hash_lock
);
993 inode
= find_inode(sb
, head
, test
, data
);
994 spin_unlock(&inode_hash_lock
);
997 wait_on_inode(inode
);
1001 inode
= alloc_inode(sb
);
1005 spin_lock(&inode_hash_lock
);
1006 /* We released the lock, so.. */
1007 old
= find_inode(sb
, head
, test
, data
);
1009 if (set(inode
, data
))
1012 spin_lock(&inode
->i_lock
);
1013 inode
->i_state
= I_NEW
;
1014 hlist_add_head(&inode
->i_hash
, head
);
1015 spin_unlock(&inode
->i_lock
);
1016 inode_sb_list_add(inode
);
1017 spin_unlock(&inode_hash_lock
);
1019 /* Return the locked inode with I_NEW set, the
1020 * caller is responsible for filling in the contents
1026 * Uhhuh, somebody else created the same inode under
1027 * us. Use the old inode instead of the one we just
1030 spin_unlock(&inode_hash_lock
);
1031 destroy_inode(inode
);
1033 wait_on_inode(inode
);
1038 spin_unlock(&inode_hash_lock
);
1039 destroy_inode(inode
);
1042 EXPORT_SYMBOL(iget5_locked
);
1045 * iget_locked - obtain an inode from a mounted file system
1046 * @sb: super block of file system
1047 * @ino: inode number to get
1049 * Search for the inode specified by @ino in the inode cache and if present
1050 * return it with an increased reference count. This is for file systems
1051 * where the inode number is sufficient for unique identification of an inode.
1053 * If the inode is not in cache, allocate a new inode and return it locked,
1054 * hashed, and with the I_NEW flag set. The file system gets to fill it in
1055 * before unlocking it via unlock_new_inode().
1057 struct inode
*iget_locked(struct super_block
*sb
, unsigned long ino
)
1059 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1060 struct inode
*inode
;
1062 spin_lock(&inode_hash_lock
);
1063 inode
= find_inode_fast(sb
, head
, ino
);
1064 spin_unlock(&inode_hash_lock
);
1066 wait_on_inode(inode
);
1070 inode
= alloc_inode(sb
);
1074 spin_lock(&inode_hash_lock
);
1075 /* We released the lock, so.. */
1076 old
= find_inode_fast(sb
, head
, ino
);
1079 spin_lock(&inode
->i_lock
);
1080 inode
->i_state
= I_NEW
;
1081 hlist_add_head(&inode
->i_hash
, head
);
1082 spin_unlock(&inode
->i_lock
);
1083 inode_sb_list_add(inode
);
1084 spin_unlock(&inode_hash_lock
);
1086 /* Return the locked inode with I_NEW set, the
1087 * caller is responsible for filling in the contents
1093 * Uhhuh, somebody else created the same inode under
1094 * us. Use the old inode instead of the one we just
1097 spin_unlock(&inode_hash_lock
);
1098 destroy_inode(inode
);
1100 wait_on_inode(inode
);
1104 EXPORT_SYMBOL(iget_locked
);
1107 * search the inode cache for a matching inode number.
1108 * If we find one, then the inode number we are trying to
1109 * allocate is not unique and so we should not use it.
1111 * Returns 1 if the inode number is unique, 0 if it is not.
1113 static int test_inode_iunique(struct super_block
*sb
, unsigned long ino
)
1115 struct hlist_head
*b
= inode_hashtable
+ hash(sb
, ino
);
1116 struct hlist_node
*node
;
1117 struct inode
*inode
;
1119 spin_lock(&inode_hash_lock
);
1120 hlist_for_each_entry(inode
, node
, b
, i_hash
) {
1121 if (inode
->i_ino
== ino
&& inode
->i_sb
== sb
) {
1122 spin_unlock(&inode_hash_lock
);
1126 spin_unlock(&inode_hash_lock
);
1132 * iunique - get a unique inode number
1134 * @max_reserved: highest reserved inode number
1136 * Obtain an inode number that is unique on the system for a given
1137 * superblock. This is used by file systems that have no natural
1138 * permanent inode numbering system. An inode number is returned that
1139 * is higher than the reserved limit but unique.
1142 * With a large number of inodes live on the file system this function
1143 * currently becomes quite slow.
1145 ino_t
iunique(struct super_block
*sb
, ino_t max_reserved
)
1148 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1149 * error if st_ino won't fit in target struct field. Use 32bit counter
1150 * here to attempt to avoid that.
1152 static DEFINE_SPINLOCK(iunique_lock
);
1153 static unsigned int counter
;
1156 spin_lock(&iunique_lock
);
1158 if (counter
<= max_reserved
)
1159 counter
= max_reserved
+ 1;
1161 } while (!test_inode_iunique(sb
, res
));
1162 spin_unlock(&iunique_lock
);
1166 EXPORT_SYMBOL(iunique
);
1168 struct inode
*igrab(struct inode
*inode
)
1170 spin_lock(&inode
->i_lock
);
1171 if (!(inode
->i_state
& (I_FREEING
|I_WILL_FREE
))) {
1173 spin_unlock(&inode
->i_lock
);
1175 spin_unlock(&inode
->i_lock
);
1177 * Handle the case where s_op->clear_inode is not been
1178 * called yet, and somebody is calling igrab
1179 * while the inode is getting freed.
1185 EXPORT_SYMBOL(igrab
);
1188 * ilookup5_nowait - search for an inode in the inode cache
1189 * @sb: super block of file system to search
1190 * @hashval: hash value (usually inode number) to search for
1191 * @test: callback used for comparisons between inodes
1192 * @data: opaque data pointer to pass to @test
1194 * Search for the inode specified by @hashval and @data in the inode cache.
1195 * If the inode is in the cache, the inode is returned with an incremented
1198 * Note: I_NEW is not waited upon so you have to be very careful what you do
1199 * with the returned inode. You probably should be using ilookup5() instead.
1201 * Note2: @test is called with the inode_hash_lock held, so can't sleep.
1203 struct inode
*ilookup5_nowait(struct super_block
*sb
, unsigned long hashval
,
1204 int (*test
)(struct inode
*, void *), void *data
)
1206 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1207 struct inode
*inode
;
1209 spin_lock(&inode_hash_lock
);
1210 inode
= find_inode(sb
, head
, test
, data
);
1211 spin_unlock(&inode_hash_lock
);
1215 EXPORT_SYMBOL(ilookup5_nowait
);
1218 * ilookup5 - search for an inode in the inode cache
1219 * @sb: super block of file system to search
1220 * @hashval: hash value (usually inode number) to search for
1221 * @test: callback used for comparisons between inodes
1222 * @data: opaque data pointer to pass to @test
1224 * Search for the inode specified by @hashval and @data in the inode cache,
1225 * and if the inode is in the cache, return the inode with an incremented
1226 * reference count. Waits on I_NEW before returning the inode.
1227 * returned with an incremented reference count.
1229 * This is a generalized version of ilookup() for file systems where the
1230 * inode number is not sufficient for unique identification of an inode.
1232 * Note: @test is called with the inode_hash_lock held, so can't sleep.
1234 struct inode
*ilookup5(struct super_block
*sb
, unsigned long hashval
,
1235 int (*test
)(struct inode
*, void *), void *data
)
1237 struct inode
*inode
= ilookup5_nowait(sb
, hashval
, test
, data
);
1240 wait_on_inode(inode
);
1243 EXPORT_SYMBOL(ilookup5
);
1246 * ilookup - search for an inode in the inode cache
1247 * @sb: super block of file system to search
1248 * @ino: inode number to search for
1250 * Search for the inode @ino in the inode cache, and if the inode is in the
1251 * cache, the inode is returned with an incremented reference count.
1253 struct inode
*ilookup(struct super_block
*sb
, unsigned long ino
)
1255 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1256 struct inode
*inode
;
1258 spin_lock(&inode_hash_lock
);
1259 inode
= find_inode_fast(sb
, head
, ino
);
1260 spin_unlock(&inode_hash_lock
);
1263 wait_on_inode(inode
);
1266 EXPORT_SYMBOL(ilookup
);
1268 int insert_inode_locked(struct inode
*inode
)
1270 struct super_block
*sb
= inode
->i_sb
;
1271 ino_t ino
= inode
->i_ino
;
1272 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1275 struct hlist_node
*node
;
1276 struct inode
*old
= NULL
;
1277 spin_lock(&inode_hash_lock
);
1278 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1279 if (old
->i_ino
!= ino
)
1281 if (old
->i_sb
!= sb
)
1283 spin_lock(&old
->i_lock
);
1284 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
1285 spin_unlock(&old
->i_lock
);
1290 if (likely(!node
)) {
1291 spin_lock(&inode
->i_lock
);
1292 inode
->i_state
|= I_NEW
;
1293 hlist_add_head(&inode
->i_hash
, head
);
1294 spin_unlock(&inode
->i_lock
);
1295 spin_unlock(&inode_hash_lock
);
1299 spin_unlock(&old
->i_lock
);
1300 spin_unlock(&inode_hash_lock
);
1302 if (unlikely(!inode_unhashed(old
))) {
1309 EXPORT_SYMBOL(insert_inode_locked
);
1311 int insert_inode_locked4(struct inode
*inode
, unsigned long hashval
,
1312 int (*test
)(struct inode
*, void *), void *data
)
1314 struct super_block
*sb
= inode
->i_sb
;
1315 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1318 struct hlist_node
*node
;
1319 struct inode
*old
= NULL
;
1321 spin_lock(&inode_hash_lock
);
1322 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1323 if (old
->i_sb
!= sb
)
1325 if (!test(old
, data
))
1327 spin_lock(&old
->i_lock
);
1328 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
1329 spin_unlock(&old
->i_lock
);
1334 if (likely(!node
)) {
1335 spin_lock(&inode
->i_lock
);
1336 inode
->i_state
|= I_NEW
;
1337 hlist_add_head(&inode
->i_hash
, head
);
1338 spin_unlock(&inode
->i_lock
);
1339 spin_unlock(&inode_hash_lock
);
1343 spin_unlock(&old
->i_lock
);
1344 spin_unlock(&inode_hash_lock
);
1346 if (unlikely(!inode_unhashed(old
))) {
1353 EXPORT_SYMBOL(insert_inode_locked4
);
1356 int generic_delete_inode(struct inode
*inode
)
1360 EXPORT_SYMBOL(generic_delete_inode
);
1363 * Called when we're dropping the last reference
1366 * Call the FS "drop_inode()" function, defaulting to
1367 * the legacy UNIX filesystem behaviour. If it tells
1368 * us to evict inode, do so. Otherwise, retain inode
1369 * in cache if fs is alive, sync and evict if fs is
1372 static void iput_final(struct inode
*inode
)
1374 struct super_block
*sb
= inode
->i_sb
;
1375 const struct super_operations
*op
= inode
->i_sb
->s_op
;
1378 WARN_ON(inode
->i_state
& I_NEW
);
1381 drop
= op
->drop_inode(inode
);
1383 drop
= generic_drop_inode(inode
);
1385 if (!drop
&& (sb
->s_flags
& MS_ACTIVE
)) {
1386 inode
->i_state
|= I_REFERENCED
;
1387 if (!(inode
->i_state
& (I_DIRTY
|I_SYNC
)))
1388 inode_lru_list_add(inode
);
1389 spin_unlock(&inode
->i_lock
);
1394 inode
->i_state
|= I_WILL_FREE
;
1395 spin_unlock(&inode
->i_lock
);
1396 write_inode_now(inode
, 1);
1397 spin_lock(&inode
->i_lock
);
1398 WARN_ON(inode
->i_state
& I_NEW
);
1399 inode
->i_state
&= ~I_WILL_FREE
;
1402 inode
->i_state
|= I_FREEING
;
1403 if (!list_empty(&inode
->i_lru
))
1404 inode_lru_list_del(inode
);
1405 spin_unlock(&inode
->i_lock
);
1411 * iput - put an inode
1412 * @inode: inode to put
1414 * Puts an inode, dropping its usage count. If the inode use count hits
1415 * zero, the inode is then freed and may also be destroyed.
1417 * Consequently, iput() can sleep.
1419 void iput(struct inode
*inode
)
1422 BUG_ON(inode
->i_state
& I_CLEAR
);
1424 if (atomic_dec_and_lock(&inode
->i_count
, &inode
->i_lock
))
1428 EXPORT_SYMBOL(iput
);
1431 * bmap - find a block number in a file
1432 * @inode: inode of file
1433 * @block: block to find
1435 * Returns the block number on the device holding the inode that
1436 * is the disk block number for the block of the file requested.
1437 * That is, asked for block 4 of inode 1 the function will return the
1438 * disk block relative to the disk start that holds that block of the
1441 sector_t
bmap(struct inode
*inode
, sector_t block
)
1444 if (inode
->i_mapping
->a_ops
->bmap
)
1445 res
= inode
->i_mapping
->a_ops
->bmap(inode
->i_mapping
, block
);
1448 EXPORT_SYMBOL(bmap
);
1451 * With relative atime, only update atime if the previous atime is
1452 * earlier than either the ctime or mtime or if at least a day has
1453 * passed since the last atime update.
1455 static int relatime_need_update(struct vfsmount
*mnt
, struct inode
*inode
,
1456 struct timespec now
)
1459 if (!(mnt
->mnt_flags
& MNT_RELATIME
))
1462 * Is mtime younger than atime? If yes, update atime:
1464 if (timespec_compare(&inode
->i_mtime
, &inode
->i_atime
) >= 0)
1467 * Is ctime younger than atime? If yes, update atime:
1469 if (timespec_compare(&inode
->i_ctime
, &inode
->i_atime
) >= 0)
1473 * Is the previous atime value older than a day? If yes,
1476 if ((long)(now
.tv_sec
- inode
->i_atime
.tv_sec
) >= 24*60*60)
1479 * Good, we can skip the atime update:
1485 * touch_atime - update the access time
1486 * @mnt: mount the inode is accessed on
1487 * @dentry: dentry accessed
1489 * Update the accessed time on an inode and mark it for writeback.
1490 * This function automatically handles read only file systems and media,
1491 * as well as the "noatime" flag and inode specific "noatime" markers.
1493 void touch_atime(struct path
*path
)
1495 struct vfsmount
*mnt
= path
->mnt
;
1496 struct inode
*inode
= path
->dentry
->d_inode
;
1497 struct timespec now
;
1499 if (inode
->i_flags
& S_NOATIME
)
1501 if (IS_NOATIME(inode
))
1503 if ((inode
->i_sb
->s_flags
& MS_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1506 if (mnt
->mnt_flags
& MNT_NOATIME
)
1508 if ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1511 now
= current_fs_time(inode
->i_sb
);
1513 if (!relatime_need_update(mnt
, inode
, now
))
1516 if (timespec_equal(&inode
->i_atime
, &now
))
1519 if (mnt_want_write(mnt
))
1522 inode
->i_atime
= now
;
1523 mark_inode_dirty_sync(inode
);
1524 mnt_drop_write(mnt
);
1526 EXPORT_SYMBOL(touch_atime
);
1529 * file_update_time - update mtime and ctime time
1530 * @file: file accessed
1532 * Update the mtime and ctime members of an inode and mark the inode
1533 * for writeback. Note that this function is meant exclusively for
1534 * usage in the file write path of filesystems, and filesystems may
1535 * choose to explicitly ignore update via this function with the
1536 * S_NOCMTIME inode flag, e.g. for network filesystem where these
1537 * timestamps are handled by the server.
1540 void file_update_time(struct file
*file
)
1542 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1543 struct timespec now
;
1544 enum { S_MTIME
= 1, S_CTIME
= 2, S_VERSION
= 4 } sync_it
= 0;
1546 /* First try to exhaust all avenues to not sync */
1547 if (IS_NOCMTIME(inode
))
1550 now
= current_fs_time(inode
->i_sb
);
1551 if (!timespec_equal(&inode
->i_mtime
, &now
))
1554 if (!timespec_equal(&inode
->i_ctime
, &now
))
1557 if (IS_I_VERSION(inode
))
1558 sync_it
|= S_VERSION
;
1563 /* Finally allowed to write? Takes lock. */
1564 if (mnt_want_write_file(file
))
1567 /* Only change inode inside the lock region */
1568 if (sync_it
& S_VERSION
)
1569 inode_inc_iversion(inode
);
1570 if (sync_it
& S_CTIME
)
1571 inode
->i_ctime
= now
;
1572 if (sync_it
& S_MTIME
)
1573 inode
->i_mtime
= now
;
1574 mark_inode_dirty_sync(inode
);
1575 mnt_drop_write_file(file
);
1577 EXPORT_SYMBOL(file_update_time
);
1579 int inode_needs_sync(struct inode
*inode
)
1583 if (S_ISDIR(inode
->i_mode
) && IS_DIRSYNC(inode
))
1587 EXPORT_SYMBOL(inode_needs_sync
);
1589 int inode_wait(void *word
)
1594 EXPORT_SYMBOL(inode_wait
);
1597 * If we try to find an inode in the inode hash while it is being
1598 * deleted, we have to wait until the filesystem completes its
1599 * deletion before reporting that it isn't found. This function waits
1600 * until the deletion _might_ have completed. Callers are responsible
1601 * to recheck inode state.
1603 * It doesn't matter if I_NEW is not set initially, a call to
1604 * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
1607 static void __wait_on_freeing_inode(struct inode
*inode
)
1609 wait_queue_head_t
*wq
;
1610 DEFINE_WAIT_BIT(wait
, &inode
->i_state
, __I_NEW
);
1611 wq
= bit_waitqueue(&inode
->i_state
, __I_NEW
);
1612 prepare_to_wait(wq
, &wait
.wait
, TASK_UNINTERRUPTIBLE
);
1613 spin_unlock(&inode
->i_lock
);
1614 spin_unlock(&inode_hash_lock
);
1616 finish_wait(wq
, &wait
.wait
);
1617 spin_lock(&inode_hash_lock
);
1620 static __initdata
unsigned long ihash_entries
;
1621 static int __init
set_ihash_entries(char *str
)
1625 ihash_entries
= simple_strtoul(str
, &str
, 0);
1628 __setup("ihash_entries=", set_ihash_entries
);
1631 * Initialize the waitqueues and inode hash table.
1633 void __init
inode_init_early(void)
1637 /* If hashes are distributed across NUMA nodes, defer
1638 * hash allocation until vmalloc space is available.
1644 alloc_large_system_hash("Inode-cache",
1645 sizeof(struct hlist_head
),
1653 for (loop
= 0; loop
< (1U << i_hash_shift
); loop
++)
1654 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1657 void __init
inode_init(void)
1661 /* inode slab cache */
1662 inode_cachep
= kmem_cache_create("inode_cache",
1663 sizeof(struct inode
),
1665 (SLAB_RECLAIM_ACCOUNT
|SLAB_PANIC
|
1669 /* Hash may have been set up in inode_init_early */
1674 alloc_large_system_hash("Inode-cache",
1675 sizeof(struct hlist_head
),
1683 for (loop
= 0; loop
< (1U << i_hash_shift
); loop
++)
1684 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1687 void init_special_inode(struct inode
*inode
, umode_t mode
, dev_t rdev
)
1689 inode
->i_mode
= mode
;
1690 if (S_ISCHR(mode
)) {
1691 inode
->i_fop
= &def_chr_fops
;
1692 inode
->i_rdev
= rdev
;
1693 } else if (S_ISBLK(mode
)) {
1694 inode
->i_fop
= &def_blk_fops
;
1695 inode
->i_rdev
= rdev
;
1696 } else if (S_ISFIFO(mode
))
1697 inode
->i_fop
= &def_fifo_fops
;
1698 else if (S_ISSOCK(mode
))
1699 inode
->i_fop
= &bad_sock_fops
;
1701 printk(KERN_DEBUG
"init_special_inode: bogus i_mode (%o) for"
1702 " inode %s:%lu\n", mode
, inode
->i_sb
->s_id
,
1705 EXPORT_SYMBOL(init_special_inode
);
1708 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
1710 * @dir: Directory inode
1711 * @mode: mode of the new inode
1713 void inode_init_owner(struct inode
*inode
, const struct inode
*dir
,
1716 inode
->i_uid
= current_fsuid();
1717 if (dir
&& dir
->i_mode
& S_ISGID
) {
1718 inode
->i_gid
= dir
->i_gid
;
1722 inode
->i_gid
= current_fsgid();
1723 inode
->i_mode
= mode
;
1725 EXPORT_SYMBOL(inode_init_owner
);
1728 * inode_owner_or_capable - check current task permissions to inode
1729 * @inode: inode being checked
1731 * Return true if current either has CAP_FOWNER to the inode, or
1734 bool inode_owner_or_capable(const struct inode
*inode
)
1736 struct user_namespace
*ns
= inode_userns(inode
);
1738 if (current_user_ns() == ns
&& current_fsuid() == inode
->i_uid
)
1740 if (ns_capable(ns
, CAP_FOWNER
))
1744 EXPORT_SYMBOL(inode_owner_or_capable
);