4 * (C) 1997 Linus Torvalds
9 #include <linux/dcache.h>
10 #include <linux/init.h>
11 #include <linux/slab.h>
12 #include <linux/writeback.h>
13 #include <linux/module.h>
14 #include <linux/backing-dev.h>
15 #include <linux/wait.h>
16 #include <linux/rwsem.h>
17 #include <linux/hash.h>
18 #include <linux/swap.h>
19 #include <linux/security.h>
20 #include <linux/pagemap.h>
21 #include <linux/cdev.h>
22 #include <linux/bootmem.h>
23 #include <linux/fsnotify.h>
24 #include <linux/mount.h>
25 #include <linux/async.h>
26 #include <linux/posix_acl.h>
27 #include <linux/ima.h>
30 * This is needed for the following functions:
34 * FIXME: remove all knowledge of the buffer layer from this file
36 #include <linux/buffer_head.h>
39 * New inode.c implementation.
41 * This implementation has the basic premise of trying
42 * to be extremely low-overhead and SMP-safe, yet be
43 * simple enough to be "obviously correct".
48 /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */
50 /* #define INODE_PARANOIA 1 */
51 /* #define INODE_DEBUG 1 */
54 * Inode lookup is no longer as critical as it used to be:
55 * most of the lookups are going to be through the dcache.
57 #define I_HASHBITS i_hash_shift
58 #define I_HASHMASK i_hash_mask
60 static unsigned int i_hash_mask __read_mostly
;
61 static unsigned int i_hash_shift __read_mostly
;
64 * Each inode can be on two separate lists. One is
65 * the hash list of the inode, used for lookups. The
66 * other linked list is the "type" list:
67 * "in_use" - valid inode, i_count > 0, i_nlink > 0
68 * "dirty" - as "in_use" but also dirty
69 * "unused" - valid inode, i_count = 0
71 * A "dirty" list is maintained for each super block,
72 * allowing for low-overhead inode sync() operations.
75 static LIST_HEAD(inode_lru
);
76 static struct hlist_head
*inode_hashtable __read_mostly
;
79 * A simple spinlock to protect the list manipulations.
81 * NOTE! You also have to own the lock if you change
82 * the i_state of an inode while it is in use..
84 DEFINE_SPINLOCK(inode_lock
);
87 * iprune_sem provides exclusion between the icache shrinking and the
90 * We don't actually need it to protect anything in the umount path,
91 * but only need to cycle through it to make sure any inode that
92 * prune_icache took off the LRU list has been fully torn down by the
93 * time we are past evict_inodes.
95 static DECLARE_RWSEM(iprune_sem
);
98 * Statistics gathering..
100 struct inodes_stat_t inodes_stat
;
102 static DEFINE_PER_CPU(unsigned int, nr_inodes
);
104 static struct kmem_cache
*inode_cachep __read_mostly
;
106 static int get_nr_inodes(void)
110 for_each_possible_cpu(i
)
111 sum
+= per_cpu(nr_inodes
, i
);
112 return sum
< 0 ? 0 : sum
;
115 static inline int get_nr_inodes_unused(void)
117 return inodes_stat
.nr_unused
;
120 int get_nr_dirty_inodes(void)
122 /* not actually dirty inodes, but a wild approximation */
123 int nr_dirty
= get_nr_inodes() - get_nr_inodes_unused();
124 return nr_dirty
> 0 ? nr_dirty
: 0;
128 * Handle nr_inode sysctl
131 int proc_nr_inodes(ctl_table
*table
, int write
,
132 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
134 inodes_stat
.nr_inodes
= get_nr_inodes();
135 return proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
139 static void wake_up_inode(struct inode
*inode
)
142 * Prevent speculative execution through spin_unlock(&inode_lock);
145 wake_up_bit(&inode
->i_state
, __I_NEW
);
149 * inode_init_always - perform inode structure intialisation
150 * @sb: superblock inode belongs to
151 * @inode: inode to initialise
153 * These are initializations that need to be done on every inode
154 * allocation as the fields are not initialised by slab allocation.
156 int inode_init_always(struct super_block
*sb
, struct inode
*inode
)
158 static const struct address_space_operations empty_aops
;
159 static const struct inode_operations empty_iops
;
160 static const struct file_operations empty_fops
;
161 struct address_space
*const mapping
= &inode
->i_data
;
164 inode
->i_blkbits
= sb
->s_blocksize_bits
;
166 atomic_set(&inode
->i_count
, 1);
167 inode
->i_op
= &empty_iops
;
168 inode
->i_fop
= &empty_fops
;
172 atomic_set(&inode
->i_writecount
, 0);
176 inode
->i_generation
= 0;
178 memset(&inode
->i_dquot
, 0, sizeof(inode
->i_dquot
));
180 inode
->i_pipe
= NULL
;
181 inode
->i_bdev
= NULL
;
182 inode
->i_cdev
= NULL
;
184 inode
->dirtied_when
= 0;
186 if (security_inode_alloc(inode
))
188 spin_lock_init(&inode
->i_lock
);
189 lockdep_set_class(&inode
->i_lock
, &sb
->s_type
->i_lock_key
);
191 mutex_init(&inode
->i_mutex
);
192 lockdep_set_class(&inode
->i_mutex
, &sb
->s_type
->i_mutex_key
);
194 init_rwsem(&inode
->i_alloc_sem
);
195 lockdep_set_class(&inode
->i_alloc_sem
, &sb
->s_type
->i_alloc_sem_key
);
197 mapping
->a_ops
= &empty_aops
;
198 mapping
->host
= inode
;
200 mapping_set_gfp_mask(mapping
, GFP_HIGHUSER_MOVABLE
);
201 mapping
->assoc_mapping
= NULL
;
202 mapping
->backing_dev_info
= &default_backing_dev_info
;
203 mapping
->writeback_index
= 0;
206 * If the block_device provides a backing_dev_info for client
207 * inodes then use that. Otherwise the inode share the bdev's
211 struct backing_dev_info
*bdi
;
213 bdi
= sb
->s_bdev
->bd_inode
->i_mapping
->backing_dev_info
;
214 mapping
->backing_dev_info
= bdi
;
216 inode
->i_private
= NULL
;
217 inode
->i_mapping
= mapping
;
218 #ifdef CONFIG_FS_POSIX_ACL
219 inode
->i_acl
= inode
->i_default_acl
= ACL_NOT_CACHED
;
222 #ifdef CONFIG_FSNOTIFY
223 inode
->i_fsnotify_mask
= 0;
226 this_cpu_inc(nr_inodes
);
232 EXPORT_SYMBOL(inode_init_always
);
234 static struct inode
*alloc_inode(struct super_block
*sb
)
238 if (sb
->s_op
->alloc_inode
)
239 inode
= sb
->s_op
->alloc_inode(sb
);
241 inode
= kmem_cache_alloc(inode_cachep
, GFP_KERNEL
);
246 if (unlikely(inode_init_always(sb
, inode
))) {
247 if (inode
->i_sb
->s_op
->destroy_inode
)
248 inode
->i_sb
->s_op
->destroy_inode(inode
);
250 kmem_cache_free(inode_cachep
, inode
);
257 void free_inode_nonrcu(struct inode
*inode
)
259 kmem_cache_free(inode_cachep
, inode
);
261 EXPORT_SYMBOL(free_inode_nonrcu
);
263 void __destroy_inode(struct inode
*inode
)
265 BUG_ON(inode_has_buffers(inode
));
266 security_inode_free(inode
);
267 fsnotify_inode_delete(inode
);
268 #ifdef CONFIG_FS_POSIX_ACL
269 if (inode
->i_acl
&& inode
->i_acl
!= ACL_NOT_CACHED
)
270 posix_acl_release(inode
->i_acl
);
271 if (inode
->i_default_acl
&& inode
->i_default_acl
!= ACL_NOT_CACHED
)
272 posix_acl_release(inode
->i_default_acl
);
274 this_cpu_dec(nr_inodes
);
276 EXPORT_SYMBOL(__destroy_inode
);
278 static void i_callback(struct rcu_head
*head
)
280 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
281 INIT_LIST_HEAD(&inode
->i_dentry
);
282 kmem_cache_free(inode_cachep
, inode
);
285 static void destroy_inode(struct inode
*inode
)
287 BUG_ON(!list_empty(&inode
->i_lru
));
288 __destroy_inode(inode
);
289 if (inode
->i_sb
->s_op
->destroy_inode
)
290 inode
->i_sb
->s_op
->destroy_inode(inode
);
292 call_rcu(&inode
->i_rcu
, i_callback
);
295 void address_space_init_once(struct address_space
*mapping
)
297 memset(mapping
, 0, sizeof(*mapping
));
298 INIT_RADIX_TREE(&mapping
->page_tree
, GFP_ATOMIC
);
299 spin_lock_init(&mapping
->tree_lock
);
300 spin_lock_init(&mapping
->i_mmap_lock
);
301 INIT_LIST_HEAD(&mapping
->private_list
);
302 spin_lock_init(&mapping
->private_lock
);
303 INIT_RAW_PRIO_TREE_ROOT(&mapping
->i_mmap
);
304 INIT_LIST_HEAD(&mapping
->i_mmap_nonlinear
);
305 mutex_init(&mapping
->unmap_mutex
);
307 EXPORT_SYMBOL(address_space_init_once
);
310 * These are initializations that only need to be done
311 * once, because the fields are idempotent across use
312 * of the inode, so let the slab aware of that.
314 void inode_init_once(struct inode
*inode
)
316 memset(inode
, 0, sizeof(*inode
));
317 INIT_HLIST_NODE(&inode
->i_hash
);
318 INIT_LIST_HEAD(&inode
->i_dentry
);
319 INIT_LIST_HEAD(&inode
->i_devices
);
320 INIT_LIST_HEAD(&inode
->i_wb_list
);
321 INIT_LIST_HEAD(&inode
->i_lru
);
322 address_space_init_once(&inode
->i_data
);
323 i_size_ordered_init(inode
);
324 #ifdef CONFIG_FSNOTIFY
325 INIT_HLIST_HEAD(&inode
->i_fsnotify_marks
);
328 EXPORT_SYMBOL(inode_init_once
);
330 static void init_once(void *foo
)
332 struct inode
*inode
= (struct inode
*) foo
;
334 inode_init_once(inode
);
338 * inode_lock must be held
340 void __iget(struct inode
*inode
)
342 atomic_inc(&inode
->i_count
);
346 * get additional reference to inode; caller must already hold one.
348 void ihold(struct inode
*inode
)
350 WARN_ON(atomic_inc_return(&inode
->i_count
) < 2);
352 EXPORT_SYMBOL(ihold
);
354 static void inode_lru_list_add(struct inode
*inode
)
356 if (list_empty(&inode
->i_lru
)) {
357 list_add(&inode
->i_lru
, &inode_lru
);
358 inodes_stat
.nr_unused
++;
362 static void inode_lru_list_del(struct inode
*inode
)
364 if (!list_empty(&inode
->i_lru
)) {
365 list_del_init(&inode
->i_lru
);
366 inodes_stat
.nr_unused
--;
370 static inline void __inode_sb_list_add(struct inode
*inode
)
372 list_add(&inode
->i_sb_list
, &inode
->i_sb
->s_inodes
);
376 * inode_sb_list_add - add inode to the superblock list of inodes
377 * @inode: inode to add
379 void inode_sb_list_add(struct inode
*inode
)
381 spin_lock(&inode_lock
);
382 __inode_sb_list_add(inode
);
383 spin_unlock(&inode_lock
);
385 EXPORT_SYMBOL_GPL(inode_sb_list_add
);
387 static inline void __inode_sb_list_del(struct inode
*inode
)
389 list_del_init(&inode
->i_sb_list
);
392 static unsigned long hash(struct super_block
*sb
, unsigned long hashval
)
396 tmp
= (hashval
* (unsigned long)sb
) ^ (GOLDEN_RATIO_PRIME
+ hashval
) /
398 tmp
= tmp
^ ((tmp
^ GOLDEN_RATIO_PRIME
) >> I_HASHBITS
);
399 return tmp
& I_HASHMASK
;
403 * __insert_inode_hash - hash an inode
404 * @inode: unhashed inode
405 * @hashval: unsigned long value used to locate this object in the
408 * Add an inode to the inode hash for this superblock.
410 void __insert_inode_hash(struct inode
*inode
, unsigned long hashval
)
412 struct hlist_head
*b
= inode_hashtable
+ hash(inode
->i_sb
, hashval
);
414 spin_lock(&inode_lock
);
415 hlist_add_head(&inode
->i_hash
, b
);
416 spin_unlock(&inode_lock
);
418 EXPORT_SYMBOL(__insert_inode_hash
);
421 * __remove_inode_hash - remove an inode from the hash
422 * @inode: inode to unhash
424 * Remove an inode from the superblock.
426 static void __remove_inode_hash(struct inode
*inode
)
428 hlist_del_init(&inode
->i_hash
);
432 * remove_inode_hash - remove an inode from the hash
433 * @inode: inode to unhash
435 * Remove an inode from the superblock.
437 void remove_inode_hash(struct inode
*inode
)
439 spin_lock(&inode_lock
);
440 hlist_del_init(&inode
->i_hash
);
441 spin_unlock(&inode_lock
);
443 EXPORT_SYMBOL(remove_inode_hash
);
445 void end_writeback(struct inode
*inode
)
448 BUG_ON(inode
->i_data
.nrpages
);
449 BUG_ON(!list_empty(&inode
->i_data
.private_list
));
450 BUG_ON(!(inode
->i_state
& I_FREEING
));
451 BUG_ON(inode
->i_state
& I_CLEAR
);
452 inode_sync_wait(inode
);
453 /* don't need i_lock here, no concurrent mods to i_state */
454 inode
->i_state
= I_FREEING
| I_CLEAR
;
456 EXPORT_SYMBOL(end_writeback
);
458 static void evict(struct inode
*inode
)
460 const struct super_operations
*op
= inode
->i_sb
->s_op
;
462 if (op
->evict_inode
) {
463 op
->evict_inode(inode
);
465 if (inode
->i_data
.nrpages
)
466 truncate_inode_pages(&inode
->i_data
, 0);
467 end_writeback(inode
);
469 if (S_ISBLK(inode
->i_mode
) && inode
->i_bdev
)
471 if (S_ISCHR(inode
->i_mode
) && inode
->i_cdev
)
476 * dispose_list - dispose of the contents of a local list
477 * @head: the head of the list to free
479 * Dispose-list gets a local list with local inodes in it, so it doesn't
480 * need to worry about list corruption and SMP locks.
482 static void dispose_list(struct list_head
*head
)
484 while (!list_empty(head
)) {
487 inode
= list_first_entry(head
, struct inode
, i_lru
);
488 list_del_init(&inode
->i_lru
);
492 spin_lock(&inode_lock
);
493 __remove_inode_hash(inode
);
494 __inode_sb_list_del(inode
);
495 spin_unlock(&inode_lock
);
497 wake_up_inode(inode
);
498 destroy_inode(inode
);
503 * evict_inodes - evict all evictable inodes for a superblock
504 * @sb: superblock to operate on
506 * Make sure that no inodes with zero refcount are retained. This is
507 * called by superblock shutdown after having MS_ACTIVE flag removed,
508 * so any inode reaching zero refcount during or after that call will
509 * be immediately evicted.
511 void evict_inodes(struct super_block
*sb
)
513 struct inode
*inode
, *next
;
516 spin_lock(&inode_lock
);
517 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
518 if (atomic_read(&inode
->i_count
))
520 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
))
523 inode
->i_state
|= I_FREEING
;
526 * Move the inode off the IO lists and LRU once I_FREEING is
527 * set so that it won't get moved back on there if it is dirty.
529 list_move(&inode
->i_lru
, &dispose
);
530 list_del_init(&inode
->i_wb_list
);
531 if (!(inode
->i_state
& (I_DIRTY
| I_SYNC
)))
532 inodes_stat
.nr_unused
--;
534 spin_unlock(&inode_lock
);
536 dispose_list(&dispose
);
539 * Cycle through iprune_sem to make sure any inode that prune_icache
540 * moved off the list before we took the lock has been fully torn
543 down_write(&iprune_sem
);
544 up_write(&iprune_sem
);
548 * invalidate_inodes - attempt to free all inodes on a superblock
549 * @sb: superblock to operate on
550 * @kill_dirty: flag to guide handling of dirty inodes
552 * Attempts to free all inodes for a given superblock. If there were any
553 * busy inodes return a non-zero value, else zero.
554 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
557 int invalidate_inodes(struct super_block
*sb
, bool kill_dirty
)
560 struct inode
*inode
, *next
;
563 spin_lock(&inode_lock
);
564 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
565 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
))
567 if (inode
->i_state
& I_DIRTY
&& !kill_dirty
) {
571 if (atomic_read(&inode
->i_count
)) {
576 inode
->i_state
|= I_FREEING
;
579 * Move the inode off the IO lists and LRU once I_FREEING is
580 * set so that it won't get moved back on there if it is dirty.
582 list_move(&inode
->i_lru
, &dispose
);
583 list_del_init(&inode
->i_wb_list
);
584 if (!(inode
->i_state
& (I_DIRTY
| I_SYNC
)))
585 inodes_stat
.nr_unused
--;
587 spin_unlock(&inode_lock
);
589 dispose_list(&dispose
);
594 static int can_unuse(struct inode
*inode
)
596 if (inode
->i_state
& ~I_REFERENCED
)
598 if (inode_has_buffers(inode
))
600 if (atomic_read(&inode
->i_count
))
602 if (inode
->i_data
.nrpages
)
608 * Scan `goal' inodes on the unused list for freeable ones. They are moved to a
609 * temporary list and then are freed outside inode_lock by dispose_list().
611 * Any inodes which are pinned purely because of attached pagecache have their
612 * pagecache removed. If the inode has metadata buffers attached to
613 * mapping->private_list then try to remove them.
615 * If the inode has the I_REFERENCED flag set, then it means that it has been
616 * used recently - the flag is set in iput_final(). When we encounter such an
617 * inode, clear the flag and move it to the back of the LRU so it gets another
618 * pass through the LRU before it gets reclaimed. This is necessary because of
619 * the fact we are doing lazy LRU updates to minimise lock contention so the
620 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
621 * with this flag set because they are the inodes that are out of order.
623 static void prune_icache(int nr_to_scan
)
627 unsigned long reap
= 0;
629 down_read(&iprune_sem
);
630 spin_lock(&inode_lock
);
631 for (nr_scanned
= 0; nr_scanned
< nr_to_scan
; nr_scanned
++) {
634 if (list_empty(&inode_lru
))
637 inode
= list_entry(inode_lru
.prev
, struct inode
, i_lru
);
640 * Referenced or dirty inodes are still in use. Give them
641 * another pass through the LRU as we canot reclaim them now.
643 if (atomic_read(&inode
->i_count
) ||
644 (inode
->i_state
& ~I_REFERENCED
)) {
645 list_del_init(&inode
->i_lru
);
646 inodes_stat
.nr_unused
--;
650 /* recently referenced inodes get one more pass */
651 if (inode
->i_state
& I_REFERENCED
) {
652 list_move(&inode
->i_lru
, &inode_lru
);
653 inode
->i_state
&= ~I_REFERENCED
;
656 if (inode_has_buffers(inode
) || inode
->i_data
.nrpages
) {
658 spin_unlock(&inode_lock
);
659 if (remove_inode_buffers(inode
))
660 reap
+= invalidate_mapping_pages(&inode
->i_data
,
663 spin_lock(&inode_lock
);
665 if (inode
!= list_entry(inode_lru
.next
,
666 struct inode
, i_lru
))
667 continue; /* wrong inode or list_empty */
668 if (!can_unuse(inode
))
671 WARN_ON(inode
->i_state
& I_NEW
);
672 inode
->i_state
|= I_FREEING
;
675 * Move the inode off the IO lists and LRU once I_FREEING is
676 * set so that it won't get moved back on there if it is dirty.
678 list_move(&inode
->i_lru
, &freeable
);
679 list_del_init(&inode
->i_wb_list
);
680 inodes_stat
.nr_unused
--;
682 if (current_is_kswapd())
683 __count_vm_events(KSWAPD_INODESTEAL
, reap
);
685 __count_vm_events(PGINODESTEAL
, reap
);
686 spin_unlock(&inode_lock
);
688 dispose_list(&freeable
);
689 up_read(&iprune_sem
);
693 * shrink_icache_memory() will attempt to reclaim some unused inodes. Here,
694 * "unused" means that no dentries are referring to the inodes: the files are
695 * not open and the dcache references to those inodes have already been
698 * This function is passed the number of inodes to scan, and it returns the
699 * total number of remaining possibly-reclaimable inodes.
701 static int shrink_icache_memory(struct shrinker
*shrink
, int nr
, gfp_t gfp_mask
)
705 * Nasty deadlock avoidance. We may hold various FS locks,
706 * and we don't want to recurse into the FS that called us
707 * in clear_inode() and friends..
709 if (!(gfp_mask
& __GFP_FS
))
713 return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure
;
716 static struct shrinker icache_shrinker
= {
717 .shrink
= shrink_icache_memory
,
718 .seeks
= DEFAULT_SEEKS
,
721 static void __wait_on_freeing_inode(struct inode
*inode
);
723 * Called with the inode lock held.
725 static struct inode
*find_inode(struct super_block
*sb
,
726 struct hlist_head
*head
,
727 int (*test
)(struct inode
*, void *),
730 struct hlist_node
*node
;
731 struct inode
*inode
= NULL
;
734 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
735 if (inode
->i_sb
!= sb
)
737 if (!test(inode
, data
))
739 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
740 __wait_on_freeing_inode(inode
);
750 * find_inode_fast is the fast path version of find_inode, see the comment at
751 * iget_locked for details.
753 static struct inode
*find_inode_fast(struct super_block
*sb
,
754 struct hlist_head
*head
, unsigned long ino
)
756 struct hlist_node
*node
;
757 struct inode
*inode
= NULL
;
760 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
761 if (inode
->i_ino
!= ino
)
763 if (inode
->i_sb
!= sb
)
765 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
766 __wait_on_freeing_inode(inode
);
776 * Each cpu owns a range of LAST_INO_BATCH numbers.
777 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
778 * to renew the exhausted range.
780 * This does not significantly increase overflow rate because every CPU can
781 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
782 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
783 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
784 * overflow rate by 2x, which does not seem too significant.
786 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
787 * error if st_ino won't fit in target struct field. Use 32bit counter
788 * here to attempt to avoid that.
790 #define LAST_INO_BATCH 1024
791 static DEFINE_PER_CPU(unsigned int, last_ino
);
793 unsigned int get_next_ino(void)
795 unsigned int *p
= &get_cpu_var(last_ino
);
796 unsigned int res
= *p
;
799 if (unlikely((res
& (LAST_INO_BATCH
-1)) == 0)) {
800 static atomic_t shared_last_ino
;
801 int next
= atomic_add_return(LAST_INO_BATCH
, &shared_last_ino
);
803 res
= next
- LAST_INO_BATCH
;
808 put_cpu_var(last_ino
);
811 EXPORT_SYMBOL(get_next_ino
);
814 * new_inode - obtain an inode
817 * Allocates a new inode for given superblock. The default gfp_mask
818 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
819 * If HIGHMEM pages are unsuitable or it is known that pages allocated
820 * for the page cache are not reclaimable or migratable,
821 * mapping_set_gfp_mask() must be called with suitable flags on the
822 * newly created inode's mapping
825 struct inode
*new_inode(struct super_block
*sb
)
829 spin_lock_prefetch(&inode_lock
);
831 inode
= alloc_inode(sb
);
833 spin_lock(&inode_lock
);
834 __inode_sb_list_add(inode
);
836 spin_unlock(&inode_lock
);
840 EXPORT_SYMBOL(new_inode
);
842 void unlock_new_inode(struct inode
*inode
)
844 #ifdef CONFIG_DEBUG_LOCK_ALLOC
845 if (S_ISDIR(inode
->i_mode
)) {
846 struct file_system_type
*type
= inode
->i_sb
->s_type
;
848 /* Set new key only if filesystem hasn't already changed it */
849 if (!lockdep_match_class(&inode
->i_mutex
,
850 &type
->i_mutex_key
)) {
852 * ensure nobody is actually holding i_mutex
854 mutex_destroy(&inode
->i_mutex
);
855 mutex_init(&inode
->i_mutex
);
856 lockdep_set_class(&inode
->i_mutex
,
857 &type
->i_mutex_dir_key
);
862 * This is special! We do not need the spinlock when clearing I_NEW,
863 * because we're guaranteed that nobody else tries to do anything about
864 * the state of the inode when it is locked, as we just created it (so
865 * there can be no old holders that haven't tested I_NEW).
866 * However we must emit the memory barrier so that other CPUs reliably
867 * see the clearing of I_NEW after the other inode initialisation has
871 WARN_ON(!(inode
->i_state
& I_NEW
));
872 inode
->i_state
&= ~I_NEW
;
873 wake_up_inode(inode
);
875 EXPORT_SYMBOL(unlock_new_inode
);
878 * This is called without the inode lock held.. Be careful.
880 * We no longer cache the sb_flags in i_flags - see fs.h
881 * -- rmk@arm.uk.linux.org
883 static struct inode
*get_new_inode(struct super_block
*sb
,
884 struct hlist_head
*head
,
885 int (*test
)(struct inode
*, void *),
886 int (*set
)(struct inode
*, void *),
891 inode
= alloc_inode(sb
);
895 spin_lock(&inode_lock
);
896 /* We released the lock, so.. */
897 old
= find_inode(sb
, head
, test
, data
);
899 if (set(inode
, data
))
902 hlist_add_head(&inode
->i_hash
, head
);
903 __inode_sb_list_add(inode
);
904 inode
->i_state
= I_NEW
;
905 spin_unlock(&inode_lock
);
907 /* Return the locked inode with I_NEW set, the
908 * caller is responsible for filling in the contents
914 * Uhhuh, somebody else created the same inode under
915 * us. Use the old inode instead of the one we just
918 spin_unlock(&inode_lock
);
919 destroy_inode(inode
);
921 wait_on_inode(inode
);
926 spin_unlock(&inode_lock
);
927 destroy_inode(inode
);
932 * get_new_inode_fast is the fast path version of get_new_inode, see the
933 * comment at iget_locked for details.
935 static struct inode
*get_new_inode_fast(struct super_block
*sb
,
936 struct hlist_head
*head
, unsigned long ino
)
940 inode
= alloc_inode(sb
);
944 spin_lock(&inode_lock
);
945 /* We released the lock, so.. */
946 old
= find_inode_fast(sb
, head
, ino
);
949 hlist_add_head(&inode
->i_hash
, head
);
950 __inode_sb_list_add(inode
);
951 inode
->i_state
= I_NEW
;
952 spin_unlock(&inode_lock
);
954 /* Return the locked inode with I_NEW set, the
955 * caller is responsible for filling in the contents
961 * Uhhuh, somebody else created the same inode under
962 * us. Use the old inode instead of the one we just
965 spin_unlock(&inode_lock
);
966 destroy_inode(inode
);
968 wait_on_inode(inode
);
974 * search the inode cache for a matching inode number.
975 * If we find one, then the inode number we are trying to
976 * allocate is not unique and so we should not use it.
978 * Returns 1 if the inode number is unique, 0 if it is not.
980 static int test_inode_iunique(struct super_block
*sb
, unsigned long ino
)
982 struct hlist_head
*b
= inode_hashtable
+ hash(sb
, ino
);
983 struct hlist_node
*node
;
986 hlist_for_each_entry(inode
, node
, b
, i_hash
) {
987 if (inode
->i_ino
== ino
&& inode
->i_sb
== sb
)
995 * iunique - get a unique inode number
997 * @max_reserved: highest reserved inode number
999 * Obtain an inode number that is unique on the system for a given
1000 * superblock. This is used by file systems that have no natural
1001 * permanent inode numbering system. An inode number is returned that
1002 * is higher than the reserved limit but unique.
1005 * With a large number of inodes live on the file system this function
1006 * currently becomes quite slow.
1008 ino_t
iunique(struct super_block
*sb
, ino_t max_reserved
)
1011 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1012 * error if st_ino won't fit in target struct field. Use 32bit counter
1013 * here to attempt to avoid that.
1015 static DEFINE_SPINLOCK(iunique_lock
);
1016 static unsigned int counter
;
1019 spin_lock(&inode_lock
);
1020 spin_lock(&iunique_lock
);
1022 if (counter
<= max_reserved
)
1023 counter
= max_reserved
+ 1;
1025 } while (!test_inode_iunique(sb
, res
));
1026 spin_unlock(&iunique_lock
);
1027 spin_unlock(&inode_lock
);
1031 EXPORT_SYMBOL(iunique
);
1033 struct inode
*igrab(struct inode
*inode
)
1035 spin_lock(&inode_lock
);
1036 if (!(inode
->i_state
& (I_FREEING
|I_WILL_FREE
)))
1040 * Handle the case where s_op->clear_inode is not been
1041 * called yet, and somebody is calling igrab
1042 * while the inode is getting freed.
1045 spin_unlock(&inode_lock
);
1048 EXPORT_SYMBOL(igrab
);
1051 * ifind - internal function, you want ilookup5() or iget5().
1052 * @sb: super block of file system to search
1053 * @head: the head of the list to search
1054 * @test: callback used for comparisons between inodes
1055 * @data: opaque data pointer to pass to @test
1056 * @wait: if true wait for the inode to be unlocked, if false do not
1058 * ifind() searches for the inode specified by @data in the inode
1059 * cache. This is a generalized version of ifind_fast() for file systems where
1060 * the inode number is not sufficient for unique identification of an inode.
1062 * If the inode is in the cache, the inode is returned with an incremented
1065 * Otherwise NULL is returned.
1067 * Note, @test is called with the inode_lock held, so can't sleep.
1069 static struct inode
*ifind(struct super_block
*sb
,
1070 struct hlist_head
*head
, int (*test
)(struct inode
*, void *),
1071 void *data
, const int wait
)
1073 struct inode
*inode
;
1075 spin_lock(&inode_lock
);
1076 inode
= find_inode(sb
, head
, test
, data
);
1078 spin_unlock(&inode_lock
);
1080 wait_on_inode(inode
);
1083 spin_unlock(&inode_lock
);
1088 * ifind_fast - internal function, you want ilookup() or iget().
1089 * @sb: super block of file system to search
1090 * @head: head of the list to search
1091 * @ino: inode number to search for
1093 * ifind_fast() searches for the inode @ino in the inode cache. This is for
1094 * file systems where the inode number is sufficient for unique identification
1097 * If the inode is in the cache, the inode is returned with an incremented
1100 * Otherwise NULL is returned.
1102 static struct inode
*ifind_fast(struct super_block
*sb
,
1103 struct hlist_head
*head
, unsigned long ino
)
1105 struct inode
*inode
;
1107 spin_lock(&inode_lock
);
1108 inode
= find_inode_fast(sb
, head
, ino
);
1110 spin_unlock(&inode_lock
);
1111 wait_on_inode(inode
);
1114 spin_unlock(&inode_lock
);
1119 * ilookup5_nowait - search for an inode in the inode cache
1120 * @sb: super block of file system to search
1121 * @hashval: hash value (usually inode number) to search for
1122 * @test: callback used for comparisons between inodes
1123 * @data: opaque data pointer to pass to @test
1125 * ilookup5() uses ifind() to search for the inode specified by @hashval and
1126 * @data in the inode cache. This is a generalized version of ilookup() for
1127 * file systems where the inode number is not sufficient for unique
1128 * identification of an inode.
1130 * If the inode is in the cache, the inode is returned with an incremented
1131 * reference count. Note, the inode lock is not waited upon so you have to be
1132 * very careful what you do with the returned inode. You probably should be
1133 * using ilookup5() instead.
1135 * Otherwise NULL is returned.
1137 * Note, @test is called with the inode_lock held, so can't sleep.
1139 struct inode
*ilookup5_nowait(struct super_block
*sb
, unsigned long hashval
,
1140 int (*test
)(struct inode
*, void *), void *data
)
1142 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1144 return ifind(sb
, head
, test
, data
, 0);
1146 EXPORT_SYMBOL(ilookup5_nowait
);
1149 * ilookup5 - search for an inode in the inode cache
1150 * @sb: super block of file system to search
1151 * @hashval: hash value (usually inode number) to search for
1152 * @test: callback used for comparisons between inodes
1153 * @data: opaque data pointer to pass to @test
1155 * ilookup5() uses ifind() to search for the inode specified by @hashval and
1156 * @data in the inode cache. This is a generalized version of ilookup() for
1157 * file systems where the inode number is not sufficient for unique
1158 * identification of an inode.
1160 * If the inode is in the cache, the inode lock is waited upon and the inode is
1161 * returned with an incremented reference count.
1163 * Otherwise NULL is returned.
1165 * Note, @test is called with the inode_lock held, so can't sleep.
1167 struct inode
*ilookup5(struct super_block
*sb
, unsigned long hashval
,
1168 int (*test
)(struct inode
*, void *), void *data
)
1170 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1172 return ifind(sb
, head
, test
, data
, 1);
1174 EXPORT_SYMBOL(ilookup5
);
1177 * ilookup - search for an inode in the inode cache
1178 * @sb: super block of file system to search
1179 * @ino: inode number to search for
1181 * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache.
1182 * This is for file systems where the inode number is sufficient for unique
1183 * identification of an inode.
1185 * If the inode is in the cache, the inode is returned with an incremented
1188 * Otherwise NULL is returned.
1190 struct inode
*ilookup(struct super_block
*sb
, unsigned long ino
)
1192 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1194 return ifind_fast(sb
, head
, ino
);
1196 EXPORT_SYMBOL(ilookup
);
1199 * iget5_locked - obtain an inode from a mounted file system
1200 * @sb: super block of file system
1201 * @hashval: hash value (usually inode number) to get
1202 * @test: callback used for comparisons between inodes
1203 * @set: callback used to initialize a new struct inode
1204 * @data: opaque data pointer to pass to @test and @set
1206 * iget5_locked() uses ifind() to search for the inode specified by @hashval
1207 * and @data in the inode cache and if present it is returned with an increased
1208 * reference count. This is a generalized version of iget_locked() for file
1209 * systems where the inode number is not sufficient for unique identification
1212 * If the inode is not in cache, get_new_inode() is called to allocate a new
1213 * inode and this is returned locked, hashed, and with the I_NEW flag set. The
1214 * file system gets to fill it in before unlocking it via unlock_new_inode().
1216 * Note both @test and @set are called with the inode_lock held, so can't sleep.
1218 struct inode
*iget5_locked(struct super_block
*sb
, unsigned long hashval
,
1219 int (*test
)(struct inode
*, void *),
1220 int (*set
)(struct inode
*, void *), void *data
)
1222 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1223 struct inode
*inode
;
1225 inode
= ifind(sb
, head
, test
, data
, 1);
1229 * get_new_inode() will do the right thing, re-trying the search
1230 * in case it had to block at any point.
1232 return get_new_inode(sb
, head
, test
, set
, data
);
1234 EXPORT_SYMBOL(iget5_locked
);
1237 * iget_locked - obtain an inode from a mounted file system
1238 * @sb: super block of file system
1239 * @ino: inode number to get
1241 * iget_locked() uses ifind_fast() to search for the inode specified by @ino in
1242 * the inode cache and if present it is returned with an increased reference
1243 * count. This is for file systems where the inode number is sufficient for
1244 * unique identification of an inode.
1246 * If the inode is not in cache, get_new_inode_fast() is called to allocate a
1247 * new inode and this is returned locked, hashed, and with the I_NEW flag set.
1248 * The file system gets to fill it in before unlocking it via
1249 * unlock_new_inode().
1251 struct inode
*iget_locked(struct super_block
*sb
, unsigned long ino
)
1253 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1254 struct inode
*inode
;
1256 inode
= ifind_fast(sb
, head
, ino
);
1260 * get_new_inode_fast() will do the right thing, re-trying the search
1261 * in case it had to block at any point.
1263 return get_new_inode_fast(sb
, head
, ino
);
1265 EXPORT_SYMBOL(iget_locked
);
1267 int insert_inode_locked(struct inode
*inode
)
1269 struct super_block
*sb
= inode
->i_sb
;
1270 ino_t ino
= inode
->i_ino
;
1271 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1273 inode
->i_state
|= I_NEW
;
1275 struct hlist_node
*node
;
1276 struct inode
*old
= NULL
;
1277 spin_lock(&inode_lock
);
1278 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1279 if (old
->i_ino
!= ino
)
1281 if (old
->i_sb
!= sb
)
1283 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
))
1287 if (likely(!node
)) {
1288 hlist_add_head(&inode
->i_hash
, head
);
1289 spin_unlock(&inode_lock
);
1293 spin_unlock(&inode_lock
);
1295 if (unlikely(!inode_unhashed(old
))) {
1302 EXPORT_SYMBOL(insert_inode_locked
);
1304 int insert_inode_locked4(struct inode
*inode
, unsigned long hashval
,
1305 int (*test
)(struct inode
*, void *), void *data
)
1307 struct super_block
*sb
= inode
->i_sb
;
1308 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1310 inode
->i_state
|= I_NEW
;
1313 struct hlist_node
*node
;
1314 struct inode
*old
= NULL
;
1316 spin_lock(&inode_lock
);
1317 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1318 if (old
->i_sb
!= sb
)
1320 if (!test(old
, data
))
1322 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
))
1326 if (likely(!node
)) {
1327 hlist_add_head(&inode
->i_hash
, head
);
1328 spin_unlock(&inode_lock
);
1332 spin_unlock(&inode_lock
);
1334 if (unlikely(!inode_unhashed(old
))) {
1341 EXPORT_SYMBOL(insert_inode_locked4
);
1344 int generic_delete_inode(struct inode
*inode
)
1348 EXPORT_SYMBOL(generic_delete_inode
);
1351 * Normal UNIX filesystem behaviour: delete the
1352 * inode when the usage count drops to zero, and
1355 int generic_drop_inode(struct inode
*inode
)
1357 return !inode
->i_nlink
|| inode_unhashed(inode
);
1359 EXPORT_SYMBOL_GPL(generic_drop_inode
);
1362 * Called when we're dropping the last reference
1365 * Call the FS "drop_inode()" function, defaulting to
1366 * the legacy UNIX filesystem behaviour. If it tells
1367 * us to evict inode, do so. Otherwise, retain inode
1368 * in cache if fs is alive, sync and evict if fs is
1371 static void iput_final(struct inode
*inode
)
1373 struct super_block
*sb
= inode
->i_sb
;
1374 const struct super_operations
*op
= inode
->i_sb
->s_op
;
1377 if (op
&& op
->drop_inode
)
1378 drop
= op
->drop_inode(inode
);
1380 drop
= generic_drop_inode(inode
);
1383 if (sb
->s_flags
& MS_ACTIVE
) {
1384 inode
->i_state
|= I_REFERENCED
;
1385 if (!(inode
->i_state
& (I_DIRTY
|I_SYNC
))) {
1386 inode_lru_list_add(inode
);
1388 spin_unlock(&inode_lock
);
1391 WARN_ON(inode
->i_state
& I_NEW
);
1392 inode
->i_state
|= I_WILL_FREE
;
1393 spin_unlock(&inode_lock
);
1394 write_inode_now(inode
, 1);
1395 spin_lock(&inode_lock
);
1396 WARN_ON(inode
->i_state
& I_NEW
);
1397 inode
->i_state
&= ~I_WILL_FREE
;
1398 __remove_inode_hash(inode
);
1401 WARN_ON(inode
->i_state
& I_NEW
);
1402 inode
->i_state
|= I_FREEING
;
1405 * Move the inode off the IO lists and LRU once I_FREEING is
1406 * set so that it won't get moved back on there if it is dirty.
1408 inode_lru_list_del(inode
);
1409 list_del_init(&inode
->i_wb_list
);
1411 __inode_sb_list_del(inode
);
1412 spin_unlock(&inode_lock
);
1414 remove_inode_hash(inode
);
1415 wake_up_inode(inode
);
1416 BUG_ON(inode
->i_state
!= (I_FREEING
| I_CLEAR
));
1417 destroy_inode(inode
);
1421 * iput - put an inode
1422 * @inode: inode to put
1424 * Puts an inode, dropping its usage count. If the inode use count hits
1425 * zero, the inode is then freed and may also be destroyed.
1427 * Consequently, iput() can sleep.
1429 void iput(struct inode
*inode
)
1432 BUG_ON(inode
->i_state
& I_CLEAR
);
1434 if (atomic_dec_and_lock(&inode
->i_count
, &inode_lock
))
1438 EXPORT_SYMBOL(iput
);
1441 * bmap - find a block number in a file
1442 * @inode: inode of file
1443 * @block: block to find
1445 * Returns the block number on the device holding the inode that
1446 * is the disk block number for the block of the file requested.
1447 * That is, asked for block 4 of inode 1 the function will return the
1448 * disk block relative to the disk start that holds that block of the
1451 sector_t
bmap(struct inode
*inode
, sector_t block
)
1454 if (inode
->i_mapping
->a_ops
->bmap
)
1455 res
= inode
->i_mapping
->a_ops
->bmap(inode
->i_mapping
, block
);
1458 EXPORT_SYMBOL(bmap
);
1461 * With relative atime, only update atime if the previous atime is
1462 * earlier than either the ctime or mtime or if at least a day has
1463 * passed since the last atime update.
1465 static int relatime_need_update(struct vfsmount
*mnt
, struct inode
*inode
,
1466 struct timespec now
)
1469 if (!(mnt
->mnt_flags
& MNT_RELATIME
))
1472 * Is mtime younger than atime? If yes, update atime:
1474 if (timespec_compare(&inode
->i_mtime
, &inode
->i_atime
) >= 0)
1477 * Is ctime younger than atime? If yes, update atime:
1479 if (timespec_compare(&inode
->i_ctime
, &inode
->i_atime
) >= 0)
1483 * Is the previous atime value older than a day? If yes,
1486 if ((long)(now
.tv_sec
- inode
->i_atime
.tv_sec
) >= 24*60*60)
1489 * Good, we can skip the atime update:
1495 * touch_atime - update the access time
1496 * @mnt: mount the inode is accessed on
1497 * @dentry: dentry accessed
1499 * Update the accessed time on an inode and mark it for writeback.
1500 * This function automatically handles read only file systems and media,
1501 * as well as the "noatime" flag and inode specific "noatime" markers.
1503 void touch_atime(struct vfsmount
*mnt
, struct dentry
*dentry
)
1505 struct inode
*inode
= dentry
->d_inode
;
1506 struct timespec now
;
1508 if (inode
->i_flags
& S_NOATIME
)
1510 if (IS_NOATIME(inode
))
1512 if ((inode
->i_sb
->s_flags
& MS_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1515 if (mnt
->mnt_flags
& MNT_NOATIME
)
1517 if ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1520 now
= current_fs_time(inode
->i_sb
);
1522 if (!relatime_need_update(mnt
, inode
, now
))
1525 if (timespec_equal(&inode
->i_atime
, &now
))
1528 if (mnt_want_write(mnt
))
1531 inode
->i_atime
= now
;
1532 mark_inode_dirty_sync(inode
);
1533 mnt_drop_write(mnt
);
1535 EXPORT_SYMBOL(touch_atime
);
1538 * file_update_time - update mtime and ctime time
1539 * @file: file accessed
1541 * Update the mtime and ctime members of an inode and mark the inode
1542 * for writeback. Note that this function is meant exclusively for
1543 * usage in the file write path of filesystems, and filesystems may
1544 * choose to explicitly ignore update via this function with the
1545 * S_NOCMTIME inode flag, e.g. for network filesystem where these
1546 * timestamps are handled by the server.
1549 void file_update_time(struct file
*file
)
1551 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1552 struct timespec now
;
1553 enum { S_MTIME
= 1, S_CTIME
= 2, S_VERSION
= 4 } sync_it
= 0;
1555 /* First try to exhaust all avenues to not sync */
1556 if (IS_NOCMTIME(inode
))
1559 now
= current_fs_time(inode
->i_sb
);
1560 if (!timespec_equal(&inode
->i_mtime
, &now
))
1563 if (!timespec_equal(&inode
->i_ctime
, &now
))
1566 if (IS_I_VERSION(inode
))
1567 sync_it
|= S_VERSION
;
1572 /* Finally allowed to write? Takes lock. */
1573 if (mnt_want_write_file(file
))
1576 /* Only change inode inside the lock region */
1577 if (sync_it
& S_VERSION
)
1578 inode_inc_iversion(inode
);
1579 if (sync_it
& S_CTIME
)
1580 inode
->i_ctime
= now
;
1581 if (sync_it
& S_MTIME
)
1582 inode
->i_mtime
= now
;
1583 mark_inode_dirty_sync(inode
);
1584 mnt_drop_write(file
->f_path
.mnt
);
1586 EXPORT_SYMBOL(file_update_time
);
1588 int inode_needs_sync(struct inode
*inode
)
1592 if (S_ISDIR(inode
->i_mode
) && IS_DIRSYNC(inode
))
1596 EXPORT_SYMBOL(inode_needs_sync
);
1598 int inode_wait(void *word
)
1603 EXPORT_SYMBOL(inode_wait
);
1606 * If we try to find an inode in the inode hash while it is being
1607 * deleted, we have to wait until the filesystem completes its
1608 * deletion before reporting that it isn't found. This function waits
1609 * until the deletion _might_ have completed. Callers are responsible
1610 * to recheck inode state.
1612 * It doesn't matter if I_NEW is not set initially, a call to
1613 * wake_up_inode() after removing from the hash list will DTRT.
1615 * This is called with inode_lock held.
1617 static void __wait_on_freeing_inode(struct inode
*inode
)
1619 wait_queue_head_t
*wq
;
1620 DEFINE_WAIT_BIT(wait
, &inode
->i_state
, __I_NEW
);
1621 wq
= bit_waitqueue(&inode
->i_state
, __I_NEW
);
1622 prepare_to_wait(wq
, &wait
.wait
, TASK_UNINTERRUPTIBLE
);
1623 spin_unlock(&inode_lock
);
1625 finish_wait(wq
, &wait
.wait
);
1626 spin_lock(&inode_lock
);
1629 static __initdata
unsigned long ihash_entries
;
1630 static int __init
set_ihash_entries(char *str
)
1634 ihash_entries
= simple_strtoul(str
, &str
, 0);
1637 __setup("ihash_entries=", set_ihash_entries
);
1640 * Initialize the waitqueues and inode hash table.
1642 void __init
inode_init_early(void)
1646 /* If hashes are distributed across NUMA nodes, defer
1647 * hash allocation until vmalloc space is available.
1653 alloc_large_system_hash("Inode-cache",
1654 sizeof(struct hlist_head
),
1662 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1663 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1666 void __init
inode_init(void)
1670 /* inode slab cache */
1671 inode_cachep
= kmem_cache_create("inode_cache",
1672 sizeof(struct inode
),
1674 (SLAB_RECLAIM_ACCOUNT
|SLAB_PANIC
|
1677 register_shrinker(&icache_shrinker
);
1679 /* Hash may have been set up in inode_init_early */
1684 alloc_large_system_hash("Inode-cache",
1685 sizeof(struct hlist_head
),
1693 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1694 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1697 void init_special_inode(struct inode
*inode
, umode_t mode
, dev_t rdev
)
1699 inode
->i_mode
= mode
;
1700 if (S_ISCHR(mode
)) {
1701 inode
->i_fop
= &def_chr_fops
;
1702 inode
->i_rdev
= rdev
;
1703 } else if (S_ISBLK(mode
)) {
1704 inode
->i_fop
= &def_blk_fops
;
1705 inode
->i_rdev
= rdev
;
1706 } else if (S_ISFIFO(mode
))
1707 inode
->i_fop
= &def_fifo_fops
;
1708 else if (S_ISSOCK(mode
))
1709 inode
->i_fop
= &bad_sock_fops
;
1711 printk(KERN_DEBUG
"init_special_inode: bogus i_mode (%o) for"
1712 " inode %s:%lu\n", mode
, inode
->i_sb
->s_id
,
1715 EXPORT_SYMBOL(init_special_inode
);
1718 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
1720 * @dir: Directory inode
1721 * @mode: mode of the new inode
1723 void inode_init_owner(struct inode
*inode
, const struct inode
*dir
,
1726 inode
->i_uid
= current_fsuid();
1727 if (dir
&& dir
->i_mode
& S_ISGID
) {
1728 inode
->i_gid
= dir
->i_gid
;
1732 inode
->i_gid
= current_fsgid();
1733 inode
->i_mode
= mode
;
1735 EXPORT_SYMBOL(inode_init_owner
);