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 kswapd or try_to_free_pages
88 * icache shrinking path, and the umount path. Without this exclusion,
89 * by the time prune_icache calls iput for the inode whose pages it has
90 * been invalidating, or by the time it calls clear_inode & destroy_inode
91 * from its final dispose_list, the struct super_block they refer to
92 * (for inode->i_sb->s_op) may already have been freed and reused.
94 * We make this an rwsem because the fastpath is icache shrinking. In
95 * some cases a filesystem may be doing a significant amount of work in
96 * its inode reclaim code, so this should improve parallelism.
98 static DECLARE_RWSEM(iprune_sem
);
101 * Statistics gathering..
103 struct inodes_stat_t inodes_stat
;
105 static DEFINE_PER_CPU(unsigned int, nr_inodes
);
107 static struct kmem_cache
*inode_cachep __read_mostly
;
109 static int get_nr_inodes(void)
113 for_each_possible_cpu(i
)
114 sum
+= per_cpu(nr_inodes
, i
);
115 return sum
< 0 ? 0 : sum
;
118 static inline int get_nr_inodes_unused(void)
120 return inodes_stat
.nr_unused
;
123 int get_nr_dirty_inodes(void)
125 /* not actually dirty inodes, but a wild approximation */
126 int nr_dirty
= get_nr_inodes() - get_nr_inodes_unused();
127 return nr_dirty
> 0 ? nr_dirty
: 0;
131 * Handle nr_inode sysctl
134 int proc_nr_inodes(ctl_table
*table
, int write
,
135 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
137 inodes_stat
.nr_inodes
= get_nr_inodes();
138 return proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
142 static void wake_up_inode(struct inode
*inode
)
145 * Prevent speculative execution through spin_unlock(&inode_lock);
148 wake_up_bit(&inode
->i_state
, __I_NEW
);
152 * inode_init_always - perform inode structure intialisation
153 * @sb: superblock inode belongs to
154 * @inode: inode to initialise
156 * These are initializations that need to be done on every inode
157 * allocation as the fields are not initialised by slab allocation.
159 int inode_init_always(struct super_block
*sb
, struct inode
*inode
)
161 static const struct address_space_operations empty_aops
;
162 static const struct inode_operations empty_iops
;
163 static const struct file_operations empty_fops
;
164 struct address_space
*const mapping
= &inode
->i_data
;
167 inode
->i_blkbits
= sb
->s_blocksize_bits
;
169 atomic_set(&inode
->i_count
, 1);
170 inode
->i_op
= &empty_iops
;
171 inode
->i_fop
= &empty_fops
;
175 atomic_set(&inode
->i_writecount
, 0);
179 inode
->i_generation
= 0;
181 memset(&inode
->i_dquot
, 0, sizeof(inode
->i_dquot
));
183 inode
->i_pipe
= NULL
;
184 inode
->i_bdev
= NULL
;
185 inode
->i_cdev
= NULL
;
187 inode
->dirtied_when
= 0;
189 if (security_inode_alloc(inode
))
191 spin_lock_init(&inode
->i_lock
);
192 lockdep_set_class(&inode
->i_lock
, &sb
->s_type
->i_lock_key
);
194 mutex_init(&inode
->i_mutex
);
195 lockdep_set_class(&inode
->i_mutex
, &sb
->s_type
->i_mutex_key
);
197 init_rwsem(&inode
->i_alloc_sem
);
198 lockdep_set_class(&inode
->i_alloc_sem
, &sb
->s_type
->i_alloc_sem_key
);
200 mapping
->a_ops
= &empty_aops
;
201 mapping
->host
= inode
;
203 mapping_set_gfp_mask(mapping
, GFP_HIGHUSER_MOVABLE
);
204 mapping
->assoc_mapping
= NULL
;
205 mapping
->backing_dev_info
= &default_backing_dev_info
;
206 mapping
->writeback_index
= 0;
209 * If the block_device provides a backing_dev_info for client
210 * inodes then use that. Otherwise the inode share the bdev's
214 struct backing_dev_info
*bdi
;
216 bdi
= sb
->s_bdev
->bd_inode
->i_mapping
->backing_dev_info
;
217 mapping
->backing_dev_info
= bdi
;
219 inode
->i_private
= NULL
;
220 inode
->i_mapping
= mapping
;
221 #ifdef CONFIG_FS_POSIX_ACL
222 inode
->i_acl
= inode
->i_default_acl
= ACL_NOT_CACHED
;
225 #ifdef CONFIG_FSNOTIFY
226 inode
->i_fsnotify_mask
= 0;
229 this_cpu_inc(nr_inodes
);
235 EXPORT_SYMBOL(inode_init_always
);
237 static struct inode
*alloc_inode(struct super_block
*sb
)
241 if (sb
->s_op
->alloc_inode
)
242 inode
= sb
->s_op
->alloc_inode(sb
);
244 inode
= kmem_cache_alloc(inode_cachep
, GFP_KERNEL
);
249 if (unlikely(inode_init_always(sb
, inode
))) {
250 if (inode
->i_sb
->s_op
->destroy_inode
)
251 inode
->i_sb
->s_op
->destroy_inode(inode
);
253 kmem_cache_free(inode_cachep
, inode
);
260 void free_inode_nonrcu(struct inode
*inode
)
262 kmem_cache_free(inode_cachep
, inode
);
264 EXPORT_SYMBOL(free_inode_nonrcu
);
266 void __destroy_inode(struct inode
*inode
)
268 BUG_ON(inode_has_buffers(inode
));
269 security_inode_free(inode
);
270 fsnotify_inode_delete(inode
);
271 #ifdef CONFIG_FS_POSIX_ACL
272 if (inode
->i_acl
&& inode
->i_acl
!= ACL_NOT_CACHED
)
273 posix_acl_release(inode
->i_acl
);
274 if (inode
->i_default_acl
&& inode
->i_default_acl
!= ACL_NOT_CACHED
)
275 posix_acl_release(inode
->i_default_acl
);
277 this_cpu_dec(nr_inodes
);
279 EXPORT_SYMBOL(__destroy_inode
);
281 static void i_callback(struct rcu_head
*head
)
283 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
284 INIT_LIST_HEAD(&inode
->i_dentry
);
285 kmem_cache_free(inode_cachep
, inode
);
288 static void destroy_inode(struct inode
*inode
)
290 BUG_ON(!list_empty(&inode
->i_lru
));
291 __destroy_inode(inode
);
292 if (inode
->i_sb
->s_op
->destroy_inode
)
293 inode
->i_sb
->s_op
->destroy_inode(inode
);
295 call_rcu(&inode
->i_rcu
, i_callback
);
298 void address_space_init_once(struct address_space
*mapping
)
300 memset(mapping
, 0, sizeof(*mapping
));
301 INIT_RADIX_TREE(&mapping
->page_tree
, GFP_ATOMIC
);
302 spin_lock_init(&mapping
->tree_lock
);
303 spin_lock_init(&mapping
->i_mmap_lock
);
304 INIT_LIST_HEAD(&mapping
->private_list
);
305 spin_lock_init(&mapping
->private_lock
);
306 INIT_RAW_PRIO_TREE_ROOT(&mapping
->i_mmap
);
307 INIT_LIST_HEAD(&mapping
->i_mmap_nonlinear
);
308 mutex_init(&mapping
->unmap_mutex
);
310 EXPORT_SYMBOL(address_space_init_once
);
313 * These are initializations that only need to be done
314 * once, because the fields are idempotent across use
315 * of the inode, so let the slab aware of that.
317 void inode_init_once(struct inode
*inode
)
319 memset(inode
, 0, sizeof(*inode
));
320 INIT_HLIST_NODE(&inode
->i_hash
);
321 INIT_LIST_HEAD(&inode
->i_dentry
);
322 INIT_LIST_HEAD(&inode
->i_devices
);
323 INIT_LIST_HEAD(&inode
->i_wb_list
);
324 INIT_LIST_HEAD(&inode
->i_lru
);
325 address_space_init_once(&inode
->i_data
);
326 i_size_ordered_init(inode
);
327 #ifdef CONFIG_FSNOTIFY
328 INIT_HLIST_HEAD(&inode
->i_fsnotify_marks
);
331 EXPORT_SYMBOL(inode_init_once
);
333 static void init_once(void *foo
)
335 struct inode
*inode
= (struct inode
*) foo
;
337 inode_init_once(inode
);
341 * inode_lock must be held
343 void __iget(struct inode
*inode
)
345 atomic_inc(&inode
->i_count
);
349 * get additional reference to inode; caller must already hold one.
351 void ihold(struct inode
*inode
)
353 WARN_ON(atomic_inc_return(&inode
->i_count
) < 2);
355 EXPORT_SYMBOL(ihold
);
357 static void inode_lru_list_add(struct inode
*inode
)
359 if (list_empty(&inode
->i_lru
)) {
360 list_add(&inode
->i_lru
, &inode_lru
);
361 inodes_stat
.nr_unused
++;
365 static void inode_lru_list_del(struct inode
*inode
)
367 if (!list_empty(&inode
->i_lru
)) {
368 list_del_init(&inode
->i_lru
);
369 inodes_stat
.nr_unused
--;
373 static inline void __inode_sb_list_add(struct inode
*inode
)
375 list_add(&inode
->i_sb_list
, &inode
->i_sb
->s_inodes
);
379 * inode_sb_list_add - add inode to the superblock list of inodes
380 * @inode: inode to add
382 void inode_sb_list_add(struct inode
*inode
)
384 spin_lock(&inode_lock
);
385 __inode_sb_list_add(inode
);
386 spin_unlock(&inode_lock
);
388 EXPORT_SYMBOL_GPL(inode_sb_list_add
);
390 static inline void __inode_sb_list_del(struct inode
*inode
)
392 list_del_init(&inode
->i_sb_list
);
395 static unsigned long hash(struct super_block
*sb
, unsigned long hashval
)
399 tmp
= (hashval
* (unsigned long)sb
) ^ (GOLDEN_RATIO_PRIME
+ hashval
) /
401 tmp
= tmp
^ ((tmp
^ GOLDEN_RATIO_PRIME
) >> I_HASHBITS
);
402 return tmp
& I_HASHMASK
;
406 * __insert_inode_hash - hash an inode
407 * @inode: unhashed inode
408 * @hashval: unsigned long value used to locate this object in the
411 * Add an inode to the inode hash for this superblock.
413 void __insert_inode_hash(struct inode
*inode
, unsigned long hashval
)
415 struct hlist_head
*b
= inode_hashtable
+ hash(inode
->i_sb
, hashval
);
417 spin_lock(&inode_lock
);
418 hlist_add_head(&inode
->i_hash
, b
);
419 spin_unlock(&inode_lock
);
421 EXPORT_SYMBOL(__insert_inode_hash
);
424 * __remove_inode_hash - remove an inode from the hash
425 * @inode: inode to unhash
427 * Remove an inode from the superblock.
429 static void __remove_inode_hash(struct inode
*inode
)
431 hlist_del_init(&inode
->i_hash
);
435 * remove_inode_hash - remove an inode from the hash
436 * @inode: inode to unhash
438 * Remove an inode from the superblock.
440 void remove_inode_hash(struct inode
*inode
)
442 spin_lock(&inode_lock
);
443 hlist_del_init(&inode
->i_hash
);
444 spin_unlock(&inode_lock
);
446 EXPORT_SYMBOL(remove_inode_hash
);
448 void end_writeback(struct inode
*inode
)
451 BUG_ON(inode
->i_data
.nrpages
);
452 BUG_ON(!list_empty(&inode
->i_data
.private_list
));
453 BUG_ON(!(inode
->i_state
& I_FREEING
));
454 BUG_ON(inode
->i_state
& I_CLEAR
);
455 inode_sync_wait(inode
);
456 /* don't need i_lock here, no concurrent mods to i_state */
457 inode
->i_state
= I_FREEING
| I_CLEAR
;
459 EXPORT_SYMBOL(end_writeback
);
461 static void evict(struct inode
*inode
)
463 const struct super_operations
*op
= inode
->i_sb
->s_op
;
465 if (op
->evict_inode
) {
466 op
->evict_inode(inode
);
468 if (inode
->i_data
.nrpages
)
469 truncate_inode_pages(&inode
->i_data
, 0);
470 end_writeback(inode
);
472 if (S_ISBLK(inode
->i_mode
) && inode
->i_bdev
)
474 if (S_ISCHR(inode
->i_mode
) && inode
->i_cdev
)
479 * dispose_list - dispose of the contents of a local list
480 * @head: the head of the list to free
482 * Dispose-list gets a local list with local inodes in it, so it doesn't
483 * need to worry about list corruption and SMP locks.
485 static void dispose_list(struct list_head
*head
)
487 while (!list_empty(head
)) {
490 inode
= list_first_entry(head
, struct inode
, i_lru
);
491 list_del_init(&inode
->i_lru
);
495 spin_lock(&inode_lock
);
496 __remove_inode_hash(inode
);
497 __inode_sb_list_del(inode
);
498 spin_unlock(&inode_lock
);
500 wake_up_inode(inode
);
501 destroy_inode(inode
);
506 * evict_inodes - evict all evictable inodes for a superblock
507 * @sb: superblock to operate on
509 * Make sure that no inodes with zero refcount are retained. This is
510 * called by superblock shutdown after having MS_ACTIVE flag removed,
511 * so any inode reaching zero refcount during or after that call will
512 * be immediately evicted.
514 void evict_inodes(struct super_block
*sb
)
516 struct inode
*inode
, *next
;
519 down_write(&iprune_sem
);
521 spin_lock(&inode_lock
);
522 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
523 if (atomic_read(&inode
->i_count
))
526 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
531 inode
->i_state
|= I_FREEING
;
534 * Move the inode off the IO lists and LRU once I_FREEING is
535 * set so that it won't get moved back on there if it is dirty.
537 list_move(&inode
->i_lru
, &dispose
);
538 list_del_init(&inode
->i_wb_list
);
539 if (!(inode
->i_state
& (I_DIRTY
| I_SYNC
)))
540 inodes_stat
.nr_unused
--;
542 spin_unlock(&inode_lock
);
544 dispose_list(&dispose
);
545 up_write(&iprune_sem
);
549 * invalidate_inodes - attempt to free all inodes on a superblock
550 * @sb: superblock to operate on
552 * Attempts to free all inodes for a given superblock. If there were any
553 * busy inodes return a non-zero value, else zero.
555 int invalidate_inodes(struct super_block
*sb
)
558 struct inode
*inode
, *next
;
561 down_write(&iprune_sem
);
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 (atomic_read(&inode
->i_count
)) {
572 inode
->i_state
|= I_FREEING
;
575 * Move the inode off the IO lists and LRU once I_FREEING is
576 * set so that it won't get moved back on there if it is dirty.
578 list_move(&inode
->i_lru
, &dispose
);
579 list_del_init(&inode
->i_wb_list
);
580 if (!(inode
->i_state
& (I_DIRTY
| I_SYNC
)))
581 inodes_stat
.nr_unused
--;
583 spin_unlock(&inode_lock
);
585 dispose_list(&dispose
);
586 up_write(&iprune_sem
);
591 static int can_unuse(struct inode
*inode
)
593 if (inode
->i_state
& ~I_REFERENCED
)
595 if (inode_has_buffers(inode
))
597 if (atomic_read(&inode
->i_count
))
599 if (inode
->i_data
.nrpages
)
605 * Scan `goal' inodes on the unused list for freeable ones. They are moved to a
606 * temporary list and then are freed outside inode_lock by dispose_list().
608 * Any inodes which are pinned purely because of attached pagecache have their
609 * pagecache removed. If the inode has metadata buffers attached to
610 * mapping->private_list then try to remove them.
612 * If the inode has the I_REFERENCED flag set, then it means that it has been
613 * used recently - the flag is set in iput_final(). When we encounter such an
614 * inode, clear the flag and move it to the back of the LRU so it gets another
615 * pass through the LRU before it gets reclaimed. This is necessary because of
616 * the fact we are doing lazy LRU updates to minimise lock contention so the
617 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
618 * with this flag set because they are the inodes that are out of order.
620 static void prune_icache(int nr_to_scan
)
624 unsigned long reap
= 0;
626 down_read(&iprune_sem
);
627 spin_lock(&inode_lock
);
628 for (nr_scanned
= 0; nr_scanned
< nr_to_scan
; nr_scanned
++) {
631 if (list_empty(&inode_lru
))
634 inode
= list_entry(inode_lru
.prev
, struct inode
, i_lru
);
637 * Referenced or dirty inodes are still in use. Give them
638 * another pass through the LRU as we canot reclaim them now.
640 if (atomic_read(&inode
->i_count
) ||
641 (inode
->i_state
& ~I_REFERENCED
)) {
642 list_del_init(&inode
->i_lru
);
643 inodes_stat
.nr_unused
--;
647 /* recently referenced inodes get one more pass */
648 if (inode
->i_state
& I_REFERENCED
) {
649 list_move(&inode
->i_lru
, &inode_lru
);
650 inode
->i_state
&= ~I_REFERENCED
;
653 if (inode_has_buffers(inode
) || inode
->i_data
.nrpages
) {
655 spin_unlock(&inode_lock
);
656 if (remove_inode_buffers(inode
))
657 reap
+= invalidate_mapping_pages(&inode
->i_data
,
660 spin_lock(&inode_lock
);
662 if (inode
!= list_entry(inode_lru
.next
,
663 struct inode
, i_lru
))
664 continue; /* wrong inode or list_empty */
665 if (!can_unuse(inode
))
668 WARN_ON(inode
->i_state
& I_NEW
);
669 inode
->i_state
|= I_FREEING
;
672 * Move the inode off the IO lists and LRU once I_FREEING is
673 * set so that it won't get moved back on there if it is dirty.
675 list_move(&inode
->i_lru
, &freeable
);
676 list_del_init(&inode
->i_wb_list
);
677 inodes_stat
.nr_unused
--;
679 if (current_is_kswapd())
680 __count_vm_events(KSWAPD_INODESTEAL
, reap
);
682 __count_vm_events(PGINODESTEAL
, reap
);
683 spin_unlock(&inode_lock
);
685 dispose_list(&freeable
);
686 up_read(&iprune_sem
);
690 * shrink_icache_memory() will attempt to reclaim some unused inodes. Here,
691 * "unused" means that no dentries are referring to the inodes: the files are
692 * not open and the dcache references to those inodes have already been
695 * This function is passed the number of inodes to scan, and it returns the
696 * total number of remaining possibly-reclaimable inodes.
698 static int shrink_icache_memory(struct shrinker
*shrink
, int nr
, gfp_t gfp_mask
)
702 * Nasty deadlock avoidance. We may hold various FS locks,
703 * and we don't want to recurse into the FS that called us
704 * in clear_inode() and friends..
706 if (!(gfp_mask
& __GFP_FS
))
710 return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure
;
713 static struct shrinker icache_shrinker
= {
714 .shrink
= shrink_icache_memory
,
715 .seeks
= DEFAULT_SEEKS
,
718 static void __wait_on_freeing_inode(struct inode
*inode
);
720 * Called with the inode lock held.
722 static struct inode
*find_inode(struct super_block
*sb
,
723 struct hlist_head
*head
,
724 int (*test
)(struct inode
*, void *),
727 struct hlist_node
*node
;
728 struct inode
*inode
= NULL
;
731 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
732 if (inode
->i_sb
!= sb
)
734 if (!test(inode
, data
))
736 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
737 __wait_on_freeing_inode(inode
);
747 * find_inode_fast is the fast path version of find_inode, see the comment at
748 * iget_locked for details.
750 static struct inode
*find_inode_fast(struct super_block
*sb
,
751 struct hlist_head
*head
, unsigned long ino
)
753 struct hlist_node
*node
;
754 struct inode
*inode
= NULL
;
757 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
758 if (inode
->i_ino
!= ino
)
760 if (inode
->i_sb
!= sb
)
762 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
763 __wait_on_freeing_inode(inode
);
773 * Each cpu owns a range of LAST_INO_BATCH numbers.
774 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
775 * to renew the exhausted range.
777 * This does not significantly increase overflow rate because every CPU can
778 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
779 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
780 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
781 * overflow rate by 2x, which does not seem too significant.
783 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
784 * error if st_ino won't fit in target struct field. Use 32bit counter
785 * here to attempt to avoid that.
787 #define LAST_INO_BATCH 1024
788 static DEFINE_PER_CPU(unsigned int, last_ino
);
790 unsigned int get_next_ino(void)
792 unsigned int *p
= &get_cpu_var(last_ino
);
793 unsigned int res
= *p
;
796 if (unlikely((res
& (LAST_INO_BATCH
-1)) == 0)) {
797 static atomic_t shared_last_ino
;
798 int next
= atomic_add_return(LAST_INO_BATCH
, &shared_last_ino
);
800 res
= next
- LAST_INO_BATCH
;
805 put_cpu_var(last_ino
);
808 EXPORT_SYMBOL(get_next_ino
);
811 * new_inode - obtain an inode
814 * Allocates a new inode for given superblock. The default gfp_mask
815 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
816 * If HIGHMEM pages are unsuitable or it is known that pages allocated
817 * for the page cache are not reclaimable or migratable,
818 * mapping_set_gfp_mask() must be called with suitable flags on the
819 * newly created inode's mapping
822 struct inode
*new_inode(struct super_block
*sb
)
826 spin_lock_prefetch(&inode_lock
);
828 inode
= alloc_inode(sb
);
830 spin_lock(&inode_lock
);
831 __inode_sb_list_add(inode
);
833 spin_unlock(&inode_lock
);
837 EXPORT_SYMBOL(new_inode
);
839 void unlock_new_inode(struct inode
*inode
)
841 #ifdef CONFIG_DEBUG_LOCK_ALLOC
842 if (S_ISDIR(inode
->i_mode
)) {
843 struct file_system_type
*type
= inode
->i_sb
->s_type
;
845 /* Set new key only if filesystem hasn't already changed it */
846 if (!lockdep_match_class(&inode
->i_mutex
,
847 &type
->i_mutex_key
)) {
849 * ensure nobody is actually holding i_mutex
851 mutex_destroy(&inode
->i_mutex
);
852 mutex_init(&inode
->i_mutex
);
853 lockdep_set_class(&inode
->i_mutex
,
854 &type
->i_mutex_dir_key
);
859 * This is special! We do not need the spinlock when clearing I_NEW,
860 * because we're guaranteed that nobody else tries to do anything about
861 * the state of the inode when it is locked, as we just created it (so
862 * there can be no old holders that haven't tested I_NEW).
863 * However we must emit the memory barrier so that other CPUs reliably
864 * see the clearing of I_NEW after the other inode initialisation has
868 WARN_ON(!(inode
->i_state
& I_NEW
));
869 inode
->i_state
&= ~I_NEW
;
870 wake_up_inode(inode
);
872 EXPORT_SYMBOL(unlock_new_inode
);
875 * This is called without the inode lock held.. Be careful.
877 * We no longer cache the sb_flags in i_flags - see fs.h
878 * -- rmk@arm.uk.linux.org
880 static struct inode
*get_new_inode(struct super_block
*sb
,
881 struct hlist_head
*head
,
882 int (*test
)(struct inode
*, void *),
883 int (*set
)(struct inode
*, void *),
888 inode
= alloc_inode(sb
);
892 spin_lock(&inode_lock
);
893 /* We released the lock, so.. */
894 old
= find_inode(sb
, head
, test
, data
);
896 if (set(inode
, data
))
899 hlist_add_head(&inode
->i_hash
, head
);
900 __inode_sb_list_add(inode
);
901 inode
->i_state
= I_NEW
;
902 spin_unlock(&inode_lock
);
904 /* Return the locked inode with I_NEW set, the
905 * caller is responsible for filling in the contents
911 * Uhhuh, somebody else created the same inode under
912 * us. Use the old inode instead of the one we just
915 spin_unlock(&inode_lock
);
916 destroy_inode(inode
);
918 wait_on_inode(inode
);
923 spin_unlock(&inode_lock
);
924 destroy_inode(inode
);
929 * get_new_inode_fast is the fast path version of get_new_inode, see the
930 * comment at iget_locked for details.
932 static struct inode
*get_new_inode_fast(struct super_block
*sb
,
933 struct hlist_head
*head
, unsigned long ino
)
937 inode
= alloc_inode(sb
);
941 spin_lock(&inode_lock
);
942 /* We released the lock, so.. */
943 old
= find_inode_fast(sb
, head
, ino
);
946 hlist_add_head(&inode
->i_hash
, head
);
947 __inode_sb_list_add(inode
);
948 inode
->i_state
= I_NEW
;
949 spin_unlock(&inode_lock
);
951 /* Return the locked inode with I_NEW set, the
952 * caller is responsible for filling in the contents
958 * Uhhuh, somebody else created the same inode under
959 * us. Use the old inode instead of the one we just
962 spin_unlock(&inode_lock
);
963 destroy_inode(inode
);
965 wait_on_inode(inode
);
971 * search the inode cache for a matching inode number.
972 * If we find one, then the inode number we are trying to
973 * allocate is not unique and so we should not use it.
975 * Returns 1 if the inode number is unique, 0 if it is not.
977 static int test_inode_iunique(struct super_block
*sb
, unsigned long ino
)
979 struct hlist_head
*b
= inode_hashtable
+ hash(sb
, ino
);
980 struct hlist_node
*node
;
983 hlist_for_each_entry(inode
, node
, b
, i_hash
) {
984 if (inode
->i_ino
== ino
&& inode
->i_sb
== sb
)
992 * iunique - get a unique inode number
994 * @max_reserved: highest reserved inode number
996 * Obtain an inode number that is unique on the system for a given
997 * superblock. This is used by file systems that have no natural
998 * permanent inode numbering system. An inode number is returned that
999 * is higher than the reserved limit but unique.
1002 * With a large number of inodes live on the file system this function
1003 * currently becomes quite slow.
1005 ino_t
iunique(struct super_block
*sb
, ino_t max_reserved
)
1008 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1009 * error if st_ino won't fit in target struct field. Use 32bit counter
1010 * here to attempt to avoid that.
1012 static DEFINE_SPINLOCK(iunique_lock
);
1013 static unsigned int counter
;
1016 spin_lock(&inode_lock
);
1017 spin_lock(&iunique_lock
);
1019 if (counter
<= max_reserved
)
1020 counter
= max_reserved
+ 1;
1022 } while (!test_inode_iunique(sb
, res
));
1023 spin_unlock(&iunique_lock
);
1024 spin_unlock(&inode_lock
);
1028 EXPORT_SYMBOL(iunique
);
1030 struct inode
*igrab(struct inode
*inode
)
1032 spin_lock(&inode_lock
);
1033 if (!(inode
->i_state
& (I_FREEING
|I_WILL_FREE
)))
1037 * Handle the case where s_op->clear_inode is not been
1038 * called yet, and somebody is calling igrab
1039 * while the inode is getting freed.
1042 spin_unlock(&inode_lock
);
1045 EXPORT_SYMBOL(igrab
);
1048 * ifind - internal function, you want ilookup5() or iget5().
1049 * @sb: super block of file system to search
1050 * @head: the head of the list to search
1051 * @test: callback used for comparisons between inodes
1052 * @data: opaque data pointer to pass to @test
1053 * @wait: if true wait for the inode to be unlocked, if false do not
1055 * ifind() searches for the inode specified by @data in the inode
1056 * cache. This is a generalized version of ifind_fast() for file systems where
1057 * the inode number is not sufficient for unique identification of an inode.
1059 * If the inode is in the cache, the inode is returned with an incremented
1062 * Otherwise NULL is returned.
1064 * Note, @test is called with the inode_lock held, so can't sleep.
1066 static struct inode
*ifind(struct super_block
*sb
,
1067 struct hlist_head
*head
, int (*test
)(struct inode
*, void *),
1068 void *data
, const int wait
)
1070 struct inode
*inode
;
1072 spin_lock(&inode_lock
);
1073 inode
= find_inode(sb
, head
, test
, data
);
1075 spin_unlock(&inode_lock
);
1077 wait_on_inode(inode
);
1080 spin_unlock(&inode_lock
);
1085 * ifind_fast - internal function, you want ilookup() or iget().
1086 * @sb: super block of file system to search
1087 * @head: head of the list to search
1088 * @ino: inode number to search for
1090 * ifind_fast() searches for the inode @ino in the inode cache. This is for
1091 * file systems where the inode number is sufficient for unique identification
1094 * If the inode is in the cache, the inode is returned with an incremented
1097 * Otherwise NULL is returned.
1099 static struct inode
*ifind_fast(struct super_block
*sb
,
1100 struct hlist_head
*head
, unsigned long ino
)
1102 struct inode
*inode
;
1104 spin_lock(&inode_lock
);
1105 inode
= find_inode_fast(sb
, head
, ino
);
1107 spin_unlock(&inode_lock
);
1108 wait_on_inode(inode
);
1111 spin_unlock(&inode_lock
);
1116 * ilookup5_nowait - search for an inode in the inode cache
1117 * @sb: super block of file system to search
1118 * @hashval: hash value (usually inode number) to search for
1119 * @test: callback used for comparisons between inodes
1120 * @data: opaque data pointer to pass to @test
1122 * ilookup5() uses ifind() to search for the inode specified by @hashval and
1123 * @data in the inode cache. This is a generalized version of ilookup() for
1124 * file systems where the inode number is not sufficient for unique
1125 * identification of an inode.
1127 * If the inode is in the cache, the inode is returned with an incremented
1128 * reference count. Note, the inode lock is not waited upon so you have to be
1129 * very careful what you do with the returned inode. You probably should be
1130 * using ilookup5() instead.
1132 * Otherwise NULL is returned.
1134 * Note, @test is called with the inode_lock held, so can't sleep.
1136 struct inode
*ilookup5_nowait(struct super_block
*sb
, unsigned long hashval
,
1137 int (*test
)(struct inode
*, void *), void *data
)
1139 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1141 return ifind(sb
, head
, test
, data
, 0);
1143 EXPORT_SYMBOL(ilookup5_nowait
);
1146 * ilookup5 - search for an inode in the inode cache
1147 * @sb: super block of file system to search
1148 * @hashval: hash value (usually inode number) to search for
1149 * @test: callback used for comparisons between inodes
1150 * @data: opaque data pointer to pass to @test
1152 * ilookup5() uses ifind() to search for the inode specified by @hashval and
1153 * @data in the inode cache. This is a generalized version of ilookup() for
1154 * file systems where the inode number is not sufficient for unique
1155 * identification of an inode.
1157 * If the inode is in the cache, the inode lock is waited upon and the inode is
1158 * returned with an incremented reference count.
1160 * Otherwise NULL is returned.
1162 * Note, @test is called with the inode_lock held, so can't sleep.
1164 struct inode
*ilookup5(struct super_block
*sb
, unsigned long hashval
,
1165 int (*test
)(struct inode
*, void *), void *data
)
1167 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1169 return ifind(sb
, head
, test
, data
, 1);
1171 EXPORT_SYMBOL(ilookup5
);
1174 * ilookup - search for an inode in the inode cache
1175 * @sb: super block of file system to search
1176 * @ino: inode number to search for
1178 * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache.
1179 * This is for file systems where the inode number is sufficient for unique
1180 * identification of an inode.
1182 * If the inode is in the cache, the inode is returned with an incremented
1185 * Otherwise NULL is returned.
1187 struct inode
*ilookup(struct super_block
*sb
, unsigned long ino
)
1189 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1191 return ifind_fast(sb
, head
, ino
);
1193 EXPORT_SYMBOL(ilookup
);
1196 * iget5_locked - obtain an inode from a mounted file system
1197 * @sb: super block of file system
1198 * @hashval: hash value (usually inode number) to get
1199 * @test: callback used for comparisons between inodes
1200 * @set: callback used to initialize a new struct inode
1201 * @data: opaque data pointer to pass to @test and @set
1203 * iget5_locked() uses ifind() to search for the inode specified by @hashval
1204 * and @data in the inode cache and if present it is returned with an increased
1205 * reference count. This is a generalized version of iget_locked() for file
1206 * systems where the inode number is not sufficient for unique identification
1209 * If the inode is not in cache, get_new_inode() is called to allocate a new
1210 * inode and this is returned locked, hashed, and with the I_NEW flag set. The
1211 * file system gets to fill it in before unlocking it via unlock_new_inode().
1213 * Note both @test and @set are called with the inode_lock held, so can't sleep.
1215 struct inode
*iget5_locked(struct super_block
*sb
, unsigned long hashval
,
1216 int (*test
)(struct inode
*, void *),
1217 int (*set
)(struct inode
*, void *), void *data
)
1219 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1220 struct inode
*inode
;
1222 inode
= ifind(sb
, head
, test
, data
, 1);
1226 * get_new_inode() will do the right thing, re-trying the search
1227 * in case it had to block at any point.
1229 return get_new_inode(sb
, head
, test
, set
, data
);
1231 EXPORT_SYMBOL(iget5_locked
);
1234 * iget_locked - obtain an inode from a mounted file system
1235 * @sb: super block of file system
1236 * @ino: inode number to get
1238 * iget_locked() uses ifind_fast() to search for the inode specified by @ino in
1239 * the inode cache and if present it is returned with an increased reference
1240 * count. This is for file systems where the inode number is sufficient for
1241 * unique identification of an inode.
1243 * If the inode is not in cache, get_new_inode_fast() is called to allocate a
1244 * new inode and this is returned locked, hashed, and with the I_NEW flag set.
1245 * The file system gets to fill it in before unlocking it via
1246 * unlock_new_inode().
1248 struct inode
*iget_locked(struct super_block
*sb
, unsigned long ino
)
1250 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1251 struct inode
*inode
;
1253 inode
= ifind_fast(sb
, head
, ino
);
1257 * get_new_inode_fast() will do the right thing, re-trying the search
1258 * in case it had to block at any point.
1260 return get_new_inode_fast(sb
, head
, ino
);
1262 EXPORT_SYMBOL(iget_locked
);
1264 int insert_inode_locked(struct inode
*inode
)
1266 struct super_block
*sb
= inode
->i_sb
;
1267 ino_t ino
= inode
->i_ino
;
1268 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1270 inode
->i_state
|= I_NEW
;
1272 struct hlist_node
*node
;
1273 struct inode
*old
= NULL
;
1274 spin_lock(&inode_lock
);
1275 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1276 if (old
->i_ino
!= ino
)
1278 if (old
->i_sb
!= sb
)
1280 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
))
1284 if (likely(!node
)) {
1285 hlist_add_head(&inode
->i_hash
, head
);
1286 spin_unlock(&inode_lock
);
1290 spin_unlock(&inode_lock
);
1292 if (unlikely(!inode_unhashed(old
))) {
1299 EXPORT_SYMBOL(insert_inode_locked
);
1301 int insert_inode_locked4(struct inode
*inode
, unsigned long hashval
,
1302 int (*test
)(struct inode
*, void *), void *data
)
1304 struct super_block
*sb
= inode
->i_sb
;
1305 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1307 inode
->i_state
|= I_NEW
;
1310 struct hlist_node
*node
;
1311 struct inode
*old
= NULL
;
1313 spin_lock(&inode_lock
);
1314 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1315 if (old
->i_sb
!= sb
)
1317 if (!test(old
, data
))
1319 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
))
1323 if (likely(!node
)) {
1324 hlist_add_head(&inode
->i_hash
, head
);
1325 spin_unlock(&inode_lock
);
1329 spin_unlock(&inode_lock
);
1331 if (unlikely(!inode_unhashed(old
))) {
1338 EXPORT_SYMBOL(insert_inode_locked4
);
1341 int generic_delete_inode(struct inode
*inode
)
1345 EXPORT_SYMBOL(generic_delete_inode
);
1348 * Normal UNIX filesystem behaviour: delete the
1349 * inode when the usage count drops to zero, and
1352 int generic_drop_inode(struct inode
*inode
)
1354 return !inode
->i_nlink
|| inode_unhashed(inode
);
1356 EXPORT_SYMBOL_GPL(generic_drop_inode
);
1359 * Called when we're dropping the last reference
1362 * Call the FS "drop_inode()" function, defaulting to
1363 * the legacy UNIX filesystem behaviour. If it tells
1364 * us to evict inode, do so. Otherwise, retain inode
1365 * in cache if fs is alive, sync and evict if fs is
1368 static void iput_final(struct inode
*inode
)
1370 struct super_block
*sb
= inode
->i_sb
;
1371 const struct super_operations
*op
= inode
->i_sb
->s_op
;
1374 if (op
&& op
->drop_inode
)
1375 drop
= op
->drop_inode(inode
);
1377 drop
= generic_drop_inode(inode
);
1380 if (sb
->s_flags
& MS_ACTIVE
) {
1381 inode
->i_state
|= I_REFERENCED
;
1382 if (!(inode
->i_state
& (I_DIRTY
|I_SYNC
))) {
1383 inode_lru_list_add(inode
);
1385 spin_unlock(&inode_lock
);
1388 WARN_ON(inode
->i_state
& I_NEW
);
1389 inode
->i_state
|= I_WILL_FREE
;
1390 spin_unlock(&inode_lock
);
1391 write_inode_now(inode
, 1);
1392 spin_lock(&inode_lock
);
1393 WARN_ON(inode
->i_state
& I_NEW
);
1394 inode
->i_state
&= ~I_WILL_FREE
;
1395 __remove_inode_hash(inode
);
1398 WARN_ON(inode
->i_state
& I_NEW
);
1399 inode
->i_state
|= I_FREEING
;
1402 * Move the inode off the IO lists and LRU once I_FREEING is
1403 * set so that it won't get moved back on there if it is dirty.
1405 inode_lru_list_del(inode
);
1406 list_del_init(&inode
->i_wb_list
);
1408 __inode_sb_list_del(inode
);
1409 spin_unlock(&inode_lock
);
1411 remove_inode_hash(inode
);
1412 wake_up_inode(inode
);
1413 BUG_ON(inode
->i_state
!= (I_FREEING
| I_CLEAR
));
1414 destroy_inode(inode
);
1418 * iput - put an inode
1419 * @inode: inode to put
1421 * Puts an inode, dropping its usage count. If the inode use count hits
1422 * zero, the inode is then freed and may also be destroyed.
1424 * Consequently, iput() can sleep.
1426 void iput(struct inode
*inode
)
1429 BUG_ON(inode
->i_state
& I_CLEAR
);
1431 if (atomic_dec_and_lock(&inode
->i_count
, &inode_lock
))
1435 EXPORT_SYMBOL(iput
);
1438 * bmap - find a block number in a file
1439 * @inode: inode of file
1440 * @block: block to find
1442 * Returns the block number on the device holding the inode that
1443 * is the disk block number for the block of the file requested.
1444 * That is, asked for block 4 of inode 1 the function will return the
1445 * disk block relative to the disk start that holds that block of the
1448 sector_t
bmap(struct inode
*inode
, sector_t block
)
1451 if (inode
->i_mapping
->a_ops
->bmap
)
1452 res
= inode
->i_mapping
->a_ops
->bmap(inode
->i_mapping
, block
);
1455 EXPORT_SYMBOL(bmap
);
1458 * With relative atime, only update atime if the previous atime is
1459 * earlier than either the ctime or mtime or if at least a day has
1460 * passed since the last atime update.
1462 static int relatime_need_update(struct vfsmount
*mnt
, struct inode
*inode
,
1463 struct timespec now
)
1466 if (!(mnt
->mnt_flags
& MNT_RELATIME
))
1469 * Is mtime younger than atime? If yes, update atime:
1471 if (timespec_compare(&inode
->i_mtime
, &inode
->i_atime
) >= 0)
1474 * Is ctime younger than atime? If yes, update atime:
1476 if (timespec_compare(&inode
->i_ctime
, &inode
->i_atime
) >= 0)
1480 * Is the previous atime value older than a day? If yes,
1483 if ((long)(now
.tv_sec
- inode
->i_atime
.tv_sec
) >= 24*60*60)
1486 * Good, we can skip the atime update:
1492 * touch_atime - update the access time
1493 * @mnt: mount the inode is accessed on
1494 * @dentry: dentry accessed
1496 * Update the accessed time on an inode and mark it for writeback.
1497 * This function automatically handles read only file systems and media,
1498 * as well as the "noatime" flag and inode specific "noatime" markers.
1500 void touch_atime(struct vfsmount
*mnt
, struct dentry
*dentry
)
1502 struct inode
*inode
= dentry
->d_inode
;
1503 struct timespec now
;
1505 if (inode
->i_flags
& S_NOATIME
)
1507 if (IS_NOATIME(inode
))
1509 if ((inode
->i_sb
->s_flags
& MS_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1512 if (mnt
->mnt_flags
& MNT_NOATIME
)
1514 if ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1517 now
= current_fs_time(inode
->i_sb
);
1519 if (!relatime_need_update(mnt
, inode
, now
))
1522 if (timespec_equal(&inode
->i_atime
, &now
))
1525 if (mnt_want_write(mnt
))
1528 inode
->i_atime
= now
;
1529 mark_inode_dirty_sync(inode
);
1530 mnt_drop_write(mnt
);
1532 EXPORT_SYMBOL(touch_atime
);
1535 * file_update_time - update mtime and ctime time
1536 * @file: file accessed
1538 * Update the mtime and ctime members of an inode and mark the inode
1539 * for writeback. Note that this function is meant exclusively for
1540 * usage in the file write path of filesystems, and filesystems may
1541 * choose to explicitly ignore update via this function with the
1542 * S_NOCMTIME inode flag, e.g. for network filesystem where these
1543 * timestamps are handled by the server.
1546 void file_update_time(struct file
*file
)
1548 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1549 struct timespec now
;
1550 enum { S_MTIME
= 1, S_CTIME
= 2, S_VERSION
= 4 } sync_it
= 0;
1552 /* First try to exhaust all avenues to not sync */
1553 if (IS_NOCMTIME(inode
))
1556 now
= current_fs_time(inode
->i_sb
);
1557 if (!timespec_equal(&inode
->i_mtime
, &now
))
1560 if (!timespec_equal(&inode
->i_ctime
, &now
))
1563 if (IS_I_VERSION(inode
))
1564 sync_it
|= S_VERSION
;
1569 /* Finally allowed to write? Takes lock. */
1570 if (mnt_want_write_file(file
))
1573 /* Only change inode inside the lock region */
1574 if (sync_it
& S_VERSION
)
1575 inode_inc_iversion(inode
);
1576 if (sync_it
& S_CTIME
)
1577 inode
->i_ctime
= now
;
1578 if (sync_it
& S_MTIME
)
1579 inode
->i_mtime
= now
;
1580 mark_inode_dirty_sync(inode
);
1581 mnt_drop_write(file
->f_path
.mnt
);
1583 EXPORT_SYMBOL(file_update_time
);
1585 int inode_needs_sync(struct inode
*inode
)
1589 if (S_ISDIR(inode
->i_mode
) && IS_DIRSYNC(inode
))
1593 EXPORT_SYMBOL(inode_needs_sync
);
1595 int inode_wait(void *word
)
1600 EXPORT_SYMBOL(inode_wait
);
1603 * If we try to find an inode in the inode hash while it is being
1604 * deleted, we have to wait until the filesystem completes its
1605 * deletion before reporting that it isn't found. This function waits
1606 * until the deletion _might_ have completed. Callers are responsible
1607 * to recheck inode state.
1609 * It doesn't matter if I_NEW is not set initially, a call to
1610 * wake_up_inode() after removing from the hash list will DTRT.
1612 * This is called with inode_lock held.
1614 static void __wait_on_freeing_inode(struct inode
*inode
)
1616 wait_queue_head_t
*wq
;
1617 DEFINE_WAIT_BIT(wait
, &inode
->i_state
, __I_NEW
);
1618 wq
= bit_waitqueue(&inode
->i_state
, __I_NEW
);
1619 prepare_to_wait(wq
, &wait
.wait
, TASK_UNINTERRUPTIBLE
);
1620 spin_unlock(&inode_lock
);
1622 finish_wait(wq
, &wait
.wait
);
1623 spin_lock(&inode_lock
);
1626 static __initdata
unsigned long ihash_entries
;
1627 static int __init
set_ihash_entries(char *str
)
1631 ihash_entries
= simple_strtoul(str
, &str
, 0);
1634 __setup("ihash_entries=", set_ihash_entries
);
1637 * Initialize the waitqueues and inode hash table.
1639 void __init
inode_init_early(void)
1643 /* If hashes are distributed across NUMA nodes, defer
1644 * hash allocation until vmalloc space is available.
1650 alloc_large_system_hash("Inode-cache",
1651 sizeof(struct hlist_head
),
1659 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1660 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1663 void __init
inode_init(void)
1667 /* inode slab cache */
1668 inode_cachep
= kmem_cache_create("inode_cache",
1669 sizeof(struct inode
),
1671 (SLAB_RECLAIM_ACCOUNT
|SLAB_PANIC
|
1674 register_shrinker(&icache_shrinker
);
1676 /* Hash may have been set up in inode_init_early */
1681 alloc_large_system_hash("Inode-cache",
1682 sizeof(struct hlist_head
),
1690 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1691 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1694 void init_special_inode(struct inode
*inode
, umode_t mode
, dev_t rdev
)
1696 inode
->i_mode
= mode
;
1697 if (S_ISCHR(mode
)) {
1698 inode
->i_fop
= &def_chr_fops
;
1699 inode
->i_rdev
= rdev
;
1700 } else if (S_ISBLK(mode
)) {
1701 inode
->i_fop
= &def_blk_fops
;
1702 inode
->i_rdev
= rdev
;
1703 } else if (S_ISFIFO(mode
))
1704 inode
->i_fop
= &def_fifo_fops
;
1705 else if (S_ISSOCK(mode
))
1706 inode
->i_fop
= &bad_sock_fops
;
1708 printk(KERN_DEBUG
"init_special_inode: bogus i_mode (%o) for"
1709 " inode %s:%lu\n", mode
, inode
->i_sb
->s_id
,
1712 EXPORT_SYMBOL(init_special_inode
);
1715 * Init uid,gid,mode for new inode according to posix standards
1717 * @dir: Directory inode
1718 * @mode: mode of the new inode
1720 void inode_init_owner(struct inode
*inode
, const struct inode
*dir
,
1723 inode
->i_uid
= current_fsuid();
1724 if (dir
&& dir
->i_mode
& S_ISGID
) {
1725 inode
->i_gid
= dir
->i_gid
;
1729 inode
->i_gid
= current_fsgid();
1730 inode
->i_mode
= mode
;
1732 EXPORT_SYMBOL(inode_init_owner
);