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 struct percpu_counter nr_inodes __cacheline_aligned_in_smp
;
106 static struct percpu_counter nr_inodes_unused __cacheline_aligned_in_smp
;
108 static struct kmem_cache
*inode_cachep __read_mostly
;
110 static inline int get_nr_inodes(void)
112 return percpu_counter_sum_positive(&nr_inodes
);
115 static inline int get_nr_inodes_unused(void)
117 return percpu_counter_sum_positive(&nr_inodes_unused
);
120 int get_nr_dirty_inodes(void)
122 int nr_dirty
= get_nr_inodes() - get_nr_inodes_unused();
123 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 inodes_stat
.nr_unused
= get_nr_inodes_unused();
136 return proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
140 static void wake_up_inode(struct inode
*inode
)
143 * Prevent speculative execution through spin_unlock(&inode_lock);
146 wake_up_bit(&inode
->i_state
, __I_NEW
);
150 * inode_init_always - perform inode structure intialisation
151 * @sb: superblock inode belongs to
152 * @inode: inode to initialise
154 * These are initializations that need to be done on every inode
155 * allocation as the fields are not initialised by slab allocation.
157 int inode_init_always(struct super_block
*sb
, struct inode
*inode
)
159 static const struct address_space_operations empty_aops
;
160 static const struct inode_operations empty_iops
;
161 static const struct file_operations empty_fops
;
162 struct address_space
*const mapping
= &inode
->i_data
;
165 inode
->i_blkbits
= sb
->s_blocksize_bits
;
167 atomic_set(&inode
->i_count
, 1);
168 inode
->i_op
= &empty_iops
;
169 inode
->i_fop
= &empty_fops
;
173 atomic_set(&inode
->i_writecount
, 0);
177 inode
->i_generation
= 0;
179 memset(&inode
->i_dquot
, 0, sizeof(inode
->i_dquot
));
181 inode
->i_pipe
= NULL
;
182 inode
->i_bdev
= NULL
;
183 inode
->i_cdev
= NULL
;
185 inode
->dirtied_when
= 0;
187 if (security_inode_alloc(inode
))
189 spin_lock_init(&inode
->i_lock
);
190 lockdep_set_class(&inode
->i_lock
, &sb
->s_type
->i_lock_key
);
192 mutex_init(&inode
->i_mutex
);
193 lockdep_set_class(&inode
->i_mutex
, &sb
->s_type
->i_mutex_key
);
195 init_rwsem(&inode
->i_alloc_sem
);
196 lockdep_set_class(&inode
->i_alloc_sem
, &sb
->s_type
->i_alloc_sem_key
);
198 mapping
->a_ops
= &empty_aops
;
199 mapping
->host
= inode
;
201 mapping_set_gfp_mask(mapping
, GFP_HIGHUSER_MOVABLE
);
202 mapping
->assoc_mapping
= NULL
;
203 mapping
->backing_dev_info
= &default_backing_dev_info
;
204 mapping
->writeback_index
= 0;
207 * If the block_device provides a backing_dev_info for client
208 * inodes then use that. Otherwise the inode share the bdev's
212 struct backing_dev_info
*bdi
;
214 bdi
= sb
->s_bdev
->bd_inode
->i_mapping
->backing_dev_info
;
215 mapping
->backing_dev_info
= bdi
;
217 inode
->i_private
= NULL
;
218 inode
->i_mapping
= mapping
;
219 #ifdef CONFIG_FS_POSIX_ACL
220 inode
->i_acl
= inode
->i_default_acl
= ACL_NOT_CACHED
;
223 #ifdef CONFIG_FSNOTIFY
224 inode
->i_fsnotify_mask
= 0;
227 percpu_counter_inc(&nr_inodes
);
233 EXPORT_SYMBOL(inode_init_always
);
235 static struct inode
*alloc_inode(struct super_block
*sb
)
239 if (sb
->s_op
->alloc_inode
)
240 inode
= sb
->s_op
->alloc_inode(sb
);
242 inode
= kmem_cache_alloc(inode_cachep
, GFP_KERNEL
);
247 if (unlikely(inode_init_always(sb
, inode
))) {
248 if (inode
->i_sb
->s_op
->destroy_inode
)
249 inode
->i_sb
->s_op
->destroy_inode(inode
);
251 kmem_cache_free(inode_cachep
, inode
);
258 void __destroy_inode(struct inode
*inode
)
260 BUG_ON(inode_has_buffers(inode
));
261 security_inode_free(inode
);
262 fsnotify_inode_delete(inode
);
263 #ifdef CONFIG_FS_POSIX_ACL
264 if (inode
->i_acl
&& inode
->i_acl
!= ACL_NOT_CACHED
)
265 posix_acl_release(inode
->i_acl
);
266 if (inode
->i_default_acl
&& inode
->i_default_acl
!= ACL_NOT_CACHED
)
267 posix_acl_release(inode
->i_default_acl
);
269 percpu_counter_dec(&nr_inodes
);
271 EXPORT_SYMBOL(__destroy_inode
);
273 static void destroy_inode(struct inode
*inode
)
275 BUG_ON(!list_empty(&inode
->i_lru
));
276 __destroy_inode(inode
);
277 if (inode
->i_sb
->s_op
->destroy_inode
)
278 inode
->i_sb
->s_op
->destroy_inode(inode
);
280 kmem_cache_free(inode_cachep
, (inode
));
283 void address_space_init_once(struct address_space
*mapping
)
285 memset(mapping
, 0, sizeof(*mapping
));
286 INIT_RADIX_TREE(&mapping
->page_tree
, GFP_ATOMIC
);
287 spin_lock_init(&mapping
->tree_lock
);
288 spin_lock_init(&mapping
->i_mmap_lock
);
289 INIT_LIST_HEAD(&mapping
->private_list
);
290 spin_lock_init(&mapping
->private_lock
);
291 INIT_RAW_PRIO_TREE_ROOT(&mapping
->i_mmap
);
292 INIT_LIST_HEAD(&mapping
->i_mmap_nonlinear
);
293 mutex_init(&mapping
->unmap_mutex
);
295 EXPORT_SYMBOL(address_space_init_once
);
298 * These are initializations that only need to be done
299 * once, because the fields are idempotent across use
300 * of the inode, so let the slab aware of that.
302 void inode_init_once(struct inode
*inode
)
304 memset(inode
, 0, sizeof(*inode
));
305 INIT_HLIST_NODE(&inode
->i_hash
);
306 INIT_LIST_HEAD(&inode
->i_dentry
);
307 INIT_LIST_HEAD(&inode
->i_devices
);
308 INIT_LIST_HEAD(&inode
->i_wb_list
);
309 INIT_LIST_HEAD(&inode
->i_lru
);
310 address_space_init_once(&inode
->i_data
);
311 i_size_ordered_init(inode
);
312 #ifdef CONFIG_FSNOTIFY
313 INIT_HLIST_HEAD(&inode
->i_fsnotify_marks
);
316 EXPORT_SYMBOL(inode_init_once
);
318 static void init_once(void *foo
)
320 struct inode
*inode
= (struct inode
*) foo
;
322 inode_init_once(inode
);
326 * inode_lock must be held
328 void __iget(struct inode
*inode
)
330 atomic_inc(&inode
->i_count
);
334 * get additional reference to inode; caller must already hold one.
336 void ihold(struct inode
*inode
)
338 WARN_ON(atomic_inc_return(&inode
->i_count
) < 2);
340 EXPORT_SYMBOL(ihold
);
342 static void inode_lru_list_add(struct inode
*inode
)
344 if (list_empty(&inode
->i_lru
)) {
345 list_add(&inode
->i_lru
, &inode_lru
);
346 percpu_counter_inc(&nr_inodes_unused
);
350 static void inode_lru_list_del(struct inode
*inode
)
352 if (!list_empty(&inode
->i_lru
)) {
353 list_del_init(&inode
->i_lru
);
354 percpu_counter_dec(&nr_inodes_unused
);
358 static inline void __inode_sb_list_add(struct inode
*inode
)
360 list_add(&inode
->i_sb_list
, &inode
->i_sb
->s_inodes
);
364 * inode_sb_list_add - add inode to the superblock list of inodes
365 * @inode: inode to add
367 void inode_sb_list_add(struct inode
*inode
)
369 spin_lock(&inode_lock
);
370 __inode_sb_list_add(inode
);
371 spin_unlock(&inode_lock
);
373 EXPORT_SYMBOL_GPL(inode_sb_list_add
);
375 static inline void __inode_sb_list_del(struct inode
*inode
)
377 list_del_init(&inode
->i_sb_list
);
380 static unsigned long hash(struct super_block
*sb
, unsigned long hashval
)
384 tmp
= (hashval
* (unsigned long)sb
) ^ (GOLDEN_RATIO_PRIME
+ hashval
) /
386 tmp
= tmp
^ ((tmp
^ GOLDEN_RATIO_PRIME
) >> I_HASHBITS
);
387 return tmp
& I_HASHMASK
;
391 * __insert_inode_hash - hash an inode
392 * @inode: unhashed inode
393 * @hashval: unsigned long value used to locate this object in the
396 * Add an inode to the inode hash for this superblock.
398 void __insert_inode_hash(struct inode
*inode
, unsigned long hashval
)
400 struct hlist_head
*b
= inode_hashtable
+ hash(inode
->i_sb
, hashval
);
402 spin_lock(&inode_lock
);
403 hlist_add_head(&inode
->i_hash
, b
);
404 spin_unlock(&inode_lock
);
406 EXPORT_SYMBOL(__insert_inode_hash
);
409 * __remove_inode_hash - remove an inode from the hash
410 * @inode: inode to unhash
412 * Remove an inode from the superblock.
414 static void __remove_inode_hash(struct inode
*inode
)
416 hlist_del_init(&inode
->i_hash
);
420 * remove_inode_hash - remove an inode from the hash
421 * @inode: inode to unhash
423 * Remove an inode from the superblock.
425 void remove_inode_hash(struct inode
*inode
)
427 spin_lock(&inode_lock
);
428 hlist_del_init(&inode
->i_hash
);
429 spin_unlock(&inode_lock
);
431 EXPORT_SYMBOL(remove_inode_hash
);
433 void end_writeback(struct inode
*inode
)
436 BUG_ON(inode
->i_data
.nrpages
);
437 BUG_ON(!list_empty(&inode
->i_data
.private_list
));
438 BUG_ON(!(inode
->i_state
& I_FREEING
));
439 BUG_ON(inode
->i_state
& I_CLEAR
);
440 inode_sync_wait(inode
);
441 inode
->i_state
= I_FREEING
| I_CLEAR
;
443 EXPORT_SYMBOL(end_writeback
);
445 static void evict(struct inode
*inode
)
447 const struct super_operations
*op
= inode
->i_sb
->s_op
;
449 if (op
->evict_inode
) {
450 op
->evict_inode(inode
);
452 if (inode
->i_data
.nrpages
)
453 truncate_inode_pages(&inode
->i_data
, 0);
454 end_writeback(inode
);
456 if (S_ISBLK(inode
->i_mode
) && inode
->i_bdev
)
458 if (S_ISCHR(inode
->i_mode
) && inode
->i_cdev
)
463 * dispose_list - dispose of the contents of a local list
464 * @head: the head of the list to free
466 * Dispose-list gets a local list with local inodes in it, so it doesn't
467 * need to worry about list corruption and SMP locks.
469 static void dispose_list(struct list_head
*head
)
471 while (!list_empty(head
)) {
474 inode
= list_first_entry(head
, struct inode
, i_lru
);
475 list_del_init(&inode
->i_lru
);
479 spin_lock(&inode_lock
);
480 __remove_inode_hash(inode
);
481 __inode_sb_list_del(inode
);
482 spin_unlock(&inode_lock
);
484 wake_up_inode(inode
);
485 destroy_inode(inode
);
490 * evict_inodes - evict all evictable inodes for a superblock
491 * @sb: superblock to operate on
493 * Make sure that no inodes with zero refcount are retained. This is
494 * called by superblock shutdown after having MS_ACTIVE flag removed,
495 * so any inode reaching zero refcount during or after that call will
496 * be immediately evicted.
498 void evict_inodes(struct super_block
*sb
)
500 struct inode
*inode
, *next
;
503 down_write(&iprune_sem
);
505 spin_lock(&inode_lock
);
506 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
507 if (atomic_read(&inode
->i_count
))
510 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
515 inode
->i_state
|= I_FREEING
;
518 * Move the inode off the IO lists and LRU once I_FREEING is
519 * set so that it won't get moved back on there if it is dirty.
521 list_move(&inode
->i_lru
, &dispose
);
522 list_del_init(&inode
->i_wb_list
);
523 if (!(inode
->i_state
& (I_DIRTY
| I_SYNC
)))
524 percpu_counter_dec(&nr_inodes_unused
);
526 spin_unlock(&inode_lock
);
528 dispose_list(&dispose
);
529 up_write(&iprune_sem
);
533 * invalidate_inodes - attempt to free all inodes on a superblock
534 * @sb: superblock to operate on
535 * @kill_dirty: flag to guide handling of dirty inodes
537 * Attempts to free all inodes for a given superblock. If there were any
538 * busy inodes return a non-zero value, else zero.
539 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
542 int invalidate_inodes(struct super_block
*sb
, bool kill_dirty
)
545 struct inode
*inode
, *next
;
548 down_write(&iprune_sem
);
550 spin_lock(&inode_lock
);
551 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
552 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
))
554 if (inode
->i_state
& I_DIRTY
&& !kill_dirty
) {
558 if (atomic_read(&inode
->i_count
)) {
563 inode
->i_state
|= I_FREEING
;
566 * Move the inode off the IO lists and LRU once I_FREEING is
567 * set so that it won't get moved back on there if it is dirty.
569 list_move(&inode
->i_lru
, &dispose
);
570 list_del_init(&inode
->i_wb_list
);
571 if (!(inode
->i_state
& (I_DIRTY
| I_SYNC
)))
572 percpu_counter_dec(&nr_inodes_unused
);
574 spin_unlock(&inode_lock
);
576 dispose_list(&dispose
);
577 up_write(&iprune_sem
);
582 static int can_unuse(struct inode
*inode
)
584 if (inode
->i_state
& ~I_REFERENCED
)
586 if (inode_has_buffers(inode
))
588 if (atomic_read(&inode
->i_count
))
590 if (inode
->i_data
.nrpages
)
596 * Scan `goal' inodes on the unused list for freeable ones. They are moved to a
597 * temporary list and then are freed outside inode_lock by dispose_list().
599 * Any inodes which are pinned purely because of attached pagecache have their
600 * pagecache removed. If the inode has metadata buffers attached to
601 * mapping->private_list then try to remove them.
603 * If the inode has the I_REFERENCED flag set, then it means that it has been
604 * used recently - the flag is set in iput_final(). When we encounter such an
605 * inode, clear the flag and move it to the back of the LRU so it gets another
606 * pass through the LRU before it gets reclaimed. This is necessary because of
607 * the fact we are doing lazy LRU updates to minimise lock contention so the
608 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
609 * with this flag set because they are the inodes that are out of order.
611 static void prune_icache(int nr_to_scan
)
615 unsigned long reap
= 0;
617 down_read(&iprune_sem
);
618 spin_lock(&inode_lock
);
619 for (nr_scanned
= 0; nr_scanned
< nr_to_scan
; nr_scanned
++) {
622 if (list_empty(&inode_lru
))
625 inode
= list_entry(inode_lru
.prev
, struct inode
, i_lru
);
628 * Referenced or dirty inodes are still in use. Give them
629 * another pass through the LRU as we canot reclaim them now.
631 if (atomic_read(&inode
->i_count
) ||
632 (inode
->i_state
& ~I_REFERENCED
)) {
633 list_del_init(&inode
->i_lru
);
634 percpu_counter_dec(&nr_inodes_unused
);
638 /* recently referenced inodes get one more pass */
639 if (inode
->i_state
& I_REFERENCED
) {
640 list_move(&inode
->i_lru
, &inode_lru
);
641 inode
->i_state
&= ~I_REFERENCED
;
644 if (inode_has_buffers(inode
) || inode
->i_data
.nrpages
) {
646 spin_unlock(&inode_lock
);
647 if (remove_inode_buffers(inode
))
648 reap
+= invalidate_mapping_pages(&inode
->i_data
,
651 spin_lock(&inode_lock
);
653 if (inode
!= list_entry(inode_lru
.next
,
654 struct inode
, i_lru
))
655 continue; /* wrong inode or list_empty */
656 if (!can_unuse(inode
))
659 WARN_ON(inode
->i_state
& I_NEW
);
660 inode
->i_state
|= I_FREEING
;
663 * Move the inode off the IO lists and LRU once I_FREEING is
664 * set so that it won't get moved back on there if it is dirty.
666 list_move(&inode
->i_lru
, &freeable
);
667 list_del_init(&inode
->i_wb_list
);
668 percpu_counter_dec(&nr_inodes_unused
);
670 if (current_is_kswapd())
671 __count_vm_events(KSWAPD_INODESTEAL
, reap
);
673 __count_vm_events(PGINODESTEAL
, reap
);
674 spin_unlock(&inode_lock
);
676 dispose_list(&freeable
);
677 up_read(&iprune_sem
);
681 * shrink_icache_memory() will attempt to reclaim some unused inodes. Here,
682 * "unused" means that no dentries are referring to the inodes: the files are
683 * not open and the dcache references to those inodes have already been
686 * This function is passed the number of inodes to scan, and it returns the
687 * total number of remaining possibly-reclaimable inodes.
689 static int shrink_icache_memory(struct shrinker
*shrink
, int nr
, gfp_t gfp_mask
)
693 * Nasty deadlock avoidance. We may hold various FS locks,
694 * and we don't want to recurse into the FS that called us
695 * in clear_inode() and friends..
697 if (!(gfp_mask
& __GFP_FS
))
701 return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure
;
704 static struct shrinker icache_shrinker
= {
705 .shrink
= shrink_icache_memory
,
706 .seeks
= DEFAULT_SEEKS
,
709 static void __wait_on_freeing_inode(struct inode
*inode
);
711 * Called with the inode lock held.
713 static struct inode
*find_inode(struct super_block
*sb
,
714 struct hlist_head
*head
,
715 int (*test
)(struct inode
*, void *),
718 struct hlist_node
*node
;
719 struct inode
*inode
= NULL
;
722 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
723 if (inode
->i_sb
!= sb
)
725 if (!test(inode
, data
))
727 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
728 __wait_on_freeing_inode(inode
);
738 * find_inode_fast is the fast path version of find_inode, see the comment at
739 * iget_locked for details.
741 static struct inode
*find_inode_fast(struct super_block
*sb
,
742 struct hlist_head
*head
, unsigned long ino
)
744 struct hlist_node
*node
;
745 struct inode
*inode
= NULL
;
748 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
749 if (inode
->i_ino
!= ino
)
751 if (inode
->i_sb
!= sb
)
753 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
754 __wait_on_freeing_inode(inode
);
764 * Each cpu owns a range of LAST_INO_BATCH numbers.
765 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
766 * to renew the exhausted range.
768 * This does not significantly increase overflow rate because every CPU can
769 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
770 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
771 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
772 * overflow rate by 2x, which does not seem too significant.
774 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
775 * error if st_ino won't fit in target struct field. Use 32bit counter
776 * here to attempt to avoid that.
778 #define LAST_INO_BATCH 1024
779 static DEFINE_PER_CPU(unsigned int, last_ino
);
781 unsigned int get_next_ino(void)
783 unsigned int *p
= &get_cpu_var(last_ino
);
784 unsigned int res
= *p
;
787 if (unlikely((res
& (LAST_INO_BATCH
-1)) == 0)) {
788 static atomic_t shared_last_ino
;
789 int next
= atomic_add_return(LAST_INO_BATCH
, &shared_last_ino
);
791 res
= next
- LAST_INO_BATCH
;
796 put_cpu_var(last_ino
);
799 EXPORT_SYMBOL(get_next_ino
);
802 * new_inode - obtain an inode
805 * Allocates a new inode for given superblock. The default gfp_mask
806 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
807 * If HIGHMEM pages are unsuitable or it is known that pages allocated
808 * for the page cache are not reclaimable or migratable,
809 * mapping_set_gfp_mask() must be called with suitable flags on the
810 * newly created inode's mapping
813 struct inode
*new_inode(struct super_block
*sb
)
817 spin_lock_prefetch(&inode_lock
);
819 inode
= alloc_inode(sb
);
821 spin_lock(&inode_lock
);
822 __inode_sb_list_add(inode
);
824 spin_unlock(&inode_lock
);
828 EXPORT_SYMBOL(new_inode
);
830 void unlock_new_inode(struct inode
*inode
)
832 #ifdef CONFIG_DEBUG_LOCK_ALLOC
833 if (S_ISDIR(inode
->i_mode
)) {
834 struct file_system_type
*type
= inode
->i_sb
->s_type
;
836 /* Set new key only if filesystem hasn't already changed it */
837 if (!lockdep_match_class(&inode
->i_mutex
,
838 &type
->i_mutex_key
)) {
840 * ensure nobody is actually holding i_mutex
842 mutex_destroy(&inode
->i_mutex
);
843 mutex_init(&inode
->i_mutex
);
844 lockdep_set_class(&inode
->i_mutex
,
845 &type
->i_mutex_dir_key
);
850 * This is special! We do not need the spinlock when clearing I_NEW,
851 * because we're guaranteed that nobody else tries to do anything about
852 * the state of the inode when it is locked, as we just created it (so
853 * there can be no old holders that haven't tested I_NEW).
854 * However we must emit the memory barrier so that other CPUs reliably
855 * see the clearing of I_NEW after the other inode initialisation has
859 WARN_ON(!(inode
->i_state
& I_NEW
));
860 inode
->i_state
&= ~I_NEW
;
861 wake_up_inode(inode
);
863 EXPORT_SYMBOL(unlock_new_inode
);
866 * This is called without the inode lock held.. Be careful.
868 * We no longer cache the sb_flags in i_flags - see fs.h
869 * -- rmk@arm.uk.linux.org
871 static struct inode
*get_new_inode(struct super_block
*sb
,
872 struct hlist_head
*head
,
873 int (*test
)(struct inode
*, void *),
874 int (*set
)(struct inode
*, void *),
879 inode
= alloc_inode(sb
);
883 spin_lock(&inode_lock
);
884 /* We released the lock, so.. */
885 old
= find_inode(sb
, head
, test
, data
);
887 if (set(inode
, data
))
890 hlist_add_head(&inode
->i_hash
, head
);
891 __inode_sb_list_add(inode
);
892 inode
->i_state
= I_NEW
;
893 spin_unlock(&inode_lock
);
895 /* Return the locked inode with I_NEW set, the
896 * caller is responsible for filling in the contents
902 * Uhhuh, somebody else created the same inode under
903 * us. Use the old inode instead of the one we just
906 spin_unlock(&inode_lock
);
907 destroy_inode(inode
);
909 wait_on_inode(inode
);
914 spin_unlock(&inode_lock
);
915 destroy_inode(inode
);
920 * get_new_inode_fast is the fast path version of get_new_inode, see the
921 * comment at iget_locked for details.
923 static struct inode
*get_new_inode_fast(struct super_block
*sb
,
924 struct hlist_head
*head
, unsigned long ino
)
928 inode
= alloc_inode(sb
);
932 spin_lock(&inode_lock
);
933 /* We released the lock, so.. */
934 old
= find_inode_fast(sb
, head
, ino
);
937 hlist_add_head(&inode
->i_hash
, head
);
938 __inode_sb_list_add(inode
);
939 inode
->i_state
= I_NEW
;
940 spin_unlock(&inode_lock
);
942 /* Return the locked inode with I_NEW set, the
943 * caller is responsible for filling in the contents
949 * Uhhuh, somebody else created the same inode under
950 * us. Use the old inode instead of the one we just
953 spin_unlock(&inode_lock
);
954 destroy_inode(inode
);
956 wait_on_inode(inode
);
962 * search the inode cache for a matching inode number.
963 * If we find one, then the inode number we are trying to
964 * allocate is not unique and so we should not use it.
966 * Returns 1 if the inode number is unique, 0 if it is not.
968 static int test_inode_iunique(struct super_block
*sb
, unsigned long ino
)
970 struct hlist_head
*b
= inode_hashtable
+ hash(sb
, ino
);
971 struct hlist_node
*node
;
974 hlist_for_each_entry(inode
, node
, b
, i_hash
) {
975 if (inode
->i_ino
== ino
&& inode
->i_sb
== sb
)
983 * iunique - get a unique inode number
985 * @max_reserved: highest reserved inode number
987 * Obtain an inode number that is unique on the system for a given
988 * superblock. This is used by file systems that have no natural
989 * permanent inode numbering system. An inode number is returned that
990 * is higher than the reserved limit but unique.
993 * With a large number of inodes live on the file system this function
994 * currently becomes quite slow.
996 ino_t
iunique(struct super_block
*sb
, ino_t max_reserved
)
999 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1000 * error if st_ino won't fit in target struct field. Use 32bit counter
1001 * here to attempt to avoid that.
1003 static DEFINE_SPINLOCK(iunique_lock
);
1004 static unsigned int counter
;
1007 spin_lock(&inode_lock
);
1008 spin_lock(&iunique_lock
);
1010 if (counter
<= max_reserved
)
1011 counter
= max_reserved
+ 1;
1013 } while (!test_inode_iunique(sb
, res
));
1014 spin_unlock(&iunique_lock
);
1015 spin_unlock(&inode_lock
);
1019 EXPORT_SYMBOL(iunique
);
1021 struct inode
*igrab(struct inode
*inode
)
1023 spin_lock(&inode_lock
);
1024 if (!(inode
->i_state
& (I_FREEING
|I_WILL_FREE
)))
1028 * Handle the case where s_op->clear_inode is not been
1029 * called yet, and somebody is calling igrab
1030 * while the inode is getting freed.
1033 spin_unlock(&inode_lock
);
1036 EXPORT_SYMBOL(igrab
);
1039 * ifind - internal function, you want ilookup5() or iget5().
1040 * @sb: super block of file system to search
1041 * @head: the head of the list to search
1042 * @test: callback used for comparisons between inodes
1043 * @data: opaque data pointer to pass to @test
1044 * @wait: if true wait for the inode to be unlocked, if false do not
1046 * ifind() searches for the inode specified by @data in the inode
1047 * cache. This is a generalized version of ifind_fast() for file systems where
1048 * the inode number is not sufficient for unique identification of an inode.
1050 * If the inode is in the cache, the inode is returned with an incremented
1053 * Otherwise NULL is returned.
1055 * Note, @test is called with the inode_lock held, so can't sleep.
1057 static struct inode
*ifind(struct super_block
*sb
,
1058 struct hlist_head
*head
, int (*test
)(struct inode
*, void *),
1059 void *data
, const int wait
)
1061 struct inode
*inode
;
1063 spin_lock(&inode_lock
);
1064 inode
= find_inode(sb
, head
, test
, data
);
1066 spin_unlock(&inode_lock
);
1068 wait_on_inode(inode
);
1071 spin_unlock(&inode_lock
);
1076 * ifind_fast - internal function, you want ilookup() or iget().
1077 * @sb: super block of file system to search
1078 * @head: head of the list to search
1079 * @ino: inode number to search for
1081 * ifind_fast() searches for the inode @ino in the inode cache. This is for
1082 * file systems where the inode number is sufficient for unique identification
1085 * If the inode is in the cache, the inode is returned with an incremented
1088 * Otherwise NULL is returned.
1090 static struct inode
*ifind_fast(struct super_block
*sb
,
1091 struct hlist_head
*head
, unsigned long ino
)
1093 struct inode
*inode
;
1095 spin_lock(&inode_lock
);
1096 inode
= find_inode_fast(sb
, head
, ino
);
1098 spin_unlock(&inode_lock
);
1099 wait_on_inode(inode
);
1102 spin_unlock(&inode_lock
);
1107 * ilookup5_nowait - search for an inode in the inode cache
1108 * @sb: super block of file system to search
1109 * @hashval: hash value (usually inode number) to search for
1110 * @test: callback used for comparisons between inodes
1111 * @data: opaque data pointer to pass to @test
1113 * ilookup5() uses ifind() to search for the inode specified by @hashval and
1114 * @data in the inode cache. This is a generalized version of ilookup() for
1115 * file systems where the inode number is not sufficient for unique
1116 * identification of an inode.
1118 * If the inode is in the cache, the inode is returned with an incremented
1119 * reference count. Note, the inode lock is not waited upon so you have to be
1120 * very careful what you do with the returned inode. You probably should be
1121 * using ilookup5() instead.
1123 * Otherwise NULL is returned.
1125 * Note, @test is called with the inode_lock held, so can't sleep.
1127 struct inode
*ilookup5_nowait(struct super_block
*sb
, unsigned long hashval
,
1128 int (*test
)(struct inode
*, void *), void *data
)
1130 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1132 return ifind(sb
, head
, test
, data
, 0);
1134 EXPORT_SYMBOL(ilookup5_nowait
);
1137 * ilookup5 - search for an inode in the inode cache
1138 * @sb: super block of file system to search
1139 * @hashval: hash value (usually inode number) to search for
1140 * @test: callback used for comparisons between inodes
1141 * @data: opaque data pointer to pass to @test
1143 * ilookup5() uses ifind() to search for the inode specified by @hashval and
1144 * @data in the inode cache. This is a generalized version of ilookup() for
1145 * file systems where the inode number is not sufficient for unique
1146 * identification of an inode.
1148 * If the inode is in the cache, the inode lock is waited upon and the inode is
1149 * returned with an incremented reference count.
1151 * Otherwise NULL is returned.
1153 * Note, @test is called with the inode_lock held, so can't sleep.
1155 struct inode
*ilookup5(struct super_block
*sb
, unsigned long hashval
,
1156 int (*test
)(struct inode
*, void *), void *data
)
1158 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1160 return ifind(sb
, head
, test
, data
, 1);
1162 EXPORT_SYMBOL(ilookup5
);
1165 * ilookup - search for an inode in the inode cache
1166 * @sb: super block of file system to search
1167 * @ino: inode number to search for
1169 * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache.
1170 * This is for file systems where the inode number is sufficient for unique
1171 * identification of an inode.
1173 * If the inode is in the cache, the inode is returned with an incremented
1176 * Otherwise NULL is returned.
1178 struct inode
*ilookup(struct super_block
*sb
, unsigned long ino
)
1180 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1182 return ifind_fast(sb
, head
, ino
);
1184 EXPORT_SYMBOL(ilookup
);
1187 * iget5_locked - obtain an inode from a mounted file system
1188 * @sb: super block of file system
1189 * @hashval: hash value (usually inode number) to get
1190 * @test: callback used for comparisons between inodes
1191 * @set: callback used to initialize a new struct inode
1192 * @data: opaque data pointer to pass to @test and @set
1194 * iget5_locked() uses ifind() to search for the inode specified by @hashval
1195 * and @data in the inode cache and if present it is returned with an increased
1196 * reference count. This is a generalized version of iget_locked() for file
1197 * systems where the inode number is not sufficient for unique identification
1200 * If the inode is not in cache, get_new_inode() is called to allocate a new
1201 * inode and this is returned locked, hashed, and with the I_NEW flag set. The
1202 * file system gets to fill it in before unlocking it via unlock_new_inode().
1204 * Note both @test and @set are called with the inode_lock held, so can't sleep.
1206 struct inode
*iget5_locked(struct super_block
*sb
, unsigned long hashval
,
1207 int (*test
)(struct inode
*, void *),
1208 int (*set
)(struct inode
*, void *), void *data
)
1210 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1211 struct inode
*inode
;
1213 inode
= ifind(sb
, head
, test
, data
, 1);
1217 * get_new_inode() will do the right thing, re-trying the search
1218 * in case it had to block at any point.
1220 return get_new_inode(sb
, head
, test
, set
, data
);
1222 EXPORT_SYMBOL(iget5_locked
);
1225 * iget_locked - obtain an inode from a mounted file system
1226 * @sb: super block of file system
1227 * @ino: inode number to get
1229 * iget_locked() uses ifind_fast() to search for the inode specified by @ino in
1230 * the inode cache and if present it is returned with an increased reference
1231 * count. This is for file systems where the inode number is sufficient for
1232 * unique identification of an inode.
1234 * If the inode is not in cache, get_new_inode_fast() is called to allocate a
1235 * new inode and this is returned locked, hashed, and with the I_NEW flag set.
1236 * The file system gets to fill it in before unlocking it via
1237 * unlock_new_inode().
1239 struct inode
*iget_locked(struct super_block
*sb
, unsigned long ino
)
1241 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1242 struct inode
*inode
;
1244 inode
= ifind_fast(sb
, head
, ino
);
1248 * get_new_inode_fast() will do the right thing, re-trying the search
1249 * in case it had to block at any point.
1251 return get_new_inode_fast(sb
, head
, ino
);
1253 EXPORT_SYMBOL(iget_locked
);
1255 int insert_inode_locked(struct inode
*inode
)
1257 struct super_block
*sb
= inode
->i_sb
;
1258 ino_t ino
= inode
->i_ino
;
1259 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1261 inode
->i_state
|= I_NEW
;
1263 struct hlist_node
*node
;
1264 struct inode
*old
= NULL
;
1265 spin_lock(&inode_lock
);
1266 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1267 if (old
->i_ino
!= ino
)
1269 if (old
->i_sb
!= sb
)
1271 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
))
1275 if (likely(!node
)) {
1276 hlist_add_head(&inode
->i_hash
, head
);
1277 spin_unlock(&inode_lock
);
1281 spin_unlock(&inode_lock
);
1283 if (unlikely(!inode_unhashed(old
))) {
1290 EXPORT_SYMBOL(insert_inode_locked
);
1292 int insert_inode_locked4(struct inode
*inode
, unsigned long hashval
,
1293 int (*test
)(struct inode
*, void *), void *data
)
1295 struct super_block
*sb
= inode
->i_sb
;
1296 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1298 inode
->i_state
|= I_NEW
;
1301 struct hlist_node
*node
;
1302 struct inode
*old
= NULL
;
1304 spin_lock(&inode_lock
);
1305 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1306 if (old
->i_sb
!= sb
)
1308 if (!test(old
, data
))
1310 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
))
1314 if (likely(!node
)) {
1315 hlist_add_head(&inode
->i_hash
, head
);
1316 spin_unlock(&inode_lock
);
1320 spin_unlock(&inode_lock
);
1322 if (unlikely(!inode_unhashed(old
))) {
1329 EXPORT_SYMBOL(insert_inode_locked4
);
1332 int generic_delete_inode(struct inode
*inode
)
1336 EXPORT_SYMBOL(generic_delete_inode
);
1339 * Normal UNIX filesystem behaviour: delete the
1340 * inode when the usage count drops to zero, and
1343 int generic_drop_inode(struct inode
*inode
)
1345 return !inode
->i_nlink
|| inode_unhashed(inode
);
1347 EXPORT_SYMBOL_GPL(generic_drop_inode
);
1350 * Called when we're dropping the last reference
1353 * Call the FS "drop_inode()" function, defaulting to
1354 * the legacy UNIX filesystem behaviour. If it tells
1355 * us to evict inode, do so. Otherwise, retain inode
1356 * in cache if fs is alive, sync and evict if fs is
1359 static void iput_final(struct inode
*inode
)
1361 struct super_block
*sb
= inode
->i_sb
;
1362 const struct super_operations
*op
= inode
->i_sb
->s_op
;
1365 if (op
&& op
->drop_inode
)
1366 drop
= op
->drop_inode(inode
);
1368 drop
= generic_drop_inode(inode
);
1371 if (sb
->s_flags
& MS_ACTIVE
) {
1372 inode
->i_state
|= I_REFERENCED
;
1373 if (!(inode
->i_state
& (I_DIRTY
|I_SYNC
))) {
1374 inode_lru_list_add(inode
);
1376 spin_unlock(&inode_lock
);
1379 WARN_ON(inode
->i_state
& I_NEW
);
1380 inode
->i_state
|= I_WILL_FREE
;
1381 spin_unlock(&inode_lock
);
1382 write_inode_now(inode
, 1);
1383 spin_lock(&inode_lock
);
1384 WARN_ON(inode
->i_state
& I_NEW
);
1385 inode
->i_state
&= ~I_WILL_FREE
;
1386 __remove_inode_hash(inode
);
1389 WARN_ON(inode
->i_state
& I_NEW
);
1390 inode
->i_state
|= I_FREEING
;
1393 * Move the inode off the IO lists and LRU once I_FREEING is
1394 * set so that it won't get moved back on there if it is dirty.
1396 inode_lru_list_del(inode
);
1397 list_del_init(&inode
->i_wb_list
);
1399 __inode_sb_list_del(inode
);
1400 spin_unlock(&inode_lock
);
1402 remove_inode_hash(inode
);
1403 wake_up_inode(inode
);
1404 BUG_ON(inode
->i_state
!= (I_FREEING
| I_CLEAR
));
1405 destroy_inode(inode
);
1409 * iput - put an inode
1410 * @inode: inode to put
1412 * Puts an inode, dropping its usage count. If the inode use count hits
1413 * zero, the inode is then freed and may also be destroyed.
1415 * Consequently, iput() can sleep.
1417 void iput(struct inode
*inode
)
1420 BUG_ON(inode
->i_state
& I_CLEAR
);
1422 if (atomic_dec_and_lock(&inode
->i_count
, &inode_lock
))
1426 EXPORT_SYMBOL(iput
);
1429 * bmap - find a block number in a file
1430 * @inode: inode of file
1431 * @block: block to find
1433 * Returns the block number on the device holding the inode that
1434 * is the disk block number for the block of the file requested.
1435 * That is, asked for block 4 of inode 1 the function will return the
1436 * disk block relative to the disk start that holds that block of the
1439 sector_t
bmap(struct inode
*inode
, sector_t block
)
1442 if (inode
->i_mapping
->a_ops
->bmap
)
1443 res
= inode
->i_mapping
->a_ops
->bmap(inode
->i_mapping
, block
);
1446 EXPORT_SYMBOL(bmap
);
1449 * With relative atime, only update atime if the previous atime is
1450 * earlier than either the ctime or mtime or if at least a day has
1451 * passed since the last atime update.
1453 static int relatime_need_update(struct vfsmount
*mnt
, struct inode
*inode
,
1454 struct timespec now
)
1457 if (!(mnt
->mnt_flags
& MNT_RELATIME
))
1460 * Is mtime younger than atime? If yes, update atime:
1462 if (timespec_compare(&inode
->i_mtime
, &inode
->i_atime
) >= 0)
1465 * Is ctime younger than atime? If yes, update atime:
1467 if (timespec_compare(&inode
->i_ctime
, &inode
->i_atime
) >= 0)
1471 * Is the previous atime value older than a day? If yes,
1474 if ((long)(now
.tv_sec
- inode
->i_atime
.tv_sec
) >= 24*60*60)
1477 * Good, we can skip the atime update:
1483 * touch_atime - update the access time
1484 * @mnt: mount the inode is accessed on
1485 * @dentry: dentry accessed
1487 * Update the accessed time on an inode and mark it for writeback.
1488 * This function automatically handles read only file systems and media,
1489 * as well as the "noatime" flag and inode specific "noatime" markers.
1491 void touch_atime(struct vfsmount
*mnt
, struct dentry
*dentry
)
1493 struct inode
*inode
= dentry
->d_inode
;
1494 struct timespec now
;
1496 if (inode
->i_flags
& S_NOATIME
)
1498 if (IS_NOATIME(inode
))
1500 if ((inode
->i_sb
->s_flags
& MS_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1503 if (mnt
->mnt_flags
& MNT_NOATIME
)
1505 if ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1508 now
= current_fs_time(inode
->i_sb
);
1510 if (!relatime_need_update(mnt
, inode
, now
))
1513 if (timespec_equal(&inode
->i_atime
, &now
))
1516 if (mnt_want_write(mnt
))
1519 inode
->i_atime
= now
;
1520 mark_inode_dirty_sync(inode
);
1521 mnt_drop_write(mnt
);
1523 EXPORT_SYMBOL(touch_atime
);
1526 * file_update_time - update mtime and ctime time
1527 * @file: file accessed
1529 * Update the mtime and ctime members of an inode and mark the inode
1530 * for writeback. Note that this function is meant exclusively for
1531 * usage in the file write path of filesystems, and filesystems may
1532 * choose to explicitly ignore update via this function with the
1533 * S_NOCMTIME inode flag, e.g. for network filesystem where these
1534 * timestamps are handled by the server.
1537 void file_update_time(struct file
*file
)
1539 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1540 struct timespec now
;
1541 enum { S_MTIME
= 1, S_CTIME
= 2, S_VERSION
= 4 } sync_it
= 0;
1543 /* First try to exhaust all avenues to not sync */
1544 if (IS_NOCMTIME(inode
))
1547 now
= current_fs_time(inode
->i_sb
);
1548 if (!timespec_equal(&inode
->i_mtime
, &now
))
1551 if (!timespec_equal(&inode
->i_ctime
, &now
))
1554 if (IS_I_VERSION(inode
))
1555 sync_it
|= S_VERSION
;
1560 /* Finally allowed to write? Takes lock. */
1561 if (mnt_want_write_file(file
))
1564 /* Only change inode inside the lock region */
1565 if (sync_it
& S_VERSION
)
1566 inode_inc_iversion(inode
);
1567 if (sync_it
& S_CTIME
)
1568 inode
->i_ctime
= now
;
1569 if (sync_it
& S_MTIME
)
1570 inode
->i_mtime
= now
;
1571 mark_inode_dirty_sync(inode
);
1572 mnt_drop_write(file
->f_path
.mnt
);
1574 EXPORT_SYMBOL(file_update_time
);
1576 int inode_needs_sync(struct inode
*inode
)
1580 if (S_ISDIR(inode
->i_mode
) && IS_DIRSYNC(inode
))
1584 EXPORT_SYMBOL(inode_needs_sync
);
1586 int inode_wait(void *word
)
1591 EXPORT_SYMBOL(inode_wait
);
1594 * If we try to find an inode in the inode hash while it is being
1595 * deleted, we have to wait until the filesystem completes its
1596 * deletion before reporting that it isn't found. This function waits
1597 * until the deletion _might_ have completed. Callers are responsible
1598 * to recheck inode state.
1600 * It doesn't matter if I_NEW is not set initially, a call to
1601 * wake_up_inode() after removing from the hash list will DTRT.
1603 * This is called with inode_lock held.
1605 static void __wait_on_freeing_inode(struct inode
*inode
)
1607 wait_queue_head_t
*wq
;
1608 DEFINE_WAIT_BIT(wait
, &inode
->i_state
, __I_NEW
);
1609 wq
= bit_waitqueue(&inode
->i_state
, __I_NEW
);
1610 prepare_to_wait(wq
, &wait
.wait
, TASK_UNINTERRUPTIBLE
);
1611 spin_unlock(&inode_lock
);
1613 finish_wait(wq
, &wait
.wait
);
1614 spin_lock(&inode_lock
);
1617 static __initdata
unsigned long ihash_entries
;
1618 static int __init
set_ihash_entries(char *str
)
1622 ihash_entries
= simple_strtoul(str
, &str
, 0);
1625 __setup("ihash_entries=", set_ihash_entries
);
1628 * Initialize the waitqueues and inode hash table.
1630 void __init
inode_init_early(void)
1634 /* If hashes are distributed across NUMA nodes, defer
1635 * hash allocation until vmalloc space is available.
1641 alloc_large_system_hash("Inode-cache",
1642 sizeof(struct hlist_head
),
1650 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1651 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1654 void __init
inode_init(void)
1658 /* inode slab cache */
1659 inode_cachep
= kmem_cache_create("inode_cache",
1660 sizeof(struct inode
),
1662 (SLAB_RECLAIM_ACCOUNT
|SLAB_PANIC
|
1665 register_shrinker(&icache_shrinker
);
1666 percpu_counter_init(&nr_inodes
, 0);
1667 percpu_counter_init(&nr_inodes_unused
, 0);
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
< (1 << 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 * 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
);