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>
28 #include <linux/cred.h>
31 * This is needed for the following functions:
35 * FIXME: remove all knowledge of the buffer layer from this file
37 #include <linux/buffer_head.h>
40 * New inode.c implementation.
42 * This implementation has the basic premise of trying
43 * to be extremely low-overhead and SMP-safe, yet be
44 * simple enough to be "obviously correct".
49 /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */
51 /* #define INODE_PARANOIA 1 */
52 /* #define INODE_DEBUG 1 */
55 * Inode lookup is no longer as critical as it used to be:
56 * most of the lookups are going to be through the dcache.
58 #define I_HASHBITS i_hash_shift
59 #define I_HASHMASK i_hash_mask
61 static unsigned int i_hash_mask __read_mostly
;
62 static unsigned int i_hash_shift __read_mostly
;
65 * Each inode can be on two separate lists. One is
66 * the hash list of the inode, used for lookups. The
67 * other linked list is the "type" list:
68 * "in_use" - valid inode, i_count > 0, i_nlink > 0
69 * "dirty" - as "in_use" but also dirty
70 * "unused" - valid inode, i_count = 0
72 * A "dirty" list is maintained for each super block,
73 * allowing for low-overhead inode sync() operations.
76 static LIST_HEAD(inode_lru
);
77 static struct hlist_head
*inode_hashtable __read_mostly
;
80 * A simple spinlock to protect the list manipulations.
82 * NOTE! You also have to own the lock if you change
83 * the i_state of an inode while it is in use..
85 DEFINE_SPINLOCK(inode_lock
);
88 * iprune_sem provides exclusion between the icache shrinking and the
91 * We don't actually need it to protect anything in the umount path,
92 * but only need to cycle through it to make sure any inode that
93 * prune_icache took off the LRU list has been fully torn down by the
94 * time we are past evict_inodes.
96 static DECLARE_RWSEM(iprune_sem
);
99 * Statistics gathering..
101 struct inodes_stat_t inodes_stat
;
103 static DEFINE_PER_CPU(unsigned int, nr_inodes
);
105 static struct kmem_cache
*inode_cachep __read_mostly
;
107 static int get_nr_inodes(void)
111 for_each_possible_cpu(i
)
112 sum
+= per_cpu(nr_inodes
, i
);
113 return sum
< 0 ? 0 : sum
;
116 static inline int get_nr_inodes_unused(void)
118 return inodes_stat
.nr_unused
;
121 int get_nr_dirty_inodes(void)
123 /* not actually dirty inodes, but a wild approximation */
124 int nr_dirty
= get_nr_inodes() - get_nr_inodes_unused();
125 return nr_dirty
> 0 ? nr_dirty
: 0;
129 * Handle nr_inode sysctl
132 int proc_nr_inodes(ctl_table
*table
, int write
,
133 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
135 inodes_stat
.nr_inodes
= get_nr_inodes();
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 this_cpu_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 free_inode_nonrcu(struct inode
*inode
)
260 kmem_cache_free(inode_cachep
, inode
);
262 EXPORT_SYMBOL(free_inode_nonrcu
);
264 void __destroy_inode(struct inode
*inode
)
266 BUG_ON(inode_has_buffers(inode
));
267 security_inode_free(inode
);
268 fsnotify_inode_delete(inode
);
269 #ifdef CONFIG_FS_POSIX_ACL
270 if (inode
->i_acl
&& inode
->i_acl
!= ACL_NOT_CACHED
)
271 posix_acl_release(inode
->i_acl
);
272 if (inode
->i_default_acl
&& inode
->i_default_acl
!= ACL_NOT_CACHED
)
273 posix_acl_release(inode
->i_default_acl
);
275 this_cpu_dec(nr_inodes
);
277 EXPORT_SYMBOL(__destroy_inode
);
279 static void i_callback(struct rcu_head
*head
)
281 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
282 INIT_LIST_HEAD(&inode
->i_dentry
);
283 kmem_cache_free(inode_cachep
, inode
);
286 static void destroy_inode(struct inode
*inode
)
288 BUG_ON(!list_empty(&inode
->i_lru
));
289 __destroy_inode(inode
);
290 if (inode
->i_sb
->s_op
->destroy_inode
)
291 inode
->i_sb
->s_op
->destroy_inode(inode
);
293 call_rcu(&inode
->i_rcu
, i_callback
);
296 void address_space_init_once(struct address_space
*mapping
)
298 memset(mapping
, 0, sizeof(*mapping
));
299 INIT_RADIX_TREE(&mapping
->page_tree
, GFP_ATOMIC
);
300 spin_lock_init(&mapping
->tree_lock
);
301 spin_lock_init(&mapping
->i_mmap_lock
);
302 INIT_LIST_HEAD(&mapping
->private_list
);
303 spin_lock_init(&mapping
->private_lock
);
304 INIT_RAW_PRIO_TREE_ROOT(&mapping
->i_mmap
);
305 INIT_LIST_HEAD(&mapping
->i_mmap_nonlinear
);
306 mutex_init(&mapping
->unmap_mutex
);
308 EXPORT_SYMBOL(address_space_init_once
);
311 * These are initializations that only need to be done
312 * once, because the fields are idempotent across use
313 * of the inode, so let the slab aware of that.
315 void inode_init_once(struct inode
*inode
)
317 memset(inode
, 0, sizeof(*inode
));
318 INIT_HLIST_NODE(&inode
->i_hash
);
319 INIT_LIST_HEAD(&inode
->i_dentry
);
320 INIT_LIST_HEAD(&inode
->i_devices
);
321 INIT_LIST_HEAD(&inode
->i_wb_list
);
322 INIT_LIST_HEAD(&inode
->i_lru
);
323 address_space_init_once(&inode
->i_data
);
324 i_size_ordered_init(inode
);
325 #ifdef CONFIG_FSNOTIFY
326 INIT_HLIST_HEAD(&inode
->i_fsnotify_marks
);
329 EXPORT_SYMBOL(inode_init_once
);
331 static void init_once(void *foo
)
333 struct inode
*inode
= (struct inode
*) foo
;
335 inode_init_once(inode
);
339 * inode_lock must be held
341 void __iget(struct inode
*inode
)
343 atomic_inc(&inode
->i_count
);
347 * get additional reference to inode; caller must already hold one.
349 void ihold(struct inode
*inode
)
351 WARN_ON(atomic_inc_return(&inode
->i_count
) < 2);
353 EXPORT_SYMBOL(ihold
);
355 static void inode_lru_list_add(struct inode
*inode
)
357 if (list_empty(&inode
->i_lru
)) {
358 list_add(&inode
->i_lru
, &inode_lru
);
359 inodes_stat
.nr_unused
++;
363 static void inode_lru_list_del(struct inode
*inode
)
365 if (!list_empty(&inode
->i_lru
)) {
366 list_del_init(&inode
->i_lru
);
367 inodes_stat
.nr_unused
--;
371 static inline void __inode_sb_list_add(struct inode
*inode
)
373 list_add(&inode
->i_sb_list
, &inode
->i_sb
->s_inodes
);
377 * inode_sb_list_add - add inode to the superblock list of inodes
378 * @inode: inode to add
380 void inode_sb_list_add(struct inode
*inode
)
382 spin_lock(&inode_lock
);
383 __inode_sb_list_add(inode
);
384 spin_unlock(&inode_lock
);
386 EXPORT_SYMBOL_GPL(inode_sb_list_add
);
388 static inline void __inode_sb_list_del(struct inode
*inode
)
390 list_del_init(&inode
->i_sb_list
);
393 static unsigned long hash(struct super_block
*sb
, unsigned long hashval
)
397 tmp
= (hashval
* (unsigned long)sb
) ^ (GOLDEN_RATIO_PRIME
+ hashval
) /
399 tmp
= tmp
^ ((tmp
^ GOLDEN_RATIO_PRIME
) >> I_HASHBITS
);
400 return tmp
& I_HASHMASK
;
404 * __insert_inode_hash - hash an inode
405 * @inode: unhashed inode
406 * @hashval: unsigned long value used to locate this object in the
409 * Add an inode to the inode hash for this superblock.
411 void __insert_inode_hash(struct inode
*inode
, unsigned long hashval
)
413 struct hlist_head
*b
= inode_hashtable
+ hash(inode
->i_sb
, hashval
);
415 spin_lock(&inode_lock
);
416 hlist_add_head(&inode
->i_hash
, b
);
417 spin_unlock(&inode_lock
);
419 EXPORT_SYMBOL(__insert_inode_hash
);
422 * __remove_inode_hash - remove an inode from the hash
423 * @inode: inode to unhash
425 * Remove an inode from the superblock.
427 static void __remove_inode_hash(struct inode
*inode
)
429 hlist_del_init(&inode
->i_hash
);
433 * remove_inode_hash - remove an inode from the hash
434 * @inode: inode to unhash
436 * Remove an inode from the superblock.
438 void remove_inode_hash(struct inode
*inode
)
440 spin_lock(&inode_lock
);
441 hlist_del_init(&inode
->i_hash
);
442 spin_unlock(&inode_lock
);
444 EXPORT_SYMBOL(remove_inode_hash
);
446 void end_writeback(struct inode
*inode
)
449 BUG_ON(inode
->i_data
.nrpages
);
450 BUG_ON(!list_empty(&inode
->i_data
.private_list
));
451 BUG_ON(!(inode
->i_state
& I_FREEING
));
452 BUG_ON(inode
->i_state
& I_CLEAR
);
453 inode_sync_wait(inode
);
454 /* don't need i_lock here, no concurrent mods to i_state */
455 inode
->i_state
= I_FREEING
| I_CLEAR
;
457 EXPORT_SYMBOL(end_writeback
);
459 static void evict(struct inode
*inode
)
461 const struct super_operations
*op
= inode
->i_sb
->s_op
;
463 if (op
->evict_inode
) {
464 op
->evict_inode(inode
);
466 if (inode
->i_data
.nrpages
)
467 truncate_inode_pages(&inode
->i_data
, 0);
468 end_writeback(inode
);
470 if (S_ISBLK(inode
->i_mode
) && inode
->i_bdev
)
472 if (S_ISCHR(inode
->i_mode
) && inode
->i_cdev
)
477 * dispose_list - dispose of the contents of a local list
478 * @head: the head of the list to free
480 * Dispose-list gets a local list with local inodes in it, so it doesn't
481 * need to worry about list corruption and SMP locks.
483 static void dispose_list(struct list_head
*head
)
485 while (!list_empty(head
)) {
488 inode
= list_first_entry(head
, struct inode
, i_lru
);
489 list_del_init(&inode
->i_lru
);
493 spin_lock(&inode_lock
);
494 __remove_inode_hash(inode
);
495 __inode_sb_list_del(inode
);
496 spin_unlock(&inode_lock
);
498 wake_up_inode(inode
);
499 destroy_inode(inode
);
504 * evict_inodes - evict all evictable inodes for a superblock
505 * @sb: superblock to operate on
507 * Make sure that no inodes with zero refcount are retained. This is
508 * called by superblock shutdown after having MS_ACTIVE flag removed,
509 * so any inode reaching zero refcount during or after that call will
510 * be immediately evicted.
512 void evict_inodes(struct super_block
*sb
)
514 struct inode
*inode
, *next
;
517 spin_lock(&inode_lock
);
518 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
519 if (atomic_read(&inode
->i_count
))
521 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
))
524 inode
->i_state
|= I_FREEING
;
527 * Move the inode off the IO lists and LRU once I_FREEING is
528 * set so that it won't get moved back on there if it is dirty.
530 list_move(&inode
->i_lru
, &dispose
);
531 list_del_init(&inode
->i_wb_list
);
532 if (!(inode
->i_state
& (I_DIRTY
| I_SYNC
)))
533 inodes_stat
.nr_unused
--;
535 spin_unlock(&inode_lock
);
537 dispose_list(&dispose
);
540 * Cycle through iprune_sem to make sure any inode that prune_icache
541 * moved off the list before we took the lock has been fully torn
544 down_write(&iprune_sem
);
545 up_write(&iprune_sem
);
549 * invalidate_inodes - attempt to free all inodes on a superblock
550 * @sb: superblock to operate on
551 * @kill_dirty: flag to guide handling of dirty inodes
553 * Attempts to free all inodes for a given superblock. If there were any
554 * busy inodes return a non-zero value, else zero.
555 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
558 int invalidate_inodes(struct super_block
*sb
, bool kill_dirty
)
561 struct inode
*inode
, *next
;
564 spin_lock(&inode_lock
);
565 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
566 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
))
568 if (inode
->i_state
& I_DIRTY
&& !kill_dirty
) {
572 if (atomic_read(&inode
->i_count
)) {
577 inode
->i_state
|= I_FREEING
;
580 * Move the inode off the IO lists and LRU once I_FREEING is
581 * set so that it won't get moved back on there if it is dirty.
583 list_move(&inode
->i_lru
, &dispose
);
584 list_del_init(&inode
->i_wb_list
);
585 if (!(inode
->i_state
& (I_DIRTY
| I_SYNC
)))
586 inodes_stat
.nr_unused
--;
588 spin_unlock(&inode_lock
);
590 dispose_list(&dispose
);
595 static int can_unuse(struct inode
*inode
)
597 if (inode
->i_state
& ~I_REFERENCED
)
599 if (inode_has_buffers(inode
))
601 if (atomic_read(&inode
->i_count
))
603 if (inode
->i_data
.nrpages
)
609 * Scan `goal' inodes on the unused list for freeable ones. They are moved to a
610 * temporary list and then are freed outside inode_lock by dispose_list().
612 * Any inodes which are pinned purely because of attached pagecache have their
613 * pagecache removed. If the inode has metadata buffers attached to
614 * mapping->private_list then try to remove them.
616 * If the inode has the I_REFERENCED flag set, then it means that it has been
617 * used recently - the flag is set in iput_final(). When we encounter such an
618 * inode, clear the flag and move it to the back of the LRU so it gets another
619 * pass through the LRU before it gets reclaimed. This is necessary because of
620 * the fact we are doing lazy LRU updates to minimise lock contention so the
621 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
622 * with this flag set because they are the inodes that are out of order.
624 static void prune_icache(int nr_to_scan
)
628 unsigned long reap
= 0;
630 down_read(&iprune_sem
);
631 spin_lock(&inode_lock
);
632 for (nr_scanned
= 0; nr_scanned
< nr_to_scan
; nr_scanned
++) {
635 if (list_empty(&inode_lru
))
638 inode
= list_entry(inode_lru
.prev
, struct inode
, i_lru
);
641 * Referenced or dirty inodes are still in use. Give them
642 * another pass through the LRU as we canot reclaim them now.
644 if (atomic_read(&inode
->i_count
) ||
645 (inode
->i_state
& ~I_REFERENCED
)) {
646 list_del_init(&inode
->i_lru
);
647 inodes_stat
.nr_unused
--;
651 /* recently referenced inodes get one more pass */
652 if (inode
->i_state
& I_REFERENCED
) {
653 list_move(&inode
->i_lru
, &inode_lru
);
654 inode
->i_state
&= ~I_REFERENCED
;
657 if (inode_has_buffers(inode
) || inode
->i_data
.nrpages
) {
659 spin_unlock(&inode_lock
);
660 if (remove_inode_buffers(inode
))
661 reap
+= invalidate_mapping_pages(&inode
->i_data
,
664 spin_lock(&inode_lock
);
666 if (inode
!= list_entry(inode_lru
.next
,
667 struct inode
, i_lru
))
668 continue; /* wrong inode or list_empty */
669 if (!can_unuse(inode
))
672 WARN_ON(inode
->i_state
& I_NEW
);
673 inode
->i_state
|= I_FREEING
;
676 * Move the inode off the IO lists and LRU once I_FREEING is
677 * set so that it won't get moved back on there if it is dirty.
679 list_move(&inode
->i_lru
, &freeable
);
680 list_del_init(&inode
->i_wb_list
);
681 inodes_stat
.nr_unused
--;
683 if (current_is_kswapd())
684 __count_vm_events(KSWAPD_INODESTEAL
, reap
);
686 __count_vm_events(PGINODESTEAL
, reap
);
687 spin_unlock(&inode_lock
);
689 dispose_list(&freeable
);
690 up_read(&iprune_sem
);
694 * shrink_icache_memory() will attempt to reclaim some unused inodes. Here,
695 * "unused" means that no dentries are referring to the inodes: the files are
696 * not open and the dcache references to those inodes have already been
699 * This function is passed the number of inodes to scan, and it returns the
700 * total number of remaining possibly-reclaimable inodes.
702 static int shrink_icache_memory(struct shrinker
*shrink
, int nr
, gfp_t gfp_mask
)
706 * Nasty deadlock avoidance. We may hold various FS locks,
707 * and we don't want to recurse into the FS that called us
708 * in clear_inode() and friends..
710 if (!(gfp_mask
& __GFP_FS
))
714 return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure
;
717 static struct shrinker icache_shrinker
= {
718 .shrink
= shrink_icache_memory
,
719 .seeks
= DEFAULT_SEEKS
,
722 static void __wait_on_freeing_inode(struct inode
*inode
);
724 * Called with the inode lock held.
726 static struct inode
*find_inode(struct super_block
*sb
,
727 struct hlist_head
*head
,
728 int (*test
)(struct inode
*, void *),
731 struct hlist_node
*node
;
732 struct inode
*inode
= NULL
;
735 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
736 if (inode
->i_sb
!= sb
)
738 if (!test(inode
, data
))
740 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
741 __wait_on_freeing_inode(inode
);
751 * find_inode_fast is the fast path version of find_inode, see the comment at
752 * iget_locked for details.
754 static struct inode
*find_inode_fast(struct super_block
*sb
,
755 struct hlist_head
*head
, unsigned long ino
)
757 struct hlist_node
*node
;
758 struct inode
*inode
= NULL
;
761 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
762 if (inode
->i_ino
!= ino
)
764 if (inode
->i_sb
!= sb
)
766 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
767 __wait_on_freeing_inode(inode
);
777 * Each cpu owns a range of LAST_INO_BATCH numbers.
778 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
779 * to renew the exhausted range.
781 * This does not significantly increase overflow rate because every CPU can
782 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
783 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
784 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
785 * overflow rate by 2x, which does not seem too significant.
787 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
788 * error if st_ino won't fit in target struct field. Use 32bit counter
789 * here to attempt to avoid that.
791 #define LAST_INO_BATCH 1024
792 static DEFINE_PER_CPU(unsigned int, last_ino
);
794 unsigned int get_next_ino(void)
796 unsigned int *p
= &get_cpu_var(last_ino
);
797 unsigned int res
= *p
;
800 if (unlikely((res
& (LAST_INO_BATCH
-1)) == 0)) {
801 static atomic_t shared_last_ino
;
802 int next
= atomic_add_return(LAST_INO_BATCH
, &shared_last_ino
);
804 res
= next
- LAST_INO_BATCH
;
809 put_cpu_var(last_ino
);
812 EXPORT_SYMBOL(get_next_ino
);
815 * new_inode - obtain an inode
818 * Allocates a new inode for given superblock. The default gfp_mask
819 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
820 * If HIGHMEM pages are unsuitable or it is known that pages allocated
821 * for the page cache are not reclaimable or migratable,
822 * mapping_set_gfp_mask() must be called with suitable flags on the
823 * newly created inode's mapping
826 struct inode
*new_inode(struct super_block
*sb
)
830 spin_lock_prefetch(&inode_lock
);
832 inode
= alloc_inode(sb
);
834 spin_lock(&inode_lock
);
835 __inode_sb_list_add(inode
);
837 spin_unlock(&inode_lock
);
841 EXPORT_SYMBOL(new_inode
);
843 void unlock_new_inode(struct inode
*inode
)
845 #ifdef CONFIG_DEBUG_LOCK_ALLOC
846 if (S_ISDIR(inode
->i_mode
)) {
847 struct file_system_type
*type
= inode
->i_sb
->s_type
;
849 /* Set new key only if filesystem hasn't already changed it */
850 if (!lockdep_match_class(&inode
->i_mutex
,
851 &type
->i_mutex_key
)) {
853 * ensure nobody is actually holding i_mutex
855 mutex_destroy(&inode
->i_mutex
);
856 mutex_init(&inode
->i_mutex
);
857 lockdep_set_class(&inode
->i_mutex
,
858 &type
->i_mutex_dir_key
);
863 * This is special! We do not need the spinlock when clearing I_NEW,
864 * because we're guaranteed that nobody else tries to do anything about
865 * the state of the inode when it is locked, as we just created it (so
866 * there can be no old holders that haven't tested I_NEW).
867 * However we must emit the memory barrier so that other CPUs reliably
868 * see the clearing of I_NEW after the other inode initialisation has
872 WARN_ON(!(inode
->i_state
& I_NEW
));
873 inode
->i_state
&= ~I_NEW
;
874 wake_up_inode(inode
);
876 EXPORT_SYMBOL(unlock_new_inode
);
879 * This is called without the inode lock held.. Be careful.
881 * We no longer cache the sb_flags in i_flags - see fs.h
882 * -- rmk@arm.uk.linux.org
884 static struct inode
*get_new_inode(struct super_block
*sb
,
885 struct hlist_head
*head
,
886 int (*test
)(struct inode
*, void *),
887 int (*set
)(struct inode
*, void *),
892 inode
= alloc_inode(sb
);
896 spin_lock(&inode_lock
);
897 /* We released the lock, so.. */
898 old
= find_inode(sb
, head
, test
, data
);
900 if (set(inode
, data
))
903 hlist_add_head(&inode
->i_hash
, head
);
904 __inode_sb_list_add(inode
);
905 inode
->i_state
= I_NEW
;
906 spin_unlock(&inode_lock
);
908 /* Return the locked inode with I_NEW set, the
909 * caller is responsible for filling in the contents
915 * Uhhuh, somebody else created the same inode under
916 * us. Use the old inode instead of the one we just
919 spin_unlock(&inode_lock
);
920 destroy_inode(inode
);
922 wait_on_inode(inode
);
927 spin_unlock(&inode_lock
);
928 destroy_inode(inode
);
933 * get_new_inode_fast is the fast path version of get_new_inode, see the
934 * comment at iget_locked for details.
936 static struct inode
*get_new_inode_fast(struct super_block
*sb
,
937 struct hlist_head
*head
, unsigned long ino
)
941 inode
= alloc_inode(sb
);
945 spin_lock(&inode_lock
);
946 /* We released the lock, so.. */
947 old
= find_inode_fast(sb
, head
, ino
);
950 hlist_add_head(&inode
->i_hash
, head
);
951 __inode_sb_list_add(inode
);
952 inode
->i_state
= I_NEW
;
953 spin_unlock(&inode_lock
);
955 /* Return the locked inode with I_NEW set, the
956 * caller is responsible for filling in the contents
962 * Uhhuh, somebody else created the same inode under
963 * us. Use the old inode instead of the one we just
966 spin_unlock(&inode_lock
);
967 destroy_inode(inode
);
969 wait_on_inode(inode
);
975 * search the inode cache for a matching inode number.
976 * If we find one, then the inode number we are trying to
977 * allocate is not unique and so we should not use it.
979 * Returns 1 if the inode number is unique, 0 if it is not.
981 static int test_inode_iunique(struct super_block
*sb
, unsigned long ino
)
983 struct hlist_head
*b
= inode_hashtable
+ hash(sb
, ino
);
984 struct hlist_node
*node
;
987 hlist_for_each_entry(inode
, node
, b
, i_hash
) {
988 if (inode
->i_ino
== ino
&& inode
->i_sb
== sb
)
996 * iunique - get a unique inode number
998 * @max_reserved: highest reserved inode number
1000 * Obtain an inode number that is unique on the system for a given
1001 * superblock. This is used by file systems that have no natural
1002 * permanent inode numbering system. An inode number is returned that
1003 * is higher than the reserved limit but unique.
1006 * With a large number of inodes live on the file system this function
1007 * currently becomes quite slow.
1009 ino_t
iunique(struct super_block
*sb
, ino_t max_reserved
)
1012 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1013 * error if st_ino won't fit in target struct field. Use 32bit counter
1014 * here to attempt to avoid that.
1016 static DEFINE_SPINLOCK(iunique_lock
);
1017 static unsigned int counter
;
1020 spin_lock(&inode_lock
);
1021 spin_lock(&iunique_lock
);
1023 if (counter
<= max_reserved
)
1024 counter
= max_reserved
+ 1;
1026 } while (!test_inode_iunique(sb
, res
));
1027 spin_unlock(&iunique_lock
);
1028 spin_unlock(&inode_lock
);
1032 EXPORT_SYMBOL(iunique
);
1034 struct inode
*igrab(struct inode
*inode
)
1036 spin_lock(&inode_lock
);
1037 if (!(inode
->i_state
& (I_FREEING
|I_WILL_FREE
)))
1041 * Handle the case where s_op->clear_inode is not been
1042 * called yet, and somebody is calling igrab
1043 * while the inode is getting freed.
1046 spin_unlock(&inode_lock
);
1049 EXPORT_SYMBOL(igrab
);
1052 * ifind - internal function, you want ilookup5() or iget5().
1053 * @sb: super block of file system to search
1054 * @head: the head of the list to search
1055 * @test: callback used for comparisons between inodes
1056 * @data: opaque data pointer to pass to @test
1057 * @wait: if true wait for the inode to be unlocked, if false do not
1059 * ifind() searches for the inode specified by @data in the inode
1060 * cache. This is a generalized version of ifind_fast() for file systems where
1061 * the inode number is not sufficient for unique identification of an inode.
1063 * If the inode is in the cache, the inode is returned with an incremented
1066 * Otherwise NULL is returned.
1068 * Note, @test is called with the inode_lock held, so can't sleep.
1070 static struct inode
*ifind(struct super_block
*sb
,
1071 struct hlist_head
*head
, int (*test
)(struct inode
*, void *),
1072 void *data
, const int wait
)
1074 struct inode
*inode
;
1076 spin_lock(&inode_lock
);
1077 inode
= find_inode(sb
, head
, test
, data
);
1079 spin_unlock(&inode_lock
);
1081 wait_on_inode(inode
);
1084 spin_unlock(&inode_lock
);
1089 * ifind_fast - internal function, you want ilookup() or iget().
1090 * @sb: super block of file system to search
1091 * @head: head of the list to search
1092 * @ino: inode number to search for
1094 * ifind_fast() searches for the inode @ino in the inode cache. This is for
1095 * file systems where the inode number is sufficient for unique identification
1098 * If the inode is in the cache, the inode is returned with an incremented
1101 * Otherwise NULL is returned.
1103 static struct inode
*ifind_fast(struct super_block
*sb
,
1104 struct hlist_head
*head
, unsigned long ino
)
1106 struct inode
*inode
;
1108 spin_lock(&inode_lock
);
1109 inode
= find_inode_fast(sb
, head
, ino
);
1111 spin_unlock(&inode_lock
);
1112 wait_on_inode(inode
);
1115 spin_unlock(&inode_lock
);
1120 * ilookup5_nowait - search for an inode in the inode cache
1121 * @sb: super block of file system to search
1122 * @hashval: hash value (usually inode number) to search for
1123 * @test: callback used for comparisons between inodes
1124 * @data: opaque data pointer to pass to @test
1126 * ilookup5() uses ifind() to search for the inode specified by @hashval and
1127 * @data in the inode cache. This is a generalized version of ilookup() for
1128 * file systems where the inode number is not sufficient for unique
1129 * identification of an inode.
1131 * If the inode is in the cache, the inode is returned with an incremented
1132 * reference count. Note, the inode lock is not waited upon so you have to be
1133 * very careful what you do with the returned inode. You probably should be
1134 * using ilookup5() instead.
1136 * Otherwise NULL is returned.
1138 * Note, @test is called with the inode_lock held, so can't sleep.
1140 struct inode
*ilookup5_nowait(struct super_block
*sb
, unsigned long hashval
,
1141 int (*test
)(struct inode
*, void *), void *data
)
1143 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1145 return ifind(sb
, head
, test
, data
, 0);
1147 EXPORT_SYMBOL(ilookup5_nowait
);
1150 * ilookup5 - search for an inode in the inode cache
1151 * @sb: super block of file system to search
1152 * @hashval: hash value (usually inode number) to search for
1153 * @test: callback used for comparisons between inodes
1154 * @data: opaque data pointer to pass to @test
1156 * ilookup5() uses ifind() to search for the inode specified by @hashval and
1157 * @data in the inode cache. This is a generalized version of ilookup() for
1158 * file systems where the inode number is not sufficient for unique
1159 * identification of an inode.
1161 * If the inode is in the cache, the inode lock is waited upon and the inode is
1162 * returned with an incremented reference count.
1164 * Otherwise NULL is returned.
1166 * Note, @test is called with the inode_lock held, so can't sleep.
1168 struct inode
*ilookup5(struct super_block
*sb
, unsigned long hashval
,
1169 int (*test
)(struct inode
*, void *), void *data
)
1171 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1173 return ifind(sb
, head
, test
, data
, 1);
1175 EXPORT_SYMBOL(ilookup5
);
1178 * ilookup - search for an inode in the inode cache
1179 * @sb: super block of file system to search
1180 * @ino: inode number to search for
1182 * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache.
1183 * This is for file systems where the inode number is sufficient for unique
1184 * identification of an inode.
1186 * If the inode is in the cache, the inode is returned with an incremented
1189 * Otherwise NULL is returned.
1191 struct inode
*ilookup(struct super_block
*sb
, unsigned long ino
)
1193 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1195 return ifind_fast(sb
, head
, ino
);
1197 EXPORT_SYMBOL(ilookup
);
1200 * iget5_locked - obtain an inode from a mounted file system
1201 * @sb: super block of file system
1202 * @hashval: hash value (usually inode number) to get
1203 * @test: callback used for comparisons between inodes
1204 * @set: callback used to initialize a new struct inode
1205 * @data: opaque data pointer to pass to @test and @set
1207 * iget5_locked() uses ifind() to search for the inode specified by @hashval
1208 * and @data in the inode cache and if present it is returned with an increased
1209 * reference count. This is a generalized version of iget_locked() for file
1210 * systems where the inode number is not sufficient for unique identification
1213 * If the inode is not in cache, get_new_inode() is called to allocate a new
1214 * inode and this is returned locked, hashed, and with the I_NEW flag set. The
1215 * file system gets to fill it in before unlocking it via unlock_new_inode().
1217 * Note both @test and @set are called with the inode_lock held, so can't sleep.
1219 struct inode
*iget5_locked(struct super_block
*sb
, unsigned long hashval
,
1220 int (*test
)(struct inode
*, void *),
1221 int (*set
)(struct inode
*, void *), void *data
)
1223 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1224 struct inode
*inode
;
1226 inode
= ifind(sb
, head
, test
, data
, 1);
1230 * get_new_inode() will do the right thing, re-trying the search
1231 * in case it had to block at any point.
1233 return get_new_inode(sb
, head
, test
, set
, data
);
1235 EXPORT_SYMBOL(iget5_locked
);
1238 * iget_locked - obtain an inode from a mounted file system
1239 * @sb: super block of file system
1240 * @ino: inode number to get
1242 * iget_locked() uses ifind_fast() to search for the inode specified by @ino in
1243 * the inode cache and if present it is returned with an increased reference
1244 * count. This is for file systems where the inode number is sufficient for
1245 * unique identification of an inode.
1247 * If the inode is not in cache, get_new_inode_fast() is called to allocate a
1248 * new inode and this is returned locked, hashed, and with the I_NEW flag set.
1249 * The file system gets to fill it in before unlocking it via
1250 * unlock_new_inode().
1252 struct inode
*iget_locked(struct super_block
*sb
, unsigned long ino
)
1254 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1255 struct inode
*inode
;
1257 inode
= ifind_fast(sb
, head
, ino
);
1261 * get_new_inode_fast() will do the right thing, re-trying the search
1262 * in case it had to block at any point.
1264 return get_new_inode_fast(sb
, head
, ino
);
1266 EXPORT_SYMBOL(iget_locked
);
1268 int insert_inode_locked(struct inode
*inode
)
1270 struct super_block
*sb
= inode
->i_sb
;
1271 ino_t ino
= inode
->i_ino
;
1272 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1274 inode
->i_state
|= I_NEW
;
1276 struct hlist_node
*node
;
1277 struct inode
*old
= NULL
;
1278 spin_lock(&inode_lock
);
1279 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1280 if (old
->i_ino
!= ino
)
1282 if (old
->i_sb
!= sb
)
1284 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
))
1288 if (likely(!node
)) {
1289 hlist_add_head(&inode
->i_hash
, head
);
1290 spin_unlock(&inode_lock
);
1294 spin_unlock(&inode_lock
);
1296 if (unlikely(!inode_unhashed(old
))) {
1303 EXPORT_SYMBOL(insert_inode_locked
);
1305 int insert_inode_locked4(struct inode
*inode
, unsigned long hashval
,
1306 int (*test
)(struct inode
*, void *), void *data
)
1308 struct super_block
*sb
= inode
->i_sb
;
1309 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1311 inode
->i_state
|= I_NEW
;
1314 struct hlist_node
*node
;
1315 struct inode
*old
= NULL
;
1317 spin_lock(&inode_lock
);
1318 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1319 if (old
->i_sb
!= sb
)
1321 if (!test(old
, data
))
1323 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
))
1327 if (likely(!node
)) {
1328 hlist_add_head(&inode
->i_hash
, head
);
1329 spin_unlock(&inode_lock
);
1333 spin_unlock(&inode_lock
);
1335 if (unlikely(!inode_unhashed(old
))) {
1342 EXPORT_SYMBOL(insert_inode_locked4
);
1345 int generic_delete_inode(struct inode
*inode
)
1349 EXPORT_SYMBOL(generic_delete_inode
);
1352 * Normal UNIX filesystem behaviour: delete the
1353 * inode when the usage count drops to zero, and
1356 int generic_drop_inode(struct inode
*inode
)
1358 return !inode
->i_nlink
|| inode_unhashed(inode
);
1360 EXPORT_SYMBOL_GPL(generic_drop_inode
);
1363 * Called when we're dropping the last reference
1366 * Call the FS "drop_inode()" function, defaulting to
1367 * the legacy UNIX filesystem behaviour. If it tells
1368 * us to evict inode, do so. Otherwise, retain inode
1369 * in cache if fs is alive, sync and evict if fs is
1372 static void iput_final(struct inode
*inode
)
1374 struct super_block
*sb
= inode
->i_sb
;
1375 const struct super_operations
*op
= inode
->i_sb
->s_op
;
1378 if (op
&& op
->drop_inode
)
1379 drop
= op
->drop_inode(inode
);
1381 drop
= generic_drop_inode(inode
);
1384 if (sb
->s_flags
& MS_ACTIVE
) {
1385 inode
->i_state
|= I_REFERENCED
;
1386 if (!(inode
->i_state
& (I_DIRTY
|I_SYNC
))) {
1387 inode_lru_list_add(inode
);
1389 spin_unlock(&inode_lock
);
1392 WARN_ON(inode
->i_state
& I_NEW
);
1393 inode
->i_state
|= I_WILL_FREE
;
1394 spin_unlock(&inode_lock
);
1395 write_inode_now(inode
, 1);
1396 spin_lock(&inode_lock
);
1397 WARN_ON(inode
->i_state
& I_NEW
);
1398 inode
->i_state
&= ~I_WILL_FREE
;
1399 __remove_inode_hash(inode
);
1402 WARN_ON(inode
->i_state
& I_NEW
);
1403 inode
->i_state
|= I_FREEING
;
1406 * Move the inode off the IO lists and LRU once I_FREEING is
1407 * set so that it won't get moved back on there if it is dirty.
1409 inode_lru_list_del(inode
);
1410 list_del_init(&inode
->i_wb_list
);
1412 __inode_sb_list_del(inode
);
1413 spin_unlock(&inode_lock
);
1415 remove_inode_hash(inode
);
1416 wake_up_inode(inode
);
1417 BUG_ON(inode
->i_state
!= (I_FREEING
| I_CLEAR
));
1418 destroy_inode(inode
);
1422 * iput - put an inode
1423 * @inode: inode to put
1425 * Puts an inode, dropping its usage count. If the inode use count hits
1426 * zero, the inode is then freed and may also be destroyed.
1428 * Consequently, iput() can sleep.
1430 void iput(struct inode
*inode
)
1433 BUG_ON(inode
->i_state
& I_CLEAR
);
1435 if (atomic_dec_and_lock(&inode
->i_count
, &inode_lock
))
1439 EXPORT_SYMBOL(iput
);
1442 * bmap - find a block number in a file
1443 * @inode: inode of file
1444 * @block: block to find
1446 * Returns the block number on the device holding the inode that
1447 * is the disk block number for the block of the file requested.
1448 * That is, asked for block 4 of inode 1 the function will return the
1449 * disk block relative to the disk start that holds that block of the
1452 sector_t
bmap(struct inode
*inode
, sector_t block
)
1455 if (inode
->i_mapping
->a_ops
->bmap
)
1456 res
= inode
->i_mapping
->a_ops
->bmap(inode
->i_mapping
, block
);
1459 EXPORT_SYMBOL(bmap
);
1462 * With relative atime, only update atime if the previous atime is
1463 * earlier than either the ctime or mtime or if at least a day has
1464 * passed since the last atime update.
1466 static int relatime_need_update(struct vfsmount
*mnt
, struct inode
*inode
,
1467 struct timespec now
)
1470 if (!(mnt
->mnt_flags
& MNT_RELATIME
))
1473 * Is mtime younger than atime? If yes, update atime:
1475 if (timespec_compare(&inode
->i_mtime
, &inode
->i_atime
) >= 0)
1478 * Is ctime younger than atime? If yes, update atime:
1480 if (timespec_compare(&inode
->i_ctime
, &inode
->i_atime
) >= 0)
1484 * Is the previous atime value older than a day? If yes,
1487 if ((long)(now
.tv_sec
- inode
->i_atime
.tv_sec
) >= 24*60*60)
1490 * Good, we can skip the atime update:
1496 * touch_atime - update the access time
1497 * @mnt: mount the inode is accessed on
1498 * @dentry: dentry accessed
1500 * Update the accessed time on an inode and mark it for writeback.
1501 * This function automatically handles read only file systems and media,
1502 * as well as the "noatime" flag and inode specific "noatime" markers.
1504 void touch_atime(struct vfsmount
*mnt
, struct dentry
*dentry
)
1506 struct inode
*inode
= dentry
->d_inode
;
1507 struct timespec now
;
1509 if (inode
->i_flags
& S_NOATIME
)
1511 if (IS_NOATIME(inode
))
1513 if ((inode
->i_sb
->s_flags
& MS_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1516 if (mnt
->mnt_flags
& MNT_NOATIME
)
1518 if ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1521 now
= current_fs_time(inode
->i_sb
);
1523 if (!relatime_need_update(mnt
, inode
, now
))
1526 if (timespec_equal(&inode
->i_atime
, &now
))
1529 if (mnt_want_write(mnt
))
1532 inode
->i_atime
= now
;
1533 mark_inode_dirty_sync(inode
);
1534 mnt_drop_write(mnt
);
1536 EXPORT_SYMBOL(touch_atime
);
1539 * file_update_time - update mtime and ctime time
1540 * @file: file accessed
1542 * Update the mtime and ctime members of an inode and mark the inode
1543 * for writeback. Note that this function is meant exclusively for
1544 * usage in the file write path of filesystems, and filesystems may
1545 * choose to explicitly ignore update via this function with the
1546 * S_NOCMTIME inode flag, e.g. for network filesystem where these
1547 * timestamps are handled by the server.
1550 void file_update_time(struct file
*file
)
1552 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1553 struct timespec now
;
1554 enum { S_MTIME
= 1, S_CTIME
= 2, S_VERSION
= 4 } sync_it
= 0;
1556 /* First try to exhaust all avenues to not sync */
1557 if (IS_NOCMTIME(inode
))
1560 now
= current_fs_time(inode
->i_sb
);
1561 if (!timespec_equal(&inode
->i_mtime
, &now
))
1564 if (!timespec_equal(&inode
->i_ctime
, &now
))
1567 if (IS_I_VERSION(inode
))
1568 sync_it
|= S_VERSION
;
1573 /* Finally allowed to write? Takes lock. */
1574 if (mnt_want_write_file(file
))
1577 /* Only change inode inside the lock region */
1578 if (sync_it
& S_VERSION
)
1579 inode_inc_iversion(inode
);
1580 if (sync_it
& S_CTIME
)
1581 inode
->i_ctime
= now
;
1582 if (sync_it
& S_MTIME
)
1583 inode
->i_mtime
= now
;
1584 mark_inode_dirty_sync(inode
);
1585 mnt_drop_write(file
->f_path
.mnt
);
1587 EXPORT_SYMBOL(file_update_time
);
1589 int inode_needs_sync(struct inode
*inode
)
1593 if (S_ISDIR(inode
->i_mode
) && IS_DIRSYNC(inode
))
1597 EXPORT_SYMBOL(inode_needs_sync
);
1599 int inode_wait(void *word
)
1604 EXPORT_SYMBOL(inode_wait
);
1607 * If we try to find an inode in the inode hash while it is being
1608 * deleted, we have to wait until the filesystem completes its
1609 * deletion before reporting that it isn't found. This function waits
1610 * until the deletion _might_ have completed. Callers are responsible
1611 * to recheck inode state.
1613 * It doesn't matter if I_NEW is not set initially, a call to
1614 * wake_up_inode() after removing from the hash list will DTRT.
1616 * This is called with inode_lock held.
1618 static void __wait_on_freeing_inode(struct inode
*inode
)
1620 wait_queue_head_t
*wq
;
1621 DEFINE_WAIT_BIT(wait
, &inode
->i_state
, __I_NEW
);
1622 wq
= bit_waitqueue(&inode
->i_state
, __I_NEW
);
1623 prepare_to_wait(wq
, &wait
.wait
, TASK_UNINTERRUPTIBLE
);
1624 spin_unlock(&inode_lock
);
1626 finish_wait(wq
, &wait
.wait
);
1627 spin_lock(&inode_lock
);
1630 static __initdata
unsigned long ihash_entries
;
1631 static int __init
set_ihash_entries(char *str
)
1635 ihash_entries
= simple_strtoul(str
, &str
, 0);
1638 __setup("ihash_entries=", set_ihash_entries
);
1641 * Initialize the waitqueues and inode hash table.
1643 void __init
inode_init_early(void)
1647 /* If hashes are distributed across NUMA nodes, defer
1648 * hash allocation until vmalloc space is available.
1654 alloc_large_system_hash("Inode-cache",
1655 sizeof(struct hlist_head
),
1663 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1664 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1667 void __init
inode_init(void)
1671 /* inode slab cache */
1672 inode_cachep
= kmem_cache_create("inode_cache",
1673 sizeof(struct inode
),
1675 (SLAB_RECLAIM_ACCOUNT
|SLAB_PANIC
|
1678 register_shrinker(&icache_shrinker
);
1680 /* Hash may have been set up in inode_init_early */
1685 alloc_large_system_hash("Inode-cache",
1686 sizeof(struct hlist_head
),
1694 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1695 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1698 void init_special_inode(struct inode
*inode
, umode_t mode
, dev_t rdev
)
1700 inode
->i_mode
= mode
;
1701 if (S_ISCHR(mode
)) {
1702 inode
->i_fop
= &def_chr_fops
;
1703 inode
->i_rdev
= rdev
;
1704 } else if (S_ISBLK(mode
)) {
1705 inode
->i_fop
= &def_blk_fops
;
1706 inode
->i_rdev
= rdev
;
1707 } else if (S_ISFIFO(mode
))
1708 inode
->i_fop
= &def_fifo_fops
;
1709 else if (S_ISSOCK(mode
))
1710 inode
->i_fop
= &bad_sock_fops
;
1712 printk(KERN_DEBUG
"init_special_inode: bogus i_mode (%o) for"
1713 " inode %s:%lu\n", mode
, inode
->i_sb
->s_id
,
1716 EXPORT_SYMBOL(init_special_inode
);
1719 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
1721 * @dir: Directory inode
1722 * @mode: mode of the new inode
1724 void inode_init_owner(struct inode
*inode
, const struct inode
*dir
,
1727 inode
->i_uid
= current_fsuid();
1728 if (dir
&& dir
->i_mode
& S_ISGID
) {
1729 inode
->i_gid
= dir
->i_gid
;
1733 inode
->i_gid
= current_fsgid();
1734 inode
->i_mode
= mode
;
1736 EXPORT_SYMBOL(inode_init_owner
);
1739 * return true if current either has CAP_FOWNER to the
1740 * file, or owns the file.
1742 bool is_owner_or_cap(const struct inode
*inode
)
1744 struct user_namespace
*ns
= inode_userns(inode
);
1746 if (current_user_ns() == ns
&& current_fsuid() == inode
->i_uid
)
1748 if (ns_capable(ns
, CAP_FOWNER
))
1752 EXPORT_SYMBOL(is_owner_or_cap
);