2 * (C) 1997 Linus Torvalds
3 * (C) 1999 Andrea Arcangeli <andrea@suse.de> (dynamic inode allocation)
5 #include <linux/export.h>
8 #include <linux/backing-dev.h>
9 #include <linux/hash.h>
10 #include <linux/swap.h>
11 #include <linux/security.h>
12 #include <linux/cdev.h>
13 #include <linux/bootmem.h>
14 #include <linux/fsnotify.h>
15 #include <linux/mount.h>
16 #include <linux/posix_acl.h>
17 #include <linux/prefetch.h>
18 #include <linux/buffer_head.h> /* for inode_has_buffers */
19 #include <linux/ratelimit.h>
23 * Inode locking rules:
25 * inode->i_lock protects:
26 * inode->i_state, inode->i_hash, __iget()
27 * inode->i_sb->s_inode_lru_lock protects:
28 * inode->i_sb->s_inode_lru, inode->i_lru
29 * inode_sb_list_lock protects:
30 * sb->s_inodes, inode->i_sb_list
31 * bdi->wb.list_lock protects:
32 * bdi->wb.b_{dirty,io,more_io}, inode->i_wb_list
33 * inode_hash_lock protects:
34 * inode_hashtable, inode->i_hash
40 * inode->i_sb->s_inode_lru_lock
53 static unsigned int i_hash_mask __read_mostly;
54 static unsigned int i_hash_shift __read_mostly;
55 static struct hlist_head *inode_hashtable __read_mostly;
56 static __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_hash_lock);
58 __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_sb_list_lock);
61 * Empty aops. Can be used for the cases where the user does not
62 * define any of the address_space operations.
64 const struct address_space_operations empty_aops = {
66 EXPORT_SYMBOL(empty_aops);
69 * Statistics gathering..
71 struct inodes_stat_t inodes_stat;
73 static DEFINE_PER_CPU(unsigned int, nr_inodes);
74 static DEFINE_PER_CPU(unsigned int, nr_unused);
76 static struct kmem_cache *inode_cachep __read_mostly;
78 static int get_nr_inodes(void)
82 for_each_possible_cpu(i)
83 sum += per_cpu(nr_inodes, i);
84 return sum < 0 ? 0 : sum;
87 static inline int get_nr_inodes_unused(void)
91 for_each_possible_cpu(i)
92 sum += per_cpu(nr_unused, i);
93 return sum < 0 ? 0 : sum;
96 int get_nr_dirty_inodes(void)
98 /* not actually dirty inodes, but a wild approximation */
99 int nr_dirty = get_nr_inodes() - get_nr_inodes_unused();
100 return nr_dirty > 0 ? nr_dirty : 0;
104 * Handle nr_inode sysctl
107 int proc_nr_inodes(ctl_table *table, int write,
108 void __user *buffer, size_t *lenp, loff_t *ppos)
110 inodes_stat.nr_inodes = get_nr_inodes();
111 inodes_stat.nr_unused = get_nr_inodes_unused();
112 return proc_dointvec(table, write, buffer, lenp, ppos);
117 * inode_init_always - perform inode structure intialisation
118 * @sb: superblock inode belongs to
119 * @inode: inode to initialise
121 * These are initializations that need to be done on every inode
122 * allocation as the fields are not initialised by slab allocation.
124 int inode_init_always(struct super_block *sb, struct inode *inode)
126 static const struct inode_operations empty_iops;
127 static const struct file_operations empty_fops;
128 struct address_space *const mapping = &inode->i_data;
131 inode->i_blkbits = sb->s_blocksize_bits;
133 atomic_set(&inode->i_count, 1);
134 inode->i_op = &empty_iops;
135 inode->i_fop = &empty_fops;
136 inode->__i_nlink = 1;
137 inode->i_opflags = 0;
138 i_uid_write(inode, 0);
139 i_gid_write(inode, 0);
140 atomic_set(&inode->i_writecount, 0);
144 inode->i_generation = 0;
146 memset(&inode->i_dquot, 0, sizeof(inode->i_dquot));
148 inode->i_pipe = NULL;
149 inode->i_bdev = NULL;
150 inode->i_cdev = NULL;
152 inode->dirtied_when = 0;
154 if (security_inode_alloc(inode))
156 spin_lock_init(&inode->i_lock);
157 lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key);
159 mutex_init(&inode->i_mutex);
160 lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key);
162 atomic_set(&inode->i_dio_count, 0);
164 mapping->a_ops = &empty_aops;
165 mapping->host = inode;
167 mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE);
168 mapping->assoc_mapping = NULL;
169 mapping->backing_dev_info = &default_backing_dev_info;
170 mapping->writeback_index = 0;
173 * If the block_device provides a backing_dev_info for client
174 * inodes then use that. Otherwise the inode share the bdev's
178 struct backing_dev_info *bdi;
180 bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
181 mapping->backing_dev_info = bdi;
183 inode->i_private = NULL;
184 inode->i_mapping = mapping;
185 INIT_LIST_HEAD(&inode->i_dentry); /* buggered by rcu freeing */
186 #ifdef CONFIG_FS_POSIX_ACL
187 inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED;
190 #ifdef CONFIG_FSNOTIFY
191 inode->i_fsnotify_mask = 0;
194 this_cpu_inc(nr_inodes);
200 EXPORT_SYMBOL(inode_init_always);
202 static struct inode *alloc_inode(struct super_block *sb)
206 if (sb->s_op->alloc_inode)
207 inode = sb->s_op->alloc_inode(sb);
209 inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL);
214 if (unlikely(inode_init_always(sb, inode))) {
215 if (inode->i_sb->s_op->destroy_inode)
216 inode->i_sb->s_op->destroy_inode(inode);
218 kmem_cache_free(inode_cachep, inode);
225 void free_inode_nonrcu(struct inode *inode)
227 kmem_cache_free(inode_cachep, inode);
229 EXPORT_SYMBOL(free_inode_nonrcu);
231 void __destroy_inode(struct inode *inode)
233 BUG_ON(inode_has_buffers(inode));
234 security_inode_free(inode);
235 fsnotify_inode_delete(inode);
236 if (!inode->i_nlink) {
237 WARN_ON(atomic_long_read(&inode->i_sb->s_remove_count) == 0);
238 atomic_long_dec(&inode->i_sb->s_remove_count);
241 #ifdef CONFIG_FS_POSIX_ACL
242 if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED)
243 posix_acl_release(inode->i_acl);
244 if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED)
245 posix_acl_release(inode->i_default_acl);
247 this_cpu_dec(nr_inodes);
249 EXPORT_SYMBOL(__destroy_inode);
251 static void i_callback(struct rcu_head *head)
253 struct inode *inode = container_of(head, struct inode, i_rcu);
254 kmem_cache_free(inode_cachep, inode);
257 static void destroy_inode(struct inode *inode)
259 BUG_ON(!list_empty(&inode->i_lru));
260 __destroy_inode(inode);
261 if (inode->i_sb->s_op->destroy_inode)
262 inode->i_sb->s_op->destroy_inode(inode);
264 call_rcu(&inode->i_rcu, i_callback);
268 * drop_nlink - directly drop an inode's link count
271 * This is a low-level filesystem helper to replace any
272 * direct filesystem manipulation of i_nlink. In cases
273 * where we are attempting to track writes to the
274 * filesystem, a decrement to zero means an imminent
275 * write when the file is truncated and actually unlinked
278 void drop_nlink(struct inode *inode)
280 WARN_ON(inode->i_nlink == 0);
283 atomic_long_inc(&inode->i_sb->s_remove_count);
285 EXPORT_SYMBOL(drop_nlink);
288 * clear_nlink - directly zero an inode's link count
291 * This is a low-level filesystem helper to replace any
292 * direct filesystem manipulation of i_nlink. See
293 * drop_nlink() for why we care about i_nlink hitting zero.
295 void clear_nlink(struct inode *inode)
297 if (inode->i_nlink) {
298 inode->__i_nlink = 0;
299 atomic_long_inc(&inode->i_sb->s_remove_count);
302 EXPORT_SYMBOL(clear_nlink);
305 * set_nlink - directly set an inode's link count
307 * @nlink: new nlink (should be non-zero)
309 * This is a low-level filesystem helper to replace any
310 * direct filesystem manipulation of i_nlink.
312 void set_nlink(struct inode *inode, unsigned int nlink)
317 /* Yes, some filesystems do change nlink from zero to one */
318 if (inode->i_nlink == 0)
319 atomic_long_dec(&inode->i_sb->s_remove_count);
321 inode->__i_nlink = nlink;
324 EXPORT_SYMBOL(set_nlink);
327 * inc_nlink - directly increment an inode's link count
330 * This is a low-level filesystem helper to replace any
331 * direct filesystem manipulation of i_nlink. Currently,
332 * it is only here for parity with dec_nlink().
334 void inc_nlink(struct inode *inode)
336 if (WARN_ON(inode->i_nlink == 0))
337 atomic_long_dec(&inode->i_sb->s_remove_count);
341 EXPORT_SYMBOL(inc_nlink);
343 void address_space_init_once(struct address_space *mapping)
345 memset(mapping, 0, sizeof(*mapping));
346 INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC);
347 spin_lock_init(&mapping->tree_lock);
348 mutex_init(&mapping->i_mmap_mutex);
349 INIT_LIST_HEAD(&mapping->private_list);
350 spin_lock_init(&mapping->private_lock);
351 INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap);
352 INIT_LIST_HEAD(&mapping->i_mmap_nonlinear);
354 EXPORT_SYMBOL(address_space_init_once);
357 * These are initializations that only need to be done
358 * once, because the fields are idempotent across use
359 * of the inode, so let the slab aware of that.
361 void inode_init_once(struct inode *inode)
363 memset(inode, 0, sizeof(*inode));
364 INIT_HLIST_NODE(&inode->i_hash);
365 INIT_LIST_HEAD(&inode->i_devices);
366 INIT_LIST_HEAD(&inode->i_wb_list);
367 INIT_LIST_HEAD(&inode->i_lru);
368 address_space_init_once(&inode->i_data);
369 i_size_ordered_init(inode);
370 #ifdef CONFIG_FSNOTIFY
371 INIT_HLIST_HEAD(&inode->i_fsnotify_marks);
374 EXPORT_SYMBOL(inode_init_once);
376 static void init_once(void *foo)
378 struct inode *inode = (struct inode *) foo;
380 inode_init_once(inode);
384 * inode->i_lock must be held
386 void __iget(struct inode *inode)
388 atomic_inc(&inode->i_count);
392 * get additional reference to inode; caller must already hold one.
394 void ihold(struct inode *inode)
396 WARN_ON(atomic_inc_return(&inode->i_count) < 2);
398 EXPORT_SYMBOL(ihold);
400 static void inode_lru_list_add(struct inode *inode)
402 spin_lock(&inode->i_sb->s_inode_lru_lock);
403 if (list_empty(&inode->i_lru)) {
404 list_add(&inode->i_lru, &inode->i_sb->s_inode_lru);
405 inode->i_sb->s_nr_inodes_unused++;
406 this_cpu_inc(nr_unused);
408 spin_unlock(&inode->i_sb->s_inode_lru_lock);
411 static void inode_lru_list_del(struct inode *inode)
413 spin_lock(&inode->i_sb->s_inode_lru_lock);
414 if (!list_empty(&inode->i_lru)) {
415 list_del_init(&inode->i_lru);
416 inode->i_sb->s_nr_inodes_unused--;
417 this_cpu_dec(nr_unused);
419 spin_unlock(&inode->i_sb->s_inode_lru_lock);
423 * inode_sb_list_add - add inode to the superblock list of inodes
424 * @inode: inode to add
426 void inode_sb_list_add(struct inode *inode)
428 spin_lock(&inode_sb_list_lock);
429 list_add(&inode->i_sb_list, &inode->i_sb->s_inodes);
430 spin_unlock(&inode_sb_list_lock);
432 EXPORT_SYMBOL_GPL(inode_sb_list_add);
434 static inline void inode_sb_list_del(struct inode *inode)
436 if (!list_empty(&inode->i_sb_list)) {
437 spin_lock(&inode_sb_list_lock);
438 list_del_init(&inode->i_sb_list);
439 spin_unlock(&inode_sb_list_lock);
443 static unsigned long hash(struct super_block *sb, unsigned long hashval)
447 tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) /
449 tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> i_hash_shift);
450 return tmp & i_hash_mask;
454 * __insert_inode_hash - hash an inode
455 * @inode: unhashed inode
456 * @hashval: unsigned long value used to locate this object in the
459 * Add an inode to the inode hash for this superblock.
461 void __insert_inode_hash(struct inode *inode, unsigned long hashval)
463 struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval);
465 spin_lock(&inode_hash_lock);
466 spin_lock(&inode->i_lock);
467 hlist_add_head(&inode->i_hash, b);
468 spin_unlock(&inode->i_lock);
469 spin_unlock(&inode_hash_lock);
471 EXPORT_SYMBOL(__insert_inode_hash);
474 * __remove_inode_hash - remove an inode from the hash
475 * @inode: inode to unhash
477 * Remove an inode from the superblock.
479 void __remove_inode_hash(struct inode *inode)
481 spin_lock(&inode_hash_lock);
482 spin_lock(&inode->i_lock);
483 hlist_del_init(&inode->i_hash);
484 spin_unlock(&inode->i_lock);
485 spin_unlock(&inode_hash_lock);
487 EXPORT_SYMBOL(__remove_inode_hash);
489 void end_writeback(struct inode *inode)
493 * We have to cycle tree_lock here because reclaim can be still in the
494 * process of removing the last page (in __delete_from_page_cache())
495 * and we must not free mapping under it.
497 spin_lock_irq(&inode->i_data.tree_lock);
498 BUG_ON(inode->i_data.nrpages);
499 spin_unlock_irq(&inode->i_data.tree_lock);
500 BUG_ON(!list_empty(&inode->i_data.private_list));
501 BUG_ON(!(inode->i_state & I_FREEING));
502 BUG_ON(inode->i_state & I_CLEAR);
503 inode_sync_wait(inode);
504 /* don't need i_lock here, no concurrent mods to i_state */
505 inode->i_state = I_FREEING | I_CLEAR;
507 EXPORT_SYMBOL(end_writeback);
510 * Free the inode passed in, removing it from the lists it is still connected
511 * to. We remove any pages still attached to the inode and wait for any IO that
512 * is still in progress before finally destroying the inode.
514 * An inode must already be marked I_FREEING so that we avoid the inode being
515 * moved back onto lists if we race with other code that manipulates the lists
516 * (e.g. writeback_single_inode). The caller is responsible for setting this.
518 * An inode must already be removed from the LRU list before being evicted from
519 * the cache. This should occur atomically with setting the I_FREEING state
520 * flag, so no inodes here should ever be on the LRU when being evicted.
522 static void evict(struct inode *inode)
524 const struct super_operations *op = inode->i_sb->s_op;
526 BUG_ON(!(inode->i_state & I_FREEING));
527 BUG_ON(!list_empty(&inode->i_lru));
529 if (!list_empty(&inode->i_wb_list))
530 inode_wb_list_del(inode);
532 inode_sb_list_del(inode);
534 if (op->evict_inode) {
535 op->evict_inode(inode);
537 if (inode->i_data.nrpages)
538 truncate_inode_pages(&inode->i_data, 0);
539 end_writeback(inode);
541 if (S_ISBLK(inode->i_mode) && inode->i_bdev)
543 if (S_ISCHR(inode->i_mode) && inode->i_cdev)
546 remove_inode_hash(inode);
548 spin_lock(&inode->i_lock);
549 wake_up_bit(&inode->i_state, __I_NEW);
550 BUG_ON(inode->i_state != (I_FREEING | I_CLEAR));
551 spin_unlock(&inode->i_lock);
553 destroy_inode(inode);
557 * dispose_list - dispose of the contents of a local list
558 * @head: the head of the list to free
560 * Dispose-list gets a local list with local inodes in it, so it doesn't
561 * need to worry about list corruption and SMP locks.
563 static void dispose_list(struct list_head *head)
565 while (!list_empty(head)) {
568 inode = list_first_entry(head, struct inode, i_lru);
569 list_del_init(&inode->i_lru);
576 * evict_inodes - evict all evictable inodes for a superblock
577 * @sb: superblock to operate on
579 * Make sure that no inodes with zero refcount are retained. This is
580 * called by superblock shutdown after having MS_ACTIVE flag removed,
581 * so any inode reaching zero refcount during or after that call will
582 * be immediately evicted.
584 void evict_inodes(struct super_block *sb)
586 struct inode *inode, *next;
589 spin_lock(&inode_sb_list_lock);
590 list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
591 if (atomic_read(&inode->i_count))
594 spin_lock(&inode->i_lock);
595 if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
596 spin_unlock(&inode->i_lock);
600 inode->i_state |= I_FREEING;
601 inode_lru_list_del(inode);
602 spin_unlock(&inode->i_lock);
603 list_add(&inode->i_lru, &dispose);
605 spin_unlock(&inode_sb_list_lock);
607 dispose_list(&dispose);
611 * invalidate_inodes - attempt to free all inodes on a superblock
612 * @sb: superblock to operate on
613 * @kill_dirty: flag to guide handling of dirty inodes
615 * Attempts to free all inodes for a given superblock. If there were any
616 * busy inodes return a non-zero value, else zero.
617 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
620 int invalidate_inodes(struct super_block *sb, bool kill_dirty)
623 struct inode *inode, *next;
626 spin_lock(&inode_sb_list_lock);
627 list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
628 spin_lock(&inode->i_lock);
629 if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
630 spin_unlock(&inode->i_lock);
633 if (inode->i_state & I_DIRTY && !kill_dirty) {
634 spin_unlock(&inode->i_lock);
638 if (atomic_read(&inode->i_count)) {
639 spin_unlock(&inode->i_lock);
644 inode->i_state |= I_FREEING;
645 inode_lru_list_del(inode);
646 spin_unlock(&inode->i_lock);
647 list_add(&inode->i_lru, &dispose);
649 spin_unlock(&inode_sb_list_lock);
651 dispose_list(&dispose);
656 static int can_unuse(struct inode *inode)
658 if (inode->i_state & ~I_REFERENCED)
660 if (inode_has_buffers(inode))
662 if (atomic_read(&inode->i_count))
664 if (inode->i_data.nrpages)
670 * Walk the superblock inode LRU for freeable inodes and attempt to free them.
671 * This is called from the superblock shrinker function with a number of inodes
672 * to trim from the LRU. Inodes to be freed are moved to a temporary list and
673 * then are freed outside inode_lock by dispose_list().
675 * Any inodes which are pinned purely because of attached pagecache have their
676 * pagecache removed. If the inode has metadata buffers attached to
677 * mapping->private_list then try to remove them.
679 * If the inode has the I_REFERENCED flag set, then it means that it has been
680 * used recently - the flag is set in iput_final(). When we encounter such an
681 * inode, clear the flag and move it to the back of the LRU so it gets another
682 * pass through the LRU before it gets reclaimed. This is necessary because of
683 * the fact we are doing lazy LRU updates to minimise lock contention so the
684 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
685 * with this flag set because they are the inodes that are out of order.
687 void prune_icache_sb(struct super_block *sb, int nr_to_scan)
691 unsigned long reap = 0;
693 spin_lock(&sb->s_inode_lru_lock);
694 for (nr_scanned = nr_to_scan; nr_scanned >= 0; nr_scanned--) {
697 if (list_empty(&sb->s_inode_lru))
700 inode = list_entry(sb->s_inode_lru.prev, struct inode, i_lru);
703 * we are inverting the sb->s_inode_lru_lock/inode->i_lock here,
704 * so use a trylock. If we fail to get the lock, just move the
705 * inode to the back of the list so we don't spin on it.
707 if (!spin_trylock(&inode->i_lock)) {
708 list_move_tail(&inode->i_lru, &sb->s_inode_lru);
713 * Referenced or dirty inodes are still in use. Give them
714 * another pass through the LRU as we canot reclaim them now.
716 if (atomic_read(&inode->i_count) ||
717 (inode->i_state & ~I_REFERENCED)) {
718 list_del_init(&inode->i_lru);
719 spin_unlock(&inode->i_lock);
720 sb->s_nr_inodes_unused--;
721 this_cpu_dec(nr_unused);
725 /* recently referenced inodes get one more pass */
726 if (inode->i_state & I_REFERENCED) {
727 inode->i_state &= ~I_REFERENCED;
728 list_move(&inode->i_lru, &sb->s_inode_lru);
729 spin_unlock(&inode->i_lock);
732 if (inode_has_buffers(inode) || inode->i_data.nrpages) {
734 spin_unlock(&inode->i_lock);
735 spin_unlock(&sb->s_inode_lru_lock);
736 if (remove_inode_buffers(inode))
737 reap += invalidate_mapping_pages(&inode->i_data,
740 spin_lock(&sb->s_inode_lru_lock);
742 if (inode != list_entry(sb->s_inode_lru.next,
743 struct inode, i_lru))
744 continue; /* wrong inode or list_empty */
745 /* avoid lock inversions with trylock */
746 if (!spin_trylock(&inode->i_lock))
748 if (!can_unuse(inode)) {
749 spin_unlock(&inode->i_lock);
753 WARN_ON(inode->i_state & I_NEW);
754 inode->i_state |= I_FREEING;
755 spin_unlock(&inode->i_lock);
757 list_move(&inode->i_lru, &freeable);
758 sb->s_nr_inodes_unused--;
759 this_cpu_dec(nr_unused);
761 if (current_is_kswapd())
762 __count_vm_events(KSWAPD_INODESTEAL, reap);
764 __count_vm_events(PGINODESTEAL, reap);
765 spin_unlock(&sb->s_inode_lru_lock);
766 if (current->reclaim_state)
767 current->reclaim_state->reclaimed_slab += reap;
769 dispose_list(&freeable);
772 static void __wait_on_freeing_inode(struct inode *inode);
774 * Called with the inode lock held.
776 static struct inode *find_inode(struct super_block *sb,
777 struct hlist_head *head,
778 int (*test)(struct inode *, void *),
781 struct hlist_node *node;
782 struct inode *inode = NULL;
785 hlist_for_each_entry(inode, node, head, i_hash) {
786 spin_lock(&inode->i_lock);
787 if (inode->i_sb != sb) {
788 spin_unlock(&inode->i_lock);
791 if (!test(inode, data)) {
792 spin_unlock(&inode->i_lock);
795 if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
796 __wait_on_freeing_inode(inode);
800 spin_unlock(&inode->i_lock);
807 * find_inode_fast is the fast path version of find_inode, see the comment at
808 * iget_locked for details.
810 static struct inode *find_inode_fast(struct super_block *sb,
811 struct hlist_head *head, unsigned long ino)
813 struct hlist_node *node;
814 struct inode *inode = NULL;
817 hlist_for_each_entry(inode, node, head, i_hash) {
818 spin_lock(&inode->i_lock);
819 if (inode->i_ino != ino) {
820 spin_unlock(&inode->i_lock);
823 if (inode->i_sb != sb) {
824 spin_unlock(&inode->i_lock);
827 if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
828 __wait_on_freeing_inode(inode);
832 spin_unlock(&inode->i_lock);
839 * Each cpu owns a range of LAST_INO_BATCH numbers.
840 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
841 * to renew the exhausted range.
843 * This does not significantly increase overflow rate because every CPU can
844 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
845 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
846 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
847 * overflow rate by 2x, which does not seem too significant.
849 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
850 * error if st_ino won't fit in target struct field. Use 32bit counter
851 * here to attempt to avoid that.
853 #define LAST_INO_BATCH 1024
854 static DEFINE_PER_CPU(unsigned int, last_ino);
856 unsigned int get_next_ino(void)
858 unsigned int *p = &get_cpu_var(last_ino);
859 unsigned int res = *p;
862 if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) {
863 static atomic_t shared_last_ino;
864 int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino);
866 res = next - LAST_INO_BATCH;
871 put_cpu_var(last_ino);
874 EXPORT_SYMBOL(get_next_ino);
877 * new_inode_pseudo - obtain an inode
880 * Allocates a new inode for given superblock.
881 * Inode wont be chained in superblock s_inodes list
883 * - fs can't be unmount
884 * - quotas, fsnotify, writeback can't work
886 struct inode *new_inode_pseudo(struct super_block *sb)
888 struct inode *inode = alloc_inode(sb);
891 spin_lock(&inode->i_lock);
893 spin_unlock(&inode->i_lock);
894 INIT_LIST_HEAD(&inode->i_sb_list);
900 * new_inode - obtain an inode
903 * Allocates a new inode for given superblock. The default gfp_mask
904 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
905 * If HIGHMEM pages are unsuitable or it is known that pages allocated
906 * for the page cache are not reclaimable or migratable,
907 * mapping_set_gfp_mask() must be called with suitable flags on the
908 * newly created inode's mapping
911 struct inode *new_inode(struct super_block *sb)
915 spin_lock_prefetch(&inode_sb_list_lock);
917 inode = new_inode_pseudo(sb);
919 inode_sb_list_add(inode);
922 EXPORT_SYMBOL(new_inode);
924 #ifdef CONFIG_DEBUG_LOCK_ALLOC
925 void lockdep_annotate_inode_mutex_key(struct inode *inode)
927 if (S_ISDIR(inode->i_mode)) {
928 struct file_system_type *type = inode->i_sb->s_type;
930 /* Set new key only if filesystem hasn't already changed it */
931 if (lockdep_match_class(&inode->i_mutex, &type->i_mutex_key)) {
933 * ensure nobody is actually holding i_mutex
935 mutex_destroy(&inode->i_mutex);
936 mutex_init(&inode->i_mutex);
937 lockdep_set_class(&inode->i_mutex,
938 &type->i_mutex_dir_key);
942 EXPORT_SYMBOL(lockdep_annotate_inode_mutex_key);
946 * unlock_new_inode - clear the I_NEW state and wake up any waiters
947 * @inode: new inode to unlock
949 * Called when the inode is fully initialised to clear the new state of the
950 * inode and wake up anyone waiting for the inode to finish initialisation.
952 void unlock_new_inode(struct inode *inode)
954 lockdep_annotate_inode_mutex_key(inode);
955 spin_lock(&inode->i_lock);
956 WARN_ON(!(inode->i_state & I_NEW));
957 inode->i_state &= ~I_NEW;
959 wake_up_bit(&inode->i_state, __I_NEW);
960 spin_unlock(&inode->i_lock);
962 EXPORT_SYMBOL(unlock_new_inode);
965 * iget5_locked - obtain an inode from a mounted file system
966 * @sb: super block of file system
967 * @hashval: hash value (usually inode number) to get
968 * @test: callback used for comparisons between inodes
969 * @set: callback used to initialize a new struct inode
970 * @data: opaque data pointer to pass to @test and @set
972 * Search for the inode specified by @hashval and @data in the inode cache,
973 * and if present it is return it with an increased reference count. This is
974 * a generalized version of iget_locked() for file systems where the inode
975 * number is not sufficient for unique identification of an inode.
977 * If the inode is not in cache, allocate a new inode and return it locked,
978 * hashed, and with the I_NEW flag set. The file system gets to fill it in
979 * before unlocking it via unlock_new_inode().
981 * Note both @test and @set are called with the inode_hash_lock held, so can't
984 struct inode *iget5_locked(struct super_block *sb, unsigned long hashval,
985 int (*test)(struct inode *, void *),
986 int (*set)(struct inode *, void *), void *data)
988 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
991 spin_lock(&inode_hash_lock);
992 inode = find_inode(sb, head, test, data);
993 spin_unlock(&inode_hash_lock);
996 wait_on_inode(inode);
1000 inode = alloc_inode(sb);
1004 spin_lock(&inode_hash_lock);
1005 /* We released the lock, so.. */
1006 old = find_inode(sb, head, test, data);
1008 if (set(inode, data))
1011 spin_lock(&inode->i_lock);
1012 inode->i_state = I_NEW;
1013 hlist_add_head(&inode->i_hash, head);
1014 spin_unlock(&inode->i_lock);
1015 inode_sb_list_add(inode);
1016 spin_unlock(&inode_hash_lock);
1018 /* Return the locked inode with I_NEW set, the
1019 * caller is responsible for filling in the contents
1025 * Uhhuh, somebody else created the same inode under
1026 * us. Use the old inode instead of the one we just
1029 spin_unlock(&inode_hash_lock);
1030 destroy_inode(inode);
1032 wait_on_inode(inode);
1037 spin_unlock(&inode_hash_lock);
1038 destroy_inode(inode);
1041 EXPORT_SYMBOL(iget5_locked);
1044 * iget_locked - obtain an inode from a mounted file system
1045 * @sb: super block of file system
1046 * @ino: inode number to get
1048 * Search for the inode specified by @ino in the inode cache and if present
1049 * return it with an increased reference count. This is for file systems
1050 * where the inode number is sufficient for unique identification of an inode.
1052 * If the inode is not in cache, allocate a new inode and return it locked,
1053 * hashed, and with the I_NEW flag set. The file system gets to fill it in
1054 * before unlocking it via unlock_new_inode().
1056 struct inode *iget_locked(struct super_block *sb, unsigned long ino)
1058 struct hlist_head *head = inode_hashtable + hash(sb, ino);
1059 struct inode *inode;
1061 spin_lock(&inode_hash_lock);
1062 inode = find_inode_fast(sb, head, ino);
1063 spin_unlock(&inode_hash_lock);
1065 wait_on_inode(inode);
1069 inode = alloc_inode(sb);
1073 spin_lock(&inode_hash_lock);
1074 /* We released the lock, so.. */
1075 old = find_inode_fast(sb, head, ino);
1078 spin_lock(&inode->i_lock);
1079 inode->i_state = I_NEW;
1080 hlist_add_head(&inode->i_hash, head);
1081 spin_unlock(&inode->i_lock);
1082 inode_sb_list_add(inode);
1083 spin_unlock(&inode_hash_lock);
1085 /* Return the locked inode with I_NEW set, the
1086 * caller is responsible for filling in the contents
1092 * Uhhuh, somebody else created the same inode under
1093 * us. Use the old inode instead of the one we just
1096 spin_unlock(&inode_hash_lock);
1097 destroy_inode(inode);
1099 wait_on_inode(inode);
1103 EXPORT_SYMBOL(iget_locked);
1106 * search the inode cache for a matching inode number.
1107 * If we find one, then the inode number we are trying to
1108 * allocate is not unique and so we should not use it.
1110 * Returns 1 if the inode number is unique, 0 if it is not.
1112 static int test_inode_iunique(struct super_block *sb, unsigned long ino)
1114 struct hlist_head *b = inode_hashtable + hash(sb, ino);
1115 struct hlist_node *node;
1116 struct inode *inode;
1118 spin_lock(&inode_hash_lock);
1119 hlist_for_each_entry(inode, node, b, i_hash) {
1120 if (inode->i_ino == ino && inode->i_sb == sb) {
1121 spin_unlock(&inode_hash_lock);
1125 spin_unlock(&inode_hash_lock);
1131 * iunique - get a unique inode number
1133 * @max_reserved: highest reserved inode number
1135 * Obtain an inode number that is unique on the system for a given
1136 * superblock. This is used by file systems that have no natural
1137 * permanent inode numbering system. An inode number is returned that
1138 * is higher than the reserved limit but unique.
1141 * With a large number of inodes live on the file system this function
1142 * currently becomes quite slow.
1144 ino_t iunique(struct super_block *sb, ino_t max_reserved)
1147 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1148 * error if st_ino won't fit in target struct field. Use 32bit counter
1149 * here to attempt to avoid that.
1151 static DEFINE_SPINLOCK(iunique_lock);
1152 static unsigned int counter;
1155 spin_lock(&iunique_lock);
1157 if (counter <= max_reserved)
1158 counter = max_reserved + 1;
1160 } while (!test_inode_iunique(sb, res));
1161 spin_unlock(&iunique_lock);
1165 EXPORT_SYMBOL(iunique);
1167 struct inode *igrab(struct inode *inode)
1169 spin_lock(&inode->i_lock);
1170 if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) {
1172 spin_unlock(&inode->i_lock);
1174 spin_unlock(&inode->i_lock);
1176 * Handle the case where s_op->clear_inode is not been
1177 * called yet, and somebody is calling igrab
1178 * while the inode is getting freed.
1184 EXPORT_SYMBOL(igrab);
1187 * ilookup5_nowait - search for an inode in the inode cache
1188 * @sb: super block of file system to search
1189 * @hashval: hash value (usually inode number) to search for
1190 * @test: callback used for comparisons between inodes
1191 * @data: opaque data pointer to pass to @test
1193 * Search for the inode specified by @hashval and @data in the inode cache.
1194 * If the inode is in the cache, the inode is returned with an incremented
1197 * Note: I_NEW is not waited upon so you have to be very careful what you do
1198 * with the returned inode. You probably should be using ilookup5() instead.
1200 * Note2: @test is called with the inode_hash_lock held, so can't sleep.
1202 struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval,
1203 int (*test)(struct inode *, void *), void *data)
1205 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1206 struct inode *inode;
1208 spin_lock(&inode_hash_lock);
1209 inode = find_inode(sb, head, test, data);
1210 spin_unlock(&inode_hash_lock);
1214 EXPORT_SYMBOL(ilookup5_nowait);
1217 * ilookup5 - search for an inode in the inode cache
1218 * @sb: super block of file system to search
1219 * @hashval: hash value (usually inode number) to search for
1220 * @test: callback used for comparisons between inodes
1221 * @data: opaque data pointer to pass to @test
1223 * Search for the inode specified by @hashval and @data in the inode cache,
1224 * and if the inode is in the cache, return the inode with an incremented
1225 * reference count. Waits on I_NEW before returning the inode.
1226 * returned with an incremented reference count.
1228 * This is a generalized version of ilookup() for file systems where the
1229 * inode number is not sufficient for unique identification of an inode.
1231 * Note: @test is called with the inode_hash_lock held, so can't sleep.
1233 struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
1234 int (*test)(struct inode *, void *), void *data)
1236 struct inode *inode = ilookup5_nowait(sb, hashval, test, data);
1239 wait_on_inode(inode);
1242 EXPORT_SYMBOL(ilookup5);
1245 * ilookup - search for an inode in the inode cache
1246 * @sb: super block of file system to search
1247 * @ino: inode number to search for
1249 * Search for the inode @ino in the inode cache, and if the inode is in the
1250 * cache, the inode is returned with an incremented reference count.
1252 struct inode *ilookup(struct super_block *sb, unsigned long ino)
1254 struct hlist_head *head = inode_hashtable + hash(sb, ino);
1255 struct inode *inode;
1257 spin_lock(&inode_hash_lock);
1258 inode = find_inode_fast(sb, head, ino);
1259 spin_unlock(&inode_hash_lock);
1262 wait_on_inode(inode);
1265 EXPORT_SYMBOL(ilookup);
1267 int insert_inode_locked(struct inode *inode)
1269 struct super_block *sb = inode->i_sb;
1270 ino_t ino = inode->i_ino;
1271 struct hlist_head *head = inode_hashtable + hash(sb, ino);
1274 struct hlist_node *node;
1275 struct inode *old = NULL;
1276 spin_lock(&inode_hash_lock);
1277 hlist_for_each_entry(old, node, head, i_hash) {
1278 if (old->i_ino != ino)
1280 if (old->i_sb != sb)
1282 spin_lock(&old->i_lock);
1283 if (old->i_state & (I_FREEING|I_WILL_FREE)) {
1284 spin_unlock(&old->i_lock);
1289 if (likely(!node)) {
1290 spin_lock(&inode->i_lock);
1291 inode->i_state |= I_NEW;
1292 hlist_add_head(&inode->i_hash, head);
1293 spin_unlock(&inode->i_lock);
1294 spin_unlock(&inode_hash_lock);
1298 spin_unlock(&old->i_lock);
1299 spin_unlock(&inode_hash_lock);
1301 if (unlikely(!inode_unhashed(old))) {
1308 EXPORT_SYMBOL(insert_inode_locked);
1310 int insert_inode_locked4(struct inode *inode, unsigned long hashval,
1311 int (*test)(struct inode *, void *), void *data)
1313 struct super_block *sb = inode->i_sb;
1314 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1317 struct hlist_node *node;
1318 struct inode *old = NULL;
1320 spin_lock(&inode_hash_lock);
1321 hlist_for_each_entry(old, node, head, i_hash) {
1322 if (old->i_sb != sb)
1324 if (!test(old, data))
1326 spin_lock(&old->i_lock);
1327 if (old->i_state & (I_FREEING|I_WILL_FREE)) {
1328 spin_unlock(&old->i_lock);
1333 if (likely(!node)) {
1334 spin_lock(&inode->i_lock);
1335 inode->i_state |= I_NEW;
1336 hlist_add_head(&inode->i_hash, head);
1337 spin_unlock(&inode->i_lock);
1338 spin_unlock(&inode_hash_lock);
1342 spin_unlock(&old->i_lock);
1343 spin_unlock(&inode_hash_lock);
1345 if (unlikely(!inode_unhashed(old))) {
1352 EXPORT_SYMBOL(insert_inode_locked4);
1355 int generic_delete_inode(struct inode *inode)
1359 EXPORT_SYMBOL(generic_delete_inode);
1362 * Called when we're dropping the last reference
1365 * Call the FS "drop_inode()" function, defaulting to
1366 * the legacy UNIX filesystem behaviour. If it tells
1367 * us to evict inode, do so. Otherwise, retain inode
1368 * in cache if fs is alive, sync and evict if fs is
1371 static void iput_final(struct inode *inode)
1373 struct super_block *sb = inode->i_sb;
1374 const struct super_operations *op = inode->i_sb->s_op;
1377 WARN_ON(inode->i_state & I_NEW);
1380 drop = op->drop_inode(inode);
1382 drop = generic_drop_inode(inode);
1384 if (!drop && (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);
1388 spin_unlock(&inode->i_lock);
1393 inode->i_state |= I_WILL_FREE;
1394 spin_unlock(&inode->i_lock);
1395 write_inode_now(inode, 1);
1396 spin_lock(&inode->i_lock);
1397 WARN_ON(inode->i_state & I_NEW);
1398 inode->i_state &= ~I_WILL_FREE;
1401 inode->i_state |= I_FREEING;
1402 if (!list_empty(&inode->i_lru))
1403 inode_lru_list_del(inode);
1404 spin_unlock(&inode->i_lock);
1410 * iput - put an inode
1411 * @inode: inode to put
1413 * Puts an inode, dropping its usage count. If the inode use count hits
1414 * zero, the inode is then freed and may also be destroyed.
1416 * Consequently, iput() can sleep.
1418 void iput(struct inode *inode)
1421 BUG_ON(inode->i_state & I_CLEAR);
1423 if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock))
1427 EXPORT_SYMBOL(iput);
1430 * bmap - find a block number in a file
1431 * @inode: inode of file
1432 * @block: block to find
1434 * Returns the block number on the device holding the inode that
1435 * is the disk block number for the block of the file requested.
1436 * That is, asked for block 4 of inode 1 the function will return the
1437 * disk block relative to the disk start that holds that block of the
1440 sector_t bmap(struct inode *inode, sector_t block)
1443 if (inode->i_mapping->a_ops->bmap)
1444 res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block);
1447 EXPORT_SYMBOL(bmap);
1450 * With relative atime, only update atime if the previous atime is
1451 * earlier than either the ctime or mtime or if at least a day has
1452 * passed since the last atime update.
1454 static int relatime_need_update(struct vfsmount *mnt, struct inode *inode,
1455 struct timespec now)
1458 if (!(mnt->mnt_flags & MNT_RELATIME))
1461 * Is mtime younger than atime? If yes, update atime:
1463 if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0)
1466 * Is ctime younger than atime? If yes, update atime:
1468 if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0)
1472 * Is the previous atime value older than a day? If yes,
1475 if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60)
1478 * Good, we can skip the atime update:
1484 * touch_atime - update the access time
1485 * @mnt: mount the inode is accessed on
1486 * @dentry: dentry accessed
1488 * Update the accessed time on an inode and mark it for writeback.
1489 * This function automatically handles read only file systems and media,
1490 * as well as the "noatime" flag and inode specific "noatime" markers.
1492 void touch_atime(struct path *path)
1494 struct vfsmount *mnt = path->mnt;
1495 struct inode *inode = path->dentry->d_inode;
1496 struct timespec now;
1498 if (inode->i_flags & S_NOATIME)
1500 if (IS_NOATIME(inode))
1502 if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode))
1505 if (mnt->mnt_flags & MNT_NOATIME)
1507 if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode))
1510 now = current_fs_time(inode->i_sb);
1512 if (!relatime_need_update(mnt, inode, now))
1515 if (timespec_equal(&inode->i_atime, &now))
1518 if (mnt_want_write(mnt))
1521 inode->i_atime = now;
1522 mark_inode_dirty_sync(inode);
1523 mnt_drop_write(mnt);
1525 EXPORT_SYMBOL(touch_atime);
1528 * file_update_time - update mtime and ctime time
1529 * @file: file accessed
1531 * Update the mtime and ctime members of an inode and mark the inode
1532 * for writeback. Note that this function is meant exclusively for
1533 * usage in the file write path of filesystems, and filesystems may
1534 * choose to explicitly ignore update via this function with the
1535 * S_NOCMTIME inode flag, e.g. for network filesystem where these
1536 * timestamps are handled by the server.
1539 void file_update_time(struct file *file)
1541 struct inode *inode = file->f_path.dentry->d_inode;
1542 struct timespec now;
1543 enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0;
1545 /* First try to exhaust all avenues to not sync */
1546 if (IS_NOCMTIME(inode))
1549 now = current_fs_time(inode->i_sb);
1550 if (!timespec_equal(&inode->i_mtime, &now))
1553 if (!timespec_equal(&inode->i_ctime, &now))
1556 if (IS_I_VERSION(inode))
1557 sync_it |= S_VERSION;
1562 /* Finally allowed to write? Takes lock. */
1563 if (mnt_want_write_file(file))
1566 /* Only change inode inside the lock region */
1567 if (sync_it & S_VERSION)
1568 inode_inc_iversion(inode);
1569 if (sync_it & S_CTIME)
1570 inode->i_ctime = now;
1571 if (sync_it & S_MTIME)
1572 inode->i_mtime = now;
1573 mark_inode_dirty_sync(inode);
1574 mnt_drop_write_file(file);
1576 EXPORT_SYMBOL(file_update_time);
1578 int inode_needs_sync(struct inode *inode)
1582 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
1586 EXPORT_SYMBOL(inode_needs_sync);
1588 int inode_wait(void *word)
1593 EXPORT_SYMBOL(inode_wait);
1596 * If we try to find an inode in the inode hash while it is being
1597 * deleted, we have to wait until the filesystem completes its
1598 * deletion before reporting that it isn't found. This function waits
1599 * until the deletion _might_ have completed. Callers are responsible
1600 * to recheck inode state.
1602 * It doesn't matter if I_NEW is not set initially, a call to
1603 * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
1606 static void __wait_on_freeing_inode(struct inode *inode)
1608 wait_queue_head_t *wq;
1609 DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW);
1610 wq = bit_waitqueue(&inode->i_state, __I_NEW);
1611 prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE);
1612 spin_unlock(&inode->i_lock);
1613 spin_unlock(&inode_hash_lock);
1615 finish_wait(wq, &wait.wait);
1616 spin_lock(&inode_hash_lock);
1619 static __initdata unsigned long ihash_entries;
1620 static int __init set_ihash_entries(char *str)
1624 ihash_entries = simple_strtoul(str, &str, 0);
1627 __setup("ihash_entries=", set_ihash_entries);
1630 * Initialize the waitqueues and inode hash table.
1632 void __init inode_init_early(void)
1636 /* If hashes are distributed across NUMA nodes, defer
1637 * hash allocation until vmalloc space is available.
1643 alloc_large_system_hash("Inode-cache",
1644 sizeof(struct hlist_head),
1652 for (loop = 0; loop < (1U << i_hash_shift); loop++)
1653 INIT_HLIST_HEAD(&inode_hashtable[loop]);
1656 void __init inode_init(void)
1660 /* inode slab cache */
1661 inode_cachep = kmem_cache_create("inode_cache",
1662 sizeof(struct inode),
1664 (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|
1668 /* Hash may have been set up in inode_init_early */
1673 alloc_large_system_hash("Inode-cache",
1674 sizeof(struct hlist_head),
1682 for (loop = 0; loop < (1U << i_hash_shift); loop++)
1683 INIT_HLIST_HEAD(&inode_hashtable[loop]);
1686 void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev)
1688 inode->i_mode = mode;
1689 if (S_ISCHR(mode)) {
1690 inode->i_fop = &def_chr_fops;
1691 inode->i_rdev = rdev;
1692 } else if (S_ISBLK(mode)) {
1693 inode->i_fop = &def_blk_fops;
1694 inode->i_rdev = rdev;
1695 } else if (S_ISFIFO(mode))
1696 inode->i_fop = &def_fifo_fops;
1697 else if (S_ISSOCK(mode))
1698 inode->i_fop = &bad_sock_fops;
1700 printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for"
1701 " inode %s:%lu\n", mode, inode->i_sb->s_id,
1704 EXPORT_SYMBOL(init_special_inode);
1707 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
1709 * @dir: Directory inode
1710 * @mode: mode of the new inode
1712 void inode_init_owner(struct inode *inode, const struct inode *dir,
1715 inode->i_uid = current_fsuid();
1716 if (dir && dir->i_mode & S_ISGID) {
1717 inode->i_gid = dir->i_gid;
1721 inode->i_gid = current_fsgid();
1722 inode->i_mode = mode;
1724 EXPORT_SYMBOL(inode_init_owner);
1727 * inode_owner_or_capable - check current task permissions to inode
1728 * @inode: inode being checked
1730 * Return true if current either has CAP_FOWNER to the inode, or
1733 bool inode_owner_or_capable(const struct inode *inode)
1735 if (uid_eq(current_fsuid(), inode->i_uid))
1737 if (inode_capable(inode, CAP_FOWNER))
1741 EXPORT_SYMBOL(inode_owner_or_capable);