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 * inode locking rules.
33 * inode->i_lock protects:
34 * inode->i_state, inode->i_hash, __iget()
42 * This is needed for the following functions:
46 * FIXME: remove all knowledge of the buffer layer from this file
48 #include <linux/buffer_head.h>
51 * New inode.c implementation.
53 * This implementation has the basic premise of trying
54 * to be extremely low-overhead and SMP-safe, yet be
55 * simple enough to be "obviously correct".
60 /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */
62 /* #define INODE_PARANOIA 1 */
63 /* #define INODE_DEBUG 1 */
66 * Inode lookup is no longer as critical as it used to be:
67 * most of the lookups are going to be through the dcache.
69 #define I_HASHBITS i_hash_shift
70 #define I_HASHMASK i_hash_mask
72 static unsigned int i_hash_mask __read_mostly;
73 static unsigned int i_hash_shift __read_mostly;
76 * Each inode can be on two separate lists. One is
77 * the hash list of the inode, used for lookups. The
78 * other linked list is the "type" list:
79 * "in_use" - valid inode, i_count > 0, i_nlink > 0
80 * "dirty" - as "in_use" but also dirty
81 * "unused" - valid inode, i_count = 0
83 * A "dirty" list is maintained for each super block,
84 * allowing for low-overhead inode sync() operations.
87 static LIST_HEAD(inode_lru);
88 static struct hlist_head *inode_hashtable __read_mostly;
91 * A simple spinlock to protect the list manipulations.
93 * NOTE! You also have to own the lock if you change
94 * the i_state of an inode while it is in use..
96 DEFINE_SPINLOCK(inode_lock);
99 * iprune_sem provides exclusion between the icache shrinking and the
102 * We don't actually need it to protect anything in the umount path,
103 * but only need to cycle through it to make sure any inode that
104 * prune_icache took off the LRU list has been fully torn down by the
105 * time we are past evict_inodes.
107 static DECLARE_RWSEM(iprune_sem);
110 * Statistics gathering..
112 struct inodes_stat_t inodes_stat;
114 static DEFINE_PER_CPU(unsigned int, nr_inodes);
116 static struct kmem_cache *inode_cachep __read_mostly;
118 static int get_nr_inodes(void)
122 for_each_possible_cpu(i)
123 sum += per_cpu(nr_inodes, i);
124 return sum < 0 ? 0 : sum;
127 static inline int get_nr_inodes_unused(void)
129 return inodes_stat.nr_unused;
132 int get_nr_dirty_inodes(void)
134 /* not actually dirty inodes, but a wild approximation */
135 int nr_dirty = get_nr_inodes() - get_nr_inodes_unused();
136 return nr_dirty > 0 ? nr_dirty : 0;
140 * Handle nr_inode sysctl
143 int proc_nr_inodes(ctl_table *table, int write,
144 void __user *buffer, size_t *lenp, loff_t *ppos)
146 inodes_stat.nr_inodes = get_nr_inodes();
147 return proc_dointvec(table, write, buffer, lenp, ppos);
152 * inode_init_always - perform inode structure intialisation
153 * @sb: superblock inode belongs to
154 * @inode: inode to initialise
156 * These are initializations that need to be done on every inode
157 * allocation as the fields are not initialised by slab allocation.
159 int inode_init_always(struct super_block *sb, struct inode *inode)
161 static const struct address_space_operations empty_aops;
162 static const struct inode_operations empty_iops;
163 static const struct file_operations empty_fops;
164 struct address_space *const mapping = &inode->i_data;
167 inode->i_blkbits = sb->s_blocksize_bits;
169 atomic_set(&inode->i_count, 1);
170 inode->i_op = &empty_iops;
171 inode->i_fop = &empty_fops;
175 atomic_set(&inode->i_writecount, 0);
179 inode->i_generation = 0;
181 memset(&inode->i_dquot, 0, sizeof(inode->i_dquot));
183 inode->i_pipe = NULL;
184 inode->i_bdev = NULL;
185 inode->i_cdev = NULL;
187 inode->dirtied_when = 0;
189 if (security_inode_alloc(inode))
191 spin_lock_init(&inode->i_lock);
192 lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key);
194 mutex_init(&inode->i_mutex);
195 lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key);
197 init_rwsem(&inode->i_alloc_sem);
198 lockdep_set_class(&inode->i_alloc_sem, &sb->s_type->i_alloc_sem_key);
200 mapping->a_ops = &empty_aops;
201 mapping->host = inode;
203 mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE);
204 mapping->assoc_mapping = NULL;
205 mapping->backing_dev_info = &default_backing_dev_info;
206 mapping->writeback_index = 0;
209 * If the block_device provides a backing_dev_info for client
210 * inodes then use that. Otherwise the inode share the bdev's
214 struct backing_dev_info *bdi;
216 bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
217 mapping->backing_dev_info = bdi;
219 inode->i_private = NULL;
220 inode->i_mapping = mapping;
221 #ifdef CONFIG_FS_POSIX_ACL
222 inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED;
225 #ifdef CONFIG_FSNOTIFY
226 inode->i_fsnotify_mask = 0;
229 this_cpu_inc(nr_inodes);
235 EXPORT_SYMBOL(inode_init_always);
237 static struct inode *alloc_inode(struct super_block *sb)
241 if (sb->s_op->alloc_inode)
242 inode = sb->s_op->alloc_inode(sb);
244 inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL);
249 if (unlikely(inode_init_always(sb, inode))) {
250 if (inode->i_sb->s_op->destroy_inode)
251 inode->i_sb->s_op->destroy_inode(inode);
253 kmem_cache_free(inode_cachep, inode);
260 void free_inode_nonrcu(struct inode *inode)
262 kmem_cache_free(inode_cachep, inode);
264 EXPORT_SYMBOL(free_inode_nonrcu);
266 void __destroy_inode(struct inode *inode)
268 BUG_ON(inode_has_buffers(inode));
269 security_inode_free(inode);
270 fsnotify_inode_delete(inode);
271 #ifdef CONFIG_FS_POSIX_ACL
272 if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED)
273 posix_acl_release(inode->i_acl);
274 if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED)
275 posix_acl_release(inode->i_default_acl);
277 this_cpu_dec(nr_inodes);
279 EXPORT_SYMBOL(__destroy_inode);
281 static void i_callback(struct rcu_head *head)
283 struct inode *inode = container_of(head, struct inode, i_rcu);
284 INIT_LIST_HEAD(&inode->i_dentry);
285 kmem_cache_free(inode_cachep, inode);
288 static void destroy_inode(struct inode *inode)
290 BUG_ON(!list_empty(&inode->i_lru));
291 __destroy_inode(inode);
292 if (inode->i_sb->s_op->destroy_inode)
293 inode->i_sb->s_op->destroy_inode(inode);
295 call_rcu(&inode->i_rcu, i_callback);
298 void address_space_init_once(struct address_space *mapping)
300 memset(mapping, 0, sizeof(*mapping));
301 INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC);
302 spin_lock_init(&mapping->tree_lock);
303 spin_lock_init(&mapping->i_mmap_lock);
304 INIT_LIST_HEAD(&mapping->private_list);
305 spin_lock_init(&mapping->private_lock);
306 INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap);
307 INIT_LIST_HEAD(&mapping->i_mmap_nonlinear);
308 mutex_init(&mapping->unmap_mutex);
310 EXPORT_SYMBOL(address_space_init_once);
313 * These are initializations that only need to be done
314 * once, because the fields are idempotent across use
315 * of the inode, so let the slab aware of that.
317 void inode_init_once(struct inode *inode)
319 memset(inode, 0, sizeof(*inode));
320 INIT_HLIST_NODE(&inode->i_hash);
321 INIT_LIST_HEAD(&inode->i_dentry);
322 INIT_LIST_HEAD(&inode->i_devices);
323 INIT_LIST_HEAD(&inode->i_wb_list);
324 INIT_LIST_HEAD(&inode->i_lru);
325 address_space_init_once(&inode->i_data);
326 i_size_ordered_init(inode);
327 #ifdef CONFIG_FSNOTIFY
328 INIT_HLIST_HEAD(&inode->i_fsnotify_marks);
331 EXPORT_SYMBOL(inode_init_once);
333 static void init_once(void *foo)
335 struct inode *inode = (struct inode *) foo;
337 inode_init_once(inode);
341 * inode->i_lock must be held
343 void __iget(struct inode *inode)
345 atomic_inc(&inode->i_count);
349 * get additional reference to inode; caller must already hold one.
351 void ihold(struct inode *inode)
353 WARN_ON(atomic_inc_return(&inode->i_count) < 2);
355 EXPORT_SYMBOL(ihold);
357 static void inode_lru_list_add(struct inode *inode)
359 if (list_empty(&inode->i_lru)) {
360 list_add(&inode->i_lru, &inode_lru);
361 inodes_stat.nr_unused++;
365 static void inode_lru_list_del(struct inode *inode)
367 if (!list_empty(&inode->i_lru)) {
368 list_del_init(&inode->i_lru);
369 inodes_stat.nr_unused--;
373 static inline void __inode_sb_list_add(struct inode *inode)
375 list_add(&inode->i_sb_list, &inode->i_sb->s_inodes);
379 * inode_sb_list_add - add inode to the superblock list of inodes
380 * @inode: inode to add
382 void inode_sb_list_add(struct inode *inode)
384 spin_lock(&inode_lock);
385 __inode_sb_list_add(inode);
386 spin_unlock(&inode_lock);
388 EXPORT_SYMBOL_GPL(inode_sb_list_add);
390 static inline void __inode_sb_list_del(struct inode *inode)
392 list_del_init(&inode->i_sb_list);
395 static unsigned long hash(struct super_block *sb, unsigned long hashval)
399 tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) /
401 tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> I_HASHBITS);
402 return tmp & I_HASHMASK;
406 * __insert_inode_hash - hash an inode
407 * @inode: unhashed inode
408 * @hashval: unsigned long value used to locate this object in the
411 * Add an inode to the inode hash for this superblock.
413 void __insert_inode_hash(struct inode *inode, unsigned long hashval)
415 struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval);
417 spin_lock(&inode_lock);
418 spin_lock(&inode->i_lock);
419 hlist_add_head(&inode->i_hash, b);
420 spin_unlock(&inode->i_lock);
421 spin_unlock(&inode_lock);
423 EXPORT_SYMBOL(__insert_inode_hash);
426 * remove_inode_hash - remove an inode from the hash
427 * @inode: inode to unhash
429 * Remove an inode from the superblock.
431 void remove_inode_hash(struct inode *inode)
433 spin_lock(&inode_lock);
434 spin_lock(&inode->i_lock);
435 hlist_del_init(&inode->i_hash);
436 spin_unlock(&inode->i_lock);
437 spin_unlock(&inode_lock);
439 EXPORT_SYMBOL(remove_inode_hash);
441 void end_writeback(struct inode *inode)
444 BUG_ON(inode->i_data.nrpages);
445 BUG_ON(!list_empty(&inode->i_data.private_list));
446 BUG_ON(!(inode->i_state & I_FREEING));
447 BUG_ON(inode->i_state & I_CLEAR);
448 inode_sync_wait(inode);
449 /* don't need i_lock here, no concurrent mods to i_state */
450 inode->i_state = I_FREEING | I_CLEAR;
452 EXPORT_SYMBOL(end_writeback);
455 * Free the inode passed in, removing it from the lists it is still connected
456 * to. We remove any pages still attached to the inode and wait for any IO that
457 * is still in progress before finally destroying the inode.
459 * An inode must already be marked I_FREEING so that we avoid the inode being
460 * moved back onto lists if we race with other code that manipulates the lists
461 * (e.g. writeback_single_inode). The caller is responsible for setting this.
463 * An inode must already be removed from the LRU list before being evicted from
464 * the cache. This should occur atomically with setting the I_FREEING state
465 * flag, so no inodes here should ever be on the LRU when being evicted.
467 static void evict(struct inode *inode)
469 const struct super_operations *op = inode->i_sb->s_op;
471 BUG_ON(!(inode->i_state & I_FREEING));
472 BUG_ON(!list_empty(&inode->i_lru));
474 spin_lock(&inode_lock);
475 list_del_init(&inode->i_wb_list);
476 __inode_sb_list_del(inode);
477 spin_unlock(&inode_lock);
479 if (op->evict_inode) {
480 op->evict_inode(inode);
482 if (inode->i_data.nrpages)
483 truncate_inode_pages(&inode->i_data, 0);
484 end_writeback(inode);
486 if (S_ISBLK(inode->i_mode) && inode->i_bdev)
488 if (S_ISCHR(inode->i_mode) && inode->i_cdev)
491 remove_inode_hash(inode);
493 spin_lock(&inode->i_lock);
494 wake_up_bit(&inode->i_state, __I_NEW);
495 BUG_ON(inode->i_state != (I_FREEING | I_CLEAR));
496 spin_unlock(&inode->i_lock);
498 destroy_inode(inode);
502 * dispose_list - dispose of the contents of a local list
503 * @head: the head of the list to free
505 * Dispose-list gets a local list with local inodes in it, so it doesn't
506 * need to worry about list corruption and SMP locks.
508 static void dispose_list(struct list_head *head)
510 while (!list_empty(head)) {
513 inode = list_first_entry(head, struct inode, i_lru);
514 list_del_init(&inode->i_lru);
521 * evict_inodes - evict all evictable inodes for a superblock
522 * @sb: superblock to operate on
524 * Make sure that no inodes with zero refcount are retained. This is
525 * called by superblock shutdown after having MS_ACTIVE flag removed,
526 * so any inode reaching zero refcount during or after that call will
527 * be immediately evicted.
529 void evict_inodes(struct super_block *sb)
531 struct inode *inode, *next;
534 spin_lock(&inode_lock);
535 list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
536 if (atomic_read(&inode->i_count))
539 spin_lock(&inode->i_lock);
540 if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
541 spin_unlock(&inode->i_lock);
545 inode->i_state |= I_FREEING;
546 if (!(inode->i_state & (I_DIRTY | I_SYNC)))
547 inodes_stat.nr_unused--;
548 spin_unlock(&inode->i_lock);
549 list_move(&inode->i_lru, &dispose);
551 spin_unlock(&inode_lock);
553 dispose_list(&dispose);
556 * Cycle through iprune_sem to make sure any inode that prune_icache
557 * moved off the list before we took the lock has been fully torn
560 down_write(&iprune_sem);
561 up_write(&iprune_sem);
565 * invalidate_inodes - attempt to free all inodes on a superblock
566 * @sb: superblock to operate on
567 * @kill_dirty: flag to guide handling of dirty inodes
569 * Attempts to free all inodes for a given superblock. If there were any
570 * busy inodes return a non-zero value, else zero.
571 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
574 int invalidate_inodes(struct super_block *sb, bool kill_dirty)
577 struct inode *inode, *next;
580 spin_lock(&inode_lock);
581 list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
582 spin_lock(&inode->i_lock);
583 if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
584 spin_unlock(&inode->i_lock);
587 if (inode->i_state & I_DIRTY && !kill_dirty) {
588 spin_unlock(&inode->i_lock);
592 if (atomic_read(&inode->i_count)) {
593 spin_unlock(&inode->i_lock);
598 inode->i_state |= I_FREEING;
599 if (!(inode->i_state & (I_DIRTY | I_SYNC)))
600 inodes_stat.nr_unused--;
601 spin_unlock(&inode->i_lock);
602 list_move(&inode->i_lru, &dispose);
604 spin_unlock(&inode_lock);
606 dispose_list(&dispose);
611 static int can_unuse(struct inode *inode)
613 if (inode->i_state & ~I_REFERENCED)
615 if (inode_has_buffers(inode))
617 if (atomic_read(&inode->i_count))
619 if (inode->i_data.nrpages)
625 * Scan `goal' inodes on the unused list for freeable ones. They are moved to a
626 * temporary list and then are freed outside inode_lock by dispose_list().
628 * Any inodes which are pinned purely because of attached pagecache have their
629 * pagecache removed. If the inode has metadata buffers attached to
630 * mapping->private_list then try to remove them.
632 * If the inode has the I_REFERENCED flag set, then it means that it has been
633 * used recently - the flag is set in iput_final(). When we encounter such an
634 * inode, clear the flag and move it to the back of the LRU so it gets another
635 * pass through the LRU before it gets reclaimed. This is necessary because of
636 * the fact we are doing lazy LRU updates to minimise lock contention so the
637 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
638 * with this flag set because they are the inodes that are out of order.
640 static void prune_icache(int nr_to_scan)
644 unsigned long reap = 0;
646 down_read(&iprune_sem);
647 spin_lock(&inode_lock);
648 for (nr_scanned = 0; nr_scanned < nr_to_scan; nr_scanned++) {
651 if (list_empty(&inode_lru))
654 inode = list_entry(inode_lru.prev, struct inode, i_lru);
657 * Referenced or dirty inodes are still in use. Give them
658 * another pass through the LRU as we canot reclaim them now.
660 spin_lock(&inode->i_lock);
661 if (atomic_read(&inode->i_count) ||
662 (inode->i_state & ~I_REFERENCED)) {
663 spin_unlock(&inode->i_lock);
664 list_del_init(&inode->i_lru);
665 inodes_stat.nr_unused--;
669 /* recently referenced inodes get one more pass */
670 if (inode->i_state & I_REFERENCED) {
671 inode->i_state &= ~I_REFERENCED;
672 spin_unlock(&inode->i_lock);
673 list_move(&inode->i_lru, &inode_lru);
676 if (inode_has_buffers(inode) || inode->i_data.nrpages) {
678 spin_unlock(&inode->i_lock);
679 spin_unlock(&inode_lock);
680 if (remove_inode_buffers(inode))
681 reap += invalidate_mapping_pages(&inode->i_data,
684 spin_lock(&inode_lock);
686 if (inode != list_entry(inode_lru.next,
687 struct inode, i_lru))
688 continue; /* wrong inode or list_empty */
689 spin_lock(&inode->i_lock);
690 if (!can_unuse(inode)) {
691 spin_unlock(&inode->i_lock);
695 WARN_ON(inode->i_state & I_NEW);
696 inode->i_state |= I_FREEING;
697 spin_unlock(&inode->i_lock);
699 list_move(&inode->i_lru, &freeable);
700 inodes_stat.nr_unused--;
702 if (current_is_kswapd())
703 __count_vm_events(KSWAPD_INODESTEAL, reap);
705 __count_vm_events(PGINODESTEAL, reap);
706 spin_unlock(&inode_lock);
708 dispose_list(&freeable);
709 up_read(&iprune_sem);
713 * shrink_icache_memory() will attempt to reclaim some unused inodes. Here,
714 * "unused" means that no dentries are referring to the inodes: the files are
715 * not open and the dcache references to those inodes have already been
718 * This function is passed the number of inodes to scan, and it returns the
719 * total number of remaining possibly-reclaimable inodes.
721 static int shrink_icache_memory(struct shrinker *shrink, int nr, gfp_t gfp_mask)
725 * Nasty deadlock avoidance. We may hold various FS locks,
726 * and we don't want to recurse into the FS that called us
727 * in clear_inode() and friends..
729 if (!(gfp_mask & __GFP_FS))
733 return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure;
736 static struct shrinker icache_shrinker = {
737 .shrink = shrink_icache_memory,
738 .seeks = DEFAULT_SEEKS,
741 static void __wait_on_freeing_inode(struct inode *inode);
743 * Called with the inode lock held.
745 static struct inode *find_inode(struct super_block *sb,
746 struct hlist_head *head,
747 int (*test)(struct inode *, void *),
750 struct hlist_node *node;
751 struct inode *inode = NULL;
754 hlist_for_each_entry(inode, node, head, i_hash) {
755 if (inode->i_sb != sb)
757 if (!test(inode, data))
759 spin_lock(&inode->i_lock);
760 if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
761 __wait_on_freeing_inode(inode);
765 spin_unlock(&inode->i_lock);
772 * find_inode_fast is the fast path version of find_inode, see the comment at
773 * iget_locked for details.
775 static struct inode *find_inode_fast(struct super_block *sb,
776 struct hlist_head *head, unsigned long ino)
778 struct hlist_node *node;
779 struct inode *inode = NULL;
782 hlist_for_each_entry(inode, node, head, i_hash) {
783 if (inode->i_ino != ino)
785 if (inode->i_sb != sb)
787 spin_lock(&inode->i_lock);
788 if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
789 __wait_on_freeing_inode(inode);
793 spin_unlock(&inode->i_lock);
800 * Each cpu owns a range of LAST_INO_BATCH numbers.
801 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
802 * to renew the exhausted range.
804 * This does not significantly increase overflow rate because every CPU can
805 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
806 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
807 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
808 * overflow rate by 2x, which does not seem too significant.
810 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
811 * error if st_ino won't fit in target struct field. Use 32bit counter
812 * here to attempt to avoid that.
814 #define LAST_INO_BATCH 1024
815 static DEFINE_PER_CPU(unsigned int, last_ino);
817 unsigned int get_next_ino(void)
819 unsigned int *p = &get_cpu_var(last_ino);
820 unsigned int res = *p;
823 if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) {
824 static atomic_t shared_last_ino;
825 int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino);
827 res = next - LAST_INO_BATCH;
832 put_cpu_var(last_ino);
835 EXPORT_SYMBOL(get_next_ino);
838 * new_inode - obtain an inode
841 * Allocates a new inode for given superblock. The default gfp_mask
842 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
843 * If HIGHMEM pages are unsuitable or it is known that pages allocated
844 * for the page cache are not reclaimable or migratable,
845 * mapping_set_gfp_mask() must be called with suitable flags on the
846 * newly created inode's mapping
849 struct inode *new_inode(struct super_block *sb)
853 spin_lock_prefetch(&inode_lock);
855 inode = alloc_inode(sb);
857 spin_lock(&inode_lock);
858 spin_lock(&inode->i_lock);
860 spin_unlock(&inode->i_lock);
861 __inode_sb_list_add(inode);
862 spin_unlock(&inode_lock);
866 EXPORT_SYMBOL(new_inode);
869 * unlock_new_inode - clear the I_NEW state and wake up any waiters
870 * @inode: new inode to unlock
872 * Called when the inode is fully initialised to clear the new state of the
873 * inode and wake up anyone waiting for the inode to finish initialisation.
875 void unlock_new_inode(struct inode *inode)
877 #ifdef CONFIG_DEBUG_LOCK_ALLOC
878 if (S_ISDIR(inode->i_mode)) {
879 struct file_system_type *type = inode->i_sb->s_type;
881 /* Set new key only if filesystem hasn't already changed it */
882 if (!lockdep_match_class(&inode->i_mutex,
883 &type->i_mutex_key)) {
885 * ensure nobody is actually holding i_mutex
887 mutex_destroy(&inode->i_mutex);
888 mutex_init(&inode->i_mutex);
889 lockdep_set_class(&inode->i_mutex,
890 &type->i_mutex_dir_key);
894 spin_lock(&inode->i_lock);
895 WARN_ON(!(inode->i_state & I_NEW));
896 inode->i_state &= ~I_NEW;
897 wake_up_bit(&inode->i_state, __I_NEW);
898 spin_unlock(&inode->i_lock);
900 EXPORT_SYMBOL(unlock_new_inode);
903 * This is called without the inode lock held.. Be careful.
905 * We no longer cache the sb_flags in i_flags - see fs.h
906 * -- rmk@arm.uk.linux.org
908 static struct inode *get_new_inode(struct super_block *sb,
909 struct hlist_head *head,
910 int (*test)(struct inode *, void *),
911 int (*set)(struct inode *, void *),
916 inode = alloc_inode(sb);
920 spin_lock(&inode_lock);
921 /* We released the lock, so.. */
922 old = find_inode(sb, head, test, data);
924 if (set(inode, data))
927 spin_lock(&inode->i_lock);
928 inode->i_state = I_NEW;
929 hlist_add_head(&inode->i_hash, head);
930 spin_unlock(&inode->i_lock);
931 __inode_sb_list_add(inode);
932 spin_unlock(&inode_lock);
934 /* Return the locked inode with I_NEW set, the
935 * caller is responsible for filling in the contents
941 * Uhhuh, somebody else created the same inode under
942 * us. Use the old inode instead of the one we just
945 spin_unlock(&inode_lock);
946 destroy_inode(inode);
948 wait_on_inode(inode);
953 spin_unlock(&inode_lock);
954 destroy_inode(inode);
959 * get_new_inode_fast is the fast path version of get_new_inode, see the
960 * comment at iget_locked for details.
962 static struct inode *get_new_inode_fast(struct super_block *sb,
963 struct hlist_head *head, unsigned long ino)
967 inode = alloc_inode(sb);
971 spin_lock(&inode_lock);
972 /* We released the lock, so.. */
973 old = find_inode_fast(sb, head, ino);
976 spin_lock(&inode->i_lock);
977 inode->i_state = I_NEW;
978 hlist_add_head(&inode->i_hash, head);
979 spin_unlock(&inode->i_lock);
980 __inode_sb_list_add(inode);
981 spin_unlock(&inode_lock);
983 /* Return the locked inode with I_NEW set, the
984 * caller is responsible for filling in the contents
990 * Uhhuh, somebody else created the same inode under
991 * us. Use the old inode instead of the one we just
994 spin_unlock(&inode_lock);
995 destroy_inode(inode);
997 wait_on_inode(inode);
1003 * search the inode cache for a matching inode number.
1004 * If we find one, then the inode number we are trying to
1005 * allocate is not unique and so we should not use it.
1007 * Returns 1 if the inode number is unique, 0 if it is not.
1009 static int test_inode_iunique(struct super_block *sb, unsigned long ino)
1011 struct hlist_head *b = inode_hashtable + hash(sb, ino);
1012 struct hlist_node *node;
1013 struct inode *inode;
1015 hlist_for_each_entry(inode, node, b, i_hash) {
1016 if (inode->i_ino == ino && inode->i_sb == sb)
1024 * iunique - get a unique inode number
1026 * @max_reserved: highest reserved inode number
1028 * Obtain an inode number that is unique on the system for a given
1029 * superblock. This is used by file systems that have no natural
1030 * permanent inode numbering system. An inode number is returned that
1031 * is higher than the reserved limit but unique.
1034 * With a large number of inodes live on the file system this function
1035 * currently becomes quite slow.
1037 ino_t iunique(struct super_block *sb, ino_t max_reserved)
1040 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1041 * error if st_ino won't fit in target struct field. Use 32bit counter
1042 * here to attempt to avoid that.
1044 static DEFINE_SPINLOCK(iunique_lock);
1045 static unsigned int counter;
1048 spin_lock(&inode_lock);
1049 spin_lock(&iunique_lock);
1051 if (counter <= max_reserved)
1052 counter = max_reserved + 1;
1054 } while (!test_inode_iunique(sb, res));
1055 spin_unlock(&iunique_lock);
1056 spin_unlock(&inode_lock);
1060 EXPORT_SYMBOL(iunique);
1062 struct inode *igrab(struct inode *inode)
1064 spin_lock(&inode_lock);
1065 spin_lock(&inode->i_lock);
1066 if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) {
1068 spin_unlock(&inode->i_lock);
1070 spin_unlock(&inode->i_lock);
1072 * Handle the case where s_op->clear_inode is not been
1073 * called yet, and somebody is calling igrab
1074 * while the inode is getting freed.
1078 spin_unlock(&inode_lock);
1081 EXPORT_SYMBOL(igrab);
1084 * ifind - internal function, you want ilookup5() or iget5().
1085 * @sb: super block of file system to search
1086 * @head: the head of the list to search
1087 * @test: callback used for comparisons between inodes
1088 * @data: opaque data pointer to pass to @test
1089 * @wait: if true wait for the inode to be unlocked, if false do not
1091 * ifind() searches for the inode specified by @data in the inode
1092 * cache. This is a generalized version of ifind_fast() for file systems where
1093 * the inode number is not sufficient for unique identification of an inode.
1095 * If the inode is in the cache, the inode is returned with an incremented
1098 * Otherwise NULL is returned.
1100 * Note, @test is called with the inode_lock held, so can't sleep.
1102 static struct inode *ifind(struct super_block *sb,
1103 struct hlist_head *head, int (*test)(struct inode *, void *),
1104 void *data, const int wait)
1106 struct inode *inode;
1108 spin_lock(&inode_lock);
1109 inode = find_inode(sb, head, test, data);
1111 spin_unlock(&inode_lock);
1113 wait_on_inode(inode);
1116 spin_unlock(&inode_lock);
1121 * ifind_fast - internal function, you want ilookup() or iget().
1122 * @sb: super block of file system to search
1123 * @head: head of the list to search
1124 * @ino: inode number to search for
1126 * ifind_fast() searches for the inode @ino in the inode cache. This is for
1127 * file systems where the inode number is sufficient for unique identification
1130 * If the inode is in the cache, the inode is returned with an incremented
1133 * Otherwise NULL is returned.
1135 static struct inode *ifind_fast(struct super_block *sb,
1136 struct hlist_head *head, unsigned long ino)
1138 struct inode *inode;
1140 spin_lock(&inode_lock);
1141 inode = find_inode_fast(sb, head, ino);
1143 spin_unlock(&inode_lock);
1144 wait_on_inode(inode);
1147 spin_unlock(&inode_lock);
1152 * ilookup5_nowait - search for an inode in the inode cache
1153 * @sb: super block of file system to search
1154 * @hashval: hash value (usually inode number) to search for
1155 * @test: callback used for comparisons between inodes
1156 * @data: opaque data pointer to pass to @test
1158 * ilookup5() uses ifind() to search for the inode specified by @hashval and
1159 * @data in the inode cache. This is a generalized version of ilookup() for
1160 * file systems where the inode number is not sufficient for unique
1161 * identification of an inode.
1163 * If the inode is in the cache, the inode is returned with an incremented
1164 * reference count. Note, the inode lock is not waited upon so you have to be
1165 * very careful what you do with the returned inode. You probably should be
1166 * using ilookup5() instead.
1168 * Otherwise NULL is returned.
1170 * Note, @test is called with the inode_lock held, so can't sleep.
1172 struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval,
1173 int (*test)(struct inode *, void *), void *data)
1175 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1177 return ifind(sb, head, test, data, 0);
1179 EXPORT_SYMBOL(ilookup5_nowait);
1182 * ilookup5 - search for an inode in the inode cache
1183 * @sb: super block of file system to search
1184 * @hashval: hash value (usually inode number) to search for
1185 * @test: callback used for comparisons between inodes
1186 * @data: opaque data pointer to pass to @test
1188 * ilookup5() uses ifind() to search for the inode specified by @hashval and
1189 * @data in the inode cache. This is a generalized version of ilookup() for
1190 * file systems where the inode number is not sufficient for unique
1191 * identification of an inode.
1193 * If the inode is in the cache, the inode lock is waited upon and the inode is
1194 * returned with an incremented reference count.
1196 * Otherwise NULL is returned.
1198 * Note, @test is called with the inode_lock held, so can't sleep.
1200 struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
1201 int (*test)(struct inode *, void *), void *data)
1203 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1205 return ifind(sb, head, test, data, 1);
1207 EXPORT_SYMBOL(ilookup5);
1210 * ilookup - search for an inode in the inode cache
1211 * @sb: super block of file system to search
1212 * @ino: inode number to search for
1214 * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache.
1215 * This is for file systems where the inode number is sufficient for unique
1216 * identification of an inode.
1218 * If the inode is in the cache, the inode is returned with an incremented
1221 * Otherwise NULL is returned.
1223 struct inode *ilookup(struct super_block *sb, unsigned long ino)
1225 struct hlist_head *head = inode_hashtable + hash(sb, ino);
1227 return ifind_fast(sb, head, ino);
1229 EXPORT_SYMBOL(ilookup);
1232 * iget5_locked - obtain an inode from a mounted file system
1233 * @sb: super block of file system
1234 * @hashval: hash value (usually inode number) to get
1235 * @test: callback used for comparisons between inodes
1236 * @set: callback used to initialize a new struct inode
1237 * @data: opaque data pointer to pass to @test and @set
1239 * iget5_locked() uses ifind() to search for the inode specified by @hashval
1240 * and @data in the inode cache and if present it is returned with an increased
1241 * reference count. This is a generalized version of iget_locked() for file
1242 * systems where the inode number is not sufficient for unique identification
1245 * If the inode is not in cache, get_new_inode() is called to allocate a new
1246 * inode and this is returned locked, hashed, and with the I_NEW flag set. The
1247 * file system gets to fill it in before unlocking it via unlock_new_inode().
1249 * Note both @test and @set are called with the inode_lock held, so can't sleep.
1251 struct inode *iget5_locked(struct super_block *sb, unsigned long hashval,
1252 int (*test)(struct inode *, void *),
1253 int (*set)(struct inode *, void *), void *data)
1255 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1256 struct inode *inode;
1258 inode = ifind(sb, head, test, data, 1);
1262 * get_new_inode() will do the right thing, re-trying the search
1263 * in case it had to block at any point.
1265 return get_new_inode(sb, head, test, set, data);
1267 EXPORT_SYMBOL(iget5_locked);
1270 * iget_locked - obtain an inode from a mounted file system
1271 * @sb: super block of file system
1272 * @ino: inode number to get
1274 * iget_locked() uses ifind_fast() to search for the inode specified by @ino in
1275 * the inode cache and if present it is returned with an increased reference
1276 * count. This is for file systems where the inode number is sufficient for
1277 * unique identification of an inode.
1279 * If the inode is not in cache, get_new_inode_fast() is called to allocate a
1280 * new inode and this is returned locked, hashed, and with the I_NEW flag set.
1281 * The file system gets to fill it in before unlocking it via
1282 * unlock_new_inode().
1284 struct inode *iget_locked(struct super_block *sb, unsigned long ino)
1286 struct hlist_head *head = inode_hashtable + hash(sb, ino);
1287 struct inode *inode;
1289 inode = ifind_fast(sb, head, ino);
1293 * get_new_inode_fast() will do the right thing, re-trying the search
1294 * in case it had to block at any point.
1296 return get_new_inode_fast(sb, head, ino);
1298 EXPORT_SYMBOL(iget_locked);
1300 int insert_inode_locked(struct inode *inode)
1302 struct super_block *sb = inode->i_sb;
1303 ino_t ino = inode->i_ino;
1304 struct hlist_head *head = inode_hashtable + hash(sb, ino);
1307 struct hlist_node *node;
1308 struct inode *old = NULL;
1309 spin_lock(&inode_lock);
1310 hlist_for_each_entry(old, node, head, i_hash) {
1311 if (old->i_ino != ino)
1313 if (old->i_sb != sb)
1315 spin_lock(&old->i_lock);
1316 if (old->i_state & (I_FREEING|I_WILL_FREE)) {
1317 spin_unlock(&old->i_lock);
1322 if (likely(!node)) {
1323 spin_lock(&inode->i_lock);
1324 inode->i_state |= I_NEW;
1325 hlist_add_head(&inode->i_hash, head);
1326 spin_unlock(&inode->i_lock);
1327 spin_unlock(&inode_lock);
1331 spin_unlock(&old->i_lock);
1332 spin_unlock(&inode_lock);
1334 if (unlikely(!inode_unhashed(old))) {
1341 EXPORT_SYMBOL(insert_inode_locked);
1343 int insert_inode_locked4(struct inode *inode, unsigned long hashval,
1344 int (*test)(struct inode *, void *), void *data)
1346 struct super_block *sb = inode->i_sb;
1347 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1350 struct hlist_node *node;
1351 struct inode *old = NULL;
1353 spin_lock(&inode_lock);
1354 hlist_for_each_entry(old, node, head, i_hash) {
1355 if (old->i_sb != sb)
1357 if (!test(old, data))
1359 spin_lock(&old->i_lock);
1360 if (old->i_state & (I_FREEING|I_WILL_FREE)) {
1361 spin_unlock(&old->i_lock);
1366 if (likely(!node)) {
1367 spin_lock(&inode->i_lock);
1368 inode->i_state |= I_NEW;
1369 hlist_add_head(&inode->i_hash, head);
1370 spin_unlock(&inode->i_lock);
1371 spin_unlock(&inode_lock);
1375 spin_unlock(&old->i_lock);
1376 spin_unlock(&inode_lock);
1378 if (unlikely(!inode_unhashed(old))) {
1385 EXPORT_SYMBOL(insert_inode_locked4);
1388 int generic_delete_inode(struct inode *inode)
1392 EXPORT_SYMBOL(generic_delete_inode);
1395 * Normal UNIX filesystem behaviour: delete the
1396 * inode when the usage count drops to zero, and
1399 int generic_drop_inode(struct inode *inode)
1401 return !inode->i_nlink || inode_unhashed(inode);
1403 EXPORT_SYMBOL_GPL(generic_drop_inode);
1406 * Called when we're dropping the last reference
1409 * Call the FS "drop_inode()" function, defaulting to
1410 * the legacy UNIX filesystem behaviour. If it tells
1411 * us to evict inode, do so. Otherwise, retain inode
1412 * in cache if fs is alive, sync and evict if fs is
1415 static void iput_final(struct inode *inode)
1417 struct super_block *sb = inode->i_sb;
1418 const struct super_operations *op = inode->i_sb->s_op;
1421 spin_lock(&inode->i_lock);
1422 WARN_ON(inode->i_state & I_NEW);
1424 if (op && op->drop_inode)
1425 drop = op->drop_inode(inode);
1427 drop = generic_drop_inode(inode);
1429 if (!drop && (sb->s_flags & MS_ACTIVE)) {
1430 inode->i_state |= I_REFERENCED;
1431 if (!(inode->i_state & (I_DIRTY|I_SYNC)))
1432 inode_lru_list_add(inode);
1433 spin_unlock(&inode->i_lock);
1434 spin_unlock(&inode_lock);
1439 inode->i_state |= I_WILL_FREE;
1440 spin_unlock(&inode->i_lock);
1441 spin_unlock(&inode_lock);
1442 write_inode_now(inode, 1);
1443 spin_lock(&inode_lock);
1444 spin_lock(&inode->i_lock);
1445 WARN_ON(inode->i_state & I_NEW);
1446 inode->i_state &= ~I_WILL_FREE;
1449 inode->i_state |= I_FREEING;
1450 inode_lru_list_del(inode);
1451 spin_unlock(&inode->i_lock);
1452 spin_unlock(&inode_lock);
1458 * iput - put an inode
1459 * @inode: inode to put
1461 * Puts an inode, dropping its usage count. If the inode use count hits
1462 * zero, the inode is then freed and may also be destroyed.
1464 * Consequently, iput() can sleep.
1466 void iput(struct inode *inode)
1469 BUG_ON(inode->i_state & I_CLEAR);
1471 if (atomic_dec_and_lock(&inode->i_count, &inode_lock))
1475 EXPORT_SYMBOL(iput);
1478 * bmap - find a block number in a file
1479 * @inode: inode of file
1480 * @block: block to find
1482 * Returns the block number on the device holding the inode that
1483 * is the disk block number for the block of the file requested.
1484 * That is, asked for block 4 of inode 1 the function will return the
1485 * disk block relative to the disk start that holds that block of the
1488 sector_t bmap(struct inode *inode, sector_t block)
1491 if (inode->i_mapping->a_ops->bmap)
1492 res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block);
1495 EXPORT_SYMBOL(bmap);
1498 * With relative atime, only update atime if the previous atime is
1499 * earlier than either the ctime or mtime or if at least a day has
1500 * passed since the last atime update.
1502 static int relatime_need_update(struct vfsmount *mnt, struct inode *inode,
1503 struct timespec now)
1506 if (!(mnt->mnt_flags & MNT_RELATIME))
1509 * Is mtime younger than atime? If yes, update atime:
1511 if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0)
1514 * Is ctime younger than atime? If yes, update atime:
1516 if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0)
1520 * Is the previous atime value older than a day? If yes,
1523 if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60)
1526 * Good, we can skip the atime update:
1532 * touch_atime - update the access time
1533 * @mnt: mount the inode is accessed on
1534 * @dentry: dentry accessed
1536 * Update the accessed time on an inode and mark it for writeback.
1537 * This function automatically handles read only file systems and media,
1538 * as well as the "noatime" flag and inode specific "noatime" markers.
1540 void touch_atime(struct vfsmount *mnt, struct dentry *dentry)
1542 struct inode *inode = dentry->d_inode;
1543 struct timespec now;
1545 if (inode->i_flags & S_NOATIME)
1547 if (IS_NOATIME(inode))
1549 if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode))
1552 if (mnt->mnt_flags & MNT_NOATIME)
1554 if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode))
1557 now = current_fs_time(inode->i_sb);
1559 if (!relatime_need_update(mnt, inode, now))
1562 if (timespec_equal(&inode->i_atime, &now))
1565 if (mnt_want_write(mnt))
1568 inode->i_atime = now;
1569 mark_inode_dirty_sync(inode);
1570 mnt_drop_write(mnt);
1572 EXPORT_SYMBOL(touch_atime);
1575 * file_update_time - update mtime and ctime time
1576 * @file: file accessed
1578 * Update the mtime and ctime members of an inode and mark the inode
1579 * for writeback. Note that this function is meant exclusively for
1580 * usage in the file write path of filesystems, and filesystems may
1581 * choose to explicitly ignore update via this function with the
1582 * S_NOCMTIME inode flag, e.g. for network filesystem where these
1583 * timestamps are handled by the server.
1586 void file_update_time(struct file *file)
1588 struct inode *inode = file->f_path.dentry->d_inode;
1589 struct timespec now;
1590 enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0;
1592 /* First try to exhaust all avenues to not sync */
1593 if (IS_NOCMTIME(inode))
1596 now = current_fs_time(inode->i_sb);
1597 if (!timespec_equal(&inode->i_mtime, &now))
1600 if (!timespec_equal(&inode->i_ctime, &now))
1603 if (IS_I_VERSION(inode))
1604 sync_it |= S_VERSION;
1609 /* Finally allowed to write? Takes lock. */
1610 if (mnt_want_write_file(file))
1613 /* Only change inode inside the lock region */
1614 if (sync_it & S_VERSION)
1615 inode_inc_iversion(inode);
1616 if (sync_it & S_CTIME)
1617 inode->i_ctime = now;
1618 if (sync_it & S_MTIME)
1619 inode->i_mtime = now;
1620 mark_inode_dirty_sync(inode);
1621 mnt_drop_write(file->f_path.mnt);
1623 EXPORT_SYMBOL(file_update_time);
1625 int inode_needs_sync(struct inode *inode)
1629 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
1633 EXPORT_SYMBOL(inode_needs_sync);
1635 int inode_wait(void *word)
1640 EXPORT_SYMBOL(inode_wait);
1643 * If we try to find an inode in the inode hash while it is being
1644 * deleted, we have to wait until the filesystem completes its
1645 * deletion before reporting that it isn't found. This function waits
1646 * until the deletion _might_ have completed. Callers are responsible
1647 * to recheck inode state.
1649 * It doesn't matter if I_NEW is not set initially, a call to
1650 * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
1653 static void __wait_on_freeing_inode(struct inode *inode)
1655 wait_queue_head_t *wq;
1656 DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW);
1657 wq = bit_waitqueue(&inode->i_state, __I_NEW);
1658 prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE);
1659 spin_unlock(&inode->i_lock);
1660 spin_unlock(&inode_lock);
1662 finish_wait(wq, &wait.wait);
1663 spin_lock(&inode_lock);
1666 static __initdata unsigned long ihash_entries;
1667 static int __init set_ihash_entries(char *str)
1671 ihash_entries = simple_strtoul(str, &str, 0);
1674 __setup("ihash_entries=", set_ihash_entries);
1677 * Initialize the waitqueues and inode hash table.
1679 void __init inode_init_early(void)
1683 /* If hashes are distributed across NUMA nodes, defer
1684 * hash allocation until vmalloc space is available.
1690 alloc_large_system_hash("Inode-cache",
1691 sizeof(struct hlist_head),
1699 for (loop = 0; loop < (1 << i_hash_shift); loop++)
1700 INIT_HLIST_HEAD(&inode_hashtable[loop]);
1703 void __init inode_init(void)
1707 /* inode slab cache */
1708 inode_cachep = kmem_cache_create("inode_cache",
1709 sizeof(struct inode),
1711 (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|
1714 register_shrinker(&icache_shrinker);
1716 /* Hash may have been set up in inode_init_early */
1721 alloc_large_system_hash("Inode-cache",
1722 sizeof(struct hlist_head),
1730 for (loop = 0; loop < (1 << i_hash_shift); loop++)
1731 INIT_HLIST_HEAD(&inode_hashtable[loop]);
1734 void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev)
1736 inode->i_mode = mode;
1737 if (S_ISCHR(mode)) {
1738 inode->i_fop = &def_chr_fops;
1739 inode->i_rdev = rdev;
1740 } else if (S_ISBLK(mode)) {
1741 inode->i_fop = &def_blk_fops;
1742 inode->i_rdev = rdev;
1743 } else if (S_ISFIFO(mode))
1744 inode->i_fop = &def_fifo_fops;
1745 else if (S_ISSOCK(mode))
1746 inode->i_fop = &bad_sock_fops;
1748 printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for"
1749 " inode %s:%lu\n", mode, inode->i_sb->s_id,
1752 EXPORT_SYMBOL(init_special_inode);
1755 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
1757 * @dir: Directory inode
1758 * @mode: mode of the new inode
1760 void inode_init_owner(struct inode *inode, const struct inode *dir,
1763 inode->i_uid = current_fsuid();
1764 if (dir && dir->i_mode & S_ISGID) {
1765 inode->i_gid = dir->i_gid;
1769 inode->i_gid = current_fsgid();
1770 inode->i_mode = mode;
1772 EXPORT_SYMBOL(inode_init_owner);
1775 * inode_owner_or_capable - check current task permissions to inode
1776 * @inode: inode being checked
1778 * Return true if current either has CAP_FOWNER to the inode, or
1781 bool inode_owner_or_capable(const struct inode *inode)
1783 struct user_namespace *ns = inode_userns(inode);
1785 if (current_user_ns() == ns && current_fsuid() == inode->i_uid)
1787 if (ns_capable(ns, CAP_FOWNER))
1791 EXPORT_SYMBOL(inode_owner_or_capable);