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>
32 * inode locking rules.
34 * inode->i_lock protects:
35 * inode->i_state, inode->i_hash, __iget()
36 * inode_lru_lock protects:
37 * inode_lru, inode->i_lru
38 * inode_sb_list_lock protects:
39 * sb->s_inodes, inode->i_sb_list
40 * inode_wb_list_lock protects:
41 * bdi->wb.b_{dirty,io,more_io}, inode->i_wb_list
56 * This is needed for the following functions:
60 * FIXME: remove all knowledge of the buffer layer from this file
62 #include <linux/buffer_head.h>
65 * New inode.c implementation.
67 * This implementation has the basic premise of trying
68 * to be extremely low-overhead and SMP-safe, yet be
69 * simple enough to be "obviously correct".
74 /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */
76 /* #define INODE_PARANOIA 1 */
77 /* #define INODE_DEBUG 1 */
80 * Inode lookup is no longer as critical as it used to be:
81 * most of the lookups are going to be through the dcache.
83 #define I_HASHBITS i_hash_shift
84 #define I_HASHMASK i_hash_mask
86 static unsigned int i_hash_mask __read_mostly;
87 static unsigned int i_hash_shift __read_mostly;
90 * Each inode can be on two separate lists. One is
91 * the hash list of the inode, used for lookups. The
92 * other linked list is the "type" list:
93 * "in_use" - valid inode, i_count > 0, i_nlink > 0
94 * "dirty" - as "in_use" but also dirty
95 * "unused" - valid inode, i_count = 0
97 * A "dirty" list is maintained for each super block,
98 * allowing for low-overhead inode sync() operations.
101 static LIST_HEAD(inode_lru);
102 static DEFINE_SPINLOCK(inode_lru_lock);
103 static struct hlist_head *inode_hashtable __read_mostly;
106 * A simple spinlock to protect the list manipulations.
108 * NOTE! You also have to own the lock if you change
109 * the i_state of an inode while it is in use..
111 DEFINE_SPINLOCK(inode_lock);
113 __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_sb_list_lock);
114 __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_wb_list_lock);
117 * iprune_sem provides exclusion between the icache shrinking and the
120 * We don't actually need it to protect anything in the umount path,
121 * but only need to cycle through it to make sure any inode that
122 * prune_icache took off the LRU list has been fully torn down by the
123 * time we are past evict_inodes.
125 static DECLARE_RWSEM(iprune_sem);
128 * Statistics gathering..
130 struct inodes_stat_t inodes_stat;
132 static DEFINE_PER_CPU(unsigned int, nr_inodes);
134 static struct kmem_cache *inode_cachep __read_mostly;
136 static int get_nr_inodes(void)
140 for_each_possible_cpu(i)
141 sum += per_cpu(nr_inodes, i);
142 return sum < 0 ? 0 : sum;
145 static inline int get_nr_inodes_unused(void)
147 return inodes_stat.nr_unused;
150 int get_nr_dirty_inodes(void)
152 /* not actually dirty inodes, but a wild approximation */
153 int nr_dirty = get_nr_inodes() - get_nr_inodes_unused();
154 return nr_dirty > 0 ? nr_dirty : 0;
158 * Handle nr_inode sysctl
161 int proc_nr_inodes(ctl_table *table, int write,
162 void __user *buffer, size_t *lenp, loff_t *ppos)
164 inodes_stat.nr_inodes = get_nr_inodes();
165 return proc_dointvec(table, write, buffer, lenp, ppos);
170 * inode_init_always - perform inode structure intialisation
171 * @sb: superblock inode belongs to
172 * @inode: inode to initialise
174 * These are initializations that need to be done on every inode
175 * allocation as the fields are not initialised by slab allocation.
177 int inode_init_always(struct super_block *sb, struct inode *inode)
179 static const struct address_space_operations empty_aops;
180 static const struct inode_operations empty_iops;
181 static const struct file_operations empty_fops;
182 struct address_space *const mapping = &inode->i_data;
185 inode->i_blkbits = sb->s_blocksize_bits;
187 atomic_set(&inode->i_count, 1);
188 inode->i_op = &empty_iops;
189 inode->i_fop = &empty_fops;
193 atomic_set(&inode->i_writecount, 0);
197 inode->i_generation = 0;
199 memset(&inode->i_dquot, 0, sizeof(inode->i_dquot));
201 inode->i_pipe = NULL;
202 inode->i_bdev = NULL;
203 inode->i_cdev = NULL;
205 inode->dirtied_when = 0;
207 if (security_inode_alloc(inode))
209 spin_lock_init(&inode->i_lock);
210 lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key);
212 mutex_init(&inode->i_mutex);
213 lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key);
215 init_rwsem(&inode->i_alloc_sem);
216 lockdep_set_class(&inode->i_alloc_sem, &sb->s_type->i_alloc_sem_key);
218 mapping->a_ops = &empty_aops;
219 mapping->host = inode;
221 mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE);
222 mapping->assoc_mapping = NULL;
223 mapping->backing_dev_info = &default_backing_dev_info;
224 mapping->writeback_index = 0;
227 * If the block_device provides a backing_dev_info for client
228 * inodes then use that. Otherwise the inode share the bdev's
232 struct backing_dev_info *bdi;
234 bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
235 mapping->backing_dev_info = bdi;
237 inode->i_private = NULL;
238 inode->i_mapping = mapping;
239 #ifdef CONFIG_FS_POSIX_ACL
240 inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED;
243 #ifdef CONFIG_FSNOTIFY
244 inode->i_fsnotify_mask = 0;
247 this_cpu_inc(nr_inodes);
253 EXPORT_SYMBOL(inode_init_always);
255 static struct inode *alloc_inode(struct super_block *sb)
259 if (sb->s_op->alloc_inode)
260 inode = sb->s_op->alloc_inode(sb);
262 inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL);
267 if (unlikely(inode_init_always(sb, inode))) {
268 if (inode->i_sb->s_op->destroy_inode)
269 inode->i_sb->s_op->destroy_inode(inode);
271 kmem_cache_free(inode_cachep, inode);
278 void free_inode_nonrcu(struct inode *inode)
280 kmem_cache_free(inode_cachep, inode);
282 EXPORT_SYMBOL(free_inode_nonrcu);
284 void __destroy_inode(struct inode *inode)
286 BUG_ON(inode_has_buffers(inode));
287 security_inode_free(inode);
288 fsnotify_inode_delete(inode);
289 #ifdef CONFIG_FS_POSIX_ACL
290 if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED)
291 posix_acl_release(inode->i_acl);
292 if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED)
293 posix_acl_release(inode->i_default_acl);
295 this_cpu_dec(nr_inodes);
297 EXPORT_SYMBOL(__destroy_inode);
299 static void i_callback(struct rcu_head *head)
301 struct inode *inode = container_of(head, struct inode, i_rcu);
302 INIT_LIST_HEAD(&inode->i_dentry);
303 kmem_cache_free(inode_cachep, inode);
306 static void destroy_inode(struct inode *inode)
308 BUG_ON(!list_empty(&inode->i_lru));
309 __destroy_inode(inode);
310 if (inode->i_sb->s_op->destroy_inode)
311 inode->i_sb->s_op->destroy_inode(inode);
313 call_rcu(&inode->i_rcu, i_callback);
316 void address_space_init_once(struct address_space *mapping)
318 memset(mapping, 0, sizeof(*mapping));
319 INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC);
320 spin_lock_init(&mapping->tree_lock);
321 spin_lock_init(&mapping->i_mmap_lock);
322 INIT_LIST_HEAD(&mapping->private_list);
323 spin_lock_init(&mapping->private_lock);
324 INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap);
325 INIT_LIST_HEAD(&mapping->i_mmap_nonlinear);
326 mutex_init(&mapping->unmap_mutex);
328 EXPORT_SYMBOL(address_space_init_once);
331 * These are initializations that only need to be done
332 * once, because the fields are idempotent across use
333 * of the inode, so let the slab aware of that.
335 void inode_init_once(struct inode *inode)
337 memset(inode, 0, sizeof(*inode));
338 INIT_HLIST_NODE(&inode->i_hash);
339 INIT_LIST_HEAD(&inode->i_dentry);
340 INIT_LIST_HEAD(&inode->i_devices);
341 INIT_LIST_HEAD(&inode->i_wb_list);
342 INIT_LIST_HEAD(&inode->i_lru);
343 address_space_init_once(&inode->i_data);
344 i_size_ordered_init(inode);
345 #ifdef CONFIG_FSNOTIFY
346 INIT_HLIST_HEAD(&inode->i_fsnotify_marks);
349 EXPORT_SYMBOL(inode_init_once);
351 static void init_once(void *foo)
353 struct inode *inode = (struct inode *) foo;
355 inode_init_once(inode);
359 * inode->i_lock must be held
361 void __iget(struct inode *inode)
363 atomic_inc(&inode->i_count);
367 * get additional reference to inode; caller must already hold one.
369 void ihold(struct inode *inode)
371 WARN_ON(atomic_inc_return(&inode->i_count) < 2);
373 EXPORT_SYMBOL(ihold);
375 static void inode_lru_list_add(struct inode *inode)
377 spin_lock(&inode_lru_lock);
378 if (list_empty(&inode->i_lru)) {
379 list_add(&inode->i_lru, &inode_lru);
380 inodes_stat.nr_unused++;
382 spin_unlock(&inode_lru_lock);
385 static void inode_lru_list_del(struct inode *inode)
387 spin_lock(&inode_lru_lock);
388 if (!list_empty(&inode->i_lru)) {
389 list_del_init(&inode->i_lru);
390 inodes_stat.nr_unused--;
392 spin_unlock(&inode_lru_lock);
396 * inode_sb_list_add - add inode to the superblock list of inodes
397 * @inode: inode to add
399 void inode_sb_list_add(struct inode *inode)
401 spin_lock(&inode_sb_list_lock);
402 list_add(&inode->i_sb_list, &inode->i_sb->s_inodes);
403 spin_unlock(&inode_sb_list_lock);
405 EXPORT_SYMBOL_GPL(inode_sb_list_add);
407 static inline void inode_sb_list_del(struct inode *inode)
409 spin_lock(&inode_sb_list_lock);
410 list_del_init(&inode->i_sb_list);
411 spin_unlock(&inode_sb_list_lock);
414 static unsigned long hash(struct super_block *sb, unsigned long hashval)
418 tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) /
420 tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> I_HASHBITS);
421 return tmp & I_HASHMASK;
425 * __insert_inode_hash - hash an inode
426 * @inode: unhashed inode
427 * @hashval: unsigned long value used to locate this object in the
430 * Add an inode to the inode hash for this superblock.
432 void __insert_inode_hash(struct inode *inode, unsigned long hashval)
434 struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval);
436 spin_lock(&inode_lock);
437 spin_lock(&inode->i_lock);
438 hlist_add_head(&inode->i_hash, b);
439 spin_unlock(&inode->i_lock);
440 spin_unlock(&inode_lock);
442 EXPORT_SYMBOL(__insert_inode_hash);
445 * remove_inode_hash - remove an inode from the hash
446 * @inode: inode to unhash
448 * Remove an inode from the superblock.
450 void remove_inode_hash(struct inode *inode)
452 spin_lock(&inode_lock);
453 spin_lock(&inode->i_lock);
454 hlist_del_init(&inode->i_hash);
455 spin_unlock(&inode->i_lock);
456 spin_unlock(&inode_lock);
458 EXPORT_SYMBOL(remove_inode_hash);
460 void end_writeback(struct inode *inode)
463 BUG_ON(inode->i_data.nrpages);
464 BUG_ON(!list_empty(&inode->i_data.private_list));
465 BUG_ON(!(inode->i_state & I_FREEING));
466 BUG_ON(inode->i_state & I_CLEAR);
467 inode_sync_wait(inode);
468 /* don't need i_lock here, no concurrent mods to i_state */
469 inode->i_state = I_FREEING | I_CLEAR;
471 EXPORT_SYMBOL(end_writeback);
474 * Free the inode passed in, removing it from the lists it is still connected
475 * to. We remove any pages still attached to the inode and wait for any IO that
476 * is still in progress before finally destroying the inode.
478 * An inode must already be marked I_FREEING so that we avoid the inode being
479 * moved back onto lists if we race with other code that manipulates the lists
480 * (e.g. writeback_single_inode). The caller is responsible for setting this.
482 * An inode must already be removed from the LRU list before being evicted from
483 * the cache. This should occur atomically with setting the I_FREEING state
484 * flag, so no inodes here should ever be on the LRU when being evicted.
486 static void evict(struct inode *inode)
488 const struct super_operations *op = inode->i_sb->s_op;
490 BUG_ON(!(inode->i_state & I_FREEING));
491 BUG_ON(!list_empty(&inode->i_lru));
493 inode_wb_list_del(inode);
494 inode_sb_list_del(inode);
496 if (op->evict_inode) {
497 op->evict_inode(inode);
499 if (inode->i_data.nrpages)
500 truncate_inode_pages(&inode->i_data, 0);
501 end_writeback(inode);
503 if (S_ISBLK(inode->i_mode) && inode->i_bdev)
505 if (S_ISCHR(inode->i_mode) && inode->i_cdev)
508 remove_inode_hash(inode);
510 spin_lock(&inode->i_lock);
511 wake_up_bit(&inode->i_state, __I_NEW);
512 BUG_ON(inode->i_state != (I_FREEING | I_CLEAR));
513 spin_unlock(&inode->i_lock);
515 destroy_inode(inode);
519 * dispose_list - dispose of the contents of a local list
520 * @head: the head of the list to free
522 * Dispose-list gets a local list with local inodes in it, so it doesn't
523 * need to worry about list corruption and SMP locks.
525 static void dispose_list(struct list_head *head)
527 while (!list_empty(head)) {
530 inode = list_first_entry(head, struct inode, i_lru);
531 list_del_init(&inode->i_lru);
538 * evict_inodes - evict all evictable inodes for a superblock
539 * @sb: superblock to operate on
541 * Make sure that no inodes with zero refcount are retained. This is
542 * called by superblock shutdown after having MS_ACTIVE flag removed,
543 * so any inode reaching zero refcount during or after that call will
544 * be immediately evicted.
546 void evict_inodes(struct super_block *sb)
548 struct inode *inode, *next;
551 spin_lock(&inode_sb_list_lock);
552 list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
553 if (atomic_read(&inode->i_count))
556 spin_lock(&inode->i_lock);
557 if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
558 spin_unlock(&inode->i_lock);
562 inode->i_state |= I_FREEING;
563 inode_lru_list_del(inode);
564 spin_unlock(&inode->i_lock);
565 list_add(&inode->i_lru, &dispose);
567 spin_unlock(&inode_sb_list_lock);
569 dispose_list(&dispose);
572 * Cycle through iprune_sem to make sure any inode that prune_icache
573 * moved off the list before we took the lock has been fully torn
576 down_write(&iprune_sem);
577 up_write(&iprune_sem);
581 * invalidate_inodes - attempt to free all inodes on a superblock
582 * @sb: superblock to operate on
583 * @kill_dirty: flag to guide handling of dirty inodes
585 * Attempts to free all inodes for a given superblock. If there were any
586 * busy inodes return a non-zero value, else zero.
587 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
590 int invalidate_inodes(struct super_block *sb, bool kill_dirty)
593 struct inode *inode, *next;
596 spin_lock(&inode_sb_list_lock);
597 list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
598 spin_lock(&inode->i_lock);
599 if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
600 spin_unlock(&inode->i_lock);
603 if (inode->i_state & I_DIRTY && !kill_dirty) {
604 spin_unlock(&inode->i_lock);
608 if (atomic_read(&inode->i_count)) {
609 spin_unlock(&inode->i_lock);
614 inode->i_state |= I_FREEING;
615 inode_lru_list_del(inode);
616 spin_unlock(&inode->i_lock);
617 list_add(&inode->i_lru, &dispose);
619 spin_unlock(&inode_sb_list_lock);
621 dispose_list(&dispose);
626 static int can_unuse(struct inode *inode)
628 if (inode->i_state & ~I_REFERENCED)
630 if (inode_has_buffers(inode))
632 if (atomic_read(&inode->i_count))
634 if (inode->i_data.nrpages)
640 * Scan `goal' inodes on the unused list for freeable ones. They are moved to a
641 * temporary list and then are freed outside inode_lru_lock by dispose_list().
643 * Any inodes which are pinned purely because of attached pagecache have their
644 * pagecache removed. If the inode has metadata buffers attached to
645 * mapping->private_list then try to remove them.
647 * If the inode has the I_REFERENCED flag set, then it means that it has been
648 * used recently - the flag is set in iput_final(). When we encounter such an
649 * inode, clear the flag and move it to the back of the LRU so it gets another
650 * pass through the LRU before it gets reclaimed. This is necessary because of
651 * the fact we are doing lazy LRU updates to minimise lock contention so the
652 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
653 * with this flag set because they are the inodes that are out of order.
655 static void prune_icache(int nr_to_scan)
659 unsigned long reap = 0;
661 down_read(&iprune_sem);
662 spin_lock(&inode_lru_lock);
663 for (nr_scanned = 0; nr_scanned < nr_to_scan; nr_scanned++) {
666 if (list_empty(&inode_lru))
669 inode = list_entry(inode_lru.prev, struct inode, i_lru);
672 * we are inverting the inode_lru_lock/inode->i_lock here,
673 * so use a trylock. If we fail to get the lock, just move the
674 * inode to the back of the list so we don't spin on it.
676 if (!spin_trylock(&inode->i_lock)) {
677 list_move(&inode->i_lru, &inode_lru);
682 * Referenced or dirty inodes are still in use. Give them
683 * another pass through the LRU as we canot reclaim them now.
685 if (atomic_read(&inode->i_count) ||
686 (inode->i_state & ~I_REFERENCED)) {
687 list_del_init(&inode->i_lru);
688 spin_unlock(&inode->i_lock);
689 inodes_stat.nr_unused--;
693 /* recently referenced inodes get one more pass */
694 if (inode->i_state & I_REFERENCED) {
695 inode->i_state &= ~I_REFERENCED;
696 list_move(&inode->i_lru, &inode_lru);
697 spin_unlock(&inode->i_lock);
700 if (inode_has_buffers(inode) || inode->i_data.nrpages) {
702 spin_unlock(&inode->i_lock);
703 spin_unlock(&inode_lru_lock);
704 if (remove_inode_buffers(inode))
705 reap += invalidate_mapping_pages(&inode->i_data,
708 spin_lock(&inode_lru_lock);
710 if (inode != list_entry(inode_lru.next,
711 struct inode, i_lru))
712 continue; /* wrong inode or list_empty */
713 /* avoid lock inversions with trylock */
714 if (!spin_trylock(&inode->i_lock))
716 if (!can_unuse(inode)) {
717 spin_unlock(&inode->i_lock);
721 WARN_ON(inode->i_state & I_NEW);
722 inode->i_state |= I_FREEING;
723 spin_unlock(&inode->i_lock);
725 list_move(&inode->i_lru, &freeable);
726 inodes_stat.nr_unused--;
728 if (current_is_kswapd())
729 __count_vm_events(KSWAPD_INODESTEAL, reap);
731 __count_vm_events(PGINODESTEAL, reap);
732 spin_unlock(&inode_lru_lock);
734 dispose_list(&freeable);
735 up_read(&iprune_sem);
739 * shrink_icache_memory() will attempt to reclaim some unused inodes. Here,
740 * "unused" means that no dentries are referring to the inodes: the files are
741 * not open and the dcache references to those inodes have already been
744 * This function is passed the number of inodes to scan, and it returns the
745 * total number of remaining possibly-reclaimable inodes.
747 static int shrink_icache_memory(struct shrinker *shrink, int nr, gfp_t gfp_mask)
751 * Nasty deadlock avoidance. We may hold various FS locks,
752 * and we don't want to recurse into the FS that called us
753 * in clear_inode() and friends..
755 if (!(gfp_mask & __GFP_FS))
759 return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure;
762 static struct shrinker icache_shrinker = {
763 .shrink = shrink_icache_memory,
764 .seeks = DEFAULT_SEEKS,
767 static void __wait_on_freeing_inode(struct inode *inode);
769 * Called with the inode lock held.
771 static struct inode *find_inode(struct super_block *sb,
772 struct hlist_head *head,
773 int (*test)(struct inode *, void *),
776 struct hlist_node *node;
777 struct inode *inode = NULL;
780 hlist_for_each_entry(inode, node, head, i_hash) {
781 if (inode->i_sb != sb)
783 if (!test(inode, data))
785 spin_lock(&inode->i_lock);
786 if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
787 __wait_on_freeing_inode(inode);
791 spin_unlock(&inode->i_lock);
798 * find_inode_fast is the fast path version of find_inode, see the comment at
799 * iget_locked for details.
801 static struct inode *find_inode_fast(struct super_block *sb,
802 struct hlist_head *head, unsigned long ino)
804 struct hlist_node *node;
805 struct inode *inode = NULL;
808 hlist_for_each_entry(inode, node, head, i_hash) {
809 if (inode->i_ino != ino)
811 if (inode->i_sb != sb)
813 spin_lock(&inode->i_lock);
814 if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
815 __wait_on_freeing_inode(inode);
819 spin_unlock(&inode->i_lock);
826 * Each cpu owns a range of LAST_INO_BATCH numbers.
827 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
828 * to renew the exhausted range.
830 * This does not significantly increase overflow rate because every CPU can
831 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
832 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
833 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
834 * overflow rate by 2x, which does not seem too significant.
836 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
837 * error if st_ino won't fit in target struct field. Use 32bit counter
838 * here to attempt to avoid that.
840 #define LAST_INO_BATCH 1024
841 static DEFINE_PER_CPU(unsigned int, last_ino);
843 unsigned int get_next_ino(void)
845 unsigned int *p = &get_cpu_var(last_ino);
846 unsigned int res = *p;
849 if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) {
850 static atomic_t shared_last_ino;
851 int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino);
853 res = next - LAST_INO_BATCH;
858 put_cpu_var(last_ino);
861 EXPORT_SYMBOL(get_next_ino);
864 * new_inode - obtain an inode
867 * Allocates a new inode for given superblock. The default gfp_mask
868 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
869 * If HIGHMEM pages are unsuitable or it is known that pages allocated
870 * for the page cache are not reclaimable or migratable,
871 * mapping_set_gfp_mask() must be called with suitable flags on the
872 * newly created inode's mapping
875 struct inode *new_inode(struct super_block *sb)
879 spin_lock_prefetch(&inode_sb_list_lock);
881 inode = alloc_inode(sb);
883 spin_lock(&inode->i_lock);
885 spin_unlock(&inode->i_lock);
886 inode_sb_list_add(inode);
890 EXPORT_SYMBOL(new_inode);
893 * unlock_new_inode - clear the I_NEW state and wake up any waiters
894 * @inode: new inode to unlock
896 * Called when the inode is fully initialised to clear the new state of the
897 * inode and wake up anyone waiting for the inode to finish initialisation.
899 void unlock_new_inode(struct inode *inode)
901 #ifdef CONFIG_DEBUG_LOCK_ALLOC
902 if (S_ISDIR(inode->i_mode)) {
903 struct file_system_type *type = inode->i_sb->s_type;
905 /* Set new key only if filesystem hasn't already changed it */
906 if (!lockdep_match_class(&inode->i_mutex,
907 &type->i_mutex_key)) {
909 * ensure nobody is actually holding i_mutex
911 mutex_destroy(&inode->i_mutex);
912 mutex_init(&inode->i_mutex);
913 lockdep_set_class(&inode->i_mutex,
914 &type->i_mutex_dir_key);
918 spin_lock(&inode->i_lock);
919 WARN_ON(!(inode->i_state & I_NEW));
920 inode->i_state &= ~I_NEW;
921 wake_up_bit(&inode->i_state, __I_NEW);
922 spin_unlock(&inode->i_lock);
924 EXPORT_SYMBOL(unlock_new_inode);
927 * This is called without the inode lock held.. Be careful.
929 * We no longer cache the sb_flags in i_flags - see fs.h
930 * -- rmk@arm.uk.linux.org
932 static struct inode *get_new_inode(struct super_block *sb,
933 struct hlist_head *head,
934 int (*test)(struct inode *, void *),
935 int (*set)(struct inode *, void *),
940 inode = alloc_inode(sb);
944 spin_lock(&inode_lock);
945 /* We released the lock, so.. */
946 old = find_inode(sb, head, test, data);
948 if (set(inode, data))
951 spin_lock(&inode->i_lock);
952 inode->i_state = I_NEW;
953 hlist_add_head(&inode->i_hash, head);
954 spin_unlock(&inode->i_lock);
955 inode_sb_list_add(inode);
956 spin_unlock(&inode_lock);
958 /* Return the locked inode with I_NEW set, the
959 * caller is responsible for filling in the contents
965 * Uhhuh, somebody else created the same inode under
966 * us. Use the old inode instead of the one we just
969 spin_unlock(&inode_lock);
970 destroy_inode(inode);
972 wait_on_inode(inode);
977 spin_unlock(&inode_lock);
978 destroy_inode(inode);
983 * get_new_inode_fast is the fast path version of get_new_inode, see the
984 * comment at iget_locked for details.
986 static struct inode *get_new_inode_fast(struct super_block *sb,
987 struct hlist_head *head, unsigned long ino)
991 inode = alloc_inode(sb);
995 spin_lock(&inode_lock);
996 /* We released the lock, so.. */
997 old = find_inode_fast(sb, head, ino);
1000 spin_lock(&inode->i_lock);
1001 inode->i_state = I_NEW;
1002 hlist_add_head(&inode->i_hash, head);
1003 spin_unlock(&inode->i_lock);
1004 inode_sb_list_add(inode);
1005 spin_unlock(&inode_lock);
1007 /* Return the locked inode with I_NEW set, the
1008 * caller is responsible for filling in the contents
1014 * Uhhuh, somebody else created the same inode under
1015 * us. Use the old inode instead of the one we just
1018 spin_unlock(&inode_lock);
1019 destroy_inode(inode);
1021 wait_on_inode(inode);
1027 * search the inode cache for a matching inode number.
1028 * If we find one, then the inode number we are trying to
1029 * allocate is not unique and so we should not use it.
1031 * Returns 1 if the inode number is unique, 0 if it is not.
1033 static int test_inode_iunique(struct super_block *sb, unsigned long ino)
1035 struct hlist_head *b = inode_hashtable + hash(sb, ino);
1036 struct hlist_node *node;
1037 struct inode *inode;
1039 hlist_for_each_entry(inode, node, b, i_hash) {
1040 if (inode->i_ino == ino && inode->i_sb == sb)
1048 * iunique - get a unique inode number
1050 * @max_reserved: highest reserved inode number
1052 * Obtain an inode number that is unique on the system for a given
1053 * superblock. This is used by file systems that have no natural
1054 * permanent inode numbering system. An inode number is returned that
1055 * is higher than the reserved limit but unique.
1058 * With a large number of inodes live on the file system this function
1059 * currently becomes quite slow.
1061 ino_t iunique(struct super_block *sb, ino_t max_reserved)
1064 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1065 * error if st_ino won't fit in target struct field. Use 32bit counter
1066 * here to attempt to avoid that.
1068 static DEFINE_SPINLOCK(iunique_lock);
1069 static unsigned int counter;
1072 spin_lock(&inode_lock);
1073 spin_lock(&iunique_lock);
1075 if (counter <= max_reserved)
1076 counter = max_reserved + 1;
1078 } while (!test_inode_iunique(sb, res));
1079 spin_unlock(&iunique_lock);
1080 spin_unlock(&inode_lock);
1084 EXPORT_SYMBOL(iunique);
1086 struct inode *igrab(struct inode *inode)
1088 spin_lock(&inode->i_lock);
1089 if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) {
1091 spin_unlock(&inode->i_lock);
1093 spin_unlock(&inode->i_lock);
1095 * Handle the case where s_op->clear_inode is not been
1096 * called yet, and somebody is calling igrab
1097 * while the inode is getting freed.
1103 EXPORT_SYMBOL(igrab);
1106 * ifind - internal function, you want ilookup5() or iget5().
1107 * @sb: super block of file system to search
1108 * @head: the head of the list to search
1109 * @test: callback used for comparisons between inodes
1110 * @data: opaque data pointer to pass to @test
1111 * @wait: if true wait for the inode to be unlocked, if false do not
1113 * ifind() searches for the inode specified by @data in the inode
1114 * cache. This is a generalized version of ifind_fast() for file systems where
1115 * the inode number is not sufficient for unique identification of an inode.
1117 * If the inode is in the cache, the inode is returned with an incremented
1120 * Otherwise NULL is returned.
1122 * Note, @test is called with the inode_lock held, so can't sleep.
1124 static struct inode *ifind(struct super_block *sb,
1125 struct hlist_head *head, int (*test)(struct inode *, void *),
1126 void *data, const int wait)
1128 struct inode *inode;
1130 spin_lock(&inode_lock);
1131 inode = find_inode(sb, head, test, data);
1133 spin_unlock(&inode_lock);
1135 wait_on_inode(inode);
1138 spin_unlock(&inode_lock);
1143 * ifind_fast - internal function, you want ilookup() or iget().
1144 * @sb: super block of file system to search
1145 * @head: head of the list to search
1146 * @ino: inode number to search for
1148 * ifind_fast() searches for the inode @ino in the inode cache. This is for
1149 * file systems where the inode number is sufficient for unique identification
1152 * If the inode is in the cache, the inode is returned with an incremented
1155 * Otherwise NULL is returned.
1157 static struct inode *ifind_fast(struct super_block *sb,
1158 struct hlist_head *head, unsigned long ino)
1160 struct inode *inode;
1162 spin_lock(&inode_lock);
1163 inode = find_inode_fast(sb, head, ino);
1165 spin_unlock(&inode_lock);
1166 wait_on_inode(inode);
1169 spin_unlock(&inode_lock);
1174 * ilookup5_nowait - search for an inode in the inode cache
1175 * @sb: super block of file system to search
1176 * @hashval: hash value (usually inode number) to search for
1177 * @test: callback used for comparisons between inodes
1178 * @data: opaque data pointer to pass to @test
1180 * ilookup5() uses ifind() to search for the inode specified by @hashval and
1181 * @data in the inode cache. This is a generalized version of ilookup() for
1182 * file systems where the inode number is not sufficient for unique
1183 * identification of an inode.
1185 * If the inode is in the cache, the inode is returned with an incremented
1186 * reference count. Note, the inode lock is not waited upon so you have to be
1187 * very careful what you do with the returned inode. You probably should be
1188 * using ilookup5() instead.
1190 * Otherwise NULL is returned.
1192 * Note, @test is called with the inode_lock held, so can't sleep.
1194 struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval,
1195 int (*test)(struct inode *, void *), void *data)
1197 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1199 return ifind(sb, head, test, data, 0);
1201 EXPORT_SYMBOL(ilookup5_nowait);
1204 * ilookup5 - search for an inode in the inode cache
1205 * @sb: super block of file system to search
1206 * @hashval: hash value (usually inode number) to search for
1207 * @test: callback used for comparisons between inodes
1208 * @data: opaque data pointer to pass to @test
1210 * ilookup5() uses ifind() to search for the inode specified by @hashval and
1211 * @data in the inode cache. This is a generalized version of ilookup() for
1212 * file systems where the inode number is not sufficient for unique
1213 * identification of an inode.
1215 * If the inode is in the cache, the inode lock is waited upon and the inode is
1216 * returned with an incremented reference count.
1218 * Otherwise NULL is returned.
1220 * Note, @test is called with the inode_lock held, so can't sleep.
1222 struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
1223 int (*test)(struct inode *, void *), void *data)
1225 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1227 return ifind(sb, head, test, data, 1);
1229 EXPORT_SYMBOL(ilookup5);
1232 * ilookup - search for an inode in the inode cache
1233 * @sb: super block of file system to search
1234 * @ino: inode number to search for
1236 * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache.
1237 * This is for file systems where the inode number is sufficient for unique
1238 * identification of an inode.
1240 * If the inode is in the cache, the inode is returned with an incremented
1243 * Otherwise NULL is returned.
1245 struct inode *ilookup(struct super_block *sb, unsigned long ino)
1247 struct hlist_head *head = inode_hashtable + hash(sb, ino);
1249 return ifind_fast(sb, head, ino);
1251 EXPORT_SYMBOL(ilookup);
1254 * iget5_locked - obtain an inode from a mounted file system
1255 * @sb: super block of file system
1256 * @hashval: hash value (usually inode number) to get
1257 * @test: callback used for comparisons between inodes
1258 * @set: callback used to initialize a new struct inode
1259 * @data: opaque data pointer to pass to @test and @set
1261 * iget5_locked() uses ifind() to search for the inode specified by @hashval
1262 * and @data in the inode cache and if present it is returned with an increased
1263 * reference count. This is a generalized version of iget_locked() for file
1264 * systems where the inode number is not sufficient for unique identification
1267 * If the inode is not in cache, get_new_inode() is called to allocate a new
1268 * inode and this is returned locked, hashed, and with the I_NEW flag set. The
1269 * file system gets to fill it in before unlocking it via unlock_new_inode().
1271 * Note both @test and @set are called with the inode_lock held, so can't sleep.
1273 struct inode *iget5_locked(struct super_block *sb, unsigned long hashval,
1274 int (*test)(struct inode *, void *),
1275 int (*set)(struct inode *, void *), void *data)
1277 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1278 struct inode *inode;
1280 inode = ifind(sb, head, test, data, 1);
1284 * get_new_inode() will do the right thing, re-trying the search
1285 * in case it had to block at any point.
1287 return get_new_inode(sb, head, test, set, data);
1289 EXPORT_SYMBOL(iget5_locked);
1292 * iget_locked - obtain an inode from a mounted file system
1293 * @sb: super block of file system
1294 * @ino: inode number to get
1296 * iget_locked() uses ifind_fast() to search for the inode specified by @ino in
1297 * the inode cache and if present it is returned with an increased reference
1298 * count. This is for file systems where the inode number is sufficient for
1299 * unique identification of an inode.
1301 * If the inode is not in cache, get_new_inode_fast() is called to allocate a
1302 * new inode and this is returned locked, hashed, and with the I_NEW flag set.
1303 * The file system gets to fill it in before unlocking it via
1304 * unlock_new_inode().
1306 struct inode *iget_locked(struct super_block *sb, unsigned long ino)
1308 struct hlist_head *head = inode_hashtable + hash(sb, ino);
1309 struct inode *inode;
1311 inode = ifind_fast(sb, head, ino);
1315 * get_new_inode_fast() will do the right thing, re-trying the search
1316 * in case it had to block at any point.
1318 return get_new_inode_fast(sb, head, ino);
1320 EXPORT_SYMBOL(iget_locked);
1322 int insert_inode_locked(struct inode *inode)
1324 struct super_block *sb = inode->i_sb;
1325 ino_t ino = inode->i_ino;
1326 struct hlist_head *head = inode_hashtable + hash(sb, ino);
1329 struct hlist_node *node;
1330 struct inode *old = NULL;
1331 spin_lock(&inode_lock);
1332 hlist_for_each_entry(old, node, head, i_hash) {
1333 if (old->i_ino != ino)
1335 if (old->i_sb != sb)
1337 spin_lock(&old->i_lock);
1338 if (old->i_state & (I_FREEING|I_WILL_FREE)) {
1339 spin_unlock(&old->i_lock);
1344 if (likely(!node)) {
1345 spin_lock(&inode->i_lock);
1346 inode->i_state |= I_NEW;
1347 hlist_add_head(&inode->i_hash, head);
1348 spin_unlock(&inode->i_lock);
1349 spin_unlock(&inode_lock);
1353 spin_unlock(&old->i_lock);
1354 spin_unlock(&inode_lock);
1356 if (unlikely(!inode_unhashed(old))) {
1363 EXPORT_SYMBOL(insert_inode_locked);
1365 int insert_inode_locked4(struct inode *inode, unsigned long hashval,
1366 int (*test)(struct inode *, void *), void *data)
1368 struct super_block *sb = inode->i_sb;
1369 struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1372 struct hlist_node *node;
1373 struct inode *old = NULL;
1375 spin_lock(&inode_lock);
1376 hlist_for_each_entry(old, node, head, i_hash) {
1377 if (old->i_sb != sb)
1379 if (!test(old, data))
1381 spin_lock(&old->i_lock);
1382 if (old->i_state & (I_FREEING|I_WILL_FREE)) {
1383 spin_unlock(&old->i_lock);
1388 if (likely(!node)) {
1389 spin_lock(&inode->i_lock);
1390 inode->i_state |= I_NEW;
1391 hlist_add_head(&inode->i_hash, head);
1392 spin_unlock(&inode->i_lock);
1393 spin_unlock(&inode_lock);
1397 spin_unlock(&old->i_lock);
1398 spin_unlock(&inode_lock);
1400 if (unlikely(!inode_unhashed(old))) {
1407 EXPORT_SYMBOL(insert_inode_locked4);
1410 int generic_delete_inode(struct inode *inode)
1414 EXPORT_SYMBOL(generic_delete_inode);
1417 * Normal UNIX filesystem behaviour: delete the
1418 * inode when the usage count drops to zero, and
1421 int generic_drop_inode(struct inode *inode)
1423 return !inode->i_nlink || inode_unhashed(inode);
1425 EXPORT_SYMBOL_GPL(generic_drop_inode);
1428 * Called when we're dropping the last reference
1431 * Call the FS "drop_inode()" function, defaulting to
1432 * the legacy UNIX filesystem behaviour. If it tells
1433 * us to evict inode, do so. Otherwise, retain inode
1434 * in cache if fs is alive, sync and evict if fs is
1437 static void iput_final(struct inode *inode)
1439 struct super_block *sb = inode->i_sb;
1440 const struct super_operations *op = inode->i_sb->s_op;
1443 WARN_ON(inode->i_state & I_NEW);
1445 if (op && op->drop_inode)
1446 drop = op->drop_inode(inode);
1448 drop = generic_drop_inode(inode);
1450 if (!drop && (sb->s_flags & MS_ACTIVE)) {
1451 inode->i_state |= I_REFERENCED;
1452 if (!(inode->i_state & (I_DIRTY|I_SYNC)))
1453 inode_lru_list_add(inode);
1454 spin_unlock(&inode->i_lock);
1459 inode->i_state |= I_WILL_FREE;
1460 spin_unlock(&inode->i_lock);
1461 write_inode_now(inode, 1);
1462 spin_lock(&inode->i_lock);
1463 WARN_ON(inode->i_state & I_NEW);
1464 inode->i_state &= ~I_WILL_FREE;
1467 inode->i_state |= I_FREEING;
1468 inode_lru_list_del(inode);
1469 spin_unlock(&inode->i_lock);
1475 * iput - put an inode
1476 * @inode: inode to put
1478 * Puts an inode, dropping its usage count. If the inode use count hits
1479 * zero, the inode is then freed and may also be destroyed.
1481 * Consequently, iput() can sleep.
1483 void iput(struct inode *inode)
1486 BUG_ON(inode->i_state & I_CLEAR);
1488 if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock))
1492 EXPORT_SYMBOL(iput);
1495 * bmap - find a block number in a file
1496 * @inode: inode of file
1497 * @block: block to find
1499 * Returns the block number on the device holding the inode that
1500 * is the disk block number for the block of the file requested.
1501 * That is, asked for block 4 of inode 1 the function will return the
1502 * disk block relative to the disk start that holds that block of the
1505 sector_t bmap(struct inode *inode, sector_t block)
1508 if (inode->i_mapping->a_ops->bmap)
1509 res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block);
1512 EXPORT_SYMBOL(bmap);
1515 * With relative atime, only update atime if the previous atime is
1516 * earlier than either the ctime or mtime or if at least a day has
1517 * passed since the last atime update.
1519 static int relatime_need_update(struct vfsmount *mnt, struct inode *inode,
1520 struct timespec now)
1523 if (!(mnt->mnt_flags & MNT_RELATIME))
1526 * Is mtime younger than atime? If yes, update atime:
1528 if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0)
1531 * Is ctime younger than atime? If yes, update atime:
1533 if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0)
1537 * Is the previous atime value older than a day? If yes,
1540 if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60)
1543 * Good, we can skip the atime update:
1549 * touch_atime - update the access time
1550 * @mnt: mount the inode is accessed on
1551 * @dentry: dentry accessed
1553 * Update the accessed time on an inode and mark it for writeback.
1554 * This function automatically handles read only file systems and media,
1555 * as well as the "noatime" flag and inode specific "noatime" markers.
1557 void touch_atime(struct vfsmount *mnt, struct dentry *dentry)
1559 struct inode *inode = dentry->d_inode;
1560 struct timespec now;
1562 if (inode->i_flags & S_NOATIME)
1564 if (IS_NOATIME(inode))
1566 if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode))
1569 if (mnt->mnt_flags & MNT_NOATIME)
1571 if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode))
1574 now = current_fs_time(inode->i_sb);
1576 if (!relatime_need_update(mnt, inode, now))
1579 if (timespec_equal(&inode->i_atime, &now))
1582 if (mnt_want_write(mnt))
1585 inode->i_atime = now;
1586 mark_inode_dirty_sync(inode);
1587 mnt_drop_write(mnt);
1589 EXPORT_SYMBOL(touch_atime);
1592 * file_update_time - update mtime and ctime time
1593 * @file: file accessed
1595 * Update the mtime and ctime members of an inode and mark the inode
1596 * for writeback. Note that this function is meant exclusively for
1597 * usage in the file write path of filesystems, and filesystems may
1598 * choose to explicitly ignore update via this function with the
1599 * S_NOCMTIME inode flag, e.g. for network filesystem where these
1600 * timestamps are handled by the server.
1603 void file_update_time(struct file *file)
1605 struct inode *inode = file->f_path.dentry->d_inode;
1606 struct timespec now;
1607 enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0;
1609 /* First try to exhaust all avenues to not sync */
1610 if (IS_NOCMTIME(inode))
1613 now = current_fs_time(inode->i_sb);
1614 if (!timespec_equal(&inode->i_mtime, &now))
1617 if (!timespec_equal(&inode->i_ctime, &now))
1620 if (IS_I_VERSION(inode))
1621 sync_it |= S_VERSION;
1626 /* Finally allowed to write? Takes lock. */
1627 if (mnt_want_write_file(file))
1630 /* Only change inode inside the lock region */
1631 if (sync_it & S_VERSION)
1632 inode_inc_iversion(inode);
1633 if (sync_it & S_CTIME)
1634 inode->i_ctime = now;
1635 if (sync_it & S_MTIME)
1636 inode->i_mtime = now;
1637 mark_inode_dirty_sync(inode);
1638 mnt_drop_write(file->f_path.mnt);
1640 EXPORT_SYMBOL(file_update_time);
1642 int inode_needs_sync(struct inode *inode)
1646 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
1650 EXPORT_SYMBOL(inode_needs_sync);
1652 int inode_wait(void *word)
1657 EXPORT_SYMBOL(inode_wait);
1660 * If we try to find an inode in the inode hash while it is being
1661 * deleted, we have to wait until the filesystem completes its
1662 * deletion before reporting that it isn't found. This function waits
1663 * until the deletion _might_ have completed. Callers are responsible
1664 * to recheck inode state.
1666 * It doesn't matter if I_NEW is not set initially, a call to
1667 * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
1670 static void __wait_on_freeing_inode(struct inode *inode)
1672 wait_queue_head_t *wq;
1673 DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW);
1674 wq = bit_waitqueue(&inode->i_state, __I_NEW);
1675 prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE);
1676 spin_unlock(&inode->i_lock);
1677 spin_unlock(&inode_lock);
1679 finish_wait(wq, &wait.wait);
1680 spin_lock(&inode_lock);
1683 static __initdata unsigned long ihash_entries;
1684 static int __init set_ihash_entries(char *str)
1688 ihash_entries = simple_strtoul(str, &str, 0);
1691 __setup("ihash_entries=", set_ihash_entries);
1694 * Initialize the waitqueues and inode hash table.
1696 void __init inode_init_early(void)
1700 /* If hashes are distributed across NUMA nodes, defer
1701 * hash allocation until vmalloc space is available.
1707 alloc_large_system_hash("Inode-cache",
1708 sizeof(struct hlist_head),
1716 for (loop = 0; loop < (1 << i_hash_shift); loop++)
1717 INIT_HLIST_HEAD(&inode_hashtable[loop]);
1720 void __init inode_init(void)
1724 /* inode slab cache */
1725 inode_cachep = kmem_cache_create("inode_cache",
1726 sizeof(struct inode),
1728 (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|
1731 register_shrinker(&icache_shrinker);
1733 /* Hash may have been set up in inode_init_early */
1738 alloc_large_system_hash("Inode-cache",
1739 sizeof(struct hlist_head),
1747 for (loop = 0; loop < (1 << i_hash_shift); loop++)
1748 INIT_HLIST_HEAD(&inode_hashtable[loop]);
1751 void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev)
1753 inode->i_mode = mode;
1754 if (S_ISCHR(mode)) {
1755 inode->i_fop = &def_chr_fops;
1756 inode->i_rdev = rdev;
1757 } else if (S_ISBLK(mode)) {
1758 inode->i_fop = &def_blk_fops;
1759 inode->i_rdev = rdev;
1760 } else if (S_ISFIFO(mode))
1761 inode->i_fop = &def_fifo_fops;
1762 else if (S_ISSOCK(mode))
1763 inode->i_fop = &bad_sock_fops;
1765 printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for"
1766 " inode %s:%lu\n", mode, inode->i_sb->s_id,
1769 EXPORT_SYMBOL(init_special_inode);
1772 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
1774 * @dir: Directory inode
1775 * @mode: mode of the new inode
1777 void inode_init_owner(struct inode *inode, const struct inode *dir,
1780 inode->i_uid = current_fsuid();
1781 if (dir && dir->i_mode & S_ISGID) {
1782 inode->i_gid = dir->i_gid;
1786 inode->i_gid = current_fsgid();
1787 inode->i_mode = mode;
1789 EXPORT_SYMBOL(inode_init_owner);
1792 * inode_owner_or_capable - check current task permissions to inode
1793 * @inode: inode being checked
1795 * Return true if current either has CAP_FOWNER to the inode, or
1798 bool inode_owner_or_capable(const struct inode *inode)
1800 struct user_namespace *ns = inode_userns(inode);
1802 if (current_user_ns() == ns && current_fsuid() == inode->i_uid)
1804 if (ns_capable(ns, CAP_FOWNER))
1808 EXPORT_SYMBOL(inode_owner_or_capable);