1 // SPDX-License-Identifier: GPL-2.0-only
3 * (C) 1997 Linus Torvalds
4 * (C) 1999 Andrea Arcangeli <andrea@suse.de> (dynamic inode allocation)
6 #include <linux/export.h>
8 #include <linux/filelock.h>
10 #include <linux/backing-dev.h>
11 #include <linux/hash.h>
12 #include <linux/swap.h>
13 #include <linux/security.h>
14 #include <linux/cdev.h>
15 #include <linux/memblock.h>
16 #include <linux/fsnotify.h>
17 #include <linux/mount.h>
18 #include <linux/posix_acl.h>
19 #include <linux/prefetch.h>
20 #include <linux/buffer_head.h> /* for inode_has_buffers */
21 #include <linux/ratelimit.h>
22 #include <linux/list_lru.h>
23 #include <linux/iversion.h>
24 #include <trace/events/writeback.h>
28 * Inode locking rules:
30 * inode->i_lock protects:
31 * inode->i_state, inode->i_hash, __iget(), inode->i_io_list
32 * Inode LRU list locks protect:
33 * inode->i_sb->s_inode_lru, inode->i_lru
34 * inode->i_sb->s_inode_list_lock protects:
35 * inode->i_sb->s_inodes, inode->i_sb_list
36 * bdi->wb.list_lock protects:
37 * bdi->wb.b_{dirty,io,more_io,dirty_time}, inode->i_io_list
38 * inode_hash_lock protects:
39 * inode_hashtable, inode->i_hash
43 * inode->i_sb->s_inode_list_lock
45 * Inode LRU list locks
51 * inode->i_sb->s_inode_list_lock
58 static unsigned int i_hash_mask __read_mostly
;
59 static unsigned int i_hash_shift __read_mostly
;
60 static struct hlist_head
*inode_hashtable __read_mostly
;
61 static __cacheline_aligned_in_smp
DEFINE_SPINLOCK(inode_hash_lock
);
64 * Empty aops. Can be used for the cases where the user does not
65 * define any of the address_space operations.
67 const struct address_space_operations empty_aops
= {
69 EXPORT_SYMBOL(empty_aops
);
71 static DEFINE_PER_CPU(unsigned long, nr_inodes
);
72 static DEFINE_PER_CPU(unsigned long, nr_unused
);
74 static struct kmem_cache
*inode_cachep __read_mostly
;
76 static long get_nr_inodes(void)
80 for_each_possible_cpu(i
)
81 sum
+= per_cpu(nr_inodes
, i
);
82 return sum
< 0 ? 0 : sum
;
85 static inline long get_nr_inodes_unused(void)
89 for_each_possible_cpu(i
)
90 sum
+= per_cpu(nr_unused
, i
);
91 return sum
< 0 ? 0 : sum
;
94 long get_nr_dirty_inodes(void)
96 /* not actually dirty inodes, but a wild approximation */
97 long nr_dirty
= get_nr_inodes() - get_nr_inodes_unused();
98 return nr_dirty
> 0 ? nr_dirty
: 0;
102 * Handle nr_inode sysctl
106 * Statistics gathering..
108 static struct inodes_stat_t inodes_stat
;
110 static int proc_nr_inodes(struct ctl_table
*table
, int write
, void *buffer
,
111 size_t *lenp
, loff_t
*ppos
)
113 inodes_stat
.nr_inodes
= get_nr_inodes();
114 inodes_stat
.nr_unused
= get_nr_inodes_unused();
115 return proc_doulongvec_minmax(table
, write
, buffer
, lenp
, ppos
);
118 static struct ctl_table inodes_sysctls
[] = {
120 .procname
= "inode-nr",
121 .data
= &inodes_stat
,
122 .maxlen
= 2*sizeof(long),
124 .proc_handler
= proc_nr_inodes
,
127 .procname
= "inode-state",
128 .data
= &inodes_stat
,
129 .maxlen
= 7*sizeof(long),
131 .proc_handler
= proc_nr_inodes
,
136 static int __init
init_fs_inode_sysctls(void)
138 register_sysctl_init("fs", inodes_sysctls
);
141 early_initcall(init_fs_inode_sysctls
);
144 static int no_open(struct inode
*inode
, struct file
*file
)
150 * inode_init_always - perform inode structure initialisation
151 * @sb: superblock inode belongs to
152 * @inode: inode to initialise
154 * These are initializations that need to be done on every inode
155 * allocation as the fields are not initialised by slab allocation.
157 int inode_init_always(struct super_block
*sb
, struct inode
*inode
)
159 static const struct inode_operations empty_iops
;
160 static const struct file_operations no_open_fops
= {.open
= no_open
};
161 struct address_space
*const mapping
= &inode
->i_data
;
164 inode
->i_blkbits
= sb
->s_blocksize_bits
;
166 atomic64_set(&inode
->i_sequence
, 0);
167 atomic_set(&inode
->i_count
, 1);
168 inode
->i_op
= &empty_iops
;
169 inode
->i_fop
= &no_open_fops
;
171 inode
->__i_nlink
= 1;
172 inode
->i_opflags
= 0;
174 inode
->i_opflags
|= IOP_XATTR
;
175 i_uid_write(inode
, 0);
176 i_gid_write(inode
, 0);
177 atomic_set(&inode
->i_writecount
, 0);
179 inode
->i_write_hint
= WRITE_LIFE_NOT_SET
;
182 inode
->i_generation
= 0;
183 inode
->i_pipe
= NULL
;
184 inode
->i_cdev
= NULL
;
185 inode
->i_link
= NULL
;
186 inode
->i_dir_seq
= 0;
188 inode
->dirtied_when
= 0;
190 #ifdef CONFIG_CGROUP_WRITEBACK
191 inode
->i_wb_frn_winner
= 0;
192 inode
->i_wb_frn_avg_time
= 0;
193 inode
->i_wb_frn_history
= 0;
196 spin_lock_init(&inode
->i_lock
);
197 lockdep_set_class(&inode
->i_lock
, &sb
->s_type
->i_lock_key
);
199 init_rwsem(&inode
->i_rwsem
);
200 lockdep_set_class(&inode
->i_rwsem
, &sb
->s_type
->i_mutex_key
);
202 atomic_set(&inode
->i_dio_count
, 0);
204 mapping
->a_ops
= &empty_aops
;
205 mapping
->host
= inode
;
208 atomic_set(&mapping
->i_mmap_writable
, 0);
209 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
210 atomic_set(&mapping
->nr_thps
, 0);
212 mapping_set_gfp_mask(mapping
, GFP_HIGHUSER_MOVABLE
);
213 mapping
->private_data
= NULL
;
214 mapping
->writeback_index
= 0;
215 init_rwsem(&mapping
->invalidate_lock
);
216 lockdep_set_class_and_name(&mapping
->invalidate_lock
,
217 &sb
->s_type
->invalidate_lock_key
,
218 "mapping.invalidate_lock");
219 inode
->i_private
= NULL
;
220 inode
->i_mapping
= mapping
;
221 INIT_HLIST_HEAD(&inode
->i_dentry
); /* buggered by rcu freeing */
222 #ifdef CONFIG_FS_POSIX_ACL
223 inode
->i_acl
= inode
->i_default_acl
= ACL_NOT_CACHED
;
226 #ifdef CONFIG_FSNOTIFY
227 inode
->i_fsnotify_mask
= 0;
229 inode
->i_flctx
= NULL
;
231 if (unlikely(security_inode_alloc(inode
)))
233 this_cpu_inc(nr_inodes
);
237 EXPORT_SYMBOL(inode_init_always
);
239 void free_inode_nonrcu(struct inode
*inode
)
241 kmem_cache_free(inode_cachep
, inode
);
243 EXPORT_SYMBOL(free_inode_nonrcu
);
245 static void i_callback(struct rcu_head
*head
)
247 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
248 if (inode
->free_inode
)
249 inode
->free_inode(inode
);
251 free_inode_nonrcu(inode
);
254 static struct inode
*alloc_inode(struct super_block
*sb
)
256 const struct super_operations
*ops
= sb
->s_op
;
259 if (ops
->alloc_inode
)
260 inode
= ops
->alloc_inode(sb
);
262 inode
= alloc_inode_sb(sb
, inode_cachep
, GFP_KERNEL
);
267 if (unlikely(inode_init_always(sb
, inode
))) {
268 if (ops
->destroy_inode
) {
269 ops
->destroy_inode(inode
);
270 if (!ops
->free_inode
)
273 inode
->free_inode
= ops
->free_inode
;
274 i_callback(&inode
->i_rcu
);
281 void __destroy_inode(struct inode
*inode
)
283 BUG_ON(inode_has_buffers(inode
));
284 inode_detach_wb(inode
);
285 security_inode_free(inode
);
286 fsnotify_inode_delete(inode
);
287 locks_free_lock_context(inode
);
288 if (!inode
->i_nlink
) {
289 WARN_ON(atomic_long_read(&inode
->i_sb
->s_remove_count
) == 0);
290 atomic_long_dec(&inode
->i_sb
->s_remove_count
);
293 #ifdef CONFIG_FS_POSIX_ACL
294 if (inode
->i_acl
&& !is_uncached_acl(inode
->i_acl
))
295 posix_acl_release(inode
->i_acl
);
296 if (inode
->i_default_acl
&& !is_uncached_acl(inode
->i_default_acl
))
297 posix_acl_release(inode
->i_default_acl
);
299 this_cpu_dec(nr_inodes
);
301 EXPORT_SYMBOL(__destroy_inode
);
303 static void destroy_inode(struct inode
*inode
)
305 const struct super_operations
*ops
= inode
->i_sb
->s_op
;
307 BUG_ON(!list_empty(&inode
->i_lru
));
308 __destroy_inode(inode
);
309 if (ops
->destroy_inode
) {
310 ops
->destroy_inode(inode
);
311 if (!ops
->free_inode
)
314 inode
->free_inode
= ops
->free_inode
;
315 call_rcu(&inode
->i_rcu
, i_callback
);
319 * drop_nlink - directly drop an inode's link count
322 * This is a low-level filesystem helper to replace any
323 * direct filesystem manipulation of i_nlink. In cases
324 * where we are attempting to track writes to the
325 * filesystem, a decrement to zero means an imminent
326 * write when the file is truncated and actually unlinked
329 void drop_nlink(struct inode
*inode
)
331 WARN_ON(inode
->i_nlink
== 0);
334 atomic_long_inc(&inode
->i_sb
->s_remove_count
);
336 EXPORT_SYMBOL(drop_nlink
);
339 * clear_nlink - directly zero an inode's link count
342 * This is a low-level filesystem helper to replace any
343 * direct filesystem manipulation of i_nlink. See
344 * drop_nlink() for why we care about i_nlink hitting zero.
346 void clear_nlink(struct inode
*inode
)
348 if (inode
->i_nlink
) {
349 inode
->__i_nlink
= 0;
350 atomic_long_inc(&inode
->i_sb
->s_remove_count
);
353 EXPORT_SYMBOL(clear_nlink
);
356 * set_nlink - directly set an inode's link count
358 * @nlink: new nlink (should be non-zero)
360 * This is a low-level filesystem helper to replace any
361 * direct filesystem manipulation of i_nlink.
363 void set_nlink(struct inode
*inode
, unsigned int nlink
)
368 /* Yes, some filesystems do change nlink from zero to one */
369 if (inode
->i_nlink
== 0)
370 atomic_long_dec(&inode
->i_sb
->s_remove_count
);
372 inode
->__i_nlink
= nlink
;
375 EXPORT_SYMBOL(set_nlink
);
378 * inc_nlink - directly increment an inode's link count
381 * This is a low-level filesystem helper to replace any
382 * direct filesystem manipulation of i_nlink. Currently,
383 * it is only here for parity with dec_nlink().
385 void inc_nlink(struct inode
*inode
)
387 if (unlikely(inode
->i_nlink
== 0)) {
388 WARN_ON(!(inode
->i_state
& I_LINKABLE
));
389 atomic_long_dec(&inode
->i_sb
->s_remove_count
);
394 EXPORT_SYMBOL(inc_nlink
);
396 static void __address_space_init_once(struct address_space
*mapping
)
398 xa_init_flags(&mapping
->i_pages
, XA_FLAGS_LOCK_IRQ
| XA_FLAGS_ACCOUNT
);
399 init_rwsem(&mapping
->i_mmap_rwsem
);
400 INIT_LIST_HEAD(&mapping
->private_list
);
401 spin_lock_init(&mapping
->private_lock
);
402 mapping
->i_mmap
= RB_ROOT_CACHED
;
405 void address_space_init_once(struct address_space
*mapping
)
407 memset(mapping
, 0, sizeof(*mapping
));
408 __address_space_init_once(mapping
);
410 EXPORT_SYMBOL(address_space_init_once
);
413 * These are initializations that only need to be done
414 * once, because the fields are idempotent across use
415 * of the inode, so let the slab aware of that.
417 void inode_init_once(struct inode
*inode
)
419 memset(inode
, 0, sizeof(*inode
));
420 INIT_HLIST_NODE(&inode
->i_hash
);
421 INIT_LIST_HEAD(&inode
->i_devices
);
422 INIT_LIST_HEAD(&inode
->i_io_list
);
423 INIT_LIST_HEAD(&inode
->i_wb_list
);
424 INIT_LIST_HEAD(&inode
->i_lru
);
425 INIT_LIST_HEAD(&inode
->i_sb_list
);
426 __address_space_init_once(&inode
->i_data
);
427 i_size_ordered_init(inode
);
429 EXPORT_SYMBOL(inode_init_once
);
431 static void init_once(void *foo
)
433 struct inode
*inode
= (struct inode
*) foo
;
435 inode_init_once(inode
);
439 * inode->i_lock must be held
441 void __iget(struct inode
*inode
)
443 atomic_inc(&inode
->i_count
);
447 * get additional reference to inode; caller must already hold one.
449 void ihold(struct inode
*inode
)
451 WARN_ON(atomic_inc_return(&inode
->i_count
) < 2);
453 EXPORT_SYMBOL(ihold
);
455 static void __inode_add_lru(struct inode
*inode
, bool rotate
)
457 if (inode
->i_state
& (I_DIRTY_ALL
| I_SYNC
| I_FREEING
| I_WILL_FREE
))
459 if (atomic_read(&inode
->i_count
))
461 if (!(inode
->i_sb
->s_flags
& SB_ACTIVE
))
463 if (!mapping_shrinkable(&inode
->i_data
))
466 if (list_lru_add(&inode
->i_sb
->s_inode_lru
, &inode
->i_lru
))
467 this_cpu_inc(nr_unused
);
469 inode
->i_state
|= I_REFERENCED
;
473 * Add inode to LRU if needed (inode is unused and clean).
475 * Needs inode->i_lock held.
477 void inode_add_lru(struct inode
*inode
)
479 __inode_add_lru(inode
, false);
482 static void inode_lru_list_del(struct inode
*inode
)
484 if (list_lru_del(&inode
->i_sb
->s_inode_lru
, &inode
->i_lru
))
485 this_cpu_dec(nr_unused
);
489 * inode_sb_list_add - add inode to the superblock list of inodes
490 * @inode: inode to add
492 void inode_sb_list_add(struct inode
*inode
)
494 spin_lock(&inode
->i_sb
->s_inode_list_lock
);
495 list_add(&inode
->i_sb_list
, &inode
->i_sb
->s_inodes
);
496 spin_unlock(&inode
->i_sb
->s_inode_list_lock
);
498 EXPORT_SYMBOL_GPL(inode_sb_list_add
);
500 static inline void inode_sb_list_del(struct inode
*inode
)
502 if (!list_empty(&inode
->i_sb_list
)) {
503 spin_lock(&inode
->i_sb
->s_inode_list_lock
);
504 list_del_init(&inode
->i_sb_list
);
505 spin_unlock(&inode
->i_sb
->s_inode_list_lock
);
509 static unsigned long hash(struct super_block
*sb
, unsigned long hashval
)
513 tmp
= (hashval
* (unsigned long)sb
) ^ (GOLDEN_RATIO_PRIME
+ hashval
) /
515 tmp
= tmp
^ ((tmp
^ GOLDEN_RATIO_PRIME
) >> i_hash_shift
);
516 return tmp
& i_hash_mask
;
520 * __insert_inode_hash - hash an inode
521 * @inode: unhashed inode
522 * @hashval: unsigned long value used to locate this object in the
525 * Add an inode to the inode hash for this superblock.
527 void __insert_inode_hash(struct inode
*inode
, unsigned long hashval
)
529 struct hlist_head
*b
= inode_hashtable
+ hash(inode
->i_sb
, hashval
);
531 spin_lock(&inode_hash_lock
);
532 spin_lock(&inode
->i_lock
);
533 hlist_add_head_rcu(&inode
->i_hash
, b
);
534 spin_unlock(&inode
->i_lock
);
535 spin_unlock(&inode_hash_lock
);
537 EXPORT_SYMBOL(__insert_inode_hash
);
540 * __remove_inode_hash - remove an inode from the hash
541 * @inode: inode to unhash
543 * Remove an inode from the superblock.
545 void __remove_inode_hash(struct inode
*inode
)
547 spin_lock(&inode_hash_lock
);
548 spin_lock(&inode
->i_lock
);
549 hlist_del_init_rcu(&inode
->i_hash
);
550 spin_unlock(&inode
->i_lock
);
551 spin_unlock(&inode_hash_lock
);
553 EXPORT_SYMBOL(__remove_inode_hash
);
555 void dump_mapping(const struct address_space
*mapping
)
558 const struct address_space_operations
*a_ops
;
559 struct hlist_node
*dentry_first
;
560 struct dentry
*dentry_ptr
;
561 struct dentry dentry
;
565 * If mapping is an invalid pointer, we don't want to crash
566 * accessing it, so probe everything depending on it carefully.
568 if (get_kernel_nofault(host
, &mapping
->host
) ||
569 get_kernel_nofault(a_ops
, &mapping
->a_ops
)) {
570 pr_warn("invalid mapping:%px\n", mapping
);
575 pr_warn("aops:%ps\n", a_ops
);
579 if (get_kernel_nofault(dentry_first
, &host
->i_dentry
.first
) ||
580 get_kernel_nofault(ino
, &host
->i_ino
)) {
581 pr_warn("aops:%ps invalid inode:%px\n", a_ops
, host
);
586 pr_warn("aops:%ps ino:%lx\n", a_ops
, ino
);
590 dentry_ptr
= container_of(dentry_first
, struct dentry
, d_u
.d_alias
);
591 if (get_kernel_nofault(dentry
, dentry_ptr
)) {
592 pr_warn("aops:%ps ino:%lx invalid dentry:%px\n",
593 a_ops
, ino
, dentry_ptr
);
598 * if dentry is corrupted, the %pd handler may still crash,
599 * but it's unlikely that we reach here with a corrupt mapping
601 pr_warn("aops:%ps ino:%lx dentry name:\"%pd\"\n", a_ops
, ino
, &dentry
);
604 void clear_inode(struct inode
*inode
)
607 * We have to cycle the i_pages lock here because reclaim can be in the
608 * process of removing the last page (in __filemap_remove_folio())
609 * and we must not free the mapping under it.
611 xa_lock_irq(&inode
->i_data
.i_pages
);
612 BUG_ON(inode
->i_data
.nrpages
);
614 * Almost always, mapping_empty(&inode->i_data) here; but there are
615 * two known and long-standing ways in which nodes may get left behind
616 * (when deep radix-tree node allocation failed partway; or when THP
617 * collapse_file() failed). Until those two known cases are cleaned up,
618 * or a cleanup function is called here, do not BUG_ON(!mapping_empty),
619 * nor even WARN_ON(!mapping_empty).
621 xa_unlock_irq(&inode
->i_data
.i_pages
);
622 BUG_ON(!list_empty(&inode
->i_data
.private_list
));
623 BUG_ON(!(inode
->i_state
& I_FREEING
));
624 BUG_ON(inode
->i_state
& I_CLEAR
);
625 BUG_ON(!list_empty(&inode
->i_wb_list
));
626 /* don't need i_lock here, no concurrent mods to i_state */
627 inode
->i_state
= I_FREEING
| I_CLEAR
;
629 EXPORT_SYMBOL(clear_inode
);
632 * Free the inode passed in, removing it from the lists it is still connected
633 * to. We remove any pages still attached to the inode and wait for any IO that
634 * is still in progress before finally destroying the inode.
636 * An inode must already be marked I_FREEING so that we avoid the inode being
637 * moved back onto lists if we race with other code that manipulates the lists
638 * (e.g. writeback_single_inode). The caller is responsible for setting this.
640 * An inode must already be removed from the LRU list before being evicted from
641 * the cache. This should occur atomically with setting the I_FREEING state
642 * flag, so no inodes here should ever be on the LRU when being evicted.
644 static void evict(struct inode
*inode
)
646 const struct super_operations
*op
= inode
->i_sb
->s_op
;
648 BUG_ON(!(inode
->i_state
& I_FREEING
));
649 BUG_ON(!list_empty(&inode
->i_lru
));
651 if (!list_empty(&inode
->i_io_list
))
652 inode_io_list_del(inode
);
654 inode_sb_list_del(inode
);
657 * Wait for flusher thread to be done with the inode so that filesystem
658 * does not start destroying it while writeback is still running. Since
659 * the inode has I_FREEING set, flusher thread won't start new work on
660 * the inode. We just have to wait for running writeback to finish.
662 inode_wait_for_writeback(inode
);
664 if (op
->evict_inode
) {
665 op
->evict_inode(inode
);
667 truncate_inode_pages_final(&inode
->i_data
);
670 if (S_ISCHR(inode
->i_mode
) && inode
->i_cdev
)
673 remove_inode_hash(inode
);
675 spin_lock(&inode
->i_lock
);
676 wake_up_bit(&inode
->i_state
, __I_NEW
);
677 BUG_ON(inode
->i_state
!= (I_FREEING
| I_CLEAR
));
678 spin_unlock(&inode
->i_lock
);
680 destroy_inode(inode
);
684 * dispose_list - dispose of the contents of a local list
685 * @head: the head of the list to free
687 * Dispose-list gets a local list with local inodes in it, so it doesn't
688 * need to worry about list corruption and SMP locks.
690 static void dispose_list(struct list_head
*head
)
692 while (!list_empty(head
)) {
695 inode
= list_first_entry(head
, struct inode
, i_lru
);
696 list_del_init(&inode
->i_lru
);
704 * evict_inodes - evict all evictable inodes for a superblock
705 * @sb: superblock to operate on
707 * Make sure that no inodes with zero refcount are retained. This is
708 * called by superblock shutdown after having SB_ACTIVE flag removed,
709 * so any inode reaching zero refcount during or after that call will
710 * be immediately evicted.
712 void evict_inodes(struct super_block
*sb
)
714 struct inode
*inode
, *next
;
718 spin_lock(&sb
->s_inode_list_lock
);
719 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
720 if (atomic_read(&inode
->i_count
))
723 spin_lock(&inode
->i_lock
);
724 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
725 spin_unlock(&inode
->i_lock
);
729 inode
->i_state
|= I_FREEING
;
730 inode_lru_list_del(inode
);
731 spin_unlock(&inode
->i_lock
);
732 list_add(&inode
->i_lru
, &dispose
);
735 * We can have a ton of inodes to evict at unmount time given
736 * enough memory, check to see if we need to go to sleep for a
737 * bit so we don't livelock.
739 if (need_resched()) {
740 spin_unlock(&sb
->s_inode_list_lock
);
742 dispose_list(&dispose
);
746 spin_unlock(&sb
->s_inode_list_lock
);
748 dispose_list(&dispose
);
750 EXPORT_SYMBOL_GPL(evict_inodes
);
753 * invalidate_inodes - attempt to free all inodes on a superblock
754 * @sb: superblock to operate on
755 * @kill_dirty: flag to guide handling of dirty inodes
757 * Attempts to free all inodes for a given superblock. If there were any
758 * busy inodes return a non-zero value, else zero.
759 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
762 int invalidate_inodes(struct super_block
*sb
, bool kill_dirty
)
765 struct inode
*inode
, *next
;
769 spin_lock(&sb
->s_inode_list_lock
);
770 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
771 spin_lock(&inode
->i_lock
);
772 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
773 spin_unlock(&inode
->i_lock
);
776 if (inode
->i_state
& I_DIRTY_ALL
&& !kill_dirty
) {
777 spin_unlock(&inode
->i_lock
);
781 if (atomic_read(&inode
->i_count
)) {
782 spin_unlock(&inode
->i_lock
);
787 inode
->i_state
|= I_FREEING
;
788 inode_lru_list_del(inode
);
789 spin_unlock(&inode
->i_lock
);
790 list_add(&inode
->i_lru
, &dispose
);
791 if (need_resched()) {
792 spin_unlock(&sb
->s_inode_list_lock
);
794 dispose_list(&dispose
);
798 spin_unlock(&sb
->s_inode_list_lock
);
800 dispose_list(&dispose
);
806 * Isolate the inode from the LRU in preparation for freeing it.
808 * If the inode has the I_REFERENCED flag set, then it means that it has been
809 * used recently - the flag is set in iput_final(). When we encounter such an
810 * inode, clear the flag and move it to the back of the LRU so it gets another
811 * pass through the LRU before it gets reclaimed. This is necessary because of
812 * the fact we are doing lazy LRU updates to minimise lock contention so the
813 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
814 * with this flag set because they are the inodes that are out of order.
816 static enum lru_status
inode_lru_isolate(struct list_head
*item
,
817 struct list_lru_one
*lru
, spinlock_t
*lru_lock
, void *arg
)
819 struct list_head
*freeable
= arg
;
820 struct inode
*inode
= container_of(item
, struct inode
, i_lru
);
823 * We are inverting the lru lock/inode->i_lock here, so use a
824 * trylock. If we fail to get the lock, just skip it.
826 if (!spin_trylock(&inode
->i_lock
))
830 * Inodes can get referenced, redirtied, or repopulated while
831 * they're already on the LRU, and this can make them
832 * unreclaimable for a while. Remove them lazily here; iput,
833 * sync, or the last page cache deletion will requeue them.
835 if (atomic_read(&inode
->i_count
) ||
836 (inode
->i_state
& ~I_REFERENCED
) ||
837 !mapping_shrinkable(&inode
->i_data
)) {
838 list_lru_isolate(lru
, &inode
->i_lru
);
839 spin_unlock(&inode
->i_lock
);
840 this_cpu_dec(nr_unused
);
844 /* Recently referenced inodes get one more pass */
845 if (inode
->i_state
& I_REFERENCED
) {
846 inode
->i_state
&= ~I_REFERENCED
;
847 spin_unlock(&inode
->i_lock
);
852 * On highmem systems, mapping_shrinkable() permits dropping
853 * page cache in order to free up struct inodes: lowmem might
854 * be under pressure before the cache inside the highmem zone.
856 if (inode_has_buffers(inode
) || !mapping_empty(&inode
->i_data
)) {
858 spin_unlock(&inode
->i_lock
);
859 spin_unlock(lru_lock
);
860 if (remove_inode_buffers(inode
)) {
862 reap
= invalidate_mapping_pages(&inode
->i_data
, 0, -1);
863 if (current_is_kswapd())
864 __count_vm_events(KSWAPD_INODESTEAL
, reap
);
866 __count_vm_events(PGINODESTEAL
, reap
);
867 if (current
->reclaim_state
)
868 current
->reclaim_state
->reclaimed_slab
+= reap
;
875 WARN_ON(inode
->i_state
& I_NEW
);
876 inode
->i_state
|= I_FREEING
;
877 list_lru_isolate_move(lru
, &inode
->i_lru
, freeable
);
878 spin_unlock(&inode
->i_lock
);
880 this_cpu_dec(nr_unused
);
885 * Walk the superblock inode LRU for freeable inodes and attempt to free them.
886 * This is called from the superblock shrinker function with a number of inodes
887 * to trim from the LRU. Inodes to be freed are moved to a temporary list and
888 * then are freed outside inode_lock by dispose_list().
890 long prune_icache_sb(struct super_block
*sb
, struct shrink_control
*sc
)
895 freed
= list_lru_shrink_walk(&sb
->s_inode_lru
, sc
,
896 inode_lru_isolate
, &freeable
);
897 dispose_list(&freeable
);
901 static void __wait_on_freeing_inode(struct inode
*inode
);
903 * Called with the inode lock held.
905 static struct inode
*find_inode(struct super_block
*sb
,
906 struct hlist_head
*head
,
907 int (*test
)(struct inode
*, void *),
910 struct inode
*inode
= NULL
;
913 hlist_for_each_entry(inode
, head
, i_hash
) {
914 if (inode
->i_sb
!= sb
)
916 if (!test(inode
, data
))
918 spin_lock(&inode
->i_lock
);
919 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
920 __wait_on_freeing_inode(inode
);
923 if (unlikely(inode
->i_state
& I_CREATING
)) {
924 spin_unlock(&inode
->i_lock
);
925 return ERR_PTR(-ESTALE
);
928 spin_unlock(&inode
->i_lock
);
935 * find_inode_fast is the fast path version of find_inode, see the comment at
936 * iget_locked for details.
938 static struct inode
*find_inode_fast(struct super_block
*sb
,
939 struct hlist_head
*head
, unsigned long ino
)
941 struct inode
*inode
= NULL
;
944 hlist_for_each_entry(inode
, head
, i_hash
) {
945 if (inode
->i_ino
!= ino
)
947 if (inode
->i_sb
!= sb
)
949 spin_lock(&inode
->i_lock
);
950 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
951 __wait_on_freeing_inode(inode
);
954 if (unlikely(inode
->i_state
& I_CREATING
)) {
955 spin_unlock(&inode
->i_lock
);
956 return ERR_PTR(-ESTALE
);
959 spin_unlock(&inode
->i_lock
);
966 * Each cpu owns a range of LAST_INO_BATCH numbers.
967 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
968 * to renew the exhausted range.
970 * This does not significantly increase overflow rate because every CPU can
971 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
972 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
973 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
974 * overflow rate by 2x, which does not seem too significant.
976 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
977 * error if st_ino won't fit in target struct field. Use 32bit counter
978 * here to attempt to avoid that.
980 #define LAST_INO_BATCH 1024
981 static DEFINE_PER_CPU(unsigned int, last_ino
);
983 unsigned int get_next_ino(void)
985 unsigned int *p
= &get_cpu_var(last_ino
);
986 unsigned int res
= *p
;
989 if (unlikely((res
& (LAST_INO_BATCH
-1)) == 0)) {
990 static atomic_t shared_last_ino
;
991 int next
= atomic_add_return(LAST_INO_BATCH
, &shared_last_ino
);
993 res
= next
- LAST_INO_BATCH
;
998 /* get_next_ino should not provide a 0 inode number */
1002 put_cpu_var(last_ino
);
1005 EXPORT_SYMBOL(get_next_ino
);
1008 * new_inode_pseudo - obtain an inode
1011 * Allocates a new inode for given superblock.
1012 * Inode wont be chained in superblock s_inodes list
1014 * - fs can't be unmount
1015 * - quotas, fsnotify, writeback can't work
1017 struct inode
*new_inode_pseudo(struct super_block
*sb
)
1019 struct inode
*inode
= alloc_inode(sb
);
1022 spin_lock(&inode
->i_lock
);
1024 spin_unlock(&inode
->i_lock
);
1030 * new_inode - obtain an inode
1033 * Allocates a new inode for given superblock. The default gfp_mask
1034 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
1035 * If HIGHMEM pages are unsuitable or it is known that pages allocated
1036 * for the page cache are not reclaimable or migratable,
1037 * mapping_set_gfp_mask() must be called with suitable flags on the
1038 * newly created inode's mapping
1041 struct inode
*new_inode(struct super_block
*sb
)
1043 struct inode
*inode
;
1045 spin_lock_prefetch(&sb
->s_inode_list_lock
);
1047 inode
= new_inode_pseudo(sb
);
1049 inode_sb_list_add(inode
);
1052 EXPORT_SYMBOL(new_inode
);
1054 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1055 void lockdep_annotate_inode_mutex_key(struct inode
*inode
)
1057 if (S_ISDIR(inode
->i_mode
)) {
1058 struct file_system_type
*type
= inode
->i_sb
->s_type
;
1060 /* Set new key only if filesystem hasn't already changed it */
1061 if (lockdep_match_class(&inode
->i_rwsem
, &type
->i_mutex_key
)) {
1063 * ensure nobody is actually holding i_mutex
1065 // mutex_destroy(&inode->i_mutex);
1066 init_rwsem(&inode
->i_rwsem
);
1067 lockdep_set_class(&inode
->i_rwsem
,
1068 &type
->i_mutex_dir_key
);
1072 EXPORT_SYMBOL(lockdep_annotate_inode_mutex_key
);
1076 * unlock_new_inode - clear the I_NEW state and wake up any waiters
1077 * @inode: new inode to unlock
1079 * Called when the inode is fully initialised to clear the new state of the
1080 * inode and wake up anyone waiting for the inode to finish initialisation.
1082 void unlock_new_inode(struct inode
*inode
)
1084 lockdep_annotate_inode_mutex_key(inode
);
1085 spin_lock(&inode
->i_lock
);
1086 WARN_ON(!(inode
->i_state
& I_NEW
));
1087 inode
->i_state
&= ~I_NEW
& ~I_CREATING
;
1089 wake_up_bit(&inode
->i_state
, __I_NEW
);
1090 spin_unlock(&inode
->i_lock
);
1092 EXPORT_SYMBOL(unlock_new_inode
);
1094 void discard_new_inode(struct inode
*inode
)
1096 lockdep_annotate_inode_mutex_key(inode
);
1097 spin_lock(&inode
->i_lock
);
1098 WARN_ON(!(inode
->i_state
& I_NEW
));
1099 inode
->i_state
&= ~I_NEW
;
1101 wake_up_bit(&inode
->i_state
, __I_NEW
);
1102 spin_unlock(&inode
->i_lock
);
1105 EXPORT_SYMBOL(discard_new_inode
);
1108 * lock_two_nondirectories - take two i_mutexes on non-directory objects
1110 * Lock any non-NULL argument that is not a directory.
1111 * Zero, one or two objects may be locked by this function.
1113 * @inode1: first inode to lock
1114 * @inode2: second inode to lock
1116 void lock_two_nondirectories(struct inode
*inode1
, struct inode
*inode2
)
1118 if (inode1
> inode2
)
1119 swap(inode1
, inode2
);
1121 if (inode1
&& !S_ISDIR(inode1
->i_mode
))
1123 if (inode2
&& !S_ISDIR(inode2
->i_mode
) && inode2
!= inode1
)
1124 inode_lock_nested(inode2
, I_MUTEX_NONDIR2
);
1126 EXPORT_SYMBOL(lock_two_nondirectories
);
1129 * unlock_two_nondirectories - release locks from lock_two_nondirectories()
1130 * @inode1: first inode to unlock
1131 * @inode2: second inode to unlock
1133 void unlock_two_nondirectories(struct inode
*inode1
, struct inode
*inode2
)
1135 if (inode1
&& !S_ISDIR(inode1
->i_mode
))
1136 inode_unlock(inode1
);
1137 if (inode2
&& !S_ISDIR(inode2
->i_mode
) && inode2
!= inode1
)
1138 inode_unlock(inode2
);
1140 EXPORT_SYMBOL(unlock_two_nondirectories
);
1143 * inode_insert5 - obtain an inode from a mounted file system
1144 * @inode: pre-allocated inode to use for insert to cache
1145 * @hashval: hash value (usually inode number) to get
1146 * @test: callback used for comparisons between inodes
1147 * @set: callback used to initialize a new struct inode
1148 * @data: opaque data pointer to pass to @test and @set
1150 * Search for the inode specified by @hashval and @data in the inode cache,
1151 * and if present it is return it with an increased reference count. This is
1152 * a variant of iget5_locked() for callers that don't want to fail on memory
1153 * allocation of inode.
1155 * If the inode is not in cache, insert the pre-allocated inode to cache and
1156 * return it locked, hashed, and with the I_NEW flag set. The file system gets
1157 * to fill it in before unlocking it via unlock_new_inode().
1159 * Note both @test and @set are called with the inode_hash_lock held, so can't
1162 struct inode
*inode_insert5(struct inode
*inode
, unsigned long hashval
,
1163 int (*test
)(struct inode
*, void *),
1164 int (*set
)(struct inode
*, void *), void *data
)
1166 struct hlist_head
*head
= inode_hashtable
+ hash(inode
->i_sb
, hashval
);
1170 spin_lock(&inode_hash_lock
);
1171 old
= find_inode(inode
->i_sb
, head
, test
, data
);
1172 if (unlikely(old
)) {
1174 * Uhhuh, somebody else created the same inode under us.
1175 * Use the old inode instead of the preallocated one.
1177 spin_unlock(&inode_hash_lock
);
1181 if (unlikely(inode_unhashed(old
))) {
1188 if (set
&& unlikely(set(inode
, data
))) {
1194 * Return the locked inode with I_NEW set, the
1195 * caller is responsible for filling in the contents
1197 spin_lock(&inode
->i_lock
);
1198 inode
->i_state
|= I_NEW
;
1199 hlist_add_head_rcu(&inode
->i_hash
, head
);
1200 spin_unlock(&inode
->i_lock
);
1203 * Add inode to the sb list if it's not already. It has I_NEW at this
1204 * point, so it should be safe to test i_sb_list locklessly.
1206 if (list_empty(&inode
->i_sb_list
))
1207 inode_sb_list_add(inode
);
1209 spin_unlock(&inode_hash_lock
);
1213 EXPORT_SYMBOL(inode_insert5
);
1216 * iget5_locked - obtain an inode from a mounted file system
1217 * @sb: super block of file system
1218 * @hashval: hash value (usually inode number) to get
1219 * @test: callback used for comparisons between inodes
1220 * @set: callback used to initialize a new struct inode
1221 * @data: opaque data pointer to pass to @test and @set
1223 * Search for the inode specified by @hashval and @data in the inode cache,
1224 * and if present it is return it with an increased reference count. This is
1225 * a generalized version of iget_locked() for file systems where the inode
1226 * number is not sufficient for unique identification of an inode.
1228 * If the inode is not in cache, allocate a new inode and return it locked,
1229 * hashed, and with the I_NEW flag set. The file system gets to fill it in
1230 * before unlocking it via unlock_new_inode().
1232 * Note both @test and @set are called with the inode_hash_lock held, so can't
1235 struct inode
*iget5_locked(struct super_block
*sb
, unsigned long hashval
,
1236 int (*test
)(struct inode
*, void *),
1237 int (*set
)(struct inode
*, void *), void *data
)
1239 struct inode
*inode
= ilookup5(sb
, hashval
, test
, data
);
1242 struct inode
*new = alloc_inode(sb
);
1246 inode
= inode_insert5(new, hashval
, test
, set
, data
);
1247 if (unlikely(inode
!= new))
1253 EXPORT_SYMBOL(iget5_locked
);
1256 * iget_locked - obtain an inode from a mounted file system
1257 * @sb: super block of file system
1258 * @ino: inode number to get
1260 * Search for the inode specified by @ino in the inode cache and if present
1261 * return it with an increased reference count. This is for file systems
1262 * where the inode number is sufficient for unique identification of an inode.
1264 * If the inode is not in cache, allocate a new inode and return it locked,
1265 * hashed, and with the I_NEW flag set. The file system gets to fill it in
1266 * before unlocking it via unlock_new_inode().
1268 struct inode
*iget_locked(struct super_block
*sb
, unsigned long ino
)
1270 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1271 struct inode
*inode
;
1273 spin_lock(&inode_hash_lock
);
1274 inode
= find_inode_fast(sb
, head
, ino
);
1275 spin_unlock(&inode_hash_lock
);
1279 wait_on_inode(inode
);
1280 if (unlikely(inode_unhashed(inode
))) {
1287 inode
= alloc_inode(sb
);
1291 spin_lock(&inode_hash_lock
);
1292 /* We released the lock, so.. */
1293 old
= find_inode_fast(sb
, head
, ino
);
1296 spin_lock(&inode
->i_lock
);
1297 inode
->i_state
= I_NEW
;
1298 hlist_add_head_rcu(&inode
->i_hash
, head
);
1299 spin_unlock(&inode
->i_lock
);
1300 inode_sb_list_add(inode
);
1301 spin_unlock(&inode_hash_lock
);
1303 /* Return the locked inode with I_NEW set, the
1304 * caller is responsible for filling in the contents
1310 * Uhhuh, somebody else created the same inode under
1311 * us. Use the old inode instead of the one we just
1314 spin_unlock(&inode_hash_lock
);
1315 destroy_inode(inode
);
1319 wait_on_inode(inode
);
1320 if (unlikely(inode_unhashed(inode
))) {
1327 EXPORT_SYMBOL(iget_locked
);
1330 * search the inode cache for a matching inode number.
1331 * If we find one, then the inode number we are trying to
1332 * allocate is not unique and so we should not use it.
1334 * Returns 1 if the inode number is unique, 0 if it is not.
1336 static int test_inode_iunique(struct super_block
*sb
, unsigned long ino
)
1338 struct hlist_head
*b
= inode_hashtable
+ hash(sb
, ino
);
1339 struct inode
*inode
;
1341 hlist_for_each_entry_rcu(inode
, b
, i_hash
) {
1342 if (inode
->i_ino
== ino
&& inode
->i_sb
== sb
)
1349 * iunique - get a unique inode number
1351 * @max_reserved: highest reserved inode number
1353 * Obtain an inode number that is unique on the system for a given
1354 * superblock. This is used by file systems that have no natural
1355 * permanent inode numbering system. An inode number is returned that
1356 * is higher than the reserved limit but unique.
1359 * With a large number of inodes live on the file system this function
1360 * currently becomes quite slow.
1362 ino_t
iunique(struct super_block
*sb
, ino_t max_reserved
)
1365 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1366 * error if st_ino won't fit in target struct field. Use 32bit counter
1367 * here to attempt to avoid that.
1369 static DEFINE_SPINLOCK(iunique_lock
);
1370 static unsigned int counter
;
1374 spin_lock(&iunique_lock
);
1376 if (counter
<= max_reserved
)
1377 counter
= max_reserved
+ 1;
1379 } while (!test_inode_iunique(sb
, res
));
1380 spin_unlock(&iunique_lock
);
1385 EXPORT_SYMBOL(iunique
);
1387 struct inode
*igrab(struct inode
*inode
)
1389 spin_lock(&inode
->i_lock
);
1390 if (!(inode
->i_state
& (I_FREEING
|I_WILL_FREE
))) {
1392 spin_unlock(&inode
->i_lock
);
1394 spin_unlock(&inode
->i_lock
);
1396 * Handle the case where s_op->clear_inode is not been
1397 * called yet, and somebody is calling igrab
1398 * while the inode is getting freed.
1404 EXPORT_SYMBOL(igrab
);
1407 * ilookup5_nowait - search for an inode in the inode cache
1408 * @sb: super block of file system to search
1409 * @hashval: hash value (usually inode number) to search for
1410 * @test: callback used for comparisons between inodes
1411 * @data: opaque data pointer to pass to @test
1413 * Search for the inode specified by @hashval and @data in the inode cache.
1414 * If the inode is in the cache, the inode is returned with an incremented
1417 * Note: I_NEW is not waited upon so you have to be very careful what you do
1418 * with the returned inode. You probably should be using ilookup5() instead.
1420 * Note2: @test is called with the inode_hash_lock held, so can't sleep.
1422 struct inode
*ilookup5_nowait(struct super_block
*sb
, unsigned long hashval
,
1423 int (*test
)(struct inode
*, void *), void *data
)
1425 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1426 struct inode
*inode
;
1428 spin_lock(&inode_hash_lock
);
1429 inode
= find_inode(sb
, head
, test
, data
);
1430 spin_unlock(&inode_hash_lock
);
1432 return IS_ERR(inode
) ? NULL
: inode
;
1434 EXPORT_SYMBOL(ilookup5_nowait
);
1437 * ilookup5 - search for an inode in the inode cache
1438 * @sb: super block of file system to search
1439 * @hashval: hash value (usually inode number) to search for
1440 * @test: callback used for comparisons between inodes
1441 * @data: opaque data pointer to pass to @test
1443 * Search for the inode specified by @hashval and @data in the inode cache,
1444 * and if the inode is in the cache, return the inode with an incremented
1445 * reference count. Waits on I_NEW before returning the inode.
1446 * returned with an incremented reference count.
1448 * This is a generalized version of ilookup() for file systems where the
1449 * inode number is not sufficient for unique identification of an inode.
1451 * Note: @test is called with the inode_hash_lock held, so can't sleep.
1453 struct inode
*ilookup5(struct super_block
*sb
, unsigned long hashval
,
1454 int (*test
)(struct inode
*, void *), void *data
)
1456 struct inode
*inode
;
1458 inode
= ilookup5_nowait(sb
, hashval
, test
, data
);
1460 wait_on_inode(inode
);
1461 if (unlikely(inode_unhashed(inode
))) {
1468 EXPORT_SYMBOL(ilookup5
);
1471 * ilookup - search for an inode in the inode cache
1472 * @sb: super block of file system to search
1473 * @ino: inode number to search for
1475 * Search for the inode @ino in the inode cache, and if the inode is in the
1476 * cache, the inode is returned with an incremented reference count.
1478 struct inode
*ilookup(struct super_block
*sb
, unsigned long ino
)
1480 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1481 struct inode
*inode
;
1483 spin_lock(&inode_hash_lock
);
1484 inode
= find_inode_fast(sb
, head
, ino
);
1485 spin_unlock(&inode_hash_lock
);
1490 wait_on_inode(inode
);
1491 if (unlikely(inode_unhashed(inode
))) {
1498 EXPORT_SYMBOL(ilookup
);
1501 * find_inode_nowait - find an inode in the inode cache
1502 * @sb: super block of file system to search
1503 * @hashval: hash value (usually inode number) to search for
1504 * @match: callback used for comparisons between inodes
1505 * @data: opaque data pointer to pass to @match
1507 * Search for the inode specified by @hashval and @data in the inode
1508 * cache, where the helper function @match will return 0 if the inode
1509 * does not match, 1 if the inode does match, and -1 if the search
1510 * should be stopped. The @match function must be responsible for
1511 * taking the i_lock spin_lock and checking i_state for an inode being
1512 * freed or being initialized, and incrementing the reference count
1513 * before returning 1. It also must not sleep, since it is called with
1514 * the inode_hash_lock spinlock held.
1516 * This is a even more generalized version of ilookup5() when the
1517 * function must never block --- find_inode() can block in
1518 * __wait_on_freeing_inode() --- or when the caller can not increment
1519 * the reference count because the resulting iput() might cause an
1520 * inode eviction. The tradeoff is that the @match funtion must be
1521 * very carefully implemented.
1523 struct inode
*find_inode_nowait(struct super_block
*sb
,
1524 unsigned long hashval
,
1525 int (*match
)(struct inode
*, unsigned long,
1529 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1530 struct inode
*inode
, *ret_inode
= NULL
;
1533 spin_lock(&inode_hash_lock
);
1534 hlist_for_each_entry(inode
, head
, i_hash
) {
1535 if (inode
->i_sb
!= sb
)
1537 mval
= match(inode
, hashval
, data
);
1545 spin_unlock(&inode_hash_lock
);
1548 EXPORT_SYMBOL(find_inode_nowait
);
1551 * find_inode_rcu - find an inode in the inode cache
1552 * @sb: Super block of file system to search
1553 * @hashval: Key to hash
1554 * @test: Function to test match on an inode
1555 * @data: Data for test function
1557 * Search for the inode specified by @hashval and @data in the inode cache,
1558 * where the helper function @test will return 0 if the inode does not match
1559 * and 1 if it does. The @test function must be responsible for taking the
1560 * i_lock spin_lock and checking i_state for an inode being freed or being
1563 * If successful, this will return the inode for which the @test function
1564 * returned 1 and NULL otherwise.
1566 * The @test function is not permitted to take a ref on any inode presented.
1567 * It is also not permitted to sleep.
1569 * The caller must hold the RCU read lock.
1571 struct inode
*find_inode_rcu(struct super_block
*sb
, unsigned long hashval
,
1572 int (*test
)(struct inode
*, void *), void *data
)
1574 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1575 struct inode
*inode
;
1577 RCU_LOCKDEP_WARN(!rcu_read_lock_held(),
1578 "suspicious find_inode_rcu() usage");
1580 hlist_for_each_entry_rcu(inode
, head
, i_hash
) {
1581 if (inode
->i_sb
== sb
&&
1582 !(READ_ONCE(inode
->i_state
) & (I_FREEING
| I_WILL_FREE
)) &&
1588 EXPORT_SYMBOL(find_inode_rcu
);
1591 * find_inode_by_ino_rcu - Find an inode in the inode cache
1592 * @sb: Super block of file system to search
1593 * @ino: The inode number to match
1595 * Search for the inode specified by @hashval and @data in the inode cache,
1596 * where the helper function @test will return 0 if the inode does not match
1597 * and 1 if it does. The @test function must be responsible for taking the
1598 * i_lock spin_lock and checking i_state for an inode being freed or being
1601 * If successful, this will return the inode for which the @test function
1602 * returned 1 and NULL otherwise.
1604 * The @test function is not permitted to take a ref on any inode presented.
1605 * It is also not permitted to sleep.
1607 * The caller must hold the RCU read lock.
1609 struct inode
*find_inode_by_ino_rcu(struct super_block
*sb
,
1612 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1613 struct inode
*inode
;
1615 RCU_LOCKDEP_WARN(!rcu_read_lock_held(),
1616 "suspicious find_inode_by_ino_rcu() usage");
1618 hlist_for_each_entry_rcu(inode
, head
, i_hash
) {
1619 if (inode
->i_ino
== ino
&&
1620 inode
->i_sb
== sb
&&
1621 !(READ_ONCE(inode
->i_state
) & (I_FREEING
| I_WILL_FREE
)))
1626 EXPORT_SYMBOL(find_inode_by_ino_rcu
);
1628 int insert_inode_locked(struct inode
*inode
)
1630 struct super_block
*sb
= inode
->i_sb
;
1631 ino_t ino
= inode
->i_ino
;
1632 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1635 struct inode
*old
= NULL
;
1636 spin_lock(&inode_hash_lock
);
1637 hlist_for_each_entry(old
, head
, i_hash
) {
1638 if (old
->i_ino
!= ino
)
1640 if (old
->i_sb
!= sb
)
1642 spin_lock(&old
->i_lock
);
1643 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
1644 spin_unlock(&old
->i_lock
);
1650 spin_lock(&inode
->i_lock
);
1651 inode
->i_state
|= I_NEW
| I_CREATING
;
1652 hlist_add_head_rcu(&inode
->i_hash
, head
);
1653 spin_unlock(&inode
->i_lock
);
1654 spin_unlock(&inode_hash_lock
);
1657 if (unlikely(old
->i_state
& I_CREATING
)) {
1658 spin_unlock(&old
->i_lock
);
1659 spin_unlock(&inode_hash_lock
);
1663 spin_unlock(&old
->i_lock
);
1664 spin_unlock(&inode_hash_lock
);
1666 if (unlikely(!inode_unhashed(old
))) {
1673 EXPORT_SYMBOL(insert_inode_locked
);
1675 int insert_inode_locked4(struct inode
*inode
, unsigned long hashval
,
1676 int (*test
)(struct inode
*, void *), void *data
)
1680 inode
->i_state
|= I_CREATING
;
1681 old
= inode_insert5(inode
, hashval
, test
, NULL
, data
);
1689 EXPORT_SYMBOL(insert_inode_locked4
);
1692 int generic_delete_inode(struct inode
*inode
)
1696 EXPORT_SYMBOL(generic_delete_inode
);
1699 * Called when we're dropping the last reference
1702 * Call the FS "drop_inode()" function, defaulting to
1703 * the legacy UNIX filesystem behaviour. If it tells
1704 * us to evict inode, do so. Otherwise, retain inode
1705 * in cache if fs is alive, sync and evict if fs is
1708 static void iput_final(struct inode
*inode
)
1710 struct super_block
*sb
= inode
->i_sb
;
1711 const struct super_operations
*op
= inode
->i_sb
->s_op
;
1712 unsigned long state
;
1715 WARN_ON(inode
->i_state
& I_NEW
);
1718 drop
= op
->drop_inode(inode
);
1720 drop
= generic_drop_inode(inode
);
1723 !(inode
->i_state
& I_DONTCACHE
) &&
1724 (sb
->s_flags
& SB_ACTIVE
)) {
1725 __inode_add_lru(inode
, true);
1726 spin_unlock(&inode
->i_lock
);
1730 state
= inode
->i_state
;
1732 WRITE_ONCE(inode
->i_state
, state
| I_WILL_FREE
);
1733 spin_unlock(&inode
->i_lock
);
1735 write_inode_now(inode
, 1);
1737 spin_lock(&inode
->i_lock
);
1738 state
= inode
->i_state
;
1739 WARN_ON(state
& I_NEW
);
1740 state
&= ~I_WILL_FREE
;
1743 WRITE_ONCE(inode
->i_state
, state
| I_FREEING
);
1744 if (!list_empty(&inode
->i_lru
))
1745 inode_lru_list_del(inode
);
1746 spin_unlock(&inode
->i_lock
);
1752 * iput - put an inode
1753 * @inode: inode to put
1755 * Puts an inode, dropping its usage count. If the inode use count hits
1756 * zero, the inode is then freed and may also be destroyed.
1758 * Consequently, iput() can sleep.
1760 void iput(struct inode
*inode
)
1764 BUG_ON(inode
->i_state
& I_CLEAR
);
1766 if (atomic_dec_and_lock(&inode
->i_count
, &inode
->i_lock
)) {
1767 if (inode
->i_nlink
&& (inode
->i_state
& I_DIRTY_TIME
)) {
1768 atomic_inc(&inode
->i_count
);
1769 spin_unlock(&inode
->i_lock
);
1770 trace_writeback_lazytime_iput(inode
);
1771 mark_inode_dirty_sync(inode
);
1777 EXPORT_SYMBOL(iput
);
1781 * bmap - find a block number in a file
1782 * @inode: inode owning the block number being requested
1783 * @block: pointer containing the block to find
1785 * Replaces the value in ``*block`` with the block number on the device holding
1786 * corresponding to the requested block number in the file.
1787 * That is, asked for block 4 of inode 1 the function will replace the
1788 * 4 in ``*block``, with disk block relative to the disk start that holds that
1789 * block of the file.
1791 * Returns -EINVAL in case of error, 0 otherwise. If mapping falls into a
1792 * hole, returns 0 and ``*block`` is also set to 0.
1794 int bmap(struct inode
*inode
, sector_t
*block
)
1796 if (!inode
->i_mapping
->a_ops
->bmap
)
1799 *block
= inode
->i_mapping
->a_ops
->bmap(inode
->i_mapping
, *block
);
1802 EXPORT_SYMBOL(bmap
);
1806 * With relative atime, only update atime if the previous atime is
1807 * earlier than either the ctime or mtime or if at least a day has
1808 * passed since the last atime update.
1810 static int relatime_need_update(struct vfsmount
*mnt
, struct inode
*inode
,
1811 struct timespec64 now
)
1814 if (!(mnt
->mnt_flags
& MNT_RELATIME
))
1817 * Is mtime younger than atime? If yes, update atime:
1819 if (timespec64_compare(&inode
->i_mtime
, &inode
->i_atime
) >= 0)
1822 * Is ctime younger than atime? If yes, update atime:
1824 if (timespec64_compare(&inode
->i_ctime
, &inode
->i_atime
) >= 0)
1828 * Is the previous atime value older than a day? If yes,
1831 if ((long)(now
.tv_sec
- inode
->i_atime
.tv_sec
) >= 24*60*60)
1834 * Good, we can skip the atime update:
1839 int generic_update_time(struct inode
*inode
, struct timespec64
*time
, int flags
)
1841 int dirty_flags
= 0;
1843 if (flags
& (S_ATIME
| S_CTIME
| S_MTIME
)) {
1844 if (flags
& S_ATIME
)
1845 inode
->i_atime
= *time
;
1846 if (flags
& S_CTIME
)
1847 inode
->i_ctime
= *time
;
1848 if (flags
& S_MTIME
)
1849 inode
->i_mtime
= *time
;
1851 if (inode
->i_sb
->s_flags
& SB_LAZYTIME
)
1852 dirty_flags
|= I_DIRTY_TIME
;
1854 dirty_flags
|= I_DIRTY_SYNC
;
1857 if ((flags
& S_VERSION
) && inode_maybe_inc_iversion(inode
, false))
1858 dirty_flags
|= I_DIRTY_SYNC
;
1860 __mark_inode_dirty(inode
, dirty_flags
);
1863 EXPORT_SYMBOL(generic_update_time
);
1866 * This does the actual work of updating an inodes time or version. Must have
1867 * had called mnt_want_write() before calling this.
1869 int inode_update_time(struct inode
*inode
, struct timespec64
*time
, int flags
)
1871 if (inode
->i_op
->update_time
)
1872 return inode
->i_op
->update_time(inode
, time
, flags
);
1873 return generic_update_time(inode
, time
, flags
);
1875 EXPORT_SYMBOL(inode_update_time
);
1878 * atime_needs_update - update the access time
1879 * @path: the &struct path to update
1880 * @inode: inode to update
1882 * Update the accessed time on an inode and mark it for writeback.
1883 * This function automatically handles read only file systems and media,
1884 * as well as the "noatime" flag and inode specific "noatime" markers.
1886 bool atime_needs_update(const struct path
*path
, struct inode
*inode
)
1888 struct vfsmount
*mnt
= path
->mnt
;
1889 struct timespec64 now
;
1891 if (inode
->i_flags
& S_NOATIME
)
1894 /* Atime updates will likely cause i_uid and i_gid to be written
1895 * back improprely if their true value is unknown to the vfs.
1897 if (HAS_UNMAPPED_ID(mnt_idmap(mnt
), inode
))
1900 if (IS_NOATIME(inode
))
1902 if ((inode
->i_sb
->s_flags
& SB_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1905 if (mnt
->mnt_flags
& MNT_NOATIME
)
1907 if ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1910 now
= current_time(inode
);
1912 if (!relatime_need_update(mnt
, inode
, now
))
1915 if (timespec64_equal(&inode
->i_atime
, &now
))
1921 void touch_atime(const struct path
*path
)
1923 struct vfsmount
*mnt
= path
->mnt
;
1924 struct inode
*inode
= d_inode(path
->dentry
);
1925 struct timespec64 now
;
1927 if (!atime_needs_update(path
, inode
))
1930 if (!sb_start_write_trylock(inode
->i_sb
))
1933 if (__mnt_want_write(mnt
) != 0)
1936 * File systems can error out when updating inodes if they need to
1937 * allocate new space to modify an inode (such is the case for
1938 * Btrfs), but since we touch atime while walking down the path we
1939 * really don't care if we failed to update the atime of the file,
1940 * so just ignore the return value.
1941 * We may also fail on filesystems that have the ability to make parts
1942 * of the fs read only, e.g. subvolumes in Btrfs.
1944 now
= current_time(inode
);
1945 inode_update_time(inode
, &now
, S_ATIME
);
1946 __mnt_drop_write(mnt
);
1948 sb_end_write(inode
->i_sb
);
1950 EXPORT_SYMBOL(touch_atime
);
1953 * Return mask of changes for notify_change() that need to be done as a
1954 * response to write or truncate. Return 0 if nothing has to be changed.
1955 * Negative value on error (change should be denied).
1957 int dentry_needs_remove_privs(struct mnt_idmap
*idmap
,
1958 struct dentry
*dentry
)
1960 struct inode
*inode
= d_inode(dentry
);
1964 if (IS_NOSEC(inode
))
1967 mask
= setattr_should_drop_suidgid(idmap
, inode
);
1968 ret
= security_inode_need_killpriv(dentry
);
1972 mask
|= ATTR_KILL_PRIV
;
1976 static int __remove_privs(struct mnt_idmap
*idmap
,
1977 struct dentry
*dentry
, int kill
)
1979 struct iattr newattrs
;
1981 newattrs
.ia_valid
= ATTR_FORCE
| kill
;
1983 * Note we call this on write, so notify_change will not
1984 * encounter any conflicting delegations:
1986 return notify_change(idmap
, dentry
, &newattrs
, NULL
);
1989 static int __file_remove_privs(struct file
*file
, unsigned int flags
)
1991 struct dentry
*dentry
= file_dentry(file
);
1992 struct inode
*inode
= file_inode(file
);
1996 if (IS_NOSEC(inode
) || !S_ISREG(inode
->i_mode
))
1999 kill
= dentry_needs_remove_privs(file_mnt_idmap(file
), dentry
);
2004 if (flags
& IOCB_NOWAIT
)
2007 error
= __remove_privs(file_mnt_idmap(file
), dentry
, kill
);
2011 inode_has_no_xattr(inode
);
2016 * file_remove_privs - remove special file privileges (suid, capabilities)
2017 * @file: file to remove privileges from
2019 * When file is modified by a write or truncation ensure that special
2020 * file privileges are removed.
2022 * Return: 0 on success, negative errno on failure.
2024 int file_remove_privs(struct file
*file
)
2026 return __file_remove_privs(file
, 0);
2028 EXPORT_SYMBOL(file_remove_privs
);
2030 static int inode_needs_update_time(struct inode
*inode
, struct timespec64
*now
)
2034 /* First try to exhaust all avenues to not sync */
2035 if (IS_NOCMTIME(inode
))
2038 if (!timespec64_equal(&inode
->i_mtime
, now
))
2041 if (!timespec64_equal(&inode
->i_ctime
, now
))
2044 if (IS_I_VERSION(inode
) && inode_iversion_need_inc(inode
))
2045 sync_it
|= S_VERSION
;
2050 static int __file_update_time(struct file
*file
, struct timespec64
*now
,
2054 struct inode
*inode
= file_inode(file
);
2056 /* try to update time settings */
2057 if (!__mnt_want_write_file(file
)) {
2058 ret
= inode_update_time(inode
, now
, sync_mode
);
2059 __mnt_drop_write_file(file
);
2066 * file_update_time - update mtime and ctime time
2067 * @file: file accessed
2069 * Update the mtime and ctime members of an inode and mark the inode for
2070 * writeback. Note that this function is meant exclusively for usage in
2071 * the file write path of filesystems, and filesystems may choose to
2072 * explicitly ignore updates via this function with the _NOCMTIME inode
2073 * flag, e.g. for network filesystem where these imestamps are handled
2074 * by the server. This can return an error for file systems who need to
2075 * allocate space in order to update an inode.
2077 * Return: 0 on success, negative errno on failure.
2079 int file_update_time(struct file
*file
)
2082 struct inode
*inode
= file_inode(file
);
2083 struct timespec64 now
= current_time(inode
);
2085 ret
= inode_needs_update_time(inode
, &now
);
2089 return __file_update_time(file
, &now
, ret
);
2091 EXPORT_SYMBOL(file_update_time
);
2094 * file_modified_flags - handle mandated vfs changes when modifying a file
2095 * @file: file that was modified
2096 * @flags: kiocb flags
2098 * When file has been modified ensure that special
2099 * file privileges are removed and time settings are updated.
2101 * If IOCB_NOWAIT is set, special file privileges will not be removed and
2102 * time settings will not be updated. It will return -EAGAIN.
2104 * Context: Caller must hold the file's inode lock.
2106 * Return: 0 on success, negative errno on failure.
2108 static int file_modified_flags(struct file
*file
, int flags
)
2111 struct inode
*inode
= file_inode(file
);
2112 struct timespec64 now
= current_time(inode
);
2115 * Clear the security bits if the process is not being run by root.
2116 * This keeps people from modifying setuid and setgid binaries.
2118 ret
= __file_remove_privs(file
, flags
);
2122 if (unlikely(file
->f_mode
& FMODE_NOCMTIME
))
2125 ret
= inode_needs_update_time(inode
, &now
);
2128 if (flags
& IOCB_NOWAIT
)
2131 return __file_update_time(file
, &now
, ret
);
2135 * file_modified - handle mandated vfs changes when modifying a file
2136 * @file: file that was modified
2138 * When file has been modified ensure that special
2139 * file privileges are removed and time settings are updated.
2141 * Context: Caller must hold the file's inode lock.
2143 * Return: 0 on success, negative errno on failure.
2145 int file_modified(struct file
*file
)
2147 return file_modified_flags(file
, 0);
2149 EXPORT_SYMBOL(file_modified
);
2152 * kiocb_modified - handle mandated vfs changes when modifying a file
2153 * @iocb: iocb that was modified
2155 * When file has been modified ensure that special
2156 * file privileges are removed and time settings are updated.
2158 * Context: Caller must hold the file's inode lock.
2160 * Return: 0 on success, negative errno on failure.
2162 int kiocb_modified(struct kiocb
*iocb
)
2164 return file_modified_flags(iocb
->ki_filp
, iocb
->ki_flags
);
2166 EXPORT_SYMBOL_GPL(kiocb_modified
);
2168 int inode_needs_sync(struct inode
*inode
)
2172 if (S_ISDIR(inode
->i_mode
) && IS_DIRSYNC(inode
))
2176 EXPORT_SYMBOL(inode_needs_sync
);
2179 * If we try to find an inode in the inode hash while it is being
2180 * deleted, we have to wait until the filesystem completes its
2181 * deletion before reporting that it isn't found. This function waits
2182 * until the deletion _might_ have completed. Callers are responsible
2183 * to recheck inode state.
2185 * It doesn't matter if I_NEW is not set initially, a call to
2186 * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
2189 static void __wait_on_freeing_inode(struct inode
*inode
)
2191 wait_queue_head_t
*wq
;
2192 DEFINE_WAIT_BIT(wait
, &inode
->i_state
, __I_NEW
);
2193 wq
= bit_waitqueue(&inode
->i_state
, __I_NEW
);
2194 prepare_to_wait(wq
, &wait
.wq_entry
, TASK_UNINTERRUPTIBLE
);
2195 spin_unlock(&inode
->i_lock
);
2196 spin_unlock(&inode_hash_lock
);
2198 finish_wait(wq
, &wait
.wq_entry
);
2199 spin_lock(&inode_hash_lock
);
2202 static __initdata
unsigned long ihash_entries
;
2203 static int __init
set_ihash_entries(char *str
)
2207 ihash_entries
= simple_strtoul(str
, &str
, 0);
2210 __setup("ihash_entries=", set_ihash_entries
);
2213 * Initialize the waitqueues and inode hash table.
2215 void __init
inode_init_early(void)
2217 /* If hashes are distributed across NUMA nodes, defer
2218 * hash allocation until vmalloc space is available.
2224 alloc_large_system_hash("Inode-cache",
2225 sizeof(struct hlist_head
),
2228 HASH_EARLY
| HASH_ZERO
,
2235 void __init
inode_init(void)
2237 /* inode slab cache */
2238 inode_cachep
= kmem_cache_create("inode_cache",
2239 sizeof(struct inode
),
2241 (SLAB_RECLAIM_ACCOUNT
|SLAB_PANIC
|
2242 SLAB_MEM_SPREAD
|SLAB_ACCOUNT
),
2245 /* Hash may have been set up in inode_init_early */
2250 alloc_large_system_hash("Inode-cache",
2251 sizeof(struct hlist_head
),
2261 void init_special_inode(struct inode
*inode
, umode_t mode
, dev_t rdev
)
2263 inode
->i_mode
= mode
;
2264 if (S_ISCHR(mode
)) {
2265 inode
->i_fop
= &def_chr_fops
;
2266 inode
->i_rdev
= rdev
;
2267 } else if (S_ISBLK(mode
)) {
2268 inode
->i_fop
= &def_blk_fops
;
2269 inode
->i_rdev
= rdev
;
2270 } else if (S_ISFIFO(mode
))
2271 inode
->i_fop
= &pipefifo_fops
;
2272 else if (S_ISSOCK(mode
))
2273 ; /* leave it no_open_fops */
2275 printk(KERN_DEBUG
"init_special_inode: bogus i_mode (%o) for"
2276 " inode %s:%lu\n", mode
, inode
->i_sb
->s_id
,
2279 EXPORT_SYMBOL(init_special_inode
);
2282 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
2283 * @idmap: idmap of the mount the inode was created from
2285 * @dir: Directory inode
2286 * @mode: mode of the new inode
2288 * If the inode has been created through an idmapped mount the idmap of
2289 * the vfsmount must be passed through @idmap. This function will then take
2290 * care to map the inode according to @idmap before checking permissions
2291 * and initializing i_uid and i_gid. On non-idmapped mounts or if permission
2292 * checking is to be performed on the raw inode simply pass @nop_mnt_idmap.
2294 void inode_init_owner(struct mnt_idmap
*idmap
, struct inode
*inode
,
2295 const struct inode
*dir
, umode_t mode
)
2297 inode_fsuid_set(inode
, idmap
);
2298 if (dir
&& dir
->i_mode
& S_ISGID
) {
2299 inode
->i_gid
= dir
->i_gid
;
2301 /* Directories are special, and always inherit S_ISGID */
2305 inode_fsgid_set(inode
, idmap
);
2306 inode
->i_mode
= mode
;
2308 EXPORT_SYMBOL(inode_init_owner
);
2311 * inode_owner_or_capable - check current task permissions to inode
2312 * @idmap: idmap of the mount the inode was found from
2313 * @inode: inode being checked
2315 * Return true if current either has CAP_FOWNER in a namespace with the
2316 * inode owner uid mapped, or owns the file.
2318 * If the inode has been found through an idmapped mount the idmap of
2319 * the vfsmount must be passed through @idmap. This function will then take
2320 * care to map the inode according to @idmap before checking permissions.
2321 * On non-idmapped mounts or if permission checking is to be performed on the
2322 * raw inode simply passs @nop_mnt_idmap.
2324 bool inode_owner_or_capable(struct mnt_idmap
*idmap
,
2325 const struct inode
*inode
)
2328 struct user_namespace
*ns
;
2330 vfsuid
= i_uid_into_vfsuid(idmap
, inode
);
2331 if (vfsuid_eq_kuid(vfsuid
, current_fsuid()))
2334 ns
= current_user_ns();
2335 if (vfsuid_has_mapping(ns
, vfsuid
) && ns_capable(ns
, CAP_FOWNER
))
2339 EXPORT_SYMBOL(inode_owner_or_capable
);
2342 * Direct i/o helper functions
2344 static void __inode_dio_wait(struct inode
*inode
)
2346 wait_queue_head_t
*wq
= bit_waitqueue(&inode
->i_state
, __I_DIO_WAKEUP
);
2347 DEFINE_WAIT_BIT(q
, &inode
->i_state
, __I_DIO_WAKEUP
);
2350 prepare_to_wait(wq
, &q
.wq_entry
, TASK_UNINTERRUPTIBLE
);
2351 if (atomic_read(&inode
->i_dio_count
))
2353 } while (atomic_read(&inode
->i_dio_count
));
2354 finish_wait(wq
, &q
.wq_entry
);
2358 * inode_dio_wait - wait for outstanding DIO requests to finish
2359 * @inode: inode to wait for
2361 * Waits for all pending direct I/O requests to finish so that we can
2362 * proceed with a truncate or equivalent operation.
2364 * Must be called under a lock that serializes taking new references
2365 * to i_dio_count, usually by inode->i_mutex.
2367 void inode_dio_wait(struct inode
*inode
)
2369 if (atomic_read(&inode
->i_dio_count
))
2370 __inode_dio_wait(inode
);
2372 EXPORT_SYMBOL(inode_dio_wait
);
2375 * inode_set_flags - atomically set some inode flags
2377 * Note: the caller should be holding i_mutex, or else be sure that
2378 * they have exclusive access to the inode structure (i.e., while the
2379 * inode is being instantiated). The reason for the cmpxchg() loop
2380 * --- which wouldn't be necessary if all code paths which modify
2381 * i_flags actually followed this rule, is that there is at least one
2382 * code path which doesn't today so we use cmpxchg() out of an abundance
2385 * In the long run, i_mutex is overkill, and we should probably look
2386 * at using the i_lock spinlock to protect i_flags, and then make sure
2387 * it is so documented in include/linux/fs.h and that all code follows
2388 * the locking convention!!
2390 void inode_set_flags(struct inode
*inode
, unsigned int flags
,
2393 WARN_ON_ONCE(flags
& ~mask
);
2394 set_mask_bits(&inode
->i_flags
, mask
, flags
);
2396 EXPORT_SYMBOL(inode_set_flags
);
2398 void inode_nohighmem(struct inode
*inode
)
2400 mapping_set_gfp_mask(inode
->i_mapping
, GFP_USER
);
2402 EXPORT_SYMBOL(inode_nohighmem
);
2405 * timestamp_truncate - Truncate timespec to a granularity
2407 * @inode: inode being updated
2409 * Truncate a timespec to the granularity supported by the fs
2410 * containing the inode. Always rounds down. gran must
2411 * not be 0 nor greater than a second (NSEC_PER_SEC, or 10^9 ns).
2413 struct timespec64
timestamp_truncate(struct timespec64 t
, struct inode
*inode
)
2415 struct super_block
*sb
= inode
->i_sb
;
2416 unsigned int gran
= sb
->s_time_gran
;
2418 t
.tv_sec
= clamp(t
.tv_sec
, sb
->s_time_min
, sb
->s_time_max
);
2419 if (unlikely(t
.tv_sec
== sb
->s_time_max
|| t
.tv_sec
== sb
->s_time_min
))
2422 /* Avoid division in the common cases 1 ns and 1 s. */
2425 else if (gran
== NSEC_PER_SEC
)
2427 else if (gran
> 1 && gran
< NSEC_PER_SEC
)
2428 t
.tv_nsec
-= t
.tv_nsec
% gran
;
2430 WARN(1, "invalid file time granularity: %u", gran
);
2433 EXPORT_SYMBOL(timestamp_truncate
);
2436 * current_time - Return FS time
2439 * Return the current time truncated to the time granularity supported by
2442 * Note that inode and inode->sb cannot be NULL.
2443 * Otherwise, the function warns and returns time without truncation.
2445 struct timespec64
current_time(struct inode
*inode
)
2447 struct timespec64 now
;
2449 ktime_get_coarse_real_ts64(&now
);
2451 if (unlikely(!inode
->i_sb
)) {
2452 WARN(1, "current_time() called with uninitialized super_block in the inode");
2456 return timestamp_truncate(now
, inode
);
2458 EXPORT_SYMBOL(current_time
);
2461 * in_group_or_capable - check whether caller is CAP_FSETID privileged
2462 * @idmap: idmap of the mount @inode was found from
2463 * @inode: inode to check
2464 * @vfsgid: the new/current vfsgid of @inode
2466 * Check wether @vfsgid is in the caller's group list or if the caller is
2467 * privileged with CAP_FSETID over @inode. This can be used to determine
2468 * whether the setgid bit can be kept or must be dropped.
2470 * Return: true if the caller is sufficiently privileged, false if not.
2472 bool in_group_or_capable(struct mnt_idmap
*idmap
,
2473 const struct inode
*inode
, vfsgid_t vfsgid
)
2475 if (vfsgid_in_group_p(vfsgid
))
2477 if (capable_wrt_inode_uidgid(idmap
, inode
, CAP_FSETID
))
2483 * mode_strip_sgid - handle the sgid bit for non-directories
2484 * @idmap: idmap of the mount the inode was created from
2485 * @dir: parent directory inode
2486 * @mode: mode of the file to be created in @dir
2488 * If the @mode of the new file has both the S_ISGID and S_IXGRP bit
2489 * raised and @dir has the S_ISGID bit raised ensure that the caller is
2490 * either in the group of the parent directory or they have CAP_FSETID
2491 * in their user namespace and are privileged over the parent directory.
2492 * In all other cases, strip the S_ISGID bit from @mode.
2494 * Return: the new mode to use for the file
2496 umode_t
mode_strip_sgid(struct mnt_idmap
*idmap
,
2497 const struct inode
*dir
, umode_t mode
)
2499 if ((mode
& (S_ISGID
| S_IXGRP
)) != (S_ISGID
| S_IXGRP
))
2501 if (S_ISDIR(mode
) || !dir
|| !(dir
->i_mode
& S_ISGID
))
2503 if (in_group_or_capable(idmap
, dir
, i_gid_into_vfsgid(idmap
, dir
)))
2505 return mode
& ~S_ISGID
;
2507 EXPORT_SYMBOL(mode_strip_sgid
);