]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/inode.c | |
3 | * | |
4 | * (C) 1997 Linus Torvalds | |
5 | */ | |
6 | ||
1da177e4 LT |
7 | #include <linux/fs.h> |
8 | #include <linux/mm.h> | |
9 | #include <linux/dcache.h> | |
10 | #include <linux/init.h> | |
1da177e4 LT |
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> | |
88e0fbc4 | 16 | #include <linux/rwsem.h> |
1da177e4 LT |
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> | |
3be25f49 | 23 | #include <linux/fsnotify.h> |
fc33a7bb | 24 | #include <linux/mount.h> |
efaee192 | 25 | #include <linux/async.h> |
f19d4a8f | 26 | #include <linux/posix_acl.h> |
1da177e4 LT |
27 | |
28 | /* | |
29 | * This is needed for the following functions: | |
30 | * - inode_has_buffers | |
31 | * - invalidate_inode_buffers | |
1da177e4 LT |
32 | * - invalidate_bdev |
33 | * | |
34 | * FIXME: remove all knowledge of the buffer layer from this file | |
35 | */ | |
36 | #include <linux/buffer_head.h> | |
37 | ||
38 | /* | |
39 | * New inode.c implementation. | |
40 | * | |
41 | * This implementation has the basic premise of trying | |
42 | * to be extremely low-overhead and SMP-safe, yet be | |
43 | * simple enough to be "obviously correct". | |
44 | * | |
45 | * Famous last words. | |
46 | */ | |
47 | ||
48 | /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */ | |
49 | ||
50 | /* #define INODE_PARANOIA 1 */ | |
51 | /* #define INODE_DEBUG 1 */ | |
52 | ||
53 | /* | |
54 | * Inode lookup is no longer as critical as it used to be: | |
55 | * most of the lookups are going to be through the dcache. | |
56 | */ | |
57 | #define I_HASHBITS i_hash_shift | |
58 | #define I_HASHMASK i_hash_mask | |
59 | ||
fa3536cc ED |
60 | static unsigned int i_hash_mask __read_mostly; |
61 | static unsigned int i_hash_shift __read_mostly; | |
1da177e4 LT |
62 | |
63 | /* | |
64 | * Each inode can be on two separate lists. One is | |
65 | * the hash list of the inode, used for lookups. The | |
66 | * other linked list is the "type" list: | |
67 | * "in_use" - valid inode, i_count > 0, i_nlink > 0 | |
68 | * "dirty" - as "in_use" but also dirty | |
69 | * "unused" - valid inode, i_count = 0 | |
70 | * | |
71 | * A "dirty" list is maintained for each super block, | |
72 | * allowing for low-overhead inode sync() operations. | |
73 | */ | |
74 | ||
75 | LIST_HEAD(inode_in_use); | |
76 | LIST_HEAD(inode_unused); | |
fa3536cc | 77 | static struct hlist_head *inode_hashtable __read_mostly; |
1da177e4 LT |
78 | |
79 | /* | |
80 | * A simple spinlock to protect the list manipulations. | |
81 | * | |
82 | * NOTE! You also have to own the lock if you change | |
83 | * the i_state of an inode while it is in use.. | |
84 | */ | |
85 | DEFINE_SPINLOCK(inode_lock); | |
86 | ||
87 | /* | |
88e0fbc4 | 88 | * iprune_sem provides exclusion between the kswapd or try_to_free_pages |
1da177e4 LT |
89 | * icache shrinking path, and the umount path. Without this exclusion, |
90 | * by the time prune_icache calls iput for the inode whose pages it has | |
91 | * been invalidating, or by the time it calls clear_inode & destroy_inode | |
92 | * from its final dispose_list, the struct super_block they refer to | |
93 | * (for inode->i_sb->s_op) may already have been freed and reused. | |
88e0fbc4 NP |
94 | * |
95 | * We make this an rwsem because the fastpath is icache shrinking. In | |
96 | * some cases a filesystem may be doing a significant amount of work in | |
97 | * its inode reclaim code, so this should improve parallelism. | |
1da177e4 | 98 | */ |
88e0fbc4 | 99 | static DECLARE_RWSEM(iprune_sem); |
1da177e4 LT |
100 | |
101 | /* | |
102 | * Statistics gathering.. | |
103 | */ | |
104 | struct inodes_stat_t inodes_stat; | |
105 | ||
cffbc8aa DC |
106 | static struct percpu_counter nr_inodes __cacheline_aligned_in_smp; |
107 | static struct percpu_counter nr_inodes_unused __cacheline_aligned_in_smp; | |
108 | ||
6b3304b5 | 109 | static struct kmem_cache *inode_cachep __read_mostly; |
1da177e4 | 110 | |
cffbc8aa DC |
111 | static inline int get_nr_inodes(void) |
112 | { | |
113 | return percpu_counter_sum_positive(&nr_inodes); | |
114 | } | |
115 | ||
116 | static inline int get_nr_inodes_unused(void) | |
117 | { | |
118 | return percpu_counter_sum_positive(&nr_inodes_unused); | |
119 | } | |
120 | ||
121 | int get_nr_dirty_inodes(void) | |
122 | { | |
123 | int nr_dirty = get_nr_inodes() - get_nr_inodes_unused(); | |
124 | return nr_dirty > 0 ? nr_dirty : 0; | |
125 | ||
126 | } | |
127 | ||
128 | /* | |
129 | * Handle nr_inode sysctl | |
130 | */ | |
131 | #ifdef CONFIG_SYSCTL | |
132 | int proc_nr_inodes(ctl_table *table, int write, | |
133 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
134 | { | |
135 | inodes_stat.nr_inodes = get_nr_inodes(); | |
136 | inodes_stat.nr_unused = get_nr_inodes_unused(); | |
137 | return proc_dointvec(table, write, buffer, lenp, ppos); | |
138 | } | |
139 | #endif | |
140 | ||
1c0eeaf5 JE |
141 | static void wake_up_inode(struct inode *inode) |
142 | { | |
143 | /* | |
144 | * Prevent speculative execution through spin_unlock(&inode_lock); | |
145 | */ | |
146 | smp_mb(); | |
eaff8079 | 147 | wake_up_bit(&inode->i_state, __I_NEW); |
1c0eeaf5 JE |
148 | } |
149 | ||
2cb1599f DC |
150 | /** |
151 | * inode_init_always - perform inode structure intialisation | |
0bc02f3f RD |
152 | * @sb: superblock inode belongs to |
153 | * @inode: inode to initialise | |
2cb1599f DC |
154 | * |
155 | * These are initializations that need to be done on every inode | |
156 | * allocation as the fields are not initialised by slab allocation. | |
157 | */ | |
54e34621 | 158 | int inode_init_always(struct super_block *sb, struct inode *inode) |
1da177e4 | 159 | { |
f5e54d6e | 160 | static const struct address_space_operations empty_aops; |
6e1d5dcc | 161 | static const struct inode_operations empty_iops; |
99ac48f5 | 162 | static const struct file_operations empty_fops; |
6b3304b5 | 163 | struct address_space *const mapping = &inode->i_data; |
2cb1599f DC |
164 | |
165 | inode->i_sb = sb; | |
166 | inode->i_blkbits = sb->s_blocksize_bits; | |
167 | inode->i_flags = 0; | |
168 | atomic_set(&inode->i_count, 1); | |
169 | inode->i_op = &empty_iops; | |
170 | inode->i_fop = &empty_fops; | |
171 | inode->i_nlink = 1; | |
56ff5efa AV |
172 | inode->i_uid = 0; |
173 | inode->i_gid = 0; | |
2cb1599f DC |
174 | atomic_set(&inode->i_writecount, 0); |
175 | inode->i_size = 0; | |
176 | inode->i_blocks = 0; | |
177 | inode->i_bytes = 0; | |
178 | inode->i_generation = 0; | |
1da177e4 | 179 | #ifdef CONFIG_QUOTA |
2cb1599f | 180 | memset(&inode->i_dquot, 0, sizeof(inode->i_dquot)); |
1da177e4 | 181 | #endif |
2cb1599f DC |
182 | inode->i_pipe = NULL; |
183 | inode->i_bdev = NULL; | |
184 | inode->i_cdev = NULL; | |
185 | inode->i_rdev = 0; | |
186 | inode->dirtied_when = 0; | |
6146f0d5 MZ |
187 | |
188 | if (security_inode_alloc(inode)) | |
54e34621 | 189 | goto out; |
2cb1599f DC |
190 | spin_lock_init(&inode->i_lock); |
191 | lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key); | |
192 | ||
193 | mutex_init(&inode->i_mutex); | |
194 | lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key); | |
195 | ||
196 | init_rwsem(&inode->i_alloc_sem); | |
197 | lockdep_set_class(&inode->i_alloc_sem, &sb->s_type->i_alloc_sem_key); | |
198 | ||
199 | mapping->a_ops = &empty_aops; | |
200 | mapping->host = inode; | |
201 | mapping->flags = 0; | |
3c1d4378 | 202 | mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE); |
2cb1599f DC |
203 | mapping->assoc_mapping = NULL; |
204 | mapping->backing_dev_info = &default_backing_dev_info; | |
205 | mapping->writeback_index = 0; | |
206 | ||
207 | /* | |
208 | * If the block_device provides a backing_dev_info for client | |
209 | * inodes then use that. Otherwise the inode share the bdev's | |
210 | * backing_dev_info. | |
211 | */ | |
212 | if (sb->s_bdev) { | |
213 | struct backing_dev_info *bdi; | |
214 | ||
2c96ce9f | 215 | bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info; |
2cb1599f DC |
216 | mapping->backing_dev_info = bdi; |
217 | } | |
218 | inode->i_private = NULL; | |
219 | inode->i_mapping = mapping; | |
f19d4a8f AV |
220 | #ifdef CONFIG_FS_POSIX_ACL |
221 | inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED; | |
222 | #endif | |
2cb1599f | 223 | |
3be25f49 EP |
224 | #ifdef CONFIG_FSNOTIFY |
225 | inode->i_fsnotify_mask = 0; | |
226 | #endif | |
227 | ||
cffbc8aa DC |
228 | percpu_counter_inc(&nr_inodes); |
229 | ||
54e34621 | 230 | return 0; |
54e34621 CH |
231 | out: |
232 | return -ENOMEM; | |
1da177e4 | 233 | } |
2cb1599f DC |
234 | EXPORT_SYMBOL(inode_init_always); |
235 | ||
236 | static struct inode *alloc_inode(struct super_block *sb) | |
237 | { | |
238 | struct inode *inode; | |
239 | ||
240 | if (sb->s_op->alloc_inode) | |
241 | inode = sb->s_op->alloc_inode(sb); | |
242 | else | |
243 | inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL); | |
244 | ||
54e34621 CH |
245 | if (!inode) |
246 | return NULL; | |
247 | ||
248 | if (unlikely(inode_init_always(sb, inode))) { | |
249 | if (inode->i_sb->s_op->destroy_inode) | |
250 | inode->i_sb->s_op->destroy_inode(inode); | |
251 | else | |
252 | kmem_cache_free(inode_cachep, inode); | |
253 | return NULL; | |
254 | } | |
255 | ||
256 | return inode; | |
2cb1599f | 257 | } |
1da177e4 | 258 | |
2e00c97e | 259 | void __destroy_inode(struct inode *inode) |
1da177e4 | 260 | { |
b7542f8c | 261 | BUG_ON(inode_has_buffers(inode)); |
1da177e4 | 262 | security_inode_free(inode); |
3be25f49 | 263 | fsnotify_inode_delete(inode); |
f19d4a8f AV |
264 | #ifdef CONFIG_FS_POSIX_ACL |
265 | if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED) | |
266 | posix_acl_release(inode->i_acl); | |
267 | if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED) | |
268 | posix_acl_release(inode->i_default_acl); | |
269 | #endif | |
cffbc8aa | 270 | percpu_counter_dec(&nr_inodes); |
2e00c97e CH |
271 | } |
272 | EXPORT_SYMBOL(__destroy_inode); | |
273 | ||
56b0dacf | 274 | static void destroy_inode(struct inode *inode) |
2e00c97e CH |
275 | { |
276 | __destroy_inode(inode); | |
1da177e4 LT |
277 | if (inode->i_sb->s_op->destroy_inode) |
278 | inode->i_sb->s_op->destroy_inode(inode); | |
279 | else | |
280 | kmem_cache_free(inode_cachep, (inode)); | |
281 | } | |
1da177e4 LT |
282 | |
283 | /* | |
284 | * These are initializations that only need to be done | |
285 | * once, because the fields are idempotent across use | |
286 | * of the inode, so let the slab aware of that. | |
287 | */ | |
288 | void inode_init_once(struct inode *inode) | |
289 | { | |
290 | memset(inode, 0, sizeof(*inode)); | |
291 | INIT_HLIST_NODE(&inode->i_hash); | |
292 | INIT_LIST_HEAD(&inode->i_dentry); | |
293 | INIT_LIST_HEAD(&inode->i_devices); | |
1da177e4 | 294 | INIT_RADIX_TREE(&inode->i_data.page_tree, GFP_ATOMIC); |
19fd6231 | 295 | spin_lock_init(&inode->i_data.tree_lock); |
1da177e4 LT |
296 | spin_lock_init(&inode->i_data.i_mmap_lock); |
297 | INIT_LIST_HEAD(&inode->i_data.private_list); | |
298 | spin_lock_init(&inode->i_data.private_lock); | |
299 | INIT_RAW_PRIO_TREE_ROOT(&inode->i_data.i_mmap); | |
300 | INIT_LIST_HEAD(&inode->i_data.i_mmap_nonlinear); | |
1da177e4 | 301 | i_size_ordered_init(inode); |
3be25f49 | 302 | #ifdef CONFIG_FSNOTIFY |
e61ce867 | 303 | INIT_HLIST_HEAD(&inode->i_fsnotify_marks); |
3be25f49 | 304 | #endif |
1da177e4 | 305 | } |
1da177e4 LT |
306 | EXPORT_SYMBOL(inode_init_once); |
307 | ||
51cc5068 | 308 | static void init_once(void *foo) |
1da177e4 | 309 | { |
6b3304b5 | 310 | struct inode *inode = (struct inode *) foo; |
1da177e4 | 311 | |
a35afb83 | 312 | inode_init_once(inode); |
1da177e4 LT |
313 | } |
314 | ||
315 | /* | |
316 | * inode_lock must be held | |
317 | */ | |
6b3304b5 | 318 | void __iget(struct inode *inode) |
1da177e4 | 319 | { |
2e147f1e | 320 | if (atomic_inc_return(&inode->i_count) != 1) |
1da177e4 | 321 | return; |
2e147f1e | 322 | |
1c0eeaf5 | 323 | if (!(inode->i_state & (I_DIRTY|I_SYNC))) |
1da177e4 | 324 | list_move(&inode->i_list, &inode_in_use); |
cffbc8aa | 325 | percpu_counter_dec(&nr_inodes_unused); |
1da177e4 LT |
326 | } |
327 | ||
b0683aa6 AV |
328 | void end_writeback(struct inode *inode) |
329 | { | |
330 | might_sleep(); | |
331 | BUG_ON(inode->i_data.nrpages); | |
332 | BUG_ON(!list_empty(&inode->i_data.private_list)); | |
333 | BUG_ON(!(inode->i_state & I_FREEING)); | |
334 | BUG_ON(inode->i_state & I_CLEAR); | |
335 | inode_sync_wait(inode); | |
336 | inode->i_state = I_FREEING | I_CLEAR; | |
337 | } | |
338 | EXPORT_SYMBOL(end_writeback); | |
339 | ||
644da596 | 340 | static void evict(struct inode *inode) |
b4272d4c AV |
341 | { |
342 | const struct super_operations *op = inode->i_sb->s_op; | |
343 | ||
be7ce416 AV |
344 | if (op->evict_inode) { |
345 | op->evict_inode(inode); | |
b4272d4c AV |
346 | } else { |
347 | if (inode->i_data.nrpages) | |
348 | truncate_inode_pages(&inode->i_data, 0); | |
30140837 | 349 | end_writeback(inode); |
b4272d4c | 350 | } |
661074e9 AV |
351 | if (S_ISBLK(inode->i_mode) && inode->i_bdev) |
352 | bd_forget(inode); | |
353 | if (S_ISCHR(inode->i_mode) && inode->i_cdev) | |
354 | cd_forget(inode); | |
b4272d4c AV |
355 | } |
356 | ||
1da177e4 LT |
357 | /* |
358 | * dispose_list - dispose of the contents of a local list | |
359 | * @head: the head of the list to free | |
360 | * | |
361 | * Dispose-list gets a local list with local inodes in it, so it doesn't | |
362 | * need to worry about list corruption and SMP locks. | |
363 | */ | |
364 | static void dispose_list(struct list_head *head) | |
365 | { | |
1da177e4 LT |
366 | while (!list_empty(head)) { |
367 | struct inode *inode; | |
368 | ||
b5e61818 | 369 | inode = list_first_entry(head, struct inode, i_list); |
1da177e4 LT |
370 | list_del(&inode->i_list); |
371 | ||
644da596 | 372 | evict(inode); |
4120db47 AB |
373 | |
374 | spin_lock(&inode_lock); | |
375 | hlist_del_init(&inode->i_hash); | |
376 | list_del_init(&inode->i_sb_list); | |
377 | spin_unlock(&inode_lock); | |
378 | ||
379 | wake_up_inode(inode); | |
1da177e4 | 380 | destroy_inode(inode); |
1da177e4 | 381 | } |
1da177e4 LT |
382 | } |
383 | ||
384 | /* | |
385 | * Invalidate all inodes for a device. | |
386 | */ | |
387 | static int invalidate_list(struct list_head *head, struct list_head *dispose) | |
388 | { | |
389 | struct list_head *next; | |
cffbc8aa | 390 | int busy = 0; |
1da177e4 LT |
391 | |
392 | next = head->next; | |
393 | for (;;) { | |
6b3304b5 MK |
394 | struct list_head *tmp = next; |
395 | struct inode *inode; | |
1da177e4 LT |
396 | |
397 | /* | |
398 | * We can reschedule here without worrying about the list's | |
399 | * consistency because the per-sb list of inodes must not | |
88e0fbc4 | 400 | * change during umount anymore, and because iprune_sem keeps |
1da177e4 LT |
401 | * shrink_icache_memory() away. |
402 | */ | |
403 | cond_resched_lock(&inode_lock); | |
404 | ||
405 | next = next->next; | |
406 | if (tmp == head) | |
407 | break; | |
408 | inode = list_entry(tmp, struct inode, i_sb_list); | |
aabb8fdb NP |
409 | if (inode->i_state & I_NEW) |
410 | continue; | |
1da177e4 LT |
411 | invalidate_inode_buffers(inode); |
412 | if (!atomic_read(&inode->i_count)) { | |
1da177e4 | 413 | list_move(&inode->i_list, dispose); |
7ef0d737 | 414 | WARN_ON(inode->i_state & I_NEW); |
1da177e4 | 415 | inode->i_state |= I_FREEING; |
cffbc8aa | 416 | percpu_counter_dec(&nr_inodes_unused); |
1da177e4 LT |
417 | continue; |
418 | } | |
419 | busy = 1; | |
420 | } | |
1da177e4 LT |
421 | return busy; |
422 | } | |
423 | ||
1da177e4 LT |
424 | /** |
425 | * invalidate_inodes - discard the inodes on a device | |
426 | * @sb: superblock | |
427 | * | |
428 | * Discard all of the inodes for a given superblock. If the discard | |
429 | * fails because there are busy inodes then a non zero value is returned. | |
430 | * If the discard is successful all the inodes have been discarded. | |
431 | */ | |
6b3304b5 | 432 | int invalidate_inodes(struct super_block *sb) |
1da177e4 LT |
433 | { |
434 | int busy; | |
435 | LIST_HEAD(throw_away); | |
436 | ||
88e0fbc4 | 437 | down_write(&iprune_sem); |
1da177e4 | 438 | spin_lock(&inode_lock); |
164bc619 | 439 | fsnotify_unmount_inodes(&sb->s_inodes); |
1da177e4 LT |
440 | busy = invalidate_list(&sb->s_inodes, &throw_away); |
441 | spin_unlock(&inode_lock); | |
442 | ||
443 | dispose_list(&throw_away); | |
88e0fbc4 | 444 | up_write(&iprune_sem); |
1da177e4 LT |
445 | |
446 | return busy; | |
447 | } | |
1da177e4 LT |
448 | |
449 | static int can_unuse(struct inode *inode) | |
450 | { | |
451 | if (inode->i_state) | |
452 | return 0; | |
453 | if (inode_has_buffers(inode)) | |
454 | return 0; | |
455 | if (atomic_read(&inode->i_count)) | |
456 | return 0; | |
457 | if (inode->i_data.nrpages) | |
458 | return 0; | |
459 | return 1; | |
460 | } | |
461 | ||
462 | /* | |
463 | * Scan `goal' inodes on the unused list for freeable ones. They are moved to | |
464 | * a temporary list and then are freed outside inode_lock by dispose_list(). | |
465 | * | |
466 | * Any inodes which are pinned purely because of attached pagecache have their | |
467 | * pagecache removed. We expect the final iput() on that inode to add it to | |
468 | * the front of the inode_unused list. So look for it there and if the | |
469 | * inode is still freeable, proceed. The right inode is found 99.9% of the | |
470 | * time in testing on a 4-way. | |
471 | * | |
472 | * If the inode has metadata buffers attached to mapping->private_list then | |
473 | * try to remove them. | |
474 | */ | |
475 | static void prune_icache(int nr_to_scan) | |
476 | { | |
477 | LIST_HEAD(freeable); | |
1da177e4 LT |
478 | int nr_scanned; |
479 | unsigned long reap = 0; | |
480 | ||
88e0fbc4 | 481 | down_read(&iprune_sem); |
1da177e4 LT |
482 | spin_lock(&inode_lock); |
483 | for (nr_scanned = 0; nr_scanned < nr_to_scan; nr_scanned++) { | |
484 | struct inode *inode; | |
485 | ||
486 | if (list_empty(&inode_unused)) | |
487 | break; | |
488 | ||
489 | inode = list_entry(inode_unused.prev, struct inode, i_list); | |
490 | ||
491 | if (inode->i_state || atomic_read(&inode->i_count)) { | |
492 | list_move(&inode->i_list, &inode_unused); | |
493 | continue; | |
494 | } | |
495 | if (inode_has_buffers(inode) || inode->i_data.nrpages) { | |
496 | __iget(inode); | |
497 | spin_unlock(&inode_lock); | |
498 | if (remove_inode_buffers(inode)) | |
fc0ecff6 AM |
499 | reap += invalidate_mapping_pages(&inode->i_data, |
500 | 0, -1); | |
1da177e4 LT |
501 | iput(inode); |
502 | spin_lock(&inode_lock); | |
503 | ||
504 | if (inode != list_entry(inode_unused.next, | |
505 | struct inode, i_list)) | |
506 | continue; /* wrong inode or list_empty */ | |
507 | if (!can_unuse(inode)) | |
508 | continue; | |
509 | } | |
1da177e4 | 510 | list_move(&inode->i_list, &freeable); |
7ef0d737 | 511 | WARN_ON(inode->i_state & I_NEW); |
1da177e4 | 512 | inode->i_state |= I_FREEING; |
cffbc8aa | 513 | percpu_counter_dec(&nr_inodes_unused); |
1da177e4 | 514 | } |
f8891e5e CL |
515 | if (current_is_kswapd()) |
516 | __count_vm_events(KSWAPD_INODESTEAL, reap); | |
517 | else | |
518 | __count_vm_events(PGINODESTEAL, reap); | |
1da177e4 LT |
519 | spin_unlock(&inode_lock); |
520 | ||
521 | dispose_list(&freeable); | |
88e0fbc4 | 522 | up_read(&iprune_sem); |
1da177e4 LT |
523 | } |
524 | ||
525 | /* | |
526 | * shrink_icache_memory() will attempt to reclaim some unused inodes. Here, | |
527 | * "unused" means that no dentries are referring to the inodes: the files are | |
528 | * not open and the dcache references to those inodes have already been | |
529 | * reclaimed. | |
530 | * | |
531 | * This function is passed the number of inodes to scan, and it returns the | |
532 | * total number of remaining possibly-reclaimable inodes. | |
533 | */ | |
7f8275d0 | 534 | static int shrink_icache_memory(struct shrinker *shrink, int nr, gfp_t gfp_mask) |
1da177e4 LT |
535 | { |
536 | if (nr) { | |
537 | /* | |
538 | * Nasty deadlock avoidance. We may hold various FS locks, | |
539 | * and we don't want to recurse into the FS that called us | |
540 | * in clear_inode() and friends.. | |
6b3304b5 | 541 | */ |
1da177e4 LT |
542 | if (!(gfp_mask & __GFP_FS)) |
543 | return -1; | |
544 | prune_icache(nr); | |
545 | } | |
cffbc8aa | 546 | return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure; |
1da177e4 LT |
547 | } |
548 | ||
8e1f936b RR |
549 | static struct shrinker icache_shrinker = { |
550 | .shrink = shrink_icache_memory, | |
551 | .seeks = DEFAULT_SEEKS, | |
552 | }; | |
553 | ||
1da177e4 LT |
554 | static void __wait_on_freeing_inode(struct inode *inode); |
555 | /* | |
556 | * Called with the inode lock held. | |
557 | * NOTE: we are not increasing the inode-refcount, you must call __iget() | |
558 | * by hand after calling find_inode now! This simplifies iunique and won't | |
559 | * add any additional branch in the common code. | |
560 | */ | |
6b3304b5 MK |
561 | static struct inode *find_inode(struct super_block *sb, |
562 | struct hlist_head *head, | |
563 | int (*test)(struct inode *, void *), | |
564 | void *data) | |
1da177e4 LT |
565 | { |
566 | struct hlist_node *node; | |
6b3304b5 | 567 | struct inode *inode = NULL; |
1da177e4 LT |
568 | |
569 | repeat: | |
c5c8be3c | 570 | hlist_for_each_entry(inode, node, head, i_hash) { |
1da177e4 LT |
571 | if (inode->i_sb != sb) |
572 | continue; | |
573 | if (!test(inode, data)) | |
574 | continue; | |
a4ffdde6 | 575 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
576 | __wait_on_freeing_inode(inode); |
577 | goto repeat; | |
578 | } | |
579 | break; | |
580 | } | |
581 | return node ? inode : NULL; | |
582 | } | |
583 | ||
584 | /* | |
585 | * find_inode_fast is the fast path version of find_inode, see the comment at | |
586 | * iget_locked for details. | |
587 | */ | |
6b3304b5 MK |
588 | static struct inode *find_inode_fast(struct super_block *sb, |
589 | struct hlist_head *head, unsigned long ino) | |
1da177e4 LT |
590 | { |
591 | struct hlist_node *node; | |
6b3304b5 | 592 | struct inode *inode = NULL; |
1da177e4 LT |
593 | |
594 | repeat: | |
c5c8be3c | 595 | hlist_for_each_entry(inode, node, head, i_hash) { |
1da177e4 LT |
596 | if (inode->i_ino != ino) |
597 | continue; | |
598 | if (inode->i_sb != sb) | |
599 | continue; | |
a4ffdde6 | 600 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
601 | __wait_on_freeing_inode(inode); |
602 | goto repeat; | |
603 | } | |
604 | break; | |
605 | } | |
606 | return node ? inode : NULL; | |
607 | } | |
608 | ||
8290c35f DC |
609 | static unsigned long hash(struct super_block *sb, unsigned long hashval) |
610 | { | |
611 | unsigned long tmp; | |
612 | ||
613 | tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) / | |
614 | L1_CACHE_BYTES; | |
615 | tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> I_HASHBITS); | |
616 | return tmp & I_HASHMASK; | |
617 | } | |
618 | ||
619 | static inline void | |
620 | __inode_add_to_lists(struct super_block *sb, struct hlist_head *head, | |
621 | struct inode *inode) | |
622 | { | |
8290c35f DC |
623 | list_add(&inode->i_list, &inode_in_use); |
624 | list_add(&inode->i_sb_list, &sb->s_inodes); | |
625 | if (head) | |
626 | hlist_add_head(&inode->i_hash, head); | |
627 | } | |
628 | ||
629 | /** | |
630 | * inode_add_to_lists - add a new inode to relevant lists | |
0bc02f3f RD |
631 | * @sb: superblock inode belongs to |
632 | * @inode: inode to mark in use | |
8290c35f DC |
633 | * |
634 | * When an inode is allocated it needs to be accounted for, added to the in use | |
635 | * list, the owning superblock and the inode hash. This needs to be done under | |
636 | * the inode_lock, so export a function to do this rather than the inode lock | |
637 | * itself. We calculate the hash list to add to here so it is all internal | |
638 | * which requires the caller to have already set up the inode number in the | |
639 | * inode to add. | |
640 | */ | |
641 | void inode_add_to_lists(struct super_block *sb, struct inode *inode) | |
642 | { | |
643 | struct hlist_head *head = inode_hashtable + hash(sb, inode->i_ino); | |
644 | ||
645 | spin_lock(&inode_lock); | |
646 | __inode_add_to_lists(sb, head, inode); | |
647 | spin_unlock(&inode_lock); | |
648 | } | |
649 | EXPORT_SYMBOL_GPL(inode_add_to_lists); | |
650 | ||
1da177e4 LT |
651 | /** |
652 | * new_inode - obtain an inode | |
653 | * @sb: superblock | |
654 | * | |
769848c0 | 655 | * Allocates a new inode for given superblock. The default gfp_mask |
3c1d4378 | 656 | * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE. |
769848c0 MG |
657 | * If HIGHMEM pages are unsuitable or it is known that pages allocated |
658 | * for the page cache are not reclaimable or migratable, | |
659 | * mapping_set_gfp_mask() must be called with suitable flags on the | |
660 | * newly created inode's mapping | |
661 | * | |
1da177e4 LT |
662 | */ |
663 | struct inode *new_inode(struct super_block *sb) | |
664 | { | |
866b04fc JL |
665 | /* |
666 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
667 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
668 | * here to attempt to avoid that. | |
669 | */ | |
670 | static unsigned int last_ino; | |
6b3304b5 | 671 | struct inode *inode; |
1da177e4 LT |
672 | |
673 | spin_lock_prefetch(&inode_lock); | |
6b3304b5 | 674 | |
1da177e4 LT |
675 | inode = alloc_inode(sb); |
676 | if (inode) { | |
677 | spin_lock(&inode_lock); | |
8290c35f | 678 | __inode_add_to_lists(sb, NULL, inode); |
1da177e4 LT |
679 | inode->i_ino = ++last_ino; |
680 | inode->i_state = 0; | |
681 | spin_unlock(&inode_lock); | |
682 | } | |
683 | return inode; | |
684 | } | |
1da177e4 LT |
685 | EXPORT_SYMBOL(new_inode); |
686 | ||
687 | void unlock_new_inode(struct inode *inode) | |
688 | { | |
14358e6d | 689 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
a3314a0e | 690 | if (S_ISDIR(inode->i_mode)) { |
1e89a5e1 PZ |
691 | struct file_system_type *type = inode->i_sb->s_type; |
692 | ||
9a7aa12f JK |
693 | /* Set new key only if filesystem hasn't already changed it */ |
694 | if (!lockdep_match_class(&inode->i_mutex, | |
695 | &type->i_mutex_key)) { | |
696 | /* | |
697 | * ensure nobody is actually holding i_mutex | |
698 | */ | |
699 | mutex_destroy(&inode->i_mutex); | |
700 | mutex_init(&inode->i_mutex); | |
701 | lockdep_set_class(&inode->i_mutex, | |
702 | &type->i_mutex_dir_key); | |
703 | } | |
1e89a5e1 | 704 | } |
14358e6d | 705 | #endif |
1da177e4 | 706 | /* |
eaff8079 | 707 | * This is special! We do not need the spinlock when clearing I_NEW, |
580be083 JK |
708 | * because we're guaranteed that nobody else tries to do anything about |
709 | * the state of the inode when it is locked, as we just created it (so | |
eaff8079 | 710 | * there can be no old holders that haven't tested I_NEW). |
580be083 | 711 | * However we must emit the memory barrier so that other CPUs reliably |
eaff8079 | 712 | * see the clearing of I_NEW after the other inode initialisation has |
580be083 | 713 | * completed. |
1da177e4 | 714 | */ |
580be083 | 715 | smp_mb(); |
eaff8079 CH |
716 | WARN_ON(!(inode->i_state & I_NEW)); |
717 | inode->i_state &= ~I_NEW; | |
1da177e4 LT |
718 | wake_up_inode(inode); |
719 | } | |
1da177e4 LT |
720 | EXPORT_SYMBOL(unlock_new_inode); |
721 | ||
722 | /* | |
723 | * This is called without the inode lock held.. Be careful. | |
724 | * | |
725 | * We no longer cache the sb_flags in i_flags - see fs.h | |
726 | * -- rmk@arm.uk.linux.org | |
727 | */ | |
6b3304b5 MK |
728 | static struct inode *get_new_inode(struct super_block *sb, |
729 | struct hlist_head *head, | |
730 | int (*test)(struct inode *, void *), | |
731 | int (*set)(struct inode *, void *), | |
732 | void *data) | |
1da177e4 | 733 | { |
6b3304b5 | 734 | struct inode *inode; |
1da177e4 LT |
735 | |
736 | inode = alloc_inode(sb); | |
737 | if (inode) { | |
6b3304b5 | 738 | struct inode *old; |
1da177e4 LT |
739 | |
740 | spin_lock(&inode_lock); | |
741 | /* We released the lock, so.. */ | |
742 | old = find_inode(sb, head, test, data); | |
743 | if (!old) { | |
744 | if (set(inode, data)) | |
745 | goto set_failed; | |
746 | ||
8290c35f | 747 | __inode_add_to_lists(sb, head, inode); |
eaff8079 | 748 | inode->i_state = I_NEW; |
1da177e4 LT |
749 | spin_unlock(&inode_lock); |
750 | ||
751 | /* Return the locked inode with I_NEW set, the | |
752 | * caller is responsible for filling in the contents | |
753 | */ | |
754 | return inode; | |
755 | } | |
756 | ||
757 | /* | |
758 | * Uhhuh, somebody else created the same inode under | |
759 | * us. Use the old inode instead of the one we just | |
760 | * allocated. | |
761 | */ | |
762 | __iget(old); | |
763 | spin_unlock(&inode_lock); | |
764 | destroy_inode(inode); | |
765 | inode = old; | |
766 | wait_on_inode(inode); | |
767 | } | |
768 | return inode; | |
769 | ||
770 | set_failed: | |
771 | spin_unlock(&inode_lock); | |
772 | destroy_inode(inode); | |
773 | return NULL; | |
774 | } | |
775 | ||
776 | /* | |
777 | * get_new_inode_fast is the fast path version of get_new_inode, see the | |
778 | * comment at iget_locked for details. | |
779 | */ | |
6b3304b5 MK |
780 | static struct inode *get_new_inode_fast(struct super_block *sb, |
781 | struct hlist_head *head, unsigned long ino) | |
1da177e4 | 782 | { |
6b3304b5 | 783 | struct inode *inode; |
1da177e4 LT |
784 | |
785 | inode = alloc_inode(sb); | |
786 | if (inode) { | |
6b3304b5 | 787 | struct inode *old; |
1da177e4 LT |
788 | |
789 | spin_lock(&inode_lock); | |
790 | /* We released the lock, so.. */ | |
791 | old = find_inode_fast(sb, head, ino); | |
792 | if (!old) { | |
793 | inode->i_ino = ino; | |
8290c35f | 794 | __inode_add_to_lists(sb, head, inode); |
eaff8079 | 795 | inode->i_state = I_NEW; |
1da177e4 LT |
796 | spin_unlock(&inode_lock); |
797 | ||
798 | /* Return the locked inode with I_NEW set, the | |
799 | * caller is responsible for filling in the contents | |
800 | */ | |
801 | return inode; | |
802 | } | |
803 | ||
804 | /* | |
805 | * Uhhuh, somebody else created the same inode under | |
806 | * us. Use the old inode instead of the one we just | |
807 | * allocated. | |
808 | */ | |
809 | __iget(old); | |
810 | spin_unlock(&inode_lock); | |
811 | destroy_inode(inode); | |
812 | inode = old; | |
813 | wait_on_inode(inode); | |
814 | } | |
815 | return inode; | |
816 | } | |
817 | ||
1da177e4 LT |
818 | /** |
819 | * iunique - get a unique inode number | |
820 | * @sb: superblock | |
821 | * @max_reserved: highest reserved inode number | |
822 | * | |
823 | * Obtain an inode number that is unique on the system for a given | |
824 | * superblock. This is used by file systems that have no natural | |
825 | * permanent inode numbering system. An inode number is returned that | |
826 | * is higher than the reserved limit but unique. | |
827 | * | |
828 | * BUGS: | |
829 | * With a large number of inodes live on the file system this function | |
830 | * currently becomes quite slow. | |
831 | */ | |
832 | ino_t iunique(struct super_block *sb, ino_t max_reserved) | |
833 | { | |
866b04fc JL |
834 | /* |
835 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
836 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
837 | * here to attempt to avoid that. | |
838 | */ | |
839 | static unsigned int counter; | |
1da177e4 | 840 | struct inode *inode; |
3361c7be | 841 | struct hlist_head *head; |
1da177e4 | 842 | ino_t res; |
3361c7be | 843 | |
1da177e4 | 844 | spin_lock(&inode_lock); |
3361c7be JL |
845 | do { |
846 | if (counter <= max_reserved) | |
847 | counter = max_reserved + 1; | |
1da177e4 | 848 | res = counter++; |
3361c7be | 849 | head = inode_hashtable + hash(sb, res); |
1da177e4 | 850 | inode = find_inode_fast(sb, head, res); |
3361c7be JL |
851 | } while (inode != NULL); |
852 | spin_unlock(&inode_lock); | |
1da177e4 | 853 | |
3361c7be JL |
854 | return res; |
855 | } | |
1da177e4 LT |
856 | EXPORT_SYMBOL(iunique); |
857 | ||
858 | struct inode *igrab(struct inode *inode) | |
859 | { | |
860 | spin_lock(&inode_lock); | |
a4ffdde6 | 861 | if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) |
1da177e4 LT |
862 | __iget(inode); |
863 | else | |
864 | /* | |
865 | * Handle the case where s_op->clear_inode is not been | |
866 | * called yet, and somebody is calling igrab | |
867 | * while the inode is getting freed. | |
868 | */ | |
869 | inode = NULL; | |
870 | spin_unlock(&inode_lock); | |
871 | return inode; | |
872 | } | |
1da177e4 LT |
873 | EXPORT_SYMBOL(igrab); |
874 | ||
875 | /** | |
876 | * ifind - internal function, you want ilookup5() or iget5(). | |
877 | * @sb: super block of file system to search | |
878 | * @head: the head of the list to search | |
879 | * @test: callback used for comparisons between inodes | |
880 | * @data: opaque data pointer to pass to @test | |
88bd5121 | 881 | * @wait: if true wait for the inode to be unlocked, if false do not |
1da177e4 LT |
882 | * |
883 | * ifind() searches for the inode specified by @data in the inode | |
884 | * cache. This is a generalized version of ifind_fast() for file systems where | |
885 | * the inode number is not sufficient for unique identification of an inode. | |
886 | * | |
887 | * If the inode is in the cache, the inode is returned with an incremented | |
888 | * reference count. | |
889 | * | |
890 | * Otherwise NULL is returned. | |
891 | * | |
892 | * Note, @test is called with the inode_lock held, so can't sleep. | |
893 | */ | |
5d2bea45 | 894 | static struct inode *ifind(struct super_block *sb, |
1da177e4 | 895 | struct hlist_head *head, int (*test)(struct inode *, void *), |
88bd5121 | 896 | void *data, const int wait) |
1da177e4 LT |
897 | { |
898 | struct inode *inode; | |
899 | ||
900 | spin_lock(&inode_lock); | |
901 | inode = find_inode(sb, head, test, data); | |
902 | if (inode) { | |
903 | __iget(inode); | |
904 | spin_unlock(&inode_lock); | |
88bd5121 AA |
905 | if (likely(wait)) |
906 | wait_on_inode(inode); | |
1da177e4 LT |
907 | return inode; |
908 | } | |
909 | spin_unlock(&inode_lock); | |
910 | return NULL; | |
911 | } | |
912 | ||
913 | /** | |
914 | * ifind_fast - internal function, you want ilookup() or iget(). | |
915 | * @sb: super block of file system to search | |
916 | * @head: head of the list to search | |
917 | * @ino: inode number to search for | |
918 | * | |
919 | * ifind_fast() searches for the inode @ino in the inode cache. This is for | |
920 | * file systems where the inode number is sufficient for unique identification | |
921 | * of an inode. | |
922 | * | |
923 | * If the inode is in the cache, the inode is returned with an incremented | |
924 | * reference count. | |
925 | * | |
926 | * Otherwise NULL is returned. | |
927 | */ | |
5d2bea45 | 928 | static struct inode *ifind_fast(struct super_block *sb, |
1da177e4 LT |
929 | struct hlist_head *head, unsigned long ino) |
930 | { | |
931 | struct inode *inode; | |
932 | ||
933 | spin_lock(&inode_lock); | |
934 | inode = find_inode_fast(sb, head, ino); | |
935 | if (inode) { | |
936 | __iget(inode); | |
937 | spin_unlock(&inode_lock); | |
938 | wait_on_inode(inode); | |
939 | return inode; | |
940 | } | |
941 | spin_unlock(&inode_lock); | |
942 | return NULL; | |
943 | } | |
944 | ||
945 | /** | |
88bd5121 | 946 | * ilookup5_nowait - search for an inode in the inode cache |
1da177e4 LT |
947 | * @sb: super block of file system to search |
948 | * @hashval: hash value (usually inode number) to search for | |
949 | * @test: callback used for comparisons between inodes | |
950 | * @data: opaque data pointer to pass to @test | |
951 | * | |
952 | * ilookup5() uses ifind() to search for the inode specified by @hashval and | |
953 | * @data in the inode cache. This is a generalized version of ilookup() for | |
954 | * file systems where the inode number is not sufficient for unique | |
955 | * identification of an inode. | |
956 | * | |
957 | * If the inode is in the cache, the inode is returned with an incremented | |
88bd5121 AA |
958 | * reference count. Note, the inode lock is not waited upon so you have to be |
959 | * very careful what you do with the returned inode. You probably should be | |
960 | * using ilookup5() instead. | |
961 | * | |
962 | * Otherwise NULL is returned. | |
963 | * | |
964 | * Note, @test is called with the inode_lock held, so can't sleep. | |
965 | */ | |
966 | struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval, | |
967 | int (*test)(struct inode *, void *), void *data) | |
968 | { | |
969 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
970 | ||
971 | return ifind(sb, head, test, data, 0); | |
972 | } | |
88bd5121 AA |
973 | EXPORT_SYMBOL(ilookup5_nowait); |
974 | ||
975 | /** | |
976 | * ilookup5 - search for an inode in the inode cache | |
977 | * @sb: super block of file system to search | |
978 | * @hashval: hash value (usually inode number) to search for | |
979 | * @test: callback used for comparisons between inodes | |
980 | * @data: opaque data pointer to pass to @test | |
981 | * | |
982 | * ilookup5() uses ifind() to search for the inode specified by @hashval and | |
983 | * @data in the inode cache. This is a generalized version of ilookup() for | |
984 | * file systems where the inode number is not sufficient for unique | |
985 | * identification of an inode. | |
986 | * | |
987 | * If the inode is in the cache, the inode lock is waited upon and the inode is | |
988 | * returned with an incremented reference count. | |
1da177e4 LT |
989 | * |
990 | * Otherwise NULL is returned. | |
991 | * | |
992 | * Note, @test is called with the inode_lock held, so can't sleep. | |
993 | */ | |
994 | struct inode *ilookup5(struct super_block *sb, unsigned long hashval, | |
995 | int (*test)(struct inode *, void *), void *data) | |
996 | { | |
997 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
998 | ||
88bd5121 | 999 | return ifind(sb, head, test, data, 1); |
1da177e4 | 1000 | } |
1da177e4 LT |
1001 | EXPORT_SYMBOL(ilookup5); |
1002 | ||
1003 | /** | |
1004 | * ilookup - search for an inode in the inode cache | |
1005 | * @sb: super block of file system to search | |
1006 | * @ino: inode number to search for | |
1007 | * | |
1008 | * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache. | |
1009 | * This is for file systems where the inode number is sufficient for unique | |
1010 | * identification of an inode. | |
1011 | * | |
1012 | * If the inode is in the cache, the inode is returned with an incremented | |
1013 | * reference count. | |
1014 | * | |
1015 | * Otherwise NULL is returned. | |
1016 | */ | |
1017 | struct inode *ilookup(struct super_block *sb, unsigned long ino) | |
1018 | { | |
1019 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1020 | ||
1021 | return ifind_fast(sb, head, ino); | |
1022 | } | |
1da177e4 LT |
1023 | EXPORT_SYMBOL(ilookup); |
1024 | ||
1025 | /** | |
1026 | * iget5_locked - obtain an inode from a mounted file system | |
1027 | * @sb: super block of file system | |
1028 | * @hashval: hash value (usually inode number) to get | |
1029 | * @test: callback used for comparisons between inodes | |
1030 | * @set: callback used to initialize a new struct inode | |
1031 | * @data: opaque data pointer to pass to @test and @set | |
1032 | * | |
1da177e4 LT |
1033 | * iget5_locked() uses ifind() to search for the inode specified by @hashval |
1034 | * and @data in the inode cache and if present it is returned with an increased | |
1035 | * reference count. This is a generalized version of iget_locked() for file | |
1036 | * systems where the inode number is not sufficient for unique identification | |
1037 | * of an inode. | |
1038 | * | |
1039 | * If the inode is not in cache, get_new_inode() is called to allocate a new | |
1040 | * inode and this is returned locked, hashed, and with the I_NEW flag set. The | |
1041 | * file system gets to fill it in before unlocking it via unlock_new_inode(). | |
1042 | * | |
1043 | * Note both @test and @set are called with the inode_lock held, so can't sleep. | |
1044 | */ | |
1045 | struct inode *iget5_locked(struct super_block *sb, unsigned long hashval, | |
1046 | int (*test)(struct inode *, void *), | |
1047 | int (*set)(struct inode *, void *), void *data) | |
1048 | { | |
1049 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1050 | struct inode *inode; | |
1051 | ||
88bd5121 | 1052 | inode = ifind(sb, head, test, data, 1); |
1da177e4 LT |
1053 | if (inode) |
1054 | return inode; | |
1055 | /* | |
1056 | * get_new_inode() will do the right thing, re-trying the search | |
1057 | * in case it had to block at any point. | |
1058 | */ | |
1059 | return get_new_inode(sb, head, test, set, data); | |
1060 | } | |
1da177e4 LT |
1061 | EXPORT_SYMBOL(iget5_locked); |
1062 | ||
1063 | /** | |
1064 | * iget_locked - obtain an inode from a mounted file system | |
1065 | * @sb: super block of file system | |
1066 | * @ino: inode number to get | |
1067 | * | |
1da177e4 LT |
1068 | * iget_locked() uses ifind_fast() to search for the inode specified by @ino in |
1069 | * the inode cache and if present it is returned with an increased reference | |
1070 | * count. This is for file systems where the inode number is sufficient for | |
1071 | * unique identification of an inode. | |
1072 | * | |
1073 | * If the inode is not in cache, get_new_inode_fast() is called to allocate a | |
1074 | * new inode and this is returned locked, hashed, and with the I_NEW flag set. | |
1075 | * The file system gets to fill it in before unlocking it via | |
1076 | * unlock_new_inode(). | |
1077 | */ | |
1078 | struct inode *iget_locked(struct super_block *sb, unsigned long ino) | |
1079 | { | |
1080 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1081 | struct inode *inode; | |
1082 | ||
1083 | inode = ifind_fast(sb, head, ino); | |
1084 | if (inode) | |
1085 | return inode; | |
1086 | /* | |
1087 | * get_new_inode_fast() will do the right thing, re-trying the search | |
1088 | * in case it had to block at any point. | |
1089 | */ | |
1090 | return get_new_inode_fast(sb, head, ino); | |
1091 | } | |
1da177e4 LT |
1092 | EXPORT_SYMBOL(iget_locked); |
1093 | ||
261bca86 AV |
1094 | int insert_inode_locked(struct inode *inode) |
1095 | { | |
1096 | struct super_block *sb = inode->i_sb; | |
1097 | ino_t ino = inode->i_ino; | |
1098 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
261bca86 | 1099 | |
eaff8079 | 1100 | inode->i_state |= I_NEW; |
261bca86 | 1101 | while (1) { |
72a43d63 AV |
1102 | struct hlist_node *node; |
1103 | struct inode *old = NULL; | |
261bca86 | 1104 | spin_lock(&inode_lock); |
72a43d63 AV |
1105 | hlist_for_each_entry(old, node, head, i_hash) { |
1106 | if (old->i_ino != ino) | |
1107 | continue; | |
1108 | if (old->i_sb != sb) | |
1109 | continue; | |
a4ffdde6 | 1110 | if (old->i_state & (I_FREEING|I_WILL_FREE)) |
72a43d63 AV |
1111 | continue; |
1112 | break; | |
1113 | } | |
1114 | if (likely(!node)) { | |
261bca86 AV |
1115 | hlist_add_head(&inode->i_hash, head); |
1116 | spin_unlock(&inode_lock); | |
1117 | return 0; | |
1118 | } | |
1119 | __iget(old); | |
1120 | spin_unlock(&inode_lock); | |
1121 | wait_on_inode(old); | |
1d3382cb | 1122 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1123 | iput(old); |
1124 | return -EBUSY; | |
1125 | } | |
1126 | iput(old); | |
1127 | } | |
1128 | } | |
261bca86 AV |
1129 | EXPORT_SYMBOL(insert_inode_locked); |
1130 | ||
1131 | int insert_inode_locked4(struct inode *inode, unsigned long hashval, | |
1132 | int (*test)(struct inode *, void *), void *data) | |
1133 | { | |
1134 | struct super_block *sb = inode->i_sb; | |
1135 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
261bca86 | 1136 | |
eaff8079 | 1137 | inode->i_state |= I_NEW; |
261bca86 AV |
1138 | |
1139 | while (1) { | |
72a43d63 AV |
1140 | struct hlist_node *node; |
1141 | struct inode *old = NULL; | |
1142 | ||
261bca86 | 1143 | spin_lock(&inode_lock); |
72a43d63 AV |
1144 | hlist_for_each_entry(old, node, head, i_hash) { |
1145 | if (old->i_sb != sb) | |
1146 | continue; | |
1147 | if (!test(old, data)) | |
1148 | continue; | |
a4ffdde6 | 1149 | if (old->i_state & (I_FREEING|I_WILL_FREE)) |
72a43d63 AV |
1150 | continue; |
1151 | break; | |
1152 | } | |
1153 | if (likely(!node)) { | |
261bca86 AV |
1154 | hlist_add_head(&inode->i_hash, head); |
1155 | spin_unlock(&inode_lock); | |
1156 | return 0; | |
1157 | } | |
1158 | __iget(old); | |
1159 | spin_unlock(&inode_lock); | |
1160 | wait_on_inode(old); | |
1d3382cb | 1161 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1162 | iput(old); |
1163 | return -EBUSY; | |
1164 | } | |
1165 | iput(old); | |
1166 | } | |
1167 | } | |
261bca86 AV |
1168 | EXPORT_SYMBOL(insert_inode_locked4); |
1169 | ||
1da177e4 LT |
1170 | /** |
1171 | * __insert_inode_hash - hash an inode | |
1172 | * @inode: unhashed inode | |
1173 | * @hashval: unsigned long value used to locate this object in the | |
1174 | * inode_hashtable. | |
1175 | * | |
1176 | * Add an inode to the inode hash for this superblock. | |
1177 | */ | |
1178 | void __insert_inode_hash(struct inode *inode, unsigned long hashval) | |
1179 | { | |
1180 | struct hlist_head *head = inode_hashtable + hash(inode->i_sb, hashval); | |
1181 | spin_lock(&inode_lock); | |
1182 | hlist_add_head(&inode->i_hash, head); | |
1183 | spin_unlock(&inode_lock); | |
1184 | } | |
1da177e4 LT |
1185 | EXPORT_SYMBOL(__insert_inode_hash); |
1186 | ||
1187 | /** | |
1188 | * remove_inode_hash - remove an inode from the hash | |
1189 | * @inode: inode to unhash | |
1190 | * | |
1191 | * Remove an inode from the superblock. | |
1192 | */ | |
1193 | void remove_inode_hash(struct inode *inode) | |
1194 | { | |
1195 | spin_lock(&inode_lock); | |
1196 | hlist_del_init(&inode->i_hash); | |
1197 | spin_unlock(&inode_lock); | |
1198 | } | |
1da177e4 LT |
1199 | EXPORT_SYMBOL(remove_inode_hash); |
1200 | ||
45321ac5 AV |
1201 | int generic_delete_inode(struct inode *inode) |
1202 | { | |
1203 | return 1; | |
1204 | } | |
1205 | EXPORT_SYMBOL(generic_delete_inode); | |
1206 | ||
1da177e4 | 1207 | /* |
45321ac5 AV |
1208 | * Normal UNIX filesystem behaviour: delete the |
1209 | * inode when the usage count drops to zero, and | |
1210 | * i_nlink is zero. | |
1da177e4 | 1211 | */ |
45321ac5 | 1212 | int generic_drop_inode(struct inode *inode) |
1da177e4 | 1213 | { |
1d3382cb | 1214 | return !inode->i_nlink || inode_unhashed(inode); |
1da177e4 | 1215 | } |
45321ac5 | 1216 | EXPORT_SYMBOL_GPL(generic_drop_inode); |
1da177e4 | 1217 | |
45321ac5 AV |
1218 | /* |
1219 | * Called when we're dropping the last reference | |
1220 | * to an inode. | |
22fe4042 | 1221 | * |
45321ac5 AV |
1222 | * Call the FS "drop_inode()" function, defaulting to |
1223 | * the legacy UNIX filesystem behaviour. If it tells | |
1224 | * us to evict inode, do so. Otherwise, retain inode | |
1225 | * in cache if fs is alive, sync and evict if fs is | |
1226 | * shutting down. | |
22fe4042 | 1227 | */ |
45321ac5 | 1228 | static void iput_final(struct inode *inode) |
1da177e4 LT |
1229 | { |
1230 | struct super_block *sb = inode->i_sb; | |
45321ac5 AV |
1231 | const struct super_operations *op = inode->i_sb->s_op; |
1232 | int drop; | |
1233 | ||
1234 | if (op && op->drop_inode) | |
1235 | drop = op->drop_inode(inode); | |
1236 | else | |
1237 | drop = generic_drop_inode(inode); | |
1da177e4 | 1238 | |
45321ac5 | 1239 | if (!drop) { |
1c0eeaf5 | 1240 | if (!(inode->i_state & (I_DIRTY|I_SYNC))) |
1da177e4 | 1241 | list_move(&inode->i_list, &inode_unused); |
cffbc8aa | 1242 | percpu_counter_inc(&nr_inodes_unused); |
acb0c854 | 1243 | if (sb->s_flags & MS_ACTIVE) { |
991114c6 | 1244 | spin_unlock(&inode_lock); |
45321ac5 | 1245 | return; |
991114c6 | 1246 | } |
7ef0d737 | 1247 | WARN_ON(inode->i_state & I_NEW); |
991114c6 AV |
1248 | inode->i_state |= I_WILL_FREE; |
1249 | spin_unlock(&inode_lock); | |
1da177e4 LT |
1250 | write_inode_now(inode, 1); |
1251 | spin_lock(&inode_lock); | |
7ef0d737 | 1252 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1253 | inode->i_state &= ~I_WILL_FREE; |
cffbc8aa | 1254 | percpu_counter_dec(&nr_inodes_unused); |
1da177e4 LT |
1255 | hlist_del_init(&inode->i_hash); |
1256 | } | |
1257 | list_del_init(&inode->i_list); | |
1258 | list_del_init(&inode->i_sb_list); | |
7ef0d737 | 1259 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1260 | inode->i_state |= I_FREEING; |
1da177e4 | 1261 | spin_unlock(&inode_lock); |
644da596 | 1262 | evict(inode); |
45321ac5 AV |
1263 | spin_lock(&inode_lock); |
1264 | hlist_del_init(&inode->i_hash); | |
1265 | spin_unlock(&inode_lock); | |
7f04c26d | 1266 | wake_up_inode(inode); |
45321ac5 | 1267 | BUG_ON(inode->i_state != (I_FREEING | I_CLEAR)); |
1da177e4 LT |
1268 | destroy_inode(inode); |
1269 | } | |
1270 | ||
1da177e4 | 1271 | /** |
6b3304b5 | 1272 | * iput - put an inode |
1da177e4 LT |
1273 | * @inode: inode to put |
1274 | * | |
1275 | * Puts an inode, dropping its usage count. If the inode use count hits | |
1276 | * zero, the inode is then freed and may also be destroyed. | |
1277 | * | |
1278 | * Consequently, iput() can sleep. | |
1279 | */ | |
1280 | void iput(struct inode *inode) | |
1281 | { | |
1282 | if (inode) { | |
a4ffdde6 | 1283 | BUG_ON(inode->i_state & I_CLEAR); |
1da177e4 | 1284 | |
1da177e4 LT |
1285 | if (atomic_dec_and_lock(&inode->i_count, &inode_lock)) |
1286 | iput_final(inode); | |
1287 | } | |
1288 | } | |
1da177e4 LT |
1289 | EXPORT_SYMBOL(iput); |
1290 | ||
1291 | /** | |
1292 | * bmap - find a block number in a file | |
1293 | * @inode: inode of file | |
1294 | * @block: block to find | |
1295 | * | |
1296 | * Returns the block number on the device holding the inode that | |
1297 | * is the disk block number for the block of the file requested. | |
1298 | * That is, asked for block 4 of inode 1 the function will return the | |
6b3304b5 | 1299 | * disk block relative to the disk start that holds that block of the |
1da177e4 LT |
1300 | * file. |
1301 | */ | |
6b3304b5 | 1302 | sector_t bmap(struct inode *inode, sector_t block) |
1da177e4 LT |
1303 | { |
1304 | sector_t res = 0; | |
1305 | if (inode->i_mapping->a_ops->bmap) | |
1306 | res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block); | |
1307 | return res; | |
1308 | } | |
1da177e4 LT |
1309 | EXPORT_SYMBOL(bmap); |
1310 | ||
11ff6f05 MG |
1311 | /* |
1312 | * With relative atime, only update atime if the previous atime is | |
1313 | * earlier than either the ctime or mtime or if at least a day has | |
1314 | * passed since the last atime update. | |
1315 | */ | |
1316 | static int relatime_need_update(struct vfsmount *mnt, struct inode *inode, | |
1317 | struct timespec now) | |
1318 | { | |
1319 | ||
1320 | if (!(mnt->mnt_flags & MNT_RELATIME)) | |
1321 | return 1; | |
1322 | /* | |
1323 | * Is mtime younger than atime? If yes, update atime: | |
1324 | */ | |
1325 | if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0) | |
1326 | return 1; | |
1327 | /* | |
1328 | * Is ctime younger than atime? If yes, update atime: | |
1329 | */ | |
1330 | if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0) | |
1331 | return 1; | |
1332 | ||
1333 | /* | |
1334 | * Is the previous atime value older than a day? If yes, | |
1335 | * update atime: | |
1336 | */ | |
1337 | if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60) | |
1338 | return 1; | |
1339 | /* | |
1340 | * Good, we can skip the atime update: | |
1341 | */ | |
1342 | return 0; | |
1343 | } | |
1344 | ||
1da177e4 | 1345 | /** |
869243a0 CH |
1346 | * touch_atime - update the access time |
1347 | * @mnt: mount the inode is accessed on | |
7045f37b | 1348 | * @dentry: dentry accessed |
1da177e4 LT |
1349 | * |
1350 | * Update the accessed time on an inode and mark it for writeback. | |
1351 | * This function automatically handles read only file systems and media, | |
1352 | * as well as the "noatime" flag and inode specific "noatime" markers. | |
1353 | */ | |
869243a0 | 1354 | void touch_atime(struct vfsmount *mnt, struct dentry *dentry) |
1da177e4 | 1355 | { |
869243a0 | 1356 | struct inode *inode = dentry->d_inode; |
1da177e4 LT |
1357 | struct timespec now; |
1358 | ||
cdb70f3f | 1359 | if (inode->i_flags & S_NOATIME) |
b12536c2 | 1360 | return; |
37756ced | 1361 | if (IS_NOATIME(inode)) |
b12536c2 | 1362 | return; |
b2276138 | 1363 | if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1364 | return; |
47ae32d6 | 1365 | |
cdb70f3f | 1366 | if (mnt->mnt_flags & MNT_NOATIME) |
b12536c2 | 1367 | return; |
cdb70f3f | 1368 | if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1369 | return; |
1da177e4 LT |
1370 | |
1371 | now = current_fs_time(inode->i_sb); | |
11ff6f05 MG |
1372 | |
1373 | if (!relatime_need_update(mnt, inode, now)) | |
b12536c2 | 1374 | return; |
11ff6f05 | 1375 | |
47ae32d6 | 1376 | if (timespec_equal(&inode->i_atime, &now)) |
b12536c2 AK |
1377 | return; |
1378 | ||
1379 | if (mnt_want_write(mnt)) | |
1380 | return; | |
47ae32d6 VH |
1381 | |
1382 | inode->i_atime = now; | |
1383 | mark_inode_dirty_sync(inode); | |
cdb70f3f | 1384 | mnt_drop_write(mnt); |
1da177e4 | 1385 | } |
869243a0 | 1386 | EXPORT_SYMBOL(touch_atime); |
1da177e4 LT |
1387 | |
1388 | /** | |
870f4817 CH |
1389 | * file_update_time - update mtime and ctime time |
1390 | * @file: file accessed | |
1da177e4 | 1391 | * |
870f4817 CH |
1392 | * Update the mtime and ctime members of an inode and mark the inode |
1393 | * for writeback. Note that this function is meant exclusively for | |
1394 | * usage in the file write path of filesystems, and filesystems may | |
1395 | * choose to explicitly ignore update via this function with the | |
2eadfc0e | 1396 | * S_NOCMTIME inode flag, e.g. for network filesystem where these |
870f4817 | 1397 | * timestamps are handled by the server. |
1da177e4 LT |
1398 | */ |
1399 | ||
870f4817 | 1400 | void file_update_time(struct file *file) |
1da177e4 | 1401 | { |
0f7fc9e4 | 1402 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 | 1403 | struct timespec now; |
ce06e0b2 | 1404 | enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0; |
1da177e4 | 1405 | |
ce06e0b2 | 1406 | /* First try to exhaust all avenues to not sync */ |
1da177e4 LT |
1407 | if (IS_NOCMTIME(inode)) |
1408 | return; | |
20ddee2c | 1409 | |
1da177e4 | 1410 | now = current_fs_time(inode->i_sb); |
ce06e0b2 AK |
1411 | if (!timespec_equal(&inode->i_mtime, &now)) |
1412 | sync_it = S_MTIME; | |
1da177e4 | 1413 | |
ce06e0b2 AK |
1414 | if (!timespec_equal(&inode->i_ctime, &now)) |
1415 | sync_it |= S_CTIME; | |
870f4817 | 1416 | |
ce06e0b2 AK |
1417 | if (IS_I_VERSION(inode)) |
1418 | sync_it |= S_VERSION; | |
7a224228 | 1419 | |
ce06e0b2 AK |
1420 | if (!sync_it) |
1421 | return; | |
1422 | ||
1423 | /* Finally allowed to write? Takes lock. */ | |
1424 | if (mnt_want_write_file(file)) | |
1425 | return; | |
1426 | ||
1427 | /* Only change inode inside the lock region */ | |
1428 | if (sync_it & S_VERSION) | |
1429 | inode_inc_iversion(inode); | |
1430 | if (sync_it & S_CTIME) | |
1431 | inode->i_ctime = now; | |
1432 | if (sync_it & S_MTIME) | |
1433 | inode->i_mtime = now; | |
1434 | mark_inode_dirty_sync(inode); | |
20ddee2c | 1435 | mnt_drop_write(file->f_path.mnt); |
1da177e4 | 1436 | } |
870f4817 | 1437 | EXPORT_SYMBOL(file_update_time); |
1da177e4 LT |
1438 | |
1439 | int inode_needs_sync(struct inode *inode) | |
1440 | { | |
1441 | if (IS_SYNC(inode)) | |
1442 | return 1; | |
1443 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) | |
1444 | return 1; | |
1445 | return 0; | |
1446 | } | |
1da177e4 LT |
1447 | EXPORT_SYMBOL(inode_needs_sync); |
1448 | ||
1da177e4 LT |
1449 | int inode_wait(void *word) |
1450 | { | |
1451 | schedule(); | |
1452 | return 0; | |
1453 | } | |
d44dab8d | 1454 | EXPORT_SYMBOL(inode_wait); |
1da177e4 LT |
1455 | |
1456 | /* | |
168a9fd6 MS |
1457 | * If we try to find an inode in the inode hash while it is being |
1458 | * deleted, we have to wait until the filesystem completes its | |
1459 | * deletion before reporting that it isn't found. This function waits | |
1460 | * until the deletion _might_ have completed. Callers are responsible | |
1461 | * to recheck inode state. | |
1462 | * | |
eaff8079 | 1463 | * It doesn't matter if I_NEW is not set initially, a call to |
168a9fd6 MS |
1464 | * wake_up_inode() after removing from the hash list will DTRT. |
1465 | * | |
1da177e4 LT |
1466 | * This is called with inode_lock held. |
1467 | */ | |
1468 | static void __wait_on_freeing_inode(struct inode *inode) | |
1469 | { | |
1470 | wait_queue_head_t *wq; | |
eaff8079 CH |
1471 | DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW); |
1472 | wq = bit_waitqueue(&inode->i_state, __I_NEW); | |
1da177e4 LT |
1473 | prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE); |
1474 | spin_unlock(&inode_lock); | |
1475 | schedule(); | |
1476 | finish_wait(wq, &wait.wait); | |
1477 | spin_lock(&inode_lock); | |
1478 | } | |
1479 | ||
1da177e4 LT |
1480 | static __initdata unsigned long ihash_entries; |
1481 | static int __init set_ihash_entries(char *str) | |
1482 | { | |
1483 | if (!str) | |
1484 | return 0; | |
1485 | ihash_entries = simple_strtoul(str, &str, 0); | |
1486 | return 1; | |
1487 | } | |
1488 | __setup("ihash_entries=", set_ihash_entries); | |
1489 | ||
1490 | /* | |
1491 | * Initialize the waitqueues and inode hash table. | |
1492 | */ | |
1493 | void __init inode_init_early(void) | |
1494 | { | |
1495 | int loop; | |
1496 | ||
1497 | /* If hashes are distributed across NUMA nodes, defer | |
1498 | * hash allocation until vmalloc space is available. | |
1499 | */ | |
1500 | if (hashdist) | |
1501 | return; | |
1502 | ||
1503 | inode_hashtable = | |
1504 | alloc_large_system_hash("Inode-cache", | |
1505 | sizeof(struct hlist_head), | |
1506 | ihash_entries, | |
1507 | 14, | |
1508 | HASH_EARLY, | |
1509 | &i_hash_shift, | |
1510 | &i_hash_mask, | |
1511 | 0); | |
1512 | ||
1513 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1514 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1515 | } | |
1516 | ||
74bf17cf | 1517 | void __init inode_init(void) |
1da177e4 LT |
1518 | { |
1519 | int loop; | |
1520 | ||
1521 | /* inode slab cache */ | |
b0196009 PJ |
1522 | inode_cachep = kmem_cache_create("inode_cache", |
1523 | sizeof(struct inode), | |
1524 | 0, | |
1525 | (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC| | |
1526 | SLAB_MEM_SPREAD), | |
20c2df83 | 1527 | init_once); |
8e1f936b | 1528 | register_shrinker(&icache_shrinker); |
cffbc8aa DC |
1529 | percpu_counter_init(&nr_inodes, 0); |
1530 | percpu_counter_init(&nr_inodes_unused, 0); | |
1da177e4 LT |
1531 | |
1532 | /* Hash may have been set up in inode_init_early */ | |
1533 | if (!hashdist) | |
1534 | return; | |
1535 | ||
1536 | inode_hashtable = | |
1537 | alloc_large_system_hash("Inode-cache", | |
1538 | sizeof(struct hlist_head), | |
1539 | ihash_entries, | |
1540 | 14, | |
1541 | 0, | |
1542 | &i_hash_shift, | |
1543 | &i_hash_mask, | |
1544 | 0); | |
1545 | ||
1546 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1547 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1548 | } | |
1549 | ||
1550 | void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev) | |
1551 | { | |
1552 | inode->i_mode = mode; | |
1553 | if (S_ISCHR(mode)) { | |
1554 | inode->i_fop = &def_chr_fops; | |
1555 | inode->i_rdev = rdev; | |
1556 | } else if (S_ISBLK(mode)) { | |
1557 | inode->i_fop = &def_blk_fops; | |
1558 | inode->i_rdev = rdev; | |
1559 | } else if (S_ISFIFO(mode)) | |
1560 | inode->i_fop = &def_fifo_fops; | |
1561 | else if (S_ISSOCK(mode)) | |
1562 | inode->i_fop = &bad_sock_fops; | |
1563 | else | |
af0d9ae8 MK |
1564 | printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for" |
1565 | " inode %s:%lu\n", mode, inode->i_sb->s_id, | |
1566 | inode->i_ino); | |
1da177e4 LT |
1567 | } |
1568 | EXPORT_SYMBOL(init_special_inode); | |
a1bd120d DM |
1569 | |
1570 | /** | |
1571 | * Init uid,gid,mode for new inode according to posix standards | |
1572 | * @inode: New inode | |
1573 | * @dir: Directory inode | |
1574 | * @mode: mode of the new inode | |
1575 | */ | |
1576 | void inode_init_owner(struct inode *inode, const struct inode *dir, | |
1577 | mode_t mode) | |
1578 | { | |
1579 | inode->i_uid = current_fsuid(); | |
1580 | if (dir && dir->i_mode & S_ISGID) { | |
1581 | inode->i_gid = dir->i_gid; | |
1582 | if (S_ISDIR(mode)) | |
1583 | mode |= S_ISGID; | |
1584 | } else | |
1585 | inode->i_gid = current_fsgid(); | |
1586 | inode->i_mode = mode; | |
1587 | } | |
1588 | EXPORT_SYMBOL(inode_init_owner); |