]>
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> |
a178d202 | 27 | #include <linux/ima.h> |
e795b717 | 28 | #include <linux/cred.h> |
1da177e4 LT |
29 | |
30 | /* | |
31 | * This is needed for the following functions: | |
32 | * - inode_has_buffers | |
1da177e4 LT |
33 | * - invalidate_bdev |
34 | * | |
35 | * FIXME: remove all knowledge of the buffer layer from this file | |
36 | */ | |
37 | #include <linux/buffer_head.h> | |
38 | ||
39 | /* | |
40 | * New inode.c implementation. | |
41 | * | |
42 | * This implementation has the basic premise of trying | |
43 | * to be extremely low-overhead and SMP-safe, yet be | |
44 | * simple enough to be "obviously correct". | |
45 | * | |
46 | * Famous last words. | |
47 | */ | |
48 | ||
49 | /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */ | |
50 | ||
51 | /* #define INODE_PARANOIA 1 */ | |
52 | /* #define INODE_DEBUG 1 */ | |
53 | ||
54 | /* | |
55 | * Inode lookup is no longer as critical as it used to be: | |
56 | * most of the lookups are going to be through the dcache. | |
57 | */ | |
58 | #define I_HASHBITS i_hash_shift | |
59 | #define I_HASHMASK i_hash_mask | |
60 | ||
fa3536cc ED |
61 | static unsigned int i_hash_mask __read_mostly; |
62 | static unsigned int i_hash_shift __read_mostly; | |
1da177e4 LT |
63 | |
64 | /* | |
65 | * Each inode can be on two separate lists. One is | |
66 | * the hash list of the inode, used for lookups. The | |
67 | * other linked list is the "type" list: | |
68 | * "in_use" - valid inode, i_count > 0, i_nlink > 0 | |
69 | * "dirty" - as "in_use" but also dirty | |
70 | * "unused" - valid inode, i_count = 0 | |
71 | * | |
72 | * A "dirty" list is maintained for each super block, | |
73 | * allowing for low-overhead inode sync() operations. | |
74 | */ | |
75 | ||
7ccf19a8 | 76 | static LIST_HEAD(inode_lru); |
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 | /* | |
bab1d944 CH |
88 | * iprune_sem provides exclusion between the icache shrinking and the |
89 | * umount path. | |
88e0fbc4 | 90 | * |
bab1d944 CH |
91 | * We don't actually need it to protect anything in the umount path, |
92 | * but only need to cycle through it to make sure any inode that | |
93 | * prune_icache took off the LRU list has been fully torn down by the | |
94 | * time we are past evict_inodes. | |
1da177e4 | 95 | */ |
88e0fbc4 | 96 | static DECLARE_RWSEM(iprune_sem); |
1da177e4 LT |
97 | |
98 | /* | |
99 | * Statistics gathering.. | |
100 | */ | |
101 | struct inodes_stat_t inodes_stat; | |
102 | ||
3e880fb5 | 103 | static DEFINE_PER_CPU(unsigned int, nr_inodes); |
cffbc8aa | 104 | |
6b3304b5 | 105 | static struct kmem_cache *inode_cachep __read_mostly; |
1da177e4 | 106 | |
3e880fb5 | 107 | static int get_nr_inodes(void) |
cffbc8aa | 108 | { |
3e880fb5 NP |
109 | int i; |
110 | int sum = 0; | |
111 | for_each_possible_cpu(i) | |
112 | sum += per_cpu(nr_inodes, i); | |
113 | return sum < 0 ? 0 : sum; | |
cffbc8aa DC |
114 | } |
115 | ||
116 | static inline int get_nr_inodes_unused(void) | |
117 | { | |
86c8749e | 118 | return inodes_stat.nr_unused; |
cffbc8aa DC |
119 | } |
120 | ||
121 | int get_nr_dirty_inodes(void) | |
122 | { | |
3e880fb5 | 123 | /* not actually dirty inodes, but a wild approximation */ |
cffbc8aa DC |
124 | int nr_dirty = get_nr_inodes() - get_nr_inodes_unused(); |
125 | return nr_dirty > 0 ? nr_dirty : 0; | |
cffbc8aa DC |
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(); | |
cffbc8aa DC |
136 | return proc_dointvec(table, write, buffer, lenp, ppos); |
137 | } | |
138 | #endif | |
139 | ||
1c0eeaf5 JE |
140 | static void wake_up_inode(struct inode *inode) |
141 | { | |
142 | /* | |
143 | * Prevent speculative execution through spin_unlock(&inode_lock); | |
144 | */ | |
145 | smp_mb(); | |
eaff8079 | 146 | wake_up_bit(&inode->i_state, __I_NEW); |
1c0eeaf5 JE |
147 | } |
148 | ||
2cb1599f DC |
149 | /** |
150 | * inode_init_always - perform inode structure intialisation | |
0bc02f3f RD |
151 | * @sb: superblock inode belongs to |
152 | * @inode: inode to initialise | |
2cb1599f DC |
153 | * |
154 | * These are initializations that need to be done on every inode | |
155 | * allocation as the fields are not initialised by slab allocation. | |
156 | */ | |
54e34621 | 157 | int inode_init_always(struct super_block *sb, struct inode *inode) |
1da177e4 | 158 | { |
f5e54d6e | 159 | static const struct address_space_operations empty_aops; |
6e1d5dcc | 160 | static const struct inode_operations empty_iops; |
99ac48f5 | 161 | static const struct file_operations empty_fops; |
6b3304b5 | 162 | struct address_space *const mapping = &inode->i_data; |
2cb1599f DC |
163 | |
164 | inode->i_sb = sb; | |
165 | inode->i_blkbits = sb->s_blocksize_bits; | |
166 | inode->i_flags = 0; | |
167 | atomic_set(&inode->i_count, 1); | |
168 | inode->i_op = &empty_iops; | |
169 | inode->i_fop = &empty_fops; | |
170 | inode->i_nlink = 1; | |
56ff5efa AV |
171 | inode->i_uid = 0; |
172 | inode->i_gid = 0; | |
2cb1599f DC |
173 | atomic_set(&inode->i_writecount, 0); |
174 | inode->i_size = 0; | |
175 | inode->i_blocks = 0; | |
176 | inode->i_bytes = 0; | |
177 | inode->i_generation = 0; | |
1da177e4 | 178 | #ifdef CONFIG_QUOTA |
2cb1599f | 179 | memset(&inode->i_dquot, 0, sizeof(inode->i_dquot)); |
1da177e4 | 180 | #endif |
2cb1599f DC |
181 | inode->i_pipe = NULL; |
182 | inode->i_bdev = NULL; | |
183 | inode->i_cdev = NULL; | |
184 | inode->i_rdev = 0; | |
185 | inode->dirtied_when = 0; | |
6146f0d5 MZ |
186 | |
187 | if (security_inode_alloc(inode)) | |
54e34621 | 188 | goto out; |
2cb1599f DC |
189 | spin_lock_init(&inode->i_lock); |
190 | lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key); | |
191 | ||
192 | mutex_init(&inode->i_mutex); | |
193 | lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key); | |
194 | ||
195 | init_rwsem(&inode->i_alloc_sem); | |
196 | lockdep_set_class(&inode->i_alloc_sem, &sb->s_type->i_alloc_sem_key); | |
197 | ||
198 | mapping->a_ops = &empty_aops; | |
199 | mapping->host = inode; | |
200 | mapping->flags = 0; | |
3c1d4378 | 201 | mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE); |
2cb1599f DC |
202 | mapping->assoc_mapping = NULL; |
203 | mapping->backing_dev_info = &default_backing_dev_info; | |
204 | mapping->writeback_index = 0; | |
205 | ||
206 | /* | |
207 | * If the block_device provides a backing_dev_info for client | |
208 | * inodes then use that. Otherwise the inode share the bdev's | |
209 | * backing_dev_info. | |
210 | */ | |
211 | if (sb->s_bdev) { | |
212 | struct backing_dev_info *bdi; | |
213 | ||
2c96ce9f | 214 | bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info; |
2cb1599f DC |
215 | mapping->backing_dev_info = bdi; |
216 | } | |
217 | inode->i_private = NULL; | |
218 | inode->i_mapping = mapping; | |
f19d4a8f AV |
219 | #ifdef CONFIG_FS_POSIX_ACL |
220 | inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED; | |
221 | #endif | |
2cb1599f | 222 | |
3be25f49 EP |
223 | #ifdef CONFIG_FSNOTIFY |
224 | inode->i_fsnotify_mask = 0; | |
225 | #endif | |
226 | ||
3e880fb5 | 227 | this_cpu_inc(nr_inodes); |
cffbc8aa | 228 | |
54e34621 | 229 | return 0; |
54e34621 CH |
230 | out: |
231 | return -ENOMEM; | |
1da177e4 | 232 | } |
2cb1599f DC |
233 | EXPORT_SYMBOL(inode_init_always); |
234 | ||
235 | static struct inode *alloc_inode(struct super_block *sb) | |
236 | { | |
237 | struct inode *inode; | |
238 | ||
239 | if (sb->s_op->alloc_inode) | |
240 | inode = sb->s_op->alloc_inode(sb); | |
241 | else | |
242 | inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL); | |
243 | ||
54e34621 CH |
244 | if (!inode) |
245 | return NULL; | |
246 | ||
247 | if (unlikely(inode_init_always(sb, inode))) { | |
248 | if (inode->i_sb->s_op->destroy_inode) | |
249 | inode->i_sb->s_op->destroy_inode(inode); | |
250 | else | |
251 | kmem_cache_free(inode_cachep, inode); | |
252 | return NULL; | |
253 | } | |
254 | ||
255 | return inode; | |
2cb1599f | 256 | } |
1da177e4 | 257 | |
ff0c7d15 NP |
258 | void free_inode_nonrcu(struct inode *inode) |
259 | { | |
260 | kmem_cache_free(inode_cachep, inode); | |
261 | } | |
262 | EXPORT_SYMBOL(free_inode_nonrcu); | |
263 | ||
2e00c97e | 264 | void __destroy_inode(struct inode *inode) |
1da177e4 | 265 | { |
b7542f8c | 266 | BUG_ON(inode_has_buffers(inode)); |
1da177e4 | 267 | security_inode_free(inode); |
3be25f49 | 268 | fsnotify_inode_delete(inode); |
f19d4a8f AV |
269 | #ifdef CONFIG_FS_POSIX_ACL |
270 | if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED) | |
271 | posix_acl_release(inode->i_acl); | |
272 | if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED) | |
273 | posix_acl_release(inode->i_default_acl); | |
274 | #endif | |
3e880fb5 | 275 | this_cpu_dec(nr_inodes); |
2e00c97e CH |
276 | } |
277 | EXPORT_SYMBOL(__destroy_inode); | |
278 | ||
fa0d7e3d NP |
279 | static void i_callback(struct rcu_head *head) |
280 | { | |
281 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
282 | INIT_LIST_HEAD(&inode->i_dentry); | |
283 | kmem_cache_free(inode_cachep, inode); | |
284 | } | |
285 | ||
56b0dacf | 286 | static void destroy_inode(struct inode *inode) |
2e00c97e | 287 | { |
7ccf19a8 | 288 | BUG_ON(!list_empty(&inode->i_lru)); |
2e00c97e | 289 | __destroy_inode(inode); |
1da177e4 LT |
290 | if (inode->i_sb->s_op->destroy_inode) |
291 | inode->i_sb->s_op->destroy_inode(inode); | |
292 | else | |
fa0d7e3d | 293 | call_rcu(&inode->i_rcu, i_callback); |
1da177e4 | 294 | } |
1da177e4 | 295 | |
2aa15890 MS |
296 | void address_space_init_once(struct address_space *mapping) |
297 | { | |
298 | memset(mapping, 0, sizeof(*mapping)); | |
299 | INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC); | |
300 | spin_lock_init(&mapping->tree_lock); | |
301 | spin_lock_init(&mapping->i_mmap_lock); | |
302 | INIT_LIST_HEAD(&mapping->private_list); | |
303 | spin_lock_init(&mapping->private_lock); | |
304 | INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap); | |
305 | INIT_LIST_HEAD(&mapping->i_mmap_nonlinear); | |
306 | mutex_init(&mapping->unmap_mutex); | |
307 | } | |
308 | EXPORT_SYMBOL(address_space_init_once); | |
309 | ||
1da177e4 LT |
310 | /* |
311 | * These are initializations that only need to be done | |
312 | * once, because the fields are idempotent across use | |
313 | * of the inode, so let the slab aware of that. | |
314 | */ | |
315 | void inode_init_once(struct inode *inode) | |
316 | { | |
317 | memset(inode, 0, sizeof(*inode)); | |
318 | INIT_HLIST_NODE(&inode->i_hash); | |
319 | INIT_LIST_HEAD(&inode->i_dentry); | |
320 | INIT_LIST_HEAD(&inode->i_devices); | |
7ccf19a8 NP |
321 | INIT_LIST_HEAD(&inode->i_wb_list); |
322 | INIT_LIST_HEAD(&inode->i_lru); | |
2aa15890 | 323 | address_space_init_once(&inode->i_data); |
1da177e4 | 324 | i_size_ordered_init(inode); |
3be25f49 | 325 | #ifdef CONFIG_FSNOTIFY |
e61ce867 | 326 | INIT_HLIST_HEAD(&inode->i_fsnotify_marks); |
3be25f49 | 327 | #endif |
1da177e4 | 328 | } |
1da177e4 LT |
329 | EXPORT_SYMBOL(inode_init_once); |
330 | ||
51cc5068 | 331 | static void init_once(void *foo) |
1da177e4 | 332 | { |
6b3304b5 | 333 | struct inode *inode = (struct inode *) foo; |
1da177e4 | 334 | |
a35afb83 | 335 | inode_init_once(inode); |
1da177e4 LT |
336 | } |
337 | ||
338 | /* | |
339 | * inode_lock must be held | |
340 | */ | |
6b3304b5 | 341 | void __iget(struct inode *inode) |
1da177e4 | 342 | { |
9e38d86f NP |
343 | atomic_inc(&inode->i_count); |
344 | } | |
2e147f1e | 345 | |
7de9c6ee AV |
346 | /* |
347 | * get additional reference to inode; caller must already hold one. | |
348 | */ | |
349 | void ihold(struct inode *inode) | |
350 | { | |
351 | WARN_ON(atomic_inc_return(&inode->i_count) < 2); | |
352 | } | |
353 | EXPORT_SYMBOL(ihold); | |
354 | ||
9e38d86f NP |
355 | static void inode_lru_list_add(struct inode *inode) |
356 | { | |
7ccf19a8 NP |
357 | if (list_empty(&inode->i_lru)) { |
358 | list_add(&inode->i_lru, &inode_lru); | |
86c8749e | 359 | inodes_stat.nr_unused++; |
9e38d86f NP |
360 | } |
361 | } | |
2e147f1e | 362 | |
9e38d86f NP |
363 | static void inode_lru_list_del(struct inode *inode) |
364 | { | |
7ccf19a8 NP |
365 | if (!list_empty(&inode->i_lru)) { |
366 | list_del_init(&inode->i_lru); | |
86c8749e | 367 | inodes_stat.nr_unused--; |
9e38d86f | 368 | } |
1da177e4 LT |
369 | } |
370 | ||
646ec461 CH |
371 | static inline void __inode_sb_list_add(struct inode *inode) |
372 | { | |
373 | list_add(&inode->i_sb_list, &inode->i_sb->s_inodes); | |
1da177e4 LT |
374 | } |
375 | ||
646ec461 CH |
376 | /** |
377 | * inode_sb_list_add - add inode to the superblock list of inodes | |
378 | * @inode: inode to add | |
379 | */ | |
380 | void inode_sb_list_add(struct inode *inode) | |
381 | { | |
382 | spin_lock(&inode_lock); | |
383 | __inode_sb_list_add(inode); | |
384 | spin_unlock(&inode_lock); | |
385 | } | |
386 | EXPORT_SYMBOL_GPL(inode_sb_list_add); | |
387 | ||
388 | static inline void __inode_sb_list_del(struct inode *inode) | |
389 | { | |
390 | list_del_init(&inode->i_sb_list); | |
391 | } | |
392 | ||
4c51acbc DC |
393 | static unsigned long hash(struct super_block *sb, unsigned long hashval) |
394 | { | |
395 | unsigned long tmp; | |
396 | ||
397 | tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) / | |
398 | L1_CACHE_BYTES; | |
399 | tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> I_HASHBITS); | |
400 | return tmp & I_HASHMASK; | |
401 | } | |
402 | ||
403 | /** | |
404 | * __insert_inode_hash - hash an inode | |
405 | * @inode: unhashed inode | |
406 | * @hashval: unsigned long value used to locate this object in the | |
407 | * inode_hashtable. | |
408 | * | |
409 | * Add an inode to the inode hash for this superblock. | |
410 | */ | |
411 | void __insert_inode_hash(struct inode *inode, unsigned long hashval) | |
412 | { | |
646ec461 CH |
413 | struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval); |
414 | ||
4c51acbc | 415 | spin_lock(&inode_lock); |
646ec461 | 416 | hlist_add_head(&inode->i_hash, b); |
4c51acbc DC |
417 | spin_unlock(&inode_lock); |
418 | } | |
419 | EXPORT_SYMBOL(__insert_inode_hash); | |
420 | ||
421 | /** | |
422 | * __remove_inode_hash - remove an inode from the hash | |
423 | * @inode: inode to unhash | |
424 | * | |
425 | * Remove an inode from the superblock. | |
426 | */ | |
427 | static void __remove_inode_hash(struct inode *inode) | |
428 | { | |
429 | hlist_del_init(&inode->i_hash); | |
430 | } | |
431 | ||
432 | /** | |
433 | * remove_inode_hash - remove an inode from the hash | |
434 | * @inode: inode to unhash | |
435 | * | |
436 | * Remove an inode from the superblock. | |
437 | */ | |
438 | void remove_inode_hash(struct inode *inode) | |
439 | { | |
440 | spin_lock(&inode_lock); | |
441 | hlist_del_init(&inode->i_hash); | |
442 | spin_unlock(&inode_lock); | |
443 | } | |
444 | EXPORT_SYMBOL(remove_inode_hash); | |
445 | ||
b0683aa6 AV |
446 | void end_writeback(struct inode *inode) |
447 | { | |
448 | might_sleep(); | |
449 | BUG_ON(inode->i_data.nrpages); | |
450 | BUG_ON(!list_empty(&inode->i_data.private_list)); | |
451 | BUG_ON(!(inode->i_state & I_FREEING)); | |
452 | BUG_ON(inode->i_state & I_CLEAR); | |
453 | inode_sync_wait(inode); | |
fa0d7e3d | 454 | /* don't need i_lock here, no concurrent mods to i_state */ |
b0683aa6 AV |
455 | inode->i_state = I_FREEING | I_CLEAR; |
456 | } | |
457 | EXPORT_SYMBOL(end_writeback); | |
458 | ||
644da596 | 459 | static void evict(struct inode *inode) |
b4272d4c AV |
460 | { |
461 | const struct super_operations *op = inode->i_sb->s_op; | |
462 | ||
be7ce416 AV |
463 | if (op->evict_inode) { |
464 | op->evict_inode(inode); | |
b4272d4c AV |
465 | } else { |
466 | if (inode->i_data.nrpages) | |
467 | truncate_inode_pages(&inode->i_data, 0); | |
30140837 | 468 | end_writeback(inode); |
b4272d4c | 469 | } |
661074e9 AV |
470 | if (S_ISBLK(inode->i_mode) && inode->i_bdev) |
471 | bd_forget(inode); | |
472 | if (S_ISCHR(inode->i_mode) && inode->i_cdev) | |
473 | cd_forget(inode); | |
b4272d4c AV |
474 | } |
475 | ||
1da177e4 LT |
476 | /* |
477 | * dispose_list - dispose of the contents of a local list | |
478 | * @head: the head of the list to free | |
479 | * | |
480 | * Dispose-list gets a local list with local inodes in it, so it doesn't | |
481 | * need to worry about list corruption and SMP locks. | |
482 | */ | |
483 | static void dispose_list(struct list_head *head) | |
484 | { | |
1da177e4 LT |
485 | while (!list_empty(head)) { |
486 | struct inode *inode; | |
487 | ||
7ccf19a8 NP |
488 | inode = list_first_entry(head, struct inode, i_lru); |
489 | list_del_init(&inode->i_lru); | |
1da177e4 | 490 | |
644da596 | 491 | evict(inode); |
4120db47 AB |
492 | |
493 | spin_lock(&inode_lock); | |
4c51acbc | 494 | __remove_inode_hash(inode); |
646ec461 | 495 | __inode_sb_list_del(inode); |
4120db47 AB |
496 | spin_unlock(&inode_lock); |
497 | ||
498 | wake_up_inode(inode); | |
1da177e4 | 499 | destroy_inode(inode); |
1da177e4 | 500 | } |
1da177e4 LT |
501 | } |
502 | ||
63997e98 AV |
503 | /** |
504 | * evict_inodes - evict all evictable inodes for a superblock | |
505 | * @sb: superblock to operate on | |
506 | * | |
507 | * Make sure that no inodes with zero refcount are retained. This is | |
508 | * called by superblock shutdown after having MS_ACTIVE flag removed, | |
509 | * so any inode reaching zero refcount during or after that call will | |
510 | * be immediately evicted. | |
1da177e4 | 511 | */ |
63997e98 | 512 | void evict_inodes(struct super_block *sb) |
1da177e4 | 513 | { |
63997e98 AV |
514 | struct inode *inode, *next; |
515 | LIST_HEAD(dispose); | |
1da177e4 | 516 | |
63997e98 AV |
517 | spin_lock(&inode_lock); |
518 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { | |
519 | if (atomic_read(&inode->i_count)) | |
aabb8fdb | 520 | continue; |
bab1d944 | 521 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) |
1da177e4 | 522 | continue; |
63997e98 AV |
523 | |
524 | inode->i_state |= I_FREEING; | |
525 | ||
526 | /* | |
527 | * Move the inode off the IO lists and LRU once I_FREEING is | |
528 | * set so that it won't get moved back on there if it is dirty. | |
529 | */ | |
530 | list_move(&inode->i_lru, &dispose); | |
531 | list_del_init(&inode->i_wb_list); | |
532 | if (!(inode->i_state & (I_DIRTY | I_SYNC))) | |
86c8749e | 533 | inodes_stat.nr_unused--; |
1da177e4 | 534 | } |
63997e98 AV |
535 | spin_unlock(&inode_lock); |
536 | ||
537 | dispose_list(&dispose); | |
bab1d944 CH |
538 | |
539 | /* | |
540 | * Cycle through iprune_sem to make sure any inode that prune_icache | |
541 | * moved off the list before we took the lock has been fully torn | |
542 | * down. | |
543 | */ | |
544 | down_write(&iprune_sem); | |
63997e98 | 545 | up_write(&iprune_sem); |
1da177e4 LT |
546 | } |
547 | ||
1da177e4 | 548 | /** |
a0318786 CH |
549 | * invalidate_inodes - attempt to free all inodes on a superblock |
550 | * @sb: superblock to operate on | |
93b270f7 | 551 | * @kill_dirty: flag to guide handling of dirty inodes |
1da177e4 | 552 | * |
a0318786 CH |
553 | * Attempts to free all inodes for a given superblock. If there were any |
554 | * busy inodes return a non-zero value, else zero. | |
93b270f7 N |
555 | * If @kill_dirty is set, discard dirty inodes too, otherwise treat |
556 | * them as busy. | |
1da177e4 | 557 | */ |
93b270f7 | 558 | int invalidate_inodes(struct super_block *sb, bool kill_dirty) |
1da177e4 | 559 | { |
cffbc8aa | 560 | int busy = 0; |
a0318786 CH |
561 | struct inode *inode, *next; |
562 | LIST_HEAD(dispose); | |
1da177e4 | 563 | |
1da177e4 | 564 | spin_lock(&inode_lock); |
a0318786 | 565 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
63997e98 | 566 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) |
aabb8fdb | 567 | continue; |
93b270f7 N |
568 | if (inode->i_state & I_DIRTY && !kill_dirty) { |
569 | busy = 1; | |
570 | continue; | |
571 | } | |
99a38919 CH |
572 | if (atomic_read(&inode->i_count)) { |
573 | busy = 1; | |
1da177e4 LT |
574 | continue; |
575 | } | |
99a38919 | 576 | |
99a38919 | 577 | inode->i_state |= I_FREEING; |
7ccf19a8 NP |
578 | |
579 | /* | |
580 | * Move the inode off the IO lists and LRU once I_FREEING is | |
581 | * set so that it won't get moved back on there if it is dirty. | |
582 | */ | |
a0318786 | 583 | list_move(&inode->i_lru, &dispose); |
7ccf19a8 | 584 | list_del_init(&inode->i_wb_list); |
99a38919 | 585 | if (!(inode->i_state & (I_DIRTY | I_SYNC))) |
86c8749e | 586 | inodes_stat.nr_unused--; |
1da177e4 | 587 | } |
1da177e4 LT |
588 | spin_unlock(&inode_lock); |
589 | ||
a0318786 | 590 | dispose_list(&dispose); |
1da177e4 LT |
591 | |
592 | return busy; | |
593 | } | |
1da177e4 LT |
594 | |
595 | static int can_unuse(struct inode *inode) | |
596 | { | |
9e38d86f | 597 | if (inode->i_state & ~I_REFERENCED) |
1da177e4 LT |
598 | return 0; |
599 | if (inode_has_buffers(inode)) | |
600 | return 0; | |
601 | if (atomic_read(&inode->i_count)) | |
602 | return 0; | |
603 | if (inode->i_data.nrpages) | |
604 | return 0; | |
605 | return 1; | |
606 | } | |
607 | ||
608 | /* | |
9e38d86f NP |
609 | * Scan `goal' inodes on the unused list for freeable ones. They are moved to a |
610 | * temporary list and then are freed outside inode_lock by dispose_list(). | |
1da177e4 LT |
611 | * |
612 | * Any inodes which are pinned purely because of attached pagecache have their | |
9e38d86f NP |
613 | * pagecache removed. If the inode has metadata buffers attached to |
614 | * mapping->private_list then try to remove them. | |
1da177e4 | 615 | * |
9e38d86f NP |
616 | * If the inode has the I_REFERENCED flag set, then it means that it has been |
617 | * used recently - the flag is set in iput_final(). When we encounter such an | |
618 | * inode, clear the flag and move it to the back of the LRU so it gets another | |
619 | * pass through the LRU before it gets reclaimed. This is necessary because of | |
620 | * the fact we are doing lazy LRU updates to minimise lock contention so the | |
621 | * LRU does not have strict ordering. Hence we don't want to reclaim inodes | |
622 | * with this flag set because they are the inodes that are out of order. | |
1da177e4 LT |
623 | */ |
624 | static void prune_icache(int nr_to_scan) | |
625 | { | |
626 | LIST_HEAD(freeable); | |
1da177e4 LT |
627 | int nr_scanned; |
628 | unsigned long reap = 0; | |
629 | ||
88e0fbc4 | 630 | down_read(&iprune_sem); |
1da177e4 LT |
631 | spin_lock(&inode_lock); |
632 | for (nr_scanned = 0; nr_scanned < nr_to_scan; nr_scanned++) { | |
633 | struct inode *inode; | |
634 | ||
7ccf19a8 | 635 | if (list_empty(&inode_lru)) |
1da177e4 LT |
636 | break; |
637 | ||
7ccf19a8 | 638 | inode = list_entry(inode_lru.prev, struct inode, i_lru); |
1da177e4 | 639 | |
9e38d86f NP |
640 | /* |
641 | * Referenced or dirty inodes are still in use. Give them | |
642 | * another pass through the LRU as we canot reclaim them now. | |
643 | */ | |
644 | if (atomic_read(&inode->i_count) || | |
645 | (inode->i_state & ~I_REFERENCED)) { | |
7ccf19a8 | 646 | list_del_init(&inode->i_lru); |
86c8749e | 647 | inodes_stat.nr_unused--; |
9e38d86f NP |
648 | continue; |
649 | } | |
650 | ||
651 | /* recently referenced inodes get one more pass */ | |
652 | if (inode->i_state & I_REFERENCED) { | |
7ccf19a8 | 653 | list_move(&inode->i_lru, &inode_lru); |
9e38d86f | 654 | inode->i_state &= ~I_REFERENCED; |
1da177e4 LT |
655 | continue; |
656 | } | |
657 | if (inode_has_buffers(inode) || inode->i_data.nrpages) { | |
658 | __iget(inode); | |
659 | spin_unlock(&inode_lock); | |
660 | if (remove_inode_buffers(inode)) | |
fc0ecff6 AM |
661 | reap += invalidate_mapping_pages(&inode->i_data, |
662 | 0, -1); | |
1da177e4 LT |
663 | iput(inode); |
664 | spin_lock(&inode_lock); | |
665 | ||
7ccf19a8 NP |
666 | if (inode != list_entry(inode_lru.next, |
667 | struct inode, i_lru)) | |
1da177e4 LT |
668 | continue; /* wrong inode or list_empty */ |
669 | if (!can_unuse(inode)) | |
670 | continue; | |
671 | } | |
7ef0d737 | 672 | WARN_ON(inode->i_state & I_NEW); |
1da177e4 | 673 | inode->i_state |= I_FREEING; |
7ccf19a8 NP |
674 | |
675 | /* | |
676 | * Move the inode off the IO lists and LRU once I_FREEING is | |
677 | * set so that it won't get moved back on there if it is dirty. | |
678 | */ | |
679 | list_move(&inode->i_lru, &freeable); | |
680 | list_del_init(&inode->i_wb_list); | |
86c8749e | 681 | inodes_stat.nr_unused--; |
1da177e4 | 682 | } |
f8891e5e CL |
683 | if (current_is_kswapd()) |
684 | __count_vm_events(KSWAPD_INODESTEAL, reap); | |
685 | else | |
686 | __count_vm_events(PGINODESTEAL, reap); | |
1da177e4 LT |
687 | spin_unlock(&inode_lock); |
688 | ||
689 | dispose_list(&freeable); | |
88e0fbc4 | 690 | up_read(&iprune_sem); |
1da177e4 LT |
691 | } |
692 | ||
693 | /* | |
694 | * shrink_icache_memory() will attempt to reclaim some unused inodes. Here, | |
695 | * "unused" means that no dentries are referring to the inodes: the files are | |
696 | * not open and the dcache references to those inodes have already been | |
697 | * reclaimed. | |
698 | * | |
699 | * This function is passed the number of inodes to scan, and it returns the | |
700 | * total number of remaining possibly-reclaimable inodes. | |
701 | */ | |
7f8275d0 | 702 | static int shrink_icache_memory(struct shrinker *shrink, int nr, gfp_t gfp_mask) |
1da177e4 LT |
703 | { |
704 | if (nr) { | |
705 | /* | |
706 | * Nasty deadlock avoidance. We may hold various FS locks, | |
707 | * and we don't want to recurse into the FS that called us | |
708 | * in clear_inode() and friends.. | |
6b3304b5 | 709 | */ |
1da177e4 LT |
710 | if (!(gfp_mask & __GFP_FS)) |
711 | return -1; | |
712 | prune_icache(nr); | |
713 | } | |
cffbc8aa | 714 | return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure; |
1da177e4 LT |
715 | } |
716 | ||
8e1f936b RR |
717 | static struct shrinker icache_shrinker = { |
718 | .shrink = shrink_icache_memory, | |
719 | .seeks = DEFAULT_SEEKS, | |
720 | }; | |
721 | ||
1da177e4 LT |
722 | static void __wait_on_freeing_inode(struct inode *inode); |
723 | /* | |
724 | * Called with the inode lock held. | |
1da177e4 | 725 | */ |
6b3304b5 MK |
726 | static struct inode *find_inode(struct super_block *sb, |
727 | struct hlist_head *head, | |
728 | int (*test)(struct inode *, void *), | |
729 | void *data) | |
1da177e4 LT |
730 | { |
731 | struct hlist_node *node; | |
6b3304b5 | 732 | struct inode *inode = NULL; |
1da177e4 LT |
733 | |
734 | repeat: | |
c5c8be3c | 735 | hlist_for_each_entry(inode, node, head, i_hash) { |
1da177e4 LT |
736 | if (inode->i_sb != sb) |
737 | continue; | |
738 | if (!test(inode, data)) | |
739 | continue; | |
a4ffdde6 | 740 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
741 | __wait_on_freeing_inode(inode); |
742 | goto repeat; | |
743 | } | |
f7899bd5 CH |
744 | __iget(inode); |
745 | return inode; | |
1da177e4 | 746 | } |
f7899bd5 | 747 | return NULL; |
1da177e4 LT |
748 | } |
749 | ||
750 | /* | |
751 | * find_inode_fast is the fast path version of find_inode, see the comment at | |
752 | * iget_locked for details. | |
753 | */ | |
6b3304b5 MK |
754 | static struct inode *find_inode_fast(struct super_block *sb, |
755 | struct hlist_head *head, unsigned long ino) | |
1da177e4 LT |
756 | { |
757 | struct hlist_node *node; | |
6b3304b5 | 758 | struct inode *inode = NULL; |
1da177e4 LT |
759 | |
760 | repeat: | |
c5c8be3c | 761 | hlist_for_each_entry(inode, node, head, i_hash) { |
1da177e4 LT |
762 | if (inode->i_ino != ino) |
763 | continue; | |
764 | if (inode->i_sb != sb) | |
765 | continue; | |
a4ffdde6 | 766 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
767 | __wait_on_freeing_inode(inode); |
768 | goto repeat; | |
769 | } | |
f7899bd5 CH |
770 | __iget(inode); |
771 | return inode; | |
1da177e4 | 772 | } |
f7899bd5 | 773 | return NULL; |
8290c35f DC |
774 | } |
775 | ||
f991bd2e ED |
776 | /* |
777 | * Each cpu owns a range of LAST_INO_BATCH numbers. | |
778 | * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations, | |
779 | * to renew the exhausted range. | |
8290c35f | 780 | * |
f991bd2e ED |
781 | * This does not significantly increase overflow rate because every CPU can |
782 | * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is | |
783 | * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the | |
784 | * 2^32 range, and is a worst-case. Even a 50% wastage would only increase | |
785 | * overflow rate by 2x, which does not seem too significant. | |
786 | * | |
787 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
788 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
789 | * here to attempt to avoid that. | |
8290c35f | 790 | */ |
f991bd2e ED |
791 | #define LAST_INO_BATCH 1024 |
792 | static DEFINE_PER_CPU(unsigned int, last_ino); | |
793 | ||
85fe4025 | 794 | unsigned int get_next_ino(void) |
8290c35f | 795 | { |
f991bd2e ED |
796 | unsigned int *p = &get_cpu_var(last_ino); |
797 | unsigned int res = *p; | |
8290c35f | 798 | |
f991bd2e ED |
799 | #ifdef CONFIG_SMP |
800 | if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) { | |
801 | static atomic_t shared_last_ino; | |
802 | int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino); | |
803 | ||
804 | res = next - LAST_INO_BATCH; | |
805 | } | |
806 | #endif | |
807 | ||
808 | *p = ++res; | |
809 | put_cpu_var(last_ino); | |
810 | return res; | |
8290c35f | 811 | } |
85fe4025 | 812 | EXPORT_SYMBOL(get_next_ino); |
8290c35f | 813 | |
1da177e4 LT |
814 | /** |
815 | * new_inode - obtain an inode | |
816 | * @sb: superblock | |
817 | * | |
769848c0 | 818 | * Allocates a new inode for given superblock. The default gfp_mask |
3c1d4378 | 819 | * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE. |
769848c0 MG |
820 | * If HIGHMEM pages are unsuitable or it is known that pages allocated |
821 | * for the page cache are not reclaimable or migratable, | |
822 | * mapping_set_gfp_mask() must be called with suitable flags on the | |
823 | * newly created inode's mapping | |
824 | * | |
1da177e4 LT |
825 | */ |
826 | struct inode *new_inode(struct super_block *sb) | |
827 | { | |
6b3304b5 | 828 | struct inode *inode; |
1da177e4 LT |
829 | |
830 | spin_lock_prefetch(&inode_lock); | |
6b3304b5 | 831 | |
1da177e4 LT |
832 | inode = alloc_inode(sb); |
833 | if (inode) { | |
834 | spin_lock(&inode_lock); | |
646ec461 | 835 | __inode_sb_list_add(inode); |
1da177e4 LT |
836 | inode->i_state = 0; |
837 | spin_unlock(&inode_lock); | |
838 | } | |
839 | return inode; | |
840 | } | |
1da177e4 LT |
841 | EXPORT_SYMBOL(new_inode); |
842 | ||
843 | void unlock_new_inode(struct inode *inode) | |
844 | { | |
14358e6d | 845 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
a3314a0e | 846 | if (S_ISDIR(inode->i_mode)) { |
1e89a5e1 PZ |
847 | struct file_system_type *type = inode->i_sb->s_type; |
848 | ||
9a7aa12f JK |
849 | /* Set new key only if filesystem hasn't already changed it */ |
850 | if (!lockdep_match_class(&inode->i_mutex, | |
851 | &type->i_mutex_key)) { | |
852 | /* | |
853 | * ensure nobody is actually holding i_mutex | |
854 | */ | |
855 | mutex_destroy(&inode->i_mutex); | |
856 | mutex_init(&inode->i_mutex); | |
857 | lockdep_set_class(&inode->i_mutex, | |
858 | &type->i_mutex_dir_key); | |
859 | } | |
1e89a5e1 | 860 | } |
14358e6d | 861 | #endif |
1da177e4 | 862 | /* |
eaff8079 | 863 | * This is special! We do not need the spinlock when clearing I_NEW, |
580be083 JK |
864 | * because we're guaranteed that nobody else tries to do anything about |
865 | * the state of the inode when it is locked, as we just created it (so | |
eaff8079 | 866 | * there can be no old holders that haven't tested I_NEW). |
580be083 | 867 | * However we must emit the memory barrier so that other CPUs reliably |
eaff8079 | 868 | * see the clearing of I_NEW after the other inode initialisation has |
580be083 | 869 | * completed. |
1da177e4 | 870 | */ |
580be083 | 871 | smp_mb(); |
eaff8079 CH |
872 | WARN_ON(!(inode->i_state & I_NEW)); |
873 | inode->i_state &= ~I_NEW; | |
1da177e4 LT |
874 | wake_up_inode(inode); |
875 | } | |
1da177e4 LT |
876 | EXPORT_SYMBOL(unlock_new_inode); |
877 | ||
878 | /* | |
879 | * This is called without the inode lock held.. Be careful. | |
880 | * | |
881 | * We no longer cache the sb_flags in i_flags - see fs.h | |
882 | * -- rmk@arm.uk.linux.org | |
883 | */ | |
6b3304b5 MK |
884 | static struct inode *get_new_inode(struct super_block *sb, |
885 | struct hlist_head *head, | |
886 | int (*test)(struct inode *, void *), | |
887 | int (*set)(struct inode *, void *), | |
888 | void *data) | |
1da177e4 | 889 | { |
6b3304b5 | 890 | struct inode *inode; |
1da177e4 LT |
891 | |
892 | inode = alloc_inode(sb); | |
893 | if (inode) { | |
6b3304b5 | 894 | struct inode *old; |
1da177e4 LT |
895 | |
896 | spin_lock(&inode_lock); | |
897 | /* We released the lock, so.. */ | |
898 | old = find_inode(sb, head, test, data); | |
899 | if (!old) { | |
900 | if (set(inode, data)) | |
901 | goto set_failed; | |
902 | ||
646ec461 CH |
903 | hlist_add_head(&inode->i_hash, head); |
904 | __inode_sb_list_add(inode); | |
eaff8079 | 905 | inode->i_state = I_NEW; |
1da177e4 LT |
906 | spin_unlock(&inode_lock); |
907 | ||
908 | /* Return the locked inode with I_NEW set, the | |
909 | * caller is responsible for filling in the contents | |
910 | */ | |
911 | return inode; | |
912 | } | |
913 | ||
914 | /* | |
915 | * Uhhuh, somebody else created the same inode under | |
916 | * us. Use the old inode instead of the one we just | |
917 | * allocated. | |
918 | */ | |
1da177e4 LT |
919 | spin_unlock(&inode_lock); |
920 | destroy_inode(inode); | |
921 | inode = old; | |
922 | wait_on_inode(inode); | |
923 | } | |
924 | return inode; | |
925 | ||
926 | set_failed: | |
927 | spin_unlock(&inode_lock); | |
928 | destroy_inode(inode); | |
929 | return NULL; | |
930 | } | |
931 | ||
932 | /* | |
933 | * get_new_inode_fast is the fast path version of get_new_inode, see the | |
934 | * comment at iget_locked for details. | |
935 | */ | |
6b3304b5 MK |
936 | static struct inode *get_new_inode_fast(struct super_block *sb, |
937 | struct hlist_head *head, unsigned long ino) | |
1da177e4 | 938 | { |
6b3304b5 | 939 | struct inode *inode; |
1da177e4 LT |
940 | |
941 | inode = alloc_inode(sb); | |
942 | if (inode) { | |
6b3304b5 | 943 | struct inode *old; |
1da177e4 LT |
944 | |
945 | spin_lock(&inode_lock); | |
946 | /* We released the lock, so.. */ | |
947 | old = find_inode_fast(sb, head, ino); | |
948 | if (!old) { | |
949 | inode->i_ino = ino; | |
646ec461 CH |
950 | hlist_add_head(&inode->i_hash, head); |
951 | __inode_sb_list_add(inode); | |
eaff8079 | 952 | inode->i_state = I_NEW; |
1da177e4 LT |
953 | spin_unlock(&inode_lock); |
954 | ||
955 | /* Return the locked inode with I_NEW set, the | |
956 | * caller is responsible for filling in the contents | |
957 | */ | |
958 | return inode; | |
959 | } | |
960 | ||
961 | /* | |
962 | * Uhhuh, somebody else created the same inode under | |
963 | * us. Use the old inode instead of the one we just | |
964 | * allocated. | |
965 | */ | |
1da177e4 LT |
966 | spin_unlock(&inode_lock); |
967 | destroy_inode(inode); | |
968 | inode = old; | |
969 | wait_on_inode(inode); | |
970 | } | |
971 | return inode; | |
972 | } | |
973 | ||
ad5e195a CH |
974 | /* |
975 | * search the inode cache for a matching inode number. | |
976 | * If we find one, then the inode number we are trying to | |
977 | * allocate is not unique and so we should not use it. | |
978 | * | |
979 | * Returns 1 if the inode number is unique, 0 if it is not. | |
980 | */ | |
981 | static int test_inode_iunique(struct super_block *sb, unsigned long ino) | |
982 | { | |
983 | struct hlist_head *b = inode_hashtable + hash(sb, ino); | |
984 | struct hlist_node *node; | |
985 | struct inode *inode; | |
986 | ||
987 | hlist_for_each_entry(inode, node, b, i_hash) { | |
988 | if (inode->i_ino == ino && inode->i_sb == sb) | |
989 | return 0; | |
990 | } | |
991 | ||
992 | return 1; | |
993 | } | |
994 | ||
1da177e4 LT |
995 | /** |
996 | * iunique - get a unique inode number | |
997 | * @sb: superblock | |
998 | * @max_reserved: highest reserved inode number | |
999 | * | |
1000 | * Obtain an inode number that is unique on the system for a given | |
1001 | * superblock. This is used by file systems that have no natural | |
1002 | * permanent inode numbering system. An inode number is returned that | |
1003 | * is higher than the reserved limit but unique. | |
1004 | * | |
1005 | * BUGS: | |
1006 | * With a large number of inodes live on the file system this function | |
1007 | * currently becomes quite slow. | |
1008 | */ | |
1009 | ino_t iunique(struct super_block *sb, ino_t max_reserved) | |
1010 | { | |
866b04fc JL |
1011 | /* |
1012 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
1013 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
1014 | * here to attempt to avoid that. | |
1015 | */ | |
ad5e195a | 1016 | static DEFINE_SPINLOCK(iunique_lock); |
866b04fc | 1017 | static unsigned int counter; |
1da177e4 | 1018 | ino_t res; |
3361c7be | 1019 | |
1da177e4 | 1020 | spin_lock(&inode_lock); |
ad5e195a | 1021 | spin_lock(&iunique_lock); |
3361c7be JL |
1022 | do { |
1023 | if (counter <= max_reserved) | |
1024 | counter = max_reserved + 1; | |
1da177e4 | 1025 | res = counter++; |
ad5e195a CH |
1026 | } while (!test_inode_iunique(sb, res)); |
1027 | spin_unlock(&iunique_lock); | |
3361c7be | 1028 | spin_unlock(&inode_lock); |
1da177e4 | 1029 | |
3361c7be JL |
1030 | return res; |
1031 | } | |
1da177e4 LT |
1032 | EXPORT_SYMBOL(iunique); |
1033 | ||
1034 | struct inode *igrab(struct inode *inode) | |
1035 | { | |
1036 | spin_lock(&inode_lock); | |
a4ffdde6 | 1037 | if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) |
1da177e4 LT |
1038 | __iget(inode); |
1039 | else | |
1040 | /* | |
1041 | * Handle the case where s_op->clear_inode is not been | |
1042 | * called yet, and somebody is calling igrab | |
1043 | * while the inode is getting freed. | |
1044 | */ | |
1045 | inode = NULL; | |
1046 | spin_unlock(&inode_lock); | |
1047 | return inode; | |
1048 | } | |
1da177e4 LT |
1049 | EXPORT_SYMBOL(igrab); |
1050 | ||
1051 | /** | |
1052 | * ifind - internal function, you want ilookup5() or iget5(). | |
1053 | * @sb: super block of file system to search | |
1054 | * @head: the head of the list to search | |
1055 | * @test: callback used for comparisons between inodes | |
1056 | * @data: opaque data pointer to pass to @test | |
88bd5121 | 1057 | * @wait: if true wait for the inode to be unlocked, if false do not |
1da177e4 LT |
1058 | * |
1059 | * ifind() searches for the inode specified by @data in the inode | |
1060 | * cache. This is a generalized version of ifind_fast() for file systems where | |
1061 | * the inode number is not sufficient for unique identification of an inode. | |
1062 | * | |
1063 | * If the inode is in the cache, the inode is returned with an incremented | |
1064 | * reference count. | |
1065 | * | |
1066 | * Otherwise NULL is returned. | |
1067 | * | |
1068 | * Note, @test is called with the inode_lock held, so can't sleep. | |
1069 | */ | |
5d2bea45 | 1070 | static struct inode *ifind(struct super_block *sb, |
1da177e4 | 1071 | struct hlist_head *head, int (*test)(struct inode *, void *), |
88bd5121 | 1072 | void *data, const int wait) |
1da177e4 LT |
1073 | { |
1074 | struct inode *inode; | |
1075 | ||
1076 | spin_lock(&inode_lock); | |
1077 | inode = find_inode(sb, head, test, data); | |
1078 | if (inode) { | |
1da177e4 | 1079 | spin_unlock(&inode_lock); |
88bd5121 AA |
1080 | if (likely(wait)) |
1081 | wait_on_inode(inode); | |
1da177e4 LT |
1082 | return inode; |
1083 | } | |
1084 | spin_unlock(&inode_lock); | |
1085 | return NULL; | |
1086 | } | |
1087 | ||
1088 | /** | |
1089 | * ifind_fast - internal function, you want ilookup() or iget(). | |
1090 | * @sb: super block of file system to search | |
1091 | * @head: head of the list to search | |
1092 | * @ino: inode number to search for | |
1093 | * | |
1094 | * ifind_fast() searches for the inode @ino in the inode cache. This is for | |
1095 | * file systems where the inode number is sufficient for unique identification | |
1096 | * of an inode. | |
1097 | * | |
1098 | * If the inode is in the cache, the inode is returned with an incremented | |
1099 | * reference count. | |
1100 | * | |
1101 | * Otherwise NULL is returned. | |
1102 | */ | |
5d2bea45 | 1103 | static struct inode *ifind_fast(struct super_block *sb, |
1da177e4 LT |
1104 | struct hlist_head *head, unsigned long ino) |
1105 | { | |
1106 | struct inode *inode; | |
1107 | ||
1108 | spin_lock(&inode_lock); | |
1109 | inode = find_inode_fast(sb, head, ino); | |
1110 | if (inode) { | |
1da177e4 LT |
1111 | spin_unlock(&inode_lock); |
1112 | wait_on_inode(inode); | |
1113 | return inode; | |
1114 | } | |
1115 | spin_unlock(&inode_lock); | |
1116 | return NULL; | |
1117 | } | |
1118 | ||
1119 | /** | |
88bd5121 | 1120 | * ilookup5_nowait - search for an inode in the inode cache |
1da177e4 LT |
1121 | * @sb: super block of file system to search |
1122 | * @hashval: hash value (usually inode number) to search for | |
1123 | * @test: callback used for comparisons between inodes | |
1124 | * @data: opaque data pointer to pass to @test | |
1125 | * | |
1126 | * ilookup5() uses ifind() to search for the inode specified by @hashval and | |
1127 | * @data in the inode cache. This is a generalized version of ilookup() for | |
1128 | * file systems where the inode number is not sufficient for unique | |
1129 | * identification of an inode. | |
1130 | * | |
1131 | * If the inode is in the cache, the inode is returned with an incremented | |
88bd5121 AA |
1132 | * reference count. Note, the inode lock is not waited upon so you have to be |
1133 | * very careful what you do with the returned inode. You probably should be | |
1134 | * using ilookup5() instead. | |
1135 | * | |
1136 | * Otherwise NULL is returned. | |
1137 | * | |
1138 | * Note, @test is called with the inode_lock held, so can't sleep. | |
1139 | */ | |
1140 | struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval, | |
1141 | int (*test)(struct inode *, void *), void *data) | |
1142 | { | |
1143 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1144 | ||
1145 | return ifind(sb, head, test, data, 0); | |
1146 | } | |
88bd5121 AA |
1147 | EXPORT_SYMBOL(ilookup5_nowait); |
1148 | ||
1149 | /** | |
1150 | * ilookup5 - search for an inode in the inode cache | |
1151 | * @sb: super block of file system to search | |
1152 | * @hashval: hash value (usually inode number) to search for | |
1153 | * @test: callback used for comparisons between inodes | |
1154 | * @data: opaque data pointer to pass to @test | |
1155 | * | |
1156 | * ilookup5() uses ifind() to search for the inode specified by @hashval and | |
1157 | * @data in the inode cache. This is a generalized version of ilookup() for | |
1158 | * file systems where the inode number is not sufficient for unique | |
1159 | * identification of an inode. | |
1160 | * | |
1161 | * If the inode is in the cache, the inode lock is waited upon and the inode is | |
1162 | * returned with an incremented reference count. | |
1da177e4 LT |
1163 | * |
1164 | * Otherwise NULL is returned. | |
1165 | * | |
1166 | * Note, @test is called with the inode_lock held, so can't sleep. | |
1167 | */ | |
1168 | struct inode *ilookup5(struct super_block *sb, unsigned long hashval, | |
1169 | int (*test)(struct inode *, void *), void *data) | |
1170 | { | |
1171 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1172 | ||
88bd5121 | 1173 | return ifind(sb, head, test, data, 1); |
1da177e4 | 1174 | } |
1da177e4 LT |
1175 | EXPORT_SYMBOL(ilookup5); |
1176 | ||
1177 | /** | |
1178 | * ilookup - search for an inode in the inode cache | |
1179 | * @sb: super block of file system to search | |
1180 | * @ino: inode number to search for | |
1181 | * | |
1182 | * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache. | |
1183 | * This is for file systems where the inode number is sufficient for unique | |
1184 | * identification of an inode. | |
1185 | * | |
1186 | * If the inode is in the cache, the inode is returned with an incremented | |
1187 | * reference count. | |
1188 | * | |
1189 | * Otherwise NULL is returned. | |
1190 | */ | |
1191 | struct inode *ilookup(struct super_block *sb, unsigned long ino) | |
1192 | { | |
1193 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1194 | ||
1195 | return ifind_fast(sb, head, ino); | |
1196 | } | |
1da177e4 LT |
1197 | EXPORT_SYMBOL(ilookup); |
1198 | ||
1199 | /** | |
1200 | * iget5_locked - obtain an inode from a mounted file system | |
1201 | * @sb: super block of file system | |
1202 | * @hashval: hash value (usually inode number) to get | |
1203 | * @test: callback used for comparisons between inodes | |
1204 | * @set: callback used to initialize a new struct inode | |
1205 | * @data: opaque data pointer to pass to @test and @set | |
1206 | * | |
1da177e4 LT |
1207 | * iget5_locked() uses ifind() to search for the inode specified by @hashval |
1208 | * and @data in the inode cache and if present it is returned with an increased | |
1209 | * reference count. This is a generalized version of iget_locked() for file | |
1210 | * systems where the inode number is not sufficient for unique identification | |
1211 | * of an inode. | |
1212 | * | |
1213 | * If the inode is not in cache, get_new_inode() is called to allocate a new | |
1214 | * inode and this is returned locked, hashed, and with the I_NEW flag set. The | |
1215 | * file system gets to fill it in before unlocking it via unlock_new_inode(). | |
1216 | * | |
1217 | * Note both @test and @set are called with the inode_lock held, so can't sleep. | |
1218 | */ | |
1219 | struct inode *iget5_locked(struct super_block *sb, unsigned long hashval, | |
1220 | int (*test)(struct inode *, void *), | |
1221 | int (*set)(struct inode *, void *), void *data) | |
1222 | { | |
1223 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1224 | struct inode *inode; | |
1225 | ||
88bd5121 | 1226 | inode = ifind(sb, head, test, data, 1); |
1da177e4 LT |
1227 | if (inode) |
1228 | return inode; | |
1229 | /* | |
1230 | * get_new_inode() will do the right thing, re-trying the search | |
1231 | * in case it had to block at any point. | |
1232 | */ | |
1233 | return get_new_inode(sb, head, test, set, data); | |
1234 | } | |
1da177e4 LT |
1235 | EXPORT_SYMBOL(iget5_locked); |
1236 | ||
1237 | /** | |
1238 | * iget_locked - obtain an inode from a mounted file system | |
1239 | * @sb: super block of file system | |
1240 | * @ino: inode number to get | |
1241 | * | |
1da177e4 LT |
1242 | * iget_locked() uses ifind_fast() to search for the inode specified by @ino in |
1243 | * the inode cache and if present it is returned with an increased reference | |
1244 | * count. This is for file systems where the inode number is sufficient for | |
1245 | * unique identification of an inode. | |
1246 | * | |
1247 | * If the inode is not in cache, get_new_inode_fast() is called to allocate a | |
1248 | * new inode and this is returned locked, hashed, and with the I_NEW flag set. | |
1249 | * The file system gets to fill it in before unlocking it via | |
1250 | * unlock_new_inode(). | |
1251 | */ | |
1252 | struct inode *iget_locked(struct super_block *sb, unsigned long ino) | |
1253 | { | |
1254 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1255 | struct inode *inode; | |
1256 | ||
1257 | inode = ifind_fast(sb, head, ino); | |
1258 | if (inode) | |
1259 | return inode; | |
1260 | /* | |
1261 | * get_new_inode_fast() will do the right thing, re-trying the search | |
1262 | * in case it had to block at any point. | |
1263 | */ | |
1264 | return get_new_inode_fast(sb, head, ino); | |
1265 | } | |
1da177e4 LT |
1266 | EXPORT_SYMBOL(iget_locked); |
1267 | ||
261bca86 AV |
1268 | int insert_inode_locked(struct inode *inode) |
1269 | { | |
1270 | struct super_block *sb = inode->i_sb; | |
1271 | ino_t ino = inode->i_ino; | |
1272 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
261bca86 | 1273 | |
eaff8079 | 1274 | inode->i_state |= I_NEW; |
261bca86 | 1275 | while (1) { |
72a43d63 AV |
1276 | struct hlist_node *node; |
1277 | struct inode *old = NULL; | |
261bca86 | 1278 | spin_lock(&inode_lock); |
72a43d63 AV |
1279 | hlist_for_each_entry(old, node, head, i_hash) { |
1280 | if (old->i_ino != ino) | |
1281 | continue; | |
1282 | if (old->i_sb != sb) | |
1283 | continue; | |
a4ffdde6 | 1284 | if (old->i_state & (I_FREEING|I_WILL_FREE)) |
72a43d63 AV |
1285 | continue; |
1286 | break; | |
1287 | } | |
1288 | if (likely(!node)) { | |
261bca86 AV |
1289 | hlist_add_head(&inode->i_hash, head); |
1290 | spin_unlock(&inode_lock); | |
1291 | return 0; | |
1292 | } | |
1293 | __iget(old); | |
1294 | spin_unlock(&inode_lock); | |
1295 | wait_on_inode(old); | |
1d3382cb | 1296 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1297 | iput(old); |
1298 | return -EBUSY; | |
1299 | } | |
1300 | iput(old); | |
1301 | } | |
1302 | } | |
261bca86 AV |
1303 | EXPORT_SYMBOL(insert_inode_locked); |
1304 | ||
1305 | int insert_inode_locked4(struct inode *inode, unsigned long hashval, | |
1306 | int (*test)(struct inode *, void *), void *data) | |
1307 | { | |
1308 | struct super_block *sb = inode->i_sb; | |
1309 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
261bca86 | 1310 | |
eaff8079 | 1311 | inode->i_state |= I_NEW; |
261bca86 AV |
1312 | |
1313 | while (1) { | |
72a43d63 AV |
1314 | struct hlist_node *node; |
1315 | struct inode *old = NULL; | |
1316 | ||
261bca86 | 1317 | spin_lock(&inode_lock); |
72a43d63 AV |
1318 | hlist_for_each_entry(old, node, head, i_hash) { |
1319 | if (old->i_sb != sb) | |
1320 | continue; | |
1321 | if (!test(old, data)) | |
1322 | continue; | |
a4ffdde6 | 1323 | if (old->i_state & (I_FREEING|I_WILL_FREE)) |
72a43d63 AV |
1324 | continue; |
1325 | break; | |
1326 | } | |
1327 | if (likely(!node)) { | |
261bca86 AV |
1328 | hlist_add_head(&inode->i_hash, head); |
1329 | spin_unlock(&inode_lock); | |
1330 | return 0; | |
1331 | } | |
1332 | __iget(old); | |
1333 | spin_unlock(&inode_lock); | |
1334 | wait_on_inode(old); | |
1d3382cb | 1335 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1336 | iput(old); |
1337 | return -EBUSY; | |
1338 | } | |
1339 | iput(old); | |
1340 | } | |
1341 | } | |
261bca86 AV |
1342 | EXPORT_SYMBOL(insert_inode_locked4); |
1343 | ||
1da177e4 | 1344 | |
45321ac5 AV |
1345 | int generic_delete_inode(struct inode *inode) |
1346 | { | |
1347 | return 1; | |
1348 | } | |
1349 | EXPORT_SYMBOL(generic_delete_inode); | |
1350 | ||
1da177e4 | 1351 | /* |
45321ac5 AV |
1352 | * Normal UNIX filesystem behaviour: delete the |
1353 | * inode when the usage count drops to zero, and | |
1354 | * i_nlink is zero. | |
1da177e4 | 1355 | */ |
45321ac5 | 1356 | int generic_drop_inode(struct inode *inode) |
1da177e4 | 1357 | { |
1d3382cb | 1358 | return !inode->i_nlink || inode_unhashed(inode); |
1da177e4 | 1359 | } |
45321ac5 | 1360 | EXPORT_SYMBOL_GPL(generic_drop_inode); |
1da177e4 | 1361 | |
45321ac5 AV |
1362 | /* |
1363 | * Called when we're dropping the last reference | |
1364 | * to an inode. | |
22fe4042 | 1365 | * |
45321ac5 AV |
1366 | * Call the FS "drop_inode()" function, defaulting to |
1367 | * the legacy UNIX filesystem behaviour. If it tells | |
1368 | * us to evict inode, do so. Otherwise, retain inode | |
1369 | * in cache if fs is alive, sync and evict if fs is | |
1370 | * shutting down. | |
22fe4042 | 1371 | */ |
45321ac5 | 1372 | static void iput_final(struct inode *inode) |
1da177e4 LT |
1373 | { |
1374 | struct super_block *sb = inode->i_sb; | |
45321ac5 AV |
1375 | const struct super_operations *op = inode->i_sb->s_op; |
1376 | int drop; | |
1377 | ||
1378 | if (op && op->drop_inode) | |
1379 | drop = op->drop_inode(inode); | |
1380 | else | |
1381 | drop = generic_drop_inode(inode); | |
1da177e4 | 1382 | |
45321ac5 | 1383 | if (!drop) { |
acb0c854 | 1384 | if (sb->s_flags & MS_ACTIVE) { |
9e38d86f NP |
1385 | inode->i_state |= I_REFERENCED; |
1386 | if (!(inode->i_state & (I_DIRTY|I_SYNC))) { | |
1387 | inode_lru_list_add(inode); | |
1388 | } | |
991114c6 | 1389 | spin_unlock(&inode_lock); |
45321ac5 | 1390 | return; |
991114c6 | 1391 | } |
7ef0d737 | 1392 | WARN_ON(inode->i_state & I_NEW); |
991114c6 AV |
1393 | inode->i_state |= I_WILL_FREE; |
1394 | spin_unlock(&inode_lock); | |
1da177e4 LT |
1395 | write_inode_now(inode, 1); |
1396 | spin_lock(&inode_lock); | |
7ef0d737 | 1397 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1398 | inode->i_state &= ~I_WILL_FREE; |
4c51acbc | 1399 | __remove_inode_hash(inode); |
1da177e4 | 1400 | } |
7ccf19a8 | 1401 | |
7ef0d737 | 1402 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1403 | inode->i_state |= I_FREEING; |
9e38d86f NP |
1404 | |
1405 | /* | |
7ccf19a8 NP |
1406 | * Move the inode off the IO lists and LRU once I_FREEING is |
1407 | * set so that it won't get moved back on there if it is dirty. | |
9e38d86f NP |
1408 | */ |
1409 | inode_lru_list_del(inode); | |
7ccf19a8 | 1410 | list_del_init(&inode->i_wb_list); |
9e38d86f | 1411 | |
646ec461 | 1412 | __inode_sb_list_del(inode); |
1da177e4 | 1413 | spin_unlock(&inode_lock); |
644da596 | 1414 | evict(inode); |
4c51acbc | 1415 | remove_inode_hash(inode); |
7f04c26d | 1416 | wake_up_inode(inode); |
45321ac5 | 1417 | BUG_ON(inode->i_state != (I_FREEING | I_CLEAR)); |
1da177e4 LT |
1418 | destroy_inode(inode); |
1419 | } | |
1420 | ||
1da177e4 | 1421 | /** |
6b3304b5 | 1422 | * iput - put an inode |
1da177e4 LT |
1423 | * @inode: inode to put |
1424 | * | |
1425 | * Puts an inode, dropping its usage count. If the inode use count hits | |
1426 | * zero, the inode is then freed and may also be destroyed. | |
1427 | * | |
1428 | * Consequently, iput() can sleep. | |
1429 | */ | |
1430 | void iput(struct inode *inode) | |
1431 | { | |
1432 | if (inode) { | |
a4ffdde6 | 1433 | BUG_ON(inode->i_state & I_CLEAR); |
1da177e4 | 1434 | |
1da177e4 LT |
1435 | if (atomic_dec_and_lock(&inode->i_count, &inode_lock)) |
1436 | iput_final(inode); | |
1437 | } | |
1438 | } | |
1da177e4 LT |
1439 | EXPORT_SYMBOL(iput); |
1440 | ||
1441 | /** | |
1442 | * bmap - find a block number in a file | |
1443 | * @inode: inode of file | |
1444 | * @block: block to find | |
1445 | * | |
1446 | * Returns the block number on the device holding the inode that | |
1447 | * is the disk block number for the block of the file requested. | |
1448 | * That is, asked for block 4 of inode 1 the function will return the | |
6b3304b5 | 1449 | * disk block relative to the disk start that holds that block of the |
1da177e4 LT |
1450 | * file. |
1451 | */ | |
6b3304b5 | 1452 | sector_t bmap(struct inode *inode, sector_t block) |
1da177e4 LT |
1453 | { |
1454 | sector_t res = 0; | |
1455 | if (inode->i_mapping->a_ops->bmap) | |
1456 | res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block); | |
1457 | return res; | |
1458 | } | |
1da177e4 LT |
1459 | EXPORT_SYMBOL(bmap); |
1460 | ||
11ff6f05 MG |
1461 | /* |
1462 | * With relative atime, only update atime if the previous atime is | |
1463 | * earlier than either the ctime or mtime or if at least a day has | |
1464 | * passed since the last atime update. | |
1465 | */ | |
1466 | static int relatime_need_update(struct vfsmount *mnt, struct inode *inode, | |
1467 | struct timespec now) | |
1468 | { | |
1469 | ||
1470 | if (!(mnt->mnt_flags & MNT_RELATIME)) | |
1471 | return 1; | |
1472 | /* | |
1473 | * Is mtime younger than atime? If yes, update atime: | |
1474 | */ | |
1475 | if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0) | |
1476 | return 1; | |
1477 | /* | |
1478 | * Is ctime younger than atime? If yes, update atime: | |
1479 | */ | |
1480 | if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0) | |
1481 | return 1; | |
1482 | ||
1483 | /* | |
1484 | * Is the previous atime value older than a day? If yes, | |
1485 | * update atime: | |
1486 | */ | |
1487 | if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60) | |
1488 | return 1; | |
1489 | /* | |
1490 | * Good, we can skip the atime update: | |
1491 | */ | |
1492 | return 0; | |
1493 | } | |
1494 | ||
1da177e4 | 1495 | /** |
869243a0 CH |
1496 | * touch_atime - update the access time |
1497 | * @mnt: mount the inode is accessed on | |
7045f37b | 1498 | * @dentry: dentry accessed |
1da177e4 LT |
1499 | * |
1500 | * Update the accessed time on an inode and mark it for writeback. | |
1501 | * This function automatically handles read only file systems and media, | |
1502 | * as well as the "noatime" flag and inode specific "noatime" markers. | |
1503 | */ | |
869243a0 | 1504 | void touch_atime(struct vfsmount *mnt, struct dentry *dentry) |
1da177e4 | 1505 | { |
869243a0 | 1506 | struct inode *inode = dentry->d_inode; |
1da177e4 LT |
1507 | struct timespec now; |
1508 | ||
cdb70f3f | 1509 | if (inode->i_flags & S_NOATIME) |
b12536c2 | 1510 | return; |
37756ced | 1511 | if (IS_NOATIME(inode)) |
b12536c2 | 1512 | return; |
b2276138 | 1513 | if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1514 | return; |
47ae32d6 | 1515 | |
cdb70f3f | 1516 | if (mnt->mnt_flags & MNT_NOATIME) |
b12536c2 | 1517 | return; |
cdb70f3f | 1518 | if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1519 | return; |
1da177e4 LT |
1520 | |
1521 | now = current_fs_time(inode->i_sb); | |
11ff6f05 MG |
1522 | |
1523 | if (!relatime_need_update(mnt, inode, now)) | |
b12536c2 | 1524 | return; |
11ff6f05 | 1525 | |
47ae32d6 | 1526 | if (timespec_equal(&inode->i_atime, &now)) |
b12536c2 AK |
1527 | return; |
1528 | ||
1529 | if (mnt_want_write(mnt)) | |
1530 | return; | |
47ae32d6 VH |
1531 | |
1532 | inode->i_atime = now; | |
1533 | mark_inode_dirty_sync(inode); | |
cdb70f3f | 1534 | mnt_drop_write(mnt); |
1da177e4 | 1535 | } |
869243a0 | 1536 | EXPORT_SYMBOL(touch_atime); |
1da177e4 LT |
1537 | |
1538 | /** | |
870f4817 CH |
1539 | * file_update_time - update mtime and ctime time |
1540 | * @file: file accessed | |
1da177e4 | 1541 | * |
870f4817 CH |
1542 | * Update the mtime and ctime members of an inode and mark the inode |
1543 | * for writeback. Note that this function is meant exclusively for | |
1544 | * usage in the file write path of filesystems, and filesystems may | |
1545 | * choose to explicitly ignore update via this function with the | |
2eadfc0e | 1546 | * S_NOCMTIME inode flag, e.g. for network filesystem where these |
870f4817 | 1547 | * timestamps are handled by the server. |
1da177e4 LT |
1548 | */ |
1549 | ||
870f4817 | 1550 | void file_update_time(struct file *file) |
1da177e4 | 1551 | { |
0f7fc9e4 | 1552 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 | 1553 | struct timespec now; |
ce06e0b2 | 1554 | enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0; |
1da177e4 | 1555 | |
ce06e0b2 | 1556 | /* First try to exhaust all avenues to not sync */ |
1da177e4 LT |
1557 | if (IS_NOCMTIME(inode)) |
1558 | return; | |
20ddee2c | 1559 | |
1da177e4 | 1560 | now = current_fs_time(inode->i_sb); |
ce06e0b2 AK |
1561 | if (!timespec_equal(&inode->i_mtime, &now)) |
1562 | sync_it = S_MTIME; | |
1da177e4 | 1563 | |
ce06e0b2 AK |
1564 | if (!timespec_equal(&inode->i_ctime, &now)) |
1565 | sync_it |= S_CTIME; | |
870f4817 | 1566 | |
ce06e0b2 AK |
1567 | if (IS_I_VERSION(inode)) |
1568 | sync_it |= S_VERSION; | |
7a224228 | 1569 | |
ce06e0b2 AK |
1570 | if (!sync_it) |
1571 | return; | |
1572 | ||
1573 | /* Finally allowed to write? Takes lock. */ | |
1574 | if (mnt_want_write_file(file)) | |
1575 | return; | |
1576 | ||
1577 | /* Only change inode inside the lock region */ | |
1578 | if (sync_it & S_VERSION) | |
1579 | inode_inc_iversion(inode); | |
1580 | if (sync_it & S_CTIME) | |
1581 | inode->i_ctime = now; | |
1582 | if (sync_it & S_MTIME) | |
1583 | inode->i_mtime = now; | |
1584 | mark_inode_dirty_sync(inode); | |
20ddee2c | 1585 | mnt_drop_write(file->f_path.mnt); |
1da177e4 | 1586 | } |
870f4817 | 1587 | EXPORT_SYMBOL(file_update_time); |
1da177e4 LT |
1588 | |
1589 | int inode_needs_sync(struct inode *inode) | |
1590 | { | |
1591 | if (IS_SYNC(inode)) | |
1592 | return 1; | |
1593 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) | |
1594 | return 1; | |
1595 | return 0; | |
1596 | } | |
1da177e4 LT |
1597 | EXPORT_SYMBOL(inode_needs_sync); |
1598 | ||
1da177e4 LT |
1599 | int inode_wait(void *word) |
1600 | { | |
1601 | schedule(); | |
1602 | return 0; | |
1603 | } | |
d44dab8d | 1604 | EXPORT_SYMBOL(inode_wait); |
1da177e4 LT |
1605 | |
1606 | /* | |
168a9fd6 MS |
1607 | * If we try to find an inode in the inode hash while it is being |
1608 | * deleted, we have to wait until the filesystem completes its | |
1609 | * deletion before reporting that it isn't found. This function waits | |
1610 | * until the deletion _might_ have completed. Callers are responsible | |
1611 | * to recheck inode state. | |
1612 | * | |
eaff8079 | 1613 | * It doesn't matter if I_NEW is not set initially, a call to |
168a9fd6 MS |
1614 | * wake_up_inode() after removing from the hash list will DTRT. |
1615 | * | |
1da177e4 LT |
1616 | * This is called with inode_lock held. |
1617 | */ | |
1618 | static void __wait_on_freeing_inode(struct inode *inode) | |
1619 | { | |
1620 | wait_queue_head_t *wq; | |
eaff8079 CH |
1621 | DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW); |
1622 | wq = bit_waitqueue(&inode->i_state, __I_NEW); | |
1da177e4 LT |
1623 | prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE); |
1624 | spin_unlock(&inode_lock); | |
1625 | schedule(); | |
1626 | finish_wait(wq, &wait.wait); | |
1627 | spin_lock(&inode_lock); | |
1628 | } | |
1629 | ||
1da177e4 LT |
1630 | static __initdata unsigned long ihash_entries; |
1631 | static int __init set_ihash_entries(char *str) | |
1632 | { | |
1633 | if (!str) | |
1634 | return 0; | |
1635 | ihash_entries = simple_strtoul(str, &str, 0); | |
1636 | return 1; | |
1637 | } | |
1638 | __setup("ihash_entries=", set_ihash_entries); | |
1639 | ||
1640 | /* | |
1641 | * Initialize the waitqueues and inode hash table. | |
1642 | */ | |
1643 | void __init inode_init_early(void) | |
1644 | { | |
1645 | int loop; | |
1646 | ||
1647 | /* If hashes are distributed across NUMA nodes, defer | |
1648 | * hash allocation until vmalloc space is available. | |
1649 | */ | |
1650 | if (hashdist) | |
1651 | return; | |
1652 | ||
1653 | inode_hashtable = | |
1654 | alloc_large_system_hash("Inode-cache", | |
1655 | sizeof(struct hlist_head), | |
1656 | ihash_entries, | |
1657 | 14, | |
1658 | HASH_EARLY, | |
1659 | &i_hash_shift, | |
1660 | &i_hash_mask, | |
1661 | 0); | |
1662 | ||
1663 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1664 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1665 | } | |
1666 | ||
74bf17cf | 1667 | void __init inode_init(void) |
1da177e4 LT |
1668 | { |
1669 | int loop; | |
1670 | ||
1671 | /* inode slab cache */ | |
b0196009 PJ |
1672 | inode_cachep = kmem_cache_create("inode_cache", |
1673 | sizeof(struct inode), | |
1674 | 0, | |
1675 | (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC| | |
1676 | SLAB_MEM_SPREAD), | |
20c2df83 | 1677 | init_once); |
8e1f936b | 1678 | register_shrinker(&icache_shrinker); |
1da177e4 LT |
1679 | |
1680 | /* Hash may have been set up in inode_init_early */ | |
1681 | if (!hashdist) | |
1682 | return; | |
1683 | ||
1684 | inode_hashtable = | |
1685 | alloc_large_system_hash("Inode-cache", | |
1686 | sizeof(struct hlist_head), | |
1687 | ihash_entries, | |
1688 | 14, | |
1689 | 0, | |
1690 | &i_hash_shift, | |
1691 | &i_hash_mask, | |
1692 | 0); | |
1693 | ||
1694 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1695 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1696 | } | |
1697 | ||
1698 | void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev) | |
1699 | { | |
1700 | inode->i_mode = mode; | |
1701 | if (S_ISCHR(mode)) { | |
1702 | inode->i_fop = &def_chr_fops; | |
1703 | inode->i_rdev = rdev; | |
1704 | } else if (S_ISBLK(mode)) { | |
1705 | inode->i_fop = &def_blk_fops; | |
1706 | inode->i_rdev = rdev; | |
1707 | } else if (S_ISFIFO(mode)) | |
1708 | inode->i_fop = &def_fifo_fops; | |
1709 | else if (S_ISSOCK(mode)) | |
1710 | inode->i_fop = &bad_sock_fops; | |
1711 | else | |
af0d9ae8 MK |
1712 | printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for" |
1713 | " inode %s:%lu\n", mode, inode->i_sb->s_id, | |
1714 | inode->i_ino); | |
1da177e4 LT |
1715 | } |
1716 | EXPORT_SYMBOL(init_special_inode); | |
a1bd120d DM |
1717 | |
1718 | /** | |
eaae668d | 1719 | * inode_init_owner - Init uid,gid,mode for new inode according to posix standards |
a1bd120d DM |
1720 | * @inode: New inode |
1721 | * @dir: Directory inode | |
1722 | * @mode: mode of the new inode | |
1723 | */ | |
1724 | void inode_init_owner(struct inode *inode, const struct inode *dir, | |
1725 | mode_t mode) | |
1726 | { | |
1727 | inode->i_uid = current_fsuid(); | |
1728 | if (dir && dir->i_mode & S_ISGID) { | |
1729 | inode->i_gid = dir->i_gid; | |
1730 | if (S_ISDIR(mode)) | |
1731 | mode |= S_ISGID; | |
1732 | } else | |
1733 | inode->i_gid = current_fsgid(); | |
1734 | inode->i_mode = mode; | |
1735 | } | |
1736 | EXPORT_SYMBOL(inode_init_owner); | |
e795b717 | 1737 | |
2e149670 SH |
1738 | /** |
1739 | * inode_owner_or_capable - check current task permissions to inode | |
1740 | * @inode: inode being checked | |
1741 | * | |
1742 | * Return true if current either has CAP_FOWNER to the inode, or | |
1743 | * owns the file. | |
e795b717 | 1744 | */ |
2e149670 | 1745 | bool inode_owner_or_capable(const struct inode *inode) |
e795b717 SH |
1746 | { |
1747 | struct user_namespace *ns = inode_userns(inode); | |
1748 | ||
1749 | if (current_user_ns() == ns && current_fsuid() == inode->i_uid) | |
1750 | return true; | |
1751 | if (ns_capable(ns, CAP_FOWNER)) | |
1752 | return true; | |
1753 | return false; | |
1754 | } | |
2e149670 | 1755 | EXPORT_SYMBOL(inode_owner_or_capable); |