]>
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 | 29 | |
250df6ed DC |
30 | /* |
31 | * inode locking rules. | |
32 | * | |
33 | * inode->i_lock protects: | |
34 | * inode->i_state, inode->i_hash, __iget() | |
35 | * | |
36 | * Lock ordering: | |
37 | * inode_lock | |
38 | * inode->i_lock | |
39 | */ | |
40 | ||
1da177e4 LT |
41 | /* |
42 | * This is needed for the following functions: | |
43 | * - inode_has_buffers | |
1da177e4 LT |
44 | * - invalidate_bdev |
45 | * | |
46 | * FIXME: remove all knowledge of the buffer layer from this file | |
47 | */ | |
48 | #include <linux/buffer_head.h> | |
49 | ||
50 | /* | |
51 | * New inode.c implementation. | |
52 | * | |
53 | * This implementation has the basic premise of trying | |
54 | * to be extremely low-overhead and SMP-safe, yet be | |
55 | * simple enough to be "obviously correct". | |
56 | * | |
57 | * Famous last words. | |
58 | */ | |
59 | ||
60 | /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */ | |
61 | ||
62 | /* #define INODE_PARANOIA 1 */ | |
63 | /* #define INODE_DEBUG 1 */ | |
64 | ||
65 | /* | |
66 | * Inode lookup is no longer as critical as it used to be: | |
67 | * most of the lookups are going to be through the dcache. | |
68 | */ | |
69 | #define I_HASHBITS i_hash_shift | |
70 | #define I_HASHMASK i_hash_mask | |
71 | ||
fa3536cc ED |
72 | static unsigned int i_hash_mask __read_mostly; |
73 | static unsigned int i_hash_shift __read_mostly; | |
1da177e4 LT |
74 | |
75 | /* | |
76 | * Each inode can be on two separate lists. One is | |
77 | * the hash list of the inode, used for lookups. The | |
78 | * other linked list is the "type" list: | |
79 | * "in_use" - valid inode, i_count > 0, i_nlink > 0 | |
80 | * "dirty" - as "in_use" but also dirty | |
81 | * "unused" - valid inode, i_count = 0 | |
82 | * | |
83 | * A "dirty" list is maintained for each super block, | |
84 | * allowing for low-overhead inode sync() operations. | |
85 | */ | |
86 | ||
7ccf19a8 | 87 | static LIST_HEAD(inode_lru); |
fa3536cc | 88 | static struct hlist_head *inode_hashtable __read_mostly; |
1da177e4 LT |
89 | |
90 | /* | |
91 | * A simple spinlock to protect the list manipulations. | |
92 | * | |
93 | * NOTE! You also have to own the lock if you change | |
94 | * the i_state of an inode while it is in use.. | |
95 | */ | |
96 | DEFINE_SPINLOCK(inode_lock); | |
97 | ||
98 | /* | |
bab1d944 CH |
99 | * iprune_sem provides exclusion between the icache shrinking and the |
100 | * umount path. | |
88e0fbc4 | 101 | * |
bab1d944 CH |
102 | * We don't actually need it to protect anything in the umount path, |
103 | * but only need to cycle through it to make sure any inode that | |
104 | * prune_icache took off the LRU list has been fully torn down by the | |
105 | * time we are past evict_inodes. | |
1da177e4 | 106 | */ |
88e0fbc4 | 107 | static DECLARE_RWSEM(iprune_sem); |
1da177e4 LT |
108 | |
109 | /* | |
110 | * Statistics gathering.. | |
111 | */ | |
112 | struct inodes_stat_t inodes_stat; | |
113 | ||
3e880fb5 | 114 | static DEFINE_PER_CPU(unsigned int, nr_inodes); |
cffbc8aa | 115 | |
6b3304b5 | 116 | static struct kmem_cache *inode_cachep __read_mostly; |
1da177e4 | 117 | |
3e880fb5 | 118 | static int get_nr_inodes(void) |
cffbc8aa | 119 | { |
3e880fb5 NP |
120 | int i; |
121 | int sum = 0; | |
122 | for_each_possible_cpu(i) | |
123 | sum += per_cpu(nr_inodes, i); | |
124 | return sum < 0 ? 0 : sum; | |
cffbc8aa DC |
125 | } |
126 | ||
127 | static inline int get_nr_inodes_unused(void) | |
128 | { | |
86c8749e | 129 | return inodes_stat.nr_unused; |
cffbc8aa DC |
130 | } |
131 | ||
132 | int get_nr_dirty_inodes(void) | |
133 | { | |
3e880fb5 | 134 | /* not actually dirty inodes, but a wild approximation */ |
cffbc8aa DC |
135 | int nr_dirty = get_nr_inodes() - get_nr_inodes_unused(); |
136 | return nr_dirty > 0 ? nr_dirty : 0; | |
cffbc8aa DC |
137 | } |
138 | ||
139 | /* | |
140 | * Handle nr_inode sysctl | |
141 | */ | |
142 | #ifdef CONFIG_SYSCTL | |
143 | int proc_nr_inodes(ctl_table *table, int write, | |
144 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
145 | { | |
146 | inodes_stat.nr_inodes = get_nr_inodes(); | |
cffbc8aa DC |
147 | return proc_dointvec(table, write, buffer, lenp, ppos); |
148 | } | |
149 | #endif | |
150 | ||
2cb1599f DC |
151 | /** |
152 | * inode_init_always - perform inode structure intialisation | |
0bc02f3f RD |
153 | * @sb: superblock inode belongs to |
154 | * @inode: inode to initialise | |
2cb1599f DC |
155 | * |
156 | * These are initializations that need to be done on every inode | |
157 | * allocation as the fields are not initialised by slab allocation. | |
158 | */ | |
54e34621 | 159 | int inode_init_always(struct super_block *sb, struct inode *inode) |
1da177e4 | 160 | { |
f5e54d6e | 161 | static const struct address_space_operations empty_aops; |
6e1d5dcc | 162 | static const struct inode_operations empty_iops; |
99ac48f5 | 163 | static const struct file_operations empty_fops; |
6b3304b5 | 164 | struct address_space *const mapping = &inode->i_data; |
2cb1599f DC |
165 | |
166 | inode->i_sb = sb; | |
167 | inode->i_blkbits = sb->s_blocksize_bits; | |
168 | inode->i_flags = 0; | |
169 | atomic_set(&inode->i_count, 1); | |
170 | inode->i_op = &empty_iops; | |
171 | inode->i_fop = &empty_fops; | |
172 | inode->i_nlink = 1; | |
56ff5efa AV |
173 | inode->i_uid = 0; |
174 | inode->i_gid = 0; | |
2cb1599f DC |
175 | atomic_set(&inode->i_writecount, 0); |
176 | inode->i_size = 0; | |
177 | inode->i_blocks = 0; | |
178 | inode->i_bytes = 0; | |
179 | inode->i_generation = 0; | |
1da177e4 | 180 | #ifdef CONFIG_QUOTA |
2cb1599f | 181 | memset(&inode->i_dquot, 0, sizeof(inode->i_dquot)); |
1da177e4 | 182 | #endif |
2cb1599f DC |
183 | inode->i_pipe = NULL; |
184 | inode->i_bdev = NULL; | |
185 | inode->i_cdev = NULL; | |
186 | inode->i_rdev = 0; | |
187 | inode->dirtied_when = 0; | |
6146f0d5 MZ |
188 | |
189 | if (security_inode_alloc(inode)) | |
54e34621 | 190 | goto out; |
2cb1599f DC |
191 | spin_lock_init(&inode->i_lock); |
192 | lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key); | |
193 | ||
194 | mutex_init(&inode->i_mutex); | |
195 | lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key); | |
196 | ||
197 | init_rwsem(&inode->i_alloc_sem); | |
198 | lockdep_set_class(&inode->i_alloc_sem, &sb->s_type->i_alloc_sem_key); | |
199 | ||
200 | mapping->a_ops = &empty_aops; | |
201 | mapping->host = inode; | |
202 | mapping->flags = 0; | |
3c1d4378 | 203 | mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE); |
2cb1599f DC |
204 | mapping->assoc_mapping = NULL; |
205 | mapping->backing_dev_info = &default_backing_dev_info; | |
206 | mapping->writeback_index = 0; | |
207 | ||
208 | /* | |
209 | * If the block_device provides a backing_dev_info for client | |
210 | * inodes then use that. Otherwise the inode share the bdev's | |
211 | * backing_dev_info. | |
212 | */ | |
213 | if (sb->s_bdev) { | |
214 | struct backing_dev_info *bdi; | |
215 | ||
2c96ce9f | 216 | bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info; |
2cb1599f DC |
217 | mapping->backing_dev_info = bdi; |
218 | } | |
219 | inode->i_private = NULL; | |
220 | inode->i_mapping = mapping; | |
f19d4a8f AV |
221 | #ifdef CONFIG_FS_POSIX_ACL |
222 | inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED; | |
223 | #endif | |
2cb1599f | 224 | |
3be25f49 EP |
225 | #ifdef CONFIG_FSNOTIFY |
226 | inode->i_fsnotify_mask = 0; | |
227 | #endif | |
228 | ||
3e880fb5 | 229 | this_cpu_inc(nr_inodes); |
cffbc8aa | 230 | |
54e34621 | 231 | return 0; |
54e34621 CH |
232 | out: |
233 | return -ENOMEM; | |
1da177e4 | 234 | } |
2cb1599f DC |
235 | EXPORT_SYMBOL(inode_init_always); |
236 | ||
237 | static struct inode *alloc_inode(struct super_block *sb) | |
238 | { | |
239 | struct inode *inode; | |
240 | ||
241 | if (sb->s_op->alloc_inode) | |
242 | inode = sb->s_op->alloc_inode(sb); | |
243 | else | |
244 | inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL); | |
245 | ||
54e34621 CH |
246 | if (!inode) |
247 | return NULL; | |
248 | ||
249 | if (unlikely(inode_init_always(sb, inode))) { | |
250 | if (inode->i_sb->s_op->destroy_inode) | |
251 | inode->i_sb->s_op->destroy_inode(inode); | |
252 | else | |
253 | kmem_cache_free(inode_cachep, inode); | |
254 | return NULL; | |
255 | } | |
256 | ||
257 | return inode; | |
2cb1599f | 258 | } |
1da177e4 | 259 | |
ff0c7d15 NP |
260 | void free_inode_nonrcu(struct inode *inode) |
261 | { | |
262 | kmem_cache_free(inode_cachep, inode); | |
263 | } | |
264 | EXPORT_SYMBOL(free_inode_nonrcu); | |
265 | ||
2e00c97e | 266 | void __destroy_inode(struct inode *inode) |
1da177e4 | 267 | { |
b7542f8c | 268 | BUG_ON(inode_has_buffers(inode)); |
1da177e4 | 269 | security_inode_free(inode); |
3be25f49 | 270 | fsnotify_inode_delete(inode); |
f19d4a8f AV |
271 | #ifdef CONFIG_FS_POSIX_ACL |
272 | if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED) | |
273 | posix_acl_release(inode->i_acl); | |
274 | if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED) | |
275 | posix_acl_release(inode->i_default_acl); | |
276 | #endif | |
3e880fb5 | 277 | this_cpu_dec(nr_inodes); |
2e00c97e CH |
278 | } |
279 | EXPORT_SYMBOL(__destroy_inode); | |
280 | ||
fa0d7e3d NP |
281 | static void i_callback(struct rcu_head *head) |
282 | { | |
283 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
284 | INIT_LIST_HEAD(&inode->i_dentry); | |
285 | kmem_cache_free(inode_cachep, inode); | |
286 | } | |
287 | ||
56b0dacf | 288 | static void destroy_inode(struct inode *inode) |
2e00c97e | 289 | { |
7ccf19a8 | 290 | BUG_ON(!list_empty(&inode->i_lru)); |
2e00c97e | 291 | __destroy_inode(inode); |
1da177e4 LT |
292 | if (inode->i_sb->s_op->destroy_inode) |
293 | inode->i_sb->s_op->destroy_inode(inode); | |
294 | else | |
fa0d7e3d | 295 | call_rcu(&inode->i_rcu, i_callback); |
1da177e4 | 296 | } |
1da177e4 | 297 | |
2aa15890 MS |
298 | void address_space_init_once(struct address_space *mapping) |
299 | { | |
300 | memset(mapping, 0, sizeof(*mapping)); | |
301 | INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC); | |
302 | spin_lock_init(&mapping->tree_lock); | |
303 | spin_lock_init(&mapping->i_mmap_lock); | |
304 | INIT_LIST_HEAD(&mapping->private_list); | |
305 | spin_lock_init(&mapping->private_lock); | |
306 | INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap); | |
307 | INIT_LIST_HEAD(&mapping->i_mmap_nonlinear); | |
308 | mutex_init(&mapping->unmap_mutex); | |
309 | } | |
310 | EXPORT_SYMBOL(address_space_init_once); | |
311 | ||
1da177e4 LT |
312 | /* |
313 | * These are initializations that only need to be done | |
314 | * once, because the fields are idempotent across use | |
315 | * of the inode, so let the slab aware of that. | |
316 | */ | |
317 | void inode_init_once(struct inode *inode) | |
318 | { | |
319 | memset(inode, 0, sizeof(*inode)); | |
320 | INIT_HLIST_NODE(&inode->i_hash); | |
321 | INIT_LIST_HEAD(&inode->i_dentry); | |
322 | INIT_LIST_HEAD(&inode->i_devices); | |
7ccf19a8 NP |
323 | INIT_LIST_HEAD(&inode->i_wb_list); |
324 | INIT_LIST_HEAD(&inode->i_lru); | |
2aa15890 | 325 | address_space_init_once(&inode->i_data); |
1da177e4 | 326 | i_size_ordered_init(inode); |
3be25f49 | 327 | #ifdef CONFIG_FSNOTIFY |
e61ce867 | 328 | INIT_HLIST_HEAD(&inode->i_fsnotify_marks); |
3be25f49 | 329 | #endif |
1da177e4 | 330 | } |
1da177e4 LT |
331 | EXPORT_SYMBOL(inode_init_once); |
332 | ||
51cc5068 | 333 | static void init_once(void *foo) |
1da177e4 | 334 | { |
6b3304b5 | 335 | struct inode *inode = (struct inode *) foo; |
1da177e4 | 336 | |
a35afb83 | 337 | inode_init_once(inode); |
1da177e4 LT |
338 | } |
339 | ||
340 | /* | |
250df6ed | 341 | * inode->i_lock must be held |
1da177e4 | 342 | */ |
6b3304b5 | 343 | void __iget(struct inode *inode) |
1da177e4 | 344 | { |
9e38d86f NP |
345 | atomic_inc(&inode->i_count); |
346 | } | |
2e147f1e | 347 | |
7de9c6ee AV |
348 | /* |
349 | * get additional reference to inode; caller must already hold one. | |
350 | */ | |
351 | void ihold(struct inode *inode) | |
352 | { | |
353 | WARN_ON(atomic_inc_return(&inode->i_count) < 2); | |
354 | } | |
355 | EXPORT_SYMBOL(ihold); | |
356 | ||
9e38d86f NP |
357 | static void inode_lru_list_add(struct inode *inode) |
358 | { | |
7ccf19a8 NP |
359 | if (list_empty(&inode->i_lru)) { |
360 | list_add(&inode->i_lru, &inode_lru); | |
86c8749e | 361 | inodes_stat.nr_unused++; |
9e38d86f NP |
362 | } |
363 | } | |
2e147f1e | 364 | |
9e38d86f NP |
365 | static void inode_lru_list_del(struct inode *inode) |
366 | { | |
7ccf19a8 NP |
367 | if (!list_empty(&inode->i_lru)) { |
368 | list_del_init(&inode->i_lru); | |
86c8749e | 369 | inodes_stat.nr_unused--; |
9e38d86f | 370 | } |
1da177e4 LT |
371 | } |
372 | ||
646ec461 CH |
373 | static inline void __inode_sb_list_add(struct inode *inode) |
374 | { | |
375 | list_add(&inode->i_sb_list, &inode->i_sb->s_inodes); | |
1da177e4 LT |
376 | } |
377 | ||
646ec461 CH |
378 | /** |
379 | * inode_sb_list_add - add inode to the superblock list of inodes | |
380 | * @inode: inode to add | |
381 | */ | |
382 | void inode_sb_list_add(struct inode *inode) | |
383 | { | |
384 | spin_lock(&inode_lock); | |
385 | __inode_sb_list_add(inode); | |
386 | spin_unlock(&inode_lock); | |
387 | } | |
388 | EXPORT_SYMBOL_GPL(inode_sb_list_add); | |
389 | ||
390 | static inline void __inode_sb_list_del(struct inode *inode) | |
391 | { | |
392 | list_del_init(&inode->i_sb_list); | |
393 | } | |
394 | ||
4c51acbc DC |
395 | static unsigned long hash(struct super_block *sb, unsigned long hashval) |
396 | { | |
397 | unsigned long tmp; | |
398 | ||
399 | tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) / | |
400 | L1_CACHE_BYTES; | |
401 | tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> I_HASHBITS); | |
402 | return tmp & I_HASHMASK; | |
403 | } | |
404 | ||
405 | /** | |
406 | * __insert_inode_hash - hash an inode | |
407 | * @inode: unhashed inode | |
408 | * @hashval: unsigned long value used to locate this object in the | |
409 | * inode_hashtable. | |
410 | * | |
411 | * Add an inode to the inode hash for this superblock. | |
412 | */ | |
413 | void __insert_inode_hash(struct inode *inode, unsigned long hashval) | |
414 | { | |
646ec461 CH |
415 | struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval); |
416 | ||
4c51acbc | 417 | spin_lock(&inode_lock); |
250df6ed | 418 | spin_lock(&inode->i_lock); |
646ec461 | 419 | hlist_add_head(&inode->i_hash, b); |
250df6ed | 420 | spin_unlock(&inode->i_lock); |
4c51acbc DC |
421 | spin_unlock(&inode_lock); |
422 | } | |
423 | EXPORT_SYMBOL(__insert_inode_hash); | |
424 | ||
425 | /** | |
426 | * __remove_inode_hash - remove an inode from the hash | |
427 | * @inode: inode to unhash | |
428 | * | |
429 | * Remove an inode from the superblock. | |
430 | */ | |
431 | static void __remove_inode_hash(struct inode *inode) | |
432 | { | |
433 | hlist_del_init(&inode->i_hash); | |
434 | } | |
435 | ||
436 | /** | |
437 | * remove_inode_hash - remove an inode from the hash | |
438 | * @inode: inode to unhash | |
439 | * | |
440 | * Remove an inode from the superblock. | |
441 | */ | |
442 | void remove_inode_hash(struct inode *inode) | |
443 | { | |
444 | spin_lock(&inode_lock); | |
250df6ed | 445 | spin_lock(&inode->i_lock); |
4c51acbc | 446 | hlist_del_init(&inode->i_hash); |
250df6ed | 447 | spin_unlock(&inode->i_lock); |
4c51acbc DC |
448 | spin_unlock(&inode_lock); |
449 | } | |
450 | EXPORT_SYMBOL(remove_inode_hash); | |
451 | ||
b0683aa6 AV |
452 | void end_writeback(struct inode *inode) |
453 | { | |
454 | might_sleep(); | |
455 | BUG_ON(inode->i_data.nrpages); | |
456 | BUG_ON(!list_empty(&inode->i_data.private_list)); | |
457 | BUG_ON(!(inode->i_state & I_FREEING)); | |
458 | BUG_ON(inode->i_state & I_CLEAR); | |
459 | inode_sync_wait(inode); | |
fa0d7e3d | 460 | /* don't need i_lock here, no concurrent mods to i_state */ |
b0683aa6 AV |
461 | inode->i_state = I_FREEING | I_CLEAR; |
462 | } | |
463 | EXPORT_SYMBOL(end_writeback); | |
464 | ||
644da596 | 465 | static void evict(struct inode *inode) |
b4272d4c AV |
466 | { |
467 | const struct super_operations *op = inode->i_sb->s_op; | |
468 | ||
be7ce416 AV |
469 | if (op->evict_inode) { |
470 | op->evict_inode(inode); | |
b4272d4c AV |
471 | } else { |
472 | if (inode->i_data.nrpages) | |
473 | truncate_inode_pages(&inode->i_data, 0); | |
30140837 | 474 | end_writeback(inode); |
b4272d4c | 475 | } |
661074e9 AV |
476 | if (S_ISBLK(inode->i_mode) && inode->i_bdev) |
477 | bd_forget(inode); | |
478 | if (S_ISCHR(inode->i_mode) && inode->i_cdev) | |
479 | cd_forget(inode); | |
b4272d4c AV |
480 | } |
481 | ||
1da177e4 LT |
482 | /* |
483 | * dispose_list - dispose of the contents of a local list | |
484 | * @head: the head of the list to free | |
485 | * | |
486 | * Dispose-list gets a local list with local inodes in it, so it doesn't | |
487 | * need to worry about list corruption and SMP locks. | |
488 | */ | |
489 | static void dispose_list(struct list_head *head) | |
490 | { | |
1da177e4 LT |
491 | while (!list_empty(head)) { |
492 | struct inode *inode; | |
493 | ||
7ccf19a8 NP |
494 | inode = list_first_entry(head, struct inode, i_lru); |
495 | list_del_init(&inode->i_lru); | |
1da177e4 | 496 | |
644da596 | 497 | evict(inode); |
4120db47 AB |
498 | |
499 | spin_lock(&inode_lock); | |
4c51acbc | 500 | __remove_inode_hash(inode); |
646ec461 | 501 | __inode_sb_list_del(inode); |
4120db47 AB |
502 | spin_unlock(&inode_lock); |
503 | ||
250df6ed DC |
504 | spin_lock(&inode->i_lock); |
505 | wake_up_bit(&inode->i_state, __I_NEW); | |
506 | spin_unlock(&inode->i_lock); | |
1da177e4 | 507 | destroy_inode(inode); |
1da177e4 | 508 | } |
1da177e4 LT |
509 | } |
510 | ||
63997e98 AV |
511 | /** |
512 | * evict_inodes - evict all evictable inodes for a superblock | |
513 | * @sb: superblock to operate on | |
514 | * | |
515 | * Make sure that no inodes with zero refcount are retained. This is | |
516 | * called by superblock shutdown after having MS_ACTIVE flag removed, | |
517 | * so any inode reaching zero refcount during or after that call will | |
518 | * be immediately evicted. | |
1da177e4 | 519 | */ |
63997e98 | 520 | void evict_inodes(struct super_block *sb) |
1da177e4 | 521 | { |
63997e98 AV |
522 | struct inode *inode, *next; |
523 | LIST_HEAD(dispose); | |
1da177e4 | 524 | |
63997e98 AV |
525 | spin_lock(&inode_lock); |
526 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { | |
527 | if (atomic_read(&inode->i_count)) | |
aabb8fdb | 528 | continue; |
250df6ed DC |
529 | |
530 | spin_lock(&inode->i_lock); | |
531 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
532 | spin_unlock(&inode->i_lock); | |
1da177e4 | 533 | continue; |
250df6ed | 534 | } |
63997e98 AV |
535 | |
536 | inode->i_state |= I_FREEING; | |
250df6ed DC |
537 | if (!(inode->i_state & (I_DIRTY | I_SYNC))) |
538 | inodes_stat.nr_unused--; | |
539 | spin_unlock(&inode->i_lock); | |
63997e98 AV |
540 | |
541 | /* | |
542 | * Move the inode off the IO lists and LRU once I_FREEING is | |
543 | * set so that it won't get moved back on there if it is dirty. | |
544 | */ | |
545 | list_move(&inode->i_lru, &dispose); | |
546 | list_del_init(&inode->i_wb_list); | |
1da177e4 | 547 | } |
63997e98 AV |
548 | spin_unlock(&inode_lock); |
549 | ||
550 | dispose_list(&dispose); | |
bab1d944 CH |
551 | |
552 | /* | |
553 | * Cycle through iprune_sem to make sure any inode that prune_icache | |
554 | * moved off the list before we took the lock has been fully torn | |
555 | * down. | |
556 | */ | |
557 | down_write(&iprune_sem); | |
63997e98 | 558 | up_write(&iprune_sem); |
1da177e4 LT |
559 | } |
560 | ||
1da177e4 | 561 | /** |
a0318786 CH |
562 | * invalidate_inodes - attempt to free all inodes on a superblock |
563 | * @sb: superblock to operate on | |
93b270f7 | 564 | * @kill_dirty: flag to guide handling of dirty inodes |
1da177e4 | 565 | * |
a0318786 CH |
566 | * Attempts to free all inodes for a given superblock. If there were any |
567 | * busy inodes return a non-zero value, else zero. | |
93b270f7 N |
568 | * If @kill_dirty is set, discard dirty inodes too, otherwise treat |
569 | * them as busy. | |
1da177e4 | 570 | */ |
93b270f7 | 571 | int invalidate_inodes(struct super_block *sb, bool kill_dirty) |
1da177e4 | 572 | { |
cffbc8aa | 573 | int busy = 0; |
a0318786 CH |
574 | struct inode *inode, *next; |
575 | LIST_HEAD(dispose); | |
1da177e4 | 576 | |
1da177e4 | 577 | spin_lock(&inode_lock); |
a0318786 | 578 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
250df6ed DC |
579 | spin_lock(&inode->i_lock); |
580 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
581 | spin_unlock(&inode->i_lock); | |
aabb8fdb | 582 | continue; |
250df6ed | 583 | } |
93b270f7 | 584 | if (inode->i_state & I_DIRTY && !kill_dirty) { |
250df6ed | 585 | spin_unlock(&inode->i_lock); |
93b270f7 N |
586 | busy = 1; |
587 | continue; | |
588 | } | |
99a38919 | 589 | if (atomic_read(&inode->i_count)) { |
250df6ed | 590 | spin_unlock(&inode->i_lock); |
99a38919 | 591 | busy = 1; |
1da177e4 LT |
592 | continue; |
593 | } | |
99a38919 | 594 | |
99a38919 | 595 | inode->i_state |= I_FREEING; |
250df6ed DC |
596 | if (!(inode->i_state & (I_DIRTY | I_SYNC))) |
597 | inodes_stat.nr_unused--; | |
598 | spin_unlock(&inode->i_lock); | |
7ccf19a8 NP |
599 | |
600 | /* | |
601 | * Move the inode off the IO lists and LRU once I_FREEING is | |
602 | * set so that it won't get moved back on there if it is dirty. | |
603 | */ | |
a0318786 | 604 | list_move(&inode->i_lru, &dispose); |
7ccf19a8 | 605 | list_del_init(&inode->i_wb_list); |
1da177e4 | 606 | } |
1da177e4 LT |
607 | spin_unlock(&inode_lock); |
608 | ||
a0318786 | 609 | dispose_list(&dispose); |
1da177e4 LT |
610 | |
611 | return busy; | |
612 | } | |
1da177e4 LT |
613 | |
614 | static int can_unuse(struct inode *inode) | |
615 | { | |
9e38d86f | 616 | if (inode->i_state & ~I_REFERENCED) |
1da177e4 LT |
617 | return 0; |
618 | if (inode_has_buffers(inode)) | |
619 | return 0; | |
620 | if (atomic_read(&inode->i_count)) | |
621 | return 0; | |
622 | if (inode->i_data.nrpages) | |
623 | return 0; | |
624 | return 1; | |
625 | } | |
626 | ||
627 | /* | |
9e38d86f NP |
628 | * Scan `goal' inodes on the unused list for freeable ones. They are moved to a |
629 | * temporary list and then are freed outside inode_lock by dispose_list(). | |
1da177e4 LT |
630 | * |
631 | * Any inodes which are pinned purely because of attached pagecache have their | |
9e38d86f NP |
632 | * pagecache removed. If the inode has metadata buffers attached to |
633 | * mapping->private_list then try to remove them. | |
1da177e4 | 634 | * |
9e38d86f NP |
635 | * If the inode has the I_REFERENCED flag set, then it means that it has been |
636 | * used recently - the flag is set in iput_final(). When we encounter such an | |
637 | * inode, clear the flag and move it to the back of the LRU so it gets another | |
638 | * pass through the LRU before it gets reclaimed. This is necessary because of | |
639 | * the fact we are doing lazy LRU updates to minimise lock contention so the | |
640 | * LRU does not have strict ordering. Hence we don't want to reclaim inodes | |
641 | * with this flag set because they are the inodes that are out of order. | |
1da177e4 LT |
642 | */ |
643 | static void prune_icache(int nr_to_scan) | |
644 | { | |
645 | LIST_HEAD(freeable); | |
1da177e4 LT |
646 | int nr_scanned; |
647 | unsigned long reap = 0; | |
648 | ||
88e0fbc4 | 649 | down_read(&iprune_sem); |
1da177e4 LT |
650 | spin_lock(&inode_lock); |
651 | for (nr_scanned = 0; nr_scanned < nr_to_scan; nr_scanned++) { | |
652 | struct inode *inode; | |
653 | ||
7ccf19a8 | 654 | if (list_empty(&inode_lru)) |
1da177e4 LT |
655 | break; |
656 | ||
7ccf19a8 | 657 | inode = list_entry(inode_lru.prev, struct inode, i_lru); |
1da177e4 | 658 | |
9e38d86f NP |
659 | /* |
660 | * Referenced or dirty inodes are still in use. Give them | |
661 | * another pass through the LRU as we canot reclaim them now. | |
662 | */ | |
250df6ed | 663 | spin_lock(&inode->i_lock); |
9e38d86f NP |
664 | if (atomic_read(&inode->i_count) || |
665 | (inode->i_state & ~I_REFERENCED)) { | |
250df6ed | 666 | spin_unlock(&inode->i_lock); |
7ccf19a8 | 667 | list_del_init(&inode->i_lru); |
86c8749e | 668 | inodes_stat.nr_unused--; |
9e38d86f NP |
669 | continue; |
670 | } | |
671 | ||
672 | /* recently referenced inodes get one more pass */ | |
673 | if (inode->i_state & I_REFERENCED) { | |
9e38d86f | 674 | inode->i_state &= ~I_REFERENCED; |
250df6ed DC |
675 | spin_unlock(&inode->i_lock); |
676 | list_move(&inode->i_lru, &inode_lru); | |
1da177e4 LT |
677 | continue; |
678 | } | |
679 | if (inode_has_buffers(inode) || inode->i_data.nrpages) { | |
680 | __iget(inode); | |
250df6ed | 681 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
682 | spin_unlock(&inode_lock); |
683 | if (remove_inode_buffers(inode)) | |
fc0ecff6 AM |
684 | reap += invalidate_mapping_pages(&inode->i_data, |
685 | 0, -1); | |
1da177e4 LT |
686 | iput(inode); |
687 | spin_lock(&inode_lock); | |
688 | ||
7ccf19a8 NP |
689 | if (inode != list_entry(inode_lru.next, |
690 | struct inode, i_lru)) | |
1da177e4 | 691 | continue; /* wrong inode or list_empty */ |
250df6ed DC |
692 | spin_lock(&inode->i_lock); |
693 | if (!can_unuse(inode)) { | |
694 | spin_unlock(&inode->i_lock); | |
1da177e4 | 695 | continue; |
250df6ed | 696 | } |
1da177e4 | 697 | } |
7ef0d737 | 698 | WARN_ON(inode->i_state & I_NEW); |
1da177e4 | 699 | inode->i_state |= I_FREEING; |
250df6ed | 700 | spin_unlock(&inode->i_lock); |
7ccf19a8 NP |
701 | |
702 | /* | |
703 | * Move the inode off the IO lists and LRU once I_FREEING is | |
704 | * set so that it won't get moved back on there if it is dirty. | |
705 | */ | |
706 | list_move(&inode->i_lru, &freeable); | |
707 | list_del_init(&inode->i_wb_list); | |
86c8749e | 708 | inodes_stat.nr_unused--; |
1da177e4 | 709 | } |
f8891e5e CL |
710 | if (current_is_kswapd()) |
711 | __count_vm_events(KSWAPD_INODESTEAL, reap); | |
712 | else | |
713 | __count_vm_events(PGINODESTEAL, reap); | |
1da177e4 LT |
714 | spin_unlock(&inode_lock); |
715 | ||
716 | dispose_list(&freeable); | |
88e0fbc4 | 717 | up_read(&iprune_sem); |
1da177e4 LT |
718 | } |
719 | ||
720 | /* | |
721 | * shrink_icache_memory() will attempt to reclaim some unused inodes. Here, | |
722 | * "unused" means that no dentries are referring to the inodes: the files are | |
723 | * not open and the dcache references to those inodes have already been | |
724 | * reclaimed. | |
725 | * | |
726 | * This function is passed the number of inodes to scan, and it returns the | |
727 | * total number of remaining possibly-reclaimable inodes. | |
728 | */ | |
7f8275d0 | 729 | static int shrink_icache_memory(struct shrinker *shrink, int nr, gfp_t gfp_mask) |
1da177e4 LT |
730 | { |
731 | if (nr) { | |
732 | /* | |
733 | * Nasty deadlock avoidance. We may hold various FS locks, | |
734 | * and we don't want to recurse into the FS that called us | |
735 | * in clear_inode() and friends.. | |
6b3304b5 | 736 | */ |
1da177e4 LT |
737 | if (!(gfp_mask & __GFP_FS)) |
738 | return -1; | |
739 | prune_icache(nr); | |
740 | } | |
cffbc8aa | 741 | return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure; |
1da177e4 LT |
742 | } |
743 | ||
8e1f936b RR |
744 | static struct shrinker icache_shrinker = { |
745 | .shrink = shrink_icache_memory, | |
746 | .seeks = DEFAULT_SEEKS, | |
747 | }; | |
748 | ||
1da177e4 LT |
749 | static void __wait_on_freeing_inode(struct inode *inode); |
750 | /* | |
751 | * Called with the inode lock held. | |
1da177e4 | 752 | */ |
6b3304b5 MK |
753 | static struct inode *find_inode(struct super_block *sb, |
754 | struct hlist_head *head, | |
755 | int (*test)(struct inode *, void *), | |
756 | void *data) | |
1da177e4 LT |
757 | { |
758 | struct hlist_node *node; | |
6b3304b5 | 759 | struct inode *inode = NULL; |
1da177e4 LT |
760 | |
761 | repeat: | |
c5c8be3c | 762 | hlist_for_each_entry(inode, node, head, i_hash) { |
1da177e4 LT |
763 | if (inode->i_sb != sb) |
764 | continue; | |
765 | if (!test(inode, data)) | |
766 | continue; | |
250df6ed | 767 | spin_lock(&inode->i_lock); |
a4ffdde6 | 768 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
769 | __wait_on_freeing_inode(inode); |
770 | goto repeat; | |
771 | } | |
f7899bd5 | 772 | __iget(inode); |
250df6ed | 773 | spin_unlock(&inode->i_lock); |
f7899bd5 | 774 | return inode; |
1da177e4 | 775 | } |
f7899bd5 | 776 | return NULL; |
1da177e4 LT |
777 | } |
778 | ||
779 | /* | |
780 | * find_inode_fast is the fast path version of find_inode, see the comment at | |
781 | * iget_locked for details. | |
782 | */ | |
6b3304b5 MK |
783 | static struct inode *find_inode_fast(struct super_block *sb, |
784 | struct hlist_head *head, unsigned long ino) | |
1da177e4 LT |
785 | { |
786 | struct hlist_node *node; | |
6b3304b5 | 787 | struct inode *inode = NULL; |
1da177e4 LT |
788 | |
789 | repeat: | |
c5c8be3c | 790 | hlist_for_each_entry(inode, node, head, i_hash) { |
1da177e4 LT |
791 | if (inode->i_ino != ino) |
792 | continue; | |
793 | if (inode->i_sb != sb) | |
794 | continue; | |
250df6ed | 795 | spin_lock(&inode->i_lock); |
a4ffdde6 | 796 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
797 | __wait_on_freeing_inode(inode); |
798 | goto repeat; | |
799 | } | |
f7899bd5 | 800 | __iget(inode); |
250df6ed | 801 | spin_unlock(&inode->i_lock); |
f7899bd5 | 802 | return inode; |
1da177e4 | 803 | } |
f7899bd5 | 804 | return NULL; |
8290c35f DC |
805 | } |
806 | ||
f991bd2e ED |
807 | /* |
808 | * Each cpu owns a range of LAST_INO_BATCH numbers. | |
809 | * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations, | |
810 | * to renew the exhausted range. | |
8290c35f | 811 | * |
f991bd2e ED |
812 | * This does not significantly increase overflow rate because every CPU can |
813 | * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is | |
814 | * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the | |
815 | * 2^32 range, and is a worst-case. Even a 50% wastage would only increase | |
816 | * overflow rate by 2x, which does not seem too significant. | |
817 | * | |
818 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
819 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
820 | * here to attempt to avoid that. | |
8290c35f | 821 | */ |
f991bd2e ED |
822 | #define LAST_INO_BATCH 1024 |
823 | static DEFINE_PER_CPU(unsigned int, last_ino); | |
824 | ||
85fe4025 | 825 | unsigned int get_next_ino(void) |
8290c35f | 826 | { |
f991bd2e ED |
827 | unsigned int *p = &get_cpu_var(last_ino); |
828 | unsigned int res = *p; | |
8290c35f | 829 | |
f991bd2e ED |
830 | #ifdef CONFIG_SMP |
831 | if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) { | |
832 | static atomic_t shared_last_ino; | |
833 | int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino); | |
834 | ||
835 | res = next - LAST_INO_BATCH; | |
836 | } | |
837 | #endif | |
838 | ||
839 | *p = ++res; | |
840 | put_cpu_var(last_ino); | |
841 | return res; | |
8290c35f | 842 | } |
85fe4025 | 843 | EXPORT_SYMBOL(get_next_ino); |
8290c35f | 844 | |
1da177e4 LT |
845 | /** |
846 | * new_inode - obtain an inode | |
847 | * @sb: superblock | |
848 | * | |
769848c0 | 849 | * Allocates a new inode for given superblock. The default gfp_mask |
3c1d4378 | 850 | * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE. |
769848c0 MG |
851 | * If HIGHMEM pages are unsuitable or it is known that pages allocated |
852 | * for the page cache are not reclaimable or migratable, | |
853 | * mapping_set_gfp_mask() must be called with suitable flags on the | |
854 | * newly created inode's mapping | |
855 | * | |
1da177e4 LT |
856 | */ |
857 | struct inode *new_inode(struct super_block *sb) | |
858 | { | |
6b3304b5 | 859 | struct inode *inode; |
1da177e4 LT |
860 | |
861 | spin_lock_prefetch(&inode_lock); | |
6b3304b5 | 862 | |
1da177e4 LT |
863 | inode = alloc_inode(sb); |
864 | if (inode) { | |
865 | spin_lock(&inode_lock); | |
250df6ed | 866 | spin_lock(&inode->i_lock); |
1da177e4 | 867 | inode->i_state = 0; |
250df6ed DC |
868 | spin_unlock(&inode->i_lock); |
869 | __inode_sb_list_add(inode); | |
1da177e4 LT |
870 | spin_unlock(&inode_lock); |
871 | } | |
872 | return inode; | |
873 | } | |
1da177e4 LT |
874 | EXPORT_SYMBOL(new_inode); |
875 | ||
250df6ed DC |
876 | /** |
877 | * unlock_new_inode - clear the I_NEW state and wake up any waiters | |
878 | * @inode: new inode to unlock | |
879 | * | |
880 | * Called when the inode is fully initialised to clear the new state of the | |
881 | * inode and wake up anyone waiting for the inode to finish initialisation. | |
882 | */ | |
1da177e4 LT |
883 | void unlock_new_inode(struct inode *inode) |
884 | { | |
14358e6d | 885 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
a3314a0e | 886 | if (S_ISDIR(inode->i_mode)) { |
1e89a5e1 PZ |
887 | struct file_system_type *type = inode->i_sb->s_type; |
888 | ||
9a7aa12f JK |
889 | /* Set new key only if filesystem hasn't already changed it */ |
890 | if (!lockdep_match_class(&inode->i_mutex, | |
891 | &type->i_mutex_key)) { | |
892 | /* | |
893 | * ensure nobody is actually holding i_mutex | |
894 | */ | |
895 | mutex_destroy(&inode->i_mutex); | |
896 | mutex_init(&inode->i_mutex); | |
897 | lockdep_set_class(&inode->i_mutex, | |
898 | &type->i_mutex_dir_key); | |
899 | } | |
1e89a5e1 | 900 | } |
14358e6d | 901 | #endif |
250df6ed | 902 | spin_lock(&inode->i_lock); |
eaff8079 CH |
903 | WARN_ON(!(inode->i_state & I_NEW)); |
904 | inode->i_state &= ~I_NEW; | |
250df6ed DC |
905 | wake_up_bit(&inode->i_state, __I_NEW); |
906 | spin_unlock(&inode->i_lock); | |
1da177e4 | 907 | } |
1da177e4 LT |
908 | EXPORT_SYMBOL(unlock_new_inode); |
909 | ||
910 | /* | |
911 | * This is called without the inode lock held.. Be careful. | |
912 | * | |
913 | * We no longer cache the sb_flags in i_flags - see fs.h | |
914 | * -- rmk@arm.uk.linux.org | |
915 | */ | |
6b3304b5 MK |
916 | static struct inode *get_new_inode(struct super_block *sb, |
917 | struct hlist_head *head, | |
918 | int (*test)(struct inode *, void *), | |
919 | int (*set)(struct inode *, void *), | |
920 | void *data) | |
1da177e4 | 921 | { |
6b3304b5 | 922 | struct inode *inode; |
1da177e4 LT |
923 | |
924 | inode = alloc_inode(sb); | |
925 | if (inode) { | |
6b3304b5 | 926 | struct inode *old; |
1da177e4 LT |
927 | |
928 | spin_lock(&inode_lock); | |
929 | /* We released the lock, so.. */ | |
930 | old = find_inode(sb, head, test, data); | |
931 | if (!old) { | |
932 | if (set(inode, data)) | |
933 | goto set_failed; | |
934 | ||
250df6ed DC |
935 | spin_lock(&inode->i_lock); |
936 | inode->i_state = I_NEW; | |
646ec461 | 937 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 938 | spin_unlock(&inode->i_lock); |
646ec461 | 939 | __inode_sb_list_add(inode); |
1da177e4 LT |
940 | spin_unlock(&inode_lock); |
941 | ||
942 | /* Return the locked inode with I_NEW set, the | |
943 | * caller is responsible for filling in the contents | |
944 | */ | |
945 | return inode; | |
946 | } | |
947 | ||
948 | /* | |
949 | * Uhhuh, somebody else created the same inode under | |
950 | * us. Use the old inode instead of the one we just | |
951 | * allocated. | |
952 | */ | |
1da177e4 LT |
953 | spin_unlock(&inode_lock); |
954 | destroy_inode(inode); | |
955 | inode = old; | |
956 | wait_on_inode(inode); | |
957 | } | |
958 | return inode; | |
959 | ||
960 | set_failed: | |
961 | spin_unlock(&inode_lock); | |
962 | destroy_inode(inode); | |
963 | return NULL; | |
964 | } | |
965 | ||
966 | /* | |
967 | * get_new_inode_fast is the fast path version of get_new_inode, see the | |
968 | * comment at iget_locked for details. | |
969 | */ | |
6b3304b5 MK |
970 | static struct inode *get_new_inode_fast(struct super_block *sb, |
971 | struct hlist_head *head, unsigned long ino) | |
1da177e4 | 972 | { |
6b3304b5 | 973 | struct inode *inode; |
1da177e4 LT |
974 | |
975 | inode = alloc_inode(sb); | |
976 | if (inode) { | |
6b3304b5 | 977 | struct inode *old; |
1da177e4 LT |
978 | |
979 | spin_lock(&inode_lock); | |
980 | /* We released the lock, so.. */ | |
981 | old = find_inode_fast(sb, head, ino); | |
982 | if (!old) { | |
983 | inode->i_ino = ino; | |
250df6ed DC |
984 | spin_lock(&inode->i_lock); |
985 | inode->i_state = I_NEW; | |
646ec461 | 986 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 987 | spin_unlock(&inode->i_lock); |
646ec461 | 988 | __inode_sb_list_add(inode); |
1da177e4 LT |
989 | spin_unlock(&inode_lock); |
990 | ||
991 | /* Return the locked inode with I_NEW set, the | |
992 | * caller is responsible for filling in the contents | |
993 | */ | |
994 | return inode; | |
995 | } | |
996 | ||
997 | /* | |
998 | * Uhhuh, somebody else created the same inode under | |
999 | * us. Use the old inode instead of the one we just | |
1000 | * allocated. | |
1001 | */ | |
1da177e4 LT |
1002 | spin_unlock(&inode_lock); |
1003 | destroy_inode(inode); | |
1004 | inode = old; | |
1005 | wait_on_inode(inode); | |
1006 | } | |
1007 | return inode; | |
1008 | } | |
1009 | ||
ad5e195a CH |
1010 | /* |
1011 | * search the inode cache for a matching inode number. | |
1012 | * If we find one, then the inode number we are trying to | |
1013 | * allocate is not unique and so we should not use it. | |
1014 | * | |
1015 | * Returns 1 if the inode number is unique, 0 if it is not. | |
1016 | */ | |
1017 | static int test_inode_iunique(struct super_block *sb, unsigned long ino) | |
1018 | { | |
1019 | struct hlist_head *b = inode_hashtable + hash(sb, ino); | |
1020 | struct hlist_node *node; | |
1021 | struct inode *inode; | |
1022 | ||
1023 | hlist_for_each_entry(inode, node, b, i_hash) { | |
1024 | if (inode->i_ino == ino && inode->i_sb == sb) | |
1025 | return 0; | |
1026 | } | |
1027 | ||
1028 | return 1; | |
1029 | } | |
1030 | ||
1da177e4 LT |
1031 | /** |
1032 | * iunique - get a unique inode number | |
1033 | * @sb: superblock | |
1034 | * @max_reserved: highest reserved inode number | |
1035 | * | |
1036 | * Obtain an inode number that is unique on the system for a given | |
1037 | * superblock. This is used by file systems that have no natural | |
1038 | * permanent inode numbering system. An inode number is returned that | |
1039 | * is higher than the reserved limit but unique. | |
1040 | * | |
1041 | * BUGS: | |
1042 | * With a large number of inodes live on the file system this function | |
1043 | * currently becomes quite slow. | |
1044 | */ | |
1045 | ino_t iunique(struct super_block *sb, ino_t max_reserved) | |
1046 | { | |
866b04fc JL |
1047 | /* |
1048 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
1049 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
1050 | * here to attempt to avoid that. | |
1051 | */ | |
ad5e195a | 1052 | static DEFINE_SPINLOCK(iunique_lock); |
866b04fc | 1053 | static unsigned int counter; |
1da177e4 | 1054 | ino_t res; |
3361c7be | 1055 | |
1da177e4 | 1056 | spin_lock(&inode_lock); |
ad5e195a | 1057 | spin_lock(&iunique_lock); |
3361c7be JL |
1058 | do { |
1059 | if (counter <= max_reserved) | |
1060 | counter = max_reserved + 1; | |
1da177e4 | 1061 | res = counter++; |
ad5e195a CH |
1062 | } while (!test_inode_iunique(sb, res)); |
1063 | spin_unlock(&iunique_lock); | |
3361c7be | 1064 | spin_unlock(&inode_lock); |
1da177e4 | 1065 | |
3361c7be JL |
1066 | return res; |
1067 | } | |
1da177e4 LT |
1068 | EXPORT_SYMBOL(iunique); |
1069 | ||
1070 | struct inode *igrab(struct inode *inode) | |
1071 | { | |
1072 | spin_lock(&inode_lock); | |
250df6ed DC |
1073 | spin_lock(&inode->i_lock); |
1074 | if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) { | |
1da177e4 | 1075 | __iget(inode); |
250df6ed DC |
1076 | spin_unlock(&inode->i_lock); |
1077 | } else { | |
1078 | spin_unlock(&inode->i_lock); | |
1da177e4 LT |
1079 | /* |
1080 | * Handle the case where s_op->clear_inode is not been | |
1081 | * called yet, and somebody is calling igrab | |
1082 | * while the inode is getting freed. | |
1083 | */ | |
1084 | inode = NULL; | |
250df6ed | 1085 | } |
1da177e4 LT |
1086 | spin_unlock(&inode_lock); |
1087 | return inode; | |
1088 | } | |
1da177e4 LT |
1089 | EXPORT_SYMBOL(igrab); |
1090 | ||
1091 | /** | |
1092 | * ifind - internal function, you want ilookup5() or iget5(). | |
1093 | * @sb: super block of file system to search | |
1094 | * @head: the head of the list to search | |
1095 | * @test: callback used for comparisons between inodes | |
1096 | * @data: opaque data pointer to pass to @test | |
88bd5121 | 1097 | * @wait: if true wait for the inode to be unlocked, if false do not |
1da177e4 LT |
1098 | * |
1099 | * ifind() searches for the inode specified by @data in the inode | |
1100 | * cache. This is a generalized version of ifind_fast() for file systems where | |
1101 | * the inode number is not sufficient for unique identification of an inode. | |
1102 | * | |
1103 | * If the inode is in the cache, the inode is returned with an incremented | |
1104 | * reference count. | |
1105 | * | |
1106 | * Otherwise NULL is returned. | |
1107 | * | |
1108 | * Note, @test is called with the inode_lock held, so can't sleep. | |
1109 | */ | |
5d2bea45 | 1110 | static struct inode *ifind(struct super_block *sb, |
1da177e4 | 1111 | struct hlist_head *head, int (*test)(struct inode *, void *), |
88bd5121 | 1112 | void *data, const int wait) |
1da177e4 LT |
1113 | { |
1114 | struct inode *inode; | |
1115 | ||
1116 | spin_lock(&inode_lock); | |
1117 | inode = find_inode(sb, head, test, data); | |
1118 | if (inode) { | |
1da177e4 | 1119 | spin_unlock(&inode_lock); |
88bd5121 AA |
1120 | if (likely(wait)) |
1121 | wait_on_inode(inode); | |
1da177e4 LT |
1122 | return inode; |
1123 | } | |
1124 | spin_unlock(&inode_lock); | |
1125 | return NULL; | |
1126 | } | |
1127 | ||
1128 | /** | |
1129 | * ifind_fast - internal function, you want ilookup() or iget(). | |
1130 | * @sb: super block of file system to search | |
1131 | * @head: head of the list to search | |
1132 | * @ino: inode number to search for | |
1133 | * | |
1134 | * ifind_fast() searches for the inode @ino in the inode cache. This is for | |
1135 | * file systems where the inode number is sufficient for unique identification | |
1136 | * of an inode. | |
1137 | * | |
1138 | * If the inode is in the cache, the inode is returned with an incremented | |
1139 | * reference count. | |
1140 | * | |
1141 | * Otherwise NULL is returned. | |
1142 | */ | |
5d2bea45 | 1143 | static struct inode *ifind_fast(struct super_block *sb, |
1da177e4 LT |
1144 | struct hlist_head *head, unsigned long ino) |
1145 | { | |
1146 | struct inode *inode; | |
1147 | ||
1148 | spin_lock(&inode_lock); | |
1149 | inode = find_inode_fast(sb, head, ino); | |
1150 | if (inode) { | |
1da177e4 LT |
1151 | spin_unlock(&inode_lock); |
1152 | wait_on_inode(inode); | |
1153 | return inode; | |
1154 | } | |
1155 | spin_unlock(&inode_lock); | |
1156 | return NULL; | |
1157 | } | |
1158 | ||
1159 | /** | |
88bd5121 | 1160 | * ilookup5_nowait - search for an inode in the inode cache |
1da177e4 LT |
1161 | * @sb: super block of file system to search |
1162 | * @hashval: hash value (usually inode number) to search for | |
1163 | * @test: callback used for comparisons between inodes | |
1164 | * @data: opaque data pointer to pass to @test | |
1165 | * | |
1166 | * ilookup5() uses ifind() to search for the inode specified by @hashval and | |
1167 | * @data in the inode cache. This is a generalized version of ilookup() for | |
1168 | * file systems where the inode number is not sufficient for unique | |
1169 | * identification of an inode. | |
1170 | * | |
1171 | * If the inode is in the cache, the inode is returned with an incremented | |
88bd5121 AA |
1172 | * reference count. Note, the inode lock is not waited upon so you have to be |
1173 | * very careful what you do with the returned inode. You probably should be | |
1174 | * using ilookup5() instead. | |
1175 | * | |
1176 | * Otherwise NULL is returned. | |
1177 | * | |
1178 | * Note, @test is called with the inode_lock held, so can't sleep. | |
1179 | */ | |
1180 | struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval, | |
1181 | int (*test)(struct inode *, void *), void *data) | |
1182 | { | |
1183 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1184 | ||
1185 | return ifind(sb, head, test, data, 0); | |
1186 | } | |
88bd5121 AA |
1187 | EXPORT_SYMBOL(ilookup5_nowait); |
1188 | ||
1189 | /** | |
1190 | * ilookup5 - search for an inode in the inode cache | |
1191 | * @sb: super block of file system to search | |
1192 | * @hashval: hash value (usually inode number) to search for | |
1193 | * @test: callback used for comparisons between inodes | |
1194 | * @data: opaque data pointer to pass to @test | |
1195 | * | |
1196 | * ilookup5() uses ifind() to search for the inode specified by @hashval and | |
1197 | * @data in the inode cache. This is a generalized version of ilookup() for | |
1198 | * file systems where the inode number is not sufficient for unique | |
1199 | * identification of an inode. | |
1200 | * | |
1201 | * If the inode is in the cache, the inode lock is waited upon and the inode is | |
1202 | * returned with an incremented reference count. | |
1da177e4 LT |
1203 | * |
1204 | * Otherwise NULL is returned. | |
1205 | * | |
1206 | * Note, @test is called with the inode_lock held, so can't sleep. | |
1207 | */ | |
1208 | struct inode *ilookup5(struct super_block *sb, unsigned long hashval, | |
1209 | int (*test)(struct inode *, void *), void *data) | |
1210 | { | |
1211 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1212 | ||
88bd5121 | 1213 | return ifind(sb, head, test, data, 1); |
1da177e4 | 1214 | } |
1da177e4 LT |
1215 | EXPORT_SYMBOL(ilookup5); |
1216 | ||
1217 | /** | |
1218 | * ilookup - search for an inode in the inode cache | |
1219 | * @sb: super block of file system to search | |
1220 | * @ino: inode number to search for | |
1221 | * | |
1222 | * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache. | |
1223 | * This is for file systems where the inode number is sufficient for unique | |
1224 | * identification of an inode. | |
1225 | * | |
1226 | * If the inode is in the cache, the inode is returned with an incremented | |
1227 | * reference count. | |
1228 | * | |
1229 | * Otherwise NULL is returned. | |
1230 | */ | |
1231 | struct inode *ilookup(struct super_block *sb, unsigned long ino) | |
1232 | { | |
1233 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1234 | ||
1235 | return ifind_fast(sb, head, ino); | |
1236 | } | |
1da177e4 LT |
1237 | EXPORT_SYMBOL(ilookup); |
1238 | ||
1239 | /** | |
1240 | * iget5_locked - obtain an inode from a mounted file system | |
1241 | * @sb: super block of file system | |
1242 | * @hashval: hash value (usually inode number) to get | |
1243 | * @test: callback used for comparisons between inodes | |
1244 | * @set: callback used to initialize a new struct inode | |
1245 | * @data: opaque data pointer to pass to @test and @set | |
1246 | * | |
1da177e4 LT |
1247 | * iget5_locked() uses ifind() to search for the inode specified by @hashval |
1248 | * and @data in the inode cache and if present it is returned with an increased | |
1249 | * reference count. This is a generalized version of iget_locked() for file | |
1250 | * systems where the inode number is not sufficient for unique identification | |
1251 | * of an inode. | |
1252 | * | |
1253 | * If the inode is not in cache, get_new_inode() is called to allocate a new | |
1254 | * inode and this is returned locked, hashed, and with the I_NEW flag set. The | |
1255 | * file system gets to fill it in before unlocking it via unlock_new_inode(). | |
1256 | * | |
1257 | * Note both @test and @set are called with the inode_lock held, so can't sleep. | |
1258 | */ | |
1259 | struct inode *iget5_locked(struct super_block *sb, unsigned long hashval, | |
1260 | int (*test)(struct inode *, void *), | |
1261 | int (*set)(struct inode *, void *), void *data) | |
1262 | { | |
1263 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1264 | struct inode *inode; | |
1265 | ||
88bd5121 | 1266 | inode = ifind(sb, head, test, data, 1); |
1da177e4 LT |
1267 | if (inode) |
1268 | return inode; | |
1269 | /* | |
1270 | * get_new_inode() will do the right thing, re-trying the search | |
1271 | * in case it had to block at any point. | |
1272 | */ | |
1273 | return get_new_inode(sb, head, test, set, data); | |
1274 | } | |
1da177e4 LT |
1275 | EXPORT_SYMBOL(iget5_locked); |
1276 | ||
1277 | /** | |
1278 | * iget_locked - obtain an inode from a mounted file system | |
1279 | * @sb: super block of file system | |
1280 | * @ino: inode number to get | |
1281 | * | |
1da177e4 LT |
1282 | * iget_locked() uses ifind_fast() to search for the inode specified by @ino in |
1283 | * the inode cache and if present it is returned with an increased reference | |
1284 | * count. This is for file systems where the inode number is sufficient for | |
1285 | * unique identification of an inode. | |
1286 | * | |
1287 | * If the inode is not in cache, get_new_inode_fast() is called to allocate a | |
1288 | * new inode and this is returned locked, hashed, and with the I_NEW flag set. | |
1289 | * The file system gets to fill it in before unlocking it via | |
1290 | * unlock_new_inode(). | |
1291 | */ | |
1292 | struct inode *iget_locked(struct super_block *sb, unsigned long ino) | |
1293 | { | |
1294 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1295 | struct inode *inode; | |
1296 | ||
1297 | inode = ifind_fast(sb, head, ino); | |
1298 | if (inode) | |
1299 | return inode; | |
1300 | /* | |
1301 | * get_new_inode_fast() will do the right thing, re-trying the search | |
1302 | * in case it had to block at any point. | |
1303 | */ | |
1304 | return get_new_inode_fast(sb, head, ino); | |
1305 | } | |
1da177e4 LT |
1306 | EXPORT_SYMBOL(iget_locked); |
1307 | ||
261bca86 AV |
1308 | int insert_inode_locked(struct inode *inode) |
1309 | { | |
1310 | struct super_block *sb = inode->i_sb; | |
1311 | ino_t ino = inode->i_ino; | |
1312 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
261bca86 | 1313 | |
261bca86 | 1314 | while (1) { |
72a43d63 AV |
1315 | struct hlist_node *node; |
1316 | struct inode *old = NULL; | |
261bca86 | 1317 | spin_lock(&inode_lock); |
72a43d63 AV |
1318 | hlist_for_each_entry(old, node, head, i_hash) { |
1319 | if (old->i_ino != ino) | |
1320 | continue; | |
1321 | if (old->i_sb != sb) | |
1322 | continue; | |
250df6ed DC |
1323 | spin_lock(&old->i_lock); |
1324 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1325 | spin_unlock(&old->i_lock); | |
72a43d63 | 1326 | continue; |
250df6ed | 1327 | } |
72a43d63 AV |
1328 | break; |
1329 | } | |
1330 | if (likely(!node)) { | |
250df6ed DC |
1331 | spin_lock(&inode->i_lock); |
1332 | inode->i_state |= I_NEW; | |
261bca86 | 1333 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1334 | spin_unlock(&inode->i_lock); |
261bca86 AV |
1335 | spin_unlock(&inode_lock); |
1336 | return 0; | |
1337 | } | |
1338 | __iget(old); | |
250df6ed | 1339 | spin_unlock(&old->i_lock); |
261bca86 AV |
1340 | spin_unlock(&inode_lock); |
1341 | wait_on_inode(old); | |
1d3382cb | 1342 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1343 | iput(old); |
1344 | return -EBUSY; | |
1345 | } | |
1346 | iput(old); | |
1347 | } | |
1348 | } | |
261bca86 AV |
1349 | EXPORT_SYMBOL(insert_inode_locked); |
1350 | ||
1351 | int insert_inode_locked4(struct inode *inode, unsigned long hashval, | |
1352 | int (*test)(struct inode *, void *), void *data) | |
1353 | { | |
1354 | struct super_block *sb = inode->i_sb; | |
1355 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
261bca86 | 1356 | |
261bca86 | 1357 | while (1) { |
72a43d63 AV |
1358 | struct hlist_node *node; |
1359 | struct inode *old = NULL; | |
1360 | ||
261bca86 | 1361 | spin_lock(&inode_lock); |
72a43d63 AV |
1362 | hlist_for_each_entry(old, node, head, i_hash) { |
1363 | if (old->i_sb != sb) | |
1364 | continue; | |
1365 | if (!test(old, data)) | |
1366 | continue; | |
250df6ed DC |
1367 | spin_lock(&old->i_lock); |
1368 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1369 | spin_unlock(&old->i_lock); | |
72a43d63 | 1370 | continue; |
250df6ed | 1371 | } |
72a43d63 AV |
1372 | break; |
1373 | } | |
1374 | if (likely(!node)) { | |
250df6ed DC |
1375 | spin_lock(&inode->i_lock); |
1376 | inode->i_state |= I_NEW; | |
261bca86 | 1377 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1378 | spin_unlock(&inode->i_lock); |
261bca86 AV |
1379 | spin_unlock(&inode_lock); |
1380 | return 0; | |
1381 | } | |
1382 | __iget(old); | |
250df6ed | 1383 | spin_unlock(&old->i_lock); |
261bca86 AV |
1384 | spin_unlock(&inode_lock); |
1385 | wait_on_inode(old); | |
1d3382cb | 1386 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1387 | iput(old); |
1388 | return -EBUSY; | |
1389 | } | |
1390 | iput(old); | |
1391 | } | |
1392 | } | |
261bca86 AV |
1393 | EXPORT_SYMBOL(insert_inode_locked4); |
1394 | ||
1da177e4 | 1395 | |
45321ac5 AV |
1396 | int generic_delete_inode(struct inode *inode) |
1397 | { | |
1398 | return 1; | |
1399 | } | |
1400 | EXPORT_SYMBOL(generic_delete_inode); | |
1401 | ||
1da177e4 | 1402 | /* |
45321ac5 AV |
1403 | * Normal UNIX filesystem behaviour: delete the |
1404 | * inode when the usage count drops to zero, and | |
1405 | * i_nlink is zero. | |
1da177e4 | 1406 | */ |
45321ac5 | 1407 | int generic_drop_inode(struct inode *inode) |
1da177e4 | 1408 | { |
1d3382cb | 1409 | return !inode->i_nlink || inode_unhashed(inode); |
1da177e4 | 1410 | } |
45321ac5 | 1411 | EXPORT_SYMBOL_GPL(generic_drop_inode); |
1da177e4 | 1412 | |
45321ac5 AV |
1413 | /* |
1414 | * Called when we're dropping the last reference | |
1415 | * to an inode. | |
22fe4042 | 1416 | * |
45321ac5 AV |
1417 | * Call the FS "drop_inode()" function, defaulting to |
1418 | * the legacy UNIX filesystem behaviour. If it tells | |
1419 | * us to evict inode, do so. Otherwise, retain inode | |
1420 | * in cache if fs is alive, sync and evict if fs is | |
1421 | * shutting down. | |
22fe4042 | 1422 | */ |
45321ac5 | 1423 | static void iput_final(struct inode *inode) |
1da177e4 LT |
1424 | { |
1425 | struct super_block *sb = inode->i_sb; | |
45321ac5 AV |
1426 | const struct super_operations *op = inode->i_sb->s_op; |
1427 | int drop; | |
1428 | ||
250df6ed DC |
1429 | spin_lock(&inode->i_lock); |
1430 | WARN_ON(inode->i_state & I_NEW); | |
1431 | ||
45321ac5 AV |
1432 | if (op && op->drop_inode) |
1433 | drop = op->drop_inode(inode); | |
1434 | else | |
1435 | drop = generic_drop_inode(inode); | |
1da177e4 | 1436 | |
45321ac5 | 1437 | if (!drop) { |
acb0c854 | 1438 | if (sb->s_flags & MS_ACTIVE) { |
9e38d86f NP |
1439 | inode->i_state |= I_REFERENCED; |
1440 | if (!(inode->i_state & (I_DIRTY|I_SYNC))) { | |
1441 | inode_lru_list_add(inode); | |
1442 | } | |
250df6ed | 1443 | spin_unlock(&inode->i_lock); |
991114c6 | 1444 | spin_unlock(&inode_lock); |
45321ac5 | 1445 | return; |
991114c6 AV |
1446 | } |
1447 | inode->i_state |= I_WILL_FREE; | |
250df6ed | 1448 | spin_unlock(&inode->i_lock); |
991114c6 | 1449 | spin_unlock(&inode_lock); |
1da177e4 LT |
1450 | write_inode_now(inode, 1); |
1451 | spin_lock(&inode_lock); | |
250df6ed | 1452 | spin_lock(&inode->i_lock); |
7ef0d737 | 1453 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1454 | inode->i_state &= ~I_WILL_FREE; |
4c51acbc | 1455 | __remove_inode_hash(inode); |
1da177e4 | 1456 | } |
7ccf19a8 | 1457 | |
991114c6 | 1458 | inode->i_state |= I_FREEING; |
250df6ed | 1459 | spin_unlock(&inode->i_lock); |
9e38d86f NP |
1460 | |
1461 | /* | |
7ccf19a8 NP |
1462 | * Move the inode off the IO lists and LRU once I_FREEING is |
1463 | * set so that it won't get moved back on there if it is dirty. | |
9e38d86f NP |
1464 | */ |
1465 | inode_lru_list_del(inode); | |
7ccf19a8 | 1466 | list_del_init(&inode->i_wb_list); |
9e38d86f | 1467 | |
646ec461 | 1468 | __inode_sb_list_del(inode); |
1da177e4 | 1469 | spin_unlock(&inode_lock); |
644da596 | 1470 | evict(inode); |
4c51acbc | 1471 | remove_inode_hash(inode); |
250df6ed DC |
1472 | spin_lock(&inode->i_lock); |
1473 | wake_up_bit(&inode->i_state, __I_NEW); | |
45321ac5 | 1474 | BUG_ON(inode->i_state != (I_FREEING | I_CLEAR)); |
250df6ed | 1475 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1476 | destroy_inode(inode); |
1477 | } | |
1478 | ||
1da177e4 | 1479 | /** |
6b3304b5 | 1480 | * iput - put an inode |
1da177e4 LT |
1481 | * @inode: inode to put |
1482 | * | |
1483 | * Puts an inode, dropping its usage count. If the inode use count hits | |
1484 | * zero, the inode is then freed and may also be destroyed. | |
1485 | * | |
1486 | * Consequently, iput() can sleep. | |
1487 | */ | |
1488 | void iput(struct inode *inode) | |
1489 | { | |
1490 | if (inode) { | |
a4ffdde6 | 1491 | BUG_ON(inode->i_state & I_CLEAR); |
1da177e4 | 1492 | |
1da177e4 LT |
1493 | if (atomic_dec_and_lock(&inode->i_count, &inode_lock)) |
1494 | iput_final(inode); | |
1495 | } | |
1496 | } | |
1da177e4 LT |
1497 | EXPORT_SYMBOL(iput); |
1498 | ||
1499 | /** | |
1500 | * bmap - find a block number in a file | |
1501 | * @inode: inode of file | |
1502 | * @block: block to find | |
1503 | * | |
1504 | * Returns the block number on the device holding the inode that | |
1505 | * is the disk block number for the block of the file requested. | |
1506 | * That is, asked for block 4 of inode 1 the function will return the | |
6b3304b5 | 1507 | * disk block relative to the disk start that holds that block of the |
1da177e4 LT |
1508 | * file. |
1509 | */ | |
6b3304b5 | 1510 | sector_t bmap(struct inode *inode, sector_t block) |
1da177e4 LT |
1511 | { |
1512 | sector_t res = 0; | |
1513 | if (inode->i_mapping->a_ops->bmap) | |
1514 | res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block); | |
1515 | return res; | |
1516 | } | |
1da177e4 LT |
1517 | EXPORT_SYMBOL(bmap); |
1518 | ||
11ff6f05 MG |
1519 | /* |
1520 | * With relative atime, only update atime if the previous atime is | |
1521 | * earlier than either the ctime or mtime or if at least a day has | |
1522 | * passed since the last atime update. | |
1523 | */ | |
1524 | static int relatime_need_update(struct vfsmount *mnt, struct inode *inode, | |
1525 | struct timespec now) | |
1526 | { | |
1527 | ||
1528 | if (!(mnt->mnt_flags & MNT_RELATIME)) | |
1529 | return 1; | |
1530 | /* | |
1531 | * Is mtime younger than atime? If yes, update atime: | |
1532 | */ | |
1533 | if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0) | |
1534 | return 1; | |
1535 | /* | |
1536 | * Is ctime younger than atime? If yes, update atime: | |
1537 | */ | |
1538 | if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0) | |
1539 | return 1; | |
1540 | ||
1541 | /* | |
1542 | * Is the previous atime value older than a day? If yes, | |
1543 | * update atime: | |
1544 | */ | |
1545 | if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60) | |
1546 | return 1; | |
1547 | /* | |
1548 | * Good, we can skip the atime update: | |
1549 | */ | |
1550 | return 0; | |
1551 | } | |
1552 | ||
1da177e4 | 1553 | /** |
869243a0 CH |
1554 | * touch_atime - update the access time |
1555 | * @mnt: mount the inode is accessed on | |
7045f37b | 1556 | * @dentry: dentry accessed |
1da177e4 LT |
1557 | * |
1558 | * Update the accessed time on an inode and mark it for writeback. | |
1559 | * This function automatically handles read only file systems and media, | |
1560 | * as well as the "noatime" flag and inode specific "noatime" markers. | |
1561 | */ | |
869243a0 | 1562 | void touch_atime(struct vfsmount *mnt, struct dentry *dentry) |
1da177e4 | 1563 | { |
869243a0 | 1564 | struct inode *inode = dentry->d_inode; |
1da177e4 LT |
1565 | struct timespec now; |
1566 | ||
cdb70f3f | 1567 | if (inode->i_flags & S_NOATIME) |
b12536c2 | 1568 | return; |
37756ced | 1569 | if (IS_NOATIME(inode)) |
b12536c2 | 1570 | return; |
b2276138 | 1571 | if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1572 | return; |
47ae32d6 | 1573 | |
cdb70f3f | 1574 | if (mnt->mnt_flags & MNT_NOATIME) |
b12536c2 | 1575 | return; |
cdb70f3f | 1576 | if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1577 | return; |
1da177e4 LT |
1578 | |
1579 | now = current_fs_time(inode->i_sb); | |
11ff6f05 MG |
1580 | |
1581 | if (!relatime_need_update(mnt, inode, now)) | |
b12536c2 | 1582 | return; |
11ff6f05 | 1583 | |
47ae32d6 | 1584 | if (timespec_equal(&inode->i_atime, &now)) |
b12536c2 AK |
1585 | return; |
1586 | ||
1587 | if (mnt_want_write(mnt)) | |
1588 | return; | |
47ae32d6 VH |
1589 | |
1590 | inode->i_atime = now; | |
1591 | mark_inode_dirty_sync(inode); | |
cdb70f3f | 1592 | mnt_drop_write(mnt); |
1da177e4 | 1593 | } |
869243a0 | 1594 | EXPORT_SYMBOL(touch_atime); |
1da177e4 LT |
1595 | |
1596 | /** | |
870f4817 CH |
1597 | * file_update_time - update mtime and ctime time |
1598 | * @file: file accessed | |
1da177e4 | 1599 | * |
870f4817 CH |
1600 | * Update the mtime and ctime members of an inode and mark the inode |
1601 | * for writeback. Note that this function is meant exclusively for | |
1602 | * usage in the file write path of filesystems, and filesystems may | |
1603 | * choose to explicitly ignore update via this function with the | |
2eadfc0e | 1604 | * S_NOCMTIME inode flag, e.g. for network filesystem where these |
870f4817 | 1605 | * timestamps are handled by the server. |
1da177e4 LT |
1606 | */ |
1607 | ||
870f4817 | 1608 | void file_update_time(struct file *file) |
1da177e4 | 1609 | { |
0f7fc9e4 | 1610 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 | 1611 | struct timespec now; |
ce06e0b2 | 1612 | enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0; |
1da177e4 | 1613 | |
ce06e0b2 | 1614 | /* First try to exhaust all avenues to not sync */ |
1da177e4 LT |
1615 | if (IS_NOCMTIME(inode)) |
1616 | return; | |
20ddee2c | 1617 | |
1da177e4 | 1618 | now = current_fs_time(inode->i_sb); |
ce06e0b2 AK |
1619 | if (!timespec_equal(&inode->i_mtime, &now)) |
1620 | sync_it = S_MTIME; | |
1da177e4 | 1621 | |
ce06e0b2 AK |
1622 | if (!timespec_equal(&inode->i_ctime, &now)) |
1623 | sync_it |= S_CTIME; | |
870f4817 | 1624 | |
ce06e0b2 AK |
1625 | if (IS_I_VERSION(inode)) |
1626 | sync_it |= S_VERSION; | |
7a224228 | 1627 | |
ce06e0b2 AK |
1628 | if (!sync_it) |
1629 | return; | |
1630 | ||
1631 | /* Finally allowed to write? Takes lock. */ | |
1632 | if (mnt_want_write_file(file)) | |
1633 | return; | |
1634 | ||
1635 | /* Only change inode inside the lock region */ | |
1636 | if (sync_it & S_VERSION) | |
1637 | inode_inc_iversion(inode); | |
1638 | if (sync_it & S_CTIME) | |
1639 | inode->i_ctime = now; | |
1640 | if (sync_it & S_MTIME) | |
1641 | inode->i_mtime = now; | |
1642 | mark_inode_dirty_sync(inode); | |
20ddee2c | 1643 | mnt_drop_write(file->f_path.mnt); |
1da177e4 | 1644 | } |
870f4817 | 1645 | EXPORT_SYMBOL(file_update_time); |
1da177e4 LT |
1646 | |
1647 | int inode_needs_sync(struct inode *inode) | |
1648 | { | |
1649 | if (IS_SYNC(inode)) | |
1650 | return 1; | |
1651 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) | |
1652 | return 1; | |
1653 | return 0; | |
1654 | } | |
1da177e4 LT |
1655 | EXPORT_SYMBOL(inode_needs_sync); |
1656 | ||
1da177e4 LT |
1657 | int inode_wait(void *word) |
1658 | { | |
1659 | schedule(); | |
1660 | return 0; | |
1661 | } | |
d44dab8d | 1662 | EXPORT_SYMBOL(inode_wait); |
1da177e4 LT |
1663 | |
1664 | /* | |
168a9fd6 MS |
1665 | * If we try to find an inode in the inode hash while it is being |
1666 | * deleted, we have to wait until the filesystem completes its | |
1667 | * deletion before reporting that it isn't found. This function waits | |
1668 | * until the deletion _might_ have completed. Callers are responsible | |
1669 | * to recheck inode state. | |
1670 | * | |
eaff8079 | 1671 | * It doesn't matter if I_NEW is not set initially, a call to |
250df6ed DC |
1672 | * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list |
1673 | * will DTRT. | |
1da177e4 LT |
1674 | */ |
1675 | static void __wait_on_freeing_inode(struct inode *inode) | |
1676 | { | |
1677 | wait_queue_head_t *wq; | |
eaff8079 CH |
1678 | DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW); |
1679 | wq = bit_waitqueue(&inode->i_state, __I_NEW); | |
1da177e4 | 1680 | prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE); |
250df6ed | 1681 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1682 | spin_unlock(&inode_lock); |
1683 | schedule(); | |
1684 | finish_wait(wq, &wait.wait); | |
1685 | spin_lock(&inode_lock); | |
1686 | } | |
1687 | ||
1da177e4 LT |
1688 | static __initdata unsigned long ihash_entries; |
1689 | static int __init set_ihash_entries(char *str) | |
1690 | { | |
1691 | if (!str) | |
1692 | return 0; | |
1693 | ihash_entries = simple_strtoul(str, &str, 0); | |
1694 | return 1; | |
1695 | } | |
1696 | __setup("ihash_entries=", set_ihash_entries); | |
1697 | ||
1698 | /* | |
1699 | * Initialize the waitqueues and inode hash table. | |
1700 | */ | |
1701 | void __init inode_init_early(void) | |
1702 | { | |
1703 | int loop; | |
1704 | ||
1705 | /* If hashes are distributed across NUMA nodes, defer | |
1706 | * hash allocation until vmalloc space is available. | |
1707 | */ | |
1708 | if (hashdist) | |
1709 | return; | |
1710 | ||
1711 | inode_hashtable = | |
1712 | alloc_large_system_hash("Inode-cache", | |
1713 | sizeof(struct hlist_head), | |
1714 | ihash_entries, | |
1715 | 14, | |
1716 | HASH_EARLY, | |
1717 | &i_hash_shift, | |
1718 | &i_hash_mask, | |
1719 | 0); | |
1720 | ||
1721 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1722 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1723 | } | |
1724 | ||
74bf17cf | 1725 | void __init inode_init(void) |
1da177e4 LT |
1726 | { |
1727 | int loop; | |
1728 | ||
1729 | /* inode slab cache */ | |
b0196009 PJ |
1730 | inode_cachep = kmem_cache_create("inode_cache", |
1731 | sizeof(struct inode), | |
1732 | 0, | |
1733 | (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC| | |
1734 | SLAB_MEM_SPREAD), | |
20c2df83 | 1735 | init_once); |
8e1f936b | 1736 | register_shrinker(&icache_shrinker); |
1da177e4 LT |
1737 | |
1738 | /* Hash may have been set up in inode_init_early */ | |
1739 | if (!hashdist) | |
1740 | return; | |
1741 | ||
1742 | inode_hashtable = | |
1743 | alloc_large_system_hash("Inode-cache", | |
1744 | sizeof(struct hlist_head), | |
1745 | ihash_entries, | |
1746 | 14, | |
1747 | 0, | |
1748 | &i_hash_shift, | |
1749 | &i_hash_mask, | |
1750 | 0); | |
1751 | ||
1752 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1753 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1754 | } | |
1755 | ||
1756 | void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev) | |
1757 | { | |
1758 | inode->i_mode = mode; | |
1759 | if (S_ISCHR(mode)) { | |
1760 | inode->i_fop = &def_chr_fops; | |
1761 | inode->i_rdev = rdev; | |
1762 | } else if (S_ISBLK(mode)) { | |
1763 | inode->i_fop = &def_blk_fops; | |
1764 | inode->i_rdev = rdev; | |
1765 | } else if (S_ISFIFO(mode)) | |
1766 | inode->i_fop = &def_fifo_fops; | |
1767 | else if (S_ISSOCK(mode)) | |
1768 | inode->i_fop = &bad_sock_fops; | |
1769 | else | |
af0d9ae8 MK |
1770 | printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for" |
1771 | " inode %s:%lu\n", mode, inode->i_sb->s_id, | |
1772 | inode->i_ino); | |
1da177e4 LT |
1773 | } |
1774 | EXPORT_SYMBOL(init_special_inode); | |
a1bd120d DM |
1775 | |
1776 | /** | |
eaae668d | 1777 | * inode_init_owner - Init uid,gid,mode for new inode according to posix standards |
a1bd120d DM |
1778 | * @inode: New inode |
1779 | * @dir: Directory inode | |
1780 | * @mode: mode of the new inode | |
1781 | */ | |
1782 | void inode_init_owner(struct inode *inode, const struct inode *dir, | |
1783 | mode_t mode) | |
1784 | { | |
1785 | inode->i_uid = current_fsuid(); | |
1786 | if (dir && dir->i_mode & S_ISGID) { | |
1787 | inode->i_gid = dir->i_gid; | |
1788 | if (S_ISDIR(mode)) | |
1789 | mode |= S_ISGID; | |
1790 | } else | |
1791 | inode->i_gid = current_fsgid(); | |
1792 | inode->i_mode = mode; | |
1793 | } | |
1794 | EXPORT_SYMBOL(inode_init_owner); | |
e795b717 | 1795 | |
2e149670 SH |
1796 | /** |
1797 | * inode_owner_or_capable - check current task permissions to inode | |
1798 | * @inode: inode being checked | |
1799 | * | |
1800 | * Return true if current either has CAP_FOWNER to the inode, or | |
1801 | * owns the file. | |
e795b717 | 1802 | */ |
2e149670 | 1803 | bool inode_owner_or_capable(const struct inode *inode) |
e795b717 SH |
1804 | { |
1805 | struct user_namespace *ns = inode_userns(inode); | |
1806 | ||
1807 | if (current_user_ns() == ns && current_fsuid() == inode->i_uid) | |
1808 | return true; | |
1809 | if (ns_capable(ns, CAP_FOWNER)) | |
1810 | return true; | |
1811 | return false; | |
1812 | } | |
2e149670 | 1813 | EXPORT_SYMBOL(inode_owner_or_capable); |