]>
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> |
1da177e4 LT |
28 | |
29 | /* | |
30 | * This is needed for the following functions: | |
31 | * - inode_has_buffers | |
1da177e4 LT |
32 | * - invalidate_bdev |
33 | * | |
34 | * FIXME: remove all knowledge of the buffer layer from this file | |
35 | */ | |
36 | #include <linux/buffer_head.h> | |
37 | ||
38 | /* | |
39 | * New inode.c implementation. | |
40 | * | |
41 | * This implementation has the basic premise of trying | |
42 | * to be extremely low-overhead and SMP-safe, yet be | |
43 | * simple enough to be "obviously correct". | |
44 | * | |
45 | * Famous last words. | |
46 | */ | |
47 | ||
48 | /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */ | |
49 | ||
50 | /* #define INODE_PARANOIA 1 */ | |
51 | /* #define INODE_DEBUG 1 */ | |
52 | ||
53 | /* | |
54 | * Inode lookup is no longer as critical as it used to be: | |
55 | * most of the lookups are going to be through the dcache. | |
56 | */ | |
57 | #define I_HASHBITS i_hash_shift | |
58 | #define I_HASHMASK i_hash_mask | |
59 | ||
fa3536cc ED |
60 | static unsigned int i_hash_mask __read_mostly; |
61 | static unsigned int i_hash_shift __read_mostly; | |
1da177e4 LT |
62 | |
63 | /* | |
64 | * Each inode can be on two separate lists. One is | |
65 | * the hash list of the inode, used for lookups. The | |
66 | * other linked list is the "type" list: | |
67 | * "in_use" - valid inode, i_count > 0, i_nlink > 0 | |
68 | * "dirty" - as "in_use" but also dirty | |
69 | * "unused" - valid inode, i_count = 0 | |
70 | * | |
71 | * A "dirty" list is maintained for each super block, | |
72 | * allowing for low-overhead inode sync() operations. | |
73 | */ | |
74 | ||
7ccf19a8 | 75 | static LIST_HEAD(inode_lru); |
fa3536cc | 76 | static struct hlist_head *inode_hashtable __read_mostly; |
1da177e4 LT |
77 | |
78 | /* | |
79 | * A simple spinlock to protect the list manipulations. | |
80 | * | |
81 | * NOTE! You also have to own the lock if you change | |
82 | * the i_state of an inode while it is in use.. | |
83 | */ | |
84 | DEFINE_SPINLOCK(inode_lock); | |
85 | ||
86 | /* | |
88e0fbc4 | 87 | * iprune_sem provides exclusion between the kswapd or try_to_free_pages |
1da177e4 LT |
88 | * icache shrinking path, and the umount path. Without this exclusion, |
89 | * by the time prune_icache calls iput for the inode whose pages it has | |
90 | * been invalidating, or by the time it calls clear_inode & destroy_inode | |
91 | * from its final dispose_list, the struct super_block they refer to | |
92 | * (for inode->i_sb->s_op) may already have been freed and reused. | |
88e0fbc4 NP |
93 | * |
94 | * We make this an rwsem because the fastpath is icache shrinking. In | |
95 | * some cases a filesystem may be doing a significant amount of work in | |
96 | * its inode reclaim code, so this should improve parallelism. | |
1da177e4 | 97 | */ |
88e0fbc4 | 98 | static DECLARE_RWSEM(iprune_sem); |
1da177e4 LT |
99 | |
100 | /* | |
101 | * Statistics gathering.. | |
102 | */ | |
103 | struct inodes_stat_t inodes_stat; | |
104 | ||
3e880fb5 | 105 | static DEFINE_PER_CPU(unsigned int, nr_inodes); |
cffbc8aa | 106 | |
6b3304b5 | 107 | static struct kmem_cache *inode_cachep __read_mostly; |
1da177e4 | 108 | |
3e880fb5 | 109 | static int get_nr_inodes(void) |
cffbc8aa | 110 | { |
3e880fb5 NP |
111 | int i; |
112 | int sum = 0; | |
113 | for_each_possible_cpu(i) | |
114 | sum += per_cpu(nr_inodes, i); | |
115 | return sum < 0 ? 0 : sum; | |
cffbc8aa DC |
116 | } |
117 | ||
118 | static inline int get_nr_inodes_unused(void) | |
119 | { | |
86c8749e | 120 | return inodes_stat.nr_unused; |
cffbc8aa DC |
121 | } |
122 | ||
123 | int get_nr_dirty_inodes(void) | |
124 | { | |
3e880fb5 | 125 | /* not actually dirty inodes, but a wild approximation */ |
cffbc8aa DC |
126 | int nr_dirty = get_nr_inodes() - get_nr_inodes_unused(); |
127 | return nr_dirty > 0 ? nr_dirty : 0; | |
cffbc8aa DC |
128 | } |
129 | ||
130 | /* | |
131 | * Handle nr_inode sysctl | |
132 | */ | |
133 | #ifdef CONFIG_SYSCTL | |
134 | int proc_nr_inodes(ctl_table *table, int write, | |
135 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
136 | { | |
137 | inodes_stat.nr_inodes = get_nr_inodes(); | |
cffbc8aa DC |
138 | return proc_dointvec(table, write, buffer, lenp, ppos); |
139 | } | |
140 | #endif | |
141 | ||
1c0eeaf5 JE |
142 | static void wake_up_inode(struct inode *inode) |
143 | { | |
144 | /* | |
145 | * Prevent speculative execution through spin_unlock(&inode_lock); | |
146 | */ | |
147 | smp_mb(); | |
eaff8079 | 148 | wake_up_bit(&inode->i_state, __I_NEW); |
1c0eeaf5 JE |
149 | } |
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 | |
2e00c97e | 260 | void __destroy_inode(struct inode *inode) |
1da177e4 | 261 | { |
b7542f8c | 262 | BUG_ON(inode_has_buffers(inode)); |
1da177e4 | 263 | security_inode_free(inode); |
3be25f49 | 264 | fsnotify_inode_delete(inode); |
f19d4a8f AV |
265 | #ifdef CONFIG_FS_POSIX_ACL |
266 | if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED) | |
267 | posix_acl_release(inode->i_acl); | |
268 | if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED) | |
269 | posix_acl_release(inode->i_default_acl); | |
270 | #endif | |
3e880fb5 | 271 | this_cpu_dec(nr_inodes); |
2e00c97e CH |
272 | } |
273 | EXPORT_SYMBOL(__destroy_inode); | |
274 | ||
fa0d7e3d NP |
275 | static void i_callback(struct rcu_head *head) |
276 | { | |
277 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
278 | INIT_LIST_HEAD(&inode->i_dentry); | |
279 | kmem_cache_free(inode_cachep, inode); | |
280 | } | |
281 | ||
56b0dacf | 282 | static void destroy_inode(struct inode *inode) |
2e00c97e | 283 | { |
7ccf19a8 | 284 | BUG_ON(!list_empty(&inode->i_lru)); |
2e00c97e | 285 | __destroy_inode(inode); |
1da177e4 LT |
286 | if (inode->i_sb->s_op->destroy_inode) |
287 | inode->i_sb->s_op->destroy_inode(inode); | |
288 | else | |
fa0d7e3d | 289 | call_rcu(&inode->i_rcu, i_callback); |
1da177e4 | 290 | } |
1da177e4 LT |
291 | |
292 | /* | |
293 | * These are initializations that only need to be done | |
294 | * once, because the fields are idempotent across use | |
295 | * of the inode, so let the slab aware of that. | |
296 | */ | |
297 | void inode_init_once(struct inode *inode) | |
298 | { | |
299 | memset(inode, 0, sizeof(*inode)); | |
300 | INIT_HLIST_NODE(&inode->i_hash); | |
301 | INIT_LIST_HEAD(&inode->i_dentry); | |
302 | INIT_LIST_HEAD(&inode->i_devices); | |
7ccf19a8 NP |
303 | INIT_LIST_HEAD(&inode->i_wb_list); |
304 | INIT_LIST_HEAD(&inode->i_lru); | |
1da177e4 | 305 | INIT_RADIX_TREE(&inode->i_data.page_tree, GFP_ATOMIC); |
19fd6231 | 306 | spin_lock_init(&inode->i_data.tree_lock); |
1da177e4 LT |
307 | spin_lock_init(&inode->i_data.i_mmap_lock); |
308 | INIT_LIST_HEAD(&inode->i_data.private_list); | |
309 | spin_lock_init(&inode->i_data.private_lock); | |
310 | INIT_RAW_PRIO_TREE_ROOT(&inode->i_data.i_mmap); | |
311 | INIT_LIST_HEAD(&inode->i_data.i_mmap_nonlinear); | |
1da177e4 | 312 | i_size_ordered_init(inode); |
3be25f49 | 313 | #ifdef CONFIG_FSNOTIFY |
e61ce867 | 314 | INIT_HLIST_HEAD(&inode->i_fsnotify_marks); |
3be25f49 | 315 | #endif |
1da177e4 | 316 | } |
1da177e4 LT |
317 | EXPORT_SYMBOL(inode_init_once); |
318 | ||
51cc5068 | 319 | static void init_once(void *foo) |
1da177e4 | 320 | { |
6b3304b5 | 321 | struct inode *inode = (struct inode *) foo; |
1da177e4 | 322 | |
a35afb83 | 323 | inode_init_once(inode); |
1da177e4 LT |
324 | } |
325 | ||
326 | /* | |
327 | * inode_lock must be held | |
328 | */ | |
6b3304b5 | 329 | void __iget(struct inode *inode) |
1da177e4 | 330 | { |
9e38d86f NP |
331 | atomic_inc(&inode->i_count); |
332 | } | |
2e147f1e | 333 | |
7de9c6ee AV |
334 | /* |
335 | * get additional reference to inode; caller must already hold one. | |
336 | */ | |
337 | void ihold(struct inode *inode) | |
338 | { | |
339 | WARN_ON(atomic_inc_return(&inode->i_count) < 2); | |
340 | } | |
341 | EXPORT_SYMBOL(ihold); | |
342 | ||
9e38d86f NP |
343 | static void inode_lru_list_add(struct inode *inode) |
344 | { | |
7ccf19a8 NP |
345 | if (list_empty(&inode->i_lru)) { |
346 | list_add(&inode->i_lru, &inode_lru); | |
86c8749e | 347 | inodes_stat.nr_unused++; |
9e38d86f NP |
348 | } |
349 | } | |
2e147f1e | 350 | |
9e38d86f NP |
351 | static void inode_lru_list_del(struct inode *inode) |
352 | { | |
7ccf19a8 NP |
353 | if (!list_empty(&inode->i_lru)) { |
354 | list_del_init(&inode->i_lru); | |
86c8749e | 355 | inodes_stat.nr_unused--; |
9e38d86f | 356 | } |
1da177e4 LT |
357 | } |
358 | ||
646ec461 CH |
359 | static inline void __inode_sb_list_add(struct inode *inode) |
360 | { | |
361 | list_add(&inode->i_sb_list, &inode->i_sb->s_inodes); | |
1da177e4 LT |
362 | } |
363 | ||
646ec461 CH |
364 | /** |
365 | * inode_sb_list_add - add inode to the superblock list of inodes | |
366 | * @inode: inode to add | |
367 | */ | |
368 | void inode_sb_list_add(struct inode *inode) | |
369 | { | |
370 | spin_lock(&inode_lock); | |
371 | __inode_sb_list_add(inode); | |
372 | spin_unlock(&inode_lock); | |
373 | } | |
374 | EXPORT_SYMBOL_GPL(inode_sb_list_add); | |
375 | ||
376 | static inline void __inode_sb_list_del(struct inode *inode) | |
377 | { | |
378 | list_del_init(&inode->i_sb_list); | |
379 | } | |
380 | ||
4c51acbc DC |
381 | static unsigned long hash(struct super_block *sb, unsigned long hashval) |
382 | { | |
383 | unsigned long tmp; | |
384 | ||
385 | tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) / | |
386 | L1_CACHE_BYTES; | |
387 | tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> I_HASHBITS); | |
388 | return tmp & I_HASHMASK; | |
389 | } | |
390 | ||
391 | /** | |
392 | * __insert_inode_hash - hash an inode | |
393 | * @inode: unhashed inode | |
394 | * @hashval: unsigned long value used to locate this object in the | |
395 | * inode_hashtable. | |
396 | * | |
397 | * Add an inode to the inode hash for this superblock. | |
398 | */ | |
399 | void __insert_inode_hash(struct inode *inode, unsigned long hashval) | |
400 | { | |
646ec461 CH |
401 | struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval); |
402 | ||
4c51acbc | 403 | spin_lock(&inode_lock); |
646ec461 | 404 | hlist_add_head(&inode->i_hash, b); |
4c51acbc DC |
405 | spin_unlock(&inode_lock); |
406 | } | |
407 | EXPORT_SYMBOL(__insert_inode_hash); | |
408 | ||
409 | /** | |
410 | * __remove_inode_hash - remove an inode from the hash | |
411 | * @inode: inode to unhash | |
412 | * | |
413 | * Remove an inode from the superblock. | |
414 | */ | |
415 | static void __remove_inode_hash(struct inode *inode) | |
416 | { | |
417 | hlist_del_init(&inode->i_hash); | |
418 | } | |
419 | ||
420 | /** | |
421 | * remove_inode_hash - remove an inode from the hash | |
422 | * @inode: inode to unhash | |
423 | * | |
424 | * Remove an inode from the superblock. | |
425 | */ | |
426 | void remove_inode_hash(struct inode *inode) | |
427 | { | |
428 | spin_lock(&inode_lock); | |
429 | hlist_del_init(&inode->i_hash); | |
430 | spin_unlock(&inode_lock); | |
431 | } | |
432 | EXPORT_SYMBOL(remove_inode_hash); | |
433 | ||
b0683aa6 AV |
434 | void end_writeback(struct inode *inode) |
435 | { | |
436 | might_sleep(); | |
437 | BUG_ON(inode->i_data.nrpages); | |
438 | BUG_ON(!list_empty(&inode->i_data.private_list)); | |
439 | BUG_ON(!(inode->i_state & I_FREEING)); | |
440 | BUG_ON(inode->i_state & I_CLEAR); | |
441 | inode_sync_wait(inode); | |
fa0d7e3d | 442 | /* don't need i_lock here, no concurrent mods to i_state */ |
b0683aa6 AV |
443 | inode->i_state = I_FREEING | I_CLEAR; |
444 | } | |
445 | EXPORT_SYMBOL(end_writeback); | |
446 | ||
644da596 | 447 | static void evict(struct inode *inode) |
b4272d4c AV |
448 | { |
449 | const struct super_operations *op = inode->i_sb->s_op; | |
450 | ||
be7ce416 AV |
451 | if (op->evict_inode) { |
452 | op->evict_inode(inode); | |
b4272d4c AV |
453 | } else { |
454 | if (inode->i_data.nrpages) | |
455 | truncate_inode_pages(&inode->i_data, 0); | |
30140837 | 456 | end_writeback(inode); |
b4272d4c | 457 | } |
661074e9 AV |
458 | if (S_ISBLK(inode->i_mode) && inode->i_bdev) |
459 | bd_forget(inode); | |
460 | if (S_ISCHR(inode->i_mode) && inode->i_cdev) | |
461 | cd_forget(inode); | |
b4272d4c AV |
462 | } |
463 | ||
1da177e4 LT |
464 | /* |
465 | * dispose_list - dispose of the contents of a local list | |
466 | * @head: the head of the list to free | |
467 | * | |
468 | * Dispose-list gets a local list with local inodes in it, so it doesn't | |
469 | * need to worry about list corruption and SMP locks. | |
470 | */ | |
471 | static void dispose_list(struct list_head *head) | |
472 | { | |
1da177e4 LT |
473 | while (!list_empty(head)) { |
474 | struct inode *inode; | |
475 | ||
7ccf19a8 NP |
476 | inode = list_first_entry(head, struct inode, i_lru); |
477 | list_del_init(&inode->i_lru); | |
1da177e4 | 478 | |
644da596 | 479 | evict(inode); |
4120db47 AB |
480 | |
481 | spin_lock(&inode_lock); | |
4c51acbc | 482 | __remove_inode_hash(inode); |
646ec461 | 483 | __inode_sb_list_del(inode); |
4120db47 AB |
484 | spin_unlock(&inode_lock); |
485 | ||
486 | wake_up_inode(inode); | |
1da177e4 | 487 | destroy_inode(inode); |
1da177e4 | 488 | } |
1da177e4 LT |
489 | } |
490 | ||
63997e98 AV |
491 | /** |
492 | * evict_inodes - evict all evictable inodes for a superblock | |
493 | * @sb: superblock to operate on | |
494 | * | |
495 | * Make sure that no inodes with zero refcount are retained. This is | |
496 | * called by superblock shutdown after having MS_ACTIVE flag removed, | |
497 | * so any inode reaching zero refcount during or after that call will | |
498 | * be immediately evicted. | |
1da177e4 | 499 | */ |
63997e98 | 500 | void evict_inodes(struct super_block *sb) |
1da177e4 | 501 | { |
63997e98 AV |
502 | struct inode *inode, *next; |
503 | LIST_HEAD(dispose); | |
1da177e4 | 504 | |
63997e98 | 505 | down_write(&iprune_sem); |
1da177e4 | 506 | |
63997e98 AV |
507 | spin_lock(&inode_lock); |
508 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { | |
509 | if (atomic_read(&inode->i_count)) | |
aabb8fdb | 510 | continue; |
63997e98 AV |
511 | |
512 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
513 | WARN_ON(1); | |
1da177e4 LT |
514 | continue; |
515 | } | |
63997e98 AV |
516 | |
517 | inode->i_state |= I_FREEING; | |
518 | ||
519 | /* | |
520 | * Move the inode off the IO lists and LRU once I_FREEING is | |
521 | * set so that it won't get moved back on there if it is dirty. | |
522 | */ | |
523 | list_move(&inode->i_lru, &dispose); | |
524 | list_del_init(&inode->i_wb_list); | |
525 | if (!(inode->i_state & (I_DIRTY | I_SYNC))) | |
86c8749e | 526 | inodes_stat.nr_unused--; |
1da177e4 | 527 | } |
63997e98 AV |
528 | spin_unlock(&inode_lock); |
529 | ||
530 | dispose_list(&dispose); | |
531 | up_write(&iprune_sem); | |
1da177e4 LT |
532 | } |
533 | ||
1da177e4 | 534 | /** |
a0318786 CH |
535 | * invalidate_inodes - attempt to free all inodes on a superblock |
536 | * @sb: superblock to operate on | |
1da177e4 | 537 | * |
a0318786 CH |
538 | * Attempts to free all inodes for a given superblock. If there were any |
539 | * busy inodes return a non-zero value, else zero. | |
1da177e4 | 540 | */ |
6b3304b5 | 541 | int invalidate_inodes(struct super_block *sb) |
1da177e4 | 542 | { |
cffbc8aa | 543 | int busy = 0; |
a0318786 CH |
544 | struct inode *inode, *next; |
545 | LIST_HEAD(dispose); | |
1da177e4 | 546 | |
88e0fbc4 | 547 | down_write(&iprune_sem); |
a0318786 | 548 | |
1da177e4 | 549 | spin_lock(&inode_lock); |
a0318786 | 550 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
63997e98 | 551 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) |
aabb8fdb | 552 | continue; |
99a38919 CH |
553 | if (atomic_read(&inode->i_count)) { |
554 | busy = 1; | |
1da177e4 LT |
555 | continue; |
556 | } | |
99a38919 | 557 | |
99a38919 | 558 | inode->i_state |= I_FREEING; |
7ccf19a8 NP |
559 | |
560 | /* | |
561 | * Move the inode off the IO lists and LRU once I_FREEING is | |
562 | * set so that it won't get moved back on there if it is dirty. | |
563 | */ | |
a0318786 | 564 | list_move(&inode->i_lru, &dispose); |
7ccf19a8 | 565 | list_del_init(&inode->i_wb_list); |
99a38919 | 566 | if (!(inode->i_state & (I_DIRTY | I_SYNC))) |
86c8749e | 567 | inodes_stat.nr_unused--; |
1da177e4 | 568 | } |
1da177e4 LT |
569 | spin_unlock(&inode_lock); |
570 | ||
a0318786 | 571 | dispose_list(&dispose); |
88e0fbc4 | 572 | up_write(&iprune_sem); |
1da177e4 LT |
573 | |
574 | return busy; | |
575 | } | |
1da177e4 LT |
576 | |
577 | static int can_unuse(struct inode *inode) | |
578 | { | |
9e38d86f | 579 | if (inode->i_state & ~I_REFERENCED) |
1da177e4 LT |
580 | return 0; |
581 | if (inode_has_buffers(inode)) | |
582 | return 0; | |
583 | if (atomic_read(&inode->i_count)) | |
584 | return 0; | |
585 | if (inode->i_data.nrpages) | |
586 | return 0; | |
587 | return 1; | |
588 | } | |
589 | ||
590 | /* | |
9e38d86f NP |
591 | * Scan `goal' inodes on the unused list for freeable ones. They are moved to a |
592 | * temporary list and then are freed outside inode_lock by dispose_list(). | |
1da177e4 LT |
593 | * |
594 | * Any inodes which are pinned purely because of attached pagecache have their | |
9e38d86f NP |
595 | * pagecache removed. If the inode has metadata buffers attached to |
596 | * mapping->private_list then try to remove them. | |
1da177e4 | 597 | * |
9e38d86f NP |
598 | * If the inode has the I_REFERENCED flag set, then it means that it has been |
599 | * used recently - the flag is set in iput_final(). When we encounter such an | |
600 | * inode, clear the flag and move it to the back of the LRU so it gets another | |
601 | * pass through the LRU before it gets reclaimed. This is necessary because of | |
602 | * the fact we are doing lazy LRU updates to minimise lock contention so the | |
603 | * LRU does not have strict ordering. Hence we don't want to reclaim inodes | |
604 | * with this flag set because they are the inodes that are out of order. | |
1da177e4 LT |
605 | */ |
606 | static void prune_icache(int nr_to_scan) | |
607 | { | |
608 | LIST_HEAD(freeable); | |
1da177e4 LT |
609 | int nr_scanned; |
610 | unsigned long reap = 0; | |
611 | ||
88e0fbc4 | 612 | down_read(&iprune_sem); |
1da177e4 LT |
613 | spin_lock(&inode_lock); |
614 | for (nr_scanned = 0; nr_scanned < nr_to_scan; nr_scanned++) { | |
615 | struct inode *inode; | |
616 | ||
7ccf19a8 | 617 | if (list_empty(&inode_lru)) |
1da177e4 LT |
618 | break; |
619 | ||
7ccf19a8 | 620 | inode = list_entry(inode_lru.prev, struct inode, i_lru); |
1da177e4 | 621 | |
9e38d86f NP |
622 | /* |
623 | * Referenced or dirty inodes are still in use. Give them | |
624 | * another pass through the LRU as we canot reclaim them now. | |
625 | */ | |
626 | if (atomic_read(&inode->i_count) || | |
627 | (inode->i_state & ~I_REFERENCED)) { | |
7ccf19a8 | 628 | list_del_init(&inode->i_lru); |
86c8749e | 629 | inodes_stat.nr_unused--; |
9e38d86f NP |
630 | continue; |
631 | } | |
632 | ||
633 | /* recently referenced inodes get one more pass */ | |
634 | if (inode->i_state & I_REFERENCED) { | |
7ccf19a8 | 635 | list_move(&inode->i_lru, &inode_lru); |
9e38d86f | 636 | inode->i_state &= ~I_REFERENCED; |
1da177e4 LT |
637 | continue; |
638 | } | |
639 | if (inode_has_buffers(inode) || inode->i_data.nrpages) { | |
640 | __iget(inode); | |
641 | spin_unlock(&inode_lock); | |
642 | if (remove_inode_buffers(inode)) | |
fc0ecff6 AM |
643 | reap += invalidate_mapping_pages(&inode->i_data, |
644 | 0, -1); | |
1da177e4 LT |
645 | iput(inode); |
646 | spin_lock(&inode_lock); | |
647 | ||
7ccf19a8 NP |
648 | if (inode != list_entry(inode_lru.next, |
649 | struct inode, i_lru)) | |
1da177e4 LT |
650 | continue; /* wrong inode or list_empty */ |
651 | if (!can_unuse(inode)) | |
652 | continue; | |
653 | } | |
7ef0d737 | 654 | WARN_ON(inode->i_state & I_NEW); |
1da177e4 | 655 | inode->i_state |= I_FREEING; |
7ccf19a8 NP |
656 | |
657 | /* | |
658 | * Move the inode off the IO lists and LRU once I_FREEING is | |
659 | * set so that it won't get moved back on there if it is dirty. | |
660 | */ | |
661 | list_move(&inode->i_lru, &freeable); | |
662 | list_del_init(&inode->i_wb_list); | |
86c8749e | 663 | inodes_stat.nr_unused--; |
1da177e4 | 664 | } |
f8891e5e CL |
665 | if (current_is_kswapd()) |
666 | __count_vm_events(KSWAPD_INODESTEAL, reap); | |
667 | else | |
668 | __count_vm_events(PGINODESTEAL, reap); | |
1da177e4 LT |
669 | spin_unlock(&inode_lock); |
670 | ||
671 | dispose_list(&freeable); | |
88e0fbc4 | 672 | up_read(&iprune_sem); |
1da177e4 LT |
673 | } |
674 | ||
675 | /* | |
676 | * shrink_icache_memory() will attempt to reclaim some unused inodes. Here, | |
677 | * "unused" means that no dentries are referring to the inodes: the files are | |
678 | * not open and the dcache references to those inodes have already been | |
679 | * reclaimed. | |
680 | * | |
681 | * This function is passed the number of inodes to scan, and it returns the | |
682 | * total number of remaining possibly-reclaimable inodes. | |
683 | */ | |
7f8275d0 | 684 | static int shrink_icache_memory(struct shrinker *shrink, int nr, gfp_t gfp_mask) |
1da177e4 LT |
685 | { |
686 | if (nr) { | |
687 | /* | |
688 | * Nasty deadlock avoidance. We may hold various FS locks, | |
689 | * and we don't want to recurse into the FS that called us | |
690 | * in clear_inode() and friends.. | |
6b3304b5 | 691 | */ |
1da177e4 LT |
692 | if (!(gfp_mask & __GFP_FS)) |
693 | return -1; | |
694 | prune_icache(nr); | |
695 | } | |
cffbc8aa | 696 | return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure; |
1da177e4 LT |
697 | } |
698 | ||
8e1f936b RR |
699 | static struct shrinker icache_shrinker = { |
700 | .shrink = shrink_icache_memory, | |
701 | .seeks = DEFAULT_SEEKS, | |
702 | }; | |
703 | ||
1da177e4 LT |
704 | static void __wait_on_freeing_inode(struct inode *inode); |
705 | /* | |
706 | * Called with the inode lock held. | |
1da177e4 | 707 | */ |
6b3304b5 MK |
708 | static struct inode *find_inode(struct super_block *sb, |
709 | struct hlist_head *head, | |
710 | int (*test)(struct inode *, void *), | |
711 | void *data) | |
1da177e4 LT |
712 | { |
713 | struct hlist_node *node; | |
6b3304b5 | 714 | struct inode *inode = NULL; |
1da177e4 LT |
715 | |
716 | repeat: | |
c5c8be3c | 717 | hlist_for_each_entry(inode, node, head, i_hash) { |
1da177e4 LT |
718 | if (inode->i_sb != sb) |
719 | continue; | |
720 | if (!test(inode, data)) | |
721 | continue; | |
a4ffdde6 | 722 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
723 | __wait_on_freeing_inode(inode); |
724 | goto repeat; | |
725 | } | |
f7899bd5 CH |
726 | __iget(inode); |
727 | return inode; | |
1da177e4 | 728 | } |
f7899bd5 | 729 | return NULL; |
1da177e4 LT |
730 | } |
731 | ||
732 | /* | |
733 | * find_inode_fast is the fast path version of find_inode, see the comment at | |
734 | * iget_locked for details. | |
735 | */ | |
6b3304b5 MK |
736 | static struct inode *find_inode_fast(struct super_block *sb, |
737 | struct hlist_head *head, unsigned long ino) | |
1da177e4 LT |
738 | { |
739 | struct hlist_node *node; | |
6b3304b5 | 740 | struct inode *inode = NULL; |
1da177e4 LT |
741 | |
742 | repeat: | |
c5c8be3c | 743 | hlist_for_each_entry(inode, node, head, i_hash) { |
1da177e4 LT |
744 | if (inode->i_ino != ino) |
745 | continue; | |
746 | if (inode->i_sb != sb) | |
747 | continue; | |
a4ffdde6 | 748 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
749 | __wait_on_freeing_inode(inode); |
750 | goto repeat; | |
751 | } | |
f7899bd5 CH |
752 | __iget(inode); |
753 | return inode; | |
1da177e4 | 754 | } |
f7899bd5 | 755 | return NULL; |
8290c35f DC |
756 | } |
757 | ||
f991bd2e ED |
758 | /* |
759 | * Each cpu owns a range of LAST_INO_BATCH numbers. | |
760 | * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations, | |
761 | * to renew the exhausted range. | |
8290c35f | 762 | * |
f991bd2e ED |
763 | * This does not significantly increase overflow rate because every CPU can |
764 | * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is | |
765 | * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the | |
766 | * 2^32 range, and is a worst-case. Even a 50% wastage would only increase | |
767 | * overflow rate by 2x, which does not seem too significant. | |
768 | * | |
769 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
770 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
771 | * here to attempt to avoid that. | |
8290c35f | 772 | */ |
f991bd2e ED |
773 | #define LAST_INO_BATCH 1024 |
774 | static DEFINE_PER_CPU(unsigned int, last_ino); | |
775 | ||
85fe4025 | 776 | unsigned int get_next_ino(void) |
8290c35f | 777 | { |
f991bd2e ED |
778 | unsigned int *p = &get_cpu_var(last_ino); |
779 | unsigned int res = *p; | |
8290c35f | 780 | |
f991bd2e ED |
781 | #ifdef CONFIG_SMP |
782 | if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) { | |
783 | static atomic_t shared_last_ino; | |
784 | int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino); | |
785 | ||
786 | res = next - LAST_INO_BATCH; | |
787 | } | |
788 | #endif | |
789 | ||
790 | *p = ++res; | |
791 | put_cpu_var(last_ino); | |
792 | return res; | |
8290c35f | 793 | } |
85fe4025 | 794 | EXPORT_SYMBOL(get_next_ino); |
8290c35f | 795 | |
1da177e4 LT |
796 | /** |
797 | * new_inode - obtain an inode | |
798 | * @sb: superblock | |
799 | * | |
769848c0 | 800 | * Allocates a new inode for given superblock. The default gfp_mask |
3c1d4378 | 801 | * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE. |
769848c0 MG |
802 | * If HIGHMEM pages are unsuitable or it is known that pages allocated |
803 | * for the page cache are not reclaimable or migratable, | |
804 | * mapping_set_gfp_mask() must be called with suitable flags on the | |
805 | * newly created inode's mapping | |
806 | * | |
1da177e4 LT |
807 | */ |
808 | struct inode *new_inode(struct super_block *sb) | |
809 | { | |
6b3304b5 | 810 | struct inode *inode; |
1da177e4 LT |
811 | |
812 | spin_lock_prefetch(&inode_lock); | |
6b3304b5 | 813 | |
1da177e4 LT |
814 | inode = alloc_inode(sb); |
815 | if (inode) { | |
816 | spin_lock(&inode_lock); | |
646ec461 | 817 | __inode_sb_list_add(inode); |
1da177e4 LT |
818 | inode->i_state = 0; |
819 | spin_unlock(&inode_lock); | |
820 | } | |
821 | return inode; | |
822 | } | |
1da177e4 LT |
823 | EXPORT_SYMBOL(new_inode); |
824 | ||
825 | void unlock_new_inode(struct inode *inode) | |
826 | { | |
14358e6d | 827 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
a3314a0e | 828 | if (S_ISDIR(inode->i_mode)) { |
1e89a5e1 PZ |
829 | struct file_system_type *type = inode->i_sb->s_type; |
830 | ||
9a7aa12f JK |
831 | /* Set new key only if filesystem hasn't already changed it */ |
832 | if (!lockdep_match_class(&inode->i_mutex, | |
833 | &type->i_mutex_key)) { | |
834 | /* | |
835 | * ensure nobody is actually holding i_mutex | |
836 | */ | |
837 | mutex_destroy(&inode->i_mutex); | |
838 | mutex_init(&inode->i_mutex); | |
839 | lockdep_set_class(&inode->i_mutex, | |
840 | &type->i_mutex_dir_key); | |
841 | } | |
1e89a5e1 | 842 | } |
14358e6d | 843 | #endif |
1da177e4 | 844 | /* |
eaff8079 | 845 | * This is special! We do not need the spinlock when clearing I_NEW, |
580be083 JK |
846 | * because we're guaranteed that nobody else tries to do anything about |
847 | * the state of the inode when it is locked, as we just created it (so | |
eaff8079 | 848 | * there can be no old holders that haven't tested I_NEW). |
580be083 | 849 | * However we must emit the memory barrier so that other CPUs reliably |
eaff8079 | 850 | * see the clearing of I_NEW after the other inode initialisation has |
580be083 | 851 | * completed. |
1da177e4 | 852 | */ |
580be083 | 853 | smp_mb(); |
eaff8079 CH |
854 | WARN_ON(!(inode->i_state & I_NEW)); |
855 | inode->i_state &= ~I_NEW; | |
1da177e4 LT |
856 | wake_up_inode(inode); |
857 | } | |
1da177e4 LT |
858 | EXPORT_SYMBOL(unlock_new_inode); |
859 | ||
860 | /* | |
861 | * This is called without the inode lock held.. Be careful. | |
862 | * | |
863 | * We no longer cache the sb_flags in i_flags - see fs.h | |
864 | * -- rmk@arm.uk.linux.org | |
865 | */ | |
6b3304b5 MK |
866 | static struct inode *get_new_inode(struct super_block *sb, |
867 | struct hlist_head *head, | |
868 | int (*test)(struct inode *, void *), | |
869 | int (*set)(struct inode *, void *), | |
870 | void *data) | |
1da177e4 | 871 | { |
6b3304b5 | 872 | struct inode *inode; |
1da177e4 LT |
873 | |
874 | inode = alloc_inode(sb); | |
875 | if (inode) { | |
6b3304b5 | 876 | struct inode *old; |
1da177e4 LT |
877 | |
878 | spin_lock(&inode_lock); | |
879 | /* We released the lock, so.. */ | |
880 | old = find_inode(sb, head, test, data); | |
881 | if (!old) { | |
882 | if (set(inode, data)) | |
883 | goto set_failed; | |
884 | ||
646ec461 CH |
885 | hlist_add_head(&inode->i_hash, head); |
886 | __inode_sb_list_add(inode); | |
eaff8079 | 887 | inode->i_state = I_NEW; |
1da177e4 LT |
888 | spin_unlock(&inode_lock); |
889 | ||
890 | /* Return the locked inode with I_NEW set, the | |
891 | * caller is responsible for filling in the contents | |
892 | */ | |
893 | return inode; | |
894 | } | |
895 | ||
896 | /* | |
897 | * Uhhuh, somebody else created the same inode under | |
898 | * us. Use the old inode instead of the one we just | |
899 | * allocated. | |
900 | */ | |
1da177e4 LT |
901 | spin_unlock(&inode_lock); |
902 | destroy_inode(inode); | |
903 | inode = old; | |
904 | wait_on_inode(inode); | |
905 | } | |
906 | return inode; | |
907 | ||
908 | set_failed: | |
909 | spin_unlock(&inode_lock); | |
910 | destroy_inode(inode); | |
911 | return NULL; | |
912 | } | |
913 | ||
914 | /* | |
915 | * get_new_inode_fast is the fast path version of get_new_inode, see the | |
916 | * comment at iget_locked for details. | |
917 | */ | |
6b3304b5 MK |
918 | static struct inode *get_new_inode_fast(struct super_block *sb, |
919 | struct hlist_head *head, unsigned long ino) | |
1da177e4 | 920 | { |
6b3304b5 | 921 | struct inode *inode; |
1da177e4 LT |
922 | |
923 | inode = alloc_inode(sb); | |
924 | if (inode) { | |
6b3304b5 | 925 | struct inode *old; |
1da177e4 LT |
926 | |
927 | spin_lock(&inode_lock); | |
928 | /* We released the lock, so.. */ | |
929 | old = find_inode_fast(sb, head, ino); | |
930 | if (!old) { | |
931 | inode->i_ino = ino; | |
646ec461 CH |
932 | hlist_add_head(&inode->i_hash, head); |
933 | __inode_sb_list_add(inode); | |
eaff8079 | 934 | inode->i_state = I_NEW; |
1da177e4 LT |
935 | spin_unlock(&inode_lock); |
936 | ||
937 | /* Return the locked inode with I_NEW set, the | |
938 | * caller is responsible for filling in the contents | |
939 | */ | |
940 | return inode; | |
941 | } | |
942 | ||
943 | /* | |
944 | * Uhhuh, somebody else created the same inode under | |
945 | * us. Use the old inode instead of the one we just | |
946 | * allocated. | |
947 | */ | |
1da177e4 LT |
948 | spin_unlock(&inode_lock); |
949 | destroy_inode(inode); | |
950 | inode = old; | |
951 | wait_on_inode(inode); | |
952 | } | |
953 | return inode; | |
954 | } | |
955 | ||
ad5e195a CH |
956 | /* |
957 | * search the inode cache for a matching inode number. | |
958 | * If we find one, then the inode number we are trying to | |
959 | * allocate is not unique and so we should not use it. | |
960 | * | |
961 | * Returns 1 if the inode number is unique, 0 if it is not. | |
962 | */ | |
963 | static int test_inode_iunique(struct super_block *sb, unsigned long ino) | |
964 | { | |
965 | struct hlist_head *b = inode_hashtable + hash(sb, ino); | |
966 | struct hlist_node *node; | |
967 | struct inode *inode; | |
968 | ||
969 | hlist_for_each_entry(inode, node, b, i_hash) { | |
970 | if (inode->i_ino == ino && inode->i_sb == sb) | |
971 | return 0; | |
972 | } | |
973 | ||
974 | return 1; | |
975 | } | |
976 | ||
1da177e4 LT |
977 | /** |
978 | * iunique - get a unique inode number | |
979 | * @sb: superblock | |
980 | * @max_reserved: highest reserved inode number | |
981 | * | |
982 | * Obtain an inode number that is unique on the system for a given | |
983 | * superblock. This is used by file systems that have no natural | |
984 | * permanent inode numbering system. An inode number is returned that | |
985 | * is higher than the reserved limit but unique. | |
986 | * | |
987 | * BUGS: | |
988 | * With a large number of inodes live on the file system this function | |
989 | * currently becomes quite slow. | |
990 | */ | |
991 | ino_t iunique(struct super_block *sb, ino_t max_reserved) | |
992 | { | |
866b04fc JL |
993 | /* |
994 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
995 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
996 | * here to attempt to avoid that. | |
997 | */ | |
ad5e195a | 998 | static DEFINE_SPINLOCK(iunique_lock); |
866b04fc | 999 | static unsigned int counter; |
1da177e4 | 1000 | ino_t res; |
3361c7be | 1001 | |
1da177e4 | 1002 | spin_lock(&inode_lock); |
ad5e195a | 1003 | spin_lock(&iunique_lock); |
3361c7be JL |
1004 | do { |
1005 | if (counter <= max_reserved) | |
1006 | counter = max_reserved + 1; | |
1da177e4 | 1007 | res = counter++; |
ad5e195a CH |
1008 | } while (!test_inode_iunique(sb, res)); |
1009 | spin_unlock(&iunique_lock); | |
3361c7be | 1010 | spin_unlock(&inode_lock); |
1da177e4 | 1011 | |
3361c7be JL |
1012 | return res; |
1013 | } | |
1da177e4 LT |
1014 | EXPORT_SYMBOL(iunique); |
1015 | ||
1016 | struct inode *igrab(struct inode *inode) | |
1017 | { | |
1018 | spin_lock(&inode_lock); | |
a4ffdde6 | 1019 | if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) |
1da177e4 LT |
1020 | __iget(inode); |
1021 | else | |
1022 | /* | |
1023 | * Handle the case where s_op->clear_inode is not been | |
1024 | * called yet, and somebody is calling igrab | |
1025 | * while the inode is getting freed. | |
1026 | */ | |
1027 | inode = NULL; | |
1028 | spin_unlock(&inode_lock); | |
1029 | return inode; | |
1030 | } | |
1da177e4 LT |
1031 | EXPORT_SYMBOL(igrab); |
1032 | ||
1033 | /** | |
1034 | * ifind - internal function, you want ilookup5() or iget5(). | |
1035 | * @sb: super block of file system to search | |
1036 | * @head: the head of the list to search | |
1037 | * @test: callback used for comparisons between inodes | |
1038 | * @data: opaque data pointer to pass to @test | |
88bd5121 | 1039 | * @wait: if true wait for the inode to be unlocked, if false do not |
1da177e4 LT |
1040 | * |
1041 | * ifind() searches for the inode specified by @data in the inode | |
1042 | * cache. This is a generalized version of ifind_fast() for file systems where | |
1043 | * the inode number is not sufficient for unique identification of an inode. | |
1044 | * | |
1045 | * If the inode is in the cache, the inode is returned with an incremented | |
1046 | * reference count. | |
1047 | * | |
1048 | * Otherwise NULL is returned. | |
1049 | * | |
1050 | * Note, @test is called with the inode_lock held, so can't sleep. | |
1051 | */ | |
5d2bea45 | 1052 | static struct inode *ifind(struct super_block *sb, |
1da177e4 | 1053 | struct hlist_head *head, int (*test)(struct inode *, void *), |
88bd5121 | 1054 | void *data, const int wait) |
1da177e4 LT |
1055 | { |
1056 | struct inode *inode; | |
1057 | ||
1058 | spin_lock(&inode_lock); | |
1059 | inode = find_inode(sb, head, test, data); | |
1060 | if (inode) { | |
1da177e4 | 1061 | spin_unlock(&inode_lock); |
88bd5121 AA |
1062 | if (likely(wait)) |
1063 | wait_on_inode(inode); | |
1da177e4 LT |
1064 | return inode; |
1065 | } | |
1066 | spin_unlock(&inode_lock); | |
1067 | return NULL; | |
1068 | } | |
1069 | ||
1070 | /** | |
1071 | * ifind_fast - internal function, you want ilookup() or iget(). | |
1072 | * @sb: super block of file system to search | |
1073 | * @head: head of the list to search | |
1074 | * @ino: inode number to search for | |
1075 | * | |
1076 | * ifind_fast() searches for the inode @ino in the inode cache. This is for | |
1077 | * file systems where the inode number is sufficient for unique identification | |
1078 | * of an inode. | |
1079 | * | |
1080 | * If the inode is in the cache, the inode is returned with an incremented | |
1081 | * reference count. | |
1082 | * | |
1083 | * Otherwise NULL is returned. | |
1084 | */ | |
5d2bea45 | 1085 | static struct inode *ifind_fast(struct super_block *sb, |
1da177e4 LT |
1086 | struct hlist_head *head, unsigned long ino) |
1087 | { | |
1088 | struct inode *inode; | |
1089 | ||
1090 | spin_lock(&inode_lock); | |
1091 | inode = find_inode_fast(sb, head, ino); | |
1092 | if (inode) { | |
1da177e4 LT |
1093 | spin_unlock(&inode_lock); |
1094 | wait_on_inode(inode); | |
1095 | return inode; | |
1096 | } | |
1097 | spin_unlock(&inode_lock); | |
1098 | return NULL; | |
1099 | } | |
1100 | ||
1101 | /** | |
88bd5121 | 1102 | * ilookup5_nowait - search for an inode in the inode cache |
1da177e4 LT |
1103 | * @sb: super block of file system to search |
1104 | * @hashval: hash value (usually inode number) to search for | |
1105 | * @test: callback used for comparisons between inodes | |
1106 | * @data: opaque data pointer to pass to @test | |
1107 | * | |
1108 | * ilookup5() uses ifind() to search for the inode specified by @hashval and | |
1109 | * @data in the inode cache. This is a generalized version of ilookup() for | |
1110 | * file systems where the inode number is not sufficient for unique | |
1111 | * identification of an inode. | |
1112 | * | |
1113 | * If the inode is in the cache, the inode is returned with an incremented | |
88bd5121 AA |
1114 | * reference count. Note, the inode lock is not waited upon so you have to be |
1115 | * very careful what you do with the returned inode. You probably should be | |
1116 | * using ilookup5() instead. | |
1117 | * | |
1118 | * Otherwise NULL is returned. | |
1119 | * | |
1120 | * Note, @test is called with the inode_lock held, so can't sleep. | |
1121 | */ | |
1122 | struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval, | |
1123 | int (*test)(struct inode *, void *), void *data) | |
1124 | { | |
1125 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1126 | ||
1127 | return ifind(sb, head, test, data, 0); | |
1128 | } | |
88bd5121 AA |
1129 | EXPORT_SYMBOL(ilookup5_nowait); |
1130 | ||
1131 | /** | |
1132 | * ilookup5 - search for an inode in the inode cache | |
1133 | * @sb: super block of file system to search | |
1134 | * @hashval: hash value (usually inode number) to search for | |
1135 | * @test: callback used for comparisons between inodes | |
1136 | * @data: opaque data pointer to pass to @test | |
1137 | * | |
1138 | * ilookup5() uses ifind() to search for the inode specified by @hashval and | |
1139 | * @data in the inode cache. This is a generalized version of ilookup() for | |
1140 | * file systems where the inode number is not sufficient for unique | |
1141 | * identification of an inode. | |
1142 | * | |
1143 | * If the inode is in the cache, the inode lock is waited upon and the inode is | |
1144 | * returned with an incremented reference count. | |
1da177e4 LT |
1145 | * |
1146 | * Otherwise NULL is returned. | |
1147 | * | |
1148 | * Note, @test is called with the inode_lock held, so can't sleep. | |
1149 | */ | |
1150 | struct inode *ilookup5(struct super_block *sb, unsigned long hashval, | |
1151 | int (*test)(struct inode *, void *), void *data) | |
1152 | { | |
1153 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1154 | ||
88bd5121 | 1155 | return ifind(sb, head, test, data, 1); |
1da177e4 | 1156 | } |
1da177e4 LT |
1157 | EXPORT_SYMBOL(ilookup5); |
1158 | ||
1159 | /** | |
1160 | * ilookup - search for an inode in the inode cache | |
1161 | * @sb: super block of file system to search | |
1162 | * @ino: inode number to search for | |
1163 | * | |
1164 | * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache. | |
1165 | * This is for file systems where the inode number is sufficient for unique | |
1166 | * identification of an inode. | |
1167 | * | |
1168 | * If the inode is in the cache, the inode is returned with an incremented | |
1169 | * reference count. | |
1170 | * | |
1171 | * Otherwise NULL is returned. | |
1172 | */ | |
1173 | struct inode *ilookup(struct super_block *sb, unsigned long ino) | |
1174 | { | |
1175 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1176 | ||
1177 | return ifind_fast(sb, head, ino); | |
1178 | } | |
1da177e4 LT |
1179 | EXPORT_SYMBOL(ilookup); |
1180 | ||
1181 | /** | |
1182 | * iget5_locked - obtain an inode from a mounted file system | |
1183 | * @sb: super block of file system | |
1184 | * @hashval: hash value (usually inode number) to get | |
1185 | * @test: callback used for comparisons between inodes | |
1186 | * @set: callback used to initialize a new struct inode | |
1187 | * @data: opaque data pointer to pass to @test and @set | |
1188 | * | |
1da177e4 LT |
1189 | * iget5_locked() uses ifind() to search for the inode specified by @hashval |
1190 | * and @data in the inode cache and if present it is returned with an increased | |
1191 | * reference count. This is a generalized version of iget_locked() for file | |
1192 | * systems where the inode number is not sufficient for unique identification | |
1193 | * of an inode. | |
1194 | * | |
1195 | * If the inode is not in cache, get_new_inode() is called to allocate a new | |
1196 | * inode and this is returned locked, hashed, and with the I_NEW flag set. The | |
1197 | * file system gets to fill it in before unlocking it via unlock_new_inode(). | |
1198 | * | |
1199 | * Note both @test and @set are called with the inode_lock held, so can't sleep. | |
1200 | */ | |
1201 | struct inode *iget5_locked(struct super_block *sb, unsigned long hashval, | |
1202 | int (*test)(struct inode *, void *), | |
1203 | int (*set)(struct inode *, void *), void *data) | |
1204 | { | |
1205 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1206 | struct inode *inode; | |
1207 | ||
88bd5121 | 1208 | inode = ifind(sb, head, test, data, 1); |
1da177e4 LT |
1209 | if (inode) |
1210 | return inode; | |
1211 | /* | |
1212 | * get_new_inode() will do the right thing, re-trying the search | |
1213 | * in case it had to block at any point. | |
1214 | */ | |
1215 | return get_new_inode(sb, head, test, set, data); | |
1216 | } | |
1da177e4 LT |
1217 | EXPORT_SYMBOL(iget5_locked); |
1218 | ||
1219 | /** | |
1220 | * iget_locked - obtain an inode from a mounted file system | |
1221 | * @sb: super block of file system | |
1222 | * @ino: inode number to get | |
1223 | * | |
1da177e4 LT |
1224 | * iget_locked() uses ifind_fast() to search for the inode specified by @ino in |
1225 | * the inode cache and if present it is returned with an increased reference | |
1226 | * count. This is for file systems where the inode number is sufficient for | |
1227 | * unique identification of an inode. | |
1228 | * | |
1229 | * If the inode is not in cache, get_new_inode_fast() is called to allocate a | |
1230 | * new inode and this is returned locked, hashed, and with the I_NEW flag set. | |
1231 | * The file system gets to fill it in before unlocking it via | |
1232 | * unlock_new_inode(). | |
1233 | */ | |
1234 | struct inode *iget_locked(struct super_block *sb, unsigned long ino) | |
1235 | { | |
1236 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1237 | struct inode *inode; | |
1238 | ||
1239 | inode = ifind_fast(sb, head, ino); | |
1240 | if (inode) | |
1241 | return inode; | |
1242 | /* | |
1243 | * get_new_inode_fast() will do the right thing, re-trying the search | |
1244 | * in case it had to block at any point. | |
1245 | */ | |
1246 | return get_new_inode_fast(sb, head, ino); | |
1247 | } | |
1da177e4 LT |
1248 | EXPORT_SYMBOL(iget_locked); |
1249 | ||
261bca86 AV |
1250 | int insert_inode_locked(struct inode *inode) |
1251 | { | |
1252 | struct super_block *sb = inode->i_sb; | |
1253 | ino_t ino = inode->i_ino; | |
1254 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
261bca86 | 1255 | |
eaff8079 | 1256 | inode->i_state |= I_NEW; |
261bca86 | 1257 | while (1) { |
72a43d63 AV |
1258 | struct hlist_node *node; |
1259 | struct inode *old = NULL; | |
261bca86 | 1260 | spin_lock(&inode_lock); |
72a43d63 AV |
1261 | hlist_for_each_entry(old, node, head, i_hash) { |
1262 | if (old->i_ino != ino) | |
1263 | continue; | |
1264 | if (old->i_sb != sb) | |
1265 | continue; | |
a4ffdde6 | 1266 | if (old->i_state & (I_FREEING|I_WILL_FREE)) |
72a43d63 AV |
1267 | continue; |
1268 | break; | |
1269 | } | |
1270 | if (likely(!node)) { | |
261bca86 AV |
1271 | hlist_add_head(&inode->i_hash, head); |
1272 | spin_unlock(&inode_lock); | |
1273 | return 0; | |
1274 | } | |
1275 | __iget(old); | |
1276 | spin_unlock(&inode_lock); | |
1277 | wait_on_inode(old); | |
1d3382cb | 1278 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1279 | iput(old); |
1280 | return -EBUSY; | |
1281 | } | |
1282 | iput(old); | |
1283 | } | |
1284 | } | |
261bca86 AV |
1285 | EXPORT_SYMBOL(insert_inode_locked); |
1286 | ||
1287 | int insert_inode_locked4(struct inode *inode, unsigned long hashval, | |
1288 | int (*test)(struct inode *, void *), void *data) | |
1289 | { | |
1290 | struct super_block *sb = inode->i_sb; | |
1291 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
261bca86 | 1292 | |
eaff8079 | 1293 | inode->i_state |= I_NEW; |
261bca86 AV |
1294 | |
1295 | while (1) { | |
72a43d63 AV |
1296 | struct hlist_node *node; |
1297 | struct inode *old = NULL; | |
1298 | ||
261bca86 | 1299 | spin_lock(&inode_lock); |
72a43d63 AV |
1300 | hlist_for_each_entry(old, node, head, i_hash) { |
1301 | if (old->i_sb != sb) | |
1302 | continue; | |
1303 | if (!test(old, data)) | |
1304 | continue; | |
a4ffdde6 | 1305 | if (old->i_state & (I_FREEING|I_WILL_FREE)) |
72a43d63 AV |
1306 | continue; |
1307 | break; | |
1308 | } | |
1309 | if (likely(!node)) { | |
261bca86 AV |
1310 | hlist_add_head(&inode->i_hash, head); |
1311 | spin_unlock(&inode_lock); | |
1312 | return 0; | |
1313 | } | |
1314 | __iget(old); | |
1315 | spin_unlock(&inode_lock); | |
1316 | wait_on_inode(old); | |
1d3382cb | 1317 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1318 | iput(old); |
1319 | return -EBUSY; | |
1320 | } | |
1321 | iput(old); | |
1322 | } | |
1323 | } | |
261bca86 AV |
1324 | EXPORT_SYMBOL(insert_inode_locked4); |
1325 | ||
1da177e4 | 1326 | |
45321ac5 AV |
1327 | int generic_delete_inode(struct inode *inode) |
1328 | { | |
1329 | return 1; | |
1330 | } | |
1331 | EXPORT_SYMBOL(generic_delete_inode); | |
1332 | ||
1da177e4 | 1333 | /* |
45321ac5 AV |
1334 | * Normal UNIX filesystem behaviour: delete the |
1335 | * inode when the usage count drops to zero, and | |
1336 | * i_nlink is zero. | |
1da177e4 | 1337 | */ |
45321ac5 | 1338 | int generic_drop_inode(struct inode *inode) |
1da177e4 | 1339 | { |
1d3382cb | 1340 | return !inode->i_nlink || inode_unhashed(inode); |
1da177e4 | 1341 | } |
45321ac5 | 1342 | EXPORT_SYMBOL_GPL(generic_drop_inode); |
1da177e4 | 1343 | |
45321ac5 AV |
1344 | /* |
1345 | * Called when we're dropping the last reference | |
1346 | * to an inode. | |
22fe4042 | 1347 | * |
45321ac5 AV |
1348 | * Call the FS "drop_inode()" function, defaulting to |
1349 | * the legacy UNIX filesystem behaviour. If it tells | |
1350 | * us to evict inode, do so. Otherwise, retain inode | |
1351 | * in cache if fs is alive, sync and evict if fs is | |
1352 | * shutting down. | |
22fe4042 | 1353 | */ |
45321ac5 | 1354 | static void iput_final(struct inode *inode) |
1da177e4 LT |
1355 | { |
1356 | struct super_block *sb = inode->i_sb; | |
45321ac5 AV |
1357 | const struct super_operations *op = inode->i_sb->s_op; |
1358 | int drop; | |
1359 | ||
1360 | if (op && op->drop_inode) | |
1361 | drop = op->drop_inode(inode); | |
1362 | else | |
1363 | drop = generic_drop_inode(inode); | |
1da177e4 | 1364 | |
45321ac5 | 1365 | if (!drop) { |
acb0c854 | 1366 | if (sb->s_flags & MS_ACTIVE) { |
9e38d86f NP |
1367 | inode->i_state |= I_REFERENCED; |
1368 | if (!(inode->i_state & (I_DIRTY|I_SYNC))) { | |
1369 | inode_lru_list_add(inode); | |
1370 | } | |
991114c6 | 1371 | spin_unlock(&inode_lock); |
45321ac5 | 1372 | return; |
991114c6 | 1373 | } |
7ef0d737 | 1374 | WARN_ON(inode->i_state & I_NEW); |
991114c6 AV |
1375 | inode->i_state |= I_WILL_FREE; |
1376 | spin_unlock(&inode_lock); | |
1da177e4 LT |
1377 | write_inode_now(inode, 1); |
1378 | spin_lock(&inode_lock); | |
7ef0d737 | 1379 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1380 | inode->i_state &= ~I_WILL_FREE; |
4c51acbc | 1381 | __remove_inode_hash(inode); |
1da177e4 | 1382 | } |
7ccf19a8 | 1383 | |
7ef0d737 | 1384 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1385 | inode->i_state |= I_FREEING; |
9e38d86f NP |
1386 | |
1387 | /* | |
7ccf19a8 NP |
1388 | * Move the inode off the IO lists and LRU once I_FREEING is |
1389 | * set so that it won't get moved back on there if it is dirty. | |
9e38d86f NP |
1390 | */ |
1391 | inode_lru_list_del(inode); | |
7ccf19a8 | 1392 | list_del_init(&inode->i_wb_list); |
9e38d86f | 1393 | |
646ec461 | 1394 | __inode_sb_list_del(inode); |
1da177e4 | 1395 | spin_unlock(&inode_lock); |
644da596 | 1396 | evict(inode); |
4c51acbc | 1397 | remove_inode_hash(inode); |
7f04c26d | 1398 | wake_up_inode(inode); |
45321ac5 | 1399 | BUG_ON(inode->i_state != (I_FREEING | I_CLEAR)); |
1da177e4 LT |
1400 | destroy_inode(inode); |
1401 | } | |
1402 | ||
1da177e4 | 1403 | /** |
6b3304b5 | 1404 | * iput - put an inode |
1da177e4 LT |
1405 | * @inode: inode to put |
1406 | * | |
1407 | * Puts an inode, dropping its usage count. If the inode use count hits | |
1408 | * zero, the inode is then freed and may also be destroyed. | |
1409 | * | |
1410 | * Consequently, iput() can sleep. | |
1411 | */ | |
1412 | void iput(struct inode *inode) | |
1413 | { | |
1414 | if (inode) { | |
a4ffdde6 | 1415 | BUG_ON(inode->i_state & I_CLEAR); |
1da177e4 | 1416 | |
1da177e4 LT |
1417 | if (atomic_dec_and_lock(&inode->i_count, &inode_lock)) |
1418 | iput_final(inode); | |
1419 | } | |
1420 | } | |
1da177e4 LT |
1421 | EXPORT_SYMBOL(iput); |
1422 | ||
1423 | /** | |
1424 | * bmap - find a block number in a file | |
1425 | * @inode: inode of file | |
1426 | * @block: block to find | |
1427 | * | |
1428 | * Returns the block number on the device holding the inode that | |
1429 | * is the disk block number for the block of the file requested. | |
1430 | * That is, asked for block 4 of inode 1 the function will return the | |
6b3304b5 | 1431 | * disk block relative to the disk start that holds that block of the |
1da177e4 LT |
1432 | * file. |
1433 | */ | |
6b3304b5 | 1434 | sector_t bmap(struct inode *inode, sector_t block) |
1da177e4 LT |
1435 | { |
1436 | sector_t res = 0; | |
1437 | if (inode->i_mapping->a_ops->bmap) | |
1438 | res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block); | |
1439 | return res; | |
1440 | } | |
1da177e4 LT |
1441 | EXPORT_SYMBOL(bmap); |
1442 | ||
11ff6f05 MG |
1443 | /* |
1444 | * With relative atime, only update atime if the previous atime is | |
1445 | * earlier than either the ctime or mtime or if at least a day has | |
1446 | * passed since the last atime update. | |
1447 | */ | |
1448 | static int relatime_need_update(struct vfsmount *mnt, struct inode *inode, | |
1449 | struct timespec now) | |
1450 | { | |
1451 | ||
1452 | if (!(mnt->mnt_flags & MNT_RELATIME)) | |
1453 | return 1; | |
1454 | /* | |
1455 | * Is mtime younger than atime? If yes, update atime: | |
1456 | */ | |
1457 | if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0) | |
1458 | return 1; | |
1459 | /* | |
1460 | * Is ctime younger than atime? If yes, update atime: | |
1461 | */ | |
1462 | if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0) | |
1463 | return 1; | |
1464 | ||
1465 | /* | |
1466 | * Is the previous atime value older than a day? If yes, | |
1467 | * update atime: | |
1468 | */ | |
1469 | if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60) | |
1470 | return 1; | |
1471 | /* | |
1472 | * Good, we can skip the atime update: | |
1473 | */ | |
1474 | return 0; | |
1475 | } | |
1476 | ||
1da177e4 | 1477 | /** |
869243a0 CH |
1478 | * touch_atime - update the access time |
1479 | * @mnt: mount the inode is accessed on | |
7045f37b | 1480 | * @dentry: dentry accessed |
1da177e4 LT |
1481 | * |
1482 | * Update the accessed time on an inode and mark it for writeback. | |
1483 | * This function automatically handles read only file systems and media, | |
1484 | * as well as the "noatime" flag and inode specific "noatime" markers. | |
1485 | */ | |
869243a0 | 1486 | void touch_atime(struct vfsmount *mnt, struct dentry *dentry) |
1da177e4 | 1487 | { |
869243a0 | 1488 | struct inode *inode = dentry->d_inode; |
1da177e4 LT |
1489 | struct timespec now; |
1490 | ||
cdb70f3f | 1491 | if (inode->i_flags & S_NOATIME) |
b12536c2 | 1492 | return; |
37756ced | 1493 | if (IS_NOATIME(inode)) |
b12536c2 | 1494 | return; |
b2276138 | 1495 | if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1496 | return; |
47ae32d6 | 1497 | |
cdb70f3f | 1498 | if (mnt->mnt_flags & MNT_NOATIME) |
b12536c2 | 1499 | return; |
cdb70f3f | 1500 | if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1501 | return; |
1da177e4 LT |
1502 | |
1503 | now = current_fs_time(inode->i_sb); | |
11ff6f05 MG |
1504 | |
1505 | if (!relatime_need_update(mnt, inode, now)) | |
b12536c2 | 1506 | return; |
11ff6f05 | 1507 | |
47ae32d6 | 1508 | if (timespec_equal(&inode->i_atime, &now)) |
b12536c2 AK |
1509 | return; |
1510 | ||
1511 | if (mnt_want_write(mnt)) | |
1512 | return; | |
47ae32d6 VH |
1513 | |
1514 | inode->i_atime = now; | |
1515 | mark_inode_dirty_sync(inode); | |
cdb70f3f | 1516 | mnt_drop_write(mnt); |
1da177e4 | 1517 | } |
869243a0 | 1518 | EXPORT_SYMBOL(touch_atime); |
1da177e4 LT |
1519 | |
1520 | /** | |
870f4817 CH |
1521 | * file_update_time - update mtime and ctime time |
1522 | * @file: file accessed | |
1da177e4 | 1523 | * |
870f4817 CH |
1524 | * Update the mtime and ctime members of an inode and mark the inode |
1525 | * for writeback. Note that this function is meant exclusively for | |
1526 | * usage in the file write path of filesystems, and filesystems may | |
1527 | * choose to explicitly ignore update via this function with the | |
2eadfc0e | 1528 | * S_NOCMTIME inode flag, e.g. for network filesystem where these |
870f4817 | 1529 | * timestamps are handled by the server. |
1da177e4 LT |
1530 | */ |
1531 | ||
870f4817 | 1532 | void file_update_time(struct file *file) |
1da177e4 | 1533 | { |
0f7fc9e4 | 1534 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 | 1535 | struct timespec now; |
ce06e0b2 | 1536 | enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0; |
1da177e4 | 1537 | |
ce06e0b2 | 1538 | /* First try to exhaust all avenues to not sync */ |
1da177e4 LT |
1539 | if (IS_NOCMTIME(inode)) |
1540 | return; | |
20ddee2c | 1541 | |
1da177e4 | 1542 | now = current_fs_time(inode->i_sb); |
ce06e0b2 AK |
1543 | if (!timespec_equal(&inode->i_mtime, &now)) |
1544 | sync_it = S_MTIME; | |
1da177e4 | 1545 | |
ce06e0b2 AK |
1546 | if (!timespec_equal(&inode->i_ctime, &now)) |
1547 | sync_it |= S_CTIME; | |
870f4817 | 1548 | |
ce06e0b2 AK |
1549 | if (IS_I_VERSION(inode)) |
1550 | sync_it |= S_VERSION; | |
7a224228 | 1551 | |
ce06e0b2 AK |
1552 | if (!sync_it) |
1553 | return; | |
1554 | ||
1555 | /* Finally allowed to write? Takes lock. */ | |
1556 | if (mnt_want_write_file(file)) | |
1557 | return; | |
1558 | ||
1559 | /* Only change inode inside the lock region */ | |
1560 | if (sync_it & S_VERSION) | |
1561 | inode_inc_iversion(inode); | |
1562 | if (sync_it & S_CTIME) | |
1563 | inode->i_ctime = now; | |
1564 | if (sync_it & S_MTIME) | |
1565 | inode->i_mtime = now; | |
1566 | mark_inode_dirty_sync(inode); | |
20ddee2c | 1567 | mnt_drop_write(file->f_path.mnt); |
1da177e4 | 1568 | } |
870f4817 | 1569 | EXPORT_SYMBOL(file_update_time); |
1da177e4 LT |
1570 | |
1571 | int inode_needs_sync(struct inode *inode) | |
1572 | { | |
1573 | if (IS_SYNC(inode)) | |
1574 | return 1; | |
1575 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) | |
1576 | return 1; | |
1577 | return 0; | |
1578 | } | |
1da177e4 LT |
1579 | EXPORT_SYMBOL(inode_needs_sync); |
1580 | ||
1da177e4 LT |
1581 | int inode_wait(void *word) |
1582 | { | |
1583 | schedule(); | |
1584 | return 0; | |
1585 | } | |
d44dab8d | 1586 | EXPORT_SYMBOL(inode_wait); |
1da177e4 LT |
1587 | |
1588 | /* | |
168a9fd6 MS |
1589 | * If we try to find an inode in the inode hash while it is being |
1590 | * deleted, we have to wait until the filesystem completes its | |
1591 | * deletion before reporting that it isn't found. This function waits | |
1592 | * until the deletion _might_ have completed. Callers are responsible | |
1593 | * to recheck inode state. | |
1594 | * | |
eaff8079 | 1595 | * It doesn't matter if I_NEW is not set initially, a call to |
168a9fd6 MS |
1596 | * wake_up_inode() after removing from the hash list will DTRT. |
1597 | * | |
1da177e4 LT |
1598 | * This is called with inode_lock held. |
1599 | */ | |
1600 | static void __wait_on_freeing_inode(struct inode *inode) | |
1601 | { | |
1602 | wait_queue_head_t *wq; | |
eaff8079 CH |
1603 | DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW); |
1604 | wq = bit_waitqueue(&inode->i_state, __I_NEW); | |
1da177e4 LT |
1605 | prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE); |
1606 | spin_unlock(&inode_lock); | |
1607 | schedule(); | |
1608 | finish_wait(wq, &wait.wait); | |
1609 | spin_lock(&inode_lock); | |
1610 | } | |
1611 | ||
1da177e4 LT |
1612 | static __initdata unsigned long ihash_entries; |
1613 | static int __init set_ihash_entries(char *str) | |
1614 | { | |
1615 | if (!str) | |
1616 | return 0; | |
1617 | ihash_entries = simple_strtoul(str, &str, 0); | |
1618 | return 1; | |
1619 | } | |
1620 | __setup("ihash_entries=", set_ihash_entries); | |
1621 | ||
1622 | /* | |
1623 | * Initialize the waitqueues and inode hash table. | |
1624 | */ | |
1625 | void __init inode_init_early(void) | |
1626 | { | |
1627 | int loop; | |
1628 | ||
1629 | /* If hashes are distributed across NUMA nodes, defer | |
1630 | * hash allocation until vmalloc space is available. | |
1631 | */ | |
1632 | if (hashdist) | |
1633 | return; | |
1634 | ||
1635 | inode_hashtable = | |
1636 | alloc_large_system_hash("Inode-cache", | |
1637 | sizeof(struct hlist_head), | |
1638 | ihash_entries, | |
1639 | 14, | |
1640 | HASH_EARLY, | |
1641 | &i_hash_shift, | |
1642 | &i_hash_mask, | |
1643 | 0); | |
1644 | ||
1645 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1646 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1647 | } | |
1648 | ||
74bf17cf | 1649 | void __init inode_init(void) |
1da177e4 LT |
1650 | { |
1651 | int loop; | |
1652 | ||
1653 | /* inode slab cache */ | |
b0196009 PJ |
1654 | inode_cachep = kmem_cache_create("inode_cache", |
1655 | sizeof(struct inode), | |
1656 | 0, | |
1657 | (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC| | |
1658 | SLAB_MEM_SPREAD), | |
20c2df83 | 1659 | init_once); |
8e1f936b | 1660 | register_shrinker(&icache_shrinker); |
1da177e4 LT |
1661 | |
1662 | /* Hash may have been set up in inode_init_early */ | |
1663 | if (!hashdist) | |
1664 | return; | |
1665 | ||
1666 | inode_hashtable = | |
1667 | alloc_large_system_hash("Inode-cache", | |
1668 | sizeof(struct hlist_head), | |
1669 | ihash_entries, | |
1670 | 14, | |
1671 | 0, | |
1672 | &i_hash_shift, | |
1673 | &i_hash_mask, | |
1674 | 0); | |
1675 | ||
1676 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1677 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1678 | } | |
1679 | ||
1680 | void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev) | |
1681 | { | |
1682 | inode->i_mode = mode; | |
1683 | if (S_ISCHR(mode)) { | |
1684 | inode->i_fop = &def_chr_fops; | |
1685 | inode->i_rdev = rdev; | |
1686 | } else if (S_ISBLK(mode)) { | |
1687 | inode->i_fop = &def_blk_fops; | |
1688 | inode->i_rdev = rdev; | |
1689 | } else if (S_ISFIFO(mode)) | |
1690 | inode->i_fop = &def_fifo_fops; | |
1691 | else if (S_ISSOCK(mode)) | |
1692 | inode->i_fop = &bad_sock_fops; | |
1693 | else | |
af0d9ae8 MK |
1694 | printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for" |
1695 | " inode %s:%lu\n", mode, inode->i_sb->s_id, | |
1696 | inode->i_ino); | |
1da177e4 LT |
1697 | } |
1698 | EXPORT_SYMBOL(init_special_inode); | |
a1bd120d DM |
1699 | |
1700 | /** | |
1701 | * Init uid,gid,mode for new inode according to posix standards | |
1702 | * @inode: New inode | |
1703 | * @dir: Directory inode | |
1704 | * @mode: mode of the new inode | |
1705 | */ | |
1706 | void inode_init_owner(struct inode *inode, const struct inode *dir, | |
1707 | mode_t mode) | |
1708 | { | |
1709 | inode->i_uid = current_fsuid(); | |
1710 | if (dir && dir->i_mode & S_ISGID) { | |
1711 | inode->i_gid = dir->i_gid; | |
1712 | if (S_ISDIR(mode)) | |
1713 | mode |= S_ISGID; | |
1714 | } else | |
1715 | inode->i_gid = current_fsgid(); | |
1716 | inode->i_mode = mode; | |
1717 | } | |
1718 | EXPORT_SYMBOL(inode_init_owner); |