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> |
9ce6e0be | 27 | #include <linux/prefetch.h> |
a178d202 | 28 | #include <linux/ima.h> |
e795b717 | 29 | #include <linux/cred.h> |
a66979ab | 30 | #include "internal.h" |
1da177e4 | 31 | |
250df6ed DC |
32 | /* |
33 | * inode locking rules. | |
34 | * | |
35 | * inode->i_lock protects: | |
36 | * inode->i_state, inode->i_hash, __iget() | |
02afc410 DC |
37 | * inode_lru_lock protects: |
38 | * inode_lru, inode->i_lru | |
55fa6091 DC |
39 | * inode_sb_list_lock protects: |
40 | * sb->s_inodes, inode->i_sb_list | |
a66979ab DC |
41 | * inode_wb_list_lock protects: |
42 | * bdi->wb.b_{dirty,io,more_io}, inode->i_wb_list | |
67a23c49 DC |
43 | * inode_hash_lock protects: |
44 | * inode_hashtable, inode->i_hash | |
250df6ed DC |
45 | * |
46 | * Lock ordering: | |
55fa6091 DC |
47 | * |
48 | * inode_sb_list_lock | |
49 | * inode->i_lock | |
02afc410 | 50 | * inode_lru_lock |
a66979ab DC |
51 | * |
52 | * inode_wb_list_lock | |
53 | * inode->i_lock | |
67a23c49 DC |
54 | * |
55 | * inode_hash_lock | |
56 | * inode_sb_list_lock | |
57 | * inode->i_lock | |
58 | * | |
59 | * iunique_lock | |
60 | * inode_hash_lock | |
250df6ed DC |
61 | */ |
62 | ||
1da177e4 LT |
63 | /* |
64 | * This is needed for the following functions: | |
65 | * - inode_has_buffers | |
1da177e4 LT |
66 | * - invalidate_bdev |
67 | * | |
68 | * FIXME: remove all knowledge of the buffer layer from this file | |
69 | */ | |
70 | #include <linux/buffer_head.h> | |
71 | ||
72 | /* | |
73 | * New inode.c implementation. | |
74 | * | |
75 | * This implementation has the basic premise of trying | |
76 | * to be extremely low-overhead and SMP-safe, yet be | |
77 | * simple enough to be "obviously correct". | |
78 | * | |
79 | * Famous last words. | |
80 | */ | |
81 | ||
82 | /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */ | |
83 | ||
84 | /* #define INODE_PARANOIA 1 */ | |
85 | /* #define INODE_DEBUG 1 */ | |
86 | ||
87 | /* | |
88 | * Inode lookup is no longer as critical as it used to be: | |
89 | * most of the lookups are going to be through the dcache. | |
90 | */ | |
91 | #define I_HASHBITS i_hash_shift | |
92 | #define I_HASHMASK i_hash_mask | |
93 | ||
fa3536cc ED |
94 | static unsigned int i_hash_mask __read_mostly; |
95 | static unsigned int i_hash_shift __read_mostly; | |
67a23c49 DC |
96 | static struct hlist_head *inode_hashtable __read_mostly; |
97 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_hash_lock); | |
1da177e4 LT |
98 | |
99 | /* | |
100 | * Each inode can be on two separate lists. One is | |
101 | * the hash list of the inode, used for lookups. The | |
102 | * other linked list is the "type" list: | |
103 | * "in_use" - valid inode, i_count > 0, i_nlink > 0 | |
104 | * "dirty" - as "in_use" but also dirty | |
105 | * "unused" - valid inode, i_count = 0 | |
106 | * | |
107 | * A "dirty" list is maintained for each super block, | |
108 | * allowing for low-overhead inode sync() operations. | |
109 | */ | |
110 | ||
7ccf19a8 | 111 | static LIST_HEAD(inode_lru); |
02afc410 | 112 | static DEFINE_SPINLOCK(inode_lru_lock); |
1da177e4 | 113 | |
55fa6091 | 114 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_sb_list_lock); |
a66979ab | 115 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_wb_list_lock); |
55fa6091 | 116 | |
1da177e4 | 117 | /* |
bab1d944 CH |
118 | * iprune_sem provides exclusion between the icache shrinking and the |
119 | * umount path. | |
88e0fbc4 | 120 | * |
bab1d944 CH |
121 | * We don't actually need it to protect anything in the umount path, |
122 | * but only need to cycle through it to make sure any inode that | |
123 | * prune_icache took off the LRU list has been fully torn down by the | |
124 | * time we are past evict_inodes. | |
1da177e4 | 125 | */ |
88e0fbc4 | 126 | static DECLARE_RWSEM(iprune_sem); |
1da177e4 | 127 | |
7dcda1c9 JA |
128 | /* |
129 | * Empty aops. Can be used for the cases where the user does not | |
130 | * define any of the address_space operations. | |
131 | */ | |
132 | const struct address_space_operations empty_aops = { | |
133 | }; | |
134 | EXPORT_SYMBOL(empty_aops); | |
135 | ||
1da177e4 LT |
136 | /* |
137 | * Statistics gathering.. | |
138 | */ | |
139 | struct inodes_stat_t inodes_stat; | |
140 | ||
3e880fb5 | 141 | static DEFINE_PER_CPU(unsigned int, nr_inodes); |
cffbc8aa | 142 | |
6b3304b5 | 143 | static struct kmem_cache *inode_cachep __read_mostly; |
1da177e4 | 144 | |
3e880fb5 | 145 | static int get_nr_inodes(void) |
cffbc8aa | 146 | { |
3e880fb5 NP |
147 | int i; |
148 | int sum = 0; | |
149 | for_each_possible_cpu(i) | |
150 | sum += per_cpu(nr_inodes, i); | |
151 | return sum < 0 ? 0 : sum; | |
cffbc8aa DC |
152 | } |
153 | ||
154 | static inline int get_nr_inodes_unused(void) | |
155 | { | |
86c8749e | 156 | return inodes_stat.nr_unused; |
cffbc8aa DC |
157 | } |
158 | ||
159 | int get_nr_dirty_inodes(void) | |
160 | { | |
3e880fb5 | 161 | /* not actually dirty inodes, but a wild approximation */ |
cffbc8aa DC |
162 | int nr_dirty = get_nr_inodes() - get_nr_inodes_unused(); |
163 | return nr_dirty > 0 ? nr_dirty : 0; | |
cffbc8aa DC |
164 | } |
165 | ||
166 | /* | |
167 | * Handle nr_inode sysctl | |
168 | */ | |
169 | #ifdef CONFIG_SYSCTL | |
170 | int proc_nr_inodes(ctl_table *table, int write, | |
171 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
172 | { | |
173 | inodes_stat.nr_inodes = get_nr_inodes(); | |
cffbc8aa DC |
174 | return proc_dointvec(table, write, buffer, lenp, ppos); |
175 | } | |
176 | #endif | |
177 | ||
2cb1599f DC |
178 | /** |
179 | * inode_init_always - perform inode structure intialisation | |
0bc02f3f RD |
180 | * @sb: superblock inode belongs to |
181 | * @inode: inode to initialise | |
2cb1599f DC |
182 | * |
183 | * These are initializations that need to be done on every inode | |
184 | * allocation as the fields are not initialised by slab allocation. | |
185 | */ | |
54e34621 | 186 | int inode_init_always(struct super_block *sb, struct inode *inode) |
1da177e4 | 187 | { |
6e1d5dcc | 188 | static const struct inode_operations empty_iops; |
99ac48f5 | 189 | static const struct file_operations empty_fops; |
6b3304b5 | 190 | struct address_space *const mapping = &inode->i_data; |
2cb1599f DC |
191 | |
192 | inode->i_sb = sb; | |
193 | inode->i_blkbits = sb->s_blocksize_bits; | |
194 | inode->i_flags = 0; | |
195 | atomic_set(&inode->i_count, 1); | |
196 | inode->i_op = &empty_iops; | |
197 | inode->i_fop = &empty_fops; | |
198 | inode->i_nlink = 1; | |
56ff5efa AV |
199 | inode->i_uid = 0; |
200 | inode->i_gid = 0; | |
2cb1599f DC |
201 | atomic_set(&inode->i_writecount, 0); |
202 | inode->i_size = 0; | |
203 | inode->i_blocks = 0; | |
204 | inode->i_bytes = 0; | |
205 | inode->i_generation = 0; | |
1da177e4 | 206 | #ifdef CONFIG_QUOTA |
2cb1599f | 207 | memset(&inode->i_dquot, 0, sizeof(inode->i_dquot)); |
1da177e4 | 208 | #endif |
2cb1599f DC |
209 | inode->i_pipe = NULL; |
210 | inode->i_bdev = NULL; | |
211 | inode->i_cdev = NULL; | |
212 | inode->i_rdev = 0; | |
213 | inode->dirtied_when = 0; | |
6146f0d5 MZ |
214 | |
215 | if (security_inode_alloc(inode)) | |
54e34621 | 216 | goto out; |
2cb1599f DC |
217 | spin_lock_init(&inode->i_lock); |
218 | lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key); | |
219 | ||
220 | mutex_init(&inode->i_mutex); | |
221 | lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key); | |
222 | ||
223 | init_rwsem(&inode->i_alloc_sem); | |
224 | lockdep_set_class(&inode->i_alloc_sem, &sb->s_type->i_alloc_sem_key); | |
225 | ||
226 | mapping->a_ops = &empty_aops; | |
227 | mapping->host = inode; | |
228 | mapping->flags = 0; | |
3c1d4378 | 229 | mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE); |
2cb1599f DC |
230 | mapping->assoc_mapping = NULL; |
231 | mapping->backing_dev_info = &default_backing_dev_info; | |
232 | mapping->writeback_index = 0; | |
233 | ||
234 | /* | |
235 | * If the block_device provides a backing_dev_info for client | |
236 | * inodes then use that. Otherwise the inode share the bdev's | |
237 | * backing_dev_info. | |
238 | */ | |
239 | if (sb->s_bdev) { | |
240 | struct backing_dev_info *bdi; | |
241 | ||
2c96ce9f | 242 | bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info; |
2cb1599f DC |
243 | mapping->backing_dev_info = bdi; |
244 | } | |
245 | inode->i_private = NULL; | |
246 | inode->i_mapping = mapping; | |
f19d4a8f AV |
247 | #ifdef CONFIG_FS_POSIX_ACL |
248 | inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED; | |
249 | #endif | |
2cb1599f | 250 | |
3be25f49 EP |
251 | #ifdef CONFIG_FSNOTIFY |
252 | inode->i_fsnotify_mask = 0; | |
253 | #endif | |
254 | ||
3e880fb5 | 255 | this_cpu_inc(nr_inodes); |
cffbc8aa | 256 | |
54e34621 | 257 | return 0; |
54e34621 CH |
258 | out: |
259 | return -ENOMEM; | |
1da177e4 | 260 | } |
2cb1599f DC |
261 | EXPORT_SYMBOL(inode_init_always); |
262 | ||
263 | static struct inode *alloc_inode(struct super_block *sb) | |
264 | { | |
265 | struct inode *inode; | |
266 | ||
267 | if (sb->s_op->alloc_inode) | |
268 | inode = sb->s_op->alloc_inode(sb); | |
269 | else | |
270 | inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL); | |
271 | ||
54e34621 CH |
272 | if (!inode) |
273 | return NULL; | |
274 | ||
275 | if (unlikely(inode_init_always(sb, inode))) { | |
276 | if (inode->i_sb->s_op->destroy_inode) | |
277 | inode->i_sb->s_op->destroy_inode(inode); | |
278 | else | |
279 | kmem_cache_free(inode_cachep, inode); | |
280 | return NULL; | |
281 | } | |
282 | ||
283 | return inode; | |
2cb1599f | 284 | } |
1da177e4 | 285 | |
ff0c7d15 NP |
286 | void free_inode_nonrcu(struct inode *inode) |
287 | { | |
288 | kmem_cache_free(inode_cachep, inode); | |
289 | } | |
290 | EXPORT_SYMBOL(free_inode_nonrcu); | |
291 | ||
2e00c97e | 292 | void __destroy_inode(struct inode *inode) |
1da177e4 | 293 | { |
b7542f8c | 294 | BUG_ON(inode_has_buffers(inode)); |
1da177e4 | 295 | security_inode_free(inode); |
3be25f49 | 296 | fsnotify_inode_delete(inode); |
f19d4a8f AV |
297 | #ifdef CONFIG_FS_POSIX_ACL |
298 | if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED) | |
299 | posix_acl_release(inode->i_acl); | |
300 | if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED) | |
301 | posix_acl_release(inode->i_default_acl); | |
302 | #endif | |
3e880fb5 | 303 | this_cpu_dec(nr_inodes); |
2e00c97e CH |
304 | } |
305 | EXPORT_SYMBOL(__destroy_inode); | |
306 | ||
fa0d7e3d NP |
307 | static void i_callback(struct rcu_head *head) |
308 | { | |
309 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
310 | INIT_LIST_HEAD(&inode->i_dentry); | |
311 | kmem_cache_free(inode_cachep, inode); | |
312 | } | |
313 | ||
56b0dacf | 314 | static void destroy_inode(struct inode *inode) |
2e00c97e | 315 | { |
7ccf19a8 | 316 | BUG_ON(!list_empty(&inode->i_lru)); |
2e00c97e | 317 | __destroy_inode(inode); |
1da177e4 LT |
318 | if (inode->i_sb->s_op->destroy_inode) |
319 | inode->i_sb->s_op->destroy_inode(inode); | |
320 | else | |
fa0d7e3d | 321 | call_rcu(&inode->i_rcu, i_callback); |
1da177e4 | 322 | } |
1da177e4 | 323 | |
2aa15890 MS |
324 | void address_space_init_once(struct address_space *mapping) |
325 | { | |
326 | memset(mapping, 0, sizeof(*mapping)); | |
327 | INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC); | |
328 | spin_lock_init(&mapping->tree_lock); | |
329 | spin_lock_init(&mapping->i_mmap_lock); | |
330 | INIT_LIST_HEAD(&mapping->private_list); | |
331 | spin_lock_init(&mapping->private_lock); | |
332 | INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap); | |
333 | INIT_LIST_HEAD(&mapping->i_mmap_nonlinear); | |
334 | mutex_init(&mapping->unmap_mutex); | |
335 | } | |
336 | EXPORT_SYMBOL(address_space_init_once); | |
337 | ||
1da177e4 LT |
338 | /* |
339 | * These are initializations that only need to be done | |
340 | * once, because the fields are idempotent across use | |
341 | * of the inode, so let the slab aware of that. | |
342 | */ | |
343 | void inode_init_once(struct inode *inode) | |
344 | { | |
345 | memset(inode, 0, sizeof(*inode)); | |
346 | INIT_HLIST_NODE(&inode->i_hash); | |
347 | INIT_LIST_HEAD(&inode->i_dentry); | |
348 | INIT_LIST_HEAD(&inode->i_devices); | |
7ccf19a8 NP |
349 | INIT_LIST_HEAD(&inode->i_wb_list); |
350 | INIT_LIST_HEAD(&inode->i_lru); | |
2aa15890 | 351 | address_space_init_once(&inode->i_data); |
1da177e4 | 352 | i_size_ordered_init(inode); |
3be25f49 | 353 | #ifdef CONFIG_FSNOTIFY |
e61ce867 | 354 | INIT_HLIST_HEAD(&inode->i_fsnotify_marks); |
3be25f49 | 355 | #endif |
1da177e4 | 356 | } |
1da177e4 LT |
357 | EXPORT_SYMBOL(inode_init_once); |
358 | ||
51cc5068 | 359 | static void init_once(void *foo) |
1da177e4 | 360 | { |
6b3304b5 | 361 | struct inode *inode = (struct inode *) foo; |
1da177e4 | 362 | |
a35afb83 | 363 | inode_init_once(inode); |
1da177e4 LT |
364 | } |
365 | ||
366 | /* | |
250df6ed | 367 | * inode->i_lock must be held |
1da177e4 | 368 | */ |
6b3304b5 | 369 | void __iget(struct inode *inode) |
1da177e4 | 370 | { |
9e38d86f NP |
371 | atomic_inc(&inode->i_count); |
372 | } | |
2e147f1e | 373 | |
7de9c6ee AV |
374 | /* |
375 | * get additional reference to inode; caller must already hold one. | |
376 | */ | |
377 | void ihold(struct inode *inode) | |
378 | { | |
379 | WARN_ON(atomic_inc_return(&inode->i_count) < 2); | |
380 | } | |
381 | EXPORT_SYMBOL(ihold); | |
382 | ||
9e38d86f NP |
383 | static void inode_lru_list_add(struct inode *inode) |
384 | { | |
02afc410 | 385 | spin_lock(&inode_lru_lock); |
7ccf19a8 NP |
386 | if (list_empty(&inode->i_lru)) { |
387 | list_add(&inode->i_lru, &inode_lru); | |
86c8749e | 388 | inodes_stat.nr_unused++; |
9e38d86f | 389 | } |
02afc410 | 390 | spin_unlock(&inode_lru_lock); |
9e38d86f | 391 | } |
2e147f1e | 392 | |
9e38d86f NP |
393 | static void inode_lru_list_del(struct inode *inode) |
394 | { | |
02afc410 | 395 | spin_lock(&inode_lru_lock); |
7ccf19a8 NP |
396 | if (!list_empty(&inode->i_lru)) { |
397 | list_del_init(&inode->i_lru); | |
86c8749e | 398 | inodes_stat.nr_unused--; |
9e38d86f | 399 | } |
02afc410 | 400 | spin_unlock(&inode_lru_lock); |
1da177e4 LT |
401 | } |
402 | ||
646ec461 CH |
403 | /** |
404 | * inode_sb_list_add - add inode to the superblock list of inodes | |
405 | * @inode: inode to add | |
406 | */ | |
407 | void inode_sb_list_add(struct inode *inode) | |
408 | { | |
55fa6091 DC |
409 | spin_lock(&inode_sb_list_lock); |
410 | list_add(&inode->i_sb_list, &inode->i_sb->s_inodes); | |
411 | spin_unlock(&inode_sb_list_lock); | |
646ec461 CH |
412 | } |
413 | EXPORT_SYMBOL_GPL(inode_sb_list_add); | |
414 | ||
55fa6091 | 415 | static inline void inode_sb_list_del(struct inode *inode) |
646ec461 | 416 | { |
55fa6091 | 417 | spin_lock(&inode_sb_list_lock); |
646ec461 | 418 | list_del_init(&inode->i_sb_list); |
55fa6091 | 419 | spin_unlock(&inode_sb_list_lock); |
646ec461 CH |
420 | } |
421 | ||
4c51acbc DC |
422 | static unsigned long hash(struct super_block *sb, unsigned long hashval) |
423 | { | |
424 | unsigned long tmp; | |
425 | ||
426 | tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) / | |
427 | L1_CACHE_BYTES; | |
428 | tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> I_HASHBITS); | |
429 | return tmp & I_HASHMASK; | |
430 | } | |
431 | ||
432 | /** | |
433 | * __insert_inode_hash - hash an inode | |
434 | * @inode: unhashed inode | |
435 | * @hashval: unsigned long value used to locate this object in the | |
436 | * inode_hashtable. | |
437 | * | |
438 | * Add an inode to the inode hash for this superblock. | |
439 | */ | |
440 | void __insert_inode_hash(struct inode *inode, unsigned long hashval) | |
441 | { | |
646ec461 CH |
442 | struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval); |
443 | ||
67a23c49 | 444 | spin_lock(&inode_hash_lock); |
250df6ed | 445 | spin_lock(&inode->i_lock); |
646ec461 | 446 | hlist_add_head(&inode->i_hash, b); |
250df6ed | 447 | spin_unlock(&inode->i_lock); |
67a23c49 | 448 | spin_unlock(&inode_hash_lock); |
4c51acbc DC |
449 | } |
450 | EXPORT_SYMBOL(__insert_inode_hash); | |
451 | ||
4c51acbc DC |
452 | /** |
453 | * remove_inode_hash - remove an inode from the hash | |
454 | * @inode: inode to unhash | |
455 | * | |
456 | * Remove an inode from the superblock. | |
457 | */ | |
458 | void remove_inode_hash(struct inode *inode) | |
459 | { | |
67a23c49 | 460 | spin_lock(&inode_hash_lock); |
250df6ed | 461 | spin_lock(&inode->i_lock); |
4c51acbc | 462 | hlist_del_init(&inode->i_hash); |
250df6ed | 463 | spin_unlock(&inode->i_lock); |
67a23c49 | 464 | spin_unlock(&inode_hash_lock); |
4c51acbc DC |
465 | } |
466 | EXPORT_SYMBOL(remove_inode_hash); | |
467 | ||
b0683aa6 AV |
468 | void end_writeback(struct inode *inode) |
469 | { | |
470 | might_sleep(); | |
471 | BUG_ON(inode->i_data.nrpages); | |
472 | BUG_ON(!list_empty(&inode->i_data.private_list)); | |
473 | BUG_ON(!(inode->i_state & I_FREEING)); | |
474 | BUG_ON(inode->i_state & I_CLEAR); | |
475 | inode_sync_wait(inode); | |
fa0d7e3d | 476 | /* don't need i_lock here, no concurrent mods to i_state */ |
b0683aa6 AV |
477 | inode->i_state = I_FREEING | I_CLEAR; |
478 | } | |
479 | EXPORT_SYMBOL(end_writeback); | |
480 | ||
b2b2af8e DC |
481 | /* |
482 | * Free the inode passed in, removing it from the lists it is still connected | |
483 | * to. We remove any pages still attached to the inode and wait for any IO that | |
484 | * is still in progress before finally destroying the inode. | |
485 | * | |
486 | * An inode must already be marked I_FREEING so that we avoid the inode being | |
487 | * moved back onto lists if we race with other code that manipulates the lists | |
488 | * (e.g. writeback_single_inode). The caller is responsible for setting this. | |
489 | * | |
490 | * An inode must already be removed from the LRU list before being evicted from | |
491 | * the cache. This should occur atomically with setting the I_FREEING state | |
492 | * flag, so no inodes here should ever be on the LRU when being evicted. | |
493 | */ | |
644da596 | 494 | static void evict(struct inode *inode) |
b4272d4c AV |
495 | { |
496 | const struct super_operations *op = inode->i_sb->s_op; | |
497 | ||
b2b2af8e DC |
498 | BUG_ON(!(inode->i_state & I_FREEING)); |
499 | BUG_ON(!list_empty(&inode->i_lru)); | |
500 | ||
a66979ab | 501 | inode_wb_list_del(inode); |
55fa6091 DC |
502 | inode_sb_list_del(inode); |
503 | ||
be7ce416 AV |
504 | if (op->evict_inode) { |
505 | op->evict_inode(inode); | |
b4272d4c AV |
506 | } else { |
507 | if (inode->i_data.nrpages) | |
508 | truncate_inode_pages(&inode->i_data, 0); | |
30140837 | 509 | end_writeback(inode); |
b4272d4c | 510 | } |
661074e9 AV |
511 | if (S_ISBLK(inode->i_mode) && inode->i_bdev) |
512 | bd_forget(inode); | |
513 | if (S_ISCHR(inode->i_mode) && inode->i_cdev) | |
514 | cd_forget(inode); | |
b2b2af8e DC |
515 | |
516 | remove_inode_hash(inode); | |
517 | ||
518 | spin_lock(&inode->i_lock); | |
519 | wake_up_bit(&inode->i_state, __I_NEW); | |
520 | BUG_ON(inode->i_state != (I_FREEING | I_CLEAR)); | |
521 | spin_unlock(&inode->i_lock); | |
522 | ||
523 | destroy_inode(inode); | |
b4272d4c AV |
524 | } |
525 | ||
1da177e4 LT |
526 | /* |
527 | * dispose_list - dispose of the contents of a local list | |
528 | * @head: the head of the list to free | |
529 | * | |
530 | * Dispose-list gets a local list with local inodes in it, so it doesn't | |
531 | * need to worry about list corruption and SMP locks. | |
532 | */ | |
533 | static void dispose_list(struct list_head *head) | |
534 | { | |
1da177e4 LT |
535 | while (!list_empty(head)) { |
536 | struct inode *inode; | |
537 | ||
7ccf19a8 NP |
538 | inode = list_first_entry(head, struct inode, i_lru); |
539 | list_del_init(&inode->i_lru); | |
1da177e4 | 540 | |
644da596 | 541 | evict(inode); |
1da177e4 | 542 | } |
1da177e4 LT |
543 | } |
544 | ||
63997e98 AV |
545 | /** |
546 | * evict_inodes - evict all evictable inodes for a superblock | |
547 | * @sb: superblock to operate on | |
548 | * | |
549 | * Make sure that no inodes with zero refcount are retained. This is | |
550 | * called by superblock shutdown after having MS_ACTIVE flag removed, | |
551 | * so any inode reaching zero refcount during or after that call will | |
552 | * be immediately evicted. | |
1da177e4 | 553 | */ |
63997e98 | 554 | void evict_inodes(struct super_block *sb) |
1da177e4 | 555 | { |
63997e98 AV |
556 | struct inode *inode, *next; |
557 | LIST_HEAD(dispose); | |
1da177e4 | 558 | |
55fa6091 | 559 | spin_lock(&inode_sb_list_lock); |
63997e98 AV |
560 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
561 | if (atomic_read(&inode->i_count)) | |
aabb8fdb | 562 | continue; |
250df6ed DC |
563 | |
564 | spin_lock(&inode->i_lock); | |
565 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
566 | spin_unlock(&inode->i_lock); | |
1da177e4 | 567 | continue; |
250df6ed | 568 | } |
63997e98 AV |
569 | |
570 | inode->i_state |= I_FREEING; | |
02afc410 | 571 | inode_lru_list_del(inode); |
250df6ed | 572 | spin_unlock(&inode->i_lock); |
02afc410 | 573 | list_add(&inode->i_lru, &dispose); |
1da177e4 | 574 | } |
55fa6091 | 575 | spin_unlock(&inode_sb_list_lock); |
63997e98 AV |
576 | |
577 | dispose_list(&dispose); | |
bab1d944 CH |
578 | |
579 | /* | |
580 | * Cycle through iprune_sem to make sure any inode that prune_icache | |
581 | * moved off the list before we took the lock has been fully torn | |
582 | * down. | |
583 | */ | |
584 | down_write(&iprune_sem); | |
63997e98 | 585 | up_write(&iprune_sem); |
1da177e4 LT |
586 | } |
587 | ||
1da177e4 | 588 | /** |
a0318786 CH |
589 | * invalidate_inodes - attempt to free all inodes on a superblock |
590 | * @sb: superblock to operate on | |
93b270f7 | 591 | * @kill_dirty: flag to guide handling of dirty inodes |
1da177e4 | 592 | * |
a0318786 CH |
593 | * Attempts to free all inodes for a given superblock. If there were any |
594 | * busy inodes return a non-zero value, else zero. | |
93b270f7 N |
595 | * If @kill_dirty is set, discard dirty inodes too, otherwise treat |
596 | * them as busy. | |
1da177e4 | 597 | */ |
93b270f7 | 598 | int invalidate_inodes(struct super_block *sb, bool kill_dirty) |
1da177e4 | 599 | { |
cffbc8aa | 600 | int busy = 0; |
a0318786 CH |
601 | struct inode *inode, *next; |
602 | LIST_HEAD(dispose); | |
1da177e4 | 603 | |
55fa6091 | 604 | spin_lock(&inode_sb_list_lock); |
a0318786 | 605 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
250df6ed DC |
606 | spin_lock(&inode->i_lock); |
607 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
608 | spin_unlock(&inode->i_lock); | |
aabb8fdb | 609 | continue; |
250df6ed | 610 | } |
93b270f7 | 611 | if (inode->i_state & I_DIRTY && !kill_dirty) { |
250df6ed | 612 | spin_unlock(&inode->i_lock); |
93b270f7 N |
613 | busy = 1; |
614 | continue; | |
615 | } | |
99a38919 | 616 | if (atomic_read(&inode->i_count)) { |
250df6ed | 617 | spin_unlock(&inode->i_lock); |
99a38919 | 618 | busy = 1; |
1da177e4 LT |
619 | continue; |
620 | } | |
99a38919 | 621 | |
99a38919 | 622 | inode->i_state |= I_FREEING; |
02afc410 | 623 | inode_lru_list_del(inode); |
250df6ed | 624 | spin_unlock(&inode->i_lock); |
02afc410 | 625 | list_add(&inode->i_lru, &dispose); |
1da177e4 | 626 | } |
55fa6091 | 627 | spin_unlock(&inode_sb_list_lock); |
1da177e4 | 628 | |
a0318786 | 629 | dispose_list(&dispose); |
1da177e4 LT |
630 | |
631 | return busy; | |
632 | } | |
1da177e4 LT |
633 | |
634 | static int can_unuse(struct inode *inode) | |
635 | { | |
9e38d86f | 636 | if (inode->i_state & ~I_REFERENCED) |
1da177e4 LT |
637 | return 0; |
638 | if (inode_has_buffers(inode)) | |
639 | return 0; | |
640 | if (atomic_read(&inode->i_count)) | |
641 | return 0; | |
642 | if (inode->i_data.nrpages) | |
643 | return 0; | |
644 | return 1; | |
645 | } | |
646 | ||
647 | /* | |
9e38d86f | 648 | * Scan `goal' inodes on the unused list for freeable ones. They are moved to a |
02afc410 | 649 | * temporary list and then are freed outside inode_lru_lock by dispose_list(). |
1da177e4 LT |
650 | * |
651 | * Any inodes which are pinned purely because of attached pagecache have their | |
9e38d86f NP |
652 | * pagecache removed. If the inode has metadata buffers attached to |
653 | * mapping->private_list then try to remove them. | |
1da177e4 | 654 | * |
9e38d86f NP |
655 | * If the inode has the I_REFERENCED flag set, then it means that it has been |
656 | * used recently - the flag is set in iput_final(). When we encounter such an | |
657 | * inode, clear the flag and move it to the back of the LRU so it gets another | |
658 | * pass through the LRU before it gets reclaimed. This is necessary because of | |
659 | * the fact we are doing lazy LRU updates to minimise lock contention so the | |
660 | * LRU does not have strict ordering. Hence we don't want to reclaim inodes | |
661 | * with this flag set because they are the inodes that are out of order. | |
1da177e4 LT |
662 | */ |
663 | static void prune_icache(int nr_to_scan) | |
664 | { | |
665 | LIST_HEAD(freeable); | |
1da177e4 LT |
666 | int nr_scanned; |
667 | unsigned long reap = 0; | |
668 | ||
88e0fbc4 | 669 | down_read(&iprune_sem); |
02afc410 | 670 | spin_lock(&inode_lru_lock); |
1da177e4 LT |
671 | for (nr_scanned = 0; nr_scanned < nr_to_scan; nr_scanned++) { |
672 | struct inode *inode; | |
673 | ||
7ccf19a8 | 674 | if (list_empty(&inode_lru)) |
1da177e4 LT |
675 | break; |
676 | ||
7ccf19a8 | 677 | inode = list_entry(inode_lru.prev, struct inode, i_lru); |
1da177e4 | 678 | |
02afc410 DC |
679 | /* |
680 | * we are inverting the inode_lru_lock/inode->i_lock here, | |
681 | * so use a trylock. If we fail to get the lock, just move the | |
682 | * inode to the back of the list so we don't spin on it. | |
683 | */ | |
684 | if (!spin_trylock(&inode->i_lock)) { | |
685 | list_move(&inode->i_lru, &inode_lru); | |
686 | continue; | |
687 | } | |
688 | ||
9e38d86f NP |
689 | /* |
690 | * Referenced or dirty inodes are still in use. Give them | |
691 | * another pass through the LRU as we canot reclaim them now. | |
692 | */ | |
693 | if (atomic_read(&inode->i_count) || | |
694 | (inode->i_state & ~I_REFERENCED)) { | |
7ccf19a8 | 695 | list_del_init(&inode->i_lru); |
f283c86a | 696 | spin_unlock(&inode->i_lock); |
86c8749e | 697 | inodes_stat.nr_unused--; |
9e38d86f NP |
698 | continue; |
699 | } | |
700 | ||
701 | /* recently referenced inodes get one more pass */ | |
702 | if (inode->i_state & I_REFERENCED) { | |
9e38d86f | 703 | inode->i_state &= ~I_REFERENCED; |
250df6ed | 704 | list_move(&inode->i_lru, &inode_lru); |
f283c86a | 705 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
706 | continue; |
707 | } | |
708 | if (inode_has_buffers(inode) || inode->i_data.nrpages) { | |
709 | __iget(inode); | |
250df6ed | 710 | spin_unlock(&inode->i_lock); |
02afc410 | 711 | spin_unlock(&inode_lru_lock); |
1da177e4 | 712 | if (remove_inode_buffers(inode)) |
fc0ecff6 AM |
713 | reap += invalidate_mapping_pages(&inode->i_data, |
714 | 0, -1); | |
1da177e4 | 715 | iput(inode); |
02afc410 | 716 | spin_lock(&inode_lru_lock); |
1da177e4 | 717 | |
7ccf19a8 NP |
718 | if (inode != list_entry(inode_lru.next, |
719 | struct inode, i_lru)) | |
1da177e4 | 720 | continue; /* wrong inode or list_empty */ |
02afc410 DC |
721 | /* avoid lock inversions with trylock */ |
722 | if (!spin_trylock(&inode->i_lock)) | |
723 | continue; | |
250df6ed DC |
724 | if (!can_unuse(inode)) { |
725 | spin_unlock(&inode->i_lock); | |
1da177e4 | 726 | continue; |
250df6ed | 727 | } |
1da177e4 | 728 | } |
7ef0d737 | 729 | WARN_ON(inode->i_state & I_NEW); |
1da177e4 | 730 | inode->i_state |= I_FREEING; |
250df6ed | 731 | spin_unlock(&inode->i_lock); |
7ccf19a8 | 732 | |
7ccf19a8 | 733 | list_move(&inode->i_lru, &freeable); |
86c8749e | 734 | inodes_stat.nr_unused--; |
1da177e4 | 735 | } |
f8891e5e CL |
736 | if (current_is_kswapd()) |
737 | __count_vm_events(KSWAPD_INODESTEAL, reap); | |
738 | else | |
739 | __count_vm_events(PGINODESTEAL, reap); | |
02afc410 | 740 | spin_unlock(&inode_lru_lock); |
1da177e4 LT |
741 | |
742 | dispose_list(&freeable); | |
88e0fbc4 | 743 | up_read(&iprune_sem); |
1da177e4 LT |
744 | } |
745 | ||
746 | /* | |
747 | * shrink_icache_memory() will attempt to reclaim some unused inodes. Here, | |
748 | * "unused" means that no dentries are referring to the inodes: the files are | |
749 | * not open and the dcache references to those inodes have already been | |
750 | * reclaimed. | |
751 | * | |
752 | * This function is passed the number of inodes to scan, and it returns the | |
753 | * total number of remaining possibly-reclaimable inodes. | |
754 | */ | |
7f8275d0 | 755 | static int shrink_icache_memory(struct shrinker *shrink, int nr, gfp_t gfp_mask) |
1da177e4 LT |
756 | { |
757 | if (nr) { | |
758 | /* | |
759 | * Nasty deadlock avoidance. We may hold various FS locks, | |
760 | * and we don't want to recurse into the FS that called us | |
761 | * in clear_inode() and friends.. | |
6b3304b5 | 762 | */ |
1da177e4 LT |
763 | if (!(gfp_mask & __GFP_FS)) |
764 | return -1; | |
765 | prune_icache(nr); | |
766 | } | |
cffbc8aa | 767 | return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure; |
1da177e4 LT |
768 | } |
769 | ||
8e1f936b RR |
770 | static struct shrinker icache_shrinker = { |
771 | .shrink = shrink_icache_memory, | |
772 | .seeks = DEFAULT_SEEKS, | |
773 | }; | |
774 | ||
1da177e4 LT |
775 | static void __wait_on_freeing_inode(struct inode *inode); |
776 | /* | |
777 | * Called with the inode lock held. | |
1da177e4 | 778 | */ |
6b3304b5 MK |
779 | static struct inode *find_inode(struct super_block *sb, |
780 | struct hlist_head *head, | |
781 | int (*test)(struct inode *, void *), | |
782 | void *data) | |
1da177e4 LT |
783 | { |
784 | struct hlist_node *node; | |
6b3304b5 | 785 | struct inode *inode = NULL; |
1da177e4 LT |
786 | |
787 | repeat: | |
c5c8be3c | 788 | hlist_for_each_entry(inode, node, head, i_hash) { |
67a23c49 DC |
789 | spin_lock(&inode->i_lock); |
790 | if (inode->i_sb != sb) { | |
791 | spin_unlock(&inode->i_lock); | |
1da177e4 | 792 | continue; |
67a23c49 DC |
793 | } |
794 | if (!test(inode, data)) { | |
795 | spin_unlock(&inode->i_lock); | |
1da177e4 | 796 | continue; |
67a23c49 | 797 | } |
a4ffdde6 | 798 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
799 | __wait_on_freeing_inode(inode); |
800 | goto repeat; | |
801 | } | |
f7899bd5 | 802 | __iget(inode); |
250df6ed | 803 | spin_unlock(&inode->i_lock); |
f7899bd5 | 804 | return inode; |
1da177e4 | 805 | } |
f7899bd5 | 806 | return NULL; |
1da177e4 LT |
807 | } |
808 | ||
809 | /* | |
810 | * find_inode_fast is the fast path version of find_inode, see the comment at | |
811 | * iget_locked for details. | |
812 | */ | |
6b3304b5 MK |
813 | static struct inode *find_inode_fast(struct super_block *sb, |
814 | struct hlist_head *head, unsigned long ino) | |
1da177e4 LT |
815 | { |
816 | struct hlist_node *node; | |
6b3304b5 | 817 | struct inode *inode = NULL; |
1da177e4 LT |
818 | |
819 | repeat: | |
c5c8be3c | 820 | hlist_for_each_entry(inode, node, head, i_hash) { |
67a23c49 DC |
821 | spin_lock(&inode->i_lock); |
822 | if (inode->i_ino != ino) { | |
823 | spin_unlock(&inode->i_lock); | |
1da177e4 | 824 | continue; |
67a23c49 DC |
825 | } |
826 | if (inode->i_sb != sb) { | |
827 | spin_unlock(&inode->i_lock); | |
1da177e4 | 828 | continue; |
67a23c49 | 829 | } |
a4ffdde6 | 830 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
831 | __wait_on_freeing_inode(inode); |
832 | goto repeat; | |
833 | } | |
f7899bd5 | 834 | __iget(inode); |
250df6ed | 835 | spin_unlock(&inode->i_lock); |
f7899bd5 | 836 | return inode; |
1da177e4 | 837 | } |
f7899bd5 | 838 | return NULL; |
8290c35f DC |
839 | } |
840 | ||
f991bd2e ED |
841 | /* |
842 | * Each cpu owns a range of LAST_INO_BATCH numbers. | |
843 | * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations, | |
844 | * to renew the exhausted range. | |
8290c35f | 845 | * |
f991bd2e ED |
846 | * This does not significantly increase overflow rate because every CPU can |
847 | * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is | |
848 | * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the | |
849 | * 2^32 range, and is a worst-case. Even a 50% wastage would only increase | |
850 | * overflow rate by 2x, which does not seem too significant. | |
851 | * | |
852 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
853 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
854 | * here to attempt to avoid that. | |
8290c35f | 855 | */ |
f991bd2e ED |
856 | #define LAST_INO_BATCH 1024 |
857 | static DEFINE_PER_CPU(unsigned int, last_ino); | |
858 | ||
85fe4025 | 859 | unsigned int get_next_ino(void) |
8290c35f | 860 | { |
f991bd2e ED |
861 | unsigned int *p = &get_cpu_var(last_ino); |
862 | unsigned int res = *p; | |
8290c35f | 863 | |
f991bd2e ED |
864 | #ifdef CONFIG_SMP |
865 | if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) { | |
866 | static atomic_t shared_last_ino; | |
867 | int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino); | |
868 | ||
869 | res = next - LAST_INO_BATCH; | |
870 | } | |
871 | #endif | |
872 | ||
873 | *p = ++res; | |
874 | put_cpu_var(last_ino); | |
875 | return res; | |
8290c35f | 876 | } |
85fe4025 | 877 | EXPORT_SYMBOL(get_next_ino); |
8290c35f | 878 | |
1da177e4 LT |
879 | /** |
880 | * new_inode - obtain an inode | |
881 | * @sb: superblock | |
882 | * | |
769848c0 | 883 | * Allocates a new inode for given superblock. The default gfp_mask |
3c1d4378 | 884 | * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE. |
769848c0 MG |
885 | * If HIGHMEM pages are unsuitable or it is known that pages allocated |
886 | * for the page cache are not reclaimable or migratable, | |
887 | * mapping_set_gfp_mask() must be called with suitable flags on the | |
888 | * newly created inode's mapping | |
889 | * | |
1da177e4 LT |
890 | */ |
891 | struct inode *new_inode(struct super_block *sb) | |
892 | { | |
6b3304b5 | 893 | struct inode *inode; |
1da177e4 | 894 | |
55fa6091 | 895 | spin_lock_prefetch(&inode_sb_list_lock); |
6b3304b5 | 896 | |
1da177e4 LT |
897 | inode = alloc_inode(sb); |
898 | if (inode) { | |
250df6ed | 899 | spin_lock(&inode->i_lock); |
1da177e4 | 900 | inode->i_state = 0; |
250df6ed | 901 | spin_unlock(&inode->i_lock); |
55fa6091 | 902 | inode_sb_list_add(inode); |
1da177e4 LT |
903 | } |
904 | return inode; | |
905 | } | |
1da177e4 LT |
906 | EXPORT_SYMBOL(new_inode); |
907 | ||
250df6ed DC |
908 | /** |
909 | * unlock_new_inode - clear the I_NEW state and wake up any waiters | |
910 | * @inode: new inode to unlock | |
911 | * | |
912 | * Called when the inode is fully initialised to clear the new state of the | |
913 | * inode and wake up anyone waiting for the inode to finish initialisation. | |
914 | */ | |
1da177e4 LT |
915 | void unlock_new_inode(struct inode *inode) |
916 | { | |
14358e6d | 917 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
a3314a0e | 918 | if (S_ISDIR(inode->i_mode)) { |
1e89a5e1 PZ |
919 | struct file_system_type *type = inode->i_sb->s_type; |
920 | ||
9a7aa12f JK |
921 | /* Set new key only if filesystem hasn't already changed it */ |
922 | if (!lockdep_match_class(&inode->i_mutex, | |
923 | &type->i_mutex_key)) { | |
924 | /* | |
925 | * ensure nobody is actually holding i_mutex | |
926 | */ | |
927 | mutex_destroy(&inode->i_mutex); | |
928 | mutex_init(&inode->i_mutex); | |
929 | lockdep_set_class(&inode->i_mutex, | |
930 | &type->i_mutex_dir_key); | |
931 | } | |
1e89a5e1 | 932 | } |
14358e6d | 933 | #endif |
250df6ed | 934 | spin_lock(&inode->i_lock); |
eaff8079 CH |
935 | WARN_ON(!(inode->i_state & I_NEW)); |
936 | inode->i_state &= ~I_NEW; | |
250df6ed DC |
937 | wake_up_bit(&inode->i_state, __I_NEW); |
938 | spin_unlock(&inode->i_lock); | |
1da177e4 | 939 | } |
1da177e4 LT |
940 | EXPORT_SYMBOL(unlock_new_inode); |
941 | ||
0b2d0724 CH |
942 | /** |
943 | * iget5_locked - obtain an inode from a mounted file system | |
944 | * @sb: super block of file system | |
945 | * @hashval: hash value (usually inode number) to get | |
946 | * @test: callback used for comparisons between inodes | |
947 | * @set: callback used to initialize a new struct inode | |
948 | * @data: opaque data pointer to pass to @test and @set | |
949 | * | |
950 | * Search for the inode specified by @hashval and @data in the inode cache, | |
951 | * and if present it is return it with an increased reference count. This is | |
952 | * a generalized version of iget_locked() for file systems where the inode | |
953 | * number is not sufficient for unique identification of an inode. | |
954 | * | |
955 | * If the inode is not in cache, allocate a new inode and return it locked, | |
956 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
957 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 958 | * |
0b2d0724 CH |
959 | * Note both @test and @set are called with the inode_hash_lock held, so can't |
960 | * sleep. | |
1da177e4 | 961 | */ |
0b2d0724 CH |
962 | struct inode *iget5_locked(struct super_block *sb, unsigned long hashval, |
963 | int (*test)(struct inode *, void *), | |
964 | int (*set)(struct inode *, void *), void *data) | |
1da177e4 | 965 | { |
0b2d0724 | 966 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); |
6b3304b5 | 967 | struct inode *inode; |
1da177e4 | 968 | |
0b2d0724 CH |
969 | spin_lock(&inode_hash_lock); |
970 | inode = find_inode(sb, head, test, data); | |
971 | spin_unlock(&inode_hash_lock); | |
972 | ||
973 | if (inode) { | |
974 | wait_on_inode(inode); | |
975 | return inode; | |
976 | } | |
977 | ||
1da177e4 LT |
978 | inode = alloc_inode(sb); |
979 | if (inode) { | |
6b3304b5 | 980 | struct inode *old; |
1da177e4 | 981 | |
67a23c49 | 982 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
983 | /* We released the lock, so.. */ |
984 | old = find_inode(sb, head, test, data); | |
985 | if (!old) { | |
986 | if (set(inode, data)) | |
987 | goto set_failed; | |
988 | ||
250df6ed DC |
989 | spin_lock(&inode->i_lock); |
990 | inode->i_state = I_NEW; | |
646ec461 | 991 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 992 | spin_unlock(&inode->i_lock); |
55fa6091 | 993 | inode_sb_list_add(inode); |
67a23c49 | 994 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
995 | |
996 | /* Return the locked inode with I_NEW set, the | |
997 | * caller is responsible for filling in the contents | |
998 | */ | |
999 | return inode; | |
1000 | } | |
1001 | ||
1002 | /* | |
1003 | * Uhhuh, somebody else created the same inode under | |
1004 | * us. Use the old inode instead of the one we just | |
1005 | * allocated. | |
1006 | */ | |
67a23c49 | 1007 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1008 | destroy_inode(inode); |
1009 | inode = old; | |
1010 | wait_on_inode(inode); | |
1011 | } | |
1012 | return inode; | |
1013 | ||
1014 | set_failed: | |
67a23c49 | 1015 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1016 | destroy_inode(inode); |
1017 | return NULL; | |
1018 | } | |
0b2d0724 | 1019 | EXPORT_SYMBOL(iget5_locked); |
1da177e4 | 1020 | |
0b2d0724 CH |
1021 | /** |
1022 | * iget_locked - obtain an inode from a mounted file system | |
1023 | * @sb: super block of file system | |
1024 | * @ino: inode number to get | |
1025 | * | |
1026 | * Search for the inode specified by @ino in the inode cache and if present | |
1027 | * return it with an increased reference count. This is for file systems | |
1028 | * where the inode number is sufficient for unique identification of an inode. | |
1029 | * | |
1030 | * If the inode is not in cache, allocate a new inode and return it locked, | |
1031 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
1032 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 1033 | */ |
0b2d0724 | 1034 | struct inode *iget_locked(struct super_block *sb, unsigned long ino) |
1da177e4 | 1035 | { |
0b2d0724 | 1036 | struct hlist_head *head = inode_hashtable + hash(sb, ino); |
6b3304b5 | 1037 | struct inode *inode; |
1da177e4 | 1038 | |
0b2d0724 CH |
1039 | spin_lock(&inode_hash_lock); |
1040 | inode = find_inode_fast(sb, head, ino); | |
1041 | spin_unlock(&inode_hash_lock); | |
1042 | if (inode) { | |
1043 | wait_on_inode(inode); | |
1044 | return inode; | |
1045 | } | |
1046 | ||
1da177e4 LT |
1047 | inode = alloc_inode(sb); |
1048 | if (inode) { | |
6b3304b5 | 1049 | struct inode *old; |
1da177e4 | 1050 | |
67a23c49 | 1051 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1052 | /* We released the lock, so.. */ |
1053 | old = find_inode_fast(sb, head, ino); | |
1054 | if (!old) { | |
1055 | inode->i_ino = ino; | |
250df6ed DC |
1056 | spin_lock(&inode->i_lock); |
1057 | inode->i_state = I_NEW; | |
646ec461 | 1058 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1059 | spin_unlock(&inode->i_lock); |
55fa6091 | 1060 | inode_sb_list_add(inode); |
67a23c49 | 1061 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1062 | |
1063 | /* Return the locked inode with I_NEW set, the | |
1064 | * caller is responsible for filling in the contents | |
1065 | */ | |
1066 | return inode; | |
1067 | } | |
1068 | ||
1069 | /* | |
1070 | * Uhhuh, somebody else created the same inode under | |
1071 | * us. Use the old inode instead of the one we just | |
1072 | * allocated. | |
1073 | */ | |
67a23c49 | 1074 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1075 | destroy_inode(inode); |
1076 | inode = old; | |
1077 | wait_on_inode(inode); | |
1078 | } | |
1079 | return inode; | |
1080 | } | |
0b2d0724 | 1081 | EXPORT_SYMBOL(iget_locked); |
1da177e4 | 1082 | |
ad5e195a CH |
1083 | /* |
1084 | * search the inode cache for a matching inode number. | |
1085 | * If we find one, then the inode number we are trying to | |
1086 | * allocate is not unique and so we should not use it. | |
1087 | * | |
1088 | * Returns 1 if the inode number is unique, 0 if it is not. | |
1089 | */ | |
1090 | static int test_inode_iunique(struct super_block *sb, unsigned long ino) | |
1091 | { | |
1092 | struct hlist_head *b = inode_hashtable + hash(sb, ino); | |
1093 | struct hlist_node *node; | |
1094 | struct inode *inode; | |
1095 | ||
67a23c49 | 1096 | spin_lock(&inode_hash_lock); |
ad5e195a | 1097 | hlist_for_each_entry(inode, node, b, i_hash) { |
67a23c49 DC |
1098 | if (inode->i_ino == ino && inode->i_sb == sb) { |
1099 | spin_unlock(&inode_hash_lock); | |
ad5e195a | 1100 | return 0; |
67a23c49 | 1101 | } |
ad5e195a | 1102 | } |
67a23c49 | 1103 | spin_unlock(&inode_hash_lock); |
ad5e195a CH |
1104 | |
1105 | return 1; | |
1106 | } | |
1107 | ||
1da177e4 LT |
1108 | /** |
1109 | * iunique - get a unique inode number | |
1110 | * @sb: superblock | |
1111 | * @max_reserved: highest reserved inode number | |
1112 | * | |
1113 | * Obtain an inode number that is unique on the system for a given | |
1114 | * superblock. This is used by file systems that have no natural | |
1115 | * permanent inode numbering system. An inode number is returned that | |
1116 | * is higher than the reserved limit but unique. | |
1117 | * | |
1118 | * BUGS: | |
1119 | * With a large number of inodes live on the file system this function | |
1120 | * currently becomes quite slow. | |
1121 | */ | |
1122 | ino_t iunique(struct super_block *sb, ino_t max_reserved) | |
1123 | { | |
866b04fc JL |
1124 | /* |
1125 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
1126 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
1127 | * here to attempt to avoid that. | |
1128 | */ | |
ad5e195a | 1129 | static DEFINE_SPINLOCK(iunique_lock); |
866b04fc | 1130 | static unsigned int counter; |
1da177e4 | 1131 | ino_t res; |
3361c7be | 1132 | |
ad5e195a | 1133 | spin_lock(&iunique_lock); |
3361c7be JL |
1134 | do { |
1135 | if (counter <= max_reserved) | |
1136 | counter = max_reserved + 1; | |
1da177e4 | 1137 | res = counter++; |
ad5e195a CH |
1138 | } while (!test_inode_iunique(sb, res)); |
1139 | spin_unlock(&iunique_lock); | |
1da177e4 | 1140 | |
3361c7be JL |
1141 | return res; |
1142 | } | |
1da177e4 LT |
1143 | EXPORT_SYMBOL(iunique); |
1144 | ||
1145 | struct inode *igrab(struct inode *inode) | |
1146 | { | |
250df6ed DC |
1147 | spin_lock(&inode->i_lock); |
1148 | if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) { | |
1da177e4 | 1149 | __iget(inode); |
250df6ed DC |
1150 | spin_unlock(&inode->i_lock); |
1151 | } else { | |
1152 | spin_unlock(&inode->i_lock); | |
1da177e4 LT |
1153 | /* |
1154 | * Handle the case where s_op->clear_inode is not been | |
1155 | * called yet, and somebody is calling igrab | |
1156 | * while the inode is getting freed. | |
1157 | */ | |
1158 | inode = NULL; | |
250df6ed | 1159 | } |
1da177e4 LT |
1160 | return inode; |
1161 | } | |
1da177e4 LT |
1162 | EXPORT_SYMBOL(igrab); |
1163 | ||
1164 | /** | |
0b2d0724 | 1165 | * ilookup5_nowait - search for an inode in the inode cache |
1da177e4 | 1166 | * @sb: super block of file system to search |
0b2d0724 | 1167 | * @hashval: hash value (usually inode number) to search for |
1da177e4 LT |
1168 | * @test: callback used for comparisons between inodes |
1169 | * @data: opaque data pointer to pass to @test | |
1da177e4 | 1170 | * |
0b2d0724 | 1171 | * Search for the inode specified by @hashval and @data in the inode cache. |
1da177e4 LT |
1172 | * If the inode is in the cache, the inode is returned with an incremented |
1173 | * reference count. | |
1174 | * | |
0b2d0724 CH |
1175 | * Note: I_NEW is not waited upon so you have to be very careful what you do |
1176 | * with the returned inode. You probably should be using ilookup5() instead. | |
1da177e4 | 1177 | * |
b6d0ad68 | 1178 | * Note2: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 | 1179 | */ |
0b2d0724 CH |
1180 | struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval, |
1181 | int (*test)(struct inode *, void *), void *data) | |
1da177e4 | 1182 | { |
0b2d0724 | 1183 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); |
1da177e4 LT |
1184 | struct inode *inode; |
1185 | ||
67a23c49 | 1186 | spin_lock(&inode_hash_lock); |
1da177e4 | 1187 | inode = find_inode(sb, head, test, data); |
67a23c49 | 1188 | spin_unlock(&inode_hash_lock); |
88bd5121 | 1189 | |
0b2d0724 | 1190 | return inode; |
88bd5121 | 1191 | } |
88bd5121 AA |
1192 | EXPORT_SYMBOL(ilookup5_nowait); |
1193 | ||
1194 | /** | |
1195 | * ilookup5 - search for an inode in the inode cache | |
1196 | * @sb: super block of file system to search | |
1197 | * @hashval: hash value (usually inode number) to search for | |
1198 | * @test: callback used for comparisons between inodes | |
1199 | * @data: opaque data pointer to pass to @test | |
1200 | * | |
0b2d0724 CH |
1201 | * Search for the inode specified by @hashval and @data in the inode cache, |
1202 | * and if the inode is in the cache, return the inode with an incremented | |
1203 | * reference count. Waits on I_NEW before returning the inode. | |
88bd5121 | 1204 | * returned with an incremented reference count. |
1da177e4 | 1205 | * |
0b2d0724 CH |
1206 | * This is a generalized version of ilookup() for file systems where the |
1207 | * inode number is not sufficient for unique identification of an inode. | |
1da177e4 | 1208 | * |
0b2d0724 | 1209 | * Note: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 LT |
1210 | */ |
1211 | struct inode *ilookup5(struct super_block *sb, unsigned long hashval, | |
1212 | int (*test)(struct inode *, void *), void *data) | |
1213 | { | |
0b2d0724 | 1214 | struct inode *inode = ilookup5_nowait(sb, hashval, test, data); |
1da177e4 | 1215 | |
0b2d0724 CH |
1216 | if (inode) |
1217 | wait_on_inode(inode); | |
1218 | return inode; | |
1da177e4 | 1219 | } |
1da177e4 LT |
1220 | EXPORT_SYMBOL(ilookup5); |
1221 | ||
1222 | /** | |
1223 | * ilookup - search for an inode in the inode cache | |
1224 | * @sb: super block of file system to search | |
1225 | * @ino: inode number to search for | |
1226 | * | |
0b2d0724 CH |
1227 | * Search for the inode @ino in the inode cache, and if the inode is in the |
1228 | * cache, the inode is returned with an incremented reference count. | |
1da177e4 LT |
1229 | */ |
1230 | struct inode *ilookup(struct super_block *sb, unsigned long ino) | |
1231 | { | |
1232 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1da177e4 LT |
1233 | struct inode *inode; |
1234 | ||
0b2d0724 CH |
1235 | spin_lock(&inode_hash_lock); |
1236 | inode = find_inode_fast(sb, head, ino); | |
1237 | spin_unlock(&inode_hash_lock); | |
1da177e4 | 1238 | |
1da177e4 | 1239 | if (inode) |
0b2d0724 CH |
1240 | wait_on_inode(inode); |
1241 | return inode; | |
1da177e4 | 1242 | } |
0b2d0724 | 1243 | EXPORT_SYMBOL(ilookup); |
1da177e4 | 1244 | |
261bca86 AV |
1245 | int insert_inode_locked(struct inode *inode) |
1246 | { | |
1247 | struct super_block *sb = inode->i_sb; | |
1248 | ino_t ino = inode->i_ino; | |
1249 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
261bca86 | 1250 | |
261bca86 | 1251 | while (1) { |
72a43d63 AV |
1252 | struct hlist_node *node; |
1253 | struct inode *old = NULL; | |
67a23c49 | 1254 | spin_lock(&inode_hash_lock); |
72a43d63 AV |
1255 | hlist_for_each_entry(old, node, head, i_hash) { |
1256 | if (old->i_ino != ino) | |
1257 | continue; | |
1258 | if (old->i_sb != sb) | |
1259 | continue; | |
250df6ed DC |
1260 | spin_lock(&old->i_lock); |
1261 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1262 | spin_unlock(&old->i_lock); | |
72a43d63 | 1263 | continue; |
250df6ed | 1264 | } |
72a43d63 AV |
1265 | break; |
1266 | } | |
1267 | if (likely(!node)) { | |
250df6ed DC |
1268 | spin_lock(&inode->i_lock); |
1269 | inode->i_state |= I_NEW; | |
261bca86 | 1270 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1271 | spin_unlock(&inode->i_lock); |
67a23c49 | 1272 | spin_unlock(&inode_hash_lock); |
261bca86 AV |
1273 | return 0; |
1274 | } | |
1275 | __iget(old); | |
250df6ed | 1276 | spin_unlock(&old->i_lock); |
67a23c49 | 1277 | spin_unlock(&inode_hash_lock); |
261bca86 | 1278 | wait_on_inode(old); |
1d3382cb | 1279 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1280 | iput(old); |
1281 | return -EBUSY; | |
1282 | } | |
1283 | iput(old); | |
1284 | } | |
1285 | } | |
261bca86 AV |
1286 | EXPORT_SYMBOL(insert_inode_locked); |
1287 | ||
1288 | int insert_inode_locked4(struct inode *inode, unsigned long hashval, | |
1289 | int (*test)(struct inode *, void *), void *data) | |
1290 | { | |
1291 | struct super_block *sb = inode->i_sb; | |
1292 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
261bca86 | 1293 | |
261bca86 | 1294 | while (1) { |
72a43d63 AV |
1295 | struct hlist_node *node; |
1296 | struct inode *old = NULL; | |
1297 | ||
67a23c49 | 1298 | spin_lock(&inode_hash_lock); |
72a43d63 AV |
1299 | hlist_for_each_entry(old, node, head, i_hash) { |
1300 | if (old->i_sb != sb) | |
1301 | continue; | |
1302 | if (!test(old, data)) | |
1303 | continue; | |
250df6ed DC |
1304 | spin_lock(&old->i_lock); |
1305 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1306 | spin_unlock(&old->i_lock); | |
72a43d63 | 1307 | continue; |
250df6ed | 1308 | } |
72a43d63 AV |
1309 | break; |
1310 | } | |
1311 | if (likely(!node)) { | |
250df6ed DC |
1312 | spin_lock(&inode->i_lock); |
1313 | inode->i_state |= I_NEW; | |
261bca86 | 1314 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1315 | spin_unlock(&inode->i_lock); |
67a23c49 | 1316 | spin_unlock(&inode_hash_lock); |
261bca86 AV |
1317 | return 0; |
1318 | } | |
1319 | __iget(old); | |
250df6ed | 1320 | spin_unlock(&old->i_lock); |
67a23c49 | 1321 | spin_unlock(&inode_hash_lock); |
261bca86 | 1322 | wait_on_inode(old); |
1d3382cb | 1323 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1324 | iput(old); |
1325 | return -EBUSY; | |
1326 | } | |
1327 | iput(old); | |
1328 | } | |
1329 | } | |
261bca86 AV |
1330 | EXPORT_SYMBOL(insert_inode_locked4); |
1331 | ||
1da177e4 | 1332 | |
45321ac5 AV |
1333 | int generic_delete_inode(struct inode *inode) |
1334 | { | |
1335 | return 1; | |
1336 | } | |
1337 | EXPORT_SYMBOL(generic_delete_inode); | |
1338 | ||
1da177e4 | 1339 | /* |
45321ac5 AV |
1340 | * Normal UNIX filesystem behaviour: delete the |
1341 | * inode when the usage count drops to zero, and | |
1342 | * i_nlink is zero. | |
1da177e4 | 1343 | */ |
45321ac5 | 1344 | int generic_drop_inode(struct inode *inode) |
1da177e4 | 1345 | { |
1d3382cb | 1346 | return !inode->i_nlink || inode_unhashed(inode); |
1da177e4 | 1347 | } |
45321ac5 | 1348 | EXPORT_SYMBOL_GPL(generic_drop_inode); |
1da177e4 | 1349 | |
45321ac5 AV |
1350 | /* |
1351 | * Called when we're dropping the last reference | |
1352 | * to an inode. | |
22fe4042 | 1353 | * |
45321ac5 AV |
1354 | * Call the FS "drop_inode()" function, defaulting to |
1355 | * the legacy UNIX filesystem behaviour. If it tells | |
1356 | * us to evict inode, do so. Otherwise, retain inode | |
1357 | * in cache if fs is alive, sync and evict if fs is | |
1358 | * shutting down. | |
22fe4042 | 1359 | */ |
45321ac5 | 1360 | static void iput_final(struct inode *inode) |
1da177e4 LT |
1361 | { |
1362 | struct super_block *sb = inode->i_sb; | |
45321ac5 AV |
1363 | const struct super_operations *op = inode->i_sb->s_op; |
1364 | int drop; | |
1365 | ||
250df6ed DC |
1366 | WARN_ON(inode->i_state & I_NEW); |
1367 | ||
45321ac5 AV |
1368 | if (op && op->drop_inode) |
1369 | drop = op->drop_inode(inode); | |
1370 | else | |
1371 | drop = generic_drop_inode(inode); | |
1da177e4 | 1372 | |
b2b2af8e DC |
1373 | if (!drop && (sb->s_flags & MS_ACTIVE)) { |
1374 | inode->i_state |= I_REFERENCED; | |
1375 | if (!(inode->i_state & (I_DIRTY|I_SYNC))) | |
1376 | inode_lru_list_add(inode); | |
1377 | spin_unlock(&inode->i_lock); | |
b2b2af8e DC |
1378 | return; |
1379 | } | |
1380 | ||
45321ac5 | 1381 | if (!drop) { |
991114c6 | 1382 | inode->i_state |= I_WILL_FREE; |
250df6ed | 1383 | spin_unlock(&inode->i_lock); |
1da177e4 | 1384 | write_inode_now(inode, 1); |
250df6ed | 1385 | spin_lock(&inode->i_lock); |
7ef0d737 | 1386 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1387 | inode->i_state &= ~I_WILL_FREE; |
1da177e4 | 1388 | } |
7ccf19a8 | 1389 | |
991114c6 | 1390 | inode->i_state |= I_FREEING; |
9e38d86f | 1391 | inode_lru_list_del(inode); |
b2b2af8e | 1392 | spin_unlock(&inode->i_lock); |
b2b2af8e | 1393 | |
644da596 | 1394 | evict(inode); |
1da177e4 LT |
1395 | } |
1396 | ||
1da177e4 | 1397 | /** |
6b3304b5 | 1398 | * iput - put an inode |
1da177e4 LT |
1399 | * @inode: inode to put |
1400 | * | |
1401 | * Puts an inode, dropping its usage count. If the inode use count hits | |
1402 | * zero, the inode is then freed and may also be destroyed. | |
1403 | * | |
1404 | * Consequently, iput() can sleep. | |
1405 | */ | |
1406 | void iput(struct inode *inode) | |
1407 | { | |
1408 | if (inode) { | |
a4ffdde6 | 1409 | BUG_ON(inode->i_state & I_CLEAR); |
1da177e4 | 1410 | |
f283c86a | 1411 | if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock)) |
1da177e4 LT |
1412 | iput_final(inode); |
1413 | } | |
1414 | } | |
1da177e4 LT |
1415 | EXPORT_SYMBOL(iput); |
1416 | ||
1417 | /** | |
1418 | * bmap - find a block number in a file | |
1419 | * @inode: inode of file | |
1420 | * @block: block to find | |
1421 | * | |
1422 | * Returns the block number on the device holding the inode that | |
1423 | * is the disk block number for the block of the file requested. | |
1424 | * That is, asked for block 4 of inode 1 the function will return the | |
6b3304b5 | 1425 | * disk block relative to the disk start that holds that block of the |
1da177e4 LT |
1426 | * file. |
1427 | */ | |
6b3304b5 | 1428 | sector_t bmap(struct inode *inode, sector_t block) |
1da177e4 LT |
1429 | { |
1430 | sector_t res = 0; | |
1431 | if (inode->i_mapping->a_ops->bmap) | |
1432 | res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block); | |
1433 | return res; | |
1434 | } | |
1da177e4 LT |
1435 | EXPORT_SYMBOL(bmap); |
1436 | ||
11ff6f05 MG |
1437 | /* |
1438 | * With relative atime, only update atime if the previous atime is | |
1439 | * earlier than either the ctime or mtime or if at least a day has | |
1440 | * passed since the last atime update. | |
1441 | */ | |
1442 | static int relatime_need_update(struct vfsmount *mnt, struct inode *inode, | |
1443 | struct timespec now) | |
1444 | { | |
1445 | ||
1446 | if (!(mnt->mnt_flags & MNT_RELATIME)) | |
1447 | return 1; | |
1448 | /* | |
1449 | * Is mtime younger than atime? If yes, update atime: | |
1450 | */ | |
1451 | if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0) | |
1452 | return 1; | |
1453 | /* | |
1454 | * Is ctime younger than atime? If yes, update atime: | |
1455 | */ | |
1456 | if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0) | |
1457 | return 1; | |
1458 | ||
1459 | /* | |
1460 | * Is the previous atime value older than a day? If yes, | |
1461 | * update atime: | |
1462 | */ | |
1463 | if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60) | |
1464 | return 1; | |
1465 | /* | |
1466 | * Good, we can skip the atime update: | |
1467 | */ | |
1468 | return 0; | |
1469 | } | |
1470 | ||
1da177e4 | 1471 | /** |
869243a0 CH |
1472 | * touch_atime - update the access time |
1473 | * @mnt: mount the inode is accessed on | |
7045f37b | 1474 | * @dentry: dentry accessed |
1da177e4 LT |
1475 | * |
1476 | * Update the accessed time on an inode and mark it for writeback. | |
1477 | * This function automatically handles read only file systems and media, | |
1478 | * as well as the "noatime" flag and inode specific "noatime" markers. | |
1479 | */ | |
869243a0 | 1480 | void touch_atime(struct vfsmount *mnt, struct dentry *dentry) |
1da177e4 | 1481 | { |
869243a0 | 1482 | struct inode *inode = dentry->d_inode; |
1da177e4 LT |
1483 | struct timespec now; |
1484 | ||
cdb70f3f | 1485 | if (inode->i_flags & S_NOATIME) |
b12536c2 | 1486 | return; |
37756ced | 1487 | if (IS_NOATIME(inode)) |
b12536c2 | 1488 | return; |
b2276138 | 1489 | if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1490 | return; |
47ae32d6 | 1491 | |
cdb70f3f | 1492 | if (mnt->mnt_flags & MNT_NOATIME) |
b12536c2 | 1493 | return; |
cdb70f3f | 1494 | if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1495 | return; |
1da177e4 LT |
1496 | |
1497 | now = current_fs_time(inode->i_sb); | |
11ff6f05 MG |
1498 | |
1499 | if (!relatime_need_update(mnt, inode, now)) | |
b12536c2 | 1500 | return; |
11ff6f05 | 1501 | |
47ae32d6 | 1502 | if (timespec_equal(&inode->i_atime, &now)) |
b12536c2 AK |
1503 | return; |
1504 | ||
1505 | if (mnt_want_write(mnt)) | |
1506 | return; | |
47ae32d6 VH |
1507 | |
1508 | inode->i_atime = now; | |
1509 | mark_inode_dirty_sync(inode); | |
cdb70f3f | 1510 | mnt_drop_write(mnt); |
1da177e4 | 1511 | } |
869243a0 | 1512 | EXPORT_SYMBOL(touch_atime); |
1da177e4 LT |
1513 | |
1514 | /** | |
870f4817 CH |
1515 | * file_update_time - update mtime and ctime time |
1516 | * @file: file accessed | |
1da177e4 | 1517 | * |
870f4817 CH |
1518 | * Update the mtime and ctime members of an inode and mark the inode |
1519 | * for writeback. Note that this function is meant exclusively for | |
1520 | * usage in the file write path of filesystems, and filesystems may | |
1521 | * choose to explicitly ignore update via this function with the | |
2eadfc0e | 1522 | * S_NOCMTIME inode flag, e.g. for network filesystem where these |
870f4817 | 1523 | * timestamps are handled by the server. |
1da177e4 LT |
1524 | */ |
1525 | ||
870f4817 | 1526 | void file_update_time(struct file *file) |
1da177e4 | 1527 | { |
0f7fc9e4 | 1528 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 | 1529 | struct timespec now; |
ce06e0b2 | 1530 | enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0; |
1da177e4 | 1531 | |
ce06e0b2 | 1532 | /* First try to exhaust all avenues to not sync */ |
1da177e4 LT |
1533 | if (IS_NOCMTIME(inode)) |
1534 | return; | |
20ddee2c | 1535 | |
1da177e4 | 1536 | now = current_fs_time(inode->i_sb); |
ce06e0b2 AK |
1537 | if (!timespec_equal(&inode->i_mtime, &now)) |
1538 | sync_it = S_MTIME; | |
1da177e4 | 1539 | |
ce06e0b2 AK |
1540 | if (!timespec_equal(&inode->i_ctime, &now)) |
1541 | sync_it |= S_CTIME; | |
870f4817 | 1542 | |
ce06e0b2 AK |
1543 | if (IS_I_VERSION(inode)) |
1544 | sync_it |= S_VERSION; | |
7a224228 | 1545 | |
ce06e0b2 AK |
1546 | if (!sync_it) |
1547 | return; | |
1548 | ||
1549 | /* Finally allowed to write? Takes lock. */ | |
1550 | if (mnt_want_write_file(file)) | |
1551 | return; | |
1552 | ||
1553 | /* Only change inode inside the lock region */ | |
1554 | if (sync_it & S_VERSION) | |
1555 | inode_inc_iversion(inode); | |
1556 | if (sync_it & S_CTIME) | |
1557 | inode->i_ctime = now; | |
1558 | if (sync_it & S_MTIME) | |
1559 | inode->i_mtime = now; | |
1560 | mark_inode_dirty_sync(inode); | |
20ddee2c | 1561 | mnt_drop_write(file->f_path.mnt); |
1da177e4 | 1562 | } |
870f4817 | 1563 | EXPORT_SYMBOL(file_update_time); |
1da177e4 LT |
1564 | |
1565 | int inode_needs_sync(struct inode *inode) | |
1566 | { | |
1567 | if (IS_SYNC(inode)) | |
1568 | return 1; | |
1569 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) | |
1570 | return 1; | |
1571 | return 0; | |
1572 | } | |
1da177e4 LT |
1573 | EXPORT_SYMBOL(inode_needs_sync); |
1574 | ||
1da177e4 LT |
1575 | int inode_wait(void *word) |
1576 | { | |
1577 | schedule(); | |
1578 | return 0; | |
1579 | } | |
d44dab8d | 1580 | EXPORT_SYMBOL(inode_wait); |
1da177e4 LT |
1581 | |
1582 | /* | |
168a9fd6 MS |
1583 | * If we try to find an inode in the inode hash while it is being |
1584 | * deleted, we have to wait until the filesystem completes its | |
1585 | * deletion before reporting that it isn't found. This function waits | |
1586 | * until the deletion _might_ have completed. Callers are responsible | |
1587 | * to recheck inode state. | |
1588 | * | |
eaff8079 | 1589 | * It doesn't matter if I_NEW is not set initially, a call to |
250df6ed DC |
1590 | * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list |
1591 | * will DTRT. | |
1da177e4 LT |
1592 | */ |
1593 | static void __wait_on_freeing_inode(struct inode *inode) | |
1594 | { | |
1595 | wait_queue_head_t *wq; | |
eaff8079 CH |
1596 | DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW); |
1597 | wq = bit_waitqueue(&inode->i_state, __I_NEW); | |
1da177e4 | 1598 | prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE); |
250df6ed | 1599 | spin_unlock(&inode->i_lock); |
67a23c49 | 1600 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1601 | schedule(); |
1602 | finish_wait(wq, &wait.wait); | |
67a23c49 | 1603 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1604 | } |
1605 | ||
1da177e4 LT |
1606 | static __initdata unsigned long ihash_entries; |
1607 | static int __init set_ihash_entries(char *str) | |
1608 | { | |
1609 | if (!str) | |
1610 | return 0; | |
1611 | ihash_entries = simple_strtoul(str, &str, 0); | |
1612 | return 1; | |
1613 | } | |
1614 | __setup("ihash_entries=", set_ihash_entries); | |
1615 | ||
1616 | /* | |
1617 | * Initialize the waitqueues and inode hash table. | |
1618 | */ | |
1619 | void __init inode_init_early(void) | |
1620 | { | |
1621 | int loop; | |
1622 | ||
1623 | /* If hashes are distributed across NUMA nodes, defer | |
1624 | * hash allocation until vmalloc space is available. | |
1625 | */ | |
1626 | if (hashdist) | |
1627 | return; | |
1628 | ||
1629 | inode_hashtable = | |
1630 | alloc_large_system_hash("Inode-cache", | |
1631 | sizeof(struct hlist_head), | |
1632 | ihash_entries, | |
1633 | 14, | |
1634 | HASH_EARLY, | |
1635 | &i_hash_shift, | |
1636 | &i_hash_mask, | |
1637 | 0); | |
1638 | ||
1639 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1640 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1641 | } | |
1642 | ||
74bf17cf | 1643 | void __init inode_init(void) |
1da177e4 LT |
1644 | { |
1645 | int loop; | |
1646 | ||
1647 | /* inode slab cache */ | |
b0196009 PJ |
1648 | inode_cachep = kmem_cache_create("inode_cache", |
1649 | sizeof(struct inode), | |
1650 | 0, | |
1651 | (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC| | |
1652 | SLAB_MEM_SPREAD), | |
20c2df83 | 1653 | init_once); |
8e1f936b | 1654 | register_shrinker(&icache_shrinker); |
1da177e4 LT |
1655 | |
1656 | /* Hash may have been set up in inode_init_early */ | |
1657 | if (!hashdist) | |
1658 | return; | |
1659 | ||
1660 | inode_hashtable = | |
1661 | alloc_large_system_hash("Inode-cache", | |
1662 | sizeof(struct hlist_head), | |
1663 | ihash_entries, | |
1664 | 14, | |
1665 | 0, | |
1666 | &i_hash_shift, | |
1667 | &i_hash_mask, | |
1668 | 0); | |
1669 | ||
1670 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1671 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1672 | } | |
1673 | ||
1674 | void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev) | |
1675 | { | |
1676 | inode->i_mode = mode; | |
1677 | if (S_ISCHR(mode)) { | |
1678 | inode->i_fop = &def_chr_fops; | |
1679 | inode->i_rdev = rdev; | |
1680 | } else if (S_ISBLK(mode)) { | |
1681 | inode->i_fop = &def_blk_fops; | |
1682 | inode->i_rdev = rdev; | |
1683 | } else if (S_ISFIFO(mode)) | |
1684 | inode->i_fop = &def_fifo_fops; | |
1685 | else if (S_ISSOCK(mode)) | |
1686 | inode->i_fop = &bad_sock_fops; | |
1687 | else | |
af0d9ae8 MK |
1688 | printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for" |
1689 | " inode %s:%lu\n", mode, inode->i_sb->s_id, | |
1690 | inode->i_ino); | |
1da177e4 LT |
1691 | } |
1692 | EXPORT_SYMBOL(init_special_inode); | |
a1bd120d DM |
1693 | |
1694 | /** | |
eaae668d | 1695 | * inode_init_owner - Init uid,gid,mode for new inode according to posix standards |
a1bd120d DM |
1696 | * @inode: New inode |
1697 | * @dir: Directory inode | |
1698 | * @mode: mode of the new inode | |
1699 | */ | |
1700 | void inode_init_owner(struct inode *inode, const struct inode *dir, | |
1701 | mode_t mode) | |
1702 | { | |
1703 | inode->i_uid = current_fsuid(); | |
1704 | if (dir && dir->i_mode & S_ISGID) { | |
1705 | inode->i_gid = dir->i_gid; | |
1706 | if (S_ISDIR(mode)) | |
1707 | mode |= S_ISGID; | |
1708 | } else | |
1709 | inode->i_gid = current_fsgid(); | |
1710 | inode->i_mode = mode; | |
1711 | } | |
1712 | EXPORT_SYMBOL(inode_init_owner); | |
e795b717 | 1713 | |
2e149670 SH |
1714 | /** |
1715 | * inode_owner_or_capable - check current task permissions to inode | |
1716 | * @inode: inode being checked | |
1717 | * | |
1718 | * Return true if current either has CAP_FOWNER to the inode, or | |
1719 | * owns the file. | |
e795b717 | 1720 | */ |
2e149670 | 1721 | bool inode_owner_or_capable(const struct inode *inode) |
e795b717 SH |
1722 | { |
1723 | struct user_namespace *ns = inode_userns(inode); | |
1724 | ||
1725 | if (current_user_ns() == ns && current_fsuid() == inode->i_uid) | |
1726 | return true; | |
1727 | if (ns_capable(ns, CAP_FOWNER)) | |
1728 | return true; | |
1729 | return false; | |
1730 | } | |
2e149670 | 1731 | EXPORT_SYMBOL(inode_owner_or_capable); |