]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
1da177e4 | 2 | * (C) 1997 Linus Torvalds |
4b4563dc | 3 | * (C) 1999 Andrea Arcangeli <andrea@suse.de> (dynamic inode allocation) |
1da177e4 | 4 | */ |
1da177e4 LT |
5 | #include <linux/fs.h> |
6 | #include <linux/mm.h> | |
7 | #include <linux/dcache.h> | |
8 | #include <linux/init.h> | |
1da177e4 LT |
9 | #include <linux/slab.h> |
10 | #include <linux/writeback.h> | |
11 | #include <linux/module.h> | |
12 | #include <linux/backing-dev.h> | |
13 | #include <linux/wait.h> | |
88e0fbc4 | 14 | #include <linux/rwsem.h> |
1da177e4 LT |
15 | #include <linux/hash.h> |
16 | #include <linux/swap.h> | |
17 | #include <linux/security.h> | |
18 | #include <linux/pagemap.h> | |
19 | #include <linux/cdev.h> | |
20 | #include <linux/bootmem.h> | |
3be25f49 | 21 | #include <linux/fsnotify.h> |
fc33a7bb | 22 | #include <linux/mount.h> |
efaee192 | 23 | #include <linux/async.h> |
f19d4a8f | 24 | #include <linux/posix_acl.h> |
9ce6e0be | 25 | #include <linux/prefetch.h> |
a178d202 | 26 | #include <linux/ima.h> |
e795b717 | 27 | #include <linux/cred.h> |
4b4563dc | 28 | #include <linux/buffer_head.h> /* for inode_has_buffers */ |
a66979ab | 29 | #include "internal.h" |
1da177e4 | 30 | |
250df6ed | 31 | /* |
4b4563dc | 32 | * Inode locking rules: |
250df6ed DC |
33 | * |
34 | * inode->i_lock protects: | |
35 | * inode->i_state, inode->i_hash, __iget() | |
09cc9fc7 | 36 | * inode->i_sb->s_inode_lru_lock protects: |
98b745c6 | 37 | * inode->i_sb->s_inode_lru, inode->i_lru |
55fa6091 DC |
38 | * inode_sb_list_lock protects: |
39 | * sb->s_inodes, inode->i_sb_list | |
a66979ab DC |
40 | * inode_wb_list_lock protects: |
41 | * bdi->wb.b_{dirty,io,more_io}, inode->i_wb_list | |
67a23c49 DC |
42 | * inode_hash_lock protects: |
43 | * inode_hashtable, inode->i_hash | |
250df6ed DC |
44 | * |
45 | * Lock ordering: | |
55fa6091 DC |
46 | * |
47 | * inode_sb_list_lock | |
48 | * inode->i_lock | |
09cc9fc7 | 49 | * inode->i_sb->s_inode_lru_lock |
a66979ab DC |
50 | * |
51 | * inode_wb_list_lock | |
52 | * inode->i_lock | |
67a23c49 DC |
53 | * |
54 | * inode_hash_lock | |
55 | * inode_sb_list_lock | |
56 | * inode->i_lock | |
57 | * | |
58 | * iunique_lock | |
59 | * inode_hash_lock | |
250df6ed DC |
60 | */ |
61 | ||
fa3536cc ED |
62 | static unsigned int i_hash_mask __read_mostly; |
63 | static unsigned int i_hash_shift __read_mostly; | |
67a23c49 DC |
64 | static struct hlist_head *inode_hashtable __read_mostly; |
65 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_hash_lock); | |
1da177e4 | 66 | |
55fa6091 | 67 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_sb_list_lock); |
a66979ab | 68 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_wb_list_lock); |
55fa6091 | 69 | |
7dcda1c9 JA |
70 | /* |
71 | * Empty aops. Can be used for the cases where the user does not | |
72 | * define any of the address_space operations. | |
73 | */ | |
74 | const struct address_space_operations empty_aops = { | |
75 | }; | |
76 | EXPORT_SYMBOL(empty_aops); | |
77 | ||
1da177e4 LT |
78 | /* |
79 | * Statistics gathering.. | |
80 | */ | |
81 | struct inodes_stat_t inodes_stat; | |
82 | ||
3e880fb5 | 83 | static DEFINE_PER_CPU(unsigned int, nr_inodes); |
fcb94f72 | 84 | static DEFINE_PER_CPU(unsigned int, nr_unused); |
cffbc8aa | 85 | |
6b3304b5 | 86 | static struct kmem_cache *inode_cachep __read_mostly; |
1da177e4 | 87 | |
3e880fb5 | 88 | static int get_nr_inodes(void) |
cffbc8aa | 89 | { |
3e880fb5 NP |
90 | int i; |
91 | int sum = 0; | |
92 | for_each_possible_cpu(i) | |
93 | sum += per_cpu(nr_inodes, i); | |
94 | return sum < 0 ? 0 : sum; | |
cffbc8aa DC |
95 | } |
96 | ||
97 | static inline int get_nr_inodes_unused(void) | |
98 | { | |
fcb94f72 DC |
99 | int i; |
100 | int sum = 0; | |
101 | for_each_possible_cpu(i) | |
102 | sum += per_cpu(nr_unused, i); | |
103 | return sum < 0 ? 0 : sum; | |
cffbc8aa DC |
104 | } |
105 | ||
106 | int get_nr_dirty_inodes(void) | |
107 | { | |
3e880fb5 | 108 | /* not actually dirty inodes, but a wild approximation */ |
cffbc8aa DC |
109 | int nr_dirty = get_nr_inodes() - get_nr_inodes_unused(); |
110 | return nr_dirty > 0 ? nr_dirty : 0; | |
cffbc8aa DC |
111 | } |
112 | ||
113 | /* | |
114 | * Handle nr_inode sysctl | |
115 | */ | |
116 | #ifdef CONFIG_SYSCTL | |
117 | int proc_nr_inodes(ctl_table *table, int write, | |
118 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
119 | { | |
120 | inodes_stat.nr_inodes = get_nr_inodes(); | |
fcb94f72 | 121 | inodes_stat.nr_unused = get_nr_inodes_unused(); |
cffbc8aa DC |
122 | return proc_dointvec(table, write, buffer, lenp, ppos); |
123 | } | |
124 | #endif | |
125 | ||
2cb1599f DC |
126 | /** |
127 | * inode_init_always - perform inode structure intialisation | |
0bc02f3f RD |
128 | * @sb: superblock inode belongs to |
129 | * @inode: inode to initialise | |
2cb1599f DC |
130 | * |
131 | * These are initializations that need to be done on every inode | |
132 | * allocation as the fields are not initialised by slab allocation. | |
133 | */ | |
54e34621 | 134 | int inode_init_always(struct super_block *sb, struct inode *inode) |
1da177e4 | 135 | { |
6e1d5dcc | 136 | static const struct inode_operations empty_iops; |
99ac48f5 | 137 | static const struct file_operations empty_fops; |
6b3304b5 | 138 | struct address_space *const mapping = &inode->i_data; |
2cb1599f DC |
139 | |
140 | inode->i_sb = sb; | |
141 | inode->i_blkbits = sb->s_blocksize_bits; | |
142 | inode->i_flags = 0; | |
143 | atomic_set(&inode->i_count, 1); | |
144 | inode->i_op = &empty_iops; | |
145 | inode->i_fop = &empty_fops; | |
146 | inode->i_nlink = 1; | |
56ff5efa AV |
147 | inode->i_uid = 0; |
148 | inode->i_gid = 0; | |
2cb1599f DC |
149 | atomic_set(&inode->i_writecount, 0); |
150 | inode->i_size = 0; | |
151 | inode->i_blocks = 0; | |
152 | inode->i_bytes = 0; | |
153 | inode->i_generation = 0; | |
1da177e4 | 154 | #ifdef CONFIG_QUOTA |
2cb1599f | 155 | memset(&inode->i_dquot, 0, sizeof(inode->i_dquot)); |
1da177e4 | 156 | #endif |
2cb1599f DC |
157 | inode->i_pipe = NULL; |
158 | inode->i_bdev = NULL; | |
159 | inode->i_cdev = NULL; | |
160 | inode->i_rdev = 0; | |
161 | inode->dirtied_when = 0; | |
6146f0d5 MZ |
162 | |
163 | if (security_inode_alloc(inode)) | |
54e34621 | 164 | goto out; |
2cb1599f DC |
165 | spin_lock_init(&inode->i_lock); |
166 | lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key); | |
167 | ||
168 | mutex_init(&inode->i_mutex); | |
169 | lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key); | |
170 | ||
bd5fe6c5 | 171 | atomic_set(&inode->i_dio_count, 0); |
2cb1599f DC |
172 | |
173 | mapping->a_ops = &empty_aops; | |
174 | mapping->host = inode; | |
175 | mapping->flags = 0; | |
3c1d4378 | 176 | mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE); |
2cb1599f DC |
177 | mapping->assoc_mapping = NULL; |
178 | mapping->backing_dev_info = &default_backing_dev_info; | |
179 | mapping->writeback_index = 0; | |
180 | ||
181 | /* | |
182 | * If the block_device provides a backing_dev_info for client | |
183 | * inodes then use that. Otherwise the inode share the bdev's | |
184 | * backing_dev_info. | |
185 | */ | |
186 | if (sb->s_bdev) { | |
187 | struct backing_dev_info *bdi; | |
188 | ||
2c96ce9f | 189 | bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info; |
2cb1599f DC |
190 | mapping->backing_dev_info = bdi; |
191 | } | |
192 | inode->i_private = NULL; | |
193 | inode->i_mapping = mapping; | |
f19d4a8f AV |
194 | #ifdef CONFIG_FS_POSIX_ACL |
195 | inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED; | |
196 | #endif | |
2cb1599f | 197 | |
3be25f49 EP |
198 | #ifdef CONFIG_FSNOTIFY |
199 | inode->i_fsnotify_mask = 0; | |
200 | #endif | |
201 | ||
3e880fb5 | 202 | this_cpu_inc(nr_inodes); |
cffbc8aa | 203 | |
54e34621 | 204 | return 0; |
54e34621 CH |
205 | out: |
206 | return -ENOMEM; | |
1da177e4 | 207 | } |
2cb1599f DC |
208 | EXPORT_SYMBOL(inode_init_always); |
209 | ||
210 | static struct inode *alloc_inode(struct super_block *sb) | |
211 | { | |
212 | struct inode *inode; | |
213 | ||
214 | if (sb->s_op->alloc_inode) | |
215 | inode = sb->s_op->alloc_inode(sb); | |
216 | else | |
217 | inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL); | |
218 | ||
54e34621 CH |
219 | if (!inode) |
220 | return NULL; | |
221 | ||
222 | if (unlikely(inode_init_always(sb, inode))) { | |
223 | if (inode->i_sb->s_op->destroy_inode) | |
224 | inode->i_sb->s_op->destroy_inode(inode); | |
225 | else | |
226 | kmem_cache_free(inode_cachep, inode); | |
227 | return NULL; | |
228 | } | |
229 | ||
230 | return inode; | |
2cb1599f | 231 | } |
1da177e4 | 232 | |
ff0c7d15 NP |
233 | void free_inode_nonrcu(struct inode *inode) |
234 | { | |
235 | kmem_cache_free(inode_cachep, inode); | |
236 | } | |
237 | EXPORT_SYMBOL(free_inode_nonrcu); | |
238 | ||
2e00c97e | 239 | void __destroy_inode(struct inode *inode) |
1da177e4 | 240 | { |
b7542f8c | 241 | BUG_ON(inode_has_buffers(inode)); |
1da177e4 | 242 | security_inode_free(inode); |
3be25f49 | 243 | fsnotify_inode_delete(inode); |
f19d4a8f AV |
244 | #ifdef CONFIG_FS_POSIX_ACL |
245 | if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED) | |
246 | posix_acl_release(inode->i_acl); | |
247 | if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED) | |
248 | posix_acl_release(inode->i_default_acl); | |
249 | #endif | |
3e880fb5 | 250 | this_cpu_dec(nr_inodes); |
2e00c97e CH |
251 | } |
252 | EXPORT_SYMBOL(__destroy_inode); | |
253 | ||
fa0d7e3d NP |
254 | static void i_callback(struct rcu_head *head) |
255 | { | |
256 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
257 | INIT_LIST_HEAD(&inode->i_dentry); | |
258 | kmem_cache_free(inode_cachep, inode); | |
259 | } | |
260 | ||
56b0dacf | 261 | static void destroy_inode(struct inode *inode) |
2e00c97e | 262 | { |
7ccf19a8 | 263 | BUG_ON(!list_empty(&inode->i_lru)); |
2e00c97e | 264 | __destroy_inode(inode); |
1da177e4 LT |
265 | if (inode->i_sb->s_op->destroy_inode) |
266 | inode->i_sb->s_op->destroy_inode(inode); | |
267 | else | |
fa0d7e3d | 268 | call_rcu(&inode->i_rcu, i_callback); |
1da177e4 | 269 | } |
1da177e4 | 270 | |
2aa15890 MS |
271 | void address_space_init_once(struct address_space *mapping) |
272 | { | |
273 | memset(mapping, 0, sizeof(*mapping)); | |
274 | INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC); | |
275 | spin_lock_init(&mapping->tree_lock); | |
3d48ae45 | 276 | mutex_init(&mapping->i_mmap_mutex); |
2aa15890 MS |
277 | INIT_LIST_HEAD(&mapping->private_list); |
278 | spin_lock_init(&mapping->private_lock); | |
279 | INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap); | |
280 | INIT_LIST_HEAD(&mapping->i_mmap_nonlinear); | |
2aa15890 MS |
281 | } |
282 | EXPORT_SYMBOL(address_space_init_once); | |
283 | ||
1da177e4 LT |
284 | /* |
285 | * These are initializations that only need to be done | |
286 | * once, because the fields are idempotent across use | |
287 | * of the inode, so let the slab aware of that. | |
288 | */ | |
289 | void inode_init_once(struct inode *inode) | |
290 | { | |
291 | memset(inode, 0, sizeof(*inode)); | |
292 | INIT_HLIST_NODE(&inode->i_hash); | |
293 | INIT_LIST_HEAD(&inode->i_dentry); | |
294 | INIT_LIST_HEAD(&inode->i_devices); | |
7ccf19a8 NP |
295 | INIT_LIST_HEAD(&inode->i_wb_list); |
296 | INIT_LIST_HEAD(&inode->i_lru); | |
2aa15890 | 297 | address_space_init_once(&inode->i_data); |
1da177e4 | 298 | i_size_ordered_init(inode); |
3be25f49 | 299 | #ifdef CONFIG_FSNOTIFY |
e61ce867 | 300 | INIT_HLIST_HEAD(&inode->i_fsnotify_marks); |
3be25f49 | 301 | #endif |
1da177e4 | 302 | } |
1da177e4 LT |
303 | EXPORT_SYMBOL(inode_init_once); |
304 | ||
51cc5068 | 305 | static void init_once(void *foo) |
1da177e4 | 306 | { |
6b3304b5 | 307 | struct inode *inode = (struct inode *) foo; |
1da177e4 | 308 | |
a35afb83 | 309 | inode_init_once(inode); |
1da177e4 LT |
310 | } |
311 | ||
312 | /* | |
250df6ed | 313 | * inode->i_lock must be held |
1da177e4 | 314 | */ |
6b3304b5 | 315 | void __iget(struct inode *inode) |
1da177e4 | 316 | { |
9e38d86f NP |
317 | atomic_inc(&inode->i_count); |
318 | } | |
2e147f1e | 319 | |
7de9c6ee AV |
320 | /* |
321 | * get additional reference to inode; caller must already hold one. | |
322 | */ | |
323 | void ihold(struct inode *inode) | |
324 | { | |
325 | WARN_ON(atomic_inc_return(&inode->i_count) < 2); | |
326 | } | |
327 | EXPORT_SYMBOL(ihold); | |
328 | ||
9e38d86f NP |
329 | static void inode_lru_list_add(struct inode *inode) |
330 | { | |
09cc9fc7 | 331 | spin_lock(&inode->i_sb->s_inode_lru_lock); |
7ccf19a8 | 332 | if (list_empty(&inode->i_lru)) { |
98b745c6 DC |
333 | list_add(&inode->i_lru, &inode->i_sb->s_inode_lru); |
334 | inode->i_sb->s_nr_inodes_unused++; | |
fcb94f72 | 335 | this_cpu_inc(nr_unused); |
9e38d86f | 336 | } |
09cc9fc7 | 337 | spin_unlock(&inode->i_sb->s_inode_lru_lock); |
9e38d86f | 338 | } |
2e147f1e | 339 | |
9e38d86f NP |
340 | static void inode_lru_list_del(struct inode *inode) |
341 | { | |
09cc9fc7 | 342 | spin_lock(&inode->i_sb->s_inode_lru_lock); |
7ccf19a8 NP |
343 | if (!list_empty(&inode->i_lru)) { |
344 | list_del_init(&inode->i_lru); | |
98b745c6 | 345 | inode->i_sb->s_nr_inodes_unused--; |
fcb94f72 | 346 | this_cpu_dec(nr_unused); |
9e38d86f | 347 | } |
09cc9fc7 | 348 | spin_unlock(&inode->i_sb->s_inode_lru_lock); |
1da177e4 LT |
349 | } |
350 | ||
646ec461 CH |
351 | /** |
352 | * inode_sb_list_add - add inode to the superblock list of inodes | |
353 | * @inode: inode to add | |
354 | */ | |
355 | void inode_sb_list_add(struct inode *inode) | |
356 | { | |
55fa6091 DC |
357 | spin_lock(&inode_sb_list_lock); |
358 | list_add(&inode->i_sb_list, &inode->i_sb->s_inodes); | |
359 | spin_unlock(&inode_sb_list_lock); | |
646ec461 CH |
360 | } |
361 | EXPORT_SYMBOL_GPL(inode_sb_list_add); | |
362 | ||
55fa6091 | 363 | static inline void inode_sb_list_del(struct inode *inode) |
646ec461 | 364 | { |
55fa6091 | 365 | spin_lock(&inode_sb_list_lock); |
646ec461 | 366 | list_del_init(&inode->i_sb_list); |
55fa6091 | 367 | spin_unlock(&inode_sb_list_lock); |
646ec461 CH |
368 | } |
369 | ||
4c51acbc DC |
370 | static unsigned long hash(struct super_block *sb, unsigned long hashval) |
371 | { | |
372 | unsigned long tmp; | |
373 | ||
374 | tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) / | |
375 | L1_CACHE_BYTES; | |
4b4563dc CH |
376 | tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> i_hash_shift); |
377 | return tmp & i_hash_mask; | |
4c51acbc DC |
378 | } |
379 | ||
380 | /** | |
381 | * __insert_inode_hash - hash an inode | |
382 | * @inode: unhashed inode | |
383 | * @hashval: unsigned long value used to locate this object in the | |
384 | * inode_hashtable. | |
385 | * | |
386 | * Add an inode to the inode hash for this superblock. | |
387 | */ | |
388 | void __insert_inode_hash(struct inode *inode, unsigned long hashval) | |
389 | { | |
646ec461 CH |
390 | struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval); |
391 | ||
67a23c49 | 392 | spin_lock(&inode_hash_lock); |
250df6ed | 393 | spin_lock(&inode->i_lock); |
646ec461 | 394 | hlist_add_head(&inode->i_hash, b); |
250df6ed | 395 | spin_unlock(&inode->i_lock); |
67a23c49 | 396 | spin_unlock(&inode_hash_lock); |
4c51acbc DC |
397 | } |
398 | EXPORT_SYMBOL(__insert_inode_hash); | |
399 | ||
4c51acbc DC |
400 | /** |
401 | * remove_inode_hash - remove an inode from the hash | |
402 | * @inode: inode to unhash | |
403 | * | |
404 | * Remove an inode from the superblock. | |
405 | */ | |
406 | void remove_inode_hash(struct inode *inode) | |
407 | { | |
67a23c49 | 408 | spin_lock(&inode_hash_lock); |
250df6ed | 409 | spin_lock(&inode->i_lock); |
4c51acbc | 410 | hlist_del_init(&inode->i_hash); |
250df6ed | 411 | spin_unlock(&inode->i_lock); |
67a23c49 | 412 | spin_unlock(&inode_hash_lock); |
4c51acbc DC |
413 | } |
414 | EXPORT_SYMBOL(remove_inode_hash); | |
415 | ||
b0683aa6 AV |
416 | void end_writeback(struct inode *inode) |
417 | { | |
418 | might_sleep(); | |
08142579 JK |
419 | /* |
420 | * We have to cycle tree_lock here because reclaim can be still in the | |
421 | * process of removing the last page (in __delete_from_page_cache()) | |
422 | * and we must not free mapping under it. | |
423 | */ | |
424 | spin_lock_irq(&inode->i_data.tree_lock); | |
b0683aa6 | 425 | BUG_ON(inode->i_data.nrpages); |
08142579 | 426 | spin_unlock_irq(&inode->i_data.tree_lock); |
b0683aa6 AV |
427 | BUG_ON(!list_empty(&inode->i_data.private_list)); |
428 | BUG_ON(!(inode->i_state & I_FREEING)); | |
429 | BUG_ON(inode->i_state & I_CLEAR); | |
430 | inode_sync_wait(inode); | |
fa0d7e3d | 431 | /* don't need i_lock here, no concurrent mods to i_state */ |
b0683aa6 AV |
432 | inode->i_state = I_FREEING | I_CLEAR; |
433 | } | |
434 | EXPORT_SYMBOL(end_writeback); | |
435 | ||
b2b2af8e DC |
436 | /* |
437 | * Free the inode passed in, removing it from the lists it is still connected | |
438 | * to. We remove any pages still attached to the inode and wait for any IO that | |
439 | * is still in progress before finally destroying the inode. | |
440 | * | |
441 | * An inode must already be marked I_FREEING so that we avoid the inode being | |
442 | * moved back onto lists if we race with other code that manipulates the lists | |
443 | * (e.g. writeback_single_inode). The caller is responsible for setting this. | |
444 | * | |
445 | * An inode must already be removed from the LRU list before being evicted from | |
446 | * the cache. This should occur atomically with setting the I_FREEING state | |
447 | * flag, so no inodes here should ever be on the LRU when being evicted. | |
448 | */ | |
644da596 | 449 | static void evict(struct inode *inode) |
b4272d4c AV |
450 | { |
451 | const struct super_operations *op = inode->i_sb->s_op; | |
452 | ||
b2b2af8e DC |
453 | BUG_ON(!(inode->i_state & I_FREEING)); |
454 | BUG_ON(!list_empty(&inode->i_lru)); | |
455 | ||
a66979ab | 456 | inode_wb_list_del(inode); |
55fa6091 DC |
457 | inode_sb_list_del(inode); |
458 | ||
be7ce416 AV |
459 | if (op->evict_inode) { |
460 | op->evict_inode(inode); | |
b4272d4c AV |
461 | } else { |
462 | if (inode->i_data.nrpages) | |
463 | truncate_inode_pages(&inode->i_data, 0); | |
30140837 | 464 | end_writeback(inode); |
b4272d4c | 465 | } |
661074e9 AV |
466 | if (S_ISBLK(inode->i_mode) && inode->i_bdev) |
467 | bd_forget(inode); | |
468 | if (S_ISCHR(inode->i_mode) && inode->i_cdev) | |
469 | cd_forget(inode); | |
b2b2af8e DC |
470 | |
471 | remove_inode_hash(inode); | |
472 | ||
473 | spin_lock(&inode->i_lock); | |
474 | wake_up_bit(&inode->i_state, __I_NEW); | |
475 | BUG_ON(inode->i_state != (I_FREEING | I_CLEAR)); | |
476 | spin_unlock(&inode->i_lock); | |
477 | ||
478 | destroy_inode(inode); | |
b4272d4c AV |
479 | } |
480 | ||
1da177e4 LT |
481 | /* |
482 | * dispose_list - dispose of the contents of a local list | |
483 | * @head: the head of the list to free | |
484 | * | |
485 | * Dispose-list gets a local list with local inodes in it, so it doesn't | |
486 | * need to worry about list corruption and SMP locks. | |
487 | */ | |
488 | static void dispose_list(struct list_head *head) | |
489 | { | |
1da177e4 LT |
490 | while (!list_empty(head)) { |
491 | struct inode *inode; | |
492 | ||
7ccf19a8 NP |
493 | inode = list_first_entry(head, struct inode, i_lru); |
494 | list_del_init(&inode->i_lru); | |
1da177e4 | 495 | |
644da596 | 496 | evict(inode); |
1da177e4 | 497 | } |
1da177e4 LT |
498 | } |
499 | ||
63997e98 AV |
500 | /** |
501 | * evict_inodes - evict all evictable inodes for a superblock | |
502 | * @sb: superblock to operate on | |
503 | * | |
504 | * Make sure that no inodes with zero refcount are retained. This is | |
505 | * called by superblock shutdown after having MS_ACTIVE flag removed, | |
506 | * so any inode reaching zero refcount during or after that call will | |
507 | * be immediately evicted. | |
1da177e4 | 508 | */ |
63997e98 | 509 | void evict_inodes(struct super_block *sb) |
1da177e4 | 510 | { |
63997e98 AV |
511 | struct inode *inode, *next; |
512 | LIST_HEAD(dispose); | |
1da177e4 | 513 | |
55fa6091 | 514 | spin_lock(&inode_sb_list_lock); |
63997e98 AV |
515 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
516 | if (atomic_read(&inode->i_count)) | |
aabb8fdb | 517 | continue; |
250df6ed DC |
518 | |
519 | spin_lock(&inode->i_lock); | |
520 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
521 | spin_unlock(&inode->i_lock); | |
1da177e4 | 522 | continue; |
250df6ed | 523 | } |
63997e98 AV |
524 | |
525 | inode->i_state |= I_FREEING; | |
02afc410 | 526 | inode_lru_list_del(inode); |
250df6ed | 527 | spin_unlock(&inode->i_lock); |
02afc410 | 528 | list_add(&inode->i_lru, &dispose); |
1da177e4 | 529 | } |
55fa6091 | 530 | spin_unlock(&inode_sb_list_lock); |
63997e98 AV |
531 | |
532 | dispose_list(&dispose); | |
1da177e4 LT |
533 | } |
534 | ||
1da177e4 | 535 | /** |
a0318786 CH |
536 | * invalidate_inodes - attempt to free all inodes on a superblock |
537 | * @sb: superblock to operate on | |
93b270f7 | 538 | * @kill_dirty: flag to guide handling of dirty inodes |
1da177e4 | 539 | * |
a0318786 CH |
540 | * Attempts to free all inodes for a given superblock. If there were any |
541 | * busy inodes return a non-zero value, else zero. | |
93b270f7 N |
542 | * If @kill_dirty is set, discard dirty inodes too, otherwise treat |
543 | * them as busy. | |
1da177e4 | 544 | */ |
93b270f7 | 545 | int invalidate_inodes(struct super_block *sb, bool kill_dirty) |
1da177e4 | 546 | { |
cffbc8aa | 547 | int busy = 0; |
a0318786 CH |
548 | struct inode *inode, *next; |
549 | LIST_HEAD(dispose); | |
1da177e4 | 550 | |
55fa6091 | 551 | spin_lock(&inode_sb_list_lock); |
a0318786 | 552 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
250df6ed DC |
553 | spin_lock(&inode->i_lock); |
554 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
555 | spin_unlock(&inode->i_lock); | |
aabb8fdb | 556 | continue; |
250df6ed | 557 | } |
93b270f7 | 558 | if (inode->i_state & I_DIRTY && !kill_dirty) { |
250df6ed | 559 | spin_unlock(&inode->i_lock); |
93b270f7 N |
560 | busy = 1; |
561 | continue; | |
562 | } | |
99a38919 | 563 | if (atomic_read(&inode->i_count)) { |
250df6ed | 564 | spin_unlock(&inode->i_lock); |
99a38919 | 565 | busy = 1; |
1da177e4 LT |
566 | continue; |
567 | } | |
99a38919 | 568 | |
99a38919 | 569 | inode->i_state |= I_FREEING; |
02afc410 | 570 | inode_lru_list_del(inode); |
250df6ed | 571 | spin_unlock(&inode->i_lock); |
02afc410 | 572 | list_add(&inode->i_lru, &dispose); |
1da177e4 | 573 | } |
55fa6091 | 574 | spin_unlock(&inode_sb_list_lock); |
1da177e4 | 575 | |
a0318786 | 576 | dispose_list(&dispose); |
1da177e4 LT |
577 | |
578 | return busy; | |
579 | } | |
1da177e4 LT |
580 | |
581 | static int can_unuse(struct inode *inode) | |
582 | { | |
9e38d86f | 583 | if (inode->i_state & ~I_REFERENCED) |
1da177e4 LT |
584 | return 0; |
585 | if (inode_has_buffers(inode)) | |
586 | return 0; | |
587 | if (atomic_read(&inode->i_count)) | |
588 | return 0; | |
589 | if (inode->i_data.nrpages) | |
590 | return 0; | |
591 | return 1; | |
592 | } | |
593 | ||
594 | /* | |
b0d40c92 DC |
595 | * Walk the superblock inode LRU for freeable inodes and attempt to free them. |
596 | * This is called from the superblock shrinker function with a number of inodes | |
597 | * to trim from the LRU. Inodes to be freed are moved to a temporary list and | |
598 | * then are freed outside inode_lock by dispose_list(). | |
1da177e4 LT |
599 | * |
600 | * Any inodes which are pinned purely because of attached pagecache have their | |
9e38d86f NP |
601 | * pagecache removed. If the inode has metadata buffers attached to |
602 | * mapping->private_list then try to remove them. | |
1da177e4 | 603 | * |
9e38d86f NP |
604 | * If the inode has the I_REFERENCED flag set, then it means that it has been |
605 | * used recently - the flag is set in iput_final(). When we encounter such an | |
606 | * inode, clear the flag and move it to the back of the LRU so it gets another | |
607 | * pass through the LRU before it gets reclaimed. This is necessary because of | |
608 | * the fact we are doing lazy LRU updates to minimise lock contention so the | |
609 | * LRU does not have strict ordering. Hence we don't want to reclaim inodes | |
610 | * with this flag set because they are the inodes that are out of order. | |
1da177e4 | 611 | */ |
b0d40c92 | 612 | void prune_icache_sb(struct super_block *sb, int nr_to_scan) |
1da177e4 LT |
613 | { |
614 | LIST_HEAD(freeable); | |
1da177e4 LT |
615 | int nr_scanned; |
616 | unsigned long reap = 0; | |
617 | ||
09cc9fc7 | 618 | spin_lock(&sb->s_inode_lru_lock); |
b0d40c92 | 619 | for (nr_scanned = nr_to_scan; nr_scanned >= 0; nr_scanned--) { |
1da177e4 LT |
620 | struct inode *inode; |
621 | ||
98b745c6 | 622 | if (list_empty(&sb->s_inode_lru)) |
1da177e4 LT |
623 | break; |
624 | ||
98b745c6 | 625 | inode = list_entry(sb->s_inode_lru.prev, struct inode, i_lru); |
1da177e4 | 626 | |
02afc410 | 627 | /* |
09cc9fc7 | 628 | * we are inverting the sb->s_inode_lru_lock/inode->i_lock here, |
02afc410 DC |
629 | * so use a trylock. If we fail to get the lock, just move the |
630 | * inode to the back of the list so we don't spin on it. | |
631 | */ | |
632 | if (!spin_trylock(&inode->i_lock)) { | |
98b745c6 | 633 | list_move(&inode->i_lru, &sb->s_inode_lru); |
02afc410 DC |
634 | continue; |
635 | } | |
636 | ||
9e38d86f NP |
637 | /* |
638 | * Referenced or dirty inodes are still in use. Give them | |
639 | * another pass through the LRU as we canot reclaim them now. | |
640 | */ | |
641 | if (atomic_read(&inode->i_count) || | |
642 | (inode->i_state & ~I_REFERENCED)) { | |
7ccf19a8 | 643 | list_del_init(&inode->i_lru); |
f283c86a | 644 | spin_unlock(&inode->i_lock); |
98b745c6 | 645 | sb->s_nr_inodes_unused--; |
fcb94f72 | 646 | this_cpu_dec(nr_unused); |
9e38d86f NP |
647 | continue; |
648 | } | |
649 | ||
650 | /* recently referenced inodes get one more pass */ | |
651 | if (inode->i_state & I_REFERENCED) { | |
9e38d86f | 652 | inode->i_state &= ~I_REFERENCED; |
98b745c6 | 653 | list_move(&inode->i_lru, &sb->s_inode_lru); |
f283c86a | 654 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
655 | continue; |
656 | } | |
657 | if (inode_has_buffers(inode) || inode->i_data.nrpages) { | |
658 | __iget(inode); | |
250df6ed | 659 | spin_unlock(&inode->i_lock); |
09cc9fc7 | 660 | spin_unlock(&sb->s_inode_lru_lock); |
1da177e4 | 661 | if (remove_inode_buffers(inode)) |
fc0ecff6 AM |
662 | reap += invalidate_mapping_pages(&inode->i_data, |
663 | 0, -1); | |
1da177e4 | 664 | iput(inode); |
09cc9fc7 | 665 | spin_lock(&sb->s_inode_lru_lock); |
1da177e4 | 666 | |
98b745c6 | 667 | if (inode != list_entry(sb->s_inode_lru.next, |
7ccf19a8 | 668 | struct inode, i_lru)) |
1da177e4 | 669 | continue; /* wrong inode or list_empty */ |
02afc410 DC |
670 | /* avoid lock inversions with trylock */ |
671 | if (!spin_trylock(&inode->i_lock)) | |
672 | continue; | |
250df6ed DC |
673 | if (!can_unuse(inode)) { |
674 | spin_unlock(&inode->i_lock); | |
1da177e4 | 675 | continue; |
250df6ed | 676 | } |
1da177e4 | 677 | } |
7ef0d737 | 678 | WARN_ON(inode->i_state & I_NEW); |
1da177e4 | 679 | inode->i_state |= I_FREEING; |
250df6ed | 680 | spin_unlock(&inode->i_lock); |
7ccf19a8 | 681 | |
7ccf19a8 | 682 | list_move(&inode->i_lru, &freeable); |
98b745c6 | 683 | sb->s_nr_inodes_unused--; |
fcb94f72 | 684 | this_cpu_dec(nr_unused); |
1da177e4 | 685 | } |
f8891e5e CL |
686 | if (current_is_kswapd()) |
687 | __count_vm_events(KSWAPD_INODESTEAL, reap); | |
688 | else | |
689 | __count_vm_events(PGINODESTEAL, reap); | |
09cc9fc7 | 690 | spin_unlock(&sb->s_inode_lru_lock); |
1da177e4 LT |
691 | |
692 | dispose_list(&freeable); | |
1da177e4 LT |
693 | } |
694 | ||
1da177e4 LT |
695 | static void __wait_on_freeing_inode(struct inode *inode); |
696 | /* | |
697 | * Called with the inode lock held. | |
1da177e4 | 698 | */ |
6b3304b5 MK |
699 | static struct inode *find_inode(struct super_block *sb, |
700 | struct hlist_head *head, | |
701 | int (*test)(struct inode *, void *), | |
702 | void *data) | |
1da177e4 LT |
703 | { |
704 | struct hlist_node *node; | |
6b3304b5 | 705 | struct inode *inode = NULL; |
1da177e4 LT |
706 | |
707 | repeat: | |
c5c8be3c | 708 | hlist_for_each_entry(inode, node, head, i_hash) { |
67a23c49 DC |
709 | spin_lock(&inode->i_lock); |
710 | if (inode->i_sb != sb) { | |
711 | spin_unlock(&inode->i_lock); | |
1da177e4 | 712 | continue; |
67a23c49 DC |
713 | } |
714 | if (!test(inode, data)) { | |
715 | spin_unlock(&inode->i_lock); | |
1da177e4 | 716 | continue; |
67a23c49 | 717 | } |
a4ffdde6 | 718 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
719 | __wait_on_freeing_inode(inode); |
720 | goto repeat; | |
721 | } | |
f7899bd5 | 722 | __iget(inode); |
250df6ed | 723 | spin_unlock(&inode->i_lock); |
f7899bd5 | 724 | return inode; |
1da177e4 | 725 | } |
f7899bd5 | 726 | return NULL; |
1da177e4 LT |
727 | } |
728 | ||
729 | /* | |
730 | * find_inode_fast is the fast path version of find_inode, see the comment at | |
731 | * iget_locked for details. | |
732 | */ | |
6b3304b5 MK |
733 | static struct inode *find_inode_fast(struct super_block *sb, |
734 | struct hlist_head *head, unsigned long ino) | |
1da177e4 LT |
735 | { |
736 | struct hlist_node *node; | |
6b3304b5 | 737 | struct inode *inode = NULL; |
1da177e4 LT |
738 | |
739 | repeat: | |
c5c8be3c | 740 | hlist_for_each_entry(inode, node, head, i_hash) { |
67a23c49 DC |
741 | spin_lock(&inode->i_lock); |
742 | if (inode->i_ino != ino) { | |
743 | spin_unlock(&inode->i_lock); | |
1da177e4 | 744 | continue; |
67a23c49 DC |
745 | } |
746 | if (inode->i_sb != sb) { | |
747 | spin_unlock(&inode->i_lock); | |
1da177e4 | 748 | continue; |
67a23c49 | 749 | } |
a4ffdde6 | 750 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
751 | __wait_on_freeing_inode(inode); |
752 | goto repeat; | |
753 | } | |
f7899bd5 | 754 | __iget(inode); |
250df6ed | 755 | spin_unlock(&inode->i_lock); |
f7899bd5 | 756 | return inode; |
1da177e4 | 757 | } |
f7899bd5 | 758 | return NULL; |
8290c35f DC |
759 | } |
760 | ||
f991bd2e ED |
761 | /* |
762 | * Each cpu owns a range of LAST_INO_BATCH numbers. | |
763 | * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations, | |
764 | * to renew the exhausted range. | |
8290c35f | 765 | * |
f991bd2e ED |
766 | * This does not significantly increase overflow rate because every CPU can |
767 | * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is | |
768 | * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the | |
769 | * 2^32 range, and is a worst-case. Even a 50% wastage would only increase | |
770 | * overflow rate by 2x, which does not seem too significant. | |
771 | * | |
772 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
773 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
774 | * here to attempt to avoid that. | |
8290c35f | 775 | */ |
f991bd2e ED |
776 | #define LAST_INO_BATCH 1024 |
777 | static DEFINE_PER_CPU(unsigned int, last_ino); | |
778 | ||
85fe4025 | 779 | unsigned int get_next_ino(void) |
8290c35f | 780 | { |
f991bd2e ED |
781 | unsigned int *p = &get_cpu_var(last_ino); |
782 | unsigned int res = *p; | |
8290c35f | 783 | |
f991bd2e ED |
784 | #ifdef CONFIG_SMP |
785 | if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) { | |
786 | static atomic_t shared_last_ino; | |
787 | int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino); | |
788 | ||
789 | res = next - LAST_INO_BATCH; | |
790 | } | |
791 | #endif | |
792 | ||
793 | *p = ++res; | |
794 | put_cpu_var(last_ino); | |
795 | return res; | |
8290c35f | 796 | } |
85fe4025 | 797 | EXPORT_SYMBOL(get_next_ino); |
8290c35f | 798 | |
1da177e4 LT |
799 | /** |
800 | * new_inode - obtain an inode | |
801 | * @sb: superblock | |
802 | * | |
769848c0 | 803 | * Allocates a new inode for given superblock. The default gfp_mask |
3c1d4378 | 804 | * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE. |
769848c0 MG |
805 | * If HIGHMEM pages are unsuitable or it is known that pages allocated |
806 | * for the page cache are not reclaimable or migratable, | |
807 | * mapping_set_gfp_mask() must be called with suitable flags on the | |
808 | * newly created inode's mapping | |
809 | * | |
1da177e4 LT |
810 | */ |
811 | struct inode *new_inode(struct super_block *sb) | |
812 | { | |
6b3304b5 | 813 | struct inode *inode; |
1da177e4 | 814 | |
55fa6091 | 815 | spin_lock_prefetch(&inode_sb_list_lock); |
6b3304b5 | 816 | |
1da177e4 LT |
817 | inode = alloc_inode(sb); |
818 | if (inode) { | |
250df6ed | 819 | spin_lock(&inode->i_lock); |
1da177e4 | 820 | inode->i_state = 0; |
250df6ed | 821 | spin_unlock(&inode->i_lock); |
55fa6091 | 822 | inode_sb_list_add(inode); |
1da177e4 LT |
823 | } |
824 | return inode; | |
825 | } | |
1da177e4 LT |
826 | EXPORT_SYMBOL(new_inode); |
827 | ||
250df6ed DC |
828 | /** |
829 | * unlock_new_inode - clear the I_NEW state and wake up any waiters | |
830 | * @inode: new inode to unlock | |
831 | * | |
832 | * Called when the inode is fully initialised to clear the new state of the | |
833 | * inode and wake up anyone waiting for the inode to finish initialisation. | |
834 | */ | |
1da177e4 LT |
835 | void unlock_new_inode(struct inode *inode) |
836 | { | |
14358e6d | 837 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
a3314a0e | 838 | if (S_ISDIR(inode->i_mode)) { |
1e89a5e1 PZ |
839 | struct file_system_type *type = inode->i_sb->s_type; |
840 | ||
9a7aa12f JK |
841 | /* Set new key only if filesystem hasn't already changed it */ |
842 | if (!lockdep_match_class(&inode->i_mutex, | |
843 | &type->i_mutex_key)) { | |
844 | /* | |
845 | * ensure nobody is actually holding i_mutex | |
846 | */ | |
847 | mutex_destroy(&inode->i_mutex); | |
848 | mutex_init(&inode->i_mutex); | |
849 | lockdep_set_class(&inode->i_mutex, | |
850 | &type->i_mutex_dir_key); | |
851 | } | |
1e89a5e1 | 852 | } |
14358e6d | 853 | #endif |
250df6ed | 854 | spin_lock(&inode->i_lock); |
eaff8079 CH |
855 | WARN_ON(!(inode->i_state & I_NEW)); |
856 | inode->i_state &= ~I_NEW; | |
250df6ed DC |
857 | wake_up_bit(&inode->i_state, __I_NEW); |
858 | spin_unlock(&inode->i_lock); | |
1da177e4 | 859 | } |
1da177e4 LT |
860 | EXPORT_SYMBOL(unlock_new_inode); |
861 | ||
0b2d0724 CH |
862 | /** |
863 | * iget5_locked - obtain an inode from a mounted file system | |
864 | * @sb: super block of file system | |
865 | * @hashval: hash value (usually inode number) to get | |
866 | * @test: callback used for comparisons between inodes | |
867 | * @set: callback used to initialize a new struct inode | |
868 | * @data: opaque data pointer to pass to @test and @set | |
869 | * | |
870 | * Search for the inode specified by @hashval and @data in the inode cache, | |
871 | * and if present it is return it with an increased reference count. This is | |
872 | * a generalized version of iget_locked() for file systems where the inode | |
873 | * number is not sufficient for unique identification of an inode. | |
874 | * | |
875 | * If the inode is not in cache, allocate a new inode and return it locked, | |
876 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
877 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 878 | * |
0b2d0724 CH |
879 | * Note both @test and @set are called with the inode_hash_lock held, so can't |
880 | * sleep. | |
1da177e4 | 881 | */ |
0b2d0724 CH |
882 | struct inode *iget5_locked(struct super_block *sb, unsigned long hashval, |
883 | int (*test)(struct inode *, void *), | |
884 | int (*set)(struct inode *, void *), void *data) | |
1da177e4 | 885 | { |
0b2d0724 | 886 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); |
6b3304b5 | 887 | struct inode *inode; |
1da177e4 | 888 | |
0b2d0724 CH |
889 | spin_lock(&inode_hash_lock); |
890 | inode = find_inode(sb, head, test, data); | |
891 | spin_unlock(&inode_hash_lock); | |
892 | ||
893 | if (inode) { | |
894 | wait_on_inode(inode); | |
895 | return inode; | |
896 | } | |
897 | ||
1da177e4 LT |
898 | inode = alloc_inode(sb); |
899 | if (inode) { | |
6b3304b5 | 900 | struct inode *old; |
1da177e4 | 901 | |
67a23c49 | 902 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
903 | /* We released the lock, so.. */ |
904 | old = find_inode(sb, head, test, data); | |
905 | if (!old) { | |
906 | if (set(inode, data)) | |
907 | goto set_failed; | |
908 | ||
250df6ed DC |
909 | spin_lock(&inode->i_lock); |
910 | inode->i_state = I_NEW; | |
646ec461 | 911 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 912 | spin_unlock(&inode->i_lock); |
55fa6091 | 913 | inode_sb_list_add(inode); |
67a23c49 | 914 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
915 | |
916 | /* Return the locked inode with I_NEW set, the | |
917 | * caller is responsible for filling in the contents | |
918 | */ | |
919 | return inode; | |
920 | } | |
921 | ||
922 | /* | |
923 | * Uhhuh, somebody else created the same inode under | |
924 | * us. Use the old inode instead of the one we just | |
925 | * allocated. | |
926 | */ | |
67a23c49 | 927 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
928 | destroy_inode(inode); |
929 | inode = old; | |
930 | wait_on_inode(inode); | |
931 | } | |
932 | return inode; | |
933 | ||
934 | set_failed: | |
67a23c49 | 935 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
936 | destroy_inode(inode); |
937 | return NULL; | |
938 | } | |
0b2d0724 | 939 | EXPORT_SYMBOL(iget5_locked); |
1da177e4 | 940 | |
0b2d0724 CH |
941 | /** |
942 | * iget_locked - obtain an inode from a mounted file system | |
943 | * @sb: super block of file system | |
944 | * @ino: inode number to get | |
945 | * | |
946 | * Search for the inode specified by @ino in the inode cache and if present | |
947 | * return it with an increased reference count. This is for file systems | |
948 | * where the inode number is sufficient for unique identification of an inode. | |
949 | * | |
950 | * If the inode is not in cache, allocate a new inode and return it locked, | |
951 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
952 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 953 | */ |
0b2d0724 | 954 | struct inode *iget_locked(struct super_block *sb, unsigned long ino) |
1da177e4 | 955 | { |
0b2d0724 | 956 | struct hlist_head *head = inode_hashtable + hash(sb, ino); |
6b3304b5 | 957 | struct inode *inode; |
1da177e4 | 958 | |
0b2d0724 CH |
959 | spin_lock(&inode_hash_lock); |
960 | inode = find_inode_fast(sb, head, ino); | |
961 | spin_unlock(&inode_hash_lock); | |
962 | if (inode) { | |
963 | wait_on_inode(inode); | |
964 | return inode; | |
965 | } | |
966 | ||
1da177e4 LT |
967 | inode = alloc_inode(sb); |
968 | if (inode) { | |
6b3304b5 | 969 | struct inode *old; |
1da177e4 | 970 | |
67a23c49 | 971 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
972 | /* We released the lock, so.. */ |
973 | old = find_inode_fast(sb, head, ino); | |
974 | if (!old) { | |
975 | inode->i_ino = ino; | |
250df6ed DC |
976 | spin_lock(&inode->i_lock); |
977 | inode->i_state = I_NEW; | |
646ec461 | 978 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 979 | spin_unlock(&inode->i_lock); |
55fa6091 | 980 | inode_sb_list_add(inode); |
67a23c49 | 981 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
982 | |
983 | /* Return the locked inode with I_NEW set, the | |
984 | * caller is responsible for filling in the contents | |
985 | */ | |
986 | return inode; | |
987 | } | |
988 | ||
989 | /* | |
990 | * Uhhuh, somebody else created the same inode under | |
991 | * us. Use the old inode instead of the one we just | |
992 | * allocated. | |
993 | */ | |
67a23c49 | 994 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
995 | destroy_inode(inode); |
996 | inode = old; | |
997 | wait_on_inode(inode); | |
998 | } | |
999 | return inode; | |
1000 | } | |
0b2d0724 | 1001 | EXPORT_SYMBOL(iget_locked); |
1da177e4 | 1002 | |
ad5e195a CH |
1003 | /* |
1004 | * search the inode cache for a matching inode number. | |
1005 | * If we find one, then the inode number we are trying to | |
1006 | * allocate is not unique and so we should not use it. | |
1007 | * | |
1008 | * Returns 1 if the inode number is unique, 0 if it is not. | |
1009 | */ | |
1010 | static int test_inode_iunique(struct super_block *sb, unsigned long ino) | |
1011 | { | |
1012 | struct hlist_head *b = inode_hashtable + hash(sb, ino); | |
1013 | struct hlist_node *node; | |
1014 | struct inode *inode; | |
1015 | ||
67a23c49 | 1016 | spin_lock(&inode_hash_lock); |
ad5e195a | 1017 | hlist_for_each_entry(inode, node, b, i_hash) { |
67a23c49 DC |
1018 | if (inode->i_ino == ino && inode->i_sb == sb) { |
1019 | spin_unlock(&inode_hash_lock); | |
ad5e195a | 1020 | return 0; |
67a23c49 | 1021 | } |
ad5e195a | 1022 | } |
67a23c49 | 1023 | spin_unlock(&inode_hash_lock); |
ad5e195a CH |
1024 | |
1025 | return 1; | |
1026 | } | |
1027 | ||
1da177e4 LT |
1028 | /** |
1029 | * iunique - get a unique inode number | |
1030 | * @sb: superblock | |
1031 | * @max_reserved: highest reserved inode number | |
1032 | * | |
1033 | * Obtain an inode number that is unique on the system for a given | |
1034 | * superblock. This is used by file systems that have no natural | |
1035 | * permanent inode numbering system. An inode number is returned that | |
1036 | * is higher than the reserved limit but unique. | |
1037 | * | |
1038 | * BUGS: | |
1039 | * With a large number of inodes live on the file system this function | |
1040 | * currently becomes quite slow. | |
1041 | */ | |
1042 | ino_t iunique(struct super_block *sb, ino_t max_reserved) | |
1043 | { | |
866b04fc JL |
1044 | /* |
1045 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
1046 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
1047 | * here to attempt to avoid that. | |
1048 | */ | |
ad5e195a | 1049 | static DEFINE_SPINLOCK(iunique_lock); |
866b04fc | 1050 | static unsigned int counter; |
1da177e4 | 1051 | ino_t res; |
3361c7be | 1052 | |
ad5e195a | 1053 | spin_lock(&iunique_lock); |
3361c7be JL |
1054 | do { |
1055 | if (counter <= max_reserved) | |
1056 | counter = max_reserved + 1; | |
1da177e4 | 1057 | res = counter++; |
ad5e195a CH |
1058 | } while (!test_inode_iunique(sb, res)); |
1059 | spin_unlock(&iunique_lock); | |
1da177e4 | 1060 | |
3361c7be JL |
1061 | return res; |
1062 | } | |
1da177e4 LT |
1063 | EXPORT_SYMBOL(iunique); |
1064 | ||
1065 | struct inode *igrab(struct inode *inode) | |
1066 | { | |
250df6ed DC |
1067 | spin_lock(&inode->i_lock); |
1068 | if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) { | |
1da177e4 | 1069 | __iget(inode); |
250df6ed DC |
1070 | spin_unlock(&inode->i_lock); |
1071 | } else { | |
1072 | spin_unlock(&inode->i_lock); | |
1da177e4 LT |
1073 | /* |
1074 | * Handle the case where s_op->clear_inode is not been | |
1075 | * called yet, and somebody is calling igrab | |
1076 | * while the inode is getting freed. | |
1077 | */ | |
1078 | inode = NULL; | |
250df6ed | 1079 | } |
1da177e4 LT |
1080 | return inode; |
1081 | } | |
1da177e4 LT |
1082 | EXPORT_SYMBOL(igrab); |
1083 | ||
1084 | /** | |
0b2d0724 | 1085 | * ilookup5_nowait - search for an inode in the inode cache |
1da177e4 | 1086 | * @sb: super block of file system to search |
0b2d0724 | 1087 | * @hashval: hash value (usually inode number) to search for |
1da177e4 LT |
1088 | * @test: callback used for comparisons between inodes |
1089 | * @data: opaque data pointer to pass to @test | |
1da177e4 | 1090 | * |
0b2d0724 | 1091 | * Search for the inode specified by @hashval and @data in the inode cache. |
1da177e4 LT |
1092 | * If the inode is in the cache, the inode is returned with an incremented |
1093 | * reference count. | |
1094 | * | |
0b2d0724 CH |
1095 | * Note: I_NEW is not waited upon so you have to be very careful what you do |
1096 | * with the returned inode. You probably should be using ilookup5() instead. | |
1da177e4 | 1097 | * |
b6d0ad68 | 1098 | * Note2: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 | 1099 | */ |
0b2d0724 CH |
1100 | struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval, |
1101 | int (*test)(struct inode *, void *), void *data) | |
1da177e4 | 1102 | { |
0b2d0724 | 1103 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); |
1da177e4 LT |
1104 | struct inode *inode; |
1105 | ||
67a23c49 | 1106 | spin_lock(&inode_hash_lock); |
1da177e4 | 1107 | inode = find_inode(sb, head, test, data); |
67a23c49 | 1108 | spin_unlock(&inode_hash_lock); |
88bd5121 | 1109 | |
0b2d0724 | 1110 | return inode; |
88bd5121 | 1111 | } |
88bd5121 AA |
1112 | EXPORT_SYMBOL(ilookup5_nowait); |
1113 | ||
1114 | /** | |
1115 | * ilookup5 - search for an inode in the inode cache | |
1116 | * @sb: super block of file system to search | |
1117 | * @hashval: hash value (usually inode number) to search for | |
1118 | * @test: callback used for comparisons between inodes | |
1119 | * @data: opaque data pointer to pass to @test | |
1120 | * | |
0b2d0724 CH |
1121 | * Search for the inode specified by @hashval and @data in the inode cache, |
1122 | * and if the inode is in the cache, return the inode with an incremented | |
1123 | * reference count. Waits on I_NEW before returning the inode. | |
88bd5121 | 1124 | * returned with an incremented reference count. |
1da177e4 | 1125 | * |
0b2d0724 CH |
1126 | * This is a generalized version of ilookup() for file systems where the |
1127 | * inode number is not sufficient for unique identification of an inode. | |
1da177e4 | 1128 | * |
0b2d0724 | 1129 | * Note: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 LT |
1130 | */ |
1131 | struct inode *ilookup5(struct super_block *sb, unsigned long hashval, | |
1132 | int (*test)(struct inode *, void *), void *data) | |
1133 | { | |
0b2d0724 | 1134 | struct inode *inode = ilookup5_nowait(sb, hashval, test, data); |
1da177e4 | 1135 | |
0b2d0724 CH |
1136 | if (inode) |
1137 | wait_on_inode(inode); | |
1138 | return inode; | |
1da177e4 | 1139 | } |
1da177e4 LT |
1140 | EXPORT_SYMBOL(ilookup5); |
1141 | ||
1142 | /** | |
1143 | * ilookup - search for an inode in the inode cache | |
1144 | * @sb: super block of file system to search | |
1145 | * @ino: inode number to search for | |
1146 | * | |
0b2d0724 CH |
1147 | * Search for the inode @ino in the inode cache, and if the inode is in the |
1148 | * cache, the inode is returned with an incremented reference count. | |
1da177e4 LT |
1149 | */ |
1150 | struct inode *ilookup(struct super_block *sb, unsigned long ino) | |
1151 | { | |
1152 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1da177e4 LT |
1153 | struct inode *inode; |
1154 | ||
0b2d0724 CH |
1155 | spin_lock(&inode_hash_lock); |
1156 | inode = find_inode_fast(sb, head, ino); | |
1157 | spin_unlock(&inode_hash_lock); | |
1da177e4 | 1158 | |
1da177e4 | 1159 | if (inode) |
0b2d0724 CH |
1160 | wait_on_inode(inode); |
1161 | return inode; | |
1da177e4 | 1162 | } |
0b2d0724 | 1163 | EXPORT_SYMBOL(ilookup); |
1da177e4 | 1164 | |
261bca86 AV |
1165 | int insert_inode_locked(struct inode *inode) |
1166 | { | |
1167 | struct super_block *sb = inode->i_sb; | |
1168 | ino_t ino = inode->i_ino; | |
1169 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
261bca86 | 1170 | |
261bca86 | 1171 | while (1) { |
72a43d63 AV |
1172 | struct hlist_node *node; |
1173 | struct inode *old = NULL; | |
67a23c49 | 1174 | spin_lock(&inode_hash_lock); |
72a43d63 AV |
1175 | hlist_for_each_entry(old, node, head, i_hash) { |
1176 | if (old->i_ino != ino) | |
1177 | continue; | |
1178 | if (old->i_sb != sb) | |
1179 | continue; | |
250df6ed DC |
1180 | spin_lock(&old->i_lock); |
1181 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1182 | spin_unlock(&old->i_lock); | |
72a43d63 | 1183 | continue; |
250df6ed | 1184 | } |
72a43d63 AV |
1185 | break; |
1186 | } | |
1187 | if (likely(!node)) { | |
250df6ed DC |
1188 | spin_lock(&inode->i_lock); |
1189 | inode->i_state |= I_NEW; | |
261bca86 | 1190 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1191 | spin_unlock(&inode->i_lock); |
67a23c49 | 1192 | spin_unlock(&inode_hash_lock); |
261bca86 AV |
1193 | return 0; |
1194 | } | |
1195 | __iget(old); | |
250df6ed | 1196 | spin_unlock(&old->i_lock); |
67a23c49 | 1197 | spin_unlock(&inode_hash_lock); |
261bca86 | 1198 | wait_on_inode(old); |
1d3382cb | 1199 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1200 | iput(old); |
1201 | return -EBUSY; | |
1202 | } | |
1203 | iput(old); | |
1204 | } | |
1205 | } | |
261bca86 AV |
1206 | EXPORT_SYMBOL(insert_inode_locked); |
1207 | ||
1208 | int insert_inode_locked4(struct inode *inode, unsigned long hashval, | |
1209 | int (*test)(struct inode *, void *), void *data) | |
1210 | { | |
1211 | struct super_block *sb = inode->i_sb; | |
1212 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
261bca86 | 1213 | |
261bca86 | 1214 | while (1) { |
72a43d63 AV |
1215 | struct hlist_node *node; |
1216 | struct inode *old = NULL; | |
1217 | ||
67a23c49 | 1218 | spin_lock(&inode_hash_lock); |
72a43d63 AV |
1219 | hlist_for_each_entry(old, node, head, i_hash) { |
1220 | if (old->i_sb != sb) | |
1221 | continue; | |
1222 | if (!test(old, data)) | |
1223 | continue; | |
250df6ed DC |
1224 | spin_lock(&old->i_lock); |
1225 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1226 | spin_unlock(&old->i_lock); | |
72a43d63 | 1227 | continue; |
250df6ed | 1228 | } |
72a43d63 AV |
1229 | break; |
1230 | } | |
1231 | if (likely(!node)) { | |
250df6ed DC |
1232 | spin_lock(&inode->i_lock); |
1233 | inode->i_state |= I_NEW; | |
261bca86 | 1234 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1235 | spin_unlock(&inode->i_lock); |
67a23c49 | 1236 | spin_unlock(&inode_hash_lock); |
261bca86 AV |
1237 | return 0; |
1238 | } | |
1239 | __iget(old); | |
250df6ed | 1240 | spin_unlock(&old->i_lock); |
67a23c49 | 1241 | spin_unlock(&inode_hash_lock); |
261bca86 | 1242 | wait_on_inode(old); |
1d3382cb | 1243 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1244 | iput(old); |
1245 | return -EBUSY; | |
1246 | } | |
1247 | iput(old); | |
1248 | } | |
1249 | } | |
261bca86 AV |
1250 | EXPORT_SYMBOL(insert_inode_locked4); |
1251 | ||
1da177e4 | 1252 | |
45321ac5 AV |
1253 | int generic_delete_inode(struct inode *inode) |
1254 | { | |
1255 | return 1; | |
1256 | } | |
1257 | EXPORT_SYMBOL(generic_delete_inode); | |
1258 | ||
1da177e4 | 1259 | /* |
45321ac5 AV |
1260 | * Normal UNIX filesystem behaviour: delete the |
1261 | * inode when the usage count drops to zero, and | |
1262 | * i_nlink is zero. | |
1da177e4 | 1263 | */ |
45321ac5 | 1264 | int generic_drop_inode(struct inode *inode) |
1da177e4 | 1265 | { |
1d3382cb | 1266 | return !inode->i_nlink || inode_unhashed(inode); |
1da177e4 | 1267 | } |
45321ac5 | 1268 | EXPORT_SYMBOL_GPL(generic_drop_inode); |
1da177e4 | 1269 | |
45321ac5 AV |
1270 | /* |
1271 | * Called when we're dropping the last reference | |
1272 | * to an inode. | |
22fe4042 | 1273 | * |
45321ac5 AV |
1274 | * Call the FS "drop_inode()" function, defaulting to |
1275 | * the legacy UNIX filesystem behaviour. If it tells | |
1276 | * us to evict inode, do so. Otherwise, retain inode | |
1277 | * in cache if fs is alive, sync and evict if fs is | |
1278 | * shutting down. | |
22fe4042 | 1279 | */ |
45321ac5 | 1280 | static void iput_final(struct inode *inode) |
1da177e4 LT |
1281 | { |
1282 | struct super_block *sb = inode->i_sb; | |
45321ac5 AV |
1283 | const struct super_operations *op = inode->i_sb->s_op; |
1284 | int drop; | |
1285 | ||
250df6ed DC |
1286 | WARN_ON(inode->i_state & I_NEW); |
1287 | ||
e7f59097 | 1288 | if (op->drop_inode) |
45321ac5 AV |
1289 | drop = op->drop_inode(inode); |
1290 | else | |
1291 | drop = generic_drop_inode(inode); | |
1da177e4 | 1292 | |
b2b2af8e DC |
1293 | if (!drop && (sb->s_flags & MS_ACTIVE)) { |
1294 | inode->i_state |= I_REFERENCED; | |
1295 | if (!(inode->i_state & (I_DIRTY|I_SYNC))) | |
1296 | inode_lru_list_add(inode); | |
1297 | spin_unlock(&inode->i_lock); | |
b2b2af8e DC |
1298 | return; |
1299 | } | |
1300 | ||
45321ac5 | 1301 | if (!drop) { |
991114c6 | 1302 | inode->i_state |= I_WILL_FREE; |
250df6ed | 1303 | spin_unlock(&inode->i_lock); |
1da177e4 | 1304 | write_inode_now(inode, 1); |
250df6ed | 1305 | spin_lock(&inode->i_lock); |
7ef0d737 | 1306 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1307 | inode->i_state &= ~I_WILL_FREE; |
1da177e4 | 1308 | } |
7ccf19a8 | 1309 | |
991114c6 | 1310 | inode->i_state |= I_FREEING; |
9e38d86f | 1311 | inode_lru_list_del(inode); |
b2b2af8e | 1312 | spin_unlock(&inode->i_lock); |
b2b2af8e | 1313 | |
644da596 | 1314 | evict(inode); |
1da177e4 LT |
1315 | } |
1316 | ||
1da177e4 | 1317 | /** |
6b3304b5 | 1318 | * iput - put an inode |
1da177e4 LT |
1319 | * @inode: inode to put |
1320 | * | |
1321 | * Puts an inode, dropping its usage count. If the inode use count hits | |
1322 | * zero, the inode is then freed and may also be destroyed. | |
1323 | * | |
1324 | * Consequently, iput() can sleep. | |
1325 | */ | |
1326 | void iput(struct inode *inode) | |
1327 | { | |
1328 | if (inode) { | |
a4ffdde6 | 1329 | BUG_ON(inode->i_state & I_CLEAR); |
1da177e4 | 1330 | |
f283c86a | 1331 | if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock)) |
1da177e4 LT |
1332 | iput_final(inode); |
1333 | } | |
1334 | } | |
1da177e4 LT |
1335 | EXPORT_SYMBOL(iput); |
1336 | ||
1337 | /** | |
1338 | * bmap - find a block number in a file | |
1339 | * @inode: inode of file | |
1340 | * @block: block to find | |
1341 | * | |
1342 | * Returns the block number on the device holding the inode that | |
1343 | * is the disk block number for the block of the file requested. | |
1344 | * That is, asked for block 4 of inode 1 the function will return the | |
6b3304b5 | 1345 | * disk block relative to the disk start that holds that block of the |
1da177e4 LT |
1346 | * file. |
1347 | */ | |
6b3304b5 | 1348 | sector_t bmap(struct inode *inode, sector_t block) |
1da177e4 LT |
1349 | { |
1350 | sector_t res = 0; | |
1351 | if (inode->i_mapping->a_ops->bmap) | |
1352 | res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block); | |
1353 | return res; | |
1354 | } | |
1da177e4 LT |
1355 | EXPORT_SYMBOL(bmap); |
1356 | ||
11ff6f05 MG |
1357 | /* |
1358 | * With relative atime, only update atime if the previous atime is | |
1359 | * earlier than either the ctime or mtime or if at least a day has | |
1360 | * passed since the last atime update. | |
1361 | */ | |
1362 | static int relatime_need_update(struct vfsmount *mnt, struct inode *inode, | |
1363 | struct timespec now) | |
1364 | { | |
1365 | ||
1366 | if (!(mnt->mnt_flags & MNT_RELATIME)) | |
1367 | return 1; | |
1368 | /* | |
1369 | * Is mtime younger than atime? If yes, update atime: | |
1370 | */ | |
1371 | if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0) | |
1372 | return 1; | |
1373 | /* | |
1374 | * Is ctime younger than atime? If yes, update atime: | |
1375 | */ | |
1376 | if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0) | |
1377 | return 1; | |
1378 | ||
1379 | /* | |
1380 | * Is the previous atime value older than a day? If yes, | |
1381 | * update atime: | |
1382 | */ | |
1383 | if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60) | |
1384 | return 1; | |
1385 | /* | |
1386 | * Good, we can skip the atime update: | |
1387 | */ | |
1388 | return 0; | |
1389 | } | |
1390 | ||
1da177e4 | 1391 | /** |
869243a0 CH |
1392 | * touch_atime - update the access time |
1393 | * @mnt: mount the inode is accessed on | |
7045f37b | 1394 | * @dentry: dentry accessed |
1da177e4 LT |
1395 | * |
1396 | * Update the accessed time on an inode and mark it for writeback. | |
1397 | * This function automatically handles read only file systems and media, | |
1398 | * as well as the "noatime" flag and inode specific "noatime" markers. | |
1399 | */ | |
869243a0 | 1400 | void touch_atime(struct vfsmount *mnt, struct dentry *dentry) |
1da177e4 | 1401 | { |
869243a0 | 1402 | struct inode *inode = dentry->d_inode; |
1da177e4 LT |
1403 | struct timespec now; |
1404 | ||
cdb70f3f | 1405 | if (inode->i_flags & S_NOATIME) |
b12536c2 | 1406 | return; |
37756ced | 1407 | if (IS_NOATIME(inode)) |
b12536c2 | 1408 | return; |
b2276138 | 1409 | if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1410 | return; |
47ae32d6 | 1411 | |
cdb70f3f | 1412 | if (mnt->mnt_flags & MNT_NOATIME) |
b12536c2 | 1413 | return; |
cdb70f3f | 1414 | if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1415 | return; |
1da177e4 LT |
1416 | |
1417 | now = current_fs_time(inode->i_sb); | |
11ff6f05 MG |
1418 | |
1419 | if (!relatime_need_update(mnt, inode, now)) | |
b12536c2 | 1420 | return; |
11ff6f05 | 1421 | |
47ae32d6 | 1422 | if (timespec_equal(&inode->i_atime, &now)) |
b12536c2 AK |
1423 | return; |
1424 | ||
1425 | if (mnt_want_write(mnt)) | |
1426 | return; | |
47ae32d6 VH |
1427 | |
1428 | inode->i_atime = now; | |
1429 | mark_inode_dirty_sync(inode); | |
cdb70f3f | 1430 | mnt_drop_write(mnt); |
1da177e4 | 1431 | } |
869243a0 | 1432 | EXPORT_SYMBOL(touch_atime); |
1da177e4 LT |
1433 | |
1434 | /** | |
870f4817 CH |
1435 | * file_update_time - update mtime and ctime time |
1436 | * @file: file accessed | |
1da177e4 | 1437 | * |
870f4817 CH |
1438 | * Update the mtime and ctime members of an inode and mark the inode |
1439 | * for writeback. Note that this function is meant exclusively for | |
1440 | * usage in the file write path of filesystems, and filesystems may | |
1441 | * choose to explicitly ignore update via this function with the | |
2eadfc0e | 1442 | * S_NOCMTIME inode flag, e.g. for network filesystem where these |
870f4817 | 1443 | * timestamps are handled by the server. |
1da177e4 LT |
1444 | */ |
1445 | ||
870f4817 | 1446 | void file_update_time(struct file *file) |
1da177e4 | 1447 | { |
0f7fc9e4 | 1448 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 | 1449 | struct timespec now; |
ce06e0b2 | 1450 | enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0; |
1da177e4 | 1451 | |
ce06e0b2 | 1452 | /* First try to exhaust all avenues to not sync */ |
1da177e4 LT |
1453 | if (IS_NOCMTIME(inode)) |
1454 | return; | |
20ddee2c | 1455 | |
1da177e4 | 1456 | now = current_fs_time(inode->i_sb); |
ce06e0b2 AK |
1457 | if (!timespec_equal(&inode->i_mtime, &now)) |
1458 | sync_it = S_MTIME; | |
1da177e4 | 1459 | |
ce06e0b2 AK |
1460 | if (!timespec_equal(&inode->i_ctime, &now)) |
1461 | sync_it |= S_CTIME; | |
870f4817 | 1462 | |
ce06e0b2 AK |
1463 | if (IS_I_VERSION(inode)) |
1464 | sync_it |= S_VERSION; | |
7a224228 | 1465 | |
ce06e0b2 AK |
1466 | if (!sync_it) |
1467 | return; | |
1468 | ||
1469 | /* Finally allowed to write? Takes lock. */ | |
1470 | if (mnt_want_write_file(file)) | |
1471 | return; | |
1472 | ||
1473 | /* Only change inode inside the lock region */ | |
1474 | if (sync_it & S_VERSION) | |
1475 | inode_inc_iversion(inode); | |
1476 | if (sync_it & S_CTIME) | |
1477 | inode->i_ctime = now; | |
1478 | if (sync_it & S_MTIME) | |
1479 | inode->i_mtime = now; | |
1480 | mark_inode_dirty_sync(inode); | |
20ddee2c | 1481 | mnt_drop_write(file->f_path.mnt); |
1da177e4 | 1482 | } |
870f4817 | 1483 | EXPORT_SYMBOL(file_update_time); |
1da177e4 LT |
1484 | |
1485 | int inode_needs_sync(struct inode *inode) | |
1486 | { | |
1487 | if (IS_SYNC(inode)) | |
1488 | return 1; | |
1489 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) | |
1490 | return 1; | |
1491 | return 0; | |
1492 | } | |
1da177e4 LT |
1493 | EXPORT_SYMBOL(inode_needs_sync); |
1494 | ||
1da177e4 LT |
1495 | int inode_wait(void *word) |
1496 | { | |
1497 | schedule(); | |
1498 | return 0; | |
1499 | } | |
d44dab8d | 1500 | EXPORT_SYMBOL(inode_wait); |
1da177e4 LT |
1501 | |
1502 | /* | |
168a9fd6 MS |
1503 | * If we try to find an inode in the inode hash while it is being |
1504 | * deleted, we have to wait until the filesystem completes its | |
1505 | * deletion before reporting that it isn't found. This function waits | |
1506 | * until the deletion _might_ have completed. Callers are responsible | |
1507 | * to recheck inode state. | |
1508 | * | |
eaff8079 | 1509 | * It doesn't matter if I_NEW is not set initially, a call to |
250df6ed DC |
1510 | * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list |
1511 | * will DTRT. | |
1da177e4 LT |
1512 | */ |
1513 | static void __wait_on_freeing_inode(struct inode *inode) | |
1514 | { | |
1515 | wait_queue_head_t *wq; | |
eaff8079 CH |
1516 | DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW); |
1517 | wq = bit_waitqueue(&inode->i_state, __I_NEW); | |
1da177e4 | 1518 | prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE); |
250df6ed | 1519 | spin_unlock(&inode->i_lock); |
67a23c49 | 1520 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1521 | schedule(); |
1522 | finish_wait(wq, &wait.wait); | |
67a23c49 | 1523 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1524 | } |
1525 | ||
1da177e4 LT |
1526 | static __initdata unsigned long ihash_entries; |
1527 | static int __init set_ihash_entries(char *str) | |
1528 | { | |
1529 | if (!str) | |
1530 | return 0; | |
1531 | ihash_entries = simple_strtoul(str, &str, 0); | |
1532 | return 1; | |
1533 | } | |
1534 | __setup("ihash_entries=", set_ihash_entries); | |
1535 | ||
1536 | /* | |
1537 | * Initialize the waitqueues and inode hash table. | |
1538 | */ | |
1539 | void __init inode_init_early(void) | |
1540 | { | |
1541 | int loop; | |
1542 | ||
1543 | /* If hashes are distributed across NUMA nodes, defer | |
1544 | * hash allocation until vmalloc space is available. | |
1545 | */ | |
1546 | if (hashdist) | |
1547 | return; | |
1548 | ||
1549 | inode_hashtable = | |
1550 | alloc_large_system_hash("Inode-cache", | |
1551 | sizeof(struct hlist_head), | |
1552 | ihash_entries, | |
1553 | 14, | |
1554 | HASH_EARLY, | |
1555 | &i_hash_shift, | |
1556 | &i_hash_mask, | |
1557 | 0); | |
1558 | ||
1559 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1560 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1561 | } | |
1562 | ||
74bf17cf | 1563 | void __init inode_init(void) |
1da177e4 LT |
1564 | { |
1565 | int loop; | |
1566 | ||
1567 | /* inode slab cache */ | |
b0196009 PJ |
1568 | inode_cachep = kmem_cache_create("inode_cache", |
1569 | sizeof(struct inode), | |
1570 | 0, | |
1571 | (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC| | |
1572 | SLAB_MEM_SPREAD), | |
20c2df83 | 1573 | init_once); |
1da177e4 LT |
1574 | |
1575 | /* Hash may have been set up in inode_init_early */ | |
1576 | if (!hashdist) | |
1577 | return; | |
1578 | ||
1579 | inode_hashtable = | |
1580 | alloc_large_system_hash("Inode-cache", | |
1581 | sizeof(struct hlist_head), | |
1582 | ihash_entries, | |
1583 | 14, | |
1584 | 0, | |
1585 | &i_hash_shift, | |
1586 | &i_hash_mask, | |
1587 | 0); | |
1588 | ||
1589 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1590 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1591 | } | |
1592 | ||
1593 | void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev) | |
1594 | { | |
1595 | inode->i_mode = mode; | |
1596 | if (S_ISCHR(mode)) { | |
1597 | inode->i_fop = &def_chr_fops; | |
1598 | inode->i_rdev = rdev; | |
1599 | } else if (S_ISBLK(mode)) { | |
1600 | inode->i_fop = &def_blk_fops; | |
1601 | inode->i_rdev = rdev; | |
1602 | } else if (S_ISFIFO(mode)) | |
1603 | inode->i_fop = &def_fifo_fops; | |
1604 | else if (S_ISSOCK(mode)) | |
1605 | inode->i_fop = &bad_sock_fops; | |
1606 | else | |
af0d9ae8 MK |
1607 | printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for" |
1608 | " inode %s:%lu\n", mode, inode->i_sb->s_id, | |
1609 | inode->i_ino); | |
1da177e4 LT |
1610 | } |
1611 | EXPORT_SYMBOL(init_special_inode); | |
a1bd120d DM |
1612 | |
1613 | /** | |
eaae668d | 1614 | * inode_init_owner - Init uid,gid,mode for new inode according to posix standards |
a1bd120d DM |
1615 | * @inode: New inode |
1616 | * @dir: Directory inode | |
1617 | * @mode: mode of the new inode | |
1618 | */ | |
1619 | void inode_init_owner(struct inode *inode, const struct inode *dir, | |
1620 | mode_t mode) | |
1621 | { | |
1622 | inode->i_uid = current_fsuid(); | |
1623 | if (dir && dir->i_mode & S_ISGID) { | |
1624 | inode->i_gid = dir->i_gid; | |
1625 | if (S_ISDIR(mode)) | |
1626 | mode |= S_ISGID; | |
1627 | } else | |
1628 | inode->i_gid = current_fsgid(); | |
1629 | inode->i_mode = mode; | |
1630 | } | |
1631 | EXPORT_SYMBOL(inode_init_owner); | |
e795b717 | 1632 | |
2e149670 SH |
1633 | /** |
1634 | * inode_owner_or_capable - check current task permissions to inode | |
1635 | * @inode: inode being checked | |
1636 | * | |
1637 | * Return true if current either has CAP_FOWNER to the inode, or | |
1638 | * owns the file. | |
e795b717 | 1639 | */ |
2e149670 | 1640 | bool inode_owner_or_capable(const struct inode *inode) |
e795b717 SH |
1641 | { |
1642 | struct user_namespace *ns = inode_userns(inode); | |
1643 | ||
1644 | if (current_user_ns() == ns && current_fsuid() == inode->i_uid) | |
1645 | return true; | |
1646 | if (ns_capable(ns, CAP_FOWNER)) | |
1647 | return true; | |
1648 | return false; | |
1649 | } | |
2e149670 | 1650 | EXPORT_SYMBOL(inode_owner_or_capable); |