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> |
1da177e4 LT |
27 | |
28 | /* | |
29 | * This is needed for the following functions: | |
30 | * - inode_has_buffers | |
1da177e4 LT |
31 | * - invalidate_bdev |
32 | * | |
33 | * FIXME: remove all knowledge of the buffer layer from this file | |
34 | */ | |
35 | #include <linux/buffer_head.h> | |
36 | ||
37 | /* | |
38 | * New inode.c implementation. | |
39 | * | |
40 | * This implementation has the basic premise of trying | |
41 | * to be extremely low-overhead and SMP-safe, yet be | |
42 | * simple enough to be "obviously correct". | |
43 | * | |
44 | * Famous last words. | |
45 | */ | |
46 | ||
47 | /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */ | |
48 | ||
49 | /* #define INODE_PARANOIA 1 */ | |
50 | /* #define INODE_DEBUG 1 */ | |
51 | ||
52 | /* | |
53 | * Inode lookup is no longer as critical as it used to be: | |
54 | * most of the lookups are going to be through the dcache. | |
55 | */ | |
56 | #define I_HASHBITS i_hash_shift | |
57 | #define I_HASHMASK i_hash_mask | |
58 | ||
fa3536cc ED |
59 | static unsigned int i_hash_mask __read_mostly; |
60 | static unsigned int i_hash_shift __read_mostly; | |
1da177e4 LT |
61 | |
62 | /* | |
63 | * Each inode can be on two separate lists. One is | |
64 | * the hash list of the inode, used for lookups. The | |
65 | * other linked list is the "type" list: | |
66 | * "in_use" - valid inode, i_count > 0, i_nlink > 0 | |
67 | * "dirty" - as "in_use" but also dirty | |
68 | * "unused" - valid inode, i_count = 0 | |
69 | * | |
70 | * A "dirty" list is maintained for each super block, | |
71 | * allowing for low-overhead inode sync() operations. | |
72 | */ | |
73 | ||
7ccf19a8 | 74 | static LIST_HEAD(inode_lru); |
fa3536cc | 75 | static struct hlist_head *inode_hashtable __read_mostly; |
1da177e4 LT |
76 | |
77 | /* | |
78 | * A simple spinlock to protect the list manipulations. | |
79 | * | |
80 | * NOTE! You also have to own the lock if you change | |
81 | * the i_state of an inode while it is in use.. | |
82 | */ | |
83 | DEFINE_SPINLOCK(inode_lock); | |
84 | ||
85 | /* | |
88e0fbc4 | 86 | * iprune_sem provides exclusion between the kswapd or try_to_free_pages |
1da177e4 LT |
87 | * icache shrinking path, and the umount path. Without this exclusion, |
88 | * by the time prune_icache calls iput for the inode whose pages it has | |
89 | * been invalidating, or by the time it calls clear_inode & destroy_inode | |
90 | * from its final dispose_list, the struct super_block they refer to | |
91 | * (for inode->i_sb->s_op) may already have been freed and reused. | |
88e0fbc4 NP |
92 | * |
93 | * We make this an rwsem because the fastpath is icache shrinking. In | |
94 | * some cases a filesystem may be doing a significant amount of work in | |
95 | * its inode reclaim code, so this should improve parallelism. | |
1da177e4 | 96 | */ |
88e0fbc4 | 97 | static DECLARE_RWSEM(iprune_sem); |
1da177e4 LT |
98 | |
99 | /* | |
100 | * Statistics gathering.. | |
101 | */ | |
102 | struct inodes_stat_t inodes_stat; | |
103 | ||
cffbc8aa DC |
104 | static struct percpu_counter nr_inodes __cacheline_aligned_in_smp; |
105 | static struct percpu_counter nr_inodes_unused __cacheline_aligned_in_smp; | |
106 | ||
6b3304b5 | 107 | static struct kmem_cache *inode_cachep __read_mostly; |
1da177e4 | 108 | |
cffbc8aa DC |
109 | static inline int get_nr_inodes(void) |
110 | { | |
111 | return percpu_counter_sum_positive(&nr_inodes); | |
112 | } | |
113 | ||
114 | static inline int get_nr_inodes_unused(void) | |
115 | { | |
116 | return percpu_counter_sum_positive(&nr_inodes_unused); | |
117 | } | |
118 | ||
119 | int get_nr_dirty_inodes(void) | |
120 | { | |
121 | int nr_dirty = get_nr_inodes() - get_nr_inodes_unused(); | |
122 | return nr_dirty > 0 ? nr_dirty : 0; | |
123 | ||
124 | } | |
125 | ||
126 | /* | |
127 | * Handle nr_inode sysctl | |
128 | */ | |
129 | #ifdef CONFIG_SYSCTL | |
130 | int proc_nr_inodes(ctl_table *table, int write, | |
131 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
132 | { | |
133 | inodes_stat.nr_inodes = get_nr_inodes(); | |
134 | inodes_stat.nr_unused = get_nr_inodes_unused(); | |
135 | return proc_dointvec(table, write, buffer, lenp, ppos); | |
136 | } | |
137 | #endif | |
138 | ||
1c0eeaf5 JE |
139 | static void wake_up_inode(struct inode *inode) |
140 | { | |
141 | /* | |
142 | * Prevent speculative execution through spin_unlock(&inode_lock); | |
143 | */ | |
144 | smp_mb(); | |
eaff8079 | 145 | wake_up_bit(&inode->i_state, __I_NEW); |
1c0eeaf5 JE |
146 | } |
147 | ||
2cb1599f DC |
148 | /** |
149 | * inode_init_always - perform inode structure intialisation | |
0bc02f3f RD |
150 | * @sb: superblock inode belongs to |
151 | * @inode: inode to initialise | |
2cb1599f DC |
152 | * |
153 | * These are initializations that need to be done on every inode | |
154 | * allocation as the fields are not initialised by slab allocation. | |
155 | */ | |
54e34621 | 156 | int inode_init_always(struct super_block *sb, struct inode *inode) |
1da177e4 | 157 | { |
f5e54d6e | 158 | static const struct address_space_operations empty_aops; |
6e1d5dcc | 159 | static const struct inode_operations empty_iops; |
99ac48f5 | 160 | static const struct file_operations empty_fops; |
6b3304b5 | 161 | struct address_space *const mapping = &inode->i_data; |
2cb1599f DC |
162 | |
163 | inode->i_sb = sb; | |
164 | inode->i_blkbits = sb->s_blocksize_bits; | |
165 | inode->i_flags = 0; | |
166 | atomic_set(&inode->i_count, 1); | |
167 | inode->i_op = &empty_iops; | |
168 | inode->i_fop = &empty_fops; | |
169 | inode->i_nlink = 1; | |
56ff5efa AV |
170 | inode->i_uid = 0; |
171 | inode->i_gid = 0; | |
2cb1599f DC |
172 | atomic_set(&inode->i_writecount, 0); |
173 | inode->i_size = 0; | |
174 | inode->i_blocks = 0; | |
175 | inode->i_bytes = 0; | |
176 | inode->i_generation = 0; | |
1da177e4 | 177 | #ifdef CONFIG_QUOTA |
2cb1599f | 178 | memset(&inode->i_dquot, 0, sizeof(inode->i_dquot)); |
1da177e4 | 179 | #endif |
2cb1599f DC |
180 | inode->i_pipe = NULL; |
181 | inode->i_bdev = NULL; | |
182 | inode->i_cdev = NULL; | |
183 | inode->i_rdev = 0; | |
184 | inode->dirtied_when = 0; | |
6146f0d5 MZ |
185 | |
186 | if (security_inode_alloc(inode)) | |
54e34621 | 187 | goto out; |
2cb1599f DC |
188 | spin_lock_init(&inode->i_lock); |
189 | lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key); | |
190 | ||
191 | mutex_init(&inode->i_mutex); | |
192 | lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key); | |
193 | ||
194 | init_rwsem(&inode->i_alloc_sem); | |
195 | lockdep_set_class(&inode->i_alloc_sem, &sb->s_type->i_alloc_sem_key); | |
196 | ||
197 | mapping->a_ops = &empty_aops; | |
198 | mapping->host = inode; | |
199 | mapping->flags = 0; | |
3c1d4378 | 200 | mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE); |
2cb1599f DC |
201 | mapping->assoc_mapping = NULL; |
202 | mapping->backing_dev_info = &default_backing_dev_info; | |
203 | mapping->writeback_index = 0; | |
204 | ||
205 | /* | |
206 | * If the block_device provides a backing_dev_info for client | |
207 | * inodes then use that. Otherwise the inode share the bdev's | |
208 | * backing_dev_info. | |
209 | */ | |
210 | if (sb->s_bdev) { | |
211 | struct backing_dev_info *bdi; | |
212 | ||
2c96ce9f | 213 | bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info; |
2cb1599f DC |
214 | mapping->backing_dev_info = bdi; |
215 | } | |
216 | inode->i_private = NULL; | |
217 | inode->i_mapping = mapping; | |
f19d4a8f AV |
218 | #ifdef CONFIG_FS_POSIX_ACL |
219 | inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED; | |
220 | #endif | |
2cb1599f | 221 | |
3be25f49 EP |
222 | #ifdef CONFIG_FSNOTIFY |
223 | inode->i_fsnotify_mask = 0; | |
224 | #endif | |
225 | ||
cffbc8aa DC |
226 | percpu_counter_inc(&nr_inodes); |
227 | ||
54e34621 | 228 | return 0; |
54e34621 CH |
229 | out: |
230 | return -ENOMEM; | |
1da177e4 | 231 | } |
2cb1599f DC |
232 | EXPORT_SYMBOL(inode_init_always); |
233 | ||
234 | static struct inode *alloc_inode(struct super_block *sb) | |
235 | { | |
236 | struct inode *inode; | |
237 | ||
238 | if (sb->s_op->alloc_inode) | |
239 | inode = sb->s_op->alloc_inode(sb); | |
240 | else | |
241 | inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL); | |
242 | ||
54e34621 CH |
243 | if (!inode) |
244 | return NULL; | |
245 | ||
246 | if (unlikely(inode_init_always(sb, inode))) { | |
247 | if (inode->i_sb->s_op->destroy_inode) | |
248 | inode->i_sb->s_op->destroy_inode(inode); | |
249 | else | |
250 | kmem_cache_free(inode_cachep, inode); | |
251 | return NULL; | |
252 | } | |
253 | ||
254 | return inode; | |
2cb1599f | 255 | } |
1da177e4 | 256 | |
2e00c97e | 257 | void __destroy_inode(struct inode *inode) |
1da177e4 | 258 | { |
b7542f8c | 259 | BUG_ON(inode_has_buffers(inode)); |
1da177e4 | 260 | security_inode_free(inode); |
3be25f49 | 261 | fsnotify_inode_delete(inode); |
f19d4a8f AV |
262 | #ifdef CONFIG_FS_POSIX_ACL |
263 | if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED) | |
264 | posix_acl_release(inode->i_acl); | |
265 | if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED) | |
266 | posix_acl_release(inode->i_default_acl); | |
267 | #endif | |
cffbc8aa | 268 | percpu_counter_dec(&nr_inodes); |
2e00c97e CH |
269 | } |
270 | EXPORT_SYMBOL(__destroy_inode); | |
271 | ||
56b0dacf | 272 | static void destroy_inode(struct inode *inode) |
2e00c97e | 273 | { |
7ccf19a8 | 274 | BUG_ON(!list_empty(&inode->i_lru)); |
2e00c97e | 275 | __destroy_inode(inode); |
1da177e4 LT |
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 | } | |
1da177e4 LT |
281 | |
282 | /* | |
283 | * These are initializations that only need to be done | |
284 | * once, because the fields are idempotent across use | |
285 | * of the inode, so let the slab aware of that. | |
286 | */ | |
287 | void inode_init_once(struct inode *inode) | |
288 | { | |
289 | memset(inode, 0, sizeof(*inode)); | |
290 | INIT_HLIST_NODE(&inode->i_hash); | |
291 | INIT_LIST_HEAD(&inode->i_dentry); | |
292 | INIT_LIST_HEAD(&inode->i_devices); | |
7ccf19a8 NP |
293 | INIT_LIST_HEAD(&inode->i_wb_list); |
294 | INIT_LIST_HEAD(&inode->i_lru); | |
1da177e4 | 295 | INIT_RADIX_TREE(&inode->i_data.page_tree, GFP_ATOMIC); |
19fd6231 | 296 | spin_lock_init(&inode->i_data.tree_lock); |
1da177e4 LT |
297 | spin_lock_init(&inode->i_data.i_mmap_lock); |
298 | INIT_LIST_HEAD(&inode->i_data.private_list); | |
299 | spin_lock_init(&inode->i_data.private_lock); | |
300 | INIT_RAW_PRIO_TREE_ROOT(&inode->i_data.i_mmap); | |
301 | INIT_LIST_HEAD(&inode->i_data.i_mmap_nonlinear); | |
1da177e4 | 302 | i_size_ordered_init(inode); |
3be25f49 | 303 | #ifdef CONFIG_FSNOTIFY |
e61ce867 | 304 | INIT_HLIST_HEAD(&inode->i_fsnotify_marks); |
3be25f49 | 305 | #endif |
1da177e4 | 306 | } |
1da177e4 LT |
307 | EXPORT_SYMBOL(inode_init_once); |
308 | ||
51cc5068 | 309 | static void init_once(void *foo) |
1da177e4 | 310 | { |
6b3304b5 | 311 | struct inode *inode = (struct inode *) foo; |
1da177e4 | 312 | |
a35afb83 | 313 | inode_init_once(inode); |
1da177e4 LT |
314 | } |
315 | ||
316 | /* | |
317 | * inode_lock must be held | |
318 | */ | |
6b3304b5 | 319 | void __iget(struct inode *inode) |
1da177e4 | 320 | { |
9e38d86f NP |
321 | atomic_inc(&inode->i_count); |
322 | } | |
2e147f1e | 323 | |
7de9c6ee AV |
324 | /* |
325 | * get additional reference to inode; caller must already hold one. | |
326 | */ | |
327 | void ihold(struct inode *inode) | |
328 | { | |
329 | WARN_ON(atomic_inc_return(&inode->i_count) < 2); | |
330 | } | |
331 | EXPORT_SYMBOL(ihold); | |
332 | ||
9e38d86f NP |
333 | static void inode_lru_list_add(struct inode *inode) |
334 | { | |
7ccf19a8 NP |
335 | if (list_empty(&inode->i_lru)) { |
336 | list_add(&inode->i_lru, &inode_lru); | |
9e38d86f NP |
337 | percpu_counter_inc(&nr_inodes_unused); |
338 | } | |
339 | } | |
340 | ||
341 | static void inode_lru_list_del(struct inode *inode) | |
342 | { | |
7ccf19a8 NP |
343 | if (!list_empty(&inode->i_lru)) { |
344 | list_del_init(&inode->i_lru); | |
9e38d86f NP |
345 | percpu_counter_dec(&nr_inodes_unused); |
346 | } | |
1da177e4 LT |
347 | } |
348 | ||
646ec461 CH |
349 | static inline void __inode_sb_list_add(struct inode *inode) |
350 | { | |
351 | list_add(&inode->i_sb_list, &inode->i_sb->s_inodes); | |
352 | } | |
353 | ||
354 | /** | |
355 | * inode_sb_list_add - add inode to the superblock list of inodes | |
356 | * @inode: inode to add | |
357 | */ | |
358 | void inode_sb_list_add(struct inode *inode) | |
359 | { | |
360 | spin_lock(&inode_lock); | |
361 | __inode_sb_list_add(inode); | |
362 | spin_unlock(&inode_lock); | |
363 | } | |
364 | EXPORT_SYMBOL_GPL(inode_sb_list_add); | |
365 | ||
366 | static inline void __inode_sb_list_del(struct inode *inode) | |
367 | { | |
368 | list_del_init(&inode->i_sb_list); | |
369 | } | |
370 | ||
4c51acbc DC |
371 | static unsigned long hash(struct super_block *sb, unsigned long hashval) |
372 | { | |
373 | unsigned long tmp; | |
374 | ||
375 | tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) / | |
376 | L1_CACHE_BYTES; | |
377 | tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> I_HASHBITS); | |
378 | return tmp & I_HASHMASK; | |
379 | } | |
380 | ||
381 | /** | |
382 | * __insert_inode_hash - hash an inode | |
383 | * @inode: unhashed inode | |
384 | * @hashval: unsigned long value used to locate this object in the | |
385 | * inode_hashtable. | |
386 | * | |
387 | * Add an inode to the inode hash for this superblock. | |
388 | */ | |
389 | void __insert_inode_hash(struct inode *inode, unsigned long hashval) | |
390 | { | |
646ec461 CH |
391 | struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval); |
392 | ||
4c51acbc | 393 | spin_lock(&inode_lock); |
646ec461 | 394 | hlist_add_head(&inode->i_hash, b); |
4c51acbc DC |
395 | spin_unlock(&inode_lock); |
396 | } | |
397 | EXPORT_SYMBOL(__insert_inode_hash); | |
398 | ||
399 | /** | |
400 | * __remove_inode_hash - remove an inode from the hash | |
401 | * @inode: inode to unhash | |
402 | * | |
403 | * Remove an inode from the superblock. | |
404 | */ | |
405 | static void __remove_inode_hash(struct inode *inode) | |
406 | { | |
407 | hlist_del_init(&inode->i_hash); | |
408 | } | |
409 | ||
410 | /** | |
411 | * remove_inode_hash - remove an inode from the hash | |
412 | * @inode: inode to unhash | |
413 | * | |
414 | * Remove an inode from the superblock. | |
415 | */ | |
416 | void remove_inode_hash(struct inode *inode) | |
417 | { | |
418 | spin_lock(&inode_lock); | |
419 | hlist_del_init(&inode->i_hash); | |
420 | spin_unlock(&inode_lock); | |
421 | } | |
422 | EXPORT_SYMBOL(remove_inode_hash); | |
423 | ||
b0683aa6 AV |
424 | void end_writeback(struct inode *inode) |
425 | { | |
426 | might_sleep(); | |
427 | BUG_ON(inode->i_data.nrpages); | |
428 | BUG_ON(!list_empty(&inode->i_data.private_list)); | |
429 | BUG_ON(!(inode->i_state & I_FREEING)); | |
430 | BUG_ON(inode->i_state & I_CLEAR); | |
431 | inode_sync_wait(inode); | |
432 | inode->i_state = I_FREEING | I_CLEAR; | |
433 | } | |
434 | EXPORT_SYMBOL(end_writeback); | |
435 | ||
644da596 | 436 | static void evict(struct inode *inode) |
b4272d4c AV |
437 | { |
438 | const struct super_operations *op = inode->i_sb->s_op; | |
439 | ||
be7ce416 AV |
440 | if (op->evict_inode) { |
441 | op->evict_inode(inode); | |
b4272d4c AV |
442 | } else { |
443 | if (inode->i_data.nrpages) | |
444 | truncate_inode_pages(&inode->i_data, 0); | |
30140837 | 445 | end_writeback(inode); |
b4272d4c | 446 | } |
661074e9 AV |
447 | if (S_ISBLK(inode->i_mode) && inode->i_bdev) |
448 | bd_forget(inode); | |
449 | if (S_ISCHR(inode->i_mode) && inode->i_cdev) | |
450 | cd_forget(inode); | |
b4272d4c AV |
451 | } |
452 | ||
1da177e4 LT |
453 | /* |
454 | * dispose_list - dispose of the contents of a local list | |
455 | * @head: the head of the list to free | |
456 | * | |
457 | * Dispose-list gets a local list with local inodes in it, so it doesn't | |
458 | * need to worry about list corruption and SMP locks. | |
459 | */ | |
460 | static void dispose_list(struct list_head *head) | |
461 | { | |
1da177e4 LT |
462 | while (!list_empty(head)) { |
463 | struct inode *inode; | |
464 | ||
7ccf19a8 NP |
465 | inode = list_first_entry(head, struct inode, i_lru); |
466 | list_del_init(&inode->i_lru); | |
1da177e4 | 467 | |
644da596 | 468 | evict(inode); |
4120db47 AB |
469 | |
470 | spin_lock(&inode_lock); | |
4c51acbc | 471 | __remove_inode_hash(inode); |
646ec461 | 472 | __inode_sb_list_del(inode); |
4120db47 AB |
473 | spin_unlock(&inode_lock); |
474 | ||
475 | wake_up_inode(inode); | |
1da177e4 | 476 | destroy_inode(inode); |
1da177e4 | 477 | } |
1da177e4 LT |
478 | } |
479 | ||
a0318786 CH |
480 | /** |
481 | * invalidate_inodes - attempt to free all inodes on a superblock | |
482 | * @sb: superblock to operate on | |
483 | * | |
484 | * Attempts to free all inodes for a given superblock. If there were any | |
485 | * busy inodes return a non-zero value, else zero. | |
1da177e4 | 486 | */ |
a0318786 | 487 | int invalidate_inodes(struct super_block *sb) |
1da177e4 | 488 | { |
cffbc8aa | 489 | int busy = 0; |
a0318786 CH |
490 | struct inode *inode, *next; |
491 | LIST_HEAD(dispose); | |
1da177e4 | 492 | |
a0318786 CH |
493 | down_write(&iprune_sem); |
494 | ||
495 | spin_lock(&inode_lock); | |
496 | fsnotify_unmount_inodes(&sb->s_inodes); | |
497 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { | |
aabb8fdb NP |
498 | if (inode->i_state & I_NEW) |
499 | continue; | |
99a38919 CH |
500 | if (atomic_read(&inode->i_count)) { |
501 | busy = 1; | |
1da177e4 LT |
502 | continue; |
503 | } | |
99a38919 | 504 | |
99a38919 | 505 | inode->i_state |= I_FREEING; |
7ccf19a8 NP |
506 | |
507 | /* | |
508 | * Move the inode off the IO lists and LRU once I_FREEING is | |
509 | * set so that it won't get moved back on there if it is dirty. | |
510 | */ | |
a0318786 | 511 | list_move(&inode->i_lru, &dispose); |
7ccf19a8 | 512 | list_del_init(&inode->i_wb_list); |
99a38919 CH |
513 | if (!(inode->i_state & (I_DIRTY | I_SYNC))) |
514 | percpu_counter_dec(&nr_inodes_unused); | |
1da177e4 | 515 | } |
1da177e4 LT |
516 | spin_unlock(&inode_lock); |
517 | ||
a0318786 | 518 | dispose_list(&dispose); |
88e0fbc4 | 519 | up_write(&iprune_sem); |
1da177e4 LT |
520 | |
521 | return busy; | |
522 | } | |
1da177e4 LT |
523 | |
524 | static int can_unuse(struct inode *inode) | |
525 | { | |
9e38d86f | 526 | if (inode->i_state & ~I_REFERENCED) |
1da177e4 LT |
527 | return 0; |
528 | if (inode_has_buffers(inode)) | |
529 | return 0; | |
530 | if (atomic_read(&inode->i_count)) | |
531 | return 0; | |
532 | if (inode->i_data.nrpages) | |
533 | return 0; | |
534 | return 1; | |
535 | } | |
536 | ||
537 | /* | |
9e38d86f NP |
538 | * Scan `goal' inodes on the unused list for freeable ones. They are moved to a |
539 | * temporary list and then are freed outside inode_lock by dispose_list(). | |
1da177e4 LT |
540 | * |
541 | * Any inodes which are pinned purely because of attached pagecache have their | |
9e38d86f NP |
542 | * pagecache removed. If the inode has metadata buffers attached to |
543 | * mapping->private_list then try to remove them. | |
1da177e4 | 544 | * |
9e38d86f NP |
545 | * If the inode has the I_REFERENCED flag set, then it means that it has been |
546 | * used recently - the flag is set in iput_final(). When we encounter such an | |
547 | * inode, clear the flag and move it to the back of the LRU so it gets another | |
548 | * pass through the LRU before it gets reclaimed. This is necessary because of | |
549 | * the fact we are doing lazy LRU updates to minimise lock contention so the | |
550 | * LRU does not have strict ordering. Hence we don't want to reclaim inodes | |
551 | * with this flag set because they are the inodes that are out of order. | |
1da177e4 LT |
552 | */ |
553 | static void prune_icache(int nr_to_scan) | |
554 | { | |
555 | LIST_HEAD(freeable); | |
1da177e4 LT |
556 | int nr_scanned; |
557 | unsigned long reap = 0; | |
558 | ||
88e0fbc4 | 559 | down_read(&iprune_sem); |
1da177e4 LT |
560 | spin_lock(&inode_lock); |
561 | for (nr_scanned = 0; nr_scanned < nr_to_scan; nr_scanned++) { | |
562 | struct inode *inode; | |
563 | ||
7ccf19a8 | 564 | if (list_empty(&inode_lru)) |
1da177e4 LT |
565 | break; |
566 | ||
7ccf19a8 | 567 | inode = list_entry(inode_lru.prev, struct inode, i_lru); |
1da177e4 | 568 | |
9e38d86f NP |
569 | /* |
570 | * Referenced or dirty inodes are still in use. Give them | |
571 | * another pass through the LRU as we canot reclaim them now. | |
572 | */ | |
573 | if (atomic_read(&inode->i_count) || | |
574 | (inode->i_state & ~I_REFERENCED)) { | |
7ccf19a8 | 575 | list_del_init(&inode->i_lru); |
9e38d86f NP |
576 | percpu_counter_dec(&nr_inodes_unused); |
577 | continue; | |
578 | } | |
579 | ||
580 | /* recently referenced inodes get one more pass */ | |
581 | if (inode->i_state & I_REFERENCED) { | |
7ccf19a8 | 582 | list_move(&inode->i_lru, &inode_lru); |
9e38d86f | 583 | inode->i_state &= ~I_REFERENCED; |
1da177e4 LT |
584 | continue; |
585 | } | |
586 | if (inode_has_buffers(inode) || inode->i_data.nrpages) { | |
587 | __iget(inode); | |
588 | spin_unlock(&inode_lock); | |
589 | if (remove_inode_buffers(inode)) | |
fc0ecff6 AM |
590 | reap += invalidate_mapping_pages(&inode->i_data, |
591 | 0, -1); | |
1da177e4 LT |
592 | iput(inode); |
593 | spin_lock(&inode_lock); | |
594 | ||
7ccf19a8 NP |
595 | if (inode != list_entry(inode_lru.next, |
596 | struct inode, i_lru)) | |
1da177e4 LT |
597 | continue; /* wrong inode or list_empty */ |
598 | if (!can_unuse(inode)) | |
599 | continue; | |
600 | } | |
7ef0d737 | 601 | WARN_ON(inode->i_state & I_NEW); |
1da177e4 | 602 | inode->i_state |= I_FREEING; |
7ccf19a8 NP |
603 | |
604 | /* | |
605 | * Move the inode off the IO lists and LRU once I_FREEING is | |
606 | * set so that it won't get moved back on there if it is dirty. | |
607 | */ | |
608 | list_move(&inode->i_lru, &freeable); | |
609 | list_del_init(&inode->i_wb_list); | |
cffbc8aa | 610 | percpu_counter_dec(&nr_inodes_unused); |
1da177e4 | 611 | } |
f8891e5e CL |
612 | if (current_is_kswapd()) |
613 | __count_vm_events(KSWAPD_INODESTEAL, reap); | |
614 | else | |
615 | __count_vm_events(PGINODESTEAL, reap); | |
1da177e4 LT |
616 | spin_unlock(&inode_lock); |
617 | ||
618 | dispose_list(&freeable); | |
88e0fbc4 | 619 | up_read(&iprune_sem); |
1da177e4 LT |
620 | } |
621 | ||
622 | /* | |
623 | * shrink_icache_memory() will attempt to reclaim some unused inodes. Here, | |
624 | * "unused" means that no dentries are referring to the inodes: the files are | |
625 | * not open and the dcache references to those inodes have already been | |
626 | * reclaimed. | |
627 | * | |
628 | * This function is passed the number of inodes to scan, and it returns the | |
629 | * total number of remaining possibly-reclaimable inodes. | |
630 | */ | |
7f8275d0 | 631 | static int shrink_icache_memory(struct shrinker *shrink, int nr, gfp_t gfp_mask) |
1da177e4 LT |
632 | { |
633 | if (nr) { | |
634 | /* | |
635 | * Nasty deadlock avoidance. We may hold various FS locks, | |
636 | * and we don't want to recurse into the FS that called us | |
637 | * in clear_inode() and friends.. | |
6b3304b5 | 638 | */ |
1da177e4 LT |
639 | if (!(gfp_mask & __GFP_FS)) |
640 | return -1; | |
641 | prune_icache(nr); | |
642 | } | |
cffbc8aa | 643 | return (get_nr_inodes_unused() / 100) * sysctl_vfs_cache_pressure; |
1da177e4 LT |
644 | } |
645 | ||
8e1f936b RR |
646 | static struct shrinker icache_shrinker = { |
647 | .shrink = shrink_icache_memory, | |
648 | .seeks = DEFAULT_SEEKS, | |
649 | }; | |
650 | ||
1da177e4 LT |
651 | static void __wait_on_freeing_inode(struct inode *inode); |
652 | /* | |
653 | * Called with the inode lock held. | |
1da177e4 | 654 | */ |
6b3304b5 MK |
655 | static struct inode *find_inode(struct super_block *sb, |
656 | struct hlist_head *head, | |
657 | int (*test)(struct inode *, void *), | |
658 | void *data) | |
1da177e4 LT |
659 | { |
660 | struct hlist_node *node; | |
6b3304b5 | 661 | struct inode *inode = NULL; |
1da177e4 LT |
662 | |
663 | repeat: | |
c5c8be3c | 664 | hlist_for_each_entry(inode, node, head, i_hash) { |
1da177e4 LT |
665 | if (inode->i_sb != sb) |
666 | continue; | |
667 | if (!test(inode, data)) | |
668 | continue; | |
a4ffdde6 | 669 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
670 | __wait_on_freeing_inode(inode); |
671 | goto repeat; | |
672 | } | |
f7899bd5 CH |
673 | __iget(inode); |
674 | return inode; | |
1da177e4 | 675 | } |
f7899bd5 | 676 | return NULL; |
1da177e4 LT |
677 | } |
678 | ||
679 | /* | |
680 | * find_inode_fast is the fast path version of find_inode, see the comment at | |
681 | * iget_locked for details. | |
682 | */ | |
6b3304b5 MK |
683 | static struct inode *find_inode_fast(struct super_block *sb, |
684 | struct hlist_head *head, unsigned long ino) | |
1da177e4 LT |
685 | { |
686 | struct hlist_node *node; | |
6b3304b5 | 687 | struct inode *inode = NULL; |
1da177e4 LT |
688 | |
689 | repeat: | |
c5c8be3c | 690 | hlist_for_each_entry(inode, node, head, i_hash) { |
1da177e4 LT |
691 | if (inode->i_ino != ino) |
692 | continue; | |
693 | if (inode->i_sb != sb) | |
694 | continue; | |
a4ffdde6 | 695 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
696 | __wait_on_freeing_inode(inode); |
697 | goto repeat; | |
698 | } | |
f7899bd5 CH |
699 | __iget(inode); |
700 | return inode; | |
1da177e4 | 701 | } |
f7899bd5 | 702 | return NULL; |
1da177e4 LT |
703 | } |
704 | ||
f991bd2e ED |
705 | /* |
706 | * Each cpu owns a range of LAST_INO_BATCH numbers. | |
707 | * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations, | |
708 | * to renew the exhausted range. | |
709 | * | |
710 | * This does not significantly increase overflow rate because every CPU can | |
711 | * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is | |
712 | * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the | |
713 | * 2^32 range, and is a worst-case. Even a 50% wastage would only increase | |
714 | * overflow rate by 2x, which does not seem too significant. | |
715 | * | |
716 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
717 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
718 | * here to attempt to avoid that. | |
719 | */ | |
720 | #define LAST_INO_BATCH 1024 | |
721 | static DEFINE_PER_CPU(unsigned int, last_ino); | |
722 | ||
85fe4025 | 723 | unsigned int get_next_ino(void) |
f991bd2e ED |
724 | { |
725 | unsigned int *p = &get_cpu_var(last_ino); | |
726 | unsigned int res = *p; | |
727 | ||
728 | #ifdef CONFIG_SMP | |
729 | if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) { | |
730 | static atomic_t shared_last_ino; | |
731 | int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino); | |
732 | ||
733 | res = next - LAST_INO_BATCH; | |
734 | } | |
735 | #endif | |
736 | ||
737 | *p = ++res; | |
738 | put_cpu_var(last_ino); | |
739 | return res; | |
740 | } | |
85fe4025 | 741 | EXPORT_SYMBOL(get_next_ino); |
f991bd2e | 742 | |
1da177e4 LT |
743 | /** |
744 | * new_inode - obtain an inode | |
745 | * @sb: superblock | |
746 | * | |
769848c0 | 747 | * Allocates a new inode for given superblock. The default gfp_mask |
3c1d4378 | 748 | * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE. |
769848c0 MG |
749 | * If HIGHMEM pages are unsuitable or it is known that pages allocated |
750 | * for the page cache are not reclaimable or migratable, | |
751 | * mapping_set_gfp_mask() must be called with suitable flags on the | |
752 | * newly created inode's mapping | |
753 | * | |
1da177e4 LT |
754 | */ |
755 | struct inode *new_inode(struct super_block *sb) | |
756 | { | |
6b3304b5 | 757 | struct inode *inode; |
1da177e4 LT |
758 | |
759 | spin_lock_prefetch(&inode_lock); | |
6b3304b5 | 760 | |
1da177e4 LT |
761 | inode = alloc_inode(sb); |
762 | if (inode) { | |
763 | spin_lock(&inode_lock); | |
646ec461 | 764 | __inode_sb_list_add(inode); |
1da177e4 LT |
765 | inode->i_state = 0; |
766 | spin_unlock(&inode_lock); | |
767 | } | |
768 | return inode; | |
769 | } | |
1da177e4 LT |
770 | EXPORT_SYMBOL(new_inode); |
771 | ||
772 | void unlock_new_inode(struct inode *inode) | |
773 | { | |
14358e6d | 774 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
a3314a0e | 775 | if (S_ISDIR(inode->i_mode)) { |
1e89a5e1 PZ |
776 | struct file_system_type *type = inode->i_sb->s_type; |
777 | ||
9a7aa12f JK |
778 | /* Set new key only if filesystem hasn't already changed it */ |
779 | if (!lockdep_match_class(&inode->i_mutex, | |
780 | &type->i_mutex_key)) { | |
781 | /* | |
782 | * ensure nobody is actually holding i_mutex | |
783 | */ | |
784 | mutex_destroy(&inode->i_mutex); | |
785 | mutex_init(&inode->i_mutex); | |
786 | lockdep_set_class(&inode->i_mutex, | |
787 | &type->i_mutex_dir_key); | |
788 | } | |
1e89a5e1 | 789 | } |
14358e6d | 790 | #endif |
1da177e4 | 791 | /* |
eaff8079 | 792 | * This is special! We do not need the spinlock when clearing I_NEW, |
580be083 JK |
793 | * because we're guaranteed that nobody else tries to do anything about |
794 | * the state of the inode when it is locked, as we just created it (so | |
eaff8079 | 795 | * there can be no old holders that haven't tested I_NEW). |
580be083 | 796 | * However we must emit the memory barrier so that other CPUs reliably |
eaff8079 | 797 | * see the clearing of I_NEW after the other inode initialisation has |
580be083 | 798 | * completed. |
1da177e4 | 799 | */ |
580be083 | 800 | smp_mb(); |
eaff8079 CH |
801 | WARN_ON(!(inode->i_state & I_NEW)); |
802 | inode->i_state &= ~I_NEW; | |
1da177e4 LT |
803 | wake_up_inode(inode); |
804 | } | |
1da177e4 LT |
805 | EXPORT_SYMBOL(unlock_new_inode); |
806 | ||
807 | /* | |
808 | * This is called without the inode lock held.. Be careful. | |
809 | * | |
810 | * We no longer cache the sb_flags in i_flags - see fs.h | |
811 | * -- rmk@arm.uk.linux.org | |
812 | */ | |
6b3304b5 MK |
813 | static struct inode *get_new_inode(struct super_block *sb, |
814 | struct hlist_head *head, | |
815 | int (*test)(struct inode *, void *), | |
816 | int (*set)(struct inode *, void *), | |
817 | void *data) | |
1da177e4 | 818 | { |
6b3304b5 | 819 | struct inode *inode; |
1da177e4 LT |
820 | |
821 | inode = alloc_inode(sb); | |
822 | if (inode) { | |
6b3304b5 | 823 | struct inode *old; |
1da177e4 LT |
824 | |
825 | spin_lock(&inode_lock); | |
826 | /* We released the lock, so.. */ | |
827 | old = find_inode(sb, head, test, data); | |
828 | if (!old) { | |
829 | if (set(inode, data)) | |
830 | goto set_failed; | |
831 | ||
646ec461 CH |
832 | hlist_add_head(&inode->i_hash, head); |
833 | __inode_sb_list_add(inode); | |
eaff8079 | 834 | inode->i_state = I_NEW; |
1da177e4 LT |
835 | spin_unlock(&inode_lock); |
836 | ||
837 | /* Return the locked inode with I_NEW set, the | |
838 | * caller is responsible for filling in the contents | |
839 | */ | |
840 | return inode; | |
841 | } | |
842 | ||
843 | /* | |
844 | * Uhhuh, somebody else created the same inode under | |
845 | * us. Use the old inode instead of the one we just | |
846 | * allocated. | |
847 | */ | |
1da177e4 LT |
848 | spin_unlock(&inode_lock); |
849 | destroy_inode(inode); | |
850 | inode = old; | |
851 | wait_on_inode(inode); | |
852 | } | |
853 | return inode; | |
854 | ||
855 | set_failed: | |
856 | spin_unlock(&inode_lock); | |
857 | destroy_inode(inode); | |
858 | return NULL; | |
859 | } | |
860 | ||
861 | /* | |
862 | * get_new_inode_fast is the fast path version of get_new_inode, see the | |
863 | * comment at iget_locked for details. | |
864 | */ | |
6b3304b5 MK |
865 | static struct inode *get_new_inode_fast(struct super_block *sb, |
866 | struct hlist_head *head, unsigned long ino) | |
1da177e4 | 867 | { |
6b3304b5 | 868 | struct inode *inode; |
1da177e4 LT |
869 | |
870 | inode = alloc_inode(sb); | |
871 | if (inode) { | |
6b3304b5 | 872 | struct inode *old; |
1da177e4 LT |
873 | |
874 | spin_lock(&inode_lock); | |
875 | /* We released the lock, so.. */ | |
876 | old = find_inode_fast(sb, head, ino); | |
877 | if (!old) { | |
878 | inode->i_ino = ino; | |
646ec461 CH |
879 | hlist_add_head(&inode->i_hash, head); |
880 | __inode_sb_list_add(inode); | |
eaff8079 | 881 | inode->i_state = I_NEW; |
1da177e4 LT |
882 | spin_unlock(&inode_lock); |
883 | ||
884 | /* Return the locked inode with I_NEW set, the | |
885 | * caller is responsible for filling in the contents | |
886 | */ | |
887 | return inode; | |
888 | } | |
889 | ||
890 | /* | |
891 | * Uhhuh, somebody else created the same inode under | |
892 | * us. Use the old inode instead of the one we just | |
893 | * allocated. | |
894 | */ | |
1da177e4 LT |
895 | spin_unlock(&inode_lock); |
896 | destroy_inode(inode); | |
897 | inode = old; | |
898 | wait_on_inode(inode); | |
899 | } | |
900 | return inode; | |
901 | } | |
902 | ||
ad5e195a CH |
903 | /* |
904 | * search the inode cache for a matching inode number. | |
905 | * If we find one, then the inode number we are trying to | |
906 | * allocate is not unique and so we should not use it. | |
907 | * | |
908 | * Returns 1 if the inode number is unique, 0 if it is not. | |
909 | */ | |
910 | static int test_inode_iunique(struct super_block *sb, unsigned long ino) | |
911 | { | |
912 | struct hlist_head *b = inode_hashtable + hash(sb, ino); | |
913 | struct hlist_node *node; | |
914 | struct inode *inode; | |
915 | ||
916 | hlist_for_each_entry(inode, node, b, i_hash) { | |
917 | if (inode->i_ino == ino && inode->i_sb == sb) | |
918 | return 0; | |
919 | } | |
920 | ||
921 | return 1; | |
922 | } | |
923 | ||
1da177e4 LT |
924 | /** |
925 | * iunique - get a unique inode number | |
926 | * @sb: superblock | |
927 | * @max_reserved: highest reserved inode number | |
928 | * | |
929 | * Obtain an inode number that is unique on the system for a given | |
930 | * superblock. This is used by file systems that have no natural | |
931 | * permanent inode numbering system. An inode number is returned that | |
932 | * is higher than the reserved limit but unique. | |
933 | * | |
934 | * BUGS: | |
935 | * With a large number of inodes live on the file system this function | |
936 | * currently becomes quite slow. | |
937 | */ | |
938 | ino_t iunique(struct super_block *sb, ino_t max_reserved) | |
939 | { | |
866b04fc JL |
940 | /* |
941 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
942 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
943 | * here to attempt to avoid that. | |
944 | */ | |
ad5e195a | 945 | static DEFINE_SPINLOCK(iunique_lock); |
866b04fc | 946 | static unsigned int counter; |
1da177e4 | 947 | ino_t res; |
3361c7be | 948 | |
1da177e4 | 949 | spin_lock(&inode_lock); |
ad5e195a | 950 | spin_lock(&iunique_lock); |
3361c7be JL |
951 | do { |
952 | if (counter <= max_reserved) | |
953 | counter = max_reserved + 1; | |
1da177e4 | 954 | res = counter++; |
ad5e195a CH |
955 | } while (!test_inode_iunique(sb, res)); |
956 | spin_unlock(&iunique_lock); | |
3361c7be | 957 | spin_unlock(&inode_lock); |
1da177e4 | 958 | |
3361c7be JL |
959 | return res; |
960 | } | |
1da177e4 LT |
961 | EXPORT_SYMBOL(iunique); |
962 | ||
963 | struct inode *igrab(struct inode *inode) | |
964 | { | |
965 | spin_lock(&inode_lock); | |
a4ffdde6 | 966 | if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) |
1da177e4 LT |
967 | __iget(inode); |
968 | else | |
969 | /* | |
970 | * Handle the case where s_op->clear_inode is not been | |
971 | * called yet, and somebody is calling igrab | |
972 | * while the inode is getting freed. | |
973 | */ | |
974 | inode = NULL; | |
975 | spin_unlock(&inode_lock); | |
976 | return inode; | |
977 | } | |
1da177e4 LT |
978 | EXPORT_SYMBOL(igrab); |
979 | ||
980 | /** | |
981 | * ifind - internal function, you want ilookup5() or iget5(). | |
982 | * @sb: super block of file system to search | |
983 | * @head: the head of the list to search | |
984 | * @test: callback used for comparisons between inodes | |
985 | * @data: opaque data pointer to pass to @test | |
88bd5121 | 986 | * @wait: if true wait for the inode to be unlocked, if false do not |
1da177e4 LT |
987 | * |
988 | * ifind() searches for the inode specified by @data in the inode | |
989 | * cache. This is a generalized version of ifind_fast() for file systems where | |
990 | * the inode number is not sufficient for unique identification of an inode. | |
991 | * | |
992 | * If the inode is in the cache, the inode is returned with an incremented | |
993 | * reference count. | |
994 | * | |
995 | * Otherwise NULL is returned. | |
996 | * | |
997 | * Note, @test is called with the inode_lock held, so can't sleep. | |
998 | */ | |
5d2bea45 | 999 | static struct inode *ifind(struct super_block *sb, |
1da177e4 | 1000 | struct hlist_head *head, int (*test)(struct inode *, void *), |
88bd5121 | 1001 | void *data, const int wait) |
1da177e4 LT |
1002 | { |
1003 | struct inode *inode; | |
1004 | ||
1005 | spin_lock(&inode_lock); | |
1006 | inode = find_inode(sb, head, test, data); | |
1007 | if (inode) { | |
1da177e4 | 1008 | spin_unlock(&inode_lock); |
88bd5121 AA |
1009 | if (likely(wait)) |
1010 | wait_on_inode(inode); | |
1da177e4 LT |
1011 | return inode; |
1012 | } | |
1013 | spin_unlock(&inode_lock); | |
1014 | return NULL; | |
1015 | } | |
1016 | ||
1017 | /** | |
1018 | * ifind_fast - internal function, you want ilookup() or iget(). | |
1019 | * @sb: super block of file system to search | |
1020 | * @head: head of the list to search | |
1021 | * @ino: inode number to search for | |
1022 | * | |
1023 | * ifind_fast() searches for the inode @ino in the inode cache. This is for | |
1024 | * file systems where the inode number is sufficient for unique identification | |
1025 | * of an inode. | |
1026 | * | |
1027 | * If the inode is in the cache, the inode is returned with an incremented | |
1028 | * reference count. | |
1029 | * | |
1030 | * Otherwise NULL is returned. | |
1031 | */ | |
5d2bea45 | 1032 | static struct inode *ifind_fast(struct super_block *sb, |
1da177e4 LT |
1033 | struct hlist_head *head, unsigned long ino) |
1034 | { | |
1035 | struct inode *inode; | |
1036 | ||
1037 | spin_lock(&inode_lock); | |
1038 | inode = find_inode_fast(sb, head, ino); | |
1039 | if (inode) { | |
1da177e4 LT |
1040 | spin_unlock(&inode_lock); |
1041 | wait_on_inode(inode); | |
1042 | return inode; | |
1043 | } | |
1044 | spin_unlock(&inode_lock); | |
1045 | return NULL; | |
1046 | } | |
1047 | ||
1048 | /** | |
88bd5121 | 1049 | * ilookup5_nowait - search for an inode in the inode cache |
1da177e4 LT |
1050 | * @sb: super block of file system to search |
1051 | * @hashval: hash value (usually inode number) to search for | |
1052 | * @test: callback used for comparisons between inodes | |
1053 | * @data: opaque data pointer to pass to @test | |
1054 | * | |
1055 | * ilookup5() uses ifind() to search for the inode specified by @hashval and | |
1056 | * @data in the inode cache. This is a generalized version of ilookup() for | |
1057 | * file systems where the inode number is not sufficient for unique | |
1058 | * identification of an inode. | |
1059 | * | |
1060 | * If the inode is in the cache, the inode is returned with an incremented | |
88bd5121 AA |
1061 | * reference count. Note, the inode lock is not waited upon so you have to be |
1062 | * very careful what you do with the returned inode. You probably should be | |
1063 | * using ilookup5() instead. | |
1064 | * | |
1065 | * Otherwise NULL is returned. | |
1066 | * | |
1067 | * Note, @test is called with the inode_lock held, so can't sleep. | |
1068 | */ | |
1069 | struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval, | |
1070 | int (*test)(struct inode *, void *), void *data) | |
1071 | { | |
1072 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1073 | ||
1074 | return ifind(sb, head, test, data, 0); | |
1075 | } | |
88bd5121 AA |
1076 | EXPORT_SYMBOL(ilookup5_nowait); |
1077 | ||
1078 | /** | |
1079 | * ilookup5 - search for an inode in the inode cache | |
1080 | * @sb: super block of file system to search | |
1081 | * @hashval: hash value (usually inode number) to search for | |
1082 | * @test: callback used for comparisons between inodes | |
1083 | * @data: opaque data pointer to pass to @test | |
1084 | * | |
1085 | * ilookup5() uses ifind() to search for the inode specified by @hashval and | |
1086 | * @data in the inode cache. This is a generalized version of ilookup() for | |
1087 | * file systems where the inode number is not sufficient for unique | |
1088 | * identification of an inode. | |
1089 | * | |
1090 | * If the inode is in the cache, the inode lock is waited upon and the inode is | |
1091 | * returned with an incremented reference count. | |
1da177e4 LT |
1092 | * |
1093 | * Otherwise NULL is returned. | |
1094 | * | |
1095 | * Note, @test is called with the inode_lock held, so can't sleep. | |
1096 | */ | |
1097 | struct inode *ilookup5(struct super_block *sb, unsigned long hashval, | |
1098 | int (*test)(struct inode *, void *), void *data) | |
1099 | { | |
1100 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1101 | ||
88bd5121 | 1102 | return ifind(sb, head, test, data, 1); |
1da177e4 | 1103 | } |
1da177e4 LT |
1104 | EXPORT_SYMBOL(ilookup5); |
1105 | ||
1106 | /** | |
1107 | * ilookup - search for an inode in the inode cache | |
1108 | * @sb: super block of file system to search | |
1109 | * @ino: inode number to search for | |
1110 | * | |
1111 | * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache. | |
1112 | * This is for file systems where the inode number is sufficient for unique | |
1113 | * identification of an inode. | |
1114 | * | |
1115 | * If the inode is in the cache, the inode is returned with an incremented | |
1116 | * reference count. | |
1117 | * | |
1118 | * Otherwise NULL is returned. | |
1119 | */ | |
1120 | struct inode *ilookup(struct super_block *sb, unsigned long ino) | |
1121 | { | |
1122 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1123 | ||
1124 | return ifind_fast(sb, head, ino); | |
1125 | } | |
1da177e4 LT |
1126 | EXPORT_SYMBOL(ilookup); |
1127 | ||
1128 | /** | |
1129 | * iget5_locked - obtain an inode from a mounted file system | |
1130 | * @sb: super block of file system | |
1131 | * @hashval: hash value (usually inode number) to get | |
1132 | * @test: callback used for comparisons between inodes | |
1133 | * @set: callback used to initialize a new struct inode | |
1134 | * @data: opaque data pointer to pass to @test and @set | |
1135 | * | |
1da177e4 LT |
1136 | * iget5_locked() uses ifind() to search for the inode specified by @hashval |
1137 | * and @data in the inode cache and if present it is returned with an increased | |
1138 | * reference count. This is a generalized version of iget_locked() for file | |
1139 | * systems where the inode number is not sufficient for unique identification | |
1140 | * of an inode. | |
1141 | * | |
1142 | * If the inode is not in cache, get_new_inode() is called to allocate a new | |
1143 | * inode and this is returned locked, hashed, and with the I_NEW flag set. The | |
1144 | * file system gets to fill it in before unlocking it via unlock_new_inode(). | |
1145 | * | |
1146 | * Note both @test and @set are called with the inode_lock held, so can't sleep. | |
1147 | */ | |
1148 | struct inode *iget5_locked(struct super_block *sb, unsigned long hashval, | |
1149 | int (*test)(struct inode *, void *), | |
1150 | int (*set)(struct inode *, void *), void *data) | |
1151 | { | |
1152 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1153 | struct inode *inode; | |
1154 | ||
88bd5121 | 1155 | inode = ifind(sb, head, test, data, 1); |
1da177e4 LT |
1156 | if (inode) |
1157 | return inode; | |
1158 | /* | |
1159 | * get_new_inode() will do the right thing, re-trying the search | |
1160 | * in case it had to block at any point. | |
1161 | */ | |
1162 | return get_new_inode(sb, head, test, set, data); | |
1163 | } | |
1da177e4 LT |
1164 | EXPORT_SYMBOL(iget5_locked); |
1165 | ||
1166 | /** | |
1167 | * iget_locked - obtain an inode from a mounted file system | |
1168 | * @sb: super block of file system | |
1169 | * @ino: inode number to get | |
1170 | * | |
1da177e4 LT |
1171 | * iget_locked() uses ifind_fast() to search for the inode specified by @ino in |
1172 | * the inode cache and if present it is returned with an increased reference | |
1173 | * count. This is for file systems where the inode number is sufficient for | |
1174 | * unique identification of an inode. | |
1175 | * | |
1176 | * If the inode is not in cache, get_new_inode_fast() is called to allocate a | |
1177 | * new inode and this is returned locked, hashed, and with the I_NEW flag set. | |
1178 | * The file system gets to fill it in before unlocking it via | |
1179 | * unlock_new_inode(). | |
1180 | */ | |
1181 | struct inode *iget_locked(struct super_block *sb, unsigned long ino) | |
1182 | { | |
1183 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1184 | struct inode *inode; | |
1185 | ||
1186 | inode = ifind_fast(sb, head, ino); | |
1187 | if (inode) | |
1188 | return inode; | |
1189 | /* | |
1190 | * get_new_inode_fast() will do the right thing, re-trying the search | |
1191 | * in case it had to block at any point. | |
1192 | */ | |
1193 | return get_new_inode_fast(sb, head, ino); | |
1194 | } | |
1da177e4 LT |
1195 | EXPORT_SYMBOL(iget_locked); |
1196 | ||
261bca86 AV |
1197 | int insert_inode_locked(struct inode *inode) |
1198 | { | |
1199 | struct super_block *sb = inode->i_sb; | |
1200 | ino_t ino = inode->i_ino; | |
1201 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
261bca86 | 1202 | |
eaff8079 | 1203 | inode->i_state |= I_NEW; |
261bca86 | 1204 | while (1) { |
72a43d63 AV |
1205 | struct hlist_node *node; |
1206 | struct inode *old = NULL; | |
261bca86 | 1207 | spin_lock(&inode_lock); |
72a43d63 AV |
1208 | hlist_for_each_entry(old, node, head, i_hash) { |
1209 | if (old->i_ino != ino) | |
1210 | continue; | |
1211 | if (old->i_sb != sb) | |
1212 | continue; | |
a4ffdde6 | 1213 | if (old->i_state & (I_FREEING|I_WILL_FREE)) |
72a43d63 AV |
1214 | continue; |
1215 | break; | |
1216 | } | |
1217 | if (likely(!node)) { | |
261bca86 AV |
1218 | hlist_add_head(&inode->i_hash, head); |
1219 | spin_unlock(&inode_lock); | |
1220 | return 0; | |
1221 | } | |
1222 | __iget(old); | |
1223 | spin_unlock(&inode_lock); | |
1224 | wait_on_inode(old); | |
1d3382cb | 1225 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1226 | iput(old); |
1227 | return -EBUSY; | |
1228 | } | |
1229 | iput(old); | |
1230 | } | |
1231 | } | |
261bca86 AV |
1232 | EXPORT_SYMBOL(insert_inode_locked); |
1233 | ||
1234 | int insert_inode_locked4(struct inode *inode, unsigned long hashval, | |
1235 | int (*test)(struct inode *, void *), void *data) | |
1236 | { | |
1237 | struct super_block *sb = inode->i_sb; | |
1238 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
261bca86 | 1239 | |
eaff8079 | 1240 | inode->i_state |= I_NEW; |
261bca86 AV |
1241 | |
1242 | while (1) { | |
72a43d63 AV |
1243 | struct hlist_node *node; |
1244 | struct inode *old = NULL; | |
1245 | ||
261bca86 | 1246 | spin_lock(&inode_lock); |
72a43d63 AV |
1247 | hlist_for_each_entry(old, node, head, i_hash) { |
1248 | if (old->i_sb != sb) | |
1249 | continue; | |
1250 | if (!test(old, data)) | |
1251 | continue; | |
a4ffdde6 | 1252 | if (old->i_state & (I_FREEING|I_WILL_FREE)) |
72a43d63 AV |
1253 | continue; |
1254 | break; | |
1255 | } | |
1256 | if (likely(!node)) { | |
261bca86 AV |
1257 | hlist_add_head(&inode->i_hash, head); |
1258 | spin_unlock(&inode_lock); | |
1259 | return 0; | |
1260 | } | |
1261 | __iget(old); | |
1262 | spin_unlock(&inode_lock); | |
1263 | wait_on_inode(old); | |
1d3382cb | 1264 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1265 | iput(old); |
1266 | return -EBUSY; | |
1267 | } | |
1268 | iput(old); | |
1269 | } | |
1270 | } | |
261bca86 AV |
1271 | EXPORT_SYMBOL(insert_inode_locked4); |
1272 | ||
1da177e4 | 1273 | |
45321ac5 AV |
1274 | int generic_delete_inode(struct inode *inode) |
1275 | { | |
1276 | return 1; | |
1277 | } | |
1278 | EXPORT_SYMBOL(generic_delete_inode); | |
1279 | ||
1da177e4 | 1280 | /* |
45321ac5 AV |
1281 | * Normal UNIX filesystem behaviour: delete the |
1282 | * inode when the usage count drops to zero, and | |
1283 | * i_nlink is zero. | |
1da177e4 | 1284 | */ |
45321ac5 | 1285 | int generic_drop_inode(struct inode *inode) |
1da177e4 | 1286 | { |
1d3382cb | 1287 | return !inode->i_nlink || inode_unhashed(inode); |
1da177e4 | 1288 | } |
45321ac5 | 1289 | EXPORT_SYMBOL_GPL(generic_drop_inode); |
1da177e4 | 1290 | |
45321ac5 AV |
1291 | /* |
1292 | * Called when we're dropping the last reference | |
1293 | * to an inode. | |
22fe4042 | 1294 | * |
45321ac5 AV |
1295 | * Call the FS "drop_inode()" function, defaulting to |
1296 | * the legacy UNIX filesystem behaviour. If it tells | |
1297 | * us to evict inode, do so. Otherwise, retain inode | |
1298 | * in cache if fs is alive, sync and evict if fs is | |
1299 | * shutting down. | |
22fe4042 | 1300 | */ |
45321ac5 | 1301 | static void iput_final(struct inode *inode) |
1da177e4 LT |
1302 | { |
1303 | struct super_block *sb = inode->i_sb; | |
45321ac5 AV |
1304 | const struct super_operations *op = inode->i_sb->s_op; |
1305 | int drop; | |
1306 | ||
1307 | if (op && op->drop_inode) | |
1308 | drop = op->drop_inode(inode); | |
1309 | else | |
1310 | drop = generic_drop_inode(inode); | |
1da177e4 | 1311 | |
45321ac5 | 1312 | if (!drop) { |
acb0c854 | 1313 | if (sb->s_flags & MS_ACTIVE) { |
9e38d86f NP |
1314 | inode->i_state |= I_REFERENCED; |
1315 | if (!(inode->i_state & (I_DIRTY|I_SYNC))) { | |
1316 | inode_lru_list_add(inode); | |
1317 | } | |
991114c6 | 1318 | spin_unlock(&inode_lock); |
45321ac5 | 1319 | return; |
991114c6 | 1320 | } |
7ef0d737 | 1321 | WARN_ON(inode->i_state & I_NEW); |
991114c6 AV |
1322 | inode->i_state |= I_WILL_FREE; |
1323 | spin_unlock(&inode_lock); | |
1da177e4 LT |
1324 | write_inode_now(inode, 1); |
1325 | spin_lock(&inode_lock); | |
7ef0d737 | 1326 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1327 | inode->i_state &= ~I_WILL_FREE; |
4c51acbc | 1328 | __remove_inode_hash(inode); |
1da177e4 | 1329 | } |
7ccf19a8 | 1330 | |
7ef0d737 | 1331 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1332 | inode->i_state |= I_FREEING; |
9e38d86f NP |
1333 | |
1334 | /* | |
7ccf19a8 NP |
1335 | * Move the inode off the IO lists and LRU once I_FREEING is |
1336 | * set so that it won't get moved back on there if it is dirty. | |
9e38d86f NP |
1337 | */ |
1338 | inode_lru_list_del(inode); | |
7ccf19a8 | 1339 | list_del_init(&inode->i_wb_list); |
9e38d86f | 1340 | |
646ec461 | 1341 | __inode_sb_list_del(inode); |
1da177e4 | 1342 | spin_unlock(&inode_lock); |
644da596 | 1343 | evict(inode); |
4c51acbc | 1344 | remove_inode_hash(inode); |
7f04c26d | 1345 | wake_up_inode(inode); |
45321ac5 | 1346 | BUG_ON(inode->i_state != (I_FREEING | I_CLEAR)); |
1da177e4 LT |
1347 | destroy_inode(inode); |
1348 | } | |
1349 | ||
1da177e4 | 1350 | /** |
6b3304b5 | 1351 | * iput - put an inode |
1da177e4 LT |
1352 | * @inode: inode to put |
1353 | * | |
1354 | * Puts an inode, dropping its usage count. If the inode use count hits | |
1355 | * zero, the inode is then freed and may also be destroyed. | |
1356 | * | |
1357 | * Consequently, iput() can sleep. | |
1358 | */ | |
1359 | void iput(struct inode *inode) | |
1360 | { | |
1361 | if (inode) { | |
a4ffdde6 | 1362 | BUG_ON(inode->i_state & I_CLEAR); |
1da177e4 | 1363 | |
1da177e4 LT |
1364 | if (atomic_dec_and_lock(&inode->i_count, &inode_lock)) |
1365 | iput_final(inode); | |
1366 | } | |
1367 | } | |
1da177e4 LT |
1368 | EXPORT_SYMBOL(iput); |
1369 | ||
1370 | /** | |
1371 | * bmap - find a block number in a file | |
1372 | * @inode: inode of file | |
1373 | * @block: block to find | |
1374 | * | |
1375 | * Returns the block number on the device holding the inode that | |
1376 | * is the disk block number for the block of the file requested. | |
1377 | * That is, asked for block 4 of inode 1 the function will return the | |
6b3304b5 | 1378 | * disk block relative to the disk start that holds that block of the |
1da177e4 LT |
1379 | * file. |
1380 | */ | |
6b3304b5 | 1381 | sector_t bmap(struct inode *inode, sector_t block) |
1da177e4 LT |
1382 | { |
1383 | sector_t res = 0; | |
1384 | if (inode->i_mapping->a_ops->bmap) | |
1385 | res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block); | |
1386 | return res; | |
1387 | } | |
1da177e4 LT |
1388 | EXPORT_SYMBOL(bmap); |
1389 | ||
11ff6f05 MG |
1390 | /* |
1391 | * With relative atime, only update atime if the previous atime is | |
1392 | * earlier than either the ctime or mtime or if at least a day has | |
1393 | * passed since the last atime update. | |
1394 | */ | |
1395 | static int relatime_need_update(struct vfsmount *mnt, struct inode *inode, | |
1396 | struct timespec now) | |
1397 | { | |
1398 | ||
1399 | if (!(mnt->mnt_flags & MNT_RELATIME)) | |
1400 | return 1; | |
1401 | /* | |
1402 | * Is mtime younger than atime? If yes, update atime: | |
1403 | */ | |
1404 | if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0) | |
1405 | return 1; | |
1406 | /* | |
1407 | * Is ctime younger than atime? If yes, update atime: | |
1408 | */ | |
1409 | if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0) | |
1410 | return 1; | |
1411 | ||
1412 | /* | |
1413 | * Is the previous atime value older than a day? If yes, | |
1414 | * update atime: | |
1415 | */ | |
1416 | if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60) | |
1417 | return 1; | |
1418 | /* | |
1419 | * Good, we can skip the atime update: | |
1420 | */ | |
1421 | return 0; | |
1422 | } | |
1423 | ||
1da177e4 | 1424 | /** |
869243a0 CH |
1425 | * touch_atime - update the access time |
1426 | * @mnt: mount the inode is accessed on | |
7045f37b | 1427 | * @dentry: dentry accessed |
1da177e4 LT |
1428 | * |
1429 | * Update the accessed time on an inode and mark it for writeback. | |
1430 | * This function automatically handles read only file systems and media, | |
1431 | * as well as the "noatime" flag and inode specific "noatime" markers. | |
1432 | */ | |
869243a0 | 1433 | void touch_atime(struct vfsmount *mnt, struct dentry *dentry) |
1da177e4 | 1434 | { |
869243a0 | 1435 | struct inode *inode = dentry->d_inode; |
1da177e4 LT |
1436 | struct timespec now; |
1437 | ||
cdb70f3f | 1438 | if (inode->i_flags & S_NOATIME) |
b12536c2 | 1439 | return; |
37756ced | 1440 | if (IS_NOATIME(inode)) |
b12536c2 | 1441 | return; |
b2276138 | 1442 | if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1443 | return; |
47ae32d6 | 1444 | |
cdb70f3f | 1445 | if (mnt->mnt_flags & MNT_NOATIME) |
b12536c2 | 1446 | return; |
cdb70f3f | 1447 | if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1448 | return; |
1da177e4 LT |
1449 | |
1450 | now = current_fs_time(inode->i_sb); | |
11ff6f05 MG |
1451 | |
1452 | if (!relatime_need_update(mnt, inode, now)) | |
b12536c2 | 1453 | return; |
11ff6f05 | 1454 | |
47ae32d6 | 1455 | if (timespec_equal(&inode->i_atime, &now)) |
b12536c2 AK |
1456 | return; |
1457 | ||
1458 | if (mnt_want_write(mnt)) | |
1459 | return; | |
47ae32d6 VH |
1460 | |
1461 | inode->i_atime = now; | |
1462 | mark_inode_dirty_sync(inode); | |
cdb70f3f | 1463 | mnt_drop_write(mnt); |
1da177e4 | 1464 | } |
869243a0 | 1465 | EXPORT_SYMBOL(touch_atime); |
1da177e4 LT |
1466 | |
1467 | /** | |
870f4817 CH |
1468 | * file_update_time - update mtime and ctime time |
1469 | * @file: file accessed | |
1da177e4 | 1470 | * |
870f4817 CH |
1471 | * Update the mtime and ctime members of an inode and mark the inode |
1472 | * for writeback. Note that this function is meant exclusively for | |
1473 | * usage in the file write path of filesystems, and filesystems may | |
1474 | * choose to explicitly ignore update via this function with the | |
2eadfc0e | 1475 | * S_NOCMTIME inode flag, e.g. for network filesystem where these |
870f4817 | 1476 | * timestamps are handled by the server. |
1da177e4 LT |
1477 | */ |
1478 | ||
870f4817 | 1479 | void file_update_time(struct file *file) |
1da177e4 | 1480 | { |
0f7fc9e4 | 1481 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 | 1482 | struct timespec now; |
ce06e0b2 | 1483 | enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0; |
1da177e4 | 1484 | |
ce06e0b2 | 1485 | /* First try to exhaust all avenues to not sync */ |
1da177e4 LT |
1486 | if (IS_NOCMTIME(inode)) |
1487 | return; | |
20ddee2c | 1488 | |
1da177e4 | 1489 | now = current_fs_time(inode->i_sb); |
ce06e0b2 AK |
1490 | if (!timespec_equal(&inode->i_mtime, &now)) |
1491 | sync_it = S_MTIME; | |
1da177e4 | 1492 | |
ce06e0b2 AK |
1493 | if (!timespec_equal(&inode->i_ctime, &now)) |
1494 | sync_it |= S_CTIME; | |
870f4817 | 1495 | |
ce06e0b2 AK |
1496 | if (IS_I_VERSION(inode)) |
1497 | sync_it |= S_VERSION; | |
7a224228 | 1498 | |
ce06e0b2 AK |
1499 | if (!sync_it) |
1500 | return; | |
1501 | ||
1502 | /* Finally allowed to write? Takes lock. */ | |
1503 | if (mnt_want_write_file(file)) | |
1504 | return; | |
1505 | ||
1506 | /* Only change inode inside the lock region */ | |
1507 | if (sync_it & S_VERSION) | |
1508 | inode_inc_iversion(inode); | |
1509 | if (sync_it & S_CTIME) | |
1510 | inode->i_ctime = now; | |
1511 | if (sync_it & S_MTIME) | |
1512 | inode->i_mtime = now; | |
1513 | mark_inode_dirty_sync(inode); | |
20ddee2c | 1514 | mnt_drop_write(file->f_path.mnt); |
1da177e4 | 1515 | } |
870f4817 | 1516 | EXPORT_SYMBOL(file_update_time); |
1da177e4 LT |
1517 | |
1518 | int inode_needs_sync(struct inode *inode) | |
1519 | { | |
1520 | if (IS_SYNC(inode)) | |
1521 | return 1; | |
1522 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) | |
1523 | return 1; | |
1524 | return 0; | |
1525 | } | |
1da177e4 LT |
1526 | EXPORT_SYMBOL(inode_needs_sync); |
1527 | ||
1da177e4 LT |
1528 | int inode_wait(void *word) |
1529 | { | |
1530 | schedule(); | |
1531 | return 0; | |
1532 | } | |
d44dab8d | 1533 | EXPORT_SYMBOL(inode_wait); |
1da177e4 LT |
1534 | |
1535 | /* | |
168a9fd6 MS |
1536 | * If we try to find an inode in the inode hash while it is being |
1537 | * deleted, we have to wait until the filesystem completes its | |
1538 | * deletion before reporting that it isn't found. This function waits | |
1539 | * until the deletion _might_ have completed. Callers are responsible | |
1540 | * to recheck inode state. | |
1541 | * | |
eaff8079 | 1542 | * It doesn't matter if I_NEW is not set initially, a call to |
168a9fd6 MS |
1543 | * wake_up_inode() after removing from the hash list will DTRT. |
1544 | * | |
1da177e4 LT |
1545 | * This is called with inode_lock held. |
1546 | */ | |
1547 | static void __wait_on_freeing_inode(struct inode *inode) | |
1548 | { | |
1549 | wait_queue_head_t *wq; | |
eaff8079 CH |
1550 | DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW); |
1551 | wq = bit_waitqueue(&inode->i_state, __I_NEW); | |
1da177e4 LT |
1552 | prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE); |
1553 | spin_unlock(&inode_lock); | |
1554 | schedule(); | |
1555 | finish_wait(wq, &wait.wait); | |
1556 | spin_lock(&inode_lock); | |
1557 | } | |
1558 | ||
1da177e4 LT |
1559 | static __initdata unsigned long ihash_entries; |
1560 | static int __init set_ihash_entries(char *str) | |
1561 | { | |
1562 | if (!str) | |
1563 | return 0; | |
1564 | ihash_entries = simple_strtoul(str, &str, 0); | |
1565 | return 1; | |
1566 | } | |
1567 | __setup("ihash_entries=", set_ihash_entries); | |
1568 | ||
1569 | /* | |
1570 | * Initialize the waitqueues and inode hash table. | |
1571 | */ | |
1572 | void __init inode_init_early(void) | |
1573 | { | |
1574 | int loop; | |
1575 | ||
1576 | /* If hashes are distributed across NUMA nodes, defer | |
1577 | * hash allocation until vmalloc space is available. | |
1578 | */ | |
1579 | if (hashdist) | |
1580 | return; | |
1581 | ||
1582 | inode_hashtable = | |
1583 | alloc_large_system_hash("Inode-cache", | |
1584 | sizeof(struct hlist_head), | |
1585 | ihash_entries, | |
1586 | 14, | |
1587 | HASH_EARLY, | |
1588 | &i_hash_shift, | |
1589 | &i_hash_mask, | |
1590 | 0); | |
1591 | ||
1592 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1593 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1594 | } | |
1595 | ||
74bf17cf | 1596 | void __init inode_init(void) |
1da177e4 LT |
1597 | { |
1598 | int loop; | |
1599 | ||
1600 | /* inode slab cache */ | |
b0196009 PJ |
1601 | inode_cachep = kmem_cache_create("inode_cache", |
1602 | sizeof(struct inode), | |
1603 | 0, | |
1604 | (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC| | |
1605 | SLAB_MEM_SPREAD), | |
20c2df83 | 1606 | init_once); |
8e1f936b | 1607 | register_shrinker(&icache_shrinker); |
cffbc8aa DC |
1608 | percpu_counter_init(&nr_inodes, 0); |
1609 | percpu_counter_init(&nr_inodes_unused, 0); | |
1da177e4 LT |
1610 | |
1611 | /* Hash may have been set up in inode_init_early */ | |
1612 | if (!hashdist) | |
1613 | return; | |
1614 | ||
1615 | inode_hashtable = | |
1616 | alloc_large_system_hash("Inode-cache", | |
1617 | sizeof(struct hlist_head), | |
1618 | ihash_entries, | |
1619 | 14, | |
1620 | 0, | |
1621 | &i_hash_shift, | |
1622 | &i_hash_mask, | |
1623 | 0); | |
1624 | ||
1625 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1626 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1627 | } | |
1628 | ||
1629 | void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev) | |
1630 | { | |
1631 | inode->i_mode = mode; | |
1632 | if (S_ISCHR(mode)) { | |
1633 | inode->i_fop = &def_chr_fops; | |
1634 | inode->i_rdev = rdev; | |
1635 | } else if (S_ISBLK(mode)) { | |
1636 | inode->i_fop = &def_blk_fops; | |
1637 | inode->i_rdev = rdev; | |
1638 | } else if (S_ISFIFO(mode)) | |
1639 | inode->i_fop = &def_fifo_fops; | |
1640 | else if (S_ISSOCK(mode)) | |
1641 | inode->i_fop = &bad_sock_fops; | |
1642 | else | |
af0d9ae8 MK |
1643 | printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for" |
1644 | " inode %s:%lu\n", mode, inode->i_sb->s_id, | |
1645 | inode->i_ino); | |
1da177e4 LT |
1646 | } |
1647 | EXPORT_SYMBOL(init_special_inode); | |
a1bd120d DM |
1648 | |
1649 | /** | |
1650 | * Init uid,gid,mode for new inode according to posix standards | |
1651 | * @inode: New inode | |
1652 | * @dir: Directory inode | |
1653 | * @mode: mode of the new inode | |
1654 | */ | |
1655 | void inode_init_owner(struct inode *inode, const struct inode *dir, | |
1656 | mode_t mode) | |
1657 | { | |
1658 | inode->i_uid = current_fsuid(); | |
1659 | if (dir && dir->i_mode & S_ISGID) { | |
1660 | inode->i_gid = dir->i_gid; | |
1661 | if (S_ISDIR(mode)) | |
1662 | mode |= S_ISGID; | |
1663 | } else | |
1664 | inode->i_gid = current_fsgid(); | |
1665 | inode->i_mode = mode; | |
1666 | } | |
1667 | EXPORT_SYMBOL(inode_init_owner); |