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> | |
11 | #include <linux/quotaops.h> | |
12 | #include <linux/slab.h> | |
13 | #include <linux/writeback.h> | |
14 | #include <linux/module.h> | |
15 | #include <linux/backing-dev.h> | |
16 | #include <linux/wait.h> | |
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> | |
0eeca283 | 23 | #include <linux/inotify.h> |
fc33a7bb | 24 | #include <linux/mount.h> |
1da177e4 LT |
25 | |
26 | /* | |
27 | * This is needed for the following functions: | |
28 | * - inode_has_buffers | |
29 | * - invalidate_inode_buffers | |
1da177e4 LT |
30 | * - invalidate_bdev |
31 | * | |
32 | * FIXME: remove all knowledge of the buffer layer from this file | |
33 | */ | |
34 | #include <linux/buffer_head.h> | |
35 | ||
36 | /* | |
37 | * New inode.c implementation. | |
38 | * | |
39 | * This implementation has the basic premise of trying | |
40 | * to be extremely low-overhead and SMP-safe, yet be | |
41 | * simple enough to be "obviously correct". | |
42 | * | |
43 | * Famous last words. | |
44 | */ | |
45 | ||
46 | /* inode dynamic allocation 1999, Andrea Arcangeli <andrea@suse.de> */ | |
47 | ||
48 | /* #define INODE_PARANOIA 1 */ | |
49 | /* #define INODE_DEBUG 1 */ | |
50 | ||
51 | /* | |
52 | * Inode lookup is no longer as critical as it used to be: | |
53 | * most of the lookups are going to be through the dcache. | |
54 | */ | |
55 | #define I_HASHBITS i_hash_shift | |
56 | #define I_HASHMASK i_hash_mask | |
57 | ||
fa3536cc ED |
58 | static unsigned int i_hash_mask __read_mostly; |
59 | static unsigned int i_hash_shift __read_mostly; | |
1da177e4 LT |
60 | |
61 | /* | |
62 | * Each inode can be on two separate lists. One is | |
63 | * the hash list of the inode, used for lookups. The | |
64 | * other linked list is the "type" list: | |
65 | * "in_use" - valid inode, i_count > 0, i_nlink > 0 | |
66 | * "dirty" - as "in_use" but also dirty | |
67 | * "unused" - valid inode, i_count = 0 | |
68 | * | |
69 | * A "dirty" list is maintained for each super block, | |
70 | * allowing for low-overhead inode sync() operations. | |
71 | */ | |
72 | ||
73 | LIST_HEAD(inode_in_use); | |
74 | LIST_HEAD(inode_unused); | |
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 | /* | |
f24075bd | 86 | * iprune_mutex 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. | |
92 | */ | |
bdfc3266 | 93 | static DEFINE_MUTEX(iprune_mutex); |
1da177e4 LT |
94 | |
95 | /* | |
96 | * Statistics gathering.. | |
97 | */ | |
98 | struct inodes_stat_t inodes_stat; | |
99 | ||
e18b890b | 100 | static struct kmem_cache * inode_cachep __read_mostly; |
1da177e4 LT |
101 | |
102 | static struct inode *alloc_inode(struct super_block *sb) | |
103 | { | |
f5e54d6e | 104 | static const struct address_space_operations empty_aops; |
1da177e4 | 105 | static struct inode_operations empty_iops; |
99ac48f5 | 106 | static const struct file_operations empty_fops; |
1da177e4 LT |
107 | struct inode *inode; |
108 | ||
109 | if (sb->s_op->alloc_inode) | |
110 | inode = sb->s_op->alloc_inode(sb); | |
111 | else | |
e94b1766 | 112 | inode = (struct inode *) kmem_cache_alloc(inode_cachep, GFP_KERNEL); |
1da177e4 LT |
113 | |
114 | if (inode) { | |
115 | struct address_space * const mapping = &inode->i_data; | |
116 | ||
117 | inode->i_sb = sb; | |
118 | inode->i_blkbits = sb->s_blocksize_bits; | |
119 | inode->i_flags = 0; | |
120 | atomic_set(&inode->i_count, 1); | |
121 | inode->i_op = &empty_iops; | |
122 | inode->i_fop = &empty_fops; | |
123 | inode->i_nlink = 1; | |
124 | atomic_set(&inode->i_writecount, 0); | |
125 | inode->i_size = 0; | |
126 | inode->i_blocks = 0; | |
127 | inode->i_bytes = 0; | |
128 | inode->i_generation = 0; | |
129 | #ifdef CONFIG_QUOTA | |
130 | memset(&inode->i_dquot, 0, sizeof(inode->i_dquot)); | |
131 | #endif | |
132 | inode->i_pipe = NULL; | |
133 | inode->i_bdev = NULL; | |
134 | inode->i_cdev = NULL; | |
135 | inode->i_rdev = 0; | |
1da177e4 LT |
136 | inode->dirtied_when = 0; |
137 | if (security_inode_alloc(inode)) { | |
138 | if (inode->i_sb->s_op->destroy_inode) | |
139 | inode->i_sb->s_op->destroy_inode(inode); | |
140 | else | |
141 | kmem_cache_free(inode_cachep, (inode)); | |
142 | return NULL; | |
143 | } | |
144 | ||
145 | mapping->a_ops = &empty_aops; | |
146 | mapping->host = inode; | |
147 | mapping->flags = 0; | |
148 | mapping_set_gfp_mask(mapping, GFP_HIGHUSER); | |
149 | mapping->assoc_mapping = NULL; | |
150 | mapping->backing_dev_info = &default_backing_dev_info; | |
151 | ||
152 | /* | |
153 | * If the block_device provides a backing_dev_info for client | |
154 | * inodes then use that. Otherwise the inode share the bdev's | |
155 | * backing_dev_info. | |
156 | */ | |
157 | if (sb->s_bdev) { | |
158 | struct backing_dev_info *bdi; | |
159 | ||
160 | bdi = sb->s_bdev->bd_inode_backing_dev_info; | |
161 | if (!bdi) | |
162 | bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info; | |
163 | mapping->backing_dev_info = bdi; | |
164 | } | |
e6c6e640 | 165 | inode->i_private = NULL; |
1da177e4 LT |
166 | inode->i_mapping = mapping; |
167 | } | |
168 | return inode; | |
169 | } | |
170 | ||
171 | void destroy_inode(struct inode *inode) | |
172 | { | |
b7542f8c | 173 | BUG_ON(inode_has_buffers(inode)); |
1da177e4 LT |
174 | security_inode_free(inode); |
175 | if (inode->i_sb->s_op->destroy_inode) | |
176 | inode->i_sb->s_op->destroy_inode(inode); | |
177 | else | |
178 | kmem_cache_free(inode_cachep, (inode)); | |
179 | } | |
180 | ||
181 | ||
182 | /* | |
183 | * These are initializations that only need to be done | |
184 | * once, because the fields are idempotent across use | |
185 | * of the inode, so let the slab aware of that. | |
186 | */ | |
187 | void inode_init_once(struct inode *inode) | |
188 | { | |
189 | memset(inode, 0, sizeof(*inode)); | |
190 | INIT_HLIST_NODE(&inode->i_hash); | |
191 | INIT_LIST_HEAD(&inode->i_dentry); | |
192 | INIT_LIST_HEAD(&inode->i_devices); | |
1b1dcc1b | 193 | mutex_init(&inode->i_mutex); |
1da177e4 LT |
194 | init_rwsem(&inode->i_alloc_sem); |
195 | INIT_RADIX_TREE(&inode->i_data.page_tree, GFP_ATOMIC); | |
196 | rwlock_init(&inode->i_data.tree_lock); | |
197 | spin_lock_init(&inode->i_data.i_mmap_lock); | |
198 | INIT_LIST_HEAD(&inode->i_data.private_list); | |
199 | spin_lock_init(&inode->i_data.private_lock); | |
200 | INIT_RAW_PRIO_TREE_ROOT(&inode->i_data.i_mmap); | |
201 | INIT_LIST_HEAD(&inode->i_data.i_mmap_nonlinear); | |
202 | spin_lock_init(&inode->i_lock); | |
203 | i_size_ordered_init(inode); | |
0eeca283 RL |
204 | #ifdef CONFIG_INOTIFY |
205 | INIT_LIST_HEAD(&inode->inotify_watches); | |
d4f9af9d | 206 | mutex_init(&inode->inotify_mutex); |
0eeca283 | 207 | #endif |
1da177e4 LT |
208 | } |
209 | ||
210 | EXPORT_SYMBOL(inode_init_once); | |
211 | ||
e18b890b | 212 | static void init_once(void * foo, struct kmem_cache * cachep, unsigned long flags) |
1da177e4 LT |
213 | { |
214 | struct inode * inode = (struct inode *) foo; | |
215 | ||
216 | if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) == | |
217 | SLAB_CTOR_CONSTRUCTOR) | |
218 | inode_init_once(inode); | |
219 | } | |
220 | ||
221 | /* | |
222 | * inode_lock must be held | |
223 | */ | |
224 | void __iget(struct inode * inode) | |
225 | { | |
226 | if (atomic_read(&inode->i_count)) { | |
227 | atomic_inc(&inode->i_count); | |
228 | return; | |
229 | } | |
230 | atomic_inc(&inode->i_count); | |
231 | if (!(inode->i_state & (I_DIRTY|I_LOCK))) | |
232 | list_move(&inode->i_list, &inode_in_use); | |
233 | inodes_stat.nr_unused--; | |
234 | } | |
235 | ||
236 | /** | |
237 | * clear_inode - clear an inode | |
238 | * @inode: inode to clear | |
239 | * | |
240 | * This is called by the filesystem to tell us | |
241 | * that the inode is no longer useful. We just | |
242 | * terminate it with extreme prejudice. | |
243 | */ | |
244 | void clear_inode(struct inode *inode) | |
245 | { | |
246 | might_sleep(); | |
247 | invalidate_inode_buffers(inode); | |
248 | ||
b7542f8c ES |
249 | BUG_ON(inode->i_data.nrpages); |
250 | BUG_ON(!(inode->i_state & I_FREEING)); | |
251 | BUG_ON(inode->i_state & I_CLEAR); | |
1da177e4 LT |
252 | wait_on_inode(inode); |
253 | DQUOT_DROP(inode); | |
254 | if (inode->i_sb && inode->i_sb->s_op->clear_inode) | |
255 | inode->i_sb->s_op->clear_inode(inode); | |
eaf796e7 | 256 | if (S_ISBLK(inode->i_mode) && inode->i_bdev) |
1da177e4 | 257 | bd_forget(inode); |
577c4eb0 | 258 | if (S_ISCHR(inode->i_mode) && inode->i_cdev) |
1da177e4 LT |
259 | cd_forget(inode); |
260 | inode->i_state = I_CLEAR; | |
261 | } | |
262 | ||
263 | EXPORT_SYMBOL(clear_inode); | |
264 | ||
265 | /* | |
266 | * dispose_list - dispose of the contents of a local list | |
267 | * @head: the head of the list to free | |
268 | * | |
269 | * Dispose-list gets a local list with local inodes in it, so it doesn't | |
270 | * need to worry about list corruption and SMP locks. | |
271 | */ | |
272 | static void dispose_list(struct list_head *head) | |
273 | { | |
274 | int nr_disposed = 0; | |
275 | ||
276 | while (!list_empty(head)) { | |
277 | struct inode *inode; | |
278 | ||
279 | inode = list_entry(head->next, struct inode, i_list); | |
280 | list_del(&inode->i_list); | |
281 | ||
282 | if (inode->i_data.nrpages) | |
283 | truncate_inode_pages(&inode->i_data, 0); | |
284 | clear_inode(inode); | |
4120db47 AB |
285 | |
286 | spin_lock(&inode_lock); | |
287 | hlist_del_init(&inode->i_hash); | |
288 | list_del_init(&inode->i_sb_list); | |
289 | spin_unlock(&inode_lock); | |
290 | ||
291 | wake_up_inode(inode); | |
1da177e4 LT |
292 | destroy_inode(inode); |
293 | nr_disposed++; | |
294 | } | |
295 | spin_lock(&inode_lock); | |
296 | inodes_stat.nr_inodes -= nr_disposed; | |
297 | spin_unlock(&inode_lock); | |
298 | } | |
299 | ||
300 | /* | |
301 | * Invalidate all inodes for a device. | |
302 | */ | |
303 | static int invalidate_list(struct list_head *head, struct list_head *dispose) | |
304 | { | |
305 | struct list_head *next; | |
306 | int busy = 0, count = 0; | |
307 | ||
308 | next = head->next; | |
309 | for (;;) { | |
310 | struct list_head * tmp = next; | |
311 | struct inode * inode; | |
312 | ||
313 | /* | |
314 | * We can reschedule here without worrying about the list's | |
315 | * consistency because the per-sb list of inodes must not | |
f24075bd | 316 | * change during umount anymore, and because iprune_mutex keeps |
1da177e4 LT |
317 | * shrink_icache_memory() away. |
318 | */ | |
319 | cond_resched_lock(&inode_lock); | |
320 | ||
321 | next = next->next; | |
322 | if (tmp == head) | |
323 | break; | |
324 | inode = list_entry(tmp, struct inode, i_sb_list); | |
325 | invalidate_inode_buffers(inode); | |
326 | if (!atomic_read(&inode->i_count)) { | |
1da177e4 LT |
327 | list_move(&inode->i_list, dispose); |
328 | inode->i_state |= I_FREEING; | |
329 | count++; | |
330 | continue; | |
331 | } | |
332 | busy = 1; | |
333 | } | |
334 | /* only unused inodes may be cached with i_count zero */ | |
335 | inodes_stat.nr_unused -= count; | |
336 | return busy; | |
337 | } | |
338 | ||
1da177e4 LT |
339 | /** |
340 | * invalidate_inodes - discard the inodes on a device | |
341 | * @sb: superblock | |
342 | * | |
343 | * Discard all of the inodes for a given superblock. If the discard | |
344 | * fails because there are busy inodes then a non zero value is returned. | |
345 | * If the discard is successful all the inodes have been discarded. | |
346 | */ | |
347 | int invalidate_inodes(struct super_block * sb) | |
348 | { | |
349 | int busy; | |
350 | LIST_HEAD(throw_away); | |
351 | ||
f24075bd | 352 | mutex_lock(&iprune_mutex); |
1da177e4 | 353 | spin_lock(&inode_lock); |
0eeca283 | 354 | inotify_unmount_inodes(&sb->s_inodes); |
1da177e4 LT |
355 | busy = invalidate_list(&sb->s_inodes, &throw_away); |
356 | spin_unlock(&inode_lock); | |
357 | ||
358 | dispose_list(&throw_away); | |
f24075bd | 359 | mutex_unlock(&iprune_mutex); |
1da177e4 LT |
360 | |
361 | return busy; | |
362 | } | |
363 | ||
364 | EXPORT_SYMBOL(invalidate_inodes); | |
1da177e4 LT |
365 | |
366 | static int can_unuse(struct inode *inode) | |
367 | { | |
368 | if (inode->i_state) | |
369 | return 0; | |
370 | if (inode_has_buffers(inode)) | |
371 | return 0; | |
372 | if (atomic_read(&inode->i_count)) | |
373 | return 0; | |
374 | if (inode->i_data.nrpages) | |
375 | return 0; | |
376 | return 1; | |
377 | } | |
378 | ||
379 | /* | |
380 | * Scan `goal' inodes on the unused list for freeable ones. They are moved to | |
381 | * a temporary list and then are freed outside inode_lock by dispose_list(). | |
382 | * | |
383 | * Any inodes which are pinned purely because of attached pagecache have their | |
384 | * pagecache removed. We expect the final iput() on that inode to add it to | |
385 | * the front of the inode_unused list. So look for it there and if the | |
386 | * inode is still freeable, proceed. The right inode is found 99.9% of the | |
387 | * time in testing on a 4-way. | |
388 | * | |
389 | * If the inode has metadata buffers attached to mapping->private_list then | |
390 | * try to remove them. | |
391 | */ | |
392 | static void prune_icache(int nr_to_scan) | |
393 | { | |
394 | LIST_HEAD(freeable); | |
395 | int nr_pruned = 0; | |
396 | int nr_scanned; | |
397 | unsigned long reap = 0; | |
398 | ||
f24075bd | 399 | mutex_lock(&iprune_mutex); |
1da177e4 LT |
400 | spin_lock(&inode_lock); |
401 | for (nr_scanned = 0; nr_scanned < nr_to_scan; nr_scanned++) { | |
402 | struct inode *inode; | |
403 | ||
404 | if (list_empty(&inode_unused)) | |
405 | break; | |
406 | ||
407 | inode = list_entry(inode_unused.prev, struct inode, i_list); | |
408 | ||
409 | if (inode->i_state || atomic_read(&inode->i_count)) { | |
410 | list_move(&inode->i_list, &inode_unused); | |
411 | continue; | |
412 | } | |
413 | if (inode_has_buffers(inode) || inode->i_data.nrpages) { | |
414 | __iget(inode); | |
415 | spin_unlock(&inode_lock); | |
416 | if (remove_inode_buffers(inode)) | |
fc0ecff6 AM |
417 | reap += invalidate_mapping_pages(&inode->i_data, |
418 | 0, -1); | |
1da177e4 LT |
419 | iput(inode); |
420 | spin_lock(&inode_lock); | |
421 | ||
422 | if (inode != list_entry(inode_unused.next, | |
423 | struct inode, i_list)) | |
424 | continue; /* wrong inode or list_empty */ | |
425 | if (!can_unuse(inode)) | |
426 | continue; | |
427 | } | |
1da177e4 LT |
428 | list_move(&inode->i_list, &freeable); |
429 | inode->i_state |= I_FREEING; | |
430 | nr_pruned++; | |
431 | } | |
432 | inodes_stat.nr_unused -= nr_pruned; | |
f8891e5e CL |
433 | if (current_is_kswapd()) |
434 | __count_vm_events(KSWAPD_INODESTEAL, reap); | |
435 | else | |
436 | __count_vm_events(PGINODESTEAL, reap); | |
1da177e4 LT |
437 | spin_unlock(&inode_lock); |
438 | ||
439 | dispose_list(&freeable); | |
f24075bd | 440 | mutex_unlock(&iprune_mutex); |
1da177e4 LT |
441 | } |
442 | ||
443 | /* | |
444 | * shrink_icache_memory() will attempt to reclaim some unused inodes. Here, | |
445 | * "unused" means that no dentries are referring to the inodes: the files are | |
446 | * not open and the dcache references to those inodes have already been | |
447 | * reclaimed. | |
448 | * | |
449 | * This function is passed the number of inodes to scan, and it returns the | |
450 | * total number of remaining possibly-reclaimable inodes. | |
451 | */ | |
27496a8c | 452 | static int shrink_icache_memory(int nr, gfp_t gfp_mask) |
1da177e4 LT |
453 | { |
454 | if (nr) { | |
455 | /* | |
456 | * Nasty deadlock avoidance. We may hold various FS locks, | |
457 | * and we don't want to recurse into the FS that called us | |
458 | * in clear_inode() and friends.. | |
459 | */ | |
460 | if (!(gfp_mask & __GFP_FS)) | |
461 | return -1; | |
462 | prune_icache(nr); | |
463 | } | |
464 | return (inodes_stat.nr_unused / 100) * sysctl_vfs_cache_pressure; | |
465 | } | |
466 | ||
467 | static void __wait_on_freeing_inode(struct inode *inode); | |
468 | /* | |
469 | * Called with the inode lock held. | |
470 | * NOTE: we are not increasing the inode-refcount, you must call __iget() | |
471 | * by hand after calling find_inode now! This simplifies iunique and won't | |
472 | * add any additional branch in the common code. | |
473 | */ | |
474 | static struct inode * find_inode(struct super_block * sb, struct hlist_head *head, int (*test)(struct inode *, void *), void *data) | |
475 | { | |
476 | struct hlist_node *node; | |
477 | struct inode * inode = NULL; | |
478 | ||
479 | repeat: | |
480 | hlist_for_each (node, head) { | |
481 | inode = hlist_entry(node, struct inode, i_hash); | |
482 | if (inode->i_sb != sb) | |
483 | continue; | |
484 | if (!test(inode, data)) | |
485 | continue; | |
991114c6 | 486 | if (inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE)) { |
1da177e4 LT |
487 | __wait_on_freeing_inode(inode); |
488 | goto repeat; | |
489 | } | |
490 | break; | |
491 | } | |
492 | return node ? inode : NULL; | |
493 | } | |
494 | ||
495 | /* | |
496 | * find_inode_fast is the fast path version of find_inode, see the comment at | |
497 | * iget_locked for details. | |
498 | */ | |
499 | static struct inode * find_inode_fast(struct super_block * sb, struct hlist_head *head, unsigned long ino) | |
500 | { | |
501 | struct hlist_node *node; | |
502 | struct inode * inode = NULL; | |
503 | ||
504 | repeat: | |
505 | hlist_for_each (node, head) { | |
506 | inode = hlist_entry(node, struct inode, i_hash); | |
507 | if (inode->i_ino != ino) | |
508 | continue; | |
509 | if (inode->i_sb != sb) | |
510 | continue; | |
991114c6 | 511 | if (inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE)) { |
1da177e4 LT |
512 | __wait_on_freeing_inode(inode); |
513 | goto repeat; | |
514 | } | |
515 | break; | |
516 | } | |
517 | return node ? inode : NULL; | |
518 | } | |
519 | ||
520 | /** | |
521 | * new_inode - obtain an inode | |
522 | * @sb: superblock | |
523 | * | |
524 | * Allocates a new inode for given superblock. | |
525 | */ | |
526 | struct inode *new_inode(struct super_block *sb) | |
527 | { | |
528 | static unsigned long last_ino; | |
529 | struct inode * inode; | |
530 | ||
531 | spin_lock_prefetch(&inode_lock); | |
532 | ||
533 | inode = alloc_inode(sb); | |
534 | if (inode) { | |
535 | spin_lock(&inode_lock); | |
536 | inodes_stat.nr_inodes++; | |
537 | list_add(&inode->i_list, &inode_in_use); | |
538 | list_add(&inode->i_sb_list, &sb->s_inodes); | |
539 | inode->i_ino = ++last_ino; | |
540 | inode->i_state = 0; | |
541 | spin_unlock(&inode_lock); | |
542 | } | |
543 | return inode; | |
544 | } | |
545 | ||
546 | EXPORT_SYMBOL(new_inode); | |
547 | ||
548 | void unlock_new_inode(struct inode *inode) | |
549 | { | |
550 | /* | |
551 | * This is special! We do not need the spinlock | |
552 | * when clearing I_LOCK, because we're guaranteed | |
553 | * that nobody else tries to do anything about the | |
554 | * state of the inode when it is locked, as we | |
555 | * just created it (so there can be no old holders | |
556 | * that haven't tested I_LOCK). | |
557 | */ | |
558 | inode->i_state &= ~(I_LOCK|I_NEW); | |
559 | wake_up_inode(inode); | |
560 | } | |
561 | ||
562 | EXPORT_SYMBOL(unlock_new_inode); | |
563 | ||
564 | /* | |
565 | * This is called without the inode lock held.. Be careful. | |
566 | * | |
567 | * We no longer cache the sb_flags in i_flags - see fs.h | |
568 | * -- rmk@arm.uk.linux.org | |
569 | */ | |
570 | static struct inode * get_new_inode(struct super_block *sb, struct hlist_head *head, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *data) | |
571 | { | |
572 | struct inode * inode; | |
573 | ||
574 | inode = alloc_inode(sb); | |
575 | if (inode) { | |
576 | struct inode * old; | |
577 | ||
578 | spin_lock(&inode_lock); | |
579 | /* We released the lock, so.. */ | |
580 | old = find_inode(sb, head, test, data); | |
581 | if (!old) { | |
582 | if (set(inode, data)) | |
583 | goto set_failed; | |
584 | ||
585 | inodes_stat.nr_inodes++; | |
586 | list_add(&inode->i_list, &inode_in_use); | |
587 | list_add(&inode->i_sb_list, &sb->s_inodes); | |
588 | hlist_add_head(&inode->i_hash, head); | |
589 | inode->i_state = I_LOCK|I_NEW; | |
590 | spin_unlock(&inode_lock); | |
591 | ||
592 | /* Return the locked inode with I_NEW set, the | |
593 | * caller is responsible for filling in the contents | |
594 | */ | |
595 | return inode; | |
596 | } | |
597 | ||
598 | /* | |
599 | * Uhhuh, somebody else created the same inode under | |
600 | * us. Use the old inode instead of the one we just | |
601 | * allocated. | |
602 | */ | |
603 | __iget(old); | |
604 | spin_unlock(&inode_lock); | |
605 | destroy_inode(inode); | |
606 | inode = old; | |
607 | wait_on_inode(inode); | |
608 | } | |
609 | return inode; | |
610 | ||
611 | set_failed: | |
612 | spin_unlock(&inode_lock); | |
613 | destroy_inode(inode); | |
614 | return NULL; | |
615 | } | |
616 | ||
617 | /* | |
618 | * get_new_inode_fast is the fast path version of get_new_inode, see the | |
619 | * comment at iget_locked for details. | |
620 | */ | |
621 | static struct inode * get_new_inode_fast(struct super_block *sb, struct hlist_head *head, unsigned long ino) | |
622 | { | |
623 | struct inode * inode; | |
624 | ||
625 | inode = alloc_inode(sb); | |
626 | if (inode) { | |
627 | struct inode * old; | |
628 | ||
629 | spin_lock(&inode_lock); | |
630 | /* We released the lock, so.. */ | |
631 | old = find_inode_fast(sb, head, ino); | |
632 | if (!old) { | |
633 | inode->i_ino = ino; | |
634 | inodes_stat.nr_inodes++; | |
635 | list_add(&inode->i_list, &inode_in_use); | |
636 | list_add(&inode->i_sb_list, &sb->s_inodes); | |
637 | hlist_add_head(&inode->i_hash, head); | |
638 | inode->i_state = I_LOCK|I_NEW; | |
639 | spin_unlock(&inode_lock); | |
640 | ||
641 | /* Return the locked inode with I_NEW set, the | |
642 | * caller is responsible for filling in the contents | |
643 | */ | |
644 | return inode; | |
645 | } | |
646 | ||
647 | /* | |
648 | * Uhhuh, somebody else created the same inode under | |
649 | * us. Use the old inode instead of the one we just | |
650 | * allocated. | |
651 | */ | |
652 | __iget(old); | |
653 | spin_unlock(&inode_lock); | |
654 | destroy_inode(inode); | |
655 | inode = old; | |
656 | wait_on_inode(inode); | |
657 | } | |
658 | return inode; | |
659 | } | |
660 | ||
ed97bd37 | 661 | static unsigned long hash(struct super_block *sb, unsigned long hashval) |
1da177e4 LT |
662 | { |
663 | unsigned long tmp; | |
664 | ||
665 | tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) / | |
666 | L1_CACHE_BYTES; | |
667 | tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> I_HASHBITS); | |
668 | return tmp & I_HASHMASK; | |
669 | } | |
670 | ||
671 | /** | |
672 | * iunique - get a unique inode number | |
673 | * @sb: superblock | |
674 | * @max_reserved: highest reserved inode number | |
675 | * | |
676 | * Obtain an inode number that is unique on the system for a given | |
677 | * superblock. This is used by file systems that have no natural | |
678 | * permanent inode numbering system. An inode number is returned that | |
679 | * is higher than the reserved limit but unique. | |
680 | * | |
681 | * BUGS: | |
682 | * With a large number of inodes live on the file system this function | |
683 | * currently becomes quite slow. | |
684 | */ | |
685 | ino_t iunique(struct super_block *sb, ino_t max_reserved) | |
686 | { | |
687 | static ino_t counter; | |
688 | struct inode *inode; | |
689 | struct hlist_head * head; | |
690 | ino_t res; | |
691 | spin_lock(&inode_lock); | |
692 | retry: | |
693 | if (counter > max_reserved) { | |
694 | head = inode_hashtable + hash(sb,counter); | |
695 | res = counter++; | |
696 | inode = find_inode_fast(sb, head, res); | |
697 | if (!inode) { | |
698 | spin_unlock(&inode_lock); | |
699 | return res; | |
700 | } | |
701 | } else { | |
702 | counter = max_reserved + 1; | |
703 | } | |
704 | goto retry; | |
705 | ||
706 | } | |
707 | ||
708 | EXPORT_SYMBOL(iunique); | |
709 | ||
710 | struct inode *igrab(struct inode *inode) | |
711 | { | |
712 | spin_lock(&inode_lock); | |
4a3b0a49 | 713 | if (!(inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE))) |
1da177e4 LT |
714 | __iget(inode); |
715 | else | |
716 | /* | |
717 | * Handle the case where s_op->clear_inode is not been | |
718 | * called yet, and somebody is calling igrab | |
719 | * while the inode is getting freed. | |
720 | */ | |
721 | inode = NULL; | |
722 | spin_unlock(&inode_lock); | |
723 | return inode; | |
724 | } | |
725 | ||
726 | EXPORT_SYMBOL(igrab); | |
727 | ||
728 | /** | |
729 | * ifind - internal function, you want ilookup5() or iget5(). | |
730 | * @sb: super block of file system to search | |
731 | * @head: the head of the list to search | |
732 | * @test: callback used for comparisons between inodes | |
733 | * @data: opaque data pointer to pass to @test | |
88bd5121 | 734 | * @wait: if true wait for the inode to be unlocked, if false do not |
1da177e4 LT |
735 | * |
736 | * ifind() searches for the inode specified by @data in the inode | |
737 | * cache. This is a generalized version of ifind_fast() for file systems where | |
738 | * the inode number is not sufficient for unique identification of an inode. | |
739 | * | |
740 | * If the inode is in the cache, the inode is returned with an incremented | |
741 | * reference count. | |
742 | * | |
743 | * Otherwise NULL is returned. | |
744 | * | |
745 | * Note, @test is called with the inode_lock held, so can't sleep. | |
746 | */ | |
5d2bea45 | 747 | static struct inode *ifind(struct super_block *sb, |
1da177e4 | 748 | struct hlist_head *head, int (*test)(struct inode *, void *), |
88bd5121 | 749 | void *data, const int wait) |
1da177e4 LT |
750 | { |
751 | struct inode *inode; | |
752 | ||
753 | spin_lock(&inode_lock); | |
754 | inode = find_inode(sb, head, test, data); | |
755 | if (inode) { | |
756 | __iget(inode); | |
757 | spin_unlock(&inode_lock); | |
88bd5121 AA |
758 | if (likely(wait)) |
759 | wait_on_inode(inode); | |
1da177e4 LT |
760 | return inode; |
761 | } | |
762 | spin_unlock(&inode_lock); | |
763 | return NULL; | |
764 | } | |
765 | ||
766 | /** | |
767 | * ifind_fast - internal function, you want ilookup() or iget(). | |
768 | * @sb: super block of file system to search | |
769 | * @head: head of the list to search | |
770 | * @ino: inode number to search for | |
771 | * | |
772 | * ifind_fast() searches for the inode @ino in the inode cache. This is for | |
773 | * file systems where the inode number is sufficient for unique identification | |
774 | * of an inode. | |
775 | * | |
776 | * If the inode is in the cache, the inode is returned with an incremented | |
777 | * reference count. | |
778 | * | |
779 | * Otherwise NULL is returned. | |
780 | */ | |
5d2bea45 | 781 | static struct inode *ifind_fast(struct super_block *sb, |
1da177e4 LT |
782 | struct hlist_head *head, unsigned long ino) |
783 | { | |
784 | struct inode *inode; | |
785 | ||
786 | spin_lock(&inode_lock); | |
787 | inode = find_inode_fast(sb, head, ino); | |
788 | if (inode) { | |
789 | __iget(inode); | |
790 | spin_unlock(&inode_lock); | |
791 | wait_on_inode(inode); | |
792 | return inode; | |
793 | } | |
794 | spin_unlock(&inode_lock); | |
795 | return NULL; | |
796 | } | |
797 | ||
798 | /** | |
88bd5121 | 799 | * ilookup5_nowait - search for an inode in the inode cache |
1da177e4 LT |
800 | * @sb: super block of file system to search |
801 | * @hashval: hash value (usually inode number) to search for | |
802 | * @test: callback used for comparisons between inodes | |
803 | * @data: opaque data pointer to pass to @test | |
804 | * | |
805 | * ilookup5() uses ifind() to search for the inode specified by @hashval and | |
806 | * @data in the inode cache. This is a generalized version of ilookup() for | |
807 | * file systems where the inode number is not sufficient for unique | |
808 | * identification of an inode. | |
809 | * | |
810 | * If the inode is in the cache, the inode is returned with an incremented | |
88bd5121 AA |
811 | * reference count. Note, the inode lock is not waited upon so you have to be |
812 | * very careful what you do with the returned inode. You probably should be | |
813 | * using ilookup5() instead. | |
814 | * | |
815 | * Otherwise NULL is returned. | |
816 | * | |
817 | * Note, @test is called with the inode_lock held, so can't sleep. | |
818 | */ | |
819 | struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval, | |
820 | int (*test)(struct inode *, void *), void *data) | |
821 | { | |
822 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
823 | ||
824 | return ifind(sb, head, test, data, 0); | |
825 | } | |
826 | ||
827 | EXPORT_SYMBOL(ilookup5_nowait); | |
828 | ||
829 | /** | |
830 | * ilookup5 - search for an inode in the inode cache | |
831 | * @sb: super block of file system to search | |
832 | * @hashval: hash value (usually inode number) to search for | |
833 | * @test: callback used for comparisons between inodes | |
834 | * @data: opaque data pointer to pass to @test | |
835 | * | |
836 | * ilookup5() uses ifind() to search for the inode specified by @hashval and | |
837 | * @data in the inode cache. This is a generalized version of ilookup() for | |
838 | * file systems where the inode number is not sufficient for unique | |
839 | * identification of an inode. | |
840 | * | |
841 | * If the inode is in the cache, the inode lock is waited upon and the inode is | |
842 | * returned with an incremented reference count. | |
1da177e4 LT |
843 | * |
844 | * Otherwise NULL is returned. | |
845 | * | |
846 | * Note, @test is called with the inode_lock held, so can't sleep. | |
847 | */ | |
848 | struct inode *ilookup5(struct super_block *sb, unsigned long hashval, | |
849 | int (*test)(struct inode *, void *), void *data) | |
850 | { | |
851 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
852 | ||
88bd5121 | 853 | return ifind(sb, head, test, data, 1); |
1da177e4 LT |
854 | } |
855 | ||
856 | EXPORT_SYMBOL(ilookup5); | |
857 | ||
858 | /** | |
859 | * ilookup - search for an inode in the inode cache | |
860 | * @sb: super block of file system to search | |
861 | * @ino: inode number to search for | |
862 | * | |
863 | * ilookup() uses ifind_fast() to search for the inode @ino in the inode cache. | |
864 | * This is for file systems where the inode number is sufficient for unique | |
865 | * identification of an inode. | |
866 | * | |
867 | * If the inode is in the cache, the inode is returned with an incremented | |
868 | * reference count. | |
869 | * | |
870 | * Otherwise NULL is returned. | |
871 | */ | |
872 | struct inode *ilookup(struct super_block *sb, unsigned long ino) | |
873 | { | |
874 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
875 | ||
876 | return ifind_fast(sb, head, ino); | |
877 | } | |
878 | ||
879 | EXPORT_SYMBOL(ilookup); | |
880 | ||
881 | /** | |
882 | * iget5_locked - obtain an inode from a mounted file system | |
883 | * @sb: super block of file system | |
884 | * @hashval: hash value (usually inode number) to get | |
885 | * @test: callback used for comparisons between inodes | |
886 | * @set: callback used to initialize a new struct inode | |
887 | * @data: opaque data pointer to pass to @test and @set | |
888 | * | |
889 | * This is iget() without the read_inode() portion of get_new_inode(). | |
890 | * | |
891 | * iget5_locked() uses ifind() to search for the inode specified by @hashval | |
892 | * and @data in the inode cache and if present it is returned with an increased | |
893 | * reference count. This is a generalized version of iget_locked() for file | |
894 | * systems where the inode number is not sufficient for unique identification | |
895 | * of an inode. | |
896 | * | |
897 | * If the inode is not in cache, get_new_inode() is called to allocate a new | |
898 | * inode and this is returned locked, hashed, and with the I_NEW flag set. The | |
899 | * file system gets to fill it in before unlocking it via unlock_new_inode(). | |
900 | * | |
901 | * Note both @test and @set are called with the inode_lock held, so can't sleep. | |
902 | */ | |
903 | struct inode *iget5_locked(struct super_block *sb, unsigned long hashval, | |
904 | int (*test)(struct inode *, void *), | |
905 | int (*set)(struct inode *, void *), void *data) | |
906 | { | |
907 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
908 | struct inode *inode; | |
909 | ||
88bd5121 | 910 | inode = ifind(sb, head, test, data, 1); |
1da177e4 LT |
911 | if (inode) |
912 | return inode; | |
913 | /* | |
914 | * get_new_inode() will do the right thing, re-trying the search | |
915 | * in case it had to block at any point. | |
916 | */ | |
917 | return get_new_inode(sb, head, test, set, data); | |
918 | } | |
919 | ||
920 | EXPORT_SYMBOL(iget5_locked); | |
921 | ||
922 | /** | |
923 | * iget_locked - obtain an inode from a mounted file system | |
924 | * @sb: super block of file system | |
925 | * @ino: inode number to get | |
926 | * | |
927 | * This is iget() without the read_inode() portion of get_new_inode_fast(). | |
928 | * | |
929 | * iget_locked() uses ifind_fast() to search for the inode specified by @ino in | |
930 | * the inode cache and if present it is returned with an increased reference | |
931 | * count. This is for file systems where the inode number is sufficient for | |
932 | * unique identification of an inode. | |
933 | * | |
934 | * If the inode is not in cache, get_new_inode_fast() is called to allocate a | |
935 | * new inode and this is returned locked, hashed, and with the I_NEW flag set. | |
936 | * The file system gets to fill it in before unlocking it via | |
937 | * unlock_new_inode(). | |
938 | */ | |
939 | struct inode *iget_locked(struct super_block *sb, unsigned long ino) | |
940 | { | |
941 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
942 | struct inode *inode; | |
943 | ||
944 | inode = ifind_fast(sb, head, ino); | |
945 | if (inode) | |
946 | return inode; | |
947 | /* | |
948 | * get_new_inode_fast() will do the right thing, re-trying the search | |
949 | * in case it had to block at any point. | |
950 | */ | |
951 | return get_new_inode_fast(sb, head, ino); | |
952 | } | |
953 | ||
954 | EXPORT_SYMBOL(iget_locked); | |
955 | ||
956 | /** | |
957 | * __insert_inode_hash - hash an inode | |
958 | * @inode: unhashed inode | |
959 | * @hashval: unsigned long value used to locate this object in the | |
960 | * inode_hashtable. | |
961 | * | |
962 | * Add an inode to the inode hash for this superblock. | |
963 | */ | |
964 | void __insert_inode_hash(struct inode *inode, unsigned long hashval) | |
965 | { | |
966 | struct hlist_head *head = inode_hashtable + hash(inode->i_sb, hashval); | |
967 | spin_lock(&inode_lock); | |
968 | hlist_add_head(&inode->i_hash, head); | |
969 | spin_unlock(&inode_lock); | |
970 | } | |
971 | ||
972 | EXPORT_SYMBOL(__insert_inode_hash); | |
973 | ||
974 | /** | |
975 | * remove_inode_hash - remove an inode from the hash | |
976 | * @inode: inode to unhash | |
977 | * | |
978 | * Remove an inode from the superblock. | |
979 | */ | |
980 | void remove_inode_hash(struct inode *inode) | |
981 | { | |
982 | spin_lock(&inode_lock); | |
983 | hlist_del_init(&inode->i_hash); | |
984 | spin_unlock(&inode_lock); | |
985 | } | |
986 | ||
987 | EXPORT_SYMBOL(remove_inode_hash); | |
988 | ||
989 | /* | |
990 | * Tell the filesystem that this inode is no longer of any interest and should | |
991 | * be completely destroyed. | |
992 | * | |
993 | * We leave the inode in the inode hash table until *after* the filesystem's | |
994 | * ->delete_inode completes. This ensures that an iget (such as nfsd might | |
995 | * instigate) will always find up-to-date information either in the hash or on | |
996 | * disk. | |
997 | * | |
998 | * I_FREEING is set so that no-one will take a new reference to the inode while | |
999 | * it is being deleted. | |
1000 | */ | |
1001 | void generic_delete_inode(struct inode *inode) | |
1002 | { | |
ee9b6d61 | 1003 | const struct super_operations *op = inode->i_sb->s_op; |
1da177e4 LT |
1004 | |
1005 | list_del_init(&inode->i_list); | |
1006 | list_del_init(&inode->i_sb_list); | |
ed97bd37 | 1007 | inode->i_state |= I_FREEING; |
1da177e4 LT |
1008 | inodes_stat.nr_inodes--; |
1009 | spin_unlock(&inode_lock); | |
1010 | ||
1da177e4 LT |
1011 | security_inode_delete(inode); |
1012 | ||
1013 | if (op->delete_inode) { | |
1014 | void (*delete)(struct inode *) = op->delete_inode; | |
1015 | if (!is_bad_inode(inode)) | |
1016 | DQUOT_INIT(inode); | |
e85b5652 MF |
1017 | /* Filesystems implementing their own |
1018 | * s_op->delete_inode are required to call | |
1019 | * truncate_inode_pages and clear_inode() | |
1020 | * internally */ | |
1da177e4 | 1021 | delete(inode); |
e85b5652 MF |
1022 | } else { |
1023 | truncate_inode_pages(&inode->i_data, 0); | |
1da177e4 | 1024 | clear_inode(inode); |
e85b5652 | 1025 | } |
1da177e4 LT |
1026 | spin_lock(&inode_lock); |
1027 | hlist_del_init(&inode->i_hash); | |
1028 | spin_unlock(&inode_lock); | |
1029 | wake_up_inode(inode); | |
b7542f8c | 1030 | BUG_ON(inode->i_state != I_CLEAR); |
1da177e4 LT |
1031 | destroy_inode(inode); |
1032 | } | |
1033 | ||
1034 | EXPORT_SYMBOL(generic_delete_inode); | |
1035 | ||
1036 | static void generic_forget_inode(struct inode *inode) | |
1037 | { | |
1038 | struct super_block *sb = inode->i_sb; | |
1039 | ||
1040 | if (!hlist_unhashed(&inode->i_hash)) { | |
1041 | if (!(inode->i_state & (I_DIRTY|I_LOCK))) | |
1042 | list_move(&inode->i_list, &inode_unused); | |
1043 | inodes_stat.nr_unused++; | |
991114c6 AV |
1044 | if (!sb || (sb->s_flags & MS_ACTIVE)) { |
1045 | spin_unlock(&inode_lock); | |
1da177e4 | 1046 | return; |
991114c6 AV |
1047 | } |
1048 | inode->i_state |= I_WILL_FREE; | |
1049 | spin_unlock(&inode_lock); | |
1da177e4 LT |
1050 | write_inode_now(inode, 1); |
1051 | spin_lock(&inode_lock); | |
991114c6 | 1052 | inode->i_state &= ~I_WILL_FREE; |
1da177e4 LT |
1053 | inodes_stat.nr_unused--; |
1054 | hlist_del_init(&inode->i_hash); | |
1055 | } | |
1056 | list_del_init(&inode->i_list); | |
1057 | list_del_init(&inode->i_sb_list); | |
991114c6 | 1058 | inode->i_state |= I_FREEING; |
1da177e4 LT |
1059 | inodes_stat.nr_inodes--; |
1060 | spin_unlock(&inode_lock); | |
1061 | if (inode->i_data.nrpages) | |
1062 | truncate_inode_pages(&inode->i_data, 0); | |
1063 | clear_inode(inode); | |
7f04c26d | 1064 | wake_up_inode(inode); |
1da177e4 LT |
1065 | destroy_inode(inode); |
1066 | } | |
1067 | ||
1068 | /* | |
1069 | * Normal UNIX filesystem behaviour: delete the | |
1070 | * inode when the usage count drops to zero, and | |
1071 | * i_nlink is zero. | |
1072 | */ | |
cb2c0233 | 1073 | void generic_drop_inode(struct inode *inode) |
1da177e4 LT |
1074 | { |
1075 | if (!inode->i_nlink) | |
1076 | generic_delete_inode(inode); | |
1077 | else | |
1078 | generic_forget_inode(inode); | |
1079 | } | |
1080 | ||
cb2c0233 MF |
1081 | EXPORT_SYMBOL_GPL(generic_drop_inode); |
1082 | ||
1da177e4 LT |
1083 | /* |
1084 | * Called when we're dropping the last reference | |
1085 | * to an inode. | |
1086 | * | |
1087 | * Call the FS "drop()" function, defaulting to | |
1088 | * the legacy UNIX filesystem behaviour.. | |
1089 | * | |
1090 | * NOTE! NOTE! NOTE! We're called with the inode lock | |
1091 | * held, and the drop function is supposed to release | |
1092 | * the lock! | |
1093 | */ | |
1094 | static inline void iput_final(struct inode *inode) | |
1095 | { | |
ee9b6d61 | 1096 | const struct super_operations *op = inode->i_sb->s_op; |
1da177e4 LT |
1097 | void (*drop)(struct inode *) = generic_drop_inode; |
1098 | ||
1099 | if (op && op->drop_inode) | |
1100 | drop = op->drop_inode; | |
1101 | drop(inode); | |
1102 | } | |
1103 | ||
1104 | /** | |
1105 | * iput - put an inode | |
1106 | * @inode: inode to put | |
1107 | * | |
1108 | * Puts an inode, dropping its usage count. If the inode use count hits | |
1109 | * zero, the inode is then freed and may also be destroyed. | |
1110 | * | |
1111 | * Consequently, iput() can sleep. | |
1112 | */ | |
1113 | void iput(struct inode *inode) | |
1114 | { | |
1115 | if (inode) { | |
ee9b6d61 | 1116 | const struct super_operations *op = inode->i_sb->s_op; |
1da177e4 LT |
1117 | |
1118 | BUG_ON(inode->i_state == I_CLEAR); | |
1119 | ||
1120 | if (op && op->put_inode) | |
1121 | op->put_inode(inode); | |
1122 | ||
1123 | if (atomic_dec_and_lock(&inode->i_count, &inode_lock)) | |
1124 | iput_final(inode); | |
1125 | } | |
1126 | } | |
1127 | ||
1128 | EXPORT_SYMBOL(iput); | |
1129 | ||
1130 | /** | |
1131 | * bmap - find a block number in a file | |
1132 | * @inode: inode of file | |
1133 | * @block: block to find | |
1134 | * | |
1135 | * Returns the block number on the device holding the inode that | |
1136 | * is the disk block number for the block of the file requested. | |
1137 | * That is, asked for block 4 of inode 1 the function will return the | |
1138 | * disk block relative to the disk start that holds that block of the | |
1139 | * file. | |
1140 | */ | |
1141 | sector_t bmap(struct inode * inode, sector_t block) | |
1142 | { | |
1143 | sector_t res = 0; | |
1144 | if (inode->i_mapping->a_ops->bmap) | |
1145 | res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block); | |
1146 | return res; | |
1147 | } | |
1da177e4 LT |
1148 | EXPORT_SYMBOL(bmap); |
1149 | ||
1150 | /** | |
869243a0 CH |
1151 | * touch_atime - update the access time |
1152 | * @mnt: mount the inode is accessed on | |
7045f37b | 1153 | * @dentry: dentry accessed |
1da177e4 LT |
1154 | * |
1155 | * Update the accessed time on an inode and mark it for writeback. | |
1156 | * This function automatically handles read only file systems and media, | |
1157 | * as well as the "noatime" flag and inode specific "noatime" markers. | |
1158 | */ | |
869243a0 | 1159 | void touch_atime(struct vfsmount *mnt, struct dentry *dentry) |
1da177e4 | 1160 | { |
869243a0 | 1161 | struct inode *inode = dentry->d_inode; |
1da177e4 LT |
1162 | struct timespec now; |
1163 | ||
b2276138 AM |
1164 | if (inode->i_flags & S_NOATIME) |
1165 | return; | |
37756ced | 1166 | if (IS_NOATIME(inode)) |
b2276138 AM |
1167 | return; |
1168 | if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)) | |
1da177e4 | 1169 | return; |
fc33a7bb CH |
1170 | |
1171 | /* | |
1172 | * We may have a NULL vfsmount when coming from NFSD | |
1173 | */ | |
b2276138 AM |
1174 | if (mnt) { |
1175 | if (mnt->mnt_flags & MNT_NOATIME) | |
1176 | return; | |
1177 | if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) | |
1178 | return; | |
47ae32d6 VH |
1179 | |
1180 | if (mnt->mnt_flags & MNT_RELATIME) { | |
1181 | /* | |
1182 | * With relative atime, only update atime if the | |
1183 | * previous atime is earlier than either the ctime or | |
1184 | * mtime. | |
1185 | */ | |
1186 | if (timespec_compare(&inode->i_mtime, | |
1187 | &inode->i_atime) < 0 && | |
1188 | timespec_compare(&inode->i_ctime, | |
1189 | &inode->i_atime) < 0) | |
1190 | return; | |
1191 | } | |
b2276138 | 1192 | } |
1da177e4 LT |
1193 | |
1194 | now = current_fs_time(inode->i_sb); | |
47ae32d6 VH |
1195 | if (timespec_equal(&inode->i_atime, &now)) |
1196 | return; | |
1197 | ||
1198 | inode->i_atime = now; | |
1199 | mark_inode_dirty_sync(inode); | |
1da177e4 | 1200 | } |
869243a0 | 1201 | EXPORT_SYMBOL(touch_atime); |
1da177e4 LT |
1202 | |
1203 | /** | |
870f4817 CH |
1204 | * file_update_time - update mtime and ctime time |
1205 | * @file: file accessed | |
1da177e4 | 1206 | * |
870f4817 CH |
1207 | * Update the mtime and ctime members of an inode and mark the inode |
1208 | * for writeback. Note that this function is meant exclusively for | |
1209 | * usage in the file write path of filesystems, and filesystems may | |
1210 | * choose to explicitly ignore update via this function with the | |
1211 | * S_NOCTIME inode flag, e.g. for network filesystem where these | |
1212 | * timestamps are handled by the server. | |
1da177e4 LT |
1213 | */ |
1214 | ||
870f4817 | 1215 | void file_update_time(struct file *file) |
1da177e4 | 1216 | { |
0f7fc9e4 | 1217 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
1218 | struct timespec now; |
1219 | int sync_it = 0; | |
1220 | ||
1221 | if (IS_NOCMTIME(inode)) | |
1222 | return; | |
1223 | if (IS_RDONLY(inode)) | |
1224 | return; | |
1225 | ||
1226 | now = current_fs_time(inode->i_sb); | |
ed97bd37 AM |
1227 | if (!timespec_equal(&inode->i_mtime, &now)) { |
1228 | inode->i_mtime = now; | |
1da177e4 | 1229 | sync_it = 1; |
ed97bd37 | 1230 | } |
1da177e4 | 1231 | |
ed97bd37 AM |
1232 | if (!timespec_equal(&inode->i_ctime, &now)) { |
1233 | inode->i_ctime = now; | |
870f4817 | 1234 | sync_it = 1; |
ed97bd37 | 1235 | } |
870f4817 | 1236 | |
1da177e4 LT |
1237 | if (sync_it) |
1238 | mark_inode_dirty_sync(inode); | |
1239 | } | |
1240 | ||
870f4817 | 1241 | EXPORT_SYMBOL(file_update_time); |
1da177e4 LT |
1242 | |
1243 | int inode_needs_sync(struct inode *inode) | |
1244 | { | |
1245 | if (IS_SYNC(inode)) | |
1246 | return 1; | |
1247 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) | |
1248 | return 1; | |
1249 | return 0; | |
1250 | } | |
1251 | ||
1252 | EXPORT_SYMBOL(inode_needs_sync); | |
1253 | ||
1da177e4 LT |
1254 | int inode_wait(void *word) |
1255 | { | |
1256 | schedule(); | |
1257 | return 0; | |
1258 | } | |
1259 | ||
1260 | /* | |
168a9fd6 MS |
1261 | * If we try to find an inode in the inode hash while it is being |
1262 | * deleted, we have to wait until the filesystem completes its | |
1263 | * deletion before reporting that it isn't found. This function waits | |
1264 | * until the deletion _might_ have completed. Callers are responsible | |
1265 | * to recheck inode state. | |
1266 | * | |
1267 | * It doesn't matter if I_LOCK is not set initially, a call to | |
1268 | * wake_up_inode() after removing from the hash list will DTRT. | |
1269 | * | |
1da177e4 LT |
1270 | * This is called with inode_lock held. |
1271 | */ | |
1272 | static void __wait_on_freeing_inode(struct inode *inode) | |
1273 | { | |
1274 | wait_queue_head_t *wq; | |
1275 | DEFINE_WAIT_BIT(wait, &inode->i_state, __I_LOCK); | |
1da177e4 LT |
1276 | wq = bit_waitqueue(&inode->i_state, __I_LOCK); |
1277 | prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE); | |
1278 | spin_unlock(&inode_lock); | |
1279 | schedule(); | |
1280 | finish_wait(wq, &wait.wait); | |
1281 | spin_lock(&inode_lock); | |
1282 | } | |
1283 | ||
1284 | void wake_up_inode(struct inode *inode) | |
1285 | { | |
1286 | /* | |
1287 | * Prevent speculative execution through spin_unlock(&inode_lock); | |
1288 | */ | |
1289 | smp_mb(); | |
1290 | wake_up_bit(&inode->i_state, __I_LOCK); | |
1291 | } | |
1292 | ||
62752ee1 MF |
1293 | /* |
1294 | * We rarely want to lock two inodes that do not have a parent/child | |
1295 | * relationship (such as directory, child inode) simultaneously. The | |
1296 | * vast majority of file systems should be able to get along fine | |
1297 | * without this. Do not use these functions except as a last resort. | |
1298 | */ | |
1299 | void inode_double_lock(struct inode *inode1, struct inode *inode2) | |
1300 | { | |
1301 | if (inode1 == NULL || inode2 == NULL || inode1 == inode2) { | |
1302 | if (inode1) | |
1303 | mutex_lock(&inode1->i_mutex); | |
1304 | else if (inode2) | |
1305 | mutex_lock(&inode2->i_mutex); | |
1306 | return; | |
1307 | } | |
1308 | ||
1309 | if (inode1 < inode2) { | |
1310 | mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT); | |
1311 | mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD); | |
1312 | } else { | |
1313 | mutex_lock_nested(&inode2->i_mutex, I_MUTEX_PARENT); | |
1314 | mutex_lock_nested(&inode1->i_mutex, I_MUTEX_CHILD); | |
1315 | } | |
1316 | } | |
1317 | EXPORT_SYMBOL(inode_double_lock); | |
1318 | ||
1319 | void inode_double_unlock(struct inode *inode1, struct inode *inode2) | |
1320 | { | |
1321 | if (inode1) | |
1322 | mutex_unlock(&inode1->i_mutex); | |
1323 | ||
1324 | if (inode2 && inode2 != inode1) | |
1325 | mutex_unlock(&inode2->i_mutex); | |
1326 | } | |
1327 | EXPORT_SYMBOL(inode_double_unlock); | |
1328 | ||
1da177e4 LT |
1329 | static __initdata unsigned long ihash_entries; |
1330 | static int __init set_ihash_entries(char *str) | |
1331 | { | |
1332 | if (!str) | |
1333 | return 0; | |
1334 | ihash_entries = simple_strtoul(str, &str, 0); | |
1335 | return 1; | |
1336 | } | |
1337 | __setup("ihash_entries=", set_ihash_entries); | |
1338 | ||
1339 | /* | |
1340 | * Initialize the waitqueues and inode hash table. | |
1341 | */ | |
1342 | void __init inode_init_early(void) | |
1343 | { | |
1344 | int loop; | |
1345 | ||
1346 | /* If hashes are distributed across NUMA nodes, defer | |
1347 | * hash allocation until vmalloc space is available. | |
1348 | */ | |
1349 | if (hashdist) | |
1350 | return; | |
1351 | ||
1352 | inode_hashtable = | |
1353 | alloc_large_system_hash("Inode-cache", | |
1354 | sizeof(struct hlist_head), | |
1355 | ihash_entries, | |
1356 | 14, | |
1357 | HASH_EARLY, | |
1358 | &i_hash_shift, | |
1359 | &i_hash_mask, | |
1360 | 0); | |
1361 | ||
1362 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1363 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1364 | } | |
1365 | ||
1366 | void __init inode_init(unsigned long mempages) | |
1367 | { | |
1368 | int loop; | |
1369 | ||
1370 | /* inode slab cache */ | |
b0196009 PJ |
1371 | inode_cachep = kmem_cache_create("inode_cache", |
1372 | sizeof(struct inode), | |
1373 | 0, | |
1374 | (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC| | |
1375 | SLAB_MEM_SPREAD), | |
1376 | init_once, | |
1377 | NULL); | |
1da177e4 LT |
1378 | set_shrinker(DEFAULT_SEEKS, shrink_icache_memory); |
1379 | ||
1380 | /* Hash may have been set up in inode_init_early */ | |
1381 | if (!hashdist) | |
1382 | return; | |
1383 | ||
1384 | inode_hashtable = | |
1385 | alloc_large_system_hash("Inode-cache", | |
1386 | sizeof(struct hlist_head), | |
1387 | ihash_entries, | |
1388 | 14, | |
1389 | 0, | |
1390 | &i_hash_shift, | |
1391 | &i_hash_mask, | |
1392 | 0); | |
1393 | ||
1394 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1395 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1396 | } | |
1397 | ||
1398 | void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev) | |
1399 | { | |
1400 | inode->i_mode = mode; | |
1401 | if (S_ISCHR(mode)) { | |
1402 | inode->i_fop = &def_chr_fops; | |
1403 | inode->i_rdev = rdev; | |
1404 | } else if (S_ISBLK(mode)) { | |
1405 | inode->i_fop = &def_blk_fops; | |
1406 | inode->i_rdev = rdev; | |
1407 | } else if (S_ISFIFO(mode)) | |
1408 | inode->i_fop = &def_fifo_fops; | |
1409 | else if (S_ISSOCK(mode)) | |
1410 | inode->i_fop = &bad_sock_fops; | |
1411 | else | |
1412 | printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o)\n", | |
1413 | mode); | |
1414 | } | |
1415 | EXPORT_SYMBOL(init_special_inode); |