Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
1da177e4 | 2 | * (C) 1997 Linus Torvalds |
4b4563dc | 3 | * (C) 1999 Andrea Arcangeli <andrea@suse.de> (dynamic inode allocation) |
1da177e4 | 4 | */ |
1da177e4 LT |
5 | #include <linux/fs.h> |
6 | #include <linux/mm.h> | |
7 | #include <linux/dcache.h> | |
8 | #include <linux/init.h> | |
1da177e4 LT |
9 | #include <linux/slab.h> |
10 | #include <linux/writeback.h> | |
11 | #include <linux/module.h> | |
12 | #include <linux/backing-dev.h> | |
13 | #include <linux/wait.h> | |
88e0fbc4 | 14 | #include <linux/rwsem.h> |
1da177e4 LT |
15 | #include <linux/hash.h> |
16 | #include <linux/swap.h> | |
17 | #include <linux/security.h> | |
18 | #include <linux/pagemap.h> | |
19 | #include <linux/cdev.h> | |
20 | #include <linux/bootmem.h> | |
3be25f49 | 21 | #include <linux/fsnotify.h> |
fc33a7bb | 22 | #include <linux/mount.h> |
efaee192 | 23 | #include <linux/async.h> |
f19d4a8f | 24 | #include <linux/posix_acl.h> |
9ce6e0be | 25 | #include <linux/prefetch.h> |
a178d202 | 26 | #include <linux/ima.h> |
e795b717 | 27 | #include <linux/cred.h> |
4b4563dc | 28 | #include <linux/buffer_head.h> /* for inode_has_buffers */ |
7ada4db8 | 29 | #include <linux/ratelimit.h> |
a66979ab | 30 | #include "internal.h" |
1da177e4 | 31 | |
250df6ed | 32 | /* |
4b4563dc | 33 | * Inode locking rules: |
250df6ed DC |
34 | * |
35 | * inode->i_lock protects: | |
36 | * inode->i_state, inode->i_hash, __iget() | |
09cc9fc7 | 37 | * inode->i_sb->s_inode_lru_lock protects: |
98b745c6 | 38 | * inode->i_sb->s_inode_lru, inode->i_lru |
55fa6091 DC |
39 | * inode_sb_list_lock protects: |
40 | * sb->s_inodes, inode->i_sb_list | |
f758eeab | 41 | * bdi->wb.list_lock protects: |
a66979ab | 42 | * bdi->wb.b_{dirty,io,more_io}, inode->i_wb_list |
67a23c49 DC |
43 | * inode_hash_lock protects: |
44 | * inode_hashtable, inode->i_hash | |
250df6ed DC |
45 | * |
46 | * Lock ordering: | |
55fa6091 DC |
47 | * |
48 | * inode_sb_list_lock | |
49 | * inode->i_lock | |
09cc9fc7 | 50 | * inode->i_sb->s_inode_lru_lock |
a66979ab | 51 | * |
f758eeab | 52 | * bdi->wb.list_lock |
a66979ab | 53 | * inode->i_lock |
67a23c49 DC |
54 | * |
55 | * inode_hash_lock | |
56 | * inode_sb_list_lock | |
57 | * inode->i_lock | |
58 | * | |
59 | * iunique_lock | |
60 | * inode_hash_lock | |
250df6ed DC |
61 | */ |
62 | ||
fa3536cc ED |
63 | static unsigned int i_hash_mask __read_mostly; |
64 | static unsigned int i_hash_shift __read_mostly; | |
67a23c49 DC |
65 | static struct hlist_head *inode_hashtable __read_mostly; |
66 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_hash_lock); | |
1da177e4 | 67 | |
55fa6091 DC |
68 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_sb_list_lock); |
69 | ||
7dcda1c9 JA |
70 | /* |
71 | * Empty aops. Can be used for the cases where the user does not | |
72 | * define any of the address_space operations. | |
73 | */ | |
74 | const struct address_space_operations empty_aops = { | |
75 | }; | |
76 | EXPORT_SYMBOL(empty_aops); | |
77 | ||
1da177e4 LT |
78 | /* |
79 | * Statistics gathering.. | |
80 | */ | |
81 | struct inodes_stat_t inodes_stat; | |
82 | ||
3e880fb5 | 83 | static DEFINE_PER_CPU(unsigned int, nr_inodes); |
fcb94f72 | 84 | static DEFINE_PER_CPU(unsigned int, nr_unused); |
cffbc8aa | 85 | |
6b3304b5 | 86 | static struct kmem_cache *inode_cachep __read_mostly; |
1da177e4 | 87 | |
3e880fb5 | 88 | static int get_nr_inodes(void) |
cffbc8aa | 89 | { |
3e880fb5 NP |
90 | int i; |
91 | int sum = 0; | |
92 | for_each_possible_cpu(i) | |
93 | sum += per_cpu(nr_inodes, i); | |
94 | return sum < 0 ? 0 : sum; | |
cffbc8aa DC |
95 | } |
96 | ||
97 | static inline int get_nr_inodes_unused(void) | |
98 | { | |
fcb94f72 DC |
99 | int i; |
100 | int sum = 0; | |
101 | for_each_possible_cpu(i) | |
102 | sum += per_cpu(nr_unused, i); | |
103 | return sum < 0 ? 0 : sum; | |
cffbc8aa DC |
104 | } |
105 | ||
106 | int get_nr_dirty_inodes(void) | |
107 | { | |
3e880fb5 | 108 | /* not actually dirty inodes, but a wild approximation */ |
cffbc8aa DC |
109 | int nr_dirty = get_nr_inodes() - get_nr_inodes_unused(); |
110 | return nr_dirty > 0 ? nr_dirty : 0; | |
cffbc8aa DC |
111 | } |
112 | ||
113 | /* | |
114 | * Handle nr_inode sysctl | |
115 | */ | |
116 | #ifdef CONFIG_SYSCTL | |
117 | int proc_nr_inodes(ctl_table *table, int write, | |
118 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
119 | { | |
120 | inodes_stat.nr_inodes = get_nr_inodes(); | |
fcb94f72 | 121 | inodes_stat.nr_unused = get_nr_inodes_unused(); |
cffbc8aa DC |
122 | return proc_dointvec(table, write, buffer, lenp, ppos); |
123 | } | |
124 | #endif | |
125 | ||
2cb1599f DC |
126 | /** |
127 | * inode_init_always - perform inode structure intialisation | |
0bc02f3f RD |
128 | * @sb: superblock inode belongs to |
129 | * @inode: inode to initialise | |
2cb1599f DC |
130 | * |
131 | * These are initializations that need to be done on every inode | |
132 | * allocation as the fields are not initialised by slab allocation. | |
133 | */ | |
54e34621 | 134 | int inode_init_always(struct super_block *sb, struct inode *inode) |
1da177e4 | 135 | { |
6e1d5dcc | 136 | static const struct inode_operations empty_iops; |
99ac48f5 | 137 | static const struct file_operations empty_fops; |
6b3304b5 | 138 | struct address_space *const mapping = &inode->i_data; |
2cb1599f DC |
139 | |
140 | inode->i_sb = sb; | |
141 | inode->i_blkbits = sb->s_blocksize_bits; | |
142 | inode->i_flags = 0; | |
143 | atomic_set(&inode->i_count, 1); | |
144 | inode->i_op = &empty_iops; | |
145 | inode->i_fop = &empty_fops; | |
a78ef704 | 146 | inode->__i_nlink = 1; |
3ddcd056 | 147 | inode->i_opflags = 0; |
56ff5efa AV |
148 | inode->i_uid = 0; |
149 | inode->i_gid = 0; | |
2cb1599f DC |
150 | atomic_set(&inode->i_writecount, 0); |
151 | inode->i_size = 0; | |
152 | inode->i_blocks = 0; | |
153 | inode->i_bytes = 0; | |
154 | inode->i_generation = 0; | |
1da177e4 | 155 | #ifdef CONFIG_QUOTA |
2cb1599f | 156 | memset(&inode->i_dquot, 0, sizeof(inode->i_dquot)); |
1da177e4 | 157 | #endif |
2cb1599f DC |
158 | inode->i_pipe = NULL; |
159 | inode->i_bdev = NULL; | |
160 | inode->i_cdev = NULL; | |
161 | inode->i_rdev = 0; | |
162 | inode->dirtied_when = 0; | |
6146f0d5 MZ |
163 | |
164 | if (security_inode_alloc(inode)) | |
54e34621 | 165 | goto out; |
2cb1599f DC |
166 | spin_lock_init(&inode->i_lock); |
167 | lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key); | |
168 | ||
169 | mutex_init(&inode->i_mutex); | |
170 | lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key); | |
171 | ||
bd5fe6c5 | 172 | atomic_set(&inode->i_dio_count, 0); |
2cb1599f DC |
173 | |
174 | mapping->a_ops = &empty_aops; | |
175 | mapping->host = inode; | |
176 | mapping->flags = 0; | |
3c1d4378 | 177 | mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE); |
2cb1599f DC |
178 | mapping->assoc_mapping = NULL; |
179 | mapping->backing_dev_info = &default_backing_dev_info; | |
180 | mapping->writeback_index = 0; | |
181 | ||
182 | /* | |
183 | * If the block_device provides a backing_dev_info for client | |
184 | * inodes then use that. Otherwise the inode share the bdev's | |
185 | * backing_dev_info. | |
186 | */ | |
187 | if (sb->s_bdev) { | |
188 | struct backing_dev_info *bdi; | |
189 | ||
2c96ce9f | 190 | bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info; |
2cb1599f DC |
191 | mapping->backing_dev_info = bdi; |
192 | } | |
193 | inode->i_private = NULL; | |
194 | inode->i_mapping = mapping; | |
6b520e05 | 195 | INIT_LIST_HEAD(&inode->i_dentry); /* buggered by rcu freeing */ |
f19d4a8f AV |
196 | #ifdef CONFIG_FS_POSIX_ACL |
197 | inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED; | |
198 | #endif | |
2cb1599f | 199 | |
3be25f49 EP |
200 | #ifdef CONFIG_FSNOTIFY |
201 | inode->i_fsnotify_mask = 0; | |
202 | #endif | |
203 | ||
3e880fb5 | 204 | this_cpu_inc(nr_inodes); |
cffbc8aa | 205 | |
54e34621 | 206 | return 0; |
54e34621 CH |
207 | out: |
208 | return -ENOMEM; | |
1da177e4 | 209 | } |
2cb1599f DC |
210 | EXPORT_SYMBOL(inode_init_always); |
211 | ||
212 | static struct inode *alloc_inode(struct super_block *sb) | |
213 | { | |
214 | struct inode *inode; | |
215 | ||
216 | if (sb->s_op->alloc_inode) | |
217 | inode = sb->s_op->alloc_inode(sb); | |
218 | else | |
219 | inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL); | |
220 | ||
54e34621 CH |
221 | if (!inode) |
222 | return NULL; | |
223 | ||
224 | if (unlikely(inode_init_always(sb, inode))) { | |
225 | if (inode->i_sb->s_op->destroy_inode) | |
226 | inode->i_sb->s_op->destroy_inode(inode); | |
227 | else | |
228 | kmem_cache_free(inode_cachep, inode); | |
229 | return NULL; | |
230 | } | |
231 | ||
232 | return inode; | |
2cb1599f | 233 | } |
1da177e4 | 234 | |
ff0c7d15 NP |
235 | void free_inode_nonrcu(struct inode *inode) |
236 | { | |
237 | kmem_cache_free(inode_cachep, inode); | |
238 | } | |
239 | EXPORT_SYMBOL(free_inode_nonrcu); | |
240 | ||
2e00c97e | 241 | void __destroy_inode(struct inode *inode) |
1da177e4 | 242 | { |
b7542f8c | 243 | BUG_ON(inode_has_buffers(inode)); |
1da177e4 | 244 | security_inode_free(inode); |
3be25f49 | 245 | fsnotify_inode_delete(inode); |
7ada4db8 MS |
246 | if (!inode->i_nlink) { |
247 | WARN_ON(atomic_long_read(&inode->i_sb->s_remove_count) == 0); | |
248 | atomic_long_dec(&inode->i_sb->s_remove_count); | |
249 | } | |
250 | ||
f19d4a8f AV |
251 | #ifdef CONFIG_FS_POSIX_ACL |
252 | if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED) | |
253 | posix_acl_release(inode->i_acl); | |
254 | if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED) | |
255 | posix_acl_release(inode->i_default_acl); | |
256 | #endif | |
3e880fb5 | 257 | this_cpu_dec(nr_inodes); |
2e00c97e CH |
258 | } |
259 | EXPORT_SYMBOL(__destroy_inode); | |
260 | ||
fa0d7e3d NP |
261 | static void i_callback(struct rcu_head *head) |
262 | { | |
263 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
fa0d7e3d NP |
264 | kmem_cache_free(inode_cachep, inode); |
265 | } | |
266 | ||
56b0dacf | 267 | static void destroy_inode(struct inode *inode) |
2e00c97e | 268 | { |
7ccf19a8 | 269 | BUG_ON(!list_empty(&inode->i_lru)); |
2e00c97e | 270 | __destroy_inode(inode); |
1da177e4 LT |
271 | if (inode->i_sb->s_op->destroy_inode) |
272 | inode->i_sb->s_op->destroy_inode(inode); | |
273 | else | |
fa0d7e3d | 274 | call_rcu(&inode->i_rcu, i_callback); |
1da177e4 | 275 | } |
1da177e4 | 276 | |
7ada4db8 MS |
277 | /** |
278 | * drop_nlink - directly drop an inode's link count | |
279 | * @inode: inode | |
280 | * | |
281 | * This is a low-level filesystem helper to replace any | |
282 | * direct filesystem manipulation of i_nlink. In cases | |
283 | * where we are attempting to track writes to the | |
284 | * filesystem, a decrement to zero means an imminent | |
285 | * write when the file is truncated and actually unlinked | |
286 | * on the filesystem. | |
287 | */ | |
288 | void drop_nlink(struct inode *inode) | |
289 | { | |
290 | WARN_ON(inode->i_nlink == 0); | |
291 | inode->__i_nlink--; | |
292 | if (!inode->i_nlink) | |
293 | atomic_long_inc(&inode->i_sb->s_remove_count); | |
294 | } | |
295 | EXPORT_SYMBOL(drop_nlink); | |
296 | ||
297 | /** | |
298 | * clear_nlink - directly zero an inode's link count | |
299 | * @inode: inode | |
300 | * | |
301 | * This is a low-level filesystem helper to replace any | |
302 | * direct filesystem manipulation of i_nlink. See | |
303 | * drop_nlink() for why we care about i_nlink hitting zero. | |
304 | */ | |
305 | void clear_nlink(struct inode *inode) | |
306 | { | |
307 | if (inode->i_nlink) { | |
308 | inode->__i_nlink = 0; | |
309 | atomic_long_inc(&inode->i_sb->s_remove_count); | |
310 | } | |
311 | } | |
312 | EXPORT_SYMBOL(clear_nlink); | |
313 | ||
314 | /** | |
315 | * set_nlink - directly set an inode's link count | |
316 | * @inode: inode | |
317 | * @nlink: new nlink (should be non-zero) | |
318 | * | |
319 | * This is a low-level filesystem helper to replace any | |
320 | * direct filesystem manipulation of i_nlink. | |
321 | */ | |
322 | void set_nlink(struct inode *inode, unsigned int nlink) | |
323 | { | |
324 | if (!nlink) { | |
7ada4db8 MS |
325 | clear_nlink(inode); |
326 | } else { | |
327 | /* Yes, some filesystems do change nlink from zero to one */ | |
328 | if (inode->i_nlink == 0) | |
329 | atomic_long_dec(&inode->i_sb->s_remove_count); | |
330 | ||
331 | inode->__i_nlink = nlink; | |
332 | } | |
333 | } | |
334 | EXPORT_SYMBOL(set_nlink); | |
335 | ||
336 | /** | |
337 | * inc_nlink - directly increment an inode's link count | |
338 | * @inode: inode | |
339 | * | |
340 | * This is a low-level filesystem helper to replace any | |
341 | * direct filesystem manipulation of i_nlink. Currently, | |
342 | * it is only here for parity with dec_nlink(). | |
343 | */ | |
344 | void inc_nlink(struct inode *inode) | |
345 | { | |
346 | if (WARN_ON(inode->i_nlink == 0)) | |
347 | atomic_long_dec(&inode->i_sb->s_remove_count); | |
348 | ||
349 | inode->__i_nlink++; | |
350 | } | |
351 | EXPORT_SYMBOL(inc_nlink); | |
352 | ||
2aa15890 MS |
353 | void address_space_init_once(struct address_space *mapping) |
354 | { | |
355 | memset(mapping, 0, sizeof(*mapping)); | |
356 | INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC); | |
357 | spin_lock_init(&mapping->tree_lock); | |
3d48ae45 | 358 | mutex_init(&mapping->i_mmap_mutex); |
2aa15890 MS |
359 | INIT_LIST_HEAD(&mapping->private_list); |
360 | spin_lock_init(&mapping->private_lock); | |
361 | INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap); | |
362 | INIT_LIST_HEAD(&mapping->i_mmap_nonlinear); | |
2aa15890 MS |
363 | } |
364 | EXPORT_SYMBOL(address_space_init_once); | |
365 | ||
1da177e4 LT |
366 | /* |
367 | * These are initializations that only need to be done | |
368 | * once, because the fields are idempotent across use | |
369 | * of the inode, so let the slab aware of that. | |
370 | */ | |
371 | void inode_init_once(struct inode *inode) | |
372 | { | |
373 | memset(inode, 0, sizeof(*inode)); | |
374 | INIT_HLIST_NODE(&inode->i_hash); | |
1da177e4 | 375 | INIT_LIST_HEAD(&inode->i_devices); |
7ccf19a8 NP |
376 | INIT_LIST_HEAD(&inode->i_wb_list); |
377 | INIT_LIST_HEAD(&inode->i_lru); | |
2aa15890 | 378 | address_space_init_once(&inode->i_data); |
1da177e4 | 379 | i_size_ordered_init(inode); |
3be25f49 | 380 | #ifdef CONFIG_FSNOTIFY |
e61ce867 | 381 | INIT_HLIST_HEAD(&inode->i_fsnotify_marks); |
3be25f49 | 382 | #endif |
1da177e4 | 383 | } |
1da177e4 LT |
384 | EXPORT_SYMBOL(inode_init_once); |
385 | ||
51cc5068 | 386 | static void init_once(void *foo) |
1da177e4 | 387 | { |
6b3304b5 | 388 | struct inode *inode = (struct inode *) foo; |
1da177e4 | 389 | |
a35afb83 | 390 | inode_init_once(inode); |
1da177e4 LT |
391 | } |
392 | ||
393 | /* | |
250df6ed | 394 | * inode->i_lock must be held |
1da177e4 | 395 | */ |
6b3304b5 | 396 | void __iget(struct inode *inode) |
1da177e4 | 397 | { |
9e38d86f NP |
398 | atomic_inc(&inode->i_count); |
399 | } | |
2e147f1e | 400 | |
7de9c6ee AV |
401 | /* |
402 | * get additional reference to inode; caller must already hold one. | |
403 | */ | |
404 | void ihold(struct inode *inode) | |
405 | { | |
406 | WARN_ON(atomic_inc_return(&inode->i_count) < 2); | |
407 | } | |
408 | EXPORT_SYMBOL(ihold); | |
409 | ||
9e38d86f NP |
410 | static void inode_lru_list_add(struct inode *inode) |
411 | { | |
09cc9fc7 | 412 | spin_lock(&inode->i_sb->s_inode_lru_lock); |
7ccf19a8 | 413 | if (list_empty(&inode->i_lru)) { |
98b745c6 DC |
414 | list_add(&inode->i_lru, &inode->i_sb->s_inode_lru); |
415 | inode->i_sb->s_nr_inodes_unused++; | |
fcb94f72 | 416 | this_cpu_inc(nr_unused); |
9e38d86f | 417 | } |
09cc9fc7 | 418 | spin_unlock(&inode->i_sb->s_inode_lru_lock); |
9e38d86f | 419 | } |
2e147f1e | 420 | |
9e38d86f NP |
421 | static void inode_lru_list_del(struct inode *inode) |
422 | { | |
09cc9fc7 | 423 | spin_lock(&inode->i_sb->s_inode_lru_lock); |
7ccf19a8 NP |
424 | if (!list_empty(&inode->i_lru)) { |
425 | list_del_init(&inode->i_lru); | |
98b745c6 | 426 | inode->i_sb->s_nr_inodes_unused--; |
fcb94f72 | 427 | this_cpu_dec(nr_unused); |
9e38d86f | 428 | } |
09cc9fc7 | 429 | spin_unlock(&inode->i_sb->s_inode_lru_lock); |
1da177e4 LT |
430 | } |
431 | ||
646ec461 CH |
432 | /** |
433 | * inode_sb_list_add - add inode to the superblock list of inodes | |
434 | * @inode: inode to add | |
435 | */ | |
436 | void inode_sb_list_add(struct inode *inode) | |
437 | { | |
55fa6091 DC |
438 | spin_lock(&inode_sb_list_lock); |
439 | list_add(&inode->i_sb_list, &inode->i_sb->s_inodes); | |
440 | spin_unlock(&inode_sb_list_lock); | |
646ec461 CH |
441 | } |
442 | EXPORT_SYMBOL_GPL(inode_sb_list_add); | |
443 | ||
55fa6091 | 444 | static inline void inode_sb_list_del(struct inode *inode) |
646ec461 | 445 | { |
a209dfc7 ED |
446 | if (!list_empty(&inode->i_sb_list)) { |
447 | spin_lock(&inode_sb_list_lock); | |
448 | list_del_init(&inode->i_sb_list); | |
449 | spin_unlock(&inode_sb_list_lock); | |
450 | } | |
646ec461 CH |
451 | } |
452 | ||
4c51acbc DC |
453 | static unsigned long hash(struct super_block *sb, unsigned long hashval) |
454 | { | |
455 | unsigned long tmp; | |
456 | ||
457 | tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) / | |
458 | L1_CACHE_BYTES; | |
4b4563dc CH |
459 | tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> i_hash_shift); |
460 | return tmp & i_hash_mask; | |
4c51acbc DC |
461 | } |
462 | ||
463 | /** | |
464 | * __insert_inode_hash - hash an inode | |
465 | * @inode: unhashed inode | |
466 | * @hashval: unsigned long value used to locate this object in the | |
467 | * inode_hashtable. | |
468 | * | |
469 | * Add an inode to the inode hash for this superblock. | |
470 | */ | |
471 | void __insert_inode_hash(struct inode *inode, unsigned long hashval) | |
472 | { | |
646ec461 CH |
473 | struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval); |
474 | ||
67a23c49 | 475 | spin_lock(&inode_hash_lock); |
250df6ed | 476 | spin_lock(&inode->i_lock); |
646ec461 | 477 | hlist_add_head(&inode->i_hash, b); |
250df6ed | 478 | spin_unlock(&inode->i_lock); |
67a23c49 | 479 | spin_unlock(&inode_hash_lock); |
4c51acbc DC |
480 | } |
481 | EXPORT_SYMBOL(__insert_inode_hash); | |
482 | ||
4c51acbc | 483 | /** |
f2ee7abf | 484 | * __remove_inode_hash - remove an inode from the hash |
4c51acbc DC |
485 | * @inode: inode to unhash |
486 | * | |
487 | * Remove an inode from the superblock. | |
488 | */ | |
f2ee7abf | 489 | void __remove_inode_hash(struct inode *inode) |
4c51acbc | 490 | { |
67a23c49 | 491 | spin_lock(&inode_hash_lock); |
250df6ed | 492 | spin_lock(&inode->i_lock); |
4c51acbc | 493 | hlist_del_init(&inode->i_hash); |
250df6ed | 494 | spin_unlock(&inode->i_lock); |
67a23c49 | 495 | spin_unlock(&inode_hash_lock); |
4c51acbc | 496 | } |
f2ee7abf | 497 | EXPORT_SYMBOL(__remove_inode_hash); |
4c51acbc | 498 | |
b0683aa6 AV |
499 | void end_writeback(struct inode *inode) |
500 | { | |
501 | might_sleep(); | |
08142579 JK |
502 | /* |
503 | * We have to cycle tree_lock here because reclaim can be still in the | |
504 | * process of removing the last page (in __delete_from_page_cache()) | |
505 | * and we must not free mapping under it. | |
506 | */ | |
507 | spin_lock_irq(&inode->i_data.tree_lock); | |
b0683aa6 | 508 | BUG_ON(inode->i_data.nrpages); |
08142579 | 509 | spin_unlock_irq(&inode->i_data.tree_lock); |
b0683aa6 AV |
510 | BUG_ON(!list_empty(&inode->i_data.private_list)); |
511 | BUG_ON(!(inode->i_state & I_FREEING)); | |
512 | BUG_ON(inode->i_state & I_CLEAR); | |
513 | inode_sync_wait(inode); | |
fa0d7e3d | 514 | /* don't need i_lock here, no concurrent mods to i_state */ |
b0683aa6 AV |
515 | inode->i_state = I_FREEING | I_CLEAR; |
516 | } | |
517 | EXPORT_SYMBOL(end_writeback); | |
518 | ||
b2b2af8e DC |
519 | /* |
520 | * Free the inode passed in, removing it from the lists it is still connected | |
521 | * to. We remove any pages still attached to the inode and wait for any IO that | |
522 | * is still in progress before finally destroying the inode. | |
523 | * | |
524 | * An inode must already be marked I_FREEING so that we avoid the inode being | |
525 | * moved back onto lists if we race with other code that manipulates the lists | |
526 | * (e.g. writeback_single_inode). The caller is responsible for setting this. | |
527 | * | |
528 | * An inode must already be removed from the LRU list before being evicted from | |
529 | * the cache. This should occur atomically with setting the I_FREEING state | |
530 | * flag, so no inodes here should ever be on the LRU when being evicted. | |
531 | */ | |
644da596 | 532 | static void evict(struct inode *inode) |
b4272d4c AV |
533 | { |
534 | const struct super_operations *op = inode->i_sb->s_op; | |
535 | ||
b2b2af8e DC |
536 | BUG_ON(!(inode->i_state & I_FREEING)); |
537 | BUG_ON(!list_empty(&inode->i_lru)); | |
538 | ||
b12362bd ED |
539 | if (!list_empty(&inode->i_wb_list)) |
540 | inode_wb_list_del(inode); | |
541 | ||
55fa6091 DC |
542 | inode_sb_list_del(inode); |
543 | ||
be7ce416 AV |
544 | if (op->evict_inode) { |
545 | op->evict_inode(inode); | |
b4272d4c AV |
546 | } else { |
547 | if (inode->i_data.nrpages) | |
548 | truncate_inode_pages(&inode->i_data, 0); | |
30140837 | 549 | end_writeback(inode); |
b4272d4c | 550 | } |
661074e9 AV |
551 | if (S_ISBLK(inode->i_mode) && inode->i_bdev) |
552 | bd_forget(inode); | |
553 | if (S_ISCHR(inode->i_mode) && inode->i_cdev) | |
554 | cd_forget(inode); | |
b2b2af8e DC |
555 | |
556 | remove_inode_hash(inode); | |
557 | ||
558 | spin_lock(&inode->i_lock); | |
559 | wake_up_bit(&inode->i_state, __I_NEW); | |
560 | BUG_ON(inode->i_state != (I_FREEING | I_CLEAR)); | |
561 | spin_unlock(&inode->i_lock); | |
562 | ||
563 | destroy_inode(inode); | |
b4272d4c AV |
564 | } |
565 | ||
1da177e4 LT |
566 | /* |
567 | * dispose_list - dispose of the contents of a local list | |
568 | * @head: the head of the list to free | |
569 | * | |
570 | * Dispose-list gets a local list with local inodes in it, so it doesn't | |
571 | * need to worry about list corruption and SMP locks. | |
572 | */ | |
573 | static void dispose_list(struct list_head *head) | |
574 | { | |
1da177e4 LT |
575 | while (!list_empty(head)) { |
576 | struct inode *inode; | |
577 | ||
7ccf19a8 NP |
578 | inode = list_first_entry(head, struct inode, i_lru); |
579 | list_del_init(&inode->i_lru); | |
1da177e4 | 580 | |
644da596 | 581 | evict(inode); |
1da177e4 | 582 | } |
1da177e4 LT |
583 | } |
584 | ||
63997e98 AV |
585 | /** |
586 | * evict_inodes - evict all evictable inodes for a superblock | |
587 | * @sb: superblock to operate on | |
588 | * | |
589 | * Make sure that no inodes with zero refcount are retained. This is | |
590 | * called by superblock shutdown after having MS_ACTIVE flag removed, | |
591 | * so any inode reaching zero refcount during or after that call will | |
592 | * be immediately evicted. | |
1da177e4 | 593 | */ |
63997e98 | 594 | void evict_inodes(struct super_block *sb) |
1da177e4 | 595 | { |
63997e98 AV |
596 | struct inode *inode, *next; |
597 | LIST_HEAD(dispose); | |
1da177e4 | 598 | |
55fa6091 | 599 | spin_lock(&inode_sb_list_lock); |
63997e98 AV |
600 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
601 | if (atomic_read(&inode->i_count)) | |
aabb8fdb | 602 | continue; |
250df6ed DC |
603 | |
604 | spin_lock(&inode->i_lock); | |
605 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
606 | spin_unlock(&inode->i_lock); | |
1da177e4 | 607 | continue; |
250df6ed | 608 | } |
63997e98 AV |
609 | |
610 | inode->i_state |= I_FREEING; | |
02afc410 | 611 | inode_lru_list_del(inode); |
250df6ed | 612 | spin_unlock(&inode->i_lock); |
02afc410 | 613 | list_add(&inode->i_lru, &dispose); |
1da177e4 | 614 | } |
55fa6091 | 615 | spin_unlock(&inode_sb_list_lock); |
63997e98 AV |
616 | |
617 | dispose_list(&dispose); | |
1da177e4 LT |
618 | } |
619 | ||
1da177e4 | 620 | /** |
a0318786 CH |
621 | * invalidate_inodes - attempt to free all inodes on a superblock |
622 | * @sb: superblock to operate on | |
93b270f7 | 623 | * @kill_dirty: flag to guide handling of dirty inodes |
1da177e4 | 624 | * |
a0318786 CH |
625 | * Attempts to free all inodes for a given superblock. If there were any |
626 | * busy inodes return a non-zero value, else zero. | |
93b270f7 N |
627 | * If @kill_dirty is set, discard dirty inodes too, otherwise treat |
628 | * them as busy. | |
1da177e4 | 629 | */ |
93b270f7 | 630 | int invalidate_inodes(struct super_block *sb, bool kill_dirty) |
1da177e4 | 631 | { |
cffbc8aa | 632 | int busy = 0; |
a0318786 CH |
633 | struct inode *inode, *next; |
634 | LIST_HEAD(dispose); | |
1da177e4 | 635 | |
55fa6091 | 636 | spin_lock(&inode_sb_list_lock); |
a0318786 | 637 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
250df6ed DC |
638 | spin_lock(&inode->i_lock); |
639 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
640 | spin_unlock(&inode->i_lock); | |
aabb8fdb | 641 | continue; |
250df6ed | 642 | } |
93b270f7 | 643 | if (inode->i_state & I_DIRTY && !kill_dirty) { |
250df6ed | 644 | spin_unlock(&inode->i_lock); |
93b270f7 N |
645 | busy = 1; |
646 | continue; | |
647 | } | |
99a38919 | 648 | if (atomic_read(&inode->i_count)) { |
250df6ed | 649 | spin_unlock(&inode->i_lock); |
99a38919 | 650 | busy = 1; |
1da177e4 LT |
651 | continue; |
652 | } | |
99a38919 | 653 | |
99a38919 | 654 | inode->i_state |= I_FREEING; |
02afc410 | 655 | inode_lru_list_del(inode); |
250df6ed | 656 | spin_unlock(&inode->i_lock); |
02afc410 | 657 | list_add(&inode->i_lru, &dispose); |
1da177e4 | 658 | } |
55fa6091 | 659 | spin_unlock(&inode_sb_list_lock); |
1da177e4 | 660 | |
a0318786 | 661 | dispose_list(&dispose); |
1da177e4 LT |
662 | |
663 | return busy; | |
664 | } | |
1da177e4 LT |
665 | |
666 | static int can_unuse(struct inode *inode) | |
667 | { | |
9e38d86f | 668 | if (inode->i_state & ~I_REFERENCED) |
1da177e4 LT |
669 | return 0; |
670 | if (inode_has_buffers(inode)) | |
671 | return 0; | |
672 | if (atomic_read(&inode->i_count)) | |
673 | return 0; | |
674 | if (inode->i_data.nrpages) | |
675 | return 0; | |
676 | return 1; | |
677 | } | |
678 | ||
679 | /* | |
b0d40c92 DC |
680 | * Walk the superblock inode LRU for freeable inodes and attempt to free them. |
681 | * This is called from the superblock shrinker function with a number of inodes | |
682 | * to trim from the LRU. Inodes to be freed are moved to a temporary list and | |
683 | * then are freed outside inode_lock by dispose_list(). | |
1da177e4 LT |
684 | * |
685 | * Any inodes which are pinned purely because of attached pagecache have their | |
9e38d86f NP |
686 | * pagecache removed. If the inode has metadata buffers attached to |
687 | * mapping->private_list then try to remove them. | |
1da177e4 | 688 | * |
9e38d86f NP |
689 | * If the inode has the I_REFERENCED flag set, then it means that it has been |
690 | * used recently - the flag is set in iput_final(). When we encounter such an | |
691 | * inode, clear the flag and move it to the back of the LRU so it gets another | |
692 | * pass through the LRU before it gets reclaimed. This is necessary because of | |
693 | * the fact we are doing lazy LRU updates to minimise lock contention so the | |
694 | * LRU does not have strict ordering. Hence we don't want to reclaim inodes | |
695 | * with this flag set because they are the inodes that are out of order. | |
1da177e4 | 696 | */ |
b0d40c92 | 697 | void prune_icache_sb(struct super_block *sb, int nr_to_scan) |
1da177e4 LT |
698 | { |
699 | LIST_HEAD(freeable); | |
1da177e4 LT |
700 | int nr_scanned; |
701 | unsigned long reap = 0; | |
702 | ||
09cc9fc7 | 703 | spin_lock(&sb->s_inode_lru_lock); |
b0d40c92 | 704 | for (nr_scanned = nr_to_scan; nr_scanned >= 0; nr_scanned--) { |
1da177e4 LT |
705 | struct inode *inode; |
706 | ||
98b745c6 | 707 | if (list_empty(&sb->s_inode_lru)) |
1da177e4 LT |
708 | break; |
709 | ||
98b745c6 | 710 | inode = list_entry(sb->s_inode_lru.prev, struct inode, i_lru); |
1da177e4 | 711 | |
02afc410 | 712 | /* |
09cc9fc7 | 713 | * we are inverting the sb->s_inode_lru_lock/inode->i_lock here, |
02afc410 DC |
714 | * so use a trylock. If we fail to get the lock, just move the |
715 | * inode to the back of the list so we don't spin on it. | |
716 | */ | |
717 | if (!spin_trylock(&inode->i_lock)) { | |
62a3ddef | 718 | list_move_tail(&inode->i_lru, &sb->s_inode_lru); |
02afc410 DC |
719 | continue; |
720 | } | |
721 | ||
9e38d86f NP |
722 | /* |
723 | * Referenced or dirty inodes are still in use. Give them | |
724 | * another pass through the LRU as we canot reclaim them now. | |
725 | */ | |
726 | if (atomic_read(&inode->i_count) || | |
727 | (inode->i_state & ~I_REFERENCED)) { | |
7ccf19a8 | 728 | list_del_init(&inode->i_lru); |
f283c86a | 729 | spin_unlock(&inode->i_lock); |
98b745c6 | 730 | sb->s_nr_inodes_unused--; |
fcb94f72 | 731 | this_cpu_dec(nr_unused); |
9e38d86f NP |
732 | continue; |
733 | } | |
734 | ||
735 | /* recently referenced inodes get one more pass */ | |
736 | if (inode->i_state & I_REFERENCED) { | |
9e38d86f | 737 | inode->i_state &= ~I_REFERENCED; |
98b745c6 | 738 | list_move(&inode->i_lru, &sb->s_inode_lru); |
f283c86a | 739 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
740 | continue; |
741 | } | |
742 | if (inode_has_buffers(inode) || inode->i_data.nrpages) { | |
743 | __iget(inode); | |
250df6ed | 744 | spin_unlock(&inode->i_lock); |
09cc9fc7 | 745 | spin_unlock(&sb->s_inode_lru_lock); |
1da177e4 | 746 | if (remove_inode_buffers(inode)) |
fc0ecff6 AM |
747 | reap += invalidate_mapping_pages(&inode->i_data, |
748 | 0, -1); | |
1da177e4 | 749 | iput(inode); |
09cc9fc7 | 750 | spin_lock(&sb->s_inode_lru_lock); |
1da177e4 | 751 | |
98b745c6 | 752 | if (inode != list_entry(sb->s_inode_lru.next, |
7ccf19a8 | 753 | struct inode, i_lru)) |
1da177e4 | 754 | continue; /* wrong inode or list_empty */ |
02afc410 DC |
755 | /* avoid lock inversions with trylock */ |
756 | if (!spin_trylock(&inode->i_lock)) | |
757 | continue; | |
250df6ed DC |
758 | if (!can_unuse(inode)) { |
759 | spin_unlock(&inode->i_lock); | |
1da177e4 | 760 | continue; |
250df6ed | 761 | } |
1da177e4 | 762 | } |
7ef0d737 | 763 | WARN_ON(inode->i_state & I_NEW); |
1da177e4 | 764 | inode->i_state |= I_FREEING; |
250df6ed | 765 | spin_unlock(&inode->i_lock); |
7ccf19a8 | 766 | |
7ccf19a8 | 767 | list_move(&inode->i_lru, &freeable); |
98b745c6 | 768 | sb->s_nr_inodes_unused--; |
fcb94f72 | 769 | this_cpu_dec(nr_unused); |
1da177e4 | 770 | } |
f8891e5e CL |
771 | if (current_is_kswapd()) |
772 | __count_vm_events(KSWAPD_INODESTEAL, reap); | |
773 | else | |
774 | __count_vm_events(PGINODESTEAL, reap); | |
09cc9fc7 | 775 | spin_unlock(&sb->s_inode_lru_lock); |
5f8aefd4 KK |
776 | if (current->reclaim_state) |
777 | current->reclaim_state->reclaimed_slab += reap; | |
1da177e4 LT |
778 | |
779 | dispose_list(&freeable); | |
1da177e4 LT |
780 | } |
781 | ||
1da177e4 LT |
782 | static void __wait_on_freeing_inode(struct inode *inode); |
783 | /* | |
784 | * Called with the inode lock held. | |
1da177e4 | 785 | */ |
6b3304b5 MK |
786 | static struct inode *find_inode(struct super_block *sb, |
787 | struct hlist_head *head, | |
788 | int (*test)(struct inode *, void *), | |
789 | void *data) | |
1da177e4 LT |
790 | { |
791 | struct hlist_node *node; | |
6b3304b5 | 792 | struct inode *inode = NULL; |
1da177e4 LT |
793 | |
794 | repeat: | |
c5c8be3c | 795 | hlist_for_each_entry(inode, node, head, i_hash) { |
67a23c49 DC |
796 | spin_lock(&inode->i_lock); |
797 | if (inode->i_sb != sb) { | |
798 | spin_unlock(&inode->i_lock); | |
1da177e4 | 799 | continue; |
67a23c49 DC |
800 | } |
801 | if (!test(inode, data)) { | |
802 | spin_unlock(&inode->i_lock); | |
1da177e4 | 803 | continue; |
67a23c49 | 804 | } |
a4ffdde6 | 805 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
806 | __wait_on_freeing_inode(inode); |
807 | goto repeat; | |
808 | } | |
f7899bd5 | 809 | __iget(inode); |
250df6ed | 810 | spin_unlock(&inode->i_lock); |
f7899bd5 | 811 | return inode; |
1da177e4 | 812 | } |
f7899bd5 | 813 | return NULL; |
1da177e4 LT |
814 | } |
815 | ||
816 | /* | |
817 | * find_inode_fast is the fast path version of find_inode, see the comment at | |
818 | * iget_locked for details. | |
819 | */ | |
6b3304b5 MK |
820 | static struct inode *find_inode_fast(struct super_block *sb, |
821 | struct hlist_head *head, unsigned long ino) | |
1da177e4 LT |
822 | { |
823 | struct hlist_node *node; | |
6b3304b5 | 824 | struct inode *inode = NULL; |
1da177e4 LT |
825 | |
826 | repeat: | |
c5c8be3c | 827 | hlist_for_each_entry(inode, node, head, i_hash) { |
67a23c49 DC |
828 | spin_lock(&inode->i_lock); |
829 | if (inode->i_ino != ino) { | |
830 | spin_unlock(&inode->i_lock); | |
1da177e4 | 831 | continue; |
67a23c49 DC |
832 | } |
833 | if (inode->i_sb != sb) { | |
834 | spin_unlock(&inode->i_lock); | |
1da177e4 | 835 | continue; |
67a23c49 | 836 | } |
a4ffdde6 | 837 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
838 | __wait_on_freeing_inode(inode); |
839 | goto repeat; | |
840 | } | |
f7899bd5 | 841 | __iget(inode); |
250df6ed | 842 | spin_unlock(&inode->i_lock); |
f7899bd5 | 843 | return inode; |
1da177e4 | 844 | } |
f7899bd5 | 845 | return NULL; |
8290c35f DC |
846 | } |
847 | ||
f991bd2e ED |
848 | /* |
849 | * Each cpu owns a range of LAST_INO_BATCH numbers. | |
850 | * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations, | |
851 | * to renew the exhausted range. | |
8290c35f | 852 | * |
f991bd2e ED |
853 | * This does not significantly increase overflow rate because every CPU can |
854 | * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is | |
855 | * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the | |
856 | * 2^32 range, and is a worst-case. Even a 50% wastage would only increase | |
857 | * overflow rate by 2x, which does not seem too significant. | |
858 | * | |
859 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
860 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
861 | * here to attempt to avoid that. | |
8290c35f | 862 | */ |
f991bd2e ED |
863 | #define LAST_INO_BATCH 1024 |
864 | static DEFINE_PER_CPU(unsigned int, last_ino); | |
865 | ||
85fe4025 | 866 | unsigned int get_next_ino(void) |
8290c35f | 867 | { |
f991bd2e ED |
868 | unsigned int *p = &get_cpu_var(last_ino); |
869 | unsigned int res = *p; | |
8290c35f | 870 | |
f991bd2e ED |
871 | #ifdef CONFIG_SMP |
872 | if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) { | |
873 | static atomic_t shared_last_ino; | |
874 | int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino); | |
875 | ||
876 | res = next - LAST_INO_BATCH; | |
877 | } | |
878 | #endif | |
879 | ||
880 | *p = ++res; | |
881 | put_cpu_var(last_ino); | |
882 | return res; | |
8290c35f | 883 | } |
85fe4025 | 884 | EXPORT_SYMBOL(get_next_ino); |
8290c35f | 885 | |
a209dfc7 ED |
886 | /** |
887 | * new_inode_pseudo - obtain an inode | |
888 | * @sb: superblock | |
889 | * | |
890 | * Allocates a new inode for given superblock. | |
891 | * Inode wont be chained in superblock s_inodes list | |
892 | * This means : | |
893 | * - fs can't be unmount | |
894 | * - quotas, fsnotify, writeback can't work | |
895 | */ | |
896 | struct inode *new_inode_pseudo(struct super_block *sb) | |
897 | { | |
898 | struct inode *inode = alloc_inode(sb); | |
899 | ||
900 | if (inode) { | |
901 | spin_lock(&inode->i_lock); | |
902 | inode->i_state = 0; | |
903 | spin_unlock(&inode->i_lock); | |
904 | INIT_LIST_HEAD(&inode->i_sb_list); | |
905 | } | |
906 | return inode; | |
907 | } | |
908 | ||
1da177e4 LT |
909 | /** |
910 | * new_inode - obtain an inode | |
911 | * @sb: superblock | |
912 | * | |
769848c0 | 913 | * Allocates a new inode for given superblock. The default gfp_mask |
3c1d4378 | 914 | * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE. |
769848c0 MG |
915 | * If HIGHMEM pages are unsuitable or it is known that pages allocated |
916 | * for the page cache are not reclaimable or migratable, | |
917 | * mapping_set_gfp_mask() must be called with suitable flags on the | |
918 | * newly created inode's mapping | |
919 | * | |
1da177e4 LT |
920 | */ |
921 | struct inode *new_inode(struct super_block *sb) | |
922 | { | |
6b3304b5 | 923 | struct inode *inode; |
1da177e4 | 924 | |
55fa6091 | 925 | spin_lock_prefetch(&inode_sb_list_lock); |
6b3304b5 | 926 | |
a209dfc7 ED |
927 | inode = new_inode_pseudo(sb); |
928 | if (inode) | |
55fa6091 | 929 | inode_sb_list_add(inode); |
1da177e4 LT |
930 | return inode; |
931 | } | |
1da177e4 LT |
932 | EXPORT_SYMBOL(new_inode); |
933 | ||
14358e6d | 934 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
e096d0c7 JB |
935 | void lockdep_annotate_inode_mutex_key(struct inode *inode) |
936 | { | |
a3314a0e | 937 | if (S_ISDIR(inode->i_mode)) { |
1e89a5e1 PZ |
938 | struct file_system_type *type = inode->i_sb->s_type; |
939 | ||
9a7aa12f | 940 | /* Set new key only if filesystem hasn't already changed it */ |
978d6d8c | 941 | if (lockdep_match_class(&inode->i_mutex, &type->i_mutex_key)) { |
9a7aa12f JK |
942 | /* |
943 | * ensure nobody is actually holding i_mutex | |
944 | */ | |
945 | mutex_destroy(&inode->i_mutex); | |
946 | mutex_init(&inode->i_mutex); | |
947 | lockdep_set_class(&inode->i_mutex, | |
948 | &type->i_mutex_dir_key); | |
949 | } | |
1e89a5e1 | 950 | } |
e096d0c7 JB |
951 | } |
952 | EXPORT_SYMBOL(lockdep_annotate_inode_mutex_key); | |
14358e6d | 953 | #endif |
e096d0c7 JB |
954 | |
955 | /** | |
956 | * unlock_new_inode - clear the I_NEW state and wake up any waiters | |
957 | * @inode: new inode to unlock | |
958 | * | |
959 | * Called when the inode is fully initialised to clear the new state of the | |
960 | * inode and wake up anyone waiting for the inode to finish initialisation. | |
961 | */ | |
962 | void unlock_new_inode(struct inode *inode) | |
963 | { | |
964 | lockdep_annotate_inode_mutex_key(inode); | |
250df6ed | 965 | spin_lock(&inode->i_lock); |
eaff8079 CH |
966 | WARN_ON(!(inode->i_state & I_NEW)); |
967 | inode->i_state &= ~I_NEW; | |
250df6ed DC |
968 | wake_up_bit(&inode->i_state, __I_NEW); |
969 | spin_unlock(&inode->i_lock); | |
1da177e4 | 970 | } |
1da177e4 LT |
971 | EXPORT_SYMBOL(unlock_new_inode); |
972 | ||
0b2d0724 CH |
973 | /** |
974 | * iget5_locked - obtain an inode from a mounted file system | |
975 | * @sb: super block of file system | |
976 | * @hashval: hash value (usually inode number) to get | |
977 | * @test: callback used for comparisons between inodes | |
978 | * @set: callback used to initialize a new struct inode | |
979 | * @data: opaque data pointer to pass to @test and @set | |
980 | * | |
981 | * Search for the inode specified by @hashval and @data in the inode cache, | |
982 | * and if present it is return it with an increased reference count. This is | |
983 | * a generalized version of iget_locked() for file systems where the inode | |
984 | * number is not sufficient for unique identification of an inode. | |
985 | * | |
986 | * If the inode is not in cache, allocate a new inode and return it locked, | |
987 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
988 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 989 | * |
0b2d0724 CH |
990 | * Note both @test and @set are called with the inode_hash_lock held, so can't |
991 | * sleep. | |
1da177e4 | 992 | */ |
0b2d0724 CH |
993 | struct inode *iget5_locked(struct super_block *sb, unsigned long hashval, |
994 | int (*test)(struct inode *, void *), | |
995 | int (*set)(struct inode *, void *), void *data) | |
1da177e4 | 996 | { |
0b2d0724 | 997 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); |
6b3304b5 | 998 | struct inode *inode; |
1da177e4 | 999 | |
0b2d0724 CH |
1000 | spin_lock(&inode_hash_lock); |
1001 | inode = find_inode(sb, head, test, data); | |
1002 | spin_unlock(&inode_hash_lock); | |
1003 | ||
1004 | if (inode) { | |
1005 | wait_on_inode(inode); | |
1006 | return inode; | |
1007 | } | |
1008 | ||
1da177e4 LT |
1009 | inode = alloc_inode(sb); |
1010 | if (inode) { | |
6b3304b5 | 1011 | struct inode *old; |
1da177e4 | 1012 | |
67a23c49 | 1013 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1014 | /* We released the lock, so.. */ |
1015 | old = find_inode(sb, head, test, data); | |
1016 | if (!old) { | |
1017 | if (set(inode, data)) | |
1018 | goto set_failed; | |
1019 | ||
250df6ed DC |
1020 | spin_lock(&inode->i_lock); |
1021 | inode->i_state = I_NEW; | |
646ec461 | 1022 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1023 | spin_unlock(&inode->i_lock); |
55fa6091 | 1024 | inode_sb_list_add(inode); |
67a23c49 | 1025 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1026 | |
1027 | /* Return the locked inode with I_NEW set, the | |
1028 | * caller is responsible for filling in the contents | |
1029 | */ | |
1030 | return inode; | |
1031 | } | |
1032 | ||
1033 | /* | |
1034 | * Uhhuh, somebody else created the same inode under | |
1035 | * us. Use the old inode instead of the one we just | |
1036 | * allocated. | |
1037 | */ | |
67a23c49 | 1038 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1039 | destroy_inode(inode); |
1040 | inode = old; | |
1041 | wait_on_inode(inode); | |
1042 | } | |
1043 | return inode; | |
1044 | ||
1045 | set_failed: | |
67a23c49 | 1046 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1047 | destroy_inode(inode); |
1048 | return NULL; | |
1049 | } | |
0b2d0724 | 1050 | EXPORT_SYMBOL(iget5_locked); |
1da177e4 | 1051 | |
0b2d0724 CH |
1052 | /** |
1053 | * iget_locked - obtain an inode from a mounted file system | |
1054 | * @sb: super block of file system | |
1055 | * @ino: inode number to get | |
1056 | * | |
1057 | * Search for the inode specified by @ino in the inode cache and if present | |
1058 | * return it with an increased reference count. This is for file systems | |
1059 | * where the inode number is sufficient for unique identification of an inode. | |
1060 | * | |
1061 | * If the inode is not in cache, allocate a new inode and return it locked, | |
1062 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
1063 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 1064 | */ |
0b2d0724 | 1065 | struct inode *iget_locked(struct super_block *sb, unsigned long ino) |
1da177e4 | 1066 | { |
0b2d0724 | 1067 | struct hlist_head *head = inode_hashtable + hash(sb, ino); |
6b3304b5 | 1068 | struct inode *inode; |
1da177e4 | 1069 | |
0b2d0724 CH |
1070 | spin_lock(&inode_hash_lock); |
1071 | inode = find_inode_fast(sb, head, ino); | |
1072 | spin_unlock(&inode_hash_lock); | |
1073 | if (inode) { | |
1074 | wait_on_inode(inode); | |
1075 | return inode; | |
1076 | } | |
1077 | ||
1da177e4 LT |
1078 | inode = alloc_inode(sb); |
1079 | if (inode) { | |
6b3304b5 | 1080 | struct inode *old; |
1da177e4 | 1081 | |
67a23c49 | 1082 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1083 | /* We released the lock, so.. */ |
1084 | old = find_inode_fast(sb, head, ino); | |
1085 | if (!old) { | |
1086 | inode->i_ino = ino; | |
250df6ed DC |
1087 | spin_lock(&inode->i_lock); |
1088 | inode->i_state = I_NEW; | |
646ec461 | 1089 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1090 | spin_unlock(&inode->i_lock); |
55fa6091 | 1091 | inode_sb_list_add(inode); |
67a23c49 | 1092 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1093 | |
1094 | /* Return the locked inode with I_NEW set, the | |
1095 | * caller is responsible for filling in the contents | |
1096 | */ | |
1097 | return inode; | |
1098 | } | |
1099 | ||
1100 | /* | |
1101 | * Uhhuh, somebody else created the same inode under | |
1102 | * us. Use the old inode instead of the one we just | |
1103 | * allocated. | |
1104 | */ | |
67a23c49 | 1105 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1106 | destroy_inode(inode); |
1107 | inode = old; | |
1108 | wait_on_inode(inode); | |
1109 | } | |
1110 | return inode; | |
1111 | } | |
0b2d0724 | 1112 | EXPORT_SYMBOL(iget_locked); |
1da177e4 | 1113 | |
ad5e195a CH |
1114 | /* |
1115 | * search the inode cache for a matching inode number. | |
1116 | * If we find one, then the inode number we are trying to | |
1117 | * allocate is not unique and so we should not use it. | |
1118 | * | |
1119 | * Returns 1 if the inode number is unique, 0 if it is not. | |
1120 | */ | |
1121 | static int test_inode_iunique(struct super_block *sb, unsigned long ino) | |
1122 | { | |
1123 | struct hlist_head *b = inode_hashtable + hash(sb, ino); | |
1124 | struct hlist_node *node; | |
1125 | struct inode *inode; | |
1126 | ||
67a23c49 | 1127 | spin_lock(&inode_hash_lock); |
ad5e195a | 1128 | hlist_for_each_entry(inode, node, b, i_hash) { |
67a23c49 DC |
1129 | if (inode->i_ino == ino && inode->i_sb == sb) { |
1130 | spin_unlock(&inode_hash_lock); | |
ad5e195a | 1131 | return 0; |
67a23c49 | 1132 | } |
ad5e195a | 1133 | } |
67a23c49 | 1134 | spin_unlock(&inode_hash_lock); |
ad5e195a CH |
1135 | |
1136 | return 1; | |
1137 | } | |
1138 | ||
1da177e4 LT |
1139 | /** |
1140 | * iunique - get a unique inode number | |
1141 | * @sb: superblock | |
1142 | * @max_reserved: highest reserved inode number | |
1143 | * | |
1144 | * Obtain an inode number that is unique on the system for a given | |
1145 | * superblock. This is used by file systems that have no natural | |
1146 | * permanent inode numbering system. An inode number is returned that | |
1147 | * is higher than the reserved limit but unique. | |
1148 | * | |
1149 | * BUGS: | |
1150 | * With a large number of inodes live on the file system this function | |
1151 | * currently becomes quite slow. | |
1152 | */ | |
1153 | ino_t iunique(struct super_block *sb, ino_t max_reserved) | |
1154 | { | |
866b04fc JL |
1155 | /* |
1156 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
1157 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
1158 | * here to attempt to avoid that. | |
1159 | */ | |
ad5e195a | 1160 | static DEFINE_SPINLOCK(iunique_lock); |
866b04fc | 1161 | static unsigned int counter; |
1da177e4 | 1162 | ino_t res; |
3361c7be | 1163 | |
ad5e195a | 1164 | spin_lock(&iunique_lock); |
3361c7be JL |
1165 | do { |
1166 | if (counter <= max_reserved) | |
1167 | counter = max_reserved + 1; | |
1da177e4 | 1168 | res = counter++; |
ad5e195a CH |
1169 | } while (!test_inode_iunique(sb, res)); |
1170 | spin_unlock(&iunique_lock); | |
1da177e4 | 1171 | |
3361c7be JL |
1172 | return res; |
1173 | } | |
1da177e4 LT |
1174 | EXPORT_SYMBOL(iunique); |
1175 | ||
1176 | struct inode *igrab(struct inode *inode) | |
1177 | { | |
250df6ed DC |
1178 | spin_lock(&inode->i_lock); |
1179 | if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) { | |
1da177e4 | 1180 | __iget(inode); |
250df6ed DC |
1181 | spin_unlock(&inode->i_lock); |
1182 | } else { | |
1183 | spin_unlock(&inode->i_lock); | |
1da177e4 LT |
1184 | /* |
1185 | * Handle the case where s_op->clear_inode is not been | |
1186 | * called yet, and somebody is calling igrab | |
1187 | * while the inode is getting freed. | |
1188 | */ | |
1189 | inode = NULL; | |
250df6ed | 1190 | } |
1da177e4 LT |
1191 | return inode; |
1192 | } | |
1da177e4 LT |
1193 | EXPORT_SYMBOL(igrab); |
1194 | ||
1195 | /** | |
0b2d0724 | 1196 | * ilookup5_nowait - search for an inode in the inode cache |
1da177e4 | 1197 | * @sb: super block of file system to search |
0b2d0724 | 1198 | * @hashval: hash value (usually inode number) to search for |
1da177e4 LT |
1199 | * @test: callback used for comparisons between inodes |
1200 | * @data: opaque data pointer to pass to @test | |
1da177e4 | 1201 | * |
0b2d0724 | 1202 | * Search for the inode specified by @hashval and @data in the inode cache. |
1da177e4 LT |
1203 | * If the inode is in the cache, the inode is returned with an incremented |
1204 | * reference count. | |
1205 | * | |
0b2d0724 CH |
1206 | * Note: I_NEW is not waited upon so you have to be very careful what you do |
1207 | * with the returned inode. You probably should be using ilookup5() instead. | |
1da177e4 | 1208 | * |
b6d0ad68 | 1209 | * Note2: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 | 1210 | */ |
0b2d0724 CH |
1211 | struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval, |
1212 | int (*test)(struct inode *, void *), void *data) | |
1da177e4 | 1213 | { |
0b2d0724 | 1214 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); |
1da177e4 LT |
1215 | struct inode *inode; |
1216 | ||
67a23c49 | 1217 | spin_lock(&inode_hash_lock); |
1da177e4 | 1218 | inode = find_inode(sb, head, test, data); |
67a23c49 | 1219 | spin_unlock(&inode_hash_lock); |
88bd5121 | 1220 | |
0b2d0724 | 1221 | return inode; |
88bd5121 | 1222 | } |
88bd5121 AA |
1223 | EXPORT_SYMBOL(ilookup5_nowait); |
1224 | ||
1225 | /** | |
1226 | * ilookup5 - search for an inode in the inode cache | |
1227 | * @sb: super block of file system to search | |
1228 | * @hashval: hash value (usually inode number) to search for | |
1229 | * @test: callback used for comparisons between inodes | |
1230 | * @data: opaque data pointer to pass to @test | |
1231 | * | |
0b2d0724 CH |
1232 | * Search for the inode specified by @hashval and @data in the inode cache, |
1233 | * and if the inode is in the cache, return the inode with an incremented | |
1234 | * reference count. Waits on I_NEW before returning the inode. | |
88bd5121 | 1235 | * returned with an incremented reference count. |
1da177e4 | 1236 | * |
0b2d0724 CH |
1237 | * This is a generalized version of ilookup() for file systems where the |
1238 | * inode number is not sufficient for unique identification of an inode. | |
1da177e4 | 1239 | * |
0b2d0724 | 1240 | * Note: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 LT |
1241 | */ |
1242 | struct inode *ilookup5(struct super_block *sb, unsigned long hashval, | |
1243 | int (*test)(struct inode *, void *), void *data) | |
1244 | { | |
0b2d0724 | 1245 | struct inode *inode = ilookup5_nowait(sb, hashval, test, data); |
1da177e4 | 1246 | |
0b2d0724 CH |
1247 | if (inode) |
1248 | wait_on_inode(inode); | |
1249 | return inode; | |
1da177e4 | 1250 | } |
1da177e4 LT |
1251 | EXPORT_SYMBOL(ilookup5); |
1252 | ||
1253 | /** | |
1254 | * ilookup - search for an inode in the inode cache | |
1255 | * @sb: super block of file system to search | |
1256 | * @ino: inode number to search for | |
1257 | * | |
0b2d0724 CH |
1258 | * Search for the inode @ino in the inode cache, and if the inode is in the |
1259 | * cache, the inode is returned with an incremented reference count. | |
1da177e4 LT |
1260 | */ |
1261 | struct inode *ilookup(struct super_block *sb, unsigned long ino) | |
1262 | { | |
1263 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1da177e4 LT |
1264 | struct inode *inode; |
1265 | ||
0b2d0724 CH |
1266 | spin_lock(&inode_hash_lock); |
1267 | inode = find_inode_fast(sb, head, ino); | |
1268 | spin_unlock(&inode_hash_lock); | |
1da177e4 | 1269 | |
1da177e4 | 1270 | if (inode) |
0b2d0724 CH |
1271 | wait_on_inode(inode); |
1272 | return inode; | |
1da177e4 | 1273 | } |
0b2d0724 | 1274 | EXPORT_SYMBOL(ilookup); |
1da177e4 | 1275 | |
261bca86 AV |
1276 | int insert_inode_locked(struct inode *inode) |
1277 | { | |
1278 | struct super_block *sb = inode->i_sb; | |
1279 | ino_t ino = inode->i_ino; | |
1280 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
261bca86 | 1281 | |
261bca86 | 1282 | while (1) { |
72a43d63 AV |
1283 | struct hlist_node *node; |
1284 | struct inode *old = NULL; | |
67a23c49 | 1285 | spin_lock(&inode_hash_lock); |
72a43d63 AV |
1286 | hlist_for_each_entry(old, node, head, i_hash) { |
1287 | if (old->i_ino != ino) | |
1288 | continue; | |
1289 | if (old->i_sb != sb) | |
1290 | continue; | |
250df6ed DC |
1291 | spin_lock(&old->i_lock); |
1292 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1293 | spin_unlock(&old->i_lock); | |
72a43d63 | 1294 | continue; |
250df6ed | 1295 | } |
72a43d63 AV |
1296 | break; |
1297 | } | |
1298 | if (likely(!node)) { | |
250df6ed DC |
1299 | spin_lock(&inode->i_lock); |
1300 | inode->i_state |= I_NEW; | |
261bca86 | 1301 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1302 | spin_unlock(&inode->i_lock); |
67a23c49 | 1303 | spin_unlock(&inode_hash_lock); |
261bca86 AV |
1304 | return 0; |
1305 | } | |
1306 | __iget(old); | |
250df6ed | 1307 | spin_unlock(&old->i_lock); |
67a23c49 | 1308 | spin_unlock(&inode_hash_lock); |
261bca86 | 1309 | wait_on_inode(old); |
1d3382cb | 1310 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1311 | iput(old); |
1312 | return -EBUSY; | |
1313 | } | |
1314 | iput(old); | |
1315 | } | |
1316 | } | |
261bca86 AV |
1317 | EXPORT_SYMBOL(insert_inode_locked); |
1318 | ||
1319 | int insert_inode_locked4(struct inode *inode, unsigned long hashval, | |
1320 | int (*test)(struct inode *, void *), void *data) | |
1321 | { | |
1322 | struct super_block *sb = inode->i_sb; | |
1323 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
261bca86 | 1324 | |
261bca86 | 1325 | while (1) { |
72a43d63 AV |
1326 | struct hlist_node *node; |
1327 | struct inode *old = NULL; | |
1328 | ||
67a23c49 | 1329 | spin_lock(&inode_hash_lock); |
72a43d63 AV |
1330 | hlist_for_each_entry(old, node, head, i_hash) { |
1331 | if (old->i_sb != sb) | |
1332 | continue; | |
1333 | if (!test(old, data)) | |
1334 | continue; | |
250df6ed DC |
1335 | spin_lock(&old->i_lock); |
1336 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1337 | spin_unlock(&old->i_lock); | |
72a43d63 | 1338 | continue; |
250df6ed | 1339 | } |
72a43d63 AV |
1340 | break; |
1341 | } | |
1342 | if (likely(!node)) { | |
250df6ed DC |
1343 | spin_lock(&inode->i_lock); |
1344 | inode->i_state |= I_NEW; | |
261bca86 | 1345 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1346 | spin_unlock(&inode->i_lock); |
67a23c49 | 1347 | spin_unlock(&inode_hash_lock); |
261bca86 AV |
1348 | return 0; |
1349 | } | |
1350 | __iget(old); | |
250df6ed | 1351 | spin_unlock(&old->i_lock); |
67a23c49 | 1352 | spin_unlock(&inode_hash_lock); |
261bca86 | 1353 | wait_on_inode(old); |
1d3382cb | 1354 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1355 | iput(old); |
1356 | return -EBUSY; | |
1357 | } | |
1358 | iput(old); | |
1359 | } | |
1360 | } | |
261bca86 AV |
1361 | EXPORT_SYMBOL(insert_inode_locked4); |
1362 | ||
1da177e4 | 1363 | |
45321ac5 AV |
1364 | int generic_delete_inode(struct inode *inode) |
1365 | { | |
1366 | return 1; | |
1367 | } | |
1368 | EXPORT_SYMBOL(generic_delete_inode); | |
1369 | ||
1da177e4 | 1370 | /* |
45321ac5 AV |
1371 | * Normal UNIX filesystem behaviour: delete the |
1372 | * inode when the usage count drops to zero, and | |
1373 | * i_nlink is zero. | |
1da177e4 | 1374 | */ |
45321ac5 | 1375 | int generic_drop_inode(struct inode *inode) |
1da177e4 | 1376 | { |
1d3382cb | 1377 | return !inode->i_nlink || inode_unhashed(inode); |
1da177e4 | 1378 | } |
45321ac5 | 1379 | EXPORT_SYMBOL_GPL(generic_drop_inode); |
1da177e4 | 1380 | |
45321ac5 AV |
1381 | /* |
1382 | * Called when we're dropping the last reference | |
1383 | * to an inode. | |
22fe4042 | 1384 | * |
45321ac5 AV |
1385 | * Call the FS "drop_inode()" function, defaulting to |
1386 | * the legacy UNIX filesystem behaviour. If it tells | |
1387 | * us to evict inode, do so. Otherwise, retain inode | |
1388 | * in cache if fs is alive, sync and evict if fs is | |
1389 | * shutting down. | |
22fe4042 | 1390 | */ |
45321ac5 | 1391 | static void iput_final(struct inode *inode) |
1da177e4 LT |
1392 | { |
1393 | struct super_block *sb = inode->i_sb; | |
45321ac5 AV |
1394 | const struct super_operations *op = inode->i_sb->s_op; |
1395 | int drop; | |
1396 | ||
250df6ed DC |
1397 | WARN_ON(inode->i_state & I_NEW); |
1398 | ||
e7f59097 | 1399 | if (op->drop_inode) |
45321ac5 AV |
1400 | drop = op->drop_inode(inode); |
1401 | else | |
1402 | drop = generic_drop_inode(inode); | |
1da177e4 | 1403 | |
b2b2af8e DC |
1404 | if (!drop && (sb->s_flags & MS_ACTIVE)) { |
1405 | inode->i_state |= I_REFERENCED; | |
1406 | if (!(inode->i_state & (I_DIRTY|I_SYNC))) | |
1407 | inode_lru_list_add(inode); | |
1408 | spin_unlock(&inode->i_lock); | |
b2b2af8e DC |
1409 | return; |
1410 | } | |
1411 | ||
45321ac5 | 1412 | if (!drop) { |
991114c6 | 1413 | inode->i_state |= I_WILL_FREE; |
250df6ed | 1414 | spin_unlock(&inode->i_lock); |
1da177e4 | 1415 | write_inode_now(inode, 1); |
250df6ed | 1416 | spin_lock(&inode->i_lock); |
7ef0d737 | 1417 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1418 | inode->i_state &= ~I_WILL_FREE; |
1da177e4 | 1419 | } |
7ccf19a8 | 1420 | |
991114c6 | 1421 | inode->i_state |= I_FREEING; |
c4ae0c65 ED |
1422 | if (!list_empty(&inode->i_lru)) |
1423 | inode_lru_list_del(inode); | |
b2b2af8e | 1424 | spin_unlock(&inode->i_lock); |
b2b2af8e | 1425 | |
644da596 | 1426 | evict(inode); |
1da177e4 LT |
1427 | } |
1428 | ||
1da177e4 | 1429 | /** |
6b3304b5 | 1430 | * iput - put an inode |
1da177e4 LT |
1431 | * @inode: inode to put |
1432 | * | |
1433 | * Puts an inode, dropping its usage count. If the inode use count hits | |
1434 | * zero, the inode is then freed and may also be destroyed. | |
1435 | * | |
1436 | * Consequently, iput() can sleep. | |
1437 | */ | |
1438 | void iput(struct inode *inode) | |
1439 | { | |
1440 | if (inode) { | |
a4ffdde6 | 1441 | BUG_ON(inode->i_state & I_CLEAR); |
1da177e4 | 1442 | |
f283c86a | 1443 | if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock)) |
1da177e4 LT |
1444 | iput_final(inode); |
1445 | } | |
1446 | } | |
1da177e4 LT |
1447 | EXPORT_SYMBOL(iput); |
1448 | ||
1449 | /** | |
1450 | * bmap - find a block number in a file | |
1451 | * @inode: inode of file | |
1452 | * @block: block to find | |
1453 | * | |
1454 | * Returns the block number on the device holding the inode that | |
1455 | * is the disk block number for the block of the file requested. | |
1456 | * That is, asked for block 4 of inode 1 the function will return the | |
6b3304b5 | 1457 | * disk block relative to the disk start that holds that block of the |
1da177e4 LT |
1458 | * file. |
1459 | */ | |
6b3304b5 | 1460 | sector_t bmap(struct inode *inode, sector_t block) |
1da177e4 LT |
1461 | { |
1462 | sector_t res = 0; | |
1463 | if (inode->i_mapping->a_ops->bmap) | |
1464 | res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block); | |
1465 | return res; | |
1466 | } | |
1da177e4 LT |
1467 | EXPORT_SYMBOL(bmap); |
1468 | ||
11ff6f05 MG |
1469 | /* |
1470 | * With relative atime, only update atime if the previous atime is | |
1471 | * earlier than either the ctime or mtime or if at least a day has | |
1472 | * passed since the last atime update. | |
1473 | */ | |
1474 | static int relatime_need_update(struct vfsmount *mnt, struct inode *inode, | |
1475 | struct timespec now) | |
1476 | { | |
1477 | ||
1478 | if (!(mnt->mnt_flags & MNT_RELATIME)) | |
1479 | return 1; | |
1480 | /* | |
1481 | * Is mtime younger than atime? If yes, update atime: | |
1482 | */ | |
1483 | if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0) | |
1484 | return 1; | |
1485 | /* | |
1486 | * Is ctime younger than atime? If yes, update atime: | |
1487 | */ | |
1488 | if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0) | |
1489 | return 1; | |
1490 | ||
1491 | /* | |
1492 | * Is the previous atime value older than a day? If yes, | |
1493 | * update atime: | |
1494 | */ | |
1495 | if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60) | |
1496 | return 1; | |
1497 | /* | |
1498 | * Good, we can skip the atime update: | |
1499 | */ | |
1500 | return 0; | |
1501 | } | |
1502 | ||
1da177e4 | 1503 | /** |
869243a0 CH |
1504 | * touch_atime - update the access time |
1505 | * @mnt: mount the inode is accessed on | |
7045f37b | 1506 | * @dentry: dentry accessed |
1da177e4 LT |
1507 | * |
1508 | * Update the accessed time on an inode and mark it for writeback. | |
1509 | * This function automatically handles read only file systems and media, | |
1510 | * as well as the "noatime" flag and inode specific "noatime" markers. | |
1511 | */ | |
869243a0 | 1512 | void touch_atime(struct vfsmount *mnt, struct dentry *dentry) |
1da177e4 | 1513 | { |
869243a0 | 1514 | struct inode *inode = dentry->d_inode; |
1da177e4 LT |
1515 | struct timespec now; |
1516 | ||
cdb70f3f | 1517 | if (inode->i_flags & S_NOATIME) |
b12536c2 | 1518 | return; |
37756ced | 1519 | if (IS_NOATIME(inode)) |
b12536c2 | 1520 | return; |
b2276138 | 1521 | if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1522 | return; |
47ae32d6 | 1523 | |
cdb70f3f | 1524 | if (mnt->mnt_flags & MNT_NOATIME) |
b12536c2 | 1525 | return; |
cdb70f3f | 1526 | if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1527 | return; |
1da177e4 LT |
1528 | |
1529 | now = current_fs_time(inode->i_sb); | |
11ff6f05 MG |
1530 | |
1531 | if (!relatime_need_update(mnt, inode, now)) | |
b12536c2 | 1532 | return; |
11ff6f05 | 1533 | |
47ae32d6 | 1534 | if (timespec_equal(&inode->i_atime, &now)) |
b12536c2 AK |
1535 | return; |
1536 | ||
1537 | if (mnt_want_write(mnt)) | |
1538 | return; | |
47ae32d6 VH |
1539 | |
1540 | inode->i_atime = now; | |
1541 | mark_inode_dirty_sync(inode); | |
cdb70f3f | 1542 | mnt_drop_write(mnt); |
1da177e4 | 1543 | } |
869243a0 | 1544 | EXPORT_SYMBOL(touch_atime); |
1da177e4 LT |
1545 | |
1546 | /** | |
870f4817 CH |
1547 | * file_update_time - update mtime and ctime time |
1548 | * @file: file accessed | |
1da177e4 | 1549 | * |
870f4817 CH |
1550 | * Update the mtime and ctime members of an inode and mark the inode |
1551 | * for writeback. Note that this function is meant exclusively for | |
1552 | * usage in the file write path of filesystems, and filesystems may | |
1553 | * choose to explicitly ignore update via this function with the | |
2eadfc0e | 1554 | * S_NOCMTIME inode flag, e.g. for network filesystem where these |
870f4817 | 1555 | * timestamps are handled by the server. |
1da177e4 LT |
1556 | */ |
1557 | ||
870f4817 | 1558 | void file_update_time(struct file *file) |
1da177e4 | 1559 | { |
0f7fc9e4 | 1560 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 | 1561 | struct timespec now; |
ce06e0b2 | 1562 | enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0; |
1da177e4 | 1563 | |
ce06e0b2 | 1564 | /* First try to exhaust all avenues to not sync */ |
1da177e4 LT |
1565 | if (IS_NOCMTIME(inode)) |
1566 | return; | |
20ddee2c | 1567 | |
1da177e4 | 1568 | now = current_fs_time(inode->i_sb); |
ce06e0b2 AK |
1569 | if (!timespec_equal(&inode->i_mtime, &now)) |
1570 | sync_it = S_MTIME; | |
1da177e4 | 1571 | |
ce06e0b2 AK |
1572 | if (!timespec_equal(&inode->i_ctime, &now)) |
1573 | sync_it |= S_CTIME; | |
870f4817 | 1574 | |
ce06e0b2 AK |
1575 | if (IS_I_VERSION(inode)) |
1576 | sync_it |= S_VERSION; | |
7a224228 | 1577 | |
ce06e0b2 AK |
1578 | if (!sync_it) |
1579 | return; | |
1580 | ||
1581 | /* Finally allowed to write? Takes lock. */ | |
1582 | if (mnt_want_write_file(file)) | |
1583 | return; | |
1584 | ||
1585 | /* Only change inode inside the lock region */ | |
1586 | if (sync_it & S_VERSION) | |
1587 | inode_inc_iversion(inode); | |
1588 | if (sync_it & S_CTIME) | |
1589 | inode->i_ctime = now; | |
1590 | if (sync_it & S_MTIME) | |
1591 | inode->i_mtime = now; | |
1592 | mark_inode_dirty_sync(inode); | |
2a79f17e | 1593 | mnt_drop_write_file(file); |
1da177e4 | 1594 | } |
870f4817 | 1595 | EXPORT_SYMBOL(file_update_time); |
1da177e4 LT |
1596 | |
1597 | int inode_needs_sync(struct inode *inode) | |
1598 | { | |
1599 | if (IS_SYNC(inode)) | |
1600 | return 1; | |
1601 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) | |
1602 | return 1; | |
1603 | return 0; | |
1604 | } | |
1da177e4 LT |
1605 | EXPORT_SYMBOL(inode_needs_sync); |
1606 | ||
1da177e4 LT |
1607 | int inode_wait(void *word) |
1608 | { | |
1609 | schedule(); | |
1610 | return 0; | |
1611 | } | |
d44dab8d | 1612 | EXPORT_SYMBOL(inode_wait); |
1da177e4 LT |
1613 | |
1614 | /* | |
168a9fd6 MS |
1615 | * If we try to find an inode in the inode hash while it is being |
1616 | * deleted, we have to wait until the filesystem completes its | |
1617 | * deletion before reporting that it isn't found. This function waits | |
1618 | * until the deletion _might_ have completed. Callers are responsible | |
1619 | * to recheck inode state. | |
1620 | * | |
eaff8079 | 1621 | * It doesn't matter if I_NEW is not set initially, a call to |
250df6ed DC |
1622 | * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list |
1623 | * will DTRT. | |
1da177e4 LT |
1624 | */ |
1625 | static void __wait_on_freeing_inode(struct inode *inode) | |
1626 | { | |
1627 | wait_queue_head_t *wq; | |
eaff8079 CH |
1628 | DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW); |
1629 | wq = bit_waitqueue(&inode->i_state, __I_NEW); | |
1da177e4 | 1630 | prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE); |
250df6ed | 1631 | spin_unlock(&inode->i_lock); |
67a23c49 | 1632 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1633 | schedule(); |
1634 | finish_wait(wq, &wait.wait); | |
67a23c49 | 1635 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1636 | } |
1637 | ||
1da177e4 LT |
1638 | static __initdata unsigned long ihash_entries; |
1639 | static int __init set_ihash_entries(char *str) | |
1640 | { | |
1641 | if (!str) | |
1642 | return 0; | |
1643 | ihash_entries = simple_strtoul(str, &str, 0); | |
1644 | return 1; | |
1645 | } | |
1646 | __setup("ihash_entries=", set_ihash_entries); | |
1647 | ||
1648 | /* | |
1649 | * Initialize the waitqueues and inode hash table. | |
1650 | */ | |
1651 | void __init inode_init_early(void) | |
1652 | { | |
074b8517 | 1653 | unsigned int loop; |
1da177e4 LT |
1654 | |
1655 | /* If hashes are distributed across NUMA nodes, defer | |
1656 | * hash allocation until vmalloc space is available. | |
1657 | */ | |
1658 | if (hashdist) | |
1659 | return; | |
1660 | ||
1661 | inode_hashtable = | |
1662 | alloc_large_system_hash("Inode-cache", | |
1663 | sizeof(struct hlist_head), | |
1664 | ihash_entries, | |
1665 | 14, | |
1666 | HASH_EARLY, | |
1667 | &i_hash_shift, | |
1668 | &i_hash_mask, | |
1669 | 0); | |
1670 | ||
074b8517 | 1671 | for (loop = 0; loop < (1U << i_hash_shift); loop++) |
1da177e4 LT |
1672 | INIT_HLIST_HEAD(&inode_hashtable[loop]); |
1673 | } | |
1674 | ||
74bf17cf | 1675 | void __init inode_init(void) |
1da177e4 | 1676 | { |
074b8517 | 1677 | unsigned int loop; |
1da177e4 LT |
1678 | |
1679 | /* inode slab cache */ | |
b0196009 PJ |
1680 | inode_cachep = kmem_cache_create("inode_cache", |
1681 | sizeof(struct inode), | |
1682 | 0, | |
1683 | (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC| | |
1684 | SLAB_MEM_SPREAD), | |
20c2df83 | 1685 | init_once); |
1da177e4 LT |
1686 | |
1687 | /* Hash may have been set up in inode_init_early */ | |
1688 | if (!hashdist) | |
1689 | return; | |
1690 | ||
1691 | inode_hashtable = | |
1692 | alloc_large_system_hash("Inode-cache", | |
1693 | sizeof(struct hlist_head), | |
1694 | ihash_entries, | |
1695 | 14, | |
1696 | 0, | |
1697 | &i_hash_shift, | |
1698 | &i_hash_mask, | |
1699 | 0); | |
1700 | ||
074b8517 | 1701 | for (loop = 0; loop < (1U << i_hash_shift); loop++) |
1da177e4 LT |
1702 | INIT_HLIST_HEAD(&inode_hashtable[loop]); |
1703 | } | |
1704 | ||
1705 | void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev) | |
1706 | { | |
1707 | inode->i_mode = mode; | |
1708 | if (S_ISCHR(mode)) { | |
1709 | inode->i_fop = &def_chr_fops; | |
1710 | inode->i_rdev = rdev; | |
1711 | } else if (S_ISBLK(mode)) { | |
1712 | inode->i_fop = &def_blk_fops; | |
1713 | inode->i_rdev = rdev; | |
1714 | } else if (S_ISFIFO(mode)) | |
1715 | inode->i_fop = &def_fifo_fops; | |
1716 | else if (S_ISSOCK(mode)) | |
1717 | inode->i_fop = &bad_sock_fops; | |
1718 | else | |
af0d9ae8 MK |
1719 | printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for" |
1720 | " inode %s:%lu\n", mode, inode->i_sb->s_id, | |
1721 | inode->i_ino); | |
1da177e4 LT |
1722 | } |
1723 | EXPORT_SYMBOL(init_special_inode); | |
a1bd120d DM |
1724 | |
1725 | /** | |
eaae668d | 1726 | * inode_init_owner - Init uid,gid,mode for new inode according to posix standards |
a1bd120d DM |
1727 | * @inode: New inode |
1728 | * @dir: Directory inode | |
1729 | * @mode: mode of the new inode | |
1730 | */ | |
1731 | void inode_init_owner(struct inode *inode, const struct inode *dir, | |
62bb1091 | 1732 | umode_t mode) |
a1bd120d DM |
1733 | { |
1734 | inode->i_uid = current_fsuid(); | |
1735 | if (dir && dir->i_mode & S_ISGID) { | |
1736 | inode->i_gid = dir->i_gid; | |
1737 | if (S_ISDIR(mode)) | |
1738 | mode |= S_ISGID; | |
1739 | } else | |
1740 | inode->i_gid = current_fsgid(); | |
1741 | inode->i_mode = mode; | |
1742 | } | |
1743 | EXPORT_SYMBOL(inode_init_owner); | |
e795b717 | 1744 | |
2e149670 SH |
1745 | /** |
1746 | * inode_owner_or_capable - check current task permissions to inode | |
1747 | * @inode: inode being checked | |
1748 | * | |
1749 | * Return true if current either has CAP_FOWNER to the inode, or | |
1750 | * owns the file. | |
e795b717 | 1751 | */ |
2e149670 | 1752 | bool inode_owner_or_capable(const struct inode *inode) |
e795b717 SH |
1753 | { |
1754 | struct user_namespace *ns = inode_userns(inode); | |
1755 | ||
1756 | if (current_user_ns() == ns && current_fsuid() == inode->i_uid) | |
1757 | return true; | |
1758 | if (ns_capable(ns, CAP_FOWNER)) | |
1759 | return true; | |
1760 | return false; | |
1761 | } | |
2e149670 | 1762 | EXPORT_SYMBOL(inode_owner_or_capable); |