]>
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) { | |
325 | printk_ratelimited(KERN_INFO | |
326 | "set_nlink() clearing i_nlink on %s inode %li\n", | |
327 | inode->i_sb->s_type->name, inode->i_ino); | |
328 | clear_nlink(inode); | |
329 | } else { | |
330 | /* Yes, some filesystems do change nlink from zero to one */ | |
331 | if (inode->i_nlink == 0) | |
332 | atomic_long_dec(&inode->i_sb->s_remove_count); | |
333 | ||
334 | inode->__i_nlink = nlink; | |
335 | } | |
336 | } | |
337 | EXPORT_SYMBOL(set_nlink); | |
338 | ||
339 | /** | |
340 | * inc_nlink - directly increment an inode's link count | |
341 | * @inode: inode | |
342 | * | |
343 | * This is a low-level filesystem helper to replace any | |
344 | * direct filesystem manipulation of i_nlink. Currently, | |
345 | * it is only here for parity with dec_nlink(). | |
346 | */ | |
347 | void inc_nlink(struct inode *inode) | |
348 | { | |
349 | if (WARN_ON(inode->i_nlink == 0)) | |
350 | atomic_long_dec(&inode->i_sb->s_remove_count); | |
351 | ||
352 | inode->__i_nlink++; | |
353 | } | |
354 | EXPORT_SYMBOL(inc_nlink); | |
355 | ||
2aa15890 MS |
356 | void address_space_init_once(struct address_space *mapping) |
357 | { | |
358 | memset(mapping, 0, sizeof(*mapping)); | |
359 | INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC); | |
360 | spin_lock_init(&mapping->tree_lock); | |
3d48ae45 | 361 | mutex_init(&mapping->i_mmap_mutex); |
2aa15890 MS |
362 | INIT_LIST_HEAD(&mapping->private_list); |
363 | spin_lock_init(&mapping->private_lock); | |
364 | INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap); | |
365 | INIT_LIST_HEAD(&mapping->i_mmap_nonlinear); | |
2aa15890 MS |
366 | } |
367 | EXPORT_SYMBOL(address_space_init_once); | |
368 | ||
1da177e4 LT |
369 | /* |
370 | * These are initializations that only need to be done | |
371 | * once, because the fields are idempotent across use | |
372 | * of the inode, so let the slab aware of that. | |
373 | */ | |
374 | void inode_init_once(struct inode *inode) | |
375 | { | |
376 | memset(inode, 0, sizeof(*inode)); | |
377 | INIT_HLIST_NODE(&inode->i_hash); | |
1da177e4 | 378 | INIT_LIST_HEAD(&inode->i_devices); |
7ccf19a8 NP |
379 | INIT_LIST_HEAD(&inode->i_wb_list); |
380 | INIT_LIST_HEAD(&inode->i_lru); | |
2aa15890 | 381 | address_space_init_once(&inode->i_data); |
1da177e4 | 382 | i_size_ordered_init(inode); |
3be25f49 | 383 | #ifdef CONFIG_FSNOTIFY |
e61ce867 | 384 | INIT_HLIST_HEAD(&inode->i_fsnotify_marks); |
3be25f49 | 385 | #endif |
1da177e4 | 386 | } |
1da177e4 LT |
387 | EXPORT_SYMBOL(inode_init_once); |
388 | ||
51cc5068 | 389 | static void init_once(void *foo) |
1da177e4 | 390 | { |
6b3304b5 | 391 | struct inode *inode = (struct inode *) foo; |
1da177e4 | 392 | |
a35afb83 | 393 | inode_init_once(inode); |
1da177e4 LT |
394 | } |
395 | ||
396 | /* | |
250df6ed | 397 | * inode->i_lock must be held |
1da177e4 | 398 | */ |
6b3304b5 | 399 | void __iget(struct inode *inode) |
1da177e4 | 400 | { |
9e38d86f NP |
401 | atomic_inc(&inode->i_count); |
402 | } | |
2e147f1e | 403 | |
7de9c6ee AV |
404 | /* |
405 | * get additional reference to inode; caller must already hold one. | |
406 | */ | |
407 | void ihold(struct inode *inode) | |
408 | { | |
409 | WARN_ON(atomic_inc_return(&inode->i_count) < 2); | |
410 | } | |
411 | EXPORT_SYMBOL(ihold); | |
412 | ||
9e38d86f NP |
413 | static void inode_lru_list_add(struct inode *inode) |
414 | { | |
09cc9fc7 | 415 | spin_lock(&inode->i_sb->s_inode_lru_lock); |
7ccf19a8 | 416 | if (list_empty(&inode->i_lru)) { |
98b745c6 DC |
417 | list_add(&inode->i_lru, &inode->i_sb->s_inode_lru); |
418 | inode->i_sb->s_nr_inodes_unused++; | |
fcb94f72 | 419 | this_cpu_inc(nr_unused); |
9e38d86f | 420 | } |
09cc9fc7 | 421 | spin_unlock(&inode->i_sb->s_inode_lru_lock); |
9e38d86f | 422 | } |
2e147f1e | 423 | |
9e38d86f NP |
424 | static void inode_lru_list_del(struct inode *inode) |
425 | { | |
09cc9fc7 | 426 | spin_lock(&inode->i_sb->s_inode_lru_lock); |
7ccf19a8 NP |
427 | if (!list_empty(&inode->i_lru)) { |
428 | list_del_init(&inode->i_lru); | |
98b745c6 | 429 | inode->i_sb->s_nr_inodes_unused--; |
fcb94f72 | 430 | this_cpu_dec(nr_unused); |
9e38d86f | 431 | } |
09cc9fc7 | 432 | spin_unlock(&inode->i_sb->s_inode_lru_lock); |
1da177e4 LT |
433 | } |
434 | ||
646ec461 CH |
435 | /** |
436 | * inode_sb_list_add - add inode to the superblock list of inodes | |
437 | * @inode: inode to add | |
438 | */ | |
439 | void inode_sb_list_add(struct inode *inode) | |
440 | { | |
55fa6091 DC |
441 | spin_lock(&inode_sb_list_lock); |
442 | list_add(&inode->i_sb_list, &inode->i_sb->s_inodes); | |
443 | spin_unlock(&inode_sb_list_lock); | |
646ec461 CH |
444 | } |
445 | EXPORT_SYMBOL_GPL(inode_sb_list_add); | |
446 | ||
55fa6091 | 447 | static inline void inode_sb_list_del(struct inode *inode) |
646ec461 | 448 | { |
a209dfc7 ED |
449 | if (!list_empty(&inode->i_sb_list)) { |
450 | spin_lock(&inode_sb_list_lock); | |
451 | list_del_init(&inode->i_sb_list); | |
452 | spin_unlock(&inode_sb_list_lock); | |
453 | } | |
646ec461 CH |
454 | } |
455 | ||
4c51acbc DC |
456 | static unsigned long hash(struct super_block *sb, unsigned long hashval) |
457 | { | |
458 | unsigned long tmp; | |
459 | ||
460 | tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) / | |
461 | L1_CACHE_BYTES; | |
4b4563dc CH |
462 | tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> i_hash_shift); |
463 | return tmp & i_hash_mask; | |
4c51acbc DC |
464 | } |
465 | ||
466 | /** | |
467 | * __insert_inode_hash - hash an inode | |
468 | * @inode: unhashed inode | |
469 | * @hashval: unsigned long value used to locate this object in the | |
470 | * inode_hashtable. | |
471 | * | |
472 | * Add an inode to the inode hash for this superblock. | |
473 | */ | |
474 | void __insert_inode_hash(struct inode *inode, unsigned long hashval) | |
475 | { | |
646ec461 CH |
476 | struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval); |
477 | ||
67a23c49 | 478 | spin_lock(&inode_hash_lock); |
250df6ed | 479 | spin_lock(&inode->i_lock); |
646ec461 | 480 | hlist_add_head(&inode->i_hash, b); |
250df6ed | 481 | spin_unlock(&inode->i_lock); |
67a23c49 | 482 | spin_unlock(&inode_hash_lock); |
4c51acbc DC |
483 | } |
484 | EXPORT_SYMBOL(__insert_inode_hash); | |
485 | ||
4c51acbc | 486 | /** |
f2ee7abf | 487 | * __remove_inode_hash - remove an inode from the hash |
4c51acbc DC |
488 | * @inode: inode to unhash |
489 | * | |
490 | * Remove an inode from the superblock. | |
491 | */ | |
f2ee7abf | 492 | void __remove_inode_hash(struct inode *inode) |
4c51acbc | 493 | { |
67a23c49 | 494 | spin_lock(&inode_hash_lock); |
250df6ed | 495 | spin_lock(&inode->i_lock); |
4c51acbc | 496 | hlist_del_init(&inode->i_hash); |
250df6ed | 497 | spin_unlock(&inode->i_lock); |
67a23c49 | 498 | spin_unlock(&inode_hash_lock); |
4c51acbc | 499 | } |
f2ee7abf | 500 | EXPORT_SYMBOL(__remove_inode_hash); |
4c51acbc | 501 | |
b0683aa6 AV |
502 | void end_writeback(struct inode *inode) |
503 | { | |
504 | might_sleep(); | |
08142579 JK |
505 | /* |
506 | * We have to cycle tree_lock here because reclaim can be still in the | |
507 | * process of removing the last page (in __delete_from_page_cache()) | |
508 | * and we must not free mapping under it. | |
509 | */ | |
510 | spin_lock_irq(&inode->i_data.tree_lock); | |
b0683aa6 | 511 | BUG_ON(inode->i_data.nrpages); |
08142579 | 512 | spin_unlock_irq(&inode->i_data.tree_lock); |
b0683aa6 AV |
513 | BUG_ON(!list_empty(&inode->i_data.private_list)); |
514 | BUG_ON(!(inode->i_state & I_FREEING)); | |
515 | BUG_ON(inode->i_state & I_CLEAR); | |
516 | inode_sync_wait(inode); | |
fa0d7e3d | 517 | /* don't need i_lock here, no concurrent mods to i_state */ |
b0683aa6 AV |
518 | inode->i_state = I_FREEING | I_CLEAR; |
519 | } | |
520 | EXPORT_SYMBOL(end_writeback); | |
521 | ||
b2b2af8e DC |
522 | /* |
523 | * Free the inode passed in, removing it from the lists it is still connected | |
524 | * to. We remove any pages still attached to the inode and wait for any IO that | |
525 | * is still in progress before finally destroying the inode. | |
526 | * | |
527 | * An inode must already be marked I_FREEING so that we avoid the inode being | |
528 | * moved back onto lists if we race with other code that manipulates the lists | |
529 | * (e.g. writeback_single_inode). The caller is responsible for setting this. | |
530 | * | |
531 | * An inode must already be removed from the LRU list before being evicted from | |
532 | * the cache. This should occur atomically with setting the I_FREEING state | |
533 | * flag, so no inodes here should ever be on the LRU when being evicted. | |
534 | */ | |
644da596 | 535 | static void evict(struct inode *inode) |
b4272d4c AV |
536 | { |
537 | const struct super_operations *op = inode->i_sb->s_op; | |
538 | ||
b2b2af8e DC |
539 | BUG_ON(!(inode->i_state & I_FREEING)); |
540 | BUG_ON(!list_empty(&inode->i_lru)); | |
541 | ||
b12362bd ED |
542 | if (!list_empty(&inode->i_wb_list)) |
543 | inode_wb_list_del(inode); | |
544 | ||
55fa6091 DC |
545 | inode_sb_list_del(inode); |
546 | ||
be7ce416 AV |
547 | if (op->evict_inode) { |
548 | op->evict_inode(inode); | |
b4272d4c AV |
549 | } else { |
550 | if (inode->i_data.nrpages) | |
551 | truncate_inode_pages(&inode->i_data, 0); | |
30140837 | 552 | end_writeback(inode); |
b4272d4c | 553 | } |
661074e9 AV |
554 | if (S_ISBLK(inode->i_mode) && inode->i_bdev) |
555 | bd_forget(inode); | |
556 | if (S_ISCHR(inode->i_mode) && inode->i_cdev) | |
557 | cd_forget(inode); | |
b2b2af8e DC |
558 | |
559 | remove_inode_hash(inode); | |
560 | ||
561 | spin_lock(&inode->i_lock); | |
562 | wake_up_bit(&inode->i_state, __I_NEW); | |
563 | BUG_ON(inode->i_state != (I_FREEING | I_CLEAR)); | |
564 | spin_unlock(&inode->i_lock); | |
565 | ||
566 | destroy_inode(inode); | |
b4272d4c AV |
567 | } |
568 | ||
1da177e4 LT |
569 | /* |
570 | * dispose_list - dispose of the contents of a local list | |
571 | * @head: the head of the list to free | |
572 | * | |
573 | * Dispose-list gets a local list with local inodes in it, so it doesn't | |
574 | * need to worry about list corruption and SMP locks. | |
575 | */ | |
576 | static void dispose_list(struct list_head *head) | |
577 | { | |
1da177e4 LT |
578 | while (!list_empty(head)) { |
579 | struct inode *inode; | |
580 | ||
7ccf19a8 NP |
581 | inode = list_first_entry(head, struct inode, i_lru); |
582 | list_del_init(&inode->i_lru); | |
1da177e4 | 583 | |
644da596 | 584 | evict(inode); |
1da177e4 | 585 | } |
1da177e4 LT |
586 | } |
587 | ||
63997e98 AV |
588 | /** |
589 | * evict_inodes - evict all evictable inodes for a superblock | |
590 | * @sb: superblock to operate on | |
591 | * | |
592 | * Make sure that no inodes with zero refcount are retained. This is | |
593 | * called by superblock shutdown after having MS_ACTIVE flag removed, | |
594 | * so any inode reaching zero refcount during or after that call will | |
595 | * be immediately evicted. | |
1da177e4 | 596 | */ |
63997e98 | 597 | void evict_inodes(struct super_block *sb) |
1da177e4 | 598 | { |
63997e98 AV |
599 | struct inode *inode, *next; |
600 | LIST_HEAD(dispose); | |
1da177e4 | 601 | |
55fa6091 | 602 | spin_lock(&inode_sb_list_lock); |
63997e98 AV |
603 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
604 | if (atomic_read(&inode->i_count)) | |
aabb8fdb | 605 | continue; |
250df6ed DC |
606 | |
607 | spin_lock(&inode->i_lock); | |
608 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
609 | spin_unlock(&inode->i_lock); | |
1da177e4 | 610 | continue; |
250df6ed | 611 | } |
63997e98 AV |
612 | |
613 | inode->i_state |= I_FREEING; | |
02afc410 | 614 | inode_lru_list_del(inode); |
250df6ed | 615 | spin_unlock(&inode->i_lock); |
02afc410 | 616 | list_add(&inode->i_lru, &dispose); |
1da177e4 | 617 | } |
55fa6091 | 618 | spin_unlock(&inode_sb_list_lock); |
63997e98 AV |
619 | |
620 | dispose_list(&dispose); | |
1da177e4 LT |
621 | } |
622 | ||
1da177e4 | 623 | /** |
a0318786 CH |
624 | * invalidate_inodes - attempt to free all inodes on a superblock |
625 | * @sb: superblock to operate on | |
93b270f7 | 626 | * @kill_dirty: flag to guide handling of dirty inodes |
1da177e4 | 627 | * |
a0318786 CH |
628 | * Attempts to free all inodes for a given superblock. If there were any |
629 | * busy inodes return a non-zero value, else zero. | |
93b270f7 N |
630 | * If @kill_dirty is set, discard dirty inodes too, otherwise treat |
631 | * them as busy. | |
1da177e4 | 632 | */ |
93b270f7 | 633 | int invalidate_inodes(struct super_block *sb, bool kill_dirty) |
1da177e4 | 634 | { |
cffbc8aa | 635 | int busy = 0; |
a0318786 CH |
636 | struct inode *inode, *next; |
637 | LIST_HEAD(dispose); | |
1da177e4 | 638 | |
55fa6091 | 639 | spin_lock(&inode_sb_list_lock); |
a0318786 | 640 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
250df6ed DC |
641 | spin_lock(&inode->i_lock); |
642 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
643 | spin_unlock(&inode->i_lock); | |
aabb8fdb | 644 | continue; |
250df6ed | 645 | } |
93b270f7 | 646 | if (inode->i_state & I_DIRTY && !kill_dirty) { |
250df6ed | 647 | spin_unlock(&inode->i_lock); |
93b270f7 N |
648 | busy = 1; |
649 | continue; | |
650 | } | |
99a38919 | 651 | if (atomic_read(&inode->i_count)) { |
250df6ed | 652 | spin_unlock(&inode->i_lock); |
99a38919 | 653 | busy = 1; |
1da177e4 LT |
654 | continue; |
655 | } | |
99a38919 | 656 | |
99a38919 | 657 | inode->i_state |= I_FREEING; |
02afc410 | 658 | inode_lru_list_del(inode); |
250df6ed | 659 | spin_unlock(&inode->i_lock); |
02afc410 | 660 | list_add(&inode->i_lru, &dispose); |
1da177e4 | 661 | } |
55fa6091 | 662 | spin_unlock(&inode_sb_list_lock); |
1da177e4 | 663 | |
a0318786 | 664 | dispose_list(&dispose); |
1da177e4 LT |
665 | |
666 | return busy; | |
667 | } | |
1da177e4 LT |
668 | |
669 | static int can_unuse(struct inode *inode) | |
670 | { | |
9e38d86f | 671 | if (inode->i_state & ~I_REFERENCED) |
1da177e4 LT |
672 | return 0; |
673 | if (inode_has_buffers(inode)) | |
674 | return 0; | |
675 | if (atomic_read(&inode->i_count)) | |
676 | return 0; | |
677 | if (inode->i_data.nrpages) | |
678 | return 0; | |
679 | return 1; | |
680 | } | |
681 | ||
682 | /* | |
b0d40c92 DC |
683 | * Walk the superblock inode LRU for freeable inodes and attempt to free them. |
684 | * This is called from the superblock shrinker function with a number of inodes | |
685 | * to trim from the LRU. Inodes to be freed are moved to a temporary list and | |
686 | * then are freed outside inode_lock by dispose_list(). | |
1da177e4 LT |
687 | * |
688 | * Any inodes which are pinned purely because of attached pagecache have their | |
9e38d86f NP |
689 | * pagecache removed. If the inode has metadata buffers attached to |
690 | * mapping->private_list then try to remove them. | |
1da177e4 | 691 | * |
9e38d86f NP |
692 | * If the inode has the I_REFERENCED flag set, then it means that it has been |
693 | * used recently - the flag is set in iput_final(). When we encounter such an | |
694 | * inode, clear the flag and move it to the back of the LRU so it gets another | |
695 | * pass through the LRU before it gets reclaimed. This is necessary because of | |
696 | * the fact we are doing lazy LRU updates to minimise lock contention so the | |
697 | * LRU does not have strict ordering. Hence we don't want to reclaim inodes | |
698 | * with this flag set because they are the inodes that are out of order. | |
1da177e4 | 699 | */ |
b0d40c92 | 700 | void prune_icache_sb(struct super_block *sb, int nr_to_scan) |
1da177e4 LT |
701 | { |
702 | LIST_HEAD(freeable); | |
1da177e4 LT |
703 | int nr_scanned; |
704 | unsigned long reap = 0; | |
705 | ||
09cc9fc7 | 706 | spin_lock(&sb->s_inode_lru_lock); |
b0d40c92 | 707 | for (nr_scanned = nr_to_scan; nr_scanned >= 0; nr_scanned--) { |
1da177e4 LT |
708 | struct inode *inode; |
709 | ||
98b745c6 | 710 | if (list_empty(&sb->s_inode_lru)) |
1da177e4 LT |
711 | break; |
712 | ||
98b745c6 | 713 | inode = list_entry(sb->s_inode_lru.prev, struct inode, i_lru); |
1da177e4 | 714 | |
02afc410 | 715 | /* |
09cc9fc7 | 716 | * we are inverting the sb->s_inode_lru_lock/inode->i_lock here, |
02afc410 DC |
717 | * so use a trylock. If we fail to get the lock, just move the |
718 | * inode to the back of the list so we don't spin on it. | |
719 | */ | |
720 | if (!spin_trylock(&inode->i_lock)) { | |
62a3ddef | 721 | list_move_tail(&inode->i_lru, &sb->s_inode_lru); |
02afc410 DC |
722 | continue; |
723 | } | |
724 | ||
9e38d86f NP |
725 | /* |
726 | * Referenced or dirty inodes are still in use. Give them | |
727 | * another pass through the LRU as we canot reclaim them now. | |
728 | */ | |
729 | if (atomic_read(&inode->i_count) || | |
730 | (inode->i_state & ~I_REFERENCED)) { | |
7ccf19a8 | 731 | list_del_init(&inode->i_lru); |
f283c86a | 732 | spin_unlock(&inode->i_lock); |
98b745c6 | 733 | sb->s_nr_inodes_unused--; |
fcb94f72 | 734 | this_cpu_dec(nr_unused); |
9e38d86f NP |
735 | continue; |
736 | } | |
737 | ||
738 | /* recently referenced inodes get one more pass */ | |
739 | if (inode->i_state & I_REFERENCED) { | |
9e38d86f | 740 | inode->i_state &= ~I_REFERENCED; |
98b745c6 | 741 | list_move(&inode->i_lru, &sb->s_inode_lru); |
f283c86a | 742 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
743 | continue; |
744 | } | |
745 | if (inode_has_buffers(inode) || inode->i_data.nrpages) { | |
746 | __iget(inode); | |
250df6ed | 747 | spin_unlock(&inode->i_lock); |
09cc9fc7 | 748 | spin_unlock(&sb->s_inode_lru_lock); |
1da177e4 | 749 | if (remove_inode_buffers(inode)) |
fc0ecff6 AM |
750 | reap += invalidate_mapping_pages(&inode->i_data, |
751 | 0, -1); | |
1da177e4 | 752 | iput(inode); |
09cc9fc7 | 753 | spin_lock(&sb->s_inode_lru_lock); |
1da177e4 | 754 | |
98b745c6 | 755 | if (inode != list_entry(sb->s_inode_lru.next, |
7ccf19a8 | 756 | struct inode, i_lru)) |
1da177e4 | 757 | continue; /* wrong inode or list_empty */ |
02afc410 DC |
758 | /* avoid lock inversions with trylock */ |
759 | if (!spin_trylock(&inode->i_lock)) | |
760 | continue; | |
250df6ed DC |
761 | if (!can_unuse(inode)) { |
762 | spin_unlock(&inode->i_lock); | |
1da177e4 | 763 | continue; |
250df6ed | 764 | } |
1da177e4 | 765 | } |
7ef0d737 | 766 | WARN_ON(inode->i_state & I_NEW); |
1da177e4 | 767 | inode->i_state |= I_FREEING; |
250df6ed | 768 | spin_unlock(&inode->i_lock); |
7ccf19a8 | 769 | |
7ccf19a8 | 770 | list_move(&inode->i_lru, &freeable); |
98b745c6 | 771 | sb->s_nr_inodes_unused--; |
fcb94f72 | 772 | this_cpu_dec(nr_unused); |
1da177e4 | 773 | } |
f8891e5e CL |
774 | if (current_is_kswapd()) |
775 | __count_vm_events(KSWAPD_INODESTEAL, reap); | |
776 | else | |
777 | __count_vm_events(PGINODESTEAL, reap); | |
09cc9fc7 | 778 | spin_unlock(&sb->s_inode_lru_lock); |
5f8aefd4 KK |
779 | if (current->reclaim_state) |
780 | current->reclaim_state->reclaimed_slab += reap; | |
1da177e4 LT |
781 | |
782 | dispose_list(&freeable); | |
1da177e4 LT |
783 | } |
784 | ||
1da177e4 LT |
785 | static void __wait_on_freeing_inode(struct inode *inode); |
786 | /* | |
787 | * Called with the inode lock held. | |
1da177e4 | 788 | */ |
6b3304b5 MK |
789 | static struct inode *find_inode(struct super_block *sb, |
790 | struct hlist_head *head, | |
791 | int (*test)(struct inode *, void *), | |
792 | void *data) | |
1da177e4 LT |
793 | { |
794 | struct hlist_node *node; | |
6b3304b5 | 795 | struct inode *inode = NULL; |
1da177e4 LT |
796 | |
797 | repeat: | |
c5c8be3c | 798 | hlist_for_each_entry(inode, node, head, i_hash) { |
67a23c49 DC |
799 | spin_lock(&inode->i_lock); |
800 | if (inode->i_sb != sb) { | |
801 | spin_unlock(&inode->i_lock); | |
1da177e4 | 802 | continue; |
67a23c49 DC |
803 | } |
804 | if (!test(inode, data)) { | |
805 | spin_unlock(&inode->i_lock); | |
1da177e4 | 806 | continue; |
67a23c49 | 807 | } |
a4ffdde6 | 808 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
809 | __wait_on_freeing_inode(inode); |
810 | goto repeat; | |
811 | } | |
f7899bd5 | 812 | __iget(inode); |
250df6ed | 813 | spin_unlock(&inode->i_lock); |
f7899bd5 | 814 | return inode; |
1da177e4 | 815 | } |
f7899bd5 | 816 | return NULL; |
1da177e4 LT |
817 | } |
818 | ||
819 | /* | |
820 | * find_inode_fast is the fast path version of find_inode, see the comment at | |
821 | * iget_locked for details. | |
822 | */ | |
6b3304b5 MK |
823 | static struct inode *find_inode_fast(struct super_block *sb, |
824 | struct hlist_head *head, unsigned long ino) | |
1da177e4 LT |
825 | { |
826 | struct hlist_node *node; | |
6b3304b5 | 827 | struct inode *inode = NULL; |
1da177e4 LT |
828 | |
829 | repeat: | |
c5c8be3c | 830 | hlist_for_each_entry(inode, node, head, i_hash) { |
67a23c49 DC |
831 | spin_lock(&inode->i_lock); |
832 | if (inode->i_ino != ino) { | |
833 | spin_unlock(&inode->i_lock); | |
1da177e4 | 834 | continue; |
67a23c49 DC |
835 | } |
836 | if (inode->i_sb != sb) { | |
837 | spin_unlock(&inode->i_lock); | |
1da177e4 | 838 | continue; |
67a23c49 | 839 | } |
a4ffdde6 | 840 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
841 | __wait_on_freeing_inode(inode); |
842 | goto repeat; | |
843 | } | |
f7899bd5 | 844 | __iget(inode); |
250df6ed | 845 | spin_unlock(&inode->i_lock); |
f7899bd5 | 846 | return inode; |
1da177e4 | 847 | } |
f7899bd5 | 848 | return NULL; |
8290c35f DC |
849 | } |
850 | ||
f991bd2e ED |
851 | /* |
852 | * Each cpu owns a range of LAST_INO_BATCH numbers. | |
853 | * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations, | |
854 | * to renew the exhausted range. | |
8290c35f | 855 | * |
f991bd2e ED |
856 | * This does not significantly increase overflow rate because every CPU can |
857 | * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is | |
858 | * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the | |
859 | * 2^32 range, and is a worst-case. Even a 50% wastage would only increase | |
860 | * overflow rate by 2x, which does not seem too significant. | |
861 | * | |
862 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
863 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
864 | * here to attempt to avoid that. | |
8290c35f | 865 | */ |
f991bd2e ED |
866 | #define LAST_INO_BATCH 1024 |
867 | static DEFINE_PER_CPU(unsigned int, last_ino); | |
868 | ||
85fe4025 | 869 | unsigned int get_next_ino(void) |
8290c35f | 870 | { |
f991bd2e ED |
871 | unsigned int *p = &get_cpu_var(last_ino); |
872 | unsigned int res = *p; | |
8290c35f | 873 | |
f991bd2e ED |
874 | #ifdef CONFIG_SMP |
875 | if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) { | |
876 | static atomic_t shared_last_ino; | |
877 | int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino); | |
878 | ||
879 | res = next - LAST_INO_BATCH; | |
880 | } | |
881 | #endif | |
882 | ||
883 | *p = ++res; | |
884 | put_cpu_var(last_ino); | |
885 | return res; | |
8290c35f | 886 | } |
85fe4025 | 887 | EXPORT_SYMBOL(get_next_ino); |
8290c35f | 888 | |
a209dfc7 ED |
889 | /** |
890 | * new_inode_pseudo - obtain an inode | |
891 | * @sb: superblock | |
892 | * | |
893 | * Allocates a new inode for given superblock. | |
894 | * Inode wont be chained in superblock s_inodes list | |
895 | * This means : | |
896 | * - fs can't be unmount | |
897 | * - quotas, fsnotify, writeback can't work | |
898 | */ | |
899 | struct inode *new_inode_pseudo(struct super_block *sb) | |
900 | { | |
901 | struct inode *inode = alloc_inode(sb); | |
902 | ||
903 | if (inode) { | |
904 | spin_lock(&inode->i_lock); | |
905 | inode->i_state = 0; | |
906 | spin_unlock(&inode->i_lock); | |
907 | INIT_LIST_HEAD(&inode->i_sb_list); | |
908 | } | |
909 | return inode; | |
910 | } | |
911 | ||
1da177e4 LT |
912 | /** |
913 | * new_inode - obtain an inode | |
914 | * @sb: superblock | |
915 | * | |
769848c0 | 916 | * Allocates a new inode for given superblock. The default gfp_mask |
3c1d4378 | 917 | * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE. |
769848c0 MG |
918 | * If HIGHMEM pages are unsuitable or it is known that pages allocated |
919 | * for the page cache are not reclaimable or migratable, | |
920 | * mapping_set_gfp_mask() must be called with suitable flags on the | |
921 | * newly created inode's mapping | |
922 | * | |
1da177e4 LT |
923 | */ |
924 | struct inode *new_inode(struct super_block *sb) | |
925 | { | |
6b3304b5 | 926 | struct inode *inode; |
1da177e4 | 927 | |
55fa6091 | 928 | spin_lock_prefetch(&inode_sb_list_lock); |
6b3304b5 | 929 | |
a209dfc7 ED |
930 | inode = new_inode_pseudo(sb); |
931 | if (inode) | |
55fa6091 | 932 | inode_sb_list_add(inode); |
1da177e4 LT |
933 | return inode; |
934 | } | |
1da177e4 LT |
935 | EXPORT_SYMBOL(new_inode); |
936 | ||
14358e6d | 937 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
e096d0c7 JB |
938 | void lockdep_annotate_inode_mutex_key(struct inode *inode) |
939 | { | |
a3314a0e | 940 | if (S_ISDIR(inode->i_mode)) { |
1e89a5e1 PZ |
941 | struct file_system_type *type = inode->i_sb->s_type; |
942 | ||
9a7aa12f JK |
943 | /* Set new key only if filesystem hasn't already changed it */ |
944 | if (!lockdep_match_class(&inode->i_mutex, | |
945 | &type->i_mutex_key)) { | |
946 | /* | |
947 | * ensure nobody is actually holding i_mutex | |
948 | */ | |
949 | mutex_destroy(&inode->i_mutex); | |
950 | mutex_init(&inode->i_mutex); | |
951 | lockdep_set_class(&inode->i_mutex, | |
952 | &type->i_mutex_dir_key); | |
953 | } | |
1e89a5e1 | 954 | } |
e096d0c7 JB |
955 | } |
956 | EXPORT_SYMBOL(lockdep_annotate_inode_mutex_key); | |
14358e6d | 957 | #endif |
e096d0c7 JB |
958 | |
959 | /** | |
960 | * unlock_new_inode - clear the I_NEW state and wake up any waiters | |
961 | * @inode: new inode to unlock | |
962 | * | |
963 | * Called when the inode is fully initialised to clear the new state of the | |
964 | * inode and wake up anyone waiting for the inode to finish initialisation. | |
965 | */ | |
966 | void unlock_new_inode(struct inode *inode) | |
967 | { | |
968 | lockdep_annotate_inode_mutex_key(inode); | |
250df6ed | 969 | spin_lock(&inode->i_lock); |
eaff8079 CH |
970 | WARN_ON(!(inode->i_state & I_NEW)); |
971 | inode->i_state &= ~I_NEW; | |
250df6ed DC |
972 | wake_up_bit(&inode->i_state, __I_NEW); |
973 | spin_unlock(&inode->i_lock); | |
1da177e4 | 974 | } |
1da177e4 LT |
975 | EXPORT_SYMBOL(unlock_new_inode); |
976 | ||
0b2d0724 CH |
977 | /** |
978 | * iget5_locked - obtain an inode from a mounted file system | |
979 | * @sb: super block of file system | |
980 | * @hashval: hash value (usually inode number) to get | |
981 | * @test: callback used for comparisons between inodes | |
982 | * @set: callback used to initialize a new struct inode | |
983 | * @data: opaque data pointer to pass to @test and @set | |
984 | * | |
985 | * Search for the inode specified by @hashval and @data in the inode cache, | |
986 | * and if present it is return it with an increased reference count. This is | |
987 | * a generalized version of iget_locked() for file systems where the inode | |
988 | * number is not sufficient for unique identification of an inode. | |
989 | * | |
990 | * If the inode is not in cache, allocate a new inode and return it locked, | |
991 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
992 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 993 | * |
0b2d0724 CH |
994 | * Note both @test and @set are called with the inode_hash_lock held, so can't |
995 | * sleep. | |
1da177e4 | 996 | */ |
0b2d0724 CH |
997 | struct inode *iget5_locked(struct super_block *sb, unsigned long hashval, |
998 | int (*test)(struct inode *, void *), | |
999 | int (*set)(struct inode *, void *), void *data) | |
1da177e4 | 1000 | { |
0b2d0724 | 1001 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); |
6b3304b5 | 1002 | struct inode *inode; |
1da177e4 | 1003 | |
0b2d0724 CH |
1004 | spin_lock(&inode_hash_lock); |
1005 | inode = find_inode(sb, head, test, data); | |
1006 | spin_unlock(&inode_hash_lock); | |
1007 | ||
1008 | if (inode) { | |
1009 | wait_on_inode(inode); | |
1010 | return inode; | |
1011 | } | |
1012 | ||
1da177e4 LT |
1013 | inode = alloc_inode(sb); |
1014 | if (inode) { | |
6b3304b5 | 1015 | struct inode *old; |
1da177e4 | 1016 | |
67a23c49 | 1017 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1018 | /* We released the lock, so.. */ |
1019 | old = find_inode(sb, head, test, data); | |
1020 | if (!old) { | |
1021 | if (set(inode, data)) | |
1022 | goto set_failed; | |
1023 | ||
250df6ed DC |
1024 | spin_lock(&inode->i_lock); |
1025 | inode->i_state = I_NEW; | |
646ec461 | 1026 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1027 | spin_unlock(&inode->i_lock); |
55fa6091 | 1028 | inode_sb_list_add(inode); |
67a23c49 | 1029 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1030 | |
1031 | /* Return the locked inode with I_NEW set, the | |
1032 | * caller is responsible for filling in the contents | |
1033 | */ | |
1034 | return inode; | |
1035 | } | |
1036 | ||
1037 | /* | |
1038 | * Uhhuh, somebody else created the same inode under | |
1039 | * us. Use the old inode instead of the one we just | |
1040 | * allocated. | |
1041 | */ | |
67a23c49 | 1042 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1043 | destroy_inode(inode); |
1044 | inode = old; | |
1045 | wait_on_inode(inode); | |
1046 | } | |
1047 | return inode; | |
1048 | ||
1049 | set_failed: | |
67a23c49 | 1050 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1051 | destroy_inode(inode); |
1052 | return NULL; | |
1053 | } | |
0b2d0724 | 1054 | EXPORT_SYMBOL(iget5_locked); |
1da177e4 | 1055 | |
0b2d0724 CH |
1056 | /** |
1057 | * iget_locked - obtain an inode from a mounted file system | |
1058 | * @sb: super block of file system | |
1059 | * @ino: inode number to get | |
1060 | * | |
1061 | * Search for the inode specified by @ino in the inode cache and if present | |
1062 | * return it with an increased reference count. This is for file systems | |
1063 | * where the inode number is sufficient for unique identification of an inode. | |
1064 | * | |
1065 | * If the inode is not in cache, allocate a new inode and return it locked, | |
1066 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
1067 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 1068 | */ |
0b2d0724 | 1069 | struct inode *iget_locked(struct super_block *sb, unsigned long ino) |
1da177e4 | 1070 | { |
0b2d0724 | 1071 | struct hlist_head *head = inode_hashtable + hash(sb, ino); |
6b3304b5 | 1072 | struct inode *inode; |
1da177e4 | 1073 | |
0b2d0724 CH |
1074 | spin_lock(&inode_hash_lock); |
1075 | inode = find_inode_fast(sb, head, ino); | |
1076 | spin_unlock(&inode_hash_lock); | |
1077 | if (inode) { | |
1078 | wait_on_inode(inode); | |
1079 | return inode; | |
1080 | } | |
1081 | ||
1da177e4 LT |
1082 | inode = alloc_inode(sb); |
1083 | if (inode) { | |
6b3304b5 | 1084 | struct inode *old; |
1da177e4 | 1085 | |
67a23c49 | 1086 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1087 | /* We released the lock, so.. */ |
1088 | old = find_inode_fast(sb, head, ino); | |
1089 | if (!old) { | |
1090 | inode->i_ino = ino; | |
250df6ed DC |
1091 | spin_lock(&inode->i_lock); |
1092 | inode->i_state = I_NEW; | |
646ec461 | 1093 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1094 | spin_unlock(&inode->i_lock); |
55fa6091 | 1095 | inode_sb_list_add(inode); |
67a23c49 | 1096 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1097 | |
1098 | /* Return the locked inode with I_NEW set, the | |
1099 | * caller is responsible for filling in the contents | |
1100 | */ | |
1101 | return inode; | |
1102 | } | |
1103 | ||
1104 | /* | |
1105 | * Uhhuh, somebody else created the same inode under | |
1106 | * us. Use the old inode instead of the one we just | |
1107 | * allocated. | |
1108 | */ | |
67a23c49 | 1109 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1110 | destroy_inode(inode); |
1111 | inode = old; | |
1112 | wait_on_inode(inode); | |
1113 | } | |
1114 | return inode; | |
1115 | } | |
0b2d0724 | 1116 | EXPORT_SYMBOL(iget_locked); |
1da177e4 | 1117 | |
ad5e195a CH |
1118 | /* |
1119 | * search the inode cache for a matching inode number. | |
1120 | * If we find one, then the inode number we are trying to | |
1121 | * allocate is not unique and so we should not use it. | |
1122 | * | |
1123 | * Returns 1 if the inode number is unique, 0 if it is not. | |
1124 | */ | |
1125 | static int test_inode_iunique(struct super_block *sb, unsigned long ino) | |
1126 | { | |
1127 | struct hlist_head *b = inode_hashtable + hash(sb, ino); | |
1128 | struct hlist_node *node; | |
1129 | struct inode *inode; | |
1130 | ||
67a23c49 | 1131 | spin_lock(&inode_hash_lock); |
ad5e195a | 1132 | hlist_for_each_entry(inode, node, b, i_hash) { |
67a23c49 DC |
1133 | if (inode->i_ino == ino && inode->i_sb == sb) { |
1134 | spin_unlock(&inode_hash_lock); | |
ad5e195a | 1135 | return 0; |
67a23c49 | 1136 | } |
ad5e195a | 1137 | } |
67a23c49 | 1138 | spin_unlock(&inode_hash_lock); |
ad5e195a CH |
1139 | |
1140 | return 1; | |
1141 | } | |
1142 | ||
1da177e4 LT |
1143 | /** |
1144 | * iunique - get a unique inode number | |
1145 | * @sb: superblock | |
1146 | * @max_reserved: highest reserved inode number | |
1147 | * | |
1148 | * Obtain an inode number that is unique on the system for a given | |
1149 | * superblock. This is used by file systems that have no natural | |
1150 | * permanent inode numbering system. An inode number is returned that | |
1151 | * is higher than the reserved limit but unique. | |
1152 | * | |
1153 | * BUGS: | |
1154 | * With a large number of inodes live on the file system this function | |
1155 | * currently becomes quite slow. | |
1156 | */ | |
1157 | ino_t iunique(struct super_block *sb, ino_t max_reserved) | |
1158 | { | |
866b04fc JL |
1159 | /* |
1160 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
1161 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
1162 | * here to attempt to avoid that. | |
1163 | */ | |
ad5e195a | 1164 | static DEFINE_SPINLOCK(iunique_lock); |
866b04fc | 1165 | static unsigned int counter; |
1da177e4 | 1166 | ino_t res; |
3361c7be | 1167 | |
ad5e195a | 1168 | spin_lock(&iunique_lock); |
3361c7be JL |
1169 | do { |
1170 | if (counter <= max_reserved) | |
1171 | counter = max_reserved + 1; | |
1da177e4 | 1172 | res = counter++; |
ad5e195a CH |
1173 | } while (!test_inode_iunique(sb, res)); |
1174 | spin_unlock(&iunique_lock); | |
1da177e4 | 1175 | |
3361c7be JL |
1176 | return res; |
1177 | } | |
1da177e4 LT |
1178 | EXPORT_SYMBOL(iunique); |
1179 | ||
1180 | struct inode *igrab(struct inode *inode) | |
1181 | { | |
250df6ed DC |
1182 | spin_lock(&inode->i_lock); |
1183 | if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) { | |
1da177e4 | 1184 | __iget(inode); |
250df6ed DC |
1185 | spin_unlock(&inode->i_lock); |
1186 | } else { | |
1187 | spin_unlock(&inode->i_lock); | |
1da177e4 LT |
1188 | /* |
1189 | * Handle the case where s_op->clear_inode is not been | |
1190 | * called yet, and somebody is calling igrab | |
1191 | * while the inode is getting freed. | |
1192 | */ | |
1193 | inode = NULL; | |
250df6ed | 1194 | } |
1da177e4 LT |
1195 | return inode; |
1196 | } | |
1da177e4 LT |
1197 | EXPORT_SYMBOL(igrab); |
1198 | ||
1199 | /** | |
0b2d0724 | 1200 | * ilookup5_nowait - search for an inode in the inode cache |
1da177e4 | 1201 | * @sb: super block of file system to search |
0b2d0724 | 1202 | * @hashval: hash value (usually inode number) to search for |
1da177e4 LT |
1203 | * @test: callback used for comparisons between inodes |
1204 | * @data: opaque data pointer to pass to @test | |
1da177e4 | 1205 | * |
0b2d0724 | 1206 | * Search for the inode specified by @hashval and @data in the inode cache. |
1da177e4 LT |
1207 | * If the inode is in the cache, the inode is returned with an incremented |
1208 | * reference count. | |
1209 | * | |
0b2d0724 CH |
1210 | * Note: I_NEW is not waited upon so you have to be very careful what you do |
1211 | * with the returned inode. You probably should be using ilookup5() instead. | |
1da177e4 | 1212 | * |
b6d0ad68 | 1213 | * Note2: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 | 1214 | */ |
0b2d0724 CH |
1215 | struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval, |
1216 | int (*test)(struct inode *, void *), void *data) | |
1da177e4 | 1217 | { |
0b2d0724 | 1218 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); |
1da177e4 LT |
1219 | struct inode *inode; |
1220 | ||
67a23c49 | 1221 | spin_lock(&inode_hash_lock); |
1da177e4 | 1222 | inode = find_inode(sb, head, test, data); |
67a23c49 | 1223 | spin_unlock(&inode_hash_lock); |
88bd5121 | 1224 | |
0b2d0724 | 1225 | return inode; |
88bd5121 | 1226 | } |
88bd5121 AA |
1227 | EXPORT_SYMBOL(ilookup5_nowait); |
1228 | ||
1229 | /** | |
1230 | * ilookup5 - search for an inode in the inode cache | |
1231 | * @sb: super block of file system to search | |
1232 | * @hashval: hash value (usually inode number) to search for | |
1233 | * @test: callback used for comparisons between inodes | |
1234 | * @data: opaque data pointer to pass to @test | |
1235 | * | |
0b2d0724 CH |
1236 | * Search for the inode specified by @hashval and @data in the inode cache, |
1237 | * and if the inode is in the cache, return the inode with an incremented | |
1238 | * reference count. Waits on I_NEW before returning the inode. | |
88bd5121 | 1239 | * returned with an incremented reference count. |
1da177e4 | 1240 | * |
0b2d0724 CH |
1241 | * This is a generalized version of ilookup() for file systems where the |
1242 | * inode number is not sufficient for unique identification of an inode. | |
1da177e4 | 1243 | * |
0b2d0724 | 1244 | * Note: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 LT |
1245 | */ |
1246 | struct inode *ilookup5(struct super_block *sb, unsigned long hashval, | |
1247 | int (*test)(struct inode *, void *), void *data) | |
1248 | { | |
0b2d0724 | 1249 | struct inode *inode = ilookup5_nowait(sb, hashval, test, data); |
1da177e4 | 1250 | |
0b2d0724 CH |
1251 | if (inode) |
1252 | wait_on_inode(inode); | |
1253 | return inode; | |
1da177e4 | 1254 | } |
1da177e4 LT |
1255 | EXPORT_SYMBOL(ilookup5); |
1256 | ||
1257 | /** | |
1258 | * ilookup - search for an inode in the inode cache | |
1259 | * @sb: super block of file system to search | |
1260 | * @ino: inode number to search for | |
1261 | * | |
0b2d0724 CH |
1262 | * Search for the inode @ino in the inode cache, and if the inode is in the |
1263 | * cache, the inode is returned with an incremented reference count. | |
1da177e4 LT |
1264 | */ |
1265 | struct inode *ilookup(struct super_block *sb, unsigned long ino) | |
1266 | { | |
1267 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1da177e4 LT |
1268 | struct inode *inode; |
1269 | ||
0b2d0724 CH |
1270 | spin_lock(&inode_hash_lock); |
1271 | inode = find_inode_fast(sb, head, ino); | |
1272 | spin_unlock(&inode_hash_lock); | |
1da177e4 | 1273 | |
1da177e4 | 1274 | if (inode) |
0b2d0724 CH |
1275 | wait_on_inode(inode); |
1276 | return inode; | |
1da177e4 | 1277 | } |
0b2d0724 | 1278 | EXPORT_SYMBOL(ilookup); |
1da177e4 | 1279 | |
261bca86 AV |
1280 | int insert_inode_locked(struct inode *inode) |
1281 | { | |
1282 | struct super_block *sb = inode->i_sb; | |
1283 | ino_t ino = inode->i_ino; | |
1284 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
261bca86 | 1285 | |
261bca86 | 1286 | while (1) { |
72a43d63 AV |
1287 | struct hlist_node *node; |
1288 | struct inode *old = NULL; | |
67a23c49 | 1289 | spin_lock(&inode_hash_lock); |
72a43d63 AV |
1290 | hlist_for_each_entry(old, node, head, i_hash) { |
1291 | if (old->i_ino != ino) | |
1292 | continue; | |
1293 | if (old->i_sb != sb) | |
1294 | continue; | |
250df6ed DC |
1295 | spin_lock(&old->i_lock); |
1296 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1297 | spin_unlock(&old->i_lock); | |
72a43d63 | 1298 | continue; |
250df6ed | 1299 | } |
72a43d63 AV |
1300 | break; |
1301 | } | |
1302 | if (likely(!node)) { | |
250df6ed DC |
1303 | spin_lock(&inode->i_lock); |
1304 | inode->i_state |= I_NEW; | |
261bca86 | 1305 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1306 | spin_unlock(&inode->i_lock); |
67a23c49 | 1307 | spin_unlock(&inode_hash_lock); |
261bca86 AV |
1308 | return 0; |
1309 | } | |
1310 | __iget(old); | |
250df6ed | 1311 | spin_unlock(&old->i_lock); |
67a23c49 | 1312 | spin_unlock(&inode_hash_lock); |
261bca86 | 1313 | wait_on_inode(old); |
1d3382cb | 1314 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1315 | iput(old); |
1316 | return -EBUSY; | |
1317 | } | |
1318 | iput(old); | |
1319 | } | |
1320 | } | |
261bca86 AV |
1321 | EXPORT_SYMBOL(insert_inode_locked); |
1322 | ||
1323 | int insert_inode_locked4(struct inode *inode, unsigned long hashval, | |
1324 | int (*test)(struct inode *, void *), void *data) | |
1325 | { | |
1326 | struct super_block *sb = inode->i_sb; | |
1327 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
261bca86 | 1328 | |
261bca86 | 1329 | while (1) { |
72a43d63 AV |
1330 | struct hlist_node *node; |
1331 | struct inode *old = NULL; | |
1332 | ||
67a23c49 | 1333 | spin_lock(&inode_hash_lock); |
72a43d63 AV |
1334 | hlist_for_each_entry(old, node, head, i_hash) { |
1335 | if (old->i_sb != sb) | |
1336 | continue; | |
1337 | if (!test(old, data)) | |
1338 | continue; | |
250df6ed DC |
1339 | spin_lock(&old->i_lock); |
1340 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1341 | spin_unlock(&old->i_lock); | |
72a43d63 | 1342 | continue; |
250df6ed | 1343 | } |
72a43d63 AV |
1344 | break; |
1345 | } | |
1346 | if (likely(!node)) { | |
250df6ed DC |
1347 | spin_lock(&inode->i_lock); |
1348 | inode->i_state |= I_NEW; | |
261bca86 | 1349 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1350 | spin_unlock(&inode->i_lock); |
67a23c49 | 1351 | spin_unlock(&inode_hash_lock); |
261bca86 AV |
1352 | return 0; |
1353 | } | |
1354 | __iget(old); | |
250df6ed | 1355 | spin_unlock(&old->i_lock); |
67a23c49 | 1356 | spin_unlock(&inode_hash_lock); |
261bca86 | 1357 | wait_on_inode(old); |
1d3382cb | 1358 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1359 | iput(old); |
1360 | return -EBUSY; | |
1361 | } | |
1362 | iput(old); | |
1363 | } | |
1364 | } | |
261bca86 AV |
1365 | EXPORT_SYMBOL(insert_inode_locked4); |
1366 | ||
1da177e4 | 1367 | |
45321ac5 AV |
1368 | int generic_delete_inode(struct inode *inode) |
1369 | { | |
1370 | return 1; | |
1371 | } | |
1372 | EXPORT_SYMBOL(generic_delete_inode); | |
1373 | ||
1da177e4 | 1374 | /* |
45321ac5 AV |
1375 | * Normal UNIX filesystem behaviour: delete the |
1376 | * inode when the usage count drops to zero, and | |
1377 | * i_nlink is zero. | |
1da177e4 | 1378 | */ |
45321ac5 | 1379 | int generic_drop_inode(struct inode *inode) |
1da177e4 | 1380 | { |
1d3382cb | 1381 | return !inode->i_nlink || inode_unhashed(inode); |
1da177e4 | 1382 | } |
45321ac5 | 1383 | EXPORT_SYMBOL_GPL(generic_drop_inode); |
1da177e4 | 1384 | |
45321ac5 AV |
1385 | /* |
1386 | * Called when we're dropping the last reference | |
1387 | * to an inode. | |
22fe4042 | 1388 | * |
45321ac5 AV |
1389 | * Call the FS "drop_inode()" function, defaulting to |
1390 | * the legacy UNIX filesystem behaviour. If it tells | |
1391 | * us to evict inode, do so. Otherwise, retain inode | |
1392 | * in cache if fs is alive, sync and evict if fs is | |
1393 | * shutting down. | |
22fe4042 | 1394 | */ |
45321ac5 | 1395 | static void iput_final(struct inode *inode) |
1da177e4 LT |
1396 | { |
1397 | struct super_block *sb = inode->i_sb; | |
45321ac5 AV |
1398 | const struct super_operations *op = inode->i_sb->s_op; |
1399 | int drop; | |
1400 | ||
250df6ed DC |
1401 | WARN_ON(inode->i_state & I_NEW); |
1402 | ||
e7f59097 | 1403 | if (op->drop_inode) |
45321ac5 AV |
1404 | drop = op->drop_inode(inode); |
1405 | else | |
1406 | drop = generic_drop_inode(inode); | |
1da177e4 | 1407 | |
b2b2af8e DC |
1408 | if (!drop && (sb->s_flags & MS_ACTIVE)) { |
1409 | inode->i_state |= I_REFERENCED; | |
1410 | if (!(inode->i_state & (I_DIRTY|I_SYNC))) | |
1411 | inode_lru_list_add(inode); | |
1412 | spin_unlock(&inode->i_lock); | |
b2b2af8e DC |
1413 | return; |
1414 | } | |
1415 | ||
45321ac5 | 1416 | if (!drop) { |
991114c6 | 1417 | inode->i_state |= I_WILL_FREE; |
250df6ed | 1418 | spin_unlock(&inode->i_lock); |
1da177e4 | 1419 | write_inode_now(inode, 1); |
250df6ed | 1420 | spin_lock(&inode->i_lock); |
7ef0d737 | 1421 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1422 | inode->i_state &= ~I_WILL_FREE; |
1da177e4 | 1423 | } |
7ccf19a8 | 1424 | |
991114c6 | 1425 | inode->i_state |= I_FREEING; |
c4ae0c65 ED |
1426 | if (!list_empty(&inode->i_lru)) |
1427 | inode_lru_list_del(inode); | |
b2b2af8e | 1428 | spin_unlock(&inode->i_lock); |
b2b2af8e | 1429 | |
644da596 | 1430 | evict(inode); |
1da177e4 LT |
1431 | } |
1432 | ||
1da177e4 | 1433 | /** |
6b3304b5 | 1434 | * iput - put an inode |
1da177e4 LT |
1435 | * @inode: inode to put |
1436 | * | |
1437 | * Puts an inode, dropping its usage count. If the inode use count hits | |
1438 | * zero, the inode is then freed and may also be destroyed. | |
1439 | * | |
1440 | * Consequently, iput() can sleep. | |
1441 | */ | |
1442 | void iput(struct inode *inode) | |
1443 | { | |
1444 | if (inode) { | |
a4ffdde6 | 1445 | BUG_ON(inode->i_state & I_CLEAR); |
1da177e4 | 1446 | |
f283c86a | 1447 | if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock)) |
1da177e4 LT |
1448 | iput_final(inode); |
1449 | } | |
1450 | } | |
1da177e4 LT |
1451 | EXPORT_SYMBOL(iput); |
1452 | ||
1453 | /** | |
1454 | * bmap - find a block number in a file | |
1455 | * @inode: inode of file | |
1456 | * @block: block to find | |
1457 | * | |
1458 | * Returns the block number on the device holding the inode that | |
1459 | * is the disk block number for the block of the file requested. | |
1460 | * That is, asked for block 4 of inode 1 the function will return the | |
6b3304b5 | 1461 | * disk block relative to the disk start that holds that block of the |
1da177e4 LT |
1462 | * file. |
1463 | */ | |
6b3304b5 | 1464 | sector_t bmap(struct inode *inode, sector_t block) |
1da177e4 LT |
1465 | { |
1466 | sector_t res = 0; | |
1467 | if (inode->i_mapping->a_ops->bmap) | |
1468 | res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block); | |
1469 | return res; | |
1470 | } | |
1da177e4 LT |
1471 | EXPORT_SYMBOL(bmap); |
1472 | ||
11ff6f05 MG |
1473 | /* |
1474 | * With relative atime, only update atime if the previous atime is | |
1475 | * earlier than either the ctime or mtime or if at least a day has | |
1476 | * passed since the last atime update. | |
1477 | */ | |
1478 | static int relatime_need_update(struct vfsmount *mnt, struct inode *inode, | |
1479 | struct timespec now) | |
1480 | { | |
1481 | ||
1482 | if (!(mnt->mnt_flags & MNT_RELATIME)) | |
1483 | return 1; | |
1484 | /* | |
1485 | * Is mtime younger than atime? If yes, update atime: | |
1486 | */ | |
1487 | if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0) | |
1488 | return 1; | |
1489 | /* | |
1490 | * Is ctime younger than atime? If yes, update atime: | |
1491 | */ | |
1492 | if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0) | |
1493 | return 1; | |
1494 | ||
1495 | /* | |
1496 | * Is the previous atime value older than a day? If yes, | |
1497 | * update atime: | |
1498 | */ | |
1499 | if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60) | |
1500 | return 1; | |
1501 | /* | |
1502 | * Good, we can skip the atime update: | |
1503 | */ | |
1504 | return 0; | |
1505 | } | |
1506 | ||
1da177e4 | 1507 | /** |
869243a0 CH |
1508 | * touch_atime - update the access time |
1509 | * @mnt: mount the inode is accessed on | |
7045f37b | 1510 | * @dentry: dentry accessed |
1da177e4 LT |
1511 | * |
1512 | * Update the accessed time on an inode and mark it for writeback. | |
1513 | * This function automatically handles read only file systems and media, | |
1514 | * as well as the "noatime" flag and inode specific "noatime" markers. | |
1515 | */ | |
869243a0 | 1516 | void touch_atime(struct vfsmount *mnt, struct dentry *dentry) |
1da177e4 | 1517 | { |
869243a0 | 1518 | struct inode *inode = dentry->d_inode; |
1da177e4 LT |
1519 | struct timespec now; |
1520 | ||
cdb70f3f | 1521 | if (inode->i_flags & S_NOATIME) |
b12536c2 | 1522 | return; |
37756ced | 1523 | if (IS_NOATIME(inode)) |
b12536c2 | 1524 | return; |
b2276138 | 1525 | if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1526 | return; |
47ae32d6 | 1527 | |
cdb70f3f | 1528 | if (mnt->mnt_flags & MNT_NOATIME) |
b12536c2 | 1529 | return; |
cdb70f3f | 1530 | if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) |
b12536c2 | 1531 | return; |
1da177e4 LT |
1532 | |
1533 | now = current_fs_time(inode->i_sb); | |
11ff6f05 MG |
1534 | |
1535 | if (!relatime_need_update(mnt, inode, now)) | |
b12536c2 | 1536 | return; |
11ff6f05 | 1537 | |
47ae32d6 | 1538 | if (timespec_equal(&inode->i_atime, &now)) |
b12536c2 AK |
1539 | return; |
1540 | ||
1541 | if (mnt_want_write(mnt)) | |
1542 | return; | |
47ae32d6 VH |
1543 | |
1544 | inode->i_atime = now; | |
1545 | mark_inode_dirty_sync(inode); | |
cdb70f3f | 1546 | mnt_drop_write(mnt); |
1da177e4 | 1547 | } |
869243a0 | 1548 | EXPORT_SYMBOL(touch_atime); |
1da177e4 LT |
1549 | |
1550 | /** | |
870f4817 CH |
1551 | * file_update_time - update mtime and ctime time |
1552 | * @file: file accessed | |
1da177e4 | 1553 | * |
870f4817 CH |
1554 | * Update the mtime and ctime members of an inode and mark the inode |
1555 | * for writeback. Note that this function is meant exclusively for | |
1556 | * usage in the file write path of filesystems, and filesystems may | |
1557 | * choose to explicitly ignore update via this function with the | |
2eadfc0e | 1558 | * S_NOCMTIME inode flag, e.g. for network filesystem where these |
870f4817 | 1559 | * timestamps are handled by the server. |
1da177e4 LT |
1560 | */ |
1561 | ||
870f4817 | 1562 | void file_update_time(struct file *file) |
1da177e4 | 1563 | { |
0f7fc9e4 | 1564 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 | 1565 | struct timespec now; |
ce06e0b2 | 1566 | enum { S_MTIME = 1, S_CTIME = 2, S_VERSION = 4 } sync_it = 0; |
1da177e4 | 1567 | |
ce06e0b2 | 1568 | /* First try to exhaust all avenues to not sync */ |
1da177e4 LT |
1569 | if (IS_NOCMTIME(inode)) |
1570 | return; | |
20ddee2c | 1571 | |
1da177e4 | 1572 | now = current_fs_time(inode->i_sb); |
ce06e0b2 AK |
1573 | if (!timespec_equal(&inode->i_mtime, &now)) |
1574 | sync_it = S_MTIME; | |
1da177e4 | 1575 | |
ce06e0b2 AK |
1576 | if (!timespec_equal(&inode->i_ctime, &now)) |
1577 | sync_it |= S_CTIME; | |
870f4817 | 1578 | |
ce06e0b2 AK |
1579 | if (IS_I_VERSION(inode)) |
1580 | sync_it |= S_VERSION; | |
7a224228 | 1581 | |
ce06e0b2 AK |
1582 | if (!sync_it) |
1583 | return; | |
1584 | ||
1585 | /* Finally allowed to write? Takes lock. */ | |
1586 | if (mnt_want_write_file(file)) | |
1587 | return; | |
1588 | ||
1589 | /* Only change inode inside the lock region */ | |
1590 | if (sync_it & S_VERSION) | |
1591 | inode_inc_iversion(inode); | |
1592 | if (sync_it & S_CTIME) | |
1593 | inode->i_ctime = now; | |
1594 | if (sync_it & S_MTIME) | |
1595 | inode->i_mtime = now; | |
1596 | mark_inode_dirty_sync(inode); | |
2a79f17e | 1597 | mnt_drop_write_file(file); |
1da177e4 | 1598 | } |
870f4817 | 1599 | EXPORT_SYMBOL(file_update_time); |
1da177e4 LT |
1600 | |
1601 | int inode_needs_sync(struct inode *inode) | |
1602 | { | |
1603 | if (IS_SYNC(inode)) | |
1604 | return 1; | |
1605 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) | |
1606 | return 1; | |
1607 | return 0; | |
1608 | } | |
1da177e4 LT |
1609 | EXPORT_SYMBOL(inode_needs_sync); |
1610 | ||
1da177e4 LT |
1611 | int inode_wait(void *word) |
1612 | { | |
1613 | schedule(); | |
1614 | return 0; | |
1615 | } | |
d44dab8d | 1616 | EXPORT_SYMBOL(inode_wait); |
1da177e4 LT |
1617 | |
1618 | /* | |
168a9fd6 MS |
1619 | * If we try to find an inode in the inode hash while it is being |
1620 | * deleted, we have to wait until the filesystem completes its | |
1621 | * deletion before reporting that it isn't found. This function waits | |
1622 | * until the deletion _might_ have completed. Callers are responsible | |
1623 | * to recheck inode state. | |
1624 | * | |
eaff8079 | 1625 | * It doesn't matter if I_NEW is not set initially, a call to |
250df6ed DC |
1626 | * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list |
1627 | * will DTRT. | |
1da177e4 LT |
1628 | */ |
1629 | static void __wait_on_freeing_inode(struct inode *inode) | |
1630 | { | |
1631 | wait_queue_head_t *wq; | |
eaff8079 CH |
1632 | DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW); |
1633 | wq = bit_waitqueue(&inode->i_state, __I_NEW); | |
1da177e4 | 1634 | prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE); |
250df6ed | 1635 | spin_unlock(&inode->i_lock); |
67a23c49 | 1636 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1637 | schedule(); |
1638 | finish_wait(wq, &wait.wait); | |
67a23c49 | 1639 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1640 | } |
1641 | ||
1da177e4 LT |
1642 | static __initdata unsigned long ihash_entries; |
1643 | static int __init set_ihash_entries(char *str) | |
1644 | { | |
1645 | if (!str) | |
1646 | return 0; | |
1647 | ihash_entries = simple_strtoul(str, &str, 0); | |
1648 | return 1; | |
1649 | } | |
1650 | __setup("ihash_entries=", set_ihash_entries); | |
1651 | ||
1652 | /* | |
1653 | * Initialize the waitqueues and inode hash table. | |
1654 | */ | |
1655 | void __init inode_init_early(void) | |
1656 | { | |
1657 | int loop; | |
1658 | ||
1659 | /* If hashes are distributed across NUMA nodes, defer | |
1660 | * hash allocation until vmalloc space is available. | |
1661 | */ | |
1662 | if (hashdist) | |
1663 | return; | |
1664 | ||
1665 | inode_hashtable = | |
1666 | alloc_large_system_hash("Inode-cache", | |
1667 | sizeof(struct hlist_head), | |
1668 | ihash_entries, | |
1669 | 14, | |
1670 | HASH_EARLY, | |
1671 | &i_hash_shift, | |
1672 | &i_hash_mask, | |
1673 | 0); | |
1674 | ||
1675 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1676 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1677 | } | |
1678 | ||
74bf17cf | 1679 | void __init inode_init(void) |
1da177e4 LT |
1680 | { |
1681 | int loop; | |
1682 | ||
1683 | /* inode slab cache */ | |
b0196009 PJ |
1684 | inode_cachep = kmem_cache_create("inode_cache", |
1685 | sizeof(struct inode), | |
1686 | 0, | |
1687 | (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC| | |
1688 | SLAB_MEM_SPREAD), | |
20c2df83 | 1689 | init_once); |
1da177e4 LT |
1690 | |
1691 | /* Hash may have been set up in inode_init_early */ | |
1692 | if (!hashdist) | |
1693 | return; | |
1694 | ||
1695 | inode_hashtable = | |
1696 | alloc_large_system_hash("Inode-cache", | |
1697 | sizeof(struct hlist_head), | |
1698 | ihash_entries, | |
1699 | 14, | |
1700 | 0, | |
1701 | &i_hash_shift, | |
1702 | &i_hash_mask, | |
1703 | 0); | |
1704 | ||
1705 | for (loop = 0; loop < (1 << i_hash_shift); loop++) | |
1706 | INIT_HLIST_HEAD(&inode_hashtable[loop]); | |
1707 | } | |
1708 | ||
1709 | void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev) | |
1710 | { | |
1711 | inode->i_mode = mode; | |
1712 | if (S_ISCHR(mode)) { | |
1713 | inode->i_fop = &def_chr_fops; | |
1714 | inode->i_rdev = rdev; | |
1715 | } else if (S_ISBLK(mode)) { | |
1716 | inode->i_fop = &def_blk_fops; | |
1717 | inode->i_rdev = rdev; | |
1718 | } else if (S_ISFIFO(mode)) | |
1719 | inode->i_fop = &def_fifo_fops; | |
1720 | else if (S_ISSOCK(mode)) | |
1721 | inode->i_fop = &bad_sock_fops; | |
1722 | else | |
af0d9ae8 MK |
1723 | printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for" |
1724 | " inode %s:%lu\n", mode, inode->i_sb->s_id, | |
1725 | inode->i_ino); | |
1da177e4 LT |
1726 | } |
1727 | EXPORT_SYMBOL(init_special_inode); | |
a1bd120d DM |
1728 | |
1729 | /** | |
eaae668d | 1730 | * inode_init_owner - Init uid,gid,mode for new inode according to posix standards |
a1bd120d DM |
1731 | * @inode: New inode |
1732 | * @dir: Directory inode | |
1733 | * @mode: mode of the new inode | |
1734 | */ | |
1735 | void inode_init_owner(struct inode *inode, const struct inode *dir, | |
62bb1091 | 1736 | umode_t mode) |
a1bd120d DM |
1737 | { |
1738 | inode->i_uid = current_fsuid(); | |
1739 | if (dir && dir->i_mode & S_ISGID) { | |
1740 | inode->i_gid = dir->i_gid; | |
1741 | if (S_ISDIR(mode)) | |
1742 | mode |= S_ISGID; | |
1743 | } else | |
1744 | inode->i_gid = current_fsgid(); | |
1745 | inode->i_mode = mode; | |
1746 | } | |
1747 | EXPORT_SYMBOL(inode_init_owner); | |
e795b717 | 1748 | |
2e149670 SH |
1749 | /** |
1750 | * inode_owner_or_capable - check current task permissions to inode | |
1751 | * @inode: inode being checked | |
1752 | * | |
1753 | * Return true if current either has CAP_FOWNER to the inode, or | |
1754 | * owns the file. | |
e795b717 | 1755 | */ |
2e149670 | 1756 | bool inode_owner_or_capable(const struct inode *inode) |
e795b717 SH |
1757 | { |
1758 | struct user_namespace *ns = inode_userns(inode); | |
1759 | ||
1760 | if (current_user_ns() == ns && current_fsuid() == inode->i_uid) | |
1761 | return true; | |
1762 | if (ns_capable(ns, CAP_FOWNER)) | |
1763 | return true; | |
1764 | return false; | |
1765 | } | |
2e149670 | 1766 | EXPORT_SYMBOL(inode_owner_or_capable); |