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 | */ |
e59cc473 | 5 | #include <linux/export.h> |
1da177e4 LT |
6 | #include <linux/fs.h> |
7 | #include <linux/mm.h> | |
1da177e4 | 8 | #include <linux/backing-dev.h> |
1da177e4 LT |
9 | #include <linux/hash.h> |
10 | #include <linux/swap.h> | |
11 | #include <linux/security.h> | |
1da177e4 LT |
12 | #include <linux/cdev.h> |
13 | #include <linux/bootmem.h> | |
3be25f49 | 14 | #include <linux/fsnotify.h> |
fc33a7bb | 15 | #include <linux/mount.h> |
f19d4a8f | 16 | #include <linux/posix_acl.h> |
9ce6e0be | 17 | #include <linux/prefetch.h> |
4b4563dc | 18 | #include <linux/buffer_head.h> /* for inode_has_buffers */ |
7ada4db8 | 19 | #include <linux/ratelimit.h> |
bc3b14cb | 20 | #include <linux/list_lru.h> |
0ae45f63 | 21 | #include <trace/events/writeback.h> |
a66979ab | 22 | #include "internal.h" |
1da177e4 | 23 | |
250df6ed | 24 | /* |
4b4563dc | 25 | * Inode locking rules: |
250df6ed DC |
26 | * |
27 | * inode->i_lock protects: | |
28 | * inode->i_state, inode->i_hash, __iget() | |
bc3b14cb | 29 | * Inode LRU list locks protect: |
98b745c6 | 30 | * inode->i_sb->s_inode_lru, inode->i_lru |
74278da9 DC |
31 | * inode->i_sb->s_inode_list_lock protects: |
32 | * inode->i_sb->s_inodes, inode->i_sb_list | |
f758eeab | 33 | * bdi->wb.list_lock protects: |
c7f54084 | 34 | * bdi->wb.b_{dirty,io,more_io,dirty_time}, inode->i_io_list |
67a23c49 DC |
35 | * inode_hash_lock protects: |
36 | * inode_hashtable, inode->i_hash | |
250df6ed DC |
37 | * |
38 | * Lock ordering: | |
55fa6091 | 39 | * |
74278da9 | 40 | * inode->i_sb->s_inode_list_lock |
55fa6091 | 41 | * inode->i_lock |
bc3b14cb | 42 | * Inode LRU list locks |
a66979ab | 43 | * |
f758eeab | 44 | * bdi->wb.list_lock |
a66979ab | 45 | * inode->i_lock |
67a23c49 DC |
46 | * |
47 | * inode_hash_lock | |
74278da9 | 48 | * inode->i_sb->s_inode_list_lock |
67a23c49 DC |
49 | * inode->i_lock |
50 | * | |
51 | * iunique_lock | |
52 | * inode_hash_lock | |
250df6ed DC |
53 | */ |
54 | ||
fa3536cc ED |
55 | static unsigned int i_hash_mask __read_mostly; |
56 | static unsigned int i_hash_shift __read_mostly; | |
67a23c49 DC |
57 | static struct hlist_head *inode_hashtable __read_mostly; |
58 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_hash_lock); | |
1da177e4 | 59 | |
7dcda1c9 JA |
60 | /* |
61 | * Empty aops. Can be used for the cases where the user does not | |
62 | * define any of the address_space operations. | |
63 | */ | |
64 | const struct address_space_operations empty_aops = { | |
65 | }; | |
66 | EXPORT_SYMBOL(empty_aops); | |
67 | ||
1da177e4 LT |
68 | /* |
69 | * Statistics gathering.. | |
70 | */ | |
71 | struct inodes_stat_t inodes_stat; | |
72 | ||
3942c07c GC |
73 | static DEFINE_PER_CPU(unsigned long, nr_inodes); |
74 | static DEFINE_PER_CPU(unsigned long, nr_unused); | |
cffbc8aa | 75 | |
6b3304b5 | 76 | static struct kmem_cache *inode_cachep __read_mostly; |
1da177e4 | 77 | |
3942c07c | 78 | static long get_nr_inodes(void) |
cffbc8aa | 79 | { |
3e880fb5 | 80 | int i; |
3942c07c | 81 | long sum = 0; |
3e880fb5 NP |
82 | for_each_possible_cpu(i) |
83 | sum += per_cpu(nr_inodes, i); | |
84 | return sum < 0 ? 0 : sum; | |
cffbc8aa DC |
85 | } |
86 | ||
3942c07c | 87 | static inline long get_nr_inodes_unused(void) |
cffbc8aa | 88 | { |
fcb94f72 | 89 | int i; |
3942c07c | 90 | long sum = 0; |
fcb94f72 DC |
91 | for_each_possible_cpu(i) |
92 | sum += per_cpu(nr_unused, i); | |
93 | return sum < 0 ? 0 : sum; | |
cffbc8aa DC |
94 | } |
95 | ||
3942c07c | 96 | long get_nr_dirty_inodes(void) |
cffbc8aa | 97 | { |
3e880fb5 | 98 | /* not actually dirty inodes, but a wild approximation */ |
3942c07c | 99 | long nr_dirty = get_nr_inodes() - get_nr_inodes_unused(); |
cffbc8aa | 100 | return nr_dirty > 0 ? nr_dirty : 0; |
cffbc8aa DC |
101 | } |
102 | ||
103 | /* | |
104 | * Handle nr_inode sysctl | |
105 | */ | |
106 | #ifdef CONFIG_SYSCTL | |
1f7e0616 | 107 | int proc_nr_inodes(struct ctl_table *table, int write, |
cffbc8aa DC |
108 | void __user *buffer, size_t *lenp, loff_t *ppos) |
109 | { | |
110 | inodes_stat.nr_inodes = get_nr_inodes(); | |
fcb94f72 | 111 | inodes_stat.nr_unused = get_nr_inodes_unused(); |
3942c07c | 112 | return proc_doulongvec_minmax(table, write, buffer, lenp, ppos); |
cffbc8aa DC |
113 | } |
114 | #endif | |
115 | ||
bd9b51e7 AV |
116 | static int no_open(struct inode *inode, struct file *file) |
117 | { | |
118 | return -ENXIO; | |
119 | } | |
120 | ||
2cb1599f DC |
121 | /** |
122 | * inode_init_always - perform inode structure intialisation | |
0bc02f3f RD |
123 | * @sb: superblock inode belongs to |
124 | * @inode: inode to initialise | |
2cb1599f DC |
125 | * |
126 | * These are initializations that need to be done on every inode | |
127 | * allocation as the fields are not initialised by slab allocation. | |
128 | */ | |
54e34621 | 129 | int inode_init_always(struct super_block *sb, struct inode *inode) |
1da177e4 | 130 | { |
6e1d5dcc | 131 | static const struct inode_operations empty_iops; |
bd9b51e7 | 132 | static const struct file_operations no_open_fops = {.open = no_open}; |
6b3304b5 | 133 | struct address_space *const mapping = &inode->i_data; |
2cb1599f DC |
134 | |
135 | inode->i_sb = sb; | |
136 | inode->i_blkbits = sb->s_blocksize_bits; | |
137 | inode->i_flags = 0; | |
138 | atomic_set(&inode->i_count, 1); | |
139 | inode->i_op = &empty_iops; | |
bd9b51e7 | 140 | inode->i_fop = &no_open_fops; |
a78ef704 | 141 | inode->__i_nlink = 1; |
3ddcd056 | 142 | inode->i_opflags = 0; |
92361636 EB |
143 | i_uid_write(inode, 0); |
144 | i_gid_write(inode, 0); | |
2cb1599f DC |
145 | atomic_set(&inode->i_writecount, 0); |
146 | inode->i_size = 0; | |
147 | inode->i_blocks = 0; | |
148 | inode->i_bytes = 0; | |
149 | inode->i_generation = 0; | |
2cb1599f DC |
150 | inode->i_pipe = NULL; |
151 | inode->i_bdev = NULL; | |
152 | inode->i_cdev = NULL; | |
61ba64fc | 153 | inode->i_link = NULL; |
2cb1599f DC |
154 | inode->i_rdev = 0; |
155 | inode->dirtied_when = 0; | |
6146f0d5 MZ |
156 | |
157 | if (security_inode_alloc(inode)) | |
54e34621 | 158 | goto out; |
2cb1599f DC |
159 | spin_lock_init(&inode->i_lock); |
160 | lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key); | |
161 | ||
162 | mutex_init(&inode->i_mutex); | |
163 | lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key); | |
164 | ||
bd5fe6c5 | 165 | atomic_set(&inode->i_dio_count, 0); |
2cb1599f DC |
166 | |
167 | mapping->a_ops = &empty_aops; | |
168 | mapping->host = inode; | |
169 | mapping->flags = 0; | |
4bb5f5d9 | 170 | atomic_set(&mapping->i_mmap_writable, 0); |
3c1d4378 | 171 | mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE); |
252aa6f5 | 172 | mapping->private_data = NULL; |
2cb1599f | 173 | mapping->writeback_index = 0; |
1cac41cb MB |
174 | #if defined(CONFIG_FMP_ECRYPT_FS) || defined(CONFIG_FMP_EXT4CRYPT_FS) |
175 | mapping->iv = NULL; | |
176 | memset(mapping->key, 0, KEY_MAX_SIZE); | |
177 | mapping->key_length = 0; | |
178 | mapping->alg = NULL; | |
179 | mapping->sensitive_data_index = 0; | |
180 | mapping->hash_tfm = NULL; | |
181 | #ifdef CONFIG_CRYPTO_FIPS | |
182 | mapping->cc_enable = 0; | |
183 | #endif | |
184 | mapping->private_enc_mode = 0; | |
185 | mapping->private_algo_mode = 0; | |
186 | mapping->plain_text = 0; | |
187 | #endif | |
188 | #ifdef CONFIG_SDP | |
189 | mapping->userid = 0; | |
190 | #endif | |
2cb1599f DC |
191 | inode->i_private = NULL; |
192 | inode->i_mapping = mapping; | |
b3d9b7a3 | 193 | INIT_HLIST_HEAD(&inode->i_dentry); /* buggered by rcu freeing */ |
f19d4a8f AV |
194 | #ifdef CONFIG_FS_POSIX_ACL |
195 | inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED; | |
196 | #endif | |
2cb1599f | 197 | |
3be25f49 EP |
198 | #ifdef CONFIG_FSNOTIFY |
199 | inode->i_fsnotify_mask = 0; | |
200 | #endif | |
4a075e39 | 201 | inode->i_flctx = NULL; |
3e880fb5 | 202 | this_cpu_inc(nr_inodes); |
cffbc8aa | 203 | |
54e34621 | 204 | return 0; |
54e34621 CH |
205 | out: |
206 | return -ENOMEM; | |
1da177e4 | 207 | } |
2cb1599f DC |
208 | EXPORT_SYMBOL(inode_init_always); |
209 | ||
210 | static struct inode *alloc_inode(struct super_block *sb) | |
211 | { | |
212 | struct inode *inode; | |
213 | ||
214 | if (sb->s_op->alloc_inode) | |
215 | inode = sb->s_op->alloc_inode(sb); | |
216 | else | |
217 | inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL); | |
218 | ||
54e34621 CH |
219 | if (!inode) |
220 | return NULL; | |
221 | ||
222 | if (unlikely(inode_init_always(sb, inode))) { | |
223 | if (inode->i_sb->s_op->destroy_inode) | |
224 | inode->i_sb->s_op->destroy_inode(inode); | |
225 | else | |
226 | kmem_cache_free(inode_cachep, inode); | |
227 | return NULL; | |
228 | } | |
229 | ||
230 | return inode; | |
2cb1599f | 231 | } |
1da177e4 | 232 | |
ff0c7d15 NP |
233 | void free_inode_nonrcu(struct inode *inode) |
234 | { | |
235 | kmem_cache_free(inode_cachep, inode); | |
236 | } | |
237 | EXPORT_SYMBOL(free_inode_nonrcu); | |
238 | ||
2e00c97e | 239 | void __destroy_inode(struct inode *inode) |
1da177e4 | 240 | { |
b7542f8c | 241 | BUG_ON(inode_has_buffers(inode)); |
52ebea74 | 242 | inode_detach_wb(inode); |
1da177e4 | 243 | security_inode_free(inode); |
3be25f49 | 244 | fsnotify_inode_delete(inode); |
4a075e39 | 245 | locks_free_lock_context(inode->i_flctx); |
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 | { | |
f4e0c30c AV |
346 | if (unlikely(inode->i_nlink == 0)) { |
347 | WARN_ON(!(inode->i_state & I_LINKABLE)); | |
7ada4db8 | 348 | atomic_long_dec(&inode->i_sb->s_remove_count); |
f4e0c30c | 349 | } |
7ada4db8 MS |
350 | |
351 | inode->__i_nlink++; | |
352 | } | |
353 | EXPORT_SYMBOL(inc_nlink); | |
354 | ||
2aa15890 MS |
355 | void address_space_init_once(struct address_space *mapping) |
356 | { | |
357 | memset(mapping, 0, sizeof(*mapping)); | |
358 | INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC); | |
359 | spin_lock_init(&mapping->tree_lock); | |
c8c06efa | 360 | init_rwsem(&mapping->i_mmap_rwsem); |
2aa15890 MS |
361 | INIT_LIST_HEAD(&mapping->private_list); |
362 | spin_lock_init(&mapping->private_lock); | |
6b2dbba8 | 363 | mapping->i_mmap = RB_ROOT; |
2aa15890 MS |
364 | } |
365 | EXPORT_SYMBOL(address_space_init_once); | |
366 | ||
1da177e4 LT |
367 | /* |
368 | * These are initializations that only need to be done | |
369 | * once, because the fields are idempotent across use | |
370 | * of the inode, so let the slab aware of that. | |
371 | */ | |
372 | void inode_init_once(struct inode *inode) | |
373 | { | |
374 | memset(inode, 0, sizeof(*inode)); | |
375 | INIT_HLIST_NODE(&inode->i_hash); | |
1da177e4 | 376 | INIT_LIST_HEAD(&inode->i_devices); |
c7f54084 | 377 | INIT_LIST_HEAD(&inode->i_io_list); |
7ccf19a8 | 378 | INIT_LIST_HEAD(&inode->i_lru); |
2aa15890 | 379 | address_space_init_once(&inode->i_data); |
1da177e4 | 380 | i_size_ordered_init(inode); |
3be25f49 | 381 | #ifdef CONFIG_FSNOTIFY |
e61ce867 | 382 | INIT_HLIST_HEAD(&inode->i_fsnotify_marks); |
3be25f49 | 383 | #endif |
1da177e4 | 384 | } |
1da177e4 LT |
385 | EXPORT_SYMBOL(inode_init_once); |
386 | ||
51cc5068 | 387 | static void init_once(void *foo) |
1da177e4 | 388 | { |
6b3304b5 | 389 | struct inode *inode = (struct inode *) foo; |
1da177e4 | 390 | |
a35afb83 | 391 | inode_init_once(inode); |
1da177e4 LT |
392 | } |
393 | ||
394 | /* | |
250df6ed | 395 | * inode->i_lock must be held |
1da177e4 | 396 | */ |
6b3304b5 | 397 | void __iget(struct inode *inode) |
1da177e4 | 398 | { |
9e38d86f NP |
399 | atomic_inc(&inode->i_count); |
400 | } | |
2e147f1e | 401 | |
7de9c6ee AV |
402 | /* |
403 | * get additional reference to inode; caller must already hold one. | |
404 | */ | |
405 | void ihold(struct inode *inode) | |
406 | { | |
407 | WARN_ON(atomic_inc_return(&inode->i_count) < 2); | |
408 | } | |
409 | EXPORT_SYMBOL(ihold); | |
410 | ||
9e38d86f NP |
411 | static void inode_lru_list_add(struct inode *inode) |
412 | { | |
bc3b14cb | 413 | if (list_lru_add(&inode->i_sb->s_inode_lru, &inode->i_lru)) |
fcb94f72 | 414 | this_cpu_inc(nr_unused); |
9e38d86f | 415 | } |
2e147f1e | 416 | |
4eff96dd JK |
417 | /* |
418 | * Add inode to LRU if needed (inode is unused and clean). | |
419 | * | |
420 | * Needs inode->i_lock held. | |
421 | */ | |
422 | void inode_add_lru(struct inode *inode) | |
423 | { | |
0ae45f63 TT |
424 | if (!(inode->i_state & (I_DIRTY_ALL | I_SYNC | |
425 | I_FREEING | I_WILL_FREE)) && | |
4eff96dd JK |
426 | !atomic_read(&inode->i_count) && inode->i_sb->s_flags & MS_ACTIVE) |
427 | inode_lru_list_add(inode); | |
428 | } | |
429 | ||
430 | ||
9e38d86f NP |
431 | static void inode_lru_list_del(struct inode *inode) |
432 | { | |
bc3b14cb DC |
433 | |
434 | if (list_lru_del(&inode->i_sb->s_inode_lru, &inode->i_lru)) | |
fcb94f72 | 435 | this_cpu_dec(nr_unused); |
1da177e4 LT |
436 | } |
437 | ||
646ec461 CH |
438 | /** |
439 | * inode_sb_list_add - add inode to the superblock list of inodes | |
440 | * @inode: inode to add | |
441 | */ | |
442 | void inode_sb_list_add(struct inode *inode) | |
443 | { | |
74278da9 | 444 | spin_lock(&inode->i_sb->s_inode_list_lock); |
55fa6091 | 445 | list_add(&inode->i_sb_list, &inode->i_sb->s_inodes); |
74278da9 | 446 | spin_unlock(&inode->i_sb->s_inode_list_lock); |
646ec461 CH |
447 | } |
448 | EXPORT_SYMBOL_GPL(inode_sb_list_add); | |
449 | ||
55fa6091 | 450 | static inline void inode_sb_list_del(struct inode *inode) |
646ec461 | 451 | { |
a209dfc7 | 452 | if (!list_empty(&inode->i_sb_list)) { |
74278da9 | 453 | spin_lock(&inode->i_sb->s_inode_list_lock); |
a209dfc7 | 454 | list_del_init(&inode->i_sb_list); |
74278da9 | 455 | spin_unlock(&inode->i_sb->s_inode_list_lock); |
a209dfc7 | 456 | } |
646ec461 CH |
457 | } |
458 | ||
4c51acbc DC |
459 | static unsigned long hash(struct super_block *sb, unsigned long hashval) |
460 | { | |
461 | unsigned long tmp; | |
462 | ||
463 | tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) / | |
464 | L1_CACHE_BYTES; | |
4b4563dc CH |
465 | tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> i_hash_shift); |
466 | return tmp & i_hash_mask; | |
4c51acbc DC |
467 | } |
468 | ||
469 | /** | |
470 | * __insert_inode_hash - hash an inode | |
471 | * @inode: unhashed inode | |
472 | * @hashval: unsigned long value used to locate this object in the | |
473 | * inode_hashtable. | |
474 | * | |
475 | * Add an inode to the inode hash for this superblock. | |
476 | */ | |
477 | void __insert_inode_hash(struct inode *inode, unsigned long hashval) | |
478 | { | |
646ec461 CH |
479 | struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval); |
480 | ||
67a23c49 | 481 | spin_lock(&inode_hash_lock); |
250df6ed | 482 | spin_lock(&inode->i_lock); |
646ec461 | 483 | hlist_add_head(&inode->i_hash, b); |
250df6ed | 484 | spin_unlock(&inode->i_lock); |
67a23c49 | 485 | spin_unlock(&inode_hash_lock); |
4c51acbc DC |
486 | } |
487 | EXPORT_SYMBOL(__insert_inode_hash); | |
488 | ||
4c51acbc | 489 | /** |
f2ee7abf | 490 | * __remove_inode_hash - remove an inode from the hash |
4c51acbc DC |
491 | * @inode: inode to unhash |
492 | * | |
493 | * Remove an inode from the superblock. | |
494 | */ | |
f2ee7abf | 495 | void __remove_inode_hash(struct inode *inode) |
4c51acbc | 496 | { |
67a23c49 | 497 | spin_lock(&inode_hash_lock); |
250df6ed | 498 | spin_lock(&inode->i_lock); |
4c51acbc | 499 | hlist_del_init(&inode->i_hash); |
250df6ed | 500 | spin_unlock(&inode->i_lock); |
67a23c49 | 501 | spin_unlock(&inode_hash_lock); |
4c51acbc | 502 | } |
f2ee7abf | 503 | EXPORT_SYMBOL(__remove_inode_hash); |
4c51acbc | 504 | |
dbd5768f | 505 | void clear_inode(struct inode *inode) |
b0683aa6 AV |
506 | { |
507 | might_sleep(); | |
08142579 JK |
508 | /* |
509 | * We have to cycle tree_lock here because reclaim can be still in the | |
510 | * process of removing the last page (in __delete_from_page_cache()) | |
511 | * and we must not free mapping under it. | |
512 | */ | |
513 | spin_lock_irq(&inode->i_data.tree_lock); | |
b0683aa6 | 514 | BUG_ON(inode->i_data.nrpages); |
91b0abe3 | 515 | BUG_ON(inode->i_data.nrshadows); |
08142579 | 516 | spin_unlock_irq(&inode->i_data.tree_lock); |
b0683aa6 AV |
517 | BUG_ON(!list_empty(&inode->i_data.private_list)); |
518 | BUG_ON(!(inode->i_state & I_FREEING)); | |
519 | BUG_ON(inode->i_state & I_CLEAR); | |
fa0d7e3d | 520 | /* don't need i_lock here, no concurrent mods to i_state */ |
b0683aa6 AV |
521 | inode->i_state = I_FREEING | I_CLEAR; |
522 | } | |
dbd5768f | 523 | EXPORT_SYMBOL(clear_inode); |
b0683aa6 | 524 | |
b2b2af8e DC |
525 | /* |
526 | * Free the inode passed in, removing it from the lists it is still connected | |
527 | * to. We remove any pages still attached to the inode and wait for any IO that | |
528 | * is still in progress before finally destroying the inode. | |
529 | * | |
530 | * An inode must already be marked I_FREEING so that we avoid the inode being | |
531 | * moved back onto lists if we race with other code that manipulates the lists | |
532 | * (e.g. writeback_single_inode). The caller is responsible for setting this. | |
533 | * | |
534 | * An inode must already be removed from the LRU list before being evicted from | |
535 | * the cache. This should occur atomically with setting the I_FREEING state | |
536 | * flag, so no inodes here should ever be on the LRU when being evicted. | |
537 | */ | |
644da596 | 538 | static void evict(struct inode *inode) |
b4272d4c AV |
539 | { |
540 | const struct super_operations *op = inode->i_sb->s_op; | |
541 | ||
b2b2af8e DC |
542 | BUG_ON(!(inode->i_state & I_FREEING)); |
543 | BUG_ON(!list_empty(&inode->i_lru)); | |
544 | ||
c7f54084 DC |
545 | if (!list_empty(&inode->i_io_list)) |
546 | inode_io_list_del(inode); | |
b12362bd | 547 | |
55fa6091 DC |
548 | inode_sb_list_del(inode); |
549 | ||
169ebd90 JK |
550 | /* |
551 | * Wait for flusher thread to be done with the inode so that filesystem | |
552 | * does not start destroying it while writeback is still running. Since | |
553 | * the inode has I_FREEING set, flusher thread won't start new work on | |
554 | * the inode. We just have to wait for running writeback to finish. | |
555 | */ | |
556 | inode_wait_for_writeback(inode); | |
7994e6f7 | 557 | |
be7ce416 AV |
558 | if (op->evict_inode) { |
559 | op->evict_inode(inode); | |
b4272d4c | 560 | } else { |
91b0abe3 | 561 | truncate_inode_pages_final(&inode->i_data); |
dbd5768f | 562 | clear_inode(inode); |
b4272d4c | 563 | } |
661074e9 AV |
564 | if (S_ISBLK(inode->i_mode) && inode->i_bdev) |
565 | bd_forget(inode); | |
566 | if (S_ISCHR(inode->i_mode) && inode->i_cdev) | |
567 | cd_forget(inode); | |
b2b2af8e DC |
568 | |
569 | remove_inode_hash(inode); | |
570 | ||
571 | spin_lock(&inode->i_lock); | |
572 | wake_up_bit(&inode->i_state, __I_NEW); | |
573 | BUG_ON(inode->i_state != (I_FREEING | I_CLEAR)); | |
574 | spin_unlock(&inode->i_lock); | |
575 | ||
576 | destroy_inode(inode); | |
b4272d4c AV |
577 | } |
578 | ||
1da177e4 LT |
579 | /* |
580 | * dispose_list - dispose of the contents of a local list | |
581 | * @head: the head of the list to free | |
582 | * | |
583 | * Dispose-list gets a local list with local inodes in it, so it doesn't | |
584 | * need to worry about list corruption and SMP locks. | |
585 | */ | |
586 | static void dispose_list(struct list_head *head) | |
587 | { | |
1da177e4 LT |
588 | while (!list_empty(head)) { |
589 | struct inode *inode; | |
590 | ||
7ccf19a8 NP |
591 | inode = list_first_entry(head, struct inode, i_lru); |
592 | list_del_init(&inode->i_lru); | |
1da177e4 | 593 | |
644da596 | 594 | evict(inode); |
ac05fbb4 | 595 | cond_resched(); |
1da177e4 | 596 | } |
1da177e4 LT |
597 | } |
598 | ||
63997e98 AV |
599 | /** |
600 | * evict_inodes - evict all evictable inodes for a superblock | |
601 | * @sb: superblock to operate on | |
602 | * | |
603 | * Make sure that no inodes with zero refcount are retained. This is | |
604 | * called by superblock shutdown after having MS_ACTIVE flag removed, | |
605 | * so any inode reaching zero refcount during or after that call will | |
606 | * be immediately evicted. | |
1da177e4 | 607 | */ |
63997e98 | 608 | void evict_inodes(struct super_block *sb) |
1da177e4 | 609 | { |
63997e98 AV |
610 | struct inode *inode, *next; |
611 | LIST_HEAD(dispose); | |
1da177e4 | 612 | |
ac05fbb4 | 613 | again: |
74278da9 | 614 | spin_lock(&sb->s_inode_list_lock); |
63997e98 AV |
615 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
616 | if (atomic_read(&inode->i_count)) | |
aabb8fdb | 617 | continue; |
250df6ed DC |
618 | |
619 | spin_lock(&inode->i_lock); | |
620 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
621 | spin_unlock(&inode->i_lock); | |
1da177e4 | 622 | continue; |
250df6ed | 623 | } |
63997e98 AV |
624 | |
625 | inode->i_state |= I_FREEING; | |
02afc410 | 626 | inode_lru_list_del(inode); |
250df6ed | 627 | spin_unlock(&inode->i_lock); |
02afc410 | 628 | list_add(&inode->i_lru, &dispose); |
ac05fbb4 JB |
629 | |
630 | /* | |
631 | * We can have a ton of inodes to evict at unmount time given | |
632 | * enough memory, check to see if we need to go to sleep for a | |
633 | * bit so we don't livelock. | |
634 | */ | |
635 | if (need_resched()) { | |
636 | spin_unlock(&sb->s_inode_list_lock); | |
637 | cond_resched(); | |
638 | dispose_list(&dispose); | |
639 | goto again; | |
640 | } | |
1da177e4 | 641 | } |
74278da9 | 642 | spin_unlock(&sb->s_inode_list_lock); |
63997e98 AV |
643 | |
644 | dispose_list(&dispose); | |
1da177e4 LT |
645 | } |
646 | ||
1da177e4 | 647 | /** |
a0318786 CH |
648 | * invalidate_inodes - attempt to free all inodes on a superblock |
649 | * @sb: superblock to operate on | |
93b270f7 | 650 | * @kill_dirty: flag to guide handling of dirty inodes |
1da177e4 | 651 | * |
a0318786 CH |
652 | * Attempts to free all inodes for a given superblock. If there were any |
653 | * busy inodes return a non-zero value, else zero. | |
93b270f7 N |
654 | * If @kill_dirty is set, discard dirty inodes too, otherwise treat |
655 | * them as busy. | |
1da177e4 | 656 | */ |
93b270f7 | 657 | int invalidate_inodes(struct super_block *sb, bool kill_dirty) |
1da177e4 | 658 | { |
cffbc8aa | 659 | int busy = 0; |
a0318786 CH |
660 | struct inode *inode, *next; |
661 | LIST_HEAD(dispose); | |
1da177e4 | 662 | |
74278da9 | 663 | spin_lock(&sb->s_inode_list_lock); |
a0318786 | 664 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
250df6ed DC |
665 | spin_lock(&inode->i_lock); |
666 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
667 | spin_unlock(&inode->i_lock); | |
aabb8fdb | 668 | continue; |
250df6ed | 669 | } |
0ae45f63 | 670 | if (inode->i_state & I_DIRTY_ALL && !kill_dirty) { |
250df6ed | 671 | spin_unlock(&inode->i_lock); |
93b270f7 N |
672 | busy = 1; |
673 | continue; | |
674 | } | |
99a38919 | 675 | if (atomic_read(&inode->i_count)) { |
250df6ed | 676 | spin_unlock(&inode->i_lock); |
99a38919 | 677 | busy = 1; |
1da177e4 LT |
678 | continue; |
679 | } | |
99a38919 | 680 | |
99a38919 | 681 | inode->i_state |= I_FREEING; |
02afc410 | 682 | inode_lru_list_del(inode); |
250df6ed | 683 | spin_unlock(&inode->i_lock); |
02afc410 | 684 | list_add(&inode->i_lru, &dispose); |
1da177e4 | 685 | } |
74278da9 | 686 | spin_unlock(&sb->s_inode_list_lock); |
1da177e4 | 687 | |
a0318786 | 688 | dispose_list(&dispose); |
1da177e4 LT |
689 | |
690 | return busy; | |
691 | } | |
1da177e4 | 692 | |
1da177e4 | 693 | /* |
bc3b14cb | 694 | * Isolate the inode from the LRU in preparation for freeing it. |
1da177e4 LT |
695 | * |
696 | * Any inodes which are pinned purely because of attached pagecache have their | |
9e38d86f NP |
697 | * pagecache removed. If the inode has metadata buffers attached to |
698 | * mapping->private_list then try to remove them. | |
1da177e4 | 699 | * |
9e38d86f NP |
700 | * If the inode has the I_REFERENCED flag set, then it means that it has been |
701 | * used recently - the flag is set in iput_final(). When we encounter such an | |
702 | * inode, clear the flag and move it to the back of the LRU so it gets another | |
703 | * pass through the LRU before it gets reclaimed. This is necessary because of | |
704 | * the fact we are doing lazy LRU updates to minimise lock contention so the | |
705 | * LRU does not have strict ordering. Hence we don't want to reclaim inodes | |
706 | * with this flag set because they are the inodes that are out of order. | |
1da177e4 | 707 | */ |
3f97b163 VD |
708 | static enum lru_status inode_lru_isolate(struct list_head *item, |
709 | struct list_lru_one *lru, spinlock_t *lru_lock, void *arg) | |
1da177e4 | 710 | { |
bc3b14cb DC |
711 | struct list_head *freeable = arg; |
712 | struct inode *inode = container_of(item, struct inode, i_lru); | |
1da177e4 | 713 | |
bc3b14cb DC |
714 | /* |
715 | * we are inverting the lru lock/inode->i_lock here, so use a trylock. | |
716 | * If we fail to get the lock, just skip it. | |
717 | */ | |
718 | if (!spin_trylock(&inode->i_lock)) | |
719 | return LRU_SKIP; | |
1da177e4 | 720 | |
bc3b14cb DC |
721 | /* |
722 | * Referenced or dirty inodes are still in use. Give them another pass | |
723 | * through the LRU as we canot reclaim them now. | |
724 | */ | |
725 | if (atomic_read(&inode->i_count) || | |
726 | (inode->i_state & ~I_REFERENCED)) { | |
3f97b163 | 727 | list_lru_isolate(lru, &inode->i_lru); |
bc3b14cb DC |
728 | spin_unlock(&inode->i_lock); |
729 | this_cpu_dec(nr_unused); | |
730 | return LRU_REMOVED; | |
731 | } | |
1da177e4 | 732 | |
bc3b14cb DC |
733 | /* recently referenced inodes get one more pass */ |
734 | if (inode->i_state & I_REFERENCED) { | |
735 | inode->i_state &= ~I_REFERENCED; | |
736 | spin_unlock(&inode->i_lock); | |
737 | return LRU_ROTATE; | |
738 | } | |
1da177e4 | 739 | |
bc3b14cb DC |
740 | if (inode_has_buffers(inode) || inode->i_data.nrpages) { |
741 | __iget(inode); | |
742 | spin_unlock(&inode->i_lock); | |
743 | spin_unlock(lru_lock); | |
744 | if (remove_inode_buffers(inode)) { | |
745 | unsigned long reap; | |
746 | reap = invalidate_mapping_pages(&inode->i_data, 0, -1); | |
747 | if (current_is_kswapd()) | |
748 | __count_vm_events(KSWAPD_INODESTEAL, reap); | |
749 | else | |
750 | __count_vm_events(PGINODESTEAL, reap); | |
751 | if (current->reclaim_state) | |
752 | current->reclaim_state->reclaimed_slab += reap; | |
02afc410 | 753 | } |
bc3b14cb DC |
754 | iput(inode); |
755 | spin_lock(lru_lock); | |
756 | return LRU_RETRY; | |
757 | } | |
02afc410 | 758 | |
bc3b14cb DC |
759 | WARN_ON(inode->i_state & I_NEW); |
760 | inode->i_state |= I_FREEING; | |
3f97b163 | 761 | list_lru_isolate_move(lru, &inode->i_lru, freeable); |
bc3b14cb | 762 | spin_unlock(&inode->i_lock); |
9e38d86f | 763 | |
bc3b14cb DC |
764 | this_cpu_dec(nr_unused); |
765 | return LRU_REMOVED; | |
766 | } | |
7ccf19a8 | 767 | |
bc3b14cb DC |
768 | /* |
769 | * Walk the superblock inode LRU for freeable inodes and attempt to free them. | |
770 | * This is called from the superblock shrinker function with a number of inodes | |
771 | * to trim from the LRU. Inodes to be freed are moved to a temporary list and | |
772 | * then are freed outside inode_lock by dispose_list(). | |
773 | */ | |
503c358c | 774 | long prune_icache_sb(struct super_block *sb, struct shrink_control *sc) |
bc3b14cb DC |
775 | { |
776 | LIST_HEAD(freeable); | |
777 | long freed; | |
1da177e4 | 778 | |
503c358c VD |
779 | freed = list_lru_shrink_walk(&sb->s_inode_lru, sc, |
780 | inode_lru_isolate, &freeable); | |
1da177e4 | 781 | dispose_list(&freeable); |
0a234c6d | 782 | return freed; |
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 | 793 | { |
6b3304b5 | 794 | struct inode *inode = NULL; |
1da177e4 LT |
795 | |
796 | repeat: | |
b67bfe0d | 797 | hlist_for_each_entry(inode, head, i_hash) { |
5a3cd992 | 798 | if (inode->i_sb != sb) |
1da177e4 | 799 | continue; |
5a3cd992 | 800 | if (!test(inode, data)) |
1da177e4 | 801 | continue; |
5a3cd992 | 802 | spin_lock(&inode->i_lock); |
a4ffdde6 | 803 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
804 | __wait_on_freeing_inode(inode); |
805 | goto repeat; | |
806 | } | |
f7899bd5 | 807 | __iget(inode); |
250df6ed | 808 | spin_unlock(&inode->i_lock); |
f7899bd5 | 809 | return inode; |
1da177e4 | 810 | } |
f7899bd5 | 811 | return NULL; |
1da177e4 LT |
812 | } |
813 | ||
814 | /* | |
815 | * find_inode_fast is the fast path version of find_inode, see the comment at | |
816 | * iget_locked for details. | |
817 | */ | |
6b3304b5 MK |
818 | static struct inode *find_inode_fast(struct super_block *sb, |
819 | struct hlist_head *head, unsigned long ino) | |
1da177e4 | 820 | { |
6b3304b5 | 821 | struct inode *inode = NULL; |
1da177e4 LT |
822 | |
823 | repeat: | |
b67bfe0d | 824 | hlist_for_each_entry(inode, head, i_hash) { |
5a3cd992 | 825 | if (inode->i_ino != ino) |
1da177e4 | 826 | continue; |
5a3cd992 | 827 | if (inode->i_sb != sb) |
1da177e4 | 828 | continue; |
5a3cd992 | 829 | spin_lock(&inode->i_lock); |
a4ffdde6 | 830 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
831 | __wait_on_freeing_inode(inode); |
832 | goto repeat; | |
833 | } | |
f7899bd5 | 834 | __iget(inode); |
250df6ed | 835 | spin_unlock(&inode->i_lock); |
f7899bd5 | 836 | return inode; |
1da177e4 | 837 | } |
f7899bd5 | 838 | return NULL; |
8290c35f DC |
839 | } |
840 | ||
f991bd2e ED |
841 | /* |
842 | * Each cpu owns a range of LAST_INO_BATCH numbers. | |
843 | * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations, | |
844 | * to renew the exhausted range. | |
8290c35f | 845 | * |
f991bd2e ED |
846 | * This does not significantly increase overflow rate because every CPU can |
847 | * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is | |
848 | * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the | |
849 | * 2^32 range, and is a worst-case. Even a 50% wastage would only increase | |
850 | * overflow rate by 2x, which does not seem too significant. | |
851 | * | |
852 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
853 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
854 | * here to attempt to avoid that. | |
8290c35f | 855 | */ |
f991bd2e ED |
856 | #define LAST_INO_BATCH 1024 |
857 | static DEFINE_PER_CPU(unsigned int, last_ino); | |
858 | ||
85fe4025 | 859 | unsigned int get_next_ino(void) |
8290c35f | 860 | { |
f991bd2e ED |
861 | unsigned int *p = &get_cpu_var(last_ino); |
862 | unsigned int res = *p; | |
8290c35f | 863 | |
f991bd2e ED |
864 | #ifdef CONFIG_SMP |
865 | if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) { | |
866 | static atomic_t shared_last_ino; | |
867 | int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino); | |
868 | ||
869 | res = next - LAST_INO_BATCH; | |
870 | } | |
871 | #endif | |
872 | ||
2adc376c CM |
873 | res++; |
874 | /* get_next_ino should not provide a 0 inode number */ | |
875 | if (unlikely(!res)) | |
876 | res++; | |
877 | *p = res; | |
f991bd2e ED |
878 | put_cpu_var(last_ino); |
879 | return res; | |
8290c35f | 880 | } |
85fe4025 | 881 | EXPORT_SYMBOL(get_next_ino); |
8290c35f | 882 | |
a209dfc7 ED |
883 | /** |
884 | * new_inode_pseudo - obtain an inode | |
885 | * @sb: superblock | |
886 | * | |
887 | * Allocates a new inode for given superblock. | |
888 | * Inode wont be chained in superblock s_inodes list | |
889 | * This means : | |
890 | * - fs can't be unmount | |
891 | * - quotas, fsnotify, writeback can't work | |
892 | */ | |
893 | struct inode *new_inode_pseudo(struct super_block *sb) | |
894 | { | |
895 | struct inode *inode = alloc_inode(sb); | |
896 | ||
897 | if (inode) { | |
898 | spin_lock(&inode->i_lock); | |
899 | inode->i_state = 0; | |
900 | spin_unlock(&inode->i_lock); | |
901 | INIT_LIST_HEAD(&inode->i_sb_list); | |
902 | } | |
903 | return inode; | |
904 | } | |
905 | ||
1da177e4 LT |
906 | /** |
907 | * new_inode - obtain an inode | |
908 | * @sb: superblock | |
909 | * | |
769848c0 | 910 | * Allocates a new inode for given superblock. The default gfp_mask |
3c1d4378 | 911 | * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE. |
769848c0 MG |
912 | * If HIGHMEM pages are unsuitable or it is known that pages allocated |
913 | * for the page cache are not reclaimable or migratable, | |
914 | * mapping_set_gfp_mask() must be called with suitable flags on the | |
915 | * newly created inode's mapping | |
916 | * | |
1da177e4 LT |
917 | */ |
918 | struct inode *new_inode(struct super_block *sb) | |
919 | { | |
6b3304b5 | 920 | struct inode *inode; |
1da177e4 | 921 | |
74278da9 | 922 | spin_lock_prefetch(&sb->s_inode_list_lock); |
6b3304b5 | 923 | |
a209dfc7 ED |
924 | inode = new_inode_pseudo(sb); |
925 | if (inode) | |
55fa6091 | 926 | inode_sb_list_add(inode); |
1da177e4 LT |
927 | return inode; |
928 | } | |
1da177e4 LT |
929 | EXPORT_SYMBOL(new_inode); |
930 | ||
14358e6d | 931 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
e096d0c7 JB |
932 | void lockdep_annotate_inode_mutex_key(struct inode *inode) |
933 | { | |
a3314a0e | 934 | if (S_ISDIR(inode->i_mode)) { |
1e89a5e1 PZ |
935 | struct file_system_type *type = inode->i_sb->s_type; |
936 | ||
9a7aa12f | 937 | /* Set new key only if filesystem hasn't already changed it */ |
978d6d8c | 938 | if (lockdep_match_class(&inode->i_mutex, &type->i_mutex_key)) { |
9a7aa12f JK |
939 | /* |
940 | * ensure nobody is actually holding i_mutex | |
941 | */ | |
942 | mutex_destroy(&inode->i_mutex); | |
943 | mutex_init(&inode->i_mutex); | |
944 | lockdep_set_class(&inode->i_mutex, | |
945 | &type->i_mutex_dir_key); | |
946 | } | |
1e89a5e1 | 947 | } |
e096d0c7 JB |
948 | } |
949 | EXPORT_SYMBOL(lockdep_annotate_inode_mutex_key); | |
14358e6d | 950 | #endif |
e096d0c7 JB |
951 | |
952 | /** | |
953 | * unlock_new_inode - clear the I_NEW state and wake up any waiters | |
954 | * @inode: new inode to unlock | |
955 | * | |
956 | * Called when the inode is fully initialised to clear the new state of the | |
957 | * inode and wake up anyone waiting for the inode to finish initialisation. | |
958 | */ | |
959 | void unlock_new_inode(struct inode *inode) | |
960 | { | |
961 | lockdep_annotate_inode_mutex_key(inode); | |
250df6ed | 962 | spin_lock(&inode->i_lock); |
eaff8079 CH |
963 | WARN_ON(!(inode->i_state & I_NEW)); |
964 | inode->i_state &= ~I_NEW; | |
310fa7a3 | 965 | smp_mb(); |
250df6ed DC |
966 | wake_up_bit(&inode->i_state, __I_NEW); |
967 | spin_unlock(&inode->i_lock); | |
1da177e4 | 968 | } |
1da177e4 LT |
969 | EXPORT_SYMBOL(unlock_new_inode); |
970 | ||
375e289e BF |
971 | /** |
972 | * lock_two_nondirectories - take two i_mutexes on non-directory objects | |
4fd699ae BF |
973 | * |
974 | * Lock any non-NULL argument that is not a directory. | |
975 | * Zero, one or two objects may be locked by this function. | |
976 | * | |
375e289e BF |
977 | * @inode1: first inode to lock |
978 | * @inode2: second inode to lock | |
979 | */ | |
980 | void lock_two_nondirectories(struct inode *inode1, struct inode *inode2) | |
981 | { | |
4fd699ae BF |
982 | if (inode1 > inode2) |
983 | swap(inode1, inode2); | |
984 | ||
985 | if (inode1 && !S_ISDIR(inode1->i_mode)) | |
27555516 | 986 | mutex_lock(&inode1->i_mutex); |
4fd699ae | 987 | if (inode2 && !S_ISDIR(inode2->i_mode) && inode2 != inode1) |
40bd22c9 | 988 | mutex_lock_nested(&inode2->i_mutex, I_MUTEX_NONDIR2); |
375e289e BF |
989 | } |
990 | EXPORT_SYMBOL(lock_two_nondirectories); | |
991 | ||
992 | /** | |
993 | * unlock_two_nondirectories - release locks from lock_two_nondirectories() | |
994 | * @inode1: first inode to unlock | |
995 | * @inode2: second inode to unlock | |
996 | */ | |
997 | void unlock_two_nondirectories(struct inode *inode1, struct inode *inode2) | |
998 | { | |
4fd699ae BF |
999 | if (inode1 && !S_ISDIR(inode1->i_mode)) |
1000 | mutex_unlock(&inode1->i_mutex); | |
1001 | if (inode2 && !S_ISDIR(inode2->i_mode) && inode2 != inode1) | |
375e289e BF |
1002 | mutex_unlock(&inode2->i_mutex); |
1003 | } | |
1004 | EXPORT_SYMBOL(unlock_two_nondirectories); | |
1005 | ||
0b2d0724 CH |
1006 | /** |
1007 | * iget5_locked - obtain an inode from a mounted file system | |
1008 | * @sb: super block of file system | |
1009 | * @hashval: hash value (usually inode number) to get | |
1010 | * @test: callback used for comparisons between inodes | |
1011 | * @set: callback used to initialize a new struct inode | |
1012 | * @data: opaque data pointer to pass to @test and @set | |
1013 | * | |
1014 | * Search for the inode specified by @hashval and @data in the inode cache, | |
1015 | * and if present it is return it with an increased reference count. This is | |
1016 | * a generalized version of iget_locked() for file systems where the inode | |
1017 | * number is not sufficient for unique identification of an inode. | |
1018 | * | |
1019 | * If the inode is not in cache, allocate a new inode and return it locked, | |
1020 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
1021 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 1022 | * |
0b2d0724 CH |
1023 | * Note both @test and @set are called with the inode_hash_lock held, so can't |
1024 | * sleep. | |
1da177e4 | 1025 | */ |
0b2d0724 CH |
1026 | struct inode *iget5_locked(struct super_block *sb, unsigned long hashval, |
1027 | int (*test)(struct inode *, void *), | |
1028 | int (*set)(struct inode *, void *), void *data) | |
1da177e4 | 1029 | { |
0b2d0724 | 1030 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); |
6b3304b5 | 1031 | struct inode *inode; |
1da177e4 | 1032 | |
0b2d0724 CH |
1033 | spin_lock(&inode_hash_lock); |
1034 | inode = find_inode(sb, head, test, data); | |
1035 | spin_unlock(&inode_hash_lock); | |
1036 | ||
1037 | if (inode) { | |
1038 | wait_on_inode(inode); | |
1039 | return inode; | |
1040 | } | |
1041 | ||
1da177e4 LT |
1042 | inode = alloc_inode(sb); |
1043 | if (inode) { | |
6b3304b5 | 1044 | struct inode *old; |
1da177e4 | 1045 | |
67a23c49 | 1046 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1047 | /* We released the lock, so.. */ |
1048 | old = find_inode(sb, head, test, data); | |
1049 | if (!old) { | |
1050 | if (set(inode, data)) | |
1051 | goto set_failed; | |
1052 | ||
250df6ed DC |
1053 | spin_lock(&inode->i_lock); |
1054 | inode->i_state = I_NEW; | |
646ec461 | 1055 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1056 | spin_unlock(&inode->i_lock); |
55fa6091 | 1057 | inode_sb_list_add(inode); |
67a23c49 | 1058 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1059 | |
1060 | /* Return the locked inode with I_NEW set, the | |
1061 | * caller is responsible for filling in the contents | |
1062 | */ | |
1063 | return inode; | |
1064 | } | |
1065 | ||
1066 | /* | |
1067 | * Uhhuh, somebody else created the same inode under | |
1068 | * us. Use the old inode instead of the one we just | |
1069 | * allocated. | |
1070 | */ | |
67a23c49 | 1071 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1072 | destroy_inode(inode); |
1073 | inode = old; | |
1074 | wait_on_inode(inode); | |
1075 | } | |
1076 | return inode; | |
1077 | ||
1078 | set_failed: | |
67a23c49 | 1079 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1080 | destroy_inode(inode); |
1081 | return NULL; | |
1082 | } | |
0b2d0724 | 1083 | EXPORT_SYMBOL(iget5_locked); |
1da177e4 | 1084 | |
0b2d0724 CH |
1085 | /** |
1086 | * iget_locked - obtain an inode from a mounted file system | |
1087 | * @sb: super block of file system | |
1088 | * @ino: inode number to get | |
1089 | * | |
1090 | * Search for the inode specified by @ino in the inode cache and if present | |
1091 | * return it with an increased reference count. This is for file systems | |
1092 | * where the inode number is sufficient for unique identification of an inode. | |
1093 | * | |
1094 | * If the inode is not in cache, allocate a new inode and return it locked, | |
1095 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
1096 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 1097 | */ |
0b2d0724 | 1098 | struct inode *iget_locked(struct super_block *sb, unsigned long ino) |
1da177e4 | 1099 | { |
0b2d0724 | 1100 | struct hlist_head *head = inode_hashtable + hash(sb, ino); |
6b3304b5 | 1101 | struct inode *inode; |
1da177e4 | 1102 | |
0b2d0724 CH |
1103 | spin_lock(&inode_hash_lock); |
1104 | inode = find_inode_fast(sb, head, ino); | |
1105 | spin_unlock(&inode_hash_lock); | |
1106 | if (inode) { | |
1107 | wait_on_inode(inode); | |
1108 | return inode; | |
1109 | } | |
1110 | ||
1da177e4 LT |
1111 | inode = alloc_inode(sb); |
1112 | if (inode) { | |
6b3304b5 | 1113 | struct inode *old; |
1da177e4 | 1114 | |
67a23c49 | 1115 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1116 | /* We released the lock, so.. */ |
1117 | old = find_inode_fast(sb, head, ino); | |
1118 | if (!old) { | |
1119 | inode->i_ino = ino; | |
250df6ed DC |
1120 | spin_lock(&inode->i_lock); |
1121 | inode->i_state = I_NEW; | |
646ec461 | 1122 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1123 | spin_unlock(&inode->i_lock); |
55fa6091 | 1124 | inode_sb_list_add(inode); |
67a23c49 | 1125 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1126 | |
1127 | /* Return the locked inode with I_NEW set, the | |
1128 | * caller is responsible for filling in the contents | |
1129 | */ | |
1130 | return inode; | |
1131 | } | |
1132 | ||
1133 | /* | |
1134 | * Uhhuh, somebody else created the same inode under | |
1135 | * us. Use the old inode instead of the one we just | |
1136 | * allocated. | |
1137 | */ | |
67a23c49 | 1138 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1139 | destroy_inode(inode); |
1140 | inode = old; | |
1141 | wait_on_inode(inode); | |
1142 | } | |
1143 | return inode; | |
1144 | } | |
0b2d0724 | 1145 | EXPORT_SYMBOL(iget_locked); |
1da177e4 | 1146 | |
ad5e195a CH |
1147 | /* |
1148 | * search the inode cache for a matching inode number. | |
1149 | * If we find one, then the inode number we are trying to | |
1150 | * allocate is not unique and so we should not use it. | |
1151 | * | |
1152 | * Returns 1 if the inode number is unique, 0 if it is not. | |
1153 | */ | |
1154 | static int test_inode_iunique(struct super_block *sb, unsigned long ino) | |
1155 | { | |
1156 | struct hlist_head *b = inode_hashtable + hash(sb, ino); | |
ad5e195a CH |
1157 | struct inode *inode; |
1158 | ||
67a23c49 | 1159 | spin_lock(&inode_hash_lock); |
b67bfe0d | 1160 | hlist_for_each_entry(inode, b, i_hash) { |
67a23c49 DC |
1161 | if (inode->i_ino == ino && inode->i_sb == sb) { |
1162 | spin_unlock(&inode_hash_lock); | |
ad5e195a | 1163 | return 0; |
67a23c49 | 1164 | } |
ad5e195a | 1165 | } |
67a23c49 | 1166 | spin_unlock(&inode_hash_lock); |
ad5e195a CH |
1167 | |
1168 | return 1; | |
1169 | } | |
1170 | ||
1da177e4 LT |
1171 | /** |
1172 | * iunique - get a unique inode number | |
1173 | * @sb: superblock | |
1174 | * @max_reserved: highest reserved inode number | |
1175 | * | |
1176 | * Obtain an inode number that is unique on the system for a given | |
1177 | * superblock. This is used by file systems that have no natural | |
1178 | * permanent inode numbering system. An inode number is returned that | |
1179 | * is higher than the reserved limit but unique. | |
1180 | * | |
1181 | * BUGS: | |
1182 | * With a large number of inodes live on the file system this function | |
1183 | * currently becomes quite slow. | |
1184 | */ | |
1185 | ino_t iunique(struct super_block *sb, ino_t max_reserved) | |
1186 | { | |
866b04fc JL |
1187 | /* |
1188 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
1189 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
1190 | * here to attempt to avoid that. | |
1191 | */ | |
ad5e195a | 1192 | static DEFINE_SPINLOCK(iunique_lock); |
866b04fc | 1193 | static unsigned int counter; |
1da177e4 | 1194 | ino_t res; |
3361c7be | 1195 | |
ad5e195a | 1196 | spin_lock(&iunique_lock); |
3361c7be JL |
1197 | do { |
1198 | if (counter <= max_reserved) | |
1199 | counter = max_reserved + 1; | |
1da177e4 | 1200 | res = counter++; |
ad5e195a CH |
1201 | } while (!test_inode_iunique(sb, res)); |
1202 | spin_unlock(&iunique_lock); | |
1da177e4 | 1203 | |
3361c7be JL |
1204 | return res; |
1205 | } | |
1da177e4 LT |
1206 | EXPORT_SYMBOL(iunique); |
1207 | ||
1208 | struct inode *igrab(struct inode *inode) | |
1209 | { | |
250df6ed DC |
1210 | spin_lock(&inode->i_lock); |
1211 | if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) { | |
1da177e4 | 1212 | __iget(inode); |
250df6ed DC |
1213 | spin_unlock(&inode->i_lock); |
1214 | } else { | |
1215 | spin_unlock(&inode->i_lock); | |
1da177e4 LT |
1216 | /* |
1217 | * Handle the case where s_op->clear_inode is not been | |
1218 | * called yet, and somebody is calling igrab | |
1219 | * while the inode is getting freed. | |
1220 | */ | |
1221 | inode = NULL; | |
250df6ed | 1222 | } |
1da177e4 LT |
1223 | return inode; |
1224 | } | |
1da177e4 LT |
1225 | EXPORT_SYMBOL(igrab); |
1226 | ||
1227 | /** | |
0b2d0724 | 1228 | * ilookup5_nowait - search for an inode in the inode cache |
1da177e4 | 1229 | * @sb: super block of file system to search |
0b2d0724 | 1230 | * @hashval: hash value (usually inode number) to search for |
1da177e4 LT |
1231 | * @test: callback used for comparisons between inodes |
1232 | * @data: opaque data pointer to pass to @test | |
1da177e4 | 1233 | * |
0b2d0724 | 1234 | * Search for the inode specified by @hashval and @data in the inode cache. |
1da177e4 LT |
1235 | * If the inode is in the cache, the inode is returned with an incremented |
1236 | * reference count. | |
1237 | * | |
0b2d0724 CH |
1238 | * Note: I_NEW is not waited upon so you have to be very careful what you do |
1239 | * with the returned inode. You probably should be using ilookup5() instead. | |
1da177e4 | 1240 | * |
b6d0ad68 | 1241 | * Note2: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 | 1242 | */ |
0b2d0724 CH |
1243 | struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval, |
1244 | int (*test)(struct inode *, void *), void *data) | |
1da177e4 | 1245 | { |
0b2d0724 | 1246 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); |
1da177e4 LT |
1247 | struct inode *inode; |
1248 | ||
67a23c49 | 1249 | spin_lock(&inode_hash_lock); |
1da177e4 | 1250 | inode = find_inode(sb, head, test, data); |
67a23c49 | 1251 | spin_unlock(&inode_hash_lock); |
88bd5121 | 1252 | |
0b2d0724 | 1253 | return inode; |
88bd5121 | 1254 | } |
88bd5121 AA |
1255 | EXPORT_SYMBOL(ilookup5_nowait); |
1256 | ||
1257 | /** | |
1258 | * ilookup5 - search for an inode in the inode cache | |
1259 | * @sb: super block of file system to search | |
1260 | * @hashval: hash value (usually inode number) to search for | |
1261 | * @test: callback used for comparisons between inodes | |
1262 | * @data: opaque data pointer to pass to @test | |
1263 | * | |
0b2d0724 CH |
1264 | * Search for the inode specified by @hashval and @data in the inode cache, |
1265 | * and if the inode is in the cache, return the inode with an incremented | |
1266 | * reference count. Waits on I_NEW before returning the inode. | |
88bd5121 | 1267 | * returned with an incremented reference count. |
1da177e4 | 1268 | * |
0b2d0724 CH |
1269 | * This is a generalized version of ilookup() for file systems where the |
1270 | * inode number is not sufficient for unique identification of an inode. | |
1da177e4 | 1271 | * |
0b2d0724 | 1272 | * Note: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 LT |
1273 | */ |
1274 | struct inode *ilookup5(struct super_block *sb, unsigned long hashval, | |
1275 | int (*test)(struct inode *, void *), void *data) | |
1276 | { | |
0b2d0724 | 1277 | struct inode *inode = ilookup5_nowait(sb, hashval, test, data); |
1da177e4 | 1278 | |
0b2d0724 CH |
1279 | if (inode) |
1280 | wait_on_inode(inode); | |
1281 | return inode; | |
1da177e4 | 1282 | } |
1da177e4 LT |
1283 | EXPORT_SYMBOL(ilookup5); |
1284 | ||
1285 | /** | |
1286 | * ilookup - search for an inode in the inode cache | |
1287 | * @sb: super block of file system to search | |
1288 | * @ino: inode number to search for | |
1289 | * | |
0b2d0724 CH |
1290 | * Search for the inode @ino in the inode cache, and if the inode is in the |
1291 | * cache, the inode is returned with an incremented reference count. | |
1da177e4 LT |
1292 | */ |
1293 | struct inode *ilookup(struct super_block *sb, unsigned long ino) | |
1294 | { | |
1295 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1da177e4 LT |
1296 | struct inode *inode; |
1297 | ||
0b2d0724 CH |
1298 | spin_lock(&inode_hash_lock); |
1299 | inode = find_inode_fast(sb, head, ino); | |
1300 | spin_unlock(&inode_hash_lock); | |
1da177e4 | 1301 | |
1da177e4 | 1302 | if (inode) |
0b2d0724 CH |
1303 | wait_on_inode(inode); |
1304 | return inode; | |
1da177e4 | 1305 | } |
0b2d0724 | 1306 | EXPORT_SYMBOL(ilookup); |
1da177e4 | 1307 | |
fe032c42 TT |
1308 | /** |
1309 | * find_inode_nowait - find an inode in the inode cache | |
1310 | * @sb: super block of file system to search | |
1311 | * @hashval: hash value (usually inode number) to search for | |
1312 | * @match: callback used for comparisons between inodes | |
1313 | * @data: opaque data pointer to pass to @match | |
1314 | * | |
1315 | * Search for the inode specified by @hashval and @data in the inode | |
1316 | * cache, where the helper function @match will return 0 if the inode | |
1317 | * does not match, 1 if the inode does match, and -1 if the search | |
1318 | * should be stopped. The @match function must be responsible for | |
1319 | * taking the i_lock spin_lock and checking i_state for an inode being | |
1320 | * freed or being initialized, and incrementing the reference count | |
1321 | * before returning 1. It also must not sleep, since it is called with | |
1322 | * the inode_hash_lock spinlock held. | |
1323 | * | |
1324 | * This is a even more generalized version of ilookup5() when the | |
1325 | * function must never block --- find_inode() can block in | |
1326 | * __wait_on_freeing_inode() --- or when the caller can not increment | |
1327 | * the reference count because the resulting iput() might cause an | |
1328 | * inode eviction. The tradeoff is that the @match funtion must be | |
1329 | * very carefully implemented. | |
1330 | */ | |
1331 | struct inode *find_inode_nowait(struct super_block *sb, | |
1332 | unsigned long hashval, | |
1333 | int (*match)(struct inode *, unsigned long, | |
1334 | void *), | |
1335 | void *data) | |
1336 | { | |
1337 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1338 | struct inode *inode, *ret_inode = NULL; | |
1339 | int mval; | |
1340 | ||
1341 | spin_lock(&inode_hash_lock); | |
1342 | hlist_for_each_entry(inode, head, i_hash) { | |
1343 | if (inode->i_sb != sb) | |
1344 | continue; | |
1345 | mval = match(inode, hashval, data); | |
1346 | if (mval == 0) | |
1347 | continue; | |
1348 | if (mval == 1) | |
1349 | ret_inode = inode; | |
1350 | goto out; | |
1351 | } | |
1352 | out: | |
1353 | spin_unlock(&inode_hash_lock); | |
1354 | return ret_inode; | |
1355 | } | |
1356 | EXPORT_SYMBOL(find_inode_nowait); | |
1357 | ||
261bca86 AV |
1358 | int insert_inode_locked(struct inode *inode) |
1359 | { | |
1360 | struct super_block *sb = inode->i_sb; | |
1361 | ino_t ino = inode->i_ino; | |
1362 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
261bca86 | 1363 | |
261bca86 | 1364 | while (1) { |
72a43d63 | 1365 | struct inode *old = NULL; |
67a23c49 | 1366 | spin_lock(&inode_hash_lock); |
b67bfe0d | 1367 | hlist_for_each_entry(old, head, i_hash) { |
72a43d63 AV |
1368 | if (old->i_ino != ino) |
1369 | continue; | |
1370 | if (old->i_sb != sb) | |
1371 | continue; | |
250df6ed DC |
1372 | spin_lock(&old->i_lock); |
1373 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1374 | spin_unlock(&old->i_lock); | |
72a43d63 | 1375 | continue; |
250df6ed | 1376 | } |
72a43d63 AV |
1377 | break; |
1378 | } | |
b67bfe0d | 1379 | if (likely(!old)) { |
250df6ed DC |
1380 | spin_lock(&inode->i_lock); |
1381 | inode->i_state |= I_NEW; | |
261bca86 | 1382 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1383 | spin_unlock(&inode->i_lock); |
67a23c49 | 1384 | spin_unlock(&inode_hash_lock); |
261bca86 AV |
1385 | return 0; |
1386 | } | |
1387 | __iget(old); | |
250df6ed | 1388 | spin_unlock(&old->i_lock); |
67a23c49 | 1389 | spin_unlock(&inode_hash_lock); |
261bca86 | 1390 | wait_on_inode(old); |
1d3382cb | 1391 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1392 | iput(old); |
1393 | return -EBUSY; | |
1394 | } | |
1395 | iput(old); | |
1396 | } | |
1397 | } | |
261bca86 AV |
1398 | EXPORT_SYMBOL(insert_inode_locked); |
1399 | ||
1400 | int insert_inode_locked4(struct inode *inode, unsigned long hashval, | |
1401 | int (*test)(struct inode *, void *), void *data) | |
1402 | { | |
1403 | struct super_block *sb = inode->i_sb; | |
1404 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
261bca86 | 1405 | |
261bca86 | 1406 | while (1) { |
72a43d63 AV |
1407 | struct inode *old = NULL; |
1408 | ||
67a23c49 | 1409 | spin_lock(&inode_hash_lock); |
b67bfe0d | 1410 | hlist_for_each_entry(old, head, i_hash) { |
72a43d63 AV |
1411 | if (old->i_sb != sb) |
1412 | continue; | |
1413 | if (!test(old, data)) | |
1414 | continue; | |
250df6ed DC |
1415 | spin_lock(&old->i_lock); |
1416 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1417 | spin_unlock(&old->i_lock); | |
72a43d63 | 1418 | continue; |
250df6ed | 1419 | } |
72a43d63 AV |
1420 | break; |
1421 | } | |
b67bfe0d | 1422 | if (likely(!old)) { |
250df6ed DC |
1423 | spin_lock(&inode->i_lock); |
1424 | inode->i_state |= I_NEW; | |
261bca86 | 1425 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1426 | spin_unlock(&inode->i_lock); |
67a23c49 | 1427 | spin_unlock(&inode_hash_lock); |
261bca86 AV |
1428 | return 0; |
1429 | } | |
1430 | __iget(old); | |
250df6ed | 1431 | spin_unlock(&old->i_lock); |
67a23c49 | 1432 | spin_unlock(&inode_hash_lock); |
261bca86 | 1433 | wait_on_inode(old); |
1d3382cb | 1434 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1435 | iput(old); |
1436 | return -EBUSY; | |
1437 | } | |
1438 | iput(old); | |
1439 | } | |
1440 | } | |
261bca86 AV |
1441 | EXPORT_SYMBOL(insert_inode_locked4); |
1442 | ||
1da177e4 | 1443 | |
45321ac5 AV |
1444 | int generic_delete_inode(struct inode *inode) |
1445 | { | |
1446 | return 1; | |
1447 | } | |
1448 | EXPORT_SYMBOL(generic_delete_inode); | |
1449 | ||
45321ac5 AV |
1450 | /* |
1451 | * Called when we're dropping the last reference | |
1452 | * to an inode. | |
22fe4042 | 1453 | * |
45321ac5 AV |
1454 | * Call the FS "drop_inode()" function, defaulting to |
1455 | * the legacy UNIX filesystem behaviour. If it tells | |
1456 | * us to evict inode, do so. Otherwise, retain inode | |
1457 | * in cache if fs is alive, sync and evict if fs is | |
1458 | * shutting down. | |
22fe4042 | 1459 | */ |
45321ac5 | 1460 | static void iput_final(struct inode *inode) |
1da177e4 LT |
1461 | { |
1462 | struct super_block *sb = inode->i_sb; | |
45321ac5 AV |
1463 | const struct super_operations *op = inode->i_sb->s_op; |
1464 | int drop; | |
1465 | ||
250df6ed DC |
1466 | WARN_ON(inode->i_state & I_NEW); |
1467 | ||
e7f59097 | 1468 | if (op->drop_inode) |
45321ac5 AV |
1469 | drop = op->drop_inode(inode); |
1470 | else | |
1471 | drop = generic_drop_inode(inode); | |
1da177e4 | 1472 | |
1cac41cb MB |
1473 | #if defined(CONFIG_FMP_ECRYPT_FS) |
1474 | if (!drop && (sb->s_flags & MS_ACTIVE) && !inode->i_mapping->use_fmp) { | |
1475 | #else | |
b2b2af8e | 1476 | if (!drop && (sb->s_flags & MS_ACTIVE)) { |
1cac41cb | 1477 | #endif |
b2b2af8e | 1478 | inode->i_state |= I_REFERENCED; |
4eff96dd | 1479 | inode_add_lru(inode); |
b2b2af8e | 1480 | spin_unlock(&inode->i_lock); |
b2b2af8e DC |
1481 | return; |
1482 | } | |
1483 | ||
45321ac5 | 1484 | if (!drop) { |
991114c6 | 1485 | inode->i_state |= I_WILL_FREE; |
250df6ed | 1486 | spin_unlock(&inode->i_lock); |
1da177e4 | 1487 | write_inode_now(inode, 1); |
250df6ed | 1488 | spin_lock(&inode->i_lock); |
7ef0d737 | 1489 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1490 | inode->i_state &= ~I_WILL_FREE; |
1da177e4 | 1491 | } |
7ccf19a8 | 1492 | |
991114c6 | 1493 | inode->i_state |= I_FREEING; |
c4ae0c65 ED |
1494 | if (!list_empty(&inode->i_lru)) |
1495 | inode_lru_list_del(inode); | |
b2b2af8e | 1496 | spin_unlock(&inode->i_lock); |
b2b2af8e | 1497 | |
644da596 | 1498 | evict(inode); |
1da177e4 LT |
1499 | } |
1500 | ||
1da177e4 | 1501 | /** |
6b3304b5 | 1502 | * iput - put an inode |
1da177e4 LT |
1503 | * @inode: inode to put |
1504 | * | |
1505 | * Puts an inode, dropping its usage count. If the inode use count hits | |
1506 | * zero, the inode is then freed and may also be destroyed. | |
1507 | * | |
1508 | * Consequently, iput() can sleep. | |
1509 | */ | |
1510 | void iput(struct inode *inode) | |
1511 | { | |
0ae45f63 TT |
1512 | if (!inode) |
1513 | return; | |
1514 | BUG_ON(inode->i_state & I_CLEAR); | |
1515 | retry: | |
1516 | if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock)) { | |
1517 | if (inode->i_nlink && (inode->i_state & I_DIRTY_TIME)) { | |
1518 | atomic_inc(&inode->i_count); | |
1519 | inode->i_state &= ~I_DIRTY_TIME; | |
1520 | spin_unlock(&inode->i_lock); | |
1521 | trace_writeback_lazytime_iput(inode); | |
1522 | mark_inode_dirty_sync(inode); | |
1523 | goto retry; | |
1524 | } | |
1525 | iput_final(inode); | |
1da177e4 LT |
1526 | } |
1527 | } | |
1da177e4 LT |
1528 | EXPORT_SYMBOL(iput); |
1529 | ||
1530 | /** | |
1531 | * bmap - find a block number in a file | |
1532 | * @inode: inode of file | |
1533 | * @block: block to find | |
1534 | * | |
1535 | * Returns the block number on the device holding the inode that | |
1536 | * is the disk block number for the block of the file requested. | |
1537 | * That is, asked for block 4 of inode 1 the function will return the | |
6b3304b5 | 1538 | * disk block relative to the disk start that holds that block of the |
1da177e4 LT |
1539 | * file. |
1540 | */ | |
6b3304b5 | 1541 | sector_t bmap(struct inode *inode, sector_t block) |
1da177e4 LT |
1542 | { |
1543 | sector_t res = 0; | |
1544 | if (inode->i_mapping->a_ops->bmap) | |
1545 | res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block); | |
1546 | return res; | |
1547 | } | |
1da177e4 LT |
1548 | EXPORT_SYMBOL(bmap); |
1549 | ||
11ff6f05 MG |
1550 | /* |
1551 | * With relative atime, only update atime if the previous atime is | |
1552 | * earlier than either the ctime or mtime or if at least a day has | |
1553 | * passed since the last atime update. | |
1554 | */ | |
1555 | static int relatime_need_update(struct vfsmount *mnt, struct inode *inode, | |
1556 | struct timespec now) | |
1557 | { | |
1558 | ||
1559 | if (!(mnt->mnt_flags & MNT_RELATIME)) | |
1560 | return 1; | |
1561 | /* | |
1562 | * Is mtime younger than atime? If yes, update atime: | |
1563 | */ | |
1564 | if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0) | |
1565 | return 1; | |
1566 | /* | |
1567 | * Is ctime younger than atime? If yes, update atime: | |
1568 | */ | |
1569 | if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0) | |
1570 | return 1; | |
1571 | ||
1572 | /* | |
1573 | * Is the previous atime value older than a day? If yes, | |
1574 | * update atime: | |
1575 | */ | |
1576 | if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60) | |
1577 | return 1; | |
1578 | /* | |
1579 | * Good, we can skip the atime update: | |
1580 | */ | |
1581 | return 0; | |
1582 | } | |
1583 | ||
0ae45f63 | 1584 | int generic_update_time(struct inode *inode, struct timespec *time, int flags) |
c3b2da31 | 1585 | { |
0ae45f63 | 1586 | int iflags = I_DIRTY_TIME; |
c3b2da31 JB |
1587 | |
1588 | if (flags & S_ATIME) | |
1589 | inode->i_atime = *time; | |
1590 | if (flags & S_VERSION) | |
1591 | inode_inc_iversion(inode); | |
1592 | if (flags & S_CTIME) | |
1593 | inode->i_ctime = *time; | |
1594 | if (flags & S_MTIME) | |
1595 | inode->i_mtime = *time; | |
0ae45f63 TT |
1596 | |
1597 | if (!(inode->i_sb->s_flags & MS_LAZYTIME) || (flags & S_VERSION)) | |
1598 | iflags |= I_DIRTY_SYNC; | |
1599 | __mark_inode_dirty(inode, iflags); | |
c3b2da31 JB |
1600 | return 0; |
1601 | } | |
0ae45f63 TT |
1602 | EXPORT_SYMBOL(generic_update_time); |
1603 | ||
1604 | /* | |
1605 | * This does the actual work of updating an inodes time or version. Must have | |
1606 | * had called mnt_want_write() before calling this. | |
1607 | */ | |
1608 | static int update_time(struct inode *inode, struct timespec *time, int flags) | |
1609 | { | |
1610 | int (*update_time)(struct inode *, struct timespec *, int); | |
1611 | ||
1612 | update_time = inode->i_op->update_time ? inode->i_op->update_time : | |
1613 | generic_update_time; | |
1614 | ||
1615 | return update_time(inode, time, flags); | |
1616 | } | |
c3b2da31 | 1617 | |
1da177e4 | 1618 | /** |
869243a0 | 1619 | * touch_atime - update the access time |
185553b2 | 1620 | * @path: the &struct path to update |
30fdc8ee | 1621 | * @inode: inode to update |
1da177e4 LT |
1622 | * |
1623 | * Update the accessed time on an inode and mark it for writeback. | |
1624 | * This function automatically handles read only file systems and media, | |
1625 | * as well as the "noatime" flag and inode specific "noatime" markers. | |
1626 | */ | |
8fa9dd24 | 1627 | bool atime_needs_update(const struct path *path, struct inode *inode) |
1da177e4 | 1628 | { |
68ac1234 | 1629 | struct vfsmount *mnt = path->mnt; |
1da177e4 LT |
1630 | struct timespec now; |
1631 | ||
cdb70f3f | 1632 | if (inode->i_flags & S_NOATIME) |
8fa9dd24 | 1633 | return false; |
37756ced | 1634 | if (IS_NOATIME(inode)) |
8fa9dd24 | 1635 | return false; |
b2276138 | 1636 | if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)) |
8fa9dd24 | 1637 | return false; |
47ae32d6 | 1638 | |
cdb70f3f | 1639 | if (mnt->mnt_flags & MNT_NOATIME) |
8fa9dd24 | 1640 | return false; |
cdb70f3f | 1641 | if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) |
8fa9dd24 | 1642 | return false; |
1da177e4 LT |
1643 | |
1644 | now = current_fs_time(inode->i_sb); | |
11ff6f05 MG |
1645 | |
1646 | if (!relatime_need_update(mnt, inode, now)) | |
8fa9dd24 | 1647 | return false; |
11ff6f05 | 1648 | |
47ae32d6 | 1649 | if (timespec_equal(&inode->i_atime, &now)) |
8fa9dd24 N |
1650 | return false; |
1651 | ||
1652 | return true; | |
1653 | } | |
1654 | ||
1655 | void touch_atime(const struct path *path) | |
1656 | { | |
1657 | struct vfsmount *mnt = path->mnt; | |
1658 | struct inode *inode = d_inode(path->dentry); | |
1659 | struct timespec now; | |
1660 | ||
1661 | if (!atime_needs_update(path, inode)) | |
b12536c2 AK |
1662 | return; |
1663 | ||
5d37e9e6 | 1664 | if (!sb_start_write_trylock(inode->i_sb)) |
b12536c2 | 1665 | return; |
47ae32d6 | 1666 | |
8fa9dd24 | 1667 | if (__mnt_want_write(mnt) != 0) |
5d37e9e6 | 1668 | goto skip_update; |
c3b2da31 JB |
1669 | /* |
1670 | * File systems can error out when updating inodes if they need to | |
1671 | * allocate new space to modify an inode (such is the case for | |
1672 | * Btrfs), but since we touch atime while walking down the path we | |
1673 | * really don't care if we failed to update the atime of the file, | |
1674 | * so just ignore the return value. | |
2bc55652 AB |
1675 | * We may also fail on filesystems that have the ability to make parts |
1676 | * of the fs read only, e.g. subvolumes in Btrfs. | |
c3b2da31 | 1677 | */ |
8fa9dd24 | 1678 | now = current_fs_time(inode->i_sb); |
c3b2da31 | 1679 | update_time(inode, &now, S_ATIME); |
5d37e9e6 JK |
1680 | __mnt_drop_write(mnt); |
1681 | skip_update: | |
1682 | sb_end_write(inode->i_sb); | |
1da177e4 | 1683 | } |
869243a0 | 1684 | EXPORT_SYMBOL(touch_atime); |
1da177e4 | 1685 | |
3ed37648 CW |
1686 | /* |
1687 | * The logic we want is | |
1688 | * | |
1689 | * if suid or (sgid and xgrp) | |
1690 | * remove privs | |
1691 | */ | |
1692 | int should_remove_suid(struct dentry *dentry) | |
1693 | { | |
df2b1afd | 1694 | umode_t mode = d_inode(dentry)->i_mode; |
3ed37648 CW |
1695 | int kill = 0; |
1696 | ||
1697 | /* suid always must be killed */ | |
1698 | if (unlikely(mode & S_ISUID)) | |
1699 | kill = ATTR_KILL_SUID; | |
1700 | ||
1701 | /* | |
1702 | * sgid without any exec bits is just a mandatory locking mark; leave | |
1703 | * it alone. If some exec bits are set, it's a real sgid; kill it. | |
1704 | */ | |
1705 | if (unlikely((mode & S_ISGID) && (mode & S_IXGRP))) | |
1706 | kill |= ATTR_KILL_SGID; | |
1707 | ||
1708 | if (unlikely(kill && !capable(CAP_FSETID) && S_ISREG(mode))) | |
1709 | return kill; | |
1710 | ||
1711 | return 0; | |
1712 | } | |
1713 | EXPORT_SYMBOL(should_remove_suid); | |
1714 | ||
dbfae0cd JK |
1715 | /* |
1716 | * Return mask of changes for notify_change() that need to be done as a | |
1717 | * response to write or truncate. Return 0 if nothing has to be changed. | |
1718 | * Negative value on error (change should be denied). | |
1719 | */ | |
45f147a1 | 1720 | int dentry_needs_remove_privs(struct dentry *dentry) |
dbfae0cd | 1721 | { |
dbfae0cd JK |
1722 | struct inode *inode = d_inode(dentry); |
1723 | int mask = 0; | |
1724 | int ret; | |
1725 | ||
1726 | if (IS_NOSEC(inode)) | |
1727 | return 0; | |
1728 | ||
1729 | mask = should_remove_suid(dentry); | |
1730 | ret = security_inode_need_killpriv(dentry); | |
1731 | if (ret < 0) | |
1732 | return ret; | |
1733 | if (ret) | |
1734 | mask |= ATTR_KILL_PRIV; | |
1735 | return mask; | |
1736 | } | |
45f147a1 | 1737 | EXPORT_SYMBOL(dentry_needs_remove_privs); |
dbfae0cd | 1738 | |
1cbf8e31 | 1739 | static int __remove_privs(struct vfsmount *mnt, struct dentry *dentry, int kill) |
3ed37648 CW |
1740 | { |
1741 | struct iattr newattrs; | |
1742 | ||
1743 | newattrs.ia_valid = ATTR_FORCE | kill; | |
27ac0ffe BF |
1744 | /* |
1745 | * Note we call this on write, so notify_change will not | |
1746 | * encounter any conflicting delegations: | |
1747 | */ | |
1cbf8e31 | 1748 | return notify_change2(mnt, dentry, &newattrs, NULL); |
3ed37648 CW |
1749 | } |
1750 | ||
5fa8e0a1 JK |
1751 | /* |
1752 | * Remove special file priviledges (suid, capabilities) when file is written | |
1753 | * to or truncated. | |
1754 | */ | |
1755 | int file_remove_privs(struct file *file) | |
3ed37648 | 1756 | { |
8e510cd9 MS |
1757 | struct dentry *dentry = file_dentry(file); |
1758 | struct inode *inode = file_inode(file); | |
dbfae0cd | 1759 | int kill; |
3ed37648 CW |
1760 | int error = 0; |
1761 | ||
1762 | /* Fast path for nothing security related */ | |
1763 | if (IS_NOSEC(inode)) | |
1764 | return 0; | |
1765 | ||
8e510cd9 | 1766 | kill = dentry_needs_remove_privs(dentry); |
dbfae0cd JK |
1767 | if (kill < 0) |
1768 | return kill; | |
1769 | if (kill) | |
1cbf8e31 | 1770 | error = __remove_privs(file->f_path.mnt, dentry, kill); |
2426f391 JK |
1771 | if (!error) |
1772 | inode_has_no_xattr(inode); | |
3ed37648 CW |
1773 | |
1774 | return error; | |
1775 | } | |
5fa8e0a1 | 1776 | EXPORT_SYMBOL(file_remove_privs); |
3ed37648 | 1777 | |
1da177e4 | 1778 | /** |
870f4817 CH |
1779 | * file_update_time - update mtime and ctime time |
1780 | * @file: file accessed | |
1da177e4 | 1781 | * |
870f4817 CH |
1782 | * Update the mtime and ctime members of an inode and mark the inode |
1783 | * for writeback. Note that this function is meant exclusively for | |
1784 | * usage in the file write path of filesystems, and filesystems may | |
1785 | * choose to explicitly ignore update via this function with the | |
2eadfc0e | 1786 | * S_NOCMTIME inode flag, e.g. for network filesystem where these |
c3b2da31 JB |
1787 | * timestamps are handled by the server. This can return an error for |
1788 | * file systems who need to allocate space in order to update an inode. | |
1da177e4 LT |
1789 | */ |
1790 | ||
c3b2da31 | 1791 | int file_update_time(struct file *file) |
1da177e4 | 1792 | { |
496ad9aa | 1793 | struct inode *inode = file_inode(file); |
1da177e4 | 1794 | struct timespec now; |
c3b2da31 JB |
1795 | int sync_it = 0; |
1796 | int ret; | |
1da177e4 | 1797 | |
ce06e0b2 | 1798 | /* First try to exhaust all avenues to not sync */ |
1da177e4 | 1799 | if (IS_NOCMTIME(inode)) |
c3b2da31 | 1800 | return 0; |
20ddee2c | 1801 | |
1da177e4 | 1802 | now = current_fs_time(inode->i_sb); |
ce06e0b2 AK |
1803 | if (!timespec_equal(&inode->i_mtime, &now)) |
1804 | sync_it = S_MTIME; | |
1da177e4 | 1805 | |
ce06e0b2 AK |
1806 | if (!timespec_equal(&inode->i_ctime, &now)) |
1807 | sync_it |= S_CTIME; | |
870f4817 | 1808 | |
ce06e0b2 AK |
1809 | if (IS_I_VERSION(inode)) |
1810 | sync_it |= S_VERSION; | |
7a224228 | 1811 | |
ce06e0b2 | 1812 | if (!sync_it) |
c3b2da31 | 1813 | return 0; |
ce06e0b2 AK |
1814 | |
1815 | /* Finally allowed to write? Takes lock. */ | |
eb04c282 | 1816 | if (__mnt_want_write_file(file)) |
c3b2da31 | 1817 | return 0; |
ce06e0b2 | 1818 | |
c3b2da31 | 1819 | ret = update_time(inode, &now, sync_it); |
eb04c282 | 1820 | __mnt_drop_write_file(file); |
c3b2da31 JB |
1821 | |
1822 | return ret; | |
1da177e4 | 1823 | } |
870f4817 | 1824 | EXPORT_SYMBOL(file_update_time); |
1da177e4 LT |
1825 | |
1826 | int inode_needs_sync(struct inode *inode) | |
1827 | { | |
1828 | if (IS_SYNC(inode)) | |
1829 | return 1; | |
1830 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) | |
1831 | return 1; | |
1832 | return 0; | |
1833 | } | |
1da177e4 LT |
1834 | EXPORT_SYMBOL(inode_needs_sync); |
1835 | ||
1da177e4 | 1836 | /* |
168a9fd6 MS |
1837 | * If we try to find an inode in the inode hash while it is being |
1838 | * deleted, we have to wait until the filesystem completes its | |
1839 | * deletion before reporting that it isn't found. This function waits | |
1840 | * until the deletion _might_ have completed. Callers are responsible | |
1841 | * to recheck inode state. | |
1842 | * | |
eaff8079 | 1843 | * It doesn't matter if I_NEW is not set initially, a call to |
250df6ed DC |
1844 | * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list |
1845 | * will DTRT. | |
1da177e4 LT |
1846 | */ |
1847 | static void __wait_on_freeing_inode(struct inode *inode) | |
1848 | { | |
1849 | wait_queue_head_t *wq; | |
eaff8079 CH |
1850 | DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW); |
1851 | wq = bit_waitqueue(&inode->i_state, __I_NEW); | |
1da177e4 | 1852 | prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE); |
250df6ed | 1853 | spin_unlock(&inode->i_lock); |
67a23c49 | 1854 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1855 | schedule(); |
1856 | finish_wait(wq, &wait.wait); | |
67a23c49 | 1857 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1858 | } |
1859 | ||
1da177e4 LT |
1860 | static __initdata unsigned long ihash_entries; |
1861 | static int __init set_ihash_entries(char *str) | |
1862 | { | |
1863 | if (!str) | |
1864 | return 0; | |
1865 | ihash_entries = simple_strtoul(str, &str, 0); | |
1866 | return 1; | |
1867 | } | |
1868 | __setup("ihash_entries=", set_ihash_entries); | |
1869 | ||
1870 | /* | |
1871 | * Initialize the waitqueues and inode hash table. | |
1872 | */ | |
1873 | void __init inode_init_early(void) | |
1874 | { | |
074b8517 | 1875 | unsigned int loop; |
1da177e4 LT |
1876 | |
1877 | /* If hashes are distributed across NUMA nodes, defer | |
1878 | * hash allocation until vmalloc space is available. | |
1879 | */ | |
1880 | if (hashdist) | |
1881 | return; | |
1882 | ||
1883 | inode_hashtable = | |
1884 | alloc_large_system_hash("Inode-cache", | |
1885 | sizeof(struct hlist_head), | |
1886 | ihash_entries, | |
1887 | 14, | |
1888 | HASH_EARLY, | |
1889 | &i_hash_shift, | |
1890 | &i_hash_mask, | |
31fe62b9 | 1891 | 0, |
1da177e4 LT |
1892 | 0); |
1893 | ||
074b8517 | 1894 | for (loop = 0; loop < (1U << i_hash_shift); loop++) |
1da177e4 LT |
1895 | INIT_HLIST_HEAD(&inode_hashtable[loop]); |
1896 | } | |
1897 | ||
74bf17cf | 1898 | void __init inode_init(void) |
1da177e4 | 1899 | { |
074b8517 | 1900 | unsigned int loop; |
1da177e4 LT |
1901 | |
1902 | /* inode slab cache */ | |
b0196009 PJ |
1903 | inode_cachep = kmem_cache_create("inode_cache", |
1904 | sizeof(struct inode), | |
1905 | 0, | |
1906 | (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC| | |
1907 | SLAB_MEM_SPREAD), | |
20c2df83 | 1908 | init_once); |
1da177e4 LT |
1909 | |
1910 | /* Hash may have been set up in inode_init_early */ | |
1911 | if (!hashdist) | |
1912 | return; | |
1913 | ||
1914 | inode_hashtable = | |
1915 | alloc_large_system_hash("Inode-cache", | |
1916 | sizeof(struct hlist_head), | |
1917 | ihash_entries, | |
1918 | 14, | |
1919 | 0, | |
1920 | &i_hash_shift, | |
1921 | &i_hash_mask, | |
31fe62b9 | 1922 | 0, |
1da177e4 LT |
1923 | 0); |
1924 | ||
074b8517 | 1925 | for (loop = 0; loop < (1U << i_hash_shift); loop++) |
1da177e4 LT |
1926 | INIT_HLIST_HEAD(&inode_hashtable[loop]); |
1927 | } | |
1928 | ||
1929 | void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev) | |
1930 | { | |
1931 | inode->i_mode = mode; | |
1932 | if (S_ISCHR(mode)) { | |
1933 | inode->i_fop = &def_chr_fops; | |
1934 | inode->i_rdev = rdev; | |
1935 | } else if (S_ISBLK(mode)) { | |
1936 | inode->i_fop = &def_blk_fops; | |
1937 | inode->i_rdev = rdev; | |
1938 | } else if (S_ISFIFO(mode)) | |
599a0ac1 | 1939 | inode->i_fop = &pipefifo_fops; |
1da177e4 | 1940 | else if (S_ISSOCK(mode)) |
bd9b51e7 | 1941 | ; /* leave it no_open_fops */ |
1da177e4 | 1942 | else |
af0d9ae8 MK |
1943 | printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for" |
1944 | " inode %s:%lu\n", mode, inode->i_sb->s_id, | |
1945 | inode->i_ino); | |
1da177e4 LT |
1946 | } |
1947 | EXPORT_SYMBOL(init_special_inode); | |
a1bd120d DM |
1948 | |
1949 | /** | |
eaae668d | 1950 | * inode_init_owner - Init uid,gid,mode for new inode according to posix standards |
a1bd120d DM |
1951 | * @inode: New inode |
1952 | * @dir: Directory inode | |
1953 | * @mode: mode of the new inode | |
1954 | */ | |
1955 | void inode_init_owner(struct inode *inode, const struct inode *dir, | |
62bb1091 | 1956 | umode_t mode) |
a1bd120d DM |
1957 | { |
1958 | inode->i_uid = current_fsuid(); | |
1959 | if (dir && dir->i_mode & S_ISGID) { | |
1960 | inode->i_gid = dir->i_gid; | |
1cac41cb MB |
1961 | |
1962 | /* Directories are special, and always inherit S_ISGID */ | |
a1bd120d DM |
1963 | if (S_ISDIR(mode)) |
1964 | mode |= S_ISGID; | |
1cac41cb MB |
1965 | else if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP) && |
1966 | !in_group_p(inode->i_gid) && | |
1967 | !capable_wrt_inode_uidgid(dir, CAP_FSETID)) | |
1968 | mode &= ~S_ISGID; | |
a1bd120d DM |
1969 | } else |
1970 | inode->i_gid = current_fsgid(); | |
1971 | inode->i_mode = mode; | |
1972 | } | |
1973 | EXPORT_SYMBOL(inode_init_owner); | |
e795b717 | 1974 | |
2e149670 SH |
1975 | /** |
1976 | * inode_owner_or_capable - check current task permissions to inode | |
1977 | * @inode: inode being checked | |
1978 | * | |
23adbe12 AL |
1979 | * Return true if current either has CAP_FOWNER in a namespace with the |
1980 | * inode owner uid mapped, or owns the file. | |
e795b717 | 1981 | */ |
2e149670 | 1982 | bool inode_owner_or_capable(const struct inode *inode) |
e795b717 | 1983 | { |
23adbe12 AL |
1984 | struct user_namespace *ns; |
1985 | ||
92361636 | 1986 | if (uid_eq(current_fsuid(), inode->i_uid)) |
e795b717 | 1987 | return true; |
23adbe12 AL |
1988 | |
1989 | ns = current_user_ns(); | |
1990 | if (ns_capable(ns, CAP_FOWNER) && kuid_has_mapping(ns, inode->i_uid)) | |
e795b717 SH |
1991 | return true; |
1992 | return false; | |
1993 | } | |
2e149670 | 1994 | EXPORT_SYMBOL(inode_owner_or_capable); |
1d59d61f TM |
1995 | |
1996 | /* | |
1997 | * Direct i/o helper functions | |
1998 | */ | |
1999 | static void __inode_dio_wait(struct inode *inode) | |
2000 | { | |
2001 | wait_queue_head_t *wq = bit_waitqueue(&inode->i_state, __I_DIO_WAKEUP); | |
2002 | DEFINE_WAIT_BIT(q, &inode->i_state, __I_DIO_WAKEUP); | |
2003 | ||
2004 | do { | |
2005 | prepare_to_wait(wq, &q.wait, TASK_UNINTERRUPTIBLE); | |
2006 | if (atomic_read(&inode->i_dio_count)) | |
2007 | schedule(); | |
2008 | } while (atomic_read(&inode->i_dio_count)); | |
2009 | finish_wait(wq, &q.wait); | |
2010 | } | |
2011 | ||
2012 | /** | |
2013 | * inode_dio_wait - wait for outstanding DIO requests to finish | |
2014 | * @inode: inode to wait for | |
2015 | * | |
2016 | * Waits for all pending direct I/O requests to finish so that we can | |
2017 | * proceed with a truncate or equivalent operation. | |
2018 | * | |
2019 | * Must be called under a lock that serializes taking new references | |
2020 | * to i_dio_count, usually by inode->i_mutex. | |
2021 | */ | |
2022 | void inode_dio_wait(struct inode *inode) | |
2023 | { | |
2024 | if (atomic_read(&inode->i_dio_count)) | |
2025 | __inode_dio_wait(inode); | |
2026 | } | |
2027 | EXPORT_SYMBOL(inode_dio_wait); | |
2028 | ||
5f16f322 TT |
2029 | /* |
2030 | * inode_set_flags - atomically set some inode flags | |
2031 | * | |
2032 | * Note: the caller should be holding i_mutex, or else be sure that | |
2033 | * they have exclusive access to the inode structure (i.e., while the | |
2034 | * inode is being instantiated). The reason for the cmpxchg() loop | |
2035 | * --- which wouldn't be necessary if all code paths which modify | |
2036 | * i_flags actually followed this rule, is that there is at least one | |
5fa8e0a1 JK |
2037 | * code path which doesn't today so we use cmpxchg() out of an abundance |
2038 | * of caution. | |
5f16f322 TT |
2039 | * |
2040 | * In the long run, i_mutex is overkill, and we should probably look | |
2041 | * at using the i_lock spinlock to protect i_flags, and then make sure | |
2042 | * it is so documented in include/linux/fs.h and that all code follows | |
2043 | * the locking convention!! | |
2044 | */ | |
2045 | void inode_set_flags(struct inode *inode, unsigned int flags, | |
2046 | unsigned int mask) | |
2047 | { | |
2048 | unsigned int old_flags, new_flags; | |
2049 | ||
2050 | WARN_ON_ONCE(flags & ~mask); | |
2051 | do { | |
2052 | old_flags = ACCESS_ONCE(inode->i_flags); | |
2053 | new_flags = (old_flags & ~mask) | flags; | |
2054 | } while (unlikely(cmpxchg(&inode->i_flags, old_flags, | |
2055 | new_flags) != old_flags)); | |
2056 | } | |
2057 | EXPORT_SYMBOL(inode_set_flags); |