]>
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> |
ae5e165d | 21 | #include <linux/iversion.h> |
0ae45f63 | 22 | #include <trace/events/writeback.h> |
a66979ab | 23 | #include "internal.h" |
1da177e4 | 24 | |
250df6ed | 25 | /* |
4b4563dc | 26 | * Inode locking rules: |
250df6ed DC |
27 | * |
28 | * inode->i_lock protects: | |
29 | * inode->i_state, inode->i_hash, __iget() | |
bc3b14cb | 30 | * Inode LRU list locks protect: |
98b745c6 | 31 | * inode->i_sb->s_inode_lru, inode->i_lru |
74278da9 DC |
32 | * inode->i_sb->s_inode_list_lock protects: |
33 | * inode->i_sb->s_inodes, inode->i_sb_list | |
f758eeab | 34 | * bdi->wb.list_lock protects: |
c7f54084 | 35 | * bdi->wb.b_{dirty,io,more_io,dirty_time}, inode->i_io_list |
67a23c49 DC |
36 | * inode_hash_lock protects: |
37 | * inode_hashtable, inode->i_hash | |
250df6ed DC |
38 | * |
39 | * Lock ordering: | |
55fa6091 | 40 | * |
74278da9 | 41 | * inode->i_sb->s_inode_list_lock |
55fa6091 | 42 | * inode->i_lock |
bc3b14cb | 43 | * Inode LRU list locks |
a66979ab | 44 | * |
f758eeab | 45 | * bdi->wb.list_lock |
a66979ab | 46 | * inode->i_lock |
67a23c49 DC |
47 | * |
48 | * inode_hash_lock | |
74278da9 | 49 | * inode->i_sb->s_inode_list_lock |
67a23c49 DC |
50 | * inode->i_lock |
51 | * | |
52 | * iunique_lock | |
53 | * inode_hash_lock | |
250df6ed DC |
54 | */ |
55 | ||
fa3536cc ED |
56 | static unsigned int i_hash_mask __read_mostly; |
57 | static unsigned int i_hash_shift __read_mostly; | |
67a23c49 DC |
58 | static struct hlist_head *inode_hashtable __read_mostly; |
59 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_hash_lock); | |
1da177e4 | 60 | |
7dcda1c9 JA |
61 | /* |
62 | * Empty aops. Can be used for the cases where the user does not | |
63 | * define any of the address_space operations. | |
64 | */ | |
65 | const struct address_space_operations empty_aops = { | |
66 | }; | |
67 | EXPORT_SYMBOL(empty_aops); | |
68 | ||
1da177e4 LT |
69 | /* |
70 | * Statistics gathering.. | |
71 | */ | |
72 | struct inodes_stat_t inodes_stat; | |
73 | ||
3942c07c GC |
74 | static DEFINE_PER_CPU(unsigned long, nr_inodes); |
75 | static DEFINE_PER_CPU(unsigned long, nr_unused); | |
cffbc8aa | 76 | |
6b3304b5 | 77 | static struct kmem_cache *inode_cachep __read_mostly; |
1da177e4 | 78 | |
3942c07c | 79 | static long get_nr_inodes(void) |
cffbc8aa | 80 | { |
3e880fb5 | 81 | int i; |
3942c07c | 82 | long sum = 0; |
3e880fb5 NP |
83 | for_each_possible_cpu(i) |
84 | sum += per_cpu(nr_inodes, i); | |
85 | return sum < 0 ? 0 : sum; | |
cffbc8aa DC |
86 | } |
87 | ||
3942c07c | 88 | static inline long get_nr_inodes_unused(void) |
cffbc8aa | 89 | { |
fcb94f72 | 90 | int i; |
3942c07c | 91 | long sum = 0; |
fcb94f72 DC |
92 | for_each_possible_cpu(i) |
93 | sum += per_cpu(nr_unused, i); | |
94 | return sum < 0 ? 0 : sum; | |
cffbc8aa DC |
95 | } |
96 | ||
3942c07c | 97 | long get_nr_dirty_inodes(void) |
cffbc8aa | 98 | { |
3e880fb5 | 99 | /* not actually dirty inodes, but a wild approximation */ |
3942c07c | 100 | long nr_dirty = get_nr_inodes() - get_nr_inodes_unused(); |
cffbc8aa | 101 | return nr_dirty > 0 ? nr_dirty : 0; |
cffbc8aa DC |
102 | } |
103 | ||
104 | /* | |
105 | * Handle nr_inode sysctl | |
106 | */ | |
107 | #ifdef CONFIG_SYSCTL | |
1f7e0616 | 108 | int proc_nr_inodes(struct ctl_table *table, int write, |
cffbc8aa DC |
109 | void __user *buffer, size_t *lenp, loff_t *ppos) |
110 | { | |
111 | inodes_stat.nr_inodes = get_nr_inodes(); | |
fcb94f72 | 112 | inodes_stat.nr_unused = get_nr_inodes_unused(); |
3942c07c | 113 | return proc_doulongvec_minmax(table, write, buffer, lenp, ppos); |
cffbc8aa DC |
114 | } |
115 | #endif | |
116 | ||
bd9b51e7 AV |
117 | static int no_open(struct inode *inode, struct file *file) |
118 | { | |
119 | return -ENXIO; | |
120 | } | |
121 | ||
2cb1599f | 122 | /** |
6e7c2b4d | 123 | * inode_init_always - perform inode structure initialisation |
0bc02f3f RD |
124 | * @sb: superblock inode belongs to |
125 | * @inode: inode to initialise | |
2cb1599f DC |
126 | * |
127 | * These are initializations that need to be done on every inode | |
128 | * allocation as the fields are not initialised by slab allocation. | |
129 | */ | |
54e34621 | 130 | int inode_init_always(struct super_block *sb, struct inode *inode) |
1da177e4 | 131 | { |
6e1d5dcc | 132 | static const struct inode_operations empty_iops; |
bd9b51e7 | 133 | static const struct file_operations no_open_fops = {.open = no_open}; |
6b3304b5 | 134 | struct address_space *const mapping = &inode->i_data; |
2cb1599f DC |
135 | |
136 | inode->i_sb = sb; | |
137 | inode->i_blkbits = sb->s_blocksize_bits; | |
138 | inode->i_flags = 0; | |
139 | atomic_set(&inode->i_count, 1); | |
140 | inode->i_op = &empty_iops; | |
bd9b51e7 | 141 | inode->i_fop = &no_open_fops; |
a78ef704 | 142 | inode->__i_nlink = 1; |
3ddcd056 | 143 | inode->i_opflags = 0; |
d0a5b995 AG |
144 | if (sb->s_xattr) |
145 | inode->i_opflags |= IOP_XATTR; | |
92361636 EB |
146 | i_uid_write(inode, 0); |
147 | i_gid_write(inode, 0); | |
2cb1599f DC |
148 | atomic_set(&inode->i_writecount, 0); |
149 | inode->i_size = 0; | |
c75b1d94 | 150 | inode->i_write_hint = WRITE_LIFE_NOT_SET; |
2cb1599f DC |
151 | inode->i_blocks = 0; |
152 | inode->i_bytes = 0; | |
153 | inode->i_generation = 0; | |
2cb1599f DC |
154 | inode->i_pipe = NULL; |
155 | inode->i_bdev = NULL; | |
156 | inode->i_cdev = NULL; | |
61ba64fc | 157 | inode->i_link = NULL; |
84e710da | 158 | inode->i_dir_seq = 0; |
2cb1599f DC |
159 | inode->i_rdev = 0; |
160 | inode->dirtied_when = 0; | |
6146f0d5 | 161 | |
3d65ae46 TE |
162 | #ifdef CONFIG_CGROUP_WRITEBACK |
163 | inode->i_wb_frn_winner = 0; | |
164 | inode->i_wb_frn_avg_time = 0; | |
165 | inode->i_wb_frn_history = 0; | |
166 | #endif | |
167 | ||
6146f0d5 | 168 | if (security_inode_alloc(inode)) |
54e34621 | 169 | goto out; |
2cb1599f DC |
170 | spin_lock_init(&inode->i_lock); |
171 | lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key); | |
172 | ||
9902af79 AV |
173 | init_rwsem(&inode->i_rwsem); |
174 | lockdep_set_class(&inode->i_rwsem, &sb->s_type->i_mutex_key); | |
2cb1599f | 175 | |
bd5fe6c5 | 176 | atomic_set(&inode->i_dio_count, 0); |
2cb1599f DC |
177 | |
178 | mapping->a_ops = &empty_aops; | |
179 | mapping->host = inode; | |
180 | mapping->flags = 0; | |
4bb5f5d9 | 181 | atomic_set(&mapping->i_mmap_writable, 0); |
3c1d4378 | 182 | mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE); |
252aa6f5 | 183 | mapping->private_data = NULL; |
2cb1599f | 184 | mapping->writeback_index = 0; |
2cb1599f DC |
185 | inode->i_private = NULL; |
186 | inode->i_mapping = mapping; | |
b3d9b7a3 | 187 | INIT_HLIST_HEAD(&inode->i_dentry); /* buggered by rcu freeing */ |
f19d4a8f AV |
188 | #ifdef CONFIG_FS_POSIX_ACL |
189 | inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED; | |
190 | #endif | |
2cb1599f | 191 | |
3be25f49 EP |
192 | #ifdef CONFIG_FSNOTIFY |
193 | inode->i_fsnotify_mask = 0; | |
194 | #endif | |
4a075e39 | 195 | inode->i_flctx = NULL; |
3e880fb5 | 196 | this_cpu_inc(nr_inodes); |
cffbc8aa | 197 | |
54e34621 | 198 | return 0; |
54e34621 CH |
199 | out: |
200 | return -ENOMEM; | |
1da177e4 | 201 | } |
2cb1599f DC |
202 | EXPORT_SYMBOL(inode_init_always); |
203 | ||
204 | static struct inode *alloc_inode(struct super_block *sb) | |
205 | { | |
206 | struct inode *inode; | |
207 | ||
208 | if (sb->s_op->alloc_inode) | |
209 | inode = sb->s_op->alloc_inode(sb); | |
210 | else | |
211 | inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL); | |
212 | ||
54e34621 CH |
213 | if (!inode) |
214 | return NULL; | |
215 | ||
216 | if (unlikely(inode_init_always(sb, inode))) { | |
217 | if (inode->i_sb->s_op->destroy_inode) | |
218 | inode->i_sb->s_op->destroy_inode(inode); | |
219 | else | |
220 | kmem_cache_free(inode_cachep, inode); | |
221 | return NULL; | |
222 | } | |
223 | ||
224 | return inode; | |
2cb1599f | 225 | } |
1da177e4 | 226 | |
ff0c7d15 NP |
227 | void free_inode_nonrcu(struct inode *inode) |
228 | { | |
229 | kmem_cache_free(inode_cachep, inode); | |
230 | } | |
231 | EXPORT_SYMBOL(free_inode_nonrcu); | |
232 | ||
2e00c97e | 233 | void __destroy_inode(struct inode *inode) |
1da177e4 | 234 | { |
b7542f8c | 235 | BUG_ON(inode_has_buffers(inode)); |
52ebea74 | 236 | inode_detach_wb(inode); |
1da177e4 | 237 | security_inode_free(inode); |
3be25f49 | 238 | fsnotify_inode_delete(inode); |
f27a0fe0 | 239 | locks_free_lock_context(inode); |
7ada4db8 MS |
240 | if (!inode->i_nlink) { |
241 | WARN_ON(atomic_long_read(&inode->i_sb->s_remove_count) == 0); | |
242 | atomic_long_dec(&inode->i_sb->s_remove_count); | |
243 | } | |
244 | ||
f19d4a8f | 245 | #ifdef CONFIG_FS_POSIX_ACL |
b8a7a3a6 | 246 | if (inode->i_acl && !is_uncached_acl(inode->i_acl)) |
f19d4a8f | 247 | posix_acl_release(inode->i_acl); |
b8a7a3a6 | 248 | if (inode->i_default_acl && !is_uncached_acl(inode->i_default_acl)) |
f19d4a8f AV |
249 | posix_acl_release(inode->i_default_acl); |
250 | #endif | |
3e880fb5 | 251 | this_cpu_dec(nr_inodes); |
2e00c97e CH |
252 | } |
253 | EXPORT_SYMBOL(__destroy_inode); | |
254 | ||
fa0d7e3d NP |
255 | static void i_callback(struct rcu_head *head) |
256 | { | |
257 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
fa0d7e3d NP |
258 | kmem_cache_free(inode_cachep, inode); |
259 | } | |
260 | ||
56b0dacf | 261 | static void destroy_inode(struct inode *inode) |
2e00c97e | 262 | { |
7ccf19a8 | 263 | BUG_ON(!list_empty(&inode->i_lru)); |
2e00c97e | 264 | __destroy_inode(inode); |
1da177e4 LT |
265 | if (inode->i_sb->s_op->destroy_inode) |
266 | inode->i_sb->s_op->destroy_inode(inode); | |
267 | else | |
fa0d7e3d | 268 | call_rcu(&inode->i_rcu, i_callback); |
1da177e4 | 269 | } |
1da177e4 | 270 | |
7ada4db8 MS |
271 | /** |
272 | * drop_nlink - directly drop an inode's link count | |
273 | * @inode: inode | |
274 | * | |
275 | * This is a low-level filesystem helper to replace any | |
276 | * direct filesystem manipulation of i_nlink. In cases | |
277 | * where we are attempting to track writes to the | |
278 | * filesystem, a decrement to zero means an imminent | |
279 | * write when the file is truncated and actually unlinked | |
280 | * on the filesystem. | |
281 | */ | |
282 | void drop_nlink(struct inode *inode) | |
283 | { | |
284 | WARN_ON(inode->i_nlink == 0); | |
285 | inode->__i_nlink--; | |
286 | if (!inode->i_nlink) | |
287 | atomic_long_inc(&inode->i_sb->s_remove_count); | |
288 | } | |
289 | EXPORT_SYMBOL(drop_nlink); | |
290 | ||
291 | /** | |
292 | * clear_nlink - directly zero an inode's link count | |
293 | * @inode: inode | |
294 | * | |
295 | * This is a low-level filesystem helper to replace any | |
296 | * direct filesystem manipulation of i_nlink. See | |
297 | * drop_nlink() for why we care about i_nlink hitting zero. | |
298 | */ | |
299 | void clear_nlink(struct inode *inode) | |
300 | { | |
301 | if (inode->i_nlink) { | |
302 | inode->__i_nlink = 0; | |
303 | atomic_long_inc(&inode->i_sb->s_remove_count); | |
304 | } | |
305 | } | |
306 | EXPORT_SYMBOL(clear_nlink); | |
307 | ||
308 | /** | |
309 | * set_nlink - directly set an inode's link count | |
310 | * @inode: inode | |
311 | * @nlink: new nlink (should be non-zero) | |
312 | * | |
313 | * This is a low-level filesystem helper to replace any | |
314 | * direct filesystem manipulation of i_nlink. | |
315 | */ | |
316 | void set_nlink(struct inode *inode, unsigned int nlink) | |
317 | { | |
318 | if (!nlink) { | |
7ada4db8 MS |
319 | clear_nlink(inode); |
320 | } else { | |
321 | /* Yes, some filesystems do change nlink from zero to one */ | |
322 | if (inode->i_nlink == 0) | |
323 | atomic_long_dec(&inode->i_sb->s_remove_count); | |
324 | ||
325 | inode->__i_nlink = nlink; | |
326 | } | |
327 | } | |
328 | EXPORT_SYMBOL(set_nlink); | |
329 | ||
330 | /** | |
331 | * inc_nlink - directly increment an inode's link count | |
332 | * @inode: inode | |
333 | * | |
334 | * This is a low-level filesystem helper to replace any | |
335 | * direct filesystem manipulation of i_nlink. Currently, | |
336 | * it is only here for parity with dec_nlink(). | |
337 | */ | |
338 | void inc_nlink(struct inode *inode) | |
339 | { | |
f4e0c30c AV |
340 | if (unlikely(inode->i_nlink == 0)) { |
341 | WARN_ON(!(inode->i_state & I_LINKABLE)); | |
7ada4db8 | 342 | atomic_long_dec(&inode->i_sb->s_remove_count); |
f4e0c30c | 343 | } |
7ada4db8 MS |
344 | |
345 | inode->__i_nlink++; | |
346 | } | |
347 | EXPORT_SYMBOL(inc_nlink); | |
348 | ||
ae23395d | 349 | static void __address_space_init_once(struct address_space *mapping) |
2aa15890 | 350 | { |
b93b0163 | 351 | INIT_RADIX_TREE(&mapping->i_pages, GFP_ATOMIC | __GFP_ACCOUNT); |
c8c06efa | 352 | init_rwsem(&mapping->i_mmap_rwsem); |
2aa15890 MS |
353 | INIT_LIST_HEAD(&mapping->private_list); |
354 | spin_lock_init(&mapping->private_lock); | |
f808c13f | 355 | mapping->i_mmap = RB_ROOT_CACHED; |
2aa15890 | 356 | } |
ae23395d DC |
357 | |
358 | void address_space_init_once(struct address_space *mapping) | |
359 | { | |
360 | memset(mapping, 0, sizeof(*mapping)); | |
361 | __address_space_init_once(mapping); | |
362 | } | |
2aa15890 MS |
363 | EXPORT_SYMBOL(address_space_init_once); |
364 | ||
1da177e4 LT |
365 | /* |
366 | * These are initializations that only need to be done | |
367 | * once, because the fields are idempotent across use | |
368 | * of the inode, so let the slab aware of that. | |
369 | */ | |
370 | void inode_init_once(struct inode *inode) | |
371 | { | |
372 | memset(inode, 0, sizeof(*inode)); | |
373 | INIT_HLIST_NODE(&inode->i_hash); | |
1da177e4 | 374 | INIT_LIST_HEAD(&inode->i_devices); |
c7f54084 | 375 | INIT_LIST_HEAD(&inode->i_io_list); |
6c60d2b5 | 376 | INIT_LIST_HEAD(&inode->i_wb_list); |
7ccf19a8 | 377 | INIT_LIST_HEAD(&inode->i_lru); |
ae23395d | 378 | __address_space_init_once(&inode->i_data); |
1da177e4 LT |
379 | i_size_ordered_init(inode); |
380 | } | |
1da177e4 LT |
381 | EXPORT_SYMBOL(inode_init_once); |
382 | ||
51cc5068 | 383 | static void init_once(void *foo) |
1da177e4 | 384 | { |
6b3304b5 | 385 | struct inode *inode = (struct inode *) foo; |
1da177e4 | 386 | |
a35afb83 | 387 | inode_init_once(inode); |
1da177e4 LT |
388 | } |
389 | ||
390 | /* | |
250df6ed | 391 | * inode->i_lock must be held |
1da177e4 | 392 | */ |
6b3304b5 | 393 | void __iget(struct inode *inode) |
1da177e4 | 394 | { |
9e38d86f NP |
395 | atomic_inc(&inode->i_count); |
396 | } | |
2e147f1e | 397 | |
7de9c6ee AV |
398 | /* |
399 | * get additional reference to inode; caller must already hold one. | |
400 | */ | |
401 | void ihold(struct inode *inode) | |
402 | { | |
403 | WARN_ON(atomic_inc_return(&inode->i_count) < 2); | |
404 | } | |
405 | EXPORT_SYMBOL(ihold); | |
406 | ||
9e38d86f NP |
407 | static void inode_lru_list_add(struct inode *inode) |
408 | { | |
bc3b14cb | 409 | if (list_lru_add(&inode->i_sb->s_inode_lru, &inode->i_lru)) |
fcb94f72 | 410 | this_cpu_inc(nr_unused); |
563f4001 JB |
411 | else |
412 | inode->i_state |= I_REFERENCED; | |
9e38d86f | 413 | } |
2e147f1e | 414 | |
4eff96dd JK |
415 | /* |
416 | * Add inode to LRU if needed (inode is unused and clean). | |
417 | * | |
418 | * Needs inode->i_lock held. | |
419 | */ | |
420 | void inode_add_lru(struct inode *inode) | |
421 | { | |
0ae45f63 TT |
422 | if (!(inode->i_state & (I_DIRTY_ALL | I_SYNC | |
423 | I_FREEING | I_WILL_FREE)) && | |
1751e8a6 | 424 | !atomic_read(&inode->i_count) && inode->i_sb->s_flags & SB_ACTIVE) |
4eff96dd JK |
425 | inode_lru_list_add(inode); |
426 | } | |
427 | ||
428 | ||
9e38d86f NP |
429 | static void inode_lru_list_del(struct inode *inode) |
430 | { | |
bc3b14cb DC |
431 | |
432 | if (list_lru_del(&inode->i_sb->s_inode_lru, &inode->i_lru)) | |
fcb94f72 | 433 | this_cpu_dec(nr_unused); |
1da177e4 LT |
434 | } |
435 | ||
646ec461 CH |
436 | /** |
437 | * inode_sb_list_add - add inode to the superblock list of inodes | |
438 | * @inode: inode to add | |
439 | */ | |
440 | void inode_sb_list_add(struct inode *inode) | |
441 | { | |
74278da9 | 442 | spin_lock(&inode->i_sb->s_inode_list_lock); |
55fa6091 | 443 | list_add(&inode->i_sb_list, &inode->i_sb->s_inodes); |
74278da9 | 444 | spin_unlock(&inode->i_sb->s_inode_list_lock); |
646ec461 CH |
445 | } |
446 | EXPORT_SYMBOL_GPL(inode_sb_list_add); | |
447 | ||
55fa6091 | 448 | static inline void inode_sb_list_del(struct inode *inode) |
646ec461 | 449 | { |
a209dfc7 | 450 | if (!list_empty(&inode->i_sb_list)) { |
74278da9 | 451 | spin_lock(&inode->i_sb->s_inode_list_lock); |
a209dfc7 | 452 | list_del_init(&inode->i_sb_list); |
74278da9 | 453 | spin_unlock(&inode->i_sb->s_inode_list_lock); |
a209dfc7 | 454 | } |
646ec461 CH |
455 | } |
456 | ||
4c51acbc DC |
457 | static unsigned long hash(struct super_block *sb, unsigned long hashval) |
458 | { | |
459 | unsigned long tmp; | |
460 | ||
461 | tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) / | |
462 | L1_CACHE_BYTES; | |
4b4563dc CH |
463 | tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> i_hash_shift); |
464 | return tmp & i_hash_mask; | |
4c51acbc DC |
465 | } |
466 | ||
467 | /** | |
468 | * __insert_inode_hash - hash an inode | |
469 | * @inode: unhashed inode | |
470 | * @hashval: unsigned long value used to locate this object in the | |
471 | * inode_hashtable. | |
472 | * | |
473 | * Add an inode to the inode hash for this superblock. | |
474 | */ | |
475 | void __insert_inode_hash(struct inode *inode, unsigned long hashval) | |
476 | { | |
646ec461 CH |
477 | struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval); |
478 | ||
67a23c49 | 479 | spin_lock(&inode_hash_lock); |
250df6ed | 480 | spin_lock(&inode->i_lock); |
646ec461 | 481 | hlist_add_head(&inode->i_hash, b); |
250df6ed | 482 | spin_unlock(&inode->i_lock); |
67a23c49 | 483 | spin_unlock(&inode_hash_lock); |
4c51acbc DC |
484 | } |
485 | EXPORT_SYMBOL(__insert_inode_hash); | |
486 | ||
4c51acbc | 487 | /** |
f2ee7abf | 488 | * __remove_inode_hash - remove an inode from the hash |
4c51acbc DC |
489 | * @inode: inode to unhash |
490 | * | |
491 | * Remove an inode from the superblock. | |
492 | */ | |
f2ee7abf | 493 | void __remove_inode_hash(struct inode *inode) |
4c51acbc | 494 | { |
67a23c49 | 495 | spin_lock(&inode_hash_lock); |
250df6ed | 496 | spin_lock(&inode->i_lock); |
4c51acbc | 497 | hlist_del_init(&inode->i_hash); |
250df6ed | 498 | spin_unlock(&inode->i_lock); |
67a23c49 | 499 | spin_unlock(&inode_hash_lock); |
4c51acbc | 500 | } |
f2ee7abf | 501 | EXPORT_SYMBOL(__remove_inode_hash); |
4c51acbc | 502 | |
dbd5768f | 503 | void clear_inode(struct inode *inode) |
b0683aa6 | 504 | { |
08142579 | 505 | /* |
b93b0163 | 506 | * We have to cycle the i_pages lock here because reclaim can be in the |
08142579 | 507 | * process of removing the last page (in __delete_from_page_cache()) |
b93b0163 | 508 | * and we must not free the mapping under it. |
08142579 | 509 | */ |
b93b0163 | 510 | xa_lock_irq(&inode->i_data.i_pages); |
b0683aa6 | 511 | BUG_ON(inode->i_data.nrpages); |
f9fe48be | 512 | BUG_ON(inode->i_data.nrexceptional); |
b93b0163 | 513 | xa_unlock_irq(&inode->i_data.i_pages); |
b0683aa6 AV |
514 | BUG_ON(!list_empty(&inode->i_data.private_list)); |
515 | BUG_ON(!(inode->i_state & I_FREEING)); | |
516 | BUG_ON(inode->i_state & I_CLEAR); | |
6c60d2b5 | 517 | BUG_ON(!list_empty(&inode->i_wb_list)); |
fa0d7e3d | 518 | /* don't need i_lock here, no concurrent mods to i_state */ |
b0683aa6 AV |
519 | inode->i_state = I_FREEING | I_CLEAR; |
520 | } | |
dbd5768f | 521 | EXPORT_SYMBOL(clear_inode); |
b0683aa6 | 522 | |
b2b2af8e DC |
523 | /* |
524 | * Free the inode passed in, removing it from the lists it is still connected | |
525 | * to. We remove any pages still attached to the inode and wait for any IO that | |
526 | * is still in progress before finally destroying the inode. | |
527 | * | |
528 | * An inode must already be marked I_FREEING so that we avoid the inode being | |
529 | * moved back onto lists if we race with other code that manipulates the lists | |
530 | * (e.g. writeback_single_inode). The caller is responsible for setting this. | |
531 | * | |
532 | * An inode must already be removed from the LRU list before being evicted from | |
533 | * the cache. This should occur atomically with setting the I_FREEING state | |
534 | * flag, so no inodes here should ever be on the LRU when being evicted. | |
535 | */ | |
644da596 | 536 | static void evict(struct inode *inode) |
b4272d4c AV |
537 | { |
538 | const struct super_operations *op = inode->i_sb->s_op; | |
539 | ||
b2b2af8e DC |
540 | BUG_ON(!(inode->i_state & I_FREEING)); |
541 | BUG_ON(!list_empty(&inode->i_lru)); | |
542 | ||
c7f54084 DC |
543 | if (!list_empty(&inode->i_io_list)) |
544 | inode_io_list_del(inode); | |
b12362bd | 545 | |
55fa6091 DC |
546 | inode_sb_list_del(inode); |
547 | ||
169ebd90 JK |
548 | /* |
549 | * Wait for flusher thread to be done with the inode so that filesystem | |
550 | * does not start destroying it while writeback is still running. Since | |
551 | * the inode has I_FREEING set, flusher thread won't start new work on | |
552 | * the inode. We just have to wait for running writeback to finish. | |
553 | */ | |
554 | inode_wait_for_writeback(inode); | |
7994e6f7 | 555 | |
be7ce416 AV |
556 | if (op->evict_inode) { |
557 | op->evict_inode(inode); | |
b4272d4c | 558 | } else { |
91b0abe3 | 559 | truncate_inode_pages_final(&inode->i_data); |
dbd5768f | 560 | clear_inode(inode); |
b4272d4c | 561 | } |
661074e9 AV |
562 | if (S_ISBLK(inode->i_mode) && inode->i_bdev) |
563 | bd_forget(inode); | |
564 | if (S_ISCHR(inode->i_mode) && inode->i_cdev) | |
565 | cd_forget(inode); | |
b2b2af8e DC |
566 | |
567 | remove_inode_hash(inode); | |
568 | ||
569 | spin_lock(&inode->i_lock); | |
570 | wake_up_bit(&inode->i_state, __I_NEW); | |
571 | BUG_ON(inode->i_state != (I_FREEING | I_CLEAR)); | |
572 | spin_unlock(&inode->i_lock); | |
573 | ||
574 | destroy_inode(inode); | |
b4272d4c AV |
575 | } |
576 | ||
1da177e4 LT |
577 | /* |
578 | * dispose_list - dispose of the contents of a local list | |
579 | * @head: the head of the list to free | |
580 | * | |
581 | * Dispose-list gets a local list with local inodes in it, so it doesn't | |
582 | * need to worry about list corruption and SMP locks. | |
583 | */ | |
584 | static void dispose_list(struct list_head *head) | |
585 | { | |
1da177e4 LT |
586 | while (!list_empty(head)) { |
587 | struct inode *inode; | |
588 | ||
7ccf19a8 NP |
589 | inode = list_first_entry(head, struct inode, i_lru); |
590 | list_del_init(&inode->i_lru); | |
1da177e4 | 591 | |
644da596 | 592 | evict(inode); |
ac05fbb4 | 593 | cond_resched(); |
1da177e4 | 594 | } |
1da177e4 LT |
595 | } |
596 | ||
63997e98 AV |
597 | /** |
598 | * evict_inodes - evict all evictable inodes for a superblock | |
599 | * @sb: superblock to operate on | |
600 | * | |
601 | * Make sure that no inodes with zero refcount are retained. This is | |
1751e8a6 | 602 | * called by superblock shutdown after having SB_ACTIVE flag removed, |
63997e98 AV |
603 | * so any inode reaching zero refcount during or after that call will |
604 | * be immediately evicted. | |
1da177e4 | 605 | */ |
63997e98 | 606 | void evict_inodes(struct super_block *sb) |
1da177e4 | 607 | { |
63997e98 AV |
608 | struct inode *inode, *next; |
609 | LIST_HEAD(dispose); | |
1da177e4 | 610 | |
ac05fbb4 | 611 | again: |
74278da9 | 612 | spin_lock(&sb->s_inode_list_lock); |
63997e98 AV |
613 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
614 | if (atomic_read(&inode->i_count)) | |
aabb8fdb | 615 | continue; |
250df6ed DC |
616 | |
617 | spin_lock(&inode->i_lock); | |
618 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
619 | spin_unlock(&inode->i_lock); | |
1da177e4 | 620 | continue; |
250df6ed | 621 | } |
63997e98 AV |
622 | |
623 | inode->i_state |= I_FREEING; | |
02afc410 | 624 | inode_lru_list_del(inode); |
250df6ed | 625 | spin_unlock(&inode->i_lock); |
02afc410 | 626 | list_add(&inode->i_lru, &dispose); |
ac05fbb4 JB |
627 | |
628 | /* | |
629 | * We can have a ton of inodes to evict at unmount time given | |
630 | * enough memory, check to see if we need to go to sleep for a | |
631 | * bit so we don't livelock. | |
632 | */ | |
633 | if (need_resched()) { | |
634 | spin_unlock(&sb->s_inode_list_lock); | |
635 | cond_resched(); | |
636 | dispose_list(&dispose); | |
637 | goto again; | |
638 | } | |
1da177e4 | 639 | } |
74278da9 | 640 | spin_unlock(&sb->s_inode_list_lock); |
63997e98 AV |
641 | |
642 | dispose_list(&dispose); | |
1da177e4 | 643 | } |
799ea9e9 | 644 | EXPORT_SYMBOL_GPL(evict_inodes); |
1da177e4 | 645 | |
1da177e4 | 646 | /** |
a0318786 CH |
647 | * invalidate_inodes - attempt to free all inodes on a superblock |
648 | * @sb: superblock to operate on | |
93b270f7 | 649 | * @kill_dirty: flag to guide handling of dirty inodes |
1da177e4 | 650 | * |
a0318786 CH |
651 | * Attempts to free all inodes for a given superblock. If there were any |
652 | * busy inodes return a non-zero value, else zero. | |
93b270f7 N |
653 | * If @kill_dirty is set, discard dirty inodes too, otherwise treat |
654 | * them as busy. | |
1da177e4 | 655 | */ |
93b270f7 | 656 | int invalidate_inodes(struct super_block *sb, bool kill_dirty) |
1da177e4 | 657 | { |
cffbc8aa | 658 | int busy = 0; |
a0318786 CH |
659 | struct inode *inode, *next; |
660 | LIST_HEAD(dispose); | |
1da177e4 | 661 | |
74278da9 | 662 | spin_lock(&sb->s_inode_list_lock); |
a0318786 | 663 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
250df6ed DC |
664 | spin_lock(&inode->i_lock); |
665 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
666 | spin_unlock(&inode->i_lock); | |
aabb8fdb | 667 | continue; |
250df6ed | 668 | } |
0ae45f63 | 669 | if (inode->i_state & I_DIRTY_ALL && !kill_dirty) { |
250df6ed | 670 | spin_unlock(&inode->i_lock); |
93b270f7 N |
671 | busy = 1; |
672 | continue; | |
673 | } | |
99a38919 | 674 | if (atomic_read(&inode->i_count)) { |
250df6ed | 675 | spin_unlock(&inode->i_lock); |
99a38919 | 676 | busy = 1; |
1da177e4 LT |
677 | continue; |
678 | } | |
99a38919 | 679 | |
99a38919 | 680 | inode->i_state |= I_FREEING; |
02afc410 | 681 | inode_lru_list_del(inode); |
250df6ed | 682 | spin_unlock(&inode->i_lock); |
02afc410 | 683 | list_add(&inode->i_lru, &dispose); |
1da177e4 | 684 | } |
74278da9 | 685 | spin_unlock(&sb->s_inode_list_lock); |
1da177e4 | 686 | |
a0318786 | 687 | dispose_list(&dispose); |
1da177e4 LT |
688 | |
689 | return busy; | |
690 | } | |
1da177e4 | 691 | |
1da177e4 | 692 | /* |
bc3b14cb | 693 | * Isolate the inode from the LRU in preparation for freeing it. |
1da177e4 LT |
694 | * |
695 | * Any inodes which are pinned purely because of attached pagecache have their | |
9e38d86f NP |
696 | * pagecache removed. If the inode has metadata buffers attached to |
697 | * mapping->private_list then try to remove them. | |
1da177e4 | 698 | * |
9e38d86f NP |
699 | * If the inode has the I_REFERENCED flag set, then it means that it has been |
700 | * used recently - the flag is set in iput_final(). When we encounter such an | |
701 | * inode, clear the flag and move it to the back of the LRU so it gets another | |
702 | * pass through the LRU before it gets reclaimed. This is necessary because of | |
703 | * the fact we are doing lazy LRU updates to minimise lock contention so the | |
704 | * LRU does not have strict ordering. Hence we don't want to reclaim inodes | |
705 | * with this flag set because they are the inodes that are out of order. | |
1da177e4 | 706 | */ |
3f97b163 VD |
707 | static enum lru_status inode_lru_isolate(struct list_head *item, |
708 | struct list_lru_one *lru, spinlock_t *lru_lock, void *arg) | |
1da177e4 | 709 | { |
bc3b14cb DC |
710 | struct list_head *freeable = arg; |
711 | struct inode *inode = container_of(item, struct inode, i_lru); | |
1da177e4 | 712 | |
bc3b14cb DC |
713 | /* |
714 | * we are inverting the lru lock/inode->i_lock here, so use a trylock. | |
715 | * If we fail to get the lock, just skip it. | |
716 | */ | |
717 | if (!spin_trylock(&inode->i_lock)) | |
718 | return LRU_SKIP; | |
1da177e4 | 719 | |
bc3b14cb DC |
720 | /* |
721 | * Referenced or dirty inodes are still in use. Give them another pass | |
722 | * through the LRU as we canot reclaim them now. | |
723 | */ | |
724 | if (atomic_read(&inode->i_count) || | |
725 | (inode->i_state & ~I_REFERENCED)) { | |
3f97b163 | 726 | list_lru_isolate(lru, &inode->i_lru); |
bc3b14cb DC |
727 | spin_unlock(&inode->i_lock); |
728 | this_cpu_dec(nr_unused); | |
729 | return LRU_REMOVED; | |
730 | } | |
1da177e4 | 731 | |
bc3b14cb DC |
732 | /* recently referenced inodes get one more pass */ |
733 | if (inode->i_state & I_REFERENCED) { | |
734 | inode->i_state &= ~I_REFERENCED; | |
735 | spin_unlock(&inode->i_lock); | |
736 | return LRU_ROTATE; | |
737 | } | |
1da177e4 | 738 | |
bc3b14cb DC |
739 | if (inode_has_buffers(inode) || inode->i_data.nrpages) { |
740 | __iget(inode); | |
741 | spin_unlock(&inode->i_lock); | |
742 | spin_unlock(lru_lock); | |
743 | if (remove_inode_buffers(inode)) { | |
744 | unsigned long reap; | |
745 | reap = invalidate_mapping_pages(&inode->i_data, 0, -1); | |
746 | if (current_is_kswapd()) | |
747 | __count_vm_events(KSWAPD_INODESTEAL, reap); | |
748 | else | |
749 | __count_vm_events(PGINODESTEAL, reap); | |
750 | if (current->reclaim_state) | |
751 | current->reclaim_state->reclaimed_slab += reap; | |
02afc410 | 752 | } |
bc3b14cb DC |
753 | iput(inode); |
754 | spin_lock(lru_lock); | |
755 | return LRU_RETRY; | |
756 | } | |
02afc410 | 757 | |
bc3b14cb DC |
758 | WARN_ON(inode->i_state & I_NEW); |
759 | inode->i_state |= I_FREEING; | |
3f97b163 | 760 | list_lru_isolate_move(lru, &inode->i_lru, freeable); |
bc3b14cb | 761 | spin_unlock(&inode->i_lock); |
9e38d86f | 762 | |
bc3b14cb DC |
763 | this_cpu_dec(nr_unused); |
764 | return LRU_REMOVED; | |
765 | } | |
7ccf19a8 | 766 | |
bc3b14cb DC |
767 | /* |
768 | * Walk the superblock inode LRU for freeable inodes and attempt to free them. | |
769 | * This is called from the superblock shrinker function with a number of inodes | |
770 | * to trim from the LRU. Inodes to be freed are moved to a temporary list and | |
771 | * then are freed outside inode_lock by dispose_list(). | |
772 | */ | |
503c358c | 773 | long prune_icache_sb(struct super_block *sb, struct shrink_control *sc) |
bc3b14cb DC |
774 | { |
775 | LIST_HEAD(freeable); | |
776 | long freed; | |
1da177e4 | 777 | |
503c358c VD |
778 | freed = list_lru_shrink_walk(&sb->s_inode_lru, sc, |
779 | inode_lru_isolate, &freeable); | |
1da177e4 | 780 | dispose_list(&freeable); |
0a234c6d | 781 | return freed; |
1da177e4 LT |
782 | } |
783 | ||
1da177e4 LT |
784 | static void __wait_on_freeing_inode(struct inode *inode); |
785 | /* | |
786 | * Called with the inode lock held. | |
1da177e4 | 787 | */ |
6b3304b5 MK |
788 | static struct inode *find_inode(struct super_block *sb, |
789 | struct hlist_head *head, | |
790 | int (*test)(struct inode *, void *), | |
791 | void *data) | |
1da177e4 | 792 | { |
6b3304b5 | 793 | struct inode *inode = NULL; |
1da177e4 LT |
794 | |
795 | repeat: | |
b67bfe0d | 796 | hlist_for_each_entry(inode, head, i_hash) { |
5a3cd992 | 797 | if (inode->i_sb != sb) |
1da177e4 | 798 | continue; |
5a3cd992 | 799 | if (!test(inode, data)) |
1da177e4 | 800 | continue; |
5a3cd992 | 801 | spin_lock(&inode->i_lock); |
a4ffdde6 | 802 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
803 | __wait_on_freeing_inode(inode); |
804 | goto repeat; | |
805 | } | |
f7899bd5 | 806 | __iget(inode); |
250df6ed | 807 | spin_unlock(&inode->i_lock); |
f7899bd5 | 808 | return inode; |
1da177e4 | 809 | } |
f7899bd5 | 810 | return NULL; |
1da177e4 LT |
811 | } |
812 | ||
813 | /* | |
814 | * find_inode_fast is the fast path version of find_inode, see the comment at | |
815 | * iget_locked for details. | |
816 | */ | |
6b3304b5 MK |
817 | static struct inode *find_inode_fast(struct super_block *sb, |
818 | struct hlist_head *head, unsigned long ino) | |
1da177e4 | 819 | { |
6b3304b5 | 820 | struct inode *inode = NULL; |
1da177e4 LT |
821 | |
822 | repeat: | |
b67bfe0d | 823 | hlist_for_each_entry(inode, head, i_hash) { |
5a3cd992 | 824 | if (inode->i_ino != ino) |
1da177e4 | 825 | continue; |
5a3cd992 | 826 | if (inode->i_sb != sb) |
1da177e4 | 827 | continue; |
5a3cd992 | 828 | spin_lock(&inode->i_lock); |
a4ffdde6 | 829 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
830 | __wait_on_freeing_inode(inode); |
831 | goto repeat; | |
832 | } | |
f7899bd5 | 833 | __iget(inode); |
250df6ed | 834 | spin_unlock(&inode->i_lock); |
f7899bd5 | 835 | return inode; |
1da177e4 | 836 | } |
f7899bd5 | 837 | return NULL; |
8290c35f DC |
838 | } |
839 | ||
f991bd2e ED |
840 | /* |
841 | * Each cpu owns a range of LAST_INO_BATCH numbers. | |
842 | * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations, | |
843 | * to renew the exhausted range. | |
8290c35f | 844 | * |
f991bd2e ED |
845 | * This does not significantly increase overflow rate because every CPU can |
846 | * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is | |
847 | * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the | |
848 | * 2^32 range, and is a worst-case. Even a 50% wastage would only increase | |
849 | * overflow rate by 2x, which does not seem too significant. | |
850 | * | |
851 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
852 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
853 | * here to attempt to avoid that. | |
8290c35f | 854 | */ |
f991bd2e ED |
855 | #define LAST_INO_BATCH 1024 |
856 | static DEFINE_PER_CPU(unsigned int, last_ino); | |
857 | ||
85fe4025 | 858 | unsigned int get_next_ino(void) |
8290c35f | 859 | { |
f991bd2e ED |
860 | unsigned int *p = &get_cpu_var(last_ino); |
861 | unsigned int res = *p; | |
8290c35f | 862 | |
f991bd2e ED |
863 | #ifdef CONFIG_SMP |
864 | if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) { | |
865 | static atomic_t shared_last_ino; | |
866 | int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino); | |
867 | ||
868 | res = next - LAST_INO_BATCH; | |
869 | } | |
870 | #endif | |
871 | ||
2adc376c CM |
872 | res++; |
873 | /* get_next_ino should not provide a 0 inode number */ | |
874 | if (unlikely(!res)) | |
875 | res++; | |
876 | *p = res; | |
f991bd2e ED |
877 | put_cpu_var(last_ino); |
878 | return res; | |
8290c35f | 879 | } |
85fe4025 | 880 | EXPORT_SYMBOL(get_next_ino); |
8290c35f | 881 | |
a209dfc7 ED |
882 | /** |
883 | * new_inode_pseudo - obtain an inode | |
884 | * @sb: superblock | |
885 | * | |
886 | * Allocates a new inode for given superblock. | |
887 | * Inode wont be chained in superblock s_inodes list | |
888 | * This means : | |
889 | * - fs can't be unmount | |
890 | * - quotas, fsnotify, writeback can't work | |
891 | */ | |
892 | struct inode *new_inode_pseudo(struct super_block *sb) | |
893 | { | |
894 | struct inode *inode = alloc_inode(sb); | |
895 | ||
896 | if (inode) { | |
897 | spin_lock(&inode->i_lock); | |
898 | inode->i_state = 0; | |
899 | spin_unlock(&inode->i_lock); | |
900 | INIT_LIST_HEAD(&inode->i_sb_list); | |
901 | } | |
902 | return inode; | |
903 | } | |
904 | ||
1da177e4 LT |
905 | /** |
906 | * new_inode - obtain an inode | |
907 | * @sb: superblock | |
908 | * | |
769848c0 | 909 | * Allocates a new inode for given superblock. The default gfp_mask |
3c1d4378 | 910 | * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE. |
769848c0 MG |
911 | * If HIGHMEM pages are unsuitable or it is known that pages allocated |
912 | * for the page cache are not reclaimable or migratable, | |
913 | * mapping_set_gfp_mask() must be called with suitable flags on the | |
914 | * newly created inode's mapping | |
915 | * | |
1da177e4 LT |
916 | */ |
917 | struct inode *new_inode(struct super_block *sb) | |
918 | { | |
6b3304b5 | 919 | struct inode *inode; |
1da177e4 | 920 | |
74278da9 | 921 | spin_lock_prefetch(&sb->s_inode_list_lock); |
6b3304b5 | 922 | |
a209dfc7 ED |
923 | inode = new_inode_pseudo(sb); |
924 | if (inode) | |
55fa6091 | 925 | inode_sb_list_add(inode); |
1da177e4 LT |
926 | return inode; |
927 | } | |
1da177e4 LT |
928 | EXPORT_SYMBOL(new_inode); |
929 | ||
14358e6d | 930 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
e096d0c7 JB |
931 | void lockdep_annotate_inode_mutex_key(struct inode *inode) |
932 | { | |
a3314a0e | 933 | if (S_ISDIR(inode->i_mode)) { |
1e89a5e1 PZ |
934 | struct file_system_type *type = inode->i_sb->s_type; |
935 | ||
9a7aa12f | 936 | /* Set new key only if filesystem hasn't already changed it */ |
9902af79 | 937 | if (lockdep_match_class(&inode->i_rwsem, &type->i_mutex_key)) { |
9a7aa12f JK |
938 | /* |
939 | * ensure nobody is actually holding i_mutex | |
940 | */ | |
9902af79 AV |
941 | // mutex_destroy(&inode->i_mutex); |
942 | init_rwsem(&inode->i_rwsem); | |
943 | lockdep_set_class(&inode->i_rwsem, | |
9a7aa12f JK |
944 | &type->i_mutex_dir_key); |
945 | } | |
1e89a5e1 | 946 | } |
e096d0c7 JB |
947 | } |
948 | EXPORT_SYMBOL(lockdep_annotate_inode_mutex_key); | |
14358e6d | 949 | #endif |
e096d0c7 JB |
950 | |
951 | /** | |
952 | * unlock_new_inode - clear the I_NEW state and wake up any waiters | |
953 | * @inode: new inode to unlock | |
954 | * | |
955 | * Called when the inode is fully initialised to clear the new state of the | |
956 | * inode and wake up anyone waiting for the inode to finish initialisation. | |
957 | */ | |
958 | void unlock_new_inode(struct inode *inode) | |
959 | { | |
960 | lockdep_annotate_inode_mutex_key(inode); | |
250df6ed | 961 | spin_lock(&inode->i_lock); |
eaff8079 CH |
962 | WARN_ON(!(inode->i_state & I_NEW)); |
963 | inode->i_state &= ~I_NEW; | |
310fa7a3 | 964 | smp_mb(); |
250df6ed DC |
965 | wake_up_bit(&inode->i_state, __I_NEW); |
966 | spin_unlock(&inode->i_lock); | |
1da177e4 | 967 | } |
1da177e4 LT |
968 | EXPORT_SYMBOL(unlock_new_inode); |
969 | ||
375e289e BF |
970 | /** |
971 | * lock_two_nondirectories - take two i_mutexes on non-directory objects | |
4fd699ae BF |
972 | * |
973 | * Lock any non-NULL argument that is not a directory. | |
974 | * Zero, one or two objects may be locked by this function. | |
975 | * | |
375e289e BF |
976 | * @inode1: first inode to lock |
977 | * @inode2: second inode to lock | |
978 | */ | |
979 | void lock_two_nondirectories(struct inode *inode1, struct inode *inode2) | |
980 | { | |
4fd699ae BF |
981 | if (inode1 > inode2) |
982 | swap(inode1, inode2); | |
983 | ||
984 | if (inode1 && !S_ISDIR(inode1->i_mode)) | |
5955102c | 985 | inode_lock(inode1); |
4fd699ae | 986 | if (inode2 && !S_ISDIR(inode2->i_mode) && inode2 != inode1) |
5955102c | 987 | inode_lock_nested(inode2, I_MUTEX_NONDIR2); |
375e289e BF |
988 | } |
989 | EXPORT_SYMBOL(lock_two_nondirectories); | |
990 | ||
991 | /** | |
992 | * unlock_two_nondirectories - release locks from lock_two_nondirectories() | |
993 | * @inode1: first inode to unlock | |
994 | * @inode2: second inode to unlock | |
995 | */ | |
996 | void unlock_two_nondirectories(struct inode *inode1, struct inode *inode2) | |
997 | { | |
4fd699ae | 998 | if (inode1 && !S_ISDIR(inode1->i_mode)) |
5955102c | 999 | inode_unlock(inode1); |
4fd699ae | 1000 | if (inode2 && !S_ISDIR(inode2->i_mode) && inode2 != inode1) |
5955102c | 1001 | inode_unlock(inode2); |
375e289e BF |
1002 | } |
1003 | EXPORT_SYMBOL(unlock_two_nondirectories); | |
1004 | ||
80ea09a0 MS |
1005 | /** |
1006 | * inode_insert5 - obtain an inode from a mounted file system | |
1007 | * @inode: pre-allocated inode to use for insert to cache | |
1008 | * @hashval: hash value (usually inode number) to get | |
1009 | * @test: callback used for comparisons between inodes | |
1010 | * @set: callback used to initialize a new struct inode | |
1011 | * @data: opaque data pointer to pass to @test and @set | |
1012 | * | |
1013 | * Search for the inode specified by @hashval and @data in the inode cache, | |
1014 | * and if present it is return it with an increased reference count. This is | |
1015 | * a variant of iget5_locked() for callers that don't want to fail on memory | |
1016 | * allocation of inode. | |
1017 | * | |
1018 | * If the inode is not in cache, insert the pre-allocated inode to cache and | |
1019 | * return it locked, hashed, and with the I_NEW flag set. The file system gets | |
1020 | * to fill it in before unlocking it via unlock_new_inode(). | |
1021 | * | |
1022 | * Note both @test and @set are called with the inode_hash_lock held, so can't | |
1023 | * sleep. | |
1024 | */ | |
1025 | struct inode *inode_insert5(struct inode *inode, unsigned long hashval, | |
1026 | int (*test)(struct inode *, void *), | |
1027 | int (*set)(struct inode *, void *), void *data) | |
1028 | { | |
1029 | struct hlist_head *head = inode_hashtable + hash(inode->i_sb, hashval); | |
1030 | struct inode *old; | |
1031 | ||
1032 | again: | |
1033 | spin_lock(&inode_hash_lock); | |
1034 | old = find_inode(inode->i_sb, head, test, data); | |
1035 | if (unlikely(old)) { | |
1036 | /* | |
1037 | * Uhhuh, somebody else created the same inode under us. | |
1038 | * Use the old inode instead of the preallocated one. | |
1039 | */ | |
1040 | spin_unlock(&inode_hash_lock); | |
1041 | wait_on_inode(old); | |
1042 | if (unlikely(inode_unhashed(old))) { | |
1043 | iput(old); | |
1044 | goto again; | |
1045 | } | |
1046 | return old; | |
1047 | } | |
1048 | ||
1049 | if (set && unlikely(set(inode, data))) { | |
1050 | inode = NULL; | |
1051 | goto unlock; | |
1052 | } | |
1053 | ||
1054 | /* | |
1055 | * Return the locked inode with I_NEW set, the | |
1056 | * caller is responsible for filling in the contents | |
1057 | */ | |
1058 | spin_lock(&inode->i_lock); | |
1059 | inode->i_state |= I_NEW; | |
1060 | hlist_add_head(&inode->i_hash, head); | |
1061 | spin_unlock(&inode->i_lock); | |
1062 | unlock: | |
1063 | spin_unlock(&inode_hash_lock); | |
1064 | ||
1065 | return inode; | |
1066 | } | |
1067 | EXPORT_SYMBOL(inode_insert5); | |
1068 | ||
0b2d0724 CH |
1069 | /** |
1070 | * iget5_locked - obtain an inode from a mounted file system | |
1071 | * @sb: super block of file system | |
1072 | * @hashval: hash value (usually inode number) to get | |
1073 | * @test: callback used for comparisons between inodes | |
1074 | * @set: callback used to initialize a new struct inode | |
1075 | * @data: opaque data pointer to pass to @test and @set | |
1076 | * | |
1077 | * Search for the inode specified by @hashval and @data in the inode cache, | |
1078 | * and if present it is return it with an increased reference count. This is | |
1079 | * a generalized version of iget_locked() for file systems where the inode | |
1080 | * number is not sufficient for unique identification of an inode. | |
1081 | * | |
1082 | * If the inode is not in cache, allocate a new inode and return it locked, | |
1083 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
1084 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 1085 | * |
0b2d0724 CH |
1086 | * Note both @test and @set are called with the inode_hash_lock held, so can't |
1087 | * sleep. | |
1da177e4 | 1088 | */ |
0b2d0724 CH |
1089 | struct inode *iget5_locked(struct super_block *sb, unsigned long hashval, |
1090 | int (*test)(struct inode *, void *), | |
1091 | int (*set)(struct inode *, void *), void *data) | |
1da177e4 | 1092 | { |
80ea09a0 | 1093 | struct inode *inode = ilookup5(sb, hashval, test, data); |
0b2d0724 | 1094 | |
80ea09a0 MS |
1095 | if (!inode) { |
1096 | struct inode *new = new_inode(sb); | |
0b2d0724 | 1097 | |
80ea09a0 MS |
1098 | if (new) { |
1099 | inode = inode_insert5(new, hashval, test, set, data); | |
1100 | if (unlikely(inode != new)) | |
1101 | iput(new); | |
2864f301 | 1102 | } |
1da177e4 LT |
1103 | } |
1104 | return inode; | |
1da177e4 | 1105 | } |
0b2d0724 | 1106 | EXPORT_SYMBOL(iget5_locked); |
1da177e4 | 1107 | |
0b2d0724 CH |
1108 | /** |
1109 | * iget_locked - obtain an inode from a mounted file system | |
1110 | * @sb: super block of file system | |
1111 | * @ino: inode number to get | |
1112 | * | |
1113 | * Search for the inode specified by @ino in the inode cache and if present | |
1114 | * return it with an increased reference count. This is for file systems | |
1115 | * where the inode number is sufficient for unique identification of an inode. | |
1116 | * | |
1117 | * If the inode is not in cache, allocate a new inode and return it locked, | |
1118 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
1119 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 1120 | */ |
0b2d0724 | 1121 | struct inode *iget_locked(struct super_block *sb, unsigned long ino) |
1da177e4 | 1122 | { |
0b2d0724 | 1123 | struct hlist_head *head = inode_hashtable + hash(sb, ino); |
6b3304b5 | 1124 | struct inode *inode; |
2864f301 | 1125 | again: |
0b2d0724 CH |
1126 | spin_lock(&inode_hash_lock); |
1127 | inode = find_inode_fast(sb, head, ino); | |
1128 | spin_unlock(&inode_hash_lock); | |
1129 | if (inode) { | |
1130 | wait_on_inode(inode); | |
2864f301 AV |
1131 | if (unlikely(inode_unhashed(inode))) { |
1132 | iput(inode); | |
1133 | goto again; | |
1134 | } | |
0b2d0724 CH |
1135 | return inode; |
1136 | } | |
1137 | ||
1da177e4 LT |
1138 | inode = alloc_inode(sb); |
1139 | if (inode) { | |
6b3304b5 | 1140 | struct inode *old; |
1da177e4 | 1141 | |
67a23c49 | 1142 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1143 | /* We released the lock, so.. */ |
1144 | old = find_inode_fast(sb, head, ino); | |
1145 | if (!old) { | |
1146 | inode->i_ino = ino; | |
250df6ed DC |
1147 | spin_lock(&inode->i_lock); |
1148 | inode->i_state = I_NEW; | |
646ec461 | 1149 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1150 | spin_unlock(&inode->i_lock); |
55fa6091 | 1151 | inode_sb_list_add(inode); |
67a23c49 | 1152 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1153 | |
1154 | /* Return the locked inode with I_NEW set, the | |
1155 | * caller is responsible for filling in the contents | |
1156 | */ | |
1157 | return inode; | |
1158 | } | |
1159 | ||
1160 | /* | |
1161 | * Uhhuh, somebody else created the same inode under | |
1162 | * us. Use the old inode instead of the one we just | |
1163 | * allocated. | |
1164 | */ | |
67a23c49 | 1165 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1166 | destroy_inode(inode); |
1167 | inode = old; | |
1168 | wait_on_inode(inode); | |
2864f301 AV |
1169 | if (unlikely(inode_unhashed(inode))) { |
1170 | iput(inode); | |
1171 | goto again; | |
1172 | } | |
1da177e4 LT |
1173 | } |
1174 | return inode; | |
1175 | } | |
0b2d0724 | 1176 | EXPORT_SYMBOL(iget_locked); |
1da177e4 | 1177 | |
ad5e195a CH |
1178 | /* |
1179 | * search the inode cache for a matching inode number. | |
1180 | * If we find one, then the inode number we are trying to | |
1181 | * allocate is not unique and so we should not use it. | |
1182 | * | |
1183 | * Returns 1 if the inode number is unique, 0 if it is not. | |
1184 | */ | |
1185 | static int test_inode_iunique(struct super_block *sb, unsigned long ino) | |
1186 | { | |
1187 | struct hlist_head *b = inode_hashtable + hash(sb, ino); | |
ad5e195a CH |
1188 | struct inode *inode; |
1189 | ||
67a23c49 | 1190 | spin_lock(&inode_hash_lock); |
b67bfe0d | 1191 | hlist_for_each_entry(inode, b, i_hash) { |
67a23c49 DC |
1192 | if (inode->i_ino == ino && inode->i_sb == sb) { |
1193 | spin_unlock(&inode_hash_lock); | |
ad5e195a | 1194 | return 0; |
67a23c49 | 1195 | } |
ad5e195a | 1196 | } |
67a23c49 | 1197 | spin_unlock(&inode_hash_lock); |
ad5e195a CH |
1198 | |
1199 | return 1; | |
1200 | } | |
1201 | ||
1da177e4 LT |
1202 | /** |
1203 | * iunique - get a unique inode number | |
1204 | * @sb: superblock | |
1205 | * @max_reserved: highest reserved inode number | |
1206 | * | |
1207 | * Obtain an inode number that is unique on the system for a given | |
1208 | * superblock. This is used by file systems that have no natural | |
1209 | * permanent inode numbering system. An inode number is returned that | |
1210 | * is higher than the reserved limit but unique. | |
1211 | * | |
1212 | * BUGS: | |
1213 | * With a large number of inodes live on the file system this function | |
1214 | * currently becomes quite slow. | |
1215 | */ | |
1216 | ino_t iunique(struct super_block *sb, ino_t max_reserved) | |
1217 | { | |
866b04fc JL |
1218 | /* |
1219 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
1220 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
1221 | * here to attempt to avoid that. | |
1222 | */ | |
ad5e195a | 1223 | static DEFINE_SPINLOCK(iunique_lock); |
866b04fc | 1224 | static unsigned int counter; |
1da177e4 | 1225 | ino_t res; |
3361c7be | 1226 | |
ad5e195a | 1227 | spin_lock(&iunique_lock); |
3361c7be JL |
1228 | do { |
1229 | if (counter <= max_reserved) | |
1230 | counter = max_reserved + 1; | |
1da177e4 | 1231 | res = counter++; |
ad5e195a CH |
1232 | } while (!test_inode_iunique(sb, res)); |
1233 | spin_unlock(&iunique_lock); | |
1da177e4 | 1234 | |
3361c7be JL |
1235 | return res; |
1236 | } | |
1da177e4 LT |
1237 | EXPORT_SYMBOL(iunique); |
1238 | ||
1239 | struct inode *igrab(struct inode *inode) | |
1240 | { | |
250df6ed DC |
1241 | spin_lock(&inode->i_lock); |
1242 | if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) { | |
1da177e4 | 1243 | __iget(inode); |
250df6ed DC |
1244 | spin_unlock(&inode->i_lock); |
1245 | } else { | |
1246 | spin_unlock(&inode->i_lock); | |
1da177e4 LT |
1247 | /* |
1248 | * Handle the case where s_op->clear_inode is not been | |
1249 | * called yet, and somebody is calling igrab | |
1250 | * while the inode is getting freed. | |
1251 | */ | |
1252 | inode = NULL; | |
250df6ed | 1253 | } |
1da177e4 LT |
1254 | return inode; |
1255 | } | |
1da177e4 LT |
1256 | EXPORT_SYMBOL(igrab); |
1257 | ||
1258 | /** | |
0b2d0724 | 1259 | * ilookup5_nowait - search for an inode in the inode cache |
1da177e4 | 1260 | * @sb: super block of file system to search |
0b2d0724 | 1261 | * @hashval: hash value (usually inode number) to search for |
1da177e4 LT |
1262 | * @test: callback used for comparisons between inodes |
1263 | * @data: opaque data pointer to pass to @test | |
1da177e4 | 1264 | * |
0b2d0724 | 1265 | * Search for the inode specified by @hashval and @data in the inode cache. |
1da177e4 LT |
1266 | * If the inode is in the cache, the inode is returned with an incremented |
1267 | * reference count. | |
1268 | * | |
0b2d0724 CH |
1269 | * Note: I_NEW is not waited upon so you have to be very careful what you do |
1270 | * with the returned inode. You probably should be using ilookup5() instead. | |
1da177e4 | 1271 | * |
b6d0ad68 | 1272 | * Note2: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 | 1273 | */ |
0b2d0724 CH |
1274 | struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval, |
1275 | int (*test)(struct inode *, void *), void *data) | |
1da177e4 | 1276 | { |
0b2d0724 | 1277 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); |
1da177e4 LT |
1278 | struct inode *inode; |
1279 | ||
67a23c49 | 1280 | spin_lock(&inode_hash_lock); |
1da177e4 | 1281 | inode = find_inode(sb, head, test, data); |
67a23c49 | 1282 | spin_unlock(&inode_hash_lock); |
88bd5121 | 1283 | |
0b2d0724 | 1284 | return inode; |
88bd5121 | 1285 | } |
88bd5121 AA |
1286 | EXPORT_SYMBOL(ilookup5_nowait); |
1287 | ||
1288 | /** | |
1289 | * ilookup5 - search for an inode in the inode cache | |
1290 | * @sb: super block of file system to search | |
1291 | * @hashval: hash value (usually inode number) to search for | |
1292 | * @test: callback used for comparisons between inodes | |
1293 | * @data: opaque data pointer to pass to @test | |
1294 | * | |
0b2d0724 CH |
1295 | * Search for the inode specified by @hashval and @data in the inode cache, |
1296 | * and if the inode is in the cache, return the inode with an incremented | |
1297 | * reference count. Waits on I_NEW before returning the inode. | |
88bd5121 | 1298 | * returned with an incremented reference count. |
1da177e4 | 1299 | * |
0b2d0724 CH |
1300 | * This is a generalized version of ilookup() for file systems where the |
1301 | * inode number is not sufficient for unique identification of an inode. | |
1da177e4 | 1302 | * |
0b2d0724 | 1303 | * Note: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 LT |
1304 | */ |
1305 | struct inode *ilookup5(struct super_block *sb, unsigned long hashval, | |
1306 | int (*test)(struct inode *, void *), void *data) | |
1307 | { | |
2864f301 AV |
1308 | struct inode *inode; |
1309 | again: | |
1310 | inode = ilookup5_nowait(sb, hashval, test, data); | |
1311 | if (inode) { | |
0b2d0724 | 1312 | wait_on_inode(inode); |
2864f301 AV |
1313 | if (unlikely(inode_unhashed(inode))) { |
1314 | iput(inode); | |
1315 | goto again; | |
1316 | } | |
1317 | } | |
0b2d0724 | 1318 | return inode; |
1da177e4 | 1319 | } |
1da177e4 LT |
1320 | EXPORT_SYMBOL(ilookup5); |
1321 | ||
1322 | /** | |
1323 | * ilookup - search for an inode in the inode cache | |
1324 | * @sb: super block of file system to search | |
1325 | * @ino: inode number to search for | |
1326 | * | |
0b2d0724 CH |
1327 | * Search for the inode @ino in the inode cache, and if the inode is in the |
1328 | * cache, the inode is returned with an incremented reference count. | |
1da177e4 LT |
1329 | */ |
1330 | struct inode *ilookup(struct super_block *sb, unsigned long ino) | |
1331 | { | |
1332 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1da177e4 | 1333 | struct inode *inode; |
2864f301 | 1334 | again: |
0b2d0724 CH |
1335 | spin_lock(&inode_hash_lock); |
1336 | inode = find_inode_fast(sb, head, ino); | |
1337 | spin_unlock(&inode_hash_lock); | |
1da177e4 | 1338 | |
2864f301 | 1339 | if (inode) { |
0b2d0724 | 1340 | wait_on_inode(inode); |
2864f301 AV |
1341 | if (unlikely(inode_unhashed(inode))) { |
1342 | iput(inode); | |
1343 | goto again; | |
1344 | } | |
1345 | } | |
0b2d0724 | 1346 | return inode; |
1da177e4 | 1347 | } |
0b2d0724 | 1348 | EXPORT_SYMBOL(ilookup); |
1da177e4 | 1349 | |
fe032c42 TT |
1350 | /** |
1351 | * find_inode_nowait - find an inode in the inode cache | |
1352 | * @sb: super block of file system to search | |
1353 | * @hashval: hash value (usually inode number) to search for | |
1354 | * @match: callback used for comparisons between inodes | |
1355 | * @data: opaque data pointer to pass to @match | |
1356 | * | |
1357 | * Search for the inode specified by @hashval and @data in the inode | |
1358 | * cache, where the helper function @match will return 0 if the inode | |
1359 | * does not match, 1 if the inode does match, and -1 if the search | |
1360 | * should be stopped. The @match function must be responsible for | |
1361 | * taking the i_lock spin_lock and checking i_state for an inode being | |
1362 | * freed or being initialized, and incrementing the reference count | |
1363 | * before returning 1. It also must not sleep, since it is called with | |
1364 | * the inode_hash_lock spinlock held. | |
1365 | * | |
1366 | * This is a even more generalized version of ilookup5() when the | |
1367 | * function must never block --- find_inode() can block in | |
1368 | * __wait_on_freeing_inode() --- or when the caller can not increment | |
1369 | * the reference count because the resulting iput() might cause an | |
1370 | * inode eviction. The tradeoff is that the @match funtion must be | |
1371 | * very carefully implemented. | |
1372 | */ | |
1373 | struct inode *find_inode_nowait(struct super_block *sb, | |
1374 | unsigned long hashval, | |
1375 | int (*match)(struct inode *, unsigned long, | |
1376 | void *), | |
1377 | void *data) | |
1378 | { | |
1379 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1380 | struct inode *inode, *ret_inode = NULL; | |
1381 | int mval; | |
1382 | ||
1383 | spin_lock(&inode_hash_lock); | |
1384 | hlist_for_each_entry(inode, head, i_hash) { | |
1385 | if (inode->i_sb != sb) | |
1386 | continue; | |
1387 | mval = match(inode, hashval, data); | |
1388 | if (mval == 0) | |
1389 | continue; | |
1390 | if (mval == 1) | |
1391 | ret_inode = inode; | |
1392 | goto out; | |
1393 | } | |
1394 | out: | |
1395 | spin_unlock(&inode_hash_lock); | |
1396 | return ret_inode; | |
1397 | } | |
1398 | EXPORT_SYMBOL(find_inode_nowait); | |
1399 | ||
261bca86 AV |
1400 | int insert_inode_locked(struct inode *inode) |
1401 | { | |
1402 | struct super_block *sb = inode->i_sb; | |
1403 | ino_t ino = inode->i_ino; | |
1404 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
261bca86 | 1405 | |
261bca86 | 1406 | while (1) { |
72a43d63 | 1407 | struct inode *old = NULL; |
67a23c49 | 1408 | spin_lock(&inode_hash_lock); |
b67bfe0d | 1409 | hlist_for_each_entry(old, head, i_hash) { |
72a43d63 AV |
1410 | if (old->i_ino != ino) |
1411 | continue; | |
1412 | if (old->i_sb != sb) | |
1413 | continue; | |
250df6ed DC |
1414 | spin_lock(&old->i_lock); |
1415 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1416 | spin_unlock(&old->i_lock); | |
72a43d63 | 1417 | continue; |
250df6ed | 1418 | } |
72a43d63 AV |
1419 | break; |
1420 | } | |
b67bfe0d | 1421 | if (likely(!old)) { |
250df6ed DC |
1422 | spin_lock(&inode->i_lock); |
1423 | inode->i_state |= I_NEW; | |
261bca86 | 1424 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1425 | spin_unlock(&inode->i_lock); |
67a23c49 | 1426 | spin_unlock(&inode_hash_lock); |
261bca86 AV |
1427 | return 0; |
1428 | } | |
1429 | __iget(old); | |
250df6ed | 1430 | spin_unlock(&old->i_lock); |
67a23c49 | 1431 | spin_unlock(&inode_hash_lock); |
261bca86 | 1432 | wait_on_inode(old); |
1d3382cb | 1433 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1434 | iput(old); |
1435 | return -EBUSY; | |
1436 | } | |
1437 | iput(old); | |
1438 | } | |
1439 | } | |
261bca86 AV |
1440 | EXPORT_SYMBOL(insert_inode_locked); |
1441 | ||
1442 | int insert_inode_locked4(struct inode *inode, unsigned long hashval, | |
1443 | int (*test)(struct inode *, void *), void *data) | |
1444 | { | |
80ea09a0 | 1445 | struct inode *old = inode_insert5(inode, hashval, test, NULL, data); |
261bca86 | 1446 | |
80ea09a0 | 1447 | if (old != inode) { |
261bca86 | 1448 | iput(old); |
80ea09a0 | 1449 | return -EBUSY; |
261bca86 | 1450 | } |
80ea09a0 | 1451 | return 0; |
261bca86 | 1452 | } |
261bca86 AV |
1453 | EXPORT_SYMBOL(insert_inode_locked4); |
1454 | ||
1da177e4 | 1455 | |
45321ac5 AV |
1456 | int generic_delete_inode(struct inode *inode) |
1457 | { | |
1458 | return 1; | |
1459 | } | |
1460 | EXPORT_SYMBOL(generic_delete_inode); | |
1461 | ||
45321ac5 AV |
1462 | /* |
1463 | * Called when we're dropping the last reference | |
1464 | * to an inode. | |
22fe4042 | 1465 | * |
45321ac5 AV |
1466 | * Call the FS "drop_inode()" function, defaulting to |
1467 | * the legacy UNIX filesystem behaviour. If it tells | |
1468 | * us to evict inode, do so. Otherwise, retain inode | |
1469 | * in cache if fs is alive, sync and evict if fs is | |
1470 | * shutting down. | |
22fe4042 | 1471 | */ |
45321ac5 | 1472 | static void iput_final(struct inode *inode) |
1da177e4 LT |
1473 | { |
1474 | struct super_block *sb = inode->i_sb; | |
45321ac5 AV |
1475 | const struct super_operations *op = inode->i_sb->s_op; |
1476 | int drop; | |
1477 | ||
250df6ed DC |
1478 | WARN_ON(inode->i_state & I_NEW); |
1479 | ||
e7f59097 | 1480 | if (op->drop_inode) |
45321ac5 AV |
1481 | drop = op->drop_inode(inode); |
1482 | else | |
1483 | drop = generic_drop_inode(inode); | |
1da177e4 | 1484 | |
1751e8a6 | 1485 | if (!drop && (sb->s_flags & SB_ACTIVE)) { |
4eff96dd | 1486 | inode_add_lru(inode); |
b2b2af8e | 1487 | spin_unlock(&inode->i_lock); |
b2b2af8e DC |
1488 | return; |
1489 | } | |
1490 | ||
45321ac5 | 1491 | if (!drop) { |
991114c6 | 1492 | inode->i_state |= I_WILL_FREE; |
250df6ed | 1493 | spin_unlock(&inode->i_lock); |
1da177e4 | 1494 | write_inode_now(inode, 1); |
250df6ed | 1495 | spin_lock(&inode->i_lock); |
7ef0d737 | 1496 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1497 | inode->i_state &= ~I_WILL_FREE; |
1da177e4 | 1498 | } |
7ccf19a8 | 1499 | |
991114c6 | 1500 | inode->i_state |= I_FREEING; |
c4ae0c65 ED |
1501 | if (!list_empty(&inode->i_lru)) |
1502 | inode_lru_list_del(inode); | |
b2b2af8e | 1503 | spin_unlock(&inode->i_lock); |
b2b2af8e | 1504 | |
644da596 | 1505 | evict(inode); |
1da177e4 LT |
1506 | } |
1507 | ||
1da177e4 | 1508 | /** |
6b3304b5 | 1509 | * iput - put an inode |
1da177e4 LT |
1510 | * @inode: inode to put |
1511 | * | |
1512 | * Puts an inode, dropping its usage count. If the inode use count hits | |
1513 | * zero, the inode is then freed and may also be destroyed. | |
1514 | * | |
1515 | * Consequently, iput() can sleep. | |
1516 | */ | |
1517 | void iput(struct inode *inode) | |
1518 | { | |
0ae45f63 TT |
1519 | if (!inode) |
1520 | return; | |
1521 | BUG_ON(inode->i_state & I_CLEAR); | |
1522 | retry: | |
1523 | if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock)) { | |
1524 | if (inode->i_nlink && (inode->i_state & I_DIRTY_TIME)) { | |
1525 | atomic_inc(&inode->i_count); | |
0ae45f63 TT |
1526 | spin_unlock(&inode->i_lock); |
1527 | trace_writeback_lazytime_iput(inode); | |
1528 | mark_inode_dirty_sync(inode); | |
1529 | goto retry; | |
1530 | } | |
1531 | iput_final(inode); | |
1da177e4 LT |
1532 | } |
1533 | } | |
1da177e4 LT |
1534 | EXPORT_SYMBOL(iput); |
1535 | ||
1536 | /** | |
1537 | * bmap - find a block number in a file | |
1538 | * @inode: inode of file | |
1539 | * @block: block to find | |
1540 | * | |
1541 | * Returns the block number on the device holding the inode that | |
1542 | * is the disk block number for the block of the file requested. | |
1543 | * That is, asked for block 4 of inode 1 the function will return the | |
6b3304b5 | 1544 | * disk block relative to the disk start that holds that block of the |
1da177e4 LT |
1545 | * file. |
1546 | */ | |
6b3304b5 | 1547 | sector_t bmap(struct inode *inode, sector_t block) |
1da177e4 LT |
1548 | { |
1549 | sector_t res = 0; | |
1550 | if (inode->i_mapping->a_ops->bmap) | |
1551 | res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block); | |
1552 | return res; | |
1553 | } | |
1da177e4 LT |
1554 | EXPORT_SYMBOL(bmap); |
1555 | ||
598e3c8f MS |
1556 | /* |
1557 | * Update times in overlayed inode from underlying real inode | |
1558 | */ | |
1559 | static void update_ovl_inode_times(struct dentry *dentry, struct inode *inode, | |
1560 | bool rcu) | |
1561 | { | |
cd91304e | 1562 | struct dentry *upperdentry; |
598e3c8f | 1563 | |
cd91304e MS |
1564 | /* |
1565 | * Nothing to do if in rcu or if non-overlayfs | |
1566 | */ | |
1567 | if (rcu || likely(!(dentry->d_flags & DCACHE_OP_REAL))) | |
1568 | return; | |
1569 | ||
1570 | upperdentry = d_real(dentry, NULL, 0, D_REAL_UPPER); | |
1571 | ||
1572 | /* | |
1573 | * If file is on lower then we can't update atime, so no worries about | |
1574 | * stale mtime/ctime. | |
1575 | */ | |
1576 | if (upperdentry) { | |
1577 | struct inode *realinode = d_inode(upperdentry); | |
1578 | ||
1579 | if ((!timespec_equal(&inode->i_mtime, &realinode->i_mtime) || | |
598e3c8f MS |
1580 | !timespec_equal(&inode->i_ctime, &realinode->i_ctime))) { |
1581 | inode->i_mtime = realinode->i_mtime; | |
1582 | inode->i_ctime = realinode->i_ctime; | |
1583 | } | |
1584 | } | |
1585 | } | |
1586 | ||
11ff6f05 MG |
1587 | /* |
1588 | * With relative atime, only update atime if the previous atime is | |
1589 | * earlier than either the ctime or mtime or if at least a day has | |
1590 | * passed since the last atime update. | |
1591 | */ | |
598e3c8f MS |
1592 | static int relatime_need_update(const struct path *path, struct inode *inode, |
1593 | struct timespec now, bool rcu) | |
11ff6f05 MG |
1594 | { |
1595 | ||
598e3c8f | 1596 | if (!(path->mnt->mnt_flags & MNT_RELATIME)) |
11ff6f05 | 1597 | return 1; |
598e3c8f MS |
1598 | |
1599 | update_ovl_inode_times(path->dentry, inode, rcu); | |
11ff6f05 MG |
1600 | /* |
1601 | * Is mtime younger than atime? If yes, update atime: | |
1602 | */ | |
1603 | if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0) | |
1604 | return 1; | |
1605 | /* | |
1606 | * Is ctime younger than atime? If yes, update atime: | |
1607 | */ | |
1608 | if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0) | |
1609 | return 1; | |
1610 | ||
1611 | /* | |
1612 | * Is the previous atime value older than a day? If yes, | |
1613 | * update atime: | |
1614 | */ | |
1615 | if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60) | |
1616 | return 1; | |
1617 | /* | |
1618 | * Good, we can skip the atime update: | |
1619 | */ | |
1620 | return 0; | |
1621 | } | |
1622 | ||
0ae45f63 | 1623 | int generic_update_time(struct inode *inode, struct timespec *time, int flags) |
c3b2da31 | 1624 | { |
0ae45f63 | 1625 | int iflags = I_DIRTY_TIME; |
e38cf302 | 1626 | bool dirty = false; |
c3b2da31 JB |
1627 | |
1628 | if (flags & S_ATIME) | |
1629 | inode->i_atime = *time; | |
1630 | if (flags & S_VERSION) | |
e38cf302 | 1631 | dirty = inode_maybe_inc_iversion(inode, false); |
c3b2da31 JB |
1632 | if (flags & S_CTIME) |
1633 | inode->i_ctime = *time; | |
1634 | if (flags & S_MTIME) | |
1635 | inode->i_mtime = *time; | |
e38cf302 JL |
1636 | if ((flags & (S_ATIME | S_CTIME | S_MTIME)) && |
1637 | !(inode->i_sb->s_flags & SB_LAZYTIME)) | |
1638 | dirty = true; | |
0ae45f63 | 1639 | |
e38cf302 | 1640 | if (dirty) |
0ae45f63 TT |
1641 | iflags |= I_DIRTY_SYNC; |
1642 | __mark_inode_dirty(inode, iflags); | |
c3b2da31 JB |
1643 | return 0; |
1644 | } | |
0ae45f63 TT |
1645 | EXPORT_SYMBOL(generic_update_time); |
1646 | ||
1647 | /* | |
1648 | * This does the actual work of updating an inodes time or version. Must have | |
1649 | * had called mnt_want_write() before calling this. | |
1650 | */ | |
1651 | static int update_time(struct inode *inode, struct timespec *time, int flags) | |
1652 | { | |
1653 | int (*update_time)(struct inode *, struct timespec *, int); | |
1654 | ||
1655 | update_time = inode->i_op->update_time ? inode->i_op->update_time : | |
1656 | generic_update_time; | |
1657 | ||
1658 | return update_time(inode, time, flags); | |
1659 | } | |
c3b2da31 | 1660 | |
1da177e4 | 1661 | /** |
869243a0 | 1662 | * touch_atime - update the access time |
185553b2 | 1663 | * @path: the &struct path to update |
30fdc8ee | 1664 | * @inode: inode to update |
1da177e4 LT |
1665 | * |
1666 | * Update the accessed time on an inode and mark it for writeback. | |
1667 | * This function automatically handles read only file systems and media, | |
1668 | * as well as the "noatime" flag and inode specific "noatime" markers. | |
1669 | */ | |
598e3c8f MS |
1670 | bool __atime_needs_update(const struct path *path, struct inode *inode, |
1671 | bool rcu) | |
1da177e4 | 1672 | { |
68ac1234 | 1673 | struct vfsmount *mnt = path->mnt; |
1da177e4 LT |
1674 | struct timespec now; |
1675 | ||
cdb70f3f | 1676 | if (inode->i_flags & S_NOATIME) |
8fa9dd24 | 1677 | return false; |
0bd23d09 EB |
1678 | |
1679 | /* Atime updates will likely cause i_uid and i_gid to be written | |
1680 | * back improprely if their true value is unknown to the vfs. | |
1681 | */ | |
1682 | if (HAS_UNMAPPED_ID(inode)) | |
1683 | return false; | |
1684 | ||
37756ced | 1685 | if (IS_NOATIME(inode)) |
8fa9dd24 | 1686 | return false; |
1751e8a6 | 1687 | if ((inode->i_sb->s_flags & SB_NODIRATIME) && S_ISDIR(inode->i_mode)) |
8fa9dd24 | 1688 | return false; |
47ae32d6 | 1689 | |
cdb70f3f | 1690 | if (mnt->mnt_flags & MNT_NOATIME) |
8fa9dd24 | 1691 | return false; |
cdb70f3f | 1692 | if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) |
8fa9dd24 | 1693 | return false; |
1da177e4 | 1694 | |
c2050a45 | 1695 | now = current_time(inode); |
11ff6f05 | 1696 | |
598e3c8f | 1697 | if (!relatime_need_update(path, inode, now, rcu)) |
8fa9dd24 | 1698 | return false; |
11ff6f05 | 1699 | |
47ae32d6 | 1700 | if (timespec_equal(&inode->i_atime, &now)) |
8fa9dd24 N |
1701 | return false; |
1702 | ||
1703 | return true; | |
1704 | } | |
1705 | ||
1706 | void touch_atime(const struct path *path) | |
1707 | { | |
1708 | struct vfsmount *mnt = path->mnt; | |
1709 | struct inode *inode = d_inode(path->dentry); | |
1710 | struct timespec now; | |
1711 | ||
598e3c8f | 1712 | if (!__atime_needs_update(path, inode, false)) |
b12536c2 AK |
1713 | return; |
1714 | ||
5d37e9e6 | 1715 | if (!sb_start_write_trylock(inode->i_sb)) |
b12536c2 | 1716 | return; |
47ae32d6 | 1717 | |
8fa9dd24 | 1718 | if (__mnt_want_write(mnt) != 0) |
5d37e9e6 | 1719 | goto skip_update; |
c3b2da31 JB |
1720 | /* |
1721 | * File systems can error out when updating inodes if they need to | |
1722 | * allocate new space to modify an inode (such is the case for | |
1723 | * Btrfs), but since we touch atime while walking down the path we | |
1724 | * really don't care if we failed to update the atime of the file, | |
1725 | * so just ignore the return value. | |
2bc55652 AB |
1726 | * We may also fail on filesystems that have the ability to make parts |
1727 | * of the fs read only, e.g. subvolumes in Btrfs. | |
c3b2da31 | 1728 | */ |
c2050a45 | 1729 | now = current_time(inode); |
c3b2da31 | 1730 | update_time(inode, &now, S_ATIME); |
5d37e9e6 JK |
1731 | __mnt_drop_write(mnt); |
1732 | skip_update: | |
1733 | sb_end_write(inode->i_sb); | |
1da177e4 | 1734 | } |
869243a0 | 1735 | EXPORT_SYMBOL(touch_atime); |
1da177e4 | 1736 | |
3ed37648 CW |
1737 | /* |
1738 | * The logic we want is | |
1739 | * | |
1740 | * if suid or (sgid and xgrp) | |
1741 | * remove privs | |
1742 | */ | |
1743 | int should_remove_suid(struct dentry *dentry) | |
1744 | { | |
df2b1afd | 1745 | umode_t mode = d_inode(dentry)->i_mode; |
3ed37648 CW |
1746 | int kill = 0; |
1747 | ||
1748 | /* suid always must be killed */ | |
1749 | if (unlikely(mode & S_ISUID)) | |
1750 | kill = ATTR_KILL_SUID; | |
1751 | ||
1752 | /* | |
1753 | * sgid without any exec bits is just a mandatory locking mark; leave | |
1754 | * it alone. If some exec bits are set, it's a real sgid; kill it. | |
1755 | */ | |
1756 | if (unlikely((mode & S_ISGID) && (mode & S_IXGRP))) | |
1757 | kill |= ATTR_KILL_SGID; | |
1758 | ||
1759 | if (unlikely(kill && !capable(CAP_FSETID) && S_ISREG(mode))) | |
1760 | return kill; | |
1761 | ||
1762 | return 0; | |
1763 | } | |
1764 | EXPORT_SYMBOL(should_remove_suid); | |
1765 | ||
dbfae0cd JK |
1766 | /* |
1767 | * Return mask of changes for notify_change() that need to be done as a | |
1768 | * response to write or truncate. Return 0 if nothing has to be changed. | |
1769 | * Negative value on error (change should be denied). | |
1770 | */ | |
45f147a1 | 1771 | int dentry_needs_remove_privs(struct dentry *dentry) |
dbfae0cd | 1772 | { |
dbfae0cd JK |
1773 | struct inode *inode = d_inode(dentry); |
1774 | int mask = 0; | |
1775 | int ret; | |
1776 | ||
1777 | if (IS_NOSEC(inode)) | |
1778 | return 0; | |
1779 | ||
1780 | mask = should_remove_suid(dentry); | |
1781 | ret = security_inode_need_killpriv(dentry); | |
1782 | if (ret < 0) | |
1783 | return ret; | |
1784 | if (ret) | |
1785 | mask |= ATTR_KILL_PRIV; | |
1786 | return mask; | |
1787 | } | |
dbfae0cd JK |
1788 | |
1789 | static int __remove_privs(struct dentry *dentry, int kill) | |
3ed37648 CW |
1790 | { |
1791 | struct iattr newattrs; | |
1792 | ||
1793 | newattrs.ia_valid = ATTR_FORCE | kill; | |
27ac0ffe BF |
1794 | /* |
1795 | * Note we call this on write, so notify_change will not | |
1796 | * encounter any conflicting delegations: | |
1797 | */ | |
1798 | return notify_change(dentry, &newattrs, NULL); | |
3ed37648 CW |
1799 | } |
1800 | ||
5fa8e0a1 JK |
1801 | /* |
1802 | * Remove special file priviledges (suid, capabilities) when file is written | |
1803 | * to or truncated. | |
1804 | */ | |
1805 | int file_remove_privs(struct file *file) | |
3ed37648 | 1806 | { |
c1892c37 MS |
1807 | struct dentry *dentry = file_dentry(file); |
1808 | struct inode *inode = file_inode(file); | |
dbfae0cd | 1809 | int kill; |
3ed37648 CW |
1810 | int error = 0; |
1811 | ||
1812 | /* Fast path for nothing security related */ | |
1813 | if (IS_NOSEC(inode)) | |
1814 | return 0; | |
1815 | ||
c1892c37 | 1816 | kill = dentry_needs_remove_privs(dentry); |
dbfae0cd JK |
1817 | if (kill < 0) |
1818 | return kill; | |
1819 | if (kill) | |
1820 | error = __remove_privs(dentry, kill); | |
2426f391 JK |
1821 | if (!error) |
1822 | inode_has_no_xattr(inode); | |
3ed37648 CW |
1823 | |
1824 | return error; | |
1825 | } | |
5fa8e0a1 | 1826 | EXPORT_SYMBOL(file_remove_privs); |
3ed37648 | 1827 | |
1da177e4 | 1828 | /** |
870f4817 CH |
1829 | * file_update_time - update mtime and ctime time |
1830 | * @file: file accessed | |
1da177e4 | 1831 | * |
870f4817 CH |
1832 | * Update the mtime and ctime members of an inode and mark the inode |
1833 | * for writeback. Note that this function is meant exclusively for | |
1834 | * usage in the file write path of filesystems, and filesystems may | |
1835 | * choose to explicitly ignore update via this function with the | |
2eadfc0e | 1836 | * S_NOCMTIME inode flag, e.g. for network filesystem where these |
c3b2da31 JB |
1837 | * timestamps are handled by the server. This can return an error for |
1838 | * file systems who need to allocate space in order to update an inode. | |
1da177e4 LT |
1839 | */ |
1840 | ||
c3b2da31 | 1841 | int file_update_time(struct file *file) |
1da177e4 | 1842 | { |
496ad9aa | 1843 | struct inode *inode = file_inode(file); |
1da177e4 | 1844 | struct timespec now; |
c3b2da31 JB |
1845 | int sync_it = 0; |
1846 | int ret; | |
1da177e4 | 1847 | |
ce06e0b2 | 1848 | /* First try to exhaust all avenues to not sync */ |
1da177e4 | 1849 | if (IS_NOCMTIME(inode)) |
c3b2da31 | 1850 | return 0; |
20ddee2c | 1851 | |
c2050a45 | 1852 | now = current_time(inode); |
ce06e0b2 AK |
1853 | if (!timespec_equal(&inode->i_mtime, &now)) |
1854 | sync_it = S_MTIME; | |
1da177e4 | 1855 | |
ce06e0b2 AK |
1856 | if (!timespec_equal(&inode->i_ctime, &now)) |
1857 | sync_it |= S_CTIME; | |
870f4817 | 1858 | |
e38cf302 | 1859 | if (IS_I_VERSION(inode) && inode_iversion_need_inc(inode)) |
ce06e0b2 | 1860 | sync_it |= S_VERSION; |
7a224228 | 1861 | |
ce06e0b2 | 1862 | if (!sync_it) |
c3b2da31 | 1863 | return 0; |
ce06e0b2 AK |
1864 | |
1865 | /* Finally allowed to write? Takes lock. */ | |
eb04c282 | 1866 | if (__mnt_want_write_file(file)) |
c3b2da31 | 1867 | return 0; |
ce06e0b2 | 1868 | |
c3b2da31 | 1869 | ret = update_time(inode, &now, sync_it); |
eb04c282 | 1870 | __mnt_drop_write_file(file); |
c3b2da31 JB |
1871 | |
1872 | return ret; | |
1da177e4 | 1873 | } |
870f4817 | 1874 | EXPORT_SYMBOL(file_update_time); |
1da177e4 LT |
1875 | |
1876 | int inode_needs_sync(struct inode *inode) | |
1877 | { | |
1878 | if (IS_SYNC(inode)) | |
1879 | return 1; | |
1880 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) | |
1881 | return 1; | |
1882 | return 0; | |
1883 | } | |
1da177e4 LT |
1884 | EXPORT_SYMBOL(inode_needs_sync); |
1885 | ||
1da177e4 | 1886 | /* |
168a9fd6 MS |
1887 | * If we try to find an inode in the inode hash while it is being |
1888 | * deleted, we have to wait until the filesystem completes its | |
1889 | * deletion before reporting that it isn't found. This function waits | |
1890 | * until the deletion _might_ have completed. Callers are responsible | |
1891 | * to recheck inode state. | |
1892 | * | |
eaff8079 | 1893 | * It doesn't matter if I_NEW is not set initially, a call to |
250df6ed DC |
1894 | * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list |
1895 | * will DTRT. | |
1da177e4 LT |
1896 | */ |
1897 | static void __wait_on_freeing_inode(struct inode *inode) | |
1898 | { | |
1899 | wait_queue_head_t *wq; | |
eaff8079 CH |
1900 | DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW); |
1901 | wq = bit_waitqueue(&inode->i_state, __I_NEW); | |
21417136 | 1902 | prepare_to_wait(wq, &wait.wq_entry, TASK_UNINTERRUPTIBLE); |
250df6ed | 1903 | spin_unlock(&inode->i_lock); |
67a23c49 | 1904 | spin_unlock(&inode_hash_lock); |
1da177e4 | 1905 | schedule(); |
21417136 | 1906 | finish_wait(wq, &wait.wq_entry); |
67a23c49 | 1907 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1908 | } |
1909 | ||
1da177e4 LT |
1910 | static __initdata unsigned long ihash_entries; |
1911 | static int __init set_ihash_entries(char *str) | |
1912 | { | |
1913 | if (!str) | |
1914 | return 0; | |
1915 | ihash_entries = simple_strtoul(str, &str, 0); | |
1916 | return 1; | |
1917 | } | |
1918 | __setup("ihash_entries=", set_ihash_entries); | |
1919 | ||
1920 | /* | |
1921 | * Initialize the waitqueues and inode hash table. | |
1922 | */ | |
1923 | void __init inode_init_early(void) | |
1924 | { | |
1da177e4 LT |
1925 | /* If hashes are distributed across NUMA nodes, defer |
1926 | * hash allocation until vmalloc space is available. | |
1927 | */ | |
1928 | if (hashdist) | |
1929 | return; | |
1930 | ||
1931 | inode_hashtable = | |
1932 | alloc_large_system_hash("Inode-cache", | |
1933 | sizeof(struct hlist_head), | |
1934 | ihash_entries, | |
1935 | 14, | |
3d375d78 | 1936 | HASH_EARLY | HASH_ZERO, |
1da177e4 LT |
1937 | &i_hash_shift, |
1938 | &i_hash_mask, | |
31fe62b9 | 1939 | 0, |
1da177e4 | 1940 | 0); |
1da177e4 LT |
1941 | } |
1942 | ||
74bf17cf | 1943 | void __init inode_init(void) |
1da177e4 | 1944 | { |
1da177e4 | 1945 | /* inode slab cache */ |
b0196009 PJ |
1946 | inode_cachep = kmem_cache_create("inode_cache", |
1947 | sizeof(struct inode), | |
1948 | 0, | |
1949 | (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC| | |
5d097056 | 1950 | SLAB_MEM_SPREAD|SLAB_ACCOUNT), |
20c2df83 | 1951 | init_once); |
1da177e4 LT |
1952 | |
1953 | /* Hash may have been set up in inode_init_early */ | |
1954 | if (!hashdist) | |
1955 | return; | |
1956 | ||
1957 | inode_hashtable = | |
1958 | alloc_large_system_hash("Inode-cache", | |
1959 | sizeof(struct hlist_head), | |
1960 | ihash_entries, | |
1961 | 14, | |
3d375d78 | 1962 | HASH_ZERO, |
1da177e4 LT |
1963 | &i_hash_shift, |
1964 | &i_hash_mask, | |
31fe62b9 | 1965 | 0, |
1da177e4 | 1966 | 0); |
1da177e4 LT |
1967 | } |
1968 | ||
1969 | void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev) | |
1970 | { | |
1971 | inode->i_mode = mode; | |
1972 | if (S_ISCHR(mode)) { | |
1973 | inode->i_fop = &def_chr_fops; | |
1974 | inode->i_rdev = rdev; | |
1975 | } else if (S_ISBLK(mode)) { | |
1976 | inode->i_fop = &def_blk_fops; | |
1977 | inode->i_rdev = rdev; | |
1978 | } else if (S_ISFIFO(mode)) | |
599a0ac1 | 1979 | inode->i_fop = &pipefifo_fops; |
1da177e4 | 1980 | else if (S_ISSOCK(mode)) |
bd9b51e7 | 1981 | ; /* leave it no_open_fops */ |
1da177e4 | 1982 | else |
af0d9ae8 MK |
1983 | printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for" |
1984 | " inode %s:%lu\n", mode, inode->i_sb->s_id, | |
1985 | inode->i_ino); | |
1da177e4 LT |
1986 | } |
1987 | EXPORT_SYMBOL(init_special_inode); | |
a1bd120d DM |
1988 | |
1989 | /** | |
eaae668d | 1990 | * inode_init_owner - Init uid,gid,mode for new inode according to posix standards |
a1bd120d DM |
1991 | * @inode: New inode |
1992 | * @dir: Directory inode | |
1993 | * @mode: mode of the new inode | |
1994 | */ | |
1995 | void inode_init_owner(struct inode *inode, const struct inode *dir, | |
62bb1091 | 1996 | umode_t mode) |
a1bd120d DM |
1997 | { |
1998 | inode->i_uid = current_fsuid(); | |
1999 | if (dir && dir->i_mode & S_ISGID) { | |
2000 | inode->i_gid = dir->i_gid; | |
2001 | if (S_ISDIR(mode)) | |
2002 | mode |= S_ISGID; | |
2003 | } else | |
2004 | inode->i_gid = current_fsgid(); | |
2005 | inode->i_mode = mode; | |
2006 | } | |
2007 | EXPORT_SYMBOL(inode_init_owner); | |
e795b717 | 2008 | |
2e149670 SH |
2009 | /** |
2010 | * inode_owner_or_capable - check current task permissions to inode | |
2011 | * @inode: inode being checked | |
2012 | * | |
23adbe12 AL |
2013 | * Return true if current either has CAP_FOWNER in a namespace with the |
2014 | * inode owner uid mapped, or owns the file. | |
e795b717 | 2015 | */ |
2e149670 | 2016 | bool inode_owner_or_capable(const struct inode *inode) |
e795b717 | 2017 | { |
23adbe12 AL |
2018 | struct user_namespace *ns; |
2019 | ||
92361636 | 2020 | if (uid_eq(current_fsuid(), inode->i_uid)) |
e795b717 | 2021 | return true; |
23adbe12 AL |
2022 | |
2023 | ns = current_user_ns(); | |
cc658db4 | 2024 | if (kuid_has_mapping(ns, inode->i_uid) && ns_capable(ns, CAP_FOWNER)) |
e795b717 SH |
2025 | return true; |
2026 | return false; | |
2027 | } | |
2e149670 | 2028 | EXPORT_SYMBOL(inode_owner_or_capable); |
1d59d61f TM |
2029 | |
2030 | /* | |
2031 | * Direct i/o helper functions | |
2032 | */ | |
2033 | static void __inode_dio_wait(struct inode *inode) | |
2034 | { | |
2035 | wait_queue_head_t *wq = bit_waitqueue(&inode->i_state, __I_DIO_WAKEUP); | |
2036 | DEFINE_WAIT_BIT(q, &inode->i_state, __I_DIO_WAKEUP); | |
2037 | ||
2038 | do { | |
21417136 | 2039 | prepare_to_wait(wq, &q.wq_entry, TASK_UNINTERRUPTIBLE); |
1d59d61f TM |
2040 | if (atomic_read(&inode->i_dio_count)) |
2041 | schedule(); | |
2042 | } while (atomic_read(&inode->i_dio_count)); | |
21417136 | 2043 | finish_wait(wq, &q.wq_entry); |
1d59d61f TM |
2044 | } |
2045 | ||
2046 | /** | |
2047 | * inode_dio_wait - wait for outstanding DIO requests to finish | |
2048 | * @inode: inode to wait for | |
2049 | * | |
2050 | * Waits for all pending direct I/O requests to finish so that we can | |
2051 | * proceed with a truncate or equivalent operation. | |
2052 | * | |
2053 | * Must be called under a lock that serializes taking new references | |
2054 | * to i_dio_count, usually by inode->i_mutex. | |
2055 | */ | |
2056 | void inode_dio_wait(struct inode *inode) | |
2057 | { | |
2058 | if (atomic_read(&inode->i_dio_count)) | |
2059 | __inode_dio_wait(inode); | |
2060 | } | |
2061 | EXPORT_SYMBOL(inode_dio_wait); | |
2062 | ||
5f16f322 TT |
2063 | /* |
2064 | * inode_set_flags - atomically set some inode flags | |
2065 | * | |
2066 | * Note: the caller should be holding i_mutex, or else be sure that | |
2067 | * they have exclusive access to the inode structure (i.e., while the | |
2068 | * inode is being instantiated). The reason for the cmpxchg() loop | |
2069 | * --- which wouldn't be necessary if all code paths which modify | |
2070 | * i_flags actually followed this rule, is that there is at least one | |
5fa8e0a1 JK |
2071 | * code path which doesn't today so we use cmpxchg() out of an abundance |
2072 | * of caution. | |
5f16f322 TT |
2073 | * |
2074 | * In the long run, i_mutex is overkill, and we should probably look | |
2075 | * at using the i_lock spinlock to protect i_flags, and then make sure | |
2076 | * it is so documented in include/linux/fs.h and that all code follows | |
2077 | * the locking convention!! | |
2078 | */ | |
2079 | void inode_set_flags(struct inode *inode, unsigned int flags, | |
2080 | unsigned int mask) | |
2081 | { | |
2082 | unsigned int old_flags, new_flags; | |
2083 | ||
2084 | WARN_ON_ONCE(flags & ~mask); | |
2085 | do { | |
6aa7de05 | 2086 | old_flags = READ_ONCE(inode->i_flags); |
5f16f322 TT |
2087 | new_flags = (old_flags & ~mask) | flags; |
2088 | } while (unlikely(cmpxchg(&inode->i_flags, old_flags, | |
2089 | new_flags) != old_flags)); | |
2090 | } | |
2091 | EXPORT_SYMBOL(inode_set_flags); | |
21fc61c7 AV |
2092 | |
2093 | void inode_nohighmem(struct inode *inode) | |
2094 | { | |
2095 | mapping_set_gfp_mask(inode->i_mapping, GFP_USER); | |
2096 | } | |
2097 | EXPORT_SYMBOL(inode_nohighmem); | |
3cd88666 DD |
2098 | |
2099 | /** | |
2100 | * current_time - Return FS time | |
2101 | * @inode: inode. | |
2102 | * | |
2103 | * Return the current time truncated to the time granularity supported by | |
2104 | * the fs. | |
2105 | * | |
2106 | * Note that inode and inode->sb cannot be NULL. | |
2107 | * Otherwise, the function warns and returns time without truncation. | |
2108 | */ | |
2109 | struct timespec current_time(struct inode *inode) | |
2110 | { | |
2111 | struct timespec now = current_kernel_time(); | |
2112 | ||
2113 | if (unlikely(!inode->i_sb)) { | |
2114 | WARN(1, "current_time() called with uninitialized super_block in the inode"); | |
2115 | return now; | |
2116 | } | |
2117 | ||
2118 | return timespec_trunc(now, inode->i_sb->s_time_gran); | |
2119 | } | |
2120 | EXPORT_SYMBOL(current_time); |