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: |
0ae45f63 | 34 | * bdi->wb.b_{dirty,io,more_io,dirty_time}, inode->i_wb_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; |
2cb1599f DC |
174 | inode->i_private = NULL; |
175 | inode->i_mapping = mapping; | |
b3d9b7a3 | 176 | INIT_HLIST_HEAD(&inode->i_dentry); /* buggered by rcu freeing */ |
f19d4a8f AV |
177 | #ifdef CONFIG_FS_POSIX_ACL |
178 | inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED; | |
179 | #endif | |
2cb1599f | 180 | |
3be25f49 EP |
181 | #ifdef CONFIG_FSNOTIFY |
182 | inode->i_fsnotify_mask = 0; | |
183 | #endif | |
4a075e39 | 184 | inode->i_flctx = NULL; |
3e880fb5 | 185 | this_cpu_inc(nr_inodes); |
cffbc8aa | 186 | |
54e34621 | 187 | return 0; |
54e34621 CH |
188 | out: |
189 | return -ENOMEM; | |
1da177e4 | 190 | } |
2cb1599f DC |
191 | EXPORT_SYMBOL(inode_init_always); |
192 | ||
193 | static struct inode *alloc_inode(struct super_block *sb) | |
194 | { | |
195 | struct inode *inode; | |
196 | ||
197 | if (sb->s_op->alloc_inode) | |
198 | inode = sb->s_op->alloc_inode(sb); | |
199 | else | |
200 | inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL); | |
201 | ||
54e34621 CH |
202 | if (!inode) |
203 | return NULL; | |
204 | ||
205 | if (unlikely(inode_init_always(sb, inode))) { | |
206 | if (inode->i_sb->s_op->destroy_inode) | |
207 | inode->i_sb->s_op->destroy_inode(inode); | |
208 | else | |
209 | kmem_cache_free(inode_cachep, inode); | |
210 | return NULL; | |
211 | } | |
212 | ||
213 | return inode; | |
2cb1599f | 214 | } |
1da177e4 | 215 | |
ff0c7d15 NP |
216 | void free_inode_nonrcu(struct inode *inode) |
217 | { | |
218 | kmem_cache_free(inode_cachep, inode); | |
219 | } | |
220 | EXPORT_SYMBOL(free_inode_nonrcu); | |
221 | ||
2e00c97e | 222 | void __destroy_inode(struct inode *inode) |
1da177e4 | 223 | { |
b7542f8c | 224 | BUG_ON(inode_has_buffers(inode)); |
52ebea74 | 225 | inode_detach_wb(inode); |
1da177e4 | 226 | security_inode_free(inode); |
3be25f49 | 227 | fsnotify_inode_delete(inode); |
4a075e39 | 228 | locks_free_lock_context(inode->i_flctx); |
7ada4db8 MS |
229 | if (!inode->i_nlink) { |
230 | WARN_ON(atomic_long_read(&inode->i_sb->s_remove_count) == 0); | |
231 | atomic_long_dec(&inode->i_sb->s_remove_count); | |
232 | } | |
233 | ||
f19d4a8f AV |
234 | #ifdef CONFIG_FS_POSIX_ACL |
235 | if (inode->i_acl && inode->i_acl != ACL_NOT_CACHED) | |
236 | posix_acl_release(inode->i_acl); | |
237 | if (inode->i_default_acl && inode->i_default_acl != ACL_NOT_CACHED) | |
238 | posix_acl_release(inode->i_default_acl); | |
239 | #endif | |
3e880fb5 | 240 | this_cpu_dec(nr_inodes); |
2e00c97e CH |
241 | } |
242 | EXPORT_SYMBOL(__destroy_inode); | |
243 | ||
fa0d7e3d NP |
244 | static void i_callback(struct rcu_head *head) |
245 | { | |
246 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
fa0d7e3d NP |
247 | kmem_cache_free(inode_cachep, inode); |
248 | } | |
249 | ||
56b0dacf | 250 | static void destroy_inode(struct inode *inode) |
2e00c97e | 251 | { |
7ccf19a8 | 252 | BUG_ON(!list_empty(&inode->i_lru)); |
2e00c97e | 253 | __destroy_inode(inode); |
1da177e4 LT |
254 | if (inode->i_sb->s_op->destroy_inode) |
255 | inode->i_sb->s_op->destroy_inode(inode); | |
256 | else | |
fa0d7e3d | 257 | call_rcu(&inode->i_rcu, i_callback); |
1da177e4 | 258 | } |
1da177e4 | 259 | |
7ada4db8 MS |
260 | /** |
261 | * drop_nlink - directly drop an inode's link count | |
262 | * @inode: inode | |
263 | * | |
264 | * This is a low-level filesystem helper to replace any | |
265 | * direct filesystem manipulation of i_nlink. In cases | |
266 | * where we are attempting to track writes to the | |
267 | * filesystem, a decrement to zero means an imminent | |
268 | * write when the file is truncated and actually unlinked | |
269 | * on the filesystem. | |
270 | */ | |
271 | void drop_nlink(struct inode *inode) | |
272 | { | |
273 | WARN_ON(inode->i_nlink == 0); | |
274 | inode->__i_nlink--; | |
275 | if (!inode->i_nlink) | |
276 | atomic_long_inc(&inode->i_sb->s_remove_count); | |
277 | } | |
278 | EXPORT_SYMBOL(drop_nlink); | |
279 | ||
280 | /** | |
281 | * clear_nlink - directly zero an inode's link count | |
282 | * @inode: inode | |
283 | * | |
284 | * This is a low-level filesystem helper to replace any | |
285 | * direct filesystem manipulation of i_nlink. See | |
286 | * drop_nlink() for why we care about i_nlink hitting zero. | |
287 | */ | |
288 | void clear_nlink(struct inode *inode) | |
289 | { | |
290 | if (inode->i_nlink) { | |
291 | inode->__i_nlink = 0; | |
292 | atomic_long_inc(&inode->i_sb->s_remove_count); | |
293 | } | |
294 | } | |
295 | EXPORT_SYMBOL(clear_nlink); | |
296 | ||
297 | /** | |
298 | * set_nlink - directly set an inode's link count | |
299 | * @inode: inode | |
300 | * @nlink: new nlink (should be non-zero) | |
301 | * | |
302 | * This is a low-level filesystem helper to replace any | |
303 | * direct filesystem manipulation of i_nlink. | |
304 | */ | |
305 | void set_nlink(struct inode *inode, unsigned int nlink) | |
306 | { | |
307 | if (!nlink) { | |
7ada4db8 MS |
308 | clear_nlink(inode); |
309 | } else { | |
310 | /* Yes, some filesystems do change nlink from zero to one */ | |
311 | if (inode->i_nlink == 0) | |
312 | atomic_long_dec(&inode->i_sb->s_remove_count); | |
313 | ||
314 | inode->__i_nlink = nlink; | |
315 | } | |
316 | } | |
317 | EXPORT_SYMBOL(set_nlink); | |
318 | ||
319 | /** | |
320 | * inc_nlink - directly increment an inode's link count | |
321 | * @inode: inode | |
322 | * | |
323 | * This is a low-level filesystem helper to replace any | |
324 | * direct filesystem manipulation of i_nlink. Currently, | |
325 | * it is only here for parity with dec_nlink(). | |
326 | */ | |
327 | void inc_nlink(struct inode *inode) | |
328 | { | |
f4e0c30c AV |
329 | if (unlikely(inode->i_nlink == 0)) { |
330 | WARN_ON(!(inode->i_state & I_LINKABLE)); | |
7ada4db8 | 331 | atomic_long_dec(&inode->i_sb->s_remove_count); |
f4e0c30c | 332 | } |
7ada4db8 MS |
333 | |
334 | inode->__i_nlink++; | |
335 | } | |
336 | EXPORT_SYMBOL(inc_nlink); | |
337 | ||
2aa15890 MS |
338 | void address_space_init_once(struct address_space *mapping) |
339 | { | |
340 | memset(mapping, 0, sizeof(*mapping)); | |
341 | INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC); | |
342 | spin_lock_init(&mapping->tree_lock); | |
c8c06efa | 343 | init_rwsem(&mapping->i_mmap_rwsem); |
2aa15890 MS |
344 | INIT_LIST_HEAD(&mapping->private_list); |
345 | spin_lock_init(&mapping->private_lock); | |
6b2dbba8 | 346 | mapping->i_mmap = RB_ROOT; |
2aa15890 MS |
347 | } |
348 | EXPORT_SYMBOL(address_space_init_once); | |
349 | ||
1da177e4 LT |
350 | /* |
351 | * These are initializations that only need to be done | |
352 | * once, because the fields are idempotent across use | |
353 | * of the inode, so let the slab aware of that. | |
354 | */ | |
355 | void inode_init_once(struct inode *inode) | |
356 | { | |
357 | memset(inode, 0, sizeof(*inode)); | |
358 | INIT_HLIST_NODE(&inode->i_hash); | |
1da177e4 | 359 | INIT_LIST_HEAD(&inode->i_devices); |
7ccf19a8 NP |
360 | INIT_LIST_HEAD(&inode->i_wb_list); |
361 | INIT_LIST_HEAD(&inode->i_lru); | |
2aa15890 | 362 | address_space_init_once(&inode->i_data); |
1da177e4 | 363 | i_size_ordered_init(inode); |
3be25f49 | 364 | #ifdef CONFIG_FSNOTIFY |
e61ce867 | 365 | INIT_HLIST_HEAD(&inode->i_fsnotify_marks); |
3be25f49 | 366 | #endif |
1da177e4 | 367 | } |
1da177e4 LT |
368 | EXPORT_SYMBOL(inode_init_once); |
369 | ||
51cc5068 | 370 | static void init_once(void *foo) |
1da177e4 | 371 | { |
6b3304b5 | 372 | struct inode *inode = (struct inode *) foo; |
1da177e4 | 373 | |
a35afb83 | 374 | inode_init_once(inode); |
1da177e4 LT |
375 | } |
376 | ||
377 | /* | |
250df6ed | 378 | * inode->i_lock must be held |
1da177e4 | 379 | */ |
6b3304b5 | 380 | void __iget(struct inode *inode) |
1da177e4 | 381 | { |
9e38d86f NP |
382 | atomic_inc(&inode->i_count); |
383 | } | |
2e147f1e | 384 | |
7de9c6ee AV |
385 | /* |
386 | * get additional reference to inode; caller must already hold one. | |
387 | */ | |
388 | void ihold(struct inode *inode) | |
389 | { | |
390 | WARN_ON(atomic_inc_return(&inode->i_count) < 2); | |
391 | } | |
392 | EXPORT_SYMBOL(ihold); | |
393 | ||
9e38d86f NP |
394 | static void inode_lru_list_add(struct inode *inode) |
395 | { | |
bc3b14cb | 396 | if (list_lru_add(&inode->i_sb->s_inode_lru, &inode->i_lru)) |
fcb94f72 | 397 | this_cpu_inc(nr_unused); |
9e38d86f | 398 | } |
2e147f1e | 399 | |
4eff96dd JK |
400 | /* |
401 | * Add inode to LRU if needed (inode is unused and clean). | |
402 | * | |
403 | * Needs inode->i_lock held. | |
404 | */ | |
405 | void inode_add_lru(struct inode *inode) | |
406 | { | |
0ae45f63 TT |
407 | if (!(inode->i_state & (I_DIRTY_ALL | I_SYNC | |
408 | I_FREEING | I_WILL_FREE)) && | |
4eff96dd JK |
409 | !atomic_read(&inode->i_count) && inode->i_sb->s_flags & MS_ACTIVE) |
410 | inode_lru_list_add(inode); | |
411 | } | |
412 | ||
413 | ||
9e38d86f NP |
414 | static void inode_lru_list_del(struct inode *inode) |
415 | { | |
bc3b14cb DC |
416 | |
417 | if (list_lru_del(&inode->i_sb->s_inode_lru, &inode->i_lru)) | |
fcb94f72 | 418 | this_cpu_dec(nr_unused); |
1da177e4 LT |
419 | } |
420 | ||
646ec461 CH |
421 | /** |
422 | * inode_sb_list_add - add inode to the superblock list of inodes | |
423 | * @inode: inode to add | |
424 | */ | |
425 | void inode_sb_list_add(struct inode *inode) | |
426 | { | |
74278da9 | 427 | spin_lock(&inode->i_sb->s_inode_list_lock); |
55fa6091 | 428 | list_add(&inode->i_sb_list, &inode->i_sb->s_inodes); |
74278da9 | 429 | spin_unlock(&inode->i_sb->s_inode_list_lock); |
646ec461 CH |
430 | } |
431 | EXPORT_SYMBOL_GPL(inode_sb_list_add); | |
432 | ||
55fa6091 | 433 | static inline void inode_sb_list_del(struct inode *inode) |
646ec461 | 434 | { |
a209dfc7 | 435 | if (!list_empty(&inode->i_sb_list)) { |
74278da9 | 436 | spin_lock(&inode->i_sb->s_inode_list_lock); |
a209dfc7 | 437 | list_del_init(&inode->i_sb_list); |
74278da9 | 438 | spin_unlock(&inode->i_sb->s_inode_list_lock); |
a209dfc7 | 439 | } |
646ec461 CH |
440 | } |
441 | ||
4c51acbc DC |
442 | static unsigned long hash(struct super_block *sb, unsigned long hashval) |
443 | { | |
444 | unsigned long tmp; | |
445 | ||
446 | tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) / | |
447 | L1_CACHE_BYTES; | |
4b4563dc CH |
448 | tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> i_hash_shift); |
449 | return tmp & i_hash_mask; | |
4c51acbc DC |
450 | } |
451 | ||
452 | /** | |
453 | * __insert_inode_hash - hash an inode | |
454 | * @inode: unhashed inode | |
455 | * @hashval: unsigned long value used to locate this object in the | |
456 | * inode_hashtable. | |
457 | * | |
458 | * Add an inode to the inode hash for this superblock. | |
459 | */ | |
460 | void __insert_inode_hash(struct inode *inode, unsigned long hashval) | |
461 | { | |
646ec461 CH |
462 | struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval); |
463 | ||
67a23c49 | 464 | spin_lock(&inode_hash_lock); |
250df6ed | 465 | spin_lock(&inode->i_lock); |
646ec461 | 466 | hlist_add_head(&inode->i_hash, b); |
250df6ed | 467 | spin_unlock(&inode->i_lock); |
67a23c49 | 468 | spin_unlock(&inode_hash_lock); |
4c51acbc DC |
469 | } |
470 | EXPORT_SYMBOL(__insert_inode_hash); | |
471 | ||
4c51acbc | 472 | /** |
f2ee7abf | 473 | * __remove_inode_hash - remove an inode from the hash |
4c51acbc DC |
474 | * @inode: inode to unhash |
475 | * | |
476 | * Remove an inode from the superblock. | |
477 | */ | |
f2ee7abf | 478 | void __remove_inode_hash(struct inode *inode) |
4c51acbc | 479 | { |
67a23c49 | 480 | spin_lock(&inode_hash_lock); |
250df6ed | 481 | spin_lock(&inode->i_lock); |
4c51acbc | 482 | hlist_del_init(&inode->i_hash); |
250df6ed | 483 | spin_unlock(&inode->i_lock); |
67a23c49 | 484 | spin_unlock(&inode_hash_lock); |
4c51acbc | 485 | } |
f2ee7abf | 486 | EXPORT_SYMBOL(__remove_inode_hash); |
4c51acbc | 487 | |
dbd5768f | 488 | void clear_inode(struct inode *inode) |
b0683aa6 AV |
489 | { |
490 | might_sleep(); | |
08142579 JK |
491 | /* |
492 | * We have to cycle tree_lock here because reclaim can be still in the | |
493 | * process of removing the last page (in __delete_from_page_cache()) | |
494 | * and we must not free mapping under it. | |
495 | */ | |
496 | spin_lock_irq(&inode->i_data.tree_lock); | |
b0683aa6 | 497 | BUG_ON(inode->i_data.nrpages); |
91b0abe3 | 498 | BUG_ON(inode->i_data.nrshadows); |
08142579 | 499 | spin_unlock_irq(&inode->i_data.tree_lock); |
b0683aa6 AV |
500 | BUG_ON(!list_empty(&inode->i_data.private_list)); |
501 | BUG_ON(!(inode->i_state & I_FREEING)); | |
502 | BUG_ON(inode->i_state & I_CLEAR); | |
fa0d7e3d | 503 | /* don't need i_lock here, no concurrent mods to i_state */ |
b0683aa6 AV |
504 | inode->i_state = I_FREEING | I_CLEAR; |
505 | } | |
dbd5768f | 506 | EXPORT_SYMBOL(clear_inode); |
b0683aa6 | 507 | |
b2b2af8e DC |
508 | /* |
509 | * Free the inode passed in, removing it from the lists it is still connected | |
510 | * to. We remove any pages still attached to the inode and wait for any IO that | |
511 | * is still in progress before finally destroying the inode. | |
512 | * | |
513 | * An inode must already be marked I_FREEING so that we avoid the inode being | |
514 | * moved back onto lists if we race with other code that manipulates the lists | |
515 | * (e.g. writeback_single_inode). The caller is responsible for setting this. | |
516 | * | |
517 | * An inode must already be removed from the LRU list before being evicted from | |
518 | * the cache. This should occur atomically with setting the I_FREEING state | |
519 | * flag, so no inodes here should ever be on the LRU when being evicted. | |
520 | */ | |
644da596 | 521 | static void evict(struct inode *inode) |
b4272d4c AV |
522 | { |
523 | const struct super_operations *op = inode->i_sb->s_op; | |
524 | ||
b2b2af8e DC |
525 | BUG_ON(!(inode->i_state & I_FREEING)); |
526 | BUG_ON(!list_empty(&inode->i_lru)); | |
527 | ||
b12362bd ED |
528 | if (!list_empty(&inode->i_wb_list)) |
529 | inode_wb_list_del(inode); | |
530 | ||
55fa6091 DC |
531 | inode_sb_list_del(inode); |
532 | ||
169ebd90 JK |
533 | /* |
534 | * Wait for flusher thread to be done with the inode so that filesystem | |
535 | * does not start destroying it while writeback is still running. Since | |
536 | * the inode has I_FREEING set, flusher thread won't start new work on | |
537 | * the inode. We just have to wait for running writeback to finish. | |
538 | */ | |
539 | inode_wait_for_writeback(inode); | |
7994e6f7 | 540 | |
be7ce416 AV |
541 | if (op->evict_inode) { |
542 | op->evict_inode(inode); | |
b4272d4c | 543 | } else { |
91b0abe3 | 544 | truncate_inode_pages_final(&inode->i_data); |
dbd5768f | 545 | clear_inode(inode); |
b4272d4c | 546 | } |
661074e9 AV |
547 | if (S_ISBLK(inode->i_mode) && inode->i_bdev) |
548 | bd_forget(inode); | |
549 | if (S_ISCHR(inode->i_mode) && inode->i_cdev) | |
550 | cd_forget(inode); | |
b2b2af8e DC |
551 | |
552 | remove_inode_hash(inode); | |
553 | ||
554 | spin_lock(&inode->i_lock); | |
555 | wake_up_bit(&inode->i_state, __I_NEW); | |
556 | BUG_ON(inode->i_state != (I_FREEING | I_CLEAR)); | |
557 | spin_unlock(&inode->i_lock); | |
558 | ||
559 | destroy_inode(inode); | |
b4272d4c AV |
560 | } |
561 | ||
1da177e4 LT |
562 | /* |
563 | * dispose_list - dispose of the contents of a local list | |
564 | * @head: the head of the list to free | |
565 | * | |
566 | * Dispose-list gets a local list with local inodes in it, so it doesn't | |
567 | * need to worry about list corruption and SMP locks. | |
568 | */ | |
569 | static void dispose_list(struct list_head *head) | |
570 | { | |
1da177e4 LT |
571 | while (!list_empty(head)) { |
572 | struct inode *inode; | |
573 | ||
7ccf19a8 NP |
574 | inode = list_first_entry(head, struct inode, i_lru); |
575 | list_del_init(&inode->i_lru); | |
1da177e4 | 576 | |
644da596 | 577 | evict(inode); |
1da177e4 | 578 | } |
1da177e4 LT |
579 | } |
580 | ||
63997e98 AV |
581 | /** |
582 | * evict_inodes - evict all evictable inodes for a superblock | |
583 | * @sb: superblock to operate on | |
584 | * | |
585 | * Make sure that no inodes with zero refcount are retained. This is | |
586 | * called by superblock shutdown after having MS_ACTIVE flag removed, | |
587 | * so any inode reaching zero refcount during or after that call will | |
588 | * be immediately evicted. | |
1da177e4 | 589 | */ |
63997e98 | 590 | void evict_inodes(struct super_block *sb) |
1da177e4 | 591 | { |
63997e98 AV |
592 | struct inode *inode, *next; |
593 | LIST_HEAD(dispose); | |
1da177e4 | 594 | |
74278da9 | 595 | spin_lock(&sb->s_inode_list_lock); |
63997e98 AV |
596 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
597 | if (atomic_read(&inode->i_count)) | |
aabb8fdb | 598 | continue; |
250df6ed DC |
599 | |
600 | spin_lock(&inode->i_lock); | |
601 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
602 | spin_unlock(&inode->i_lock); | |
1da177e4 | 603 | continue; |
250df6ed | 604 | } |
63997e98 AV |
605 | |
606 | inode->i_state |= I_FREEING; | |
02afc410 | 607 | inode_lru_list_del(inode); |
250df6ed | 608 | spin_unlock(&inode->i_lock); |
02afc410 | 609 | list_add(&inode->i_lru, &dispose); |
1da177e4 | 610 | } |
74278da9 | 611 | spin_unlock(&sb->s_inode_list_lock); |
63997e98 AV |
612 | |
613 | dispose_list(&dispose); | |
1da177e4 LT |
614 | } |
615 | ||
1da177e4 | 616 | /** |
a0318786 CH |
617 | * invalidate_inodes - attempt to free all inodes on a superblock |
618 | * @sb: superblock to operate on | |
93b270f7 | 619 | * @kill_dirty: flag to guide handling of dirty inodes |
1da177e4 | 620 | * |
a0318786 CH |
621 | * Attempts to free all inodes for a given superblock. If there were any |
622 | * busy inodes return a non-zero value, else zero. | |
93b270f7 N |
623 | * If @kill_dirty is set, discard dirty inodes too, otherwise treat |
624 | * them as busy. | |
1da177e4 | 625 | */ |
93b270f7 | 626 | int invalidate_inodes(struct super_block *sb, bool kill_dirty) |
1da177e4 | 627 | { |
cffbc8aa | 628 | int busy = 0; |
a0318786 CH |
629 | struct inode *inode, *next; |
630 | LIST_HEAD(dispose); | |
1da177e4 | 631 | |
74278da9 | 632 | spin_lock(&sb->s_inode_list_lock); |
a0318786 | 633 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
250df6ed DC |
634 | spin_lock(&inode->i_lock); |
635 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
636 | spin_unlock(&inode->i_lock); | |
aabb8fdb | 637 | continue; |
250df6ed | 638 | } |
0ae45f63 | 639 | if (inode->i_state & I_DIRTY_ALL && !kill_dirty) { |
250df6ed | 640 | spin_unlock(&inode->i_lock); |
93b270f7 N |
641 | busy = 1; |
642 | continue; | |
643 | } | |
99a38919 | 644 | if (atomic_read(&inode->i_count)) { |
250df6ed | 645 | spin_unlock(&inode->i_lock); |
99a38919 | 646 | busy = 1; |
1da177e4 LT |
647 | continue; |
648 | } | |
99a38919 | 649 | |
99a38919 | 650 | inode->i_state |= I_FREEING; |
02afc410 | 651 | inode_lru_list_del(inode); |
250df6ed | 652 | spin_unlock(&inode->i_lock); |
02afc410 | 653 | list_add(&inode->i_lru, &dispose); |
1da177e4 | 654 | } |
74278da9 | 655 | spin_unlock(&sb->s_inode_list_lock); |
1da177e4 | 656 | |
a0318786 | 657 | dispose_list(&dispose); |
1da177e4 LT |
658 | |
659 | return busy; | |
660 | } | |
1da177e4 | 661 | |
1da177e4 | 662 | /* |
bc3b14cb | 663 | * Isolate the inode from the LRU in preparation for freeing it. |
1da177e4 LT |
664 | * |
665 | * Any inodes which are pinned purely because of attached pagecache have their | |
9e38d86f NP |
666 | * pagecache removed. If the inode has metadata buffers attached to |
667 | * mapping->private_list then try to remove them. | |
1da177e4 | 668 | * |
9e38d86f NP |
669 | * If the inode has the I_REFERENCED flag set, then it means that it has been |
670 | * used recently - the flag is set in iput_final(). When we encounter such an | |
671 | * inode, clear the flag and move it to the back of the LRU so it gets another | |
672 | * pass through the LRU before it gets reclaimed. This is necessary because of | |
673 | * the fact we are doing lazy LRU updates to minimise lock contention so the | |
674 | * LRU does not have strict ordering. Hence we don't want to reclaim inodes | |
675 | * with this flag set because they are the inodes that are out of order. | |
1da177e4 | 676 | */ |
3f97b163 VD |
677 | static enum lru_status inode_lru_isolate(struct list_head *item, |
678 | struct list_lru_one *lru, spinlock_t *lru_lock, void *arg) | |
1da177e4 | 679 | { |
bc3b14cb DC |
680 | struct list_head *freeable = arg; |
681 | struct inode *inode = container_of(item, struct inode, i_lru); | |
1da177e4 | 682 | |
bc3b14cb DC |
683 | /* |
684 | * we are inverting the lru lock/inode->i_lock here, so use a trylock. | |
685 | * If we fail to get the lock, just skip it. | |
686 | */ | |
687 | if (!spin_trylock(&inode->i_lock)) | |
688 | return LRU_SKIP; | |
1da177e4 | 689 | |
bc3b14cb DC |
690 | /* |
691 | * Referenced or dirty inodes are still in use. Give them another pass | |
692 | * through the LRU as we canot reclaim them now. | |
693 | */ | |
694 | if (atomic_read(&inode->i_count) || | |
695 | (inode->i_state & ~I_REFERENCED)) { | |
3f97b163 | 696 | list_lru_isolate(lru, &inode->i_lru); |
bc3b14cb DC |
697 | spin_unlock(&inode->i_lock); |
698 | this_cpu_dec(nr_unused); | |
699 | return LRU_REMOVED; | |
700 | } | |
1da177e4 | 701 | |
bc3b14cb DC |
702 | /* recently referenced inodes get one more pass */ |
703 | if (inode->i_state & I_REFERENCED) { | |
704 | inode->i_state &= ~I_REFERENCED; | |
705 | spin_unlock(&inode->i_lock); | |
706 | return LRU_ROTATE; | |
707 | } | |
1da177e4 | 708 | |
bc3b14cb DC |
709 | if (inode_has_buffers(inode) || inode->i_data.nrpages) { |
710 | __iget(inode); | |
711 | spin_unlock(&inode->i_lock); | |
712 | spin_unlock(lru_lock); | |
713 | if (remove_inode_buffers(inode)) { | |
714 | unsigned long reap; | |
715 | reap = invalidate_mapping_pages(&inode->i_data, 0, -1); | |
716 | if (current_is_kswapd()) | |
717 | __count_vm_events(KSWAPD_INODESTEAL, reap); | |
718 | else | |
719 | __count_vm_events(PGINODESTEAL, reap); | |
720 | if (current->reclaim_state) | |
721 | current->reclaim_state->reclaimed_slab += reap; | |
02afc410 | 722 | } |
bc3b14cb DC |
723 | iput(inode); |
724 | spin_lock(lru_lock); | |
725 | return LRU_RETRY; | |
726 | } | |
02afc410 | 727 | |
bc3b14cb DC |
728 | WARN_ON(inode->i_state & I_NEW); |
729 | inode->i_state |= I_FREEING; | |
3f97b163 | 730 | list_lru_isolate_move(lru, &inode->i_lru, freeable); |
bc3b14cb | 731 | spin_unlock(&inode->i_lock); |
9e38d86f | 732 | |
bc3b14cb DC |
733 | this_cpu_dec(nr_unused); |
734 | return LRU_REMOVED; | |
735 | } | |
7ccf19a8 | 736 | |
bc3b14cb DC |
737 | /* |
738 | * Walk the superblock inode LRU for freeable inodes and attempt to free them. | |
739 | * This is called from the superblock shrinker function with a number of inodes | |
740 | * to trim from the LRU. Inodes to be freed are moved to a temporary list and | |
741 | * then are freed outside inode_lock by dispose_list(). | |
742 | */ | |
503c358c | 743 | long prune_icache_sb(struct super_block *sb, struct shrink_control *sc) |
bc3b14cb DC |
744 | { |
745 | LIST_HEAD(freeable); | |
746 | long freed; | |
1da177e4 | 747 | |
503c358c VD |
748 | freed = list_lru_shrink_walk(&sb->s_inode_lru, sc, |
749 | inode_lru_isolate, &freeable); | |
1da177e4 | 750 | dispose_list(&freeable); |
0a234c6d | 751 | return freed; |
1da177e4 LT |
752 | } |
753 | ||
1da177e4 LT |
754 | static void __wait_on_freeing_inode(struct inode *inode); |
755 | /* | |
756 | * Called with the inode lock held. | |
1da177e4 | 757 | */ |
6b3304b5 MK |
758 | static struct inode *find_inode(struct super_block *sb, |
759 | struct hlist_head *head, | |
760 | int (*test)(struct inode *, void *), | |
761 | void *data) | |
1da177e4 | 762 | { |
6b3304b5 | 763 | struct inode *inode = NULL; |
1da177e4 LT |
764 | |
765 | repeat: | |
b67bfe0d | 766 | hlist_for_each_entry(inode, head, i_hash) { |
5a3cd992 | 767 | if (inode->i_sb != sb) |
1da177e4 | 768 | continue; |
5a3cd992 | 769 | if (!test(inode, data)) |
1da177e4 | 770 | continue; |
5a3cd992 | 771 | spin_lock(&inode->i_lock); |
a4ffdde6 | 772 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
773 | __wait_on_freeing_inode(inode); |
774 | goto repeat; | |
775 | } | |
f7899bd5 | 776 | __iget(inode); |
250df6ed | 777 | spin_unlock(&inode->i_lock); |
f7899bd5 | 778 | return inode; |
1da177e4 | 779 | } |
f7899bd5 | 780 | return NULL; |
1da177e4 LT |
781 | } |
782 | ||
783 | /* | |
784 | * find_inode_fast is the fast path version of find_inode, see the comment at | |
785 | * iget_locked for details. | |
786 | */ | |
6b3304b5 MK |
787 | static struct inode *find_inode_fast(struct super_block *sb, |
788 | struct hlist_head *head, unsigned long ino) | |
1da177e4 | 789 | { |
6b3304b5 | 790 | struct inode *inode = NULL; |
1da177e4 LT |
791 | |
792 | repeat: | |
b67bfe0d | 793 | hlist_for_each_entry(inode, head, i_hash) { |
5a3cd992 | 794 | if (inode->i_ino != ino) |
1da177e4 | 795 | continue; |
5a3cd992 | 796 | if (inode->i_sb != sb) |
1da177e4 | 797 | continue; |
5a3cd992 | 798 | spin_lock(&inode->i_lock); |
a4ffdde6 | 799 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
800 | __wait_on_freeing_inode(inode); |
801 | goto repeat; | |
802 | } | |
f7899bd5 | 803 | __iget(inode); |
250df6ed | 804 | spin_unlock(&inode->i_lock); |
f7899bd5 | 805 | return inode; |
1da177e4 | 806 | } |
f7899bd5 | 807 | return NULL; |
8290c35f DC |
808 | } |
809 | ||
f991bd2e ED |
810 | /* |
811 | * Each cpu owns a range of LAST_INO_BATCH numbers. | |
812 | * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations, | |
813 | * to renew the exhausted range. | |
8290c35f | 814 | * |
f991bd2e ED |
815 | * This does not significantly increase overflow rate because every CPU can |
816 | * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is | |
817 | * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the | |
818 | * 2^32 range, and is a worst-case. Even a 50% wastage would only increase | |
819 | * overflow rate by 2x, which does not seem too significant. | |
820 | * | |
821 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
822 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
823 | * here to attempt to avoid that. | |
8290c35f | 824 | */ |
f991bd2e ED |
825 | #define LAST_INO_BATCH 1024 |
826 | static DEFINE_PER_CPU(unsigned int, last_ino); | |
827 | ||
85fe4025 | 828 | unsigned int get_next_ino(void) |
8290c35f | 829 | { |
f991bd2e ED |
830 | unsigned int *p = &get_cpu_var(last_ino); |
831 | unsigned int res = *p; | |
8290c35f | 832 | |
f991bd2e ED |
833 | #ifdef CONFIG_SMP |
834 | if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) { | |
835 | static atomic_t shared_last_ino; | |
836 | int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino); | |
837 | ||
838 | res = next - LAST_INO_BATCH; | |
839 | } | |
840 | #endif | |
841 | ||
2adc376c CM |
842 | res++; |
843 | /* get_next_ino should not provide a 0 inode number */ | |
844 | if (unlikely(!res)) | |
845 | res++; | |
846 | *p = res; | |
f991bd2e ED |
847 | put_cpu_var(last_ino); |
848 | return res; | |
8290c35f | 849 | } |
85fe4025 | 850 | EXPORT_SYMBOL(get_next_ino); |
8290c35f | 851 | |
a209dfc7 ED |
852 | /** |
853 | * new_inode_pseudo - obtain an inode | |
854 | * @sb: superblock | |
855 | * | |
856 | * Allocates a new inode for given superblock. | |
857 | * Inode wont be chained in superblock s_inodes list | |
858 | * This means : | |
859 | * - fs can't be unmount | |
860 | * - quotas, fsnotify, writeback can't work | |
861 | */ | |
862 | struct inode *new_inode_pseudo(struct super_block *sb) | |
863 | { | |
864 | struct inode *inode = alloc_inode(sb); | |
865 | ||
866 | if (inode) { | |
867 | spin_lock(&inode->i_lock); | |
868 | inode->i_state = 0; | |
869 | spin_unlock(&inode->i_lock); | |
870 | INIT_LIST_HEAD(&inode->i_sb_list); | |
871 | } | |
872 | return inode; | |
873 | } | |
874 | ||
1da177e4 LT |
875 | /** |
876 | * new_inode - obtain an inode | |
877 | * @sb: superblock | |
878 | * | |
769848c0 | 879 | * Allocates a new inode for given superblock. The default gfp_mask |
3c1d4378 | 880 | * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE. |
769848c0 MG |
881 | * If HIGHMEM pages are unsuitable or it is known that pages allocated |
882 | * for the page cache are not reclaimable or migratable, | |
883 | * mapping_set_gfp_mask() must be called with suitable flags on the | |
884 | * newly created inode's mapping | |
885 | * | |
1da177e4 LT |
886 | */ |
887 | struct inode *new_inode(struct super_block *sb) | |
888 | { | |
6b3304b5 | 889 | struct inode *inode; |
1da177e4 | 890 | |
74278da9 | 891 | spin_lock_prefetch(&sb->s_inode_list_lock); |
6b3304b5 | 892 | |
a209dfc7 ED |
893 | inode = new_inode_pseudo(sb); |
894 | if (inode) | |
55fa6091 | 895 | inode_sb_list_add(inode); |
1da177e4 LT |
896 | return inode; |
897 | } | |
1da177e4 LT |
898 | EXPORT_SYMBOL(new_inode); |
899 | ||
14358e6d | 900 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
e096d0c7 JB |
901 | void lockdep_annotate_inode_mutex_key(struct inode *inode) |
902 | { | |
a3314a0e | 903 | if (S_ISDIR(inode->i_mode)) { |
1e89a5e1 PZ |
904 | struct file_system_type *type = inode->i_sb->s_type; |
905 | ||
9a7aa12f | 906 | /* Set new key only if filesystem hasn't already changed it */ |
978d6d8c | 907 | if (lockdep_match_class(&inode->i_mutex, &type->i_mutex_key)) { |
9a7aa12f JK |
908 | /* |
909 | * ensure nobody is actually holding i_mutex | |
910 | */ | |
911 | mutex_destroy(&inode->i_mutex); | |
912 | mutex_init(&inode->i_mutex); | |
913 | lockdep_set_class(&inode->i_mutex, | |
914 | &type->i_mutex_dir_key); | |
915 | } | |
1e89a5e1 | 916 | } |
e096d0c7 JB |
917 | } |
918 | EXPORT_SYMBOL(lockdep_annotate_inode_mutex_key); | |
14358e6d | 919 | #endif |
e096d0c7 JB |
920 | |
921 | /** | |
922 | * unlock_new_inode - clear the I_NEW state and wake up any waiters | |
923 | * @inode: new inode to unlock | |
924 | * | |
925 | * Called when the inode is fully initialised to clear the new state of the | |
926 | * inode and wake up anyone waiting for the inode to finish initialisation. | |
927 | */ | |
928 | void unlock_new_inode(struct inode *inode) | |
929 | { | |
930 | lockdep_annotate_inode_mutex_key(inode); | |
250df6ed | 931 | spin_lock(&inode->i_lock); |
eaff8079 CH |
932 | WARN_ON(!(inode->i_state & I_NEW)); |
933 | inode->i_state &= ~I_NEW; | |
310fa7a3 | 934 | smp_mb(); |
250df6ed DC |
935 | wake_up_bit(&inode->i_state, __I_NEW); |
936 | spin_unlock(&inode->i_lock); | |
1da177e4 | 937 | } |
1da177e4 LT |
938 | EXPORT_SYMBOL(unlock_new_inode); |
939 | ||
375e289e BF |
940 | /** |
941 | * lock_two_nondirectories - take two i_mutexes on non-directory objects | |
4fd699ae BF |
942 | * |
943 | * Lock any non-NULL argument that is not a directory. | |
944 | * Zero, one or two objects may be locked by this function. | |
945 | * | |
375e289e BF |
946 | * @inode1: first inode to lock |
947 | * @inode2: second inode to lock | |
948 | */ | |
949 | void lock_two_nondirectories(struct inode *inode1, struct inode *inode2) | |
950 | { | |
4fd699ae BF |
951 | if (inode1 > inode2) |
952 | swap(inode1, inode2); | |
953 | ||
954 | if (inode1 && !S_ISDIR(inode1->i_mode)) | |
27555516 | 955 | mutex_lock(&inode1->i_mutex); |
4fd699ae | 956 | if (inode2 && !S_ISDIR(inode2->i_mode) && inode2 != inode1) |
40bd22c9 | 957 | mutex_lock_nested(&inode2->i_mutex, I_MUTEX_NONDIR2); |
375e289e BF |
958 | } |
959 | EXPORT_SYMBOL(lock_two_nondirectories); | |
960 | ||
961 | /** | |
962 | * unlock_two_nondirectories - release locks from lock_two_nondirectories() | |
963 | * @inode1: first inode to unlock | |
964 | * @inode2: second inode to unlock | |
965 | */ | |
966 | void unlock_two_nondirectories(struct inode *inode1, struct inode *inode2) | |
967 | { | |
4fd699ae BF |
968 | if (inode1 && !S_ISDIR(inode1->i_mode)) |
969 | mutex_unlock(&inode1->i_mutex); | |
970 | if (inode2 && !S_ISDIR(inode2->i_mode) && inode2 != inode1) | |
375e289e BF |
971 | mutex_unlock(&inode2->i_mutex); |
972 | } | |
973 | EXPORT_SYMBOL(unlock_two_nondirectories); | |
974 | ||
0b2d0724 CH |
975 | /** |
976 | * iget5_locked - obtain an inode from a mounted file system | |
977 | * @sb: super block of file system | |
978 | * @hashval: hash value (usually inode number) to get | |
979 | * @test: callback used for comparisons between inodes | |
980 | * @set: callback used to initialize a new struct inode | |
981 | * @data: opaque data pointer to pass to @test and @set | |
982 | * | |
983 | * Search for the inode specified by @hashval and @data in the inode cache, | |
984 | * and if present it is return it with an increased reference count. This is | |
985 | * a generalized version of iget_locked() for file systems where the inode | |
986 | * number is not sufficient for unique identification of an inode. | |
987 | * | |
988 | * If the inode is not in cache, allocate a new inode and return it locked, | |
989 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
990 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 991 | * |
0b2d0724 CH |
992 | * Note both @test and @set are called with the inode_hash_lock held, so can't |
993 | * sleep. | |
1da177e4 | 994 | */ |
0b2d0724 CH |
995 | struct inode *iget5_locked(struct super_block *sb, unsigned long hashval, |
996 | int (*test)(struct inode *, void *), | |
997 | int (*set)(struct inode *, void *), void *data) | |
1da177e4 | 998 | { |
0b2d0724 | 999 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); |
6b3304b5 | 1000 | struct inode *inode; |
1da177e4 | 1001 | |
0b2d0724 CH |
1002 | spin_lock(&inode_hash_lock); |
1003 | inode = find_inode(sb, head, test, data); | |
1004 | spin_unlock(&inode_hash_lock); | |
1005 | ||
1006 | if (inode) { | |
1007 | wait_on_inode(inode); | |
1008 | return inode; | |
1009 | } | |
1010 | ||
1da177e4 LT |
1011 | inode = alloc_inode(sb); |
1012 | if (inode) { | |
6b3304b5 | 1013 | struct inode *old; |
1da177e4 | 1014 | |
67a23c49 | 1015 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1016 | /* We released the lock, so.. */ |
1017 | old = find_inode(sb, head, test, data); | |
1018 | if (!old) { | |
1019 | if (set(inode, data)) | |
1020 | goto set_failed; | |
1021 | ||
250df6ed DC |
1022 | spin_lock(&inode->i_lock); |
1023 | inode->i_state = I_NEW; | |
646ec461 | 1024 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1025 | spin_unlock(&inode->i_lock); |
55fa6091 | 1026 | inode_sb_list_add(inode); |
67a23c49 | 1027 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1028 | |
1029 | /* Return the locked inode with I_NEW set, the | |
1030 | * caller is responsible for filling in the contents | |
1031 | */ | |
1032 | return inode; | |
1033 | } | |
1034 | ||
1035 | /* | |
1036 | * Uhhuh, somebody else created the same inode under | |
1037 | * us. Use the old inode instead of the one we just | |
1038 | * allocated. | |
1039 | */ | |
67a23c49 | 1040 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1041 | destroy_inode(inode); |
1042 | inode = old; | |
1043 | wait_on_inode(inode); | |
1044 | } | |
1045 | return inode; | |
1046 | ||
1047 | set_failed: | |
67a23c49 | 1048 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1049 | destroy_inode(inode); |
1050 | return NULL; | |
1051 | } | |
0b2d0724 | 1052 | EXPORT_SYMBOL(iget5_locked); |
1da177e4 | 1053 | |
0b2d0724 CH |
1054 | /** |
1055 | * iget_locked - obtain an inode from a mounted file system | |
1056 | * @sb: super block of file system | |
1057 | * @ino: inode number to get | |
1058 | * | |
1059 | * Search for the inode specified by @ino in the inode cache and if present | |
1060 | * return it with an increased reference count. This is for file systems | |
1061 | * where the inode number is sufficient for unique identification of an inode. | |
1062 | * | |
1063 | * If the inode is not in cache, allocate a new inode and return it locked, | |
1064 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
1065 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 1066 | */ |
0b2d0724 | 1067 | struct inode *iget_locked(struct super_block *sb, unsigned long ino) |
1da177e4 | 1068 | { |
0b2d0724 | 1069 | struct hlist_head *head = inode_hashtable + hash(sb, ino); |
6b3304b5 | 1070 | struct inode *inode; |
1da177e4 | 1071 | |
0b2d0724 CH |
1072 | spin_lock(&inode_hash_lock); |
1073 | inode = find_inode_fast(sb, head, ino); | |
1074 | spin_unlock(&inode_hash_lock); | |
1075 | if (inode) { | |
1076 | wait_on_inode(inode); | |
1077 | return inode; | |
1078 | } | |
1079 | ||
1da177e4 LT |
1080 | inode = alloc_inode(sb); |
1081 | if (inode) { | |
6b3304b5 | 1082 | struct inode *old; |
1da177e4 | 1083 | |
67a23c49 | 1084 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1085 | /* We released the lock, so.. */ |
1086 | old = find_inode_fast(sb, head, ino); | |
1087 | if (!old) { | |
1088 | inode->i_ino = ino; | |
250df6ed DC |
1089 | spin_lock(&inode->i_lock); |
1090 | inode->i_state = I_NEW; | |
646ec461 | 1091 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1092 | spin_unlock(&inode->i_lock); |
55fa6091 | 1093 | inode_sb_list_add(inode); |
67a23c49 | 1094 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1095 | |
1096 | /* Return the locked inode with I_NEW set, the | |
1097 | * caller is responsible for filling in the contents | |
1098 | */ | |
1099 | return inode; | |
1100 | } | |
1101 | ||
1102 | /* | |
1103 | * Uhhuh, somebody else created the same inode under | |
1104 | * us. Use the old inode instead of the one we just | |
1105 | * allocated. | |
1106 | */ | |
67a23c49 | 1107 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1108 | destroy_inode(inode); |
1109 | inode = old; | |
1110 | wait_on_inode(inode); | |
1111 | } | |
1112 | return inode; | |
1113 | } | |
0b2d0724 | 1114 | EXPORT_SYMBOL(iget_locked); |
1da177e4 | 1115 | |
ad5e195a CH |
1116 | /* |
1117 | * search the inode cache for a matching inode number. | |
1118 | * If we find one, then the inode number we are trying to | |
1119 | * allocate is not unique and so we should not use it. | |
1120 | * | |
1121 | * Returns 1 if the inode number is unique, 0 if it is not. | |
1122 | */ | |
1123 | static int test_inode_iunique(struct super_block *sb, unsigned long ino) | |
1124 | { | |
1125 | struct hlist_head *b = inode_hashtable + hash(sb, ino); | |
ad5e195a CH |
1126 | struct inode *inode; |
1127 | ||
67a23c49 | 1128 | spin_lock(&inode_hash_lock); |
b67bfe0d | 1129 | hlist_for_each_entry(inode, b, i_hash) { |
67a23c49 DC |
1130 | if (inode->i_ino == ino && inode->i_sb == sb) { |
1131 | spin_unlock(&inode_hash_lock); | |
ad5e195a | 1132 | return 0; |
67a23c49 | 1133 | } |
ad5e195a | 1134 | } |
67a23c49 | 1135 | spin_unlock(&inode_hash_lock); |
ad5e195a CH |
1136 | |
1137 | return 1; | |
1138 | } | |
1139 | ||
1da177e4 LT |
1140 | /** |
1141 | * iunique - get a unique inode number | |
1142 | * @sb: superblock | |
1143 | * @max_reserved: highest reserved inode number | |
1144 | * | |
1145 | * Obtain an inode number that is unique on the system for a given | |
1146 | * superblock. This is used by file systems that have no natural | |
1147 | * permanent inode numbering system. An inode number is returned that | |
1148 | * is higher than the reserved limit but unique. | |
1149 | * | |
1150 | * BUGS: | |
1151 | * With a large number of inodes live on the file system this function | |
1152 | * currently becomes quite slow. | |
1153 | */ | |
1154 | ino_t iunique(struct super_block *sb, ino_t max_reserved) | |
1155 | { | |
866b04fc JL |
1156 | /* |
1157 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
1158 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
1159 | * here to attempt to avoid that. | |
1160 | */ | |
ad5e195a | 1161 | static DEFINE_SPINLOCK(iunique_lock); |
866b04fc | 1162 | static unsigned int counter; |
1da177e4 | 1163 | ino_t res; |
3361c7be | 1164 | |
ad5e195a | 1165 | spin_lock(&iunique_lock); |
3361c7be JL |
1166 | do { |
1167 | if (counter <= max_reserved) | |
1168 | counter = max_reserved + 1; | |
1da177e4 | 1169 | res = counter++; |
ad5e195a CH |
1170 | } while (!test_inode_iunique(sb, res)); |
1171 | spin_unlock(&iunique_lock); | |
1da177e4 | 1172 | |
3361c7be JL |
1173 | return res; |
1174 | } | |
1da177e4 LT |
1175 | EXPORT_SYMBOL(iunique); |
1176 | ||
1177 | struct inode *igrab(struct inode *inode) | |
1178 | { | |
250df6ed DC |
1179 | spin_lock(&inode->i_lock); |
1180 | if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) { | |
1da177e4 | 1181 | __iget(inode); |
250df6ed DC |
1182 | spin_unlock(&inode->i_lock); |
1183 | } else { | |
1184 | spin_unlock(&inode->i_lock); | |
1da177e4 LT |
1185 | /* |
1186 | * Handle the case where s_op->clear_inode is not been | |
1187 | * called yet, and somebody is calling igrab | |
1188 | * while the inode is getting freed. | |
1189 | */ | |
1190 | inode = NULL; | |
250df6ed | 1191 | } |
1da177e4 LT |
1192 | return inode; |
1193 | } | |
1da177e4 LT |
1194 | EXPORT_SYMBOL(igrab); |
1195 | ||
1196 | /** | |
0b2d0724 | 1197 | * ilookup5_nowait - search for an inode in the inode cache |
1da177e4 | 1198 | * @sb: super block of file system to search |
0b2d0724 | 1199 | * @hashval: hash value (usually inode number) to search for |
1da177e4 LT |
1200 | * @test: callback used for comparisons between inodes |
1201 | * @data: opaque data pointer to pass to @test | |
1da177e4 | 1202 | * |
0b2d0724 | 1203 | * Search for the inode specified by @hashval and @data in the inode cache. |
1da177e4 LT |
1204 | * If the inode is in the cache, the inode is returned with an incremented |
1205 | * reference count. | |
1206 | * | |
0b2d0724 CH |
1207 | * Note: I_NEW is not waited upon so you have to be very careful what you do |
1208 | * with the returned inode. You probably should be using ilookup5() instead. | |
1da177e4 | 1209 | * |
b6d0ad68 | 1210 | * Note2: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 | 1211 | */ |
0b2d0724 CH |
1212 | struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval, |
1213 | int (*test)(struct inode *, void *), void *data) | |
1da177e4 | 1214 | { |
0b2d0724 | 1215 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); |
1da177e4 LT |
1216 | struct inode *inode; |
1217 | ||
67a23c49 | 1218 | spin_lock(&inode_hash_lock); |
1da177e4 | 1219 | inode = find_inode(sb, head, test, data); |
67a23c49 | 1220 | spin_unlock(&inode_hash_lock); |
88bd5121 | 1221 | |
0b2d0724 | 1222 | return inode; |
88bd5121 | 1223 | } |
88bd5121 AA |
1224 | EXPORT_SYMBOL(ilookup5_nowait); |
1225 | ||
1226 | /** | |
1227 | * ilookup5 - search for an inode in the inode cache | |
1228 | * @sb: super block of file system to search | |
1229 | * @hashval: hash value (usually inode number) to search for | |
1230 | * @test: callback used for comparisons between inodes | |
1231 | * @data: opaque data pointer to pass to @test | |
1232 | * | |
0b2d0724 CH |
1233 | * Search for the inode specified by @hashval and @data in the inode cache, |
1234 | * and if the inode is in the cache, return the inode with an incremented | |
1235 | * reference count. Waits on I_NEW before returning the inode. | |
88bd5121 | 1236 | * returned with an incremented reference count. |
1da177e4 | 1237 | * |
0b2d0724 CH |
1238 | * This is a generalized version of ilookup() for file systems where the |
1239 | * inode number is not sufficient for unique identification of an inode. | |
1da177e4 | 1240 | * |
0b2d0724 | 1241 | * Note: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 LT |
1242 | */ |
1243 | struct inode *ilookup5(struct super_block *sb, unsigned long hashval, | |
1244 | int (*test)(struct inode *, void *), void *data) | |
1245 | { | |
0b2d0724 | 1246 | struct inode *inode = ilookup5_nowait(sb, hashval, test, data); |
1da177e4 | 1247 | |
0b2d0724 CH |
1248 | if (inode) |
1249 | wait_on_inode(inode); | |
1250 | return inode; | |
1da177e4 | 1251 | } |
1da177e4 LT |
1252 | EXPORT_SYMBOL(ilookup5); |
1253 | ||
1254 | /** | |
1255 | * ilookup - search for an inode in the inode cache | |
1256 | * @sb: super block of file system to search | |
1257 | * @ino: inode number to search for | |
1258 | * | |
0b2d0724 CH |
1259 | * Search for the inode @ino in the inode cache, and if the inode is in the |
1260 | * cache, the inode is returned with an incremented reference count. | |
1da177e4 LT |
1261 | */ |
1262 | struct inode *ilookup(struct super_block *sb, unsigned long ino) | |
1263 | { | |
1264 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1da177e4 LT |
1265 | struct inode *inode; |
1266 | ||
0b2d0724 CH |
1267 | spin_lock(&inode_hash_lock); |
1268 | inode = find_inode_fast(sb, head, ino); | |
1269 | spin_unlock(&inode_hash_lock); | |
1da177e4 | 1270 | |
1da177e4 | 1271 | if (inode) |
0b2d0724 CH |
1272 | wait_on_inode(inode); |
1273 | return inode; | |
1da177e4 | 1274 | } |
0b2d0724 | 1275 | EXPORT_SYMBOL(ilookup); |
1da177e4 | 1276 | |
fe032c42 TT |
1277 | /** |
1278 | * find_inode_nowait - find an inode in the inode cache | |
1279 | * @sb: super block of file system to search | |
1280 | * @hashval: hash value (usually inode number) to search for | |
1281 | * @match: callback used for comparisons between inodes | |
1282 | * @data: opaque data pointer to pass to @match | |
1283 | * | |
1284 | * Search for the inode specified by @hashval and @data in the inode | |
1285 | * cache, where the helper function @match will return 0 if the inode | |
1286 | * does not match, 1 if the inode does match, and -1 if the search | |
1287 | * should be stopped. The @match function must be responsible for | |
1288 | * taking the i_lock spin_lock and checking i_state for an inode being | |
1289 | * freed or being initialized, and incrementing the reference count | |
1290 | * before returning 1. It also must not sleep, since it is called with | |
1291 | * the inode_hash_lock spinlock held. | |
1292 | * | |
1293 | * This is a even more generalized version of ilookup5() when the | |
1294 | * function must never block --- find_inode() can block in | |
1295 | * __wait_on_freeing_inode() --- or when the caller can not increment | |
1296 | * the reference count because the resulting iput() might cause an | |
1297 | * inode eviction. The tradeoff is that the @match funtion must be | |
1298 | * very carefully implemented. | |
1299 | */ | |
1300 | struct inode *find_inode_nowait(struct super_block *sb, | |
1301 | unsigned long hashval, | |
1302 | int (*match)(struct inode *, unsigned long, | |
1303 | void *), | |
1304 | void *data) | |
1305 | { | |
1306 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1307 | struct inode *inode, *ret_inode = NULL; | |
1308 | int mval; | |
1309 | ||
1310 | spin_lock(&inode_hash_lock); | |
1311 | hlist_for_each_entry(inode, head, i_hash) { | |
1312 | if (inode->i_sb != sb) | |
1313 | continue; | |
1314 | mval = match(inode, hashval, data); | |
1315 | if (mval == 0) | |
1316 | continue; | |
1317 | if (mval == 1) | |
1318 | ret_inode = inode; | |
1319 | goto out; | |
1320 | } | |
1321 | out: | |
1322 | spin_unlock(&inode_hash_lock); | |
1323 | return ret_inode; | |
1324 | } | |
1325 | EXPORT_SYMBOL(find_inode_nowait); | |
1326 | ||
261bca86 AV |
1327 | int insert_inode_locked(struct inode *inode) |
1328 | { | |
1329 | struct super_block *sb = inode->i_sb; | |
1330 | ino_t ino = inode->i_ino; | |
1331 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
261bca86 | 1332 | |
261bca86 | 1333 | while (1) { |
72a43d63 | 1334 | struct inode *old = NULL; |
67a23c49 | 1335 | spin_lock(&inode_hash_lock); |
b67bfe0d | 1336 | hlist_for_each_entry(old, head, i_hash) { |
72a43d63 AV |
1337 | if (old->i_ino != ino) |
1338 | continue; | |
1339 | if (old->i_sb != sb) | |
1340 | continue; | |
250df6ed DC |
1341 | spin_lock(&old->i_lock); |
1342 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1343 | spin_unlock(&old->i_lock); | |
72a43d63 | 1344 | continue; |
250df6ed | 1345 | } |
72a43d63 AV |
1346 | break; |
1347 | } | |
b67bfe0d | 1348 | if (likely(!old)) { |
250df6ed DC |
1349 | spin_lock(&inode->i_lock); |
1350 | inode->i_state |= I_NEW; | |
261bca86 | 1351 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1352 | spin_unlock(&inode->i_lock); |
67a23c49 | 1353 | spin_unlock(&inode_hash_lock); |
261bca86 AV |
1354 | return 0; |
1355 | } | |
1356 | __iget(old); | |
250df6ed | 1357 | spin_unlock(&old->i_lock); |
67a23c49 | 1358 | spin_unlock(&inode_hash_lock); |
261bca86 | 1359 | wait_on_inode(old); |
1d3382cb | 1360 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1361 | iput(old); |
1362 | return -EBUSY; | |
1363 | } | |
1364 | iput(old); | |
1365 | } | |
1366 | } | |
261bca86 AV |
1367 | EXPORT_SYMBOL(insert_inode_locked); |
1368 | ||
1369 | int insert_inode_locked4(struct inode *inode, unsigned long hashval, | |
1370 | int (*test)(struct inode *, void *), void *data) | |
1371 | { | |
1372 | struct super_block *sb = inode->i_sb; | |
1373 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
261bca86 | 1374 | |
261bca86 | 1375 | while (1) { |
72a43d63 AV |
1376 | struct inode *old = NULL; |
1377 | ||
67a23c49 | 1378 | spin_lock(&inode_hash_lock); |
b67bfe0d | 1379 | hlist_for_each_entry(old, head, i_hash) { |
72a43d63 AV |
1380 | if (old->i_sb != sb) |
1381 | continue; | |
1382 | if (!test(old, data)) | |
1383 | continue; | |
250df6ed DC |
1384 | spin_lock(&old->i_lock); |
1385 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1386 | spin_unlock(&old->i_lock); | |
72a43d63 | 1387 | continue; |
250df6ed | 1388 | } |
72a43d63 AV |
1389 | break; |
1390 | } | |
b67bfe0d | 1391 | if (likely(!old)) { |
250df6ed DC |
1392 | spin_lock(&inode->i_lock); |
1393 | inode->i_state |= I_NEW; | |
261bca86 | 1394 | hlist_add_head(&inode->i_hash, head); |
250df6ed | 1395 | spin_unlock(&inode->i_lock); |
67a23c49 | 1396 | spin_unlock(&inode_hash_lock); |
261bca86 AV |
1397 | return 0; |
1398 | } | |
1399 | __iget(old); | |
250df6ed | 1400 | spin_unlock(&old->i_lock); |
67a23c49 | 1401 | spin_unlock(&inode_hash_lock); |
261bca86 | 1402 | wait_on_inode(old); |
1d3382cb | 1403 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1404 | iput(old); |
1405 | return -EBUSY; | |
1406 | } | |
1407 | iput(old); | |
1408 | } | |
1409 | } | |
261bca86 AV |
1410 | EXPORT_SYMBOL(insert_inode_locked4); |
1411 | ||
1da177e4 | 1412 | |
45321ac5 AV |
1413 | int generic_delete_inode(struct inode *inode) |
1414 | { | |
1415 | return 1; | |
1416 | } | |
1417 | EXPORT_SYMBOL(generic_delete_inode); | |
1418 | ||
45321ac5 AV |
1419 | /* |
1420 | * Called when we're dropping the last reference | |
1421 | * to an inode. | |
22fe4042 | 1422 | * |
45321ac5 AV |
1423 | * Call the FS "drop_inode()" function, defaulting to |
1424 | * the legacy UNIX filesystem behaviour. If it tells | |
1425 | * us to evict inode, do so. Otherwise, retain inode | |
1426 | * in cache if fs is alive, sync and evict if fs is | |
1427 | * shutting down. | |
22fe4042 | 1428 | */ |
45321ac5 | 1429 | static void iput_final(struct inode *inode) |
1da177e4 LT |
1430 | { |
1431 | struct super_block *sb = inode->i_sb; | |
45321ac5 AV |
1432 | const struct super_operations *op = inode->i_sb->s_op; |
1433 | int drop; | |
1434 | ||
250df6ed DC |
1435 | WARN_ON(inode->i_state & I_NEW); |
1436 | ||
e7f59097 | 1437 | if (op->drop_inode) |
45321ac5 AV |
1438 | drop = op->drop_inode(inode); |
1439 | else | |
1440 | drop = generic_drop_inode(inode); | |
1da177e4 | 1441 | |
b2b2af8e DC |
1442 | if (!drop && (sb->s_flags & MS_ACTIVE)) { |
1443 | inode->i_state |= I_REFERENCED; | |
4eff96dd | 1444 | inode_add_lru(inode); |
b2b2af8e | 1445 | spin_unlock(&inode->i_lock); |
b2b2af8e DC |
1446 | return; |
1447 | } | |
1448 | ||
45321ac5 | 1449 | if (!drop) { |
991114c6 | 1450 | inode->i_state |= I_WILL_FREE; |
250df6ed | 1451 | spin_unlock(&inode->i_lock); |
1da177e4 | 1452 | write_inode_now(inode, 1); |
250df6ed | 1453 | spin_lock(&inode->i_lock); |
7ef0d737 | 1454 | WARN_ON(inode->i_state & I_NEW); |
991114c6 | 1455 | inode->i_state &= ~I_WILL_FREE; |
1da177e4 | 1456 | } |
7ccf19a8 | 1457 | |
991114c6 | 1458 | inode->i_state |= I_FREEING; |
c4ae0c65 ED |
1459 | if (!list_empty(&inode->i_lru)) |
1460 | inode_lru_list_del(inode); | |
b2b2af8e | 1461 | spin_unlock(&inode->i_lock); |
b2b2af8e | 1462 | |
644da596 | 1463 | evict(inode); |
1da177e4 LT |
1464 | } |
1465 | ||
1da177e4 | 1466 | /** |
6b3304b5 | 1467 | * iput - put an inode |
1da177e4 LT |
1468 | * @inode: inode to put |
1469 | * | |
1470 | * Puts an inode, dropping its usage count. If the inode use count hits | |
1471 | * zero, the inode is then freed and may also be destroyed. | |
1472 | * | |
1473 | * Consequently, iput() can sleep. | |
1474 | */ | |
1475 | void iput(struct inode *inode) | |
1476 | { | |
0ae45f63 TT |
1477 | if (!inode) |
1478 | return; | |
1479 | BUG_ON(inode->i_state & I_CLEAR); | |
1480 | retry: | |
1481 | if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock)) { | |
1482 | if (inode->i_nlink && (inode->i_state & I_DIRTY_TIME)) { | |
1483 | atomic_inc(&inode->i_count); | |
1484 | inode->i_state &= ~I_DIRTY_TIME; | |
1485 | spin_unlock(&inode->i_lock); | |
1486 | trace_writeback_lazytime_iput(inode); | |
1487 | mark_inode_dirty_sync(inode); | |
1488 | goto retry; | |
1489 | } | |
1490 | iput_final(inode); | |
1da177e4 LT |
1491 | } |
1492 | } | |
1da177e4 LT |
1493 | EXPORT_SYMBOL(iput); |
1494 | ||
1495 | /** | |
1496 | * bmap - find a block number in a file | |
1497 | * @inode: inode of file | |
1498 | * @block: block to find | |
1499 | * | |
1500 | * Returns the block number on the device holding the inode that | |
1501 | * is the disk block number for the block of the file requested. | |
1502 | * That is, asked for block 4 of inode 1 the function will return the | |
6b3304b5 | 1503 | * disk block relative to the disk start that holds that block of the |
1da177e4 LT |
1504 | * file. |
1505 | */ | |
6b3304b5 | 1506 | sector_t bmap(struct inode *inode, sector_t block) |
1da177e4 LT |
1507 | { |
1508 | sector_t res = 0; | |
1509 | if (inode->i_mapping->a_ops->bmap) | |
1510 | res = inode->i_mapping->a_ops->bmap(inode->i_mapping, block); | |
1511 | return res; | |
1512 | } | |
1da177e4 LT |
1513 | EXPORT_SYMBOL(bmap); |
1514 | ||
11ff6f05 MG |
1515 | /* |
1516 | * With relative atime, only update atime if the previous atime is | |
1517 | * earlier than either the ctime or mtime or if at least a day has | |
1518 | * passed since the last atime update. | |
1519 | */ | |
1520 | static int relatime_need_update(struct vfsmount *mnt, struct inode *inode, | |
1521 | struct timespec now) | |
1522 | { | |
1523 | ||
1524 | if (!(mnt->mnt_flags & MNT_RELATIME)) | |
1525 | return 1; | |
1526 | /* | |
1527 | * Is mtime younger than atime? If yes, update atime: | |
1528 | */ | |
1529 | if (timespec_compare(&inode->i_mtime, &inode->i_atime) >= 0) | |
1530 | return 1; | |
1531 | /* | |
1532 | * Is ctime younger than atime? If yes, update atime: | |
1533 | */ | |
1534 | if (timespec_compare(&inode->i_ctime, &inode->i_atime) >= 0) | |
1535 | return 1; | |
1536 | ||
1537 | /* | |
1538 | * Is the previous atime value older than a day? If yes, | |
1539 | * update atime: | |
1540 | */ | |
1541 | if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60) | |
1542 | return 1; | |
1543 | /* | |
1544 | * Good, we can skip the atime update: | |
1545 | */ | |
1546 | return 0; | |
1547 | } | |
1548 | ||
0ae45f63 | 1549 | int generic_update_time(struct inode *inode, struct timespec *time, int flags) |
c3b2da31 | 1550 | { |
0ae45f63 | 1551 | int iflags = I_DIRTY_TIME; |
c3b2da31 JB |
1552 | |
1553 | if (flags & S_ATIME) | |
1554 | inode->i_atime = *time; | |
1555 | if (flags & S_VERSION) | |
1556 | inode_inc_iversion(inode); | |
1557 | if (flags & S_CTIME) | |
1558 | inode->i_ctime = *time; | |
1559 | if (flags & S_MTIME) | |
1560 | inode->i_mtime = *time; | |
0ae45f63 TT |
1561 | |
1562 | if (!(inode->i_sb->s_flags & MS_LAZYTIME) || (flags & S_VERSION)) | |
1563 | iflags |= I_DIRTY_SYNC; | |
1564 | __mark_inode_dirty(inode, iflags); | |
c3b2da31 JB |
1565 | return 0; |
1566 | } | |
0ae45f63 TT |
1567 | EXPORT_SYMBOL(generic_update_time); |
1568 | ||
1569 | /* | |
1570 | * This does the actual work of updating an inodes time or version. Must have | |
1571 | * had called mnt_want_write() before calling this. | |
1572 | */ | |
1573 | static int update_time(struct inode *inode, struct timespec *time, int flags) | |
1574 | { | |
1575 | int (*update_time)(struct inode *, struct timespec *, int); | |
1576 | ||
1577 | update_time = inode->i_op->update_time ? inode->i_op->update_time : | |
1578 | generic_update_time; | |
1579 | ||
1580 | return update_time(inode, time, flags); | |
1581 | } | |
c3b2da31 | 1582 | |
1da177e4 | 1583 | /** |
869243a0 | 1584 | * touch_atime - update the access time |
185553b2 | 1585 | * @path: the &struct path to update |
1da177e4 LT |
1586 | * |
1587 | * Update the accessed time on an inode and mark it for writeback. | |
1588 | * This function automatically handles read only file systems and media, | |
1589 | * as well as the "noatime" flag and inode specific "noatime" markers. | |
1590 | */ | |
8fa9dd24 | 1591 | bool atime_needs_update(const struct path *path, struct inode *inode) |
1da177e4 | 1592 | { |
68ac1234 | 1593 | struct vfsmount *mnt = path->mnt; |
1da177e4 LT |
1594 | struct timespec now; |
1595 | ||
cdb70f3f | 1596 | if (inode->i_flags & S_NOATIME) |
8fa9dd24 | 1597 | return false; |
37756ced | 1598 | if (IS_NOATIME(inode)) |
8fa9dd24 | 1599 | return false; |
b2276138 | 1600 | if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)) |
8fa9dd24 | 1601 | return false; |
47ae32d6 | 1602 | |
cdb70f3f | 1603 | if (mnt->mnt_flags & MNT_NOATIME) |
8fa9dd24 | 1604 | return false; |
cdb70f3f | 1605 | if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) |
8fa9dd24 | 1606 | return false; |
1da177e4 LT |
1607 | |
1608 | now = current_fs_time(inode->i_sb); | |
11ff6f05 MG |
1609 | |
1610 | if (!relatime_need_update(mnt, inode, now)) | |
8fa9dd24 | 1611 | return false; |
11ff6f05 | 1612 | |
47ae32d6 | 1613 | if (timespec_equal(&inode->i_atime, &now)) |
8fa9dd24 N |
1614 | return false; |
1615 | ||
1616 | return true; | |
1617 | } | |
1618 | ||
1619 | void touch_atime(const struct path *path) | |
1620 | { | |
1621 | struct vfsmount *mnt = path->mnt; | |
1622 | struct inode *inode = d_inode(path->dentry); | |
1623 | struct timespec now; | |
1624 | ||
1625 | if (!atime_needs_update(path, inode)) | |
b12536c2 AK |
1626 | return; |
1627 | ||
5d37e9e6 | 1628 | if (!sb_start_write_trylock(inode->i_sb)) |
b12536c2 | 1629 | return; |
47ae32d6 | 1630 | |
8fa9dd24 | 1631 | if (__mnt_want_write(mnt) != 0) |
5d37e9e6 | 1632 | goto skip_update; |
c3b2da31 JB |
1633 | /* |
1634 | * File systems can error out when updating inodes if they need to | |
1635 | * allocate new space to modify an inode (such is the case for | |
1636 | * Btrfs), but since we touch atime while walking down the path we | |
1637 | * really don't care if we failed to update the atime of the file, | |
1638 | * so just ignore the return value. | |
2bc55652 AB |
1639 | * We may also fail on filesystems that have the ability to make parts |
1640 | * of the fs read only, e.g. subvolumes in Btrfs. | |
c3b2da31 | 1641 | */ |
8fa9dd24 | 1642 | now = current_fs_time(inode->i_sb); |
c3b2da31 | 1643 | update_time(inode, &now, S_ATIME); |
5d37e9e6 JK |
1644 | __mnt_drop_write(mnt); |
1645 | skip_update: | |
1646 | sb_end_write(inode->i_sb); | |
1da177e4 | 1647 | } |
869243a0 | 1648 | EXPORT_SYMBOL(touch_atime); |
1da177e4 | 1649 | |
3ed37648 CW |
1650 | /* |
1651 | * The logic we want is | |
1652 | * | |
1653 | * if suid or (sgid and xgrp) | |
1654 | * remove privs | |
1655 | */ | |
1656 | int should_remove_suid(struct dentry *dentry) | |
1657 | { | |
df2b1afd | 1658 | umode_t mode = d_inode(dentry)->i_mode; |
3ed37648 CW |
1659 | int kill = 0; |
1660 | ||
1661 | /* suid always must be killed */ | |
1662 | if (unlikely(mode & S_ISUID)) | |
1663 | kill = ATTR_KILL_SUID; | |
1664 | ||
1665 | /* | |
1666 | * sgid without any exec bits is just a mandatory locking mark; leave | |
1667 | * it alone. If some exec bits are set, it's a real sgid; kill it. | |
1668 | */ | |
1669 | if (unlikely((mode & S_ISGID) && (mode & S_IXGRP))) | |
1670 | kill |= ATTR_KILL_SGID; | |
1671 | ||
1672 | if (unlikely(kill && !capable(CAP_FSETID) && S_ISREG(mode))) | |
1673 | return kill; | |
1674 | ||
1675 | return 0; | |
1676 | } | |
1677 | EXPORT_SYMBOL(should_remove_suid); | |
1678 | ||
dbfae0cd JK |
1679 | /* |
1680 | * Return mask of changes for notify_change() that need to be done as a | |
1681 | * response to write or truncate. Return 0 if nothing has to be changed. | |
1682 | * Negative value on error (change should be denied). | |
1683 | */ | |
45f147a1 | 1684 | int dentry_needs_remove_privs(struct dentry *dentry) |
dbfae0cd | 1685 | { |
dbfae0cd JK |
1686 | struct inode *inode = d_inode(dentry); |
1687 | int mask = 0; | |
1688 | int ret; | |
1689 | ||
1690 | if (IS_NOSEC(inode)) | |
1691 | return 0; | |
1692 | ||
1693 | mask = should_remove_suid(dentry); | |
1694 | ret = security_inode_need_killpriv(dentry); | |
1695 | if (ret < 0) | |
1696 | return ret; | |
1697 | if (ret) | |
1698 | mask |= ATTR_KILL_PRIV; | |
1699 | return mask; | |
1700 | } | |
45f147a1 | 1701 | EXPORT_SYMBOL(dentry_needs_remove_privs); |
dbfae0cd JK |
1702 | |
1703 | static int __remove_privs(struct dentry *dentry, int kill) | |
3ed37648 CW |
1704 | { |
1705 | struct iattr newattrs; | |
1706 | ||
1707 | newattrs.ia_valid = ATTR_FORCE | kill; | |
27ac0ffe BF |
1708 | /* |
1709 | * Note we call this on write, so notify_change will not | |
1710 | * encounter any conflicting delegations: | |
1711 | */ | |
1712 | return notify_change(dentry, &newattrs, NULL); | |
3ed37648 CW |
1713 | } |
1714 | ||
5fa8e0a1 JK |
1715 | /* |
1716 | * Remove special file priviledges (suid, capabilities) when file is written | |
1717 | * to or truncated. | |
1718 | */ | |
1719 | int file_remove_privs(struct file *file) | |
3ed37648 CW |
1720 | { |
1721 | struct dentry *dentry = file->f_path.dentry; | |
df2b1afd | 1722 | struct inode *inode = d_inode(dentry); |
dbfae0cd | 1723 | int kill; |
3ed37648 CW |
1724 | int error = 0; |
1725 | ||
1726 | /* Fast path for nothing security related */ | |
1727 | if (IS_NOSEC(inode)) | |
1728 | return 0; | |
1729 | ||
dbfae0cd JK |
1730 | kill = file_needs_remove_privs(file); |
1731 | if (kill < 0) | |
1732 | return kill; | |
1733 | if (kill) | |
1734 | error = __remove_privs(dentry, kill); | |
2426f391 JK |
1735 | if (!error) |
1736 | inode_has_no_xattr(inode); | |
3ed37648 CW |
1737 | |
1738 | return error; | |
1739 | } | |
5fa8e0a1 | 1740 | EXPORT_SYMBOL(file_remove_privs); |
3ed37648 | 1741 | |
1da177e4 | 1742 | /** |
870f4817 CH |
1743 | * file_update_time - update mtime and ctime time |
1744 | * @file: file accessed | |
1da177e4 | 1745 | * |
870f4817 CH |
1746 | * Update the mtime and ctime members of an inode and mark the inode |
1747 | * for writeback. Note that this function is meant exclusively for | |
1748 | * usage in the file write path of filesystems, and filesystems may | |
1749 | * choose to explicitly ignore update via this function with the | |
2eadfc0e | 1750 | * S_NOCMTIME inode flag, e.g. for network filesystem where these |
c3b2da31 JB |
1751 | * timestamps are handled by the server. This can return an error for |
1752 | * file systems who need to allocate space in order to update an inode. | |
1da177e4 LT |
1753 | */ |
1754 | ||
c3b2da31 | 1755 | int file_update_time(struct file *file) |
1da177e4 | 1756 | { |
496ad9aa | 1757 | struct inode *inode = file_inode(file); |
1da177e4 | 1758 | struct timespec now; |
c3b2da31 JB |
1759 | int sync_it = 0; |
1760 | int ret; | |
1da177e4 | 1761 | |
ce06e0b2 | 1762 | /* First try to exhaust all avenues to not sync */ |
1da177e4 | 1763 | if (IS_NOCMTIME(inode)) |
c3b2da31 | 1764 | return 0; |
20ddee2c | 1765 | |
1da177e4 | 1766 | now = current_fs_time(inode->i_sb); |
ce06e0b2 AK |
1767 | if (!timespec_equal(&inode->i_mtime, &now)) |
1768 | sync_it = S_MTIME; | |
1da177e4 | 1769 | |
ce06e0b2 AK |
1770 | if (!timespec_equal(&inode->i_ctime, &now)) |
1771 | sync_it |= S_CTIME; | |
870f4817 | 1772 | |
ce06e0b2 AK |
1773 | if (IS_I_VERSION(inode)) |
1774 | sync_it |= S_VERSION; | |
7a224228 | 1775 | |
ce06e0b2 | 1776 | if (!sync_it) |
c3b2da31 | 1777 | return 0; |
ce06e0b2 AK |
1778 | |
1779 | /* Finally allowed to write? Takes lock. */ | |
eb04c282 | 1780 | if (__mnt_want_write_file(file)) |
c3b2da31 | 1781 | return 0; |
ce06e0b2 | 1782 | |
c3b2da31 | 1783 | ret = update_time(inode, &now, sync_it); |
eb04c282 | 1784 | __mnt_drop_write_file(file); |
c3b2da31 JB |
1785 | |
1786 | return ret; | |
1da177e4 | 1787 | } |
870f4817 | 1788 | EXPORT_SYMBOL(file_update_time); |
1da177e4 LT |
1789 | |
1790 | int inode_needs_sync(struct inode *inode) | |
1791 | { | |
1792 | if (IS_SYNC(inode)) | |
1793 | return 1; | |
1794 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) | |
1795 | return 1; | |
1796 | return 0; | |
1797 | } | |
1da177e4 LT |
1798 | EXPORT_SYMBOL(inode_needs_sync); |
1799 | ||
1da177e4 | 1800 | /* |
168a9fd6 MS |
1801 | * If we try to find an inode in the inode hash while it is being |
1802 | * deleted, we have to wait until the filesystem completes its | |
1803 | * deletion before reporting that it isn't found. This function waits | |
1804 | * until the deletion _might_ have completed. Callers are responsible | |
1805 | * to recheck inode state. | |
1806 | * | |
eaff8079 | 1807 | * It doesn't matter if I_NEW is not set initially, a call to |
250df6ed DC |
1808 | * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list |
1809 | * will DTRT. | |
1da177e4 LT |
1810 | */ |
1811 | static void __wait_on_freeing_inode(struct inode *inode) | |
1812 | { | |
1813 | wait_queue_head_t *wq; | |
eaff8079 CH |
1814 | DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW); |
1815 | wq = bit_waitqueue(&inode->i_state, __I_NEW); | |
1da177e4 | 1816 | prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE); |
250df6ed | 1817 | spin_unlock(&inode->i_lock); |
67a23c49 | 1818 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1819 | schedule(); |
1820 | finish_wait(wq, &wait.wait); | |
67a23c49 | 1821 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1822 | } |
1823 | ||
1da177e4 LT |
1824 | static __initdata unsigned long ihash_entries; |
1825 | static int __init set_ihash_entries(char *str) | |
1826 | { | |
1827 | if (!str) | |
1828 | return 0; | |
1829 | ihash_entries = simple_strtoul(str, &str, 0); | |
1830 | return 1; | |
1831 | } | |
1832 | __setup("ihash_entries=", set_ihash_entries); | |
1833 | ||
1834 | /* | |
1835 | * Initialize the waitqueues and inode hash table. | |
1836 | */ | |
1837 | void __init inode_init_early(void) | |
1838 | { | |
074b8517 | 1839 | unsigned int loop; |
1da177e4 LT |
1840 | |
1841 | /* If hashes are distributed across NUMA nodes, defer | |
1842 | * hash allocation until vmalloc space is available. | |
1843 | */ | |
1844 | if (hashdist) | |
1845 | return; | |
1846 | ||
1847 | inode_hashtable = | |
1848 | alloc_large_system_hash("Inode-cache", | |
1849 | sizeof(struct hlist_head), | |
1850 | ihash_entries, | |
1851 | 14, | |
1852 | HASH_EARLY, | |
1853 | &i_hash_shift, | |
1854 | &i_hash_mask, | |
31fe62b9 | 1855 | 0, |
1da177e4 LT |
1856 | 0); |
1857 | ||
074b8517 | 1858 | for (loop = 0; loop < (1U << i_hash_shift); loop++) |
1da177e4 LT |
1859 | INIT_HLIST_HEAD(&inode_hashtable[loop]); |
1860 | } | |
1861 | ||
74bf17cf | 1862 | void __init inode_init(void) |
1da177e4 | 1863 | { |
074b8517 | 1864 | unsigned int loop; |
1da177e4 LT |
1865 | |
1866 | /* inode slab cache */ | |
b0196009 PJ |
1867 | inode_cachep = kmem_cache_create("inode_cache", |
1868 | sizeof(struct inode), | |
1869 | 0, | |
1870 | (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC| | |
1871 | SLAB_MEM_SPREAD), | |
20c2df83 | 1872 | init_once); |
1da177e4 LT |
1873 | |
1874 | /* Hash may have been set up in inode_init_early */ | |
1875 | if (!hashdist) | |
1876 | return; | |
1877 | ||
1878 | inode_hashtable = | |
1879 | alloc_large_system_hash("Inode-cache", | |
1880 | sizeof(struct hlist_head), | |
1881 | ihash_entries, | |
1882 | 14, | |
1883 | 0, | |
1884 | &i_hash_shift, | |
1885 | &i_hash_mask, | |
31fe62b9 | 1886 | 0, |
1da177e4 LT |
1887 | 0); |
1888 | ||
074b8517 | 1889 | for (loop = 0; loop < (1U << i_hash_shift); loop++) |
1da177e4 LT |
1890 | INIT_HLIST_HEAD(&inode_hashtable[loop]); |
1891 | } | |
1892 | ||
1893 | void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev) | |
1894 | { | |
1895 | inode->i_mode = mode; | |
1896 | if (S_ISCHR(mode)) { | |
1897 | inode->i_fop = &def_chr_fops; | |
1898 | inode->i_rdev = rdev; | |
1899 | } else if (S_ISBLK(mode)) { | |
1900 | inode->i_fop = &def_blk_fops; | |
1901 | inode->i_rdev = rdev; | |
1902 | } else if (S_ISFIFO(mode)) | |
599a0ac1 | 1903 | inode->i_fop = &pipefifo_fops; |
1da177e4 | 1904 | else if (S_ISSOCK(mode)) |
bd9b51e7 | 1905 | ; /* leave it no_open_fops */ |
1da177e4 | 1906 | else |
af0d9ae8 MK |
1907 | printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for" |
1908 | " inode %s:%lu\n", mode, inode->i_sb->s_id, | |
1909 | inode->i_ino); | |
1da177e4 LT |
1910 | } |
1911 | EXPORT_SYMBOL(init_special_inode); | |
a1bd120d DM |
1912 | |
1913 | /** | |
eaae668d | 1914 | * inode_init_owner - Init uid,gid,mode for new inode according to posix standards |
a1bd120d DM |
1915 | * @inode: New inode |
1916 | * @dir: Directory inode | |
1917 | * @mode: mode of the new inode | |
1918 | */ | |
1919 | void inode_init_owner(struct inode *inode, const struct inode *dir, | |
62bb1091 | 1920 | umode_t mode) |
a1bd120d DM |
1921 | { |
1922 | inode->i_uid = current_fsuid(); | |
1923 | if (dir && dir->i_mode & S_ISGID) { | |
1924 | inode->i_gid = dir->i_gid; | |
1925 | if (S_ISDIR(mode)) | |
1926 | mode |= S_ISGID; | |
1927 | } else | |
1928 | inode->i_gid = current_fsgid(); | |
1929 | inode->i_mode = mode; | |
1930 | } | |
1931 | EXPORT_SYMBOL(inode_init_owner); | |
e795b717 | 1932 | |
2e149670 SH |
1933 | /** |
1934 | * inode_owner_or_capable - check current task permissions to inode | |
1935 | * @inode: inode being checked | |
1936 | * | |
23adbe12 AL |
1937 | * Return true if current either has CAP_FOWNER in a namespace with the |
1938 | * inode owner uid mapped, or owns the file. | |
e795b717 | 1939 | */ |
2e149670 | 1940 | bool inode_owner_or_capable(const struct inode *inode) |
e795b717 | 1941 | { |
23adbe12 AL |
1942 | struct user_namespace *ns; |
1943 | ||
92361636 | 1944 | if (uid_eq(current_fsuid(), inode->i_uid)) |
e795b717 | 1945 | return true; |
23adbe12 AL |
1946 | |
1947 | ns = current_user_ns(); | |
1948 | if (ns_capable(ns, CAP_FOWNER) && kuid_has_mapping(ns, inode->i_uid)) | |
e795b717 SH |
1949 | return true; |
1950 | return false; | |
1951 | } | |
2e149670 | 1952 | EXPORT_SYMBOL(inode_owner_or_capable); |
1d59d61f TM |
1953 | |
1954 | /* | |
1955 | * Direct i/o helper functions | |
1956 | */ | |
1957 | static void __inode_dio_wait(struct inode *inode) | |
1958 | { | |
1959 | wait_queue_head_t *wq = bit_waitqueue(&inode->i_state, __I_DIO_WAKEUP); | |
1960 | DEFINE_WAIT_BIT(q, &inode->i_state, __I_DIO_WAKEUP); | |
1961 | ||
1962 | do { | |
1963 | prepare_to_wait(wq, &q.wait, TASK_UNINTERRUPTIBLE); | |
1964 | if (atomic_read(&inode->i_dio_count)) | |
1965 | schedule(); | |
1966 | } while (atomic_read(&inode->i_dio_count)); | |
1967 | finish_wait(wq, &q.wait); | |
1968 | } | |
1969 | ||
1970 | /** | |
1971 | * inode_dio_wait - wait for outstanding DIO requests to finish | |
1972 | * @inode: inode to wait for | |
1973 | * | |
1974 | * Waits for all pending direct I/O requests to finish so that we can | |
1975 | * proceed with a truncate or equivalent operation. | |
1976 | * | |
1977 | * Must be called under a lock that serializes taking new references | |
1978 | * to i_dio_count, usually by inode->i_mutex. | |
1979 | */ | |
1980 | void inode_dio_wait(struct inode *inode) | |
1981 | { | |
1982 | if (atomic_read(&inode->i_dio_count)) | |
1983 | __inode_dio_wait(inode); | |
1984 | } | |
1985 | EXPORT_SYMBOL(inode_dio_wait); | |
1986 | ||
5f16f322 TT |
1987 | /* |
1988 | * inode_set_flags - atomically set some inode flags | |
1989 | * | |
1990 | * Note: the caller should be holding i_mutex, or else be sure that | |
1991 | * they have exclusive access to the inode structure (i.e., while the | |
1992 | * inode is being instantiated). The reason for the cmpxchg() loop | |
1993 | * --- which wouldn't be necessary if all code paths which modify | |
1994 | * i_flags actually followed this rule, is that there is at least one | |
5fa8e0a1 JK |
1995 | * code path which doesn't today so we use cmpxchg() out of an abundance |
1996 | * of caution. | |
5f16f322 TT |
1997 | * |
1998 | * In the long run, i_mutex is overkill, and we should probably look | |
1999 | * at using the i_lock spinlock to protect i_flags, and then make sure | |
2000 | * it is so documented in include/linux/fs.h and that all code follows | |
2001 | * the locking convention!! | |
2002 | */ | |
2003 | void inode_set_flags(struct inode *inode, unsigned int flags, | |
2004 | unsigned int mask) | |
2005 | { | |
2006 | unsigned int old_flags, new_flags; | |
2007 | ||
2008 | WARN_ON_ONCE(flags & ~mask); | |
2009 | do { | |
2010 | old_flags = ACCESS_ONCE(inode->i_flags); | |
2011 | new_flags = (old_flags & ~mask) | flags; | |
2012 | } while (unlikely(cmpxchg(&inode->i_flags, old_flags, | |
2013 | new_flags) != old_flags)); | |
2014 | } | |
2015 | EXPORT_SYMBOL(inode_set_flags); |