Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 | 2 | /* |
1da177e4 | 3 | * (C) 1997 Linus Torvalds |
4b4563dc | 4 | * (C) 1999 Andrea Arcangeli <andrea@suse.de> (dynamic inode allocation) |
1da177e4 | 5 | */ |
e59cc473 | 6 | #include <linux/export.h> |
1da177e4 | 7 | #include <linux/fs.h> |
5970e15d | 8 | #include <linux/filelock.h> |
1da177e4 | 9 | #include <linux/mm.h> |
1da177e4 | 10 | #include <linux/backing-dev.h> |
1da177e4 LT |
11 | #include <linux/hash.h> |
12 | #include <linux/swap.h> | |
13 | #include <linux/security.h> | |
1da177e4 | 14 | #include <linux/cdev.h> |
57c8a661 | 15 | #include <linux/memblock.h> |
3be25f49 | 16 | #include <linux/fsnotify.h> |
fc33a7bb | 17 | #include <linux/mount.h> |
f19d4a8f | 18 | #include <linux/posix_acl.h> |
9ce6e0be | 19 | #include <linux/prefetch.h> |
4b4563dc | 20 | #include <linux/buffer_head.h> /* for inode_has_buffers */ |
7ada4db8 | 21 | #include <linux/ratelimit.h> |
bc3b14cb | 22 | #include <linux/list_lru.h> |
ae5e165d | 23 | #include <linux/iversion.h> |
0ae45f63 | 24 | #include <trace/events/writeback.h> |
a66979ab | 25 | #include "internal.h" |
1da177e4 | 26 | |
250df6ed | 27 | /* |
4b4563dc | 28 | * Inode locking rules: |
250df6ed DC |
29 | * |
30 | * inode->i_lock protects: | |
10e14073 | 31 | * inode->i_state, inode->i_hash, __iget(), inode->i_io_list |
bc3b14cb | 32 | * Inode LRU list locks protect: |
98b745c6 | 33 | * inode->i_sb->s_inode_lru, inode->i_lru |
74278da9 DC |
34 | * inode->i_sb->s_inode_list_lock protects: |
35 | * inode->i_sb->s_inodes, inode->i_sb_list | |
f758eeab | 36 | * bdi->wb.list_lock protects: |
c7f54084 | 37 | * bdi->wb.b_{dirty,io,more_io,dirty_time}, inode->i_io_list |
67a23c49 DC |
38 | * inode_hash_lock protects: |
39 | * inode_hashtable, inode->i_hash | |
250df6ed DC |
40 | * |
41 | * Lock ordering: | |
55fa6091 | 42 | * |
74278da9 | 43 | * inode->i_sb->s_inode_list_lock |
55fa6091 | 44 | * inode->i_lock |
bc3b14cb | 45 | * Inode LRU list locks |
a66979ab | 46 | * |
f758eeab | 47 | * bdi->wb.list_lock |
a66979ab | 48 | * inode->i_lock |
67a23c49 DC |
49 | * |
50 | * inode_hash_lock | |
74278da9 | 51 | * inode->i_sb->s_inode_list_lock |
67a23c49 DC |
52 | * inode->i_lock |
53 | * | |
54 | * iunique_lock | |
55 | * inode_hash_lock | |
250df6ed DC |
56 | */ |
57 | ||
fa3536cc ED |
58 | static unsigned int i_hash_mask __read_mostly; |
59 | static unsigned int i_hash_shift __read_mostly; | |
67a23c49 DC |
60 | static struct hlist_head *inode_hashtable __read_mostly; |
61 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_hash_lock); | |
1da177e4 | 62 | |
7dcda1c9 JA |
63 | /* |
64 | * Empty aops. Can be used for the cases where the user does not | |
65 | * define any of the address_space operations. | |
66 | */ | |
67 | const struct address_space_operations empty_aops = { | |
68 | }; | |
69 | EXPORT_SYMBOL(empty_aops); | |
70 | ||
3942c07c GC |
71 | static DEFINE_PER_CPU(unsigned long, nr_inodes); |
72 | static DEFINE_PER_CPU(unsigned long, nr_unused); | |
cffbc8aa | 73 | |
6b3304b5 | 74 | static struct kmem_cache *inode_cachep __read_mostly; |
1da177e4 | 75 | |
3942c07c | 76 | static long get_nr_inodes(void) |
cffbc8aa | 77 | { |
3e880fb5 | 78 | int i; |
3942c07c | 79 | long sum = 0; |
3e880fb5 NP |
80 | for_each_possible_cpu(i) |
81 | sum += per_cpu(nr_inodes, i); | |
82 | return sum < 0 ? 0 : sum; | |
cffbc8aa DC |
83 | } |
84 | ||
3942c07c | 85 | static inline long get_nr_inodes_unused(void) |
cffbc8aa | 86 | { |
fcb94f72 | 87 | int i; |
3942c07c | 88 | long sum = 0; |
fcb94f72 DC |
89 | for_each_possible_cpu(i) |
90 | sum += per_cpu(nr_unused, i); | |
91 | return sum < 0 ? 0 : sum; | |
cffbc8aa DC |
92 | } |
93 | ||
3942c07c | 94 | long get_nr_dirty_inodes(void) |
cffbc8aa | 95 | { |
3e880fb5 | 96 | /* not actually dirty inodes, but a wild approximation */ |
3942c07c | 97 | long nr_dirty = get_nr_inodes() - get_nr_inodes_unused(); |
cffbc8aa | 98 | return nr_dirty > 0 ? nr_dirty : 0; |
cffbc8aa DC |
99 | } |
100 | ||
101 | /* | |
102 | * Handle nr_inode sysctl | |
103 | */ | |
104 | #ifdef CONFIG_SYSCTL | |
1d67fe58 LC |
105 | /* |
106 | * Statistics gathering.. | |
107 | */ | |
108 | static struct inodes_stat_t inodes_stat; | |
109 | ||
110 | static int proc_nr_inodes(struct ctl_table *table, int write, void *buffer, | |
111 | size_t *lenp, loff_t *ppos) | |
cffbc8aa DC |
112 | { |
113 | inodes_stat.nr_inodes = get_nr_inodes(); | |
fcb94f72 | 114 | inodes_stat.nr_unused = get_nr_inodes_unused(); |
3942c07c | 115 | return proc_doulongvec_minmax(table, write, buffer, lenp, ppos); |
cffbc8aa | 116 | } |
1d67fe58 LC |
117 | |
118 | static struct ctl_table inodes_sysctls[] = { | |
119 | { | |
120 | .procname = "inode-nr", | |
121 | .data = &inodes_stat, | |
122 | .maxlen = 2*sizeof(long), | |
123 | .mode = 0444, | |
124 | .proc_handler = proc_nr_inodes, | |
125 | }, | |
126 | { | |
127 | .procname = "inode-state", | |
128 | .data = &inodes_stat, | |
129 | .maxlen = 7*sizeof(long), | |
130 | .mode = 0444, | |
131 | .proc_handler = proc_nr_inodes, | |
132 | }, | |
133 | { } | |
134 | }; | |
135 | ||
136 | static int __init init_fs_inode_sysctls(void) | |
137 | { | |
138 | register_sysctl_init("fs", inodes_sysctls); | |
139 | return 0; | |
140 | } | |
141 | early_initcall(init_fs_inode_sysctls); | |
cffbc8aa DC |
142 | #endif |
143 | ||
bd9b51e7 AV |
144 | static int no_open(struct inode *inode, struct file *file) |
145 | { | |
146 | return -ENXIO; | |
147 | } | |
148 | ||
2cb1599f | 149 | /** |
6e7c2b4d | 150 | * inode_init_always - perform inode structure initialisation |
0bc02f3f RD |
151 | * @sb: superblock inode belongs to |
152 | * @inode: inode to initialise | |
2cb1599f DC |
153 | * |
154 | * These are initializations that need to be done on every inode | |
155 | * allocation as the fields are not initialised by slab allocation. | |
156 | */ | |
54e34621 | 157 | int inode_init_always(struct super_block *sb, struct inode *inode) |
1da177e4 | 158 | { |
6e1d5dcc | 159 | static const struct inode_operations empty_iops; |
bd9b51e7 | 160 | static const struct file_operations no_open_fops = {.open = no_open}; |
6b3304b5 | 161 | struct address_space *const mapping = &inode->i_data; |
2cb1599f DC |
162 | |
163 | inode->i_sb = sb; | |
164 | inode->i_blkbits = sb->s_blocksize_bits; | |
165 | inode->i_flags = 0; | |
8019ad13 | 166 | atomic64_set(&inode->i_sequence, 0); |
2cb1599f DC |
167 | atomic_set(&inode->i_count, 1); |
168 | inode->i_op = &empty_iops; | |
bd9b51e7 | 169 | inode->i_fop = &no_open_fops; |
edbb35cc | 170 | inode->i_ino = 0; |
a78ef704 | 171 | inode->__i_nlink = 1; |
3ddcd056 | 172 | inode->i_opflags = 0; |
d0a5b995 AG |
173 | if (sb->s_xattr) |
174 | inode->i_opflags |= IOP_XATTR; | |
92361636 EB |
175 | i_uid_write(inode, 0); |
176 | i_gid_write(inode, 0); | |
2cb1599f DC |
177 | atomic_set(&inode->i_writecount, 0); |
178 | inode->i_size = 0; | |
c75b1d94 | 179 | inode->i_write_hint = WRITE_LIFE_NOT_SET; |
2cb1599f DC |
180 | inode->i_blocks = 0; |
181 | inode->i_bytes = 0; | |
182 | inode->i_generation = 0; | |
2cb1599f | 183 | inode->i_pipe = NULL; |
2cb1599f | 184 | inode->i_cdev = NULL; |
61ba64fc | 185 | inode->i_link = NULL; |
84e710da | 186 | inode->i_dir_seq = 0; |
2cb1599f DC |
187 | inode->i_rdev = 0; |
188 | inode->dirtied_when = 0; | |
6146f0d5 | 189 | |
3d65ae46 TE |
190 | #ifdef CONFIG_CGROUP_WRITEBACK |
191 | inode->i_wb_frn_winner = 0; | |
192 | inode->i_wb_frn_avg_time = 0; | |
193 | inode->i_wb_frn_history = 0; | |
194 | #endif | |
195 | ||
2cb1599f DC |
196 | spin_lock_init(&inode->i_lock); |
197 | lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key); | |
198 | ||
9902af79 AV |
199 | init_rwsem(&inode->i_rwsem); |
200 | lockdep_set_class(&inode->i_rwsem, &sb->s_type->i_mutex_key); | |
2cb1599f | 201 | |
bd5fe6c5 | 202 | atomic_set(&inode->i_dio_count, 0); |
2cb1599f DC |
203 | |
204 | mapping->a_ops = &empty_aops; | |
205 | mapping->host = inode; | |
206 | mapping->flags = 0; | |
829bc787 | 207 | mapping->wb_err = 0; |
4bb5f5d9 | 208 | atomic_set(&mapping->i_mmap_writable, 0); |
09d91cda SL |
209 | #ifdef CONFIG_READ_ONLY_THP_FOR_FS |
210 | atomic_set(&mapping->nr_thps, 0); | |
211 | #endif | |
3c1d4378 | 212 | mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE); |
252aa6f5 | 213 | mapping->private_data = NULL; |
2cb1599f | 214 | mapping->writeback_index = 0; |
23ca067b SAS |
215 | init_rwsem(&mapping->invalidate_lock); |
216 | lockdep_set_class_and_name(&mapping->invalidate_lock, | |
217 | &sb->s_type->invalidate_lock_key, | |
218 | "mapping.invalidate_lock"); | |
2cb1599f DC |
219 | inode->i_private = NULL; |
220 | inode->i_mapping = mapping; | |
b3d9b7a3 | 221 | INIT_HLIST_HEAD(&inode->i_dentry); /* buggered by rcu freeing */ |
f19d4a8f AV |
222 | #ifdef CONFIG_FS_POSIX_ACL |
223 | inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED; | |
224 | #endif | |
2cb1599f | 225 | |
3be25f49 EP |
226 | #ifdef CONFIG_FSNOTIFY |
227 | inode->i_fsnotify_mask = 0; | |
228 | #endif | |
4a075e39 | 229 | inode->i_flctx = NULL; |
2e488f13 DM |
230 | |
231 | if (unlikely(security_inode_alloc(inode))) | |
232 | return -ENOMEM; | |
3e880fb5 | 233 | this_cpu_inc(nr_inodes); |
cffbc8aa | 234 | |
54e34621 | 235 | return 0; |
1da177e4 | 236 | } |
2cb1599f DC |
237 | EXPORT_SYMBOL(inode_init_always); |
238 | ||
fdb0da89 AV |
239 | void free_inode_nonrcu(struct inode *inode) |
240 | { | |
241 | kmem_cache_free(inode_cachep, inode); | |
242 | } | |
243 | EXPORT_SYMBOL(free_inode_nonrcu); | |
244 | ||
245 | static void i_callback(struct rcu_head *head) | |
246 | { | |
247 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
248 | if (inode->free_inode) | |
249 | inode->free_inode(inode); | |
250 | else | |
251 | free_inode_nonrcu(inode); | |
252 | } | |
253 | ||
2cb1599f DC |
254 | static struct inode *alloc_inode(struct super_block *sb) |
255 | { | |
fdb0da89 | 256 | const struct super_operations *ops = sb->s_op; |
2cb1599f DC |
257 | struct inode *inode; |
258 | ||
fdb0da89 AV |
259 | if (ops->alloc_inode) |
260 | inode = ops->alloc_inode(sb); | |
2cb1599f | 261 | else |
8b9f3ac5 | 262 | inode = alloc_inode_sb(sb, inode_cachep, GFP_KERNEL); |
2cb1599f | 263 | |
54e34621 CH |
264 | if (!inode) |
265 | return NULL; | |
266 | ||
267 | if (unlikely(inode_init_always(sb, inode))) { | |
fdb0da89 AV |
268 | if (ops->destroy_inode) { |
269 | ops->destroy_inode(inode); | |
270 | if (!ops->free_inode) | |
271 | return NULL; | |
272 | } | |
273 | inode->free_inode = ops->free_inode; | |
274 | i_callback(&inode->i_rcu); | |
54e34621 CH |
275 | return NULL; |
276 | } | |
277 | ||
278 | return inode; | |
2cb1599f | 279 | } |
1da177e4 | 280 | |
2e00c97e | 281 | void __destroy_inode(struct inode *inode) |
1da177e4 | 282 | { |
b7542f8c | 283 | BUG_ON(inode_has_buffers(inode)); |
52ebea74 | 284 | inode_detach_wb(inode); |
1da177e4 | 285 | security_inode_free(inode); |
3be25f49 | 286 | fsnotify_inode_delete(inode); |
f27a0fe0 | 287 | locks_free_lock_context(inode); |
7ada4db8 MS |
288 | if (!inode->i_nlink) { |
289 | WARN_ON(atomic_long_read(&inode->i_sb->s_remove_count) == 0); | |
290 | atomic_long_dec(&inode->i_sb->s_remove_count); | |
291 | } | |
292 | ||
f19d4a8f | 293 | #ifdef CONFIG_FS_POSIX_ACL |
b8a7a3a6 | 294 | if (inode->i_acl && !is_uncached_acl(inode->i_acl)) |
f19d4a8f | 295 | posix_acl_release(inode->i_acl); |
b8a7a3a6 | 296 | if (inode->i_default_acl && !is_uncached_acl(inode->i_default_acl)) |
f19d4a8f AV |
297 | posix_acl_release(inode->i_default_acl); |
298 | #endif | |
3e880fb5 | 299 | this_cpu_dec(nr_inodes); |
2e00c97e CH |
300 | } |
301 | EXPORT_SYMBOL(__destroy_inode); | |
302 | ||
56b0dacf | 303 | static void destroy_inode(struct inode *inode) |
2e00c97e | 304 | { |
fdb0da89 AV |
305 | const struct super_operations *ops = inode->i_sb->s_op; |
306 | ||
7ccf19a8 | 307 | BUG_ON(!list_empty(&inode->i_lru)); |
2e00c97e | 308 | __destroy_inode(inode); |
fdb0da89 AV |
309 | if (ops->destroy_inode) { |
310 | ops->destroy_inode(inode); | |
311 | if (!ops->free_inode) | |
312 | return; | |
313 | } | |
314 | inode->free_inode = ops->free_inode; | |
315 | call_rcu(&inode->i_rcu, i_callback); | |
1da177e4 | 316 | } |
1da177e4 | 317 | |
7ada4db8 MS |
318 | /** |
319 | * drop_nlink - directly drop an inode's link count | |
320 | * @inode: inode | |
321 | * | |
322 | * This is a low-level filesystem helper to replace any | |
323 | * direct filesystem manipulation of i_nlink. In cases | |
324 | * where we are attempting to track writes to the | |
325 | * filesystem, a decrement to zero means an imminent | |
326 | * write when the file is truncated and actually unlinked | |
327 | * on the filesystem. | |
328 | */ | |
329 | void drop_nlink(struct inode *inode) | |
330 | { | |
331 | WARN_ON(inode->i_nlink == 0); | |
332 | inode->__i_nlink--; | |
333 | if (!inode->i_nlink) | |
334 | atomic_long_inc(&inode->i_sb->s_remove_count); | |
335 | } | |
336 | EXPORT_SYMBOL(drop_nlink); | |
337 | ||
338 | /** | |
339 | * clear_nlink - directly zero an inode's link count | |
340 | * @inode: inode | |
341 | * | |
342 | * This is a low-level filesystem helper to replace any | |
343 | * direct filesystem manipulation of i_nlink. See | |
344 | * drop_nlink() for why we care about i_nlink hitting zero. | |
345 | */ | |
346 | void clear_nlink(struct inode *inode) | |
347 | { | |
348 | if (inode->i_nlink) { | |
349 | inode->__i_nlink = 0; | |
350 | atomic_long_inc(&inode->i_sb->s_remove_count); | |
351 | } | |
352 | } | |
353 | EXPORT_SYMBOL(clear_nlink); | |
354 | ||
355 | /** | |
356 | * set_nlink - directly set an inode's link count | |
357 | * @inode: inode | |
358 | * @nlink: new nlink (should be non-zero) | |
359 | * | |
360 | * This is a low-level filesystem helper to replace any | |
361 | * direct filesystem manipulation of i_nlink. | |
362 | */ | |
363 | void set_nlink(struct inode *inode, unsigned int nlink) | |
364 | { | |
365 | if (!nlink) { | |
7ada4db8 MS |
366 | clear_nlink(inode); |
367 | } else { | |
368 | /* Yes, some filesystems do change nlink from zero to one */ | |
369 | if (inode->i_nlink == 0) | |
370 | atomic_long_dec(&inode->i_sb->s_remove_count); | |
371 | ||
372 | inode->__i_nlink = nlink; | |
373 | } | |
374 | } | |
375 | EXPORT_SYMBOL(set_nlink); | |
376 | ||
377 | /** | |
378 | * inc_nlink - directly increment an inode's link count | |
379 | * @inode: inode | |
380 | * | |
381 | * This is a low-level filesystem helper to replace any | |
382 | * direct filesystem manipulation of i_nlink. Currently, | |
383 | * it is only here for parity with dec_nlink(). | |
384 | */ | |
385 | void inc_nlink(struct inode *inode) | |
386 | { | |
f4e0c30c AV |
387 | if (unlikely(inode->i_nlink == 0)) { |
388 | WARN_ON(!(inode->i_state & I_LINKABLE)); | |
7ada4db8 | 389 | atomic_long_dec(&inode->i_sb->s_remove_count); |
f4e0c30c | 390 | } |
7ada4db8 MS |
391 | |
392 | inode->__i_nlink++; | |
393 | } | |
394 | EXPORT_SYMBOL(inc_nlink); | |
395 | ||
ae23395d | 396 | static void __address_space_init_once(struct address_space *mapping) |
2aa15890 | 397 | { |
7b785645 | 398 | xa_init_flags(&mapping->i_pages, XA_FLAGS_LOCK_IRQ | XA_FLAGS_ACCOUNT); |
c8c06efa | 399 | init_rwsem(&mapping->i_mmap_rwsem); |
2aa15890 MS |
400 | INIT_LIST_HEAD(&mapping->private_list); |
401 | spin_lock_init(&mapping->private_lock); | |
f808c13f | 402 | mapping->i_mmap = RB_ROOT_CACHED; |
2aa15890 | 403 | } |
ae23395d DC |
404 | |
405 | void address_space_init_once(struct address_space *mapping) | |
406 | { | |
407 | memset(mapping, 0, sizeof(*mapping)); | |
408 | __address_space_init_once(mapping); | |
409 | } | |
2aa15890 MS |
410 | EXPORT_SYMBOL(address_space_init_once); |
411 | ||
1da177e4 LT |
412 | /* |
413 | * These are initializations that only need to be done | |
414 | * once, because the fields are idempotent across use | |
415 | * of the inode, so let the slab aware of that. | |
416 | */ | |
417 | void inode_init_once(struct inode *inode) | |
418 | { | |
419 | memset(inode, 0, sizeof(*inode)); | |
420 | INIT_HLIST_NODE(&inode->i_hash); | |
1da177e4 | 421 | INIT_LIST_HEAD(&inode->i_devices); |
c7f54084 | 422 | INIT_LIST_HEAD(&inode->i_io_list); |
6c60d2b5 | 423 | INIT_LIST_HEAD(&inode->i_wb_list); |
7ccf19a8 | 424 | INIT_LIST_HEAD(&inode->i_lru); |
18cc912b | 425 | INIT_LIST_HEAD(&inode->i_sb_list); |
ae23395d | 426 | __address_space_init_once(&inode->i_data); |
1da177e4 LT |
427 | i_size_ordered_init(inode); |
428 | } | |
1da177e4 LT |
429 | EXPORT_SYMBOL(inode_init_once); |
430 | ||
51cc5068 | 431 | static void init_once(void *foo) |
1da177e4 | 432 | { |
6b3304b5 | 433 | struct inode *inode = (struct inode *) foo; |
1da177e4 | 434 | |
a35afb83 | 435 | inode_init_once(inode); |
1da177e4 LT |
436 | } |
437 | ||
438 | /* | |
250df6ed | 439 | * inode->i_lock must be held |
1da177e4 | 440 | */ |
6b3304b5 | 441 | void __iget(struct inode *inode) |
1da177e4 | 442 | { |
9e38d86f NP |
443 | atomic_inc(&inode->i_count); |
444 | } | |
2e147f1e | 445 | |
7de9c6ee AV |
446 | /* |
447 | * get additional reference to inode; caller must already hold one. | |
448 | */ | |
449 | void ihold(struct inode *inode) | |
450 | { | |
451 | WARN_ON(atomic_inc_return(&inode->i_count) < 2); | |
452 | } | |
453 | EXPORT_SYMBOL(ihold); | |
454 | ||
51b8c1fe | 455 | static void __inode_add_lru(struct inode *inode, bool rotate) |
9e38d86f | 456 | { |
51b8c1fe JW |
457 | if (inode->i_state & (I_DIRTY_ALL | I_SYNC | I_FREEING | I_WILL_FREE)) |
458 | return; | |
459 | if (atomic_read(&inode->i_count)) | |
460 | return; | |
461 | if (!(inode->i_sb->s_flags & SB_ACTIVE)) | |
462 | return; | |
463 | if (!mapping_shrinkable(&inode->i_data)) | |
464 | return; | |
465 | ||
bc3b14cb | 466 | if (list_lru_add(&inode->i_sb->s_inode_lru, &inode->i_lru)) |
fcb94f72 | 467 | this_cpu_inc(nr_unused); |
51b8c1fe | 468 | else if (rotate) |
563f4001 | 469 | inode->i_state |= I_REFERENCED; |
9e38d86f | 470 | } |
2e147f1e | 471 | |
4eff96dd JK |
472 | /* |
473 | * Add inode to LRU if needed (inode is unused and clean). | |
474 | * | |
475 | * Needs inode->i_lock held. | |
476 | */ | |
477 | void inode_add_lru(struct inode *inode) | |
478 | { | |
51b8c1fe | 479 | __inode_add_lru(inode, false); |
4eff96dd JK |
480 | } |
481 | ||
9e38d86f NP |
482 | static void inode_lru_list_del(struct inode *inode) |
483 | { | |
bc3b14cb | 484 | if (list_lru_del(&inode->i_sb->s_inode_lru, &inode->i_lru)) |
fcb94f72 | 485 | this_cpu_dec(nr_unused); |
1da177e4 LT |
486 | } |
487 | ||
646ec461 CH |
488 | /** |
489 | * inode_sb_list_add - add inode to the superblock list of inodes | |
490 | * @inode: inode to add | |
491 | */ | |
492 | void inode_sb_list_add(struct inode *inode) | |
493 | { | |
74278da9 | 494 | spin_lock(&inode->i_sb->s_inode_list_lock); |
55fa6091 | 495 | list_add(&inode->i_sb_list, &inode->i_sb->s_inodes); |
74278da9 | 496 | spin_unlock(&inode->i_sb->s_inode_list_lock); |
646ec461 CH |
497 | } |
498 | EXPORT_SYMBOL_GPL(inode_sb_list_add); | |
499 | ||
55fa6091 | 500 | static inline void inode_sb_list_del(struct inode *inode) |
646ec461 | 501 | { |
a209dfc7 | 502 | if (!list_empty(&inode->i_sb_list)) { |
74278da9 | 503 | spin_lock(&inode->i_sb->s_inode_list_lock); |
a209dfc7 | 504 | list_del_init(&inode->i_sb_list); |
74278da9 | 505 | spin_unlock(&inode->i_sb->s_inode_list_lock); |
a209dfc7 | 506 | } |
646ec461 CH |
507 | } |
508 | ||
4c51acbc DC |
509 | static unsigned long hash(struct super_block *sb, unsigned long hashval) |
510 | { | |
511 | unsigned long tmp; | |
512 | ||
513 | tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) / | |
514 | L1_CACHE_BYTES; | |
4b4563dc CH |
515 | tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> i_hash_shift); |
516 | return tmp & i_hash_mask; | |
4c51acbc DC |
517 | } |
518 | ||
519 | /** | |
520 | * __insert_inode_hash - hash an inode | |
521 | * @inode: unhashed inode | |
522 | * @hashval: unsigned long value used to locate this object in the | |
523 | * inode_hashtable. | |
524 | * | |
525 | * Add an inode to the inode hash for this superblock. | |
526 | */ | |
527 | void __insert_inode_hash(struct inode *inode, unsigned long hashval) | |
528 | { | |
646ec461 CH |
529 | struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval); |
530 | ||
67a23c49 | 531 | spin_lock(&inode_hash_lock); |
250df6ed | 532 | spin_lock(&inode->i_lock); |
3f19b2ab | 533 | hlist_add_head_rcu(&inode->i_hash, b); |
250df6ed | 534 | spin_unlock(&inode->i_lock); |
67a23c49 | 535 | spin_unlock(&inode_hash_lock); |
4c51acbc DC |
536 | } |
537 | EXPORT_SYMBOL(__insert_inode_hash); | |
538 | ||
4c51acbc | 539 | /** |
f2ee7abf | 540 | * __remove_inode_hash - remove an inode from the hash |
4c51acbc DC |
541 | * @inode: inode to unhash |
542 | * | |
543 | * Remove an inode from the superblock. | |
544 | */ | |
f2ee7abf | 545 | void __remove_inode_hash(struct inode *inode) |
4c51acbc | 546 | { |
67a23c49 | 547 | spin_lock(&inode_hash_lock); |
250df6ed | 548 | spin_lock(&inode->i_lock); |
3f19b2ab | 549 | hlist_del_init_rcu(&inode->i_hash); |
250df6ed | 550 | spin_unlock(&inode->i_lock); |
67a23c49 | 551 | spin_unlock(&inode_hash_lock); |
4c51acbc | 552 | } |
f2ee7abf | 553 | EXPORT_SYMBOL(__remove_inode_hash); |
4c51acbc | 554 | |
3e9d80a8 MWO |
555 | void dump_mapping(const struct address_space *mapping) |
556 | { | |
557 | struct inode *host; | |
558 | const struct address_space_operations *a_ops; | |
559 | struct hlist_node *dentry_first; | |
560 | struct dentry *dentry_ptr; | |
561 | struct dentry dentry; | |
562 | unsigned long ino; | |
563 | ||
564 | /* | |
565 | * If mapping is an invalid pointer, we don't want to crash | |
566 | * accessing it, so probe everything depending on it carefully. | |
567 | */ | |
568 | if (get_kernel_nofault(host, &mapping->host) || | |
569 | get_kernel_nofault(a_ops, &mapping->a_ops)) { | |
570 | pr_warn("invalid mapping:%px\n", mapping); | |
571 | return; | |
572 | } | |
573 | ||
574 | if (!host) { | |
575 | pr_warn("aops:%ps\n", a_ops); | |
576 | return; | |
577 | } | |
578 | ||
579 | if (get_kernel_nofault(dentry_first, &host->i_dentry.first) || | |
580 | get_kernel_nofault(ino, &host->i_ino)) { | |
581 | pr_warn("aops:%ps invalid inode:%px\n", a_ops, host); | |
582 | return; | |
583 | } | |
584 | ||
585 | if (!dentry_first) { | |
586 | pr_warn("aops:%ps ino:%lx\n", a_ops, ino); | |
587 | return; | |
588 | } | |
589 | ||
590 | dentry_ptr = container_of(dentry_first, struct dentry, d_u.d_alias); | |
591 | if (get_kernel_nofault(dentry, dentry_ptr)) { | |
592 | pr_warn("aops:%ps ino:%lx invalid dentry:%px\n", | |
593 | a_ops, ino, dentry_ptr); | |
594 | return; | |
595 | } | |
596 | ||
597 | /* | |
598 | * if dentry is corrupted, the %pd handler may still crash, | |
599 | * but it's unlikely that we reach here with a corrupt mapping | |
600 | */ | |
601 | pr_warn("aops:%ps ino:%lx dentry name:\"%pd\"\n", a_ops, ino, &dentry); | |
602 | } | |
603 | ||
dbd5768f | 604 | void clear_inode(struct inode *inode) |
b0683aa6 | 605 | { |
08142579 | 606 | /* |
b93b0163 | 607 | * We have to cycle the i_pages lock here because reclaim can be in the |
6ffcd825 | 608 | * process of removing the last page (in __filemap_remove_folio()) |
b93b0163 | 609 | * and we must not free the mapping under it. |
08142579 | 610 | */ |
b93b0163 | 611 | xa_lock_irq(&inode->i_data.i_pages); |
b0683aa6 | 612 | BUG_ON(inode->i_data.nrpages); |
786b3112 HD |
613 | /* |
614 | * Almost always, mapping_empty(&inode->i_data) here; but there are | |
615 | * two known and long-standing ways in which nodes may get left behind | |
616 | * (when deep radix-tree node allocation failed partway; or when THP | |
617 | * collapse_file() failed). Until those two known cases are cleaned up, | |
618 | * or a cleanup function is called here, do not BUG_ON(!mapping_empty), | |
619 | * nor even WARN_ON(!mapping_empty). | |
620 | */ | |
b93b0163 | 621 | xa_unlock_irq(&inode->i_data.i_pages); |
b0683aa6 AV |
622 | BUG_ON(!list_empty(&inode->i_data.private_list)); |
623 | BUG_ON(!(inode->i_state & I_FREEING)); | |
624 | BUG_ON(inode->i_state & I_CLEAR); | |
6c60d2b5 | 625 | BUG_ON(!list_empty(&inode->i_wb_list)); |
fa0d7e3d | 626 | /* don't need i_lock here, no concurrent mods to i_state */ |
b0683aa6 AV |
627 | inode->i_state = I_FREEING | I_CLEAR; |
628 | } | |
dbd5768f | 629 | EXPORT_SYMBOL(clear_inode); |
b0683aa6 | 630 | |
b2b2af8e DC |
631 | /* |
632 | * Free the inode passed in, removing it from the lists it is still connected | |
633 | * to. We remove any pages still attached to the inode and wait for any IO that | |
634 | * is still in progress before finally destroying the inode. | |
635 | * | |
636 | * An inode must already be marked I_FREEING so that we avoid the inode being | |
637 | * moved back onto lists if we race with other code that manipulates the lists | |
638 | * (e.g. writeback_single_inode). The caller is responsible for setting this. | |
639 | * | |
640 | * An inode must already be removed from the LRU list before being evicted from | |
641 | * the cache. This should occur atomically with setting the I_FREEING state | |
642 | * flag, so no inodes here should ever be on the LRU when being evicted. | |
643 | */ | |
644da596 | 644 | static void evict(struct inode *inode) |
b4272d4c AV |
645 | { |
646 | const struct super_operations *op = inode->i_sb->s_op; | |
647 | ||
b2b2af8e DC |
648 | BUG_ON(!(inode->i_state & I_FREEING)); |
649 | BUG_ON(!list_empty(&inode->i_lru)); | |
650 | ||
c7f54084 DC |
651 | if (!list_empty(&inode->i_io_list)) |
652 | inode_io_list_del(inode); | |
b12362bd | 653 | |
55fa6091 DC |
654 | inode_sb_list_del(inode); |
655 | ||
169ebd90 JK |
656 | /* |
657 | * Wait for flusher thread to be done with the inode so that filesystem | |
658 | * does not start destroying it while writeback is still running. Since | |
659 | * the inode has I_FREEING set, flusher thread won't start new work on | |
660 | * the inode. We just have to wait for running writeback to finish. | |
661 | */ | |
662 | inode_wait_for_writeback(inode); | |
7994e6f7 | 663 | |
be7ce416 AV |
664 | if (op->evict_inode) { |
665 | op->evict_inode(inode); | |
b4272d4c | 666 | } else { |
91b0abe3 | 667 | truncate_inode_pages_final(&inode->i_data); |
dbd5768f | 668 | clear_inode(inode); |
b4272d4c | 669 | } |
661074e9 AV |
670 | if (S_ISCHR(inode->i_mode) && inode->i_cdev) |
671 | cd_forget(inode); | |
b2b2af8e DC |
672 | |
673 | remove_inode_hash(inode); | |
674 | ||
675 | spin_lock(&inode->i_lock); | |
676 | wake_up_bit(&inode->i_state, __I_NEW); | |
677 | BUG_ON(inode->i_state != (I_FREEING | I_CLEAR)); | |
678 | spin_unlock(&inode->i_lock); | |
679 | ||
680 | destroy_inode(inode); | |
b4272d4c AV |
681 | } |
682 | ||
1da177e4 LT |
683 | /* |
684 | * dispose_list - dispose of the contents of a local list | |
685 | * @head: the head of the list to free | |
686 | * | |
687 | * Dispose-list gets a local list with local inodes in it, so it doesn't | |
688 | * need to worry about list corruption and SMP locks. | |
689 | */ | |
690 | static void dispose_list(struct list_head *head) | |
691 | { | |
1da177e4 LT |
692 | while (!list_empty(head)) { |
693 | struct inode *inode; | |
694 | ||
7ccf19a8 NP |
695 | inode = list_first_entry(head, struct inode, i_lru); |
696 | list_del_init(&inode->i_lru); | |
1da177e4 | 697 | |
644da596 | 698 | evict(inode); |
ac05fbb4 | 699 | cond_resched(); |
1da177e4 | 700 | } |
1da177e4 LT |
701 | } |
702 | ||
63997e98 AV |
703 | /** |
704 | * evict_inodes - evict all evictable inodes for a superblock | |
705 | * @sb: superblock to operate on | |
706 | * | |
707 | * Make sure that no inodes with zero refcount are retained. This is | |
1751e8a6 | 708 | * called by superblock shutdown after having SB_ACTIVE flag removed, |
63997e98 AV |
709 | * so any inode reaching zero refcount during or after that call will |
710 | * be immediately evicted. | |
1da177e4 | 711 | */ |
63997e98 | 712 | void evict_inodes(struct super_block *sb) |
1da177e4 | 713 | { |
63997e98 AV |
714 | struct inode *inode, *next; |
715 | LIST_HEAD(dispose); | |
1da177e4 | 716 | |
ac05fbb4 | 717 | again: |
74278da9 | 718 | spin_lock(&sb->s_inode_list_lock); |
63997e98 AV |
719 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
720 | if (atomic_read(&inode->i_count)) | |
aabb8fdb | 721 | continue; |
250df6ed DC |
722 | |
723 | spin_lock(&inode->i_lock); | |
724 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
725 | spin_unlock(&inode->i_lock); | |
1da177e4 | 726 | continue; |
250df6ed | 727 | } |
63997e98 AV |
728 | |
729 | inode->i_state |= I_FREEING; | |
02afc410 | 730 | inode_lru_list_del(inode); |
250df6ed | 731 | spin_unlock(&inode->i_lock); |
02afc410 | 732 | list_add(&inode->i_lru, &dispose); |
ac05fbb4 JB |
733 | |
734 | /* | |
735 | * We can have a ton of inodes to evict at unmount time given | |
736 | * enough memory, check to see if we need to go to sleep for a | |
737 | * bit so we don't livelock. | |
738 | */ | |
739 | if (need_resched()) { | |
740 | spin_unlock(&sb->s_inode_list_lock); | |
741 | cond_resched(); | |
742 | dispose_list(&dispose); | |
743 | goto again; | |
744 | } | |
1da177e4 | 745 | } |
74278da9 | 746 | spin_unlock(&sb->s_inode_list_lock); |
63997e98 AV |
747 | |
748 | dispose_list(&dispose); | |
1da177e4 | 749 | } |
799ea9e9 | 750 | EXPORT_SYMBOL_GPL(evict_inodes); |
1da177e4 | 751 | |
1da177e4 | 752 | /** |
a0318786 CH |
753 | * invalidate_inodes - attempt to free all inodes on a superblock |
754 | * @sb: superblock to operate on | |
93b270f7 | 755 | * @kill_dirty: flag to guide handling of dirty inodes |
1da177e4 | 756 | * |
a0318786 CH |
757 | * Attempts to free all inodes for a given superblock. If there were any |
758 | * busy inodes return a non-zero value, else zero. | |
93b270f7 N |
759 | * If @kill_dirty is set, discard dirty inodes too, otherwise treat |
760 | * them as busy. | |
1da177e4 | 761 | */ |
93b270f7 | 762 | int invalidate_inodes(struct super_block *sb, bool kill_dirty) |
1da177e4 | 763 | { |
cffbc8aa | 764 | int busy = 0; |
a0318786 CH |
765 | struct inode *inode, *next; |
766 | LIST_HEAD(dispose); | |
1da177e4 | 767 | |
04646aeb | 768 | again: |
74278da9 | 769 | spin_lock(&sb->s_inode_list_lock); |
a0318786 | 770 | list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) { |
250df6ed DC |
771 | spin_lock(&inode->i_lock); |
772 | if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) { | |
773 | spin_unlock(&inode->i_lock); | |
aabb8fdb | 774 | continue; |
250df6ed | 775 | } |
0ae45f63 | 776 | if (inode->i_state & I_DIRTY_ALL && !kill_dirty) { |
250df6ed | 777 | spin_unlock(&inode->i_lock); |
93b270f7 N |
778 | busy = 1; |
779 | continue; | |
780 | } | |
99a38919 | 781 | if (atomic_read(&inode->i_count)) { |
250df6ed | 782 | spin_unlock(&inode->i_lock); |
99a38919 | 783 | busy = 1; |
1da177e4 LT |
784 | continue; |
785 | } | |
99a38919 | 786 | |
99a38919 | 787 | inode->i_state |= I_FREEING; |
02afc410 | 788 | inode_lru_list_del(inode); |
250df6ed | 789 | spin_unlock(&inode->i_lock); |
02afc410 | 790 | list_add(&inode->i_lru, &dispose); |
04646aeb ES |
791 | if (need_resched()) { |
792 | spin_unlock(&sb->s_inode_list_lock); | |
793 | cond_resched(); | |
794 | dispose_list(&dispose); | |
795 | goto again; | |
796 | } | |
1da177e4 | 797 | } |
74278da9 | 798 | spin_unlock(&sb->s_inode_list_lock); |
1da177e4 | 799 | |
a0318786 | 800 | dispose_list(&dispose); |
1da177e4 LT |
801 | |
802 | return busy; | |
803 | } | |
1da177e4 | 804 | |
1da177e4 | 805 | /* |
bc3b14cb | 806 | * Isolate the inode from the LRU in preparation for freeing it. |
1da177e4 | 807 | * |
9e38d86f NP |
808 | * If the inode has the I_REFERENCED flag set, then it means that it has been |
809 | * used recently - the flag is set in iput_final(). When we encounter such an | |
810 | * inode, clear the flag and move it to the back of the LRU so it gets another | |
811 | * pass through the LRU before it gets reclaimed. This is necessary because of | |
812 | * the fact we are doing lazy LRU updates to minimise lock contention so the | |
813 | * LRU does not have strict ordering. Hence we don't want to reclaim inodes | |
814 | * with this flag set because they are the inodes that are out of order. | |
1da177e4 | 815 | */ |
3f97b163 VD |
816 | static enum lru_status inode_lru_isolate(struct list_head *item, |
817 | struct list_lru_one *lru, spinlock_t *lru_lock, void *arg) | |
1da177e4 | 818 | { |
bc3b14cb DC |
819 | struct list_head *freeable = arg; |
820 | struct inode *inode = container_of(item, struct inode, i_lru); | |
1da177e4 | 821 | |
bc3b14cb | 822 | /* |
51b8c1fe JW |
823 | * We are inverting the lru lock/inode->i_lock here, so use a |
824 | * trylock. If we fail to get the lock, just skip it. | |
bc3b14cb DC |
825 | */ |
826 | if (!spin_trylock(&inode->i_lock)) | |
827 | return LRU_SKIP; | |
1da177e4 | 828 | |
bc3b14cb | 829 | /* |
51b8c1fe JW |
830 | * Inodes can get referenced, redirtied, or repopulated while |
831 | * they're already on the LRU, and this can make them | |
832 | * unreclaimable for a while. Remove them lazily here; iput, | |
833 | * sync, or the last page cache deletion will requeue them. | |
bc3b14cb DC |
834 | */ |
835 | if (atomic_read(&inode->i_count) || | |
51b8c1fe JW |
836 | (inode->i_state & ~I_REFERENCED) || |
837 | !mapping_shrinkable(&inode->i_data)) { | |
3f97b163 | 838 | list_lru_isolate(lru, &inode->i_lru); |
bc3b14cb DC |
839 | spin_unlock(&inode->i_lock); |
840 | this_cpu_dec(nr_unused); | |
841 | return LRU_REMOVED; | |
842 | } | |
1da177e4 | 843 | |
51b8c1fe | 844 | /* Recently referenced inodes get one more pass */ |
69056ee6 | 845 | if (inode->i_state & I_REFERENCED) { |
bc3b14cb DC |
846 | inode->i_state &= ~I_REFERENCED; |
847 | spin_unlock(&inode->i_lock); | |
848 | return LRU_ROTATE; | |
849 | } | |
1da177e4 | 850 | |
51b8c1fe JW |
851 | /* |
852 | * On highmem systems, mapping_shrinkable() permits dropping | |
853 | * page cache in order to free up struct inodes: lowmem might | |
854 | * be under pressure before the cache inside the highmem zone. | |
855 | */ | |
7ae12c80 | 856 | if (inode_has_buffers(inode) || !mapping_empty(&inode->i_data)) { |
bc3b14cb DC |
857 | __iget(inode); |
858 | spin_unlock(&inode->i_lock); | |
859 | spin_unlock(lru_lock); | |
860 | if (remove_inode_buffers(inode)) { | |
861 | unsigned long reap; | |
862 | reap = invalidate_mapping_pages(&inode->i_data, 0, -1); | |
863 | if (current_is_kswapd()) | |
864 | __count_vm_events(KSWAPD_INODESTEAL, reap); | |
865 | else | |
866 | __count_vm_events(PGINODESTEAL, reap); | |
867 | if (current->reclaim_state) | |
868 | current->reclaim_state->reclaimed_slab += reap; | |
02afc410 | 869 | } |
bc3b14cb DC |
870 | iput(inode); |
871 | spin_lock(lru_lock); | |
872 | return LRU_RETRY; | |
873 | } | |
02afc410 | 874 | |
bc3b14cb DC |
875 | WARN_ON(inode->i_state & I_NEW); |
876 | inode->i_state |= I_FREEING; | |
3f97b163 | 877 | list_lru_isolate_move(lru, &inode->i_lru, freeable); |
bc3b14cb | 878 | spin_unlock(&inode->i_lock); |
9e38d86f | 879 | |
bc3b14cb DC |
880 | this_cpu_dec(nr_unused); |
881 | return LRU_REMOVED; | |
882 | } | |
7ccf19a8 | 883 | |
bc3b14cb DC |
884 | /* |
885 | * Walk the superblock inode LRU for freeable inodes and attempt to free them. | |
886 | * This is called from the superblock shrinker function with a number of inodes | |
887 | * to trim from the LRU. Inodes to be freed are moved to a temporary list and | |
888 | * then are freed outside inode_lock by dispose_list(). | |
889 | */ | |
503c358c | 890 | long prune_icache_sb(struct super_block *sb, struct shrink_control *sc) |
bc3b14cb DC |
891 | { |
892 | LIST_HEAD(freeable); | |
893 | long freed; | |
1da177e4 | 894 | |
503c358c VD |
895 | freed = list_lru_shrink_walk(&sb->s_inode_lru, sc, |
896 | inode_lru_isolate, &freeable); | |
1da177e4 | 897 | dispose_list(&freeable); |
0a234c6d | 898 | return freed; |
1da177e4 LT |
899 | } |
900 | ||
1da177e4 LT |
901 | static void __wait_on_freeing_inode(struct inode *inode); |
902 | /* | |
903 | * Called with the inode lock held. | |
1da177e4 | 904 | */ |
6b3304b5 MK |
905 | static struct inode *find_inode(struct super_block *sb, |
906 | struct hlist_head *head, | |
907 | int (*test)(struct inode *, void *), | |
908 | void *data) | |
1da177e4 | 909 | { |
6b3304b5 | 910 | struct inode *inode = NULL; |
1da177e4 LT |
911 | |
912 | repeat: | |
b67bfe0d | 913 | hlist_for_each_entry(inode, head, i_hash) { |
5a3cd992 | 914 | if (inode->i_sb != sb) |
1da177e4 | 915 | continue; |
5a3cd992 | 916 | if (!test(inode, data)) |
1da177e4 | 917 | continue; |
5a3cd992 | 918 | spin_lock(&inode->i_lock); |
a4ffdde6 | 919 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
920 | __wait_on_freeing_inode(inode); |
921 | goto repeat; | |
922 | } | |
c2b6d621 AV |
923 | if (unlikely(inode->i_state & I_CREATING)) { |
924 | spin_unlock(&inode->i_lock); | |
925 | return ERR_PTR(-ESTALE); | |
926 | } | |
f7899bd5 | 927 | __iget(inode); |
250df6ed | 928 | spin_unlock(&inode->i_lock); |
f7899bd5 | 929 | return inode; |
1da177e4 | 930 | } |
f7899bd5 | 931 | return NULL; |
1da177e4 LT |
932 | } |
933 | ||
934 | /* | |
935 | * find_inode_fast is the fast path version of find_inode, see the comment at | |
936 | * iget_locked for details. | |
937 | */ | |
6b3304b5 MK |
938 | static struct inode *find_inode_fast(struct super_block *sb, |
939 | struct hlist_head *head, unsigned long ino) | |
1da177e4 | 940 | { |
6b3304b5 | 941 | struct inode *inode = NULL; |
1da177e4 LT |
942 | |
943 | repeat: | |
b67bfe0d | 944 | hlist_for_each_entry(inode, head, i_hash) { |
5a3cd992 | 945 | if (inode->i_ino != ino) |
1da177e4 | 946 | continue; |
5a3cd992 | 947 | if (inode->i_sb != sb) |
1da177e4 | 948 | continue; |
5a3cd992 | 949 | spin_lock(&inode->i_lock); |
a4ffdde6 | 950 | if (inode->i_state & (I_FREEING|I_WILL_FREE)) { |
1da177e4 LT |
951 | __wait_on_freeing_inode(inode); |
952 | goto repeat; | |
953 | } | |
c2b6d621 AV |
954 | if (unlikely(inode->i_state & I_CREATING)) { |
955 | spin_unlock(&inode->i_lock); | |
956 | return ERR_PTR(-ESTALE); | |
957 | } | |
f7899bd5 | 958 | __iget(inode); |
250df6ed | 959 | spin_unlock(&inode->i_lock); |
f7899bd5 | 960 | return inode; |
1da177e4 | 961 | } |
f7899bd5 | 962 | return NULL; |
8290c35f DC |
963 | } |
964 | ||
f991bd2e ED |
965 | /* |
966 | * Each cpu owns a range of LAST_INO_BATCH numbers. | |
967 | * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations, | |
968 | * to renew the exhausted range. | |
8290c35f | 969 | * |
f991bd2e ED |
970 | * This does not significantly increase overflow rate because every CPU can |
971 | * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is | |
972 | * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the | |
973 | * 2^32 range, and is a worst-case. Even a 50% wastage would only increase | |
974 | * overflow rate by 2x, which does not seem too significant. | |
975 | * | |
976 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
977 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
978 | * here to attempt to avoid that. | |
8290c35f | 979 | */ |
f991bd2e ED |
980 | #define LAST_INO_BATCH 1024 |
981 | static DEFINE_PER_CPU(unsigned int, last_ino); | |
982 | ||
85fe4025 | 983 | unsigned int get_next_ino(void) |
8290c35f | 984 | { |
f991bd2e ED |
985 | unsigned int *p = &get_cpu_var(last_ino); |
986 | unsigned int res = *p; | |
8290c35f | 987 | |
f991bd2e ED |
988 | #ifdef CONFIG_SMP |
989 | if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) { | |
990 | static atomic_t shared_last_ino; | |
991 | int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino); | |
992 | ||
993 | res = next - LAST_INO_BATCH; | |
994 | } | |
995 | #endif | |
996 | ||
2adc376c CM |
997 | res++; |
998 | /* get_next_ino should not provide a 0 inode number */ | |
999 | if (unlikely(!res)) | |
1000 | res++; | |
1001 | *p = res; | |
f991bd2e ED |
1002 | put_cpu_var(last_ino); |
1003 | return res; | |
8290c35f | 1004 | } |
85fe4025 | 1005 | EXPORT_SYMBOL(get_next_ino); |
8290c35f | 1006 | |
a209dfc7 ED |
1007 | /** |
1008 | * new_inode_pseudo - obtain an inode | |
1009 | * @sb: superblock | |
1010 | * | |
1011 | * Allocates a new inode for given superblock. | |
1012 | * Inode wont be chained in superblock s_inodes list | |
1013 | * This means : | |
1014 | * - fs can't be unmount | |
1015 | * - quotas, fsnotify, writeback can't work | |
1016 | */ | |
1017 | struct inode *new_inode_pseudo(struct super_block *sb) | |
1018 | { | |
1019 | struct inode *inode = alloc_inode(sb); | |
1020 | ||
1021 | if (inode) { | |
1022 | spin_lock(&inode->i_lock); | |
1023 | inode->i_state = 0; | |
1024 | spin_unlock(&inode->i_lock); | |
a209dfc7 ED |
1025 | } |
1026 | return inode; | |
1027 | } | |
1028 | ||
1da177e4 LT |
1029 | /** |
1030 | * new_inode - obtain an inode | |
1031 | * @sb: superblock | |
1032 | * | |
769848c0 | 1033 | * Allocates a new inode for given superblock. The default gfp_mask |
3c1d4378 | 1034 | * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE. |
769848c0 MG |
1035 | * If HIGHMEM pages are unsuitable or it is known that pages allocated |
1036 | * for the page cache are not reclaimable or migratable, | |
1037 | * mapping_set_gfp_mask() must be called with suitable flags on the | |
1038 | * newly created inode's mapping | |
1039 | * | |
1da177e4 LT |
1040 | */ |
1041 | struct inode *new_inode(struct super_block *sb) | |
1042 | { | |
6b3304b5 | 1043 | struct inode *inode; |
1da177e4 | 1044 | |
74278da9 | 1045 | spin_lock_prefetch(&sb->s_inode_list_lock); |
6b3304b5 | 1046 | |
a209dfc7 ED |
1047 | inode = new_inode_pseudo(sb); |
1048 | if (inode) | |
55fa6091 | 1049 | inode_sb_list_add(inode); |
1da177e4 LT |
1050 | return inode; |
1051 | } | |
1da177e4 LT |
1052 | EXPORT_SYMBOL(new_inode); |
1053 | ||
14358e6d | 1054 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
e096d0c7 JB |
1055 | void lockdep_annotate_inode_mutex_key(struct inode *inode) |
1056 | { | |
a3314a0e | 1057 | if (S_ISDIR(inode->i_mode)) { |
1e89a5e1 PZ |
1058 | struct file_system_type *type = inode->i_sb->s_type; |
1059 | ||
9a7aa12f | 1060 | /* Set new key only if filesystem hasn't already changed it */ |
9902af79 | 1061 | if (lockdep_match_class(&inode->i_rwsem, &type->i_mutex_key)) { |
9a7aa12f JK |
1062 | /* |
1063 | * ensure nobody is actually holding i_mutex | |
1064 | */ | |
9902af79 AV |
1065 | // mutex_destroy(&inode->i_mutex); |
1066 | init_rwsem(&inode->i_rwsem); | |
1067 | lockdep_set_class(&inode->i_rwsem, | |
9a7aa12f JK |
1068 | &type->i_mutex_dir_key); |
1069 | } | |
1e89a5e1 | 1070 | } |
e096d0c7 JB |
1071 | } |
1072 | EXPORT_SYMBOL(lockdep_annotate_inode_mutex_key); | |
14358e6d | 1073 | #endif |
e096d0c7 JB |
1074 | |
1075 | /** | |
1076 | * unlock_new_inode - clear the I_NEW state and wake up any waiters | |
1077 | * @inode: new inode to unlock | |
1078 | * | |
1079 | * Called when the inode is fully initialised to clear the new state of the | |
1080 | * inode and wake up anyone waiting for the inode to finish initialisation. | |
1081 | */ | |
1082 | void unlock_new_inode(struct inode *inode) | |
1083 | { | |
1084 | lockdep_annotate_inode_mutex_key(inode); | |
250df6ed | 1085 | spin_lock(&inode->i_lock); |
eaff8079 | 1086 | WARN_ON(!(inode->i_state & I_NEW)); |
c2b6d621 | 1087 | inode->i_state &= ~I_NEW & ~I_CREATING; |
310fa7a3 | 1088 | smp_mb(); |
250df6ed DC |
1089 | wake_up_bit(&inode->i_state, __I_NEW); |
1090 | spin_unlock(&inode->i_lock); | |
1da177e4 | 1091 | } |
1da177e4 LT |
1092 | EXPORT_SYMBOL(unlock_new_inode); |
1093 | ||
c2b6d621 AV |
1094 | void discard_new_inode(struct inode *inode) |
1095 | { | |
1096 | lockdep_annotate_inode_mutex_key(inode); | |
1097 | spin_lock(&inode->i_lock); | |
1098 | WARN_ON(!(inode->i_state & I_NEW)); | |
1099 | inode->i_state &= ~I_NEW; | |
1100 | smp_mb(); | |
1101 | wake_up_bit(&inode->i_state, __I_NEW); | |
1102 | spin_unlock(&inode->i_lock); | |
1103 | iput(inode); | |
1104 | } | |
1105 | EXPORT_SYMBOL(discard_new_inode); | |
1106 | ||
375e289e BF |
1107 | /** |
1108 | * lock_two_nondirectories - take two i_mutexes on non-directory objects | |
4fd699ae BF |
1109 | * |
1110 | * Lock any non-NULL argument that is not a directory. | |
1111 | * Zero, one or two objects may be locked by this function. | |
1112 | * | |
375e289e BF |
1113 | * @inode1: first inode to lock |
1114 | * @inode2: second inode to lock | |
1115 | */ | |
1116 | void lock_two_nondirectories(struct inode *inode1, struct inode *inode2) | |
1117 | { | |
4fd699ae BF |
1118 | if (inode1 > inode2) |
1119 | swap(inode1, inode2); | |
1120 | ||
1121 | if (inode1 && !S_ISDIR(inode1->i_mode)) | |
5955102c | 1122 | inode_lock(inode1); |
4fd699ae | 1123 | if (inode2 && !S_ISDIR(inode2->i_mode) && inode2 != inode1) |
5955102c | 1124 | inode_lock_nested(inode2, I_MUTEX_NONDIR2); |
375e289e BF |
1125 | } |
1126 | EXPORT_SYMBOL(lock_two_nondirectories); | |
1127 | ||
1128 | /** | |
1129 | * unlock_two_nondirectories - release locks from lock_two_nondirectories() | |
1130 | * @inode1: first inode to unlock | |
1131 | * @inode2: second inode to unlock | |
1132 | */ | |
1133 | void unlock_two_nondirectories(struct inode *inode1, struct inode *inode2) | |
1134 | { | |
4fd699ae | 1135 | if (inode1 && !S_ISDIR(inode1->i_mode)) |
5955102c | 1136 | inode_unlock(inode1); |
4fd699ae | 1137 | if (inode2 && !S_ISDIR(inode2->i_mode) && inode2 != inode1) |
5955102c | 1138 | inode_unlock(inode2); |
375e289e BF |
1139 | } |
1140 | EXPORT_SYMBOL(unlock_two_nondirectories); | |
1141 | ||
80ea09a0 MS |
1142 | /** |
1143 | * inode_insert5 - obtain an inode from a mounted file system | |
1144 | * @inode: pre-allocated inode to use for insert to cache | |
1145 | * @hashval: hash value (usually inode number) to get | |
1146 | * @test: callback used for comparisons between inodes | |
1147 | * @set: callback used to initialize a new struct inode | |
1148 | * @data: opaque data pointer to pass to @test and @set | |
1149 | * | |
1150 | * Search for the inode specified by @hashval and @data in the inode cache, | |
1151 | * and if present it is return it with an increased reference count. This is | |
1152 | * a variant of iget5_locked() for callers that don't want to fail on memory | |
1153 | * allocation of inode. | |
1154 | * | |
1155 | * If the inode is not in cache, insert the pre-allocated inode to cache and | |
1156 | * return it locked, hashed, and with the I_NEW flag set. The file system gets | |
1157 | * to fill it in before unlocking it via unlock_new_inode(). | |
1158 | * | |
1159 | * Note both @test and @set are called with the inode_hash_lock held, so can't | |
1160 | * sleep. | |
1161 | */ | |
1162 | struct inode *inode_insert5(struct inode *inode, unsigned long hashval, | |
1163 | int (*test)(struct inode *, void *), | |
1164 | int (*set)(struct inode *, void *), void *data) | |
1165 | { | |
1166 | struct hlist_head *head = inode_hashtable + hash(inode->i_sb, hashval); | |
1167 | struct inode *old; | |
1168 | ||
1169 | again: | |
1170 | spin_lock(&inode_hash_lock); | |
1171 | old = find_inode(inode->i_sb, head, test, data); | |
1172 | if (unlikely(old)) { | |
1173 | /* | |
1174 | * Uhhuh, somebody else created the same inode under us. | |
1175 | * Use the old inode instead of the preallocated one. | |
1176 | */ | |
1177 | spin_unlock(&inode_hash_lock); | |
c2b6d621 AV |
1178 | if (IS_ERR(old)) |
1179 | return NULL; | |
80ea09a0 MS |
1180 | wait_on_inode(old); |
1181 | if (unlikely(inode_unhashed(old))) { | |
1182 | iput(old); | |
1183 | goto again; | |
1184 | } | |
1185 | return old; | |
1186 | } | |
1187 | ||
1188 | if (set && unlikely(set(inode, data))) { | |
1189 | inode = NULL; | |
1190 | goto unlock; | |
1191 | } | |
1192 | ||
1193 | /* | |
1194 | * Return the locked inode with I_NEW set, the | |
1195 | * caller is responsible for filling in the contents | |
1196 | */ | |
1197 | spin_lock(&inode->i_lock); | |
1198 | inode->i_state |= I_NEW; | |
3f19b2ab | 1199 | hlist_add_head_rcu(&inode->i_hash, head); |
80ea09a0 | 1200 | spin_unlock(&inode->i_lock); |
18cc912b JL |
1201 | |
1202 | /* | |
1203 | * Add inode to the sb list if it's not already. It has I_NEW at this | |
1204 | * point, so it should be safe to test i_sb_list locklessly. | |
1205 | */ | |
1206 | if (list_empty(&inode->i_sb_list)) | |
e950564b | 1207 | inode_sb_list_add(inode); |
80ea09a0 MS |
1208 | unlock: |
1209 | spin_unlock(&inode_hash_lock); | |
1210 | ||
1211 | return inode; | |
1212 | } | |
1213 | EXPORT_SYMBOL(inode_insert5); | |
1214 | ||
0b2d0724 CH |
1215 | /** |
1216 | * iget5_locked - obtain an inode from a mounted file system | |
1217 | * @sb: super block of file system | |
1218 | * @hashval: hash value (usually inode number) to get | |
1219 | * @test: callback used for comparisons between inodes | |
1220 | * @set: callback used to initialize a new struct inode | |
1221 | * @data: opaque data pointer to pass to @test and @set | |
1222 | * | |
1223 | * Search for the inode specified by @hashval and @data in the inode cache, | |
1224 | * and if present it is return it with an increased reference count. This is | |
1225 | * a generalized version of iget_locked() for file systems where the inode | |
1226 | * number is not sufficient for unique identification of an inode. | |
1227 | * | |
1228 | * If the inode is not in cache, allocate a new inode and return it locked, | |
1229 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
1230 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 1231 | * |
0b2d0724 CH |
1232 | * Note both @test and @set are called with the inode_hash_lock held, so can't |
1233 | * sleep. | |
1da177e4 | 1234 | */ |
0b2d0724 CH |
1235 | struct inode *iget5_locked(struct super_block *sb, unsigned long hashval, |
1236 | int (*test)(struct inode *, void *), | |
1237 | int (*set)(struct inode *, void *), void *data) | |
1da177e4 | 1238 | { |
80ea09a0 | 1239 | struct inode *inode = ilookup5(sb, hashval, test, data); |
0b2d0724 | 1240 | |
80ea09a0 | 1241 | if (!inode) { |
e950564b | 1242 | struct inode *new = alloc_inode(sb); |
0b2d0724 | 1243 | |
80ea09a0 | 1244 | if (new) { |
e950564b | 1245 | new->i_state = 0; |
80ea09a0 MS |
1246 | inode = inode_insert5(new, hashval, test, set, data); |
1247 | if (unlikely(inode != new)) | |
e950564b | 1248 | destroy_inode(new); |
2864f301 | 1249 | } |
1da177e4 LT |
1250 | } |
1251 | return inode; | |
1da177e4 | 1252 | } |
0b2d0724 | 1253 | EXPORT_SYMBOL(iget5_locked); |
1da177e4 | 1254 | |
0b2d0724 CH |
1255 | /** |
1256 | * iget_locked - obtain an inode from a mounted file system | |
1257 | * @sb: super block of file system | |
1258 | * @ino: inode number to get | |
1259 | * | |
1260 | * Search for the inode specified by @ino in the inode cache and if present | |
1261 | * return it with an increased reference count. This is for file systems | |
1262 | * where the inode number is sufficient for unique identification of an inode. | |
1263 | * | |
1264 | * If the inode is not in cache, allocate a new inode and return it locked, | |
1265 | * hashed, and with the I_NEW flag set. The file system gets to fill it in | |
1266 | * before unlocking it via unlock_new_inode(). | |
1da177e4 | 1267 | */ |
0b2d0724 | 1268 | struct inode *iget_locked(struct super_block *sb, unsigned long ino) |
1da177e4 | 1269 | { |
0b2d0724 | 1270 | struct hlist_head *head = inode_hashtable + hash(sb, ino); |
6b3304b5 | 1271 | struct inode *inode; |
2864f301 | 1272 | again: |
0b2d0724 CH |
1273 | spin_lock(&inode_hash_lock); |
1274 | inode = find_inode_fast(sb, head, ino); | |
1275 | spin_unlock(&inode_hash_lock); | |
1276 | if (inode) { | |
c2b6d621 AV |
1277 | if (IS_ERR(inode)) |
1278 | return NULL; | |
0b2d0724 | 1279 | wait_on_inode(inode); |
2864f301 AV |
1280 | if (unlikely(inode_unhashed(inode))) { |
1281 | iput(inode); | |
1282 | goto again; | |
1283 | } | |
0b2d0724 CH |
1284 | return inode; |
1285 | } | |
1286 | ||
1da177e4 LT |
1287 | inode = alloc_inode(sb); |
1288 | if (inode) { | |
6b3304b5 | 1289 | struct inode *old; |
1da177e4 | 1290 | |
67a23c49 | 1291 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
1292 | /* We released the lock, so.. */ |
1293 | old = find_inode_fast(sb, head, ino); | |
1294 | if (!old) { | |
1295 | inode->i_ino = ino; | |
250df6ed DC |
1296 | spin_lock(&inode->i_lock); |
1297 | inode->i_state = I_NEW; | |
3f19b2ab | 1298 | hlist_add_head_rcu(&inode->i_hash, head); |
250df6ed | 1299 | spin_unlock(&inode->i_lock); |
55fa6091 | 1300 | inode_sb_list_add(inode); |
67a23c49 | 1301 | spin_unlock(&inode_hash_lock); |
1da177e4 LT |
1302 | |
1303 | /* Return the locked inode with I_NEW set, the | |
1304 | * caller is responsible for filling in the contents | |
1305 | */ | |
1306 | return inode; | |
1307 | } | |
1308 | ||
1309 | /* | |
1310 | * Uhhuh, somebody else created the same inode under | |
1311 | * us. Use the old inode instead of the one we just | |
1312 | * allocated. | |
1313 | */ | |
67a23c49 | 1314 | spin_unlock(&inode_hash_lock); |
1da177e4 | 1315 | destroy_inode(inode); |
c2b6d621 AV |
1316 | if (IS_ERR(old)) |
1317 | return NULL; | |
1da177e4 LT |
1318 | inode = old; |
1319 | wait_on_inode(inode); | |
2864f301 AV |
1320 | if (unlikely(inode_unhashed(inode))) { |
1321 | iput(inode); | |
1322 | goto again; | |
1323 | } | |
1da177e4 LT |
1324 | } |
1325 | return inode; | |
1326 | } | |
0b2d0724 | 1327 | EXPORT_SYMBOL(iget_locked); |
1da177e4 | 1328 | |
ad5e195a CH |
1329 | /* |
1330 | * search the inode cache for a matching inode number. | |
1331 | * If we find one, then the inode number we are trying to | |
1332 | * allocate is not unique and so we should not use it. | |
1333 | * | |
1334 | * Returns 1 if the inode number is unique, 0 if it is not. | |
1335 | */ | |
1336 | static int test_inode_iunique(struct super_block *sb, unsigned long ino) | |
1337 | { | |
1338 | struct hlist_head *b = inode_hashtable + hash(sb, ino); | |
ad5e195a CH |
1339 | struct inode *inode; |
1340 | ||
3f19b2ab DH |
1341 | hlist_for_each_entry_rcu(inode, b, i_hash) { |
1342 | if (inode->i_ino == ino && inode->i_sb == sb) | |
ad5e195a CH |
1343 | return 0; |
1344 | } | |
ad5e195a CH |
1345 | return 1; |
1346 | } | |
1347 | ||
1da177e4 LT |
1348 | /** |
1349 | * iunique - get a unique inode number | |
1350 | * @sb: superblock | |
1351 | * @max_reserved: highest reserved inode number | |
1352 | * | |
1353 | * Obtain an inode number that is unique on the system for a given | |
1354 | * superblock. This is used by file systems that have no natural | |
1355 | * permanent inode numbering system. An inode number is returned that | |
1356 | * is higher than the reserved limit but unique. | |
1357 | * | |
1358 | * BUGS: | |
1359 | * With a large number of inodes live on the file system this function | |
1360 | * currently becomes quite slow. | |
1361 | */ | |
1362 | ino_t iunique(struct super_block *sb, ino_t max_reserved) | |
1363 | { | |
866b04fc JL |
1364 | /* |
1365 | * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW | |
1366 | * error if st_ino won't fit in target struct field. Use 32bit counter | |
1367 | * here to attempt to avoid that. | |
1368 | */ | |
ad5e195a | 1369 | static DEFINE_SPINLOCK(iunique_lock); |
866b04fc | 1370 | static unsigned int counter; |
1da177e4 | 1371 | ino_t res; |
3361c7be | 1372 | |
3f19b2ab | 1373 | rcu_read_lock(); |
ad5e195a | 1374 | spin_lock(&iunique_lock); |
3361c7be JL |
1375 | do { |
1376 | if (counter <= max_reserved) | |
1377 | counter = max_reserved + 1; | |
1da177e4 | 1378 | res = counter++; |
ad5e195a CH |
1379 | } while (!test_inode_iunique(sb, res)); |
1380 | spin_unlock(&iunique_lock); | |
3f19b2ab | 1381 | rcu_read_unlock(); |
1da177e4 | 1382 | |
3361c7be JL |
1383 | return res; |
1384 | } | |
1da177e4 LT |
1385 | EXPORT_SYMBOL(iunique); |
1386 | ||
1387 | struct inode *igrab(struct inode *inode) | |
1388 | { | |
250df6ed DC |
1389 | spin_lock(&inode->i_lock); |
1390 | if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) { | |
1da177e4 | 1391 | __iget(inode); |
250df6ed DC |
1392 | spin_unlock(&inode->i_lock); |
1393 | } else { | |
1394 | spin_unlock(&inode->i_lock); | |
1da177e4 LT |
1395 | /* |
1396 | * Handle the case where s_op->clear_inode is not been | |
1397 | * called yet, and somebody is calling igrab | |
1398 | * while the inode is getting freed. | |
1399 | */ | |
1400 | inode = NULL; | |
250df6ed | 1401 | } |
1da177e4 LT |
1402 | return inode; |
1403 | } | |
1da177e4 LT |
1404 | EXPORT_SYMBOL(igrab); |
1405 | ||
1406 | /** | |
0b2d0724 | 1407 | * ilookup5_nowait - search for an inode in the inode cache |
1da177e4 | 1408 | * @sb: super block of file system to search |
0b2d0724 | 1409 | * @hashval: hash value (usually inode number) to search for |
1da177e4 LT |
1410 | * @test: callback used for comparisons between inodes |
1411 | * @data: opaque data pointer to pass to @test | |
1da177e4 | 1412 | * |
0b2d0724 | 1413 | * Search for the inode specified by @hashval and @data in the inode cache. |
1da177e4 LT |
1414 | * If the inode is in the cache, the inode is returned with an incremented |
1415 | * reference count. | |
1416 | * | |
0b2d0724 CH |
1417 | * Note: I_NEW is not waited upon so you have to be very careful what you do |
1418 | * with the returned inode. You probably should be using ilookup5() instead. | |
1da177e4 | 1419 | * |
b6d0ad68 | 1420 | * Note2: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 | 1421 | */ |
0b2d0724 CH |
1422 | struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval, |
1423 | int (*test)(struct inode *, void *), void *data) | |
1da177e4 | 1424 | { |
0b2d0724 | 1425 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); |
1da177e4 LT |
1426 | struct inode *inode; |
1427 | ||
67a23c49 | 1428 | spin_lock(&inode_hash_lock); |
1da177e4 | 1429 | inode = find_inode(sb, head, test, data); |
67a23c49 | 1430 | spin_unlock(&inode_hash_lock); |
88bd5121 | 1431 | |
c2b6d621 | 1432 | return IS_ERR(inode) ? NULL : inode; |
88bd5121 | 1433 | } |
88bd5121 AA |
1434 | EXPORT_SYMBOL(ilookup5_nowait); |
1435 | ||
1436 | /** | |
1437 | * ilookup5 - search for an inode in the inode cache | |
1438 | * @sb: super block of file system to search | |
1439 | * @hashval: hash value (usually inode number) to search for | |
1440 | * @test: callback used for comparisons between inodes | |
1441 | * @data: opaque data pointer to pass to @test | |
1442 | * | |
0b2d0724 CH |
1443 | * Search for the inode specified by @hashval and @data in the inode cache, |
1444 | * and if the inode is in the cache, return the inode with an incremented | |
1445 | * reference count. Waits on I_NEW before returning the inode. | |
88bd5121 | 1446 | * returned with an incremented reference count. |
1da177e4 | 1447 | * |
0b2d0724 CH |
1448 | * This is a generalized version of ilookup() for file systems where the |
1449 | * inode number is not sufficient for unique identification of an inode. | |
1da177e4 | 1450 | * |
0b2d0724 | 1451 | * Note: @test is called with the inode_hash_lock held, so can't sleep. |
1da177e4 LT |
1452 | */ |
1453 | struct inode *ilookup5(struct super_block *sb, unsigned long hashval, | |
1454 | int (*test)(struct inode *, void *), void *data) | |
1455 | { | |
2864f301 AV |
1456 | struct inode *inode; |
1457 | again: | |
1458 | inode = ilookup5_nowait(sb, hashval, test, data); | |
1459 | if (inode) { | |
0b2d0724 | 1460 | wait_on_inode(inode); |
2864f301 AV |
1461 | if (unlikely(inode_unhashed(inode))) { |
1462 | iput(inode); | |
1463 | goto again; | |
1464 | } | |
1465 | } | |
0b2d0724 | 1466 | return inode; |
1da177e4 | 1467 | } |
1da177e4 LT |
1468 | EXPORT_SYMBOL(ilookup5); |
1469 | ||
1470 | /** | |
1471 | * ilookup - search for an inode in the inode cache | |
1472 | * @sb: super block of file system to search | |
1473 | * @ino: inode number to search for | |
1474 | * | |
0b2d0724 CH |
1475 | * Search for the inode @ino in the inode cache, and if the inode is in the |
1476 | * cache, the inode is returned with an incremented reference count. | |
1da177e4 LT |
1477 | */ |
1478 | struct inode *ilookup(struct super_block *sb, unsigned long ino) | |
1479 | { | |
1480 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1da177e4 | 1481 | struct inode *inode; |
2864f301 | 1482 | again: |
0b2d0724 CH |
1483 | spin_lock(&inode_hash_lock); |
1484 | inode = find_inode_fast(sb, head, ino); | |
1485 | spin_unlock(&inode_hash_lock); | |
1da177e4 | 1486 | |
2864f301 | 1487 | if (inode) { |
c2b6d621 AV |
1488 | if (IS_ERR(inode)) |
1489 | return NULL; | |
0b2d0724 | 1490 | wait_on_inode(inode); |
2864f301 AV |
1491 | if (unlikely(inode_unhashed(inode))) { |
1492 | iput(inode); | |
1493 | goto again; | |
1494 | } | |
1495 | } | |
0b2d0724 | 1496 | return inode; |
1da177e4 | 1497 | } |
0b2d0724 | 1498 | EXPORT_SYMBOL(ilookup); |
1da177e4 | 1499 | |
fe032c42 TT |
1500 | /** |
1501 | * find_inode_nowait - find an inode in the inode cache | |
1502 | * @sb: super block of file system to search | |
1503 | * @hashval: hash value (usually inode number) to search for | |
1504 | * @match: callback used for comparisons between inodes | |
1505 | * @data: opaque data pointer to pass to @match | |
1506 | * | |
1507 | * Search for the inode specified by @hashval and @data in the inode | |
1508 | * cache, where the helper function @match will return 0 if the inode | |
1509 | * does not match, 1 if the inode does match, and -1 if the search | |
1510 | * should be stopped. The @match function must be responsible for | |
1511 | * taking the i_lock spin_lock and checking i_state for an inode being | |
1512 | * freed or being initialized, and incrementing the reference count | |
1513 | * before returning 1. It also must not sleep, since it is called with | |
1514 | * the inode_hash_lock spinlock held. | |
1515 | * | |
1516 | * This is a even more generalized version of ilookup5() when the | |
1517 | * function must never block --- find_inode() can block in | |
1518 | * __wait_on_freeing_inode() --- or when the caller can not increment | |
1519 | * the reference count because the resulting iput() might cause an | |
1520 | * inode eviction. The tradeoff is that the @match funtion must be | |
1521 | * very carefully implemented. | |
1522 | */ | |
1523 | struct inode *find_inode_nowait(struct super_block *sb, | |
1524 | unsigned long hashval, | |
1525 | int (*match)(struct inode *, unsigned long, | |
1526 | void *), | |
1527 | void *data) | |
1528 | { | |
1529 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1530 | struct inode *inode, *ret_inode = NULL; | |
1531 | int mval; | |
1532 | ||
1533 | spin_lock(&inode_hash_lock); | |
1534 | hlist_for_each_entry(inode, head, i_hash) { | |
1535 | if (inode->i_sb != sb) | |
1536 | continue; | |
1537 | mval = match(inode, hashval, data); | |
1538 | if (mval == 0) | |
1539 | continue; | |
1540 | if (mval == 1) | |
1541 | ret_inode = inode; | |
1542 | goto out; | |
1543 | } | |
1544 | out: | |
1545 | spin_unlock(&inode_hash_lock); | |
1546 | return ret_inode; | |
1547 | } | |
1548 | EXPORT_SYMBOL(find_inode_nowait); | |
1549 | ||
3f19b2ab DH |
1550 | /** |
1551 | * find_inode_rcu - find an inode in the inode cache | |
1552 | * @sb: Super block of file system to search | |
1553 | * @hashval: Key to hash | |
1554 | * @test: Function to test match on an inode | |
1555 | * @data: Data for test function | |
1556 | * | |
1557 | * Search for the inode specified by @hashval and @data in the inode cache, | |
1558 | * where the helper function @test will return 0 if the inode does not match | |
1559 | * and 1 if it does. The @test function must be responsible for taking the | |
1560 | * i_lock spin_lock and checking i_state for an inode being freed or being | |
1561 | * initialized. | |
1562 | * | |
1563 | * If successful, this will return the inode for which the @test function | |
1564 | * returned 1 and NULL otherwise. | |
1565 | * | |
1566 | * The @test function is not permitted to take a ref on any inode presented. | |
1567 | * It is also not permitted to sleep. | |
1568 | * | |
1569 | * The caller must hold the RCU read lock. | |
1570 | */ | |
1571 | struct inode *find_inode_rcu(struct super_block *sb, unsigned long hashval, | |
1572 | int (*test)(struct inode *, void *), void *data) | |
1573 | { | |
1574 | struct hlist_head *head = inode_hashtable + hash(sb, hashval); | |
1575 | struct inode *inode; | |
1576 | ||
1577 | RCU_LOCKDEP_WARN(!rcu_read_lock_held(), | |
1578 | "suspicious find_inode_rcu() usage"); | |
1579 | ||
1580 | hlist_for_each_entry_rcu(inode, head, i_hash) { | |
1581 | if (inode->i_sb == sb && | |
1582 | !(READ_ONCE(inode->i_state) & (I_FREEING | I_WILL_FREE)) && | |
1583 | test(inode, data)) | |
1584 | return inode; | |
1585 | } | |
1586 | return NULL; | |
1587 | } | |
1588 | EXPORT_SYMBOL(find_inode_rcu); | |
1589 | ||
1590 | /** | |
961f3c89 | 1591 | * find_inode_by_ino_rcu - Find an inode in the inode cache |
3f19b2ab DH |
1592 | * @sb: Super block of file system to search |
1593 | * @ino: The inode number to match | |
1594 | * | |
1595 | * Search for the inode specified by @hashval and @data in the inode cache, | |
1596 | * where the helper function @test will return 0 if the inode does not match | |
1597 | * and 1 if it does. The @test function must be responsible for taking the | |
1598 | * i_lock spin_lock and checking i_state for an inode being freed or being | |
1599 | * initialized. | |
1600 | * | |
1601 | * If successful, this will return the inode for which the @test function | |
1602 | * returned 1 and NULL otherwise. | |
1603 | * | |
1604 | * The @test function is not permitted to take a ref on any inode presented. | |
1605 | * It is also not permitted to sleep. | |
1606 | * | |
1607 | * The caller must hold the RCU read lock. | |
1608 | */ | |
1609 | struct inode *find_inode_by_ino_rcu(struct super_block *sb, | |
1610 | unsigned long ino) | |
1611 | { | |
1612 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
1613 | struct inode *inode; | |
1614 | ||
1615 | RCU_LOCKDEP_WARN(!rcu_read_lock_held(), | |
1616 | "suspicious find_inode_by_ino_rcu() usage"); | |
1617 | ||
1618 | hlist_for_each_entry_rcu(inode, head, i_hash) { | |
1619 | if (inode->i_ino == ino && | |
1620 | inode->i_sb == sb && | |
1621 | !(READ_ONCE(inode->i_state) & (I_FREEING | I_WILL_FREE))) | |
1622 | return inode; | |
1623 | } | |
1624 | return NULL; | |
1625 | } | |
1626 | EXPORT_SYMBOL(find_inode_by_ino_rcu); | |
1627 | ||
261bca86 AV |
1628 | int insert_inode_locked(struct inode *inode) |
1629 | { | |
1630 | struct super_block *sb = inode->i_sb; | |
1631 | ino_t ino = inode->i_ino; | |
1632 | struct hlist_head *head = inode_hashtable + hash(sb, ino); | |
261bca86 | 1633 | |
261bca86 | 1634 | while (1) { |
72a43d63 | 1635 | struct inode *old = NULL; |
67a23c49 | 1636 | spin_lock(&inode_hash_lock); |
b67bfe0d | 1637 | hlist_for_each_entry(old, head, i_hash) { |
72a43d63 AV |
1638 | if (old->i_ino != ino) |
1639 | continue; | |
1640 | if (old->i_sb != sb) | |
1641 | continue; | |
250df6ed DC |
1642 | spin_lock(&old->i_lock); |
1643 | if (old->i_state & (I_FREEING|I_WILL_FREE)) { | |
1644 | spin_unlock(&old->i_lock); | |
72a43d63 | 1645 | continue; |
250df6ed | 1646 | } |
72a43d63 AV |
1647 | break; |
1648 | } | |
b67bfe0d | 1649 | if (likely(!old)) { |
250df6ed | 1650 | spin_lock(&inode->i_lock); |
c2b6d621 | 1651 | inode->i_state |= I_NEW | I_CREATING; |
3f19b2ab | 1652 | hlist_add_head_rcu(&inode->i_hash, head); |
250df6ed | 1653 | spin_unlock(&inode->i_lock); |
67a23c49 | 1654 | spin_unlock(&inode_hash_lock); |
261bca86 AV |
1655 | return 0; |
1656 | } | |
c2b6d621 AV |
1657 | if (unlikely(old->i_state & I_CREATING)) { |
1658 | spin_unlock(&old->i_lock); | |
1659 | spin_unlock(&inode_hash_lock); | |
1660 | return -EBUSY; | |
1661 | } | |
261bca86 | 1662 | __iget(old); |
250df6ed | 1663 | spin_unlock(&old->i_lock); |
67a23c49 | 1664 | spin_unlock(&inode_hash_lock); |
261bca86 | 1665 | wait_on_inode(old); |
1d3382cb | 1666 | if (unlikely(!inode_unhashed(old))) { |
261bca86 AV |
1667 | iput(old); |
1668 | return -EBUSY; | |
1669 | } | |
1670 | iput(old); | |
1671 | } | |
1672 | } | |
261bca86 AV |
1673 | EXPORT_SYMBOL(insert_inode_locked); |
1674 | ||
1675 | int insert_inode_locked4(struct inode *inode, unsigned long hashval, | |
1676 | int (*test)(struct inode *, void *), void *data) | |
1677 | { | |
c2b6d621 AV |
1678 | struct inode *old; |
1679 | ||
1680 | inode->i_state |= I_CREATING; | |
1681 | old = inode_insert5(inode, hashval, test, NULL, data); | |
261bca86 | 1682 | |
80ea09a0 | 1683 | if (old != inode) { |
261bca86 | 1684 | iput(old); |
80ea09a0 | 1685 | return -EBUSY; |
261bca86 | 1686 | } |
80ea09a0 | 1687 | return 0; |
261bca86 | 1688 | } |
261bca86 AV |
1689 | EXPORT_SYMBOL(insert_inode_locked4); |
1690 | ||
1da177e4 | 1691 | |
45321ac5 AV |
1692 | int generic_delete_inode(struct inode *inode) |
1693 | { | |
1694 | return 1; | |
1695 | } | |
1696 | EXPORT_SYMBOL(generic_delete_inode); | |
1697 | ||
45321ac5 AV |
1698 | /* |
1699 | * Called when we're dropping the last reference | |
1700 | * to an inode. | |
22fe4042 | 1701 | * |
45321ac5 AV |
1702 | * Call the FS "drop_inode()" function, defaulting to |
1703 | * the legacy UNIX filesystem behaviour. If it tells | |
1704 | * us to evict inode, do so. Otherwise, retain inode | |
1705 | * in cache if fs is alive, sync and evict if fs is | |
1706 | * shutting down. | |
22fe4042 | 1707 | */ |
45321ac5 | 1708 | static void iput_final(struct inode *inode) |
1da177e4 LT |
1709 | { |
1710 | struct super_block *sb = inode->i_sb; | |
45321ac5 | 1711 | const struct super_operations *op = inode->i_sb->s_op; |
3f19b2ab | 1712 | unsigned long state; |
45321ac5 AV |
1713 | int drop; |
1714 | ||
250df6ed DC |
1715 | WARN_ON(inode->i_state & I_NEW); |
1716 | ||
e7f59097 | 1717 | if (op->drop_inode) |
45321ac5 AV |
1718 | drop = op->drop_inode(inode); |
1719 | else | |
1720 | drop = generic_drop_inode(inode); | |
1da177e4 | 1721 | |
88149082 HL |
1722 | if (!drop && |
1723 | !(inode->i_state & I_DONTCACHE) && | |
1724 | (sb->s_flags & SB_ACTIVE)) { | |
51b8c1fe | 1725 | __inode_add_lru(inode, true); |
b2b2af8e | 1726 | spin_unlock(&inode->i_lock); |
b2b2af8e DC |
1727 | return; |
1728 | } | |
1729 | ||
3f19b2ab | 1730 | state = inode->i_state; |
45321ac5 | 1731 | if (!drop) { |
3f19b2ab | 1732 | WRITE_ONCE(inode->i_state, state | I_WILL_FREE); |
250df6ed | 1733 | spin_unlock(&inode->i_lock); |
3f19b2ab | 1734 | |
1da177e4 | 1735 | write_inode_now(inode, 1); |
3f19b2ab | 1736 | |
250df6ed | 1737 | spin_lock(&inode->i_lock); |
3f19b2ab DH |
1738 | state = inode->i_state; |
1739 | WARN_ON(state & I_NEW); | |
1740 | state &= ~I_WILL_FREE; | |
1da177e4 | 1741 | } |
7ccf19a8 | 1742 | |
3f19b2ab | 1743 | WRITE_ONCE(inode->i_state, state | I_FREEING); |
c4ae0c65 ED |
1744 | if (!list_empty(&inode->i_lru)) |
1745 | inode_lru_list_del(inode); | |
b2b2af8e | 1746 | spin_unlock(&inode->i_lock); |
b2b2af8e | 1747 | |
644da596 | 1748 | evict(inode); |
1da177e4 LT |
1749 | } |
1750 | ||
1da177e4 | 1751 | /** |
6b3304b5 | 1752 | * iput - put an inode |
1da177e4 LT |
1753 | * @inode: inode to put |
1754 | * | |
1755 | * Puts an inode, dropping its usage count. If the inode use count hits | |
1756 | * zero, the inode is then freed and may also be destroyed. | |
1757 | * | |
1758 | * Consequently, iput() can sleep. | |
1759 | */ | |
1760 | void iput(struct inode *inode) | |
1761 | { | |
0ae45f63 TT |
1762 | if (!inode) |
1763 | return; | |
1764 | BUG_ON(inode->i_state & I_CLEAR); | |
1765 | retry: | |
1766 | if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock)) { | |
1767 | if (inode->i_nlink && (inode->i_state & I_DIRTY_TIME)) { | |
1768 | atomic_inc(&inode->i_count); | |
0ae45f63 TT |
1769 | spin_unlock(&inode->i_lock); |
1770 | trace_writeback_lazytime_iput(inode); | |
1771 | mark_inode_dirty_sync(inode); | |
1772 | goto retry; | |
1773 | } | |
1774 | iput_final(inode); | |
1da177e4 LT |
1775 | } |
1776 | } | |
1da177e4 LT |
1777 | EXPORT_SYMBOL(iput); |
1778 | ||
30460e1e | 1779 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
1780 | /** |
1781 | * bmap - find a block number in a file | |
30460e1e CM |
1782 | * @inode: inode owning the block number being requested |
1783 | * @block: pointer containing the block to find | |
1da177e4 | 1784 | * |
2b8e8b55 | 1785 | * Replaces the value in ``*block`` with the block number on the device holding |
30460e1e CM |
1786 | * corresponding to the requested block number in the file. |
1787 | * That is, asked for block 4 of inode 1 the function will replace the | |
2b8e8b55 | 1788 | * 4 in ``*block``, with disk block relative to the disk start that holds that |
30460e1e CM |
1789 | * block of the file. |
1790 | * | |
1791 | * Returns -EINVAL in case of error, 0 otherwise. If mapping falls into a | |
2b8e8b55 | 1792 | * hole, returns 0 and ``*block`` is also set to 0. |
1da177e4 | 1793 | */ |
30460e1e | 1794 | int bmap(struct inode *inode, sector_t *block) |
1da177e4 | 1795 | { |
30460e1e CM |
1796 | if (!inode->i_mapping->a_ops->bmap) |
1797 | return -EINVAL; | |
1798 | ||
1799 | *block = inode->i_mapping->a_ops->bmap(inode->i_mapping, *block); | |
1800 | return 0; | |
1da177e4 | 1801 | } |
1da177e4 | 1802 | EXPORT_SYMBOL(bmap); |
30460e1e | 1803 | #endif |
1da177e4 | 1804 | |
11ff6f05 MG |
1805 | /* |
1806 | * With relative atime, only update atime if the previous atime is | |
d98ffa1a SK |
1807 | * earlier than or equal to either the ctime or mtime, |
1808 | * or if at least a day has passed since the last atime update. | |
11ff6f05 | 1809 | */ |
c6718543 | 1810 | static int relatime_need_update(struct vfsmount *mnt, struct inode *inode, |
6f22b664 | 1811 | struct timespec64 now) |
11ff6f05 MG |
1812 | { |
1813 | ||
c6718543 | 1814 | if (!(mnt->mnt_flags & MNT_RELATIME)) |
11ff6f05 MG |
1815 | return 1; |
1816 | /* | |
d98ffa1a | 1817 | * Is mtime younger than or equal to atime? If yes, update atime: |
11ff6f05 | 1818 | */ |
95582b00 | 1819 | if (timespec64_compare(&inode->i_mtime, &inode->i_atime) >= 0) |
11ff6f05 MG |
1820 | return 1; |
1821 | /* | |
d98ffa1a | 1822 | * Is ctime younger than or equal to atime? If yes, update atime: |
11ff6f05 | 1823 | */ |
95582b00 | 1824 | if (timespec64_compare(&inode->i_ctime, &inode->i_atime) >= 0) |
11ff6f05 MG |
1825 | return 1; |
1826 | ||
1827 | /* | |
1828 | * Is the previous atime value older than a day? If yes, | |
1829 | * update atime: | |
1830 | */ | |
1831 | if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60) | |
1832 | return 1; | |
1833 | /* | |
1834 | * Good, we can skip the atime update: | |
1835 | */ | |
1836 | return 0; | |
1837 | } | |
1838 | ||
95582b00 | 1839 | int generic_update_time(struct inode *inode, struct timespec64 *time, int flags) |
c3b2da31 | 1840 | { |
e20b14db EB |
1841 | int dirty_flags = 0; |
1842 | ||
1843 | if (flags & (S_ATIME | S_CTIME | S_MTIME)) { | |
1844 | if (flags & S_ATIME) | |
1845 | inode->i_atime = *time; | |
1846 | if (flags & S_CTIME) | |
1847 | inode->i_ctime = *time; | |
1848 | if (flags & S_MTIME) | |
1849 | inode->i_mtime = *time; | |
1850 | ||
1851 | if (inode->i_sb->s_flags & SB_LAZYTIME) | |
1852 | dirty_flags |= I_DIRTY_TIME; | |
1853 | else | |
1854 | dirty_flags |= I_DIRTY_SYNC; | |
1855 | } | |
1856 | ||
1857 | if ((flags & S_VERSION) && inode_maybe_inc_iversion(inode, false)) | |
1858 | dirty_flags |= I_DIRTY_SYNC; | |
1859 | ||
1860 | __mark_inode_dirty(inode, dirty_flags); | |
c3b2da31 JB |
1861 | return 0; |
1862 | } | |
0ae45f63 TT |
1863 | EXPORT_SYMBOL(generic_update_time); |
1864 | ||
1865 | /* | |
1866 | * This does the actual work of updating an inodes time or version. Must have | |
1867 | * had called mnt_want_write() before calling this. | |
1868 | */ | |
e60feb44 | 1869 | int inode_update_time(struct inode *inode, struct timespec64 *time, int flags) |
0ae45f63 | 1870 | { |
23b424d9 DD |
1871 | if (inode->i_op->update_time) |
1872 | return inode->i_op->update_time(inode, time, flags); | |
1873 | return generic_update_time(inode, time, flags); | |
0ae45f63 | 1874 | } |
e60feb44 | 1875 | EXPORT_SYMBOL(inode_update_time); |
c3b2da31 | 1876 | |
1da177e4 | 1877 | /** |
961f3c89 | 1878 | * atime_needs_update - update the access time |
185553b2 | 1879 | * @path: the &struct path to update |
30fdc8ee | 1880 | * @inode: inode to update |
1da177e4 LT |
1881 | * |
1882 | * Update the accessed time on an inode and mark it for writeback. | |
1883 | * This function automatically handles read only file systems and media, | |
1884 | * as well as the "noatime" flag and inode specific "noatime" markers. | |
1885 | */ | |
c6718543 | 1886 | bool atime_needs_update(const struct path *path, struct inode *inode) |
1da177e4 | 1887 | { |
68ac1234 | 1888 | struct vfsmount *mnt = path->mnt; |
95582b00 | 1889 | struct timespec64 now; |
1da177e4 | 1890 | |
cdb70f3f | 1891 | if (inode->i_flags & S_NOATIME) |
8fa9dd24 | 1892 | return false; |
0bd23d09 EB |
1893 | |
1894 | /* Atime updates will likely cause i_uid and i_gid to be written | |
1895 | * back improprely if their true value is unknown to the vfs. | |
1896 | */ | |
4609e1f1 | 1897 | if (HAS_UNMAPPED_ID(mnt_idmap(mnt), inode)) |
0bd23d09 EB |
1898 | return false; |
1899 | ||
37756ced | 1900 | if (IS_NOATIME(inode)) |
8fa9dd24 | 1901 | return false; |
1751e8a6 | 1902 | if ((inode->i_sb->s_flags & SB_NODIRATIME) && S_ISDIR(inode->i_mode)) |
8fa9dd24 | 1903 | return false; |
47ae32d6 | 1904 | |
cdb70f3f | 1905 | if (mnt->mnt_flags & MNT_NOATIME) |
8fa9dd24 | 1906 | return false; |
cdb70f3f | 1907 | if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) |
8fa9dd24 | 1908 | return false; |
1da177e4 | 1909 | |
c2050a45 | 1910 | now = current_time(inode); |
11ff6f05 | 1911 | |
6f22b664 | 1912 | if (!relatime_need_update(mnt, inode, now)) |
8fa9dd24 | 1913 | return false; |
11ff6f05 | 1914 | |
95582b00 | 1915 | if (timespec64_equal(&inode->i_atime, &now)) |
8fa9dd24 N |
1916 | return false; |
1917 | ||
1918 | return true; | |
1919 | } | |
1920 | ||
1921 | void touch_atime(const struct path *path) | |
1922 | { | |
1923 | struct vfsmount *mnt = path->mnt; | |
1924 | struct inode *inode = d_inode(path->dentry); | |
95582b00 | 1925 | struct timespec64 now; |
8fa9dd24 | 1926 | |
c6718543 | 1927 | if (!atime_needs_update(path, inode)) |
b12536c2 AK |
1928 | return; |
1929 | ||
5d37e9e6 | 1930 | if (!sb_start_write_trylock(inode->i_sb)) |
b12536c2 | 1931 | return; |
47ae32d6 | 1932 | |
8fa9dd24 | 1933 | if (__mnt_want_write(mnt) != 0) |
5d37e9e6 | 1934 | goto skip_update; |
c3b2da31 JB |
1935 | /* |
1936 | * File systems can error out when updating inodes if they need to | |
1937 | * allocate new space to modify an inode (such is the case for | |
1938 | * Btrfs), but since we touch atime while walking down the path we | |
1939 | * really don't care if we failed to update the atime of the file, | |
1940 | * so just ignore the return value. | |
2bc55652 AB |
1941 | * We may also fail on filesystems that have the ability to make parts |
1942 | * of the fs read only, e.g. subvolumes in Btrfs. | |
c3b2da31 | 1943 | */ |
c2050a45 | 1944 | now = current_time(inode); |
e60feb44 | 1945 | inode_update_time(inode, &now, S_ATIME); |
5d37e9e6 JK |
1946 | __mnt_drop_write(mnt); |
1947 | skip_update: | |
1948 | sb_end_write(inode->i_sb); | |
1da177e4 | 1949 | } |
869243a0 | 1950 | EXPORT_SYMBOL(touch_atime); |
1da177e4 | 1951 | |
dbfae0cd JK |
1952 | /* |
1953 | * Return mask of changes for notify_change() that need to be done as a | |
1954 | * response to write or truncate. Return 0 if nothing has to be changed. | |
1955 | * Negative value on error (change should be denied). | |
1956 | */ | |
9452e93e | 1957 | int dentry_needs_remove_privs(struct mnt_idmap *idmap, |
ed5a7047 | 1958 | struct dentry *dentry) |
dbfae0cd | 1959 | { |
dbfae0cd JK |
1960 | struct inode *inode = d_inode(dentry); |
1961 | int mask = 0; | |
1962 | int ret; | |
1963 | ||
1964 | if (IS_NOSEC(inode)) | |
1965 | return 0; | |
1966 | ||
9452e93e | 1967 | mask = setattr_should_drop_suidgid(idmap, inode); |
dbfae0cd JK |
1968 | ret = security_inode_need_killpriv(dentry); |
1969 | if (ret < 0) | |
1970 | return ret; | |
1971 | if (ret) | |
1972 | mask |= ATTR_KILL_PRIV; | |
1973 | return mask; | |
1974 | } | |
dbfae0cd | 1975 | |
abf08576 | 1976 | static int __remove_privs(struct mnt_idmap *idmap, |
643fe55a | 1977 | struct dentry *dentry, int kill) |
3ed37648 CW |
1978 | { |
1979 | struct iattr newattrs; | |
1980 | ||
1981 | newattrs.ia_valid = ATTR_FORCE | kill; | |
27ac0ffe BF |
1982 | /* |
1983 | * Note we call this on write, so notify_change will not | |
1984 | * encounter any conflicting delegations: | |
1985 | */ | |
abf08576 | 1986 | return notify_change(idmap, dentry, &newattrs, NULL); |
3ed37648 CW |
1987 | } |
1988 | ||
faf99b56 | 1989 | static int __file_remove_privs(struct file *file, unsigned int flags) |
3ed37648 | 1990 | { |
c1892c37 MS |
1991 | struct dentry *dentry = file_dentry(file); |
1992 | struct inode *inode = file_inode(file); | |
41191cf6 | 1993 | int error = 0; |
dbfae0cd | 1994 | int kill; |
3ed37648 | 1995 | |
f69e749a | 1996 | if (IS_NOSEC(inode) || !S_ISREG(inode->i_mode)) |
3ed37648 CW |
1997 | return 0; |
1998 | ||
9452e93e | 1999 | kill = dentry_needs_remove_privs(file_mnt_idmap(file), dentry); |
41191cf6 | 2000 | if (kill < 0) |
dbfae0cd | 2001 | return kill; |
faf99b56 | 2002 | |
41191cf6 SR |
2003 | if (kill) { |
2004 | if (flags & IOCB_NOWAIT) | |
2005 | return -EAGAIN; | |
2006 | ||
abf08576 | 2007 | error = __remove_privs(file_mnt_idmap(file), dentry, kill); |
41191cf6 | 2008 | } |
faf99b56 | 2009 | |
2426f391 JK |
2010 | if (!error) |
2011 | inode_has_no_xattr(inode); | |
3ed37648 CW |
2012 | return error; |
2013 | } | |
faf99b56 SR |
2014 | |
2015 | /** | |
2016 | * file_remove_privs - remove special file privileges (suid, capabilities) | |
2017 | * @file: file to remove privileges from | |
2018 | * | |
2019 | * When file is modified by a write or truncation ensure that special | |
2020 | * file privileges are removed. | |
2021 | * | |
2022 | * Return: 0 on success, negative errno on failure. | |
2023 | */ | |
2024 | int file_remove_privs(struct file *file) | |
2025 | { | |
2026 | return __file_remove_privs(file, 0); | |
2027 | } | |
5fa8e0a1 | 2028 | EXPORT_SYMBOL(file_remove_privs); |
3ed37648 | 2029 | |
6a2aa5d8 | 2030 | static int inode_needs_update_time(struct inode *inode, struct timespec64 *now) |
1da177e4 | 2031 | { |
c3b2da31 | 2032 | int sync_it = 0; |
1da177e4 | 2033 | |
ce06e0b2 | 2034 | /* First try to exhaust all avenues to not sync */ |
1da177e4 | 2035 | if (IS_NOCMTIME(inode)) |
c3b2da31 | 2036 | return 0; |
20ddee2c | 2037 | |
6a2aa5d8 | 2038 | if (!timespec64_equal(&inode->i_mtime, now)) |
ce06e0b2 | 2039 | sync_it = S_MTIME; |
1da177e4 | 2040 | |
6a2aa5d8 | 2041 | if (!timespec64_equal(&inode->i_ctime, now)) |
ce06e0b2 | 2042 | sync_it |= S_CTIME; |
870f4817 | 2043 | |
e38cf302 | 2044 | if (IS_I_VERSION(inode) && inode_iversion_need_inc(inode)) |
ce06e0b2 | 2045 | sync_it |= S_VERSION; |
7a224228 | 2046 | |
6a2aa5d8 SR |
2047 | return sync_it; |
2048 | } | |
2049 | ||
2050 | static int __file_update_time(struct file *file, struct timespec64 *now, | |
2051 | int sync_mode) | |
2052 | { | |
2053 | int ret = 0; | |
2054 | struct inode *inode = file_inode(file); | |
ce06e0b2 | 2055 | |
6a2aa5d8 SR |
2056 | /* try to update time settings */ |
2057 | if (!__mnt_want_write_file(file)) { | |
2058 | ret = inode_update_time(inode, now, sync_mode); | |
2059 | __mnt_drop_write_file(file); | |
2060 | } | |
c3b2da31 JB |
2061 | |
2062 | return ret; | |
1da177e4 | 2063 | } |
6a2aa5d8 SR |
2064 | |
2065 | /** | |
2066 | * file_update_time - update mtime and ctime time | |
2067 | * @file: file accessed | |
2068 | * | |
2069 | * Update the mtime and ctime members of an inode and mark the inode for | |
2070 | * writeback. Note that this function is meant exclusively for usage in | |
2071 | * the file write path of filesystems, and filesystems may choose to | |
2072 | * explicitly ignore updates via this function with the _NOCMTIME inode | |
2073 | * flag, e.g. for network filesystem where these imestamps are handled | |
2074 | * by the server. This can return an error for file systems who need to | |
2075 | * allocate space in order to update an inode. | |
2076 | * | |
2077 | * Return: 0 on success, negative errno on failure. | |
2078 | */ | |
2079 | int file_update_time(struct file *file) | |
2080 | { | |
2081 | int ret; | |
2082 | struct inode *inode = file_inode(file); | |
2083 | struct timespec64 now = current_time(inode); | |
2084 | ||
2085 | ret = inode_needs_update_time(inode, &now); | |
2086 | if (ret <= 0) | |
2087 | return ret; | |
2088 | ||
2089 | return __file_update_time(file, &now, ret); | |
2090 | } | |
870f4817 | 2091 | EXPORT_SYMBOL(file_update_time); |
1da177e4 | 2092 | |
faf99b56 | 2093 | /** |
66fa3ced | 2094 | * file_modified_flags - handle mandated vfs changes when modifying a file |
faf99b56 | 2095 | * @file: file that was modified |
66fa3ced | 2096 | * @flags: kiocb flags |
faf99b56 SR |
2097 | * |
2098 | * When file has been modified ensure that special | |
2099 | * file privileges are removed and time settings are updated. | |
2100 | * | |
66fa3ced SR |
2101 | * If IOCB_NOWAIT is set, special file privileges will not be removed and |
2102 | * time settings will not be updated. It will return -EAGAIN. | |
2103 | * | |
faf99b56 SR |
2104 | * Context: Caller must hold the file's inode lock. |
2105 | * | |
2106 | * Return: 0 on success, negative errno on failure. | |
2107 | */ | |
66fa3ced | 2108 | static int file_modified_flags(struct file *file, int flags) |
e38f7f53 | 2109 | { |
faf99b56 | 2110 | int ret; |
6a2aa5d8 SR |
2111 | struct inode *inode = file_inode(file); |
2112 | struct timespec64 now = current_time(inode); | |
e38f7f53 AG |
2113 | |
2114 | /* | |
2115 | * Clear the security bits if the process is not being run by root. | |
2116 | * This keeps people from modifying setuid and setgid binaries. | |
2117 | */ | |
66fa3ced | 2118 | ret = __file_remove_privs(file, flags); |
faf99b56 SR |
2119 | if (ret) |
2120 | return ret; | |
e38f7f53 AG |
2121 | |
2122 | if (unlikely(file->f_mode & FMODE_NOCMTIME)) | |
2123 | return 0; | |
2124 | ||
6a2aa5d8 SR |
2125 | ret = inode_needs_update_time(inode, &now); |
2126 | if (ret <= 0) | |
2127 | return ret; | |
66fa3ced SR |
2128 | if (flags & IOCB_NOWAIT) |
2129 | return -EAGAIN; | |
6a2aa5d8 SR |
2130 | |
2131 | return __file_update_time(file, &now, ret); | |
e38f7f53 | 2132 | } |
66fa3ced SR |
2133 | |
2134 | /** | |
2135 | * file_modified - handle mandated vfs changes when modifying a file | |
2136 | * @file: file that was modified | |
2137 | * | |
2138 | * When file has been modified ensure that special | |
2139 | * file privileges are removed and time settings are updated. | |
2140 | * | |
2141 | * Context: Caller must hold the file's inode lock. | |
2142 | * | |
2143 | * Return: 0 on success, negative errno on failure. | |
2144 | */ | |
2145 | int file_modified(struct file *file) | |
2146 | { | |
2147 | return file_modified_flags(file, 0); | |
2148 | } | |
e38f7f53 AG |
2149 | EXPORT_SYMBOL(file_modified); |
2150 | ||
66fa3ced SR |
2151 | /** |
2152 | * kiocb_modified - handle mandated vfs changes when modifying a file | |
2153 | * @iocb: iocb that was modified | |
2154 | * | |
2155 | * When file has been modified ensure that special | |
2156 | * file privileges are removed and time settings are updated. | |
2157 | * | |
2158 | * Context: Caller must hold the file's inode lock. | |
2159 | * | |
2160 | * Return: 0 on success, negative errno on failure. | |
2161 | */ | |
2162 | int kiocb_modified(struct kiocb *iocb) | |
2163 | { | |
2164 | return file_modified_flags(iocb->ki_filp, iocb->ki_flags); | |
2165 | } | |
2166 | EXPORT_SYMBOL_GPL(kiocb_modified); | |
2167 | ||
1da177e4 LT |
2168 | int inode_needs_sync(struct inode *inode) |
2169 | { | |
2170 | if (IS_SYNC(inode)) | |
2171 | return 1; | |
2172 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) | |
2173 | return 1; | |
2174 | return 0; | |
2175 | } | |
1da177e4 LT |
2176 | EXPORT_SYMBOL(inode_needs_sync); |
2177 | ||
1da177e4 | 2178 | /* |
168a9fd6 MS |
2179 | * If we try to find an inode in the inode hash while it is being |
2180 | * deleted, we have to wait until the filesystem completes its | |
2181 | * deletion before reporting that it isn't found. This function waits | |
2182 | * until the deletion _might_ have completed. Callers are responsible | |
2183 | * to recheck inode state. | |
2184 | * | |
eaff8079 | 2185 | * It doesn't matter if I_NEW is not set initially, a call to |
250df6ed DC |
2186 | * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list |
2187 | * will DTRT. | |
1da177e4 LT |
2188 | */ |
2189 | static void __wait_on_freeing_inode(struct inode *inode) | |
2190 | { | |
2191 | wait_queue_head_t *wq; | |
eaff8079 CH |
2192 | DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW); |
2193 | wq = bit_waitqueue(&inode->i_state, __I_NEW); | |
21417136 | 2194 | prepare_to_wait(wq, &wait.wq_entry, TASK_UNINTERRUPTIBLE); |
250df6ed | 2195 | spin_unlock(&inode->i_lock); |
67a23c49 | 2196 | spin_unlock(&inode_hash_lock); |
1da177e4 | 2197 | schedule(); |
21417136 | 2198 | finish_wait(wq, &wait.wq_entry); |
67a23c49 | 2199 | spin_lock(&inode_hash_lock); |
1da177e4 LT |
2200 | } |
2201 | ||
1da177e4 LT |
2202 | static __initdata unsigned long ihash_entries; |
2203 | static int __init set_ihash_entries(char *str) | |
2204 | { | |
2205 | if (!str) | |
2206 | return 0; | |
2207 | ihash_entries = simple_strtoul(str, &str, 0); | |
2208 | return 1; | |
2209 | } | |
2210 | __setup("ihash_entries=", set_ihash_entries); | |
2211 | ||
2212 | /* | |
2213 | * Initialize the waitqueues and inode hash table. | |
2214 | */ | |
2215 | void __init inode_init_early(void) | |
2216 | { | |
1da177e4 LT |
2217 | /* If hashes are distributed across NUMA nodes, defer |
2218 | * hash allocation until vmalloc space is available. | |
2219 | */ | |
2220 | if (hashdist) | |
2221 | return; | |
2222 | ||
2223 | inode_hashtable = | |
2224 | alloc_large_system_hash("Inode-cache", | |
2225 | sizeof(struct hlist_head), | |
2226 | ihash_entries, | |
2227 | 14, | |
3d375d78 | 2228 | HASH_EARLY | HASH_ZERO, |
1da177e4 LT |
2229 | &i_hash_shift, |
2230 | &i_hash_mask, | |
31fe62b9 | 2231 | 0, |
1da177e4 | 2232 | 0); |
1da177e4 LT |
2233 | } |
2234 | ||
74bf17cf | 2235 | void __init inode_init(void) |
1da177e4 | 2236 | { |
1da177e4 | 2237 | /* inode slab cache */ |
b0196009 PJ |
2238 | inode_cachep = kmem_cache_create("inode_cache", |
2239 | sizeof(struct inode), | |
2240 | 0, | |
2241 | (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC| | |
5d097056 | 2242 | SLAB_MEM_SPREAD|SLAB_ACCOUNT), |
20c2df83 | 2243 | init_once); |
1da177e4 LT |
2244 | |
2245 | /* Hash may have been set up in inode_init_early */ | |
2246 | if (!hashdist) | |
2247 | return; | |
2248 | ||
2249 | inode_hashtable = | |
2250 | alloc_large_system_hash("Inode-cache", | |
2251 | sizeof(struct hlist_head), | |
2252 | ihash_entries, | |
2253 | 14, | |
3d375d78 | 2254 | HASH_ZERO, |
1da177e4 LT |
2255 | &i_hash_shift, |
2256 | &i_hash_mask, | |
31fe62b9 | 2257 | 0, |
1da177e4 | 2258 | 0); |
1da177e4 LT |
2259 | } |
2260 | ||
2261 | void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev) | |
2262 | { | |
2263 | inode->i_mode = mode; | |
2264 | if (S_ISCHR(mode)) { | |
2265 | inode->i_fop = &def_chr_fops; | |
2266 | inode->i_rdev = rdev; | |
2267 | } else if (S_ISBLK(mode)) { | |
2268 | inode->i_fop = &def_blk_fops; | |
2269 | inode->i_rdev = rdev; | |
2270 | } else if (S_ISFIFO(mode)) | |
599a0ac1 | 2271 | inode->i_fop = &pipefifo_fops; |
1da177e4 | 2272 | else if (S_ISSOCK(mode)) |
bd9b51e7 | 2273 | ; /* leave it no_open_fops */ |
1da177e4 | 2274 | else |
af0d9ae8 MK |
2275 | printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for" |
2276 | " inode %s:%lu\n", mode, inode->i_sb->s_id, | |
2277 | inode->i_ino); | |
1da177e4 LT |
2278 | } |
2279 | EXPORT_SYMBOL(init_special_inode); | |
a1bd120d DM |
2280 | |
2281 | /** | |
eaae668d | 2282 | * inode_init_owner - Init uid,gid,mode for new inode according to posix standards |
f2d40141 | 2283 | * @idmap: idmap of the mount the inode was created from |
a1bd120d DM |
2284 | * @inode: New inode |
2285 | * @dir: Directory inode | |
2286 | * @mode: mode of the new inode | |
21cb47be | 2287 | * |
f2d40141 CB |
2288 | * If the inode has been created through an idmapped mount the idmap of |
2289 | * the vfsmount must be passed through @idmap. This function will then take | |
2290 | * care to map the inode according to @idmap before checking permissions | |
21cb47be | 2291 | * and initializing i_uid and i_gid. On non-idmapped mounts or if permission |
f2d40141 | 2292 | * checking is to be performed on the raw inode simply pass @nop_mnt_idmap. |
a1bd120d | 2293 | */ |
f2d40141 | 2294 | void inode_init_owner(struct mnt_idmap *idmap, struct inode *inode, |
21cb47be | 2295 | const struct inode *dir, umode_t mode) |
a1bd120d | 2296 | { |
c14329d3 | 2297 | inode_fsuid_set(inode, idmap); |
a1bd120d DM |
2298 | if (dir && dir->i_mode & S_ISGID) { |
2299 | inode->i_gid = dir->i_gid; | |
0fa3ecd8 LT |
2300 | |
2301 | /* Directories are special, and always inherit S_ISGID */ | |
a1bd120d DM |
2302 | if (S_ISDIR(mode)) |
2303 | mode |= S_ISGID; | |
2304 | } else | |
c14329d3 | 2305 | inode_fsgid_set(inode, idmap); |
a1bd120d DM |
2306 | inode->i_mode = mode; |
2307 | } | |
2308 | EXPORT_SYMBOL(inode_init_owner); | |
e795b717 | 2309 | |
2e149670 SH |
2310 | /** |
2311 | * inode_owner_or_capable - check current task permissions to inode | |
01beba79 | 2312 | * @idmap: idmap of the mount the inode was found from |
2e149670 SH |
2313 | * @inode: inode being checked |
2314 | * | |
23adbe12 AL |
2315 | * Return true if current either has CAP_FOWNER in a namespace with the |
2316 | * inode owner uid mapped, or owns the file. | |
21cb47be | 2317 | * |
01beba79 CB |
2318 | * If the inode has been found through an idmapped mount the idmap of |
2319 | * the vfsmount must be passed through @idmap. This function will then take | |
2320 | * care to map the inode according to @idmap before checking permissions. | |
21cb47be | 2321 | * On non-idmapped mounts or if permission checking is to be performed on the |
01beba79 | 2322 | * raw inode simply passs @nop_mnt_idmap. |
e795b717 | 2323 | */ |
01beba79 | 2324 | bool inode_owner_or_capable(struct mnt_idmap *idmap, |
21cb47be | 2325 | const struct inode *inode) |
e795b717 | 2326 | { |
a2bd096f | 2327 | vfsuid_t vfsuid; |
23adbe12 AL |
2328 | struct user_namespace *ns; |
2329 | ||
e67fe633 | 2330 | vfsuid = i_uid_into_vfsuid(idmap, inode); |
a2bd096f | 2331 | if (vfsuid_eq_kuid(vfsuid, current_fsuid())) |
e795b717 | 2332 | return true; |
23adbe12 AL |
2333 | |
2334 | ns = current_user_ns(); | |
a2bd096f | 2335 | if (vfsuid_has_mapping(ns, vfsuid) && ns_capable(ns, CAP_FOWNER)) |
e795b717 SH |
2336 | return true; |
2337 | return false; | |
2338 | } | |
2e149670 | 2339 | EXPORT_SYMBOL(inode_owner_or_capable); |
1d59d61f TM |
2340 | |
2341 | /* | |
2342 | * Direct i/o helper functions | |
2343 | */ | |
2344 | static void __inode_dio_wait(struct inode *inode) | |
2345 | { | |
2346 | wait_queue_head_t *wq = bit_waitqueue(&inode->i_state, __I_DIO_WAKEUP); | |
2347 | DEFINE_WAIT_BIT(q, &inode->i_state, __I_DIO_WAKEUP); | |
2348 | ||
2349 | do { | |
21417136 | 2350 | prepare_to_wait(wq, &q.wq_entry, TASK_UNINTERRUPTIBLE); |
1d59d61f TM |
2351 | if (atomic_read(&inode->i_dio_count)) |
2352 | schedule(); | |
2353 | } while (atomic_read(&inode->i_dio_count)); | |
21417136 | 2354 | finish_wait(wq, &q.wq_entry); |
1d59d61f TM |
2355 | } |
2356 | ||
2357 | /** | |
2358 | * inode_dio_wait - wait for outstanding DIO requests to finish | |
2359 | * @inode: inode to wait for | |
2360 | * | |
2361 | * Waits for all pending direct I/O requests to finish so that we can | |
2362 | * proceed with a truncate or equivalent operation. | |
2363 | * | |
2364 | * Must be called under a lock that serializes taking new references | |
2365 | * to i_dio_count, usually by inode->i_mutex. | |
2366 | */ | |
2367 | void inode_dio_wait(struct inode *inode) | |
2368 | { | |
2369 | if (atomic_read(&inode->i_dio_count)) | |
2370 | __inode_dio_wait(inode); | |
2371 | } | |
2372 | EXPORT_SYMBOL(inode_dio_wait); | |
2373 | ||
5f16f322 TT |
2374 | /* |
2375 | * inode_set_flags - atomically set some inode flags | |
2376 | * | |
2377 | * Note: the caller should be holding i_mutex, or else be sure that | |
2378 | * they have exclusive access to the inode structure (i.e., while the | |
2379 | * inode is being instantiated). The reason for the cmpxchg() loop | |
2380 | * --- which wouldn't be necessary if all code paths which modify | |
2381 | * i_flags actually followed this rule, is that there is at least one | |
5fa8e0a1 JK |
2382 | * code path which doesn't today so we use cmpxchg() out of an abundance |
2383 | * of caution. | |
5f16f322 TT |
2384 | * |
2385 | * In the long run, i_mutex is overkill, and we should probably look | |
2386 | * at using the i_lock spinlock to protect i_flags, and then make sure | |
2387 | * it is so documented in include/linux/fs.h and that all code follows | |
2388 | * the locking convention!! | |
2389 | */ | |
2390 | void inode_set_flags(struct inode *inode, unsigned int flags, | |
2391 | unsigned int mask) | |
2392 | { | |
5f16f322 | 2393 | WARN_ON_ONCE(flags & ~mask); |
a905737f | 2394 | set_mask_bits(&inode->i_flags, mask, flags); |
5f16f322 TT |
2395 | } |
2396 | EXPORT_SYMBOL(inode_set_flags); | |
21fc61c7 AV |
2397 | |
2398 | void inode_nohighmem(struct inode *inode) | |
2399 | { | |
2400 | mapping_set_gfp_mask(inode->i_mapping, GFP_USER); | |
2401 | } | |
2402 | EXPORT_SYMBOL(inode_nohighmem); | |
3cd88666 | 2403 | |
50e17c00 DD |
2404 | /** |
2405 | * timestamp_truncate - Truncate timespec to a granularity | |
2406 | * @t: Timespec | |
2407 | * @inode: inode being updated | |
2408 | * | |
2409 | * Truncate a timespec to the granularity supported by the fs | |
2410 | * containing the inode. Always rounds down. gran must | |
2411 | * not be 0 nor greater than a second (NSEC_PER_SEC, or 10^9 ns). | |
2412 | */ | |
2413 | struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode) | |
2414 | { | |
2415 | struct super_block *sb = inode->i_sb; | |
2416 | unsigned int gran = sb->s_time_gran; | |
2417 | ||
2418 | t.tv_sec = clamp(t.tv_sec, sb->s_time_min, sb->s_time_max); | |
2419 | if (unlikely(t.tv_sec == sb->s_time_max || t.tv_sec == sb->s_time_min)) | |
2420 | t.tv_nsec = 0; | |
2421 | ||
2422 | /* Avoid division in the common cases 1 ns and 1 s. */ | |
2423 | if (gran == 1) | |
2424 | ; /* nothing */ | |
2425 | else if (gran == NSEC_PER_SEC) | |
2426 | t.tv_nsec = 0; | |
2427 | else if (gran > 1 && gran < NSEC_PER_SEC) | |
2428 | t.tv_nsec -= t.tv_nsec % gran; | |
2429 | else | |
2430 | WARN(1, "invalid file time granularity: %u", gran); | |
2431 | return t; | |
2432 | } | |
2433 | EXPORT_SYMBOL(timestamp_truncate); | |
2434 | ||
3cd88666 DD |
2435 | /** |
2436 | * current_time - Return FS time | |
2437 | * @inode: inode. | |
2438 | * | |
2439 | * Return the current time truncated to the time granularity supported by | |
2440 | * the fs. | |
2441 | * | |
2442 | * Note that inode and inode->sb cannot be NULL. | |
2443 | * Otherwise, the function warns and returns time without truncation. | |
2444 | */ | |
95582b00 | 2445 | struct timespec64 current_time(struct inode *inode) |
3cd88666 | 2446 | { |
d651d160 AB |
2447 | struct timespec64 now; |
2448 | ||
2449 | ktime_get_coarse_real_ts64(&now); | |
3cd88666 DD |
2450 | |
2451 | if (unlikely(!inode->i_sb)) { | |
2452 | WARN(1, "current_time() called with uninitialized super_block in the inode"); | |
2453 | return now; | |
2454 | } | |
2455 | ||
50e17c00 | 2456 | return timestamp_truncate(now, inode); |
3cd88666 DD |
2457 | } |
2458 | EXPORT_SYMBOL(current_time); | |
2b3416ce | 2459 | |
11c2a870 CB |
2460 | /** |
2461 | * in_group_or_capable - check whether caller is CAP_FSETID privileged | |
9452e93e | 2462 | * @idmap: idmap of the mount @inode was found from |
11c2a870 CB |
2463 | * @inode: inode to check |
2464 | * @vfsgid: the new/current vfsgid of @inode | |
2465 | * | |
2466 | * Check wether @vfsgid is in the caller's group list or if the caller is | |
2467 | * privileged with CAP_FSETID over @inode. This can be used to determine | |
2468 | * whether the setgid bit can be kept or must be dropped. | |
2469 | * | |
2470 | * Return: true if the caller is sufficiently privileged, false if not. | |
2471 | */ | |
9452e93e | 2472 | bool in_group_or_capable(struct mnt_idmap *idmap, |
11c2a870 CB |
2473 | const struct inode *inode, vfsgid_t vfsgid) |
2474 | { | |
2475 | if (vfsgid_in_group_p(vfsgid)) | |
2476 | return true; | |
9452e93e | 2477 | if (capable_wrt_inode_uidgid(idmap, inode, CAP_FSETID)) |
11c2a870 CB |
2478 | return true; |
2479 | return false; | |
2480 | } | |
2481 | ||
2b3416ce YX |
2482 | /** |
2483 | * mode_strip_sgid - handle the sgid bit for non-directories | |
9452e93e | 2484 | * @idmap: idmap of the mount the inode was created from |
2b3416ce YX |
2485 | * @dir: parent directory inode |
2486 | * @mode: mode of the file to be created in @dir | |
2487 | * | |
2488 | * If the @mode of the new file has both the S_ISGID and S_IXGRP bit | |
2489 | * raised and @dir has the S_ISGID bit raised ensure that the caller is | |
2490 | * either in the group of the parent directory or they have CAP_FSETID | |
2491 | * in their user namespace and are privileged over the parent directory. | |
2492 | * In all other cases, strip the S_ISGID bit from @mode. | |
2493 | * | |
2494 | * Return: the new mode to use for the file | |
2495 | */ | |
9452e93e | 2496 | umode_t mode_strip_sgid(struct mnt_idmap *idmap, |
2b3416ce YX |
2497 | const struct inode *dir, umode_t mode) |
2498 | { | |
2499 | if ((mode & (S_ISGID | S_IXGRP)) != (S_ISGID | S_IXGRP)) | |
2500 | return mode; | |
2501 | if (S_ISDIR(mode) || !dir || !(dir->i_mode & S_ISGID)) | |
2502 | return mode; | |
e67fe633 | 2503 | if (in_group_or_capable(idmap, dir, i_gid_into_vfsgid(idmap, dir))) |
2b3416ce | 2504 | return mode; |
2b3416ce YX |
2505 | return mode & ~S_ISGID; |
2506 | } | |
2507 | EXPORT_SYMBOL(mode_strip_sgid); |