Line data Source code
1 : // SPDX-License-Identifier: GPL-2.0-only
2 : /*
3 : * (C) 1997 Linus Torvalds
4 : * (C) 1999 Andrea Arcangeli <andrea@suse.de> (dynamic inode allocation)
5 : */
6 : #include <linux/export.h>
7 : #include <linux/fs.h>
8 : #include <linux/filelock.h>
9 : #include <linux/mm.h>
10 : #include <linux/backing-dev.h>
11 : #include <linux/hash.h>
12 : #include <linux/swap.h>
13 : #include <linux/security.h>
14 : #include <linux/cdev.h>
15 : #include <linux/memblock.h>
16 : #include <linux/fsnotify.h>
17 : #include <linux/mount.h>
18 : #include <linux/posix_acl.h>
19 : #include <linux/prefetch.h>
20 : #include <linux/buffer_head.h> /* for inode_has_buffers */
21 : #include <linux/ratelimit.h>
22 : #include <linux/list_lru.h>
23 : #include <linux/iversion.h>
24 : #include <trace/events/writeback.h>
25 : #include "internal.h"
26 :
27 : /*
28 : * Inode locking rules:
29 : *
30 : * inode->i_lock protects:
31 : * inode->i_state, inode->i_hash, __iget(), inode->i_io_list
32 : * Inode LRU list locks protect:
33 : * inode->i_sb->s_inode_lru, inode->i_lru
34 : * inode->i_sb->s_inode_list_lock protects:
35 : * inode->i_sb->s_inodes, inode->i_sb_list
36 : * bdi->wb.list_lock protects:
37 : * bdi->wb.b_{dirty,io,more_io,dirty_time}, inode->i_io_list
38 : * inode_hash_lock protects:
39 : * inode_hashtable, inode->i_hash
40 : *
41 : * Lock ordering:
42 : *
43 : * inode->i_sb->s_inode_list_lock
44 : * inode->i_lock
45 : * Inode LRU list locks
46 : *
47 : * bdi->wb.list_lock
48 : * inode->i_lock
49 : *
50 : * inode_hash_lock
51 : * inode->i_sb->s_inode_list_lock
52 : * inode->i_lock
53 : *
54 : * iunique_lock
55 : * inode_hash_lock
56 : */
57 :
58 : static unsigned int i_hash_mask __read_mostly;
59 : static unsigned int i_hash_shift __read_mostly;
60 : static struct hlist_head *inode_hashtable __read_mostly;
61 : static __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_hash_lock);
62 :
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 :
71 : static DEFINE_PER_CPU(unsigned long, nr_inodes);
72 : static DEFINE_PER_CPU(unsigned long, nr_unused);
73 :
74 : static struct kmem_cache *inode_cachep __read_mostly;
75 :
76 2237340 : static long get_nr_inodes(void)
77 : {
78 2237340 : int i;
79 2237340 : long sum = 0;
80 11175288 : for_each_possible_cpu(i)
81 8938811 : sum += per_cpu(nr_inodes, i);
82 2233941 : return sum < 0 ? 0 : sum;
83 : }
84 :
85 2234607 : static inline long get_nr_inodes_unused(void)
86 : {
87 2234607 : int i;
88 2234607 : long sum = 0;
89 11166128 : for_each_possible_cpu(i)
90 8932099 : sum += per_cpu(nr_unused, i);
91 2234345 : return sum < 0 ? 0 : sum;
92 : }
93 :
94 2236575 : long get_nr_dirty_inodes(void)
95 : {
96 : /* not actually dirty inodes, but a wild approximation */
97 2236575 : long nr_dirty = get_nr_inodes() - get_nr_inodes_unused();
98 2233499 : return nr_dirty > 0 ? nr_dirty : 0;
99 : }
100 :
101 : /*
102 : * Handle nr_inode sysctl
103 : */
104 : #ifdef CONFIG_SYSCTL
105 : /*
106 : * Statistics gathering..
107 : */
108 : static struct inodes_stat_t inodes_stat;
109 :
110 903 : static int proc_nr_inodes(struct ctl_table *table, int write, void *buffer,
111 : size_t *lenp, loff_t *ppos)
112 : {
113 903 : inodes_stat.nr_inodes = get_nr_inodes();
114 903 : inodes_stat.nr_unused = get_nr_inodes_unused();
115 903 : return proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
116 : }
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 0 : static int __init init_fs_inode_sysctls(void)
137 : {
138 0 : register_sysctl_init("fs", inodes_sysctls);
139 0 : return 0;
140 : }
141 : early_initcall(init_fs_inode_sysctls);
142 : #endif
143 :
144 0 : static int no_open(struct inode *inode, struct file *file)
145 : {
146 0 : return -ENXIO;
147 : }
148 :
149 : /**
150 : * inode_init_always - perform inode structure initialisation
151 : * @sb: superblock inode belongs to
152 : * @inode: inode to initialise
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 : */
157 1626311208 : int inode_init_always(struct super_block *sb, struct inode *inode)
158 : {
159 1626311208 : static const struct inode_operations empty_iops;
160 1626311208 : static const struct file_operations no_open_fops = {.open = no_open};
161 1626311208 : struct address_space *const mapping = &inode->i_data;
162 :
163 1626311208 : inode->i_sb = sb;
164 1626311208 : inode->i_blkbits = sb->s_blocksize_bits;
165 1626311208 : inode->i_flags = 0;
166 1626311208 : atomic64_set(&inode->i_sequence, 0);
167 1626311208 : atomic_set(&inode->i_count, 1);
168 1626311208 : inode->i_op = &empty_iops;
169 1626311208 : inode->i_fop = &no_open_fops;
170 1626311208 : inode->i_ino = 0;
171 1626311208 : inode->__i_nlink = 1;
172 1626311208 : inode->i_opflags = 0;
173 1626311208 : if (sb->s_xattr)
174 1581479879 : inode->i_opflags |= IOP_XATTR;
175 1626311208 : i_uid_write(inode, 0);
176 1625747761 : i_gid_write(inode, 0);
177 1625786581 : atomic_set(&inode->i_writecount, 0);
178 1625786581 : inode->i_size = 0;
179 1625786581 : inode->i_write_hint = WRITE_LIFE_NOT_SET;
180 1625786581 : inode->i_blocks = 0;
181 1625786581 : inode->i_bytes = 0;
182 1625786581 : inode->i_generation = 0;
183 1625786581 : inode->i_pipe = NULL;
184 1625786581 : inode->i_cdev = NULL;
185 1625786581 : inode->i_link = NULL;
186 1625786581 : inode->i_dir_seq = 0;
187 1625786581 : inode->i_rdev = 0;
188 1625786581 : inode->dirtied_when = 0;
189 :
190 : #ifdef CONFIG_CGROUP_WRITEBACK
191 1625786581 : inode->i_wb_frn_winner = 0;
192 1625786581 : inode->i_wb_frn_avg_time = 0;
193 1625786581 : inode->i_wb_frn_history = 0;
194 : #endif
195 :
196 1625786581 : spin_lock_init(&inode->i_lock);
197 1625858532 : lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key);
198 :
199 1625858532 : init_rwsem(&inode->i_rwsem);
200 1625832369 : lockdep_set_class(&inode->i_rwsem, &sb->s_type->i_mutex_key);
201 :
202 1625832369 : atomic_set(&inode->i_dio_count, 0);
203 :
204 1625832369 : mapping->a_ops = &empty_aops;
205 1625832369 : mapping->host = inode;
206 1625832369 : mapping->flags = 0;
207 1625832369 : mapping->wb_err = 0;
208 1625832369 : atomic_set(&mapping->i_mmap_writable, 0);
209 : #ifdef CONFIG_READ_ONLY_THP_FOR_FS
210 : atomic_set(&mapping->nr_thps, 0);
211 : #endif
212 1625832369 : mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE);
213 1625832369 : mapping->private_data = NULL;
214 1625832369 : mapping->writeback_index = 0;
215 1625832369 : init_rwsem(&mapping->invalidate_lock);
216 1626859451 : lockdep_set_class_and_name(&mapping->invalidate_lock,
217 : &sb->s_type->invalidate_lock_key,
218 : "mapping.invalidate_lock");
219 1626859451 : inode->i_private = NULL;
220 1626859451 : inode->i_mapping = mapping;
221 1626859451 : INIT_HLIST_HEAD(&inode->i_dentry); /* buggered by rcu freeing */
222 : #ifdef CONFIG_FS_POSIX_ACL
223 1626859451 : inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED;
224 : #endif
225 :
226 : #ifdef CONFIG_FSNOTIFY
227 1626859451 : inode->i_fsnotify_mask = 0;
228 : #endif
229 1626859451 : inode->i_flctx = NULL;
230 :
231 1626859451 : if (unlikely(security_inode_alloc(inode)))
232 : return -ENOMEM;
233 1626859451 : this_cpu_inc(nr_inodes);
234 :
235 1628022828 : return 0;
236 : }
237 : EXPORT_SYMBOL(inode_init_always);
238 :
239 21887250 : void free_inode_nonrcu(struct inode *inode)
240 : {
241 21887250 : kmem_cache_free(inode_cachep, inode);
242 5113153 : }
243 : EXPORT_SYMBOL(free_inode_nonrcu);
244 :
245 489639231 : static void i_callback(struct rcu_head *head)
246 : {
247 489639231 : struct inode *inode = container_of(head, struct inode, i_rcu);
248 489639231 : if (inode->free_inode)
249 484526065 : inode->free_inode(inode);
250 : else
251 5113166 : free_inode_nonrcu(inode);
252 490515293 : }
253 :
254 510058477 : static struct inode *alloc_inode(struct super_block *sb)
255 : {
256 510058477 : const struct super_operations *ops = sb->s_op;
257 510058477 : struct inode *inode;
258 :
259 510058477 : if (ops->alloc_inode)
260 483101437 : inode = ops->alloc_inode(sb);
261 : else
262 26957040 : inode = alloc_inode_sb(sb, inode_cachep, GFP_KERNEL);
263 :
264 512595539 : if (!inode)
265 : return NULL;
266 :
267 512595539 : if (unlikely(inode_init_always(sb, inode))) {
268 0 : if (ops->destroy_inode) {
269 0 : ops->destroy_inode(inode);
270 0 : if (!ops->free_inode)
271 : return NULL;
272 : }
273 0 : inode->free_inode = ops->free_inode;
274 0 : i_callback(&inode->i_rcu);
275 0 : return NULL;
276 : }
277 :
278 : return inode;
279 : }
280 :
281 1627438373 : void __destroy_inode(struct inode *inode)
282 : {
283 1627438373 : BUG_ON(inode_has_buffers(inode));
284 1627124574 : inode_detach_wb(inode);
285 1627276673 : security_inode_free(inode);
286 1627276673 : fsnotify_inode_delete(inode);
287 1628644961 : locks_free_lock_context(inode);
288 1627514712 : if (!inode->i_nlink) {
289 535174244 : WARN_ON(atomic_long_read(&inode->i_sb->s_remove_count) == 0);
290 535174244 : atomic_long_dec(&inode->i_sb->s_remove_count);
291 : }
292 :
293 : #ifdef CONFIG_FS_POSIX_ACL
294 1628723908 : if (inode->i_acl && !is_uncached_acl(inode->i_acl))
295 205488 : posix_acl_release(inode->i_acl);
296 1628723908 : if (inode->i_default_acl && !is_uncached_acl(inode->i_default_acl))
297 18985 : posix_acl_release(inode->i_default_acl);
298 : #endif
299 1628723908 : this_cpu_dec(nr_inodes);
300 1628594722 : }
301 : EXPORT_SYMBOL(__destroy_inode);
302 :
303 1627730556 : static void destroy_inode(struct inode *inode)
304 : {
305 1627730556 : const struct super_operations *ops = inode->i_sb->s_op;
306 :
307 1627730556 : BUG_ON(!list_empty(&inode->i_lru));
308 1627730556 : __destroy_inode(inode);
309 1627899932 : if (ops->destroy_inode) {
310 1587612561 : ops->destroy_inode(inode);
311 1587547913 : if (!ops->free_inode)
312 : return;
313 : }
314 491467977 : inode->free_inode = ops->free_inode;
315 491467977 : call_rcu(&inode->i_rcu, i_callback);
316 : }
317 :
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 93744073 : void drop_nlink(struct inode *inode)
330 : {
331 93744073 : WARN_ON(inode->i_nlink == 0);
332 93744073 : inode->__i_nlink--;
333 93744073 : if (!inode->i_nlink)
334 72353171 : atomic_long_inc(&inode->i_sb->s_remove_count);
335 93944232 : }
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 540031608 : void clear_nlink(struct inode *inode)
347 : {
348 540031608 : if (inode->i_nlink) {
349 519208686 : inode->__i_nlink = 0;
350 519208686 : atomic_long_inc(&inode->i_sb->s_remove_count);
351 : }
352 540373633 : }
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 1735321485 : void set_nlink(struct inode *inode, unsigned int nlink)
364 : {
365 1735321485 : if (!nlink) {
366 97567693 : clear_nlink(inode);
367 : } else {
368 : /* Yes, some filesystems do change nlink from zero to one */
369 1637753792 : if (inode->i_nlink == 0)
370 53057799 : atomic_long_dec(&inode->i_sb->s_remove_count);
371 :
372 1637948174 : inode->__i_nlink = nlink;
373 : }
374 1735589030 : }
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 40006873 : void inc_nlink(struct inode *inode)
386 : {
387 40006873 : if (unlikely(inode->i_nlink == 0)) {
388 2928010 : WARN_ON(!(inode->i_state & I_LINKABLE));
389 2928010 : atomic_long_dec(&inode->i_sb->s_remove_count);
390 : }
391 :
392 40007013 : inode->__i_nlink++;
393 40007013 : }
394 : EXPORT_SYMBOL(inc_nlink);
395 :
396 585237527 : static void __address_space_init_once(struct address_space *mapping)
397 : {
398 585237527 : xa_init_flags(&mapping->i_pages, XA_FLAGS_LOCK_IRQ | XA_FLAGS_ACCOUNT);
399 585235129 : init_rwsem(&mapping->i_mmap_rwsem);
400 585227214 : INIT_LIST_HEAD(&mapping->private_list);
401 585227214 : spin_lock_init(&mapping->private_lock);
402 585196663 : mapping->i_mmap = RB_ROOT_CACHED;
403 585196663 : }
404 :
405 0 : void address_space_init_once(struct address_space *mapping)
406 : {
407 0 : memset(mapping, 0, sizeof(*mapping));
408 0 : __address_space_init_once(mapping);
409 0 : }
410 : EXPORT_SYMBOL(address_space_init_once);
411 :
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 585220810 : void inode_init_once(struct inode *inode)
418 : {
419 585220810 : memset(inode, 0, sizeof(*inode));
420 585220810 : INIT_HLIST_NODE(&inode->i_hash);
421 585220810 : INIT_LIST_HEAD(&inode->i_devices);
422 585220810 : INIT_LIST_HEAD(&inode->i_io_list);
423 585220810 : INIT_LIST_HEAD(&inode->i_wb_list);
424 585220810 : INIT_LIST_HEAD(&inode->i_lru);
425 585220810 : INIT_LIST_HEAD(&inode->i_sb_list);
426 585220810 : __address_space_init_once(&inode->i_data);
427 585216712 : i_size_ordered_init(inode);
428 585216712 : }
429 : EXPORT_SYMBOL(inode_init_once);
430 :
431 632360 : static void init_once(void *foo)
432 : {
433 632360 : struct inode *inode = (struct inode *) foo;
434 :
435 632360 : inode_init_once(inode);
436 632360 : }
437 :
438 : /*
439 : * inode->i_lock must be held
440 : */
441 17447161 : void __iget(struct inode *inode)
442 : {
443 17447161 : atomic_inc(&inode->i_count);
444 17447349 : }
445 :
446 : /*
447 : * get additional reference to inode; caller must already hold one.
448 : */
449 237303188 : void ihold(struct inode *inode)
450 : {
451 237303188 : WARN_ON(atomic_inc_return(&inode->i_count) < 2);
452 237354872 : }
453 : EXPORT_SYMBOL(ihold);
454 :
455 66138027380 : static void __inode_add_lru(struct inode *inode, bool rotate)
456 : {
457 66138027380 : if (inode->i_state & (I_DIRTY_ALL | I_SYNC | I_FREEING | I_WILL_FREE))
458 : return;
459 66065159489 : if (atomic_read(&inode->i_count))
460 : return;
461 65967574656 : if (!(inode->i_sb->s_flags & SB_ACTIVE))
462 : return;
463 66086619330 : if (!mapping_shrinkable(&inode->i_data))
464 : return;
465 :
466 62309246788 : if (list_lru_add(&inode->i_sb->s_inode_lru, &inode->i_lru))
467 525413697 : this_cpu_inc(nr_unused);
468 62820481201 : else if (rotate)
469 62813211413 : inode->i_state |= I_REFERENCED;
470 : }
471 :
472 : /*
473 : * Add inode to LRU if needed (inode is unused and clean).
474 : *
475 : * Needs inode->i_lock held.
476 : */
477 168197438 : void inode_add_lru(struct inode *inode)
478 : {
479 168197438 : __inode_add_lru(inode, false);
480 168102940 : }
481 :
482 489978054 : static void inode_lru_list_del(struct inode *inode)
483 : {
484 489978054 : if (list_lru_del(&inode->i_sb->s_inode_lru, &inode->i_lru))
485 482194529 : this_cpu_dec(nr_unused);
486 490045601 : }
487 :
488 : /**
489 : * inode_sb_list_add - add inode to the superblock list of inodes
490 : * @inode: inode to add
491 : */
492 1594188709 : void inode_sb_list_add(struct inode *inode)
493 : {
494 1594188709 : spin_lock(&inode->i_sb->s_inode_list_lock);
495 1596132176 : list_add(&inode->i_sb_list, &inode->i_sb->s_inodes);
496 1596069278 : spin_unlock(&inode->i_sb->s_inode_list_lock);
497 1595859192 : }
498 : EXPORT_SYMBOL_GPL(inode_sb_list_add);
499 :
500 1627620942 : static inline void inode_sb_list_del(struct inode *inode)
501 : {
502 1627620942 : if (!list_empty(&inode->i_sb_list)) {
503 1594187316 : spin_lock(&inode->i_sb->s_inode_list_lock);
504 1596195912 : list_del_init(&inode->i_sb_list);
505 1596159587 : spin_unlock(&inode->i_sb->s_inode_list_lock);
506 : }
507 1629534941 : }
508 :
509 15726185 : static unsigned long hash(struct super_block *sb, unsigned long hashval)
510 : {
511 15726185 : unsigned long tmp;
512 :
513 15726185 : tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) /
514 : L1_CACHE_BYTES;
515 15726185 : tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> i_hash_shift);
516 15726185 : return tmp & i_hash_mask;
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 5471 : void __insert_inode_hash(struct inode *inode, unsigned long hashval)
528 : {
529 5471 : struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval);
530 :
531 5471 : spin_lock(&inode_hash_lock);
532 5471 : spin_lock(&inode->i_lock);
533 5471 : hlist_add_head_rcu(&inode->i_hash, b);
534 5471 : spin_unlock(&inode->i_lock);
535 5471 : spin_unlock(&inode_hash_lock);
536 5471 : }
537 : EXPORT_SYMBOL(__insert_inode_hash);
538 :
539 : /**
540 : * __remove_inode_hash - remove an inode from the hash
541 : * @inode: inode to unhash
542 : *
543 : * Remove an inode from the superblock.
544 : */
545 11644449 : void __remove_inode_hash(struct inode *inode)
546 : {
547 11644449 : spin_lock(&inode_hash_lock);
548 11646865 : spin_lock(&inode->i_lock);
549 11646865 : hlist_del_init_rcu(&inode->i_hash);
550 11646865 : spin_unlock(&inode->i_lock);
551 11646865 : spin_unlock(&inode_hash_lock);
552 11646860 : }
553 : EXPORT_SYMBOL(__remove_inode_hash);
554 :
555 0 : void dump_mapping(const struct address_space *mapping)
556 : {
557 0 : struct inode *host;
558 0 : const struct address_space_operations *a_ops;
559 0 : struct hlist_node *dentry_first;
560 0 : struct dentry *dentry_ptr;
561 0 : struct dentry dentry;
562 0 : 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 0 : if (get_kernel_nofault(host, &mapping->host) ||
569 0 : get_kernel_nofault(a_ops, &mapping->a_ops)) {
570 0 : pr_warn("invalid mapping:%px\n", mapping);
571 0 : return;
572 : }
573 :
574 0 : if (!host) {
575 0 : pr_warn("aops:%ps\n", a_ops);
576 0 : return;
577 : }
578 :
579 0 : if (get_kernel_nofault(dentry_first, &host->i_dentry.first) ||
580 0 : get_kernel_nofault(ino, &host->i_ino)) {
581 0 : pr_warn("aops:%ps invalid inode:%px\n", a_ops, host);
582 0 : return;
583 : }
584 :
585 0 : if (!dentry_first) {
586 0 : pr_warn("aops:%ps ino:%lx\n", a_ops, ino);
587 0 : return;
588 : }
589 :
590 0 : dentry_ptr = container_of(dentry_first, struct dentry, d_u.d_alias);
591 0 : if (get_kernel_nofault(dentry, dentry_ptr)) {
592 0 : pr_warn("aops:%ps ino:%lx invalid dentry:%px\n",
593 : a_ops, ino, dentry_ptr);
594 0 : 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 0 : pr_warn("aops:%ps ino:%lx dentry name:\"%pd\"\n", a_ops, ino, &dentry);
602 : }
603 :
604 1627826525 : void clear_inode(struct inode *inode)
605 : {
606 : /*
607 : * We have to cycle the i_pages lock here because reclaim can be in the
608 : * process of removing the last page (in __filemap_remove_folio())
609 : * and we must not free the mapping under it.
610 : */
611 1627826525 : xa_lock_irq(&inode->i_data.i_pages);
612 1628927695 : BUG_ON(inode->i_data.nrpages);
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 : */
621 1628927695 : xa_unlock_irq(&inode->i_data.i_pages);
622 1628608676 : BUG_ON(!list_empty(&inode->i_data.private_list));
623 1628608676 : BUG_ON(!(inode->i_state & I_FREEING));
624 1628608676 : BUG_ON(inode->i_state & I_CLEAR);
625 1628608676 : BUG_ON(!list_empty(&inode->i_wb_list));
626 : /* don't need i_lock here, no concurrent mods to i_state */
627 1628608676 : inode->i_state = I_FREEING | I_CLEAR;
628 1628608676 : }
629 : EXPORT_SYMBOL(clear_inode);
630 :
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 : */
644 1628044590 : static void evict(struct inode *inode)
645 : {
646 1628044590 : const struct super_operations *op = inode->i_sb->s_op;
647 :
648 1628044590 : BUG_ON(!(inode->i_state & I_FREEING));
649 1628044590 : BUG_ON(!list_empty(&inode->i_lru));
650 :
651 1628044590 : if (!list_empty(&inode->i_io_list))
652 30821806 : inode_io_list_del(inode);
653 :
654 1628062825 : inode_sb_list_del(inode);
655 :
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 1629499626 : inode_wait_for_writeback(inode);
663 :
664 1628634028 : if (op->evict_inode) {
665 476782676 : op->evict_inode(inode);
666 : } else {
667 1151851352 : truncate_inode_pages_final(&inode->i_data);
668 1151720408 : clear_inode(inode);
669 : }
670 1628011972 : if (S_ISCHR(inode->i_mode) && inode->i_cdev)
671 2192 : cd_forget(inode);
672 :
673 1628011972 : remove_inode_hash(inode);
674 :
675 1627276172 : spin_lock(&inode->i_lock);
676 1627829278 : wake_up_bit(&inode->i_state, __I_NEW);
677 1627415441 : BUG_ON(inode->i_state != (I_FREEING | I_CLEAR));
678 1627415441 : spin_unlock(&inode->i_lock);
679 :
680 1628324772 : destroy_inode(inode);
681 1627425458 : }
682 :
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 543133 : static void dispose_list(struct list_head *head)
691 : {
692 465904297 : while (!list_empty(head)) {
693 465361185 : struct inode *inode;
694 :
695 465361185 : inode = list_first_entry(head, struct inode, i_lru);
696 465361185 : list_del_init(&inode->i_lru);
697 :
698 465361185 : evict(inode);
699 465361185 : cond_resched();
700 : }
701 543112 : }
702 :
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
708 : * called by superblock shutdown after having SB_ACTIVE flag removed,
709 : * so any inode reaching zero refcount during or after that call will
710 : * be immediately evicted.
711 : */
712 179989 : void evict_inodes(struct super_block *sb)
713 : {
714 179989 : struct inode *inode, *next;
715 179989 : LIST_HEAD(dispose);
716 :
717 188752 : again:
718 188752 : spin_lock(&sb->s_inode_list_lock);
719 439981452 : list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
720 439801435 : if (atomic_read(&inode->i_count))
721 825087 : continue;
722 :
723 438976348 : spin_lock(&inode->i_lock);
724 438976312 : if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
725 1 : spin_unlock(&inode->i_lock);
726 1 : continue;
727 : }
728 :
729 438976311 : inode->i_state |= I_FREEING;
730 438976311 : inode_lru_list_del(inode);
731 438976311 : spin_unlock(&inode->i_lock);
732 438976311 : list_add(&inode->i_lru, &dispose);
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 438976311 : if (need_resched()) {
740 8763 : spin_unlock(&sb->s_inode_list_lock);
741 8763 : cond_resched();
742 8763 : dispose_list(&dispose);
743 8763 : goto again;
744 : }
745 : }
746 180017 : spin_unlock(&sb->s_inode_list_lock);
747 :
748 180050 : dispose_list(&dispose);
749 179983 : }
750 : EXPORT_SYMBOL_GPL(evict_inodes);
751 :
752 : /**
753 : * invalidate_inodes - attempt to free all inodes on a superblock
754 : * @sb: superblock to operate on
755 : * @kill_dirty: flag to guide handling of dirty inodes
756 : *
757 : * Attempts to free all inodes for a given superblock. If there were any
758 : * busy inodes return a non-zero value, else zero.
759 : * If @kill_dirty is set, discard dirty inodes too, otherwise treat
760 : * them as busy.
761 : */
762 0 : int invalidate_inodes(struct super_block *sb, bool kill_dirty)
763 : {
764 0 : int busy = 0;
765 0 : struct inode *inode, *next;
766 0 : LIST_HEAD(dispose);
767 :
768 0 : again:
769 0 : spin_lock(&sb->s_inode_list_lock);
770 0 : list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
771 0 : spin_lock(&inode->i_lock);
772 0 : if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
773 0 : spin_unlock(&inode->i_lock);
774 0 : continue;
775 : }
776 0 : if (inode->i_state & I_DIRTY_ALL && !kill_dirty) {
777 0 : spin_unlock(&inode->i_lock);
778 0 : busy = 1;
779 0 : continue;
780 : }
781 0 : if (atomic_read(&inode->i_count)) {
782 0 : spin_unlock(&inode->i_lock);
783 0 : busy = 1;
784 0 : continue;
785 : }
786 :
787 0 : inode->i_state |= I_FREEING;
788 0 : inode_lru_list_del(inode);
789 0 : spin_unlock(&inode->i_lock);
790 0 : list_add(&inode->i_lru, &dispose);
791 0 : if (need_resched()) {
792 0 : spin_unlock(&sb->s_inode_list_lock);
793 0 : cond_resched();
794 0 : dispose_list(&dispose);
795 0 : goto again;
796 : }
797 : }
798 0 : spin_unlock(&sb->s_inode_list_lock);
799 :
800 0 : dispose_list(&dispose);
801 :
802 0 : return busy;
803 : }
804 :
805 : /*
806 : * Isolate the inode from the LRU in preparation for freeing it.
807 : *
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.
815 : */
816 218886731 : static enum lru_status inode_lru_isolate(struct list_head *item,
817 : struct list_lru_one *lru, spinlock_t *lru_lock, void *arg)
818 : {
819 218886731 : struct list_head *freeable = arg;
820 218886731 : struct inode *inode = container_of(item, struct inode, i_lru);
821 :
822 : /*
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.
825 : */
826 218886731 : if (!spin_trylock(&inode->i_lock))
827 : return LRU_SKIP;
828 :
829 : /*
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.
834 : */
835 218882853 : if (atomic_read(&inode->i_count) ||
836 404182400 : (inode->i_state & ~I_REFERENCED) ||
837 202091022 : !mapping_shrinkable(&inode->i_data)) {
838 16809186 : list_lru_isolate(lru, &inode->i_lru);
839 16809186 : spin_unlock(&inode->i_lock);
840 16809186 : this_cpu_dec(nr_unused);
841 16809186 : return LRU_REMOVED;
842 : }
843 :
844 : /* Recently referenced inodes get one more pass */
845 202073667 : if (inode->i_state & I_REFERENCED) {
846 175660881 : inode->i_state &= ~I_REFERENCED;
847 175660881 : spin_unlock(&inode->i_lock);
848 175660881 : return LRU_ROTATE;
849 : }
850 :
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 : */
856 26412786 : if (inode_has_buffers(inode) || !mapping_empty(&inode->i_data)) {
857 27912 : __iget(inode);
858 27912 : spin_unlock(&inode->i_lock);
859 27912 : spin_unlock(lru_lock);
860 27912 : if (remove_inode_buffers(inode)) {
861 27912 : unsigned long reap;
862 27912 : reap = invalidate_mapping_pages(&inode->i_data, 0, -1);
863 27912 : if (current_is_kswapd())
864 27747 : __count_vm_events(KSWAPD_INODESTEAL, reap);
865 : else
866 165 : __count_vm_events(PGINODESTEAL, reap);
867 27912 : mm_account_reclaimed_pages(reap);
868 : }
869 27912 : iput(inode);
870 27912 : spin_lock(lru_lock);
871 27912 : return LRU_RETRY;
872 : }
873 :
874 26384874 : WARN_ON(inode->i_state & I_NEW);
875 26384874 : inode->i_state |= I_FREEING;
876 26384874 : list_lru_isolate_move(lru, &inode->i_lru, freeable);
877 26384874 : spin_unlock(&inode->i_lock);
878 :
879 26384874 : this_cpu_dec(nr_unused);
880 26384874 : return LRU_REMOVED;
881 : }
882 :
883 : /*
884 : * Walk the superblock inode LRU for freeable inodes and attempt to free them.
885 : * This is called from the superblock shrinker function with a number of inodes
886 : * to trim from the LRU. Inodes to be freed are moved to a temporary list and
887 : * then are freed outside inode_lock by dispose_list().
888 : */
889 354363 : long prune_icache_sb(struct super_block *sb, struct shrink_control *sc)
890 : {
891 354363 : LIST_HEAD(freeable);
892 354363 : long freed;
893 :
894 354363 : freed = list_lru_shrink_walk(&sb->s_inode_lru, sc,
895 : inode_lru_isolate, &freeable);
896 354363 : dispose_list(&freeable);
897 354363 : return freed;
898 : }
899 :
900 : static void __wait_on_freeing_inode(struct inode *inode);
901 : /*
902 : * Called with the inode lock held.
903 : */
904 7198431 : static struct inode *find_inode(struct super_block *sb,
905 : struct hlist_head *head,
906 : int (*test)(struct inode *, void *),
907 : void *data)
908 : {
909 7198431 : struct inode *inode = NULL;
910 :
911 7198433 : repeat:
912 14831888 : hlist_for_each_entry(inode, head, i_hash) {
913 870551 : if (inode->i_sb != sb)
914 14194 : continue;
915 856357 : if (!test(inode, data))
916 420828 : continue;
917 435529 : spin_lock(&inode->i_lock);
918 435529 : if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
919 2 : __wait_on_freeing_inode(inode);
920 2 : goto repeat;
921 : }
922 435527 : if (unlikely(inode->i_state & I_CREATING)) {
923 0 : spin_unlock(&inode->i_lock);
924 0 : return ERR_PTR(-ESTALE);
925 : }
926 435527 : __iget(inode);
927 435527 : spin_unlock(&inode->i_lock);
928 435527 : return inode;
929 : }
930 : return NULL;
931 : }
932 :
933 : /*
934 : * find_inode_fast is the fast path version of find_inode, see the comment at
935 : * iget_locked for details.
936 : */
937 9562165 : static struct inode *find_inode_fast(struct super_block *sb,
938 : struct hlist_head *head, unsigned long ino)
939 : {
940 9562165 : struct inode *inode = NULL;
941 :
942 9562165 : repeat:
943 19151642 : hlist_for_each_entry(inode, head, i_hash) {
944 1983113 : if (inode->i_ino != ino)
945 27312 : continue;
946 1955801 : if (inode->i_sb != sb)
947 0 : continue;
948 1955801 : spin_lock(&inode->i_lock);
949 1955801 : if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
950 0 : __wait_on_freeing_inode(inode);
951 0 : goto repeat;
952 : }
953 1955801 : if (unlikely(inode->i_state & I_CREATING)) {
954 0 : spin_unlock(&inode->i_lock);
955 0 : return ERR_PTR(-ESTALE);
956 : }
957 1955801 : __iget(inode);
958 1955801 : spin_unlock(&inode->i_lock);
959 1955801 : return inode;
960 : }
961 : return NULL;
962 : }
963 :
964 : /*
965 : * Each cpu owns a range of LAST_INO_BATCH numbers.
966 : * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
967 : * to renew the exhausted range.
968 : *
969 : * This does not significantly increase overflow rate because every CPU can
970 : * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
971 : * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
972 : * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
973 : * overflow rate by 2x, which does not seem too significant.
974 : *
975 : * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
976 : * error if st_ino won't fit in target struct field. Use 32bit counter
977 : * here to attempt to avoid that.
978 : */
979 : #define LAST_INO_BATCH 1024
980 : static DEFINE_PER_CPU(unsigned int, last_ino);
981 :
982 56432815 : unsigned int get_next_ino(void)
983 : {
984 56432815 : unsigned int *p = &get_cpu_var(last_ino);
985 56424817 : unsigned int res = *p;
986 :
987 : #ifdef CONFIG_SMP
988 56424817 : if (unlikely((res & (LAST_INO_BATCH-1)) == 0)) {
989 55118 : static atomic_t shared_last_ino;
990 55118 : int next = atomic_add_return(LAST_INO_BATCH, &shared_last_ino);
991 :
992 55118 : res = next - LAST_INO_BATCH;
993 : }
994 : #endif
995 :
996 56424817 : res++;
997 : /* get_next_ino should not provide a 0 inode number */
998 56424817 : if (unlikely(!res))
999 0 : res++;
1000 56424817 : *p = res;
1001 56424817 : put_cpu_var(last_ino);
1002 56420491 : return res;
1003 : }
1004 : EXPORT_SYMBOL(get_next_ino);
1005 :
1006 : /**
1007 : * new_inode_pseudo - obtain an inode
1008 : * @sb: superblock
1009 : *
1010 : * Allocates a new inode for given superblock.
1011 : * Inode wont be chained in superblock s_inodes list
1012 : * This means :
1013 : * - fs can't be unmount
1014 : * - quotas, fsnotify, writeback can't work
1015 : */
1016 505417423 : struct inode *new_inode_pseudo(struct super_block *sb)
1017 : {
1018 505417423 : struct inode *inode = alloc_inode(sb);
1019 :
1020 507309780 : if (inode) {
1021 507309780 : spin_lock(&inode->i_lock);
1022 507574211 : inode->i_state = 0;
1023 507574211 : spin_unlock(&inode->i_lock);
1024 : }
1025 507688331 : return inode;
1026 : }
1027 :
1028 : /**
1029 : * new_inode - obtain an inode
1030 : * @sb: superblock
1031 : *
1032 : * Allocates a new inode for given superblock. The default gfp_mask
1033 : * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
1034 : * If HIGHMEM pages are unsuitable or it is known that pages allocated
1035 : * for the page cache are not reclaimable or migratable,
1036 : * mapping_set_gfp_mask() must be called with suitable flags on the
1037 : * newly created inode's mapping
1038 : *
1039 : */
1040 472736008 : struct inode *new_inode(struct super_block *sb)
1041 : {
1042 472736008 : struct inode *inode;
1043 :
1044 472736008 : spin_lock_prefetch(&sb->s_inode_list_lock);
1045 :
1046 472357988 : inode = new_inode_pseudo(sb);
1047 474241297 : if (inode)
1048 474241297 : inode_sb_list_add(inode);
1049 474670927 : return inode;
1050 : }
1051 : EXPORT_SYMBOL(new_inode);
1052 :
1053 : #ifdef CONFIG_DEBUG_LOCK_ALLOC
1054 : void lockdep_annotate_inode_mutex_key(struct inode *inode)
1055 : {
1056 : if (S_ISDIR(inode->i_mode)) {
1057 : struct file_system_type *type = inode->i_sb->s_type;
1058 :
1059 : /* Set new key only if filesystem hasn't already changed it */
1060 : if (lockdep_match_class(&inode->i_rwsem, &type->i_mutex_key)) {
1061 : /*
1062 : * ensure nobody is actually holding i_mutex
1063 : */
1064 : // mutex_destroy(&inode->i_mutex);
1065 : init_rwsem(&inode->i_rwsem);
1066 : lockdep_set_class(&inode->i_rwsem,
1067 : &type->i_mutex_dir_key);
1068 : }
1069 : }
1070 : }
1071 : EXPORT_SYMBOL(lockdep_annotate_inode_mutex_key);
1072 : #endif
1073 :
1074 : /**
1075 : * unlock_new_inode - clear the I_NEW state and wake up any waiters
1076 : * @inode: new inode to unlock
1077 : *
1078 : * Called when the inode is fully initialised to clear the new state of the
1079 : * inode and wake up anyone waiting for the inode to finish initialisation.
1080 : */
1081 1121453863 : void unlock_new_inode(struct inode *inode)
1082 : {
1083 1121453863 : lockdep_annotate_inode_mutex_key(inode);
1084 1121453863 : spin_lock(&inode->i_lock);
1085 1121698241 : WARN_ON(!(inode->i_state & I_NEW));
1086 1121698241 : inode->i_state &= ~I_NEW & ~I_CREATING;
1087 1121698241 : smp_mb();
1088 1121673859 : wake_up_bit(&inode->i_state, __I_NEW);
1089 1121646203 : spin_unlock(&inode->i_lock);
1090 1121692275 : }
1091 : EXPORT_SYMBOL(unlock_new_inode);
1092 :
1093 0 : void discard_new_inode(struct inode *inode)
1094 : {
1095 0 : lockdep_annotate_inode_mutex_key(inode);
1096 0 : spin_lock(&inode->i_lock);
1097 0 : WARN_ON(!(inode->i_state & I_NEW));
1098 0 : inode->i_state &= ~I_NEW;
1099 0 : smp_mb();
1100 0 : wake_up_bit(&inode->i_state, __I_NEW);
1101 0 : spin_unlock(&inode->i_lock);
1102 0 : iput(inode);
1103 0 : }
1104 : EXPORT_SYMBOL(discard_new_inode);
1105 :
1106 : /**
1107 : * lock_two_inodes - lock two inodes (may be regular files but also dirs)
1108 : *
1109 : * Lock any non-NULL argument. The caller must make sure that if he is passing
1110 : * in two directories, one is not ancestor of the other. Zero, one or two
1111 : * objects may be locked by this function.
1112 : *
1113 : * @inode1: first inode to lock
1114 : * @inode2: second inode to lock
1115 : * @subclass1: inode lock subclass for the first lock obtained
1116 : * @subclass2: inode lock subclass for the second lock obtained
1117 : */
1118 94644744 : void lock_two_inodes(struct inode *inode1, struct inode *inode2,
1119 : unsigned subclass1, unsigned subclass2)
1120 : {
1121 94644744 : if (!inode1 || !inode2) {
1122 : /*
1123 : * Make sure @subclass1 will be used for the acquired lock.
1124 : * This is not strictly necessary (no current caller cares) but
1125 : * let's keep things consistent.
1126 : */
1127 35155724 : if (!inode1)
1128 0 : swap(inode1, inode2);
1129 35155724 : goto lock;
1130 : }
1131 :
1132 : /*
1133 : * If one object is directory and the other is not, we must make sure
1134 : * to lock directory first as the other object may be its child.
1135 : */
1136 59489020 : if (S_ISDIR(inode2->i_mode) == S_ISDIR(inode1->i_mode)) {
1137 59488804 : if (inode1 > inode2)
1138 29934794 : swap(inode1, inode2);
1139 216 : } else if (!S_ISDIR(inode1->i_mode))
1140 108 : swap(inode1, inode2);
1141 108 : lock:
1142 94644744 : if (inode1)
1143 94644607 : inode_lock_nested(inode1, subclass1);
1144 94644682 : if (inode2 && inode2 != inode1)
1145 59488766 : inode_lock_nested(inode2, subclass2);
1146 94644837 : }
1147 :
1148 : /**
1149 : * lock_two_nondirectories - take two i_mutexes on non-directory objects
1150 : *
1151 : * Lock any non-NULL argument. Passed objects must not be directories.
1152 : * Zero, one or two objects may be locked by this function.
1153 : *
1154 : * @inode1: first inode to lock
1155 : * @inode2: second inode to lock
1156 : */
1157 1853383 : void lock_two_nondirectories(struct inode *inode1, struct inode *inode2)
1158 : {
1159 1853383 : if (inode1)
1160 1853383 : WARN_ON_ONCE(S_ISDIR(inode1->i_mode));
1161 1853383 : if (inode2)
1162 1853383 : WARN_ON_ONCE(S_ISDIR(inode2->i_mode));
1163 1853383 : lock_two_inodes(inode1, inode2, I_MUTEX_NORMAL, I_MUTEX_NONDIR2);
1164 1853513 : }
1165 : EXPORT_SYMBOL(lock_two_nondirectories);
1166 :
1167 : /**
1168 : * unlock_two_nondirectories - release locks from lock_two_nondirectories()
1169 : * @inode1: first inode to unlock
1170 : * @inode2: second inode to unlock
1171 : */
1172 1853596 : void unlock_two_nondirectories(struct inode *inode1, struct inode *inode2)
1173 : {
1174 1853596 : if (inode1) {
1175 1853596 : WARN_ON_ONCE(S_ISDIR(inode1->i_mode));
1176 1853596 : inode_unlock(inode1);
1177 : }
1178 1853612 : if (inode2 && inode2 != inode1) {
1179 1853604 : WARN_ON_ONCE(S_ISDIR(inode2->i_mode));
1180 1853604 : inode_unlock(inode2);
1181 : }
1182 1853617 : }
1183 : EXPORT_SYMBOL(unlock_two_nondirectories);
1184 :
1185 : /**
1186 : * inode_insert5 - obtain an inode from a mounted file system
1187 : * @inode: pre-allocated inode to use for insert to cache
1188 : * @hashval: hash value (usually inode number) to get
1189 : * @test: callback used for comparisons between inodes
1190 : * @set: callback used to initialize a new struct inode
1191 : * @data: opaque data pointer to pass to @test and @set
1192 : *
1193 : * Search for the inode specified by @hashval and @data in the inode cache,
1194 : * and if present it is return it with an increased reference count. This is
1195 : * a variant of iget5_locked() for callers that don't want to fail on memory
1196 : * allocation of inode.
1197 : *
1198 : * If the inode is not in cache, insert the pre-allocated inode to cache and
1199 : * return it locked, hashed, and with the I_NEW flag set. The file system gets
1200 : * to fill it in before unlocking it via unlock_new_inode().
1201 : *
1202 : * Note both @test and @set are called with the inode_hash_lock held, so can't
1203 : * sleep.
1204 : */
1205 4969447 : struct inode *inode_insert5(struct inode *inode, unsigned long hashval,
1206 : int (*test)(struct inode *, void *),
1207 : int (*set)(struct inode *, void *), void *data)
1208 : {
1209 4969447 : struct hlist_head *head = inode_hashtable + hash(inode->i_sb, hashval);
1210 4963078 : struct inode *old;
1211 :
1212 4963078 : again:
1213 4963078 : spin_lock(&inode_hash_lock);
1214 4981970 : old = find_inode(inode->i_sb, head, test, data);
1215 4981970 : if (unlikely(old)) {
1216 : /*
1217 : * Uhhuh, somebody else created the same inode under us.
1218 : * Use the old inode instead of the preallocated one.
1219 : */
1220 0 : spin_unlock(&inode_hash_lock);
1221 0 : if (IS_ERR(old))
1222 : return NULL;
1223 0 : wait_on_inode(old);
1224 0 : if (unlikely(inode_unhashed(old))) {
1225 0 : iput(old);
1226 0 : goto again;
1227 : }
1228 0 : return old;
1229 : }
1230 :
1231 4981970 : if (set && unlikely(set(inode, data))) {
1232 0 : inode = NULL;
1233 0 : goto unlock;
1234 : }
1235 :
1236 : /*
1237 : * Return the locked inode with I_NEW set, the
1238 : * caller is responsible for filling in the contents
1239 : */
1240 4981970 : spin_lock(&inode->i_lock);
1241 4981970 : inode->i_state |= I_NEW;
1242 4981970 : hlist_add_head_rcu(&inode->i_hash, head);
1243 4981970 : spin_unlock(&inode->i_lock);
1244 :
1245 : /*
1246 : * Add inode to the sb list if it's not already. It has I_NEW at this
1247 : * point, so it should be safe to test i_sb_list locklessly.
1248 : */
1249 4981970 : if (list_empty(&inode->i_sb_list))
1250 1609062 : inode_sb_list_add(inode);
1251 3372908 : unlock:
1252 4981970 : spin_unlock(&inode_hash_lock);
1253 :
1254 4981970 : return inode;
1255 : }
1256 : EXPORT_SYMBOL(inode_insert5);
1257 :
1258 : /**
1259 : * iget5_locked - obtain an inode from a mounted file system
1260 : * @sb: super block of file system
1261 : * @hashval: hash value (usually inode number) to get
1262 : * @test: callback used for comparisons between inodes
1263 : * @set: callback used to initialize a new struct inode
1264 : * @data: opaque data pointer to pass to @test and @set
1265 : *
1266 : * Search for the inode specified by @hashval and @data in the inode cache,
1267 : * and if present it is return it with an increased reference count. This is
1268 : * a generalized version of iget_locked() for file systems where the inode
1269 : * number is not sufficient for unique identification of an inode.
1270 : *
1271 : * If the inode is not in cache, allocate a new inode and return it locked,
1272 : * hashed, and with the I_NEW flag set. The file system gets to fill it in
1273 : * before unlocking it via unlock_new_inode().
1274 : *
1275 : * Note both @test and @set are called with the inode_hash_lock held, so can't
1276 : * sleep.
1277 : */
1278 2044233 : struct inode *iget5_locked(struct super_block *sb, unsigned long hashval,
1279 : int (*test)(struct inode *, void *),
1280 : int (*set)(struct inode *, void *), void *data)
1281 : {
1282 2044233 : struct inode *inode = ilookup5(sb, hashval, test, data);
1283 :
1284 2044479 : if (!inode) {
1285 1609058 : struct inode *new = alloc_inode(sb);
1286 :
1287 1609011 : if (new) {
1288 1609011 : new->i_state = 0;
1289 1609011 : inode = inode_insert5(new, hashval, test, set, data);
1290 1609062 : if (unlikely(inode != new))
1291 0 : destroy_inode(new);
1292 : }
1293 : }
1294 2044483 : return inode;
1295 : }
1296 : EXPORT_SYMBOL(iget5_locked);
1297 :
1298 : /**
1299 : * iget_locked - obtain an inode from a mounted file system
1300 : * @sb: super block of file system
1301 : * @ino: inode number to get
1302 : *
1303 : * Search for the inode specified by @ino in the inode cache and if present
1304 : * return it with an increased reference count. This is for file systems
1305 : * where the inode number is sufficient for unique identification of an inode.
1306 : *
1307 : * If the inode is not in cache, allocate a new inode and return it locked,
1308 : * hashed, and with the I_NEW flag set. The file system gets to fill it in
1309 : * before unlocking it via unlock_new_inode().
1310 : */
1311 3876383 : struct inode *iget_locked(struct super_block *sb, unsigned long ino)
1312 : {
1313 3876383 : struct hlist_head *head = inode_hashtable + hash(sb, ino);
1314 3876366 : struct inode *inode;
1315 : again:
1316 3876366 : spin_lock(&inode_hash_lock);
1317 3876409 : inode = find_inode_fast(sb, head, ino);
1318 3876409 : spin_unlock(&inode_hash_lock);
1319 3876409 : if (inode) {
1320 74133 : if (IS_ERR(inode))
1321 : return NULL;
1322 74133 : wait_on_inode(inode);
1323 74130 : if (unlikely(inode_unhashed(inode))) {
1324 0 : iput(inode);
1325 0 : goto again;
1326 : }
1327 74130 : return inode;
1328 : }
1329 :
1330 3802276 : inode = alloc_inode(sb);
1331 3802274 : if (inode) {
1332 3802274 : struct inode *old;
1333 :
1334 3802274 : spin_lock(&inode_hash_lock);
1335 : /* We released the lock, so.. */
1336 3802276 : old = find_inode_fast(sb, head, ino);
1337 3802276 : if (!old) {
1338 3802276 : inode->i_ino = ino;
1339 3802276 : spin_lock(&inode->i_lock);
1340 3802276 : inode->i_state = I_NEW;
1341 3802276 : hlist_add_head_rcu(&inode->i_hash, head);
1342 3802276 : spin_unlock(&inode->i_lock);
1343 3802276 : inode_sb_list_add(inode);
1344 3802276 : spin_unlock(&inode_hash_lock);
1345 :
1346 : /* Return the locked inode with I_NEW set, the
1347 : * caller is responsible for filling in the contents
1348 : */
1349 3802276 : return inode;
1350 : }
1351 :
1352 : /*
1353 : * Uhhuh, somebody else created the same inode under
1354 : * us. Use the old inode instead of the one we just
1355 : * allocated.
1356 : */
1357 0 : spin_unlock(&inode_hash_lock);
1358 0 : destroy_inode(inode);
1359 0 : if (IS_ERR(old))
1360 : return NULL;
1361 0 : inode = old;
1362 0 : wait_on_inode(inode);
1363 0 : if (unlikely(inode_unhashed(inode))) {
1364 0 : iput(inode);
1365 0 : goto again;
1366 : }
1367 : }
1368 : return inode;
1369 : }
1370 : EXPORT_SYMBOL(iget_locked);
1371 :
1372 : /*
1373 : * search the inode cache for a matching inode number.
1374 : * If we find one, then the inode number we are trying to
1375 : * allocate is not unique and so we should not use it.
1376 : *
1377 : * Returns 1 if the inode number is unique, 0 if it is not.
1378 : */
1379 0 : static int test_inode_iunique(struct super_block *sb, unsigned long ino)
1380 : {
1381 0 : struct hlist_head *b = inode_hashtable + hash(sb, ino);
1382 0 : struct inode *inode;
1383 :
1384 0 : hlist_for_each_entry_rcu(inode, b, i_hash) {
1385 0 : if (inode->i_ino == ino && inode->i_sb == sb)
1386 : return 0;
1387 : }
1388 : return 1;
1389 : }
1390 :
1391 : /**
1392 : * iunique - get a unique inode number
1393 : * @sb: superblock
1394 : * @max_reserved: highest reserved inode number
1395 : *
1396 : * Obtain an inode number that is unique on the system for a given
1397 : * superblock. This is used by file systems that have no natural
1398 : * permanent inode numbering system. An inode number is returned that
1399 : * is higher than the reserved limit but unique.
1400 : *
1401 : * BUGS:
1402 : * With a large number of inodes live on the file system this function
1403 : * currently becomes quite slow.
1404 : */
1405 0 : ino_t iunique(struct super_block *sb, ino_t max_reserved)
1406 : {
1407 : /*
1408 : * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1409 : * error if st_ino won't fit in target struct field. Use 32bit counter
1410 : * here to attempt to avoid that.
1411 : */
1412 0 : static DEFINE_SPINLOCK(iunique_lock);
1413 0 : static unsigned int counter;
1414 0 : ino_t res;
1415 :
1416 0 : rcu_read_lock();
1417 0 : spin_lock(&iunique_lock);
1418 0 : do {
1419 0 : if (counter <= max_reserved)
1420 0 : counter = max_reserved + 1;
1421 0 : res = counter++;
1422 0 : } while (!test_inode_iunique(sb, res));
1423 0 : spin_unlock(&iunique_lock);
1424 0 : rcu_read_unlock();
1425 :
1426 0 : return res;
1427 : }
1428 : EXPORT_SYMBOL(iunique);
1429 :
1430 >11329*10^7 : struct inode *igrab(struct inode *inode)
1431 : {
1432 >11329*10^7 : spin_lock(&inode->i_lock);
1433 >11379*10^7 : if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) {
1434 >11379*10^7 : __iget(inode);
1435 >11371*10^7 : spin_unlock(&inode->i_lock);
1436 : } else {
1437 412353 : spin_unlock(&inode->i_lock);
1438 : /*
1439 : * Handle the case where s_op->clear_inode is not been
1440 : * called yet, and somebody is calling igrab
1441 : * while the inode is getting freed.
1442 : */
1443 412353 : inode = NULL;
1444 : }
1445 >11378*10^7 : return inode;
1446 : }
1447 : EXPORT_SYMBOL(igrab);
1448 :
1449 : /**
1450 : * ilookup5_nowait - search for an inode in the inode cache
1451 : * @sb: super block of file system to search
1452 : * @hashval: hash value (usually inode number) to search for
1453 : * @test: callback used for comparisons between inodes
1454 : * @data: opaque data pointer to pass to @test
1455 : *
1456 : * Search for the inode specified by @hashval and @data in the inode cache.
1457 : * If the inode is in the cache, the inode is returned with an incremented
1458 : * reference count.
1459 : *
1460 : * Note: I_NEW is not waited upon so you have to be very careful what you do
1461 : * with the returned inode. You probably should be using ilookup5() instead.
1462 : *
1463 : * Note2: @test is called with the inode_hash_lock held, so can't sleep.
1464 : */
1465 2216013 : struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval,
1466 : int (*test)(struct inode *, void *), void *data)
1467 : {
1468 2216013 : struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1469 2215884 : struct inode *inode;
1470 :
1471 2215884 : spin_lock(&inode_hash_lock);
1472 2216461 : inode = find_inode(sb, head, test, data);
1473 2216461 : spin_unlock(&inode_hash_lock);
1474 :
1475 2216457 : return IS_ERR(inode) ? NULL : inode;
1476 : }
1477 : EXPORT_SYMBOL(ilookup5_nowait);
1478 :
1479 : /**
1480 : * ilookup5 - search for an inode in the inode cache
1481 : * @sb: super block of file system to search
1482 : * @hashval: hash value (usually inode number) to search for
1483 : * @test: callback used for comparisons between inodes
1484 : * @data: opaque data pointer to pass to @test
1485 : *
1486 : * Search for the inode specified by @hashval and @data in the inode cache,
1487 : * and if the inode is in the cache, return the inode with an incremented
1488 : * reference count. Waits on I_NEW before returning the inode.
1489 : * returned with an incremented reference count.
1490 : *
1491 : * This is a generalized version of ilookup() for file systems where the
1492 : * inode number is not sufficient for unique identification of an inode.
1493 : *
1494 : * Note: @test is called with the inode_hash_lock held, so can't sleep.
1495 : */
1496 2216033 : struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
1497 : int (*test)(struct inode *, void *), void *data)
1498 : {
1499 2216033 : struct inode *inode;
1500 2216033 : again:
1501 2216033 : inode = ilookup5_nowait(sb, hashval, test, data);
1502 2216457 : if (inode) {
1503 435527 : wait_on_inode(inode);
1504 435527 : if (unlikely(inode_unhashed(inode))) {
1505 0 : iput(inode);
1506 0 : goto again;
1507 : }
1508 : }
1509 2216457 : return inode;
1510 : }
1511 : EXPORT_SYMBOL(ilookup5);
1512 :
1513 : /**
1514 : * ilookup - search for an inode in the inode cache
1515 : * @sb: super block of file system to search
1516 : * @ino: inode number to search for
1517 : *
1518 : * Search for the inode @ino in the inode cache, and if the inode is in the
1519 : * cache, the inode is returned with an incremented reference count.
1520 : */
1521 1882231 : struct inode *ilookup(struct super_block *sb, unsigned long ino)
1522 : {
1523 1882231 : struct hlist_head *head = inode_hashtable + hash(sb, ino);
1524 1880489 : struct inode *inode;
1525 1880489 : again:
1526 1880489 : spin_lock(&inode_hash_lock);
1527 1883480 : inode = find_inode_fast(sb, head, ino);
1528 1883480 : spin_unlock(&inode_hash_lock);
1529 :
1530 1883480 : if (inode) {
1531 1881668 : if (IS_ERR(inode))
1532 : return NULL;
1533 1881668 : wait_on_inode(inode);
1534 1881614 : if (unlikely(inode_unhashed(inode))) {
1535 0 : iput(inode);
1536 0 : goto again;
1537 : }
1538 : }
1539 : return inode;
1540 : }
1541 : EXPORT_SYMBOL(ilookup);
1542 :
1543 : /**
1544 : * find_inode_nowait - find an inode in the inode cache
1545 : * @sb: super block of file system to search
1546 : * @hashval: hash value (usually inode number) to search for
1547 : * @match: callback used for comparisons between inodes
1548 : * @data: opaque data pointer to pass to @match
1549 : *
1550 : * Search for the inode specified by @hashval and @data in the inode
1551 : * cache, where the helper function @match will return 0 if the inode
1552 : * does not match, 1 if the inode does match, and -1 if the search
1553 : * should be stopped. The @match function must be responsible for
1554 : * taking the i_lock spin_lock and checking i_state for an inode being
1555 : * freed or being initialized, and incrementing the reference count
1556 : * before returning 1. It also must not sleep, since it is called with
1557 : * the inode_hash_lock spinlock held.
1558 : *
1559 : * This is a even more generalized version of ilookup5() when the
1560 : * function must never block --- find_inode() can block in
1561 : * __wait_on_freeing_inode() --- or when the caller can not increment
1562 : * the reference count because the resulting iput() might cause an
1563 : * inode eviction. The tradeoff is that the @match funtion must be
1564 : * very carefully implemented.
1565 : */
1566 0 : struct inode *find_inode_nowait(struct super_block *sb,
1567 : unsigned long hashval,
1568 : int (*match)(struct inode *, unsigned long,
1569 : void *),
1570 : void *data)
1571 : {
1572 0 : struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1573 0 : struct inode *inode, *ret_inode = NULL;
1574 0 : int mval;
1575 :
1576 0 : spin_lock(&inode_hash_lock);
1577 0 : hlist_for_each_entry(inode, head, i_hash) {
1578 0 : if (inode->i_sb != sb)
1579 0 : continue;
1580 0 : mval = match(inode, hashval, data);
1581 0 : if (mval == 0)
1582 0 : continue;
1583 0 : if (mval == 1)
1584 0 : ret_inode = inode;
1585 0 : goto out;
1586 : }
1587 0 : out:
1588 0 : spin_unlock(&inode_hash_lock);
1589 0 : return ret_inode;
1590 : }
1591 : EXPORT_SYMBOL(find_inode_nowait);
1592 :
1593 : /**
1594 : * find_inode_rcu - find an inode in the inode cache
1595 : * @sb: Super block of file system to search
1596 : * @hashval: Key to hash
1597 : * @test: Function to test match on an inode
1598 : * @data: Data for test function
1599 : *
1600 : * Search for the inode specified by @hashval and @data in the inode cache,
1601 : * where the helper function @test will return 0 if the inode does not match
1602 : * and 1 if it does. The @test function must be responsible for taking the
1603 : * i_lock spin_lock and checking i_state for an inode being freed or being
1604 : * initialized.
1605 : *
1606 : * If successful, this will return the inode for which the @test function
1607 : * returned 1 and NULL otherwise.
1608 : *
1609 : * The @test function is not permitted to take a ref on any inode presented.
1610 : * It is also not permitted to sleep.
1611 : *
1612 : * The caller must hold the RCU read lock.
1613 : */
1614 0 : struct inode *find_inode_rcu(struct super_block *sb, unsigned long hashval,
1615 : int (*test)(struct inode *, void *), void *data)
1616 : {
1617 0 : struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1618 0 : struct inode *inode;
1619 :
1620 0 : RCU_LOCKDEP_WARN(!rcu_read_lock_held(),
1621 : "suspicious find_inode_rcu() usage");
1622 :
1623 0 : hlist_for_each_entry_rcu(inode, head, i_hash) {
1624 0 : if (inode->i_sb == sb &&
1625 0 : !(READ_ONCE(inode->i_state) & (I_FREEING | I_WILL_FREE)) &&
1626 0 : test(inode, data))
1627 0 : return inode;
1628 : }
1629 : return NULL;
1630 : }
1631 : EXPORT_SYMBOL(find_inode_rcu);
1632 :
1633 : /**
1634 : * find_inode_by_ino_rcu - Find an inode in the inode cache
1635 : * @sb: Super block of file system to search
1636 : * @ino: The inode number to match
1637 : *
1638 : * Search for the inode specified by @hashval and @data in the inode cache,
1639 : * where the helper function @test will return 0 if the inode does not match
1640 : * and 1 if it does. The @test function must be responsible for taking the
1641 : * i_lock spin_lock and checking i_state for an inode being freed or being
1642 : * initialized.
1643 : *
1644 : * If successful, this will return the inode for which the @test function
1645 : * returned 1 and NULL otherwise.
1646 : *
1647 : * The @test function is not permitted to take a ref on any inode presented.
1648 : * It is also not permitted to sleep.
1649 : *
1650 : * The caller must hold the RCU read lock.
1651 : */
1652 375 : struct inode *find_inode_by_ino_rcu(struct super_block *sb,
1653 : unsigned long ino)
1654 : {
1655 375 : struct hlist_head *head = inode_hashtable + hash(sb, ino);
1656 375 : struct inode *inode;
1657 :
1658 375 : RCU_LOCKDEP_WARN(!rcu_read_lock_held(),
1659 : "suspicious find_inode_by_ino_rcu() usage");
1660 :
1661 754 : hlist_for_each_entry_rcu(inode, head, i_hash) {
1662 54 : if (inode->i_ino == ino &&
1663 50 : inode->i_sb == sb &&
1664 50 : !(READ_ONCE(inode->i_state) & (I_FREEING | I_WILL_FREE)))
1665 50 : return inode;
1666 : }
1667 : return NULL;
1668 : }
1669 : EXPORT_SYMBOL(find_inode_by_ino_rcu);
1670 :
1671 2780706 : int insert_inode_locked(struct inode *inode)
1672 : {
1673 2780706 : struct super_block *sb = inode->i_sb;
1674 2780706 : ino_t ino = inode->i_ino;
1675 2780706 : struct hlist_head *head = inode_hashtable + hash(sb, ino);
1676 :
1677 0 : while (1) {
1678 2774221 : struct inode *old = NULL;
1679 2774221 : spin_lock(&inode_hash_lock);
1680 5791333 : hlist_for_each_entry(old, head, i_hash) {
1681 211903 : if (old->i_ino != ino)
1682 211880 : continue;
1683 23 : if (old->i_sb != sb)
1684 0 : continue;
1685 23 : spin_lock(&old->i_lock);
1686 23 : if (old->i_state & (I_FREEING|I_WILL_FREE)) {
1687 23 : spin_unlock(&old->i_lock);
1688 23 : continue;
1689 : }
1690 : break;
1691 : }
1692 2789715 : if (likely(!old)) {
1693 2789715 : spin_lock(&inode->i_lock);
1694 2789715 : inode->i_state |= I_NEW | I_CREATING;
1695 2789715 : hlist_add_head_rcu(&inode->i_hash, head);
1696 2789715 : spin_unlock(&inode->i_lock);
1697 2789715 : spin_unlock(&inode_hash_lock);
1698 2789715 : return 0;
1699 : }
1700 0 : if (unlikely(old->i_state & I_CREATING)) {
1701 0 : spin_unlock(&old->i_lock);
1702 0 : spin_unlock(&inode_hash_lock);
1703 0 : return -EBUSY;
1704 : }
1705 0 : __iget(old);
1706 0 : spin_unlock(&old->i_lock);
1707 0 : spin_unlock(&inode_hash_lock);
1708 0 : wait_on_inode(old);
1709 0 : if (unlikely(!inode_unhashed(old))) {
1710 0 : iput(old);
1711 0 : return -EBUSY;
1712 : }
1713 0 : iput(old);
1714 : }
1715 : }
1716 : EXPORT_SYMBOL(insert_inode_locked);
1717 :
1718 3257597 : int insert_inode_locked4(struct inode *inode, unsigned long hashval,
1719 : int (*test)(struct inode *, void *), void *data)
1720 : {
1721 3257597 : struct inode *old;
1722 :
1723 3257597 : inode->i_state |= I_CREATING;
1724 3257597 : old = inode_insert5(inode, hashval, test, NULL, data);
1725 :
1726 3266654 : if (old != inode) {
1727 0 : iput(old);
1728 0 : return -EBUSY;
1729 : }
1730 : return 0;
1731 : }
1732 : EXPORT_SYMBOL(insert_inode_locked4);
1733 :
1734 :
1735 467735411 : int generic_delete_inode(struct inode *inode)
1736 : {
1737 467735411 : return 1;
1738 : }
1739 : EXPORT_SYMBOL(generic_delete_inode);
1740 :
1741 : /*
1742 : * Called when we're dropping the last reference
1743 : * to an inode.
1744 : *
1745 : * Call the FS "drop_inode()" function, defaulting to
1746 : * the legacy UNIX filesystem behaviour. If it tells
1747 : * us to evict inode, do so. Otherwise, retain inode
1748 : * in cache if fs is alive, sync and evict if fs is
1749 : * shutting down.
1750 : */
1751 67604083635 : static void iput_final(struct inode *inode)
1752 : {
1753 67604083635 : struct super_block *sb = inode->i_sb;
1754 67604083635 : const struct super_operations *op = inode->i_sb->s_op;
1755 67604083635 : unsigned long state;
1756 67604083635 : int drop;
1757 :
1758 67604083635 : WARN_ON(inode->i_state & I_NEW);
1759 :
1760 67604083635 : if (op->drop_inode)
1761 67568813191 : drop = op->drop_inode(inode);
1762 : else
1763 35270444 : drop = generic_drop_inode(inode);
1764 :
1765 66994487519 : if (!drop &&
1766 66468712801 : !(inode->i_state & I_DONTCACHE) &&
1767 65901992623 : (sb->s_flags & SB_ACTIVE)) {
1768 65904846385 : __inode_add_lru(inode, true);
1769 67143897342 : spin_unlock(&inode->i_lock);
1770 67143897342 : return;
1771 : }
1772 :
1773 1124911578 : state = inode->i_state;
1774 1124911578 : if (!drop) {
1775 567150543 : WRITE_ONCE(inode->i_state, state | I_WILL_FREE);
1776 567150543 : spin_unlock(&inode->i_lock);
1777 :
1778 567543207 : write_inode_now(inode, 1);
1779 :
1780 567449151 : spin_lock(&inode->i_lock);
1781 567502445 : state = inode->i_state;
1782 567502445 : WARN_ON(state & I_NEW);
1783 567502445 : state &= ~I_WILL_FREE;
1784 : }
1785 :
1786 1125263480 : WRITE_ONCE(inode->i_state, state | I_FREEING);
1787 1125263480 : if (!list_empty(&inode->i_lru))
1788 51011102 : inode_lru_list_del(inode);
1789 1125321663 : spin_unlock(&inode->i_lock);
1790 :
1791 1162969186 : evict(inode);
1792 : }
1793 :
1794 : /**
1795 : * iput - put an inode
1796 : * @inode: inode to put
1797 : *
1798 : * Puts an inode, dropping its usage count. If the inode use count hits
1799 : * zero, the inode is then freed and may also be destroyed.
1800 : *
1801 : * Consequently, iput() can sleep.
1802 : */
1803 >11373*10^7 : void iput(struct inode *inode)
1804 : {
1805 >11373*10^7 : if (!inode)
1806 : return;
1807 >11373*10^7 : BUG_ON(inode->i_state & I_CLEAR);
1808 >11373*10^7 : retry:
1809 >11373*10^7 : if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock)) {
1810 67815096978 : if (inode->i_nlink && (inode->i_state & I_DIRTY_TIME)) {
1811 24 : atomic_inc(&inode->i_count);
1812 24 : spin_unlock(&inode->i_lock);
1813 24 : trace_writeback_lazytime_iput(inode);
1814 24 : mark_inode_dirty_sync(inode);
1815 24 : goto retry;
1816 : }
1817 67815096954 : iput_final(inode);
1818 : }
1819 : }
1820 : EXPORT_SYMBOL(iput);
1821 :
1822 : #ifdef CONFIG_BLOCK
1823 : /**
1824 : * bmap - find a block number in a file
1825 : * @inode: inode owning the block number being requested
1826 : * @block: pointer containing the block to find
1827 : *
1828 : * Replaces the value in ``*block`` with the block number on the device holding
1829 : * corresponding to the requested block number in the file.
1830 : * That is, asked for block 4 of inode 1 the function will replace the
1831 : * 4 in ``*block``, with disk block relative to the disk start that holds that
1832 : * block of the file.
1833 : *
1834 : * Returns -EINVAL in case of error, 0 otherwise. If mapping falls into a
1835 : * hole, returns 0 and ``*block`` is also set to 0.
1836 : */
1837 33237 : int bmap(struct inode *inode, sector_t *block)
1838 : {
1839 33237 : if (!inode->i_mapping->a_ops->bmap)
1840 : return -EINVAL;
1841 :
1842 33235 : *block = inode->i_mapping->a_ops->bmap(inode->i_mapping, *block);
1843 33235 : return 0;
1844 : }
1845 : EXPORT_SYMBOL(bmap);
1846 : #endif
1847 :
1848 : /*
1849 : * With relative atime, only update atime if the previous atime is
1850 : * earlier than or equal to either the ctime or mtime,
1851 : * or if at least a day has passed since the last atime update.
1852 : */
1853 5407111724 : static int relatime_need_update(struct vfsmount *mnt, struct inode *inode,
1854 : struct timespec64 now)
1855 : {
1856 :
1857 5407111724 : if (!(mnt->mnt_flags & MNT_RELATIME))
1858 : return 1;
1859 : /*
1860 : * Is mtime younger than or equal to atime? If yes, update atime:
1861 : */
1862 2825049214 : if (timespec64_compare(&inode->i_mtime, &inode->i_atime) >= 0)
1863 : return 1;
1864 : /*
1865 : * Is ctime younger than or equal to atime? If yes, update atime:
1866 : */
1867 2487125594 : if (timespec64_compare(&inode->i_ctime, &inode->i_atime) >= 0)
1868 : return 1;
1869 :
1870 : /*
1871 : * Is the previous atime value older than a day? If yes,
1872 : * update atime:
1873 : */
1874 1428757898 : if ((long)(now.tv_sec - inode->i_atime.tv_sec) >= 24*60*60)
1875 2361 : return 1;
1876 : /*
1877 : * Good, we can skip the atime update:
1878 : */
1879 : return 0;
1880 : }
1881 :
1882 44077894 : int generic_update_time(struct inode *inode, struct timespec64 *time, int flags)
1883 : {
1884 44077894 : int dirty_flags = 0;
1885 :
1886 44077894 : if (flags & (S_ATIME | S_CTIME | S_MTIME)) {
1887 44076851 : if (flags & S_ATIME)
1888 21762000 : inode->i_atime = *time;
1889 44076851 : if (flags & S_CTIME)
1890 22309366 : inode->i_ctime = *time;
1891 44076851 : if (flags & S_MTIME)
1892 22313271 : inode->i_mtime = *time;
1893 :
1894 44076851 : if (inode->i_sb->s_flags & SB_LAZYTIME)
1895 : dirty_flags |= I_DIRTY_TIME;
1896 : else
1897 44071035 : dirty_flags |= I_DIRTY_SYNC;
1898 : }
1899 :
1900 44077894 : if ((flags & S_VERSION) && inode_maybe_inc_iversion(inode, false))
1901 1542 : dirty_flags |= I_DIRTY_SYNC;
1902 :
1903 44077899 : __mark_inode_dirty(inode, dirty_flags);
1904 44090760 : return 0;
1905 : }
1906 : EXPORT_SYMBOL(generic_update_time);
1907 :
1908 : /*
1909 : * This does the actual work of updating an inodes time or version. Must have
1910 : * had called mnt_want_write() before calling this.
1911 : */
1912 160574208 : int inode_update_time(struct inode *inode, struct timespec64 *time, int flags)
1913 : {
1914 160574208 : if (inode->i_op->update_time)
1915 116497770 : return inode->i_op->update_time(inode, time, flags);
1916 44076438 : return generic_update_time(inode, time, flags);
1917 : }
1918 : EXPORT_SYMBOL(inode_update_time);
1919 :
1920 : /**
1921 : * atime_needs_update - update the access time
1922 : * @path: the &struct path to update
1923 : * @inode: inode to update
1924 : *
1925 : * Update the accessed time on an inode and mark it for writeback.
1926 : * This function automatically handles read only file systems and media,
1927 : * as well as the "noatime" flag and inode specific "noatime" markers.
1928 : */
1929 6888396121 : bool atime_needs_update(const struct path *path, struct inode *inode)
1930 : {
1931 6888396121 : struct vfsmount *mnt = path->mnt;
1932 6888396121 : struct timespec64 now;
1933 :
1934 6888396121 : if (inode->i_flags & S_NOATIME)
1935 : return false;
1936 :
1937 : /* Atime updates will likely cause i_uid and i_gid to be written
1938 : * back improprely if their true value is unknown to the vfs.
1939 : */
1940 5402006813 : if (HAS_UNMAPPED_ID(mnt_idmap(mnt), inode))
1941 : return false;
1942 :
1943 5406293904 : if (IS_NOATIME(inode))
1944 : return false;
1945 5404778092 : if ((inode->i_sb->s_flags & SB_NODIRATIME) && S_ISDIR(inode->i_mode))
1946 : return false;
1947 :
1948 5404426160 : if (mnt->mnt_flags & MNT_NOATIME)
1949 : return false;
1950 5404424190 : if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode))
1951 : return false;
1952 :
1953 5404424190 : now = current_time(inode);
1954 :
1955 5406388819 : if (!relatime_need_update(mnt, inode, now))
1956 : return false;
1957 :
1958 3978001050 : if (timespec64_equal(&inode->i_atime, &now))
1959 3890942211 : return false;
1960 :
1961 : return true;
1962 : }
1963 :
1964 6517231549 : void touch_atime(const struct path *path)
1965 : {
1966 6517231549 : struct vfsmount *mnt = path->mnt;
1967 6517231549 : struct inode *inode = d_inode(path->dentry);
1968 6517231549 : struct timespec64 now;
1969 :
1970 6517231549 : if (!atime_needs_update(path, inode))
1971 6433098186 : return;
1972 :
1973 86519281 : if (!sb_start_write_trylock(inode->i_sb))
1974 : return;
1975 :
1976 86362423 : if (__mnt_want_write(mnt) != 0)
1977 94386 : goto skip_update;
1978 : /*
1979 : * File systems can error out when updating inodes if they need to
1980 : * allocate new space to modify an inode (such is the case for
1981 : * Btrfs), but since we touch atime while walking down the path we
1982 : * really don't care if we failed to update the atime of the file,
1983 : * so just ignore the return value.
1984 : * We may also fail on filesystems that have the ability to make parts
1985 : * of the fs read only, e.g. subvolumes in Btrfs.
1986 : */
1987 86274196 : now = current_time(inode);
1988 86276749 : inode_update_time(inode, &now, S_ATIME);
1989 86259131 : __mnt_drop_write(mnt);
1990 86336363 : skip_update:
1991 86336363 : sb_end_write(inode->i_sb);
1992 : }
1993 : EXPORT_SYMBOL(touch_atime);
1994 :
1995 : /*
1996 : * Return mask of changes for notify_change() that need to be done as a
1997 : * response to write or truncate. Return 0 if nothing has to be changed.
1998 : * Negative value on error (change should be denied).
1999 : */
2000 32004381 : int dentry_needs_remove_privs(struct mnt_idmap *idmap,
2001 : struct dentry *dentry)
2002 : {
2003 32004381 : struct inode *inode = d_inode(dentry);
2004 32004381 : int mask = 0;
2005 32004381 : int ret;
2006 :
2007 32004381 : if (IS_NOSEC(inode))
2008 : return 0;
2009 :
2010 23724162 : mask = setattr_should_drop_suidgid(idmap, inode);
2011 23707453 : ret = security_inode_need_killpriv(dentry);
2012 23777143 : if (ret < 0)
2013 : return ret;
2014 23777143 : if (ret)
2015 81 : mask |= ATTR_KILL_PRIV;
2016 : return mask;
2017 : }
2018 :
2019 : static int __remove_privs(struct mnt_idmap *idmap,
2020 : struct dentry *dentry, int kill)
2021 : {
2022 541 : struct iattr newattrs;
2023 :
2024 541 : newattrs.ia_valid = ATTR_FORCE | kill;
2025 : /*
2026 : * Note we call this on write, so notify_change will not
2027 : * encounter any conflicting delegations:
2028 : */
2029 541 : return notify_change(idmap, dentry, &newattrs, NULL);
2030 : }
2031 :
2032 656658463 : static int __file_remove_privs(struct file *file, unsigned int flags)
2033 : {
2034 656658463 : struct dentry *dentry = file_dentry(file);
2035 656330714 : struct inode *inode = file_inode(file);
2036 656330714 : int error = 0;
2037 656330714 : int kill;
2038 :
2039 656330714 : if (IS_NOSEC(inode) || !S_ISREG(inode->i_mode))
2040 : return 0;
2041 :
2042 22117957 : kill = dentry_needs_remove_privs(file_mnt_idmap(file), dentry);
2043 22105286 : if (kill < 0)
2044 : return kill;
2045 :
2046 22105286 : if (kill) {
2047 541 : if (flags & IOCB_NOWAIT)
2048 : return -EAGAIN;
2049 :
2050 541 : error = __remove_privs(file_mnt_idmap(file), dentry, kill);
2051 : }
2052 :
2053 541 : if (!error)
2054 22105286 : inode_has_no_xattr(inode);
2055 : return error;
2056 : }
2057 :
2058 : /**
2059 : * file_remove_privs - remove special file privileges (suid, capabilities)
2060 : * @file: file to remove privileges from
2061 : *
2062 : * When file is modified by a write or truncation ensure that special
2063 : * file privileges are removed.
2064 : *
2065 : * Return: 0 on success, negative errno on failure.
2066 : */
2067 302838932 : int file_remove_privs(struct file *file)
2068 : {
2069 302838932 : return __file_remove_privs(file, 0);
2070 : }
2071 : EXPORT_SYMBOL(file_remove_privs);
2072 :
2073 842878855 : static int inode_needs_update_time(struct inode *inode, struct timespec64 *now)
2074 : {
2075 842878855 : int sync_it = 0;
2076 :
2077 : /* First try to exhaust all avenues to not sync */
2078 842878855 : if (IS_NOCMTIME(inode))
2079 : return 0;
2080 :
2081 842878855 : if (!timespec64_equal(&inode->i_mtime, now))
2082 : sync_it = S_MTIME;
2083 :
2084 842878855 : if (!timespec64_equal(&inode->i_ctime, now))
2085 73583363 : sync_it |= S_CTIME;
2086 :
2087 842878855 : if (IS_I_VERSION(inode) && inode_iversion_need_inc(inode))
2088 322437 : sync_it |= S_VERSION;
2089 :
2090 : return sync_it;
2091 : }
2092 :
2093 74305824 : static int __file_update_time(struct file *file, struct timespec64 *now,
2094 : int sync_mode)
2095 : {
2096 74305824 : int ret = 0;
2097 74305824 : struct inode *inode = file_inode(file);
2098 :
2099 : /* try to update time settings */
2100 74305824 : if (!__mnt_want_write_file(file)) {
2101 74308624 : ret = inode_update_time(inode, now, sync_mode);
2102 74301759 : __mnt_drop_write_file(file);
2103 : }
2104 :
2105 74271690 : return ret;
2106 : }
2107 :
2108 : /**
2109 : * file_update_time - update mtime and ctime time
2110 : * @file: file accessed
2111 : *
2112 : * Update the mtime and ctime members of an inode and mark the inode for
2113 : * writeback. Note that this function is meant exclusively for usage in
2114 : * the file write path of filesystems, and filesystems may choose to
2115 : * explicitly ignore updates via this function with the _NOCMTIME inode
2116 : * flag, e.g. for network filesystem where these imestamps are handled
2117 : * by the server. This can return an error for file systems who need to
2118 : * allocate space in order to update an inode.
2119 : *
2120 : * Return: 0 on success, negative errno on failure.
2121 : */
2122 489405463 : int file_update_time(struct file *file)
2123 : {
2124 489405463 : int ret;
2125 489405463 : struct inode *inode = file_inode(file);
2126 489405463 : struct timespec64 now = current_time(inode);
2127 :
2128 489222620 : ret = inode_needs_update_time(inode, &now);
2129 489289698 : if (ret <= 0)
2130 : return ret;
2131 :
2132 24227232 : return __file_update_time(file, &now, ret);
2133 : }
2134 : EXPORT_SYMBOL(file_update_time);
2135 :
2136 : /**
2137 : * file_modified_flags - handle mandated vfs changes when modifying a file
2138 : * @file: file that was modified
2139 : * @flags: kiocb flags
2140 : *
2141 : * When file has been modified ensure that special
2142 : * file privileges are removed and time settings are updated.
2143 : *
2144 : * If IOCB_NOWAIT is set, special file privileges will not be removed and
2145 : * time settings will not be updated. It will return -EAGAIN.
2146 : *
2147 : * Context: Caller must hold the file's inode lock.
2148 : *
2149 : * Return: 0 on success, negative errno on failure.
2150 : */
2151 354038783 : static int file_modified_flags(struct file *file, int flags)
2152 : {
2153 354038783 : int ret;
2154 354038783 : struct inode *inode = file_inode(file);
2155 354038783 : struct timespec64 now = current_time(inode);
2156 :
2157 : /*
2158 : * Clear the security bits if the process is not being run by root.
2159 : * This keeps people from modifying setuid and setgid binaries.
2160 : */
2161 353891009 : ret = __file_remove_privs(file, flags);
2162 353605201 : if (ret)
2163 : return ret;
2164 :
2165 353605201 : if (unlikely(file->f_mode & FMODE_NOCMTIME))
2166 : return 0;
2167 :
2168 353605157 : ret = inode_needs_update_time(inode, &now);
2169 353712309 : if (ret <= 0)
2170 : return ret;
2171 50080169 : if (flags & IOCB_NOWAIT)
2172 : return -EAGAIN;
2173 :
2174 50080169 : return __file_update_time(file, &now, ret);
2175 : }
2176 :
2177 : /**
2178 : * file_modified - handle mandated vfs changes when modifying a file
2179 : * @file: file that was modified
2180 : *
2181 : * When file has been modified ensure that special
2182 : * file privileges are removed and time settings are updated.
2183 : *
2184 : * Context: Caller must hold the file's inode lock.
2185 : *
2186 : * Return: 0 on success, negative errno on failure.
2187 : */
2188 85264103 : int file_modified(struct file *file)
2189 : {
2190 85264103 : return file_modified_flags(file, 0);
2191 : }
2192 : EXPORT_SYMBOL(file_modified);
2193 :
2194 : /**
2195 : * kiocb_modified - handle mandated vfs changes when modifying a file
2196 : * @iocb: iocb that was modified
2197 : *
2198 : * When file has been modified ensure that special
2199 : * file privileges are removed and time settings are updated.
2200 : *
2201 : * Context: Caller must hold the file's inode lock.
2202 : *
2203 : * Return: 0 on success, negative errno on failure.
2204 : */
2205 268788815 : int kiocb_modified(struct kiocb *iocb)
2206 : {
2207 268788815 : return file_modified_flags(iocb->ki_filp, iocb->ki_flags);
2208 : }
2209 : EXPORT_SYMBOL_GPL(kiocb_modified);
2210 :
2211 664747 : int inode_needs_sync(struct inode *inode)
2212 : {
2213 664747 : if (IS_SYNC(inode))
2214 : return 1;
2215 664711 : if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
2216 0 : return 1;
2217 : return 0;
2218 : }
2219 : EXPORT_SYMBOL(inode_needs_sync);
2220 :
2221 : /*
2222 : * If we try to find an inode in the inode hash while it is being
2223 : * deleted, we have to wait until the filesystem completes its
2224 : * deletion before reporting that it isn't found. This function waits
2225 : * until the deletion _might_ have completed. Callers are responsible
2226 : * to recheck inode state.
2227 : *
2228 : * It doesn't matter if I_NEW is not set initially, a call to
2229 : * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
2230 : * will DTRT.
2231 : */
2232 2 : static void __wait_on_freeing_inode(struct inode *inode)
2233 : {
2234 2 : wait_queue_head_t *wq;
2235 2 : DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW);
2236 2 : wq = bit_waitqueue(&inode->i_state, __I_NEW);
2237 2 : prepare_to_wait(wq, &wait.wq_entry, TASK_UNINTERRUPTIBLE);
2238 2 : spin_unlock(&inode->i_lock);
2239 2 : spin_unlock(&inode_hash_lock);
2240 2 : schedule();
2241 2 : finish_wait(wq, &wait.wq_entry);
2242 2 : spin_lock(&inode_hash_lock);
2243 2 : }
2244 :
2245 : static __initdata unsigned long ihash_entries;
2246 0 : static int __init set_ihash_entries(char *str)
2247 : {
2248 0 : if (!str)
2249 : return 0;
2250 0 : ihash_entries = simple_strtoul(str, &str, 0);
2251 0 : return 1;
2252 : }
2253 : __setup("ihash_entries=", set_ihash_entries);
2254 :
2255 : /*
2256 : * Initialize the waitqueues and inode hash table.
2257 : */
2258 0 : void __init inode_init_early(void)
2259 : {
2260 : /* If hashes are distributed across NUMA nodes, defer
2261 : * hash allocation until vmalloc space is available.
2262 : */
2263 0 : if (hashdist)
2264 : return;
2265 :
2266 0 : inode_hashtable =
2267 0 : alloc_large_system_hash("Inode-cache",
2268 : sizeof(struct hlist_head),
2269 : ihash_entries,
2270 : 14,
2271 : HASH_EARLY | HASH_ZERO,
2272 : &i_hash_shift,
2273 : &i_hash_mask,
2274 : 0,
2275 : 0);
2276 : }
2277 :
2278 0 : void __init inode_init(void)
2279 : {
2280 : /* inode slab cache */
2281 0 : inode_cachep = kmem_cache_create("inode_cache",
2282 : sizeof(struct inode),
2283 : 0,
2284 : (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|
2285 : SLAB_MEM_SPREAD|SLAB_ACCOUNT),
2286 : init_once);
2287 :
2288 : /* Hash may have been set up in inode_init_early */
2289 0 : if (!hashdist)
2290 : return;
2291 :
2292 0 : inode_hashtable =
2293 0 : alloc_large_system_hash("Inode-cache",
2294 : sizeof(struct hlist_head),
2295 : ihash_entries,
2296 : 14,
2297 : HASH_ZERO,
2298 : &i_hash_shift,
2299 : &i_hash_mask,
2300 : 0,
2301 : 0);
2302 : }
2303 :
2304 355967397 : void init_special_inode(struct inode *inode, umode_t mode, dev_t rdev)
2305 : {
2306 355967397 : inode->i_mode = mode;
2307 355967397 : if (S_ISCHR(mode)) {
2308 355883718 : inode->i_fop = &def_chr_fops;
2309 355883718 : inode->i_rdev = rdev;
2310 83679 : } else if (S_ISBLK(mode)) {
2311 2886 : if (IS_ENABLED(CONFIG_BLOCK))
2312 2886 : inode->i_fop = &def_blk_fops;
2313 2886 : inode->i_rdev = rdev;
2314 80793 : } else if (S_ISFIFO(mode))
2315 74939 : inode->i_fop = &pipefifo_fops;
2316 5854 : else if (S_ISSOCK(mode))
2317 : ; /* leave it no_open_fops */
2318 : else
2319 0 : printk(KERN_DEBUG "init_special_inode: bogus i_mode (%o) for"
2320 : " inode %s:%lu\n", mode, inode->i_sb->s_id,
2321 : inode->i_ino);
2322 355967397 : }
2323 : EXPORT_SYMBOL(init_special_inode);
2324 :
2325 : /**
2326 : * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
2327 : * @idmap: idmap of the mount the inode was created from
2328 : * @inode: New inode
2329 : * @dir: Directory inode
2330 : * @mode: mode of the new inode
2331 : *
2332 : * If the inode has been created through an idmapped mount the idmap of
2333 : * the vfsmount must be passed through @idmap. This function will then take
2334 : * care to map the inode according to @idmap before checking permissions
2335 : * and initializing i_uid and i_gid. On non-idmapped mounts or if permission
2336 : * checking is to be performed on the raw inode simply pass @nop_mnt_idmap.
2337 : */
2338 578723619 : void inode_init_owner(struct mnt_idmap *idmap, struct inode *inode,
2339 : const struct inode *dir, umode_t mode)
2340 : {
2341 578723619 : inode_fsuid_set(inode, idmap);
2342 578437291 : if (dir && dir->i_mode & S_ISGID) {
2343 1233 : inode->i_gid = dir->i_gid;
2344 :
2345 : /* Directories are special, and always inherit S_ISGID */
2346 1233 : if (S_ISDIR(mode))
2347 237 : mode |= S_ISGID;
2348 : } else
2349 578436058 : inode_fsgid_set(inode, idmap);
2350 578239169 : inode->i_mode = mode;
2351 578239169 : }
2352 : EXPORT_SYMBOL(inode_init_owner);
2353 :
2354 : /**
2355 : * inode_owner_or_capable - check current task permissions to inode
2356 : * @idmap: idmap of the mount the inode was found from
2357 : * @inode: inode being checked
2358 : *
2359 : * Return true if current either has CAP_FOWNER in a namespace with the
2360 : * inode owner uid mapped, or owns the file.
2361 : *
2362 : * If the inode has been found through an idmapped mount the idmap of
2363 : * the vfsmount must be passed through @idmap. This function will then take
2364 : * care to map the inode according to @idmap before checking permissions.
2365 : * On non-idmapped mounts or if permission checking is to be performed on the
2366 : * raw inode simply passs @nop_mnt_idmap.
2367 : */
2368 73618032 : bool inode_owner_or_capable(struct mnt_idmap *idmap,
2369 : const struct inode *inode)
2370 : {
2371 73618032 : vfsuid_t vfsuid;
2372 73618032 : struct user_namespace *ns;
2373 :
2374 73618032 : vfsuid = i_uid_into_vfsuid(idmap, inode);
2375 147130942 : if (vfsuid_eq_kuid(vfsuid, current_fsuid()))
2376 : return true;
2377 :
2378 12228089 : ns = current_user_ns();
2379 12228089 : if (vfsuid_has_mapping(ns, vfsuid) && ns_capable(ns, CAP_FOWNER))
2380 12210284 : return true;
2381 : return false;
2382 : }
2383 : EXPORT_SYMBOL(inode_owner_or_capable);
2384 :
2385 : /*
2386 : * Direct i/o helper functions
2387 : */
2388 198821 : static void __inode_dio_wait(struct inode *inode)
2389 : {
2390 198821 : wait_queue_head_t *wq = bit_waitqueue(&inode->i_state, __I_DIO_WAKEUP);
2391 198714 : DEFINE_WAIT_BIT(q, &inode->i_state, __I_DIO_WAKEUP);
2392 :
2393 198716 : do {
2394 198716 : prepare_to_wait(wq, &q.wq_entry, TASK_UNINTERRUPTIBLE);
2395 198796 : if (atomic_read(&inode->i_dio_count))
2396 198797 : schedule();
2397 199341 : } while (atomic_read(&inode->i_dio_count));
2398 199339 : finish_wait(wq, &q.wq_entry);
2399 199331 : }
2400 :
2401 : /**
2402 : * inode_dio_wait - wait for outstanding DIO requests to finish
2403 : * @inode: inode to wait for
2404 : *
2405 : * Waits for all pending direct I/O requests to finish so that we can
2406 : * proceed with a truncate or equivalent operation.
2407 : *
2408 : * Must be called under a lock that serializes taking new references
2409 : * to i_dio_count, usually by inode->i_mutex.
2410 : */
2411 620903258 : void inode_dio_wait(struct inode *inode)
2412 : {
2413 620903258 : if (atomic_read(&inode->i_dio_count))
2414 198907 : __inode_dio_wait(inode);
2415 620903681 : }
2416 : EXPORT_SYMBOL(inode_dio_wait);
2417 :
2418 : /*
2419 : * inode_set_flags - atomically set some inode flags
2420 : *
2421 : * Note: the caller should be holding i_mutex, or else be sure that
2422 : * they have exclusive access to the inode structure (i.e., while the
2423 : * inode is being instantiated). The reason for the cmpxchg() loop
2424 : * --- which wouldn't be necessary if all code paths which modify
2425 : * i_flags actually followed this rule, is that there is at least one
2426 : * code path which doesn't today so we use cmpxchg() out of an abundance
2427 : * of caution.
2428 : *
2429 : * In the long run, i_mutex is overkill, and we should probably look
2430 : * at using the i_lock spinlock to protect i_flags, and then make sure
2431 : * it is so documented in include/linux/fs.h and that all code follows
2432 : * the locking convention!!
2433 : */
2434 3270602 : void inode_set_flags(struct inode *inode, unsigned int flags,
2435 : unsigned int mask)
2436 : {
2437 3270602 : WARN_ON_ONCE(flags & ~mask);
2438 3270602 : set_mask_bits(&inode->i_flags, mask, flags);
2439 3293118 : }
2440 : EXPORT_SYMBOL(inode_set_flags);
2441 :
2442 143552 : void inode_nohighmem(struct inode *inode)
2443 : {
2444 143552 : mapping_set_gfp_mask(inode->i_mapping, GFP_USER);
2445 143552 : }
2446 : EXPORT_SYMBOL(inode_nohighmem);
2447 :
2448 : /**
2449 : * timestamp_truncate - Truncate timespec to a granularity
2450 : * @t: Timespec
2451 : * @inode: inode being updated
2452 : *
2453 : * Truncate a timespec to the granularity supported by the fs
2454 : * containing the inode. Always rounds down. gran must
2455 : * not be 0 nor greater than a second (NSEC_PER_SEC, or 10^9 ns).
2456 : */
2457 8740370067 : struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode)
2458 : {
2459 8740370067 : struct super_block *sb = inode->i_sb;
2460 8740370067 : unsigned int gran = sb->s_time_gran;
2461 :
2462 8740370067 : t.tv_sec = clamp(t.tv_sec, sb->s_time_min, sb->s_time_max);
2463 8740370067 : if (unlikely(t.tv_sec == sb->s_time_max || t.tv_sec == sb->s_time_min))
2464 370 : t.tv_nsec = 0;
2465 :
2466 : /* Avoid division in the common cases 1 ns and 1 s. */
2467 8740370067 : if (gran == 1)
2468 : ; /* nothing */
2469 108672 : else if (gran == NSEC_PER_SEC)
2470 : t.tv_nsec = 0;
2471 0 : else if (gran > 1 && gran < NSEC_PER_SEC)
2472 0 : t.tv_nsec -= t.tv_nsec % gran;
2473 : else
2474 0 : WARN(1, "invalid file time granularity: %u", gran);
2475 8740370067 : return t;
2476 : }
2477 : EXPORT_SYMBOL(timestamp_truncate);
2478 :
2479 : /**
2480 : * current_time - Return FS time
2481 : * @inode: inode.
2482 : *
2483 : * Return the current time truncated to the time granularity supported by
2484 : * the fs.
2485 : *
2486 : * Note that inode and inode->sb cannot be NULL.
2487 : * Otherwise, the function warns and returns time without truncation.
2488 : */
2489 8712627949 : struct timespec64 current_time(struct inode *inode)
2490 : {
2491 8712627949 : struct timespec64 now;
2492 :
2493 8712627949 : ktime_get_coarse_real_ts64(&now);
2494 :
2495 8714267777 : if (unlikely(!inode->i_sb)) {
2496 0 : WARN(1, "current_time() called with uninitialized super_block in the inode");
2497 0 : return now;
2498 : }
2499 :
2500 8714267777 : return timestamp_truncate(now, inode);
2501 : }
2502 : EXPORT_SYMBOL(current_time);
2503 :
2504 : /**
2505 : * in_group_or_capable - check whether caller is CAP_FSETID privileged
2506 : * @idmap: idmap of the mount @inode was found from
2507 : * @inode: inode to check
2508 : * @vfsgid: the new/current vfsgid of @inode
2509 : *
2510 : * Check wether @vfsgid is in the caller's group list or if the caller is
2511 : * privileged with CAP_FSETID over @inode. This can be used to determine
2512 : * whether the setgid bit can be kept or must be dropped.
2513 : *
2514 : * Return: true if the caller is sufficiently privileged, false if not.
2515 : */
2516 3227079 : bool in_group_or_capable(struct mnt_idmap *idmap,
2517 : const struct inode *inode, vfsgid_t vfsgid)
2518 : {
2519 3227079 : if (vfsgid_in_group_p(vfsgid))
2520 : return true;
2521 19064 : if (capable_wrt_inode_uidgid(idmap, inode, CAP_FSETID))
2522 17724 : return true;
2523 : return false;
2524 : }
2525 :
2526 : /**
2527 : * mode_strip_sgid - handle the sgid bit for non-directories
2528 : * @idmap: idmap of the mount the inode was created from
2529 : * @dir: parent directory inode
2530 : * @mode: mode of the file to be created in @dir
2531 : *
2532 : * If the @mode of the new file has both the S_ISGID and S_IXGRP bit
2533 : * raised and @dir has the S_ISGID bit raised ensure that the caller is
2534 : * either in the group of the parent directory or they have CAP_FSETID
2535 : * in their user namespace and are privileged over the parent directory.
2536 : * In all other cases, strip the S_ISGID bit from @mode.
2537 : *
2538 : * Return: the new mode to use for the file
2539 : */
2540 81424120 : umode_t mode_strip_sgid(struct mnt_idmap *idmap,
2541 : const struct inode *dir, umode_t mode)
2542 : {
2543 81424120 : if ((mode & (S_ISGID | S_IXGRP)) != (S_ISGID | S_IXGRP))
2544 : return mode;
2545 977 : if (S_ISDIR(mode) || !dir || !(dir->i_mode & S_ISGID))
2546 : return mode;
2547 912 : if (in_group_or_capable(idmap, dir, i_gid_into_vfsgid(idmap, dir)))
2548 : return mode;
2549 688 : return mode & ~S_ISGID;
2550 : }
2551 : EXPORT_SYMBOL(mode_strip_sgid);
|
