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 
get_nr_inodes(void)76 static long get_nr_inodes(void)
77 {
78 	int i;
79 	long sum = 0;
80 	for_each_possible_cpu(i)
81 		sum += per_cpu(nr_inodes, i);
82 	return sum < 0 ? 0 : sum;
83 }
84 
get_nr_inodes_unused(void)85 static inline long get_nr_inodes_unused(void)
86 {
87 	int i;
88 	long sum = 0;
89 	for_each_possible_cpu(i)
90 		sum += per_cpu(nr_unused, i);
91 	return sum < 0 ? 0 : sum;
92 }
93 
get_nr_dirty_inodes(void)94 long get_nr_dirty_inodes(void)
95 {
96 	/* not actually dirty inodes, but a wild approximation */
97 	long nr_dirty = get_nr_inodes() - get_nr_inodes_unused();
98 	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 
proc_nr_inodes(struct ctl_table * table,int write,void * buffer,size_t * lenp,loff_t * ppos)110 static int proc_nr_inodes(struct ctl_table *table, int write, void *buffer,
111 			  size_t *lenp, loff_t *ppos)
112 {
113 	inodes_stat.nr_inodes = get_nr_inodes();
114 	inodes_stat.nr_unused = get_nr_inodes_unused();
115 	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 
init_fs_inode_sysctls(void)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);
142 #endif
143 
no_open(struct inode * inode,struct file * file)144 static int no_open(struct inode *inode, struct file *file)
145 {
146 	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  */
inode_init_always(struct super_block * sb,struct inode * inode)157 int inode_init_always(struct super_block *sb, struct inode *inode)
158 {
159 	static const struct inode_operations empty_iops;
160 	static const struct file_operations no_open_fops = {.open = no_open};
161 	struct address_space *const mapping = &inode->i_data;
162 
163 	inode->i_sb = sb;
164 	inode->i_blkbits = sb->s_blocksize_bits;
165 	inode->i_flags = 0;
166 	atomic64_set(&inode->i_sequence, 0);
167 	atomic_set(&inode->i_count, 1);
168 	inode->i_op = &empty_iops;
169 	inode->i_fop = &no_open_fops;
170 	inode->i_ino = 0;
171 	inode->__i_nlink = 1;
172 	inode->i_opflags = 0;
173 	if (sb->s_xattr)
174 		inode->i_opflags |= IOP_XATTR;
175 	i_uid_write(inode, 0);
176 	i_gid_write(inode, 0);
177 	atomic_set(&inode->i_writecount, 0);
178 	inode->i_size = 0;
179 	inode->i_write_hint = WRITE_LIFE_NOT_SET;
180 	inode->i_blocks = 0;
181 	inode->i_bytes = 0;
182 	inode->i_generation = 0;
183 	inode->i_pipe = NULL;
184 	inode->i_cdev = NULL;
185 	inode->i_link = NULL;
186 	inode->i_dir_seq = 0;
187 	inode->i_rdev = 0;
188 	inode->dirtied_when = 0;
189 
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 
196 	spin_lock_init(&inode->i_lock);
197 	lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key);
198 
199 	init_rwsem(&inode->i_rwsem);
200 	lockdep_set_class(&inode->i_rwsem, &sb->s_type->i_mutex_key);
201 
202 	atomic_set(&inode->i_dio_count, 0);
203 
204 	mapping->a_ops = &empty_aops;
205 	mapping->host = inode;
206 	mapping->flags = 0;
207 	mapping->wb_err = 0;
208 	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 	mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE);
213 	mapping->private_data = NULL;
214 	mapping->writeback_index = 0;
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");
219 	inode->i_private = NULL;
220 	inode->i_mapping = mapping;
221 	INIT_HLIST_HEAD(&inode->i_dentry);	/* buggered by rcu freeing */
222 #ifdef CONFIG_FS_POSIX_ACL
223 	inode->i_acl = inode->i_default_acl = ACL_NOT_CACHED;
224 #endif
225 
226 #ifdef CONFIG_FSNOTIFY
227 	inode->i_fsnotify_mask = 0;
228 #endif
229 	inode->i_flctx = NULL;
230 
231 	if (unlikely(security_inode_alloc(inode)))
232 		return -ENOMEM;
233 	this_cpu_inc(nr_inodes);
234 
235 	return 0;
236 }
237 EXPORT_SYMBOL(inode_init_always);
238 
free_inode_nonrcu(struct inode * inode)239 void free_inode_nonrcu(struct inode *inode)
240 {
241 	kmem_cache_free(inode_cachep, inode);
242 }
243 EXPORT_SYMBOL(free_inode_nonrcu);
244 
i_callback(struct rcu_head * head)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 
alloc_inode(struct super_block * sb)254 static struct inode *alloc_inode(struct super_block *sb)
255 {
256 	const struct super_operations *ops = sb->s_op;
257 	struct inode *inode;
258 
259 	if (ops->alloc_inode)
260 		inode = ops->alloc_inode(sb);
261 	else
262 		inode = alloc_inode_sb(sb, inode_cachep, GFP_KERNEL);
263 
264 	if (!inode)
265 		return NULL;
266 
267 	if (unlikely(inode_init_always(sb, inode))) {
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);
275 		return NULL;
276 	}
277 
278 	return inode;
279 }
280 
__destroy_inode(struct inode * inode)281 void __destroy_inode(struct inode *inode)
282 {
283 	BUG_ON(inode_has_buffers(inode));
284 	inode_detach_wb(inode);
285 	security_inode_free(inode);
286 	fsnotify_inode_delete(inode);
287 	locks_free_lock_context(inode);
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 
293 #ifdef CONFIG_FS_POSIX_ACL
294 	if (inode->i_acl && !is_uncached_acl(inode->i_acl))
295 		posix_acl_release(inode->i_acl);
296 	if (inode->i_default_acl && !is_uncached_acl(inode->i_default_acl))
297 		posix_acl_release(inode->i_default_acl);
298 #endif
299 	this_cpu_dec(nr_inodes);
300 }
301 EXPORT_SYMBOL(__destroy_inode);
302 
destroy_inode(struct inode * inode)303 static void destroy_inode(struct inode *inode)
304 {
305 	const struct super_operations *ops = inode->i_sb->s_op;
306 
307 	BUG_ON(!list_empty(&inode->i_lru));
308 	__destroy_inode(inode);
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);
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  */
drop_nlink(struct inode * inode)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  */
clear_nlink(struct inode * inode)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  */
set_nlink(struct inode * inode,unsigned int nlink)363 void set_nlink(struct inode *inode, unsigned int nlink)
364 {
365 	if (!nlink) {
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  */
inc_nlink(struct inode * inode)385 void inc_nlink(struct inode *inode)
386 {
387 	if (unlikely(inode->i_nlink == 0)) {
388 		WARN_ON(!(inode->i_state & I_LINKABLE));
389 		atomic_long_dec(&inode->i_sb->s_remove_count);
390 	}
391 
392 	inode->__i_nlink++;
393 }
394 EXPORT_SYMBOL(inc_nlink);
395 
__address_space_init_once(struct address_space * mapping)396 static void __address_space_init_once(struct address_space *mapping)
397 {
398 	xa_init_flags(&mapping->i_pages, XA_FLAGS_LOCK_IRQ | XA_FLAGS_ACCOUNT);
399 	init_rwsem(&mapping->i_mmap_rwsem);
400 	INIT_LIST_HEAD(&mapping->private_list);
401 	spin_lock_init(&mapping->private_lock);
402 	mapping->i_mmap = RB_ROOT_CACHED;
403 }
404 
address_space_init_once(struct address_space * mapping)405 void address_space_init_once(struct address_space *mapping)
406 {
407 	memset(mapping, 0, sizeof(*mapping));
408 	__address_space_init_once(mapping);
409 }
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  */
inode_init_once(struct inode * inode)417 void inode_init_once(struct inode *inode)
418 {
419 	memset(inode, 0, sizeof(*inode));
420 	INIT_HLIST_NODE(&inode->i_hash);
421 	INIT_LIST_HEAD(&inode->i_devices);
422 	INIT_LIST_HEAD(&inode->i_io_list);
423 	INIT_LIST_HEAD(&inode->i_wb_list);
424 	INIT_LIST_HEAD(&inode->i_lru);
425 	INIT_LIST_HEAD(&inode->i_sb_list);
426 	__address_space_init_once(&inode->i_data);
427 	i_size_ordered_init(inode);
428 }
429 EXPORT_SYMBOL(inode_init_once);
430 
init_once(void * foo)431 static void init_once(void *foo)
432 {
433 	struct inode *inode = (struct inode *) foo;
434 
435 	inode_init_once(inode);
436 }
437 
438 /*
439  * inode->i_lock must be held
440  */
__iget(struct inode * inode)441 void __iget(struct inode *inode)
442 {
443 	atomic_inc(&inode->i_count);
444 }
445 
446 /*
447  * get additional reference to inode; caller must already hold one.
448  */
ihold(struct inode * inode)449 void ihold(struct inode *inode)
450 {
451 	WARN_ON(atomic_inc_return(&inode->i_count) < 2);
452 }
453 EXPORT_SYMBOL(ihold);
454 
__inode_add_lru(struct inode * inode,bool rotate)455 static void __inode_add_lru(struct inode *inode, bool rotate)
456 {
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 
466 	if (list_lru_add(&inode->i_sb->s_inode_lru, &inode->i_lru))
467 		this_cpu_inc(nr_unused);
468 	else if (rotate)
469 		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  */
inode_add_lru(struct inode * inode)477 void inode_add_lru(struct inode *inode)
478 {
479 	__inode_add_lru(inode, false);
480 }
481 
inode_lru_list_del(struct inode * inode)482 static void inode_lru_list_del(struct inode *inode)
483 {
484 	if (list_lru_del(&inode->i_sb->s_inode_lru, &inode->i_lru))
485 		this_cpu_dec(nr_unused);
486 }
487 
488 /**
489  * inode_sb_list_add - add inode to the superblock list of inodes
490  * @inode: inode to add
491  */
inode_sb_list_add(struct inode * inode)492 void inode_sb_list_add(struct inode *inode)
493 {
494 	spin_lock(&inode->i_sb->s_inode_list_lock);
495 	list_add(&inode->i_sb_list, &inode->i_sb->s_inodes);
496 	spin_unlock(&inode->i_sb->s_inode_list_lock);
497 }
498 EXPORT_SYMBOL_GPL(inode_sb_list_add);
499 
inode_sb_list_del(struct inode * inode)500 static inline void inode_sb_list_del(struct inode *inode)
501 {
502 	if (!list_empty(&inode->i_sb_list)) {
503 		spin_lock(&inode->i_sb->s_inode_list_lock);
504 		list_del_init(&inode->i_sb_list);
505 		spin_unlock(&inode->i_sb->s_inode_list_lock);
506 	}
507 }
508 
hash(struct super_block * sb,unsigned long hashval)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;
515 	tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> i_hash_shift);
516 	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  */
__insert_inode_hash(struct inode * inode,unsigned long hashval)527 void __insert_inode_hash(struct inode *inode, unsigned long hashval)
528 {
529 	struct hlist_head *b = inode_hashtable + hash(inode->i_sb, hashval);
530 
531 	spin_lock(&inode_hash_lock);
532 	spin_lock(&inode->i_lock);
533 	hlist_add_head_rcu(&inode->i_hash, b);
534 	spin_unlock(&inode->i_lock);
535 	spin_unlock(&inode_hash_lock);
536 }
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  */
__remove_inode_hash(struct inode * inode)545 void __remove_inode_hash(struct inode *inode)
546 {
547 	spin_lock(&inode_hash_lock);
548 	spin_lock(&inode->i_lock);
549 	hlist_del_init_rcu(&inode->i_hash);
550 	spin_unlock(&inode->i_lock);
551 	spin_unlock(&inode_hash_lock);
552 }
553 EXPORT_SYMBOL(__remove_inode_hash);
554 
dump_mapping(const struct address_space * mapping)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 
clear_inode(struct inode * inode)604 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 	xa_lock_irq(&inode->i_data.i_pages);
612 	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 	xa_unlock_irq(&inode->i_data.i_pages);
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);
625 	BUG_ON(!list_empty(&inode->i_wb_list));
626 	/* don't need i_lock here, no concurrent mods to i_state */
627 	inode->i_state = I_FREEING | I_CLEAR;
628 }
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  */
evict(struct inode * inode)644 static void evict(struct inode *inode)
645 {
646 	const struct super_operations *op = inode->i_sb->s_op;
647 
648 	BUG_ON(!(inode->i_state & I_FREEING));
649 	BUG_ON(!list_empty(&inode->i_lru));
650 
651 	if (!list_empty(&inode->i_io_list))
652 		inode_io_list_del(inode);
653 
654 	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 	inode_wait_for_writeback(inode);
663 
664 	if (op->evict_inode) {
665 		op->evict_inode(inode);
666 	} else {
667 		truncate_inode_pages_final(&inode->i_data);
668 		clear_inode(inode);
669 	}
670 	if (S_ISCHR(inode->i_mode) && inode->i_cdev)
671 		cd_forget(inode);
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);
681 }
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  */
dispose_list(struct list_head * head)690 static void dispose_list(struct list_head *head)
691 {
692 	while (!list_empty(head)) {
693 		struct inode *inode;
694 
695 		inode = list_first_entry(head, struct inode, i_lru);
696 		list_del_init(&inode->i_lru);
697 
698 		evict(inode);
699 		cond_resched();
700 	}
701 }
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  */
evict_inodes(struct super_block * sb)712 void evict_inodes(struct super_block *sb)
713 {
714 	struct inode *inode, *next;
715 	LIST_HEAD(dispose);
716 
717 again:
718 	spin_lock(&sb->s_inode_list_lock);
719 	list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
720 		if (atomic_read(&inode->i_count))
721 			continue;
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);
726 			continue;
727 		}
728 
729 		inode->i_state |= I_FREEING;
730 		inode_lru_list_del(inode);
731 		spin_unlock(&inode->i_lock);
732 		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 		if (need_resched()) {
740 			spin_unlock(&sb->s_inode_list_lock);
741 			cond_resched();
742 			dispose_list(&dispose);
743 			goto again;
744 		}
745 	}
746 	spin_unlock(&sb->s_inode_list_lock);
747 
748 	dispose_list(&dispose);
749 }
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  */
invalidate_inodes(struct super_block * sb,bool kill_dirty)762 int invalidate_inodes(struct super_block *sb, bool kill_dirty)
763 {
764 	int busy = 0;
765 	struct inode *inode, *next;
766 	LIST_HEAD(dispose);
767 
768 again:
769 	spin_lock(&sb->s_inode_list_lock);
770 	list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
771 		spin_lock(&inode->i_lock);
772 		if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
773 			spin_unlock(&inode->i_lock);
774 			continue;
775 		}
776 		if (inode->i_state & I_DIRTY_ALL && !kill_dirty) {
777 			spin_unlock(&inode->i_lock);
778 			busy = 1;
779 			continue;
780 		}
781 		if (atomic_read(&inode->i_count)) {
782 			spin_unlock(&inode->i_lock);
783 			busy = 1;
784 			continue;
785 		}
786 
787 		inode->i_state |= I_FREEING;
788 		inode_lru_list_del(inode);
789 		spin_unlock(&inode->i_lock);
790 		list_add(&inode->i_lru, &dispose);
791 		if (need_resched()) {
792 			spin_unlock(&sb->s_inode_list_lock);
793 			cond_resched();
794 			dispose_list(&dispose);
795 			goto again;
796 		}
797 	}
798 	spin_unlock(&sb->s_inode_list_lock);
799 
800 	dispose_list(&dispose);
801 
802 	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  */
inode_lru_isolate(struct list_head * item,struct list_lru_one * lru,spinlock_t * lru_lock,void * arg)816 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 	struct list_head *freeable = arg;
820 	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 	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 	if (atomic_read(&inode->i_count) ||
836 	    (inode->i_state & ~I_REFERENCED) ||
837 	    !mapping_shrinkable(&inode->i_data)) {
838 		list_lru_isolate(lru, &inode->i_lru);
839 		spin_unlock(&inode->i_lock);
840 		this_cpu_dec(nr_unused);
841 		return LRU_REMOVED;
842 	}
843 
844 	/* Recently referenced inodes get one more pass */
845 	if (inode->i_state & I_REFERENCED) {
846 		inode->i_state &= ~I_REFERENCED;
847 		spin_unlock(&inode->i_lock);
848 		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 	if (inode_has_buffers(inode) || !mapping_empty(&inode->i_data)) {
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;
869 		}
870 		iput(inode);
871 		spin_lock(lru_lock);
872 		return LRU_RETRY;
873 	}
874 
875 	WARN_ON(inode->i_state & I_NEW);
876 	inode->i_state |= I_FREEING;
877 	list_lru_isolate_move(lru, &inode->i_lru, freeable);
878 	spin_unlock(&inode->i_lock);
879 
880 	this_cpu_dec(nr_unused);
881 	return LRU_REMOVED;
882 }
883 
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  */
prune_icache_sb(struct super_block * sb,struct shrink_control * sc)890 long prune_icache_sb(struct super_block *sb, struct shrink_control *sc)
891 {
892 	LIST_HEAD(freeable);
893 	long freed;
894 
895 	freed = list_lru_shrink_walk(&sb->s_inode_lru, sc,
896 				     inode_lru_isolate, &freeable);
897 	dispose_list(&freeable);
898 	return freed;
899 }
900 
901 static void __wait_on_freeing_inode(struct inode *inode);
902 /*
903  * Called with the inode lock held.
904  */
find_inode(struct super_block * sb,struct hlist_head * head,int (* test)(struct inode *,void *),void * data)905 static struct inode *find_inode(struct super_block *sb,
906 				struct hlist_head *head,
907 				int (*test)(struct inode *, void *),
908 				void *data)
909 {
910 	struct inode *inode = NULL;
911 
912 repeat:
913 	hlist_for_each_entry(inode, head, i_hash) {
914 		if (inode->i_sb != sb)
915 			continue;
916 		if (!test(inode, data))
917 			continue;
918 		spin_lock(&inode->i_lock);
919 		if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
920 			__wait_on_freeing_inode(inode);
921 			goto repeat;
922 		}
923 		if (unlikely(inode->i_state & I_CREATING)) {
924 			spin_unlock(&inode->i_lock);
925 			return ERR_PTR(-ESTALE);
926 		}
927 		__iget(inode);
928 		spin_unlock(&inode->i_lock);
929 		return inode;
930 	}
931 	return NULL;
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  */
find_inode_fast(struct super_block * sb,struct hlist_head * head,unsigned long ino)938 static struct inode *find_inode_fast(struct super_block *sb,
939 				struct hlist_head *head, unsigned long ino)
940 {
941 	struct inode *inode = NULL;
942 
943 repeat:
944 	hlist_for_each_entry(inode, head, i_hash) {
945 		if (inode->i_ino != ino)
946 			continue;
947 		if (inode->i_sb != sb)
948 			continue;
949 		spin_lock(&inode->i_lock);
950 		if (inode->i_state & (I_FREEING|I_WILL_FREE)) {
951 			__wait_on_freeing_inode(inode);
952 			goto repeat;
953 		}
954 		if (unlikely(inode->i_state & I_CREATING)) {
955 			spin_unlock(&inode->i_lock);
956 			return ERR_PTR(-ESTALE);
957 		}
958 		__iget(inode);
959 		spin_unlock(&inode->i_lock);
960 		return inode;
961 	}
962 	return NULL;
963 }
964 
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.
969  *
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.
979  */
980 #define LAST_INO_BATCH 1024
981 static DEFINE_PER_CPU(unsigned int, last_ino);
982 
get_next_ino(void)983 unsigned int get_next_ino(void)
984 {
985 	unsigned int *p = &get_cpu_var(last_ino);
986 	unsigned int res = *p;
987 
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 
997 	res++;
998 	/* get_next_ino should not provide a 0 inode number */
999 	if (unlikely(!res))
1000 		res++;
1001 	*p = res;
1002 	put_cpu_var(last_ino);
1003 	return res;
1004 }
1005 EXPORT_SYMBOL(get_next_ino);
1006 
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  */
new_inode_pseudo(struct super_block * sb)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);
1025 	}
1026 	return inode;
1027 }
1028 
1029 /**
1030  *	new_inode 	- obtain an inode
1031  *	@sb: superblock
1032  *
1033  *	Allocates a new inode for given superblock. The default gfp_mask
1034  *	for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
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  *
1040  */
new_inode(struct super_block * sb)1041 struct inode *new_inode(struct super_block *sb)
1042 {
1043 	struct inode *inode;
1044 
1045 	spin_lock_prefetch(&sb->s_inode_list_lock);
1046 
1047 	inode = new_inode_pseudo(sb);
1048 	if (inode)
1049 		inode_sb_list_add(inode);
1050 	return inode;
1051 }
1052 EXPORT_SYMBOL(new_inode);
1053 
1054 #ifdef CONFIG_DEBUG_LOCK_ALLOC
lockdep_annotate_inode_mutex_key(struct inode * inode)1055 void lockdep_annotate_inode_mutex_key(struct inode *inode)
1056 {
1057 	if (S_ISDIR(inode->i_mode)) {
1058 		struct file_system_type *type = inode->i_sb->s_type;
1059 
1060 		/* Set new key only if filesystem hasn't already changed it */
1061 		if (lockdep_match_class(&inode->i_rwsem, &type->i_mutex_key)) {
1062 			/*
1063 			 * ensure nobody is actually holding i_mutex
1064 			 */
1065 			// mutex_destroy(&inode->i_mutex);
1066 			init_rwsem(&inode->i_rwsem);
1067 			lockdep_set_class(&inode->i_rwsem,
1068 					  &type->i_mutex_dir_key);
1069 		}
1070 	}
1071 }
1072 EXPORT_SYMBOL(lockdep_annotate_inode_mutex_key);
1073 #endif
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  */
unlock_new_inode(struct inode * inode)1082 void unlock_new_inode(struct inode *inode)
1083 {
1084 	lockdep_annotate_inode_mutex_key(inode);
1085 	spin_lock(&inode->i_lock);
1086 	WARN_ON(!(inode->i_state & I_NEW));
1087 	inode->i_state &= ~I_NEW & ~I_CREATING;
1088 	smp_mb();
1089 	wake_up_bit(&inode->i_state, __I_NEW);
1090 	spin_unlock(&inode->i_lock);
1091 }
1092 EXPORT_SYMBOL(unlock_new_inode);
1093 
discard_new_inode(struct inode * inode)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 
1107 /**
1108  * lock_two_nondirectories - take two i_mutexes on non-directory objects
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  *
1113  * @inode1: first inode to lock
1114  * @inode2: second inode to lock
1115  */
lock_two_nondirectories(struct inode * inode1,struct inode * inode2)1116 void lock_two_nondirectories(struct inode *inode1, struct inode *inode2)
1117 {
1118 	if (inode1 > inode2)
1119 		swap(inode1, inode2);
1120 
1121 	if (inode1 && !S_ISDIR(inode1->i_mode))
1122 		inode_lock(inode1);
1123 	if (inode2 && !S_ISDIR(inode2->i_mode) && inode2 != inode1)
1124 		inode_lock_nested(inode2, I_MUTEX_NONDIR2);
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  */
unlock_two_nondirectories(struct inode * inode1,struct inode * inode2)1133 void unlock_two_nondirectories(struct inode *inode1, struct inode *inode2)
1134 {
1135 	if (inode1 && !S_ISDIR(inode1->i_mode))
1136 		inode_unlock(inode1);
1137 	if (inode2 && !S_ISDIR(inode2->i_mode) && inode2 != inode1)
1138 		inode_unlock(inode2);
1139 }
1140 EXPORT_SYMBOL(unlock_two_nondirectories);
1141 
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  */
inode_insert5(struct inode * inode,unsigned long hashval,int (* test)(struct inode *,void *),int (* set)(struct inode *,void *),void * data)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);
1178 		if (IS_ERR(old))
1179 			return NULL;
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;
1199 	hlist_add_head_rcu(&inode->i_hash, head);
1200 	spin_unlock(&inode->i_lock);
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))
1207 		inode_sb_list_add(inode);
1208 unlock:
1209 	spin_unlock(&inode_hash_lock);
1210 
1211 	return inode;
1212 }
1213 EXPORT_SYMBOL(inode_insert5);
1214 
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().
1231  *
1232  * Note both @test and @set are called with the inode_hash_lock held, so can't
1233  * sleep.
1234  */
iget5_locked(struct super_block * sb,unsigned long hashval,int (* test)(struct inode *,void *),int (* set)(struct inode *,void *),void * data)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)
1238 {
1239 	struct inode *inode = ilookup5(sb, hashval, test, data);
1240 
1241 	if (!inode) {
1242 		struct inode *new = alloc_inode(sb);
1243 
1244 		if (new) {
1245 			new->i_state = 0;
1246 			inode = inode_insert5(new, hashval, test, set, data);
1247 			if (unlikely(inode != new))
1248 				destroy_inode(new);
1249 		}
1250 	}
1251 	return inode;
1252 }
1253 EXPORT_SYMBOL(iget5_locked);
1254 
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().
1267  */
iget_locked(struct super_block * sb,unsigned long ino)1268 struct inode *iget_locked(struct super_block *sb, unsigned long ino)
1269 {
1270 	struct hlist_head *head = inode_hashtable + hash(sb, ino);
1271 	struct inode *inode;
1272 again:
1273 	spin_lock(&inode_hash_lock);
1274 	inode = find_inode_fast(sb, head, ino);
1275 	spin_unlock(&inode_hash_lock);
1276 	if (inode) {
1277 		if (IS_ERR(inode))
1278 			return NULL;
1279 		wait_on_inode(inode);
1280 		if (unlikely(inode_unhashed(inode))) {
1281 			iput(inode);
1282 			goto again;
1283 		}
1284 		return inode;
1285 	}
1286 
1287 	inode = alloc_inode(sb);
1288 	if (inode) {
1289 		struct inode *old;
1290 
1291 		spin_lock(&inode_hash_lock);
1292 		/* We released the lock, so.. */
1293 		old = find_inode_fast(sb, head, ino);
1294 		if (!old) {
1295 			inode->i_ino = ino;
1296 			spin_lock(&inode->i_lock);
1297 			inode->i_state = I_NEW;
1298 			hlist_add_head_rcu(&inode->i_hash, head);
1299 			spin_unlock(&inode->i_lock);
1300 			inode_sb_list_add(inode);
1301 			spin_unlock(&inode_hash_lock);
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 		 */
1314 		spin_unlock(&inode_hash_lock);
1315 		destroy_inode(inode);
1316 		if (IS_ERR(old))
1317 			return NULL;
1318 		inode = old;
1319 		wait_on_inode(inode);
1320 		if (unlikely(inode_unhashed(inode))) {
1321 			iput(inode);
1322 			goto again;
1323 		}
1324 	}
1325 	return inode;
1326 }
1327 EXPORT_SYMBOL(iget_locked);
1328 
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  */
test_inode_iunique(struct super_block * sb,unsigned long ino)1336 static int test_inode_iunique(struct super_block *sb, unsigned long ino)
1337 {
1338 	struct hlist_head *b = inode_hashtable + hash(sb, ino);
1339 	struct inode *inode;
1340 
1341 	hlist_for_each_entry_rcu(inode, b, i_hash) {
1342 		if (inode->i_ino == ino && inode->i_sb == sb)
1343 			return 0;
1344 	}
1345 	return 1;
1346 }
1347 
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  */
iunique(struct super_block * sb,ino_t max_reserved)1362 ino_t iunique(struct super_block *sb, ino_t max_reserved)
1363 {
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 	 */
1369 	static DEFINE_SPINLOCK(iunique_lock);
1370 	static unsigned int counter;
1371 	ino_t res;
1372 
1373 	rcu_read_lock();
1374 	spin_lock(&iunique_lock);
1375 	do {
1376 		if (counter <= max_reserved)
1377 			counter = max_reserved + 1;
1378 		res = counter++;
1379 	} while (!test_inode_iunique(sb, res));
1380 	spin_unlock(&iunique_lock);
1381 	rcu_read_unlock();
1382 
1383 	return res;
1384 }
1385 EXPORT_SYMBOL(iunique);
1386 
igrab(struct inode * inode)1387 struct inode *igrab(struct inode *inode)
1388 {
1389 	spin_lock(&inode->i_lock);
1390 	if (!(inode->i_state & (I_FREEING|I_WILL_FREE))) {
1391 		__iget(inode);
1392 		spin_unlock(&inode->i_lock);
1393 	} else {
1394 		spin_unlock(&inode->i_lock);
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;
1401 	}
1402 	return inode;
1403 }
1404 EXPORT_SYMBOL(igrab);
1405 
1406 /**
1407  * ilookup5_nowait - search for an inode in the inode cache
1408  * @sb:		super block of file system to search
1409  * @hashval:	hash value (usually inode number) to search for
1410  * @test:	callback used for comparisons between inodes
1411  * @data:	opaque data pointer to pass to @test
1412  *
1413  * Search for the inode specified by @hashval and @data in the inode cache.
1414  * If the inode is in the cache, the inode is returned with an incremented
1415  * reference count.
1416  *
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.
1419  *
1420  * Note2: @test is called with the inode_hash_lock held, so can't sleep.
1421  */
ilookup5_nowait(struct super_block * sb,unsigned long hashval,int (* test)(struct inode *,void *),void * data)1422 struct inode *ilookup5_nowait(struct super_block *sb, unsigned long hashval,
1423 		int (*test)(struct inode *, void *), void *data)
1424 {
1425 	struct hlist_head *head = inode_hashtable + hash(sb, hashval);
1426 	struct inode *inode;
1427 
1428 	spin_lock(&inode_hash_lock);
1429 	inode = find_inode(sb, head, test, data);
1430 	spin_unlock(&inode_hash_lock);
1431 
1432 	return IS_ERR(inode) ? NULL : inode;
1433 }
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  *
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.
1446  * returned with an incremented reference count.
1447  *
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.
1450  *
1451  * Note: @test is called with the inode_hash_lock held, so can't sleep.
1452  */
ilookup5(struct super_block * sb,unsigned long hashval,int (* test)(struct inode *,void *),void * data)1453 struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
1454 		int (*test)(struct inode *, void *), void *data)
1455 {
1456 	struct inode *inode;
1457 again:
1458 	inode = ilookup5_nowait(sb, hashval, test, data);
1459 	if (inode) {
1460 		wait_on_inode(inode);
1461 		if (unlikely(inode_unhashed(inode))) {
1462 			iput(inode);
1463 			goto again;
1464 		}
1465 	}
1466 	return inode;
1467 }
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  *
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.
1477  */
ilookup(struct super_block * sb,unsigned long ino)1478 struct inode *ilookup(struct super_block *sb, unsigned long ino)
1479 {
1480 	struct hlist_head *head = inode_hashtable + hash(sb, ino);
1481 	struct inode *inode;
1482 again:
1483 	spin_lock(&inode_hash_lock);
1484 	inode = find_inode_fast(sb, head, ino);
1485 	spin_unlock(&inode_hash_lock);
1486 
1487 	if (inode) {
1488 		if (IS_ERR(inode))
1489 			return NULL;
1490 		wait_on_inode(inode);
1491 		if (unlikely(inode_unhashed(inode))) {
1492 			iput(inode);
1493 			goto again;
1494 		}
1495 	}
1496 	return inode;
1497 }
1498 EXPORT_SYMBOL(ilookup);
1499 
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  */
find_inode_nowait(struct super_block * sb,unsigned long hashval,int (* match)(struct inode *,unsigned long,void *),void * data)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 
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  */
find_inode_rcu(struct super_block * sb,unsigned long hashval,int (* test)(struct inode *,void *),void * data)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 /**
1591  * find_inode_by_ino_rcu - Find an inode in the inode cache
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  */
find_inode_by_ino_rcu(struct super_block * sb,unsigned long ino)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 
insert_inode_locked(struct inode * inode)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);
1633 
1634 	while (1) {
1635 		struct inode *old = NULL;
1636 		spin_lock(&inode_hash_lock);
1637 		hlist_for_each_entry(old, head, i_hash) {
1638 			if (old->i_ino != ino)
1639 				continue;
1640 			if (old->i_sb != sb)
1641 				continue;
1642 			spin_lock(&old->i_lock);
1643 			if (old->i_state & (I_FREEING|I_WILL_FREE)) {
1644 				spin_unlock(&old->i_lock);
1645 				continue;
1646 			}
1647 			break;
1648 		}
1649 		if (likely(!old)) {
1650 			spin_lock(&inode->i_lock);
1651 			inode->i_state |= I_NEW | I_CREATING;
1652 			hlist_add_head_rcu(&inode->i_hash, head);
1653 			spin_unlock(&inode->i_lock);
1654 			spin_unlock(&inode_hash_lock);
1655 			return 0;
1656 		}
1657 		if (unlikely(old->i_state & I_CREATING)) {
1658 			spin_unlock(&old->i_lock);
1659 			spin_unlock(&inode_hash_lock);
1660 			return -EBUSY;
1661 		}
1662 		__iget(old);
1663 		spin_unlock(&old->i_lock);
1664 		spin_unlock(&inode_hash_lock);
1665 		wait_on_inode(old);
1666 		if (unlikely(!inode_unhashed(old))) {
1667 			iput(old);
1668 			return -EBUSY;
1669 		}
1670 		iput(old);
1671 	}
1672 }
1673 EXPORT_SYMBOL(insert_inode_locked);
1674 
insert_inode_locked4(struct inode * inode,unsigned long hashval,int (* test)(struct inode *,void *),void * data)1675 int insert_inode_locked4(struct inode *inode, unsigned long hashval,
1676 		int (*test)(struct inode *, void *), void *data)
1677 {
1678 	struct inode *old;
1679 
1680 	inode->i_state |= I_CREATING;
1681 	old = inode_insert5(inode, hashval, test, NULL, data);
1682 
1683 	if (old != inode) {
1684 		iput(old);
1685 		return -EBUSY;
1686 	}
1687 	return 0;
1688 }
1689 EXPORT_SYMBOL(insert_inode_locked4);
1690 
1691 
generic_delete_inode(struct inode * inode)1692 int generic_delete_inode(struct inode *inode)
1693 {
1694 	return 1;
1695 }
1696 EXPORT_SYMBOL(generic_delete_inode);
1697 
1698 /*
1699  * Called when we're dropping the last reference
1700  * to an inode.
1701  *
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.
1707  */
iput_final(struct inode * inode)1708 static void iput_final(struct inode *inode)
1709 {
1710 	struct super_block *sb = inode->i_sb;
1711 	const struct super_operations *op = inode->i_sb->s_op;
1712 	unsigned long state;
1713 	int drop;
1714 
1715 	WARN_ON(inode->i_state & I_NEW);
1716 
1717 	if (op->drop_inode)
1718 		drop = op->drop_inode(inode);
1719 	else
1720 		drop = generic_drop_inode(inode);
1721 
1722 	if (!drop &&
1723 	    !(inode->i_state & I_DONTCACHE) &&
1724 	    (sb->s_flags & SB_ACTIVE)) {
1725 		__inode_add_lru(inode, true);
1726 		spin_unlock(&inode->i_lock);
1727 		return;
1728 	}
1729 
1730 	state = inode->i_state;
1731 	if (!drop) {
1732 		WRITE_ONCE(inode->i_state, state | I_WILL_FREE);
1733 		spin_unlock(&inode->i_lock);
1734 
1735 		write_inode_now(inode, 1);
1736 
1737 		spin_lock(&inode->i_lock);
1738 		state = inode->i_state;
1739 		WARN_ON(state & I_NEW);
1740 		state &= ~I_WILL_FREE;
1741 	}
1742 
1743 	WRITE_ONCE(inode->i_state, state | I_FREEING);
1744 	if (!list_empty(&inode->i_lru))
1745 		inode_lru_list_del(inode);
1746 	spin_unlock(&inode->i_lock);
1747 
1748 	evict(inode);
1749 }
1750 
1751 /**
1752  *	iput	- put an inode
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  */
iput(struct inode * inode)1760 void iput(struct inode *inode)
1761 {
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);
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);
1775 	}
1776 }
1777 EXPORT_SYMBOL(iput);
1778 
1779 #ifdef CONFIG_BLOCK
1780 /**
1781  *	bmap	- find a block number in a file
1782  *	@inode:  inode owning the block number being requested
1783  *	@block: pointer containing the block to find
1784  *
1785  *	Replaces the value in ``*block`` with the block number on the device holding
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
1788  *	4 in ``*block``, with disk block relative to the disk start that holds that
1789  *	block of the file.
1790  *
1791  *	Returns -EINVAL in case of error, 0 otherwise. If mapping falls into a
1792  *	hole, returns 0 and ``*block`` is also set to 0.
1793  */
bmap(struct inode * inode,sector_t * block)1794 int bmap(struct inode *inode, sector_t *block)
1795 {
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;
1801 }
1802 EXPORT_SYMBOL(bmap);
1803 #endif
1804 
1805 /*
1806  * With relative atime, only update atime if the previous atime is
1807  * earlier than either the ctime or mtime or if at least a day has
1808  * passed since the last atime update.
1809  */
relatime_need_update(struct vfsmount * mnt,struct inode * inode,struct timespec64 now)1810 static int relatime_need_update(struct vfsmount *mnt, struct inode *inode,
1811 			     struct timespec64 now)
1812 {
1813 
1814 	if (!(mnt->mnt_flags & MNT_RELATIME))
1815 		return 1;
1816 	/*
1817 	 * Is mtime younger than atime? If yes, update atime:
1818 	 */
1819 	if (timespec64_compare(&inode->i_mtime, &inode->i_atime) >= 0)
1820 		return 1;
1821 	/*
1822 	 * Is ctime younger than atime? If yes, update atime:
1823 	 */
1824 	if (timespec64_compare(&inode->i_ctime, &inode->i_atime) >= 0)
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 
generic_update_time(struct inode * inode,struct timespec64 * time,int flags)1839 int generic_update_time(struct inode *inode, struct timespec64 *time, int flags)
1840 {
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);
1861 	return 0;
1862 }
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  */
inode_update_time(struct inode * inode,struct timespec64 * time,int flags)1869 int inode_update_time(struct inode *inode, struct timespec64 *time, int flags)
1870 {
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);
1874 }
1875 EXPORT_SYMBOL(inode_update_time);
1876 
1877 /**
1878  *	atime_needs_update	-	update the access time
1879  *	@path: the &struct path to update
1880  *	@inode: inode to update
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  */
atime_needs_update(const struct path * path,struct inode * inode)1886 bool atime_needs_update(const struct path *path, struct inode *inode)
1887 {
1888 	struct vfsmount *mnt = path->mnt;
1889 	struct timespec64 now;
1890 
1891 	if (inode->i_flags & S_NOATIME)
1892 		return false;
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 	 */
1897 	if (HAS_UNMAPPED_ID(mnt_idmap(mnt), inode))
1898 		return false;
1899 
1900 	if (IS_NOATIME(inode))
1901 		return false;
1902 	if ((inode->i_sb->s_flags & SB_NODIRATIME) && S_ISDIR(inode->i_mode))
1903 		return false;
1904 
1905 	if (mnt->mnt_flags & MNT_NOATIME)
1906 		return false;
1907 	if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode))
1908 		return false;
1909 
1910 	now = current_time(inode);
1911 
1912 	if (!relatime_need_update(mnt, inode, now))
1913 		return false;
1914 
1915 	if (timespec64_equal(&inode->i_atime, &now))
1916 		return false;
1917 
1918 	return true;
1919 }
1920 
touch_atime(const struct path * path)1921 void touch_atime(const struct path *path)
1922 {
1923 	struct vfsmount *mnt = path->mnt;
1924 	struct inode *inode = d_inode(path->dentry);
1925 	struct timespec64 now;
1926 
1927 	if (!atime_needs_update(path, inode))
1928 		return;
1929 
1930 	if (!sb_start_write_trylock(inode->i_sb))
1931 		return;
1932 
1933 	if (__mnt_want_write(mnt) != 0)
1934 		goto skip_update;
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.
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.
1943 	 */
1944 	now = current_time(inode);
1945 	inode_update_time(inode, &now, S_ATIME);
1946 	__mnt_drop_write(mnt);
1947 skip_update:
1948 	sb_end_write(inode->i_sb);
1949 }
1950 EXPORT_SYMBOL(touch_atime);
1951 
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  */
dentry_needs_remove_privs(struct mnt_idmap * idmap,struct dentry * dentry)1957 int dentry_needs_remove_privs(struct mnt_idmap *idmap,
1958 			      struct dentry *dentry)
1959 {
1960 	struct inode *inode = d_inode(dentry);
1961 	int mask = 0;
1962 	int ret;
1963 
1964 	if (IS_NOSEC(inode))
1965 		return 0;
1966 
1967 	mask = setattr_should_drop_suidgid(idmap, inode);
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 }
1975 
__remove_privs(struct mnt_idmap * idmap,struct dentry * dentry,int kill)1976 static int __remove_privs(struct mnt_idmap *idmap,
1977 			  struct dentry *dentry, int kill)
1978 {
1979 	struct iattr newattrs;
1980 
1981 	newattrs.ia_valid = ATTR_FORCE | kill;
1982 	/*
1983 	 * Note we call this on write, so notify_change will not
1984 	 * encounter any conflicting delegations:
1985 	 */
1986 	return notify_change(idmap, dentry, &newattrs, NULL);
1987 }
1988 
__file_remove_privs(struct file * file,unsigned int flags)1989 static int __file_remove_privs(struct file *file, unsigned int flags)
1990 {
1991 	struct dentry *dentry = file_dentry(file);
1992 	struct inode *inode = file_inode(file);
1993 	int error = 0;
1994 	int kill;
1995 
1996 	if (IS_NOSEC(inode) || !S_ISREG(inode->i_mode))
1997 		return 0;
1998 
1999 	kill = dentry_needs_remove_privs(file_mnt_idmap(file), dentry);
2000 	if (kill < 0)
2001 		return kill;
2002 
2003 	if (kill) {
2004 		if (flags & IOCB_NOWAIT)
2005 			return -EAGAIN;
2006 
2007 		error = __remove_privs(file_mnt_idmap(file), dentry, kill);
2008 	}
2009 
2010 	if (!error)
2011 		inode_has_no_xattr(inode);
2012 	return error;
2013 }
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  */
file_remove_privs(struct file * file)2024 int file_remove_privs(struct file *file)
2025 {
2026 	return __file_remove_privs(file, 0);
2027 }
2028 EXPORT_SYMBOL(file_remove_privs);
2029 
inode_needs_update_time(struct inode * inode,struct timespec64 * now)2030 static int inode_needs_update_time(struct inode *inode, struct timespec64 *now)
2031 {
2032 	int sync_it = 0;
2033 
2034 	/* First try to exhaust all avenues to not sync */
2035 	if (IS_NOCMTIME(inode))
2036 		return 0;
2037 
2038 	if (!timespec64_equal(&inode->i_mtime, now))
2039 		sync_it = S_MTIME;
2040 
2041 	if (!timespec64_equal(&inode->i_ctime, now))
2042 		sync_it |= S_CTIME;
2043 
2044 	if (IS_I_VERSION(inode) && inode_iversion_need_inc(inode))
2045 		sync_it |= S_VERSION;
2046 
2047 	return sync_it;
2048 }
2049 
__file_update_time(struct file * file,struct timespec64 * now,int sync_mode)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);
2055 
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 	}
2061 
2062 	return ret;
2063 }
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  */
file_update_time(struct file * file)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 }
2091 EXPORT_SYMBOL(file_update_time);
2092 
2093 /**
2094  * file_modified_flags - handle mandated vfs changes when modifying a file
2095  * @file: file that was modified
2096  * @flags: kiocb flags
2097  *
2098  * When file has been modified ensure that special
2099  * file privileges are removed and time settings are updated.
2100  *
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  *
2104  * Context: Caller must hold the file's inode lock.
2105  *
2106  * Return: 0 on success, negative errno on failure.
2107  */
file_modified_flags(struct file * file,int flags)2108 static int file_modified_flags(struct file *file, int flags)
2109 {
2110 	int ret;
2111 	struct inode *inode = file_inode(file);
2112 	struct timespec64 now = current_time(inode);
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 	 */
2118 	ret = __file_remove_privs(file, flags);
2119 	if (ret)
2120 		return ret;
2121 
2122 	if (unlikely(file->f_mode & FMODE_NOCMTIME))
2123 		return 0;
2124 
2125 	ret = inode_needs_update_time(inode, &now);
2126 	if (ret <= 0)
2127 		return ret;
2128 	if (flags & IOCB_NOWAIT)
2129 		return -EAGAIN;
2130 
2131 	return __file_update_time(file, &now, ret);
2132 }
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  */
file_modified(struct file * file)2145 int file_modified(struct file *file)
2146 {
2147 	return file_modified_flags(file, 0);
2148 }
2149 EXPORT_SYMBOL(file_modified);
2150 
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  */
kiocb_modified(struct kiocb * iocb)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 
inode_needs_sync(struct inode * inode)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 }
2176 EXPORT_SYMBOL(inode_needs_sync);
2177 
2178 /*
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  *
2185  * It doesn't matter if I_NEW is not set initially, a call to
2186  * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
2187  * will DTRT.
2188  */
__wait_on_freeing_inode(struct inode * inode)2189 static void __wait_on_freeing_inode(struct inode *inode)
2190 {
2191 	wait_queue_head_t *wq;
2192 	DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW);
2193 	wq = bit_waitqueue(&inode->i_state, __I_NEW);
2194 	prepare_to_wait(wq, &wait.wq_entry, TASK_UNINTERRUPTIBLE);
2195 	spin_unlock(&inode->i_lock);
2196 	spin_unlock(&inode_hash_lock);
2197 	schedule();
2198 	finish_wait(wq, &wait.wq_entry);
2199 	spin_lock(&inode_hash_lock);
2200 }
2201 
2202 static __initdata unsigned long ihash_entries;
set_ihash_entries(char * str)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  */
inode_init_early(void)2215 void __init inode_init_early(void)
2216 {
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,
2228 					HASH_EARLY | HASH_ZERO,
2229 					&i_hash_shift,
2230 					&i_hash_mask,
2231 					0,
2232 					0);
2233 }
2234 
inode_init(void)2235 void __init inode_init(void)
2236 {
2237 	/* inode slab cache */
2238 	inode_cachep = kmem_cache_create("inode_cache",
2239 					 sizeof(struct inode),
2240 					 0,
2241 					 (SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|
2242 					 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
2243 					 init_once);
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,
2254 					HASH_ZERO,
2255 					&i_hash_shift,
2256 					&i_hash_mask,
2257 					0,
2258 					0);
2259 }
2260 
init_special_inode(struct inode * inode,umode_t mode,dev_t rdev)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))
2271 		inode->i_fop = &pipefifo_fops;
2272 	else if (S_ISSOCK(mode))
2273 		;	/* leave it no_open_fops */
2274 	else
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);
2278 }
2279 EXPORT_SYMBOL(init_special_inode);
2280 
2281 /**
2282  * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
2283  * @idmap: idmap of the mount the inode was created from
2284  * @inode: New inode
2285  * @dir: Directory inode
2286  * @mode: mode of the new inode
2287  *
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
2291  * and initializing i_uid and i_gid. On non-idmapped mounts or if permission
2292  * checking is to be performed on the raw inode simply pass @nop_mnt_idmap.
2293  */
inode_init_owner(struct mnt_idmap * idmap,struct inode * inode,const struct inode * dir,umode_t mode)2294 void inode_init_owner(struct mnt_idmap *idmap, struct inode *inode,
2295 		      const struct inode *dir, umode_t mode)
2296 {
2297 	inode_fsuid_set(inode, idmap);
2298 	if (dir && dir->i_mode & S_ISGID) {
2299 		inode->i_gid = dir->i_gid;
2300 
2301 		/* Directories are special, and always inherit S_ISGID */
2302 		if (S_ISDIR(mode))
2303 			mode |= S_ISGID;
2304 	} else
2305 		inode_fsgid_set(inode, idmap);
2306 	inode->i_mode = mode;
2307 }
2308 EXPORT_SYMBOL(inode_init_owner);
2309 
2310 /**
2311  * inode_owner_or_capable - check current task permissions to inode
2312  * @idmap: idmap of the mount the inode was found from
2313  * @inode: inode being checked
2314  *
2315  * Return true if current either has CAP_FOWNER in a namespace with the
2316  * inode owner uid mapped, or owns the file.
2317  *
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.
2321  * On non-idmapped mounts or if permission checking is to be performed on the
2322  * raw inode simply passs @nop_mnt_idmap.
2323  */
inode_owner_or_capable(struct mnt_idmap * idmap,const struct inode * inode)2324 bool inode_owner_or_capable(struct mnt_idmap *idmap,
2325 			    const struct inode *inode)
2326 {
2327 	vfsuid_t vfsuid;
2328 	struct user_namespace *ns;
2329 
2330 	vfsuid = i_uid_into_vfsuid(idmap, inode);
2331 	if (vfsuid_eq_kuid(vfsuid, current_fsuid()))
2332 		return true;
2333 
2334 	ns = current_user_ns();
2335 	if (vfsuid_has_mapping(ns, vfsuid) && ns_capable(ns, CAP_FOWNER))
2336 		return true;
2337 	return false;
2338 }
2339 EXPORT_SYMBOL(inode_owner_or_capable);
2340 
2341 /*
2342  * Direct i/o helper functions
2343  */
__inode_dio_wait(struct inode * inode)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 {
2350 		prepare_to_wait(wq, &q.wq_entry, TASK_UNINTERRUPTIBLE);
2351 		if (atomic_read(&inode->i_dio_count))
2352 			schedule();
2353 	} while (atomic_read(&inode->i_dio_count));
2354 	finish_wait(wq, &q.wq_entry);
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  */
inode_dio_wait(struct inode * inode)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 
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
2382  * code path which doesn't today so we use cmpxchg() out of an abundance
2383  * of caution.
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  */
inode_set_flags(struct inode * inode,unsigned int flags,unsigned int mask)2390 void inode_set_flags(struct inode *inode, unsigned int flags,
2391 		     unsigned int mask)
2392 {
2393 	WARN_ON_ONCE(flags & ~mask);
2394 	set_mask_bits(&inode->i_flags, mask, flags);
2395 }
2396 EXPORT_SYMBOL(inode_set_flags);
2397 
inode_nohighmem(struct inode * inode)2398 void inode_nohighmem(struct inode *inode)
2399 {
2400 	mapping_set_gfp_mask(inode->i_mapping, GFP_USER);
2401 }
2402 EXPORT_SYMBOL(inode_nohighmem);
2403 
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  */
timestamp_truncate(struct timespec64 t,struct inode * inode)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 
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  */
current_time(struct inode * inode)2445 struct timespec64 current_time(struct inode *inode)
2446 {
2447 	struct timespec64 now;
2448 
2449 	ktime_get_coarse_real_ts64(&now);
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 
2456 	return timestamp_truncate(now, inode);
2457 }
2458 EXPORT_SYMBOL(current_time);
2459 
2460 /**
2461  * in_group_or_capable - check whether caller is CAP_FSETID privileged
2462  * @idmap:	idmap of the mount @inode was found from
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  */
in_group_or_capable(struct mnt_idmap * idmap,const struct inode * inode,vfsgid_t vfsgid)2472 bool in_group_or_capable(struct mnt_idmap *idmap,
2473 			 const struct inode *inode, vfsgid_t vfsgid)
2474 {
2475 	if (vfsgid_in_group_p(vfsgid))
2476 		return true;
2477 	if (capable_wrt_inode_uidgid(idmap, inode, CAP_FSETID))
2478 		return true;
2479 	return false;
2480 }
2481 
2482 /**
2483  * mode_strip_sgid - handle the sgid bit for non-directories
2484  * @idmap: idmap of the mount the inode was created from
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  */
mode_strip_sgid(struct mnt_idmap * idmap,const struct inode * dir,umode_t mode)2496 umode_t mode_strip_sgid(struct mnt_idmap *idmap,
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;
2503 	if (in_group_or_capable(idmap, dir, i_gid_into_vfsgid(idmap, dir)))
2504 		return mode;
2505 	return mode & ~S_ISGID;
2506 }
2507 EXPORT_SYMBOL(mode_strip_sgid);
2508