xref: /linux/fs/overlayfs/util.c (revision 7fc2cd2e4b398c57c9cf961cfea05eadbf34c05c)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Copyright (C) 2011 Novell Inc.
4  * Copyright (C) 2016 Red Hat, Inc.
5  */
6 
7 #include <linux/fs.h>
8 #include <linux/mount.h>
9 #include <linux/slab.h>
10 #include <linux/cred.h>
11 #include <linux/xattr.h>
12 #include <linux/exportfs.h>
13 #include <linux/file.h>
14 #include <linux/fileattr.h>
15 #include <linux/uuid.h>
16 #include <linux/namei.h>
17 #include <linux/ratelimit.h>
18 #include <linux/overflow.h>
19 #include "overlayfs.h"
20 
21 /* Get write access to upper mnt - may fail if upper sb was remounted ro */
22 int ovl_get_write_access(struct dentry *dentry)
23 {
24 	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);
25 	return mnt_get_write_access(ovl_upper_mnt(ofs));
26 }
27 
28 /* Get write access to upper sb - may block if upper sb is frozen */
29 void ovl_start_write(struct dentry *dentry)
30 {
31 	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);
32 	sb_start_write(ovl_upper_mnt(ofs)->mnt_sb);
33 }
34 
35 int ovl_want_write(struct dentry *dentry)
36 {
37 	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);
38 	return mnt_want_write(ovl_upper_mnt(ofs));
39 }
40 
41 void ovl_put_write_access(struct dentry *dentry)
42 {
43 	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);
44 	mnt_put_write_access(ovl_upper_mnt(ofs));
45 }
46 
47 void ovl_end_write(struct dentry *dentry)
48 {
49 	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);
50 	sb_end_write(ovl_upper_mnt(ofs)->mnt_sb);
51 }
52 
53 void ovl_drop_write(struct dentry *dentry)
54 {
55 	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);
56 	mnt_drop_write(ovl_upper_mnt(ofs));
57 }
58 
59 struct dentry *ovl_workdir(struct dentry *dentry)
60 {
61 	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);
62 	return ofs->workdir;
63 }
64 
65 const struct cred *ovl_override_creds(struct super_block *sb)
66 {
67 	struct ovl_fs *ofs = OVL_FS(sb);
68 
69 	return override_creds(ofs->creator_cred);
70 }
71 
72 /*
73  * Check if underlying fs supports file handles and try to determine encoding
74  * type, in order to deduce maximum inode number used by fs.
75  *
76  * Return 0 if file handles are not supported.
77  * Return 1 (FILEID_INO32_GEN) if fs uses the default 32bit inode encoding.
78  * Return -1 if fs uses a non default encoding with unknown inode size.
79  */
80 int ovl_can_decode_fh(struct super_block *sb)
81 {
82 	if (!capable(CAP_DAC_READ_SEARCH))
83 		return 0;
84 
85 	if (!exportfs_can_decode_fh(sb->s_export_op))
86 		return 0;
87 
88 	return sb->s_export_op->encode_fh ? -1 : FILEID_INO32_GEN;
89 }
90 
91 struct dentry *ovl_indexdir(struct super_block *sb)
92 {
93 	struct ovl_fs *ofs = OVL_FS(sb);
94 
95 	return ofs->config.index ? ofs->workdir : NULL;
96 }
97 
98 /* Index all files on copy up. For now only enabled for NFS export */
99 bool ovl_index_all(struct super_block *sb)
100 {
101 	struct ovl_fs *ofs = OVL_FS(sb);
102 
103 	return ofs->config.nfs_export && ofs->config.index;
104 }
105 
106 /* Verify lower origin on lookup. For now only enabled for NFS export */
107 bool ovl_verify_lower(struct super_block *sb)
108 {
109 	struct ovl_fs *ofs = OVL_FS(sb);
110 
111 	return ofs->config.nfs_export && ofs->config.index;
112 }
113 
114 struct ovl_path *ovl_stack_alloc(unsigned int n)
115 {
116 	return kcalloc(n, sizeof(struct ovl_path), GFP_KERNEL);
117 }
118 
119 void ovl_stack_cpy(struct ovl_path *dst, struct ovl_path *src, unsigned int n)
120 {
121 	unsigned int i;
122 
123 	memcpy(dst, src, sizeof(struct ovl_path) * n);
124 	for (i = 0; i < n; i++)
125 		dget(src[i].dentry);
126 }
127 
128 void ovl_stack_put(struct ovl_path *stack, unsigned int n)
129 {
130 	unsigned int i;
131 
132 	for (i = 0; stack && i < n; i++)
133 		dput(stack[i].dentry);
134 }
135 
136 void ovl_stack_free(struct ovl_path *stack, unsigned int n)
137 {
138 	ovl_stack_put(stack, n);
139 	kfree(stack);
140 }
141 
142 struct ovl_entry *ovl_alloc_entry(unsigned int numlower)
143 {
144 	struct ovl_entry *oe;
145 
146 	oe = kzalloc(struct_size(oe, __lowerstack, numlower), GFP_KERNEL);
147 	if (oe)
148 		oe->__numlower = numlower;
149 
150 	return oe;
151 }
152 
153 void ovl_free_entry(struct ovl_entry *oe)
154 {
155 	ovl_stack_put(ovl_lowerstack(oe), ovl_numlower(oe));
156 	kfree(oe);
157 }
158 
159 #define OVL_D_REVALIDATE (DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE)
160 
161 bool ovl_dentry_remote(struct dentry *dentry)
162 {
163 	return dentry->d_flags & OVL_D_REVALIDATE;
164 }
165 
166 void ovl_dentry_update_reval(struct dentry *dentry, struct dentry *realdentry)
167 {
168 	if (!ovl_dentry_remote(realdentry))
169 		return;
170 
171 	spin_lock(&dentry->d_lock);
172 	dentry->d_flags |= realdentry->d_flags & OVL_D_REVALIDATE;
173 	spin_unlock(&dentry->d_lock);
174 }
175 
176 void ovl_dentry_init_reval(struct dentry *dentry, struct dentry *upperdentry,
177 			   struct ovl_entry *oe)
178 {
179 	return ovl_dentry_init_flags(dentry, upperdentry, oe, OVL_D_REVALIDATE);
180 }
181 
182 void ovl_dentry_init_flags(struct dentry *dentry, struct dentry *upperdentry,
183 			   struct ovl_entry *oe, unsigned int mask)
184 {
185 	struct ovl_path *lowerstack = ovl_lowerstack(oe);
186 	unsigned int i, flags = 0;
187 
188 	if (upperdentry)
189 		flags |= upperdentry->d_flags;
190 	for (i = 0; i < ovl_numlower(oe) && lowerstack[i].dentry; i++)
191 		flags |= lowerstack[i].dentry->d_flags;
192 
193 	spin_lock(&dentry->d_lock);
194 	dentry->d_flags &= ~mask;
195 	dentry->d_flags |= flags & mask;
196 	spin_unlock(&dentry->d_lock);
197 }
198 
199 bool ovl_dentry_weird(struct dentry *dentry)
200 {
201 	if (!d_can_lookup(dentry) && !d_is_file(dentry) && !d_is_symlink(dentry))
202 		return true;
203 
204 	if (dentry->d_flags & (DCACHE_NEED_AUTOMOUNT | DCACHE_MANAGE_TRANSIT))
205 		return true;
206 
207 	/*
208 	 * Exceptionally for layers with casefold, we accept that they have
209 	 * their own hash and compare operations
210 	 */
211 	if (sb_has_encoding(dentry->d_sb))
212 		return false;
213 
214 	return dentry->d_flags & (DCACHE_OP_HASH | DCACHE_OP_COMPARE);
215 }
216 
217 enum ovl_path_type ovl_path_type(struct dentry *dentry)
218 {
219 	struct ovl_entry *oe = OVL_E(dentry);
220 	enum ovl_path_type type = 0;
221 
222 	if (ovl_dentry_upper(dentry)) {
223 		type = __OVL_PATH_UPPER;
224 
225 		/*
226 		 * Non-dir dentry can hold lower dentry of its copy up origin.
227 		 */
228 		if (ovl_numlower(oe)) {
229 			if (ovl_test_flag(OVL_CONST_INO, d_inode(dentry)))
230 				type |= __OVL_PATH_ORIGIN;
231 			if (d_is_dir(dentry) ||
232 			    !ovl_has_upperdata(d_inode(dentry)))
233 				type |= __OVL_PATH_MERGE;
234 		}
235 	} else {
236 		if (ovl_numlower(oe) > 1)
237 			type |= __OVL_PATH_MERGE;
238 	}
239 	return type;
240 }
241 
242 void ovl_path_upper(struct dentry *dentry, struct path *path)
243 {
244 	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);
245 
246 	path->mnt = ovl_upper_mnt(ofs);
247 	path->dentry = ovl_dentry_upper(dentry);
248 }
249 
250 void ovl_path_lower(struct dentry *dentry, struct path *path)
251 {
252 	struct ovl_entry *oe = OVL_E(dentry);
253 	struct ovl_path *lowerpath = ovl_lowerstack(oe);
254 
255 	if (ovl_numlower(oe)) {
256 		path->mnt = lowerpath->layer->mnt;
257 		path->dentry = lowerpath->dentry;
258 	} else {
259 		*path = (struct path) { };
260 	}
261 }
262 
263 void ovl_path_lowerdata(struct dentry *dentry, struct path *path)
264 {
265 	struct ovl_entry *oe = OVL_E(dentry);
266 	struct ovl_path *lowerdata = ovl_lowerdata(oe);
267 	struct dentry *lowerdata_dentry = ovl_lowerdata_dentry(oe);
268 
269 	if (lowerdata_dentry) {
270 		path->dentry = lowerdata_dentry;
271 		/*
272 		 * Pairs with smp_wmb() in ovl_dentry_set_lowerdata().
273 		 * Make sure that if lowerdata->dentry is visible, then
274 		 * datapath->layer is visible as well.
275 		 */
276 		smp_rmb();
277 		path->mnt = READ_ONCE(lowerdata->layer)->mnt;
278 	} else {
279 		*path = (struct path) { };
280 	}
281 }
282 
283 enum ovl_path_type ovl_path_real(struct dentry *dentry, struct path *path)
284 {
285 	enum ovl_path_type type = ovl_path_type(dentry);
286 
287 	if (!OVL_TYPE_UPPER(type))
288 		ovl_path_lower(dentry, path);
289 	else
290 		ovl_path_upper(dentry, path);
291 
292 	return type;
293 }
294 
295 enum ovl_path_type ovl_path_realdata(struct dentry *dentry, struct path *path)
296 {
297 	enum ovl_path_type type = ovl_path_type(dentry);
298 
299 	WARN_ON_ONCE(d_is_dir(dentry));
300 
301 	if (!OVL_TYPE_UPPER(type) || OVL_TYPE_MERGE(type))
302 		ovl_path_lowerdata(dentry, path);
303 	else
304 		ovl_path_upper(dentry, path);
305 
306 	return type;
307 }
308 
309 struct dentry *ovl_dentry_upper(struct dentry *dentry)
310 {
311 	struct inode *inode = d_inode(dentry);
312 
313 	return inode ? ovl_upperdentry_dereference(OVL_I(inode)) : NULL;
314 }
315 
316 struct dentry *ovl_dentry_lower(struct dentry *dentry)
317 {
318 	struct ovl_entry *oe = OVL_E(dentry);
319 
320 	return ovl_numlower(oe) ? ovl_lowerstack(oe)->dentry : NULL;
321 }
322 
323 const struct ovl_layer *ovl_layer_lower(struct dentry *dentry)
324 {
325 	struct ovl_entry *oe = OVL_E(dentry);
326 
327 	return ovl_numlower(oe) ? ovl_lowerstack(oe)->layer : NULL;
328 }
329 
330 /*
331  * ovl_dentry_lower() could return either a data dentry or metacopy dentry
332  * depending on what is stored in lowerstack[0]. At times we need to find
333  * lower dentry which has data (and not metacopy dentry). This helper
334  * returns the lower data dentry.
335  */
336 struct dentry *ovl_dentry_lowerdata(struct dentry *dentry)
337 {
338 	return ovl_lowerdata_dentry(OVL_E(dentry));
339 }
340 
341 int ovl_dentry_set_lowerdata(struct dentry *dentry, struct ovl_path *datapath)
342 {
343 	struct ovl_entry *oe = OVL_E(dentry);
344 	struct ovl_path *lowerdata = ovl_lowerdata(oe);
345 	struct dentry *datadentry = datapath->dentry;
346 
347 	if (WARN_ON_ONCE(ovl_numlower(oe) <= 1))
348 		return -EIO;
349 
350 	WRITE_ONCE(lowerdata->layer, datapath->layer);
351 	/*
352 	 * Pairs with smp_rmb() in ovl_path_lowerdata().
353 	 * Make sure that if lowerdata->dentry is visible, then
354 	 * lowerdata->layer is visible as well.
355 	 */
356 	smp_wmb();
357 	WRITE_ONCE(lowerdata->dentry, dget(datadentry));
358 
359 	ovl_dentry_update_reval(dentry, datadentry);
360 
361 	return 0;
362 }
363 
364 struct dentry *ovl_dentry_real(struct dentry *dentry)
365 {
366 	return ovl_dentry_upper(dentry) ?: ovl_dentry_lower(dentry);
367 }
368 
369 struct dentry *ovl_i_dentry_upper(struct inode *inode)
370 {
371 	return ovl_upperdentry_dereference(OVL_I(inode));
372 }
373 
374 struct inode *ovl_i_path_real(struct inode *inode, struct path *path)
375 {
376 	struct ovl_path *lowerpath = ovl_lowerpath(OVL_I_E(inode));
377 
378 	path->dentry = ovl_i_dentry_upper(inode);
379 	if (!path->dentry) {
380 		path->dentry = lowerpath->dentry;
381 		path->mnt = lowerpath->layer->mnt;
382 	} else {
383 		path->mnt = ovl_upper_mnt(OVL_FS(inode->i_sb));
384 	}
385 
386 	return path->dentry ? d_inode_rcu(path->dentry) : NULL;
387 }
388 
389 struct inode *ovl_inode_upper(struct inode *inode)
390 {
391 	struct dentry *upperdentry = ovl_i_dentry_upper(inode);
392 
393 	return upperdentry ? d_inode(upperdentry) : NULL;
394 }
395 
396 struct inode *ovl_inode_lower(struct inode *inode)
397 {
398 	struct ovl_path *lowerpath = ovl_lowerpath(OVL_I_E(inode));
399 
400 	return lowerpath ? d_inode(lowerpath->dentry) : NULL;
401 }
402 
403 struct inode *ovl_inode_real(struct inode *inode)
404 {
405 	return ovl_inode_upper(inode) ?: ovl_inode_lower(inode);
406 }
407 
408 /* Return inode which contains lower data. Do not return metacopy */
409 struct inode *ovl_inode_lowerdata(struct inode *inode)
410 {
411 	struct dentry *lowerdata = ovl_lowerdata_dentry(OVL_I_E(inode));
412 
413 	if (WARN_ON(!S_ISREG(inode->i_mode)))
414 		return NULL;
415 
416 	return lowerdata ? d_inode(lowerdata) : NULL;
417 }
418 
419 /* Return real inode which contains data. Does not return metacopy inode */
420 struct inode *ovl_inode_realdata(struct inode *inode)
421 {
422 	struct inode *upperinode;
423 
424 	upperinode = ovl_inode_upper(inode);
425 	if (upperinode && ovl_has_upperdata(inode))
426 		return upperinode;
427 
428 	return ovl_inode_lowerdata(inode);
429 }
430 
431 const char *ovl_lowerdata_redirect(struct inode *inode)
432 {
433 	return inode && S_ISREG(inode->i_mode) ?
434 		OVL_I(inode)->lowerdata_redirect : NULL;
435 }
436 
437 struct ovl_dir_cache *ovl_dir_cache(struct inode *inode)
438 {
439 	return inode && S_ISDIR(inode->i_mode) ? OVL_I(inode)->cache : NULL;
440 }
441 
442 void ovl_set_dir_cache(struct inode *inode, struct ovl_dir_cache *cache)
443 {
444 	OVL_I(inode)->cache = cache;
445 }
446 
447 void ovl_dentry_set_flag(unsigned long flag, struct dentry *dentry)
448 {
449 	set_bit(flag, OVL_E_FLAGS(dentry));
450 }
451 
452 void ovl_dentry_clear_flag(unsigned long flag, struct dentry *dentry)
453 {
454 	clear_bit(flag, OVL_E_FLAGS(dentry));
455 }
456 
457 bool ovl_dentry_test_flag(unsigned long flag, struct dentry *dentry)
458 {
459 	return test_bit(flag, OVL_E_FLAGS(dentry));
460 }
461 
462 bool ovl_dentry_is_opaque(struct dentry *dentry)
463 {
464 	return ovl_dentry_test_flag(OVL_E_OPAQUE, dentry);
465 }
466 
467 bool ovl_dentry_is_whiteout(struct dentry *dentry)
468 {
469 	return !dentry->d_inode && ovl_dentry_is_opaque(dentry);
470 }
471 
472 void ovl_dentry_set_opaque(struct dentry *dentry)
473 {
474 	ovl_dentry_set_flag(OVL_E_OPAQUE, dentry);
475 }
476 
477 bool ovl_dentry_has_xwhiteouts(struct dentry *dentry)
478 {
479 	return ovl_dentry_test_flag(OVL_E_XWHITEOUTS, dentry);
480 }
481 
482 void ovl_dentry_set_xwhiteouts(struct dentry *dentry)
483 {
484 	ovl_dentry_set_flag(OVL_E_XWHITEOUTS, dentry);
485 }
486 
487 /*
488  * ovl_layer_set_xwhiteouts() is called before adding the overlay dir
489  * dentry to dcache, while readdir of that same directory happens after
490  * the overlay dir dentry is in dcache, so if some cpu observes that
491  * ovl_dentry_is_xwhiteouts(), it will also observe layer->has_xwhiteouts
492  * for the layers where xwhiteouts marker was found in that merge dir.
493  */
494 void ovl_layer_set_xwhiteouts(struct ovl_fs *ofs,
495 			      const struct ovl_layer *layer)
496 {
497 	if (layer->has_xwhiteouts)
498 		return;
499 
500 	/* Write once to read-mostly layer properties */
501 	ofs->layers[layer->idx].has_xwhiteouts = true;
502 }
503 
504 /*
505  * For hard links and decoded file handles, it's possible for ovl_dentry_upper()
506  * to return positive, while there's no actual upper alias for the inode.
507  * Copy up code needs to know about the existence of the upper alias, so it
508  * can't use ovl_dentry_upper().
509  */
510 bool ovl_dentry_has_upper_alias(struct dentry *dentry)
511 {
512 	return ovl_dentry_test_flag(OVL_E_UPPER_ALIAS, dentry);
513 }
514 
515 void ovl_dentry_set_upper_alias(struct dentry *dentry)
516 {
517 	ovl_dentry_set_flag(OVL_E_UPPER_ALIAS, dentry);
518 }
519 
520 static bool ovl_should_check_upperdata(struct inode *inode)
521 {
522 	if (!S_ISREG(inode->i_mode))
523 		return false;
524 
525 	if (!ovl_inode_lower(inode))
526 		return false;
527 
528 	return true;
529 }
530 
531 bool ovl_has_upperdata(struct inode *inode)
532 {
533 	if (!ovl_should_check_upperdata(inode))
534 		return true;
535 
536 	if (!ovl_test_flag(OVL_UPPERDATA, inode))
537 		return false;
538 	/*
539 	 * Pairs with smp_wmb() in ovl_set_upperdata(). Main user of
540 	 * ovl_has_upperdata() is ovl_copy_up_meta_inode_data(). Make sure
541 	 * if setting of OVL_UPPERDATA is visible, then effects of writes
542 	 * before that are visible too.
543 	 */
544 	smp_rmb();
545 	return true;
546 }
547 
548 void ovl_set_upperdata(struct inode *inode)
549 {
550 	/*
551 	 * Pairs with smp_rmb() in ovl_has_upperdata(). Make sure
552 	 * if OVL_UPPERDATA flag is visible, then effects of write operations
553 	 * before it are visible as well.
554 	 */
555 	smp_wmb();
556 	ovl_set_flag(OVL_UPPERDATA, inode);
557 }
558 
559 /* Caller should hold ovl_inode->lock */
560 bool ovl_dentry_needs_data_copy_up_locked(struct dentry *dentry, int flags)
561 {
562 	if (!ovl_open_flags_need_copy_up(flags))
563 		return false;
564 
565 	return !ovl_test_flag(OVL_UPPERDATA, d_inode(dentry));
566 }
567 
568 bool ovl_dentry_needs_data_copy_up(struct dentry *dentry, int flags)
569 {
570 	if (!ovl_open_flags_need_copy_up(flags))
571 		return false;
572 
573 	return !ovl_has_upperdata(d_inode(dentry));
574 }
575 
576 const char *ovl_dentry_get_redirect(struct dentry *dentry)
577 {
578 	return OVL_I(d_inode(dentry))->redirect;
579 }
580 
581 void ovl_dentry_set_redirect(struct dentry *dentry, const char *redirect)
582 {
583 	struct ovl_inode *oi = OVL_I(d_inode(dentry));
584 
585 	kfree(oi->redirect);
586 	oi->redirect = redirect;
587 }
588 
589 void ovl_inode_update(struct inode *inode, struct dentry *upperdentry)
590 {
591 	struct inode *upperinode = d_inode(upperdentry);
592 
593 	WARN_ON(OVL_I(inode)->__upperdentry);
594 
595 	/*
596 	 * Make sure upperdentry is consistent before making it visible
597 	 */
598 	smp_wmb();
599 	OVL_I(inode)->__upperdentry = upperdentry;
600 	if (inode_unhashed(inode)) {
601 		inode->i_private = upperinode;
602 		__insert_inode_hash(inode, (unsigned long) upperinode);
603 	}
604 }
605 
606 static void ovl_dir_version_inc(struct dentry *dentry, bool impurity)
607 {
608 	struct inode *inode = d_inode(dentry);
609 
610 	WARN_ON(!inode_is_locked(inode));
611 	WARN_ON(!d_is_dir(dentry));
612 	/*
613 	 * Version is used by readdir code to keep cache consistent.
614 	 * For merge dirs (or dirs with origin) all changes need to be noted.
615 	 * For non-merge dirs, cache contains only impure entries (i.e. ones
616 	 * which have been copied up and have origins), so only need to note
617 	 * changes to impure entries.
618 	 */
619 	if (!ovl_dir_is_real(inode) || impurity)
620 		OVL_I(inode)->version++;
621 }
622 
623 void ovl_dir_modified(struct dentry *dentry, bool impurity)
624 {
625 	/* Copy mtime/ctime */
626 	ovl_copyattr(d_inode(dentry));
627 
628 	ovl_dir_version_inc(dentry, impurity);
629 }
630 
631 u64 ovl_inode_version_get(struct inode *inode)
632 {
633 	WARN_ON(!inode_is_locked(inode));
634 	return OVL_I(inode)->version;
635 }
636 
637 bool ovl_is_whiteout(struct dentry *dentry)
638 {
639 	struct inode *inode = dentry->d_inode;
640 
641 	return inode && IS_WHITEOUT(inode);
642 }
643 
644 /*
645  * Use this over ovl_is_whiteout for upper and lower files, as it also
646  * handles overlay.whiteout xattr whiteout files.
647  */
648 bool ovl_path_is_whiteout(struct ovl_fs *ofs, const struct path *path)
649 {
650 	return ovl_is_whiteout(path->dentry) ||
651 		ovl_path_check_xwhiteout_xattr(ofs, path);
652 }
653 
654 struct file *ovl_path_open(const struct path *path, int flags)
655 {
656 	struct inode *inode = d_inode(path->dentry);
657 	struct mnt_idmap *real_idmap = mnt_idmap(path->mnt);
658 	int err, acc_mode;
659 
660 	if (flags & ~(O_ACCMODE | O_LARGEFILE))
661 		BUG();
662 
663 	switch (flags & O_ACCMODE) {
664 	case O_RDONLY:
665 		acc_mode = MAY_READ;
666 		break;
667 	case O_WRONLY:
668 		acc_mode = MAY_WRITE;
669 		break;
670 	default:
671 		BUG();
672 	}
673 
674 	err = inode_permission(real_idmap, inode, acc_mode | MAY_OPEN);
675 	if (err)
676 		return ERR_PTR(err);
677 
678 	/* O_NOATIME is an optimization, don't fail if not permitted */
679 	if (inode_owner_or_capable(real_idmap, inode))
680 		flags |= O_NOATIME;
681 
682 	return dentry_open(path, flags, current_cred());
683 }
684 
685 /* Caller should hold ovl_inode->lock */
686 static bool ovl_already_copied_up_locked(struct dentry *dentry, int flags)
687 {
688 	bool disconnected = dentry->d_flags & DCACHE_DISCONNECTED;
689 
690 	if (ovl_dentry_upper(dentry) &&
691 	    (ovl_dentry_has_upper_alias(dentry) || disconnected) &&
692 	    !ovl_dentry_needs_data_copy_up_locked(dentry, flags))
693 		return true;
694 
695 	return false;
696 }
697 
698 bool ovl_already_copied_up(struct dentry *dentry, int flags)
699 {
700 	bool disconnected = dentry->d_flags & DCACHE_DISCONNECTED;
701 
702 	/*
703 	 * Check if copy-up has happened as well as for upper alias (in
704 	 * case of hard links) is there.
705 	 *
706 	 * Both checks are lockless:
707 	 *  - false negatives: will recheck under oi->lock
708 	 *  - false positives:
709 	 *    + ovl_dentry_upper() uses memory barriers to ensure the
710 	 *      upper dentry is up-to-date
711 	 *    + ovl_dentry_has_upper_alias() relies on locking of
712 	 *      upper parent i_rwsem to prevent reordering copy-up
713 	 *      with rename.
714 	 */
715 	if (ovl_dentry_upper(dentry) &&
716 	    (ovl_dentry_has_upper_alias(dentry) || disconnected) &&
717 	    !ovl_dentry_needs_data_copy_up(dentry, flags))
718 		return true;
719 
720 	return false;
721 }
722 
723 /*
724  * The copy up "transaction" keeps an elevated mnt write count on upper mnt,
725  * but leaves taking freeze protection on upper sb to lower level helpers.
726  */
727 int ovl_copy_up_start(struct dentry *dentry, int flags)
728 {
729 	struct inode *inode = d_inode(dentry);
730 	int err;
731 
732 	err = ovl_inode_lock_interruptible(inode);
733 	if (err)
734 		return err;
735 
736 	if (ovl_already_copied_up_locked(dentry, flags))
737 		err = 1; /* Already copied up */
738 	else
739 		err = ovl_get_write_access(dentry);
740 	if (err)
741 		goto out_unlock;
742 
743 	return 0;
744 
745 out_unlock:
746 	ovl_inode_unlock(inode);
747 	return err;
748 }
749 
750 void ovl_copy_up_end(struct dentry *dentry)
751 {
752 	ovl_put_write_access(dentry);
753 	ovl_inode_unlock(d_inode(dentry));
754 }
755 
756 bool ovl_path_check_origin_xattr(struct ovl_fs *ofs, const struct path *path)
757 {
758 	int res;
759 
760 	res = ovl_path_getxattr(ofs, path, OVL_XATTR_ORIGIN, NULL, 0);
761 
762 	/* Zero size value means "copied up but origin unknown" */
763 	if (res >= 0)
764 		return true;
765 
766 	return false;
767 }
768 
769 bool ovl_path_check_xwhiteout_xattr(struct ovl_fs *ofs, const struct path *path)
770 {
771 	struct dentry *dentry = path->dentry;
772 	int res;
773 
774 	/* xattr.whiteout must be a zero size regular file */
775 	if (!d_is_reg(dentry) || i_size_read(d_inode(dentry)) != 0)
776 		return false;
777 
778 	res = ovl_path_getxattr(ofs, path, OVL_XATTR_XWHITEOUT, NULL, 0);
779 	return res >= 0;
780 }
781 
782 /*
783  * Load persistent uuid from xattr into s_uuid if found, or store a new
784  * random generated value in s_uuid and in xattr.
785  */
786 bool ovl_init_uuid_xattr(struct super_block *sb, struct ovl_fs *ofs,
787 			 const struct path *upperpath)
788 {
789 	bool set = false;
790 	uuid_t uuid;
791 	int res;
792 
793 	/* Try to load existing persistent uuid */
794 	res = ovl_path_getxattr(ofs, upperpath, OVL_XATTR_UUID, uuid.b,
795 				UUID_SIZE);
796 	if (res == UUID_SIZE)
797 		goto set_uuid;
798 
799 	if (res != -ENODATA)
800 		goto fail;
801 
802 	/*
803 	 * With uuid=auto, if uuid xattr is found, it will be used.
804 	 * If uuid xattrs is not found, generate a persistent uuid only on mount
805 	 * of new overlays where upper root dir is not yet marked as impure.
806 	 * An upper dir is marked as impure on copy up or lookup of its subdirs.
807 	 */
808 	if (ofs->config.uuid == OVL_UUID_AUTO) {
809 		res = ovl_path_getxattr(ofs, upperpath, OVL_XATTR_IMPURE, NULL,
810 					0);
811 		if (res > 0) {
812 			/* Any mount of old overlay - downgrade to uuid=null */
813 			ofs->config.uuid = OVL_UUID_NULL;
814 			return true;
815 		} else if (res == -ENODATA) {
816 			/* First mount of new overlay - upgrade to uuid=on */
817 			ofs->config.uuid = OVL_UUID_ON;
818 		} else if (res < 0) {
819 			goto fail;
820 		}
821 
822 	}
823 
824 	/* Generate overlay instance uuid */
825 	uuid_gen(&uuid);
826 
827 	/* Try to store persistent uuid */
828 	set = true;
829 	res = ovl_setxattr(ofs, upperpath->dentry, OVL_XATTR_UUID, uuid.b,
830 			   UUID_SIZE);
831 	if (res)
832 		goto fail;
833 
834 set_uuid:
835 	super_set_uuid(sb, uuid.b, sizeof(uuid));
836 	return true;
837 
838 fail:
839 	ofs->config.uuid = OVL_UUID_NULL;
840 	pr_warn("failed to %s uuid (%pd2, err=%i); falling back to uuid=null.\n",
841 		set ? "set" : "get", upperpath->dentry, res);
842 	return false;
843 }
844 
845 char ovl_get_dir_xattr_val(struct ovl_fs *ofs, const struct path *path,
846 			   enum ovl_xattr ox)
847 {
848 	int res;
849 	char val;
850 
851 	if (!d_is_dir(path->dentry))
852 		return 0;
853 
854 	res = ovl_path_getxattr(ofs, path, ox, &val, 1);
855 	return res == 1 ? val : 0;
856 }
857 
858 #define OVL_XATTR_OPAQUE_POSTFIX	"opaque"
859 #define OVL_XATTR_REDIRECT_POSTFIX	"redirect"
860 #define OVL_XATTR_ORIGIN_POSTFIX	"origin"
861 #define OVL_XATTR_IMPURE_POSTFIX	"impure"
862 #define OVL_XATTR_NLINK_POSTFIX		"nlink"
863 #define OVL_XATTR_UPPER_POSTFIX		"upper"
864 #define OVL_XATTR_UUID_POSTFIX		"uuid"
865 #define OVL_XATTR_METACOPY_POSTFIX	"metacopy"
866 #define OVL_XATTR_PROTATTR_POSTFIX	"protattr"
867 #define OVL_XATTR_XWHITEOUT_POSTFIX	"whiteout"
868 
869 #define OVL_XATTR_TAB_ENTRY(x) \
870 	[x] = { [false] = OVL_XATTR_TRUSTED_PREFIX x ## _POSTFIX, \
871 		[true] = OVL_XATTR_USER_PREFIX x ## _POSTFIX }
872 
873 const char *const ovl_xattr_table[][2] = {
874 	OVL_XATTR_TAB_ENTRY(OVL_XATTR_OPAQUE),
875 	OVL_XATTR_TAB_ENTRY(OVL_XATTR_REDIRECT),
876 	OVL_XATTR_TAB_ENTRY(OVL_XATTR_ORIGIN),
877 	OVL_XATTR_TAB_ENTRY(OVL_XATTR_IMPURE),
878 	OVL_XATTR_TAB_ENTRY(OVL_XATTR_NLINK),
879 	OVL_XATTR_TAB_ENTRY(OVL_XATTR_UPPER),
880 	OVL_XATTR_TAB_ENTRY(OVL_XATTR_UUID),
881 	OVL_XATTR_TAB_ENTRY(OVL_XATTR_METACOPY),
882 	OVL_XATTR_TAB_ENTRY(OVL_XATTR_PROTATTR),
883 	OVL_XATTR_TAB_ENTRY(OVL_XATTR_XWHITEOUT),
884 };
885 
886 int ovl_check_setxattr(struct ovl_fs *ofs, struct dentry *upperdentry,
887 		       enum ovl_xattr ox, const void *value, size_t size,
888 		       int xerr)
889 {
890 	int err;
891 
892 	if (ofs->noxattr)
893 		return xerr;
894 
895 	err = ovl_setxattr(ofs, upperdentry, ox, value, size);
896 
897 	if (err == -EOPNOTSUPP) {
898 		pr_warn("cannot set %s xattr on upper\n", ovl_xattr(ofs, ox));
899 		ofs->noxattr = true;
900 		return xerr;
901 	}
902 
903 	return err;
904 }
905 
906 int ovl_set_impure(struct dentry *dentry, struct dentry *upperdentry)
907 {
908 	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);
909 	int err;
910 
911 	if (ovl_test_flag(OVL_IMPURE, d_inode(dentry)))
912 		return 0;
913 
914 	/*
915 	 * Do not fail when upper doesn't support xattrs.
916 	 * Upper inodes won't have origin nor redirect xattr anyway.
917 	 */
918 	err = ovl_check_setxattr(ofs, upperdentry, OVL_XATTR_IMPURE, "y", 1, 0);
919 	if (!err)
920 		ovl_set_flag(OVL_IMPURE, d_inode(dentry));
921 
922 	return err;
923 }
924 
925 
926 #define OVL_PROTATTR_MAX 32 /* Reserved for future flags */
927 
928 void ovl_check_protattr(struct inode *inode, struct dentry *upper)
929 {
930 	struct ovl_fs *ofs = OVL_FS(inode->i_sb);
931 	u32 iflags = inode->i_flags & OVL_PROT_I_FLAGS_MASK;
932 	char buf[OVL_PROTATTR_MAX+1];
933 	int res, n;
934 
935 	res = ovl_getxattr_upper(ofs, upper, OVL_XATTR_PROTATTR, buf,
936 				 OVL_PROTATTR_MAX);
937 	if (res < 0)
938 		return;
939 
940 	/*
941 	 * Initialize inode flags from overlay.protattr xattr and upper inode
942 	 * flags.  If upper inode has those fileattr flags set (i.e. from old
943 	 * kernel), we do not clear them on ovl_get_inode(), but we will clear
944 	 * them on next fileattr_set().
945 	 */
946 	for (n = 0; n < res; n++) {
947 		if (buf[n] == 'a')
948 			iflags |= S_APPEND;
949 		else if (buf[n] == 'i')
950 			iflags |= S_IMMUTABLE;
951 		else
952 			break;
953 	}
954 
955 	if (!res || n < res) {
956 		pr_warn_ratelimited("incompatible overlay.protattr format (%pd2, len=%d)\n",
957 				    upper, res);
958 	} else {
959 		inode_set_flags(inode, iflags, OVL_PROT_I_FLAGS_MASK);
960 	}
961 }
962 
963 int ovl_set_protattr(struct inode *inode, struct dentry *upper,
964 		      struct file_kattr *fa)
965 {
966 	struct ovl_fs *ofs = OVL_FS(inode->i_sb);
967 	char buf[OVL_PROTATTR_MAX];
968 	int len = 0, err = 0;
969 	u32 iflags = 0;
970 
971 	BUILD_BUG_ON(HWEIGHT32(OVL_PROT_FS_FLAGS_MASK) > OVL_PROTATTR_MAX);
972 
973 	if (fa->flags & FS_APPEND_FL) {
974 		buf[len++] = 'a';
975 		iflags |= S_APPEND;
976 	}
977 	if (fa->flags & FS_IMMUTABLE_FL) {
978 		buf[len++] = 'i';
979 		iflags |= S_IMMUTABLE;
980 	}
981 
982 	/*
983 	 * Do not allow to set protection flags when upper doesn't support
984 	 * xattrs, because we do not set those fileattr flags on upper inode.
985 	 * Remove xattr if it exist and all protection flags are cleared.
986 	 */
987 	if (len) {
988 		err = ovl_check_setxattr(ofs, upper, OVL_XATTR_PROTATTR,
989 					 buf, len, -EPERM);
990 	} else if (inode->i_flags & OVL_PROT_I_FLAGS_MASK) {
991 		err = ovl_removexattr(ofs, upper, OVL_XATTR_PROTATTR);
992 		if (err == -EOPNOTSUPP || err == -ENODATA)
993 			err = 0;
994 	}
995 	if (err)
996 		return err;
997 
998 	inode_set_flags(inode, iflags, OVL_PROT_I_FLAGS_MASK);
999 
1000 	/* Mask out the fileattr flags that should not be set in upper inode */
1001 	fa->flags &= ~OVL_PROT_FS_FLAGS_MASK;
1002 	fa->fsx_xflags &= ~OVL_PROT_FSX_FLAGS_MASK;
1003 
1004 	return 0;
1005 }
1006 
1007 /*
1008  * Caller must hold a reference to inode to prevent it from being freed while
1009  * it is marked inuse.
1010  */
1011 bool ovl_inuse_trylock(struct dentry *dentry)
1012 {
1013 	struct inode *inode = d_inode(dentry);
1014 	bool locked = false;
1015 
1016 	spin_lock(&inode->i_lock);
1017 	if (!(inode_state_read(inode) & I_OVL_INUSE)) {
1018 		inode_state_set(inode, I_OVL_INUSE);
1019 		locked = true;
1020 	}
1021 	spin_unlock(&inode->i_lock);
1022 
1023 	return locked;
1024 }
1025 
1026 void ovl_inuse_unlock(struct dentry *dentry)
1027 {
1028 	if (dentry) {
1029 		struct inode *inode = d_inode(dentry);
1030 
1031 		spin_lock(&inode->i_lock);
1032 		WARN_ON(!(inode_state_read(inode) & I_OVL_INUSE));
1033 		inode_state_clear(inode, I_OVL_INUSE);
1034 		spin_unlock(&inode->i_lock);
1035 	}
1036 }
1037 
1038 bool ovl_is_inuse(struct dentry *dentry)
1039 {
1040 	struct inode *inode = d_inode(dentry);
1041 	bool inuse;
1042 
1043 	spin_lock(&inode->i_lock);
1044 	inuse = (inode_state_read(inode) & I_OVL_INUSE);
1045 	spin_unlock(&inode->i_lock);
1046 
1047 	return inuse;
1048 }
1049 
1050 /*
1051  * Does this overlay dentry need to be indexed on copy up?
1052  */
1053 bool ovl_need_index(struct dentry *dentry)
1054 {
1055 	struct dentry *lower = ovl_dentry_lower(dentry);
1056 
1057 	if (!lower || !ovl_indexdir(dentry->d_sb))
1058 		return false;
1059 
1060 	/* Index all files for NFS export and consistency verification */
1061 	if (ovl_index_all(dentry->d_sb))
1062 		return true;
1063 
1064 	/* Index only lower hardlinks on copy up */
1065 	if (!d_is_dir(lower) && d_inode(lower)->i_nlink > 1)
1066 		return true;
1067 
1068 	return false;
1069 }
1070 
1071 /* Caller must hold OVL_I(inode)->lock */
1072 static void ovl_cleanup_index(struct dentry *dentry)
1073 {
1074 	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);
1075 	struct dentry *indexdir = ovl_indexdir(dentry->d_sb);
1076 	struct dentry *lowerdentry = ovl_dentry_lower(dentry);
1077 	struct dentry *upperdentry = ovl_dentry_upper(dentry);
1078 	struct dentry *index = NULL;
1079 	struct inode *inode;
1080 	struct qstr name = { };
1081 	bool got_write = false;
1082 	int err;
1083 
1084 	err = ovl_get_index_name(ofs, lowerdentry, &name);
1085 	if (err)
1086 		goto fail;
1087 
1088 	err = ovl_want_write(dentry);
1089 	if (err)
1090 		goto fail;
1091 
1092 	got_write = true;
1093 	inode = d_inode(upperdentry);
1094 	if (!S_ISDIR(inode->i_mode) && inode->i_nlink != 1) {
1095 		pr_warn_ratelimited("cleanup linked index (%pd2, ino=%lu, nlink=%u)\n",
1096 				    upperdentry, inode->i_ino, inode->i_nlink);
1097 		/*
1098 		 * We either have a bug with persistent union nlink or a lower
1099 		 * hardlink was added while overlay is mounted. Adding a lower
1100 		 * hardlink and then unlinking all overlay hardlinks would drop
1101 		 * overlay nlink to zero before all upper inodes are unlinked.
1102 		 * As a safety measure, when that situation is detected, set
1103 		 * the overlay nlink to the index inode nlink minus one for the
1104 		 * index entry itself.
1105 		 */
1106 		set_nlink(d_inode(dentry), inode->i_nlink - 1);
1107 		ovl_set_nlink_upper(dentry);
1108 		goto out;
1109 	}
1110 
1111 	index = ovl_lookup_upper_unlocked(ofs, name.name, indexdir, name.len);
1112 	err = PTR_ERR(index);
1113 	if (IS_ERR(index)) {
1114 		index = NULL;
1115 	} else if (ovl_index_all(dentry->d_sb)) {
1116 		/* Whiteout orphan index to block future open by handle */
1117 		err = ovl_cleanup_and_whiteout(OVL_FS(dentry->d_sb),
1118 					       indexdir, index);
1119 	} else {
1120 		/* Cleanup orphan index entries */
1121 		err = ovl_cleanup(ofs, indexdir, index);
1122 	}
1123 	if (err)
1124 		goto fail;
1125 
1126 out:
1127 	if (got_write)
1128 		ovl_drop_write(dentry);
1129 	kfree(name.name);
1130 	dput(index);
1131 	return;
1132 
1133 fail:
1134 	pr_err("cleanup index of '%pd2' failed (%i)\n", dentry, err);
1135 	goto out;
1136 }
1137 
1138 /*
1139  * Operations that change overlay inode and upper inode nlink need to be
1140  * synchronized with copy up for persistent nlink accounting.
1141  */
1142 int ovl_nlink_start(struct dentry *dentry)
1143 {
1144 	struct inode *inode = d_inode(dentry);
1145 	int err;
1146 
1147 	if (WARN_ON(!inode))
1148 		return -ENOENT;
1149 
1150 	/*
1151 	 * With inodes index is enabled, we store the union overlay nlink
1152 	 * in an xattr on the index inode. When whiting out an indexed lower,
1153 	 * we need to decrement the overlay persistent nlink, but before the
1154 	 * first copy up, we have no upper index inode to store the xattr.
1155 	 *
1156 	 * As a workaround, before whiteout/rename over an indexed lower,
1157 	 * copy up to create the upper index. Creating the upper index will
1158 	 * initialize the overlay nlink, so it could be dropped if unlink
1159 	 * or rename succeeds.
1160 	 *
1161 	 * TODO: implement metadata only index copy up when called with
1162 	 *       ovl_copy_up_flags(dentry, O_PATH).
1163 	 */
1164 	if (ovl_need_index(dentry) && !ovl_dentry_has_upper_alias(dentry)) {
1165 		err = ovl_copy_up(dentry);
1166 		if (err)
1167 			return err;
1168 	}
1169 
1170 	err = ovl_inode_lock_interruptible(inode);
1171 	if (err)
1172 		return err;
1173 
1174 	err = ovl_want_write(dentry);
1175 	if (err)
1176 		goto out_unlock;
1177 
1178 	if (d_is_dir(dentry) || !ovl_test_flag(OVL_INDEX, inode))
1179 		return 0;
1180 
1181 	/*
1182 	 * The overlay inode nlink should be incremented/decremented IFF the
1183 	 * upper operation succeeds, along with nlink change of upper inode.
1184 	 * Therefore, before link/unlink/rename, we store the union nlink
1185 	 * value relative to the upper inode nlink in an upper inode xattr.
1186 	 */
1187 	with_ovl_creds(dentry->d_sb)
1188 		err = ovl_set_nlink_upper(dentry);
1189 	if (err)
1190 		goto out_drop_write;
1191 
1192 	return 0;
1193 
1194 out_drop_write:
1195 	ovl_drop_write(dentry);
1196 out_unlock:
1197 	ovl_inode_unlock(inode);
1198 
1199 	return err;
1200 }
1201 
1202 void ovl_nlink_end(struct dentry *dentry)
1203 {
1204 	struct inode *inode = d_inode(dentry);
1205 
1206 	ovl_drop_write(dentry);
1207 
1208 	if (ovl_test_flag(OVL_INDEX, inode) && inode->i_nlink == 0) {
1209 		with_ovl_creds(dentry->d_sb)
1210 			ovl_cleanup_index(dentry);
1211 	}
1212 
1213 	ovl_inode_unlock(inode);
1214 }
1215 
1216 int ovl_lock_rename_workdir(struct dentry *workdir, struct dentry *work,
1217 			    struct dentry *upperdir, struct dentry *upper)
1218 {
1219 	struct dentry *trap;
1220 
1221 	/* Workdir should not be subdir of upperdir and vice versa */
1222 	trap = lock_rename(workdir, upperdir);
1223 	if (IS_ERR(trap))
1224 		goto err;
1225 	if (trap)
1226 		goto err_unlock;
1227 	if (work && (work->d_parent != workdir || d_unhashed(work)))
1228 		goto err_unlock;
1229 	if (upper && (upper->d_parent != upperdir || d_unhashed(upper)))
1230 		goto err_unlock;
1231 
1232 	return 0;
1233 
1234 err_unlock:
1235 	unlock_rename(workdir, upperdir);
1236 err:
1237 	pr_err("failed to lock workdir+upperdir\n");
1238 	return -EIO;
1239 }
1240 
1241 /*
1242  * err < 0, 0 if no metacopy xattr, metacopy data size if xattr found.
1243  * an empty xattr returns OVL_METACOPY_MIN_SIZE to distinguish from no xattr value.
1244  */
1245 int ovl_check_metacopy_xattr(struct ovl_fs *ofs, const struct path *path,
1246 			     struct ovl_metacopy *data)
1247 {
1248 	int res;
1249 
1250 	/* Only regular files can have metacopy xattr */
1251 	if (!S_ISREG(d_inode(path->dentry)->i_mode))
1252 		return 0;
1253 
1254 	res = ovl_path_getxattr(ofs, path, OVL_XATTR_METACOPY,
1255 				data, data ? OVL_METACOPY_MAX_SIZE : 0);
1256 	if (res < 0) {
1257 		if (res == -ENODATA || res == -EOPNOTSUPP)
1258 			return 0;
1259 		/*
1260 		 * getxattr on user.* may fail with EACCES in case there's no
1261 		 * read permission on the inode.  Not much we can do, other than
1262 		 * tell the caller that this is not a metacopy inode.
1263 		 */
1264 		if (ofs->config.userxattr && res == -EACCES)
1265 			return 0;
1266 		goto out;
1267 	}
1268 
1269 	if (res == 0) {
1270 		/* Emulate empty data for zero size metacopy xattr */
1271 		res = OVL_METACOPY_MIN_SIZE;
1272 		if (data) {
1273 			memset(data, 0, res);
1274 			data->len = res;
1275 		}
1276 	} else if (res < OVL_METACOPY_MIN_SIZE) {
1277 		pr_warn_ratelimited("metacopy file '%pd' has too small xattr\n",
1278 				    path->dentry);
1279 		return -EIO;
1280 	} else if (data) {
1281 		if (data->version != 0) {
1282 			pr_warn_ratelimited("metacopy file '%pd' has unsupported version\n",
1283 					    path->dentry);
1284 			return -EIO;
1285 		}
1286 		if (res != data->len) {
1287 			pr_warn_ratelimited("metacopy file '%pd' has invalid xattr size\n",
1288 					    path->dentry);
1289 			return -EIO;
1290 		}
1291 	}
1292 
1293 	return res;
1294 out:
1295 	pr_warn_ratelimited("failed to get metacopy (%i)\n", res);
1296 	return res;
1297 }
1298 
1299 int ovl_set_metacopy_xattr(struct ovl_fs *ofs, struct dentry *d, struct ovl_metacopy *metacopy)
1300 {
1301 	size_t len = metacopy->len;
1302 
1303 	/* If no flags or digest fall back to empty metacopy file */
1304 	if (metacopy->version == 0 && metacopy->flags == 0 && metacopy->digest_algo == 0)
1305 		len = 0;
1306 
1307 	return ovl_check_setxattr(ofs, d, OVL_XATTR_METACOPY,
1308 				  metacopy, len, -EOPNOTSUPP);
1309 }
1310 
1311 bool ovl_is_metacopy_dentry(struct dentry *dentry)
1312 {
1313 	struct ovl_entry *oe = OVL_E(dentry);
1314 
1315 	if (!d_is_reg(dentry))
1316 		return false;
1317 
1318 	if (ovl_dentry_upper(dentry)) {
1319 		if (!ovl_has_upperdata(d_inode(dentry)))
1320 			return true;
1321 		return false;
1322 	}
1323 
1324 	return (ovl_numlower(oe) > 1);
1325 }
1326 
1327 char *ovl_get_redirect_xattr(struct ovl_fs *ofs, const struct path *path, int padding)
1328 {
1329 	int res;
1330 	char *s, *next, *buf = NULL;
1331 
1332 	res = ovl_path_getxattr(ofs, path, OVL_XATTR_REDIRECT, NULL, 0);
1333 	if (res == -ENODATA || res == -EOPNOTSUPP)
1334 		return NULL;
1335 	if (res < 0)
1336 		goto fail;
1337 	if (res == 0)
1338 		goto invalid;
1339 
1340 	buf = kzalloc(res + padding + 1, GFP_KERNEL);
1341 	if (!buf)
1342 		return ERR_PTR(-ENOMEM);
1343 
1344 	res = ovl_path_getxattr(ofs, path, OVL_XATTR_REDIRECT, buf, res);
1345 	if (res < 0)
1346 		goto fail;
1347 	if (res == 0)
1348 		goto invalid;
1349 
1350 	if (buf[0] == '/') {
1351 		for (s = buf; *s++ == '/'; s = next) {
1352 			next = strchrnul(s, '/');
1353 			if (s == next)
1354 				goto invalid;
1355 		}
1356 	} else {
1357 		if (strchr(buf, '/') != NULL)
1358 			goto invalid;
1359 	}
1360 
1361 	return buf;
1362 invalid:
1363 	pr_warn_ratelimited("invalid redirect (%s)\n", buf);
1364 	res = -EINVAL;
1365 	goto err_free;
1366 fail:
1367 	pr_warn_ratelimited("failed to get redirect (%i)\n", res);
1368 err_free:
1369 	kfree(buf);
1370 	return ERR_PTR(res);
1371 }
1372 
1373 /* Call with mounter creds as it may open the file */
1374 int ovl_ensure_verity_loaded(const struct path *datapath)
1375 {
1376 	struct inode *inode = d_inode(datapath->dentry);
1377 	struct file *filp;
1378 
1379 	if (!fsverity_active(inode) && IS_VERITY(inode)) {
1380 		/*
1381 		 * If this inode was not yet opened, the verity info hasn't been
1382 		 * loaded yet, so we need to do that here to force it into memory.
1383 		 */
1384 		filp = kernel_file_open(datapath, O_RDONLY, current_cred());
1385 		if (IS_ERR(filp))
1386 			return PTR_ERR(filp);
1387 		fput(filp);
1388 	}
1389 
1390 	return 0;
1391 }
1392 
1393 int ovl_validate_verity(struct ovl_fs *ofs,
1394 			const struct path *metapath,
1395 			const struct path *datapath)
1396 {
1397 	struct ovl_metacopy metacopy_data;
1398 	u8 actual_digest[FS_VERITY_MAX_DIGEST_SIZE];
1399 	int xattr_digest_size, digest_size;
1400 	int xattr_size, err;
1401 	u8 verity_algo;
1402 
1403 	if (!ofs->config.verity_mode ||
1404 	    /* Verity only works on regular files */
1405 	    !S_ISREG(d_inode(metapath->dentry)->i_mode))
1406 		return 0;
1407 
1408 	xattr_size = ovl_check_metacopy_xattr(ofs, metapath, &metacopy_data);
1409 	if (xattr_size < 0)
1410 		return xattr_size;
1411 
1412 	if (!xattr_size || !metacopy_data.digest_algo) {
1413 		if (ofs->config.verity_mode == OVL_VERITY_REQUIRE) {
1414 			pr_warn_ratelimited("metacopy file '%pd' has no digest specified\n",
1415 					    metapath->dentry);
1416 			return -EIO;
1417 		}
1418 		return 0;
1419 	}
1420 
1421 	xattr_digest_size = ovl_metadata_digest_size(&metacopy_data);
1422 
1423 	err = ovl_ensure_verity_loaded(datapath);
1424 	if (err < 0) {
1425 		pr_warn_ratelimited("lower file '%pd' failed to load fs-verity info\n",
1426 				    datapath->dentry);
1427 		return -EIO;
1428 	}
1429 
1430 	digest_size = fsverity_get_digest(d_inode(datapath->dentry), actual_digest,
1431 					  &verity_algo, NULL);
1432 	if (digest_size == 0) {
1433 		pr_warn_ratelimited("lower file '%pd' has no fs-verity digest\n", datapath->dentry);
1434 		return -EIO;
1435 	}
1436 
1437 	if (xattr_digest_size != digest_size ||
1438 	    metacopy_data.digest_algo != verity_algo ||
1439 	    memcmp(metacopy_data.digest, actual_digest, xattr_digest_size) != 0) {
1440 		pr_warn_ratelimited("lower file '%pd' has the wrong fs-verity digest\n",
1441 				    datapath->dentry);
1442 		return -EIO;
1443 	}
1444 
1445 	return 0;
1446 }
1447 
1448 int ovl_get_verity_digest(struct ovl_fs *ofs, const struct path *src,
1449 			  struct ovl_metacopy *metacopy)
1450 {
1451 	int err, digest_size;
1452 
1453 	if (!ofs->config.verity_mode || !S_ISREG(d_inode(src->dentry)->i_mode))
1454 		return 0;
1455 
1456 	err = ovl_ensure_verity_loaded(src);
1457 	if (err < 0) {
1458 		pr_warn_ratelimited("lower file '%pd' failed to load fs-verity info\n",
1459 				    src->dentry);
1460 		return -EIO;
1461 	}
1462 
1463 	digest_size = fsverity_get_digest(d_inode(src->dentry),
1464 					  metacopy->digest, &metacopy->digest_algo, NULL);
1465 	if (digest_size == 0 ||
1466 	    WARN_ON_ONCE(digest_size > FS_VERITY_MAX_DIGEST_SIZE)) {
1467 		if (ofs->config.verity_mode == OVL_VERITY_REQUIRE) {
1468 			pr_warn_ratelimited("lower file '%pd' has no fs-verity digest\n",
1469 					    src->dentry);
1470 			return -EIO;
1471 		}
1472 		return 0;
1473 	}
1474 
1475 	metacopy->len += digest_size;
1476 	return 0;
1477 }
1478 
1479 /*
1480  * ovl_sync_status() - Check fs sync status for volatile mounts
1481  *
1482  * Returns 1 if this is not a volatile mount and a real sync is required.
1483  *
1484  * Returns 0 if syncing can be skipped because mount is volatile, and no errors
1485  * have occurred on the upperdir since the mount.
1486  *
1487  * Returns -errno if it is a volatile mount, and the error that occurred since
1488  * the last mount. If the error code changes, it'll return the latest error
1489  * code.
1490  */
1491 
1492 int ovl_sync_status(struct ovl_fs *ofs)
1493 {
1494 	struct vfsmount *mnt;
1495 
1496 	if (ovl_should_sync(ofs))
1497 		return 1;
1498 
1499 	mnt = ovl_upper_mnt(ofs);
1500 	if (!mnt)
1501 		return 0;
1502 
1503 	return errseq_check(&mnt->mnt_sb->s_wb_err, ofs->errseq);
1504 }
1505 
1506 /*
1507  * ovl_copyattr() - copy inode attributes from layer to ovl inode
1508  *
1509  * When overlay copies inode information from an upper or lower layer to the
1510  * relevant overlay inode it will apply the idmapping of the upper or lower
1511  * layer when doing so ensuring that the ovl inode ownership will correctly
1512  * reflect the ownership of the idmapped upper or lower layer. For example, an
1513  * idmapped upper or lower layer mapping id 1001 to id 1000 will take care to
1514  * map any lower or upper inode owned by id 1001 to id 1000. These mapping
1515  * helpers are nops when the relevant layer isn't idmapped.
1516  */
1517 void ovl_copyattr(struct inode *inode)
1518 {
1519 	struct path realpath;
1520 	struct inode *realinode;
1521 	struct mnt_idmap *real_idmap;
1522 	vfsuid_t vfsuid;
1523 	vfsgid_t vfsgid;
1524 
1525 	realinode = ovl_i_path_real(inode, &realpath);
1526 	real_idmap = mnt_idmap(realpath.mnt);
1527 
1528 	spin_lock(&inode->i_lock);
1529 	vfsuid = i_uid_into_vfsuid(real_idmap, realinode);
1530 	vfsgid = i_gid_into_vfsgid(real_idmap, realinode);
1531 
1532 	inode->i_uid = vfsuid_into_kuid(vfsuid);
1533 	inode->i_gid = vfsgid_into_kgid(vfsgid);
1534 	inode->i_mode = realinode->i_mode;
1535 	inode_set_atime_to_ts(inode, inode_get_atime(realinode));
1536 	inode_set_mtime_to_ts(inode, inode_get_mtime(realinode));
1537 	inode_set_ctime_to_ts(inode, inode_get_ctime(realinode));
1538 	i_size_write(inode, i_size_read(realinode));
1539 	spin_unlock(&inode->i_lock);
1540 }
1541