xref: /openbmc/linux/fs/btrfs/tree-checker.c (revision 753f1745)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (C) Qu Wenruo 2017.  All rights reserved.
4  */
5 
6 /*
7  * The module is used to catch unexpected/corrupted tree block data.
8  * Such behavior can be caused either by a fuzzed image or bugs.
9  *
10  * The objective is to do leaf/node validation checks when tree block is read
11  * from disk, and check *every* possible member, so other code won't
12  * need to checking them again.
13  *
14  * Due to the potential and unwanted damage, every checker needs to be
15  * carefully reviewed otherwise so it does not prevent mount of valid images.
16  */
17 
18 #include <linux/types.h>
19 #include <linux/stddef.h>
20 #include <linux/error-injection.h>
21 #include "messages.h"
22 #include "ctree.h"
23 #include "tree-checker.h"
24 #include "disk-io.h"
25 #include "compression.h"
26 #include "volumes.h"
27 #include "misc.h"
28 #include "fs.h"
29 #include "accessors.h"
30 #include "file-item.h"
31 #include "inode-item.h"
32 #include "extent-tree.h"
33 
34 /*
35  * Error message should follow the following format:
36  * corrupt <type>: <identifier>, <reason>[, <bad_value>]
37  *
38  * @type:	leaf or node
39  * @identifier:	the necessary info to locate the leaf/node.
40  * 		It's recommended to decode key.objecitd/offset if it's
41  * 		meaningful.
42  * @reason:	describe the error
43  * @bad_value:	optional, it's recommended to output bad value and its
44  *		expected value (range).
45  *
46  * Since comma is used to separate the components, only space is allowed
47  * inside each component.
48  */
49 
50 /*
51  * Append generic "corrupt leaf/node root=%llu block=%llu slot=%d: " to @fmt.
52  * Allows callers to customize the output.
53  */
54 __printf(3, 4)
55 __cold
56 static void generic_err(const struct extent_buffer *eb, int slot,
57 			const char *fmt, ...)
58 {
59 	const struct btrfs_fs_info *fs_info = eb->fs_info;
60 	struct va_format vaf;
61 	va_list args;
62 
63 	va_start(args, fmt);
64 
65 	vaf.fmt = fmt;
66 	vaf.va = &args;
67 
68 	btrfs_crit(fs_info,
69 		"corrupt %s: root=%llu block=%llu slot=%d, %pV",
70 		btrfs_header_level(eb) == 0 ? "leaf" : "node",
71 		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, &vaf);
72 	va_end(args);
73 }
74 
75 /*
76  * Customized reporter for extent data item, since its key objectid and
77  * offset has its own meaning.
78  */
79 __printf(3, 4)
80 __cold
81 static void file_extent_err(const struct extent_buffer *eb, int slot,
82 			    const char *fmt, ...)
83 {
84 	const struct btrfs_fs_info *fs_info = eb->fs_info;
85 	struct btrfs_key key;
86 	struct va_format vaf;
87 	va_list args;
88 
89 	btrfs_item_key_to_cpu(eb, &key, slot);
90 	va_start(args, fmt);
91 
92 	vaf.fmt = fmt;
93 	vaf.va = &args;
94 
95 	btrfs_crit(fs_info,
96 	"corrupt %s: root=%llu block=%llu slot=%d ino=%llu file_offset=%llu, %pV",
97 		btrfs_header_level(eb) == 0 ? "leaf" : "node",
98 		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
99 		key.objectid, key.offset, &vaf);
100 	va_end(args);
101 }
102 
103 /*
104  * Return 0 if the btrfs_file_extent_##name is aligned to @alignment
105  * Else return 1
106  */
107 #define CHECK_FE_ALIGNED(leaf, slot, fi, name, alignment)		      \
108 ({									      \
109 	if (unlikely(!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)),      \
110 				 (alignment))))				      \
111 		file_extent_err((leaf), (slot),				      \
112 	"invalid %s for file extent, have %llu, should be aligned to %u",     \
113 			(#name), btrfs_file_extent_##name((leaf), (fi)),      \
114 			(alignment));					      \
115 	(!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment)));   \
116 })
117 
118 static u64 file_extent_end(struct extent_buffer *leaf,
119 			   struct btrfs_key *key,
120 			   struct btrfs_file_extent_item *extent)
121 {
122 	u64 end;
123 	u64 len;
124 
125 	if (btrfs_file_extent_type(leaf, extent) == BTRFS_FILE_EXTENT_INLINE) {
126 		len = btrfs_file_extent_ram_bytes(leaf, extent);
127 		end = ALIGN(key->offset + len, leaf->fs_info->sectorsize);
128 	} else {
129 		len = btrfs_file_extent_num_bytes(leaf, extent);
130 		end = key->offset + len;
131 	}
132 	return end;
133 }
134 
135 /*
136  * Customized report for dir_item, the only new important information is
137  * key->objectid, which represents inode number
138  */
139 __printf(3, 4)
140 __cold
141 static void dir_item_err(const struct extent_buffer *eb, int slot,
142 			 const char *fmt, ...)
143 {
144 	const struct btrfs_fs_info *fs_info = eb->fs_info;
145 	struct btrfs_key key;
146 	struct va_format vaf;
147 	va_list args;
148 
149 	btrfs_item_key_to_cpu(eb, &key, slot);
150 	va_start(args, fmt);
151 
152 	vaf.fmt = fmt;
153 	vaf.va = &args;
154 
155 	btrfs_crit(fs_info,
156 		"corrupt %s: root=%llu block=%llu slot=%d ino=%llu, %pV",
157 		btrfs_header_level(eb) == 0 ? "leaf" : "node",
158 		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
159 		key.objectid, &vaf);
160 	va_end(args);
161 }
162 
163 /*
164  * This functions checks prev_key->objectid, to ensure current key and prev_key
165  * share the same objectid as inode number.
166  *
167  * This is to detect missing INODE_ITEM in subvolume trees.
168  *
169  * Return true if everything is OK or we don't need to check.
170  * Return false if anything is wrong.
171  */
172 static bool check_prev_ino(struct extent_buffer *leaf,
173 			   struct btrfs_key *key, int slot,
174 			   struct btrfs_key *prev_key)
175 {
176 	/* No prev key, skip check */
177 	if (slot == 0)
178 		return true;
179 
180 	/* Only these key->types needs to be checked */
181 	ASSERT(key->type == BTRFS_XATTR_ITEM_KEY ||
182 	       key->type == BTRFS_INODE_REF_KEY ||
183 	       key->type == BTRFS_DIR_INDEX_KEY ||
184 	       key->type == BTRFS_DIR_ITEM_KEY ||
185 	       key->type == BTRFS_EXTENT_DATA_KEY);
186 
187 	/*
188 	 * Only subvolume trees along with their reloc trees need this check.
189 	 * Things like log tree doesn't follow this ino requirement.
190 	 */
191 	if (!is_fstree(btrfs_header_owner(leaf)))
192 		return true;
193 
194 	if (key->objectid == prev_key->objectid)
195 		return true;
196 
197 	/* Error found */
198 	dir_item_err(leaf, slot,
199 		"invalid previous key objectid, have %llu expect %llu",
200 		prev_key->objectid, key->objectid);
201 	return false;
202 }
203 static int check_extent_data_item(struct extent_buffer *leaf,
204 				  struct btrfs_key *key, int slot,
205 				  struct btrfs_key *prev_key)
206 {
207 	struct btrfs_fs_info *fs_info = leaf->fs_info;
208 	struct btrfs_file_extent_item *fi;
209 	u32 sectorsize = fs_info->sectorsize;
210 	u32 item_size = btrfs_item_size(leaf, slot);
211 	u64 extent_end;
212 
213 	if (unlikely(!IS_ALIGNED(key->offset, sectorsize))) {
214 		file_extent_err(leaf, slot,
215 "unaligned file_offset for file extent, have %llu should be aligned to %u",
216 			key->offset, sectorsize);
217 		return -EUCLEAN;
218 	}
219 
220 	/*
221 	 * Previous key must have the same key->objectid (ino).
222 	 * It can be XATTR_ITEM, INODE_ITEM or just another EXTENT_DATA.
223 	 * But if objectids mismatch, it means we have a missing
224 	 * INODE_ITEM.
225 	 */
226 	if (unlikely(!check_prev_ino(leaf, key, slot, prev_key)))
227 		return -EUCLEAN;
228 
229 	fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
230 
231 	/*
232 	 * Make sure the item contains at least inline header, so the file
233 	 * extent type is not some garbage.
234 	 */
235 	if (unlikely(item_size < BTRFS_FILE_EXTENT_INLINE_DATA_START)) {
236 		file_extent_err(leaf, slot,
237 				"invalid item size, have %u expect [%zu, %u)",
238 				item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START,
239 				SZ_4K);
240 		return -EUCLEAN;
241 	}
242 	if (unlikely(btrfs_file_extent_type(leaf, fi) >=
243 		     BTRFS_NR_FILE_EXTENT_TYPES)) {
244 		file_extent_err(leaf, slot,
245 		"invalid type for file extent, have %u expect range [0, %u]",
246 			btrfs_file_extent_type(leaf, fi),
247 			BTRFS_NR_FILE_EXTENT_TYPES - 1);
248 		return -EUCLEAN;
249 	}
250 
251 	/*
252 	 * Support for new compression/encryption must introduce incompat flag,
253 	 * and must be caught in open_ctree().
254 	 */
255 	if (unlikely(btrfs_file_extent_compression(leaf, fi) >=
256 		     BTRFS_NR_COMPRESS_TYPES)) {
257 		file_extent_err(leaf, slot,
258 	"invalid compression for file extent, have %u expect range [0, %u]",
259 			btrfs_file_extent_compression(leaf, fi),
260 			BTRFS_NR_COMPRESS_TYPES - 1);
261 		return -EUCLEAN;
262 	}
263 	if (unlikely(btrfs_file_extent_encryption(leaf, fi))) {
264 		file_extent_err(leaf, slot,
265 			"invalid encryption for file extent, have %u expect 0",
266 			btrfs_file_extent_encryption(leaf, fi));
267 		return -EUCLEAN;
268 	}
269 	if (btrfs_file_extent_type(leaf, fi) == BTRFS_FILE_EXTENT_INLINE) {
270 		/* Inline extent must have 0 as key offset */
271 		if (unlikely(key->offset)) {
272 			file_extent_err(leaf, slot,
273 		"invalid file_offset for inline file extent, have %llu expect 0",
274 				key->offset);
275 			return -EUCLEAN;
276 		}
277 
278 		/* Compressed inline extent has no on-disk size, skip it */
279 		if (btrfs_file_extent_compression(leaf, fi) !=
280 		    BTRFS_COMPRESS_NONE)
281 			return 0;
282 
283 		/* Uncompressed inline extent size must match item size */
284 		if (unlikely(item_size != BTRFS_FILE_EXTENT_INLINE_DATA_START +
285 					  btrfs_file_extent_ram_bytes(leaf, fi))) {
286 			file_extent_err(leaf, slot,
287 	"invalid ram_bytes for uncompressed inline extent, have %u expect %llu",
288 				item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START +
289 				btrfs_file_extent_ram_bytes(leaf, fi));
290 			return -EUCLEAN;
291 		}
292 		return 0;
293 	}
294 
295 	/* Regular or preallocated extent has fixed item size */
296 	if (unlikely(item_size != sizeof(*fi))) {
297 		file_extent_err(leaf, slot,
298 	"invalid item size for reg/prealloc file extent, have %u expect %zu",
299 			item_size, sizeof(*fi));
300 		return -EUCLEAN;
301 	}
302 	if (unlikely(CHECK_FE_ALIGNED(leaf, slot, fi, ram_bytes, sectorsize) ||
303 		     CHECK_FE_ALIGNED(leaf, slot, fi, disk_bytenr, sectorsize) ||
304 		     CHECK_FE_ALIGNED(leaf, slot, fi, disk_num_bytes, sectorsize) ||
305 		     CHECK_FE_ALIGNED(leaf, slot, fi, offset, sectorsize) ||
306 		     CHECK_FE_ALIGNED(leaf, slot, fi, num_bytes, sectorsize)))
307 		return -EUCLEAN;
308 
309 	/* Catch extent end overflow */
310 	if (unlikely(check_add_overflow(btrfs_file_extent_num_bytes(leaf, fi),
311 					key->offset, &extent_end))) {
312 		file_extent_err(leaf, slot,
313 	"extent end overflow, have file offset %llu extent num bytes %llu",
314 				key->offset,
315 				btrfs_file_extent_num_bytes(leaf, fi));
316 		return -EUCLEAN;
317 	}
318 
319 	/*
320 	 * Check that no two consecutive file extent items, in the same leaf,
321 	 * present ranges that overlap each other.
322 	 */
323 	if (slot > 0 &&
324 	    prev_key->objectid == key->objectid &&
325 	    prev_key->type == BTRFS_EXTENT_DATA_KEY) {
326 		struct btrfs_file_extent_item *prev_fi;
327 		u64 prev_end;
328 
329 		prev_fi = btrfs_item_ptr(leaf, slot - 1,
330 					 struct btrfs_file_extent_item);
331 		prev_end = file_extent_end(leaf, prev_key, prev_fi);
332 		if (unlikely(prev_end > key->offset)) {
333 			file_extent_err(leaf, slot - 1,
334 "file extent end range (%llu) goes beyond start offset (%llu) of the next file extent",
335 					prev_end, key->offset);
336 			return -EUCLEAN;
337 		}
338 	}
339 
340 	return 0;
341 }
342 
343 static int check_csum_item(struct extent_buffer *leaf, struct btrfs_key *key,
344 			   int slot, struct btrfs_key *prev_key)
345 {
346 	struct btrfs_fs_info *fs_info = leaf->fs_info;
347 	u32 sectorsize = fs_info->sectorsize;
348 	const u32 csumsize = fs_info->csum_size;
349 
350 	if (unlikely(key->objectid != BTRFS_EXTENT_CSUM_OBJECTID)) {
351 		generic_err(leaf, slot,
352 		"invalid key objectid for csum item, have %llu expect %llu",
353 			key->objectid, BTRFS_EXTENT_CSUM_OBJECTID);
354 		return -EUCLEAN;
355 	}
356 	if (unlikely(!IS_ALIGNED(key->offset, sectorsize))) {
357 		generic_err(leaf, slot,
358 	"unaligned key offset for csum item, have %llu should be aligned to %u",
359 			key->offset, sectorsize);
360 		return -EUCLEAN;
361 	}
362 	if (unlikely(!IS_ALIGNED(btrfs_item_size(leaf, slot), csumsize))) {
363 		generic_err(leaf, slot,
364 	"unaligned item size for csum item, have %u should be aligned to %u",
365 			btrfs_item_size(leaf, slot), csumsize);
366 		return -EUCLEAN;
367 	}
368 	if (slot > 0 && prev_key->type == BTRFS_EXTENT_CSUM_KEY) {
369 		u64 prev_csum_end;
370 		u32 prev_item_size;
371 
372 		prev_item_size = btrfs_item_size(leaf, slot - 1);
373 		prev_csum_end = (prev_item_size / csumsize) * sectorsize;
374 		prev_csum_end += prev_key->offset;
375 		if (unlikely(prev_csum_end > key->offset)) {
376 			generic_err(leaf, slot - 1,
377 "csum end range (%llu) goes beyond the start range (%llu) of the next csum item",
378 				    prev_csum_end, key->offset);
379 			return -EUCLEAN;
380 		}
381 	}
382 	return 0;
383 }
384 
385 /* Inode item error output has the same format as dir_item_err() */
386 #define inode_item_err(eb, slot, fmt, ...)			\
387 	dir_item_err(eb, slot, fmt, __VA_ARGS__)
388 
389 static int check_inode_key(struct extent_buffer *leaf, struct btrfs_key *key,
390 			   int slot)
391 {
392 	struct btrfs_key item_key;
393 	bool is_inode_item;
394 
395 	btrfs_item_key_to_cpu(leaf, &item_key, slot);
396 	is_inode_item = (item_key.type == BTRFS_INODE_ITEM_KEY);
397 
398 	/* For XATTR_ITEM, location key should be all 0 */
399 	if (item_key.type == BTRFS_XATTR_ITEM_KEY) {
400 		if (unlikely(key->objectid != 0 || key->type != 0 ||
401 			     key->offset != 0))
402 			return -EUCLEAN;
403 		return 0;
404 	}
405 
406 	if (unlikely((key->objectid < BTRFS_FIRST_FREE_OBJECTID ||
407 		      key->objectid > BTRFS_LAST_FREE_OBJECTID) &&
408 		     key->objectid != BTRFS_ROOT_TREE_DIR_OBJECTID &&
409 		     key->objectid != BTRFS_FREE_INO_OBJECTID)) {
410 		if (is_inode_item) {
411 			generic_err(leaf, slot,
412 	"invalid key objectid: has %llu expect %llu or [%llu, %llu] or %llu",
413 				key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID,
414 				BTRFS_FIRST_FREE_OBJECTID,
415 				BTRFS_LAST_FREE_OBJECTID,
416 				BTRFS_FREE_INO_OBJECTID);
417 		} else {
418 			dir_item_err(leaf, slot,
419 "invalid location key objectid: has %llu expect %llu or [%llu, %llu] or %llu",
420 				key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID,
421 				BTRFS_FIRST_FREE_OBJECTID,
422 				BTRFS_LAST_FREE_OBJECTID,
423 				BTRFS_FREE_INO_OBJECTID);
424 		}
425 		return -EUCLEAN;
426 	}
427 	if (unlikely(key->offset != 0)) {
428 		if (is_inode_item)
429 			inode_item_err(leaf, slot,
430 				       "invalid key offset: has %llu expect 0",
431 				       key->offset);
432 		else
433 			dir_item_err(leaf, slot,
434 				"invalid location key offset:has %llu expect 0",
435 				key->offset);
436 		return -EUCLEAN;
437 	}
438 	return 0;
439 }
440 
441 static int check_root_key(struct extent_buffer *leaf, struct btrfs_key *key,
442 			  int slot)
443 {
444 	struct btrfs_key item_key;
445 	bool is_root_item;
446 
447 	btrfs_item_key_to_cpu(leaf, &item_key, slot);
448 	is_root_item = (item_key.type == BTRFS_ROOT_ITEM_KEY);
449 
450 	/*
451 	 * Bad rootid for reloc trees.
452 	 *
453 	 * Reloc trees are only for subvolume trees, other trees only need
454 	 * to be COWed to be relocated.
455 	 */
456 	if (unlikely(is_root_item && key->objectid == BTRFS_TREE_RELOC_OBJECTID &&
457 		     !is_fstree(key->offset))) {
458 		generic_err(leaf, slot,
459 		"invalid reloc tree for root %lld, root id is not a subvolume tree",
460 			    key->offset);
461 		return -EUCLEAN;
462 	}
463 
464 	/* No such tree id */
465 	if (unlikely(key->objectid == 0)) {
466 		if (is_root_item)
467 			generic_err(leaf, slot, "invalid root id 0");
468 		else
469 			dir_item_err(leaf, slot,
470 				     "invalid location key root id 0");
471 		return -EUCLEAN;
472 	}
473 
474 	/* DIR_ITEM/INDEX/INODE_REF is not allowed to point to non-fs trees */
475 	if (unlikely(!is_fstree(key->objectid) && !is_root_item)) {
476 		dir_item_err(leaf, slot,
477 		"invalid location key objectid, have %llu expect [%llu, %llu]",
478 				key->objectid, BTRFS_FIRST_FREE_OBJECTID,
479 				BTRFS_LAST_FREE_OBJECTID);
480 		return -EUCLEAN;
481 	}
482 
483 	/*
484 	 * ROOT_ITEM with non-zero offset means this is a snapshot, created at
485 	 * @offset transid.
486 	 * Furthermore, for location key in DIR_ITEM, its offset is always -1.
487 	 *
488 	 * So here we only check offset for reloc tree whose key->offset must
489 	 * be a valid tree.
490 	 */
491 	if (unlikely(key->objectid == BTRFS_TREE_RELOC_OBJECTID &&
492 		     key->offset == 0)) {
493 		generic_err(leaf, slot, "invalid root id 0 for reloc tree");
494 		return -EUCLEAN;
495 	}
496 	return 0;
497 }
498 
499 static int check_dir_item(struct extent_buffer *leaf,
500 			  struct btrfs_key *key, struct btrfs_key *prev_key,
501 			  int slot)
502 {
503 	struct btrfs_fs_info *fs_info = leaf->fs_info;
504 	struct btrfs_dir_item *di;
505 	u32 item_size = btrfs_item_size(leaf, slot);
506 	u32 cur = 0;
507 
508 	if (unlikely(!check_prev_ino(leaf, key, slot, prev_key)))
509 		return -EUCLEAN;
510 
511 	di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
512 	while (cur < item_size) {
513 		struct btrfs_key location_key;
514 		u32 name_len;
515 		u32 data_len;
516 		u32 max_name_len;
517 		u32 total_size;
518 		u32 name_hash;
519 		u8 dir_type;
520 		int ret;
521 
522 		/* header itself should not cross item boundary */
523 		if (unlikely(cur + sizeof(*di) > item_size)) {
524 			dir_item_err(leaf, slot,
525 		"dir item header crosses item boundary, have %zu boundary %u",
526 				cur + sizeof(*di), item_size);
527 			return -EUCLEAN;
528 		}
529 
530 		/* Location key check */
531 		btrfs_dir_item_key_to_cpu(leaf, di, &location_key);
532 		if (location_key.type == BTRFS_ROOT_ITEM_KEY) {
533 			ret = check_root_key(leaf, &location_key, slot);
534 			if (unlikely(ret < 0))
535 				return ret;
536 		} else if (location_key.type == BTRFS_INODE_ITEM_KEY ||
537 			   location_key.type == 0) {
538 			ret = check_inode_key(leaf, &location_key, slot);
539 			if (unlikely(ret < 0))
540 				return ret;
541 		} else {
542 			dir_item_err(leaf, slot,
543 			"invalid location key type, have %u, expect %u or %u",
544 				     location_key.type, BTRFS_ROOT_ITEM_KEY,
545 				     BTRFS_INODE_ITEM_KEY);
546 			return -EUCLEAN;
547 		}
548 
549 		/* dir type check */
550 		dir_type = btrfs_dir_ftype(leaf, di);
551 		if (unlikely(dir_type >= BTRFS_FT_MAX)) {
552 			dir_item_err(leaf, slot,
553 			"invalid dir item type, have %u expect [0, %u)",
554 				dir_type, BTRFS_FT_MAX);
555 			return -EUCLEAN;
556 		}
557 
558 		if (unlikely(key->type == BTRFS_XATTR_ITEM_KEY &&
559 			     dir_type != BTRFS_FT_XATTR)) {
560 			dir_item_err(leaf, slot,
561 		"invalid dir item type for XATTR key, have %u expect %u",
562 				dir_type, BTRFS_FT_XATTR);
563 			return -EUCLEAN;
564 		}
565 		if (unlikely(dir_type == BTRFS_FT_XATTR &&
566 			     key->type != BTRFS_XATTR_ITEM_KEY)) {
567 			dir_item_err(leaf, slot,
568 			"xattr dir type found for non-XATTR key");
569 			return -EUCLEAN;
570 		}
571 		if (dir_type == BTRFS_FT_XATTR)
572 			max_name_len = XATTR_NAME_MAX;
573 		else
574 			max_name_len = BTRFS_NAME_LEN;
575 
576 		/* Name/data length check */
577 		name_len = btrfs_dir_name_len(leaf, di);
578 		data_len = btrfs_dir_data_len(leaf, di);
579 		if (unlikely(name_len > max_name_len)) {
580 			dir_item_err(leaf, slot,
581 			"dir item name len too long, have %u max %u",
582 				name_len, max_name_len);
583 			return -EUCLEAN;
584 		}
585 		if (unlikely(name_len + data_len > BTRFS_MAX_XATTR_SIZE(fs_info))) {
586 			dir_item_err(leaf, slot,
587 			"dir item name and data len too long, have %u max %u",
588 				name_len + data_len,
589 				BTRFS_MAX_XATTR_SIZE(fs_info));
590 			return -EUCLEAN;
591 		}
592 
593 		if (unlikely(data_len && dir_type != BTRFS_FT_XATTR)) {
594 			dir_item_err(leaf, slot,
595 			"dir item with invalid data len, have %u expect 0",
596 				data_len);
597 			return -EUCLEAN;
598 		}
599 
600 		total_size = sizeof(*di) + name_len + data_len;
601 
602 		/* header and name/data should not cross item boundary */
603 		if (unlikely(cur + total_size > item_size)) {
604 			dir_item_err(leaf, slot,
605 		"dir item data crosses item boundary, have %u boundary %u",
606 				cur + total_size, item_size);
607 			return -EUCLEAN;
608 		}
609 
610 		/*
611 		 * Special check for XATTR/DIR_ITEM, as key->offset is name
612 		 * hash, should match its name
613 		 */
614 		if (key->type == BTRFS_DIR_ITEM_KEY ||
615 		    key->type == BTRFS_XATTR_ITEM_KEY) {
616 			char namebuf[max(BTRFS_NAME_LEN, XATTR_NAME_MAX)];
617 
618 			read_extent_buffer(leaf, namebuf,
619 					(unsigned long)(di + 1), name_len);
620 			name_hash = btrfs_name_hash(namebuf, name_len);
621 			if (unlikely(key->offset != name_hash)) {
622 				dir_item_err(leaf, slot,
623 		"name hash mismatch with key, have 0x%016x expect 0x%016llx",
624 					name_hash, key->offset);
625 				return -EUCLEAN;
626 			}
627 		}
628 		cur += total_size;
629 		di = (struct btrfs_dir_item *)((void *)di + total_size);
630 	}
631 	return 0;
632 }
633 
634 __printf(3, 4)
635 __cold
636 static void block_group_err(const struct extent_buffer *eb, int slot,
637 			    const char *fmt, ...)
638 {
639 	const struct btrfs_fs_info *fs_info = eb->fs_info;
640 	struct btrfs_key key;
641 	struct va_format vaf;
642 	va_list args;
643 
644 	btrfs_item_key_to_cpu(eb, &key, slot);
645 	va_start(args, fmt);
646 
647 	vaf.fmt = fmt;
648 	vaf.va = &args;
649 
650 	btrfs_crit(fs_info,
651 	"corrupt %s: root=%llu block=%llu slot=%d bg_start=%llu bg_len=%llu, %pV",
652 		btrfs_header_level(eb) == 0 ? "leaf" : "node",
653 		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
654 		key.objectid, key.offset, &vaf);
655 	va_end(args);
656 }
657 
658 static int check_block_group_item(struct extent_buffer *leaf,
659 				  struct btrfs_key *key, int slot)
660 {
661 	struct btrfs_fs_info *fs_info = leaf->fs_info;
662 	struct btrfs_block_group_item bgi;
663 	u32 item_size = btrfs_item_size(leaf, slot);
664 	u64 chunk_objectid;
665 	u64 flags;
666 	u64 type;
667 
668 	/*
669 	 * Here we don't really care about alignment since extent allocator can
670 	 * handle it.  We care more about the size.
671 	 */
672 	if (unlikely(key->offset == 0)) {
673 		block_group_err(leaf, slot,
674 				"invalid block group size 0");
675 		return -EUCLEAN;
676 	}
677 
678 	if (unlikely(item_size != sizeof(bgi))) {
679 		block_group_err(leaf, slot,
680 			"invalid item size, have %u expect %zu",
681 				item_size, sizeof(bgi));
682 		return -EUCLEAN;
683 	}
684 
685 	read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot),
686 			   sizeof(bgi));
687 	chunk_objectid = btrfs_stack_block_group_chunk_objectid(&bgi);
688 	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
689 		/*
690 		 * We don't init the nr_global_roots until we load the global
691 		 * roots, so this could be 0 at mount time.  If it's 0 we'll
692 		 * just assume we're fine, and later we'll check against our
693 		 * actual value.
694 		 */
695 		if (unlikely(fs_info->nr_global_roots &&
696 			     chunk_objectid >= fs_info->nr_global_roots)) {
697 			block_group_err(leaf, slot,
698 	"invalid block group global root id, have %llu, needs to be <= %llu",
699 					chunk_objectid,
700 					fs_info->nr_global_roots);
701 			return -EUCLEAN;
702 		}
703 	} else if (unlikely(chunk_objectid != BTRFS_FIRST_CHUNK_TREE_OBJECTID)) {
704 		block_group_err(leaf, slot,
705 		"invalid block group chunk objectid, have %llu expect %llu",
706 				btrfs_stack_block_group_chunk_objectid(&bgi),
707 				BTRFS_FIRST_CHUNK_TREE_OBJECTID);
708 		return -EUCLEAN;
709 	}
710 
711 	if (unlikely(btrfs_stack_block_group_used(&bgi) > key->offset)) {
712 		block_group_err(leaf, slot,
713 			"invalid block group used, have %llu expect [0, %llu)",
714 				btrfs_stack_block_group_used(&bgi), key->offset);
715 		return -EUCLEAN;
716 	}
717 
718 	flags = btrfs_stack_block_group_flags(&bgi);
719 	if (unlikely(hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) > 1)) {
720 		block_group_err(leaf, slot,
721 "invalid profile flags, have 0x%llx (%lu bits set) expect no more than 1 bit set",
722 			flags & BTRFS_BLOCK_GROUP_PROFILE_MASK,
723 			hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK));
724 		return -EUCLEAN;
725 	}
726 
727 	type = flags & BTRFS_BLOCK_GROUP_TYPE_MASK;
728 	if (unlikely(type != BTRFS_BLOCK_GROUP_DATA &&
729 		     type != BTRFS_BLOCK_GROUP_METADATA &&
730 		     type != BTRFS_BLOCK_GROUP_SYSTEM &&
731 		     type != (BTRFS_BLOCK_GROUP_METADATA |
732 			      BTRFS_BLOCK_GROUP_DATA))) {
733 		block_group_err(leaf, slot,
734 "invalid type, have 0x%llx (%lu bits set) expect either 0x%llx, 0x%llx, 0x%llx or 0x%llx",
735 			type, hweight64(type),
736 			BTRFS_BLOCK_GROUP_DATA, BTRFS_BLOCK_GROUP_METADATA,
737 			BTRFS_BLOCK_GROUP_SYSTEM,
738 			BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA);
739 		return -EUCLEAN;
740 	}
741 	return 0;
742 }
743 
744 __printf(4, 5)
745 __cold
746 static void chunk_err(const struct extent_buffer *leaf,
747 		      const struct btrfs_chunk *chunk, u64 logical,
748 		      const char *fmt, ...)
749 {
750 	const struct btrfs_fs_info *fs_info = leaf->fs_info;
751 	bool is_sb;
752 	struct va_format vaf;
753 	va_list args;
754 	int i;
755 	int slot = -1;
756 
757 	/* Only superblock eb is able to have such small offset */
758 	is_sb = (leaf->start == BTRFS_SUPER_INFO_OFFSET);
759 
760 	if (!is_sb) {
761 		/*
762 		 * Get the slot number by iterating through all slots, this
763 		 * would provide better readability.
764 		 */
765 		for (i = 0; i < btrfs_header_nritems(leaf); i++) {
766 			if (btrfs_item_ptr_offset(leaf, i) ==
767 					(unsigned long)chunk) {
768 				slot = i;
769 				break;
770 			}
771 		}
772 	}
773 	va_start(args, fmt);
774 	vaf.fmt = fmt;
775 	vaf.va = &args;
776 
777 	if (is_sb)
778 		btrfs_crit(fs_info,
779 		"corrupt superblock syschunk array: chunk_start=%llu, %pV",
780 			   logical, &vaf);
781 	else
782 		btrfs_crit(fs_info,
783 	"corrupt leaf: root=%llu block=%llu slot=%d chunk_start=%llu, %pV",
784 			   BTRFS_CHUNK_TREE_OBJECTID, leaf->start, slot,
785 			   logical, &vaf);
786 	va_end(args);
787 }
788 
789 /*
790  * The common chunk check which could also work on super block sys chunk array.
791  *
792  * Return -EUCLEAN if anything is corrupted.
793  * Return 0 if everything is OK.
794  */
795 int btrfs_check_chunk_valid(struct extent_buffer *leaf,
796 			    struct btrfs_chunk *chunk, u64 logical)
797 {
798 	struct btrfs_fs_info *fs_info = leaf->fs_info;
799 	u64 length;
800 	u64 chunk_end;
801 	u64 stripe_len;
802 	u16 num_stripes;
803 	u16 sub_stripes;
804 	u64 type;
805 	u64 features;
806 	bool mixed = false;
807 	int raid_index;
808 	int nparity;
809 	int ncopies;
810 
811 	length = btrfs_chunk_length(leaf, chunk);
812 	stripe_len = btrfs_chunk_stripe_len(leaf, chunk);
813 	num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
814 	sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk);
815 	type = btrfs_chunk_type(leaf, chunk);
816 	raid_index = btrfs_bg_flags_to_raid_index(type);
817 	ncopies = btrfs_raid_array[raid_index].ncopies;
818 	nparity = btrfs_raid_array[raid_index].nparity;
819 
820 	if (unlikely(!num_stripes)) {
821 		chunk_err(leaf, chunk, logical,
822 			  "invalid chunk num_stripes, have %u", num_stripes);
823 		return -EUCLEAN;
824 	}
825 	if (unlikely(num_stripes < ncopies)) {
826 		chunk_err(leaf, chunk, logical,
827 			  "invalid chunk num_stripes < ncopies, have %u < %d",
828 			  num_stripes, ncopies);
829 		return -EUCLEAN;
830 	}
831 	if (unlikely(nparity && num_stripes == nparity)) {
832 		chunk_err(leaf, chunk, logical,
833 			  "invalid chunk num_stripes == nparity, have %u == %d",
834 			  num_stripes, nparity);
835 		return -EUCLEAN;
836 	}
837 	if (unlikely(!IS_ALIGNED(logical, fs_info->sectorsize))) {
838 		chunk_err(leaf, chunk, logical,
839 		"invalid chunk logical, have %llu should aligned to %u",
840 			  logical, fs_info->sectorsize);
841 		return -EUCLEAN;
842 	}
843 	if (unlikely(btrfs_chunk_sector_size(leaf, chunk) != fs_info->sectorsize)) {
844 		chunk_err(leaf, chunk, logical,
845 			  "invalid chunk sectorsize, have %u expect %u",
846 			  btrfs_chunk_sector_size(leaf, chunk),
847 			  fs_info->sectorsize);
848 		return -EUCLEAN;
849 	}
850 	if (unlikely(!length || !IS_ALIGNED(length, fs_info->sectorsize))) {
851 		chunk_err(leaf, chunk, logical,
852 			  "invalid chunk length, have %llu", length);
853 		return -EUCLEAN;
854 	}
855 	if (unlikely(check_add_overflow(logical, length, &chunk_end))) {
856 		chunk_err(leaf, chunk, logical,
857 "invalid chunk logical start and length, have logical start %llu length %llu",
858 			  logical, length);
859 		return -EUCLEAN;
860 	}
861 	if (unlikely(!is_power_of_2(stripe_len) || stripe_len != BTRFS_STRIPE_LEN)) {
862 		chunk_err(leaf, chunk, logical,
863 			  "invalid chunk stripe length: %llu",
864 			  stripe_len);
865 		return -EUCLEAN;
866 	}
867 	/*
868 	 * We artificially limit the chunk size, so that the number of stripes
869 	 * inside a chunk can be fit into a U32.  The current limit (256G) is
870 	 * way too large for real world usage anyway, and it's also much larger
871 	 * than our existing limit (10G).
872 	 *
873 	 * Thus it should be a good way to catch obvious bitflips.
874 	 */
875 	if (unlikely(length >= btrfs_stripe_nr_to_offset(U32_MAX))) {
876 		chunk_err(leaf, chunk, logical,
877 			  "chunk length too large: have %llu limit %llu",
878 			  length, btrfs_stripe_nr_to_offset(U32_MAX));
879 		return -EUCLEAN;
880 	}
881 	if (unlikely(type & ~(BTRFS_BLOCK_GROUP_TYPE_MASK |
882 			      BTRFS_BLOCK_GROUP_PROFILE_MASK))) {
883 		chunk_err(leaf, chunk, logical,
884 			  "unrecognized chunk type: 0x%llx",
885 			  ~(BTRFS_BLOCK_GROUP_TYPE_MASK |
886 			    BTRFS_BLOCK_GROUP_PROFILE_MASK) &
887 			  btrfs_chunk_type(leaf, chunk));
888 		return -EUCLEAN;
889 	}
890 
891 	if (unlikely(!has_single_bit_set(type & BTRFS_BLOCK_GROUP_PROFILE_MASK) &&
892 		     (type & BTRFS_BLOCK_GROUP_PROFILE_MASK) != 0)) {
893 		chunk_err(leaf, chunk, logical,
894 		"invalid chunk profile flag: 0x%llx, expect 0 or 1 bit set",
895 			  type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
896 		return -EUCLEAN;
897 	}
898 	if (unlikely((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == 0)) {
899 		chunk_err(leaf, chunk, logical,
900 	"missing chunk type flag, have 0x%llx one bit must be set in 0x%llx",
901 			  type, BTRFS_BLOCK_GROUP_TYPE_MASK);
902 		return -EUCLEAN;
903 	}
904 
905 	if (unlikely((type & BTRFS_BLOCK_GROUP_SYSTEM) &&
906 		     (type & (BTRFS_BLOCK_GROUP_METADATA |
907 			      BTRFS_BLOCK_GROUP_DATA)))) {
908 		chunk_err(leaf, chunk, logical,
909 			  "system chunk with data or metadata type: 0x%llx",
910 			  type);
911 		return -EUCLEAN;
912 	}
913 
914 	features = btrfs_super_incompat_flags(fs_info->super_copy);
915 	if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)
916 		mixed = true;
917 
918 	if (!mixed) {
919 		if (unlikely((type & BTRFS_BLOCK_GROUP_METADATA) &&
920 			     (type & BTRFS_BLOCK_GROUP_DATA))) {
921 			chunk_err(leaf, chunk, logical,
922 			"mixed chunk type in non-mixed mode: 0x%llx", type);
923 			return -EUCLEAN;
924 		}
925 	}
926 
927 	if (unlikely((type & BTRFS_BLOCK_GROUP_RAID10 &&
928 		      sub_stripes != btrfs_raid_array[BTRFS_RAID_RAID10].sub_stripes) ||
929 		     (type & BTRFS_BLOCK_GROUP_RAID1 &&
930 		      num_stripes != btrfs_raid_array[BTRFS_RAID_RAID1].devs_min) ||
931 		     (type & BTRFS_BLOCK_GROUP_RAID1C3 &&
932 		      num_stripes != btrfs_raid_array[BTRFS_RAID_RAID1C3].devs_min) ||
933 		     (type & BTRFS_BLOCK_GROUP_RAID1C4 &&
934 		      num_stripes != btrfs_raid_array[BTRFS_RAID_RAID1C4].devs_min) ||
935 		     (type & BTRFS_BLOCK_GROUP_RAID5 &&
936 		      num_stripes < btrfs_raid_array[BTRFS_RAID_RAID5].devs_min) ||
937 		     (type & BTRFS_BLOCK_GROUP_RAID6 &&
938 		      num_stripes < btrfs_raid_array[BTRFS_RAID_RAID6].devs_min) ||
939 		     (type & BTRFS_BLOCK_GROUP_DUP &&
940 		      num_stripes != btrfs_raid_array[BTRFS_RAID_DUP].dev_stripes) ||
941 		     ((type & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 &&
942 		      num_stripes != btrfs_raid_array[BTRFS_RAID_SINGLE].dev_stripes))) {
943 		chunk_err(leaf, chunk, logical,
944 			"invalid num_stripes:sub_stripes %u:%u for profile %llu",
945 			num_stripes, sub_stripes,
946 			type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
947 		return -EUCLEAN;
948 	}
949 
950 	return 0;
951 }
952 
953 /*
954  * Enhanced version of chunk item checker.
955  *
956  * The common btrfs_check_chunk_valid() doesn't check item size since it needs
957  * to work on super block sys_chunk_array which doesn't have full item ptr.
958  */
959 static int check_leaf_chunk_item(struct extent_buffer *leaf,
960 				 struct btrfs_chunk *chunk,
961 				 struct btrfs_key *key, int slot)
962 {
963 	int num_stripes;
964 
965 	if (unlikely(btrfs_item_size(leaf, slot) < sizeof(struct btrfs_chunk))) {
966 		chunk_err(leaf, chunk, key->offset,
967 			"invalid chunk item size: have %u expect [%zu, %u)",
968 			btrfs_item_size(leaf, slot),
969 			sizeof(struct btrfs_chunk),
970 			BTRFS_LEAF_DATA_SIZE(leaf->fs_info));
971 		return -EUCLEAN;
972 	}
973 
974 	num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
975 	/* Let btrfs_check_chunk_valid() handle this error type */
976 	if (num_stripes == 0)
977 		goto out;
978 
979 	if (unlikely(btrfs_chunk_item_size(num_stripes) !=
980 		     btrfs_item_size(leaf, slot))) {
981 		chunk_err(leaf, chunk, key->offset,
982 			"invalid chunk item size: have %u expect %lu",
983 			btrfs_item_size(leaf, slot),
984 			btrfs_chunk_item_size(num_stripes));
985 		return -EUCLEAN;
986 	}
987 out:
988 	return btrfs_check_chunk_valid(leaf, chunk, key->offset);
989 }
990 
991 __printf(3, 4)
992 __cold
993 static void dev_item_err(const struct extent_buffer *eb, int slot,
994 			 const char *fmt, ...)
995 {
996 	struct btrfs_key key;
997 	struct va_format vaf;
998 	va_list args;
999 
1000 	btrfs_item_key_to_cpu(eb, &key, slot);
1001 	va_start(args, fmt);
1002 
1003 	vaf.fmt = fmt;
1004 	vaf.va = &args;
1005 
1006 	btrfs_crit(eb->fs_info,
1007 	"corrupt %s: root=%llu block=%llu slot=%d devid=%llu %pV",
1008 		btrfs_header_level(eb) == 0 ? "leaf" : "node",
1009 		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
1010 		key.objectid, &vaf);
1011 	va_end(args);
1012 }
1013 
1014 static int check_dev_item(struct extent_buffer *leaf,
1015 			  struct btrfs_key *key, int slot)
1016 {
1017 	struct btrfs_dev_item *ditem;
1018 	const u32 item_size = btrfs_item_size(leaf, slot);
1019 
1020 	if (unlikely(key->objectid != BTRFS_DEV_ITEMS_OBJECTID)) {
1021 		dev_item_err(leaf, slot,
1022 			     "invalid objectid: has=%llu expect=%llu",
1023 			     key->objectid, BTRFS_DEV_ITEMS_OBJECTID);
1024 		return -EUCLEAN;
1025 	}
1026 
1027 	if (unlikely(item_size != sizeof(*ditem))) {
1028 		dev_item_err(leaf, slot, "invalid item size: has %u expect %zu",
1029 			     item_size, sizeof(*ditem));
1030 		return -EUCLEAN;
1031 	}
1032 
1033 	ditem = btrfs_item_ptr(leaf, slot, struct btrfs_dev_item);
1034 	if (unlikely(btrfs_device_id(leaf, ditem) != key->offset)) {
1035 		dev_item_err(leaf, slot,
1036 			     "devid mismatch: key has=%llu item has=%llu",
1037 			     key->offset, btrfs_device_id(leaf, ditem));
1038 		return -EUCLEAN;
1039 	}
1040 
1041 	/*
1042 	 * For device total_bytes, we don't have reliable way to check it, as
1043 	 * it can be 0 for device removal. Device size check can only be done
1044 	 * by dev extents check.
1045 	 */
1046 	if (unlikely(btrfs_device_bytes_used(leaf, ditem) >
1047 		     btrfs_device_total_bytes(leaf, ditem))) {
1048 		dev_item_err(leaf, slot,
1049 			     "invalid bytes used: have %llu expect [0, %llu]",
1050 			     btrfs_device_bytes_used(leaf, ditem),
1051 			     btrfs_device_total_bytes(leaf, ditem));
1052 		return -EUCLEAN;
1053 	}
1054 	/*
1055 	 * Remaining members like io_align/type/gen/dev_group aren't really
1056 	 * utilized.  Skip them to make later usage of them easier.
1057 	 */
1058 	return 0;
1059 }
1060 
1061 static int check_inode_item(struct extent_buffer *leaf,
1062 			    struct btrfs_key *key, int slot)
1063 {
1064 	struct btrfs_fs_info *fs_info = leaf->fs_info;
1065 	struct btrfs_inode_item *iitem;
1066 	u64 super_gen = btrfs_super_generation(fs_info->super_copy);
1067 	u32 valid_mask = (S_IFMT | S_ISUID | S_ISGID | S_ISVTX | 0777);
1068 	const u32 item_size = btrfs_item_size(leaf, slot);
1069 	u32 mode;
1070 	int ret;
1071 	u32 flags;
1072 	u32 ro_flags;
1073 
1074 	ret = check_inode_key(leaf, key, slot);
1075 	if (unlikely(ret < 0))
1076 		return ret;
1077 
1078 	if (unlikely(item_size != sizeof(*iitem))) {
1079 		generic_err(leaf, slot, "invalid item size: has %u expect %zu",
1080 			    item_size, sizeof(*iitem));
1081 		return -EUCLEAN;
1082 	}
1083 
1084 	iitem = btrfs_item_ptr(leaf, slot, struct btrfs_inode_item);
1085 
1086 	/* Here we use super block generation + 1 to handle log tree */
1087 	if (unlikely(btrfs_inode_generation(leaf, iitem) > super_gen + 1)) {
1088 		inode_item_err(leaf, slot,
1089 			"invalid inode generation: has %llu expect (0, %llu]",
1090 			       btrfs_inode_generation(leaf, iitem),
1091 			       super_gen + 1);
1092 		return -EUCLEAN;
1093 	}
1094 	/* Note for ROOT_TREE_DIR_ITEM, mkfs could set its transid 0 */
1095 	if (unlikely(btrfs_inode_transid(leaf, iitem) > super_gen + 1)) {
1096 		inode_item_err(leaf, slot,
1097 			"invalid inode transid: has %llu expect [0, %llu]",
1098 			       btrfs_inode_transid(leaf, iitem), super_gen + 1);
1099 		return -EUCLEAN;
1100 	}
1101 
1102 	/*
1103 	 * For size and nbytes it's better not to be too strict, as for dir
1104 	 * item its size/nbytes can easily get wrong, but doesn't affect
1105 	 * anything in the fs. So here we skip the check.
1106 	 */
1107 	mode = btrfs_inode_mode(leaf, iitem);
1108 	if (unlikely(mode & ~valid_mask)) {
1109 		inode_item_err(leaf, slot,
1110 			       "unknown mode bit detected: 0x%x",
1111 			       mode & ~valid_mask);
1112 		return -EUCLEAN;
1113 	}
1114 
1115 	/*
1116 	 * S_IFMT is not bit mapped so we can't completely rely on
1117 	 * is_power_of_2/has_single_bit_set, but it can save us from checking
1118 	 * FIFO/CHR/DIR/REG.  Only needs to check BLK, LNK and SOCKS
1119 	 */
1120 	if (!has_single_bit_set(mode & S_IFMT)) {
1121 		if (unlikely(!S_ISLNK(mode) && !S_ISBLK(mode) && !S_ISSOCK(mode))) {
1122 			inode_item_err(leaf, slot,
1123 			"invalid mode: has 0%o expect valid S_IF* bit(s)",
1124 				       mode & S_IFMT);
1125 			return -EUCLEAN;
1126 		}
1127 	}
1128 	if (unlikely(S_ISDIR(mode) && btrfs_inode_nlink(leaf, iitem) > 1)) {
1129 		inode_item_err(leaf, slot,
1130 		       "invalid nlink: has %u expect no more than 1 for dir",
1131 			btrfs_inode_nlink(leaf, iitem));
1132 		return -EUCLEAN;
1133 	}
1134 	btrfs_inode_split_flags(btrfs_inode_flags(leaf, iitem), &flags, &ro_flags);
1135 	if (unlikely(flags & ~BTRFS_INODE_FLAG_MASK)) {
1136 		inode_item_err(leaf, slot,
1137 			       "unknown incompat flags detected: 0x%x", flags);
1138 		return -EUCLEAN;
1139 	}
1140 	if (unlikely(!sb_rdonly(fs_info->sb) &&
1141 		     (ro_flags & ~BTRFS_INODE_RO_FLAG_MASK))) {
1142 		inode_item_err(leaf, slot,
1143 			"unknown ro-compat flags detected on writeable mount: 0x%x",
1144 			ro_flags);
1145 		return -EUCLEAN;
1146 	}
1147 	return 0;
1148 }
1149 
1150 static int check_root_item(struct extent_buffer *leaf, struct btrfs_key *key,
1151 			   int slot)
1152 {
1153 	struct btrfs_fs_info *fs_info = leaf->fs_info;
1154 	struct btrfs_root_item ri = { 0 };
1155 	const u64 valid_root_flags = BTRFS_ROOT_SUBVOL_RDONLY |
1156 				     BTRFS_ROOT_SUBVOL_DEAD;
1157 	int ret;
1158 
1159 	ret = check_root_key(leaf, key, slot);
1160 	if (unlikely(ret < 0))
1161 		return ret;
1162 
1163 	if (unlikely(btrfs_item_size(leaf, slot) != sizeof(ri) &&
1164 		     btrfs_item_size(leaf, slot) !=
1165 		     btrfs_legacy_root_item_size())) {
1166 		generic_err(leaf, slot,
1167 			    "invalid root item size, have %u expect %zu or %u",
1168 			    btrfs_item_size(leaf, slot), sizeof(ri),
1169 			    btrfs_legacy_root_item_size());
1170 		return -EUCLEAN;
1171 	}
1172 
1173 	/*
1174 	 * For legacy root item, the members starting at generation_v2 will be
1175 	 * all filled with 0.
1176 	 * And since we allow geneartion_v2 as 0, it will still pass the check.
1177 	 */
1178 	read_extent_buffer(leaf, &ri, btrfs_item_ptr_offset(leaf, slot),
1179 			   btrfs_item_size(leaf, slot));
1180 
1181 	/* Generation related */
1182 	if (unlikely(btrfs_root_generation(&ri) >
1183 		     btrfs_super_generation(fs_info->super_copy) + 1)) {
1184 		generic_err(leaf, slot,
1185 			"invalid root generation, have %llu expect (0, %llu]",
1186 			    btrfs_root_generation(&ri),
1187 			    btrfs_super_generation(fs_info->super_copy) + 1);
1188 		return -EUCLEAN;
1189 	}
1190 	if (unlikely(btrfs_root_generation_v2(&ri) >
1191 		     btrfs_super_generation(fs_info->super_copy) + 1)) {
1192 		generic_err(leaf, slot,
1193 		"invalid root v2 generation, have %llu expect (0, %llu]",
1194 			    btrfs_root_generation_v2(&ri),
1195 			    btrfs_super_generation(fs_info->super_copy) + 1);
1196 		return -EUCLEAN;
1197 	}
1198 	if (unlikely(btrfs_root_last_snapshot(&ri) >
1199 		     btrfs_super_generation(fs_info->super_copy) + 1)) {
1200 		generic_err(leaf, slot,
1201 		"invalid root last_snapshot, have %llu expect (0, %llu]",
1202 			    btrfs_root_last_snapshot(&ri),
1203 			    btrfs_super_generation(fs_info->super_copy) + 1);
1204 		return -EUCLEAN;
1205 	}
1206 
1207 	/* Alignment and level check */
1208 	if (unlikely(!IS_ALIGNED(btrfs_root_bytenr(&ri), fs_info->sectorsize))) {
1209 		generic_err(leaf, slot,
1210 		"invalid root bytenr, have %llu expect to be aligned to %u",
1211 			    btrfs_root_bytenr(&ri), fs_info->sectorsize);
1212 		return -EUCLEAN;
1213 	}
1214 	if (unlikely(btrfs_root_level(&ri) >= BTRFS_MAX_LEVEL)) {
1215 		generic_err(leaf, slot,
1216 			    "invalid root level, have %u expect [0, %u]",
1217 			    btrfs_root_level(&ri), BTRFS_MAX_LEVEL - 1);
1218 		return -EUCLEAN;
1219 	}
1220 	if (unlikely(btrfs_root_drop_level(&ri) >= BTRFS_MAX_LEVEL)) {
1221 		generic_err(leaf, slot,
1222 			    "invalid root level, have %u expect [0, %u]",
1223 			    btrfs_root_drop_level(&ri), BTRFS_MAX_LEVEL - 1);
1224 		return -EUCLEAN;
1225 	}
1226 
1227 	/* Flags check */
1228 	if (unlikely(btrfs_root_flags(&ri) & ~valid_root_flags)) {
1229 		generic_err(leaf, slot,
1230 			    "invalid root flags, have 0x%llx expect mask 0x%llx",
1231 			    btrfs_root_flags(&ri), valid_root_flags);
1232 		return -EUCLEAN;
1233 	}
1234 	return 0;
1235 }
1236 
1237 __printf(3,4)
1238 __cold
1239 static void extent_err(const struct extent_buffer *eb, int slot,
1240 		       const char *fmt, ...)
1241 {
1242 	struct btrfs_key key;
1243 	struct va_format vaf;
1244 	va_list args;
1245 	u64 bytenr;
1246 	u64 len;
1247 
1248 	btrfs_item_key_to_cpu(eb, &key, slot);
1249 	bytenr = key.objectid;
1250 	if (key.type == BTRFS_METADATA_ITEM_KEY ||
1251 	    key.type == BTRFS_TREE_BLOCK_REF_KEY ||
1252 	    key.type == BTRFS_SHARED_BLOCK_REF_KEY)
1253 		len = eb->fs_info->nodesize;
1254 	else
1255 		len = key.offset;
1256 	va_start(args, fmt);
1257 
1258 	vaf.fmt = fmt;
1259 	vaf.va = &args;
1260 
1261 	btrfs_crit(eb->fs_info,
1262 	"corrupt %s: block=%llu slot=%d extent bytenr=%llu len=%llu %pV",
1263 		btrfs_header_level(eb) == 0 ? "leaf" : "node",
1264 		eb->start, slot, bytenr, len, &vaf);
1265 	va_end(args);
1266 }
1267 
1268 static int check_extent_item(struct extent_buffer *leaf,
1269 			     struct btrfs_key *key, int slot,
1270 			     struct btrfs_key *prev_key)
1271 {
1272 	struct btrfs_fs_info *fs_info = leaf->fs_info;
1273 	struct btrfs_extent_item *ei;
1274 	bool is_tree_block = false;
1275 	unsigned long ptr;	/* Current pointer inside inline refs */
1276 	unsigned long end;	/* Extent item end */
1277 	const u32 item_size = btrfs_item_size(leaf, slot);
1278 	u8 last_type = 0;
1279 	u64 last_seq = U64_MAX;
1280 	u64 flags;
1281 	u64 generation;
1282 	u64 total_refs;		/* Total refs in btrfs_extent_item */
1283 	u64 inline_refs = 0;	/* found total inline refs */
1284 
1285 	if (unlikely(key->type == BTRFS_METADATA_ITEM_KEY &&
1286 		     !btrfs_fs_incompat(fs_info, SKINNY_METADATA))) {
1287 		generic_err(leaf, slot,
1288 "invalid key type, METADATA_ITEM type invalid when SKINNY_METADATA feature disabled");
1289 		return -EUCLEAN;
1290 	}
1291 	/* key->objectid is the bytenr for both key types */
1292 	if (unlikely(!IS_ALIGNED(key->objectid, fs_info->sectorsize))) {
1293 		generic_err(leaf, slot,
1294 		"invalid key objectid, have %llu expect to be aligned to %u",
1295 			   key->objectid, fs_info->sectorsize);
1296 		return -EUCLEAN;
1297 	}
1298 
1299 	/* key->offset is tree level for METADATA_ITEM_KEY */
1300 	if (unlikely(key->type == BTRFS_METADATA_ITEM_KEY &&
1301 		     key->offset >= BTRFS_MAX_LEVEL)) {
1302 		extent_err(leaf, slot,
1303 			   "invalid tree level, have %llu expect [0, %u]",
1304 			   key->offset, BTRFS_MAX_LEVEL - 1);
1305 		return -EUCLEAN;
1306 	}
1307 
1308 	/*
1309 	 * EXTENT/METADATA_ITEM consists of:
1310 	 * 1) One btrfs_extent_item
1311 	 *    Records the total refs, type and generation of the extent.
1312 	 *
1313 	 * 2) One btrfs_tree_block_info (for EXTENT_ITEM and tree backref only)
1314 	 *    Records the first key and level of the tree block.
1315 	 *
1316 	 * 2) Zero or more btrfs_extent_inline_ref(s)
1317 	 *    Each inline ref has one btrfs_extent_inline_ref shows:
1318 	 *    2.1) The ref type, one of the 4
1319 	 *         TREE_BLOCK_REF	Tree block only
1320 	 *         SHARED_BLOCK_REF	Tree block only
1321 	 *         EXTENT_DATA_REF	Data only
1322 	 *         SHARED_DATA_REF	Data only
1323 	 *    2.2) Ref type specific data
1324 	 *         Either using btrfs_extent_inline_ref::offset, or specific
1325 	 *         data structure.
1326 	 *
1327 	 *    All above inline items should follow the order:
1328 	 *
1329 	 *    - All btrfs_extent_inline_ref::type should be in an ascending
1330 	 *      order
1331 	 *
1332 	 *    - Within the same type, the items should follow a descending
1333 	 *      order by their sequence number. The sequence number is
1334 	 *      determined by:
1335 	 *      * btrfs_extent_inline_ref::offset for all types  other than
1336 	 *        EXTENT_DATA_REF
1337 	 *      * hash_extent_data_ref() for EXTENT_DATA_REF
1338 	 */
1339 	if (unlikely(item_size < sizeof(*ei))) {
1340 		extent_err(leaf, slot,
1341 			   "invalid item size, have %u expect [%zu, %u)",
1342 			   item_size, sizeof(*ei),
1343 			   BTRFS_LEAF_DATA_SIZE(fs_info));
1344 		return -EUCLEAN;
1345 	}
1346 	end = item_size + btrfs_item_ptr_offset(leaf, slot);
1347 
1348 	/* Checks against extent_item */
1349 	ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
1350 	flags = btrfs_extent_flags(leaf, ei);
1351 	total_refs = btrfs_extent_refs(leaf, ei);
1352 	generation = btrfs_extent_generation(leaf, ei);
1353 	if (unlikely(generation >
1354 		     btrfs_super_generation(fs_info->super_copy) + 1)) {
1355 		extent_err(leaf, slot,
1356 			   "invalid generation, have %llu expect (0, %llu]",
1357 			   generation,
1358 			   btrfs_super_generation(fs_info->super_copy) + 1);
1359 		return -EUCLEAN;
1360 	}
1361 	if (unlikely(!has_single_bit_set(flags & (BTRFS_EXTENT_FLAG_DATA |
1362 						  BTRFS_EXTENT_FLAG_TREE_BLOCK)))) {
1363 		extent_err(leaf, slot,
1364 		"invalid extent flag, have 0x%llx expect 1 bit set in 0x%llx",
1365 			flags, BTRFS_EXTENT_FLAG_DATA |
1366 			BTRFS_EXTENT_FLAG_TREE_BLOCK);
1367 		return -EUCLEAN;
1368 	}
1369 	is_tree_block = !!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK);
1370 	if (is_tree_block) {
1371 		if (unlikely(key->type == BTRFS_EXTENT_ITEM_KEY &&
1372 			     key->offset != fs_info->nodesize)) {
1373 			extent_err(leaf, slot,
1374 				   "invalid extent length, have %llu expect %u",
1375 				   key->offset, fs_info->nodesize);
1376 			return -EUCLEAN;
1377 		}
1378 	} else {
1379 		if (unlikely(key->type != BTRFS_EXTENT_ITEM_KEY)) {
1380 			extent_err(leaf, slot,
1381 			"invalid key type, have %u expect %u for data backref",
1382 				   key->type, BTRFS_EXTENT_ITEM_KEY);
1383 			return -EUCLEAN;
1384 		}
1385 		if (unlikely(!IS_ALIGNED(key->offset, fs_info->sectorsize))) {
1386 			extent_err(leaf, slot,
1387 			"invalid extent length, have %llu expect aligned to %u",
1388 				   key->offset, fs_info->sectorsize);
1389 			return -EUCLEAN;
1390 		}
1391 		if (unlikely(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)) {
1392 			extent_err(leaf, slot,
1393 			"invalid extent flag, data has full backref set");
1394 			return -EUCLEAN;
1395 		}
1396 	}
1397 	ptr = (unsigned long)(struct btrfs_extent_item *)(ei + 1);
1398 
1399 	/* Check the special case of btrfs_tree_block_info */
1400 	if (is_tree_block && key->type != BTRFS_METADATA_ITEM_KEY) {
1401 		struct btrfs_tree_block_info *info;
1402 
1403 		info = (struct btrfs_tree_block_info *)ptr;
1404 		if (unlikely(btrfs_tree_block_level(leaf, info) >= BTRFS_MAX_LEVEL)) {
1405 			extent_err(leaf, slot,
1406 			"invalid tree block info level, have %u expect [0, %u]",
1407 				   btrfs_tree_block_level(leaf, info),
1408 				   BTRFS_MAX_LEVEL - 1);
1409 			return -EUCLEAN;
1410 		}
1411 		ptr = (unsigned long)(struct btrfs_tree_block_info *)(info + 1);
1412 	}
1413 
1414 	/* Check inline refs */
1415 	while (ptr < end) {
1416 		struct btrfs_extent_inline_ref *iref;
1417 		struct btrfs_extent_data_ref *dref;
1418 		struct btrfs_shared_data_ref *sref;
1419 		u64 seq;
1420 		u64 dref_offset;
1421 		u64 inline_offset;
1422 		u8 inline_type;
1423 
1424 		if (unlikely(ptr + sizeof(*iref) > end)) {
1425 			extent_err(leaf, slot,
1426 "inline ref item overflows extent item, ptr %lu iref size %zu end %lu",
1427 				   ptr, sizeof(*iref), end);
1428 			return -EUCLEAN;
1429 		}
1430 		iref = (struct btrfs_extent_inline_ref *)ptr;
1431 		inline_type = btrfs_extent_inline_ref_type(leaf, iref);
1432 		inline_offset = btrfs_extent_inline_ref_offset(leaf, iref);
1433 		seq = inline_offset;
1434 		if (unlikely(ptr + btrfs_extent_inline_ref_size(inline_type) > end)) {
1435 			extent_err(leaf, slot,
1436 "inline ref item overflows extent item, ptr %lu iref size %u end %lu",
1437 				   ptr, btrfs_extent_inline_ref_size(inline_type), end);
1438 			return -EUCLEAN;
1439 		}
1440 
1441 		switch (inline_type) {
1442 		/* inline_offset is subvolid of the owner, no need to check */
1443 		case BTRFS_TREE_BLOCK_REF_KEY:
1444 			inline_refs++;
1445 			break;
1446 		/* Contains parent bytenr */
1447 		case BTRFS_SHARED_BLOCK_REF_KEY:
1448 			if (unlikely(!IS_ALIGNED(inline_offset,
1449 						 fs_info->sectorsize))) {
1450 				extent_err(leaf, slot,
1451 		"invalid tree parent bytenr, have %llu expect aligned to %u",
1452 					   inline_offset, fs_info->sectorsize);
1453 				return -EUCLEAN;
1454 			}
1455 			inline_refs++;
1456 			break;
1457 		/*
1458 		 * Contains owner subvolid, owner key objectid, adjusted offset.
1459 		 * The only obvious corruption can happen in that offset.
1460 		 */
1461 		case BTRFS_EXTENT_DATA_REF_KEY:
1462 			dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1463 			dref_offset = btrfs_extent_data_ref_offset(leaf, dref);
1464 			seq = hash_extent_data_ref(
1465 					btrfs_extent_data_ref_root(leaf, dref),
1466 					btrfs_extent_data_ref_objectid(leaf, dref),
1467 					btrfs_extent_data_ref_offset(leaf, dref));
1468 			if (unlikely(!IS_ALIGNED(dref_offset,
1469 						 fs_info->sectorsize))) {
1470 				extent_err(leaf, slot,
1471 		"invalid data ref offset, have %llu expect aligned to %u",
1472 					   dref_offset, fs_info->sectorsize);
1473 				return -EUCLEAN;
1474 			}
1475 			inline_refs += btrfs_extent_data_ref_count(leaf, dref);
1476 			break;
1477 		/* Contains parent bytenr and ref count */
1478 		case BTRFS_SHARED_DATA_REF_KEY:
1479 			sref = (struct btrfs_shared_data_ref *)(iref + 1);
1480 			if (unlikely(!IS_ALIGNED(inline_offset,
1481 						 fs_info->sectorsize))) {
1482 				extent_err(leaf, slot,
1483 		"invalid data parent bytenr, have %llu expect aligned to %u",
1484 					   inline_offset, fs_info->sectorsize);
1485 				return -EUCLEAN;
1486 			}
1487 			inline_refs += btrfs_shared_data_ref_count(leaf, sref);
1488 			break;
1489 		default:
1490 			extent_err(leaf, slot, "unknown inline ref type: %u",
1491 				   inline_type);
1492 			return -EUCLEAN;
1493 		}
1494 		if (inline_type < last_type) {
1495 			extent_err(leaf, slot,
1496 				   "inline ref out-of-order: has type %u, prev type %u",
1497 				   inline_type, last_type);
1498 			return -EUCLEAN;
1499 		}
1500 		/* Type changed, allow the sequence starts from U64_MAX again. */
1501 		if (inline_type > last_type)
1502 			last_seq = U64_MAX;
1503 		if (seq > last_seq) {
1504 			extent_err(leaf, slot,
1505 "inline ref out-of-order: has type %u offset %llu seq 0x%llx, prev type %u seq 0x%llx",
1506 				   inline_type, inline_offset, seq,
1507 				   last_type, last_seq);
1508 			return -EUCLEAN;
1509 		}
1510 		last_type = inline_type;
1511 		last_seq = seq;
1512 		ptr += btrfs_extent_inline_ref_size(inline_type);
1513 	}
1514 	/* No padding is allowed */
1515 	if (unlikely(ptr != end)) {
1516 		extent_err(leaf, slot,
1517 			   "invalid extent item size, padding bytes found");
1518 		return -EUCLEAN;
1519 	}
1520 
1521 	/* Finally, check the inline refs against total refs */
1522 	if (unlikely(inline_refs > total_refs)) {
1523 		extent_err(leaf, slot,
1524 			"invalid extent refs, have %llu expect >= inline %llu",
1525 			   total_refs, inline_refs);
1526 		return -EUCLEAN;
1527 	}
1528 
1529 	if ((prev_key->type == BTRFS_EXTENT_ITEM_KEY) ||
1530 	    (prev_key->type == BTRFS_METADATA_ITEM_KEY)) {
1531 		u64 prev_end = prev_key->objectid;
1532 
1533 		if (prev_key->type == BTRFS_METADATA_ITEM_KEY)
1534 			prev_end += fs_info->nodesize;
1535 		else
1536 			prev_end += prev_key->offset;
1537 
1538 		if (unlikely(prev_end > key->objectid)) {
1539 			extent_err(leaf, slot,
1540 	"previous extent [%llu %u %llu] overlaps current extent [%llu %u %llu]",
1541 				   prev_key->objectid, prev_key->type,
1542 				   prev_key->offset, key->objectid, key->type,
1543 				   key->offset);
1544 			return -EUCLEAN;
1545 		}
1546 	}
1547 
1548 	return 0;
1549 }
1550 
1551 static int check_simple_keyed_refs(struct extent_buffer *leaf,
1552 				   struct btrfs_key *key, int slot)
1553 {
1554 	u32 expect_item_size = 0;
1555 
1556 	if (key->type == BTRFS_SHARED_DATA_REF_KEY)
1557 		expect_item_size = sizeof(struct btrfs_shared_data_ref);
1558 
1559 	if (unlikely(btrfs_item_size(leaf, slot) != expect_item_size)) {
1560 		generic_err(leaf, slot,
1561 		"invalid item size, have %u expect %u for key type %u",
1562 			    btrfs_item_size(leaf, slot),
1563 			    expect_item_size, key->type);
1564 		return -EUCLEAN;
1565 	}
1566 	if (unlikely(!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize))) {
1567 		generic_err(leaf, slot,
1568 "invalid key objectid for shared block ref, have %llu expect aligned to %u",
1569 			    key->objectid, leaf->fs_info->sectorsize);
1570 		return -EUCLEAN;
1571 	}
1572 	if (unlikely(key->type != BTRFS_TREE_BLOCK_REF_KEY &&
1573 		     !IS_ALIGNED(key->offset, leaf->fs_info->sectorsize))) {
1574 		extent_err(leaf, slot,
1575 		"invalid tree parent bytenr, have %llu expect aligned to %u",
1576 			   key->offset, leaf->fs_info->sectorsize);
1577 		return -EUCLEAN;
1578 	}
1579 	return 0;
1580 }
1581 
1582 static int check_extent_data_ref(struct extent_buffer *leaf,
1583 				 struct btrfs_key *key, int slot)
1584 {
1585 	struct btrfs_extent_data_ref *dref;
1586 	unsigned long ptr = btrfs_item_ptr_offset(leaf, slot);
1587 	const unsigned long end = ptr + btrfs_item_size(leaf, slot);
1588 
1589 	if (unlikely(btrfs_item_size(leaf, slot) % sizeof(*dref) != 0)) {
1590 		generic_err(leaf, slot,
1591 	"invalid item size, have %u expect aligned to %zu for key type %u",
1592 			    btrfs_item_size(leaf, slot),
1593 			    sizeof(*dref), key->type);
1594 		return -EUCLEAN;
1595 	}
1596 	if (unlikely(!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize))) {
1597 		generic_err(leaf, slot,
1598 "invalid key objectid for shared block ref, have %llu expect aligned to %u",
1599 			    key->objectid, leaf->fs_info->sectorsize);
1600 		return -EUCLEAN;
1601 	}
1602 	for (; ptr < end; ptr += sizeof(*dref)) {
1603 		u64 offset;
1604 
1605 		/*
1606 		 * We cannot check the extent_data_ref hash due to possible
1607 		 * overflow from the leaf due to hash collisions.
1608 		 */
1609 		dref = (struct btrfs_extent_data_ref *)ptr;
1610 		offset = btrfs_extent_data_ref_offset(leaf, dref);
1611 		if (unlikely(!IS_ALIGNED(offset, leaf->fs_info->sectorsize))) {
1612 			extent_err(leaf, slot,
1613 	"invalid extent data backref offset, have %llu expect aligned to %u",
1614 				   offset, leaf->fs_info->sectorsize);
1615 			return -EUCLEAN;
1616 		}
1617 	}
1618 	return 0;
1619 }
1620 
1621 #define inode_ref_err(eb, slot, fmt, args...)			\
1622 	inode_item_err(eb, slot, fmt, ##args)
1623 static int check_inode_ref(struct extent_buffer *leaf,
1624 			   struct btrfs_key *key, struct btrfs_key *prev_key,
1625 			   int slot)
1626 {
1627 	struct btrfs_inode_ref *iref;
1628 	unsigned long ptr;
1629 	unsigned long end;
1630 
1631 	if (unlikely(!check_prev_ino(leaf, key, slot, prev_key)))
1632 		return -EUCLEAN;
1633 	/* namelen can't be 0, so item_size == sizeof() is also invalid */
1634 	if (unlikely(btrfs_item_size(leaf, slot) <= sizeof(*iref))) {
1635 		inode_ref_err(leaf, slot,
1636 			"invalid item size, have %u expect (%zu, %u)",
1637 			btrfs_item_size(leaf, slot),
1638 			sizeof(*iref), BTRFS_LEAF_DATA_SIZE(leaf->fs_info));
1639 		return -EUCLEAN;
1640 	}
1641 
1642 	ptr = btrfs_item_ptr_offset(leaf, slot);
1643 	end = ptr + btrfs_item_size(leaf, slot);
1644 	while (ptr < end) {
1645 		u16 namelen;
1646 
1647 		if (unlikely(ptr + sizeof(iref) > end)) {
1648 			inode_ref_err(leaf, slot,
1649 			"inode ref overflow, ptr %lu end %lu inode_ref_size %zu",
1650 				ptr, end, sizeof(iref));
1651 			return -EUCLEAN;
1652 		}
1653 
1654 		iref = (struct btrfs_inode_ref *)ptr;
1655 		namelen = btrfs_inode_ref_name_len(leaf, iref);
1656 		if (unlikely(ptr + sizeof(*iref) + namelen > end)) {
1657 			inode_ref_err(leaf, slot,
1658 				"inode ref overflow, ptr %lu end %lu namelen %u",
1659 				ptr, end, namelen);
1660 			return -EUCLEAN;
1661 		}
1662 
1663 		/*
1664 		 * NOTE: In theory we should record all found index numbers
1665 		 * to find any duplicated indexes, but that will be too time
1666 		 * consuming for inodes with too many hard links.
1667 		 */
1668 		ptr += sizeof(*iref) + namelen;
1669 	}
1670 	return 0;
1671 }
1672 
1673 /*
1674  * Common point to switch the item-specific validation.
1675  */
1676 static enum btrfs_tree_block_status check_leaf_item(struct extent_buffer *leaf,
1677 						    struct btrfs_key *key,
1678 						    int slot,
1679 						    struct btrfs_key *prev_key)
1680 {
1681 	int ret = 0;
1682 	struct btrfs_chunk *chunk;
1683 
1684 	switch (key->type) {
1685 	case BTRFS_EXTENT_DATA_KEY:
1686 		ret = check_extent_data_item(leaf, key, slot, prev_key);
1687 		break;
1688 	case BTRFS_EXTENT_CSUM_KEY:
1689 		ret = check_csum_item(leaf, key, slot, prev_key);
1690 		break;
1691 	case BTRFS_DIR_ITEM_KEY:
1692 	case BTRFS_DIR_INDEX_KEY:
1693 	case BTRFS_XATTR_ITEM_KEY:
1694 		ret = check_dir_item(leaf, key, prev_key, slot);
1695 		break;
1696 	case BTRFS_INODE_REF_KEY:
1697 		ret = check_inode_ref(leaf, key, prev_key, slot);
1698 		break;
1699 	case BTRFS_BLOCK_GROUP_ITEM_KEY:
1700 		ret = check_block_group_item(leaf, key, slot);
1701 		break;
1702 	case BTRFS_CHUNK_ITEM_KEY:
1703 		chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
1704 		ret = check_leaf_chunk_item(leaf, chunk, key, slot);
1705 		break;
1706 	case BTRFS_DEV_ITEM_KEY:
1707 		ret = check_dev_item(leaf, key, slot);
1708 		break;
1709 	case BTRFS_INODE_ITEM_KEY:
1710 		ret = check_inode_item(leaf, key, slot);
1711 		break;
1712 	case BTRFS_ROOT_ITEM_KEY:
1713 		ret = check_root_item(leaf, key, slot);
1714 		break;
1715 	case BTRFS_EXTENT_ITEM_KEY:
1716 	case BTRFS_METADATA_ITEM_KEY:
1717 		ret = check_extent_item(leaf, key, slot, prev_key);
1718 		break;
1719 	case BTRFS_TREE_BLOCK_REF_KEY:
1720 	case BTRFS_SHARED_DATA_REF_KEY:
1721 	case BTRFS_SHARED_BLOCK_REF_KEY:
1722 		ret = check_simple_keyed_refs(leaf, key, slot);
1723 		break;
1724 	case BTRFS_EXTENT_DATA_REF_KEY:
1725 		ret = check_extent_data_ref(leaf, key, slot);
1726 		break;
1727 	}
1728 
1729 	if (ret)
1730 		return BTRFS_TREE_BLOCK_INVALID_ITEM;
1731 	return BTRFS_TREE_BLOCK_CLEAN;
1732 }
1733 
1734 enum btrfs_tree_block_status __btrfs_check_leaf(struct extent_buffer *leaf)
1735 {
1736 	struct btrfs_fs_info *fs_info = leaf->fs_info;
1737 	/* No valid key type is 0, so all key should be larger than this key */
1738 	struct btrfs_key prev_key = {0, 0, 0};
1739 	struct btrfs_key key;
1740 	u32 nritems = btrfs_header_nritems(leaf);
1741 	int slot;
1742 
1743 	if (unlikely(btrfs_header_level(leaf) != 0)) {
1744 		generic_err(leaf, 0,
1745 			"invalid level for leaf, have %d expect 0",
1746 			btrfs_header_level(leaf));
1747 		return BTRFS_TREE_BLOCK_INVALID_LEVEL;
1748 	}
1749 
1750 	/*
1751 	 * Extent buffers from a relocation tree have a owner field that
1752 	 * corresponds to the subvolume tree they are based on. So just from an
1753 	 * extent buffer alone we can not find out what is the id of the
1754 	 * corresponding subvolume tree, so we can not figure out if the extent
1755 	 * buffer corresponds to the root of the relocation tree or not. So
1756 	 * skip this check for relocation trees.
1757 	 */
1758 	if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) {
1759 		u64 owner = btrfs_header_owner(leaf);
1760 
1761 		/* These trees must never be empty */
1762 		if (unlikely(owner == BTRFS_ROOT_TREE_OBJECTID ||
1763 			     owner == BTRFS_CHUNK_TREE_OBJECTID ||
1764 			     owner == BTRFS_DEV_TREE_OBJECTID ||
1765 			     owner == BTRFS_FS_TREE_OBJECTID ||
1766 			     owner == BTRFS_DATA_RELOC_TREE_OBJECTID)) {
1767 			generic_err(leaf, 0,
1768 			"invalid root, root %llu must never be empty",
1769 				    owner);
1770 			return BTRFS_TREE_BLOCK_INVALID_NRITEMS;
1771 		}
1772 
1773 		/* Unknown tree */
1774 		if (unlikely(owner == 0)) {
1775 			generic_err(leaf, 0,
1776 				"invalid owner, root 0 is not defined");
1777 			return BTRFS_TREE_BLOCK_INVALID_OWNER;
1778 		}
1779 
1780 		/* EXTENT_TREE_V2 can have empty extent trees. */
1781 		if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2))
1782 			return BTRFS_TREE_BLOCK_CLEAN;
1783 
1784 		if (unlikely(owner == BTRFS_EXTENT_TREE_OBJECTID)) {
1785 			generic_err(leaf, 0,
1786 			"invalid root, root %llu must never be empty",
1787 				    owner);
1788 			return BTRFS_TREE_BLOCK_INVALID_NRITEMS;
1789 		}
1790 
1791 		return BTRFS_TREE_BLOCK_CLEAN;
1792 	}
1793 
1794 	if (unlikely(nritems == 0))
1795 		return BTRFS_TREE_BLOCK_CLEAN;
1796 
1797 	/*
1798 	 * Check the following things to make sure this is a good leaf, and
1799 	 * leaf users won't need to bother with similar sanity checks:
1800 	 *
1801 	 * 1) key ordering
1802 	 * 2) item offset and size
1803 	 *    No overlap, no hole, all inside the leaf.
1804 	 * 3) item content
1805 	 *    If possible, do comprehensive sanity check.
1806 	 *    NOTE: All checks must only rely on the item data itself.
1807 	 */
1808 	for (slot = 0; slot < nritems; slot++) {
1809 		u32 item_end_expected;
1810 		u64 item_data_end;
1811 
1812 		btrfs_item_key_to_cpu(leaf, &key, slot);
1813 
1814 		/* Make sure the keys are in the right order */
1815 		if (unlikely(btrfs_comp_cpu_keys(&prev_key, &key) >= 0)) {
1816 			generic_err(leaf, slot,
1817 	"bad key order, prev (%llu %u %llu) current (%llu %u %llu)",
1818 				prev_key.objectid, prev_key.type,
1819 				prev_key.offset, key.objectid, key.type,
1820 				key.offset);
1821 			return BTRFS_TREE_BLOCK_BAD_KEY_ORDER;
1822 		}
1823 
1824 		item_data_end = (u64)btrfs_item_offset(leaf, slot) +
1825 				btrfs_item_size(leaf, slot);
1826 		/*
1827 		 * Make sure the offset and ends are right, remember that the
1828 		 * item data starts at the end of the leaf and grows towards the
1829 		 * front.
1830 		 */
1831 		if (slot == 0)
1832 			item_end_expected = BTRFS_LEAF_DATA_SIZE(fs_info);
1833 		else
1834 			item_end_expected = btrfs_item_offset(leaf,
1835 								 slot - 1);
1836 		if (unlikely(item_data_end != item_end_expected)) {
1837 			generic_err(leaf, slot,
1838 				"unexpected item end, have %llu expect %u",
1839 				item_data_end, item_end_expected);
1840 			return BTRFS_TREE_BLOCK_INVALID_OFFSETS;
1841 		}
1842 
1843 		/*
1844 		 * Check to make sure that we don't point outside of the leaf,
1845 		 * just in case all the items are consistent to each other, but
1846 		 * all point outside of the leaf.
1847 		 */
1848 		if (unlikely(item_data_end > BTRFS_LEAF_DATA_SIZE(fs_info))) {
1849 			generic_err(leaf, slot,
1850 			"slot end outside of leaf, have %llu expect range [0, %u]",
1851 				item_data_end, BTRFS_LEAF_DATA_SIZE(fs_info));
1852 			return BTRFS_TREE_BLOCK_INVALID_OFFSETS;
1853 		}
1854 
1855 		/* Also check if the item pointer overlaps with btrfs item. */
1856 		if (unlikely(btrfs_item_ptr_offset(leaf, slot) <
1857 			     btrfs_item_nr_offset(leaf, slot) + sizeof(struct btrfs_item))) {
1858 			generic_err(leaf, slot,
1859 		"slot overlaps with its data, item end %lu data start %lu",
1860 				btrfs_item_nr_offset(leaf, slot) +
1861 				sizeof(struct btrfs_item),
1862 				btrfs_item_ptr_offset(leaf, slot));
1863 			return BTRFS_TREE_BLOCK_INVALID_OFFSETS;
1864 		}
1865 
1866 		/*
1867 		 * We only want to do this if WRITTEN is set, otherwise the leaf
1868 		 * may be in some intermediate state and won't appear valid.
1869 		 */
1870 		if (btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_WRITTEN)) {
1871 			enum btrfs_tree_block_status ret;
1872 
1873 			/*
1874 			 * Check if the item size and content meet other
1875 			 * criteria
1876 			 */
1877 			ret = check_leaf_item(leaf, &key, slot, &prev_key);
1878 			if (unlikely(ret != BTRFS_TREE_BLOCK_CLEAN))
1879 				return ret;
1880 		}
1881 
1882 		prev_key.objectid = key.objectid;
1883 		prev_key.type = key.type;
1884 		prev_key.offset = key.offset;
1885 	}
1886 
1887 	return BTRFS_TREE_BLOCK_CLEAN;
1888 }
1889 
1890 int btrfs_check_leaf(struct extent_buffer *leaf)
1891 {
1892 	enum btrfs_tree_block_status ret;
1893 
1894 	ret = __btrfs_check_leaf(leaf);
1895 	if (unlikely(ret != BTRFS_TREE_BLOCK_CLEAN))
1896 		return -EUCLEAN;
1897 	return 0;
1898 }
1899 ALLOW_ERROR_INJECTION(btrfs_check_leaf, ERRNO);
1900 
1901 enum btrfs_tree_block_status __btrfs_check_node(struct extent_buffer *node)
1902 {
1903 	struct btrfs_fs_info *fs_info = node->fs_info;
1904 	unsigned long nr = btrfs_header_nritems(node);
1905 	struct btrfs_key key, next_key;
1906 	int slot;
1907 	int level = btrfs_header_level(node);
1908 	u64 bytenr;
1909 
1910 	if (unlikely(level <= 0 || level >= BTRFS_MAX_LEVEL)) {
1911 		generic_err(node, 0,
1912 			"invalid level for node, have %d expect [1, %d]",
1913 			level, BTRFS_MAX_LEVEL - 1);
1914 		return BTRFS_TREE_BLOCK_INVALID_LEVEL;
1915 	}
1916 	if (unlikely(nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(fs_info))) {
1917 		btrfs_crit(fs_info,
1918 "corrupt node: root=%llu block=%llu, nritems too %s, have %lu expect range [1,%u]",
1919 			   btrfs_header_owner(node), node->start,
1920 			   nr == 0 ? "small" : "large", nr,
1921 			   BTRFS_NODEPTRS_PER_BLOCK(fs_info));
1922 		return BTRFS_TREE_BLOCK_INVALID_NRITEMS;
1923 	}
1924 
1925 	for (slot = 0; slot < nr - 1; slot++) {
1926 		bytenr = btrfs_node_blockptr(node, slot);
1927 		btrfs_node_key_to_cpu(node, &key, slot);
1928 		btrfs_node_key_to_cpu(node, &next_key, slot + 1);
1929 
1930 		if (unlikely(!bytenr)) {
1931 			generic_err(node, slot,
1932 				"invalid NULL node pointer");
1933 			return BTRFS_TREE_BLOCK_INVALID_BLOCKPTR;
1934 		}
1935 		if (unlikely(!IS_ALIGNED(bytenr, fs_info->sectorsize))) {
1936 			generic_err(node, slot,
1937 			"unaligned pointer, have %llu should be aligned to %u",
1938 				bytenr, fs_info->sectorsize);
1939 			return BTRFS_TREE_BLOCK_INVALID_BLOCKPTR;
1940 		}
1941 
1942 		if (unlikely(btrfs_comp_cpu_keys(&key, &next_key) >= 0)) {
1943 			generic_err(node, slot,
1944 	"bad key order, current (%llu %u %llu) next (%llu %u %llu)",
1945 				key.objectid, key.type, key.offset,
1946 				next_key.objectid, next_key.type,
1947 				next_key.offset);
1948 			return BTRFS_TREE_BLOCK_BAD_KEY_ORDER;
1949 		}
1950 	}
1951 	return BTRFS_TREE_BLOCK_CLEAN;
1952 }
1953 
1954 int btrfs_check_node(struct extent_buffer *node)
1955 {
1956 	enum btrfs_tree_block_status ret;
1957 
1958 	ret = __btrfs_check_node(node);
1959 	if (unlikely(ret != BTRFS_TREE_BLOCK_CLEAN))
1960 		return -EUCLEAN;
1961 	return 0;
1962 }
1963 ALLOW_ERROR_INJECTION(btrfs_check_node, ERRNO);
1964 
1965 int btrfs_check_eb_owner(const struct extent_buffer *eb, u64 root_owner)
1966 {
1967 	const bool is_subvol = is_fstree(root_owner);
1968 	const u64 eb_owner = btrfs_header_owner(eb);
1969 
1970 	/*
1971 	 * Skip dummy fs, as selftests don't create unique ebs for each dummy
1972 	 * root.
1973 	 */
1974 	if (test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &eb->fs_info->fs_state))
1975 		return 0;
1976 	/*
1977 	 * There are several call sites (backref walking, qgroup, and data
1978 	 * reloc) passing 0 as @root_owner, as they are not holding the
1979 	 * tree root.  In that case, we can not do a reliable ownership check,
1980 	 * so just exit.
1981 	 */
1982 	if (root_owner == 0)
1983 		return 0;
1984 	/*
1985 	 * These trees use key.offset as their owner, our callers don't have
1986 	 * the extra capacity to pass key.offset here.  So we just skip them.
1987 	 */
1988 	if (root_owner == BTRFS_TREE_LOG_OBJECTID ||
1989 	    root_owner == BTRFS_TREE_RELOC_OBJECTID)
1990 		return 0;
1991 
1992 	if (!is_subvol) {
1993 		/* For non-subvolume trees, the eb owner should match root owner */
1994 		if (unlikely(root_owner != eb_owner)) {
1995 			btrfs_crit(eb->fs_info,
1996 "corrupted %s, root=%llu block=%llu owner mismatch, have %llu expect %llu",
1997 				btrfs_header_level(eb) == 0 ? "leaf" : "node",
1998 				root_owner, btrfs_header_bytenr(eb), eb_owner,
1999 				root_owner);
2000 			return -EUCLEAN;
2001 		}
2002 		return 0;
2003 	}
2004 
2005 	/*
2006 	 * For subvolume trees, owners can mismatch, but they should all belong
2007 	 * to subvolume trees.
2008 	 */
2009 	if (unlikely(is_subvol != is_fstree(eb_owner))) {
2010 		btrfs_crit(eb->fs_info,
2011 "corrupted %s, root=%llu block=%llu owner mismatch, have %llu expect [%llu, %llu]",
2012 			btrfs_header_level(eb) == 0 ? "leaf" : "node",
2013 			root_owner, btrfs_header_bytenr(eb), eb_owner,
2014 			BTRFS_FIRST_FREE_OBJECTID, BTRFS_LAST_FREE_OBJECTID);
2015 		return -EUCLEAN;
2016 	}
2017 	return 0;
2018 }
2019 
2020 int btrfs_verify_level_key(struct extent_buffer *eb, int level,
2021 			   struct btrfs_key *first_key, u64 parent_transid)
2022 {
2023 	struct btrfs_fs_info *fs_info = eb->fs_info;
2024 	int found_level;
2025 	struct btrfs_key found_key;
2026 	int ret;
2027 
2028 	found_level = btrfs_header_level(eb);
2029 	if (found_level != level) {
2030 		WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG),
2031 		     KERN_ERR "BTRFS: tree level check failed\n");
2032 		btrfs_err(fs_info,
2033 "tree level mismatch detected, bytenr=%llu level expected=%u has=%u",
2034 			  eb->start, level, found_level);
2035 		return -EIO;
2036 	}
2037 
2038 	if (!first_key)
2039 		return 0;
2040 
2041 	/*
2042 	 * For live tree block (new tree blocks in current transaction),
2043 	 * we need proper lock context to avoid race, which is impossible here.
2044 	 * So we only checks tree blocks which is read from disk, whose
2045 	 * generation <= fs_info->last_trans_committed.
2046 	 */
2047 	if (btrfs_header_generation(eb) > fs_info->last_trans_committed)
2048 		return 0;
2049 
2050 	/* We have @first_key, so this @eb must have at least one item */
2051 	if (btrfs_header_nritems(eb) == 0) {
2052 		btrfs_err(fs_info,
2053 		"invalid tree nritems, bytenr=%llu nritems=0 expect >0",
2054 			  eb->start);
2055 		WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG));
2056 		return -EUCLEAN;
2057 	}
2058 
2059 	if (found_level)
2060 		btrfs_node_key_to_cpu(eb, &found_key, 0);
2061 	else
2062 		btrfs_item_key_to_cpu(eb, &found_key, 0);
2063 	ret = btrfs_comp_cpu_keys(first_key, &found_key);
2064 
2065 	if (ret) {
2066 		WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG),
2067 		     KERN_ERR "BTRFS: tree first key check failed\n");
2068 		btrfs_err(fs_info,
2069 "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)",
2070 			  eb->start, parent_transid, first_key->objectid,
2071 			  first_key->type, first_key->offset,
2072 			  found_key.objectid, found_key.type,
2073 			  found_key.offset);
2074 	}
2075 	return ret;
2076 }
2077