tree-checker.c revision 0ac6e06b
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 "ctree.h"
22#include "tree-checker.h"
23#include "disk-io.h"
24#include "compression.h"
25#include "volumes.h"
26#include "misc.h"
27
28/*
29 * Error message should follow the following format:
30 * corrupt <type>: <identifier>, <reason>[, <bad_value>]
31 *
32 * @type:	leaf or node
33 * @identifier:	the necessary info to locate the leaf/node.
34 * 		It's recommended to decode key.objecitd/offset if it's
35 * 		meaningful.
36 * @reason:	describe the error
37 * @bad_value:	optional, it's recommended to output bad value and its
38 *		expected value (range).
39 *
40 * Since comma is used to separate the components, only space is allowed
41 * inside each component.
42 */
43
44/*
45 * Append generic "corrupt leaf/node root=%llu block=%llu slot=%d: " to @fmt.
46 * Allows callers to customize the output.
47 */
48__printf(3, 4)
49__cold
50static void generic_err(const struct extent_buffer *eb, int slot,
51			const char *fmt, ...)
52{
53	const struct btrfs_fs_info *fs_info = eb->fs_info;
54	struct va_format vaf;
55	va_list args;
56
57	va_start(args, fmt);
58
59	vaf.fmt = fmt;
60	vaf.va = &args;
61
62	btrfs_crit(fs_info,
63		"corrupt %s: root=%llu block=%llu slot=%d, %pV",
64		btrfs_header_level(eb) == 0 ? "leaf" : "node",
65		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, &vaf);
66	va_end(args);
67}
68
69/*
70 * Customized reporter for extent data item, since its key objectid and
71 * offset has its own meaning.
72 */
73__printf(3, 4)
74__cold
75static void file_extent_err(const struct extent_buffer *eb, int slot,
76			    const char *fmt, ...)
77{
78	const struct btrfs_fs_info *fs_info = eb->fs_info;
79	struct btrfs_key key;
80	struct va_format vaf;
81	va_list args;
82
83	btrfs_item_key_to_cpu(eb, &key, slot);
84	va_start(args, fmt);
85
86	vaf.fmt = fmt;
87	vaf.va = &args;
88
89	btrfs_crit(fs_info,
90	"corrupt %s: root=%llu block=%llu slot=%d ino=%llu file_offset=%llu, %pV",
91		btrfs_header_level(eb) == 0 ? "leaf" : "node",
92		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
93		key.objectid, key.offset, &vaf);
94	va_end(args);
95}
96
97/*
98 * Return 0 if the btrfs_file_extent_##name is aligned to @alignment
99 * Else return 1
100 */
101#define CHECK_FE_ALIGNED(leaf, slot, fi, name, alignment)		      \
102({									      \
103	if (unlikely(!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)),      \
104				 (alignment))))				      \
105		file_extent_err((leaf), (slot),				      \
106	"invalid %s for file extent, have %llu, should be aligned to %u",     \
107			(#name), btrfs_file_extent_##name((leaf), (fi)),      \
108			(alignment));					      \
109	(!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment)));   \
110})
111
112static u64 file_extent_end(struct extent_buffer *leaf,
113			   struct btrfs_key *key,
114			   struct btrfs_file_extent_item *extent)
115{
116	u64 end;
117	u64 len;
118
119	if (btrfs_file_extent_type(leaf, extent) == BTRFS_FILE_EXTENT_INLINE) {
120		len = btrfs_file_extent_ram_bytes(leaf, extent);
121		end = ALIGN(key->offset + len, leaf->fs_info->sectorsize);
122	} else {
123		len = btrfs_file_extent_num_bytes(leaf, extent);
124		end = key->offset + len;
125	}
126	return end;
127}
128
129/*
130 * Customized report for dir_item, the only new important information is
131 * key->objectid, which represents inode number
132 */
133__printf(3, 4)
134__cold
135static void dir_item_err(const struct extent_buffer *eb, int slot,
136			 const char *fmt, ...)
137{
138	const struct btrfs_fs_info *fs_info = eb->fs_info;
139	struct btrfs_key key;
140	struct va_format vaf;
141	va_list args;
142
143	btrfs_item_key_to_cpu(eb, &key, slot);
144	va_start(args, fmt);
145
146	vaf.fmt = fmt;
147	vaf.va = &args;
148
149	btrfs_crit(fs_info,
150		"corrupt %s: root=%llu block=%llu slot=%d ino=%llu, %pV",
151		btrfs_header_level(eb) == 0 ? "leaf" : "node",
152		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
153		key.objectid, &vaf);
154	va_end(args);
155}
156
157/*
158 * This functions checks prev_key->objectid, to ensure current key and prev_key
159 * share the same objectid as inode number.
160 *
161 * This is to detect missing INODE_ITEM in subvolume trees.
162 *
163 * Return true if everything is OK or we don't need to check.
164 * Return false if anything is wrong.
165 */
166static bool check_prev_ino(struct extent_buffer *leaf,
167			   struct btrfs_key *key, int slot,
168			   struct btrfs_key *prev_key)
169{
170	/* No prev key, skip check */
171	if (slot == 0)
172		return true;
173
174	/* Only these key->types needs to be checked */
175	ASSERT(key->type == BTRFS_XATTR_ITEM_KEY ||
176	       key->type == BTRFS_INODE_REF_KEY ||
177	       key->type == BTRFS_DIR_INDEX_KEY ||
178	       key->type == BTRFS_DIR_ITEM_KEY ||
179	       key->type == BTRFS_EXTENT_DATA_KEY);
180
181	/*
182	 * Only subvolume trees along with their reloc trees need this check.
183	 * Things like log tree doesn't follow this ino requirement.
184	 */
185	if (!is_fstree(btrfs_header_owner(leaf)))
186		return true;
187
188	if (key->objectid == prev_key->objectid)
189		return true;
190
191	/* Error found */
192	dir_item_err(leaf, slot,
193		"invalid previous key objectid, have %llu expect %llu",
194		prev_key->objectid, key->objectid);
195	return false;
196}
197static int check_extent_data_item(struct extent_buffer *leaf,
198				  struct btrfs_key *key, int slot,
199				  struct btrfs_key *prev_key)
200{
201	struct btrfs_fs_info *fs_info = leaf->fs_info;
202	struct btrfs_file_extent_item *fi;
203	u32 sectorsize = fs_info->sectorsize;
204	u32 item_size = btrfs_item_size_nr(leaf, slot);
205	u64 extent_end;
206
207	if (unlikely(!IS_ALIGNED(key->offset, sectorsize))) {
208		file_extent_err(leaf, slot,
209"unaligned file_offset for file extent, have %llu should be aligned to %u",
210			key->offset, sectorsize);
211		return -EUCLEAN;
212	}
213
214	/*
215	 * Previous key must have the same key->objectid (ino).
216	 * It can be XATTR_ITEM, INODE_ITEM or just another EXTENT_DATA.
217	 * But if objectids mismatch, it means we have a missing
218	 * INODE_ITEM.
219	 */
220	if (unlikely(!check_prev_ino(leaf, key, slot, prev_key)))
221		return -EUCLEAN;
222
223	fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
224
225	/*
226	 * Make sure the item contains at least inline header, so the file
227	 * extent type is not some garbage.
228	 */
229	if (unlikely(item_size < BTRFS_FILE_EXTENT_INLINE_DATA_START)) {
230		file_extent_err(leaf, slot,
231				"invalid item size, have %u expect [%zu, %u)",
232				item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START,
233				SZ_4K);
234		return -EUCLEAN;
235	}
236	if (unlikely(btrfs_file_extent_type(leaf, fi) >=
237		     BTRFS_NR_FILE_EXTENT_TYPES)) {
238		file_extent_err(leaf, slot,
239		"invalid type for file extent, have %u expect range [0, %u]",
240			btrfs_file_extent_type(leaf, fi),
241			BTRFS_NR_FILE_EXTENT_TYPES - 1);
242		return -EUCLEAN;
243	}
244
245	/*
246	 * Support for new compression/encryption must introduce incompat flag,
247	 * and must be caught in open_ctree().
248	 */
249	if (unlikely(btrfs_file_extent_compression(leaf, fi) >=
250		     BTRFS_NR_COMPRESS_TYPES)) {
251		file_extent_err(leaf, slot,
252	"invalid compression for file extent, have %u expect range [0, %u]",
253			btrfs_file_extent_compression(leaf, fi),
254			BTRFS_NR_COMPRESS_TYPES - 1);
255		return -EUCLEAN;
256	}
257	if (unlikely(btrfs_file_extent_encryption(leaf, fi))) {
258		file_extent_err(leaf, slot,
259			"invalid encryption for file extent, have %u expect 0",
260			btrfs_file_extent_encryption(leaf, fi));
261		return -EUCLEAN;
262	}
263	if (btrfs_file_extent_type(leaf, fi) == BTRFS_FILE_EXTENT_INLINE) {
264		/* Inline extent must have 0 as key offset */
265		if (unlikely(key->offset)) {
266			file_extent_err(leaf, slot,
267		"invalid file_offset for inline file extent, have %llu expect 0",
268				key->offset);
269			return -EUCLEAN;
270		}
271
272		/* Compressed inline extent has no on-disk size, skip it */
273		if (btrfs_file_extent_compression(leaf, fi) !=
274		    BTRFS_COMPRESS_NONE)
275			return 0;
276
277		/* Uncompressed inline extent size must match item size */
278		if (unlikely(item_size != BTRFS_FILE_EXTENT_INLINE_DATA_START +
279					  btrfs_file_extent_ram_bytes(leaf, fi))) {
280			file_extent_err(leaf, slot,
281	"invalid ram_bytes for uncompressed inline extent, have %u expect %llu",
282				item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START +
283				btrfs_file_extent_ram_bytes(leaf, fi));
284			return -EUCLEAN;
285		}
286		return 0;
287	}
288
289	/* Regular or preallocated extent has fixed item size */
290	if (unlikely(item_size != sizeof(*fi))) {
291		file_extent_err(leaf, slot,
292	"invalid item size for reg/prealloc file extent, have %u expect %zu",
293			item_size, sizeof(*fi));
294		return -EUCLEAN;
295	}
296	if (unlikely(CHECK_FE_ALIGNED(leaf, slot, fi, ram_bytes, sectorsize) ||
297		     CHECK_FE_ALIGNED(leaf, slot, fi, disk_bytenr, sectorsize) ||
298		     CHECK_FE_ALIGNED(leaf, slot, fi, disk_num_bytes, sectorsize) ||
299		     CHECK_FE_ALIGNED(leaf, slot, fi, offset, sectorsize) ||
300		     CHECK_FE_ALIGNED(leaf, slot, fi, num_bytes, sectorsize)))
301		return -EUCLEAN;
302
303	/* Catch extent end overflow */
304	if (unlikely(check_add_overflow(btrfs_file_extent_num_bytes(leaf, fi),
305					key->offset, &extent_end))) {
306		file_extent_err(leaf, slot,
307	"extent end overflow, have file offset %llu extent num bytes %llu",
308				key->offset,
309				btrfs_file_extent_num_bytes(leaf, fi));
310		return -EUCLEAN;
311	}
312
313	/*
314	 * Check that no two consecutive file extent items, in the same leaf,
315	 * present ranges that overlap each other.
316	 */
317	if (slot > 0 &&
318	    prev_key->objectid == key->objectid &&
319	    prev_key->type == BTRFS_EXTENT_DATA_KEY) {
320		struct btrfs_file_extent_item *prev_fi;
321		u64 prev_end;
322
323		prev_fi = btrfs_item_ptr(leaf, slot - 1,
324					 struct btrfs_file_extent_item);
325		prev_end = file_extent_end(leaf, prev_key, prev_fi);
326		if (unlikely(prev_end > key->offset)) {
327			file_extent_err(leaf, slot - 1,
328"file extent end range (%llu) goes beyond start offset (%llu) of the next file extent",
329					prev_end, key->offset);
330			return -EUCLEAN;
331		}
332	}
333
334	return 0;
335}
336
337static int check_csum_item(struct extent_buffer *leaf, struct btrfs_key *key,
338			   int slot, struct btrfs_key *prev_key)
339{
340	struct btrfs_fs_info *fs_info = leaf->fs_info;
341	u32 sectorsize = fs_info->sectorsize;
342	const u32 csumsize = fs_info->csum_size;
343
344	if (unlikely(key->objectid != BTRFS_EXTENT_CSUM_OBJECTID)) {
345		generic_err(leaf, slot,
346		"invalid key objectid for csum item, have %llu expect %llu",
347			key->objectid, BTRFS_EXTENT_CSUM_OBJECTID);
348		return -EUCLEAN;
349	}
350	if (unlikely(!IS_ALIGNED(key->offset, sectorsize))) {
351		generic_err(leaf, slot,
352	"unaligned key offset for csum item, have %llu should be aligned to %u",
353			key->offset, sectorsize);
354		return -EUCLEAN;
355	}
356	if (unlikely(!IS_ALIGNED(btrfs_item_size_nr(leaf, slot), csumsize))) {
357		generic_err(leaf, slot,
358	"unaligned item size for csum item, have %u should be aligned to %u",
359			btrfs_item_size_nr(leaf, slot), csumsize);
360		return -EUCLEAN;
361	}
362	if (slot > 0 && prev_key->type == BTRFS_EXTENT_CSUM_KEY) {
363		u64 prev_csum_end;
364		u32 prev_item_size;
365
366		prev_item_size = btrfs_item_size_nr(leaf, slot - 1);
367		prev_csum_end = (prev_item_size / csumsize) * sectorsize;
368		prev_csum_end += prev_key->offset;
369		if (unlikely(prev_csum_end > key->offset)) {
370			generic_err(leaf, slot - 1,
371"csum end range (%llu) goes beyond the start range (%llu) of the next csum item",
372				    prev_csum_end, key->offset);
373			return -EUCLEAN;
374		}
375	}
376	return 0;
377}
378
379/* Inode item error output has the same format as dir_item_err() */
380#define inode_item_err(eb, slot, fmt, ...)			\
381	dir_item_err(eb, slot, fmt, __VA_ARGS__)
382
383static int check_inode_key(struct extent_buffer *leaf, struct btrfs_key *key,
384			   int slot)
385{
386	struct btrfs_key item_key;
387	bool is_inode_item;
388
389	btrfs_item_key_to_cpu(leaf, &item_key, slot);
390	is_inode_item = (item_key.type == BTRFS_INODE_ITEM_KEY);
391
392	/* For XATTR_ITEM, location key should be all 0 */
393	if (item_key.type == BTRFS_XATTR_ITEM_KEY) {
394		if (unlikely(key->objectid != 0 || key->type != 0 ||
395			     key->offset != 0))
396			return -EUCLEAN;
397		return 0;
398	}
399
400	if (unlikely((key->objectid < BTRFS_FIRST_FREE_OBJECTID ||
401		      key->objectid > BTRFS_LAST_FREE_OBJECTID) &&
402		     key->objectid != BTRFS_ROOT_TREE_DIR_OBJECTID &&
403		     key->objectid != BTRFS_FREE_INO_OBJECTID)) {
404		if (is_inode_item) {
405			generic_err(leaf, slot,
406	"invalid key objectid: has %llu expect %llu or [%llu, %llu] or %llu",
407				key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID,
408				BTRFS_FIRST_FREE_OBJECTID,
409				BTRFS_LAST_FREE_OBJECTID,
410				BTRFS_FREE_INO_OBJECTID);
411		} else {
412			dir_item_err(leaf, slot,
413"invalid location key objectid: has %llu expect %llu or [%llu, %llu] or %llu",
414				key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID,
415				BTRFS_FIRST_FREE_OBJECTID,
416				BTRFS_LAST_FREE_OBJECTID,
417				BTRFS_FREE_INO_OBJECTID);
418		}
419		return -EUCLEAN;
420	}
421	if (unlikely(key->offset != 0)) {
422		if (is_inode_item)
423			inode_item_err(leaf, slot,
424				       "invalid key offset: has %llu expect 0",
425				       key->offset);
426		else
427			dir_item_err(leaf, slot,
428				"invalid location key offset:has %llu expect 0",
429				key->offset);
430		return -EUCLEAN;
431	}
432	return 0;
433}
434
435static int check_root_key(struct extent_buffer *leaf, struct btrfs_key *key,
436			  int slot)
437{
438	struct btrfs_key item_key;
439	bool is_root_item;
440
441	btrfs_item_key_to_cpu(leaf, &item_key, slot);
442	is_root_item = (item_key.type == BTRFS_ROOT_ITEM_KEY);
443
444	/* No such tree id */
445	if (unlikely(key->objectid == 0)) {
446		if (is_root_item)
447			generic_err(leaf, slot, "invalid root id 0");
448		else
449			dir_item_err(leaf, slot,
450				     "invalid location key root id 0");
451		return -EUCLEAN;
452	}
453
454	/* DIR_ITEM/INDEX/INODE_REF is not allowed to point to non-fs trees */
455	if (unlikely(!is_fstree(key->objectid) && !is_root_item)) {
456		dir_item_err(leaf, slot,
457		"invalid location key objectid, have %llu expect [%llu, %llu]",
458				key->objectid, BTRFS_FIRST_FREE_OBJECTID,
459				BTRFS_LAST_FREE_OBJECTID);
460		return -EUCLEAN;
461	}
462
463	/*
464	 * ROOT_ITEM with non-zero offset means this is a snapshot, created at
465	 * @offset transid.
466	 * Furthermore, for location key in DIR_ITEM, its offset is always -1.
467	 *
468	 * So here we only check offset for reloc tree whose key->offset must
469	 * be a valid tree.
470	 */
471	if (unlikely(key->objectid == BTRFS_TREE_RELOC_OBJECTID &&
472		     key->offset == 0)) {
473		generic_err(leaf, slot, "invalid root id 0 for reloc tree");
474		return -EUCLEAN;
475	}
476	return 0;
477}
478
479static int check_dir_item(struct extent_buffer *leaf,
480			  struct btrfs_key *key, struct btrfs_key *prev_key,
481			  int slot)
482{
483	struct btrfs_fs_info *fs_info = leaf->fs_info;
484	struct btrfs_dir_item *di;
485	u32 item_size = btrfs_item_size_nr(leaf, slot);
486	u32 cur = 0;
487
488	if (unlikely(!check_prev_ino(leaf, key, slot, prev_key)))
489		return -EUCLEAN;
490
491	di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
492	while (cur < item_size) {
493		struct btrfs_key location_key;
494		u32 name_len;
495		u32 data_len;
496		u32 max_name_len;
497		u32 total_size;
498		u32 name_hash;
499		u8 dir_type;
500		int ret;
501
502		/* header itself should not cross item boundary */
503		if (unlikely(cur + sizeof(*di) > item_size)) {
504			dir_item_err(leaf, slot,
505		"dir item header crosses item boundary, have %zu boundary %u",
506				cur + sizeof(*di), item_size);
507			return -EUCLEAN;
508		}
509
510		/* Location key check */
511		btrfs_dir_item_key_to_cpu(leaf, di, &location_key);
512		if (location_key.type == BTRFS_ROOT_ITEM_KEY) {
513			ret = check_root_key(leaf, &location_key, slot);
514			if (unlikely(ret < 0))
515				return ret;
516		} else if (location_key.type == BTRFS_INODE_ITEM_KEY ||
517			   location_key.type == 0) {
518			ret = check_inode_key(leaf, &location_key, slot);
519			if (unlikely(ret < 0))
520				return ret;
521		} else {
522			dir_item_err(leaf, slot,
523			"invalid location key type, have %u, expect %u or %u",
524				     location_key.type, BTRFS_ROOT_ITEM_KEY,
525				     BTRFS_INODE_ITEM_KEY);
526			return -EUCLEAN;
527		}
528
529		/* dir type check */
530		dir_type = btrfs_dir_type(leaf, di);
531		if (unlikely(dir_type >= BTRFS_FT_MAX)) {
532			dir_item_err(leaf, slot,
533			"invalid dir item type, have %u expect [0, %u)",
534				dir_type, BTRFS_FT_MAX);
535			return -EUCLEAN;
536		}
537
538		if (unlikely(key->type == BTRFS_XATTR_ITEM_KEY &&
539			     dir_type != BTRFS_FT_XATTR)) {
540			dir_item_err(leaf, slot,
541		"invalid dir item type for XATTR key, have %u expect %u",
542				dir_type, BTRFS_FT_XATTR);
543			return -EUCLEAN;
544		}
545		if (unlikely(dir_type == BTRFS_FT_XATTR &&
546			     key->type != BTRFS_XATTR_ITEM_KEY)) {
547			dir_item_err(leaf, slot,
548			"xattr dir type found for non-XATTR key");
549			return -EUCLEAN;
550		}
551		if (dir_type == BTRFS_FT_XATTR)
552			max_name_len = XATTR_NAME_MAX;
553		else
554			max_name_len = BTRFS_NAME_LEN;
555
556		/* Name/data length check */
557		name_len = btrfs_dir_name_len(leaf, di);
558		data_len = btrfs_dir_data_len(leaf, di);
559		if (unlikely(name_len > max_name_len)) {
560			dir_item_err(leaf, slot,
561			"dir item name len too long, have %u max %u",
562				name_len, max_name_len);
563			return -EUCLEAN;
564		}
565		if (unlikely(name_len + data_len > BTRFS_MAX_XATTR_SIZE(fs_info))) {
566			dir_item_err(leaf, slot,
567			"dir item name and data len too long, have %u max %u",
568				name_len + data_len,
569				BTRFS_MAX_XATTR_SIZE(fs_info));
570			return -EUCLEAN;
571		}
572
573		if (unlikely(data_len && dir_type != BTRFS_FT_XATTR)) {
574			dir_item_err(leaf, slot,
575			"dir item with invalid data len, have %u expect 0",
576				data_len);
577			return -EUCLEAN;
578		}
579
580		total_size = sizeof(*di) + name_len + data_len;
581
582		/* header and name/data should not cross item boundary */
583		if (unlikely(cur + total_size > item_size)) {
584			dir_item_err(leaf, slot,
585		"dir item data crosses item boundary, have %u boundary %u",
586				cur + total_size, item_size);
587			return -EUCLEAN;
588		}
589
590		/*
591		 * Special check for XATTR/DIR_ITEM, as key->offset is name
592		 * hash, should match its name
593		 */
594		if (key->type == BTRFS_DIR_ITEM_KEY ||
595		    key->type == BTRFS_XATTR_ITEM_KEY) {
596			char namebuf[max(BTRFS_NAME_LEN, XATTR_NAME_MAX)];
597
598			read_extent_buffer(leaf, namebuf,
599					(unsigned long)(di + 1), name_len);
600			name_hash = btrfs_name_hash(namebuf, name_len);
601			if (unlikely(key->offset != name_hash)) {
602				dir_item_err(leaf, slot,
603		"name hash mismatch with key, have 0x%016x expect 0x%016llx",
604					name_hash, key->offset);
605				return -EUCLEAN;
606			}
607		}
608		cur += total_size;
609		di = (struct btrfs_dir_item *)((void *)di + total_size);
610	}
611	return 0;
612}
613
614__printf(3, 4)
615__cold
616static void block_group_err(const struct extent_buffer *eb, int slot,
617			    const char *fmt, ...)
618{
619	const struct btrfs_fs_info *fs_info = eb->fs_info;
620	struct btrfs_key key;
621	struct va_format vaf;
622	va_list args;
623
624	btrfs_item_key_to_cpu(eb, &key, slot);
625	va_start(args, fmt);
626
627	vaf.fmt = fmt;
628	vaf.va = &args;
629
630	btrfs_crit(fs_info,
631	"corrupt %s: root=%llu block=%llu slot=%d bg_start=%llu bg_len=%llu, %pV",
632		btrfs_header_level(eb) == 0 ? "leaf" : "node",
633		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
634		key.objectid, key.offset, &vaf);
635	va_end(args);
636}
637
638static int check_block_group_item(struct extent_buffer *leaf,
639				  struct btrfs_key *key, int slot)
640{
641	struct btrfs_block_group_item bgi;
642	u32 item_size = btrfs_item_size_nr(leaf, slot);
643	u64 flags;
644	u64 type;
645
646	/*
647	 * Here we don't really care about alignment since extent allocator can
648	 * handle it.  We care more about the size.
649	 */
650	if (unlikely(key->offset == 0)) {
651		block_group_err(leaf, slot,
652				"invalid block group size 0");
653		return -EUCLEAN;
654	}
655
656	if (unlikely(item_size != sizeof(bgi))) {
657		block_group_err(leaf, slot,
658			"invalid item size, have %u expect %zu",
659				item_size, sizeof(bgi));
660		return -EUCLEAN;
661	}
662
663	read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot),
664			   sizeof(bgi));
665	if (unlikely(btrfs_stack_block_group_chunk_objectid(&bgi) !=
666		     BTRFS_FIRST_CHUNK_TREE_OBJECTID)) {
667		block_group_err(leaf, slot,
668		"invalid block group chunk objectid, have %llu expect %llu",
669				btrfs_stack_block_group_chunk_objectid(&bgi),
670				BTRFS_FIRST_CHUNK_TREE_OBJECTID);
671		return -EUCLEAN;
672	}
673
674	if (unlikely(btrfs_stack_block_group_used(&bgi) > key->offset)) {
675		block_group_err(leaf, slot,
676			"invalid block group used, have %llu expect [0, %llu)",
677				btrfs_stack_block_group_used(&bgi), key->offset);
678		return -EUCLEAN;
679	}
680
681	flags = btrfs_stack_block_group_flags(&bgi);
682	if (unlikely(hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) > 1)) {
683		block_group_err(leaf, slot,
684"invalid profile flags, have 0x%llx (%lu bits set) expect no more than 1 bit set",
685			flags & BTRFS_BLOCK_GROUP_PROFILE_MASK,
686			hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK));
687		return -EUCLEAN;
688	}
689
690	type = flags & BTRFS_BLOCK_GROUP_TYPE_MASK;
691	if (unlikely(type != BTRFS_BLOCK_GROUP_DATA &&
692		     type != BTRFS_BLOCK_GROUP_METADATA &&
693		     type != BTRFS_BLOCK_GROUP_SYSTEM &&
694		     type != (BTRFS_BLOCK_GROUP_METADATA |
695			      BTRFS_BLOCK_GROUP_DATA))) {
696		block_group_err(leaf, slot,
697"invalid type, have 0x%llx (%lu bits set) expect either 0x%llx, 0x%llx, 0x%llx or 0x%llx",
698			type, hweight64(type),
699			BTRFS_BLOCK_GROUP_DATA, BTRFS_BLOCK_GROUP_METADATA,
700			BTRFS_BLOCK_GROUP_SYSTEM,
701			BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA);
702		return -EUCLEAN;
703	}
704	return 0;
705}
706
707__printf(4, 5)
708__cold
709static void chunk_err(const struct extent_buffer *leaf,
710		      const struct btrfs_chunk *chunk, u64 logical,
711		      const char *fmt, ...)
712{
713	const struct btrfs_fs_info *fs_info = leaf->fs_info;
714	bool is_sb;
715	struct va_format vaf;
716	va_list args;
717	int i;
718	int slot = -1;
719
720	/* Only superblock eb is able to have such small offset */
721	is_sb = (leaf->start == BTRFS_SUPER_INFO_OFFSET);
722
723	if (!is_sb) {
724		/*
725		 * Get the slot number by iterating through all slots, this
726		 * would provide better readability.
727		 */
728		for (i = 0; i < btrfs_header_nritems(leaf); i++) {
729			if (btrfs_item_ptr_offset(leaf, i) ==
730					(unsigned long)chunk) {
731				slot = i;
732				break;
733			}
734		}
735	}
736	va_start(args, fmt);
737	vaf.fmt = fmt;
738	vaf.va = &args;
739
740	if (is_sb)
741		btrfs_crit(fs_info,
742		"corrupt superblock syschunk array: chunk_start=%llu, %pV",
743			   logical, &vaf);
744	else
745		btrfs_crit(fs_info,
746	"corrupt leaf: root=%llu block=%llu slot=%d chunk_start=%llu, %pV",
747			   BTRFS_CHUNK_TREE_OBJECTID, leaf->start, slot,
748			   logical, &vaf);
749	va_end(args);
750}
751
752/*
753 * The common chunk check which could also work on super block sys chunk array.
754 *
755 * Return -EUCLEAN if anything is corrupted.
756 * Return 0 if everything is OK.
757 */
758int btrfs_check_chunk_valid(struct extent_buffer *leaf,
759			    struct btrfs_chunk *chunk, u64 logical)
760{
761	struct btrfs_fs_info *fs_info = leaf->fs_info;
762	u64 length;
763	u64 chunk_end;
764	u64 stripe_len;
765	u16 num_stripes;
766	u16 sub_stripes;
767	u64 type;
768	u64 features;
769	bool mixed = false;
770	int raid_index;
771	int nparity;
772	int ncopies;
773
774	length = btrfs_chunk_length(leaf, chunk);
775	stripe_len = btrfs_chunk_stripe_len(leaf, chunk);
776	num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
777	sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk);
778	type = btrfs_chunk_type(leaf, chunk);
779	raid_index = btrfs_bg_flags_to_raid_index(type);
780	ncopies = btrfs_raid_array[raid_index].ncopies;
781	nparity = btrfs_raid_array[raid_index].nparity;
782
783	if (unlikely(!num_stripes)) {
784		chunk_err(leaf, chunk, logical,
785			  "invalid chunk num_stripes, have %u", num_stripes);
786		return -EUCLEAN;
787	}
788	if (unlikely(num_stripes < ncopies)) {
789		chunk_err(leaf, chunk, logical,
790			  "invalid chunk num_stripes < ncopies, have %u < %d",
791			  num_stripes, ncopies);
792		return -EUCLEAN;
793	}
794	if (unlikely(nparity && num_stripes == nparity)) {
795		chunk_err(leaf, chunk, logical,
796			  "invalid chunk num_stripes == nparity, have %u == %d",
797			  num_stripes, nparity);
798		return -EUCLEAN;
799	}
800	if (unlikely(!IS_ALIGNED(logical, fs_info->sectorsize))) {
801		chunk_err(leaf, chunk, logical,
802		"invalid chunk logical, have %llu should aligned to %u",
803			  logical, fs_info->sectorsize);
804		return -EUCLEAN;
805	}
806	if (unlikely(btrfs_chunk_sector_size(leaf, chunk) != fs_info->sectorsize)) {
807		chunk_err(leaf, chunk, logical,
808			  "invalid chunk sectorsize, have %u expect %u",
809			  btrfs_chunk_sector_size(leaf, chunk),
810			  fs_info->sectorsize);
811		return -EUCLEAN;
812	}
813	if (unlikely(!length || !IS_ALIGNED(length, fs_info->sectorsize))) {
814		chunk_err(leaf, chunk, logical,
815			  "invalid chunk length, have %llu", length);
816		return -EUCLEAN;
817	}
818	if (unlikely(check_add_overflow(logical, length, &chunk_end))) {
819		chunk_err(leaf, chunk, logical,
820"invalid chunk logical start and length, have logical start %llu length %llu",
821			  logical, length);
822		return -EUCLEAN;
823	}
824	if (unlikely(!is_power_of_2(stripe_len) || stripe_len != BTRFS_STRIPE_LEN)) {
825		chunk_err(leaf, chunk, logical,
826			  "invalid chunk stripe length: %llu",
827			  stripe_len);
828		return -EUCLEAN;
829	}
830	if (unlikely(type & ~(BTRFS_BLOCK_GROUP_TYPE_MASK |
831			      BTRFS_BLOCK_GROUP_PROFILE_MASK))) {
832		chunk_err(leaf, chunk, logical,
833			  "unrecognized chunk type: 0x%llx",
834			  ~(BTRFS_BLOCK_GROUP_TYPE_MASK |
835			    BTRFS_BLOCK_GROUP_PROFILE_MASK) &
836			  btrfs_chunk_type(leaf, chunk));
837		return -EUCLEAN;
838	}
839
840	if (unlikely(!has_single_bit_set(type & BTRFS_BLOCK_GROUP_PROFILE_MASK) &&
841		     (type & BTRFS_BLOCK_GROUP_PROFILE_MASK) != 0)) {
842		chunk_err(leaf, chunk, logical,
843		"invalid chunk profile flag: 0x%llx, expect 0 or 1 bit set",
844			  type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
845		return -EUCLEAN;
846	}
847	if (unlikely((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == 0)) {
848		chunk_err(leaf, chunk, logical,
849	"missing chunk type flag, have 0x%llx one bit must be set in 0x%llx",
850			  type, BTRFS_BLOCK_GROUP_TYPE_MASK);
851		return -EUCLEAN;
852	}
853
854	if (unlikely((type & BTRFS_BLOCK_GROUP_SYSTEM) &&
855		     (type & (BTRFS_BLOCK_GROUP_METADATA |
856			      BTRFS_BLOCK_GROUP_DATA)))) {
857		chunk_err(leaf, chunk, logical,
858			  "system chunk with data or metadata type: 0x%llx",
859			  type);
860		return -EUCLEAN;
861	}
862
863	features = btrfs_super_incompat_flags(fs_info->super_copy);
864	if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)
865		mixed = true;
866
867	if (!mixed) {
868		if (unlikely((type & BTRFS_BLOCK_GROUP_METADATA) &&
869			     (type & BTRFS_BLOCK_GROUP_DATA))) {
870			chunk_err(leaf, chunk, logical,
871			"mixed chunk type in non-mixed mode: 0x%llx", type);
872			return -EUCLEAN;
873		}
874	}
875
876	if (unlikely((type & BTRFS_BLOCK_GROUP_RAID10 &&
877		      sub_stripes != btrfs_raid_array[BTRFS_RAID_RAID10].sub_stripes) ||
878		     (type & BTRFS_BLOCK_GROUP_RAID1 &&
879		      num_stripes != btrfs_raid_array[BTRFS_RAID_RAID1].devs_min) ||
880		     (type & BTRFS_BLOCK_GROUP_RAID5 &&
881		      num_stripes < btrfs_raid_array[BTRFS_RAID_RAID5].devs_min) ||
882		     (type & BTRFS_BLOCK_GROUP_RAID6 &&
883		      num_stripes < btrfs_raid_array[BTRFS_RAID_RAID6].devs_min) ||
884		     (type & BTRFS_BLOCK_GROUP_DUP &&
885		      num_stripes != btrfs_raid_array[BTRFS_RAID_DUP].dev_stripes) ||
886		     ((type & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 &&
887		      num_stripes != btrfs_raid_array[BTRFS_RAID_SINGLE].dev_stripes))) {
888		chunk_err(leaf, chunk, logical,
889			"invalid num_stripes:sub_stripes %u:%u for profile %llu",
890			num_stripes, sub_stripes,
891			type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
892		return -EUCLEAN;
893	}
894
895	return 0;
896}
897
898/*
899 * Enhanced version of chunk item checker.
900 *
901 * The common btrfs_check_chunk_valid() doesn't check item size since it needs
902 * to work on super block sys_chunk_array which doesn't have full item ptr.
903 */
904static int check_leaf_chunk_item(struct extent_buffer *leaf,
905				 struct btrfs_chunk *chunk,
906				 struct btrfs_key *key, int slot)
907{
908	int num_stripes;
909
910	if (unlikely(btrfs_item_size_nr(leaf, slot) < sizeof(struct btrfs_chunk))) {
911		chunk_err(leaf, chunk, key->offset,
912			"invalid chunk item size: have %u expect [%zu, %u)",
913			btrfs_item_size_nr(leaf, slot),
914			sizeof(struct btrfs_chunk),
915			BTRFS_LEAF_DATA_SIZE(leaf->fs_info));
916		return -EUCLEAN;
917	}
918
919	num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
920	/* Let btrfs_check_chunk_valid() handle this error type */
921	if (num_stripes == 0)
922		goto out;
923
924	if (unlikely(btrfs_chunk_item_size(num_stripes) !=
925		     btrfs_item_size_nr(leaf, slot))) {
926		chunk_err(leaf, chunk, key->offset,
927			"invalid chunk item size: have %u expect %lu",
928			btrfs_item_size_nr(leaf, slot),
929			btrfs_chunk_item_size(num_stripes));
930		return -EUCLEAN;
931	}
932out:
933	return btrfs_check_chunk_valid(leaf, chunk, key->offset);
934}
935
936__printf(3, 4)
937__cold
938static void dev_item_err(const struct extent_buffer *eb, int slot,
939			 const char *fmt, ...)
940{
941	struct btrfs_key key;
942	struct va_format vaf;
943	va_list args;
944
945	btrfs_item_key_to_cpu(eb, &key, slot);
946	va_start(args, fmt);
947
948	vaf.fmt = fmt;
949	vaf.va = &args;
950
951	btrfs_crit(eb->fs_info,
952	"corrupt %s: root=%llu block=%llu slot=%d devid=%llu %pV",
953		btrfs_header_level(eb) == 0 ? "leaf" : "node",
954		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
955		key.objectid, &vaf);
956	va_end(args);
957}
958
959static int check_dev_item(struct extent_buffer *leaf,
960			  struct btrfs_key *key, int slot)
961{
962	struct btrfs_dev_item *ditem;
963
964	if (unlikely(key->objectid != BTRFS_DEV_ITEMS_OBJECTID)) {
965		dev_item_err(leaf, slot,
966			     "invalid objectid: has=%llu expect=%llu",
967			     key->objectid, BTRFS_DEV_ITEMS_OBJECTID);
968		return -EUCLEAN;
969	}
970	ditem = btrfs_item_ptr(leaf, slot, struct btrfs_dev_item);
971	if (unlikely(btrfs_device_id(leaf, ditem) != key->offset)) {
972		dev_item_err(leaf, slot,
973			     "devid mismatch: key has=%llu item has=%llu",
974			     key->offset, btrfs_device_id(leaf, ditem));
975		return -EUCLEAN;
976	}
977
978	/*
979	 * For device total_bytes, we don't have reliable way to check it, as
980	 * it can be 0 for device removal. Device size check can only be done
981	 * by dev extents check.
982	 */
983	if (unlikely(btrfs_device_bytes_used(leaf, ditem) >
984		     btrfs_device_total_bytes(leaf, ditem))) {
985		dev_item_err(leaf, slot,
986			     "invalid bytes used: have %llu expect [0, %llu]",
987			     btrfs_device_bytes_used(leaf, ditem),
988			     btrfs_device_total_bytes(leaf, ditem));
989		return -EUCLEAN;
990	}
991	/*
992	 * Remaining members like io_align/type/gen/dev_group aren't really
993	 * utilized.  Skip them to make later usage of them easier.
994	 */
995	return 0;
996}
997
998static int check_inode_item(struct extent_buffer *leaf,
999			    struct btrfs_key *key, int slot)
1000{
1001	struct btrfs_fs_info *fs_info = leaf->fs_info;
1002	struct btrfs_inode_item *iitem;
1003	u64 super_gen = btrfs_super_generation(fs_info->super_copy);
1004	u32 valid_mask = (S_IFMT | S_ISUID | S_ISGID | S_ISVTX | 0777);
1005	u32 mode;
1006	int ret;
1007
1008	ret = check_inode_key(leaf, key, slot);
1009	if (unlikely(ret < 0))
1010		return ret;
1011
1012	iitem = btrfs_item_ptr(leaf, slot, struct btrfs_inode_item);
1013
1014	/* Here we use super block generation + 1 to handle log tree */
1015	if (unlikely(btrfs_inode_generation(leaf, iitem) > super_gen + 1)) {
1016		inode_item_err(leaf, slot,
1017			"invalid inode generation: has %llu expect (0, %llu]",
1018			       btrfs_inode_generation(leaf, iitem),
1019			       super_gen + 1);
1020		return -EUCLEAN;
1021	}
1022	/* Note for ROOT_TREE_DIR_ITEM, mkfs could set its transid 0 */
1023	if (unlikely(btrfs_inode_transid(leaf, iitem) > super_gen + 1)) {
1024		inode_item_err(leaf, slot,
1025			"invalid inode transid: has %llu expect [0, %llu]",
1026			       btrfs_inode_transid(leaf, iitem), super_gen + 1);
1027		return -EUCLEAN;
1028	}
1029
1030	/*
1031	 * For size and nbytes it's better not to be too strict, as for dir
1032	 * item its size/nbytes can easily get wrong, but doesn't affect
1033	 * anything in the fs. So here we skip the check.
1034	 */
1035	mode = btrfs_inode_mode(leaf, iitem);
1036	if (unlikely(mode & ~valid_mask)) {
1037		inode_item_err(leaf, slot,
1038			       "unknown mode bit detected: 0x%x",
1039			       mode & ~valid_mask);
1040		return -EUCLEAN;
1041	}
1042
1043	/*
1044	 * S_IFMT is not bit mapped so we can't completely rely on
1045	 * is_power_of_2/has_single_bit_set, but it can save us from checking
1046	 * FIFO/CHR/DIR/REG.  Only needs to check BLK, LNK and SOCKS
1047	 */
1048	if (!has_single_bit_set(mode & S_IFMT)) {
1049		if (unlikely(!S_ISLNK(mode) && !S_ISBLK(mode) && !S_ISSOCK(mode))) {
1050			inode_item_err(leaf, slot,
1051			"invalid mode: has 0%o expect valid S_IF* bit(s)",
1052				       mode & S_IFMT);
1053			return -EUCLEAN;
1054		}
1055	}
1056	if (unlikely(S_ISDIR(mode) && btrfs_inode_nlink(leaf, iitem) > 1)) {
1057		inode_item_err(leaf, slot,
1058		       "invalid nlink: has %u expect no more than 1 for dir",
1059			btrfs_inode_nlink(leaf, iitem));
1060		return -EUCLEAN;
1061	}
1062	if (unlikely(btrfs_inode_flags(leaf, iitem) & ~BTRFS_INODE_FLAG_MASK)) {
1063		inode_item_err(leaf, slot,
1064			       "unknown flags detected: 0x%llx",
1065			       btrfs_inode_flags(leaf, iitem) &
1066			       ~BTRFS_INODE_FLAG_MASK);
1067		return -EUCLEAN;
1068	}
1069	return 0;
1070}
1071
1072static int check_root_item(struct extent_buffer *leaf, struct btrfs_key *key,
1073			   int slot)
1074{
1075	struct btrfs_fs_info *fs_info = leaf->fs_info;
1076	struct btrfs_root_item ri = { 0 };
1077	const u64 valid_root_flags = BTRFS_ROOT_SUBVOL_RDONLY |
1078				     BTRFS_ROOT_SUBVOL_DEAD;
1079	int ret;
1080
1081	ret = check_root_key(leaf, key, slot);
1082	if (unlikely(ret < 0))
1083		return ret;
1084
1085	if (unlikely(btrfs_item_size_nr(leaf, slot) != sizeof(ri) &&
1086		     btrfs_item_size_nr(leaf, slot) !=
1087		     btrfs_legacy_root_item_size())) {
1088		generic_err(leaf, slot,
1089			    "invalid root item size, have %u expect %zu or %u",
1090			    btrfs_item_size_nr(leaf, slot), sizeof(ri),
1091			    btrfs_legacy_root_item_size());
1092		return -EUCLEAN;
1093	}
1094
1095	/*
1096	 * For legacy root item, the members starting at generation_v2 will be
1097	 * all filled with 0.
1098	 * And since we allow geneartion_v2 as 0, it will still pass the check.
1099	 */
1100	read_extent_buffer(leaf, &ri, btrfs_item_ptr_offset(leaf, slot),
1101			   btrfs_item_size_nr(leaf, slot));
1102
1103	/* Generation related */
1104	if (unlikely(btrfs_root_generation(&ri) >
1105		     btrfs_super_generation(fs_info->super_copy) + 1)) {
1106		generic_err(leaf, slot,
1107			"invalid root generation, have %llu expect (0, %llu]",
1108			    btrfs_root_generation(&ri),
1109			    btrfs_super_generation(fs_info->super_copy) + 1);
1110		return -EUCLEAN;
1111	}
1112	if (unlikely(btrfs_root_generation_v2(&ri) >
1113		     btrfs_super_generation(fs_info->super_copy) + 1)) {
1114		generic_err(leaf, slot,
1115		"invalid root v2 generation, have %llu expect (0, %llu]",
1116			    btrfs_root_generation_v2(&ri),
1117			    btrfs_super_generation(fs_info->super_copy) + 1);
1118		return -EUCLEAN;
1119	}
1120	if (unlikely(btrfs_root_last_snapshot(&ri) >
1121		     btrfs_super_generation(fs_info->super_copy) + 1)) {
1122		generic_err(leaf, slot,
1123		"invalid root last_snapshot, have %llu expect (0, %llu]",
1124			    btrfs_root_last_snapshot(&ri),
1125			    btrfs_super_generation(fs_info->super_copy) + 1);
1126		return -EUCLEAN;
1127	}
1128
1129	/* Alignment and level check */
1130	if (unlikely(!IS_ALIGNED(btrfs_root_bytenr(&ri), fs_info->sectorsize))) {
1131		generic_err(leaf, slot,
1132		"invalid root bytenr, have %llu expect to be aligned to %u",
1133			    btrfs_root_bytenr(&ri), fs_info->sectorsize);
1134		return -EUCLEAN;
1135	}
1136	if (unlikely(btrfs_root_level(&ri) >= BTRFS_MAX_LEVEL)) {
1137		generic_err(leaf, slot,
1138			    "invalid root level, have %u expect [0, %u]",
1139			    btrfs_root_level(&ri), BTRFS_MAX_LEVEL - 1);
1140		return -EUCLEAN;
1141	}
1142	if (unlikely(btrfs_root_drop_level(&ri) >= BTRFS_MAX_LEVEL)) {
1143		generic_err(leaf, slot,
1144			    "invalid root level, have %u expect [0, %u]",
1145			    btrfs_root_drop_level(&ri), BTRFS_MAX_LEVEL - 1);
1146		return -EUCLEAN;
1147	}
1148
1149	/* Flags check */
1150	if (unlikely(btrfs_root_flags(&ri) & ~valid_root_flags)) {
1151		generic_err(leaf, slot,
1152			    "invalid root flags, have 0x%llx expect mask 0x%llx",
1153			    btrfs_root_flags(&ri), valid_root_flags);
1154		return -EUCLEAN;
1155	}
1156	return 0;
1157}
1158
1159__printf(3,4)
1160__cold
1161static void extent_err(const struct extent_buffer *eb, int slot,
1162		       const char *fmt, ...)
1163{
1164	struct btrfs_key key;
1165	struct va_format vaf;
1166	va_list args;
1167	u64 bytenr;
1168	u64 len;
1169
1170	btrfs_item_key_to_cpu(eb, &key, slot);
1171	bytenr = key.objectid;
1172	if (key.type == BTRFS_METADATA_ITEM_KEY ||
1173	    key.type == BTRFS_TREE_BLOCK_REF_KEY ||
1174	    key.type == BTRFS_SHARED_BLOCK_REF_KEY)
1175		len = eb->fs_info->nodesize;
1176	else
1177		len = key.offset;
1178	va_start(args, fmt);
1179
1180	vaf.fmt = fmt;
1181	vaf.va = &args;
1182
1183	btrfs_crit(eb->fs_info,
1184	"corrupt %s: block=%llu slot=%d extent bytenr=%llu len=%llu %pV",
1185		btrfs_header_level(eb) == 0 ? "leaf" : "node",
1186		eb->start, slot, bytenr, len, &vaf);
1187	va_end(args);
1188}
1189
1190static int check_extent_item(struct extent_buffer *leaf,
1191			     struct btrfs_key *key, int slot)
1192{
1193	struct btrfs_fs_info *fs_info = leaf->fs_info;
1194	struct btrfs_extent_item *ei;
1195	bool is_tree_block = false;
1196	unsigned long ptr;	/* Current pointer inside inline refs */
1197	unsigned long end;	/* Extent item end */
1198	const u32 item_size = btrfs_item_size_nr(leaf, slot);
1199	u64 flags;
1200	u64 generation;
1201	u64 total_refs;		/* Total refs in btrfs_extent_item */
1202	u64 inline_refs = 0;	/* found total inline refs */
1203
1204	if (unlikely(key->type == BTRFS_METADATA_ITEM_KEY &&
1205		     !btrfs_fs_incompat(fs_info, SKINNY_METADATA))) {
1206		generic_err(leaf, slot,
1207"invalid key type, METADATA_ITEM type invalid when SKINNY_METADATA feature disabled");
1208		return -EUCLEAN;
1209	}
1210	/* key->objectid is the bytenr for both key types */
1211	if (unlikely(!IS_ALIGNED(key->objectid, fs_info->sectorsize))) {
1212		generic_err(leaf, slot,
1213		"invalid key objectid, have %llu expect to be aligned to %u",
1214			   key->objectid, fs_info->sectorsize);
1215		return -EUCLEAN;
1216	}
1217
1218	/* key->offset is tree level for METADATA_ITEM_KEY */
1219	if (unlikely(key->type == BTRFS_METADATA_ITEM_KEY &&
1220		     key->offset >= BTRFS_MAX_LEVEL)) {
1221		extent_err(leaf, slot,
1222			   "invalid tree level, have %llu expect [0, %u]",
1223			   key->offset, BTRFS_MAX_LEVEL - 1);
1224		return -EUCLEAN;
1225	}
1226
1227	/*
1228	 * EXTENT/METADATA_ITEM consists of:
1229	 * 1) One btrfs_extent_item
1230	 *    Records the total refs, type and generation of the extent.
1231	 *
1232	 * 2) One btrfs_tree_block_info (for EXTENT_ITEM and tree backref only)
1233	 *    Records the first key and level of the tree block.
1234	 *
1235	 * 2) Zero or more btrfs_extent_inline_ref(s)
1236	 *    Each inline ref has one btrfs_extent_inline_ref shows:
1237	 *    2.1) The ref type, one of the 4
1238	 *         TREE_BLOCK_REF	Tree block only
1239	 *         SHARED_BLOCK_REF	Tree block only
1240	 *         EXTENT_DATA_REF	Data only
1241	 *         SHARED_DATA_REF	Data only
1242	 *    2.2) Ref type specific data
1243	 *         Either using btrfs_extent_inline_ref::offset, or specific
1244	 *         data structure.
1245	 */
1246	if (unlikely(item_size < sizeof(*ei))) {
1247		extent_err(leaf, slot,
1248			   "invalid item size, have %u expect [%zu, %u)",
1249			   item_size, sizeof(*ei),
1250			   BTRFS_LEAF_DATA_SIZE(fs_info));
1251		return -EUCLEAN;
1252	}
1253	end = item_size + btrfs_item_ptr_offset(leaf, slot);
1254
1255	/* Checks against extent_item */
1256	ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
1257	flags = btrfs_extent_flags(leaf, ei);
1258	total_refs = btrfs_extent_refs(leaf, ei);
1259	generation = btrfs_extent_generation(leaf, ei);
1260	if (unlikely(generation >
1261		     btrfs_super_generation(fs_info->super_copy) + 1)) {
1262		extent_err(leaf, slot,
1263			   "invalid generation, have %llu expect (0, %llu]",
1264			   generation,
1265			   btrfs_super_generation(fs_info->super_copy) + 1);
1266		return -EUCLEAN;
1267	}
1268	if (unlikely(!has_single_bit_set(flags & (BTRFS_EXTENT_FLAG_DATA |
1269						  BTRFS_EXTENT_FLAG_TREE_BLOCK)))) {
1270		extent_err(leaf, slot,
1271		"invalid extent flag, have 0x%llx expect 1 bit set in 0x%llx",
1272			flags, BTRFS_EXTENT_FLAG_DATA |
1273			BTRFS_EXTENT_FLAG_TREE_BLOCK);
1274		return -EUCLEAN;
1275	}
1276	is_tree_block = !!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK);
1277	if (is_tree_block) {
1278		if (unlikely(key->type == BTRFS_EXTENT_ITEM_KEY &&
1279			     key->offset != fs_info->nodesize)) {
1280			extent_err(leaf, slot,
1281				   "invalid extent length, have %llu expect %u",
1282				   key->offset, fs_info->nodesize);
1283			return -EUCLEAN;
1284		}
1285	} else {
1286		if (unlikely(key->type != BTRFS_EXTENT_ITEM_KEY)) {
1287			extent_err(leaf, slot,
1288			"invalid key type, have %u expect %u for data backref",
1289				   key->type, BTRFS_EXTENT_ITEM_KEY);
1290			return -EUCLEAN;
1291		}
1292		if (unlikely(!IS_ALIGNED(key->offset, fs_info->sectorsize))) {
1293			extent_err(leaf, slot,
1294			"invalid extent length, have %llu expect aligned to %u",
1295				   key->offset, fs_info->sectorsize);
1296			return -EUCLEAN;
1297		}
1298		if (unlikely(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)) {
1299			extent_err(leaf, slot,
1300			"invalid extent flag, data has full backref set");
1301			return -EUCLEAN;
1302		}
1303	}
1304	ptr = (unsigned long)(struct btrfs_extent_item *)(ei + 1);
1305
1306	/* Check the special case of btrfs_tree_block_info */
1307	if (is_tree_block && key->type != BTRFS_METADATA_ITEM_KEY) {
1308		struct btrfs_tree_block_info *info;
1309
1310		info = (struct btrfs_tree_block_info *)ptr;
1311		if (unlikely(btrfs_tree_block_level(leaf, info) >= BTRFS_MAX_LEVEL)) {
1312			extent_err(leaf, slot,
1313			"invalid tree block info level, have %u expect [0, %u]",
1314				   btrfs_tree_block_level(leaf, info),
1315				   BTRFS_MAX_LEVEL - 1);
1316			return -EUCLEAN;
1317		}
1318		ptr = (unsigned long)(struct btrfs_tree_block_info *)(info + 1);
1319	}
1320
1321	/* Check inline refs */
1322	while (ptr < end) {
1323		struct btrfs_extent_inline_ref *iref;
1324		struct btrfs_extent_data_ref *dref;
1325		struct btrfs_shared_data_ref *sref;
1326		u64 dref_offset;
1327		u64 inline_offset;
1328		u8 inline_type;
1329
1330		if (unlikely(ptr + sizeof(*iref) > end)) {
1331			extent_err(leaf, slot,
1332"inline ref item overflows extent item, ptr %lu iref size %zu end %lu",
1333				   ptr, sizeof(*iref), end);
1334			return -EUCLEAN;
1335		}
1336		iref = (struct btrfs_extent_inline_ref *)ptr;
1337		inline_type = btrfs_extent_inline_ref_type(leaf, iref);
1338		inline_offset = btrfs_extent_inline_ref_offset(leaf, iref);
1339		if (unlikely(ptr + btrfs_extent_inline_ref_size(inline_type) > end)) {
1340			extent_err(leaf, slot,
1341"inline ref item overflows extent item, ptr %lu iref size %u end %lu",
1342				   ptr, inline_type, end);
1343			return -EUCLEAN;
1344		}
1345
1346		switch (inline_type) {
1347		/* inline_offset is subvolid of the owner, no need to check */
1348		case BTRFS_TREE_BLOCK_REF_KEY:
1349			inline_refs++;
1350			break;
1351		/* Contains parent bytenr */
1352		case BTRFS_SHARED_BLOCK_REF_KEY:
1353			if (unlikely(!IS_ALIGNED(inline_offset,
1354						 fs_info->sectorsize))) {
1355				extent_err(leaf, slot,
1356		"invalid tree parent bytenr, have %llu expect aligned to %u",
1357					   inline_offset, fs_info->sectorsize);
1358				return -EUCLEAN;
1359			}
1360			inline_refs++;
1361			break;
1362		/*
1363		 * Contains owner subvolid, owner key objectid, adjusted offset.
1364		 * The only obvious corruption can happen in that offset.
1365		 */
1366		case BTRFS_EXTENT_DATA_REF_KEY:
1367			dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1368			dref_offset = btrfs_extent_data_ref_offset(leaf, dref);
1369			if (unlikely(!IS_ALIGNED(dref_offset,
1370						 fs_info->sectorsize))) {
1371				extent_err(leaf, slot,
1372		"invalid data ref offset, have %llu expect aligned to %u",
1373					   dref_offset, fs_info->sectorsize);
1374				return -EUCLEAN;
1375			}
1376			inline_refs += btrfs_extent_data_ref_count(leaf, dref);
1377			break;
1378		/* Contains parent bytenr and ref count */
1379		case BTRFS_SHARED_DATA_REF_KEY:
1380			sref = (struct btrfs_shared_data_ref *)(iref + 1);
1381			if (unlikely(!IS_ALIGNED(inline_offset,
1382						 fs_info->sectorsize))) {
1383				extent_err(leaf, slot,
1384		"invalid data parent bytenr, have %llu expect aligned to %u",
1385					   inline_offset, fs_info->sectorsize);
1386				return -EUCLEAN;
1387			}
1388			inline_refs += btrfs_shared_data_ref_count(leaf, sref);
1389			break;
1390		default:
1391			extent_err(leaf, slot, "unknown inline ref type: %u",
1392				   inline_type);
1393			return -EUCLEAN;
1394		}
1395		ptr += btrfs_extent_inline_ref_size(inline_type);
1396	}
1397	/* No padding is allowed */
1398	if (unlikely(ptr != end)) {
1399		extent_err(leaf, slot,
1400			   "invalid extent item size, padding bytes found");
1401		return -EUCLEAN;
1402	}
1403
1404	/* Finally, check the inline refs against total refs */
1405	if (unlikely(inline_refs > total_refs)) {
1406		extent_err(leaf, slot,
1407			"invalid extent refs, have %llu expect >= inline %llu",
1408			   total_refs, inline_refs);
1409		return -EUCLEAN;
1410	}
1411	return 0;
1412}
1413
1414static int check_simple_keyed_refs(struct extent_buffer *leaf,
1415				   struct btrfs_key *key, int slot)
1416{
1417	u32 expect_item_size = 0;
1418
1419	if (key->type == BTRFS_SHARED_DATA_REF_KEY)
1420		expect_item_size = sizeof(struct btrfs_shared_data_ref);
1421
1422	if (unlikely(btrfs_item_size_nr(leaf, slot) != expect_item_size)) {
1423		generic_err(leaf, slot,
1424		"invalid item size, have %u expect %u for key type %u",
1425			    btrfs_item_size_nr(leaf, slot),
1426			    expect_item_size, key->type);
1427		return -EUCLEAN;
1428	}
1429	if (unlikely(!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize))) {
1430		generic_err(leaf, slot,
1431"invalid key objectid for shared block ref, have %llu expect aligned to %u",
1432			    key->objectid, leaf->fs_info->sectorsize);
1433		return -EUCLEAN;
1434	}
1435	if (unlikely(key->type != BTRFS_TREE_BLOCK_REF_KEY &&
1436		     !IS_ALIGNED(key->offset, leaf->fs_info->sectorsize))) {
1437		extent_err(leaf, slot,
1438		"invalid tree parent bytenr, have %llu expect aligned to %u",
1439			   key->offset, leaf->fs_info->sectorsize);
1440		return -EUCLEAN;
1441	}
1442	return 0;
1443}
1444
1445static int check_extent_data_ref(struct extent_buffer *leaf,
1446				 struct btrfs_key *key, int slot)
1447{
1448	struct btrfs_extent_data_ref *dref;
1449	unsigned long ptr = btrfs_item_ptr_offset(leaf, slot);
1450	const unsigned long end = ptr + btrfs_item_size_nr(leaf, slot);
1451
1452	if (unlikely(btrfs_item_size_nr(leaf, slot) % sizeof(*dref) != 0)) {
1453		generic_err(leaf, slot,
1454	"invalid item size, have %u expect aligned to %zu for key type %u",
1455			    btrfs_item_size_nr(leaf, slot),
1456			    sizeof(*dref), key->type);
1457		return -EUCLEAN;
1458	}
1459	if (unlikely(!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize))) {
1460		generic_err(leaf, slot,
1461"invalid key objectid for shared block ref, have %llu expect aligned to %u",
1462			    key->objectid, leaf->fs_info->sectorsize);
1463		return -EUCLEAN;
1464	}
1465	for (; ptr < end; ptr += sizeof(*dref)) {
1466		u64 offset;
1467
1468		/*
1469		 * We cannot check the extent_data_ref hash due to possible
1470		 * overflow from the leaf due to hash collisions.
1471		 */
1472		dref = (struct btrfs_extent_data_ref *)ptr;
1473		offset = btrfs_extent_data_ref_offset(leaf, dref);
1474		if (unlikely(!IS_ALIGNED(offset, leaf->fs_info->sectorsize))) {
1475			extent_err(leaf, slot,
1476	"invalid extent data backref offset, have %llu expect aligned to %u",
1477				   offset, leaf->fs_info->sectorsize);
1478			return -EUCLEAN;
1479		}
1480	}
1481	return 0;
1482}
1483
1484#define inode_ref_err(eb, slot, fmt, args...)			\
1485	inode_item_err(eb, slot, fmt, ##args)
1486static int check_inode_ref(struct extent_buffer *leaf,
1487			   struct btrfs_key *key, struct btrfs_key *prev_key,
1488			   int slot)
1489{
1490	struct btrfs_inode_ref *iref;
1491	unsigned long ptr;
1492	unsigned long end;
1493
1494	if (unlikely(!check_prev_ino(leaf, key, slot, prev_key)))
1495		return -EUCLEAN;
1496	/* namelen can't be 0, so item_size == sizeof() is also invalid */
1497	if (unlikely(btrfs_item_size_nr(leaf, slot) <= sizeof(*iref))) {
1498		inode_ref_err(leaf, slot,
1499			"invalid item size, have %u expect (%zu, %u)",
1500			btrfs_item_size_nr(leaf, slot),
1501			sizeof(*iref), BTRFS_LEAF_DATA_SIZE(leaf->fs_info));
1502		return -EUCLEAN;
1503	}
1504
1505	ptr = btrfs_item_ptr_offset(leaf, slot);
1506	end = ptr + btrfs_item_size_nr(leaf, slot);
1507	while (ptr < end) {
1508		u16 namelen;
1509
1510		if (unlikely(ptr + sizeof(iref) > end)) {
1511			inode_ref_err(leaf, slot,
1512			"inode ref overflow, ptr %lu end %lu inode_ref_size %zu",
1513				ptr, end, sizeof(iref));
1514			return -EUCLEAN;
1515		}
1516
1517		iref = (struct btrfs_inode_ref *)ptr;
1518		namelen = btrfs_inode_ref_name_len(leaf, iref);
1519		if (unlikely(ptr + sizeof(*iref) + namelen > end)) {
1520			inode_ref_err(leaf, slot,
1521				"inode ref overflow, ptr %lu end %lu namelen %u",
1522				ptr, end, namelen);
1523			return -EUCLEAN;
1524		}
1525
1526		/*
1527		 * NOTE: In theory we should record all found index numbers
1528		 * to find any duplicated indexes, but that will be too time
1529		 * consuming for inodes with too many hard links.
1530		 */
1531		ptr += sizeof(*iref) + namelen;
1532	}
1533	return 0;
1534}
1535
1536/*
1537 * Common point to switch the item-specific validation.
1538 */
1539static int check_leaf_item(struct extent_buffer *leaf,
1540			   struct btrfs_key *key, int slot,
1541			   struct btrfs_key *prev_key)
1542{
1543	int ret = 0;
1544	struct btrfs_chunk *chunk;
1545
1546	switch (key->type) {
1547	case BTRFS_EXTENT_DATA_KEY:
1548		ret = check_extent_data_item(leaf, key, slot, prev_key);
1549		break;
1550	case BTRFS_EXTENT_CSUM_KEY:
1551		ret = check_csum_item(leaf, key, slot, prev_key);
1552		break;
1553	case BTRFS_DIR_ITEM_KEY:
1554	case BTRFS_DIR_INDEX_KEY:
1555	case BTRFS_XATTR_ITEM_KEY:
1556		ret = check_dir_item(leaf, key, prev_key, slot);
1557		break;
1558	case BTRFS_INODE_REF_KEY:
1559		ret = check_inode_ref(leaf, key, prev_key, slot);
1560		break;
1561	case BTRFS_BLOCK_GROUP_ITEM_KEY:
1562		ret = check_block_group_item(leaf, key, slot);
1563		break;
1564	case BTRFS_CHUNK_ITEM_KEY:
1565		chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
1566		ret = check_leaf_chunk_item(leaf, chunk, key, slot);
1567		break;
1568	case BTRFS_DEV_ITEM_KEY:
1569		ret = check_dev_item(leaf, key, slot);
1570		break;
1571	case BTRFS_INODE_ITEM_KEY:
1572		ret = check_inode_item(leaf, key, slot);
1573		break;
1574	case BTRFS_ROOT_ITEM_KEY:
1575		ret = check_root_item(leaf, key, slot);
1576		break;
1577	case BTRFS_EXTENT_ITEM_KEY:
1578	case BTRFS_METADATA_ITEM_KEY:
1579		ret = check_extent_item(leaf, key, slot);
1580		break;
1581	case BTRFS_TREE_BLOCK_REF_KEY:
1582	case BTRFS_SHARED_DATA_REF_KEY:
1583	case BTRFS_SHARED_BLOCK_REF_KEY:
1584		ret = check_simple_keyed_refs(leaf, key, slot);
1585		break;
1586	case BTRFS_EXTENT_DATA_REF_KEY:
1587		ret = check_extent_data_ref(leaf, key, slot);
1588		break;
1589	}
1590	return ret;
1591}
1592
1593static int check_leaf(struct extent_buffer *leaf, bool check_item_data)
1594{
1595	struct btrfs_fs_info *fs_info = leaf->fs_info;
1596	/* No valid key type is 0, so all key should be larger than this key */
1597	struct btrfs_key prev_key = {0, 0, 0};
1598	struct btrfs_key key;
1599	u32 nritems = btrfs_header_nritems(leaf);
1600	int slot;
1601
1602	if (unlikely(btrfs_header_level(leaf) != 0)) {
1603		generic_err(leaf, 0,
1604			"invalid level for leaf, have %d expect 0",
1605			btrfs_header_level(leaf));
1606		return -EUCLEAN;
1607	}
1608
1609	/*
1610	 * Extent buffers from a relocation tree have a owner field that
1611	 * corresponds to the subvolume tree they are based on. So just from an
1612	 * extent buffer alone we can not find out what is the id of the
1613	 * corresponding subvolume tree, so we can not figure out if the extent
1614	 * buffer corresponds to the root of the relocation tree or not. So
1615	 * skip this check for relocation trees.
1616	 */
1617	if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) {
1618		u64 owner = btrfs_header_owner(leaf);
1619
1620		/* These trees must never be empty */
1621		if (unlikely(owner == BTRFS_ROOT_TREE_OBJECTID ||
1622			     owner == BTRFS_CHUNK_TREE_OBJECTID ||
1623			     owner == BTRFS_EXTENT_TREE_OBJECTID ||
1624			     owner == BTRFS_DEV_TREE_OBJECTID ||
1625			     owner == BTRFS_FS_TREE_OBJECTID ||
1626			     owner == BTRFS_DATA_RELOC_TREE_OBJECTID)) {
1627			generic_err(leaf, 0,
1628			"invalid root, root %llu must never be empty",
1629				    owner);
1630			return -EUCLEAN;
1631		}
1632		/* Unknown tree */
1633		if (unlikely(owner == 0)) {
1634			generic_err(leaf, 0,
1635				"invalid owner, root 0 is not defined");
1636			return -EUCLEAN;
1637		}
1638		return 0;
1639	}
1640
1641	if (unlikely(nritems == 0))
1642		return 0;
1643
1644	/*
1645	 * Check the following things to make sure this is a good leaf, and
1646	 * leaf users won't need to bother with similar sanity checks:
1647	 *
1648	 * 1) key ordering
1649	 * 2) item offset and size
1650	 *    No overlap, no hole, all inside the leaf.
1651	 * 3) item content
1652	 *    If possible, do comprehensive sanity check.
1653	 *    NOTE: All checks must only rely on the item data itself.
1654	 */
1655	for (slot = 0; slot < nritems; slot++) {
1656		u32 item_end_expected;
1657		int ret;
1658
1659		btrfs_item_key_to_cpu(leaf, &key, slot);
1660
1661		/* Make sure the keys are in the right order */
1662		if (unlikely(btrfs_comp_cpu_keys(&prev_key, &key) >= 0)) {
1663			generic_err(leaf, slot,
1664	"bad key order, prev (%llu %u %llu) current (%llu %u %llu)",
1665				prev_key.objectid, prev_key.type,
1666				prev_key.offset, key.objectid, key.type,
1667				key.offset);
1668			return -EUCLEAN;
1669		}
1670
1671		/*
1672		 * Make sure the offset and ends are right, remember that the
1673		 * item data starts at the end of the leaf and grows towards the
1674		 * front.
1675		 */
1676		if (slot == 0)
1677			item_end_expected = BTRFS_LEAF_DATA_SIZE(fs_info);
1678		else
1679			item_end_expected = btrfs_item_offset_nr(leaf,
1680								 slot - 1);
1681		if (unlikely(btrfs_item_end_nr(leaf, slot) != item_end_expected)) {
1682			generic_err(leaf, slot,
1683				"unexpected item end, have %u expect %u",
1684				btrfs_item_end_nr(leaf, slot),
1685				item_end_expected);
1686			return -EUCLEAN;
1687		}
1688
1689		/*
1690		 * Check to make sure that we don't point outside of the leaf,
1691		 * just in case all the items are consistent to each other, but
1692		 * all point outside of the leaf.
1693		 */
1694		if (unlikely(btrfs_item_end_nr(leaf, slot) >
1695			     BTRFS_LEAF_DATA_SIZE(fs_info))) {
1696			generic_err(leaf, slot,
1697			"slot end outside of leaf, have %u expect range [0, %u]",
1698				btrfs_item_end_nr(leaf, slot),
1699				BTRFS_LEAF_DATA_SIZE(fs_info));
1700			return -EUCLEAN;
1701		}
1702
1703		/* Also check if the item pointer overlaps with btrfs item. */
1704		if (unlikely(btrfs_item_ptr_offset(leaf, slot) <
1705			     btrfs_item_nr_offset(slot) + sizeof(struct btrfs_item))) {
1706			generic_err(leaf, slot,
1707		"slot overlaps with its data, item end %lu data start %lu",
1708				btrfs_item_nr_offset(slot) +
1709				sizeof(struct btrfs_item),
1710				btrfs_item_ptr_offset(leaf, slot));
1711			return -EUCLEAN;
1712		}
1713
1714		if (check_item_data) {
1715			/*
1716			 * Check if the item size and content meet other
1717			 * criteria
1718			 */
1719			ret = check_leaf_item(leaf, &key, slot, &prev_key);
1720			if (unlikely(ret < 0))
1721				return ret;
1722		}
1723
1724		prev_key.objectid = key.objectid;
1725		prev_key.type = key.type;
1726		prev_key.offset = key.offset;
1727	}
1728
1729	return 0;
1730}
1731
1732int btrfs_check_leaf_full(struct extent_buffer *leaf)
1733{
1734	return check_leaf(leaf, true);
1735}
1736ALLOW_ERROR_INJECTION(btrfs_check_leaf_full, ERRNO);
1737
1738int btrfs_check_leaf_relaxed(struct extent_buffer *leaf)
1739{
1740	return check_leaf(leaf, false);
1741}
1742
1743int btrfs_check_node(struct extent_buffer *node)
1744{
1745	struct btrfs_fs_info *fs_info = node->fs_info;
1746	unsigned long nr = btrfs_header_nritems(node);
1747	struct btrfs_key key, next_key;
1748	int slot;
1749	int level = btrfs_header_level(node);
1750	u64 bytenr;
1751	int ret = 0;
1752
1753	if (unlikely(level <= 0 || level >= BTRFS_MAX_LEVEL)) {
1754		generic_err(node, 0,
1755			"invalid level for node, have %d expect [1, %d]",
1756			level, BTRFS_MAX_LEVEL - 1);
1757		return -EUCLEAN;
1758	}
1759	if (unlikely(nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(fs_info))) {
1760		btrfs_crit(fs_info,
1761"corrupt node: root=%llu block=%llu, nritems too %s, have %lu expect range [1,%u]",
1762			   btrfs_header_owner(node), node->start,
1763			   nr == 0 ? "small" : "large", nr,
1764			   BTRFS_NODEPTRS_PER_BLOCK(fs_info));
1765		return -EUCLEAN;
1766	}
1767
1768	for (slot = 0; slot < nr - 1; slot++) {
1769		bytenr = btrfs_node_blockptr(node, slot);
1770		btrfs_node_key_to_cpu(node, &key, slot);
1771		btrfs_node_key_to_cpu(node, &next_key, slot + 1);
1772
1773		if (unlikely(!bytenr)) {
1774			generic_err(node, slot,
1775				"invalid NULL node pointer");
1776			ret = -EUCLEAN;
1777			goto out;
1778		}
1779		if (unlikely(!IS_ALIGNED(bytenr, fs_info->sectorsize))) {
1780			generic_err(node, slot,
1781			"unaligned pointer, have %llu should be aligned to %u",
1782				bytenr, fs_info->sectorsize);
1783			ret = -EUCLEAN;
1784			goto out;
1785		}
1786
1787		if (unlikely(btrfs_comp_cpu_keys(&key, &next_key) >= 0)) {
1788			generic_err(node, slot,
1789	"bad key order, current (%llu %u %llu) next (%llu %u %llu)",
1790				key.objectid, key.type, key.offset,
1791				next_key.objectid, next_key.type,
1792				next_key.offset);
1793			ret = -EUCLEAN;
1794			goto out;
1795		}
1796	}
1797out:
1798	return ret;
1799}
1800ALLOW_ERROR_INJECTION(btrfs_check_node, ERRNO);
1801