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