tree-checker.c

   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * Copyright (C) Qu Wenruo 2017.  All rights reserved.
   4 */
   5
   6/*
   7 * The module is used to catch unexpected/corrupted tree block data.
   8 * Such behavior can be caused either by a fuzzed image or bugs.
   9 *
  10 * The objective is to do leaf/node validation checks when tree block is read
  11 * from disk, and check *every* possible member, so other code won't
  12 * need to checking them again.
  13 *
  14 * Due to the potential and unwanted damage, every checker needs to be
  15 * carefully reviewed otherwise so it does not prevent mount of valid images.
  16 */
  17
  18#include <linux/types.h>
  19#include <linux/stddef.h>
  20#include <linux/error-injection.h>
  21#include "messages.h"
  22#include "ctree.h"
  23#include "tree-checker.h"
  24#include "compression.h"
  25#include "volumes.h"
  26#include "misc.h"
  27#include "fs.h"
  28#include "accessors.h"
  29#include "file-item.h"
  30#include "inode-item.h"
  31#include "dir-item.h"
  32#include "extent-tree.h"
  33
  34/*
  35 * Error message should follow the following format:
  36 * corrupt <type>: <identifier>, <reason>[, <bad_value>]
  37 *
  38 * @type:	leaf or node
  39 * @identifier:	the necessary info to locate the leaf/node.
  40 * 		It's recommended to decode key.objecitd/offset if it's
  41 * 		meaningful.
  42 * @reason:	describe the error
  43 * @bad_value:	optional, it's recommended to output bad value and its
  44 *		expected value (range).
  45 *
  46 * Since comma is used to separate the components, only space is allowed
  47 * inside each component.
  48 */
  49
  50/*
  51 * Append generic "corrupt leaf/node root=%llu block=%llu slot=%d: " to @fmt.
  52 * Allows callers to customize the output.
  53 */
  54__printf(3, 4)
  55__cold
  56static void generic_err(const struct extent_buffer *eb, int slot,
  57			const char *fmt, ...)
  58{
  59	const struct btrfs_fs_info *fs_info = eb->fs_info;
  60	struct va_format vaf;
  61	va_list args;
  62
  63	va_start(args, fmt);
  64
  65	vaf.fmt = fmt;
  66	vaf.va = &args;
  67
  68	dump_page(folio_page(eb->folios[0], 0), "eb page dump");
  69	btrfs_crit(fs_info,
  70		"corrupt %s: root=%llu block=%llu slot=%d, %pV",
  71		btrfs_header_level(eb) == 0 ? "leaf" : "node",
  72		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, &vaf);
  73	va_end(args);
  74}
  75
  76/*
  77 * Customized reporter for extent data item, since its key objectid and
  78 * offset has its own meaning.
  79 */
  80__printf(3, 4)
  81__cold
  82static void file_extent_err(const struct extent_buffer *eb, int slot,
  83			    const char *fmt, ...)
  84{
  85	const struct btrfs_fs_info *fs_info = eb->fs_info;
  86	struct btrfs_key key;
  87	struct va_format vaf;
  88	va_list args;
  89
  90	btrfs_item_key_to_cpu(eb, &key, slot);
  91	va_start(args, fmt);
  92
  93	vaf.fmt = fmt;
  94	vaf.va = &args;
  95
  96	dump_page(folio_page(eb->folios[0], 0), "eb page dump");
  97	btrfs_crit(fs_info,
  98	"corrupt %s: root=%llu block=%llu slot=%d ino=%llu file_offset=%llu, %pV",
  99		btrfs_header_level(eb) == 0 ? "leaf" : "node",
 100		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
 101		key.objectid, key.offset, &vaf);
 102	va_end(args);
 103}
 104
 105/*
 106 * Return 0 if the btrfs_file_extent_##name is aligned to @alignment
 107 * Else return 1
 108 */
 109#define CHECK_FE_ALIGNED(leaf, slot, fi, name, alignment)		      \
 110({									      \
 111	if (unlikely(!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)),      \
 112				 (alignment))))				      \
 113		file_extent_err((leaf), (slot),				      \
 114	"invalid %s for file extent, have %llu, should be aligned to %u",     \
 115			(#name), btrfs_file_extent_##name((leaf), (fi)),      \
 116			(alignment));					      \
 117	(!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment)));   \
 118})
 119
 120static u64 file_extent_end(struct extent_buffer *leaf,
 121			   struct btrfs_key *key,
 122			   struct btrfs_file_extent_item *extent)
 123{
 124	u64 end;
 125	u64 len;
 126
 127	if (btrfs_file_extent_type(leaf, extent) == BTRFS_FILE_EXTENT_INLINE) {
 128		len = btrfs_file_extent_ram_bytes(leaf, extent);
 129		end = ALIGN(key->offset + len, leaf->fs_info->sectorsize);
 130	} else {
 131		len = btrfs_file_extent_num_bytes(leaf, extent);
 132		end = key->offset + len;
 133	}
 134	return end;
 135}
 136
 137/*
 138 * Customized report for dir_item, the only new important information is
 139 * key->objectid, which represents inode number
 140 */
 141__printf(3, 4)
 142__cold
 143static void dir_item_err(const struct extent_buffer *eb, int slot,
 144			 const char *fmt, ...)
 145{
 146	const struct btrfs_fs_info *fs_info = eb->fs_info;
 147	struct btrfs_key key;
 148	struct va_format vaf;
 149	va_list args;
 150
 151	btrfs_item_key_to_cpu(eb, &key, slot);
 152	va_start(args, fmt);
 153
 154	vaf.fmt = fmt;
 155	vaf.va = &args;
 156
 157	dump_page(folio_page(eb->folios[0], 0), "eb page dump");
 158	btrfs_crit(fs_info,
 159		"corrupt %s: root=%llu block=%llu slot=%d ino=%llu, %pV",
 160		btrfs_header_level(eb) == 0 ? "leaf" : "node",
 161		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
 162		key.objectid, &vaf);
 163	va_end(args);
 164}
 165
 166/*
 167 * This functions checks prev_key->objectid, to ensure current key and prev_key
 168 * share the same objectid as inode number.
 169 *
 170 * This is to detect missing INODE_ITEM in subvolume trees.
 171 *
 172 * Return true if everything is OK or we don't need to check.
 173 * Return false if anything is wrong.
 174 */
 175static bool check_prev_ino(struct extent_buffer *leaf,
 176			   struct btrfs_key *key, int slot,
 177			   struct btrfs_key *prev_key)
 178{
 179	/* No prev key, skip check */
 180	if (slot == 0)
 181		return true;
 182
 183	/* Only these key->types needs to be checked */
 184	ASSERT(key->type == BTRFS_XATTR_ITEM_KEY ||
 185	       key->type == BTRFS_INODE_REF_KEY ||
 186	       key->type == BTRFS_INODE_EXTREF_KEY ||
 187	       key->type == BTRFS_DIR_INDEX_KEY ||
 188	       key->type == BTRFS_DIR_ITEM_KEY ||
 189	       key->type == BTRFS_EXTENT_DATA_KEY, "key->type=%u", key->type);
 190
 191	/*
 192	 * Only subvolume trees along with their reloc trees need this check.
 193	 * Things like log tree doesn't follow this ino requirement.
 194	 */
 195	if (!btrfs_is_fstree(btrfs_header_owner(leaf)))
 196		return true;
 197
 198	if (key->objectid == prev_key->objectid)
 199		return true;
 200
 201	/* Error found */
 202	dir_item_err(leaf, slot,
 203		"invalid previous key objectid, have %llu expect %llu",
 204		prev_key->objectid, key->objectid);
 205	return false;
 206}
 207static int check_extent_data_item(struct extent_buffer *leaf,
 208				  struct btrfs_key *key, int slot,
 209				  struct btrfs_key *prev_key)
 210{
 211	struct btrfs_fs_info *fs_info = leaf->fs_info;
 212	struct btrfs_file_extent_item *fi;
 213	u32 sectorsize = fs_info->sectorsize;
 214	u32 item_size = btrfs_item_size(leaf, slot);
 215	u64 extent_end;
 216
 217	if (unlikely(!IS_ALIGNED(key->offset, sectorsize))) {
 218		file_extent_err(leaf, slot,
 219"unaligned file_offset for file extent, have %llu should be aligned to %u",
 220			key->offset, sectorsize);
 221		return -EUCLEAN;
 222	}
 223
 224	/*
 225	 * Previous key must have the same key->objectid (ino).
 226	 * It can be XATTR_ITEM, INODE_ITEM or just another EXTENT_DATA.
 227	 * But if objectids mismatch, it means we have a missing
 228	 * INODE_ITEM.
 229	 */
 230	if (unlikely(!check_prev_ino(leaf, key, slot, prev_key)))
 231		return -EUCLEAN;
 232
 233	fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
 234
 235	/*
 236	 * Make sure the item contains at least inline header, so the file
 237	 * extent type is not some garbage.
 238	 */
 239	if (unlikely(item_size < BTRFS_FILE_EXTENT_INLINE_DATA_START)) {
 240		file_extent_err(leaf, slot,
 241				"invalid item size, have %u expect [%zu, %u)",
 242				item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START,
 243				SZ_4K);
 244		return -EUCLEAN;
 245	}
 246	if (unlikely(btrfs_file_extent_type(leaf, fi) >=
 247		     BTRFS_NR_FILE_EXTENT_TYPES)) {
 248		file_extent_err(leaf, slot,
 249		"invalid type for file extent, have %u expect range [0, %u]",
 250			btrfs_file_extent_type(leaf, fi),
 251			BTRFS_NR_FILE_EXTENT_TYPES - 1);
 252		return -EUCLEAN;
 253	}
 254
 255	/*
 256	 * Support for new compression/encryption must introduce incompat flag,
 257	 * and must be caught in open_ctree().
 258	 */
 259	if (unlikely(btrfs_file_extent_compression(leaf, fi) >=
 260		     BTRFS_NR_COMPRESS_TYPES)) {
 261		file_extent_err(leaf, slot,
 262	"invalid compression for file extent, have %u expect range [0, %u]",
 263			btrfs_file_extent_compression(leaf, fi),
 264			BTRFS_NR_COMPRESS_TYPES - 1);
 265		return -EUCLEAN;
 266	}
 267	if (unlikely(btrfs_file_extent_encryption(leaf, fi))) {
 268		file_extent_err(leaf, slot,
 269			"invalid encryption for file extent, have %u expect 0",
 270			btrfs_file_extent_encryption(leaf, fi));
 271		return -EUCLEAN;
 272	}
 273	if (btrfs_file_extent_type(leaf, fi) == BTRFS_FILE_EXTENT_INLINE) {
 274		/* Inline extent must have 0 as key offset */
 275		if (unlikely(key->offset)) {
 276			file_extent_err(leaf, slot,
 277		"invalid file_offset for inline file extent, have %llu expect 0",
 278				key->offset);
 279			return -EUCLEAN;
 280		}
 281
 282		/* Compressed inline extent has no on-disk size, skip it */
 283		if (btrfs_file_extent_compression(leaf, fi) !=
 284		    BTRFS_COMPRESS_NONE)
 285			return 0;
 286
 287		/* Uncompressed inline extent size must match item size */
 288		if (unlikely(item_size != BTRFS_FILE_EXTENT_INLINE_DATA_START +
 289					  btrfs_file_extent_ram_bytes(leaf, fi))) {
 290			file_extent_err(leaf, slot,
 291	"invalid ram_bytes for uncompressed inline extent, have %u expect %llu",
 292				item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START +
 293				btrfs_file_extent_ram_bytes(leaf, fi));
 294			return -EUCLEAN;
 295		}
 296		return 0;
 297	}
 298
 299	/* Regular or preallocated extent has fixed item size */
 300	if (unlikely(item_size != sizeof(*fi))) {
 301		file_extent_err(leaf, slot,
 302	"invalid item size for reg/prealloc file extent, have %u expect %zu",
 303			item_size, sizeof(*fi));
 304		return -EUCLEAN;
 305	}
 306	if (unlikely(CHECK_FE_ALIGNED(leaf, slot, fi, ram_bytes, sectorsize) ||
 307		     CHECK_FE_ALIGNED(leaf, slot, fi, disk_bytenr, sectorsize) ||
 308		     CHECK_FE_ALIGNED(leaf, slot, fi, disk_num_bytes, sectorsize) ||
 309		     CHECK_FE_ALIGNED(leaf, slot, fi, offset, sectorsize) ||
 310		     CHECK_FE_ALIGNED(leaf, slot, fi, num_bytes, sectorsize)))
 311		return -EUCLEAN;
 312
 313	/* Catch extent end overflow */
 314	if (unlikely(check_add_overflow(btrfs_file_extent_num_bytes(leaf, fi),
 315					key->offset, &extent_end))) {
 316		file_extent_err(leaf, slot,
 317	"extent end overflow, have file offset %llu extent num bytes %llu",
 318				key->offset,
 319				btrfs_file_extent_num_bytes(leaf, fi));
 320		return -EUCLEAN;
 321	}
 322
 323	/*
 324	 * Check that no two consecutive file extent items, in the same leaf,
 325	 * present ranges that overlap each other.
 326	 */
 327	if (slot > 0 &&
 328	    prev_key->objectid == key->objectid &&
 329	    prev_key->type == BTRFS_EXTENT_DATA_KEY) {
 330		struct btrfs_file_extent_item *prev_fi;
 331		u64 prev_end;
 332
 333		prev_fi = btrfs_item_ptr(leaf, slot - 1,
 334					 struct btrfs_file_extent_item);
 335		prev_end = file_extent_end(leaf, prev_key, prev_fi);
 336		if (unlikely(prev_end > key->offset)) {
 337			file_extent_err(leaf, slot - 1,
 338"file extent end range (%llu) goes beyond start offset (%llu) of the next file extent",
 339					prev_end, key->offset);
 340			return -EUCLEAN;
 341		}
 342	}
 343
 344	/*
 345	 * For non-compressed data extents, ram_bytes should match its
 346	 * disk_num_bytes.
 347	 * However we do not really utilize ram_bytes in this case, so this check
 348	 * is only optional for DEBUG builds for developers to catch the
 349	 * unexpected behaviors.
 350	 */
 351	if (IS_ENABLED(CONFIG_BTRFS_DEBUG) &&
 352	    btrfs_file_extent_compression(leaf, fi) == BTRFS_COMPRESS_NONE &&
 353	    btrfs_file_extent_disk_bytenr(leaf, fi)) {
 354		if (WARN_ON(btrfs_file_extent_ram_bytes(leaf, fi) !=
 355			    btrfs_file_extent_disk_num_bytes(leaf, fi)))
 356			file_extent_err(leaf, slot,
 357"mismatch ram_bytes (%llu) and disk_num_bytes (%llu) for non-compressed extent",
 358					btrfs_file_extent_ram_bytes(leaf, fi),
 359					btrfs_file_extent_disk_num_bytes(leaf, fi));
 360	}
 361
 362	return 0;
 363}
 364
 365static int check_csum_item(struct extent_buffer *leaf, struct btrfs_key *key,
 366			   int slot, struct btrfs_key *prev_key)
 367{
 368	struct btrfs_fs_info *fs_info = leaf->fs_info;
 369	u32 sectorsize = fs_info->sectorsize;
 370	const u32 csumsize = fs_info->csum_size;
 371
 372	if (unlikely(key->objectid != BTRFS_EXTENT_CSUM_OBJECTID)) {
 373		generic_err(leaf, slot,
 374		"invalid key objectid for csum item, have %llu expect %llu",
 375			key->objectid, BTRFS_EXTENT_CSUM_OBJECTID);
 376		return -EUCLEAN;
 377	}
 378	if (unlikely(!IS_ALIGNED(key->offset, sectorsize))) {
 379		generic_err(leaf, slot,
 380	"unaligned key offset for csum item, have %llu should be aligned to %u",
 381			key->offset, sectorsize);
 382		return -EUCLEAN;
 383	}
 384	if (unlikely(!IS_ALIGNED(btrfs_item_size(leaf, slot), csumsize))) {
 385		generic_err(leaf, slot,
 386	"unaligned item size for csum item, have %u should be aligned to %u",
 387			btrfs_item_size(leaf, slot), csumsize);
 388		return -EUCLEAN;
 389	}
 390	if (slot > 0 && prev_key->type == BTRFS_EXTENT_CSUM_KEY) {
 391		u64 prev_csum_end;
 392		u32 prev_item_size;
 393
 394		prev_item_size = btrfs_item_size(leaf, slot - 1);
 395		prev_csum_end = (prev_item_size / csumsize) * sectorsize;
 396		prev_csum_end += prev_key->offset;
 397		if (unlikely(prev_csum_end > key->offset)) {
 398			generic_err(leaf, slot - 1,
 399"csum end range (%llu) goes beyond the start range (%llu) of the next csum item",
 400				    prev_csum_end, key->offset);
 401			return -EUCLEAN;
 402		}
 403	}
 404	return 0;
 405}
 406
 407/* Inode item error output has the same format as dir_item_err() */
 408#define inode_item_err(eb, slot, fmt, ...)			\
 409	dir_item_err(eb, slot, fmt, __VA_ARGS__)
 410
 411static int check_inode_key(struct extent_buffer *leaf, struct btrfs_key *key,
 412			   int slot)
 413{
 414	struct btrfs_key item_key;
 415	bool is_inode_item;
 416
 417	btrfs_item_key_to_cpu(leaf, &item_key, slot);
 418	is_inode_item = (item_key.type == BTRFS_INODE_ITEM_KEY);
 419
 420	/* For XATTR_ITEM, location key should be all 0 */
 421	if (item_key.type == BTRFS_XATTR_ITEM_KEY) {
 422		if (unlikely(key->objectid != 0 || key->type != 0 ||
 423			     key->offset != 0))
 424			return -EUCLEAN;
 425		return 0;
 426	}
 427
 428	if (unlikely((key->objectid < BTRFS_FIRST_FREE_OBJECTID ||
 429		      key->objectid > BTRFS_LAST_FREE_OBJECTID) &&
 430		     key->objectid != BTRFS_ROOT_TREE_DIR_OBJECTID &&
 431		     key->objectid != BTRFS_FREE_INO_OBJECTID)) {
 432		if (is_inode_item) {
 433			generic_err(leaf, slot,
 434	"invalid key objectid: has %llu expect %llu or [%llu, %llu] or %llu",
 435				key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID,
 436				BTRFS_FIRST_FREE_OBJECTID,
 437				BTRFS_LAST_FREE_OBJECTID,
 438				BTRFS_FREE_INO_OBJECTID);
 439		} else {
 440			dir_item_err(leaf, slot,
 441"invalid location key objectid: has %llu expect %llu or [%llu, %llu] or %llu",
 442				key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID,
 443				BTRFS_FIRST_FREE_OBJECTID,
 444				BTRFS_LAST_FREE_OBJECTID,
 445				BTRFS_FREE_INO_OBJECTID);
 446		}
 447		return -EUCLEAN;
 448	}
 449	if (unlikely(key->offset != 0)) {
 450		if (is_inode_item)
 451			inode_item_err(leaf, slot,
 452				       "invalid key offset: has %llu expect 0",
 453				       key->offset);
 454		else
 455			dir_item_err(leaf, slot,
 456				"invalid location key offset:has %llu expect 0",
 457				key->offset);
 458		return -EUCLEAN;
 459	}
 460	return 0;
 461}
 462
 463static int check_root_key(struct extent_buffer *leaf, struct btrfs_key *key,
 464			  int slot)
 465{
 466	struct btrfs_key item_key;
 467	bool is_root_item;
 468
 469	btrfs_item_key_to_cpu(leaf, &item_key, slot);
 470	is_root_item = (item_key.type == BTRFS_ROOT_ITEM_KEY);
 471
 472	/*
 473	 * Bad rootid for reloc trees.
 474	 *
 475	 * Reloc trees are only for subvolume trees, other trees only need
 476	 * to be COWed to be relocated.
 477	 */
 478	if (unlikely(is_root_item && key->objectid == BTRFS_TREE_RELOC_OBJECTID &&
 479		     !btrfs_is_fstree(key->offset))) {
 480		generic_err(leaf, slot,
 481		"invalid reloc tree for root %lld, root id is not a subvolume tree",
 482			    key->offset);
 483		return -EUCLEAN;
 484	}
 485
 486	/* No such tree id */
 487	if (unlikely(key->objectid == 0)) {
 488		if (is_root_item)
 489			generic_err(leaf, slot, "invalid root id 0");
 490		else
 491			dir_item_err(leaf, slot,
 492				     "invalid location key root id 0");
 493		return -EUCLEAN;
 494	}
 495
 496	/* DIR_ITEM/INDEX/INODE_REF is not allowed to point to non-fs trees */
 497	if (unlikely(!btrfs_is_fstree(key->objectid) && !is_root_item)) {
 498		dir_item_err(leaf, slot,
 499		"invalid location key objectid, have %llu expect [%llu, %llu]",
 500				key->objectid, BTRFS_FIRST_FREE_OBJECTID,
 501				BTRFS_LAST_FREE_OBJECTID);
 502		return -EUCLEAN;
 503	}
 504
 505	/*
 506	 * ROOT_ITEM with non-zero offset means this is a snapshot, created at
 507	 * @offset transid.
 508	 * Furthermore, for location key in DIR_ITEM, its offset is always -1.
 509	 *
 510	 * So here we only check offset for reloc tree whose key->offset must
 511	 * be a valid tree.
 512	 */
 513	if (unlikely(key->objectid == BTRFS_TREE_RELOC_OBJECTID &&
 514		     key->offset == 0)) {
 515		generic_err(leaf, slot, "invalid root id 0 for reloc tree");
 516		return -EUCLEAN;
 517	}
 518	return 0;
 519}
 520
 521static int check_dir_item(struct extent_buffer *leaf,
 522			  struct btrfs_key *key, struct btrfs_key *prev_key,
 523			  int slot)
 524{
 525	struct btrfs_fs_info *fs_info = leaf->fs_info;
 526	struct btrfs_dir_item *di;
 527	u32 item_size = btrfs_item_size(leaf, slot);
 528	u32 cur = 0;
 529
 530	if (unlikely(!check_prev_ino(leaf, key, slot, prev_key)))
 531		return -EUCLEAN;
 532
 533	di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
 534	while (cur < item_size) {
 535		struct btrfs_key location_key;
 536		u32 name_len;
 537		u32 data_len;
 538		u32 max_name_len;
 539		u32 total_size;
 540		u32 name_hash;
 541		u8 dir_type;
 542		int ret;
 543
 544		/* header itself should not cross item boundary */
 545		if (unlikely(cur + sizeof(*di) > item_size)) {
 546			dir_item_err(leaf, slot,
 547		"dir item header crosses item boundary, have %zu boundary %u",
 548				cur + sizeof(*di), item_size);
 549			return -EUCLEAN;
 550		}
 551
 552		/* Location key check */
 553		btrfs_dir_item_key_to_cpu(leaf, di, &location_key);
 554		if (location_key.type == BTRFS_ROOT_ITEM_KEY) {
 555			ret = check_root_key(leaf, &location_key, slot);
 556			if (unlikely(ret < 0))
 557				return ret;
 558		} else if (location_key.type == BTRFS_INODE_ITEM_KEY ||
 559			   location_key.type == 0) {
 560			ret = check_inode_key(leaf, &location_key, slot);
 561			if (unlikely(ret < 0))
 562				return ret;
 563		} else {
 564			dir_item_err(leaf, slot,
 565			"invalid location key type, have %u, expect %u or %u",
 566				     location_key.type, BTRFS_ROOT_ITEM_KEY,
 567				     BTRFS_INODE_ITEM_KEY);
 568			return -EUCLEAN;
 569		}
 570
 571		/* dir type check */
 572		dir_type = btrfs_dir_ftype(leaf, di);
 573		if (unlikely(dir_type <= BTRFS_FT_UNKNOWN ||
 574			     dir_type >= BTRFS_FT_MAX)) {
 575			dir_item_err(leaf, slot,
 576			"invalid dir item type, have %u expect (0, %u)",
 577				dir_type, BTRFS_FT_MAX);
 578			return -EUCLEAN;
 579		}
 580
 581		if (unlikely(key->type == BTRFS_XATTR_ITEM_KEY &&
 582			     dir_type != BTRFS_FT_XATTR)) {
 583			dir_item_err(leaf, slot,
 584		"invalid dir item type for XATTR key, have %u expect %u",
 585				dir_type, BTRFS_FT_XATTR);
 586			return -EUCLEAN;
 587		}
 588		if (unlikely(dir_type == BTRFS_FT_XATTR &&
 589			     key->type != BTRFS_XATTR_ITEM_KEY)) {
 590			dir_item_err(leaf, slot,
 591			"xattr dir type found for non-XATTR key");
 592			return -EUCLEAN;
 593		}
 594		if (dir_type == BTRFS_FT_XATTR)
 595			max_name_len = XATTR_NAME_MAX;
 596		else
 597			max_name_len = BTRFS_NAME_LEN;
 598
 599		/* Name/data length check */
 600		name_len = btrfs_dir_name_len(leaf, di);
 601		data_len = btrfs_dir_data_len(leaf, di);
 602		if (unlikely(name_len > max_name_len)) {
 603			dir_item_err(leaf, slot,
 604			"dir item name len too long, have %u max %u",
 605				name_len, max_name_len);
 606			return -EUCLEAN;
 607		}
 608		if (unlikely(name_len + data_len > BTRFS_MAX_XATTR_SIZE(fs_info))) {
 609			dir_item_err(leaf, slot,
 610			"dir item name and data len too long, have %u max %u",
 611				name_len + data_len,
 612				BTRFS_MAX_XATTR_SIZE(fs_info));
 613			return -EUCLEAN;
 614		}
 615
 616		if (unlikely(data_len && dir_type != BTRFS_FT_XATTR)) {
 617			dir_item_err(leaf, slot,
 618			"dir item with invalid data len, have %u expect 0",
 619				data_len);
 620			return -EUCLEAN;
 621		}
 622
 623		total_size = sizeof(*di) + name_len + data_len;
 624
 625		/* header and name/data should not cross item boundary */
 626		if (unlikely(cur + total_size > item_size)) {
 627			dir_item_err(leaf, slot,
 628		"dir item data crosses item boundary, have %u boundary %u",
 629				cur + total_size, item_size);
 630			return -EUCLEAN;
 631		}
 632
 633		/*
 634		 * Special check for XATTR/DIR_ITEM, as key->offset is name
 635		 * hash, should match its name
 636		 */
 637		if (key->type == BTRFS_DIR_ITEM_KEY ||
 638		    key->type == BTRFS_XATTR_ITEM_KEY) {
 639			char namebuf[MAX(BTRFS_NAME_LEN, XATTR_NAME_MAX)];
 640
 641			read_extent_buffer(leaf, namebuf,
 642					(unsigned long)(di + 1), name_len);
 643			name_hash = btrfs_name_hash(namebuf, name_len);
 644			if (unlikely(key->offset != name_hash)) {
 645				dir_item_err(leaf, slot,
 646		"name hash mismatch with key, have 0x%016x expect 0x%016llx",
 647					name_hash, key->offset);
 648				return -EUCLEAN;
 649			}
 650		}
 651		cur += total_size;
 652		di = (struct btrfs_dir_item *)((void *)di + total_size);
 653	}
 654	return 0;
 655}
 656
 657__printf(3, 4)
 658__cold
 659static void block_group_err(const struct extent_buffer *eb, int slot,
 660			    const char *fmt, ...)
 661{
 662	const struct btrfs_fs_info *fs_info = eb->fs_info;
 663	struct btrfs_key key;
 664	struct va_format vaf;
 665	va_list args;
 666
 667	btrfs_item_key_to_cpu(eb, &key, slot);
 668	va_start(args, fmt);
 669
 670	vaf.fmt = fmt;
 671	vaf.va = &args;
 672
 673	dump_page(folio_page(eb->folios[0], 0), "eb page dump");
 674	btrfs_crit(fs_info,
 675	"corrupt %s: root=%llu block=%llu slot=%d bg_start=%llu bg_len=%llu, %pV",
 676		btrfs_header_level(eb) == 0 ? "leaf" : "node",
 677		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
 678		key.objectid, key.offset, &vaf);
 679	va_end(args);
 680}
 681
 682static int check_block_group_item(struct extent_buffer *leaf,
 683				  struct btrfs_key *key, int slot)
 684{
 685	struct btrfs_fs_info *fs_info = leaf->fs_info;
 686	struct btrfs_block_group_item bgi;
 687	u32 item_size = btrfs_item_size(leaf, slot);
 688	u64 chunk_objectid;
 689	u64 flags;
 690	u64 type;
 691
 692	/*
 693	 * Here we don't really care about alignment since extent allocator can
 694	 * handle it.  We care more about the size.
 695	 */
 696	if (unlikely(key->offset == 0)) {
 697		block_group_err(leaf, slot,
 698				"invalid block group size 0");
 699		return -EUCLEAN;
 700	}
 701
 702	if (unlikely(item_size != sizeof(bgi))) {
 703		block_group_err(leaf, slot,
 704			"invalid item size, have %u expect %zu",
 705				item_size, sizeof(bgi));
 706		return -EUCLEAN;
 707	}
 708
 709	read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot),
 710			   sizeof(bgi));
 711	chunk_objectid = btrfs_stack_block_group_chunk_objectid(&bgi);
 712	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
 713		/*
 714		 * We don't init the nr_global_roots until we load the global
 715		 * roots, so this could be 0 at mount time.  If it's 0 we'll
 716		 * just assume we're fine, and later we'll check against our
 717		 * actual value.
 718		 */
 719		if (unlikely(fs_info->nr_global_roots &&
 720			     chunk_objectid >= fs_info->nr_global_roots)) {
 721			block_group_err(leaf, slot,
 722	"invalid block group global root id, have %llu, needs to be <= %llu",
 723					chunk_objectid,
 724					fs_info->nr_global_roots);
 725			return -EUCLEAN;
 726		}
 727	} else if (unlikely(chunk_objectid != BTRFS_FIRST_CHUNK_TREE_OBJECTID)) {
 728		block_group_err(leaf, slot,
 729		"invalid block group chunk objectid, have %llu expect %llu",
 730				btrfs_stack_block_group_chunk_objectid(&bgi),
 731				BTRFS_FIRST_CHUNK_TREE_OBJECTID);
 732		return -EUCLEAN;
 733	}
 734
 735	if (unlikely(btrfs_stack_block_group_used(&bgi) > key->offset)) {
 736		block_group_err(leaf, slot,
 737			"invalid block group used, have %llu expect [0, %llu)",
 738				btrfs_stack_block_group_used(&bgi), key->offset);
 739		return -EUCLEAN;
 740	}
 741
 742	flags = btrfs_stack_block_group_flags(&bgi);
 743	if (unlikely(hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) > 1)) {
 744		block_group_err(leaf, slot,
 745"invalid profile flags, have 0x%llx (%lu bits set) expect no more than 1 bit set",
 746			flags & BTRFS_BLOCK_GROUP_PROFILE_MASK,
 747			hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK));
 748		return -EUCLEAN;
 749	}
 750
 751	type = flags & BTRFS_BLOCK_GROUP_TYPE_MASK;
 752	if (unlikely(type != BTRFS_BLOCK_GROUP_DATA &&
 753		     type != BTRFS_BLOCK_GROUP_METADATA &&
 754		     type != BTRFS_BLOCK_GROUP_SYSTEM &&
 755		     type != (BTRFS_BLOCK_GROUP_METADATA |
 756			      BTRFS_BLOCK_GROUP_DATA))) {
 757		block_group_err(leaf, slot,
 758"invalid type, have 0x%llx (%lu bits set) expect either 0x%llx, 0x%llx, 0x%llx or 0x%llx",
 759			type, hweight64(type),
 760			BTRFS_BLOCK_GROUP_DATA, BTRFS_BLOCK_GROUP_METADATA,
 761			BTRFS_BLOCK_GROUP_SYSTEM,
 762			BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA);
 763		return -EUCLEAN;
 764	}
 765	return 0;
 766}
 767
 768__printf(5, 6)
 769__cold
 770static void chunk_err(const struct btrfs_fs_info *fs_info,
 771		      const struct extent_buffer *leaf,
 772		      const struct btrfs_chunk *chunk, u64 logical,
 773		      const char *fmt, ...)
 774{
 775	bool is_sb = !leaf;
 776	struct va_format vaf;
 777	va_list args;
 778	int i;
 779	int slot = -1;
 780
 781	if (!is_sb) {
 782		/*
 783		 * Get the slot number by iterating through all slots, this
 784		 * would provide better readability.
 785		 */
 786		for (i = 0; i < btrfs_header_nritems(leaf); i++) {
 787			if (btrfs_item_ptr_offset(leaf, i) ==
 788					(unsigned long)chunk) {
 789				slot = i;
 790				break;
 791			}
 792		}
 793	}
 794	va_start(args, fmt);
 795	vaf.fmt = fmt;
 796	vaf.va = &args;
 797
 798	if (is_sb)
 799		btrfs_crit(fs_info,
 800		"corrupt superblock syschunk array: chunk_start=%llu, %pV",
 801			   logical, &vaf);
 802	else
 803		btrfs_crit(fs_info,
 804	"corrupt leaf: root=%llu block=%llu slot=%d chunk_start=%llu, %pV",
 805			   BTRFS_CHUNK_TREE_OBJECTID, leaf->start, slot,
 806			   logical, &vaf);
 807	va_end(args);
 808}
 809
 810/*
 811 * The common chunk check which could also work on super block sys chunk array.
 812 *
 813 * If @leaf is NULL, then @chunk must be an on-stack chunk item.
 814 * (For superblock sys_chunk array, and fs_info->sectorsize is unreliable)
 815 *
 816 * Return -EUCLEAN if anything is corrupted.
 817 * Return 0 if everything is OK.
 818 */
 819int btrfs_check_chunk_valid(const struct btrfs_fs_info *fs_info,
 820			    const struct extent_buffer *leaf,
 821			    const struct btrfs_chunk *chunk, u64 logical,
 822			    u32 sectorsize)
 823{
 824	u64 length;
 825	u64 chunk_end;
 826	u64 stripe_len;
 827	u16 num_stripes;
 828	u16 sub_stripes;
 829	u64 type;
 830	u64 features;
 831	u32 chunk_sector_size;
 832	bool mixed = false;
 833	int raid_index;
 834	int nparity;
 835	int ncopies;
 836
 837	if (leaf) {
 838		length = btrfs_chunk_length(leaf, chunk);
 839		stripe_len = btrfs_chunk_stripe_len(leaf, chunk);
 840		num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
 841		sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk);
 842		type = btrfs_chunk_type(leaf, chunk);
 843		chunk_sector_size = btrfs_chunk_sector_size(leaf, chunk);
 844	} else {
 845		length = btrfs_stack_chunk_length(chunk);
 846		stripe_len = btrfs_stack_chunk_stripe_len(chunk);
 847		num_stripes = btrfs_stack_chunk_num_stripes(chunk);
 848		sub_stripes = btrfs_stack_chunk_sub_stripes(chunk);
 849		type = btrfs_stack_chunk_type(chunk);
 850		chunk_sector_size = btrfs_stack_chunk_sector_size(chunk);
 851	}
 852	raid_index = btrfs_bg_flags_to_raid_index(type);
 853	ncopies = btrfs_raid_array[raid_index].ncopies;
 854	nparity = btrfs_raid_array[raid_index].nparity;
 855
 856	if (unlikely(!num_stripes)) {
 857		chunk_err(fs_info, leaf, chunk, logical,
 858			  "invalid chunk num_stripes, have %u", num_stripes);
 859		return -EUCLEAN;
 860	}
 861	if (unlikely(num_stripes < ncopies)) {
 862		chunk_err(fs_info, leaf, chunk, logical,
 863			  "invalid chunk num_stripes < ncopies, have %u < %d",
 864			  num_stripes, ncopies);
 865		return -EUCLEAN;
 866	}
 867	if (unlikely(nparity && num_stripes == nparity)) {
 868		chunk_err(fs_info, leaf, chunk, logical,
 869			  "invalid chunk num_stripes == nparity, have %u == %d",
 870			  num_stripes, nparity);
 871		return -EUCLEAN;
 872	}
 873	if (unlikely(!IS_ALIGNED(logical, sectorsize))) {
 874		chunk_err(fs_info, leaf, chunk, logical,
 875		"invalid chunk logical, have %llu should aligned to %u",
 876			  logical, sectorsize);
 877		return -EUCLEAN;
 878	}
 879	if (unlikely(chunk_sector_size != sectorsize)) {
 880		chunk_err(fs_info, leaf, chunk, logical,
 881			  "invalid chunk sectorsize, have %u expect %u",
 882			  chunk_sector_size, sectorsize);
 883		return -EUCLEAN;
 884	}
 885	if (unlikely(!length || !IS_ALIGNED(length, sectorsize))) {
 886		chunk_err(fs_info, leaf, chunk, logical,
 887			  "invalid chunk length, have %llu", length);
 888		return -EUCLEAN;
 889	}
 890	if (unlikely(check_add_overflow(logical, length, &chunk_end))) {
 891		chunk_err(fs_info, leaf, chunk, logical,
 892"invalid chunk logical start and length, have logical start %llu length %llu",
 893			  logical, length);
 894		return -EUCLEAN;
 895	}
 896	if (unlikely(!is_power_of_2(stripe_len) || stripe_len != BTRFS_STRIPE_LEN)) {
 897		chunk_err(fs_info, leaf, chunk, logical,
 898			  "invalid chunk stripe length: %llu",
 899			  stripe_len);
 900		return -EUCLEAN;
 901	}
 902	/*
 903	 * We artificially limit the chunk size, so that the number of stripes
 904	 * inside a chunk can be fit into a U32.  The current limit (256G) is
 905	 * way too large for real world usage anyway, and it's also much larger
 906	 * than our existing limit (10G).
 907	 *
 908	 * Thus it should be a good way to catch obvious bitflips.
 909	 */
 910	if (unlikely(length >= btrfs_stripe_nr_to_offset(U32_MAX))) {
 911		chunk_err(fs_info, leaf, chunk, logical,
 912			  "chunk length too large: have %llu limit %llu",
 913			  length, btrfs_stripe_nr_to_offset(U32_MAX));
 914		return -EUCLEAN;
 915	}
 916	if (unlikely(type & ~(BTRFS_BLOCK_GROUP_TYPE_MASK |
 917			      BTRFS_BLOCK_GROUP_PROFILE_MASK))) {
 918		chunk_err(fs_info, leaf, chunk, logical,
 919			  "unrecognized chunk type: 0x%llx",
 920			  ~(BTRFS_BLOCK_GROUP_TYPE_MASK |
 921			    BTRFS_BLOCK_GROUP_PROFILE_MASK) & type);
 922		return -EUCLEAN;
 923	}
 924
 925	if (unlikely(!has_single_bit_set(type & BTRFS_BLOCK_GROUP_PROFILE_MASK) &&
 926		     (type & BTRFS_BLOCK_GROUP_PROFILE_MASK) != 0)) {
 927		chunk_err(fs_info, leaf, chunk, logical,
 928		"invalid chunk profile flag: 0x%llx, expect 0 or 1 bit set",
 929			  type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
 930		return -EUCLEAN;
 931	}
 932	if (unlikely((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == 0)) {
 933		chunk_err(fs_info, leaf, chunk, logical,
 934	"missing chunk type flag, have 0x%llx one bit must be set in 0x%llx",
 935			  type, BTRFS_BLOCK_GROUP_TYPE_MASK);
 936		return -EUCLEAN;
 937	}
 938
 939	if (unlikely((type & BTRFS_BLOCK_GROUP_SYSTEM) &&
 940		     (type & (BTRFS_BLOCK_GROUP_METADATA |
 941			      BTRFS_BLOCK_GROUP_DATA)))) {
 942		chunk_err(fs_info, leaf, chunk, logical,
 943			  "system chunk with data or metadata type: 0x%llx",
 944			  type);
 945		return -EUCLEAN;
 946	}
 947
 948	features = btrfs_super_incompat_flags(fs_info->super_copy);
 949	if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)
 950		mixed = true;
 951
 952	if (!mixed) {
 953		if (unlikely((type & BTRFS_BLOCK_GROUP_METADATA) &&
 954			     (type & BTRFS_BLOCK_GROUP_DATA))) {
 955			chunk_err(fs_info, leaf, chunk, logical,
 956			"mixed chunk type in non-mixed mode: 0x%llx", type);
 957			return -EUCLEAN;
 958		}
 959	}
 960
 961	if (unlikely((type & BTRFS_BLOCK_GROUP_RAID10 &&
 962		      sub_stripes != btrfs_raid_array[BTRFS_RAID_RAID10].sub_stripes) ||
 963		     (type & BTRFS_BLOCK_GROUP_RAID1 &&
 964		      num_stripes != btrfs_raid_array[BTRFS_RAID_RAID1].devs_min) ||
 965		     (type & BTRFS_BLOCK_GROUP_RAID1C3 &&
 966		      num_stripes != btrfs_raid_array[BTRFS_RAID_RAID1C3].devs_min) ||
 967		     (type & BTRFS_BLOCK_GROUP_RAID1C4 &&
 968		      num_stripes != btrfs_raid_array[BTRFS_RAID_RAID1C4].devs_min) ||
 969		     (type & BTRFS_BLOCK_GROUP_RAID5 &&
 970		      num_stripes < btrfs_raid_array[BTRFS_RAID_RAID5].devs_min) ||
 971		     (type & BTRFS_BLOCK_GROUP_RAID6 &&
 972		      num_stripes < btrfs_raid_array[BTRFS_RAID_RAID6].devs_min) ||
 973		     (type & BTRFS_BLOCK_GROUP_DUP &&
 974		      num_stripes != btrfs_raid_array[BTRFS_RAID_DUP].dev_stripes) ||
 975		     ((type & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 &&
 976		      num_stripes != btrfs_raid_array[BTRFS_RAID_SINGLE].dev_stripes))) {
 977		chunk_err(fs_info, leaf, chunk, logical,
 978			"invalid num_stripes:sub_stripes %u:%u for profile %llu",
 979			num_stripes, sub_stripes,
 980			type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
 981		return -EUCLEAN;
 982	}
 983
 984	return 0;
 985}
 986
 987/*
 988 * Enhanced version of chunk item checker.
 989 *
 990 * The common btrfs_check_chunk_valid() doesn't check item size since it needs
 991 * to work on super block sys_chunk_array which doesn't have full item ptr.
 992 */
 993static int check_leaf_chunk_item(struct extent_buffer *leaf,
 994				 struct btrfs_chunk *chunk,
 995				 struct btrfs_key *key, int slot)
 996{
 997	struct btrfs_fs_info *fs_info = leaf->fs_info;
 998	int num_stripes;
 999
1000	if (unlikely(btrfs_item_size(leaf, slot) < sizeof(struct btrfs_chunk))) {
1001		chunk_err(fs_info, leaf, chunk, key->offset,
1002			"invalid chunk item size: have %u expect [%zu, %u)",
1003			btrfs_item_size(leaf, slot),
1004			sizeof(struct btrfs_chunk),
1005			BTRFS_LEAF_DATA_SIZE(fs_info));
1006		return -EUCLEAN;
1007	}
1008
1009	num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
1010	/* Let btrfs_check_chunk_valid() handle this error type */
1011	if (num_stripes == 0)
1012		goto out;
1013
1014	if (unlikely(btrfs_chunk_item_size(num_stripes) !=
1015		     btrfs_item_size(leaf, slot))) {
1016		chunk_err(fs_info, leaf, chunk, key->offset,
1017			"invalid chunk item size: have %u expect %lu",
1018			btrfs_item_size(leaf, slot),
1019			btrfs_chunk_item_size(num_stripes));
1020		return -EUCLEAN;
1021	}
1022out:
1023	return btrfs_check_chunk_valid(fs_info, leaf, chunk, key->offset,
1024				       fs_info->sectorsize);
1025}
1026
1027__printf(3, 4)
1028__cold
1029static void dev_item_err(const struct extent_buffer *eb, int slot,
1030			 const char *fmt, ...)
1031{
1032	struct btrfs_key key;
1033	struct va_format vaf;
1034	va_list args;
1035
1036	btrfs_item_key_to_cpu(eb, &key, slot);
1037	va_start(args, fmt);
1038
1039	vaf.fmt = fmt;
1040	vaf.va = &args;
1041
1042	dump_page(folio_page(eb->folios[0], 0), "eb page dump");
1043	btrfs_crit(eb->fs_info,
1044	"corrupt %s: root=%llu block=%llu slot=%d devid=%llu %pV",
1045		btrfs_header_level(eb) == 0 ? "leaf" : "node",
1046		btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
1047		key.objectid, &vaf);
1048	va_end(args);
1049}
1050
1051static int check_dev_item(struct extent_buffer *leaf,
1052			  struct btrfs_key *key, int slot)
1053{
1054	struct btrfs_dev_item *ditem;
1055	const u32 item_size = btrfs_item_size(leaf, slot);
1056
1057	if (unlikely(key->objectid != BTRFS_DEV_ITEMS_OBJECTID)) {
1058		dev_item_err(leaf, slot,
1059			     "invalid objectid: has=%llu expect=%llu",
1060			     key->objectid, BTRFS_DEV_ITEMS_OBJECTID);
1061		return -EUCLEAN;
1062	}
1063
1064	if (unlikely(item_size != sizeof(*ditem))) {
1065		dev_item_err(leaf, slot, "invalid item size: has %u expect %zu",
1066			     item_size, sizeof(*ditem));
1067		return -EUCLEAN;
1068	}
1069
1070	ditem = btrfs_item_ptr(leaf, slot, struct btrfs_dev_item);
1071	if (unlikely(btrfs_device_id(leaf, ditem) != key->offset)) {
1072		dev_item_err(leaf, slot,
1073			     "devid mismatch: key has=%llu item has=%llu",
1074			     key->offset, btrfs_device_id(leaf, ditem));
1075		return -EUCLEAN;
1076	}
1077
1078	/*
1079	 * For device total_bytes, we don't have reliable way to check it, as
1080	 * it can be 0 for device removal. Device size check can only be done
1081	 * by dev extents check.
1082	 */
1083	if (unlikely(btrfs_device_bytes_used(leaf, ditem) >
1084		     btrfs_device_total_bytes(leaf, ditem))) {
1085		dev_item_err(leaf, slot,
1086			     "invalid bytes used: have %llu expect [0, %llu]",
1087			     btrfs_device_bytes_used(leaf, ditem),
1088			     btrfs_device_total_bytes(leaf, ditem));
1089		return -EUCLEAN;
1090	}
1091	/*
1092	 * Remaining members like io_align/type/gen/dev_group aren't really
1093	 * utilized.  Skip them to make later usage of them easier.
1094	 */
1095	return 0;
1096}
1097
1098static int check_inode_item(struct extent_buffer *leaf,
1099			    struct btrfs_key *key, int slot)
1100{
1101	struct btrfs_fs_info *fs_info = leaf->fs_info;
1102	struct btrfs_inode_item *iitem;
1103	u64 super_gen = btrfs_super_generation(fs_info->super_copy);
1104	u32 valid_mask = (S_IFMT | S_ISUID | S_ISGID | S_ISVTX | 0777);
1105	const u32 item_size = btrfs_item_size(leaf, slot);
1106	u32 mode;
1107	int ret;
1108	u32 flags;
1109	u32 ro_flags;
1110
1111	ret = check_inode_key(leaf, key, slot);
1112	if (unlikely(ret < 0))
1113		return ret;
1114
1115	if (unlikely(item_size != sizeof(*iitem))) {
1116		generic_err(leaf, slot, "invalid item size: has %u expect %zu",
1117			    item_size, sizeof(*iitem));
1118		return -EUCLEAN;
1119	}
1120
1121	iitem = btrfs_item_ptr(leaf, slot, struct btrfs_inode_item);
1122
1123	/* Here we use super block generation + 1 to handle log tree */
1124	if (unlikely(btrfs_inode_generation(leaf, iitem) > super_gen + 1)) {
1125		inode_item_err(leaf, slot,
1126			"invalid inode generation: has %llu expect (0, %llu]",
1127			       btrfs_inode_generation(leaf, iitem),
1128			       super_gen + 1);
1129		return -EUCLEAN;
1130	}
1131	/* Note for ROOT_TREE_DIR_ITEM, mkfs could set its transid 0 */
1132	if (unlikely(btrfs_inode_transid(leaf, iitem) > super_gen + 1)) {
1133		inode_item_err(leaf, slot,
1134			"invalid inode transid: has %llu expect [0, %llu]",
1135			       btrfs_inode_transid(leaf, iitem), super_gen + 1);
1136		return -EUCLEAN;
1137	}
1138
1139	/*
1140	 * For size and nbytes it's better not to be too strict, as for dir
1141	 * item its size/nbytes can easily get wrong, but doesn't affect
1142	 * anything in the fs. So here we skip the check.
1143	 */
1144	mode = btrfs_inode_mode(leaf, iitem);
1145	if (unlikely(mode & ~valid_mask)) {
1146		inode_item_err(leaf, slot,
1147			       "unknown mode bit detected: 0x%x",
1148			       mode & ~valid_mask);
1149		return -EUCLEAN;
1150	}
1151
1152	/*
1153	 * S_IFMT is not bit mapped so we can't completely rely on
1154	 * is_power_of_2/has_single_bit_set, but it can save us from checking
1155	 * FIFO/CHR/DIR/REG.  Only needs to check BLK, LNK and SOCKS
1156	 */
1157	if (!has_single_bit_set(mode & S_IFMT)) {
1158		if (unlikely(!S_ISLNK(mode) && !S_ISBLK(mode) && !S_ISSOCK(mode))) {
1159			inode_item_err(leaf, slot,
1160			"invalid mode: has 0%o expect valid S_IF* bit(s)",
1161				       mode & S_IFMT);
1162			return -EUCLEAN;
1163		}
1164	}
1165	if (unlikely(S_ISDIR(mode) && btrfs_inode_nlink(leaf, iitem) > 1)) {
1166		inode_item_err(leaf, slot,
1167		       "invalid nlink: has %u expect no more than 1 for dir",
1168			btrfs_inode_nlink(leaf, iitem));
1169		return -EUCLEAN;
1170	}
1171	btrfs_inode_split_flags(btrfs_inode_flags(leaf, iitem), &flags, &ro_flags);
1172	if (unlikely(flags & ~BTRFS_INODE_FLAG_MASK)) {
1173		inode_item_err(leaf, slot,
1174			       "unknown incompat flags detected: 0x%x", flags);
1175		return -EUCLEAN;
1176	}
1177	if (unlikely(!sb_rdonly(fs_info->sb) &&
1178		     (ro_flags & ~BTRFS_INODE_RO_FLAG_MASK))) {
1179		inode_item_err(leaf, slot,
1180			"unknown ro-compat flags detected on writeable mount: 0x%x",
1181			ro_flags);
1182		return -EUCLEAN;
1183	}
1184	return 0;
1185}
1186
1187static int check_root_item(struct extent_buffer *leaf, struct btrfs_key *key,
1188			   int slot)
1189{
1190	struct btrfs_fs_info *fs_info = leaf->fs_info;
1191	struct btrfs_root_item ri = { 0 };
1192	const u64 valid_root_flags = BTRFS_ROOT_SUBVOL_RDONLY |
1193				     BTRFS_ROOT_SUBVOL_DEAD;
1194	int ret;
1195
1196	ret = check_root_key(leaf, key, slot);
1197	if (unlikely(ret < 0))
1198		return ret;
1199
1200	if (unlikely(btrfs_item_size(leaf, slot) != sizeof(ri) &&
1201		     btrfs_item_size(leaf, slot) !=
1202		     btrfs_legacy_root_item_size())) {
1203		generic_err(leaf, slot,
1204			    "invalid root item size, have %u expect %zu or %u",
1205			    btrfs_item_size(leaf, slot), sizeof(ri),
1206			    btrfs_legacy_root_item_size());
1207		return -EUCLEAN;
1208	}
1209
1210	/*
1211	 * For legacy root item, the members starting at generation_v2 will be
1212	 * all filled with 0.
1213	 * And since we allow generation_v2 as 0, it will still pass the check.
1214	 */
1215	read_extent_buffer(leaf, &ri, btrfs_item_ptr_offset(leaf, slot),
1216			   btrfs_item_size(leaf, slot));
1217
1218	/* Generation related */
1219	if (unlikely(btrfs_root_generation(&ri) >
1220		     btrfs_super_generation(fs_info->super_copy) + 1)) {
1221		generic_err(leaf, slot,
1222			"invalid root generation, have %llu expect (0, %llu]",
1223			    btrfs_root_generation(&ri),
1224			    btrfs_super_generation(fs_info->super_copy) + 1);
1225		return -EUCLEAN;
1226	}
1227	if (unlikely(btrfs_root_generation_v2(&ri) >
1228		     btrfs_super_generation(fs_info->super_copy) + 1)) {
1229		generic_err(leaf, slot,
1230		"invalid root v2 generation, have %llu expect (0, %llu]",
1231			    btrfs_root_generation_v2(&ri),
1232			    btrfs_super_generation(fs_info->super_copy) + 1);
1233		return -EUCLEAN;
1234	}
1235	if (unlikely(btrfs_root_last_snapshot(&ri) >
1236		     btrfs_super_generation(fs_info->super_copy) + 1)) {
1237		generic_err(leaf, slot,
1238		"invalid root last_snapshot, have %llu expect (0, %llu]",
1239			    btrfs_root_last_snapshot(&ri),
1240			    btrfs_super_generation(fs_info->super_copy) + 1);
1241		return -EUCLEAN;
1242	}
1243
1244	/* Alignment and level check */
1245	if (unlikely(!IS_ALIGNED(btrfs_root_bytenr(&ri), fs_info->sectorsize))) {
1246		generic_err(leaf, slot,
1247		"invalid root bytenr, have %llu expect to be aligned to %u",
1248			    btrfs_root_bytenr(&ri), fs_info->sectorsize);
1249		return -EUCLEAN;
1250	}
1251	if (unlikely(btrfs_root_level(&ri) >= BTRFS_MAX_LEVEL)) {
1252		generic_err(leaf, slot,
1253			    "invalid root level, have %u expect [0, %u]",
1254			    btrfs_root_level(&ri), BTRFS_MAX_LEVEL - 1);
1255		return -EUCLEAN;
1256	}
1257	if (unlikely(btrfs_root_drop_level(&ri) >= BTRFS_MAX_LEVEL)) {
1258		generic_err(leaf, slot,
1259			    "invalid root drop_level, have %u expect [0, %u]",
1260			    btrfs_root_drop_level(&ri), BTRFS_MAX_LEVEL - 1);
1261		return -EUCLEAN;
1262	}
1263	/*
1264	 * If drop_progress.objectid is non-zero, a btrfs_drop_snapshot() was
1265	 * interrupted and the resume point was recorded in drop_progress and
1266	 * drop_level.  In that case drop_level must be >= 1: level 0 is the
1267	 * leaf level and drop_snapshot never saves a checkpoint there (it
1268	 * only records checkpoints at internal node levels in DROP_REFERENCE
1269	 * stage).  A zero drop_level combined with a non-zero drop_progress
1270	 * objectid indicates on-disk corruption and would cause a BUG_ON in
1271	 * merge_reloc_root() and btrfs_drop_snapshot() at mount time.
1272	 */
1273	if (unlikely(btrfs_disk_key_objectid(&ri.drop_progress) != 0 &&
1274		     btrfs_root_drop_level(&ri) == 0)) {
1275		generic_err(leaf, slot,
1276			    "invalid root drop_level 0 with non-zero drop_progress objectid %llu",
1277			    btrfs_disk_key_objectid(&ri.drop_progress));
1278		return -EUCLEAN;
1279	}
1280
1281	/* Flags check */
1282	if (unlikely(btrfs_root_flags(&ri) & ~valid_root_flags)) {
1283		generic_err(leaf, slot,
1284			    "invalid root flags, have 0x%llx expect mask 0x%llx",
1285			    btrfs_root_flags(&ri), valid_root_flags);
1286		return -EUCLEAN;
1287	}
1288	return 0;
1289}
1290
1291__printf(3,4)
1292__cold
1293static void extent_err(const struct extent_buffer *eb, int slot,
1294		       const char *fmt, ...)
1295{
1296	struct btrfs_key key;
1297	struct va_format vaf;
1298	va_list args;
1299	u64 bytenr;
1300	u64 len;
1301
1302	btrfs_item_key_to_cpu(eb, &key, slot);
1303	bytenr = key.objectid;
1304	if (key.type == BTRFS_METADATA_ITEM_KEY ||
1305	    key.type == BTRFS_TREE_BLOCK_REF_KEY ||
1306	    key.type == BTRFS_SHARED_BLOCK_REF_KEY)
1307		len = eb->fs_info->nodesize;
1308	else
1309		len = key.offset;
1310	va_start(args, fmt);
1311
1312	vaf.fmt = fmt;
1313	vaf.va = &args;
1314
1315	dump_page(folio_page(eb->folios[0], 0), "eb page dump");
1316	btrfs_crit(eb->fs_info,
1317	"corrupt %s: block=%llu slot=%d extent bytenr=%llu len=%llu %pV",
1318		btrfs_header_level(eb) == 0 ? "leaf" : "node",
1319		eb->start, slot, bytenr, len, &vaf);
1320	va_end(args);
1321}
1322
1323static bool is_valid_dref_root(u64 rootid)
1324{
1325	/*
1326	 * The following tree root objectids are allowed to have a data backref:
1327	 * - subvolume trees
1328	 * - data reloc tree
1329	 * - tree root
1330	 *   For v1 space cache
1331	 */
1332	return btrfs_is_fstree(rootid) || rootid == BTRFS_DATA_RELOC_TREE_OBJECTID ||
1333	       rootid == BTRFS_ROOT_TREE_OBJECTID;
1334}
1335
1336static int check_extent_item(struct extent_buffer *leaf,
1337			     struct btrfs_key *key, int slot,
1338			     struct btrfs_key *prev_key)
1339{
1340	struct btrfs_fs_info *fs_info = leaf->fs_info;
1341	struct btrfs_extent_item *ei;
1342	bool is_tree_block = false;
1343	unsigned long ptr;	/* Current pointer inside inline refs */
1344	unsigned long end;	/* Extent item end */
1345	const u32 item_size = btrfs_item_size(leaf, slot);
1346	u8 last_type = 0;
1347	u64 last_seq = U64_MAX;
1348	u64 flags;
1349	u64 generation;
1350	u64 total_refs;		/* Total refs in btrfs_extent_item */
1351	u64 inline_refs = 0;	/* found total inline refs */
1352
1353	if (unlikely(key->type == BTRFS_METADATA_ITEM_KEY &&
1354		     !btrfs_fs_incompat(fs_info, SKINNY_METADATA))) {
1355		generic_err(leaf, slot,
1356"invalid key type, METADATA_ITEM type invalid when SKINNY_METADATA feature disabled");
1357		return -EUCLEAN;
1358	}
1359	/* key->objectid is the bytenr for both key types */
1360	if (unlikely(!IS_ALIGNED(key->objectid, fs_info->sectorsize))) {
1361		generic_err(leaf, slot,
1362		"invalid key objectid, have %llu expect to be aligned to %u",
1363			   key->objectid, fs_info->sectorsize);
1364		return -EUCLEAN;
1365	}
1366
1367	/* key->offset is tree level for METADATA_ITEM_KEY */
1368	if (unlikely(key->type == BTRFS_METADATA_ITEM_KEY &&
1369		     key->offset >= BTRFS_MAX_LEVEL)) {
1370		extent_err(leaf, slot,
1371			   "invalid tree level, have %llu expect [0, %u]",
1372			   key->offset, BTRFS_MAX_LEVEL - 1);
1373		return -EUCLEAN;
1374	}
1375
1376	/*
1377	 * EXTENT/METADATA_ITEM consists of:
1378	 * 1) One btrfs_extent_item
1379	 *    Records the total refs, type and generation of the extent.
1380	 *
1381	 * 2) One btrfs_tree_block_info (for EXTENT_ITEM and tree backref only)
1382	 *    Records the first key and level of the tree block.
1383	 *
1384	 * 2) Zero or more btrfs_extent_inline_ref(s)
1385	 *    Each inline ref has one btrfs_extent_inline_ref shows:
1386	 *    2.1) The ref type, one of the 4
1387	 *         TREE_BLOCK_REF	Tree block only
1388	 *         SHARED_BLOCK_REF	Tree block only
1389	 *         EXTENT_DATA_REF	Data only
1390	 *         SHARED_DATA_REF	Data only
1391	 *    2.2) Ref type specific data
1392	 *         Either using btrfs_extent_inline_ref::offset, or specific
1393	 *         data structure.
1394	 *
1395	 *    All above inline items should follow the order:
1396	 *
1397	 *    - All btrfs_extent_inline_ref::type should be in an ascending
1398	 *      order
1399	 *
1400	 *    - Within the same type, the items should follow a descending
1401	 *      order by their sequence number. The sequence number is
1402	 *      determined by:
1403	 *      * btrfs_extent_inline_ref::offset for all types  other than
1404	 *        EXTENT_DATA_REF
1405	 *      * hash_extent_data_ref() for EXTENT_DATA_REF
1406	 */
1407	if (unlikely(item_size < sizeof(*ei))) {
1408		extent_err(leaf, slot,
1409			   "invalid item size, have %u expect [%zu, %u)",
1410			   item_size, sizeof(*ei),
1411			   BTRFS_LEAF_DATA_SIZE(fs_info));
1412		return -EUCLEAN;
1413	}
1414	end = item_size + btrfs_item_ptr_offset(leaf, slot);
1415
1416	/* Checks against extent_item */
1417	ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
1418	flags = btrfs_extent_flags(leaf, ei);
1419	total_refs = btrfs_extent_refs(leaf, ei);
1420	generation = btrfs_extent_generation(leaf, ei);
1421	if (unlikely(generation >
1422		     btrfs_super_generation(fs_info->super_copy) + 1)) {
1423		extent_err(leaf, slot,
1424			   "invalid generation, have %llu expect (0, %llu]",
1425			   generation,
1426			   btrfs_super_generation(fs_info->super_copy) + 1);
1427		return -EUCLEAN;
1428	}
1429	if (unlikely(!has_single_bit_set(flags & (BTRFS_EXTENT_FLAG_DATA |
1430						  BTRFS_EXTENT_FLAG_TREE_BLOCK)))) {
1431		extent_err(leaf, slot,
1432		"invalid extent flag, have 0x%llx expect 1 bit set in 0x%llx",
1433			flags, BTRFS_EXTENT_FLAG_DATA |
1434			BTRFS_EXTENT_FLAG_TREE_BLOCK);
1435		return -EUCLEAN;
1436	}
1437	is_tree_block = !!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK);
1438	if (is_tree_block) {
1439		if (unlikely(key->type == BTRFS_EXTENT_ITEM_KEY &&
1440			     key->offset != fs_info->nodesize)) {
1441			extent_err(leaf, slot,
1442				   "invalid extent length, have %llu expect %u",
1443				   key->offset, fs_info->nodesize);
1444			return -EUCLEAN;
1445		}
1446	} else {
1447		if (unlikely(key->type != BTRFS_EXTENT_ITEM_KEY)) {
1448			extent_err(leaf, slot,
1449			"invalid key type, have %u expect %u for data backref",
1450				   key->type, BTRFS_EXTENT_ITEM_KEY);
1451			return -EUCLEAN;
1452		}
1453		if (unlikely(!IS_ALIGNED(key->offset, fs_info->sectorsize))) {
1454			extent_err(leaf, slot,
1455			"invalid extent length, have %llu expect aligned to %u",
1456				   key->offset, fs_info->sectorsize);
1457			return -EUCLEAN;
1458		}
1459		if (unlikely(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)) {
1460			extent_err(leaf, slot,
1461			"invalid extent flag, data has full backref set");
1462			return -EUCLEAN;
1463		}
1464	}
1465	ptr = (unsigned long)(struct btrfs_extent_item *)(ei + 1);
1466
1467	/* Check the special case of btrfs_tree_block_info */
1468	if (is_tree_block && key->type != BTRFS_METADATA_ITEM_KEY) {
1469		struct btrfs_tree_block_info *info;
1470
1471		info = (struct btrfs_tree_block_info *)ptr;
1472		if (unlikely(btrfs_tree_block_level(leaf, info) >= BTRFS_MAX_LEVEL)) {
1473			extent_err(leaf, slot,
1474			"invalid tree block info level, have %u expect [0, %u]",
1475				   btrfs_tree_block_level(leaf, info),
1476				   BTRFS_MAX_LEVEL - 1);
1477			return -EUCLEAN;
1478		}
1479		ptr = (unsigned long)(struct btrfs_tree_block_info *)(info + 1);
1480	}
1481
1482	/* Check inline refs */
1483	while (ptr < end) {
1484		struct btrfs_extent_inline_ref *iref;
1485		struct btrfs_extent_data_ref *dref;
1486		struct btrfs_shared_data_ref *sref;
1487		u64 seq;
1488		u64 dref_root;
1489		u64 dref_objectid;
1490		u64 dref_offset;
1491		u64 inline_offset;
1492		u8 inline_type;
1493
1494		if (unlikely(ptr + sizeof(*iref) > end)) {
1495			extent_err(leaf, slot,
1496"inline ref item overflows extent item, ptr %lu iref size %zu end %lu",
1497				   ptr, sizeof(*iref), end);
1498			return -EUCLEAN;
1499		}
1500		iref = (struct btrfs_extent_inline_ref *)ptr;
1501		inline_type = btrfs_extent_inline_ref_type(leaf, iref);
1502		inline_offset = btrfs_extent_inline_ref_offset(leaf, iref);
1503		seq = inline_offset;
1504		if (unlikely(ptr + btrfs_extent_inline_ref_size(inline_type) > end)) {
1505			extent_err(leaf, slot,
1506"inline ref item overflows extent item, ptr %lu iref size %u end %lu",
1507				   ptr, btrfs_extent_inline_ref_size(inline_type), end);
1508			return -EUCLEAN;
1509		}
1510
1511		switch (inline_type) {
1512		/* inline_offset is subvolid of the owner, no need to check */
1513		case BTRFS_TREE_BLOCK_REF_KEY:
1514			inline_refs++;
1515			break;
1516		/* Contains parent bytenr */
1517		case BTRFS_SHARED_BLOCK_REF_KEY:
1518			if (unlikely(!IS_ALIGNED(inline_offset,
1519						 fs_info->sectorsize))) {
1520				extent_err(leaf, slot,
1521		"invalid tree parent bytenr, have %llu expect aligned to %u",
1522					   inline_offset, fs_info->sectorsize);
1523				return -EUCLEAN;
1524			}
1525			inline_refs++;
1526			break;
1527		/*
1528		 * Contains owner subvolid, owner key objectid, adjusted offset.
1529		 * The only obvious corruption can happen in that offset.
1530		 */
1531		case BTRFS_EXTENT_DATA_REF_KEY:
1532			dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1533			dref_root = btrfs_extent_data_ref_root(leaf, dref);
1534			dref_objectid = btrfs_extent_data_ref_objectid(leaf, dref);
1535			dref_offset = btrfs_extent_data_ref_offset(leaf, dref);
1536			seq = hash_extent_data_ref(
1537					btrfs_extent_data_ref_root(leaf, dref),
1538					btrfs_extent_data_ref_objectid(leaf, dref),
1539					btrfs_extent_data_ref_offset(leaf, dref));
1540			if (unlikely(!is_valid_dref_root(dref_root))) {
1541				extent_err(leaf, slot,
1542					   "invalid data ref root value %llu",
1543					   dref_root);
1544				return -EUCLEAN;
1545			}
1546			if (unlikely(dref_objectid < BTRFS_FIRST_FREE_OBJECTID ||
1547				     dref_objectid > BTRFS_LAST_FREE_OBJECTID)) {
1548				extent_err(leaf, slot,
1549					   "invalid data ref objectid value %llu",
1550					   dref_objectid);
1551				return -EUCLEAN;
1552			}
1553			if (unlikely(!IS_ALIGNED(dref_offset,
1554						 fs_info->sectorsize))) {
1555				extent_err(leaf, slot,
1556		"invalid data ref offset, have %llu expect aligned to %u",
1557					   dref_offset, fs_info->sectorsize);
1558				return -EUCLEAN;
1559			}
1560			if (unlikely(btrfs_extent_data_ref_count(leaf, dref) == 0)) {
1561				extent_err(leaf, slot,
1562			"invalid data ref count, should have non-zero value");
1563				return -EUCLEAN;
1564			}
1565			inline_refs += btrfs_extent_data_ref_count(leaf, dref);
1566			break;
1567		/* Contains parent bytenr and ref count */
1568		case BTRFS_SHARED_DATA_REF_KEY:
1569			sref = (struct btrfs_shared_data_ref *)(iref + 1);
1570			if (unlikely(!IS_ALIGNED(inline_offset,
1571						 fs_info->sectorsize))) {
1572				extent_err(leaf, slot,
1573		"invalid data parent bytenr, have %llu expect aligned to %u",
1574					   inline_offset, fs_info->sectorsize);
1575				return -EUCLEAN;
1576			}
1577			if (unlikely(btrfs_shared_data_ref_count(leaf, sref) == 0)) {
1578				extent_err(leaf, slot,
1579			"invalid shared data ref count, should have non-zero value");
1580				return -EUCLEAN;
1581			}
1582			inline_refs += btrfs_shared_data_ref_count(leaf, sref);
1583			break;
1584		case BTRFS_EXTENT_OWNER_REF_KEY:
1585			WARN_ON(!btrfs_fs_incompat(fs_info, SIMPLE_QUOTA));
1586			break;
1587		default:
1588			extent_err(leaf, slot, "unknown inline ref type: %u",
1589				   inline_type);
1590			return -EUCLEAN;
1591		}
1592		if (unlikely(inline_type < last_type)) {
1593			extent_err(leaf, slot,
1594				   "inline ref out-of-order: has type %u, prev type %u",
1595				   inline_type, last_type);
1596			return -EUCLEAN;
1597		}
1598		/* Type changed, allow the sequence starts from U64_MAX again. */
1599		if (inline_type > last_type)
1600			last_seq = U64_MAX;
1601		if (unlikely(seq > last_seq)) {
1602			extent_err(leaf, slot,
1603"inline ref out-of-order: has type %u offset %llu seq 0x%llx, prev type %u seq 0x%llx",
1604				   inline_type, inline_offset, seq,
1605				   last_type, last_seq);
1606			return -EUCLEAN;
1607		}
1608		last_type = inline_type;
1609		last_seq = seq;
1610		ptr += btrfs_extent_inline_ref_size(inline_type);
1611	}
1612	/* No padding is allowed */
1613	if (unlikely(ptr != end)) {
1614		extent_err(leaf, slot,
1615			   "invalid extent item size, padding bytes found");
1616		return -EUCLEAN;
1617	}
1618
1619	/* Finally, check the inline refs against total refs */
1620	if (unlikely(inline_refs > total_refs)) {
1621		extent_err(leaf, slot,
1622			"invalid extent refs, have %llu expect >= inline %llu",
1623			   total_refs, inline_refs);
1624		return -EUCLEAN;
1625	}
1626
1627	if ((prev_key->type == BTRFS_EXTENT_ITEM_KEY) ||
1628	    (prev_key->type == BTRFS_METADATA_ITEM_KEY)) {
1629		u64 prev_end = prev_key->objectid;
1630
1631		if (prev_key->type == BTRFS_METADATA_ITEM_KEY)
1632			prev_end += fs_info->nodesize;
1633		else
1634			prev_end += prev_key->offset;
1635
1636		if (unlikely(prev_end > key->objectid)) {
1637			extent_err(leaf, slot,
1638	"previous extent " BTRFS_KEY_FMT " overlaps current extent " BTRFS_KEY_FMT,
1639				   BTRFS_KEY_FMT_VALUE(prev_key),
1640				   BTRFS_KEY_FMT_VALUE(key));
1641			return -EUCLEAN;
1642		}
1643	}
1644
1645	return 0;
1646}
1647
1648static int check_simple_keyed_refs(struct extent_buffer *leaf,
1649				   struct btrfs_key *key, int slot)
1650{
1651	u32 expect_item_size = 0;
1652
1653	if (key->type == BTRFS_SHARED_DATA_REF_KEY) {
1654		struct btrfs_shared_data_ref *sref;
1655
1656		sref = btrfs_item_ptr(leaf, slot, struct btrfs_shared_data_ref);
1657		if (unlikely(btrfs_shared_data_ref_count(leaf, sref) == 0)) {
1658			extent_err(leaf, slot,
1659		"invalid shared data backref count, should have non-zero value");
1660			return -EUCLEAN;
1661		}
1662
1663		expect_item_size = sizeof(struct btrfs_shared_data_ref);
1664	}
1665
1666	if (unlikely(btrfs_item_size(leaf, slot) != expect_item_size)) {
1667		generic_err(leaf, slot,
1668		"invalid item size, have %u expect %u for key type %u",
1669			    btrfs_item_size(leaf, slot),
1670			    expect_item_size, key->type);
1671		return -EUCLEAN;
1672	}
1673	if (unlikely(!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize))) {
1674		generic_err(leaf, slot,
1675"invalid key objectid for shared block ref, have %llu expect aligned to %u",
1676			    key->objectid, leaf->fs_info->sectorsize);
1677		return -EUCLEAN;
1678	}
1679	if (unlikely(key->type != BTRFS_TREE_BLOCK_REF_KEY &&
1680		     !IS_ALIGNED(key->offset, leaf->fs_info->sectorsize))) {
1681		extent_err(leaf, slot,
1682		"invalid tree parent bytenr, have %llu expect aligned to %u",
1683			   key->offset, leaf->fs_info->sectorsize);
1684		return -EUCLEAN;
1685	}
1686	return 0;
1687}
1688
1689static int check_extent_data_ref(struct extent_buffer *leaf,
1690				 struct btrfs_key *key, int slot)
1691{
1692	struct btrfs_extent_data_ref *dref;
1693	unsigned long ptr = btrfs_item_ptr_offset(leaf, slot);
1694	const unsigned long end = ptr + btrfs_item_size(leaf, slot);
1695
1696	if (unlikely(btrfs_item_size(leaf, slot) % sizeof(*dref) != 0)) {
1697		generic_err(leaf, slot,
1698	"invalid item size, have %u expect aligned to %zu for key type %u",
1699			    btrfs_item_size(leaf, slot),
1700			    sizeof(*dref), key->type);
1701		return -EUCLEAN;
1702	}
1703	if (unlikely(!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize))) {
1704		generic_err(leaf, slot,
1705"invalid key objectid for shared block ref, have %llu expect aligned to %u",
1706			    key->objectid, leaf->fs_info->sectorsize);
1707		return -EUCLEAN;
1708	}
1709	for (; ptr < end; ptr += sizeof(*dref)) {
1710		u64 root;
1711		u64 objectid;
1712		u64 offset;
1713
1714		/*
1715		 * We cannot check the extent_data_ref hash due to possible
1716		 * overflow from the leaf due to hash collisions.
1717		 */
1718		dref = (struct btrfs_extent_data_ref *)ptr;
1719		root = btrfs_extent_data_ref_root(leaf, dref);
1720		objectid = btrfs_extent_data_ref_objectid(leaf, dref);
1721		offset = btrfs_extent_data_ref_offset(leaf, dref);
1722		if (unlikely(!is_valid_dref_root(root))) {
1723			extent_err(leaf, slot,
1724				   "invalid extent data backref root value %llu",
1725				   root);
1726			return -EUCLEAN;
1727		}
1728		if (unlikely(objectid < BTRFS_FIRST_FREE_OBJECTID ||
1729			     objectid > BTRFS_LAST_FREE_OBJECTID)) {
1730			extent_err(leaf, slot,
1731				   "invalid extent data backref objectid value %llu",
1732				   objectid);
1733			return -EUCLEAN;
1734		}
1735		if (unlikely(!IS_ALIGNED(offset, leaf->fs_info->sectorsize))) {
1736			extent_err(leaf, slot,
1737	"invalid extent data backref offset, have %llu expect aligned to %u",
1738				   offset, leaf->fs_info->sectorsize);
1739			return -EUCLEAN;
1740		}
1741		if (unlikely(btrfs_extent_data_ref_count(leaf, dref) == 0)) {
1742			extent_err(leaf, slot,
1743	"invalid extent data backref count, should have non-zero value");
1744			return -EUCLEAN;
1745		}
1746	}
1747	return 0;
1748}
1749
1750#define inode_ref_err(eb, slot, fmt, args...)			\
1751	inode_item_err(eb, slot, fmt, ##args)
1752static int check_inode_ref(struct extent_buffer *leaf,
1753			   struct btrfs_key *key, struct btrfs_key *prev_key,
1754			   int slot)
1755{
1756	struct btrfs_inode_ref *iref;
1757	unsigned long ptr;
1758	unsigned long end;
1759
1760	if (unlikely(!check_prev_ino(leaf, key, slot, prev_key)))
1761		return -EUCLEAN;
1762	/* namelen can't be 0, so item_size == sizeof() is also invalid */
1763	if (unlikely(btrfs_item_size(leaf, slot) <= sizeof(*iref))) {
1764		inode_ref_err(leaf, slot,
1765			"invalid item size, have %u expect (%zu, %u)",
1766			btrfs_item_size(leaf, slot),
1767			sizeof(*iref), BTRFS_LEAF_DATA_SIZE(leaf->fs_info));
1768		return -EUCLEAN;
1769	}
1770
1771	ptr = btrfs_item_ptr_offset(leaf, slot);
1772	end = ptr + btrfs_item_size(leaf, slot);
1773	while (ptr < end) {
1774		u16 namelen;
1775
1776		if (unlikely(ptr + sizeof(*iref) > end)) {
1777			inode_ref_err(leaf, slot,
1778			"inode ref overflow, ptr %lu end %lu inode_ref_size %zu",
1779				ptr, end, sizeof(*iref));
1780			return -EUCLEAN;
1781		}
1782
1783		iref = (struct btrfs_inode_ref *)ptr;
1784		namelen = btrfs_inode_ref_name_len(leaf, iref);
1785		if (unlikely(ptr + sizeof(*iref) + namelen > end)) {
1786			inode_ref_err(leaf, slot,
1787				"inode ref overflow, ptr %lu end %lu namelen %u",
1788				ptr, end, namelen);
1789			return -EUCLEAN;
1790		}
1791
1792		/*
1793		 * NOTE: In theory we should record all found index numbers
1794		 * to find any duplicated indexes, but that will be too time
1795		 * consuming for inodes with too many hard links.
1796		 */
1797		ptr += sizeof(*iref) + namelen;
1798	}
1799	return 0;
1800}
1801
1802static int check_inode_extref(struct extent_buffer *leaf,
1803			      struct btrfs_key *key, struct btrfs_key *prev_key,
1804			      int slot)
1805{
1806	unsigned long ptr = btrfs_item_ptr_offset(leaf, slot);
1807	unsigned long end = ptr + btrfs_item_size(leaf, slot);
1808
1809	if (unlikely(!check_prev_ino(leaf, key, slot, prev_key)))
1810		return -EUCLEAN;
1811
1812	while (ptr < end) {
1813		struct btrfs_inode_extref *extref = (struct btrfs_inode_extref *)ptr;
1814		u16 namelen;
1815
1816		if (unlikely(ptr + sizeof(*extref) > end)) {
1817			inode_ref_err(leaf, slot,
1818			"inode extref overflow, ptr %lu end %lu inode_extref size %zu",
1819				      ptr, end, sizeof(*extref));
1820			return -EUCLEAN;
1821		}
1822
1823		namelen = btrfs_inode_extref_name_len(leaf, extref);
1824		if (unlikely(ptr + sizeof(*extref) + namelen > end)) {
1825			inode_ref_err(leaf, slot,
1826				"inode extref overflow, ptr %lu end %lu namelen %u",
1827				ptr, end, namelen);
1828			return -EUCLEAN;
1829		}
1830		ptr += sizeof(*extref) + namelen;
1831	}
1832	return 0;
1833}
1834
1835static int check_raid_stripe_extent(const struct extent_buffer *leaf,
1836				    const struct btrfs_key *key, int slot)
1837{
1838	if (unlikely(!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize))) {
1839		generic_err(leaf, slot,
1840"invalid key objectid for raid stripe extent, have %llu expect aligned to %u",
1841			    key->objectid, leaf->fs_info->sectorsize);
1842		return -EUCLEAN;
1843	}
1844
1845	if (unlikely(!btrfs_fs_incompat(leaf->fs_info, RAID_STRIPE_TREE))) {
1846		generic_err(leaf, slot,
1847	"RAID_STRIPE_EXTENT present but RAID_STRIPE_TREE incompat bit unset");
1848		return -EUCLEAN;
1849	}
1850
1851	return 0;
1852}
1853
1854static int check_dev_extent_item(const struct extent_buffer *leaf,
1855				 const struct btrfs_key *key,
1856				 int slot,
1857				 struct btrfs_key *prev_key)
1858{
1859	struct btrfs_dev_extent *de;
1860	const u32 sectorsize = leaf->fs_info->sectorsize;
1861
1862	de = btrfs_item_ptr(leaf, slot, struct btrfs_dev_extent);
1863	/* Basic fixed member checks. */
1864	if (unlikely(btrfs_dev_extent_chunk_tree(leaf, de) !=
1865		     BTRFS_CHUNK_TREE_OBJECTID)) {
1866		generic_err(leaf, slot,
1867			    "invalid dev extent chunk tree id, has %llu expect %llu",
1868			    btrfs_dev_extent_chunk_tree(leaf, de),
1869			    BTRFS_CHUNK_TREE_OBJECTID);
1870		return -EUCLEAN;
1871	}
1872	if (unlikely(btrfs_dev_extent_chunk_objectid(leaf, de) !=
1873		     BTRFS_FIRST_CHUNK_TREE_OBJECTID)) {
1874		generic_err(leaf, slot,
1875			    "invalid dev extent chunk objectid, has %llu expect %llu",
1876			    btrfs_dev_extent_chunk_objectid(leaf, de),
1877			    BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1878		return -EUCLEAN;
1879	}
1880	/* Alignment check. */
1881	if (unlikely(!IS_ALIGNED(key->offset, sectorsize))) {
1882		generic_err(leaf, slot,
1883			    "invalid dev extent key.offset, has %llu not aligned to %u",
1884			    key->offset, sectorsize);
1885		return -EUCLEAN;
1886	}
1887	if (unlikely(!IS_ALIGNED(btrfs_dev_extent_chunk_offset(leaf, de),
1888				 sectorsize))) {
1889		generic_err(leaf, slot,
1890			    "invalid dev extent chunk offset, has %llu not aligned to %u",
1891			    btrfs_dev_extent_chunk_objectid(leaf, de),
1892			    sectorsize);
1893		return -EUCLEAN;
1894	}
1895	if (unlikely(!IS_ALIGNED(btrfs_dev_extent_length(leaf, de),
1896				 sectorsize))) {
1897		generic_err(leaf, slot,
1898			    "invalid dev extent length, has %llu not aligned to %u",
1899			    btrfs_dev_extent_length(leaf, de), sectorsize);
1900		return -EUCLEAN;
1901	}
1902	/* Overlap check with previous dev extent. */
1903	if (slot && prev_key->objectid == key->objectid &&
1904	    prev_key->type == key->type) {
1905		struct btrfs_dev_extent *prev_de;
1906		u64 prev_len;
1907
1908		prev_de = btrfs_item_ptr(leaf, slot - 1, struct btrfs_dev_extent);
1909		prev_len = btrfs_dev_extent_length(leaf, prev_de);
1910		if (unlikely(prev_key->offset + prev_len > key->offset)) {
1911			generic_err(leaf, slot,
1912		"dev extent overlap, prev offset %llu len %llu current offset %llu",
1913				    prev_key->offset, prev_len, key->offset);
1914			return -EUCLEAN;
1915		}
1916	}
1917	return 0;
1918}
1919
1920/*
1921 * Common point to switch the item-specific validation.
1922 */
1923static enum btrfs_tree_block_status check_leaf_item(struct extent_buffer *leaf,
1924						    struct btrfs_key *key,
1925						    int slot,
1926						    struct btrfs_key *prev_key)
1927{
1928	int ret = 0;
1929	struct btrfs_chunk *chunk;
1930
1931	switch (key->type) {
1932	case BTRFS_EXTENT_DATA_KEY:
1933		ret = check_extent_data_item(leaf, key, slot, prev_key);
1934		break;
1935	case BTRFS_EXTENT_CSUM_KEY:
1936		ret = check_csum_item(leaf, key, slot, prev_key);
1937		break;
1938	case BTRFS_DIR_ITEM_KEY:
1939	case BTRFS_DIR_INDEX_KEY:
1940	case BTRFS_XATTR_ITEM_KEY:
1941		ret = check_dir_item(leaf, key, prev_key, slot);
1942		break;
1943	case BTRFS_INODE_REF_KEY:
1944		ret = check_inode_ref(leaf, key, prev_key, slot);
1945		break;
1946	case BTRFS_INODE_EXTREF_KEY:
1947		ret = check_inode_extref(leaf, key, prev_key, slot);
1948		break;
1949	case BTRFS_BLOCK_GROUP_ITEM_KEY:
1950		ret = check_block_group_item(leaf, key, slot);
1951		break;
1952	case BTRFS_CHUNK_ITEM_KEY:
1953		chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
1954		ret = check_leaf_chunk_item(leaf, chunk, key, slot);
1955		break;
1956	case BTRFS_DEV_ITEM_KEY:
1957		ret = check_dev_item(leaf, key, slot);
1958		break;
1959	case BTRFS_DEV_EXTENT_KEY:
1960		ret = check_dev_extent_item(leaf, key, slot, prev_key);
1961		break;
1962	case BTRFS_INODE_ITEM_KEY:
1963		ret = check_inode_item(leaf, key, slot);
1964		break;
1965	case BTRFS_ROOT_ITEM_KEY:
1966		ret = check_root_item(leaf, key, slot);
1967		break;
1968	case BTRFS_EXTENT_ITEM_KEY:
1969	case BTRFS_METADATA_ITEM_KEY:
1970		ret = check_extent_item(leaf, key, slot, prev_key);
1971		break;
1972	case BTRFS_TREE_BLOCK_REF_KEY:
1973	case BTRFS_SHARED_DATA_REF_KEY:
1974	case BTRFS_SHARED_BLOCK_REF_KEY:
1975		ret = check_simple_keyed_refs(leaf, key, slot);
1976		break;
1977	case BTRFS_EXTENT_DATA_REF_KEY:
1978		ret = check_extent_data_ref(leaf, key, slot);
1979		break;
1980	case BTRFS_RAID_STRIPE_KEY:
1981		ret = check_raid_stripe_extent(leaf, key, slot);
1982		break;
1983	}
1984
1985	if (unlikely(ret))
1986		return BTRFS_TREE_BLOCK_INVALID_ITEM;
1987	return BTRFS_TREE_BLOCK_CLEAN;
1988}
1989
1990enum btrfs_tree_block_status __btrfs_check_leaf(struct extent_buffer *leaf)
1991{
1992	struct btrfs_fs_info *fs_info = leaf->fs_info;
1993	/* No valid key type is 0, so all key should be larger than this key */
1994	struct btrfs_key prev_key = {0, 0, 0};
1995	struct btrfs_key key;
1996	u32 nritems = btrfs_header_nritems(leaf);
1997	int slot;
1998
1999	if (unlikely(btrfs_header_level(leaf) != 0)) {
2000		generic_err(leaf, 0,
2001			"invalid level for leaf, have %d expect 0",
2002			btrfs_header_level(leaf));
2003		return BTRFS_TREE_BLOCK_INVALID_LEVEL;
2004	}
2005
2006	if (unlikely(!btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_WRITTEN))) {
2007		generic_err(leaf, 0, "invalid flag for leaf, WRITTEN not set");
2008		return BTRFS_TREE_BLOCK_WRITTEN_NOT_SET;
2009	}
2010
2011	/*
2012	 * Extent buffers from a relocation tree have a owner field that
2013	 * corresponds to the subvolume tree they are based on. So just from an
2014	 * extent buffer alone we can not find out what is the id of the
2015	 * corresponding subvolume tree, so we can not figure out if the extent
2016	 * buffer corresponds to the root of the relocation tree or not. So
2017	 * skip this check for relocation trees.
2018	 */
2019	if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) {
2020		u64 owner = btrfs_header_owner(leaf);
2021
2022		/* These trees must never be empty */
2023		if (unlikely(owner == BTRFS_ROOT_TREE_OBJECTID ||
2024			     owner == BTRFS_CHUNK_TREE_OBJECTID ||
2025			     owner == BTRFS_DEV_TREE_OBJECTID ||
2026			     owner == BTRFS_FS_TREE_OBJECTID ||
2027			     owner == BTRFS_DATA_RELOC_TREE_OBJECTID)) {
2028			generic_err(leaf, 0,
2029			"invalid root, root %llu must never be empty",
2030				    owner);
2031			return BTRFS_TREE_BLOCK_INVALID_NRITEMS;
2032		}
2033
2034		/* Unknown tree */
2035		if (unlikely(owner == 0)) {
2036			generic_err(leaf, 0,
2037				"invalid owner, root 0 is not defined");
2038			return BTRFS_TREE_BLOCK_INVALID_OWNER;
2039		}
2040
2041		/* EXTENT_TREE_V2 can have empty extent trees. */
2042		if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2))
2043			return BTRFS_TREE_BLOCK_CLEAN;
2044
2045		if (unlikely(owner == BTRFS_EXTENT_TREE_OBJECTID)) {
2046			generic_err(leaf, 0,
2047			"invalid root, root %llu must never be empty",
2048				    owner);
2049			return BTRFS_TREE_BLOCK_INVALID_NRITEMS;
2050		}
2051
2052		return BTRFS_TREE_BLOCK_CLEAN;
2053	}
2054
2055	if (unlikely(nritems == 0))
2056		return BTRFS_TREE_BLOCK_CLEAN;
2057
2058	/*
2059	 * Check the following things to make sure this is a good leaf, and
2060	 * leaf users won't need to bother with similar sanity checks:
2061	 *
2062	 * 1) key ordering
2063	 * 2) item offset and size
2064	 *    No overlap, no hole, all inside the leaf.
2065	 * 3) item content
2066	 *    If possible, do comprehensive sanity check.
2067	 *    NOTE: All checks must only rely on the item data itself.
2068	 */
2069	for (slot = 0; slot < nritems; slot++) {
2070		u32 item_end_expected;
2071		u64 item_data_end;
2072		enum btrfs_tree_block_status ret;
2073
2074		btrfs_item_key_to_cpu(leaf, &key, slot);
2075
2076		/* Make sure the keys are in the right order */
2077		if (unlikely(btrfs_comp_cpu_keys(&prev_key, &key) >= 0)) {
2078			generic_err(leaf, slot,
2079	"bad key order, prev " BTRFS_KEY_FMT " current " BTRFS_KEY_FMT,
2080				    BTRFS_KEY_FMT_VALUE(&prev_key),
2081				    BTRFS_KEY_FMT_VALUE(&key));
2082			return BTRFS_TREE_BLOCK_BAD_KEY_ORDER;
2083		}
2084
2085		item_data_end = (u64)btrfs_item_offset(leaf, slot) +
2086				btrfs_item_size(leaf, slot);
2087		/*
2088		 * Make sure the offset and ends are right, remember that the
2089		 * item data starts at the end of the leaf and grows towards the
2090		 * front.
2091		 */
2092		if (slot == 0)
2093			item_end_expected = BTRFS_LEAF_DATA_SIZE(fs_info);
2094		else
2095			item_end_expected = btrfs_item_offset(leaf,
2096								 slot - 1);
2097		if (unlikely(item_data_end != item_end_expected)) {
2098			generic_err(leaf, slot,
2099				"unexpected item end, have %llu expect %u",
2100				item_data_end, item_end_expected);
2101			return BTRFS_TREE_BLOCK_INVALID_OFFSETS;
2102		}
2103
2104		/*
2105		 * Check to make sure that we don't point outside of the leaf,
2106		 * just in case all the items are consistent to each other, but
2107		 * all point outside of the leaf.
2108		 */
2109		if (unlikely(item_data_end > BTRFS_LEAF_DATA_SIZE(fs_info))) {
2110			generic_err(leaf, slot,
2111			"slot end outside of leaf, have %llu expect range [0, %u]",
2112				item_data_end, BTRFS_LEAF_DATA_SIZE(fs_info));
2113			return BTRFS_TREE_BLOCK_INVALID_OFFSETS;
2114		}
2115
2116		/* Also check if the item pointer overlaps with btrfs item. */
2117		if (unlikely(btrfs_item_ptr_offset(leaf, slot) <
2118			     btrfs_item_nr_offset(leaf, slot) + sizeof(struct btrfs_item))) {
2119			generic_err(leaf, slot,
2120		"slot overlaps with its data, item end %lu data start %lu",
2121				btrfs_item_nr_offset(leaf, slot) +
2122				sizeof(struct btrfs_item),
2123				btrfs_item_ptr_offset(leaf, slot));
2124			return BTRFS_TREE_BLOCK_INVALID_OFFSETS;
2125		}
2126
2127		/* Check if the item size and content meet other criteria. */
2128		ret = check_leaf_item(leaf, &key, slot, &prev_key);
2129		if (unlikely(ret != BTRFS_TREE_BLOCK_CLEAN))
2130			return ret;
2131
2132		prev_key.objectid = key.objectid;
2133		prev_key.type = key.type;
2134		prev_key.offset = key.offset;
2135	}
2136
2137	return BTRFS_TREE_BLOCK_CLEAN;
2138}
2139
2140int btrfs_check_leaf(struct extent_buffer *leaf)
2141{
2142	enum btrfs_tree_block_status ret;
2143
2144	ret = __btrfs_check_leaf(leaf);
2145	if (unlikely(ret != BTRFS_TREE_BLOCK_CLEAN))
2146		return -EUCLEAN;
2147	return 0;
2148}
2149ALLOW_ERROR_INJECTION(btrfs_check_leaf, ERRNO);
2150
2151enum btrfs_tree_block_status __btrfs_check_node(struct extent_buffer *node)
2152{
2153	struct btrfs_fs_info *fs_info = node->fs_info;
2154	unsigned long nr = btrfs_header_nritems(node);
2155	struct btrfs_key key, next_key;
2156	int slot;
2157	int level = btrfs_header_level(node);
2158	u64 bytenr;
2159
2160	if (unlikely(!btrfs_header_flag(node, BTRFS_HEADER_FLAG_WRITTEN))) {
2161		generic_err(node, 0, "invalid flag for node, WRITTEN not set");
2162		return BTRFS_TREE_BLOCK_WRITTEN_NOT_SET;
2163	}
2164
2165	if (unlikely(level <= 0 || level >= BTRFS_MAX_LEVEL)) {
2166		generic_err(node, 0,
2167			"invalid level for node, have %d expect [1, %d]",
2168			level, BTRFS_MAX_LEVEL - 1);
2169		return BTRFS_TREE_BLOCK_INVALID_LEVEL;
2170	}
2171	if (unlikely(nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(fs_info))) {
2172		btrfs_crit(fs_info,
2173"corrupt node: root=%llu block=%llu, nritems too %s, have %lu expect range [1,%u]",
2174			   btrfs_header_owner(node), node->start,
2175			   nr == 0 ? "small" : "large", nr,
2176			   BTRFS_NODEPTRS_PER_BLOCK(fs_info));
2177		return BTRFS_TREE_BLOCK_INVALID_NRITEMS;
2178	}
2179
2180	for (slot = 0; slot < nr - 1; slot++) {
2181		bytenr = btrfs_node_blockptr(node, slot);
2182		btrfs_node_key_to_cpu(node, &key, slot);
2183		btrfs_node_key_to_cpu(node, &next_key, slot + 1);
2184
2185		if (unlikely(!bytenr)) {
2186			generic_err(node, slot,
2187				"invalid NULL node pointer");
2188			return BTRFS_TREE_BLOCK_INVALID_BLOCKPTR;
2189		}
2190		if (unlikely(!IS_ALIGNED(bytenr, fs_info->sectorsize))) {
2191			generic_err(node, slot,
2192			"unaligned pointer, have %llu should be aligned to %u",
2193				bytenr, fs_info->sectorsize);
2194			return BTRFS_TREE_BLOCK_INVALID_BLOCKPTR;
2195		}
2196
2197		if (unlikely(btrfs_comp_cpu_keys(&key, &next_key) >= 0)) {
2198			generic_err(node, slot,
2199	"bad key order, current " BTRFS_KEY_FMT " next " BTRFS_KEY_FMT,
2200				    BTRFS_KEY_FMT_VALUE(&key),
2201				    BTRFS_KEY_FMT_VALUE(&next_key));
2202			return BTRFS_TREE_BLOCK_BAD_KEY_ORDER;
2203		}
2204	}
2205	return BTRFS_TREE_BLOCK_CLEAN;
2206}
2207
2208int btrfs_check_node(struct extent_buffer *node)
2209{
2210	enum btrfs_tree_block_status ret;
2211
2212	ret = __btrfs_check_node(node);
2213	if (unlikely(ret != BTRFS_TREE_BLOCK_CLEAN))
2214		return -EUCLEAN;
2215	return 0;
2216}
2217ALLOW_ERROR_INJECTION(btrfs_check_node, ERRNO);
2218
2219int btrfs_check_eb_owner(const struct extent_buffer *eb, u64 root_owner)
2220{
2221	const bool is_subvol = btrfs_is_fstree(root_owner);
2222	const u64 eb_owner = btrfs_header_owner(eb);
2223
2224	/*
2225	 * Skip dummy fs, as selftests don't create unique ebs for each dummy
2226	 * root.
2227	 */
2228	if (btrfs_is_testing(eb->fs_info))
2229		return 0;
2230	/*
2231	 * There are several call sites (backref walking, qgroup, and data
2232	 * reloc) passing 0 as @root_owner, as they are not holding the
2233	 * tree root.  In that case, we can not do a reliable ownership check,
2234	 * so just exit.
2235	 */
2236	if (root_owner == 0)
2237		return 0;
2238	/*
2239	 * These trees use key.offset as their owner, our callers don't have
2240	 * the extra capacity to pass key.offset here.  So we just skip them.
2241	 */
2242	if (root_owner == BTRFS_TREE_LOG_OBJECTID ||
2243	    root_owner == BTRFS_TREE_RELOC_OBJECTID)
2244		return 0;
2245
2246	if (!is_subvol) {
2247		/* For non-subvolume trees, the eb owner should match root owner */
2248		if (unlikely(root_owner != eb_owner)) {
2249			btrfs_crit(eb->fs_info,
2250"corrupted %s, root=%llu block=%llu owner mismatch, have %llu expect %llu",
2251				btrfs_header_level(eb) == 0 ? "leaf" : "node",
2252				root_owner, btrfs_header_bytenr(eb), eb_owner,
2253				root_owner);
2254			return -EUCLEAN;
2255		}
2256		return 0;
2257	}
2258
2259	/*
2260	 * For subvolume trees, owners can mismatch, but they should all belong
2261	 * to subvolume trees.
2262	 */
2263	if (unlikely(is_subvol != btrfs_is_fstree(eb_owner))) {
2264		btrfs_crit(eb->fs_info,
2265"corrupted %s, root=%llu block=%llu owner mismatch, have %llu expect [%llu, %llu]",
2266			btrfs_header_level(eb) == 0 ? "leaf" : "node",
2267			root_owner, btrfs_header_bytenr(eb), eb_owner,
2268			BTRFS_FIRST_FREE_OBJECTID, BTRFS_LAST_FREE_OBJECTID);
2269		return -EUCLEAN;
2270	}
2271	return 0;
2272}
2273
2274int btrfs_verify_level_key(struct extent_buffer *eb,
2275			   const struct btrfs_tree_parent_check *check)
2276{
2277	struct btrfs_fs_info *fs_info = eb->fs_info;
2278	int found_level;
2279	struct btrfs_key found_key;
2280	int ret;
2281
2282	found_level = btrfs_header_level(eb);
2283	if (unlikely(found_level != check->level)) {
2284		DEBUG_WARN();
2285		btrfs_err(fs_info,
2286"tree level mismatch detected, bytenr=%llu level expected=%u has=%u",
2287			  eb->start, check->level, found_level);
2288		return -EUCLEAN;
2289	}
2290
2291	if (!check->has_first_key)
2292		return 0;
2293
2294	/*
2295	 * For live tree block (new tree blocks in current transaction),
2296	 * we need proper lock context to avoid race, which is impossible here.
2297	 * So we only checks tree blocks which is read from disk, whose
2298	 * generation <= fs_info->last_trans_committed.
2299	 */
2300	if (btrfs_header_generation(eb) > btrfs_get_last_trans_committed(fs_info))
2301		return 0;
2302
2303	/* We have @first_key, so this @eb must have at least one item */
2304	if (unlikely(btrfs_header_nritems(eb) == 0)) {
2305		btrfs_err(fs_info,
2306		"invalid tree nritems, bytenr=%llu nritems=0 expect >0",
2307			  eb->start);
2308		DEBUG_WARN();
2309		return -EUCLEAN;
2310	}
2311
2312	if (found_level)
2313		btrfs_node_key_to_cpu(eb, &found_key, 0);
2314	else
2315		btrfs_item_key_to_cpu(eb, &found_key, 0);
2316
2317	ret = btrfs_comp_cpu_keys(&check->first_key, &found_key);
2318	if (unlikely(ret)) {
2319		DEBUG_WARN();
2320		btrfs_err(fs_info,
2321"tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)",
2322			  eb->start, check->transid, check->first_key.objectid,
2323			  check->first_key.type, check->first_key.offset,
2324			  found_key.objectid, found_key.type,
2325			  found_key.offset);
2326	}
2327	return ret;
2328}