2 * pass1.c -- pass #1 of e2fsck: sequential scan of the inode table
4 * Copyright (C) 1993, 1994, 1995, 1996, 1997 Theodore Ts'o.
7 * This file may be redistributed under the terms of the GNU Public
11 * Pass 1 of e2fsck iterates over all the inodes in the filesystems,
12 * and applies the following tests to each inode:
14 * - The mode field of the inode must be legal.
15 * - The size and block count fields of the inode are correct.
16 * - A data block must not be used by another inode
18 * Pass 1 also gathers the collects the following information:
20 * - A bitmap of which inodes are in use. (inode_used_map)
21 * - A bitmap of which inodes are directories. (inode_dir_map)
22 * - A bitmap of which inodes are regular files. (inode_reg_map)
23 * - A bitmap of which inodes have bad fields. (inode_bad_map)
24 * - A bitmap of which inodes are in bad blocks. (inode_bb_map)
25 * - A bitmap of which inodes are imagic inodes. (inode_imagic_map)
26 * - A bitmap of which inodes are casefolded. (inode_casefold_map)
27 * - A bitmap of which blocks are in use. (block_found_map)
28 * - A bitmap of which blocks are in use by two inodes (block_dup_map)
29 * - The data blocks of the directory inodes. (dir_map)
30 * - Ref counts for ea_inodes. (ea_inode_refs)
31 * - The encryption policy ID of each encrypted inode. (encrypted_files)
33 * Pass 1 is designed to stash away enough information so that the
34 * other passes should not need to read in the inode information
35 * during the normal course of a filesystem check. (Although if an
36 * inconsistency is detected, other passes may need to read in an
39 * Note that pass 1B will be invoked if there are any duplicate blocks
43 #define _GNU_SOURCE 1 /* get strnlen() */
56 #include <ext2fs/ext2_ext_attr.h>
57 /* todo remove this finally */
58 #include <ext2fs/ext2fsP.h>
63 #ifdef NO_INLINE_FUNCS
66 #define _INLINE_ inline
76 static int process_block(ext2_filsys fs, blk64_t *blocknr,
77 e2_blkcnt_t blockcnt, blk64_t ref_blk,
78 int ref_offset, void *priv_data);
79 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
80 e2_blkcnt_t blockcnt, blk64_t ref_blk,
81 int ref_offset, void *priv_data);
82 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
84 const struct ea_quota *ea_ibody_quota);
85 static void mark_table_blocks(e2fsck_t ctx);
86 static void alloc_bb_map(e2fsck_t ctx);
87 static void alloc_imagic_map(e2fsck_t ctx);
88 static void mark_inode_bad(e2fsck_t ctx, ino_t ino);
89 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino);
90 static void handle_fs_bad_blocks(e2fsck_t ctx);
91 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
92 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
93 dgrp_t group, void * priv_data);
94 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
95 char *block_buf, int adjust_sign);
96 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
98 struct process_block_struct {
100 unsigned is_dir:1, is_reg:1, clear:1, suppress:1,
101 fragmented:1, compressed:1, bbcheck:1,
106 e2_blkcnt_t last_init_lblock;
107 e2_blkcnt_t last_db_block;
108 int num_illegal_blocks;
109 blk64_t previous_block;
110 struct ext2_inode *inode;
111 struct problem_context *pctx;
112 ext2fs_block_bitmap fs_meta_blocks;
115 struct extent_tree_info eti;
118 struct process_inode_block {
120 struct ea_quota ea_ibody_quota;
121 struct ext2_inode_large inode;
124 struct scan_callback_struct {
127 struct process_inode_block *inodes_to_process;
128 int *process_inode_count;
131 static void process_inodes(e2fsck_t ctx, char *block_buf,
132 struct process_inode_block *inodes_to_process,
133 int *process_inode_count);
135 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
136 EXT2_MIN_BLOCK_LOG_SIZE + 1];
139 * Check to make sure a device inode is real. Returns 1 if the device
140 * checks out, 0 if not.
142 * Note: this routine is now also used to check FIFO's and Sockets,
143 * since they have the same requirement; the i_block fields should be
146 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
147 struct ext2_inode *inode)
152 * If the index or extents flag is set, then this is a bogus
155 if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
159 * We should be able to do the test below all the time, but
160 * because the kernel doesn't forcibly clear the device
161 * inode's additional i_block fields, there are some rare
162 * occasions when a legitimate device inode will have non-zero
163 * additional i_block fields. So for now, we only complain
164 * when the immutable flag is set, which should never happen
165 * for devices. (And that's when the problem is caused, since
166 * you can't set or clear immutable flags for devices.) Once
167 * the kernel has been fixed we can change this...
169 if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
170 for (i=4; i < EXT2_N_BLOCKS; i++)
171 if (inode->i_block[i])
178 * Check to make sure a symlink inode is real. Returns 1 if the symlink
179 * checks out, 0 if not.
181 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
182 struct ext2_inode *inode, char *buf)
187 if ((inode->i_size_high || inode->i_size == 0) ||
188 (inode->i_flags & EXT2_INDEX_FL))
191 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
194 if (inode->i_flags & EXT4_EXTENTS_FL)
196 if (ext2fs_inline_data_size(fs, ino, &inline_size))
198 if (inode->i_size != inline_size)
204 if (ext2fs_is_fast_symlink(inode)) {
205 if (inode->i_flags & EXT4_EXTENTS_FL)
207 buf = (char *)inode->i_block;
208 buflen = sizeof(inode->i_block);
210 ext2_extent_handle_t handle;
211 struct ext2_extent_info info;
212 struct ext2fs_extent extent;
216 if (inode->i_flags & EXT4_EXTENTS_FL) {
217 if (ext2fs_extent_open2(fs, ino, inode, &handle))
219 if (ext2fs_extent_get_info(handle, &info) ||
220 (info.num_entries != 1) ||
221 (info.max_depth != 0)) {
222 ext2fs_extent_free(handle);
225 if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
227 (extent.e_lblk != 0) ||
228 (extent.e_len != 1)) {
229 ext2fs_extent_free(handle);
233 ext2fs_extent_free(handle);
235 blk = inode->i_block[0];
237 for (i = 1; i < EXT2_N_BLOCKS; i++)
238 if (inode->i_block[i])
242 if (blk < fs->super->s_first_data_block ||
243 blk >= ext2fs_blocks_count(fs->super))
246 if (io_channel_read_blk64(fs->io, blk, 1, buf))
249 buflen = fs->blocksize;
252 if (inode->i_flags & EXT4_ENCRYPT_FL)
253 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
255 len = strnlen(buf, buflen);
260 if (len != inode->i_size)
266 * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
268 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
269 static void check_extents_inlinedata(e2fsck_t ctx,
270 struct problem_context *pctx)
272 if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
275 if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
278 pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
279 e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
281 #undef BAD_SPECIAL_FLAGS
284 * If the immutable (or append-only) flag is set on the inode, offer
287 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
288 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
290 if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
293 if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
296 pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
297 e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
301 * If device, fifo or socket, check size is zero -- if not offer to
304 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
306 struct ext2_inode *inode = pctx->inode;
308 if (EXT2_I_SIZE(inode) == 0)
311 if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
314 ext2fs_inode_size_set(ctx->fs, inode, 0);
315 e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
319 * For a given size, calculate how many blocks would be charged towards quota.
321 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
325 clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
326 return EXT2FS_C2B(fs, clusters);
330 * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
333 static problem_t check_large_ea_inode(e2fsck_t ctx,
334 struct ext2_ext_attr_entry *entry,
335 struct problem_context *pctx,
336 blk64_t *quota_blocks)
338 struct ext2_inode inode;
342 /* Check if inode is within valid range */
343 if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
344 (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
345 pctx->num = entry->e_value_inum;
346 return PR_1_ATTR_VALUE_EA_INODE;
349 e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
351 retval = ext2fs_ext_attr_hash_entry2(ctx->fs, entry, NULL, &hash);
353 com_err("check_large_ea_inode", retval,
354 _("while hashing entry with e_value_inum = %u"),
355 entry->e_value_inum);
359 if (hash == entry->e_hash) {
360 *quota_blocks = size_to_quota_blocks(ctx->fs,
361 entry->e_value_size);
363 /* This might be an old Lustre-style ea_inode reference. */
364 if (inode.i_mtime == pctx->ino &&
365 inode.i_generation == pctx->inode->i_generation) {
368 /* If target inode is also missing EA_INODE flag,
369 * this is likely to be a bad reference.
371 if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
372 pctx->num = entry->e_value_inum;
373 return PR_1_ATTR_VALUE_EA_INODE;
375 pctx->num = entry->e_hash;
376 return PR_1_ATTR_HASH;
381 if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
382 pctx->num = entry->e_value_inum;
383 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
384 inode.i_flags |= EXT4_EA_INODE_FL;
385 e2fsck_pass1_fix_lock(ctx);
386 ext2fs_write_inode(ctx->fs, entry->e_value_inum,
388 e2fsck_pass1_fix_unlock(ctx);
390 return PR_1_ATTR_NO_EA_INODE_FL;
396 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
397 struct ext2_ext_attr_entry *first, void *end)
399 struct ext2_ext_attr_entry *entry = first;
400 struct ext2_ext_attr_entry *np = EXT2_EXT_ATTR_NEXT(entry);
402 while ((void *) entry < end && (void *) np < end &&
403 !EXT2_EXT_IS_LAST_ENTRY(entry)) {
404 if (!entry->e_value_inum)
406 if (!ctx->ea_inode_refs) {
407 pctx->errcode = ea_refcount_create(0,
408 &ctx->ea_inode_refs);
411 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
412 ctx->flags |= E2F_FLAG_ABORT;
416 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
420 np = EXT2_EXT_ATTR_NEXT(entry);
424 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
425 struct ea_quota *ea_ibody_quota)
427 struct ext2_super_block *sb = ctx->fs->super;
428 struct ext2_inode_large *inode;
429 struct ext2_ext_attr_entry *entry;
430 char *start, *header, *end;
431 unsigned int storage_size, remain;
432 problem_t problem = 0;
435 ea_ibody_quota->blocks = 0;
436 ea_ibody_quota->inodes = 0;
438 inode = (struct ext2_inode_large *) pctx->inode;
439 storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
440 inode->i_extra_isize;
441 header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
442 inode->i_extra_isize;
443 end = header + storage_size;
444 start = header + sizeof(__u32);
445 entry = (struct ext2_ext_attr_entry *) start;
447 /* scan all entry's headers first */
449 /* take finish entry 0UL into account */
450 remain = storage_size - sizeof(__u32);
452 region = region_create(0, storage_size);
454 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
456 ctx->flags |= E2F_FLAG_ABORT;
459 if (region_allocate(region, 0, sizeof(__u32))) {
460 problem = PR_1_INODE_EA_ALLOC_COLLISION;
464 while (remain >= sizeof(struct ext2_ext_attr_entry) &&
465 !EXT2_EXT_IS_LAST_ENTRY(entry)) {
468 if (region_allocate(region, (char *)entry - (char *)header,
469 EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
470 problem = PR_1_INODE_EA_ALLOC_COLLISION;
474 /* header eats this space */
475 remain -= sizeof(struct ext2_ext_attr_entry);
477 /* is attribute name valid? */
478 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
479 pctx->num = entry->e_name_len;
480 problem = PR_1_ATTR_NAME_LEN;
484 /* attribute len eats this space */
485 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
487 if (entry->e_value_inum == 0) {
488 /* check value size */
489 if (entry->e_value_size > remain) {
490 pctx->num = entry->e_value_size;
491 problem = PR_1_ATTR_VALUE_SIZE;
495 if (entry->e_value_size &&
496 region_allocate(region,
497 sizeof(__u32) + entry->e_value_offs,
499 entry->e_value_size))) {
500 problem = PR_1_INODE_EA_ALLOC_COLLISION;
504 hash = ext2fs_ext_attr_hash_entry(entry,
505 start + entry->e_value_offs);
507 /* e_hash may be 0 in older inode's ea */
508 if (entry->e_hash != 0 && entry->e_hash != hash) {
509 pctx->num = entry->e_hash;
510 problem = PR_1_ATTR_HASH;
514 blk64_t quota_blocks;
516 problem = check_large_ea_inode(ctx, entry, pctx,
521 ea_ibody_quota->blocks += quota_blocks;
522 ea_ibody_quota->inodes++;
525 /* If EA value is stored in external inode then it does not
526 * consume space here */
527 if (entry->e_value_inum == 0)
528 remain -= entry->e_value_size;
530 entry = EXT2_EXT_ATTR_NEXT(entry);
533 if (region_allocate(region, (char *)entry - (char *)header,
535 problem = PR_1_INODE_EA_ALLOC_COLLISION;
542 * it seems like a corruption. it's very unlikely we could repair
543 * EA(s) in automatic fashion -bzzz
545 if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
546 inc_ea_inode_refs(ctx, pctx,
547 (struct ext2_ext_attr_entry *)start, end);
551 /* simply remove all possible EA(s) */
552 *((__u32 *)header) = 0UL;
553 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
554 EXT2_INODE_SIZE(sb), "pass1");
555 ea_ibody_quota->blocks = 0;
556 ea_ibody_quota->inodes = 0;
559 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
560 return (xtime & (1U << 31)) != 0 &&
561 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
564 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
565 check_inode_extra_negative_epoch(inode->i_##xtime, \
566 inode->i_##xtime##_extra)
568 /* When today's date is earlier than 2242, we assume that atimes,
569 * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
570 * actually pre-1970 dates mis-encoded.
572 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
574 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
575 struct ea_quota *ea_ibody_quota)
577 struct ext2_super_block *sb = ctx->fs->super;
578 struct ext2_inode_large *inode;
582 ea_ibody_quota->blocks = 0;
583 ea_ibody_quota->inodes = 0;
585 inode = (struct ext2_inode_large *) pctx->inode;
586 if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
587 /* this isn't large inode. so, nothing to check */
592 printf("inode #%u, i_extra_size %d\n", pctx->ino,
593 inode->i_extra_isize);
595 /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
596 min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
597 max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
599 * For now we will allow i_extra_isize to be 0, but really
600 * implementations should never allow i_extra_isize to be 0
602 if (inode->i_extra_isize &&
603 (inode->i_extra_isize < min || inode->i_extra_isize > max ||
604 inode->i_extra_isize & 3)) {
605 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
607 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
608 inode->i_extra_isize = sb->s_want_extra_isize;
610 inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
611 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
612 EXT2_INODE_SIZE(sb), "pass1");
615 /* check if there is no place for an EA header */
616 if (inode->i_extra_isize >= max - sizeof(__u32))
619 eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
620 inode->i_extra_isize);
621 if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
622 /* it seems inode has an extended attribute(s) in body */
623 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
627 * If the inode's extended atime (ctime, crtime, mtime) is stored in
628 * the old, invalid format, repair it.
630 if (((sizeof(time_t) <= 4) ||
631 (((sizeof(time_t) > 4) &&
632 ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
633 (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
634 CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
635 CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
636 CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
638 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
641 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
642 inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
643 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
644 inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
645 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
646 inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
647 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
648 inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
649 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
650 EXT2_INODE_SIZE(sb), "pass1");
655 static _INLINE_ int is_blocks_used(e2fsck_t ctx, blk64_t block,
660 /* used to avoid duplicate output from below */
661 retval = ext2fs_test_block_bitmap_range2_valid(ctx->block_found_map,
666 retval = ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num);
668 e2fsck_pass1_block_map_r_lock(ctx);
670 retval = ext2fs_test_block_bitmap_range2(
671 ctx->global_ctx->block_found_map, block, num);
672 e2fsck_pass1_block_map_r_unlock(ctx);
681 * Check to see if the inode might really be a directory, despite i_mode
683 * This is a lot of complexity for something for which I'm not really
684 * convinced happens frequently in the wild. If for any reason this
685 * causes any problems, take this code out.
686 * [tytso:20070331.0827EDT]
688 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
691 struct ext2_inode *inode = pctx->inode;
692 struct ext2_dir_entry *dirent;
695 unsigned int i, rec_len, not_device = 0;
700 * If the mode looks OK, we believe it. If the first block in
701 * the i_block array is 0, this cannot be a directory. If the
702 * inode is extent-mapped, it is still the case that the latter
703 * cannot be 0 - the magic number in the extent header would make
706 if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
707 LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
711 * Check the block numbers in the i_block array for validity:
712 * zero blocks are skipped (but the first one cannot be zero -
713 * see above), other blocks are checked against the first and
714 * max data blocks (from the the superblock) and against the
715 * block bitmap. Any invalid block found means this cannot be
718 * If there are non-zero blocks past the fourth entry, then
719 * this cannot be a device file: we remember that for the next
722 * For extent mapped files, we don't do any sanity checking:
723 * just try to get the phys block of logical block 0 and run
726 * For inline data files, we just try to get the size of inline
727 * data. If it's true, we will treat it as a directory.
730 extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
731 inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
732 if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
735 unsigned int rec_len2;
736 struct ext2_dir_entry de;
738 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
741 * If the size isn't a multiple of 4, it's probably not a
747 * If the first 10 bytes don't look like a directory entry,
748 * it's probably not a directory.
750 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
751 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
752 EXT2_DIR_REC_LEN(0));
753 dotdot = ext2fs_le32_to_cpu(dotdot);
754 de.inode = ext2fs_le32_to_cpu(de.inode);
755 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
756 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
757 if (dotdot >= ctx->fs->super->s_inodes_count ||
758 (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
759 dotdot != EXT2_ROOT_INO) ||
760 de.inode >= ctx->fs->super->s_inodes_count ||
761 (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
763 rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
764 EXT4_INLINE_DATA_DOTDOT_SIZE)
766 /* device files never have a "system.data" entry */
768 } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
770 if (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
773 /* device files are never extent mapped */
776 for (i=0; i < EXT2_N_BLOCKS; i++) {
777 blk = inode->i_block[i];
783 if (blk < ctx->fs->super->s_first_data_block ||
784 blk >= ext2fs_blocks_count(ctx->fs->super) ||
785 is_blocks_used(ctx, blk, 1))
786 return; /* Invalid block, can't be dir */
788 blk = inode->i_block[0];
792 * If the mode says this is a device file and the i_links_count field
793 * is sane and we have not ruled it out as a device file previously,
794 * we declare it a device file, not a directory.
796 if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
797 (inode->i_links_count == 1) && !not_device)
800 /* read the first block */
801 ehandler_operation(_("reading directory block"));
802 retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
803 ehandler_operation(0);
807 dirent = (struct ext2_dir_entry *) buf;
808 retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
811 if ((ext2fs_dirent_name_len(dirent) != 1) ||
812 (dirent->name[0] != '.') ||
813 (dirent->inode != pctx->ino) ||
816 (rec_len >= ctx->fs->blocksize - 12))
819 dirent = (struct ext2_dir_entry *) (buf + rec_len);
820 retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
823 if ((ext2fs_dirent_name_len(dirent) != 2) ||
824 (dirent->name[0] != '.') ||
825 (dirent->name[1] != '.') ||
831 if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
832 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
833 e2fsck_write_inode_full(ctx, pctx->ino, inode,
834 EXT2_INODE_SIZE(ctx->fs->super),
835 "check_is_really_dir");
839 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
840 int flags, ext2_icount_t hint,
843 unsigned int threshold;
844 unsigned int save_type;
852 profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
854 profile_get_uint(ctx->profile, "scratch_files",
855 "numdirs_threshold", 0, 0, &threshold);
856 profile_get_boolean(ctx->profile, "scratch_files",
857 "icount", 0, 1, &enable);
859 retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
861 num_dirs = 1024; /* Guess */
863 if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
864 (!threshold || num_dirs > threshold)) {
865 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
870 e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
872 if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
873 flags |= EXT2_ICOUNT_OPT_FULLMAP;
874 retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
875 ctx->fs->default_bitmap_type = save_type;
879 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
881 struct problem_context *pctx)
884 struct ext2_inode_large inode;
887 * Reread inode. If we don't see checksum error, then this inode
888 * has been fixed elsewhere.
890 ctx->stashed_ino = 0;
891 retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
893 if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
899 * Checksum still doesn't match. That implies that the inode passes
900 * all the sanity checks, so maybe the checksum is simply corrupt.
901 * See if the user will go for fixing that.
903 if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
907 e2fsck_pass1_fix_lock(ctx);
908 retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
910 e2fsck_pass1_fix_unlock(ctx);
914 static void reserve_block_for_root_repair(e2fsck_t ctx)
918 ext2_filsys fs = ctx->fs;
920 ctx->root_repair_block = 0;
921 if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
924 err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
927 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
928 ctx->root_repair_block = blk;
931 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
935 ext2_filsys fs = ctx->fs;
936 static const char name[] = "lost+found";
939 ctx->lnf_repair_block = 0;
940 if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
943 err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
946 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
947 ctx->lnf_repair_block = blk;
951 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
955 struct ext2_xattr_handle *handle;
958 retval = ext2fs_xattrs_open(fs, ino, &handle);
962 retval = ext2fs_xattrs_read(handle);
966 retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
971 (void) ext2fs_xattrs_close(&handle);
975 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
976 struct problem_context *pctx,
983 * If the inode failed the checksum and the user didn't
984 * clear the inode, test the checksum again -- if it still
985 * fails, ask the user if the checksum should be corrected.
987 pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
989 ctx->flags |= E2F_FLAG_ABORT;
991 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
993 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
994 if ((ctx)->flags & E2F_FLAG_ABORT) \
998 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
1003 for (i = 0; i < EXT2_N_BLOCKS; i++) {
1004 x = inode->i_block[i];
1005 #ifdef WORDS_BIGENDIAN
1006 x = ext2fs_swab32(x);
1008 if (x >= ext2fs_blocks_count(fs->super))
1016 * Figure out what to do with an inode that has both extents and inline data
1017 * inode flags set. Returns -1 if we decide to erase the inode, 0 otherwise.
1019 static int fix_inline_data_extents_file(e2fsck_t ctx,
1021 struct ext2_inode *inode,
1023 struct problem_context *pctx)
1025 size_t max_inline_ea_size;
1026 ext2_filsys fs = ctx->fs;
1029 /* Both feature flags not set? Just run the regular checks */
1030 if (!ext2fs_has_feature_extents(fs->super) &&
1031 !ext2fs_has_feature_inline_data(fs->super))
1034 /* Clear both flags if it's a special file */
1035 if (LINUX_S_ISCHR(inode->i_mode) ||
1036 LINUX_S_ISBLK(inode->i_mode) ||
1037 LINUX_S_ISFIFO(inode->i_mode) ||
1038 LINUX_S_ISSOCK(inode->i_mode)) {
1039 check_extents_inlinedata(ctx, pctx);
1043 /* If it looks like an extent tree, try to clear inlinedata */
1044 if (ext2fs_extent_header_verify(inode->i_block,
1045 sizeof(inode->i_block)) == 0 &&
1046 fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1047 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1052 /* If it looks short enough to be inline data, try to clear extents */
1053 if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1054 max_inline_ea_size = inode_size -
1055 (EXT2_GOOD_OLD_INODE_SIZE +
1056 ((struct ext2_inode_large *)inode)->i_extra_isize);
1058 max_inline_ea_size = 0;
1059 if (EXT2_I_SIZE(inode) <
1060 EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1061 fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1062 inode->i_flags &= ~EXT4_EXTENTS_FL;
1068 * Too big for inline data, but no evidence of extent tree -
1069 * maybe it's a block map file? If the mappings all look valid?
1071 if (could_be_block_map(fs, inode) &&
1072 fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1073 #ifdef WORDS_BIGENDIAN
1076 for (i = 0; i < EXT2_N_BLOCKS; i++)
1077 inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1080 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1085 /* Oh well, just clear the busted inode. */
1086 if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1087 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1093 e2fsck_write_inode(ctx, ino, inode, "pass1");
1098 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1100 ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1101 dgrp_t start = *group, grp, grp_end = ctx->fs->group_desc_count;
1102 blk64_t blocks_to_read = 0;
1103 errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1106 if (ctx->fs->fs_num_threads > 1)
1107 grp_end = ctx->thread_info.et_group_end;
1109 if (ctx->readahead_kb == 0)
1112 /* Keep iterating groups until we have enough to readahead */
1113 inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1114 for (grp = start; grp < grp_end; grp++) {
1115 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1117 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1118 ext2fs_bg_itable_unused(ctx->fs, grp);
1119 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1121 if (blocks_to_read * ctx->fs->blocksize >
1122 ctx->readahead_kb * 1024)
1126 err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1128 if (err == EAGAIN) {
1129 ctx->readahead_kb /= 2;
1135 /* Error; disable itable readahead */
1136 *group = ctx->fs->group_desc_count;
1137 *next_ino = ctx->fs->super->s_inodes_count;
1140 * Don't do more readahead until we've reached the first inode
1141 * of the last inode scan buffer block for the last group.
1144 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1145 ctx->inode_buffer_blocks :
1146 EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1147 ctx->fs->blocksize /
1148 EXT2_INODE_SIZE(ctx->fs->super);
1150 *next_ino = inodes_in_group -
1151 (inodes_in_group % inodes_per_buffer) + 1 +
1152 (grp * ctx->fs->super->s_inodes_per_group);
1157 * Check if the passed ino is one of the used superblock quota inodes.
1159 * Before the quota inodes were journaled, older superblock quota inodes
1160 * were just regular files in the filesystem and not reserved inodes. This
1161 * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1162 * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1164 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1166 enum quota_type qtype;
1168 for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1169 if (*quota_sb_inump(sb, qtype) == ino)
1176 * Check if the passed ino is one of the reserved quota inodes.
1177 * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1178 * inodes. These inodes may or may not be in use by the quota feature.
1180 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1182 enum quota_type qtype;
1184 for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1185 if (quota_type2inum(qtype, fs->super) == ino)
1191 static int e2fsck_should_abort(e2fsck_t ctx)
1193 e2fsck_t global_ctx;
1195 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1198 if (ctx->global_ctx) {
1199 global_ctx = ctx->global_ctx;
1200 if (global_ctx->flags & E2F_FLAG_SIGNAL_MASK)
1206 static void init_ext2_max_sizes()
1212 * Init ext2_max_sizes which will be immutable and shared between
1215 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1217 for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1218 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1219 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1220 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1221 max_sizes = (max_sizes * (1UL << i));
1222 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1228 /* TODO: tdb needs to be handled properly for multiple threads*/
1229 static int multiple_threads_supported(e2fsck_t ctx)
1232 unsigned int threshold;
1233 ext2_ino_t num_dirs;
1238 profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
1240 profile_get_uint(ctx->profile, "scratch_files",
1241 "numdirs_threshold", 0, 0, &threshold);
1242 profile_get_boolean(ctx->profile, "scratch_files",
1243 "icount", 0, 1, &enable);
1245 retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
1247 num_dirs = 1024; /* Guess */
1249 /* tdb is unsupported now */
1250 if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
1251 (!threshold || num_dirs > threshold))
1258 * Even though we could specify number of threads,
1259 * but it might be more than the whole filesystem
1260 * block groups, correct it here.
1262 static void e2fsck_pass1_set_thread_num(e2fsck_t ctx)
1264 unsigned flexbg_size = 1;
1265 ext2_filsys fs = ctx->fs;
1266 int num_threads = ctx->fs_num_threads;
1269 if (num_threads < 1) {
1274 if (!multiple_threads_supported(ctx)) {
1276 fprintf(stderr, "Fall through single thread for pass1 "
1277 "because tdb could not handle properly\n");
1281 if (ext2fs_has_feature_flex_bg(fs->super))
1282 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
1283 max_threads = fs->group_desc_count / flexbg_size;
1284 if (max_threads == 0)
1287 if (num_threads > max_threads) {
1288 fprintf(stderr, "Use max possible thread num: %d instead\n",
1290 num_threads = max_threads;
1293 ctx->fs_num_threads = num_threads;
1294 ctx->fs->fs_num_threads = num_threads;
1299 * We need call mark_table_blocks() before multiple
1300 * thread start, since all known system blocks should be
1301 * marked and checked later.
1303 static errcode_t e2fsck_pass1_prepare(e2fsck_t ctx)
1305 struct problem_context pctx;
1306 ext2_filsys fs = ctx->fs;
1307 unsigned long long readahead_kb;
1309 init_ext2_max_sizes();
1311 e2fsck_pass1_set_thread_num(ctx);
1313 /* If we can do readahead, figure out how many groups to pull in. */
1314 if (!e2fsck_can_readahead(ctx->fs))
1315 ctx->readahead_kb = 0;
1316 else if (ctx->readahead_kb == ~0ULL)
1317 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1320 /* don't use more than 1/10 of memory for threads checking */
1321 readahead_kb = get_memory_size() / (10 * ctx->fs_num_threads);
1322 /* maybe better disable RA if this is too small? */
1323 if (ctx->readahead_kb > readahead_kb)
1324 ctx->readahead_kb = readahead_kb;
1326 clear_problem_context(&pctx);
1327 if (!(ctx->options & E2F_OPT_PREEN))
1328 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1330 pctx.errcode = e2fsck_allocate_subcluster_bitmap(ctx->fs,
1331 _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1332 "block_found_map", &ctx->block_found_map);
1335 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1336 ctx->flags |= E2F_FLAG_ABORT;
1337 return pctx.errcode;
1339 pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
1340 _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1341 "block_metadata_map", &ctx->block_metadata_map);
1344 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1345 ctx->flags |= E2F_FLAG_ABORT;
1346 return pctx.errcode;
1349 mark_table_blocks(ctx);
1350 pctx.errcode = ext2fs_convert_subcluster_bitmap(ctx->fs,
1351 &ctx->block_found_map);
1353 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1354 ctx->flags |= E2F_FLAG_ABORT;
1355 return pctx.errcode;
1358 pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
1359 _("multiply claimed block map"),
1360 EXT2FS_BMAP64_RBTREE, "block_dup_map",
1361 &ctx->block_dup_map);
1364 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
1366 /* Should never get here */
1367 ctx->flags |= E2F_FLAG_ABORT;
1368 return pctx.errcode;
1371 if (ext2fs_has_feature_mmp(fs->super) &&
1372 fs->super->s_mmp_block > fs->super->s_first_data_block &&
1373 fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1374 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1375 fs->super->s_mmp_block);
1377 pthread_mutex_init(&ctx->fs_fix_mutex, NULL);
1378 pthread_rwlock_init(&ctx->fs_block_map_rwlock, NULL);
1384 static void e2fsck_pass1_post(e2fsck_t ctx)
1386 struct problem_context pctx;
1387 ext2_filsys fs = ctx->fs;
1390 (char *)e2fsck_allocate_memory(ctx, ctx->fs->blocksize * 3,
1391 "block interate buffer");
1392 reserve_block_for_root_repair(ctx);
1393 reserve_block_for_lnf_repair(ctx);
1396 * If any extended attribute blocks' reference counts need to
1397 * be adjusted, either up (ctx->refcount_extra), or down
1398 * (ctx->refcount), then fix them.
1400 if (ctx->refcount) {
1401 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
1402 ea_refcount_free(ctx->refcount);
1405 if (ctx->refcount_extra) {
1406 adjust_extattr_refcount(ctx, ctx->refcount_extra,
1408 ea_refcount_free(ctx->refcount_extra);
1409 ctx->refcount_extra = 0;
1412 if (ctx->invalid_bitmaps)
1413 handle_fs_bad_blocks(ctx);
1415 /* We don't need the block_ea_map any more */
1416 if (ctx->block_ea_map) {
1417 ext2fs_free_block_bitmap(ctx->block_ea_map);
1418 ctx->block_ea_map = 0;
1421 if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
1422 struct ext2_inode *inode;
1423 int inode_size = EXT2_INODE_SIZE(fs->super);
1424 inode = e2fsck_allocate_memory(ctx, inode_size,
1427 clear_problem_context(&pctx);
1428 pctx.errcode = ext2fs_create_resize_inode(fs);
1430 if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
1432 ctx->flags |= E2F_FLAG_ABORT;
1433 ext2fs_free_mem(&inode);
1434 ext2fs_free_mem(&block_buf);
1439 if (!pctx.errcode) {
1440 e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
1442 inode->i_mtime = ctx->now;
1443 e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
1446 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
1447 ext2fs_free_mem(&inode);
1450 if (ctx->flags & E2F_FLAG_RESTART) {
1451 ext2fs_free_mem(&block_buf);
1455 if (ctx->block_dup_map) {
1456 if (!(ctx->flags & E2F_FLAG_DUP_BLOCK)) {
1457 ext2fs_free_mem(&block_buf);
1460 if (ctx->options & E2F_OPT_PREEN) {
1461 clear_problem_context(&pctx);
1462 fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
1464 e2fsck_pass1_dupblocks(ctx, block_buf);
1465 ext2fs_free_mem(&block_buf);
1466 ctx->flags &= ~E2F_FLAG_DUP_BLOCK;
1471 void e2fsck_pass1_run(e2fsck_t ctx)
1474 ext2_filsys fs = ctx->fs;
1476 struct ext2_inode *inode = NULL;
1477 ext2_inode_scan scan = NULL;
1478 char *block_buf = NULL;
1479 #ifdef RESOURCE_TRACK
1480 struct resource_track rtrack;
1482 unsigned char frag, fsize;
1483 struct problem_context pctx;
1484 struct scan_callback_struct scan_struct;
1485 struct ext2_super_block *sb = ctx->fs->super;
1487 const char *eop_next_inode = _("getting next inode from scan");
1488 int imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1489 int low_dtime_check = 1;
1490 unsigned int inode_size = EXT2_INODE_SIZE(fs->super);
1491 unsigned int bufsize;
1492 int failed_csum = 0;
1493 ext2_ino_t ino_threshold = 0;
1494 dgrp_t ra_group = 0;
1495 struct ea_quota ea_ibody_quota;
1496 struct process_inode_block *inodes_to_process;
1497 int process_inode_count;
1499 init_resource_track(&rtrack, ctx->fs->io);
1500 clear_problem_context(&pctx);
1502 pass1_readahead(ctx, &ra_group, &ino_threshold);
1503 if (ext2fs_has_feature_dir_index(fs->super) &&
1504 !(ctx->options & E2F_OPT_NO)) {
1505 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1506 ctx->dirs_to_hash = 0;
1510 mtrace_print("Pass 1");
1513 imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1514 extent_fs = ext2fs_has_feature_extents(sb);
1515 inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1516 casefold_fs = ext2fs_has_feature_casefold(sb);
1519 * Allocate bitmaps structures
1521 pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1522 EXT2FS_BMAP64_RBTREE,
1524 &ctx->inode_used_map);
1527 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1528 ctx->flags |= E2F_FLAG_ABORT;
1531 pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1532 _("directory inode map"),
1533 ctx->global_ctx ? EXT2FS_BMAP64_RBTREE :
1534 EXT2FS_BMAP64_AUTODIR,
1535 "inode_dir_map", &ctx->inode_dir_map);
1538 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1539 ctx->flags |= E2F_FLAG_ABORT;
1542 pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1543 _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1544 "inode_reg_map", &ctx->inode_reg_map);
1547 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1548 ctx->flags |= E2F_FLAG_ABORT;
1553 e2fsck_allocate_inode_bitmap(fs,
1554 _("inode casefold map"),
1555 EXT2FS_BMAP64_RBTREE,
1556 "inode_casefold_map",
1557 &ctx->inode_casefold_map);
1560 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1561 ctx->flags |= E2F_FLAG_ABORT;
1565 pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1566 &ctx->inode_link_info);
1568 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1569 ctx->flags |= E2F_FLAG_ABORT;
1572 bufsize = inode_size;
1573 if (bufsize < sizeof(struct ext2_inode_large))
1574 bufsize = sizeof(struct ext2_inode_large);
1575 inode = (struct ext2_inode *)
1576 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1578 inodes_to_process = (struct process_inode_block *)
1579 e2fsck_allocate_memory(ctx,
1580 (ctx->process_inode_size *
1581 sizeof(struct process_inode_block)),
1582 "array of inodes to process");
1583 process_inode_count = 0;
1585 pctx.errcode = ext2fs_init_dblist(fs, 0);
1587 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1588 ctx->flags |= E2F_FLAG_ABORT;
1593 * If the last orphan field is set, clear it, since the pass1
1594 * processing will automatically find and clear the orphans.
1595 * In the future, we may want to try using the last_orphan
1596 * linked list ourselves, but for now, we clear it so that the
1597 * ext3 mount code won't get confused.
1599 if (!(ctx->options & E2F_OPT_READONLY)) {
1600 if (fs->super->s_last_orphan) {
1601 fs->super->s_last_orphan = 0;
1602 ext2fs_mark_super_dirty(fs);
1606 block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1607 "block interate buffer");
1608 if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1609 e2fsck_use_inode_shortcuts(ctx, 1);
1610 e2fsck_intercept_block_allocations(ctx);
1611 old_op = ehandler_operation(_("opening inode scan"));
1612 pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1614 ehandler_operation(old_op);
1616 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1617 ctx->flags |= E2F_FLAG_ABORT;
1620 ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1621 EXT2_SF_WARN_GARBAGE_INODES, 0);
1622 ctx->stashed_inode = inode;
1623 scan_struct.ctx = ctx;
1624 scan_struct.block_buf = block_buf;
1625 scan_struct.inodes_to_process = inodes_to_process;
1626 scan_struct.process_inode_count = &process_inode_count;
1627 ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1628 if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1629 ctx->fs->group_desc_count)))
1631 if ((fs->super->s_wtime &&
1632 fs->super->s_wtime < fs->super->s_inodes_count) ||
1633 (fs->super->s_mtime &&
1634 fs->super->s_mtime < fs->super->s_inodes_count) ||
1635 (fs->super->s_mkfs_time &&
1636 fs->super->s_mkfs_time < fs->super->s_inodes_count))
1637 low_dtime_check = 0;
1639 /* Set up ctx->lost_and_found if possible */
1640 (void) e2fsck_get_lost_and_found(ctx, 0);
1643 if (ctx->global_ctx) {
1644 if (ctx->options & E2F_OPT_DEBUG &&
1645 ctx->options & E2F_OPT_MULTITHREAD)
1646 fprintf(stderr, "thread %d jumping to group %d\n",
1647 ctx->thread_info.et_thread_index,
1648 ctx->thread_info.et_group_start);
1649 pctx.errcode = ext2fs_inode_scan_goto_blockgroup(scan,
1650 ctx->thread_info.et_group_start);
1652 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1653 ctx->flags |= E2F_FLAG_ABORT;
1660 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1661 if (e2fsck_mmp_update(fs))
1662 fatal_error(ctx, 0);
1664 old_op = ehandler_operation(eop_next_inode);
1665 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1667 if (ino > ino_threshold)
1668 pass1_readahead(ctx, &ra_group, &ino_threshold);
1669 ehandler_operation(old_op);
1670 if (e2fsck_should_abort(ctx))
1672 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1674 * If badblocks says badblocks is bad, offer to clear
1675 * the list, update the in-core bb list, and restart
1678 if (ino == EXT2_BAD_INO &&
1679 fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1683 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1684 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1685 ctx->fs->badblocks = NULL;
1686 err = ext2fs_read_bb_inode(ctx->fs,
1687 &ctx->fs->badblocks);
1689 fix_problem(ctx, PR_1_ISCAN_ERROR,
1691 ctx->flags |= E2F_FLAG_ABORT;
1693 ctx->flags |= E2F_FLAG_RESTART;
1696 if (!ctx->inode_bb_map)
1698 ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1699 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1702 if (pctx.errcode == EXT2_ET_SCAN_FINISHED)
1705 pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1706 pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1707 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1708 ctx->flags |= E2F_FLAG_ABORT;
1714 if (ctx->global_ctx)
1715 ctx->thread_info.et_inode_number++;
1719 ctx->stashed_ino = ino;
1721 /* Clear trashed inode? */
1722 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1723 inode->i_links_count > 0 &&
1724 fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1726 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1728 failed_csum = pctx.errcode != 0;
1731 * Check for inodes who might have been part of the
1732 * orphaned list linked list. They should have gotten
1733 * dealt with by now, unless the list had somehow been
1736 * FIXME: In the future, inodes which are still in use
1737 * (and which are therefore) pending truncation should
1738 * be handled specially. Right now we just clear the
1739 * dtime field, and the normal e2fsck handling of
1740 * inodes where i_size and the inode blocks are
1741 * inconsistent is to fix i_size, instead of releasing
1742 * the extra blocks. This won't catch the inodes that
1743 * was at the end of the orphan list, but it's better
1744 * than nothing. The right answer is that there
1745 * shouldn't be any bugs in the orphan list handling. :-)
1747 if (inode->i_dtime && low_dtime_check &&
1748 inode->i_dtime < ctx->fs->super->s_inodes_count) {
1749 if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1750 inode->i_dtime = inode->i_links_count ?
1752 e2fsck_write_inode(ctx, ino, inode,
1758 if (inode->i_links_count) {
1759 pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1760 ino, inode->i_links_count);
1762 pctx.num = inode->i_links_count;
1763 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1764 ctx->flags |= E2F_FLAG_ABORT;
1767 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1768 !quota_inum_is_reserved(fs, ino)) {
1769 if (!inode->i_dtime && inode->i_mode) {
1770 if (fix_problem(ctx,
1771 PR_1_ZERO_DTIME, &pctx)) {
1772 inode->i_dtime = ctx->now;
1773 e2fsck_write_inode(ctx, ino, inode,
1778 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1782 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1783 ((!LINUX_S_ISDIR(inode->i_mode) &&
1784 fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1786 fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1787 inode->i_flags &= ~EXT4_CASEFOLD_FL;
1788 e2fsck_write_inode(ctx, ino, inode, "pass1");
1791 /* Conflicting inlinedata/extents inode flags? */
1792 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1793 (inode->i_flags & EXT4_EXTENTS_FL)) {
1794 int res = fix_inline_data_extents_file(ctx, ino, inode,
1798 /* skip FINISH_INODE_LOOP */
1803 /* Test for incorrect inline_data flags settings. */
1804 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1805 (ino >= EXT2_FIRST_INODE(fs->super))) {
1808 pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1809 if (!pctx.errcode &&
1810 fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1811 e2fsck_pass1_fix_lock(ctx);
1812 ext2fs_set_feature_inline_data(sb);
1813 ext2fs_mark_super_dirty(fs);
1814 e2fsck_pass1_fix_unlock(ctx);
1816 } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1817 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1818 /* skip FINISH_INODE_LOOP */
1823 /* Test for inline data flag but no attr */
1824 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1825 (ino >= EXT2_FIRST_INODE(fs->super))) {
1832 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1833 err = get_inline_data_ea_size(fs, ino, &size);
1834 fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1835 (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1839 /* Everything is awesome... */
1841 case EXT2_ET_BAD_EA_BLOCK_NUM:
1842 case EXT2_ET_BAD_EA_HASH:
1843 case EXT2_ET_BAD_EA_HEADER:
1844 case EXT2_ET_EA_BAD_NAME_LEN:
1845 case EXT2_ET_EA_BAD_VALUE_SIZE:
1846 case EXT2_ET_EA_KEY_NOT_FOUND:
1847 case EXT2_ET_EA_NO_SPACE:
1848 case EXT2_ET_MISSING_EA_FEATURE:
1849 case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1850 case EXT2_ET_INLINE_DATA_NO_BLOCK:
1851 case EXT2_ET_INLINE_DATA_NO_SPACE:
1852 case EXT2_ET_NO_INLINE_DATA:
1853 case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1854 case EXT2_ET_EA_BAD_VALUE_OFFSET:
1855 case EXT2_ET_EA_INODE_CORRUPTED:
1856 /* broken EA or no system.data EA; truncate */
1857 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1859 err = ext2fs_inode_size_set(fs, inode, 0);
1862 ctx->flags |= E2F_FLAG_ABORT;
1865 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1866 memset(&inode->i_block, 0,
1867 sizeof(inode->i_block));
1868 e2fsck_write_inode(ctx, ino, inode,
1874 /* Some other kind of non-xattr error? */
1876 ctx->flags |= E2F_FLAG_ABORT;
1882 * Test for incorrect extent flag settings.
1884 * On big-endian machines we must be careful:
1885 * When the inode is read, the i_block array is not swapped
1886 * if the extent flag is set. Therefore if we are testing
1887 * for or fixing a wrongly-set flag, we must potentially
1888 * (un)swap before testing, or after fixing.
1892 * In this case the extents flag was set when read, so
1893 * extent_header_verify is ok. If the inode is cleared,
1894 * no need to swap... so no extra swapping here.
1896 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1897 (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1898 (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1899 if ((ext2fs_extent_header_verify(inode->i_block,
1900 sizeof(inode->i_block)) == 0) &&
1901 fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1902 e2fsck_pass1_fix_lock(ctx);
1903 ext2fs_set_feature_extents(sb);
1904 ext2fs_mark_super_dirty(fs);
1906 e2fsck_pass1_fix_unlock(ctx);
1907 } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1909 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1910 if (ino == EXT2_BAD_INO)
1911 ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1913 /* skip FINISH_INODE_LOOP */
1919 * For big-endian machines:
1920 * If the inode didn't have the extents flag set when it
1921 * was read, then the i_blocks array was swapped. To test
1922 * as an extents header, we must swap it back first.
1923 * IF we then set the extents flag, the entire i_block
1924 * array must be un/re-swapped to make it proper extents data.
1926 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1927 (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1928 (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1929 (LINUX_S_ISREG(inode->i_mode) ||
1930 LINUX_S_ISDIR(inode->i_mode))) {
1932 #ifdef WORDS_BIGENDIAN
1933 __u32 tmp_block[EXT2_N_BLOCKS];
1935 for (i = 0; i < EXT2_N_BLOCKS; i++)
1936 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1939 ehp = inode->i_block;
1941 if ((ext2fs_extent_header_verify(ehp,
1942 sizeof(inode->i_block)) == 0) &&
1943 (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1944 inode->i_flags |= EXT4_EXTENTS_FL;
1945 #ifdef WORDS_BIGENDIAN
1946 memcpy(inode->i_block, tmp_block,
1947 sizeof(inode->i_block));
1949 e2fsck_write_inode(ctx, ino, inode, "pass1");
1954 if (ino == EXT2_BAD_INO) {
1955 struct process_block_struct pb;
1957 if ((failed_csum || inode->i_mode || inode->i_uid ||
1958 inode->i_gid || inode->i_links_count ||
1959 (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1960 inode->i_file_acl) &&
1961 fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1962 memset(inode, 0, sizeof(struct ext2_inode));
1963 e2fsck_write_inode(ctx, ino, inode,
1968 e2fsck_pass1_block_map_r_lock(ctx);
1969 pctx.errcode = ext2fs_copy_bitmap(ctx->global_ctx ?
1970 ctx->global_ctx->block_found_map :
1971 ctx->block_found_map, &pb.fs_meta_blocks);
1972 e2fsck_pass1_block_map_r_unlock(ctx);
1975 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1976 ctx->flags |= E2F_FLAG_ABORT;
1979 pb.ino = EXT2_BAD_INO;
1980 pb.num_blocks = pb.last_block = 0;
1981 pb.last_db_block = -1;
1982 pb.num_illegal_blocks = 0;
1983 pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1984 pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1988 pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1989 block_buf, process_bad_block, &pb);
1990 ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1992 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1993 ctx->flags |= E2F_FLAG_ABORT;
1997 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1998 ctx->flags |= E2F_FLAG_ABORT;
2001 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2002 clear_problem_context(&pctx);
2003 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2005 } else if (ino == EXT2_ROOT_INO) {
2007 * Make sure the root inode is a directory; if
2008 * not, offer to clear it. It will be
2009 * regenerated in pass #3.
2011 if (!LINUX_S_ISDIR(inode->i_mode)) {
2012 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
2016 * If dtime is set, offer to clear it. mke2fs
2017 * version 0.2b created filesystems with the
2018 * dtime field set for the root and lost+found
2019 * directories. We won't worry about
2020 * /lost+found, since that can be regenerated
2021 * easily. But we will fix the root directory
2022 * as a special case.
2024 if (inode->i_dtime && inode->i_links_count) {
2025 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
2027 e2fsck_write_inode(ctx, ino, inode,
2032 } else if (ino == EXT2_JOURNAL_INO) {
2033 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2034 if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
2035 if (!LINUX_S_ISREG(inode->i_mode) &&
2036 fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
2038 inode->i_mode = LINUX_S_IFREG;
2039 e2fsck_write_inode(ctx, ino, inode,
2043 check_blocks(ctx, &pctx, block_buf, NULL);
2044 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2047 if ((inode->i_links_count ||
2048 inode->i_blocks || inode->i_block[0]) &&
2049 fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
2051 memset(inode, 0, inode_size);
2052 ext2fs_icount_store(ctx->inode_link_info,
2054 e2fsck_write_inode_full(ctx, ino, inode,
2055 inode_size, "pass1");
2058 } else if (quota_inum_is_reserved(fs, ino)) {
2059 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2060 if (ext2fs_has_feature_quota(fs->super) &&
2061 quota_inum_is_super(fs->super, ino)) {
2062 if (!LINUX_S_ISREG(inode->i_mode) &&
2063 fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
2065 inode->i_mode = LINUX_S_IFREG;
2066 e2fsck_write_inode(ctx, ino, inode,
2070 check_blocks(ctx, &pctx, block_buf, NULL);
2071 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2074 if ((inode->i_links_count ||
2075 inode->i_blocks || inode->i_block[0]) &&
2076 fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
2078 memset(inode, 0, inode_size);
2079 ext2fs_icount_store(ctx->inode_link_info,
2081 e2fsck_write_inode_full(ctx, ino, inode,
2082 inode_size, "pass1");
2085 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
2086 problem_t problem = 0;
2088 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2089 if (ino == EXT2_BOOT_LOADER_INO) {
2090 if (LINUX_S_ISDIR(inode->i_mode))
2091 problem = PR_1_RESERVED_BAD_MODE;
2092 } else if (ino == EXT2_RESIZE_INO) {
2093 if (inode->i_mode &&
2094 !LINUX_S_ISREG(inode->i_mode))
2095 problem = PR_1_RESERVED_BAD_MODE;
2097 if (inode->i_mode != 0)
2098 problem = PR_1_RESERVED_BAD_MODE;
2101 if (fix_problem(ctx, problem, &pctx)) {
2103 e2fsck_write_inode(ctx, ino, inode,
2108 check_blocks(ctx, &pctx, block_buf, NULL);
2109 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2113 if (!inode->i_links_count) {
2114 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2118 * n.b. 0.3c ext2fs code didn't clear i_links_count for
2119 * deleted files. Oops.
2121 * Since all new ext2 implementations get this right,
2122 * we now assume that the case of non-zero
2123 * i_links_count and non-zero dtime means that we
2124 * should keep the file, not delete it.
2127 if (inode->i_dtime) {
2128 if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
2130 e2fsck_write_inode(ctx, ino, inode, "pass1");
2135 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2136 switch (fs->super->s_creator_os) {
2138 frag = inode->osd2.hurd2.h_i_frag;
2139 fsize = inode->osd2.hurd2.h_i_fsize;
2145 if (inode->i_faddr || frag || fsize ||
2146 (!ext2fs_has_feature_largedir(fs->super) &&
2147 (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
2148 mark_inode_bad(ctx, ino);
2149 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
2150 !ext2fs_has_feature_64bit(fs->super) &&
2151 inode->osd2.linux2.l_i_file_acl_high != 0)
2152 mark_inode_bad(ctx, ino);
2153 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
2154 !ext2fs_has_feature_huge_file(fs->super) &&
2155 (inode->osd2.linux2.l_i_blocks_hi != 0))
2156 mark_inode_bad(ctx, ino);
2157 if (inode->i_flags & EXT2_IMAGIC_FL) {
2159 if (!ctx->inode_imagic_map)
2160 alloc_imagic_map(ctx);
2161 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
2164 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
2165 inode->i_flags &= ~EXT2_IMAGIC_FL;
2166 e2fsck_write_inode(ctx, ino,
2173 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
2174 check_is_really_dir(ctx, &pctx, block_buf);
2177 * ext2fs_inode_has_valid_blocks2 does not actually look
2178 * at i_block[] values, so not endian-sensitive here.
2180 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
2181 LINUX_S_ISLNK(inode->i_mode) &&
2182 !ext2fs_inode_has_valid_blocks2(fs, inode) &&
2183 fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
2184 inode->i_flags &= ~EXT4_EXTENTS_FL;
2185 e2fsck_write_inode(ctx, ino, inode, "pass1");
2189 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
2190 add_encrypted_file(ctx, &pctx) < 0)
2193 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
2194 ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
2196 if (LINUX_S_ISDIR(inode->i_mode)) {
2197 ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
2198 e2fsck_add_dir_info(ctx, ino, 0);
2199 ctx->fs_directory_count++;
2200 if (inode->i_flags & EXT4_CASEFOLD_FL)
2201 add_casefolded_dir(ctx, ino);
2202 } else if (LINUX_S_ISREG (inode->i_mode)) {
2203 ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
2204 ctx->fs_regular_count++;
2205 } else if (LINUX_S_ISCHR (inode->i_mode) &&
2206 e2fsck_pass1_check_device_inode(fs, inode)) {
2207 check_extents_inlinedata(ctx, &pctx);
2208 check_immutable(ctx, &pctx);
2209 check_size(ctx, &pctx);
2210 ctx->fs_chardev_count++;
2211 } else if (LINUX_S_ISBLK (inode->i_mode) &&
2212 e2fsck_pass1_check_device_inode(fs, inode)) {
2213 check_extents_inlinedata(ctx, &pctx);
2214 check_immutable(ctx, &pctx);
2215 check_size(ctx, &pctx);
2216 ctx->fs_blockdev_count++;
2217 } else if (LINUX_S_ISLNK (inode->i_mode) &&
2218 e2fsck_pass1_check_symlink(fs, ino, inode,
2220 check_immutable(ctx, &pctx);
2221 ctx->fs_symlinks_count++;
2222 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
2223 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2225 } else if (ext2fs_is_fast_symlink(inode)) {
2226 ctx->fs_fast_symlinks_count++;
2227 check_blocks(ctx, &pctx, block_buf,
2229 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2233 else if (LINUX_S_ISFIFO (inode->i_mode) &&
2234 e2fsck_pass1_check_device_inode(fs, inode)) {
2235 check_extents_inlinedata(ctx, &pctx);
2236 check_immutable(ctx, &pctx);
2237 check_size(ctx, &pctx);
2238 ctx->fs_fifo_count++;
2239 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
2240 e2fsck_pass1_check_device_inode(fs, inode)) {
2241 check_extents_inlinedata(ctx, &pctx);
2242 check_immutable(ctx, &pctx);
2243 check_size(ctx, &pctx);
2244 ctx->fs_sockets_count++;
2246 mark_inode_bad(ctx, ino);
2247 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2248 !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
2249 if (inode->i_block[EXT2_IND_BLOCK])
2250 ctx->fs_ind_count++;
2251 if (inode->i_block[EXT2_DIND_BLOCK])
2252 ctx->fs_dind_count++;
2253 if (inode->i_block[EXT2_TIND_BLOCK])
2254 ctx->fs_tind_count++;
2256 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2257 !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
2258 (inode->i_block[EXT2_IND_BLOCK] ||
2259 inode->i_block[EXT2_DIND_BLOCK] ||
2260 inode->i_block[EXT2_TIND_BLOCK] ||
2261 ext2fs_file_acl_block(fs, inode))) {
2262 struct process_inode_block *itp;
2264 itp = &inodes_to_process[process_inode_count];
2266 itp->ea_ibody_quota = ea_ibody_quota;
2267 if (inode_size < sizeof(struct ext2_inode_large))
2268 memcpy(&itp->inode, inode, inode_size);
2270 memcpy(&itp->inode, inode, sizeof(itp->inode));
2271 process_inode_count++;
2273 check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
2275 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2277 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2280 if (process_inode_count >= ctx->process_inode_size) {
2281 process_inodes(ctx, block_buf, inodes_to_process,
2282 &process_inode_count);
2284 if (e2fsck_should_abort(ctx))
2288 process_inodes(ctx, block_buf, inodes_to_process,
2289 &process_inode_count);
2290 ext2fs_close_inode_scan(scan);
2293 if (ctx->ea_block_quota_blocks) {
2294 ea_refcount_free(ctx->ea_block_quota_blocks);
2295 ctx->ea_block_quota_blocks = 0;
2298 if (ctx->ea_block_quota_inodes) {
2299 ea_refcount_free(ctx->ea_block_quota_inodes);
2300 ctx->ea_block_quota_inodes = 0;
2303 /* We don't need the encryption policy => ID map any more */
2304 destroy_encryption_policy_map(ctx);
2306 if (ctx->flags & E2F_FLAG_RESTART) {
2308 * Only the master copy of the superblock and block
2309 * group descriptors are going to be written during a
2310 * restart, so set the superblock to be used to be the
2311 * master superblock.
2313 ctx->use_superblock = 0;
2317 if (ctx->large_dirs && !ext2fs_has_feature_largedir(fs->super)) {
2318 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2319 ext2fs_set_feature_largedir(fs->super);
2320 fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2321 ext2fs_mark_super_dirty(fs);
2323 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2324 fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2325 ext2fs_update_dynamic_rev(fs);
2326 ext2fs_mark_super_dirty(fs);
2330 ctx->flags |= E2F_FLAG_ALLOC_OK;
2332 e2fsck_use_inode_shortcuts(ctx, 0);
2333 ext2fs_free_mem(&inodes_to_process);
2334 inodes_to_process = 0;
2337 ext2fs_close_inode_scan(scan);
2339 ext2fs_free_mem(&block_buf);
2341 ext2fs_free_mem(&inode);
2344 * The l+f inode may have been cleared, so zap it now and
2345 * later passes will recalculate it if necessary
2347 ctx->lost_and_found = 0;
2349 if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2350 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2352 ctx->invalid_bitmaps++;
2356 static errcode_t e2fsck_pass1_copy_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2357 ext2fs_generic_bitmap *dest)
2361 ret = ext2fs_copy_bitmap(*src, dest);
2370 static void e2fsck_pass1_free_bitmap(ext2fs_generic_bitmap *bitmap)
2373 ext2fs_free_generic_bmap(*bitmap);
2379 static errcode_t e2fsck_pass1_merge_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2380 ext2fs_generic_bitmap *dest)
2385 if (*dest == NULL) {
2389 ret = ext2fs_merge_bitmap(*src, *dest, NULL, NULL);
2399 static errcode_t e2fsck_pass1_copy_fs(ext2_filsys dest, e2fsck_t src_context,
2404 memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2405 dest->inode_map = NULL;
2406 dest->block_map = NULL;
2407 dest->badblocks = NULL;
2409 dest->dblist->fs = dest;
2410 if (src->block_map) {
2411 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2416 if (src->inode_map) {
2417 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2423 if (src->badblocks) {
2424 retval = ext2fs_badblocks_copy(src->badblocks,
2430 /* disable it for now */
2431 src_context->openfs_flags &= ~EXT2_FLAG_EXCLUSIVE;
2432 retval = ext2fs_open_channel(dest, src_context->io_options,
2433 src_context->io_manager,
2434 src_context->openfs_flags,
2435 src->io->block_size);
2439 /* Block size might not be default */
2440 io_channel_set_blksize(dest->io, src->io->block_size);
2441 ehandler_init(dest->io);
2443 assert(dest->io->magic == src->io->magic);
2444 assert(dest->io->manager == src->io->manager);
2445 assert(strcmp(dest->io->name, src->io->name) == 0);
2446 assert(dest->io->block_size == src->io->block_size);
2447 assert(dest->io->read_error == src->io->read_error);
2448 assert(dest->io->write_error == src->io->write_error);
2449 assert(dest->io->refcount == src->io->refcount);
2450 assert(dest->io->flags == src->io->flags);
2451 assert(dest->io->app_data == dest);
2452 assert(src->io->app_data == src);
2453 assert(dest->io->align == src->io->align);
2455 /* The data should be written to disk immediately */
2456 dest->io->flags |= CHANNEL_FLAGS_WRITETHROUGH;
2457 /* icache will be rebuilt if needed, so do not copy from @src */
2462 static int e2fsck_pass1_merge_fs(ext2_filsys dest, ext2_filsys src)
2464 struct ext2_inode_cache *icache = dest->icache;
2465 errcode_t retval = 0;
2467 io_channel dest_image_io;
2468 ext2fs_inode_bitmap inode_map;
2469 ext2fs_block_bitmap block_map;
2470 ext2_badblocks_list badblocks;
2473 e2fsck_t dest_ctx = dest->priv_data;
2476 dest_image_io = dest->image_io;
2477 inode_map = dest->inode_map;
2478 block_map = dest->block_map;
2479 badblocks = dest->badblocks;
2480 dblist = dest->dblist;
2481 flags = dest->flags;
2483 memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2485 dest->image_io = dest_image_io;
2486 dest->icache = icache;
2487 dest->inode_map = inode_map;
2488 dest->block_map = block_map;
2489 dest->badblocks = badblocks;
2490 dest->dblist = dblist;
2491 dest->priv_data = dest_ctx;
2493 dest->dblist->fs = dest;
2494 dest->flags = src->flags | flags;
2495 if (!(src->flags & EXT2_FLAG_VALID) || !(flags & EXT2_FLAG_VALID))
2496 ext2fs_unmark_valid(dest);
2499 ext2fs_free_inode_cache(src->icache);
2503 retval = e2fsck_pass1_merge_bitmap(dest, &src->inode_map,
2508 retval = e2fsck_pass1_merge_bitmap(dest, &src->block_map,
2515 retval = ext2fs_merge_dblist(src->dblist,
2520 dest->dblist = src->dblist;
2521 dest->dblist->fs = dest;
2526 if (src->badblocks) {
2527 if (dest->badblocks == NULL)
2528 retval = ext2fs_badblocks_copy(src->badblocks,
2531 retval = ext2fs_badblocks_merge(src->badblocks,
2535 io_channel_close(src->io);
2537 ext2fs_free_generic_bmap(src->inode_map);
2539 ext2fs_free_generic_bmap(src->block_map);
2541 ext2fs_badblocks_list_free(src->badblocks);
2543 ext2fs_free_dblist(src->dblist);
2548 static void e2fsck_pass1_copy_invalid_bitmaps(e2fsck_t global_ctx,
2549 e2fsck_t thread_ctx)
2552 dgrp_t grp_start = thread_ctx->thread_info.et_group_start;
2553 dgrp_t grp_end = thread_ctx->thread_info.et_group_end;
2554 dgrp_t total = grp_end - grp_start;
2556 thread_ctx->invalid_inode_bitmap_flag =
2557 e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2558 "invalid_inode_bitmap");
2559 thread_ctx->invalid_block_bitmap_flag =
2560 e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2561 "invalid_block_bitmap");
2562 thread_ctx->invalid_inode_table_flag =
2563 e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2564 "invalid_inode_table");
2566 memcpy(thread_ctx->invalid_block_bitmap_flag,
2567 &global_ctx->invalid_block_bitmap_flag[grp_start],
2568 total * sizeof(int));
2569 memcpy(thread_ctx->invalid_inode_bitmap_flag,
2570 &global_ctx->invalid_inode_bitmap_flag[grp_start],
2571 total * sizeof(int));
2572 memcpy(thread_ctx->invalid_inode_table_flag,
2573 &global_ctx->invalid_inode_table_flag[grp_start],
2574 total * sizeof(int));
2576 thread_ctx->invalid_bitmaps = 0;
2577 for (i = grp_start, j = 0; i < grp_end; i++, j++) {
2578 if (thread_ctx->invalid_block_bitmap_flag[j])
2579 thread_ctx->invalid_bitmaps++;
2580 if (thread_ctx->invalid_inode_bitmap_flag[j])
2581 thread_ctx->invalid_bitmaps++;
2582 if (thread_ctx->invalid_inode_table_flag[j])
2583 thread_ctx->invalid_bitmaps++;
2587 static void e2fsck_pass1_merge_invalid_bitmaps(e2fsck_t global_ctx,
2588 e2fsck_t thread_ctx)
2590 dgrp_t grp_start = thread_ctx->thread_info.et_group_start;
2591 dgrp_t grp_end = thread_ctx->thread_info.et_group_end;
2592 dgrp_t total = grp_end - grp_start;
2594 memcpy(&global_ctx->invalid_block_bitmap_flag[grp_start],
2595 thread_ctx->invalid_block_bitmap_flag, total * sizeof(int));
2596 memcpy(&global_ctx->invalid_inode_bitmap_flag[grp_start],
2597 thread_ctx->invalid_inode_bitmap_flag, total * sizeof(int));
2598 memcpy(&global_ctx->invalid_inode_table_flag[grp_start],
2599 thread_ctx->invalid_inode_table_flag, total * sizeof(int));
2600 global_ctx->invalid_bitmaps += thread_ctx->invalid_bitmaps;
2603 static errcode_t e2fsck_pass1_thread_prepare(e2fsck_t global_ctx, e2fsck_t *thread_ctx,
2604 int thread_index, int num_threads,
2605 dgrp_t average_group)
2608 e2fsck_t thread_context;
2609 ext2_filsys thread_fs;
2610 ext2_filsys global_fs = global_ctx->fs;
2611 struct e2fsck_thread *tinfo;
2613 assert(global_ctx->inode_used_map == NULL);
2614 assert(global_ctx->inode_dir_map == NULL);
2615 assert(global_ctx->inode_bb_map == NULL);
2616 assert(global_ctx->inode_imagic_map == NULL);
2617 assert(global_ctx->inode_reg_map == NULL);
2618 assert(global_ctx->inodes_to_rebuild == NULL);
2620 assert(global_ctx->block_found_map != NULL);
2621 assert(global_ctx->block_metadata_map != NULL);
2622 assert(global_ctx->block_dup_map != NULL);
2623 assert(global_ctx->block_ea_map == NULL);
2624 assert(global_ctx->fs->dblist == NULL);
2626 retval = ext2fs_get_mem(sizeof(struct e2fsck_struct), &thread_context);
2628 com_err(global_ctx->program_name, retval, "while allocating memory");
2631 memcpy(thread_context, global_ctx, sizeof(struct e2fsck_struct));
2632 thread_context->block_dup_map = NULL;
2633 thread_context->casefolded_dirs = NULL;
2635 retval = e2fsck_allocate_block_bitmap(global_ctx->fs,
2636 _("in-use block map"), EXT2FS_BMAP64_RBTREE,
2637 "block_found_map", &thread_context->block_found_map);
2641 thread_context->global_ctx = global_ctx;
2642 retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &thread_fs);
2644 com_err(global_ctx->program_name, retval, "while allocating memory");
2648 io_channel_flush_cleanup(global_fs->io);
2649 retval = e2fsck_pass1_copy_fs(thread_fs, global_ctx, global_fs);
2651 com_err(global_ctx->program_name, retval, "while copying fs");
2654 thread_fs->priv_data = thread_context;
2656 thread_context->thread_info.et_thread_index = thread_index;
2657 set_up_logging(thread_context);
2659 tinfo = &thread_context->thread_info;
2660 tinfo->et_group_start = average_group * thread_index;
2661 if (thread_index == global_fs->fs_num_threads - 1)
2662 tinfo->et_group_end = thread_fs->group_desc_count;
2664 tinfo->et_group_end = average_group * (thread_index + 1);
2665 tinfo->et_group_next = tinfo->et_group_start;
2666 tinfo->et_inode_number = 0;
2667 tinfo->et_log_buf[0] = '\0';
2668 tinfo->et_log_length = 0;
2669 if (thread_context->options & E2F_OPT_MULTITHREAD)
2670 log_out(thread_context, _("Scan group range [%d, %d)\n"),
2671 tinfo->et_group_start, tinfo->et_group_end);
2672 thread_context->fs = thread_fs;
2673 retval = quota_init_context(&thread_context->qctx, thread_fs, 0);
2675 com_err(global_ctx->program_name, retval,
2676 "while init quota context");
2679 *thread_ctx = thread_context;
2680 e2fsck_pass1_copy_invalid_bitmaps(global_ctx, thread_context);
2683 ext2fs_free_mem(&thread_fs);
2685 if (thread_context->block_found_map)
2686 ext2fs_free_mem(&thread_context->block_found_map);
2687 ext2fs_free_mem(&thread_context);
2691 static void e2fsck_pass1_merge_dir_info(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2693 if (thread_ctx->dir_info == NULL)
2696 if (global_ctx->dir_info == NULL) {
2697 global_ctx->dir_info = thread_ctx->dir_info;
2698 thread_ctx->dir_info = NULL;
2702 e2fsck_merge_dir_info(global_ctx, thread_ctx->dir_info,
2703 global_ctx->dir_info);
2706 static void e2fsck_pass1_merge_dx_dir(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2708 if (thread_ctx->dx_dir_info == NULL)
2711 if (global_ctx->dx_dir_info == NULL) {
2712 global_ctx->dx_dir_info = thread_ctx->dx_dir_info;
2713 global_ctx->dx_dir_info_size = thread_ctx->dx_dir_info_size;
2714 global_ctx->dx_dir_info_count = thread_ctx->dx_dir_info_count;
2715 thread_ctx->dx_dir_info = NULL;
2719 e2fsck_merge_dx_dir(global_ctx, thread_ctx);
2722 static inline errcode_t
2723 e2fsck_pass1_merge_icount(ext2_icount_t *dest_icount,
2724 ext2_icount_t *src_icount)
2727 if (*dest_icount == NULL) {
2728 *dest_icount = *src_icount;
2733 ret = ext2fs_icount_merge(*src_icount,
2743 static errcode_t e2fsck_pass1_merge_icounts(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2747 ret = e2fsck_pass1_merge_icount(&global_ctx->inode_count,
2748 &thread_ctx->inode_count);
2751 ret = e2fsck_pass1_merge_icount(&global_ctx->inode_link_info,
2752 &thread_ctx->inode_link_info);
2757 static errcode_t e2fsck_pass1_merge_dirs_to_hash(e2fsck_t global_ctx,
2758 e2fsck_t thread_ctx)
2760 errcode_t retval = 0;
2762 if (!thread_ctx->dirs_to_hash)
2765 if (!global_ctx->dirs_to_hash)
2766 retval = ext2fs_badblocks_copy(thread_ctx->dirs_to_hash,
2767 &global_ctx->dirs_to_hash);
2769 retval = ext2fs_badblocks_merge(thread_ctx->dirs_to_hash,
2770 global_ctx->dirs_to_hash);
2775 static errcode_t e2fsck_pass1_merge_ea_inode_refs(e2fsck_t global_ctx,
2776 e2fsck_t thread_ctx)
2782 if (!thread_ctx->ea_inode_refs)
2785 if (!global_ctx->ea_inode_refs) {
2786 global_ctx->ea_inode_refs = thread_ctx->ea_inode_refs;
2787 thread_ctx->ea_inode_refs = NULL;
2791 ea_refcount_intr_begin(thread_ctx->ea_inode_refs);
2793 if ((blk = ea_refcount_intr_next(thread_ctx->ea_inode_refs,
2796 if (!global_ctx->block_ea_map ||
2797 !ext2fs_fast_test_block_bitmap2(global_ctx->block_ea_map,
2799 retval = ea_refcount_store(global_ctx->ea_inode_refs,
2809 static ea_value_t ea_refcount_usage(e2fsck_t ctx, blk64_t blk,
2812 ea_value_t count_cur;
2813 ea_value_t count_extra = 0;
2814 ea_value_t count_orig;
2816 ea_refcount_fetch(ctx->refcount_orig, blk, &count_orig);
2817 ea_refcount_fetch(ctx->refcount, blk, &count_cur);
2818 /* most of time this is not needed */
2819 if (ctx->refcount_extra && count_cur == 0)
2820 ea_refcount_fetch(ctx->refcount_extra, blk, &count_extra);
2827 return count_orig + count_extra - count_cur;
2830 static errcode_t e2fsck_pass1_merge_ea_refcount(e2fsck_t global_ctx,
2831 e2fsck_t thread_ctx)
2835 errcode_t retval = 0;
2837 if (!thread_ctx->refcount)
2840 if (!global_ctx->refcount) {
2841 global_ctx->refcount = thread_ctx->refcount;
2842 thread_ctx->refcount = NULL;
2843 global_ctx->refcount_extra = thread_ctx->refcount;
2844 thread_ctx->refcount_extra = NULL;
2848 ea_refcount_intr_begin(thread_ctx->refcount);
2850 if ((blk = ea_refcount_intr_next(thread_ctx->refcount,
2854 * this EA has never seen before, so just store its
2855 * refcount and refcount_extra into global_ctx if needed.
2857 if (!global_ctx->block_ea_map ||
2858 !ext2fs_fast_test_block_bitmap2(global_ctx->block_ea_map,
2862 retval = ea_refcount_store(global_ctx->refcount,
2867 if (count > 0 || !thread_ctx->refcount_extra)
2869 ea_refcount_fetch(thread_ctx->refcount_extra, blk,
2874 if (!global_ctx->refcount_extra) {
2875 retval = ea_refcount_create(0,
2876 &global_ctx->refcount_extra);
2880 retval = ea_refcount_store(global_ctx->refcount_extra,
2886 ea_value_t thread_usage;
2887 ea_value_t global_usage;
2890 thread_usage = ea_refcount_usage(thread_ctx,
2892 global_usage = ea_refcount_usage(global_ctx,
2894 if (thread_usage + global_usage <= orig) {
2895 new = orig - thread_usage - global_usage;
2896 retval = ea_refcount_store(global_ctx->refcount,
2902 /* update it is as zero */
2903 retval = ea_refcount_store(global_ctx->refcount,
2907 /* Ooops, this EA was referenced more than it stated */
2908 if (!global_ctx->refcount_extra) {
2909 retval = ea_refcount_create(0,
2910 &global_ctx->refcount_extra);
2914 new = global_usage + thread_usage - orig;
2915 retval = ea_refcount_store(global_ctx->refcount_extra,
2925 static errcode_t e2fsck_pass1_merge_casefolded_dirs(e2fsck_t global_ctx,
2926 e2fsck_t thread_ctx)
2928 errcode_t retval = 0;
2930 if (!thread_ctx->casefolded_dirs)
2933 if (!global_ctx->casefolded_dirs)
2934 retval = ext2fs_badblocks_copy(thread_ctx->casefolded_dirs,
2935 &global_ctx->casefolded_dirs);
2937 retval = ext2fs_badblocks_merge(thread_ctx->casefolded_dirs,
2938 global_ctx->casefolded_dirs);
2943 static errcode_t e2fsck_pass1_merge_context(e2fsck_t global_ctx,
2944 e2fsck_t thread_ctx)
2946 ext2_filsys global_fs = global_ctx->fs;
2950 global_ctx->fs_directory_count += thread_ctx->fs_directory_count;
2951 global_ctx->fs_regular_count += thread_ctx->fs_regular_count;
2952 global_ctx->fs_blockdev_count += thread_ctx->fs_blockdev_count;
2953 global_ctx->fs_chardev_count += thread_ctx->fs_chardev_count;
2954 global_ctx->fs_links_count += thread_ctx->fs_links_count;
2955 global_ctx->fs_symlinks_count += thread_ctx->fs_symlinks_count;
2956 global_ctx->fs_fast_symlinks_count += thread_ctx->fs_fast_symlinks_count;
2957 global_ctx->fs_fifo_count += thread_ctx->fs_fifo_count;
2958 global_ctx->fs_total_count += thread_ctx->fs_total_count;
2959 global_ctx->fs_badblocks_count += thread_ctx->fs_badblocks_count;
2960 global_ctx->fs_sockets_count += thread_ctx->fs_sockets_count;
2961 global_ctx->fs_ind_count += thread_ctx->fs_ind_count;
2962 global_ctx->fs_dind_count += thread_ctx->fs_dind_count;
2963 global_ctx->fs_tind_count += thread_ctx->fs_tind_count;
2964 global_ctx->fs_fragmented += thread_ctx->fs_fragmented;
2965 global_ctx->fs_fragmented_dir += thread_ctx->fs_fragmented_dir;
2966 global_ctx->large_files += thread_ctx->large_files;
2967 /* threads might enable E2F_OPT_YES */
2968 global_ctx->options |= thread_ctx->options;
2969 global_ctx->flags |= thread_ctx->flags;
2971 * The l+f inode may have been cleared, so zap it now and
2972 * later passes will recalculate it if necessary
2974 global_ctx->lost_and_found = 0;
2975 /* merge extent depth count */
2976 for (i = 0; i < MAX_EXTENT_DEPTH_COUNT; i++)
2977 global_ctx->extent_depth_count[i] +=
2978 thread_ctx->extent_depth_count[i];
2980 e2fsck_pass1_merge_dir_info(global_ctx, thread_ctx);
2981 e2fsck_pass1_merge_dx_dir(global_ctx, thread_ctx);
2983 retval = e2fsck_pass1_merge_fs(global_ctx->fs, thread_ctx->fs);
2985 com_err(global_ctx->program_name, 0, _("while merging fs\n"));
2988 retval = e2fsck_pass1_merge_icounts(global_ctx, thread_ctx);
2990 com_err(global_ctx->program_name, 0,
2991 _("while merging icounts\n"));
2995 retval = e2fsck_pass1_merge_dirs_to_hash(global_ctx, thread_ctx);
2997 com_err(global_ctx->program_name, 0,
2998 _("while merging dirs to hash\n"));
3002 e2fsck_pass1_merge_ea_inode_refs(global_ctx, thread_ctx);
3003 e2fsck_pass1_merge_ea_refcount(global_ctx, thread_ctx);
3004 retval = quota_merge_and_update_usage(global_ctx->qctx,
3009 retval = e2fsck_pass1_merge_casefolded_dirs(global_ctx, thread_ctx);
3011 com_err(global_ctx->program_name, 0,
3012 _("while merging casefolded dirs\n"));
3016 e2fsck_pass1_merge_invalid_bitmaps(global_ctx, thread_ctx);
3018 retval = e2fsck_pass1_merge_bitmap(global_fs,
3019 &thread_ctx->inode_used_map,
3020 &global_ctx->inode_used_map);
3024 retval = e2fsck_pass1_merge_bitmap(global_fs,
3025 &thread_ctx->inode_bad_map,
3026 &global_ctx->inode_bad_map);
3029 retval = e2fsck_pass1_merge_bitmap(global_fs,
3030 &thread_ctx->inode_dir_map,
3031 &global_ctx->inode_dir_map);
3034 retval = e2fsck_pass1_merge_bitmap(global_fs,
3035 &thread_ctx->inode_bb_map,
3036 &global_ctx->inode_bb_map);
3039 retval = e2fsck_pass1_merge_bitmap(global_fs,
3040 &thread_ctx->inode_imagic_map,
3041 &global_ctx->inode_imagic_map);
3044 retval = e2fsck_pass1_merge_bitmap(global_fs,
3045 &thread_ctx->inode_reg_map,
3046 &global_ctx->inode_reg_map);
3049 retval = e2fsck_pass1_merge_bitmap(global_fs,
3050 &thread_ctx->inodes_to_rebuild,
3051 &global_ctx->inodes_to_rebuild);
3054 retval = e2fsck_pass1_merge_bitmap(global_fs,
3055 &thread_ctx->block_ea_map,
3056 &global_ctx->block_ea_map);
3060 if (ext2fs_has_feature_shared_blocks(global_fs->super) &&
3061 !(global_ctx->options & E2F_OPT_UNSHARE_BLOCKS))
3064 * This need be done after merging block_ea_map
3065 * because ea block might be shared, we need exclude
3066 * them from dup blocks.
3068 e2fsck_pass1_block_map_w_lock(thread_ctx);
3069 retval = ext2fs_merge_bitmap(thread_ctx->block_found_map,
3070 global_ctx->block_found_map,
3071 global_ctx->block_dup_map,
3072 global_ctx->block_ea_map);
3073 e2fsck_pass1_block_map_w_unlock(thread_ctx);
3074 if (retval == EEXIST)
3075 global_ctx->flags |= E2F_FLAG_DUP_BLOCK;
3080 static int e2fsck_pass1_thread_join(e2fsck_t global_ctx, e2fsck_t thread_ctx)
3084 retval = e2fsck_pass1_merge_context(global_ctx, thread_ctx);
3085 ext2fs_free_mem(&thread_ctx->fs);
3086 if (thread_ctx->logf)
3087 fclose(thread_ctx->logf);
3088 if (thread_ctx->problem_logf) {
3089 fputs("</problem_log>\n", thread_ctx->problem_logf);
3090 fclose(thread_ctx->problem_logf);
3092 e2fsck_pass1_free_bitmap(&thread_ctx->inode_used_map);
3093 e2fsck_pass1_free_bitmap(&thread_ctx->inode_bad_map);
3094 e2fsck_pass1_free_bitmap(&thread_ctx->inode_dir_map);
3095 e2fsck_pass1_free_bitmap(&thread_ctx->inode_bb_map);
3096 e2fsck_pass1_free_bitmap(&thread_ctx->inode_imagic_map);
3097 e2fsck_pass1_free_bitmap(&thread_ctx->inode_reg_map);
3098 e2fsck_pass1_free_bitmap(&thread_ctx->inodes_to_rebuild);
3099 e2fsck_pass1_free_bitmap(&thread_ctx->block_found_map);
3100 e2fsck_pass1_free_bitmap(&thread_ctx->block_ea_map);
3101 if (thread_ctx->refcount)
3102 ea_refcount_free(thread_ctx->refcount);
3103 if (thread_ctx->refcount_extra)
3104 ea_refcount_free(thread_ctx->refcount_extra);
3105 if (thread_ctx->ea_inode_refs)
3106 ea_refcount_free(thread_ctx->ea_inode_refs);
3107 if (thread_ctx->refcount_orig)
3108 ea_refcount_free(thread_ctx->refcount_orig);
3109 e2fsck_free_dir_info(thread_ctx);
3110 ext2fs_free_icount(thread_ctx->inode_count);
3111 ext2fs_free_icount(thread_ctx->inode_link_info);
3112 if (thread_ctx->dirs_to_hash)
3113 ext2fs_badblocks_list_free(thread_ctx->dirs_to_hash);
3114 quota_release_context(&thread_ctx->qctx);
3115 ext2fs_free_mem(&thread_ctx->invalid_block_bitmap_flag);
3116 ext2fs_free_mem(&thread_ctx->invalid_inode_bitmap_flag);
3117 ext2fs_free_mem(&thread_ctx->invalid_inode_table_flag);
3118 ext2fs_free_mem(&thread_ctx);
3123 static int e2fsck_pass1_threads_join(struct e2fsck_thread_info *infos,
3124 e2fsck_t global_ctx)
3129 struct e2fsck_thread_info *pinfo;
3130 int num_threads = global_ctx->fs_num_threads;
3132 /* merge invalid bitmaps will recalculate it */
3133 global_ctx->invalid_bitmaps = 0;
3134 for (i = 0; i < num_threads; i++) {
3137 if (!pinfo->eti_started)
3140 rc = pthread_join(pinfo->eti_thread_id, NULL);
3142 com_err(global_ctx->program_name, rc,
3143 _("while joining thread\n"));
3147 rc = e2fsck_pass1_thread_join(global_ctx, infos[i].eti_thread_ctx);
3149 com_err(global_ctx->program_name, rc,
3150 _("while joining pass1 thread\n"));
3160 static void *e2fsck_pass1_thread(void *arg)
3162 struct e2fsck_thread_info *info = arg;
3163 e2fsck_t thread_ctx = info->eti_thread_ctx;
3164 #ifdef DEBUG_THREADS
3165 struct e2fsck_thread_debug *thread_debug = info->eti_debug;
3168 #ifdef DEBUG_THREADS
3169 pthread_mutex_lock(&thread_debug->etd_mutex);
3170 while (info->eti_thread_index > thread_debug->etd_finished_threads) {
3171 pthread_cond_wait(&thread_debug->etd_cond,
3172 &thread_debug->etd_mutex);
3174 pthread_mutex_unlock(&thread_debug->etd_mutex);
3177 #ifdef HAVE_SETJMP_H
3179 * When fatal_error() happens, jump to here. The thread
3180 * context's flags will be saved, but its abort_loc will
3181 * be overwritten by original jump buffer for the later
3184 if (setjmp(thread_ctx->abort_loc)) {
3185 thread_ctx->flags &= ~E2F_FLAG_SETJMP_OK;
3188 thread_ctx->flags |= E2F_FLAG_SETJMP_OK;
3191 e2fsck_pass1_run(thread_ctx);
3194 if (thread_ctx->options & E2F_OPT_MULTITHREAD)
3196 _("Scanned group range [%lu, %lu), inodes %lu\n"),
3197 thread_ctx->thread_info.et_group_start,
3198 thread_ctx->thread_info.et_group_end,
3199 thread_ctx->thread_info.et_inode_number);
3201 #ifdef DEBUG_THREADS
3202 pthread_mutex_lock(&thread_debug->etd_mutex);
3203 thread_debug->etd_finished_threads++;
3204 pthread_cond_broadcast(&thread_debug->etd_cond);
3205 pthread_mutex_unlock(&thread_debug->etd_mutex);
3211 static dgrp_t ext2fs_get_avg_group(ext2_filsys fs)
3214 dgrp_t average_group;
3215 unsigned flexbg_size;
3217 if (fs->fs_num_threads <= 1)
3218 return fs->group_desc_count;
3220 average_group = fs->group_desc_count / fs->fs_num_threads;
3221 if (average_group <= 1)
3224 if (ext2fs_has_feature_flex_bg(fs->super)) {
3227 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
3228 if (average_group % flexbg_size) {
3229 times = average_group / flexbg_size;
3230 average_group = times * flexbg_size;
3234 return average_group;
3236 return fs->group_desc_count;
3240 static int e2fsck_pass1_threads_start(struct e2fsck_thread_info **pinfo,
3241 e2fsck_t global_ctx)
3243 struct e2fsck_thread_info *infos;
3244 pthread_attr_t attr;
3247 struct e2fsck_thread_info *tmp_pinfo;
3249 e2fsck_t thread_ctx;
3250 dgrp_t average_group;
3251 int num_threads = global_ctx->fs_num_threads;
3252 #ifdef DEBUG_THREADS
3253 struct e2fsck_thread_debug thread_debug =
3254 {PTHREAD_MUTEX_INITIALIZER, PTHREAD_COND_INITIALIZER, 0};
3256 thread_debug.etd_finished_threads = 0;
3259 retval = pthread_attr_init(&attr);
3261 com_err(global_ctx->program_name, retval,
3262 _("while setting pthread attribute\n"));
3266 infos = calloc(num_threads, sizeof(struct e2fsck_thread_info));
3267 if (infos == NULL) {
3269 com_err(global_ctx->program_name, retval,
3270 _("while allocating memory for threads\n"));
3271 pthread_attr_destroy(&attr);
3275 average_group = ext2fs_get_avg_group(global_ctx->fs);
3276 for (i = 0; i < num_threads; i++) {
3277 tmp_pinfo = &infos[i];
3278 tmp_pinfo->eti_thread_index = i;
3279 #ifdef DEBUG_THREADS
3280 tmp_pinfo->eti_debug = &thread_debug;
3282 retval = e2fsck_pass1_thread_prepare(global_ctx, &thread_ctx,
3286 com_err(global_ctx->program_name, retval,
3287 _("while preparing pass1 thread\n"));
3290 tmp_pinfo->eti_thread_ctx = thread_ctx;
3292 retval = pthread_create(&tmp_pinfo->eti_thread_id, &attr,
3293 &e2fsck_pass1_thread, tmp_pinfo);
3295 com_err(global_ctx->program_name, retval,
3296 _("while creating thread\n"));
3297 e2fsck_pass1_thread_join(global_ctx, thread_ctx);
3301 tmp_pinfo->eti_started = 1;
3304 /* destroy the thread attribute object, since it is no longer needed */
3305 ret = pthread_attr_destroy(&attr);
3307 com_err(global_ctx->program_name, ret,
3308 _("while destroying thread attribute\n"));
3314 e2fsck_pass1_threads_join(infos, global_ctx);
3321 static void e2fsck_pass1_multithread(e2fsck_t global_ctx)
3323 struct e2fsck_thread_info *infos = NULL;
3326 retval = e2fsck_pass1_threads_start(&infos, global_ctx);
3328 com_err(global_ctx->program_name, retval,
3329 _("while starting pass1 threads\n"));
3333 retval = e2fsck_pass1_threads_join(infos, global_ctx);
3335 com_err(global_ctx->program_name, retval,
3336 _("while joining pass1 threads\n"));
3341 global_ctx->flags |= E2F_FLAG_ABORT;
3346 void e2fsck_pass1(e2fsck_t ctx)
3349 int need_single = 1;
3351 retval = e2fsck_pass1_prepare(ctx);
3355 if (ctx->fs_num_threads > 1 ||
3356 ctx->options & E2F_OPT_MULTITHREAD) {
3358 e2fsck_pass1_multithread(ctx);
3362 e2fsck_pass1_run(ctx);
3363 e2fsck_pass1_post(ctx);
3366 #undef FINISH_INODE_LOOP
3369 * When the inode_scan routines call this callback at the end of the
3370 * glock group, call process_inodes.
3372 static errcode_t scan_callback(ext2_filsys fs,
3373 ext2_inode_scan scan EXT2FS_ATTR((unused)),
3374 dgrp_t group, void * priv_data)
3376 struct scan_callback_struct *scan_struct;
3378 struct e2fsck_thread *tinfo;
3380 scan_struct = (struct scan_callback_struct *) priv_data;
3381 ctx = scan_struct->ctx;
3383 process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf,
3384 scan_struct->inodes_to_process,
3385 scan_struct->process_inode_count);
3388 if ((ctx->progress)(ctx, 1, group+1,
3389 ctx->fs->group_desc_count))
3390 return EXT2_ET_CANCEL_REQUESTED;
3393 if (ctx->global_ctx) {
3394 tinfo = &ctx->thread_info;
3395 tinfo->et_group_next++;
3396 if (ctx->options & E2F_OPT_DEBUG &&
3397 ctx->options & E2F_OPT_MULTITHREAD)
3398 log_out(ctx, _("group %d finished\n"),
3399 tinfo->et_group_next);
3400 if (tinfo->et_group_next >= tinfo->et_group_end)
3401 return EXT2_ET_SCAN_FINISHED;
3409 * Process the inodes in the "inodes to process" list.
3411 static void process_inodes(e2fsck_t ctx, char *block_buf,
3412 struct process_inode_block *inodes_to_process,
3413 int *process_inode_count)
3416 struct ext2_inode *old_stashed_inode;
3417 ext2_ino_t old_stashed_ino;
3418 const char *old_operation;
3420 struct problem_context pctx;
3423 printf("begin process_inodes: ");
3425 if (*process_inode_count == 0)
3427 old_operation = ehandler_operation(0);
3428 old_stashed_inode = ctx->stashed_inode;
3429 old_stashed_ino = ctx->stashed_ino;
3430 qsort(inodes_to_process, *process_inode_count,
3431 sizeof(struct process_inode_block), process_inode_cmp);
3432 clear_problem_context(&pctx);
3433 for (i=0; i < *process_inode_count; i++) {
3434 pctx.inode = ctx->stashed_inode =
3435 (struct ext2_inode *) &inodes_to_process[i].inode;
3436 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
3439 printf("%u ", pctx.ino);
3441 sprintf(buf, _("reading indirect blocks of inode %u"),
3443 ehandler_operation(buf);
3444 check_blocks(ctx, &pctx, block_buf,
3445 &inodes_to_process[i].ea_ibody_quota);
3446 if (e2fsck_should_abort(ctx))
3449 ctx->stashed_inode = old_stashed_inode;
3450 ctx->stashed_ino = old_stashed_ino;
3451 *process_inode_count = 0;
3453 printf("end process inodes\n");
3455 ehandler_operation(old_operation);
3458 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
3460 const struct process_inode_block *ib_a =
3461 (const struct process_inode_block *) a;
3462 const struct process_inode_block *ib_b =
3463 (const struct process_inode_block *) b;
3466 ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
3467 ib_b->inode.i_block[EXT2_IND_BLOCK]);
3470 * We only call process_inodes() for non-extent
3471 * inodes, so it's OK to pass NULL to
3472 * ext2fs_file_acl_block() here.
3474 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
3475 ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
3477 ret = ib_a->ino - ib_b->ino;
3482 * Mark an inode as being bad in some what
3484 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
3486 struct problem_context pctx;
3488 if (!ctx->inode_bad_map) {
3489 clear_problem_context(&pctx);
3491 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3492 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
3493 "inode_bad_map", &ctx->inode_bad_map);
3496 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3497 /* Should never get here */
3498 ctx->flags |= E2F_FLAG_ABORT;
3502 ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
3505 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino)
3507 struct problem_context pctx;
3509 if (!ctx->casefolded_dirs) {
3510 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
3514 pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
3515 if (pctx.errcode == 0)
3518 fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
3519 /* Should never get here */
3520 ctx->flags |= E2F_FLAG_ABORT;
3524 * This procedure will allocate the inode "bb" (badblock) map table
3526 static void alloc_bb_map(e2fsck_t ctx)
3528 struct problem_context pctx;
3530 clear_problem_context(&pctx);
3531 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3532 _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
3533 "inode_bb_map", &ctx->inode_bb_map);
3536 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3537 /* Should never get here */
3538 ctx->flags |= E2F_FLAG_ABORT;
3544 * This procedure will allocate the inode imagic table
3546 static void alloc_imagic_map(e2fsck_t ctx)
3548 struct problem_context pctx;
3550 clear_problem_context(&pctx);
3551 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3552 _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
3553 "inode_imagic_map", &ctx->inode_imagic_map);
3556 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3557 /* Should never get here */
3558 ctx->flags |= E2F_FLAG_ABORT;
3564 * Marks a block as in use, setting the dup_map if it's been set
3565 * already. Called by process_block and process_bad_block.
3567 * WARNING: Assumes checks have already been done to make sure block
3568 * is valid. This is true in both process_block and process_bad_block.
3570 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
3572 struct problem_context pctx;
3573 e2fsck_t global_ctx = ctx->global_ctx ? ctx->global_ctx : ctx;
3575 clear_problem_context(&pctx);
3577 if (is_blocks_used(ctx, block, 1)) {
3578 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
3579 !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
3582 ctx->flags |= E2F_FLAG_DUP_BLOCK;
3583 e2fsck_pass1_block_map_w_lock(ctx);
3584 ext2fs_fast_mark_block_bitmap2(global_ctx->block_dup_map, block);
3585 e2fsck_pass1_block_map_w_unlock(ctx);
3587 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
3592 * When cluster size is greater than one block, it is caller's responsibility
3593 * to make sure block parameter starts at a cluster boundary.
3595 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
3598 if (!is_blocks_used(ctx, block, num)) {
3599 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
3603 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
3604 mark_block_used(ctx, block + i);
3608 static errcode_t _INLINE_ e2fsck_write_ext_attr3(e2fsck_t ctx, blk64_t block,
3609 void *inbuf, ext2_ino_t inum)
3612 ext2_filsys fs = ctx->fs;
3614 e2fsck_pass1_fix_lock(ctx);
3615 retval = ext2fs_write_ext_attr3(fs, block, inbuf, inum);
3616 e2fsck_pass1_fix_unlock(ctx);
3621 * Adjust the extended attribute block's reference counts at the end
3622 * of pass 1, either by subtracting out references for EA blocks that
3623 * are still referenced in ctx->refcount, or by adding references for
3624 * EA blocks that had extra references as accounted for in
3625 * ctx->refcount_extra.
3627 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
3628 char *block_buf, int adjust_sign)
3630 struct ext2_ext_attr_header *header;
3631 struct problem_context pctx;
3632 ext2_filsys fs = ctx->fs;
3637 clear_problem_context(&pctx);
3639 ea_refcount_intr_begin(refcount);
3641 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
3644 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
3647 fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
3650 header = (struct ext2_ext_attr_header *) block_buf;
3651 pctx.blkcount = header->h_refcount;
3652 should_be = header->h_refcount + adjust_sign * (int)count;
3653 pctx.num = should_be;
3654 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
3655 header->h_refcount = should_be;
3656 pctx.errcode = e2fsck_write_ext_attr3(ctx, blk,
3660 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
3669 * Handle processing the extended attribute blocks
3671 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
3672 char *block_buf, struct ea_quota *ea_block_quota)
3674 ext2_filsys fs = ctx->fs;
3675 ext2_ino_t ino = pctx->ino;
3676 struct ext2_inode *inode = pctx->inode;
3679 struct ext2_ext_attr_header *header;
3680 struct ext2_ext_attr_entry *first, *entry;
3681 blk64_t quota_blocks = EXT2FS_C2B(fs, 1);
3682 __u64 quota_inodes = 0;
3683 region_t region = 0;
3684 int failed_csum = 0;
3686 ea_block_quota->blocks = 0;
3687 ea_block_quota->inodes = 0;
3689 blk = ext2fs_file_acl_block(fs, inode);
3694 * If the Extended attribute flag isn't set, then a non-zero
3695 * file acl means that the inode is corrupted.
3697 * Or if the extended attribute block is an invalid block,
3698 * then the inode is also corrupted.
3700 if (!ext2fs_has_feature_xattr(fs->super) ||
3701 (blk < fs->super->s_first_data_block) ||
3702 (blk >= ext2fs_blocks_count(fs->super))) {
3703 mark_inode_bad(ctx, ino);
3707 /* If ea bitmap hasn't been allocated, create it */
3708 if (!ctx->block_ea_map) {
3709 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
3710 _("ext attr block map"),
3711 EXT2FS_BMAP64_RBTREE, "block_ea_map",
3712 &ctx->block_ea_map);
3713 if (pctx->errcode) {
3715 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
3716 ctx->flags |= E2F_FLAG_ABORT;
3721 /* Create the EA refcount structure if necessary */
3722 if (!ctx->refcount) {
3723 pctx->errcode = ea_refcount_create(0,
3724 &ctx->refcount_orig);
3725 if (pctx->errcode) {
3727 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3728 ctx->flags |= E2F_FLAG_ABORT;
3732 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
3733 if (pctx->errcode) {
3735 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3736 ctx->flags |= E2F_FLAG_ABORT;
3742 /* Debugging text */
3743 printf("Inode %u has EA block %u\n", ino, blk);
3746 /* Have we seen this EA block before? */
3747 if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
3748 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
3749 ea_block_quota->inodes = 0;
3751 if (ctx->ea_block_quota_blocks) {
3752 ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
3755 ea_block_quota->blocks = quota_blocks;
3758 if (ctx->ea_block_quota_inodes)
3759 ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
3760 &ea_block_quota->inodes);
3762 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
3764 /* Ooops, this EA was referenced more than it stated */
3765 if (!ctx->refcount_extra) {
3766 pctx->errcode = ea_refcount_create(0,
3767 &ctx->refcount_extra);
3768 if (pctx->errcode) {
3770 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3771 ctx->flags |= E2F_FLAG_ABORT;
3775 ea_refcount_increment(ctx->refcount_extra, blk, 0);
3780 * OK, we haven't seen this EA block yet. So we need to
3784 pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
3785 if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
3788 } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
3791 if (pctx->errcode &&
3792 fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
3796 header = (struct ext2_ext_attr_header *) block_buf;
3797 pctx->blk = ext2fs_file_acl_block(fs, inode);
3798 if (((ctx->ext_attr_ver == 1) &&
3799 (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
3800 ((ctx->ext_attr_ver == 2) &&
3801 (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
3802 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
3806 if (header->h_blocks != 1) {
3807 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
3811 if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
3814 region = region_create(0, fs->blocksize);
3816 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
3817 ctx->flags |= E2F_FLAG_ABORT;
3820 if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
3821 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3825 first = (struct ext2_ext_attr_entry *)(header+1);
3826 end = block_buf + fs->blocksize;
3828 while ((char *)entry < end && *(__u32 *)entry) {
3831 if (region_allocate(region, (char *)entry - (char *)header,
3832 EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
3833 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3837 if ((ctx->ext_attr_ver == 1 &&
3838 (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
3839 (ctx->ext_attr_ver == 2 &&
3840 entry->e_name_index == 0)) {
3841 if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
3845 if (entry->e_value_inum == 0) {
3846 if (entry->e_value_size > EXT2_XATTR_SIZE_MAX ||
3847 (entry->e_value_offs + entry->e_value_size >
3849 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
3853 if (entry->e_value_size &&
3854 region_allocate(region, entry->e_value_offs,
3855 EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
3856 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
3861 hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
3862 entry->e_value_offs);
3864 if (entry->e_hash != hash) {
3865 pctx->num = entry->e_hash;
3866 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
3868 entry->e_hash = hash;
3872 blk64_t entry_quota_blocks;
3874 problem = check_large_ea_inode(ctx, entry, pctx,
3875 &entry_quota_blocks);
3876 if (problem && fix_problem(ctx, problem, pctx))
3879 quota_blocks += entry_quota_blocks;
3883 entry = EXT2_EXT_ATTR_NEXT(entry);
3885 if (region_allocate(region, (char *)entry - (char *)header, 4)) {
3886 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3889 region_free(region);
3892 * We only get here if there was no other errors that were fixed.
3893 * If there was a checksum fail, ask to correct it.
3896 fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
3897 pctx->errcode = e2fsck_write_ext_attr3(ctx, blk, block_buf,
3903 if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
3904 if (!ctx->ea_block_quota_blocks) {
3905 pctx->errcode = ea_refcount_create(0,
3906 &ctx->ea_block_quota_blocks);
3907 if (pctx->errcode) {
3912 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
3917 if (!ctx->ea_block_quota_inodes) {
3918 pctx->errcode = ea_refcount_create(0,
3919 &ctx->ea_block_quota_inodes);
3920 if (pctx->errcode) {
3923 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3924 ctx->flags |= E2F_FLAG_ABORT;
3929 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
3932 ea_block_quota->blocks = quota_blocks;
3933 ea_block_quota->inodes = quota_inodes;
3935 inc_ea_inode_refs(ctx, pctx, first, end);
3936 ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
3937 ea_refcount_store(ctx->refcount_orig, blk, header->h_refcount);
3939 * It might be racy that this block has been merged in the
3942 if (!is_blocks_used(ctx, blk, 1))
3943 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, blk);
3944 ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
3949 region_free(region);
3950 ext2fs_file_acl_block_set(fs, inode, 0);
3951 e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
3955 /* Returns 1 if bad htree, 0 if OK */
3956 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
3957 ext2_ino_t ino, struct ext2_inode *inode,
3960 struct ext2_dx_root_info *root;
3961 ext2_filsys fs = ctx->fs;
3965 if ((!LINUX_S_ISDIR(inode->i_mode) &&
3966 fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
3967 (!ext2fs_has_feature_dir_index(fs->super) &&
3968 fix_problem(ctx, PR_1_HTREE_SET, pctx)))
3971 pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
3973 if ((pctx->errcode) ||
3975 (blk < fs->super->s_first_data_block) ||
3976 (blk >= ext2fs_blocks_count(fs->super))) {
3977 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3983 retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
3984 if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3987 /* XXX should check that beginning matches a directory */
3988 root = (struct ext2_dx_root_info *) (block_buf + 24);
3990 if ((root->reserved_zero || root->info_length < 8) &&
3991 fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3994 pctx->num = root->hash_version;
3995 if ((root->hash_version != EXT2_HASH_LEGACY) &&
3996 (root->hash_version != EXT2_HASH_HALF_MD4) &&
3997 (root->hash_version != EXT2_HASH_TEA) &&
3998 (root->hash_version != EXT2_HASH_SIPHASH) &&
3999 fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
4002 if (ext4_hash_in_dirent(inode)) {
4003 if (root->hash_version != EXT2_HASH_SIPHASH &&
4004 fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
4007 if (root->hash_version == EXT2_HASH_SIPHASH &&
4008 fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
4012 if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
4013 fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
4016 pctx->num = root->indirect_levels;
4017 /* if htree level is clearly too high, consider it to be broken */
4018 if (root->indirect_levels > EXT4_HTREE_LEVEL &&
4019 fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
4022 /* if level is only maybe too high, LARGE_DIR feature could be unset */
4023 if (root->indirect_levels > ext2_dir_htree_level(fs) &&
4024 !ext2fs_has_feature_largedir(fs->super)) {
4025 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
4026 unsigned idx_pb = 1 << (blockbits - 3);
4028 /* compare inode size/blocks vs. max-sized 2-level htree */
4029 if (EXT2_I_SIZE(pctx->inode) <
4030 (idx_pb - 1) * (idx_pb - 2) << blockbits &&
4031 pctx->inode->i_blocks <
4032 (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
4033 fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
4037 if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
4038 ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4044 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
4045 struct ext2_inode *inode, int restart_flag,
4049 inode->i_links_count = 0;
4050 ext2fs_icount_store(ctx->inode_link_info, ino, 0);
4051 inode->i_dtime = ctx->now;
4054 * If a special inode has such rotten block mappings that we
4055 * want to clear the whole inode, be sure to actually zap
4056 * the block maps because i_links_count isn't checked for
4057 * special inodes, and we'll end up right back here the next
4060 if (ino < EXT2_FIRST_INODE(ctx->fs->super))
4061 memset(inode->i_block, 0, sizeof(inode->i_block));
4063 ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
4064 ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
4065 if (ctx->inode_reg_map)
4066 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
4067 if (ctx->inode_bad_map)
4068 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
4071 * If the inode was partially accounted for before processing
4072 * was aborted, we need to restart the pass 1 scan.
4074 ctx->flags |= restart_flag;
4076 if (ino == EXT2_BAD_INO)
4077 memset(inode, 0, sizeof(struct ext2_inode));
4079 e2fsck_write_inode(ctx, ino, inode, source);
4083 * Use the multiple-blocks reclamation code to fix alignment problems in
4084 * a bigalloc filesystem. We want a logical cluster to map to *only* one
4085 * physical cluster, and we want the block offsets within that cluster to
4088 static int has_unaligned_cluster_map(e2fsck_t ctx,
4089 blk64_t last_pblk, blk64_t last_lblk,
4090 blk64_t pblk, blk64_t lblk)
4092 blk64_t cluster_mask;
4094 if (!ctx->fs->cluster_ratio_bits)
4096 cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
4099 * If the block in the logical cluster doesn't align with the block in
4100 * the physical cluster...
4102 if ((lblk & cluster_mask) != (pblk & cluster_mask))
4106 * If we cross a physical cluster boundary within a logical cluster...
4108 if (last_pblk && (lblk & cluster_mask) != 0 &&
4109 EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
4110 EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
4116 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
4117 struct process_block_struct *pb,
4118 blk64_t start_block, blk64_t end_block,
4120 ext2_extent_handle_t ehandle,
4123 struct ext2fs_extent extent;
4124 blk64_t blk, last_lblk;
4126 int is_dir, is_leaf;
4128 struct ext2_extent_info info;
4129 int failed_csum = 0;
4131 if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
4134 pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
4137 if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
4138 !pb->eti.force_rebuild &&
4139 info.curr_level < MAX_EXTENT_DEPTH_COUNT) {
4140 struct extent_tree_level *etl;
4142 etl = pb->eti.ext_info + info.curr_level;
4143 etl->num_extents += info.num_entries;
4144 etl->max_extents += info.max_entries;
4146 * Implementation wart: Splitting extent blocks when appending
4147 * will leave the old block with one free entry. Therefore
4148 * unless the node is totally full, pretend that a non-root
4149 * extent block can hold one fewer entry than it actually does,
4150 * so that we don't repeatedly rebuild the extent tree.
4152 if (info.curr_level && info.num_entries < info.max_entries)
4156 pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
4158 while ((pctx->errcode == 0 ||
4159 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
4160 info.num_entries-- > 0) {
4161 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
4162 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
4163 last_lblk = extent.e_lblk + extent.e_len - 1;
4166 pctx->blk = extent.e_pblk;
4167 pctx->blk2 = extent.e_lblk;
4168 pctx->num = extent.e_len;
4169 pctx->blkcount = extent.e_lblk + extent.e_len;
4171 if (extent.e_pblk == 0 ||
4172 extent.e_pblk < ctx->fs->super->s_first_data_block ||
4173 extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
4174 problem = PR_1_EXTENT_BAD_START_BLK;
4175 else if (extent.e_lblk < start_block)
4176 problem = PR_1_OUT_OF_ORDER_EXTENTS;
4177 else if ((end_block && last_lblk > end_block) &&
4178 !(last_lblk > eof_block &&
4179 ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
4180 (pctx->inode->i_flags & EXT4_VERITY_FL))))
4181 problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
4182 else if (is_leaf && extent.e_len == 0)
4183 problem = PR_1_EXTENT_LENGTH_ZERO;
4185 (extent.e_pblk + extent.e_len) >
4186 ext2fs_blocks_count(ctx->fs->super))
4187 problem = PR_1_EXTENT_ENDS_BEYOND;
4188 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
4189 !ext2fs_has_feature_largedir(ctx->fs->super) &&
4190 ((extent.e_lblk + extent.e_len) >
4191 (1U << (21 - ctx->fs->super->s_log_block_size))))
4192 problem = PR_1_TOOBIG_DIR;
4194 if (is_leaf && problem == 0 && extent.e_len > 0) {
4196 printf("extent_region(ino=%u, expect=%llu, "
4197 "lblk=%llu, len=%u)\n", pb->ino,
4198 (unsigned long long) pb->next_lblock,
4199 (unsigned long long) extent.e_lblk,
4202 if (extent.e_lblk < pb->next_lblock)
4203 problem = PR_1_EXTENT_COLLISION;
4204 else if (extent.e_lblk + extent.e_len > pb->next_lblock)
4205 pb->next_lblock = extent.e_lblk + extent.e_len;
4209 * Uninitialized blocks in a directory? Clear the flag and
4210 * we'll interpret the blocks later.
4212 if (try_repairs && is_dir && problem == 0 &&
4213 (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
4214 fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
4215 e2fsck_pass1_fix_lock(ctx);
4216 extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
4217 pb->inode_modified = 1;
4218 pctx->errcode = ext2fs_extent_replace(ehandle, 0,
4220 e2fsck_pass1_fix_unlock(ctx);
4225 #ifdef CONFIG_DEVELOPER_FEATURES
4226 if (try_repairs && !is_dir && problem == 0 &&
4227 (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
4228 (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
4229 fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
4230 extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
4231 pb->inode_modified = 1;
4232 pctx->errcode = ext2fs_extent_replace(ehandle, 0,
4239 if (try_repairs && problem) {
4241 if (fix_problem(ctx, problem, pctx)) {
4242 if (ctx->invalid_bitmaps) {
4244 * If fsck knows the bitmaps are bad,
4245 * skip to the next extent and
4246 * try to clear this extent again
4247 * after fixing the bitmaps, by
4250 pctx->errcode = ext2fs_extent_get(
4252 EXT2_EXTENT_NEXT_SIB,
4254 ctx->flags |= E2F_FLAG_RESTART_LATER;
4255 if (pctx->errcode ==
4256 EXT2_ET_NO_CURRENT_NODE) {
4262 e2fsck_pass1_fix_lock(ctx);
4263 e2fsck_read_bitmaps(ctx);
4264 pb->inode_modified = 1;
4266 ext2fs_extent_delete(ehandle, 0);
4267 e2fsck_pass1_fix_unlock(ctx);
4268 if (pctx->errcode) {
4269 pctx->str = "ext2fs_extent_delete";
4272 e2fsck_pass1_fix_lock(ctx);
4273 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
4274 e2fsck_pass1_fix_unlock(ctx);
4275 if (pctx->errcode &&
4276 pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
4277 pctx->str = "ext2fs_extent_fix_parents";
4280 pctx->errcode = ext2fs_extent_get(ehandle,
4281 EXT2_EXTENT_CURRENT,
4283 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
4294 blk64_t lblk = extent.e_lblk;
4295 int next_try_repairs = 1;
4297 blk = extent.e_pblk;
4300 * If this lower extent block collides with critical
4301 * metadata, don't try to repair the damage. Pass 1b
4302 * will reallocate the block; then we can try again.
4304 if (pb->ino != EXT2_RESIZE_INO &&
4305 extent.e_pblk < ctx->fs->super->s_blocks_count &&
4306 ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4308 next_try_repairs = 0;
4311 PR_1_CRITICAL_METADATA_COLLISION,
4313 if ((ctx->options & E2F_OPT_NO) == 0)
4314 ctx->flags |= E2F_FLAG_RESTART_LATER;
4316 pctx->errcode = ext2fs_extent_get(ehandle,
4317 EXT2_EXTENT_DOWN, &extent);
4318 if (pctx->errcode &&
4319 pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
4320 pctx->str = "EXT2_EXTENT_DOWN";
4321 problem = PR_1_EXTENT_HEADER_INVALID;
4322 if (!next_try_repairs)
4324 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
4325 goto report_problem;
4328 /* The next extent should match this index's logical start */
4329 if (extent.e_lblk != lblk) {
4330 struct ext2_extent_info e_info;
4332 pctx->errcode = ext2fs_extent_get_info(ehandle,
4334 if (pctx->errcode) {
4335 pctx->str = "ext2fs_extent_get_info";
4339 pctx->blk2 = extent.e_lblk;
4340 pctx->num = e_info.curr_level - 1;
4341 problem = PR_1_EXTENT_INDEX_START_INVALID;
4342 if (fix_problem(ctx, problem, pctx)) {
4343 e2fsck_pass1_fix_lock(ctx);
4344 pb->inode_modified = 1;
4346 ext2fs_extent_fix_parents(ehandle);
4347 e2fsck_pass1_fix_unlock(ctx);
4348 if (pctx->errcode) {
4349 pctx->str = "ext2fs_extent_fix_parents";
4354 scan_extent_node(ctx, pctx, pb, extent.e_lblk,
4355 last_lblk, eof_block, ehandle,
4359 pctx->errcode = ext2fs_extent_get(ehandle,
4360 EXT2_EXTENT_UP, &extent);
4361 if (pctx->errcode) {
4362 pctx->str = "EXT2_EXTENT_UP";
4365 mark_block_used(ctx, blk);
4370 if ((pb->previous_block != 0) &&
4371 (pb->previous_block+1 != extent.e_pblk)) {
4372 if (ctx->options & E2F_OPT_FRAGCHECK) {
4377 else if (pb->is_reg)
4380 printf(("%6lu(%c): expecting %6lu "
4382 "phys %6lu log %lu len %lu\n"),
4383 (unsigned long) pctx->ino, type,
4384 (unsigned long) pb->previous_block+1,
4385 (unsigned long) extent.e_pblk,
4386 (unsigned long) extent.e_lblk,
4387 (unsigned long) extent.e_len);
4392 * If we notice a gap in the logical block mappings of an
4393 * extent-mapped directory, offer to close the hole by
4394 * moving the logical block down, otherwise we'll go mad in
4395 * pass 3 allocating empty directory blocks to fill the hole.
4397 if (try_repairs && is_dir &&
4398 pb->last_block + 1 < extent.e_lblk) {
4401 new_lblk = pb->last_block + 1;
4402 if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
4403 new_lblk = ((new_lblk +
4404 EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
4405 ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
4407 EXT2FS_CLUSTER_MASK(ctx->fs));
4408 pctx->blk = extent.e_lblk;
4409 pctx->blk2 = new_lblk;
4410 if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
4411 e2fsck_pass1_fix_lock(ctx);
4412 extent.e_lblk = new_lblk;
4413 pb->inode_modified = 1;
4414 pctx->errcode = ext2fs_extent_replace(ehandle,
4416 e2fsck_pass1_fix_unlock(ctx);
4417 if (pctx->errcode) {
4421 e2fsck_pass1_fix_lock(ctx);
4422 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
4423 e2fsck_pass1_fix_unlock(ctx);
4425 goto failed_add_dir_block;
4426 pctx->errcode = ext2fs_extent_goto(ehandle,
4429 goto failed_add_dir_block;
4430 last_lblk = extent.e_lblk + extent.e_len - 1;
4436 while (++pb->last_db_block <
4437 (e2_blkcnt_t) extent.e_lblk) {
4438 pctx->errcode = ext2fs_add_dir_block2(
4442 if (pctx->errcode) {
4444 pctx->num = pb->last_db_block;
4445 goto failed_add_dir_block;
4449 for (i = 0; i < extent.e_len; i++) {
4450 pctx->errcode = ext2fs_add_dir_block2(
4455 if (pctx->errcode) {
4456 pctx->blk = extent.e_pblk + i;
4457 pctx->num = extent.e_lblk + i;
4458 failed_add_dir_block:
4459 fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4460 /* Should never get here */
4461 ctx->flags |= E2F_FLAG_ABORT;
4465 if (extent.e_len > 0)
4466 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
4468 if (has_unaligned_cluster_map(ctx, pb->previous_block,
4472 for (i = 0; i < extent.e_len; i++) {
4473 pctx->blk = extent.e_lblk + i;
4474 pctx->blk2 = extent.e_pblk + i;
4475 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4476 mark_block_used(ctx, extent.e_pblk + i);
4477 mark_block_used(ctx, extent.e_pblk + i);
4482 * Check whether first cluster got marked in previous iteration.
4484 if (ctx->fs->cluster_ratio_bits &&
4485 pb->previous_block &&
4486 (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
4487 EXT2FS_B2C(ctx->fs, pb->previous_block)))
4488 /* Set blk to the beginning of next cluster. */
4491 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
4493 /* Set blk to the beginning of current cluster. */
4494 blk = EXT2FS_C2B(ctx->fs,
4495 EXT2FS_B2C(ctx->fs, extent.e_pblk));
4497 if (blk < extent.e_pblk + extent.e_len) {
4498 mark_blocks_used(ctx, blk,
4499 extent.e_pblk + extent.e_len - blk);
4500 n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
4501 EXT2FS_CLUSTER_RATIO(ctx->fs));
4502 pb->num_blocks += n;
4504 pb->last_block = extent.e_lblk + extent.e_len - 1;
4505 pb->previous_block = extent.e_pblk + extent.e_len - 1;
4506 start_block = pb->last_block = last_lblk;
4507 if (is_leaf && !is_dir &&
4508 !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
4509 pb->last_init_lblock = last_lblk;
4511 pctx->errcode = ext2fs_extent_get(ehandle,
4512 EXT2_EXTENT_NEXT_SIB,
4516 /* Failed csum but passes checks? Ask to fix checksum. */
4518 fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
4519 e2fsck_pass1_fix_lock(ctx);
4520 pb->inode_modified = 1;
4521 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
4522 e2fsck_pass1_fix_unlock(ctx);
4527 if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
4531 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
4532 struct process_block_struct *pb)
4534 struct ext2_extent_info info;
4535 struct ext2_inode *inode = pctx->inode;
4536 ext2_extent_handle_t ehandle;
4537 ext2_filsys fs = ctx->fs;
4538 ext2_ino_t ino = pctx->ino;
4541 struct ext3_extent_header *eh;
4543 /* Check for a proper extent header... */
4544 eh = (struct ext3_extent_header *) &inode->i_block[0];
4545 retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
4547 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
4548 e2fsck_clear_inode(ctx, ino, inode, 0,
4549 "check_blocks_extents");
4554 /* ...since this function doesn't fail if i_block is zeroed. */
4555 pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
4556 if (pctx->errcode) {
4557 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
4558 e2fsck_clear_inode(ctx, ino, inode, 0,
4559 "check_blocks_extents");
4564 retval = ext2fs_extent_get_info(ehandle, &info);
4566 int max_depth = info.max_depth;
4568 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
4569 max_depth = MAX_EXTENT_DEPTH_COUNT-1;
4570 ctx->extent_depth_count[max_depth]++;
4573 /* Check maximum extent depth */
4574 pctx->blk = info.max_depth;
4575 pctx->blk2 = ext2fs_max_extent_depth(ehandle);
4576 if (pctx->blk2 < pctx->blk &&
4577 fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
4578 pb->eti.force_rebuild = 1;
4580 /* Can we collect extent tree level stats? */
4581 pctx->blk = MAX_EXTENT_DEPTH_COUNT;
4582 if (pctx->blk2 > pctx->blk)
4583 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
4584 memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
4585 pb->eti.ino = pb->ino;
4587 pb->next_lblock = 0;
4589 eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
4590 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
4591 scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
4592 if (pctx->errcode &&
4593 fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
4595 inode->i_blocks = 0;
4596 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4597 "check_blocks_extents");
4600 ext2fs_extent_free(ehandle);
4602 /* Rebuild unless it's a dir and we're rehashing it */
4603 if (LINUX_S_ISDIR(inode->i_mode) &&
4604 e2fsck_dir_will_be_rehashed(ctx, ino))
4607 if (ctx->options & E2F_OPT_CONVERT_BMAP)
4608 e2fsck_rebuild_extents_later(ctx, ino);
4610 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
4614 * In fact we don't need to check blocks for an inode with inline data
4615 * because this inode doesn't have any blocks. In this function all
4616 * we need to do is add this inode into dblist when it is a directory.
4618 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
4619 struct process_block_struct *pb)
4622 size_t inline_data_size = 0;
4629 /* Process the dirents in i_block[] as the "first" block. */
4630 pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
4634 /* Process the dirents in the EA as a "second" block. */
4635 flags = ctx->fs->flags;
4636 ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4637 pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
4639 ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4640 (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4641 if (pctx->errcode) {
4646 if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
4649 pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
4657 fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4658 ctx->flags |= E2F_FLAG_ABORT;
4662 * This subroutine is called on each inode to account for all of the
4663 * blocks used by that inode.
4665 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
4666 char *block_buf, const struct ea_quota *ea_ibody_quota)
4668 ext2_filsys fs = ctx->fs;
4669 struct process_block_struct pb;
4670 ext2_ino_t ino = pctx->ino;
4671 struct ext2_inode *inode = pctx->inode;
4672 unsigned bad_size = 0;
4673 int dirty_inode = 0;
4677 struct ea_quota ea_block_quota;
4680 pb.num_blocks = EXT2FS_B2C(ctx->fs,
4681 ea_ibody_quota ? ea_ibody_quota->blocks : 0);
4683 pb.last_init_lblock = -1;
4684 pb.last_db_block = -1;
4685 pb.num_illegal_blocks = 0;
4686 pb.suppress = 0; pb.clear = 0;
4689 pb.previous_block = 0;
4690 pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
4691 pb.is_reg = LINUX_S_ISREG(inode->i_mode);
4692 pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
4696 pb.inode_modified = 0;
4697 pb.eti.force_rebuild = 0;
4701 extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
4702 inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
4704 if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
4705 if (e2fsck_should_abort(ctx))
4707 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
4710 if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
4711 check_blocks_inline_data(ctx, pctx, &pb);
4712 else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
4713 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
4714 check_blocks_extents(ctx, pctx, &pb);
4718 * If we've modified the inode, write it out before
4719 * iterate() tries to use it.
4722 e2fsck_write_inode(ctx, ino, inode,
4727 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4728 pctx->errcode = ext2fs_block_iterate3(fs, ino,
4729 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
4730 block_buf, process_block, &pb);
4732 * We do not have uninitialized extents in non extent
4735 pb.last_init_lblock = pb.last_block;
4737 * If iterate() changed a block mapping, we have to
4738 * re-read the inode. If we decide to clear the
4739 * inode after clearing some stuff, we'll re-write the
4740 * bad mappings into the inode!
4742 if (pb.inode_modified)
4743 e2fsck_read_inode(ctx, ino, inode,
4745 fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4746 (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4748 if (ctx->options & E2F_OPT_CONVERT_BMAP) {
4750 printf("bmap rebuild ino=%d\n", ino);
4752 if (!LINUX_S_ISDIR(inode->i_mode) ||
4753 !e2fsck_dir_will_be_rehashed(ctx, ino))
4754 e2fsck_rebuild_extents_later(ctx, ino);
4758 end_problem_latch(ctx, PR_LATCH_BLOCK);
4759 end_problem_latch(ctx, PR_LATCH_TOOBIG);
4760 if (e2fsck_should_abort(ctx))
4763 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
4765 if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
4766 if (LINUX_S_ISDIR(inode->i_mode))
4767 ctx->fs_fragmented_dir++;
4769 ctx->fs_fragmented++;
4773 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4778 if (inode->i_flags & EXT2_INDEX_FL) {
4779 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
4780 inode->i_flags &= ~EXT2_INDEX_FL;
4783 e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
4787 if (!pb.num_blocks && pb.is_dir &&
4788 !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
4789 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
4790 e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
4791 ctx->fs_directory_count--;
4796 if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
4797 (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
4798 !(inode->i_flags & EXT4_EA_INODE_FL)) {
4799 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
4801 pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
4802 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
4803 ino, (ea_ibody_quota ?
4804 ea_ibody_quota->inodes : 0) +
4805 ea_block_quota.inodes + 1);
4808 if (!ext2fs_has_feature_huge_file(fs->super) ||
4809 !(inode->i_flags & EXT4_HUGE_FILE_FL))
4810 pb.num_blocks *= (fs->blocksize / 512);
4811 pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
4813 printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
4814 ino, inode->i_size, (unsigned long long) pb.last_block,
4815 (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
4816 (unsigned long long) pb.num_blocks);
4818 size = EXT2_I_SIZE(inode);
4820 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
4821 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
4826 flags = ctx->fs->flags;
4827 ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4828 err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
4830 ctx->fs->flags = (flags &
4831 EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4833 ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4834 if (err || sz != size) {
4838 } else if (size & (fs->blocksize - 1))
4840 else if (nblock > (pb.last_block + 1))
4842 else if (nblock < (pb.last_block + 1)) {
4843 if (((pb.last_block + 1) - nblock) >
4844 fs->super->s_prealloc_dir_blocks)
4848 if ((pb.last_init_lblock >= 0) &&
4849 /* Do not allow initialized allocated blocks past i_size*/
4850 (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
4851 !(inode->i_flags & EXT4_VERITY_FL))
4853 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4854 size > ext2_max_sizes[fs->super->s_log_block_size])
4855 /* too big for a direct/indirect-mapped file */
4857 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4859 ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
4860 /* too big for an extent-based file - 32bit ee_block */
4863 /* i_size for symlinks is checked elsewhere */
4864 if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
4865 /* Did inline_data set pctx->num earlier? */
4867 pctx->num = (pb.last_block + 1) * fs->blocksize;
4868 pctx->group = bad_size;
4869 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
4870 ext2fs_inode_size_set(fs, inode, pctx->num);
4871 if (EXT2_I_SIZE(inode) == 0 &&
4872 (inode->i_flags & EXT4_INLINE_DATA_FL)) {
4873 memset(inode->i_block, 0,
4874 sizeof(inode->i_block));
4875 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
4881 if (LINUX_S_ISREG(inode->i_mode) &&
4882 ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4884 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
4885 ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
4886 (ext2fs_has_feature_huge_file(fs->super) &&
4887 (inode->i_flags & EXT4_HUGE_FILE_FL) &&
4888 (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
4889 pctx->num = pb.num_blocks;
4890 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
4891 inode->i_blocks = pb.num_blocks;
4892 inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
4899 * The kernel gets mad if we ask it to allocate bigalloc clusters to
4900 * a block mapped file, so rebuild it as an extent file. We can skip
4901 * symlinks because they're never rewritten.
4903 if (ext2fs_has_feature_bigalloc(fs->super) &&
4904 (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
4905 ext2fs_inode_data_blocks2(fs, inode) > 0 &&
4906 (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
4907 !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
4908 fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
4909 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
4914 if (ctx->dirs_to_hash && pb.is_dir &&
4915 !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
4916 !(inode->i_flags & EXT2_INDEX_FL) &&
4917 ((inode->i_size / fs->blocksize) >= 3))
4918 e2fsck_rehash_dir_later(ctx, ino);
4922 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
4927 * Helper function called by process block when an illegal block is
4928 * found. It returns a description about why the block is illegal
4930 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
4934 static char problem[80];
4936 super = fs->super->s_first_data_block;
4937 strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
4938 if (block < super) {
4939 sprintf(problem, "< FIRSTBLOCK (%u)", super);
4941 } else if (block >= ext2fs_blocks_count(fs->super)) {
4942 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
4945 for (i = 0; i < fs->group_desc_count; i++) {
4946 if (block == super) {
4947 sprintf(problem, "is the superblock in group %d", i);
4950 if (block > super &&
4951 block <= (super + fs->desc_blocks)) {
4952 sprintf(problem, "is in the group descriptors "
4956 if (block == ext2fs_block_bitmap_loc(fs, i)) {
4957 sprintf(problem, "is the block bitmap of group %d", i);
4960 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
4961 sprintf(problem, "is the inode bitmap of group %d", i);
4964 if (block >= ext2fs_inode_table_loc(fs, i) &&
4965 (block < ext2fs_inode_table_loc(fs, i)
4966 + fs->inode_blocks_per_group)) {
4967 sprintf(problem, "is in the inode table of group %d",
4971 super += fs->super->s_blocks_per_group;
4978 * This is a helper function for check_blocks().
4980 static int process_block(ext2_filsys fs,
4982 e2_blkcnt_t blockcnt,
4983 blk64_t ref_block EXT2FS_ATTR((unused)),
4984 int ref_offset EXT2FS_ATTR((unused)),
4987 struct process_block_struct *p;
4988 struct problem_context *pctx;
4989 blk64_t blk = *block_nr;
4991 problem_t problem = 0;
4994 p = (struct process_block_struct *) priv_data;
4999 * For a directory, add logical block zero for processing even if it's
5000 * not mapped or we'll be perennially stuck with broken "." and ".."
5003 if (p->is_dir && blockcnt == 0 && blk == 0) {
5004 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
5005 if (pctx->errcode) {
5007 pctx->num = blockcnt;
5008 goto failed_add_dir_block;
5017 printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
5022 * Simplistic fragmentation check. We merely require that the
5023 * file be contiguous. (Which can never be true for really
5024 * big files that are greater than a block group.)
5026 if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
5027 if (p->previous_block+1 != blk) {
5028 if (ctx->options & E2F_OPT_FRAGCHECK) {
5036 printf(_("%6lu(%c): expecting %6lu "
5037 "got phys %6lu (blkcnt %lld)\n"),
5038 (unsigned long) pctx->ino, type,
5039 (unsigned long) p->previous_block+1,
5040 (unsigned long) blk,
5041 (long long) blockcnt);
5047 if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
5048 !pctx->inode->i_size_high &&
5049 blockcnt > (1 << (21 - fs->super->s_log_block_size)))
5050 problem = PR_1_TOOBIG_DIR;
5051 if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
5052 problem = PR_1_TOOBIG_DIR;
5053 if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
5054 problem = PR_1_TOOBIG_REG;
5055 if (!p->is_dir && !p->is_reg && blockcnt > 0)
5056 problem = PR_1_TOOBIG_SYMLINK;
5058 if (blk < fs->super->s_first_data_block ||
5059 blk >= ext2fs_blocks_count(fs->super))
5060 problem = PR_1_ILLEGAL_BLOCK_NUM;
5063 * If this IND/DIND/TIND block is squatting atop some critical metadata
5064 * (group descriptors, superblock, bitmap, inode table), any write to
5065 * "fix" mapping problems will destroy the metadata. We'll let pass 1b
5066 * fix that and restart fsck.
5069 p->ino != EXT2_RESIZE_INO &&
5070 blk < ctx->fs->super->s_blocks_count &&
5071 ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
5073 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
5074 if ((ctx->options & E2F_OPT_NO) == 0)
5075 ctx->flags |= E2F_FLAG_RESTART_LATER;
5079 p->num_illegal_blocks++;
5081 * A bit of subterfuge here -- we're trying to fix a block
5082 * mapping, but the IND/DIND/TIND block could have collided
5083 * with some critical metadata. So, fix the in-core mapping so
5084 * iterate won't go insane, but return 0 instead of
5085 * BLOCK_CHANGED so that it won't write the remapping out to
5086 * our multiply linked block.
5088 * Even if we previously determined that an *IND block
5089 * conflicts with critical metadata, we must still try to
5090 * iterate the *IND block as if it is an *IND block to find and
5091 * mark the blocks it points to. Better to be overly cautious
5092 * with the used_blocks map so that we don't move the *IND
5093 * block to a block that's really in use!
5095 if (p->ino != EXT2_RESIZE_INO &&
5097 ext2fs_test_block_bitmap2(ctx->block_metadata_map,
5102 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
5103 if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
5107 if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
5109 set_latch_flags(PR_LATCH_BLOCK,
5114 pctx->blkcount = blockcnt;
5115 if (fix_problem(ctx, problem, pctx)) {
5116 blk = *block_nr = 0;
5117 ret_code = BLOCK_CHANGED;
5118 p->inode_modified = 1;
5120 * If the directory block is too big and is beyond the
5121 * end of the FS, don't bother trying to add it for
5122 * processing -- the kernel would never have created a
5123 * directory this large, and we risk an ENOMEM abort.
5124 * In any case, the toobig handler for extent-based
5125 * directories also doesn't feed toobig blocks to
5128 if (problem == PR_1_TOOBIG_DIR)
5135 if (p->ino == EXT2_RESIZE_INO) {
5137 * The resize inode has already be sanity checked
5138 * during pass #0 (the superblock checks). All we
5139 * have to do is mark the double indirect block as
5140 * being in use; all of the other blocks are handled
5141 * by mark_table_blocks()).
5143 if (blockcnt == BLOCK_COUNT_DIND)
5144 mark_block_used(ctx, blk);
5146 } else if (!(ctx->fs->cluster_ratio_bits &&
5147 p->previous_block &&
5148 (EXT2FS_B2C(ctx->fs, blk) ==
5149 EXT2FS_B2C(ctx->fs, p->previous_block)) &&
5150 (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
5151 ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
5152 mark_block_used(ctx, blk);
5154 } else if (has_unaligned_cluster_map(ctx, p->previous_block,
5155 p->last_block, blk, blockcnt)) {
5156 pctx->blk = blockcnt;
5158 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
5159 mark_block_used(ctx, blk);
5160 mark_block_used(ctx, blk);
5163 p->last_block = blockcnt;
5164 p->previous_block = blk;
5166 if (p->is_dir && (blockcnt >= 0)) {
5167 while (++p->last_db_block < blockcnt) {
5168 pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
5171 if (pctx->errcode) {
5173 pctx->num = p->last_db_block;
5174 goto failed_add_dir_block;
5177 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
5179 if (pctx->errcode) {
5181 pctx->num = blockcnt;
5182 failed_add_dir_block:
5183 fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
5184 /* Should never get here */
5185 ctx->flags |= E2F_FLAG_ABORT;
5192 static int process_bad_block(ext2_filsys fs,
5194 e2_blkcnt_t blockcnt,
5195 blk64_t ref_block EXT2FS_ATTR((unused)),
5196 int ref_offset EXT2FS_ATTR((unused)),
5199 struct process_block_struct *p;
5200 blk64_t blk = *block_nr;
5201 blk64_t first_block;
5203 struct problem_context *pctx;
5209 p = (struct process_block_struct *) priv_data;
5213 pctx->ino = EXT2_BAD_INO;
5215 pctx->blkcount = blockcnt;
5217 if ((blk < fs->super->s_first_data_block) ||
5218 (blk >= ext2fs_blocks_count(fs->super))) {
5219 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
5221 return BLOCK_CHANGED;
5227 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
5229 if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
5231 return BLOCK_CHANGED;
5233 } else if (is_blocks_used(ctx, blk, 1)) {
5235 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
5238 return BLOCK_CHANGED;
5240 if (e2fsck_should_abort(ctx))
5243 mark_block_used(ctx, blk);
5248 printf ("DEBUG: Marking %u as bad.\n", blk);
5250 ctx->fs_badblocks_count++;
5252 * If the block is not used, then mark it as used and return.
5253 * If it is already marked as found, this must mean that
5254 * there's an overlap between the filesystem table blocks
5255 * (bitmaps and inode table) and the bad block list.
5257 if (!is_blocks_used(ctx, blk, 1)) {
5258 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
5262 * Try to find the where the filesystem block was used...
5264 first_block = fs->super->s_first_data_block;
5266 for (i = 0; i < fs->group_desc_count; i++ ) {
5269 if (!ext2fs_bg_has_super(fs, i))
5271 if (blk == first_block) {
5273 if (fix_problem(ctx,
5274 PR_1_BAD_PRIMARY_SUPERBLOCK,
5277 return BLOCK_CHANGED;
5281 fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
5284 if ((blk > first_block) &&
5285 (blk <= first_block + fs->desc_blocks)) {
5287 pctx->blk = *block_nr;
5288 if (fix_problem(ctx,
5289 PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
5291 return BLOCK_CHANGED;
5295 fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
5299 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
5300 if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
5301 ctx->invalid_block_bitmap_flag[i]++;
5302 ctx->invalid_bitmaps++;
5306 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
5307 if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
5308 ctx->invalid_inode_bitmap_flag[i]++;
5309 ctx->invalid_bitmaps++;
5313 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
5314 (blk < (ext2fs_inode_table_loc(fs, i) +
5315 fs->inode_blocks_per_group))) {
5317 * If there are bad blocks in the inode table,
5318 * the inode scan code will try to do
5319 * something reasonable automatically.
5323 first_block += fs->super->s_blocks_per_group;
5326 * If we've gotten to this point, then the only
5327 * possibility is that the bad block inode meta data
5328 * is using a bad block.
5330 if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
5331 (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
5332 (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
5334 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
5336 return BLOCK_CHANGED;
5338 if (e2fsck_should_abort(ctx))
5345 /* Warn user that the block wasn't claimed */
5346 fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
5351 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
5352 const char *name, int num, blk64_t *new_block)
5354 ext2_filsys fs = ctx->fs;
5356 blk64_t old_block = *new_block;
5359 unsigned flexbg_size;
5362 struct problem_context pctx;
5364 clear_problem_context(&pctx);
5367 pctx.blk = old_block;
5371 * For flex_bg filesystems, first try to allocate the metadata
5372 * within the flex_bg, and if that fails then try finding the
5373 * space anywhere in the filesystem.
5375 is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
5377 flexbg_size = 1U << fs->super->s_log_groups_per_flex;
5378 flexbg = group / flexbg_size;
5379 first_block = ext2fs_group_first_block2(fs,
5380 flexbg_size * flexbg);
5381 last_grp = group | (flexbg_size - 1);
5382 if (last_grp >= fs->group_desc_count)
5383 last_grp = fs->group_desc_count - 1;
5384 last_block = ext2fs_group_last_block2(fs, last_grp);
5386 last_block = ext2fs_group_last_block2(fs, group);
5387 pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
5388 num, ctx->block_found_map,
5390 if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
5391 pctx.errcode = ext2fs_get_free_blocks2(fs,
5392 fs->super->s_first_data_block,
5393 ext2fs_blocks_count(fs->super),
5394 num, ctx->block_found_map, new_block);
5397 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
5398 ext2fs_unmark_valid(fs);
5399 ctx->flags |= E2F_FLAG_ABORT;
5402 pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
5404 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
5405 ext2fs_unmark_valid(fs);
5406 ctx->flags |= E2F_FLAG_ABORT;
5409 ext2fs_mark_super_dirty(fs);
5410 fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
5411 pctx.blk2 = *new_block;
5412 fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
5413 PR_1_RELOC_TO), &pctx);
5415 for (i = 0; i < num; i++) {
5417 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
5419 pctx.errcode = io_channel_read_blk64(fs->io,
5420 old_block + i, 1, buf);
5422 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
5423 pctx.blk = (*new_block) + i;
5424 pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
5427 pctx.blk = (*new_block) + i;
5428 pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
5433 fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
5435 ext2fs_free_mem(&buf);
5439 * This routine gets called at the end of pass 1 if bad blocks are
5440 * detected in the superblock, group descriptors, inode_bitmaps, or
5441 * block bitmaps. At this point, all of the blocks have been mapped
5442 * out, so we can try to allocate new block(s) to replace the bad
5445 static void handle_fs_bad_blocks(e2fsck_t ctx)
5447 ext2_filsys fs = ctx->fs;
5449 blk64_t first_block;
5452 for (i = 0; i < fs->group_desc_count; i++) {
5453 first_block = ext2fs_group_first_block2(fs, i);
5455 if (ctx->invalid_block_bitmap_flag[i]) {
5456 new_blk = ext2fs_block_bitmap_loc(fs, i);
5457 new_table_block(ctx, first_block, i, _("block bitmap"),
5459 ext2fs_block_bitmap_loc_set(fs, i, new_blk);
5461 if (ctx->invalid_inode_bitmap_flag[i]) {
5462 new_blk = ext2fs_inode_bitmap_loc(fs, i);
5463 new_table_block(ctx, first_block, i, _("inode bitmap"),
5465 ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
5467 if (ctx->invalid_inode_table_flag[i]) {
5468 new_blk = ext2fs_inode_table_loc(fs, i);
5469 new_table_block(ctx, first_block, i, _("inode table"),
5470 fs->inode_blocks_per_group,
5472 ext2fs_inode_table_loc_set(fs, i, new_blk);
5473 ctx->flags |= E2F_FLAG_RESTART;
5476 ctx->invalid_bitmaps = 0;
5480 * This routine marks all blocks which are used by the superblock,
5481 * group descriptors, inode bitmaps, and block bitmaps.
5483 static void mark_table_blocks(e2fsck_t ctx)
5485 ext2_filsys fs = ctx->fs;
5489 struct problem_context pctx;
5491 clear_problem_context(&pctx);
5493 for (i = 0; i < fs->group_desc_count; i++) {
5496 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
5497 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
5500 * Mark the blocks used for the inode table
5502 if (ext2fs_inode_table_loc(fs, i)) {
5503 for (j = 0, b = ext2fs_inode_table_loc(fs, i);
5504 j < fs->inode_blocks_per_group;
5506 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5509 if (!ctx->invalid_inode_table_flag[i] &&
5511 PR_1_ITABLE_CONFLICT, &pctx)) {
5512 ctx->invalid_inode_table_flag[i]++;
5513 ctx->invalid_bitmaps++;
5516 ext2fs_mark_block_bitmap2(
5517 ctx->block_found_map, b);
5518 ext2fs_mark_block_bitmap2(
5519 ctx->block_metadata_map, b);
5525 * Mark block used for the block bitmap
5527 if (ext2fs_block_bitmap_loc(fs, i)) {
5528 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5529 ext2fs_block_bitmap_loc(fs, i))) {
5530 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
5531 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
5532 ctx->invalid_block_bitmap_flag[i]++;
5533 ctx->invalid_bitmaps++;
5536 ext2fs_mark_block_bitmap2(ctx->block_found_map,
5537 ext2fs_block_bitmap_loc(fs, i));
5538 ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5539 ext2fs_block_bitmap_loc(fs, i));
5543 * Mark block used for the inode bitmap
5545 if (ext2fs_inode_bitmap_loc(fs, i)) {
5546 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5547 ext2fs_inode_bitmap_loc(fs, i))) {
5548 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
5549 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
5550 ctx->invalid_inode_bitmap_flag[i]++;
5551 ctx->invalid_bitmaps++;
5554 ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5555 ext2fs_inode_bitmap_loc(fs, i));
5556 ext2fs_mark_block_bitmap2(ctx->block_found_map,
5557 ext2fs_inode_bitmap_loc(fs, i));
5564 * These subroutines short circuits ext2fs_get_blocks and
5565 * ext2fs_check_directory; we use them since we already have the inode
5566 * structure, so there's no point in letting the ext2fs library read
5569 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
5572 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5575 if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5576 return EXT2_ET_CALLBACK_NOTHANDLED;
5578 for (i=0; i < EXT2_N_BLOCKS; i++)
5579 blocks[i] = ctx->stashed_inode->i_block[i];
5583 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
5584 struct ext2_inode *inode)
5586 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5588 if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5589 return EXT2_ET_CALLBACK_NOTHANDLED;
5590 *inode = *ctx->stashed_inode;
5594 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
5595 struct ext2_inode *inode)
5597 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5599 if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
5600 (inode != ctx->stashed_inode))
5601 *ctx->stashed_inode = *inode;
5602 return EXT2_ET_CALLBACK_NOTHANDLED;
5605 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
5607 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5609 if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5610 return EXT2_ET_CALLBACK_NOTHANDLED;
5612 if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
5613 return EXT2_ET_NO_DIRECTORY;
5617 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
5620 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5624 if (ctx->block_found_map) {
5625 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
5629 if (fs->block_map) {
5630 ext2fs_mark_block_bitmap2(fs->block_map, new_block);
5631 ext2fs_mark_bb_dirty(fs);
5634 if (!fs->block_map) {
5635 retval = ext2fs_read_block_bitmap(fs);
5640 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
5649 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
5650 blk64_t len, blk64_t *pblk, blk64_t *plen)
5652 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5655 if (ctx->block_found_map)
5656 return ext2fs_new_range(fs, flags, goal, len,
5657 ctx->block_found_map, pblk, plen);
5659 if (!fs->block_map) {
5660 retval = ext2fs_read_block_bitmap(fs);
5665 return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
5669 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
5671 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5673 /* Never free a critical metadata block */
5674 if (ctx->block_found_map &&
5675 ctx->block_metadata_map &&
5677 ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
5680 if (ctx->block_found_map) {
5682 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
5684 ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
5688 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
5689 blk_t num, int inuse)
5691 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5693 /* Never free a critical metadata block */
5694 if (ctx->block_found_map &&
5695 ctx->block_metadata_map &&
5697 ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
5700 if (ctx->block_found_map) {
5702 ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
5705 ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
5710 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
5712 ext2_filsys fs = ctx->fs;
5714 if (use_shortcuts) {
5715 fs->get_blocks = pass1_get_blocks;
5716 fs->check_directory = pass1_check_directory;
5717 fs->read_inode = pass1_read_inode;
5718 fs->write_inode = pass1_write_inode;
5719 ctx->stashed_ino = 0;
5722 fs->check_directory = 0;
5724 fs->write_inode = 0;
5728 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
5730 ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
5731 ext2fs_set_block_alloc_stats_callback(ctx->fs,
5732 e2fsck_block_alloc_stats, 0);
5733 ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
5734 ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
5735 e2fsck_block_alloc_stats_range, NULL);