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;
402 (void *)entry < end && !EXT2_EXT_IS_LAST_ENTRY(entry);
403 entry = EXT2_EXT_ATTR_NEXT(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,
421 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
422 struct ea_quota *ea_ibody_quota)
424 struct ext2_super_block *sb = ctx->fs->super;
425 struct ext2_inode_large *inode;
426 struct ext2_ext_attr_entry *entry;
427 char *start, *header, *end;
428 unsigned int storage_size, remain;
429 problem_t problem = 0;
432 ea_ibody_quota->blocks = 0;
433 ea_ibody_quota->inodes = 0;
435 inode = (struct ext2_inode_large *) pctx->inode;
436 storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
437 inode->i_extra_isize;
438 header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
439 inode->i_extra_isize;
440 end = header + storage_size;
441 start = header + sizeof(__u32);
442 entry = (struct ext2_ext_attr_entry *) start;
444 /* scan all entry's headers first */
446 /* take finish entry 0UL into account */
447 remain = storage_size - sizeof(__u32);
449 region = region_create(0, storage_size);
451 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
453 ctx->flags |= E2F_FLAG_ABORT;
456 if (region_allocate(region, 0, sizeof(__u32))) {
457 problem = PR_1_INODE_EA_ALLOC_COLLISION;
461 while (remain >= sizeof(struct ext2_ext_attr_entry) &&
462 !EXT2_EXT_IS_LAST_ENTRY(entry)) {
465 if (region_allocate(region, (char *)entry - (char *)header,
466 EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
467 problem = PR_1_INODE_EA_ALLOC_COLLISION;
471 /* header eats this space */
472 remain -= sizeof(struct ext2_ext_attr_entry);
474 /* is attribute name valid? */
475 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
476 pctx->num = entry->e_name_len;
477 problem = PR_1_ATTR_NAME_LEN;
481 /* attribute len eats this space */
482 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
484 if (entry->e_value_inum == 0) {
485 /* check value size */
486 if (entry->e_value_size > remain) {
487 pctx->num = entry->e_value_size;
488 problem = PR_1_ATTR_VALUE_SIZE;
492 if (entry->e_value_size &&
493 region_allocate(region,
494 sizeof(__u32) + entry->e_value_offs,
496 entry->e_value_size))) {
497 problem = PR_1_INODE_EA_ALLOC_COLLISION;
501 hash = ext2fs_ext_attr_hash_entry(entry,
502 start + entry->e_value_offs);
504 /* e_hash may be 0 in older inode's ea */
505 if (entry->e_hash != 0 && entry->e_hash != hash) {
506 pctx->num = entry->e_hash;
507 problem = PR_1_ATTR_HASH;
511 blk64_t quota_blocks;
513 problem = check_large_ea_inode(ctx, entry, pctx,
518 ea_ibody_quota->blocks += quota_blocks;
519 ea_ibody_quota->inodes++;
522 /* If EA value is stored in external inode then it does not
523 * consume space here */
524 if (entry->e_value_inum == 0)
525 remain -= entry->e_value_size;
527 entry = EXT2_EXT_ATTR_NEXT(entry);
530 if (region_allocate(region, (char *)entry - (char *)header,
532 problem = PR_1_INODE_EA_ALLOC_COLLISION;
539 * it seems like a corruption. it's very unlikely we could repair
540 * EA(s) in automatic fashion -bzzz
542 if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
543 inc_ea_inode_refs(ctx, pctx,
544 (struct ext2_ext_attr_entry *)start, end);
548 /* simply remove all possible EA(s) */
549 *((__u32 *)header) = 0UL;
550 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
551 EXT2_INODE_SIZE(sb), "pass1");
552 ea_ibody_quota->blocks = 0;
553 ea_ibody_quota->inodes = 0;
556 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
557 return (xtime & (1U << 31)) != 0 &&
558 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
561 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
562 check_inode_extra_negative_epoch(inode->i_##xtime, \
563 inode->i_##xtime##_extra)
565 /* When today's date is earlier than 2242, we assume that atimes,
566 * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
567 * actually pre-1970 dates mis-encoded.
569 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
571 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
572 struct ea_quota *ea_ibody_quota)
574 struct ext2_super_block *sb = ctx->fs->super;
575 struct ext2_inode_large *inode;
579 ea_ibody_quota->blocks = 0;
580 ea_ibody_quota->inodes = 0;
582 inode = (struct ext2_inode_large *) pctx->inode;
583 if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
584 /* this isn't large inode. so, nothing to check */
589 printf("inode #%u, i_extra_size %d\n", pctx->ino,
590 inode->i_extra_isize);
592 /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
593 min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
594 max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
596 * For now we will allow i_extra_isize to be 0, but really
597 * implementations should never allow i_extra_isize to be 0
599 if (inode->i_extra_isize &&
600 (inode->i_extra_isize < min || inode->i_extra_isize > max ||
601 inode->i_extra_isize & 3)) {
602 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
604 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
605 inode->i_extra_isize = sb->s_want_extra_isize;
607 inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
608 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
609 EXT2_INODE_SIZE(sb), "pass1");
612 /* check if there is no place for an EA header */
613 if (inode->i_extra_isize >= max - sizeof(__u32))
616 eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
617 inode->i_extra_isize);
618 if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
619 /* it seems inode has an extended attribute(s) in body */
620 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
624 * If the inode's extended atime (ctime, crtime, mtime) is stored in
625 * the old, invalid format, repair it.
627 if (((sizeof(time_t) <= 4) ||
628 (((sizeof(time_t) > 4) &&
629 ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
630 (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
631 CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
632 CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
633 CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
635 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
638 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
639 inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
640 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
641 inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
642 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
643 inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
644 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
645 inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
646 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
647 EXT2_INODE_SIZE(sb), "pass1");
652 static _INLINE_ int is_blocks_used(e2fsck_t ctx, blk64_t block,
657 /* used to avoid duplicate output from below */
658 retval = ext2fs_test_block_bitmap_range2_valid(ctx->block_found_map,
663 retval = ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num);
665 e2fsck_pass1_block_map_r_lock(ctx);
667 retval = ext2fs_test_block_bitmap_range2(
668 ctx->global_ctx->block_found_map, block, num);
669 e2fsck_pass1_block_map_r_unlock(ctx);
678 * Check to see if the inode might really be a directory, despite i_mode
680 * This is a lot of complexity for something for which I'm not really
681 * convinced happens frequently in the wild. If for any reason this
682 * causes any problems, take this code out.
683 * [tytso:20070331.0827EDT]
685 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
688 struct ext2_inode *inode = pctx->inode;
689 struct ext2_dir_entry *dirent;
692 unsigned int i, rec_len, not_device = 0;
697 * If the mode looks OK, we believe it. If the first block in
698 * the i_block array is 0, this cannot be a directory. If the
699 * inode is extent-mapped, it is still the case that the latter
700 * cannot be 0 - the magic number in the extent header would make
703 if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
704 LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
708 * Check the block numbers in the i_block array for validity:
709 * zero blocks are skipped (but the first one cannot be zero -
710 * see above), other blocks are checked against the first and
711 * max data blocks (from the the superblock) and against the
712 * block bitmap. Any invalid block found means this cannot be
715 * If there are non-zero blocks past the fourth entry, then
716 * this cannot be a device file: we remember that for the next
719 * For extent mapped files, we don't do any sanity checking:
720 * just try to get the phys block of logical block 0 and run
723 * For inline data files, we just try to get the size of inline
724 * data. If it's true, we will treat it as a directory.
727 extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
728 inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
729 if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
732 unsigned int rec_len2;
733 struct ext2_dir_entry de;
735 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
738 * If the size isn't a multiple of 4, it's probably not a
744 * If the first 10 bytes don't look like a directory entry,
745 * it's probably not a directory.
747 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
748 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
749 EXT2_DIR_REC_LEN(0));
750 dotdot = ext2fs_le32_to_cpu(dotdot);
751 de.inode = ext2fs_le32_to_cpu(de.inode);
752 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
753 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
754 if (dotdot >= ctx->fs->super->s_inodes_count ||
755 (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
756 dotdot != EXT2_ROOT_INO) ||
757 de.inode >= ctx->fs->super->s_inodes_count ||
758 (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
760 rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
761 EXT4_INLINE_DATA_DOTDOT_SIZE)
763 /* device files never have a "system.data" entry */
765 } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
767 if (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
770 /* device files are never extent mapped */
773 for (i=0; i < EXT2_N_BLOCKS; i++) {
774 blk = inode->i_block[i];
780 if (blk < ctx->fs->super->s_first_data_block ||
781 blk >= ext2fs_blocks_count(ctx->fs->super) ||
782 is_blocks_used(ctx, blk, 1))
783 return; /* Invalid block, can't be dir */
785 blk = inode->i_block[0];
789 * If the mode says this is a device file and the i_links_count field
790 * is sane and we have not ruled it out as a device file previously,
791 * we declare it a device file, not a directory.
793 if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
794 (inode->i_links_count == 1) && !not_device)
797 /* read the first block */
798 ehandler_operation(_("reading directory block"));
799 retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
800 ehandler_operation(0);
804 dirent = (struct ext2_dir_entry *) buf;
805 retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
808 if ((ext2fs_dirent_name_len(dirent) != 1) ||
809 (dirent->name[0] != '.') ||
810 (dirent->inode != pctx->ino) ||
813 (rec_len >= ctx->fs->blocksize - 12))
816 dirent = (struct ext2_dir_entry *) (buf + rec_len);
817 retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
820 if ((ext2fs_dirent_name_len(dirent) != 2) ||
821 (dirent->name[0] != '.') ||
822 (dirent->name[1] != '.') ||
828 if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
829 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
830 e2fsck_write_inode_full(ctx, pctx->ino, inode,
831 EXT2_INODE_SIZE(ctx->fs->super),
832 "check_is_really_dir");
836 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
837 int flags, ext2_icount_t hint,
840 unsigned int threshold;
841 unsigned int save_type;
849 profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
851 profile_get_uint(ctx->profile, "scratch_files",
852 "numdirs_threshold", 0, 0, &threshold);
853 profile_get_boolean(ctx->profile, "scratch_files",
854 "icount", 0, 1, &enable);
856 retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
858 num_dirs = 1024; /* Guess */
860 if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
861 (!threshold || num_dirs > threshold)) {
862 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
867 e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
869 if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
870 flags |= EXT2_ICOUNT_OPT_FULLMAP;
871 retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
872 ctx->fs->default_bitmap_type = save_type;
876 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
878 struct problem_context *pctx)
881 struct ext2_inode_large inode;
884 * Reread inode. If we don't see checksum error, then this inode
885 * has been fixed elsewhere.
887 ctx->stashed_ino = 0;
888 retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
890 if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
896 * Checksum still doesn't match. That implies that the inode passes
897 * all the sanity checks, so maybe the checksum is simply corrupt.
898 * See if the user will go for fixing that.
900 if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
904 e2fsck_pass1_fix_lock(ctx);
905 retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
907 e2fsck_pass1_fix_unlock(ctx);
911 static void reserve_block_for_root_repair(e2fsck_t ctx)
915 ext2_filsys fs = ctx->fs;
917 ctx->root_repair_block = 0;
918 if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
921 err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
924 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
925 ctx->root_repair_block = blk;
928 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
932 ext2_filsys fs = ctx->fs;
933 static const char name[] = "lost+found";
936 ctx->lnf_repair_block = 0;
937 if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
940 err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
943 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
944 ctx->lnf_repair_block = blk;
948 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
952 struct ext2_xattr_handle *handle;
955 retval = ext2fs_xattrs_open(fs, ino, &handle);
959 retval = ext2fs_xattrs_read(handle);
963 retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
968 (void) ext2fs_xattrs_close(&handle);
972 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
973 struct problem_context *pctx,
980 * If the inode failed the checksum and the user didn't
981 * clear the inode, test the checksum again -- if it still
982 * fails, ask the user if the checksum should be corrected.
984 pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
986 ctx->flags |= E2F_FLAG_ABORT;
988 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
990 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
991 if ((ctx)->flags & E2F_FLAG_ABORT) { \
992 e2fsck_pass1_check_unlock(ctx); \
997 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
1002 for (i = 0; i < EXT2_N_BLOCKS; i++) {
1003 x = inode->i_block[i];
1004 #ifdef WORDS_BIGENDIAN
1005 x = ext2fs_swab32(x);
1007 if (x >= ext2fs_blocks_count(fs->super))
1015 * Figure out what to do with an inode that has both extents and inline data
1016 * inode flags set. Returns -1 if we decide to erase the inode, 0 otherwise.
1018 static int fix_inline_data_extents_file(e2fsck_t ctx,
1020 struct ext2_inode *inode,
1022 struct problem_context *pctx)
1024 size_t max_inline_ea_size;
1025 ext2_filsys fs = ctx->fs;
1028 /* Both feature flags not set? Just run the regular checks */
1029 if (!ext2fs_has_feature_extents(fs->super) &&
1030 !ext2fs_has_feature_inline_data(fs->super))
1033 /* Clear both flags if it's a special file */
1034 if (LINUX_S_ISCHR(inode->i_mode) ||
1035 LINUX_S_ISBLK(inode->i_mode) ||
1036 LINUX_S_ISFIFO(inode->i_mode) ||
1037 LINUX_S_ISSOCK(inode->i_mode)) {
1038 check_extents_inlinedata(ctx, pctx);
1042 /* If it looks like an extent tree, try to clear inlinedata */
1043 if (ext2fs_extent_header_verify(inode->i_block,
1044 sizeof(inode->i_block)) == 0 &&
1045 fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1046 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1051 /* If it looks short enough to be inline data, try to clear extents */
1052 if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1053 max_inline_ea_size = inode_size -
1054 (EXT2_GOOD_OLD_INODE_SIZE +
1055 ((struct ext2_inode_large *)inode)->i_extra_isize);
1057 max_inline_ea_size = 0;
1058 if (EXT2_I_SIZE(inode) <
1059 EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1060 fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1061 inode->i_flags &= ~EXT4_EXTENTS_FL;
1067 * Too big for inline data, but no evidence of extent tree -
1068 * maybe it's a block map file? If the mappings all look valid?
1070 if (could_be_block_map(fs, inode) &&
1071 fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1072 #ifdef WORDS_BIGENDIAN
1075 for (i = 0; i < EXT2_N_BLOCKS; i++)
1076 inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1079 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1084 /* Oh well, just clear the busted inode. */
1085 if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1086 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1092 e2fsck_write_inode(ctx, ino, inode, "pass1");
1097 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1099 ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1100 dgrp_t start = *group, grp, grp_end = ctx->fs->group_desc_count;
1101 blk64_t blocks_to_read = 0;
1102 errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1105 if (ctx->fs->fs_num_threads > 1)
1106 grp_end = ctx->thread_info.et_group_end;
1108 if (ctx->readahead_kb == 0)
1111 /* Keep iterating groups until we have enough to readahead */
1112 inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1113 for (grp = start; grp < grp_end; grp++) {
1114 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1116 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1117 ext2fs_bg_itable_unused(ctx->fs, grp);
1118 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1120 if (blocks_to_read * ctx->fs->blocksize >
1121 ctx->readahead_kb * 1024)
1125 err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1127 if (err == EAGAIN) {
1128 ctx->readahead_kb /= 2;
1134 /* Error; disable itable readahead */
1135 *group = ctx->fs->group_desc_count;
1136 *next_ino = ctx->fs->super->s_inodes_count;
1139 * Don't do more readahead until we've reached the first inode
1140 * of the last inode scan buffer block for the last group.
1143 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1144 ctx->inode_buffer_blocks :
1145 EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1146 ctx->fs->blocksize /
1147 EXT2_INODE_SIZE(ctx->fs->super);
1149 *next_ino = inodes_in_group -
1150 (inodes_in_group % inodes_per_buffer) + 1 +
1151 (grp * ctx->fs->super->s_inodes_per_group);
1156 * Check if the passed ino is one of the used superblock quota inodes.
1158 * Before the quota inodes were journaled, older superblock quota inodes
1159 * were just regular files in the filesystem and not reserved inodes. This
1160 * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1161 * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1163 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1165 enum quota_type qtype;
1167 for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1168 if (*quota_sb_inump(sb, qtype) == ino)
1175 * Check if the passed ino is one of the reserved quota inodes.
1176 * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1177 * inodes. These inodes may or may not be in use by the quota feature.
1179 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1181 enum quota_type qtype;
1183 for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1184 if (quota_type2inum(qtype, fs->super) == ino)
1190 static int e2fsck_should_abort(e2fsck_t ctx)
1192 e2fsck_t global_ctx;
1194 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1197 if (ctx->global_ctx) {
1198 global_ctx = ctx->global_ctx;
1199 if (global_ctx->flags & E2F_FLAG_SIGNAL_MASK)
1205 static void init_ext2_max_sizes()
1211 * Init ext2_max_sizes which will be immutable and shared between
1214 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1216 for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1217 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1218 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1219 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1220 max_sizes = (max_sizes * (1UL << i));
1221 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1227 /* TODO: tdb needs to be handled properly for multiple threads*/
1228 static int multiple_threads_supported(e2fsck_t ctx)
1231 unsigned int threshold;
1232 ext2_ino_t num_dirs;
1237 profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
1239 profile_get_uint(ctx->profile, "scratch_files",
1240 "numdirs_threshold", 0, 0, &threshold);
1241 profile_get_boolean(ctx->profile, "scratch_files",
1242 "icount", 0, 1, &enable);
1244 retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
1246 num_dirs = 1024; /* Guess */
1248 /* tdb is unsupported now */
1249 if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
1250 (!threshold || num_dirs > threshold))
1257 * Even though we could specify number of threads,
1258 * but it might be more than the whole filesystem
1259 * block groups, correct it here.
1261 static void e2fsck_pass1_set_thread_num(e2fsck_t ctx)
1263 unsigned flexbg_size = 1;
1264 ext2_filsys fs = ctx->fs;
1265 int num_threads = ctx->pfs_num_threads;
1268 if (num_threads < 1) {
1273 if (!multiple_threads_supported(ctx)) {
1275 fprintf(stderr, "Fall through single thread for pass1 "
1276 "because tdb could not handle properly\n");
1280 if (ext2fs_has_feature_flex_bg(fs->super))
1281 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
1282 max_threads = fs->group_desc_count / flexbg_size;
1283 if (max_threads == 0)
1285 if (max_threads > E2FSCK_MAX_THREADS)
1286 max_threads = E2FSCK_MAX_THREADS;
1288 if (num_threads > max_threads) {
1289 fprintf(stderr, "Use max possible thread num: %d instead\n",
1291 num_threads = max_threads;
1294 ctx->pfs_num_threads = num_threads;
1295 ctx->fs->fs_num_threads = num_threads;
1300 * We need call mark_table_blocks() before multiple
1301 * thread start, since all known system blocks should be
1302 * marked and checked later.
1304 static errcode_t e2fsck_pass1_prepare(e2fsck_t ctx)
1306 struct problem_context pctx;
1307 ext2_filsys fs = ctx->fs;
1308 unsigned long long readahead_kb;
1310 init_ext2_max_sizes();
1312 e2fsck_pass1_set_thread_num(ctx);
1314 /* If we can do readahead, figure out how many groups to pull in. */
1315 if (!e2fsck_can_readahead(ctx->fs))
1316 ctx->readahead_kb = 0;
1317 else if (ctx->readahead_kb == ~0ULL)
1318 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1321 /* don't use more than 1/10 of memory for threads checking */
1322 readahead_kb = get_memory_size() / (10 * ctx->pfs_num_threads);
1323 /* maybe better disable RA if this is too small? */
1324 if (ctx->readahead_kb > readahead_kb)
1325 ctx->readahead_kb = readahead_kb;
1327 clear_problem_context(&pctx);
1328 if (!(ctx->options & E2F_OPT_PREEN))
1329 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1331 pctx.errcode = e2fsck_allocate_subcluster_bitmap(ctx->fs,
1332 _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1333 "block_found_map", &ctx->block_found_map);
1336 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1337 ctx->flags |= E2F_FLAG_ABORT;
1338 return pctx.errcode;
1340 pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
1341 _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1342 "block_metadata_map", &ctx->block_metadata_map);
1345 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1346 ctx->flags |= E2F_FLAG_ABORT;
1347 return pctx.errcode;
1350 mark_table_blocks(ctx);
1351 pctx.errcode = ext2fs_convert_subcluster_bitmap(ctx->fs,
1352 &ctx->block_found_map);
1354 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1355 ctx->flags |= E2F_FLAG_ABORT;
1356 return pctx.errcode;
1359 pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
1360 _("multiply claimed block map"),
1361 EXT2FS_BMAP64_RBTREE, "block_dup_map",
1362 &ctx->block_dup_map);
1365 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
1367 /* Should never get here */
1368 ctx->flags |= E2F_FLAG_ABORT;
1369 return pctx.errcode;
1372 if (ext2fs_has_feature_mmp(fs->super) &&
1373 fs->super->s_mmp_block > fs->super->s_first_data_block &&
1374 fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1375 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1376 fs->super->s_mmp_block);
1378 pthread_rwlock_init(&ctx->fs_fix_rwlock, NULL);
1379 pthread_rwlock_init(&ctx->fs_block_map_rwlock, NULL);
1380 if (ctx->pfs_num_threads > 1)
1381 ctx->fs_need_locking = 1;
1387 static void e2fsck_pass1_post(e2fsck_t ctx)
1389 struct problem_context pctx;
1390 ext2_filsys fs = ctx->fs;
1393 if (e2fsck_should_abort(ctx))
1396 block_buf = (char *)e2fsck_allocate_memory(ctx, ctx->fs->blocksize * 3,
1397 "block interate buffer");
1398 reserve_block_for_root_repair(ctx);
1399 reserve_block_for_lnf_repair(ctx);
1402 * If any extended attribute blocks' reference counts need to
1403 * be adjusted, either up (ctx->refcount_extra), or down
1404 * (ctx->refcount), then fix them.
1406 if (ctx->refcount) {
1407 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
1408 ea_refcount_free(ctx->refcount);
1411 if (ctx->refcount_extra) {
1412 adjust_extattr_refcount(ctx, ctx->refcount_extra,
1414 ea_refcount_free(ctx->refcount_extra);
1415 ctx->refcount_extra = 0;
1418 if (ctx->invalid_bitmaps)
1419 handle_fs_bad_blocks(ctx);
1421 /* We don't need the block_ea_map any more */
1422 if (ctx->block_ea_map) {
1423 ext2fs_free_block_bitmap(ctx->block_ea_map);
1424 ctx->block_ea_map = 0;
1427 if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
1428 struct ext2_inode *inode;
1429 int inode_size = EXT2_INODE_SIZE(fs->super);
1430 inode = e2fsck_allocate_memory(ctx, inode_size,
1433 clear_problem_context(&pctx);
1434 pctx.errcode = ext2fs_create_resize_inode(fs);
1436 if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
1438 ctx->flags |= E2F_FLAG_ABORT;
1439 ext2fs_free_mem(&inode);
1440 ext2fs_free_mem(&block_buf);
1445 if (!pctx.errcode) {
1446 e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
1448 inode->i_mtime = ctx->now;
1449 e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
1452 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
1453 ext2fs_free_mem(&inode);
1456 if (ctx->flags & E2F_FLAG_RESTART) {
1457 ext2fs_free_mem(&block_buf);
1461 if (ctx->block_dup_map) {
1462 if (!(ctx->flags & E2F_FLAG_DUP_BLOCK)) {
1463 ext2fs_free_mem(&block_buf);
1466 if (ctx->options & E2F_OPT_PREEN) {
1467 clear_problem_context(&pctx);
1468 fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
1470 e2fsck_pass1_dupblocks(ctx, block_buf);
1471 ext2fs_free_mem(&block_buf);
1472 ctx->flags &= ~E2F_FLAG_DUP_BLOCK;
1475 ctx->flags |= E2F_FLAG_ALLOC_OK;
1479 void e2fsck_pass1_run(e2fsck_t ctx)
1482 ext2_filsys fs = ctx->fs;
1484 struct ext2_inode *inode = NULL;
1485 ext2_inode_scan scan = NULL;
1486 char *block_buf = NULL;
1487 #ifdef RESOURCE_TRACK
1488 struct resource_track rtrack;
1490 unsigned char frag, fsize;
1491 struct problem_context pctx;
1492 struct scan_callback_struct scan_struct;
1493 struct ext2_super_block *sb = ctx->fs->super;
1495 const char *eop_next_inode = _("getting next inode from scan");
1496 int imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1497 int low_dtime_check = 1;
1498 unsigned int inode_size = EXT2_INODE_SIZE(fs->super);
1499 unsigned int bufsize;
1500 int failed_csum = 0;
1501 ext2_ino_t ino_threshold = 0;
1502 dgrp_t ra_group = 0;
1503 struct ea_quota ea_ibody_quota;
1504 struct process_inode_block *inodes_to_process;
1505 int process_inode_count, check_mmp;
1506 e2fsck_t global_ctx = ctx->global_ctx ? ctx->global_ctx : ctx;
1508 init_resource_track(&rtrack, ctx->fs->io);
1509 clear_problem_context(&pctx);
1511 pass1_readahead(ctx, &ra_group, &ino_threshold);
1512 if (ext2fs_has_feature_dir_index(fs->super) &&
1513 !(ctx->options & E2F_OPT_NO)) {
1514 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1515 ctx->dirs_to_hash = 0;
1519 mtrace_print("Pass 1");
1522 imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1523 extent_fs = ext2fs_has_feature_extents(sb);
1524 inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1525 casefold_fs = ext2fs_has_feature_casefold(sb);
1528 * Allocate bitmaps structures
1530 pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1531 EXT2FS_BMAP64_RBTREE,
1533 &ctx->inode_used_map);
1536 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1537 ctx->flags |= E2F_FLAG_ABORT;
1540 pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1541 _("directory inode map"),
1542 ctx->global_ctx ? EXT2FS_BMAP64_RBTREE :
1543 EXT2FS_BMAP64_AUTODIR,
1544 "inode_dir_map", &ctx->inode_dir_map);
1547 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1548 ctx->flags |= E2F_FLAG_ABORT;
1551 pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1552 _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1553 "inode_reg_map", &ctx->inode_reg_map);
1556 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1557 ctx->flags |= E2F_FLAG_ABORT;
1562 e2fsck_allocate_inode_bitmap(fs,
1563 _("inode casefold map"),
1564 EXT2FS_BMAP64_RBTREE,
1565 "inode_casefold_map",
1566 &ctx->inode_casefold_map);
1569 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1570 ctx->flags |= E2F_FLAG_ABORT;
1574 pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1575 &ctx->inode_link_info);
1577 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1578 ctx->flags |= E2F_FLAG_ABORT;
1581 bufsize = inode_size;
1582 if (bufsize < sizeof(struct ext2_inode_large))
1583 bufsize = sizeof(struct ext2_inode_large);
1584 inode = (struct ext2_inode *)
1585 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1587 inodes_to_process = (struct process_inode_block *)
1588 e2fsck_allocate_memory(ctx,
1589 (ctx->process_inode_size *
1590 sizeof(struct process_inode_block)),
1591 "array of inodes to process");
1592 process_inode_count = 0;
1594 pctx.errcode = ext2fs_init_dblist(fs, 0);
1596 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1597 ctx->flags |= E2F_FLAG_ABORT;
1602 * If the last orphan field is set, clear it, since the pass1
1603 * processing will automatically find and clear the orphans.
1604 * In the future, we may want to try using the last_orphan
1605 * linked list ourselves, but for now, we clear it so that the
1606 * ext3 mount code won't get confused.
1608 if (!(ctx->options & E2F_OPT_READONLY)) {
1609 if (fs->super->s_last_orphan) {
1610 fs->super->s_last_orphan = 0;
1611 ext2fs_mark_super_dirty(fs);
1615 block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1616 "block interate buffer");
1617 if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1618 e2fsck_use_inode_shortcuts(ctx, 1);
1619 e2fsck_intercept_block_allocations(ctx);
1620 old_op = ehandler_operation(_("opening inode scan"));
1621 pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1623 ehandler_operation(old_op);
1625 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1626 ctx->flags |= E2F_FLAG_ABORT;
1629 ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1630 EXT2_SF_WARN_GARBAGE_INODES, 0);
1631 ctx->stashed_inode = inode;
1632 scan_struct.ctx = ctx;
1633 scan_struct.block_buf = block_buf;
1634 scan_struct.inodes_to_process = inodes_to_process;
1635 scan_struct.process_inode_count = &process_inode_count;
1636 ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1637 if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1638 ctx->fs->group_desc_count)))
1640 if ((fs->super->s_wtime &&
1641 fs->super->s_wtime < fs->super->s_inodes_count) ||
1642 (fs->super->s_mtime &&
1643 fs->super->s_mtime < fs->super->s_inodes_count) ||
1644 (fs->super->s_mkfs_time &&
1645 fs->super->s_mkfs_time < fs->super->s_inodes_count))
1646 low_dtime_check = 0;
1648 /* Set up ctx->lost_and_found if possible */
1649 (void) e2fsck_get_lost_and_found(ctx, 0);
1652 if (ctx->global_ctx) {
1653 if (ctx->options & E2F_OPT_DEBUG &&
1654 ctx->options & E2F_OPT_MULTITHREAD)
1655 fprintf(stderr, "thread %d jumping to group %u\n",
1656 ctx->thread_info.et_thread_index,
1657 ctx->thread_info.et_group_start);
1658 pctx.errcode = ext2fs_inode_scan_goto_blockgroup(scan,
1659 ctx->thread_info.et_group_start);
1661 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1662 ctx->flags |= E2F_FLAG_ABORT;
1670 e2fsck_pass1_check_lock(ctx);
1672 if (!global_ctx->mmp_update_thread) {
1673 e2fsck_pass1_block_map_w_lock(ctx);
1674 if (!global_ctx->mmp_update_thread) {
1675 global_ctx->mmp_update_thread =
1676 ctx->thread_info.et_thread_index + 1;
1679 e2fsck_pass1_block_map_w_unlock(ctx);
1682 /* only one active thread could update mmp block. */
1683 e2fsck_pass1_block_map_r_lock(ctx);
1684 if (global_ctx->mmp_update_thread ==
1685 ctx->thread_info.et_thread_index + 1)
1687 e2fsck_pass1_block_map_r_unlock(ctx);
1692 if (check_mmp && (ino % (fs->super->s_inodes_per_group * 4) == 1)) {
1693 if (e2fsck_mmp_update(fs))
1694 fatal_error(ctx, 0);
1696 old_op = ehandler_operation(eop_next_inode);
1697 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1699 if (ino > ino_threshold)
1700 pass1_readahead(ctx, &ra_group, &ino_threshold);
1701 ehandler_operation(old_op);
1702 if (e2fsck_should_abort(ctx)) {
1703 e2fsck_pass1_check_unlock(ctx);
1706 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1708 * If badblocks says badblocks is bad, offer to clear
1709 * the list, update the in-core bb list, and restart
1712 if (ino == EXT2_BAD_INO &&
1713 fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1717 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1718 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1719 ctx->fs->badblocks = NULL;
1720 err = ext2fs_read_bb_inode(ctx->fs,
1721 &ctx->fs->badblocks);
1723 fix_problem(ctx, PR_1_ISCAN_ERROR,
1725 ctx->flags |= E2F_FLAG_ABORT;
1726 e2fsck_pass1_check_unlock(ctx);
1729 ctx->flags |= E2F_FLAG_RESTART;
1730 err = ext2fs_inode_scan_goto_blockgroup(scan,
1733 fix_problem(ctx, PR_1_ISCAN_ERROR,
1735 ctx->flags |= E2F_FLAG_ABORT;
1736 e2fsck_pass1_check_unlock(ctx);
1739 e2fsck_pass1_check_unlock(ctx);
1742 if (!ctx->inode_bb_map)
1744 ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1745 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1746 e2fsck_pass1_check_unlock(ctx);
1749 if (pctx.errcode == EXT2_ET_SCAN_FINISHED) {
1750 e2fsck_pass1_check_unlock(ctx);
1754 pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1755 pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1756 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1757 ctx->flags |= E2F_FLAG_ABORT;
1758 e2fsck_pass1_check_unlock(ctx);
1762 e2fsck_pass1_check_unlock(ctx);
1766 if (ctx->global_ctx)
1767 ctx->thread_info.et_inode_number++;
1771 ctx->stashed_ino = ino;
1773 /* Clear trashed inode? */
1774 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1775 inode->i_links_count > 0 &&
1776 fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1778 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1780 failed_csum = pctx.errcode != 0;
1783 * Check for inodes who might have been part of the
1784 * orphaned list linked list. They should have gotten
1785 * dealt with by now, unless the list had somehow been
1788 * FIXME: In the future, inodes which are still in use
1789 * (and which are therefore) pending truncation should
1790 * be handled specially. Right now we just clear the
1791 * dtime field, and the normal e2fsck handling of
1792 * inodes where i_size and the inode blocks are
1793 * inconsistent is to fix i_size, instead of releasing
1794 * the extra blocks. This won't catch the inodes that
1795 * was at the end of the orphan list, but it's better
1796 * than nothing. The right answer is that there
1797 * shouldn't be any bugs in the orphan list handling. :-)
1799 if (inode->i_dtime && low_dtime_check &&
1800 inode->i_dtime < ctx->fs->super->s_inodes_count) {
1801 if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1802 inode->i_dtime = inode->i_links_count ?
1804 e2fsck_write_inode(ctx, ino, inode,
1810 if (inode->i_links_count) {
1811 pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1812 ino, inode->i_links_count);
1814 pctx.num = inode->i_links_count;
1815 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1816 ctx->flags |= E2F_FLAG_ABORT;
1817 e2fsck_pass1_check_unlock(ctx);
1820 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1821 !quota_inum_is_reserved(fs, ino)) {
1822 if (!inode->i_dtime && inode->i_mode) {
1823 if (fix_problem(ctx,
1824 PR_1_ZERO_DTIME, &pctx)) {
1825 inode->i_dtime = ctx->now;
1826 e2fsck_write_inode(ctx, ino, inode,
1831 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1832 e2fsck_pass1_check_unlock(ctx);
1836 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1837 ((!LINUX_S_ISDIR(inode->i_mode) &&
1838 fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1840 fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1841 inode->i_flags &= ~EXT4_CASEFOLD_FL;
1842 e2fsck_write_inode(ctx, ino, inode, "pass1");
1845 /* Conflicting inlinedata/extents inode flags? */
1846 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1847 (inode->i_flags & EXT4_EXTENTS_FL)) {
1848 int res = fix_inline_data_extents_file(ctx, ino, inode,
1852 /* skip FINISH_INODE_LOOP */
1853 e2fsck_pass1_check_unlock(ctx);
1858 /* Test for incorrect inline_data flags settings. */
1859 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1860 (ino >= EXT2_FIRST_INODE(fs->super))) {
1863 pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1864 if (!pctx.errcode &&
1865 fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1866 e2fsck_pass1_fix_lock(ctx);
1867 ext2fs_set_feature_inline_data(sb);
1868 ext2fs_mark_super_dirty(fs);
1869 e2fsck_pass1_fix_unlock(ctx);
1871 } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1872 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1873 /* skip FINISH_INODE_LOOP */
1874 e2fsck_pass1_check_unlock(ctx);
1879 /* Test for inline data flag but no attr */
1880 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1881 (ino >= EXT2_FIRST_INODE(fs->super))) {
1888 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1889 err = get_inline_data_ea_size(fs, ino, &size);
1890 fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1891 (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1895 /* Everything is awesome... */
1897 case EXT2_ET_BAD_EA_BLOCK_NUM:
1898 case EXT2_ET_BAD_EA_HASH:
1899 case EXT2_ET_BAD_EA_HEADER:
1900 case EXT2_ET_EA_BAD_NAME_LEN:
1901 case EXT2_ET_EA_BAD_VALUE_SIZE:
1902 case EXT2_ET_EA_KEY_NOT_FOUND:
1903 case EXT2_ET_EA_NO_SPACE:
1904 case EXT2_ET_MISSING_EA_FEATURE:
1905 case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1906 case EXT2_ET_INLINE_DATA_NO_BLOCK:
1907 case EXT2_ET_INLINE_DATA_NO_SPACE:
1908 case EXT2_ET_NO_INLINE_DATA:
1909 case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1910 case EXT2_ET_EA_BAD_VALUE_OFFSET:
1911 case EXT2_ET_EA_INODE_CORRUPTED:
1912 /* broken EA or no system.data EA; truncate */
1913 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1915 err = ext2fs_inode_size_set(fs, inode, 0);
1918 ctx->flags |= E2F_FLAG_ABORT;
1919 e2fsck_pass1_check_unlock(ctx);
1922 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1923 memset(&inode->i_block, 0,
1924 sizeof(inode->i_block));
1925 e2fsck_write_inode(ctx, ino, inode,
1931 /* Some other kind of non-xattr error? */
1933 ctx->flags |= E2F_FLAG_ABORT;
1934 e2fsck_pass1_check_unlock(ctx);
1940 * Test for incorrect extent flag settings.
1942 * On big-endian machines we must be careful:
1943 * When the inode is read, the i_block array is not swapped
1944 * if the extent flag is set. Therefore if we are testing
1945 * for or fixing a wrongly-set flag, we must potentially
1946 * (un)swap before testing, or after fixing.
1950 * In this case the extents flag was set when read, so
1951 * extent_header_verify is ok. If the inode is cleared,
1952 * no need to swap... so no extra swapping here.
1954 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1955 (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1956 (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1957 if ((ext2fs_extent_header_verify(inode->i_block,
1958 sizeof(inode->i_block)) == 0) &&
1959 fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1960 e2fsck_pass1_fix_lock(ctx);
1961 ext2fs_set_feature_extents(sb);
1962 ext2fs_mark_super_dirty(fs);
1964 e2fsck_pass1_fix_unlock(ctx);
1965 } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1967 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1968 if (ino == EXT2_BAD_INO)
1969 ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1971 /* skip FINISH_INODE_LOOP */
1972 e2fsck_pass1_check_unlock(ctx);
1978 * For big-endian machines:
1979 * If the inode didn't have the extents flag set when it
1980 * was read, then the i_blocks array was swapped. To test
1981 * as an extents header, we must swap it back first.
1982 * IF we then set the extents flag, the entire i_block
1983 * array must be un/re-swapped to make it proper extents data.
1985 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1986 (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1987 (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1988 (LINUX_S_ISREG(inode->i_mode) ||
1989 LINUX_S_ISDIR(inode->i_mode))) {
1991 #ifdef WORDS_BIGENDIAN
1992 __u32 tmp_block[EXT2_N_BLOCKS];
1994 for (i = 0; i < EXT2_N_BLOCKS; i++)
1995 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1998 ehp = inode->i_block;
2000 if ((ext2fs_extent_header_verify(ehp,
2001 sizeof(inode->i_block)) == 0) &&
2002 (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
2003 inode->i_flags |= EXT4_EXTENTS_FL;
2004 #ifdef WORDS_BIGENDIAN
2005 memcpy(inode->i_block, tmp_block,
2006 sizeof(inode->i_block));
2008 e2fsck_write_inode(ctx, ino, inode, "pass1");
2013 if (ino == EXT2_BAD_INO) {
2014 struct process_block_struct pb;
2016 if ((failed_csum || inode->i_mode || inode->i_uid ||
2017 inode->i_gid || inode->i_links_count ||
2018 (inode->i_flags & EXT4_INLINE_DATA_FL) ||
2019 inode->i_file_acl) &&
2020 fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
2021 memset(inode, 0, sizeof(struct ext2_inode));
2022 e2fsck_write_inode(ctx, ino, inode,
2027 e2fsck_pass1_block_map_r_lock(ctx);
2028 pctx.errcode = ext2fs_copy_bitmap(ctx->global_ctx ?
2029 ctx->global_ctx->block_found_map :
2030 ctx->block_found_map, &pb.fs_meta_blocks);
2031 e2fsck_pass1_block_map_r_unlock(ctx);
2034 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
2035 ctx->flags |= E2F_FLAG_ABORT;
2036 e2fsck_pass1_check_unlock(ctx);
2039 pb.ino = EXT2_BAD_INO;
2040 pb.num_blocks = pb.last_block = 0;
2041 pb.last_db_block = -1;
2042 pb.num_illegal_blocks = 0;
2043 pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
2044 pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
2048 pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
2049 block_buf, process_bad_block, &pb);
2050 ext2fs_free_block_bitmap(pb.fs_meta_blocks);
2052 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
2053 ctx->flags |= E2F_FLAG_ABORT;
2054 e2fsck_pass1_check_unlock(ctx);
2058 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
2059 ctx->flags |= E2F_FLAG_ABORT;
2060 e2fsck_pass1_check_unlock(ctx);
2063 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2064 clear_problem_context(&pctx);
2065 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2066 e2fsck_pass1_check_unlock(ctx);
2068 } else if (ino == EXT2_ROOT_INO) {
2070 * Make sure the root inode is a directory; if
2071 * not, offer to clear it. It will be
2072 * regenerated in pass #3.
2074 if (!LINUX_S_ISDIR(inode->i_mode)) {
2075 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
2079 * If dtime is set, offer to clear it. mke2fs
2080 * version 0.2b created filesystems with the
2081 * dtime field set for the root and lost+found
2082 * directories. We won't worry about
2083 * /lost+found, since that can be regenerated
2084 * easily. But we will fix the root directory
2085 * as a special case.
2087 if (inode->i_dtime && inode->i_links_count) {
2088 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
2090 e2fsck_write_inode(ctx, ino, inode,
2095 } else if (ino == EXT2_JOURNAL_INO) {
2096 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2097 if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
2098 if (!LINUX_S_ISREG(inode->i_mode) &&
2099 fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
2101 inode->i_mode = LINUX_S_IFREG;
2102 e2fsck_write_inode(ctx, ino, inode,
2106 check_blocks(ctx, &pctx, block_buf, NULL);
2107 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2108 e2fsck_pass1_check_unlock(ctx);
2111 if ((inode->i_links_count ||
2112 inode->i_blocks || inode->i_block[0]) &&
2113 fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
2115 memset(inode, 0, inode_size);
2116 ext2fs_icount_store(ctx->inode_link_info,
2118 e2fsck_write_inode_full(ctx, ino, inode,
2119 inode_size, "pass1");
2122 } else if (quota_inum_is_reserved(fs, ino)) {
2123 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2124 if (ext2fs_has_feature_quota(fs->super) &&
2125 quota_inum_is_super(fs->super, ino)) {
2126 if (!LINUX_S_ISREG(inode->i_mode) &&
2127 fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
2129 inode->i_mode = LINUX_S_IFREG;
2130 e2fsck_write_inode(ctx, ino, inode,
2134 check_blocks(ctx, &pctx, block_buf, NULL);
2135 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2136 e2fsck_pass1_check_unlock(ctx);
2139 if ((inode->i_links_count ||
2140 inode->i_blocks || inode->i_block[0]) &&
2141 fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
2143 memset(inode, 0, inode_size);
2144 ext2fs_icount_store(ctx->inode_link_info,
2146 e2fsck_write_inode_full(ctx, ino, inode,
2147 inode_size, "pass1");
2150 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
2151 problem_t problem = 0;
2153 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2154 if (ino == EXT2_BOOT_LOADER_INO) {
2155 if (LINUX_S_ISDIR(inode->i_mode))
2156 problem = PR_1_RESERVED_BAD_MODE;
2157 } else if (ino == EXT2_RESIZE_INO) {
2158 if (inode->i_mode &&
2159 !LINUX_S_ISREG(inode->i_mode))
2160 problem = PR_1_RESERVED_BAD_MODE;
2162 if (inode->i_mode != 0)
2163 problem = PR_1_RESERVED_BAD_MODE;
2166 if (fix_problem(ctx, problem, &pctx)) {
2168 e2fsck_write_inode(ctx, ino, inode,
2173 check_blocks(ctx, &pctx, block_buf, NULL);
2174 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2175 e2fsck_pass1_check_unlock(ctx);
2179 if (!inode->i_links_count) {
2180 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2181 e2fsck_pass1_check_unlock(ctx);
2185 * n.b. 0.3c ext2fs code didn't clear i_links_count for
2186 * deleted files. Oops.
2188 * Since all new ext2 implementations get this right,
2189 * we now assume that the case of non-zero
2190 * i_links_count and non-zero dtime means that we
2191 * should keep the file, not delete it.
2194 if (inode->i_dtime) {
2195 if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
2197 e2fsck_write_inode(ctx, ino, inode, "pass1");
2202 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2203 switch (fs->super->s_creator_os) {
2205 frag = inode->osd2.hurd2.h_i_frag;
2206 fsize = inode->osd2.hurd2.h_i_fsize;
2212 if (inode->i_faddr || frag || fsize ||
2213 (!ext2fs_has_feature_largedir(fs->super) &&
2214 (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
2215 mark_inode_bad(ctx, ino);
2216 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
2217 !ext2fs_has_feature_64bit(fs->super) &&
2218 inode->osd2.linux2.l_i_file_acl_high != 0)
2219 mark_inode_bad(ctx, ino);
2220 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
2221 !ext2fs_has_feature_huge_file(fs->super) &&
2222 (inode->osd2.linux2.l_i_blocks_hi != 0))
2223 mark_inode_bad(ctx, ino);
2224 if (inode->i_flags & EXT2_IMAGIC_FL) {
2226 if (!ctx->inode_imagic_map)
2227 alloc_imagic_map(ctx);
2228 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
2231 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
2232 inode->i_flags &= ~EXT2_IMAGIC_FL;
2233 e2fsck_write_inode(ctx, ino,
2240 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
2241 check_is_really_dir(ctx, &pctx, block_buf);
2244 * ext2fs_inode_has_valid_blocks2 does not actually look
2245 * at i_block[] values, so not endian-sensitive here.
2247 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
2248 LINUX_S_ISLNK(inode->i_mode) &&
2249 !ext2fs_inode_has_valid_blocks2(fs, inode) &&
2250 fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
2251 inode->i_flags &= ~EXT4_EXTENTS_FL;
2252 e2fsck_write_inode(ctx, ino, inode, "pass1");
2256 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
2257 add_encrypted_file(ctx, &pctx) < 0)
2260 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
2261 ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
2263 if (LINUX_S_ISDIR(inode->i_mode)) {
2264 ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
2265 e2fsck_add_dir_info(ctx, ino, 0);
2266 ctx->fs_directory_count++;
2267 if (inode->i_flags & EXT4_CASEFOLD_FL)
2268 add_casefolded_dir(ctx, ino);
2269 } else if (LINUX_S_ISREG (inode->i_mode)) {
2270 ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
2271 ctx->fs_regular_count++;
2272 } else if (LINUX_S_ISCHR (inode->i_mode) &&
2273 e2fsck_pass1_check_device_inode(fs, inode)) {
2274 check_extents_inlinedata(ctx, &pctx);
2275 check_immutable(ctx, &pctx);
2276 check_size(ctx, &pctx);
2277 ctx->fs_chardev_count++;
2278 } else if (LINUX_S_ISBLK (inode->i_mode) &&
2279 e2fsck_pass1_check_device_inode(fs, inode)) {
2280 check_extents_inlinedata(ctx, &pctx);
2281 check_immutable(ctx, &pctx);
2282 check_size(ctx, &pctx);
2283 ctx->fs_blockdev_count++;
2284 } else if (LINUX_S_ISLNK (inode->i_mode) &&
2285 e2fsck_pass1_check_symlink(fs, ino, inode,
2287 check_immutable(ctx, &pctx);
2288 ctx->fs_symlinks_count++;
2289 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
2290 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2291 e2fsck_pass1_check_unlock(ctx);
2293 } else if (ext2fs_is_fast_symlink(inode)) {
2294 ctx->fs_fast_symlinks_count++;
2295 check_blocks(ctx, &pctx, block_buf,
2297 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2298 e2fsck_pass1_check_unlock(ctx);
2302 else if (LINUX_S_ISFIFO (inode->i_mode) &&
2303 e2fsck_pass1_check_device_inode(fs, inode)) {
2304 check_extents_inlinedata(ctx, &pctx);
2305 check_immutable(ctx, &pctx);
2306 check_size(ctx, &pctx);
2307 ctx->fs_fifo_count++;
2308 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
2309 e2fsck_pass1_check_device_inode(fs, inode)) {
2310 check_extents_inlinedata(ctx, &pctx);
2311 check_immutable(ctx, &pctx);
2312 check_size(ctx, &pctx);
2313 ctx->fs_sockets_count++;
2315 mark_inode_bad(ctx, ino);
2316 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2317 !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
2318 if (inode->i_block[EXT2_IND_BLOCK])
2319 ctx->fs_ind_count++;
2320 if (inode->i_block[EXT2_DIND_BLOCK])
2321 ctx->fs_dind_count++;
2322 if (inode->i_block[EXT2_TIND_BLOCK])
2323 ctx->fs_tind_count++;
2325 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2326 !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
2327 (inode->i_block[EXT2_IND_BLOCK] ||
2328 inode->i_block[EXT2_DIND_BLOCK] ||
2329 inode->i_block[EXT2_TIND_BLOCK] ||
2330 ext2fs_file_acl_block(fs, inode))) {
2331 struct process_inode_block *itp;
2333 itp = &inodes_to_process[process_inode_count];
2335 itp->ea_ibody_quota = ea_ibody_quota;
2336 if (inode_size < sizeof(struct ext2_inode_large))
2337 memcpy(&itp->inode, inode, inode_size);
2339 memcpy(&itp->inode, inode, sizeof(itp->inode));
2340 process_inode_count++;
2342 check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
2344 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2346 if (e2fsck_should_abort(ctx)) {
2347 e2fsck_pass1_check_unlock(ctx);
2351 if (process_inode_count >= ctx->process_inode_size) {
2352 process_inodes(ctx, block_buf, inodes_to_process,
2353 &process_inode_count);
2355 if (e2fsck_should_abort(ctx)) {
2356 e2fsck_pass1_check_unlock(ctx);
2360 e2fsck_pass1_check_unlock(ctx);
2362 process_inodes(ctx, block_buf, inodes_to_process,
2363 &process_inode_count);
2364 ext2fs_close_inode_scan(scan);
2367 if (ctx->ea_block_quota_blocks) {
2368 ea_refcount_free(ctx->ea_block_quota_blocks);
2369 ctx->ea_block_quota_blocks = 0;
2372 if (ctx->ea_block_quota_inodes) {
2373 ea_refcount_free(ctx->ea_block_quota_inodes);
2374 ctx->ea_block_quota_inodes = 0;
2377 /* We don't need the encryption policy => ID map any more */
2378 destroy_encryption_policy_map(ctx);
2380 if (ctx->flags & E2F_FLAG_RESTART) {
2382 * Only the master copy of the superblock and block
2383 * group descriptors are going to be written during a
2384 * restart, so set the superblock to be used to be the
2385 * master superblock.
2387 ctx->use_superblock = 0;
2391 if (ctx->large_dirs && !ext2fs_has_feature_largedir(ctx->fs->super)) {
2392 ext2_filsys fs = ctx->fs;
2394 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2395 ext2fs_set_feature_largedir(fs->super);
2396 fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2397 ext2fs_mark_super_dirty(fs);
2399 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2400 fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2401 ext2fs_update_dynamic_rev(fs);
2402 ext2fs_mark_super_dirty(fs);
2406 ctx->flags |= E2F_FLAG_ALLOC_OK;
2407 ext2fs_free_mem(&inodes_to_process);
2409 e2fsck_use_inode_shortcuts(ctx, 0);
2410 ext2fs_free_mem(&inodes_to_process);
2411 inodes_to_process = 0;
2414 ext2fs_close_inode_scan(scan);
2416 ext2fs_free_mem(&block_buf);
2418 ext2fs_free_mem(&inode);
2421 * The l+f inode may have been cleared, so zap it now and
2422 * later passes will recalculate it if necessary
2424 ctx->lost_and_found = 0;
2426 if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2427 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2429 ctx->invalid_bitmaps++;
2431 /* reset update_thread after this thread exit */
2432 e2fsck_pass1_block_map_w_lock(ctx);
2434 global_ctx->mmp_update_thread = 0;
2435 e2fsck_pass1_block_map_w_unlock(ctx);
2440 static errcode_t e2fsck_pass1_copy_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2441 ext2fs_generic_bitmap *dest)
2445 ret = ext2fs_copy_bitmap(*src, dest);
2454 static void e2fsck_pass1_free_bitmap(ext2fs_generic_bitmap *bitmap)
2457 ext2fs_free_generic_bmap(*bitmap);
2463 static errcode_t e2fsck_pass1_merge_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2464 ext2fs_generic_bitmap *dest)
2469 if (*dest == NULL) {
2473 ret = ext2fs_merge_bitmap(*src, *dest, NULL, NULL);
2483 static errcode_t e2fsck_pass1_copy_fs(ext2_filsys dest, e2fsck_t src_context,
2488 memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2489 dest->inode_map = NULL;
2490 dest->block_map = NULL;
2491 dest->badblocks = NULL;
2493 dest->dblist->fs = dest;
2494 if (src->block_map) {
2495 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2500 if (src->inode_map) {
2501 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2507 if (src->badblocks) {
2508 retval = ext2fs_badblocks_copy(src->badblocks,
2514 /* disable it for now */
2515 src_context->openfs_flags &= ~EXT2_FLAG_EXCLUSIVE;
2516 retval = ext2fs_open_channel(dest, src_context->io_options,
2517 src_context->io_manager,
2518 src_context->openfs_flags,
2519 src->io->block_size);
2523 /* Block size might not be default */
2524 io_channel_set_blksize(dest->io, src->io->block_size);
2525 ehandler_init(dest->io);
2527 assert(dest->io->magic == src->io->magic);
2528 assert(dest->io->manager == src->io->manager);
2529 assert(strcmp(dest->io->name, src->io->name) == 0);
2530 assert(dest->io->block_size == src->io->block_size);
2531 assert(dest->io->read_error == src->io->read_error);
2532 assert(dest->io->write_error == src->io->write_error);
2533 assert(dest->io->refcount == src->io->refcount);
2534 assert(dest->io->flags == src->io->flags);
2535 assert(dest->io->app_data == dest);
2536 assert(src->io->app_data == src);
2537 assert(dest->io->align == src->io->align);
2539 /* The data should be written to disk immediately */
2540 dest->io->flags |= CHANNEL_FLAGS_WRITETHROUGH;
2541 /* icache will be rebuilt if needed, so do not copy from @src */
2546 static int e2fsck_pass1_merge_fs(ext2_filsys dest, ext2_filsys src)
2548 struct ext2_inode_cache *icache = dest->icache;
2549 errcode_t retval = 0;
2551 io_channel dest_image_io;
2552 ext2fs_inode_bitmap inode_map;
2553 ext2fs_block_bitmap block_map;
2554 ext2_badblocks_list badblocks;
2557 e2fsck_t dest_ctx = dest->priv_data;
2560 dest_image_io = dest->image_io;
2561 inode_map = dest->inode_map;
2562 block_map = dest->block_map;
2563 badblocks = dest->badblocks;
2564 dblist = dest->dblist;
2565 flags = dest->flags;
2567 memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2569 dest->image_io = dest_image_io;
2570 dest->icache = icache;
2571 dest->inode_map = inode_map;
2572 dest->block_map = block_map;
2573 dest->badblocks = badblocks;
2574 dest->dblist = dblist;
2575 dest->priv_data = dest_ctx;
2577 dest->dblist->fs = dest;
2578 dest->flags = src->flags | flags;
2579 if (!(src->flags & EXT2_FLAG_VALID) || !(flags & EXT2_FLAG_VALID))
2580 ext2fs_unmark_valid(dest);
2583 ext2fs_free_inode_cache(src->icache);
2587 retval = e2fsck_pass1_merge_bitmap(dest, &src->inode_map,
2592 retval = e2fsck_pass1_merge_bitmap(dest, &src->block_map,
2599 retval = ext2fs_merge_dblist(src->dblist,
2604 dest->dblist = src->dblist;
2605 dest->dblist->fs = dest;
2610 if (src->badblocks) {
2611 if (dest->badblocks == NULL)
2612 retval = ext2fs_badblocks_copy(src->badblocks,
2615 retval = ext2fs_badblocks_merge(src->badblocks,
2619 io_channel_close(src->io);
2621 ext2fs_free_generic_bmap(src->inode_map);
2623 ext2fs_free_generic_bmap(src->block_map);
2625 ext2fs_badblocks_list_free(src->badblocks);
2627 ext2fs_free_dblist(src->dblist);
2632 static void e2fsck_pass1_copy_invalid_bitmaps(e2fsck_t global_ctx,
2633 e2fsck_t thread_ctx)
2636 dgrp_t grp_start = thread_ctx->thread_info.et_group_start;
2637 dgrp_t grp_end = thread_ctx->thread_info.et_group_end;
2638 dgrp_t total = grp_end - grp_start;
2640 thread_ctx->invalid_inode_bitmap_flag =
2641 e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2642 "invalid_inode_bitmap");
2643 thread_ctx->invalid_block_bitmap_flag =
2644 e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2645 "invalid_block_bitmap");
2646 thread_ctx->invalid_inode_table_flag =
2647 e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2648 "invalid_inode_table");
2650 memcpy(thread_ctx->invalid_block_bitmap_flag,
2651 &global_ctx->invalid_block_bitmap_flag[grp_start],
2652 total * sizeof(int));
2653 memcpy(thread_ctx->invalid_inode_bitmap_flag,
2654 &global_ctx->invalid_inode_bitmap_flag[grp_start],
2655 total * sizeof(int));
2656 memcpy(thread_ctx->invalid_inode_table_flag,
2657 &global_ctx->invalid_inode_table_flag[grp_start],
2658 total * sizeof(int));
2660 thread_ctx->invalid_bitmaps = 0;
2661 for (i = grp_start, j = 0; i < grp_end; i++, j++) {
2662 if (thread_ctx->invalid_block_bitmap_flag[j])
2663 thread_ctx->invalid_bitmaps++;
2664 if (thread_ctx->invalid_inode_bitmap_flag[j])
2665 thread_ctx->invalid_bitmaps++;
2666 if (thread_ctx->invalid_inode_table_flag[j])
2667 thread_ctx->invalid_bitmaps++;
2671 static void e2fsck_pass1_merge_invalid_bitmaps(e2fsck_t global_ctx,
2672 e2fsck_t thread_ctx)
2675 dgrp_t grp_start = thread_ctx->thread_info.et_group_start;
2676 dgrp_t grp_end = thread_ctx->thread_info.et_group_end;
2677 dgrp_t total = grp_end - grp_start;
2679 memcpy(&global_ctx->invalid_block_bitmap_flag[grp_start],
2680 thread_ctx->invalid_block_bitmap_flag, total * sizeof(int));
2681 memcpy(&global_ctx->invalid_inode_bitmap_flag[grp_start],
2682 thread_ctx->invalid_inode_bitmap_flag, total * sizeof(int));
2683 memcpy(&global_ctx->invalid_inode_table_flag[grp_start],
2684 thread_ctx->invalid_inode_table_flag, total * sizeof(int));
2685 global_ctx->invalid_bitmaps += thread_ctx->invalid_bitmaps;
2688 static errcode_t e2fsck_pass1_thread_prepare(e2fsck_t global_ctx, e2fsck_t *thread_ctx,
2689 int thread_index, int num_threads,
2690 dgrp_t average_group)
2693 e2fsck_t thread_context;
2694 ext2_filsys thread_fs;
2695 ext2_filsys global_fs = global_ctx->fs;
2696 struct e2fsck_thread *tinfo;
2698 assert(global_ctx->inode_used_map == NULL);
2699 assert(global_ctx->inode_dir_map == NULL);
2700 assert(global_ctx->inode_bb_map == NULL);
2701 assert(global_ctx->inode_imagic_map == NULL);
2702 assert(global_ctx->inode_reg_map == NULL);
2703 assert(global_ctx->inodes_to_rebuild == NULL);
2705 assert(global_ctx->block_found_map != NULL);
2706 assert(global_ctx->block_metadata_map != NULL);
2707 assert(global_ctx->block_dup_map != NULL);
2708 assert(global_ctx->block_ea_map == NULL);
2709 assert(global_ctx->fs->dblist == NULL);
2711 retval = ext2fs_get_mem(sizeof(struct e2fsck_struct), &thread_context);
2713 com_err(global_ctx->program_name, retval, "while allocating memory");
2716 memcpy(thread_context, global_ctx, sizeof(struct e2fsck_struct));
2717 thread_context->block_dup_map = NULL;
2719 retval = e2fsck_allocate_block_bitmap(global_ctx->fs,
2720 _("in-use block map"), EXT2FS_BMAP64_RBTREE,
2721 "block_found_map", &thread_context->block_found_map);
2725 thread_context->global_ctx = global_ctx;
2726 retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &thread_fs);
2728 com_err(global_ctx->program_name, retval, "while allocating memory");
2732 io_channel_flush_cleanup(global_fs->io);
2733 retval = e2fsck_pass1_copy_fs(thread_fs, global_ctx, global_fs);
2735 com_err(global_ctx->program_name, retval, "while copying fs");
2738 thread_fs->priv_data = thread_context;
2740 thread_context->thread_info.et_thread_index = thread_index;
2741 set_up_logging(thread_context);
2743 tinfo = &thread_context->thread_info;
2744 tinfo->et_group_start = average_group * thread_index;
2745 if (thread_index == global_fs->fs_num_threads - 1)
2746 tinfo->et_group_end = thread_fs->group_desc_count;
2748 tinfo->et_group_end = average_group * (thread_index + 1);
2749 tinfo->et_group_next = tinfo->et_group_start;
2750 tinfo->et_inode_number = 0;
2751 tinfo->et_log_buf[0] = '\0';
2752 tinfo->et_log_length = 0;
2753 if (thread_context->options & E2F_OPT_MULTITHREAD)
2754 log_out(thread_context, _("Scan group range [%d, %d)\n"),
2755 tinfo->et_group_start, tinfo->et_group_end);
2756 thread_context->fs = thread_fs;
2757 retval = quota_init_context(&thread_context->qctx, thread_fs, 0);
2759 com_err(global_ctx->program_name, retval,
2760 "while init quota context");
2763 *thread_ctx = thread_context;
2764 e2fsck_pass1_copy_invalid_bitmaps(global_ctx, thread_context);
2767 ext2fs_free_mem(&thread_fs);
2769 if (thread_context->block_found_map)
2770 ext2fs_free_mem(&thread_context->block_found_map);
2771 ext2fs_free_mem(&thread_context);
2775 static void e2fsck_pass1_merge_dir_info(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2777 if (thread_ctx->dir_info == NULL)
2780 if (global_ctx->dir_info == NULL) {
2781 global_ctx->dir_info = thread_ctx->dir_info;
2782 thread_ctx->dir_info = NULL;
2786 e2fsck_merge_dir_info(global_ctx, thread_ctx->dir_info,
2787 global_ctx->dir_info);
2790 static void e2fsck_pass1_merge_dx_dir(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2792 if (thread_ctx->dx_dir_info == NULL)
2795 if (global_ctx->dx_dir_info == NULL) {
2796 global_ctx->dx_dir_info = thread_ctx->dx_dir_info;
2797 global_ctx->dx_dir_info_size = thread_ctx->dx_dir_info_size;
2798 global_ctx->dx_dir_info_count = thread_ctx->dx_dir_info_count;
2799 thread_ctx->dx_dir_info = NULL;
2803 e2fsck_merge_dx_dir(global_ctx, thread_ctx);
2806 static inline errcode_t
2807 e2fsck_pass1_merge_icount(ext2_icount_t *dest_icount,
2808 ext2_icount_t *src_icount)
2811 if (*dest_icount == NULL) {
2812 *dest_icount = *src_icount;
2817 ret = ext2fs_icount_merge(*src_icount,
2827 static errcode_t e2fsck_pass1_merge_icounts(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2831 ret = e2fsck_pass1_merge_icount(&global_ctx->inode_count,
2832 &thread_ctx->inode_count);
2835 ret = e2fsck_pass1_merge_icount(&global_ctx->inode_link_info,
2836 &thread_ctx->inode_link_info);
2841 static errcode_t e2fsck_pass1_merge_dirs_to_hash(e2fsck_t global_ctx,
2842 e2fsck_t thread_ctx)
2844 errcode_t retval = 0;
2846 if (!thread_ctx->dirs_to_hash)
2849 if (!global_ctx->dirs_to_hash)
2850 retval = ext2fs_badblocks_copy(thread_ctx->dirs_to_hash,
2851 &global_ctx->dirs_to_hash);
2853 retval = ext2fs_badblocks_merge(thread_ctx->dirs_to_hash,
2854 global_ctx->dirs_to_hash);
2859 static errcode_t e2fsck_pass1_merge_ea_inode_refs(e2fsck_t global_ctx,
2860 e2fsck_t thread_ctx)
2866 if (!thread_ctx->ea_inode_refs)
2869 if (!global_ctx->ea_inode_refs) {
2870 global_ctx->ea_inode_refs = thread_ctx->ea_inode_refs;
2871 thread_ctx->ea_inode_refs = NULL;
2875 ea_refcount_intr_begin(thread_ctx->ea_inode_refs);
2877 if ((blk = ea_refcount_intr_next(thread_ctx->ea_inode_refs,
2880 if (!global_ctx->block_ea_map ||
2881 !ext2fs_fast_test_block_bitmap2(global_ctx->block_ea_map,
2883 retval = ea_refcount_store(global_ctx->ea_inode_refs,
2893 static ea_value_t ea_refcount_usage(e2fsck_t ctx, blk64_t blk,
2896 ea_value_t count_cur;
2897 ea_value_t count_extra = 0;
2898 ea_value_t count_orig;
2900 ea_refcount_fetch(ctx->refcount_orig, blk, &count_orig);
2901 ea_refcount_fetch(ctx->refcount, blk, &count_cur);
2902 /* most of time this is not needed */
2903 if (ctx->refcount_extra && count_cur == 0)
2904 ea_refcount_fetch(ctx->refcount_extra, blk, &count_extra);
2911 return count_orig + count_extra - count_cur;
2914 static errcode_t e2fsck_pass1_merge_ea_refcount(e2fsck_t global_ctx,
2915 e2fsck_t thread_ctx)
2919 errcode_t retval = 0;
2921 if (!thread_ctx->refcount)
2924 if (!global_ctx->refcount) {
2925 global_ctx->refcount = thread_ctx->refcount;
2926 thread_ctx->refcount = NULL;
2927 global_ctx->refcount_extra = thread_ctx->refcount;
2928 thread_ctx->refcount_extra = NULL;
2932 ea_refcount_intr_begin(thread_ctx->refcount);
2934 if ((blk = ea_refcount_intr_next(thread_ctx->refcount,
2938 * this EA has never seen before, so just store its
2939 * refcount and refcount_extra into global_ctx if needed.
2941 if (!global_ctx->block_ea_map ||
2942 !ext2fs_fast_test_block_bitmap2(global_ctx->block_ea_map,
2946 retval = ea_refcount_store(global_ctx->refcount,
2951 if (count > 0 || !thread_ctx->refcount_extra)
2953 ea_refcount_fetch(thread_ctx->refcount_extra, blk,
2958 if (!global_ctx->refcount_extra) {
2959 retval = ea_refcount_create(0,
2960 &global_ctx->refcount_extra);
2964 retval = ea_refcount_store(global_ctx->refcount_extra,
2970 ea_value_t thread_usage;
2971 ea_value_t global_usage;
2974 thread_usage = ea_refcount_usage(thread_ctx,
2976 global_usage = ea_refcount_usage(global_ctx,
2978 if (thread_usage + global_usage <= orig) {
2979 new = orig - thread_usage - global_usage;
2980 retval = ea_refcount_store(global_ctx->refcount,
2986 /* update it is as zero */
2987 retval = ea_refcount_store(global_ctx->refcount,
2991 /* Ooops, this EA was referenced more than it stated */
2992 if (!global_ctx->refcount_extra) {
2993 retval = ea_refcount_create(0,
2994 &global_ctx->refcount_extra);
2998 new = global_usage + thread_usage - orig;
2999 retval = ea_refcount_store(global_ctx->refcount_extra,
3009 static errcode_t e2fsck_pass1_merge_context(e2fsck_t global_ctx,
3010 e2fsck_t thread_ctx)
3012 ext2_filsys global_fs = global_ctx->fs;
3016 global_ctx->fs_directory_count += thread_ctx->fs_directory_count;
3017 global_ctx->fs_regular_count += thread_ctx->fs_regular_count;
3018 global_ctx->fs_blockdev_count += thread_ctx->fs_blockdev_count;
3019 global_ctx->fs_chardev_count += thread_ctx->fs_chardev_count;
3020 global_ctx->fs_links_count += thread_ctx->fs_links_count;
3021 global_ctx->fs_symlinks_count += thread_ctx->fs_symlinks_count;
3022 global_ctx->fs_fast_symlinks_count += thread_ctx->fs_fast_symlinks_count;
3023 global_ctx->fs_fifo_count += thread_ctx->fs_fifo_count;
3024 global_ctx->fs_total_count += thread_ctx->fs_total_count;
3025 global_ctx->fs_badblocks_count += thread_ctx->fs_badblocks_count;
3026 global_ctx->fs_sockets_count += thread_ctx->fs_sockets_count;
3027 global_ctx->fs_ind_count += thread_ctx->fs_ind_count;
3028 global_ctx->fs_dind_count += thread_ctx->fs_dind_count;
3029 global_ctx->fs_tind_count += thread_ctx->fs_tind_count;
3030 global_ctx->fs_fragmented += thread_ctx->fs_fragmented;
3031 global_ctx->fs_fragmented_dir += thread_ctx->fs_fragmented_dir;
3032 global_ctx->large_files += thread_ctx->large_files;
3033 /* threads might enable E2F_OPT_YES */
3034 global_ctx->options |= thread_ctx->options;
3035 global_ctx->flags |= thread_ctx->flags;
3037 * The l+f inode may have been cleared, so zap it now and
3038 * later passes will recalculate it if necessary
3040 global_ctx->lost_and_found = 0;
3041 /* merge extent depth count */
3042 for (i = 0; i < MAX_EXTENT_DEPTH_COUNT; i++)
3043 global_ctx->extent_depth_count[i] +=
3044 thread_ctx->extent_depth_count[i];
3046 e2fsck_pass1_merge_dir_info(global_ctx, thread_ctx);
3047 e2fsck_pass1_merge_dx_dir(global_ctx, thread_ctx);
3049 retval = e2fsck_pass1_merge_fs(global_ctx->fs, thread_ctx->fs);
3051 com_err(global_ctx->program_name, 0, _("while merging fs\n"));
3054 retval = e2fsck_pass1_merge_icounts(global_ctx, thread_ctx);
3056 com_err(global_ctx->program_name, 0,
3057 _("while merging icounts\n"));
3061 retval = e2fsck_pass1_merge_dirs_to_hash(global_ctx, thread_ctx);
3063 com_err(global_ctx->program_name, 0,
3064 _("while merging dirs to hash\n"));
3068 e2fsck_pass1_merge_ea_inode_refs(global_ctx, thread_ctx);
3069 e2fsck_pass1_merge_ea_refcount(global_ctx, thread_ctx);
3070 retval = quota_merge_and_update_usage(global_ctx->qctx,
3075 e2fsck_pass1_merge_invalid_bitmaps(global_ctx, thread_ctx);
3077 retval = e2fsck_pass1_merge_bitmap(global_fs,
3078 &thread_ctx->inode_used_map,
3079 &global_ctx->inode_used_map);
3083 retval = e2fsck_pass1_merge_bitmap(global_fs,
3084 &thread_ctx->inode_bad_map,
3085 &global_ctx->inode_bad_map);
3088 retval = e2fsck_pass1_merge_bitmap(global_fs,
3089 &thread_ctx->inode_dir_map,
3090 &global_ctx->inode_dir_map);
3093 retval = e2fsck_pass1_merge_bitmap(global_fs,
3094 &thread_ctx->inode_bb_map,
3095 &global_ctx->inode_bb_map);
3098 retval = e2fsck_pass1_merge_bitmap(global_fs,
3099 &thread_ctx->inode_imagic_map,
3100 &global_ctx->inode_imagic_map);
3103 retval = e2fsck_pass1_merge_bitmap(global_fs,
3104 &thread_ctx->inode_reg_map,
3105 &global_ctx->inode_reg_map);
3108 retval = e2fsck_pass1_merge_bitmap(global_fs,
3109 &thread_ctx->inodes_to_rebuild,
3110 &global_ctx->inodes_to_rebuild);
3113 retval = e2fsck_pass1_merge_bitmap(global_fs,
3114 &thread_ctx->block_ea_map,
3115 &global_ctx->block_ea_map);
3119 if (ext2fs_has_feature_shared_blocks(global_fs->super) &&
3120 !(global_ctx->options & E2F_OPT_UNSHARE_BLOCKS))
3123 * This need be done after merging block_ea_map
3124 * because ea block might be shared, we need exclude
3125 * them from dup blocks.
3127 e2fsck_pass1_block_map_w_lock(thread_ctx);
3128 retval = ext2fs_merge_bitmap(thread_ctx->block_found_map,
3129 global_ctx->block_found_map,
3130 global_ctx->block_dup_map,
3131 global_ctx->block_ea_map);
3132 e2fsck_pass1_block_map_w_unlock(thread_ctx);
3133 if (retval == EEXIST)
3134 global_ctx->flags |= E2F_FLAG_DUP_BLOCK;
3139 static int e2fsck_pass1_thread_join(e2fsck_t global_ctx, e2fsck_t thread_ctx)
3143 retval = e2fsck_pass1_merge_context(global_ctx, thread_ctx);
3144 ext2fs_free_mem(&thread_ctx->fs);
3145 if (thread_ctx->logf)
3146 fclose(thread_ctx->logf);
3147 if (thread_ctx->problem_logf) {
3148 fputs("</problem_log>\n", thread_ctx->problem_logf);
3149 fclose(thread_ctx->problem_logf);
3152 quota_release_context(&thread_ctx->qctx);
3154 * @block_metadata_map and @block_dup_map are
3155 * shared, so we don't free them.
3157 thread_ctx->block_metadata_map = NULL;
3158 thread_ctx->block_dup_map = NULL;
3159 e2fsck_reset_context(thread_ctx);
3160 ext2fs_free_mem(&thread_ctx);
3165 static int e2fsck_pass1_threads_join(struct e2fsck_thread_info *infos,
3166 e2fsck_t global_ctx)
3170 struct e2fsck_thread_info *pinfo;
3171 int num_threads = global_ctx->pfs_num_threads;
3174 /* merge invalid bitmaps will recalculate it */
3175 global_ctx->invalid_bitmaps = 0;
3176 for (i = 0; i < num_threads; i++) {
3179 if (!pinfo->eti_started)
3182 rc = pthread_join(pinfo->eti_thread_id, NULL);
3184 com_err(global_ctx->program_name, rc,
3185 _("while joining thread\n"));
3189 rc = e2fsck_pass1_thread_join(global_ctx, infos[i].eti_thread_ctx);
3191 com_err(global_ctx->program_name, rc,
3192 _("while joining pass1 thread\n"));
3202 static void *e2fsck_pass1_thread(void *arg)
3204 struct e2fsck_thread_info *info = arg;
3205 e2fsck_t thread_ctx = info->eti_thread_ctx;
3206 #ifdef DEBUG_THREADS
3207 struct e2fsck_thread_debug *thread_debug = info->eti_debug;
3210 #ifdef DEBUG_THREADS
3211 pthread_mutex_lock(&thread_debug->etd_mutex);
3212 while (info->eti_thread_index > thread_debug->etd_finished_threads) {
3213 pthread_cond_wait(&thread_debug->etd_cond,
3214 &thread_debug->etd_mutex);
3216 pthread_mutex_unlock(&thread_debug->etd_mutex);
3219 #ifdef HAVE_SETJMP_H
3221 * When fatal_error() happens, jump to here. The thread
3222 * context's flags will be saved, but its abort_loc will
3223 * be overwritten by original jump buffer for the later
3226 if (setjmp(thread_ctx->abort_loc)) {
3227 thread_ctx->flags &= ~E2F_FLAG_SETJMP_OK;
3230 thread_ctx->flags |= E2F_FLAG_SETJMP_OK;
3233 e2fsck_pass1_run(thread_ctx);
3236 if (thread_ctx->options & E2F_OPT_MULTITHREAD)
3238 _("Scanned group range [%u, %u), inodes %u\n"),
3239 thread_ctx->thread_info.et_group_start,
3240 thread_ctx->thread_info.et_group_end,
3241 thread_ctx->thread_info.et_inode_number);
3243 #ifdef DEBUG_THREADS
3244 pthread_mutex_lock(&thread_debug->etd_mutex);
3245 thread_debug->etd_finished_threads++;
3246 pthread_cond_broadcast(&thread_debug->etd_cond);
3247 pthread_mutex_unlock(&thread_debug->etd_mutex);
3253 static int e2fsck_pass1_threads_start(struct e2fsck_thread_info **pinfo,
3254 e2fsck_t global_ctx)
3256 struct e2fsck_thread_info *infos;
3257 pthread_attr_t attr;
3260 struct e2fsck_thread_info *tmp_pinfo;
3262 e2fsck_t thread_ctx;
3263 dgrp_t average_group;
3264 int num_threads = global_ctx->pfs_num_threads;
3265 #ifdef DEBUG_THREADS
3266 struct e2fsck_thread_debug thread_debug =
3267 {PTHREAD_MUTEX_INITIALIZER, PTHREAD_COND_INITIALIZER, 0};
3269 thread_debug.etd_finished_threads = 0;
3272 retval = pthread_attr_init(&attr);
3274 com_err(global_ctx->program_name, retval,
3275 _("while setting pthread attribute\n"));
3279 infos = calloc(num_threads, sizeof(struct e2fsck_thread_info));
3280 if (infos == NULL) {
3282 com_err(global_ctx->program_name, retval,
3283 _("while allocating memory for threads\n"));
3284 pthread_attr_destroy(&attr);
3288 average_group = ext2fs_get_avg_group(global_ctx->fs);
3289 for (i = 0; i < num_threads; i++) {
3290 tmp_pinfo = &infos[i];
3291 tmp_pinfo->eti_thread_index = i;
3292 #ifdef DEBUG_THREADS
3293 tmp_pinfo->eti_debug = &thread_debug;
3295 retval = e2fsck_pass1_thread_prepare(global_ctx, &thread_ctx,
3299 com_err(global_ctx->program_name, retval,
3300 _("while preparing pass1 thread\n"));
3303 tmp_pinfo->eti_thread_ctx = thread_ctx;
3305 retval = pthread_create(&tmp_pinfo->eti_thread_id, &attr,
3306 &e2fsck_pass1_thread, tmp_pinfo);
3308 com_err(global_ctx->program_name, retval,
3309 _("while creating thread\n"));
3310 e2fsck_pass1_thread_join(global_ctx, thread_ctx);
3314 tmp_pinfo->eti_started = 1;
3317 /* destroy the thread attribute object, since it is no longer needed */
3318 ret = pthread_attr_destroy(&attr);
3320 com_err(global_ctx->program_name, ret,
3321 _("while destroying thread attribute\n"));
3327 e2fsck_pass1_threads_join(infos, global_ctx);
3334 static void e2fsck_pass1_multithread(e2fsck_t global_ctx)
3336 struct e2fsck_thread_info *infos = NULL;
3339 retval = e2fsck_pass1_threads_start(&infos, global_ctx);
3341 com_err(global_ctx->program_name, retval,
3342 _("while starting pass1 threads\n"));
3346 retval = e2fsck_pass1_threads_join(infos, global_ctx);
3348 com_err(global_ctx->program_name, retval,
3349 _("while joining pass1 threads\n"));
3354 global_ctx->flags |= E2F_FLAG_ABORT;
3359 void e2fsck_pass1(e2fsck_t ctx)
3362 int need_single = 1;
3364 retval = e2fsck_pass1_prepare(ctx);
3368 if (ctx->pfs_num_threads > 1 || ctx->options & E2F_OPT_MULTITHREAD) {
3370 e2fsck_pass1_multithread(ctx);
3372 /* No lock is needed at this time */
3373 ctx->fs_need_locking = 0;
3376 e2fsck_pass1_run(ctx);
3377 e2fsck_pass1_post(ctx);
3380 #undef FINISH_INODE_LOOP
3383 * When the inode_scan routines call this callback at the end of the
3384 * glock group, call process_inodes.
3386 static errcode_t scan_callback(ext2_filsys fs,
3387 ext2_inode_scan scan EXT2FS_ATTR((unused)),
3388 dgrp_t group, void * priv_data)
3390 struct scan_callback_struct *scan_struct;
3392 struct e2fsck_thread *tinfo;
3394 scan_struct = (struct scan_callback_struct *) priv_data;
3395 ctx = scan_struct->ctx;
3397 process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf,
3398 scan_struct->inodes_to_process,
3399 scan_struct->process_inode_count);
3402 if ((ctx->progress)(ctx, 1, group+1,
3403 ctx->fs->group_desc_count))
3404 return EXT2_ET_CANCEL_REQUESTED;
3407 if (ctx->global_ctx) {
3408 tinfo = &ctx->thread_info;
3409 tinfo->et_group_next++;
3410 if (ctx->options & E2F_OPT_DEBUG &&
3411 ctx->options & E2F_OPT_MULTITHREAD)
3412 log_out(ctx, _("group %d finished\n"),
3413 tinfo->et_group_next);
3414 if (tinfo->et_group_next >= tinfo->et_group_end)
3415 return EXT2_ET_SCAN_FINISHED;
3423 * Process the inodes in the "inodes to process" list.
3425 static void process_inodes(e2fsck_t ctx, char *block_buf,
3426 struct process_inode_block *inodes_to_process,
3427 int *process_inode_count)
3430 struct ext2_inode *old_stashed_inode;
3431 ext2_ino_t old_stashed_ino;
3432 const char *old_operation;
3434 struct problem_context pctx;
3437 printf("begin process_inodes: ");
3439 if (*process_inode_count == 0)
3441 old_operation = ehandler_operation(0);
3442 old_stashed_inode = ctx->stashed_inode;
3443 old_stashed_ino = ctx->stashed_ino;
3444 qsort(inodes_to_process, *process_inode_count,
3445 sizeof(struct process_inode_block), process_inode_cmp);
3446 clear_problem_context(&pctx);
3447 for (i=0; i < *process_inode_count; i++) {
3448 pctx.inode = ctx->stashed_inode =
3449 (struct ext2_inode *) &inodes_to_process[i].inode;
3450 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
3453 printf("%u ", pctx.ino);
3455 sprintf(buf, _("reading indirect blocks of inode %u"),
3457 ehandler_operation(buf);
3458 check_blocks(ctx, &pctx, block_buf,
3459 &inodes_to_process[i].ea_ibody_quota);
3460 if (e2fsck_should_abort(ctx))
3463 ctx->stashed_inode = old_stashed_inode;
3464 ctx->stashed_ino = old_stashed_ino;
3465 *process_inode_count = 0;
3467 printf("end process inodes\n");
3469 ehandler_operation(old_operation);
3472 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
3474 const struct process_inode_block *ib_a =
3475 (const struct process_inode_block *) a;
3476 const struct process_inode_block *ib_b =
3477 (const struct process_inode_block *) b;
3480 ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
3481 ib_b->inode.i_block[EXT2_IND_BLOCK]);
3484 * We only call process_inodes() for non-extent
3485 * inodes, so it's OK to pass NULL to
3486 * ext2fs_file_acl_block() here.
3488 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
3489 ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
3491 ret = ib_a->ino - ib_b->ino;
3496 * Mark an inode as being bad in some what
3498 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
3500 struct problem_context pctx;
3502 if (!ctx->inode_bad_map) {
3503 clear_problem_context(&pctx);
3505 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3506 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
3507 "inode_bad_map", &ctx->inode_bad_map);
3510 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3511 /* Should never get here */
3512 ctx->flags |= E2F_FLAG_ABORT;
3516 ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
3519 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino)
3521 struct problem_context pctx;
3523 if (!ctx->casefolded_dirs) {
3524 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
3528 pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
3529 if (pctx.errcode == 0)
3532 fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
3533 /* Should never get here */
3534 ctx->flags |= E2F_FLAG_ABORT;
3538 * This procedure will allocate the inode "bb" (badblock) map table
3540 static void alloc_bb_map(e2fsck_t ctx)
3542 struct problem_context pctx;
3544 clear_problem_context(&pctx);
3545 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3546 _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
3547 "inode_bb_map", &ctx->inode_bb_map);
3550 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3551 /* Should never get here */
3552 ctx->flags |= E2F_FLAG_ABORT;
3558 * This procedure will allocate the inode imagic table
3560 static void alloc_imagic_map(e2fsck_t ctx)
3562 struct problem_context pctx;
3564 clear_problem_context(&pctx);
3565 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3566 _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
3567 "inode_imagic_map", &ctx->inode_imagic_map);
3570 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3571 /* Should never get here */
3572 ctx->flags |= E2F_FLAG_ABORT;
3578 * Marks a block as in use, setting the dup_map if it's been set
3579 * already. Called by process_block and process_bad_block.
3581 * WARNING: Assumes checks have already been done to make sure block
3582 * is valid. This is true in both process_block and process_bad_block.
3584 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
3586 struct problem_context pctx;
3587 e2fsck_t global_ctx = ctx->global_ctx ? ctx->global_ctx : ctx;
3589 clear_problem_context(&pctx);
3591 if (is_blocks_used(ctx, block, 1)) {
3592 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
3593 !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
3596 ctx->flags |= E2F_FLAG_DUP_BLOCK;
3597 e2fsck_pass1_block_map_w_lock(ctx);
3598 ext2fs_fast_mark_block_bitmap2(global_ctx->block_dup_map, block);
3599 e2fsck_pass1_block_map_w_unlock(ctx);
3601 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
3606 * When cluster size is greater than one block, it is caller's responsibility
3607 * to make sure block parameter starts at a cluster boundary.
3609 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
3612 if (!is_blocks_used(ctx, block, num)) {
3613 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
3617 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
3618 mark_block_used(ctx, block + i);
3622 static errcode_t _INLINE_ e2fsck_write_ext_attr3(e2fsck_t ctx, blk64_t block,
3623 void *inbuf, ext2_ino_t inum)
3626 ext2_filsys fs = ctx->fs;
3628 e2fsck_pass1_fix_lock(ctx);
3629 retval = ext2fs_write_ext_attr3(fs, block, inbuf, inum);
3630 e2fsck_pass1_fix_unlock(ctx);
3635 * Adjust the extended attribute block's reference counts at the end
3636 * of pass 1, either by subtracting out references for EA blocks that
3637 * are still referenced in ctx->refcount, or by adding references for
3638 * EA blocks that had extra references as accounted for in
3639 * ctx->refcount_extra.
3641 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
3642 char *block_buf, int adjust_sign)
3644 struct ext2_ext_attr_header *header;
3645 struct problem_context pctx;
3646 ext2_filsys fs = ctx->fs;
3651 clear_problem_context(&pctx);
3653 ea_refcount_intr_begin(refcount);
3655 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
3658 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
3661 fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
3664 header = (struct ext2_ext_attr_header *) block_buf;
3665 pctx.blkcount = header->h_refcount;
3666 should_be = header->h_refcount + adjust_sign * (int)count;
3667 pctx.num = should_be;
3668 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
3669 header->h_refcount = should_be;
3670 pctx.errcode = e2fsck_write_ext_attr3(ctx, blk,
3674 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
3683 * Handle processing the extended attribute blocks
3685 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
3686 char *block_buf, struct ea_quota *ea_block_quota)
3688 ext2_filsys fs = ctx->fs;
3689 ext2_ino_t ino = pctx->ino;
3690 struct ext2_inode *inode = pctx->inode;
3693 struct ext2_ext_attr_header *header;
3694 struct ext2_ext_attr_entry *first, *entry;
3695 blk64_t quota_blocks = EXT2FS_C2B(fs, 1);
3696 __u64 quota_inodes = 0;
3697 region_t region = 0;
3698 int failed_csum = 0;
3700 ea_block_quota->blocks = 0;
3701 ea_block_quota->inodes = 0;
3703 blk = ext2fs_file_acl_block(fs, inode);
3708 * If the Extended attribute flag isn't set, then a non-zero
3709 * file acl means that the inode is corrupted.
3711 * Or if the extended attribute block is an invalid block,
3712 * then the inode is also corrupted.
3714 if (!ext2fs_has_feature_xattr(fs->super) ||
3715 (blk < fs->super->s_first_data_block) ||
3716 (blk >= ext2fs_blocks_count(fs->super))) {
3717 mark_inode_bad(ctx, ino);
3721 /* If ea bitmap hasn't been allocated, create it */
3722 if (!ctx->block_ea_map) {
3723 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
3724 _("ext attr block map"),
3725 EXT2FS_BMAP64_RBTREE, "block_ea_map",
3726 &ctx->block_ea_map);
3727 if (pctx->errcode) {
3729 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
3730 ctx->flags |= E2F_FLAG_ABORT;
3735 /* Create the EA refcount structure if necessary */
3736 if (!ctx->refcount) {
3737 pctx->errcode = ea_refcount_create(0,
3738 &ctx->refcount_orig);
3739 if (pctx->errcode) {
3741 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3742 ctx->flags |= E2F_FLAG_ABORT;
3746 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
3747 if (pctx->errcode) {
3749 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3750 ctx->flags |= E2F_FLAG_ABORT;
3756 /* Debugging text */
3757 printf("Inode %u has EA block %u\n", ino, blk);
3760 /* Have we seen this EA block before? */
3761 if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
3762 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
3763 ea_block_quota->inodes = 0;
3765 if (ctx->ea_block_quota_blocks) {
3766 ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
3769 ea_block_quota->blocks = quota_blocks;
3772 if (ctx->ea_block_quota_inodes)
3773 ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
3774 &ea_block_quota->inodes);
3776 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
3778 /* Ooops, this EA was referenced more than it stated */
3779 if (!ctx->refcount_extra) {
3780 pctx->errcode = ea_refcount_create(0,
3781 &ctx->refcount_extra);
3782 if (pctx->errcode) {
3784 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3785 ctx->flags |= E2F_FLAG_ABORT;
3789 ea_refcount_increment(ctx->refcount_extra, blk, 0);
3794 * OK, we haven't seen this EA block yet. So we need to
3798 pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
3799 if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
3802 } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
3805 if (pctx->errcode &&
3806 fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
3810 header = (struct ext2_ext_attr_header *) block_buf;
3811 pctx->blk = ext2fs_file_acl_block(fs, inode);
3812 if (((ctx->ext_attr_ver == 1) &&
3813 (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
3814 ((ctx->ext_attr_ver == 2) &&
3815 (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
3816 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
3820 if (header->h_blocks != 1) {
3821 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
3825 if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
3828 region = region_create(0, fs->blocksize);
3830 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
3831 ctx->flags |= E2F_FLAG_ABORT;
3834 if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
3835 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3839 first = (struct ext2_ext_attr_entry *)(header+1);
3840 end = block_buf + fs->blocksize;
3842 while ((char *)entry < end && *(__u32 *)entry) {
3845 if (region_allocate(region, (char *)entry - (char *)header,
3846 EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
3847 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3851 if ((ctx->ext_attr_ver == 1 &&
3852 (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
3853 (ctx->ext_attr_ver == 2 &&
3854 entry->e_name_index == 0)) {
3855 if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
3859 if (entry->e_value_inum == 0) {
3860 if (entry->e_value_offs + entry->e_value_size >
3862 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
3866 if (entry->e_value_size &&
3867 region_allocate(region, entry->e_value_offs,
3868 EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
3869 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
3874 hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
3875 entry->e_value_offs);
3877 if (entry->e_hash != hash) {
3878 pctx->num = entry->e_hash;
3879 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
3881 entry->e_hash = hash;
3885 blk64_t entry_quota_blocks;
3887 problem = check_large_ea_inode(ctx, entry, pctx,
3888 &entry_quota_blocks);
3889 if (problem && fix_problem(ctx, problem, pctx))
3892 quota_blocks += entry_quota_blocks;
3896 entry = EXT2_EXT_ATTR_NEXT(entry);
3898 if (region_allocate(region, (char *)entry - (char *)header, 4)) {
3899 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3902 region_free(region);
3905 * We only get here if there was no other errors that were fixed.
3906 * If there was a checksum fail, ask to correct it.
3909 fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
3910 pctx->errcode = e2fsck_write_ext_attr3(ctx, blk, block_buf,
3916 if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
3917 if (!ctx->ea_block_quota_blocks) {
3918 pctx->errcode = ea_refcount_create(0,
3919 &ctx->ea_block_quota_blocks);
3920 if (pctx->errcode) {
3925 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
3930 if (!ctx->ea_block_quota_inodes) {
3931 pctx->errcode = ea_refcount_create(0,
3932 &ctx->ea_block_quota_inodes);
3933 if (pctx->errcode) {
3936 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3937 ctx->flags |= E2F_FLAG_ABORT;
3942 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
3945 ea_block_quota->blocks = quota_blocks;
3946 ea_block_quota->inodes = quota_inodes;
3948 inc_ea_inode_refs(ctx, pctx, first, end);
3949 ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
3950 ea_refcount_store(ctx->refcount_orig, blk, header->h_refcount);
3952 * It might be racy that this block has been merged in the
3955 if (!is_blocks_used(ctx, blk, 1))
3956 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, blk);
3957 ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
3962 region_free(region);
3963 ext2fs_file_acl_block_set(fs, inode, 0);
3964 e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
3968 /* Returns 1 if bad htree, 0 if OK */
3969 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
3970 ext2_ino_t ino, struct ext2_inode *inode,
3973 struct ext2_dx_root_info *root;
3974 ext2_filsys fs = ctx->fs;
3978 if ((!LINUX_S_ISDIR(inode->i_mode) &&
3979 fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
3980 (!ext2fs_has_feature_dir_index(fs->super) &&
3981 fix_problem(ctx, PR_1_HTREE_SET, pctx)))
3984 pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
3986 if ((pctx->errcode) ||
3988 (blk < fs->super->s_first_data_block) ||
3989 (blk >= ext2fs_blocks_count(fs->super))) {
3990 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3996 retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
3997 if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
4000 /* XXX should check that beginning matches a directory */
4001 root = (struct ext2_dx_root_info *) (block_buf + 24);
4003 if ((root->reserved_zero || root->info_length < 8) &&
4004 fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
4007 pctx->num = root->hash_version;
4008 if ((root->hash_version != EXT2_HASH_LEGACY) &&
4009 (root->hash_version != EXT2_HASH_HALF_MD4) &&
4010 (root->hash_version != EXT2_HASH_TEA) &&
4011 (root->hash_version != EXT2_HASH_SIPHASH) &&
4012 fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
4015 if (ext4_hash_in_dirent(inode)) {
4016 if (root->hash_version != EXT2_HASH_SIPHASH &&
4017 fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
4020 if (root->hash_version == EXT2_HASH_SIPHASH &&
4021 fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
4025 if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
4026 fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
4029 pctx->num = root->indirect_levels;
4030 /* if htree level is clearly too high, consider it to be broken */
4031 if (root->indirect_levels > EXT4_HTREE_LEVEL &&
4032 fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
4035 /* if level is only maybe too high, LARGE_DIR feature could be unset */
4036 if (root->indirect_levels > ext2_dir_htree_level(fs) &&
4037 !ext2fs_has_feature_largedir(fs->super)) {
4038 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
4039 int idx_pb = 1 << (blockbits - 3);
4041 /* compare inode size/blocks vs. max-sized 2-level htree */
4042 if (EXT2_I_SIZE(pctx->inode) <
4043 (idx_pb - 1) * (idx_pb - 2) << blockbits &&
4044 pctx->inode->i_blocks <
4045 (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
4046 fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
4050 if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
4051 ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4057 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
4058 struct ext2_inode *inode, int restart_flag,
4062 inode->i_links_count = 0;
4063 ext2fs_icount_store(ctx->inode_link_info, ino, 0);
4064 inode->i_dtime = ctx->now;
4067 * If a special inode has such rotten block mappings that we
4068 * want to clear the whole inode, be sure to actually zap
4069 * the block maps because i_links_count isn't checked for
4070 * special inodes, and we'll end up right back here the next
4073 if (ino < EXT2_FIRST_INODE(ctx->fs->super))
4074 memset(inode->i_block, 0, sizeof(inode->i_block));
4076 ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
4077 ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
4078 if (ctx->inode_reg_map)
4079 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
4080 if (ctx->inode_bad_map)
4081 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
4084 * If the inode was partially accounted for before processing
4085 * was aborted, we need to restart the pass 1 scan.
4087 ctx->flags |= restart_flag;
4089 if (ino == EXT2_BAD_INO)
4090 memset(inode, 0, sizeof(struct ext2_inode));
4092 e2fsck_write_inode(ctx, ino, inode, source);
4096 * Use the multiple-blocks reclamation code to fix alignment problems in
4097 * a bigalloc filesystem. We want a logical cluster to map to *only* one
4098 * physical cluster, and we want the block offsets within that cluster to
4101 static int has_unaligned_cluster_map(e2fsck_t ctx,
4102 blk64_t last_pblk, blk64_t last_lblk,
4103 blk64_t pblk, blk64_t lblk)
4105 blk64_t cluster_mask;
4107 if (!ctx->fs->cluster_ratio_bits)
4109 cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
4112 * If the block in the logical cluster doesn't align with the block in
4113 * the physical cluster...
4115 if ((lblk & cluster_mask) != (pblk & cluster_mask))
4119 * If we cross a physical cluster boundary within a logical cluster...
4121 if (last_pblk && (lblk & cluster_mask) != 0 &&
4122 EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
4123 EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
4129 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
4130 struct process_block_struct *pb,
4131 blk64_t start_block, blk64_t end_block,
4133 ext2_extent_handle_t ehandle,
4136 struct ext2fs_extent extent;
4137 blk64_t blk, last_lblk;
4139 int is_dir, is_leaf;
4141 struct ext2_extent_info info;
4142 int failed_csum = 0;
4144 if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
4147 pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
4150 if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
4151 !pb->eti.force_rebuild) {
4152 struct extent_tree_level *etl;
4154 etl = pb->eti.ext_info + info.curr_level;
4155 etl->num_extents += info.num_entries;
4156 etl->max_extents += info.max_entries;
4158 * Implementation wart: Splitting extent blocks when appending
4159 * will leave the old block with one free entry. Therefore
4160 * unless the node is totally full, pretend that a non-root
4161 * extent block can hold one fewer entry than it actually does,
4162 * so that we don't repeatedly rebuild the extent tree.
4164 if (info.curr_level && info.num_entries < info.max_entries)
4168 pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
4170 while ((pctx->errcode == 0 ||
4171 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
4172 info.num_entries-- > 0) {
4173 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
4174 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
4175 last_lblk = extent.e_lblk + extent.e_len - 1;
4178 pctx->blk = extent.e_pblk;
4179 pctx->blk2 = extent.e_lblk;
4180 pctx->num = extent.e_len;
4181 pctx->blkcount = extent.e_lblk + extent.e_len;
4183 if (extent.e_pblk == 0 ||
4184 extent.e_pblk < ctx->fs->super->s_first_data_block ||
4185 extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
4186 problem = PR_1_EXTENT_BAD_START_BLK;
4187 else if (extent.e_lblk < start_block)
4188 problem = PR_1_OUT_OF_ORDER_EXTENTS;
4189 else if ((end_block && last_lblk > end_block) &&
4190 !(last_lblk > eof_block &&
4191 ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
4192 (pctx->inode->i_flags & EXT4_VERITY_FL))))
4193 problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
4194 else if (is_leaf && extent.e_len == 0)
4195 problem = PR_1_EXTENT_LENGTH_ZERO;
4197 (extent.e_pblk + extent.e_len) >
4198 ext2fs_blocks_count(ctx->fs->super))
4199 problem = PR_1_EXTENT_ENDS_BEYOND;
4200 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
4201 !ext2fs_has_feature_largedir(ctx->fs->super) &&
4202 ((extent.e_lblk + extent.e_len) >
4203 (1U << (21 - ctx->fs->super->s_log_block_size))))
4204 problem = PR_1_TOOBIG_DIR;
4206 if (is_leaf && problem == 0 && extent.e_len > 0) {
4208 printf("extent_region(ino=%u, expect=%llu, "
4209 "lblk=%llu, len=%u)\n", pb->ino,
4210 (unsigned long long) pb->next_lblock,
4211 (unsigned long long) extent.e_lblk,
4214 if (extent.e_lblk < pb->next_lblock)
4215 problem = PR_1_EXTENT_COLLISION;
4216 else if (extent.e_lblk + extent.e_len > pb->next_lblock)
4217 pb->next_lblock = extent.e_lblk + extent.e_len;
4221 * Uninitialized blocks in a directory? Clear the flag and
4222 * we'll interpret the blocks later.
4224 if (try_repairs && is_dir && problem == 0 &&
4225 (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
4226 fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
4227 e2fsck_pass1_fix_lock(ctx);
4228 extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
4229 pb->inode_modified = 1;
4230 pctx->errcode = ext2fs_extent_replace(ehandle, 0,
4232 e2fsck_pass1_fix_unlock(ctx);
4237 #ifdef CONFIG_DEVELOPER_FEATURES
4238 if (try_repairs && !is_dir && problem == 0 &&
4239 (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
4240 (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
4241 fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
4242 extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
4243 pb->inode_modified = 1;
4244 pctx->errcode = ext2fs_extent_replace(ehandle, 0,
4251 if (try_repairs && problem) {
4253 if (fix_problem(ctx, problem, pctx)) {
4254 if (ctx->invalid_bitmaps) {
4256 * If fsck knows the bitmaps are bad,
4257 * skip to the next extent and
4258 * try to clear this extent again
4259 * after fixing the bitmaps, by
4262 pctx->errcode = ext2fs_extent_get(
4264 EXT2_EXTENT_NEXT_SIB,
4266 ctx->flags |= E2F_FLAG_RESTART_LATER;
4267 if (pctx->errcode ==
4268 EXT2_ET_NO_CURRENT_NODE) {
4274 e2fsck_pass1_fix_lock(ctx);
4275 e2fsck_read_bitmaps(ctx);
4276 pb->inode_modified = 1;
4278 ext2fs_extent_delete(ehandle, 0);
4279 e2fsck_pass1_fix_unlock(ctx);
4280 if (pctx->errcode) {
4281 pctx->str = "ext2fs_extent_delete";
4284 e2fsck_pass1_fix_lock(ctx);
4285 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
4286 e2fsck_pass1_fix_unlock(ctx);
4287 if (pctx->errcode &&
4288 pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
4289 pctx->str = "ext2fs_extent_fix_parents";
4292 pctx->errcode = ext2fs_extent_get(ehandle,
4293 EXT2_EXTENT_CURRENT,
4295 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
4306 blk64_t lblk = extent.e_lblk;
4307 int next_try_repairs = 1;
4309 blk = extent.e_pblk;
4312 * If this lower extent block collides with critical
4313 * metadata, don't try to repair the damage. Pass 1b
4314 * will reallocate the block; then we can try again.
4316 if (pb->ino != EXT2_RESIZE_INO &&
4317 extent.e_pblk < ctx->fs->super->s_blocks_count &&
4318 ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4320 next_try_repairs = 0;
4323 PR_1_CRITICAL_METADATA_COLLISION,
4325 if ((ctx->options & E2F_OPT_NO) == 0)
4326 ctx->flags |= E2F_FLAG_RESTART_LATER;
4328 pctx->errcode = ext2fs_extent_get(ehandle,
4329 EXT2_EXTENT_DOWN, &extent);
4330 if (pctx->errcode &&
4331 pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
4332 pctx->str = "EXT2_EXTENT_DOWN";
4333 problem = PR_1_EXTENT_HEADER_INVALID;
4334 if (!next_try_repairs)
4336 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
4337 goto report_problem;
4340 /* The next extent should match this index's logical start */
4341 if (extent.e_lblk != lblk) {
4342 struct ext2_extent_info e_info;
4344 pctx->errcode = ext2fs_extent_get_info(ehandle,
4346 if (pctx->errcode) {
4347 pctx->str = "ext2fs_extent_get_info";
4351 pctx->blk2 = extent.e_lblk;
4352 pctx->num = e_info.curr_level - 1;
4353 problem = PR_1_EXTENT_INDEX_START_INVALID;
4354 if (fix_problem(ctx, problem, pctx)) {
4355 e2fsck_pass1_fix_lock(ctx);
4356 pb->inode_modified = 1;
4358 ext2fs_extent_fix_parents(ehandle);
4359 e2fsck_pass1_fix_unlock(ctx);
4360 if (pctx->errcode) {
4361 pctx->str = "ext2fs_extent_fix_parents";
4366 scan_extent_node(ctx, pctx, pb, extent.e_lblk,
4367 last_lblk, eof_block, ehandle,
4371 pctx->errcode = ext2fs_extent_get(ehandle,
4372 EXT2_EXTENT_UP, &extent);
4373 if (pctx->errcode) {
4374 pctx->str = "EXT2_EXTENT_UP";
4377 mark_block_used(ctx, blk);
4382 if ((pb->previous_block != 0) &&
4383 (pb->previous_block+1 != extent.e_pblk)) {
4384 if (ctx->options & E2F_OPT_FRAGCHECK) {
4389 else if (pb->is_reg)
4392 printf(("%6lu(%c): expecting %6lu "
4394 "phys %6lu log %lu len %lu\n"),
4395 (unsigned long) pctx->ino, type,
4396 (unsigned long) pb->previous_block+1,
4397 (unsigned long) extent.e_pblk,
4398 (unsigned long) extent.e_lblk,
4399 (unsigned long) extent.e_len);
4404 * If we notice a gap in the logical block mappings of an
4405 * extent-mapped directory, offer to close the hole by
4406 * moving the logical block down, otherwise we'll go mad in
4407 * pass 3 allocating empty directory blocks to fill the hole.
4409 if (try_repairs && is_dir &&
4410 pb->last_block + 1 < extent.e_lblk) {
4413 new_lblk = pb->last_block + 1;
4414 if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
4415 new_lblk = ((new_lblk +
4416 EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
4417 ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
4419 EXT2FS_CLUSTER_MASK(ctx->fs));
4420 pctx->blk = extent.e_lblk;
4421 pctx->blk2 = new_lblk;
4422 if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
4423 e2fsck_pass1_fix_lock(ctx);
4424 extent.e_lblk = new_lblk;
4425 pb->inode_modified = 1;
4426 pctx->errcode = ext2fs_extent_replace(ehandle,
4428 e2fsck_pass1_fix_unlock(ctx);
4429 if (pctx->errcode) {
4433 e2fsck_pass1_fix_lock(ctx);
4434 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
4435 e2fsck_pass1_fix_unlock(ctx);
4437 goto failed_add_dir_block;
4438 pctx->errcode = ext2fs_extent_goto(ehandle,
4441 goto failed_add_dir_block;
4442 last_lblk = extent.e_lblk + extent.e_len - 1;
4448 while (++pb->last_db_block <
4449 (e2_blkcnt_t) extent.e_lblk) {
4450 pctx->errcode = ext2fs_add_dir_block2(
4454 if (pctx->errcode) {
4456 pctx->num = pb->last_db_block;
4457 goto failed_add_dir_block;
4461 for (i = 0; i < extent.e_len; i++) {
4462 pctx->errcode = ext2fs_add_dir_block2(
4467 if (pctx->errcode) {
4468 pctx->blk = extent.e_pblk + i;
4469 pctx->num = extent.e_lblk + i;
4470 failed_add_dir_block:
4471 fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4472 /* Should never get here */
4473 ctx->flags |= E2F_FLAG_ABORT;
4477 if (extent.e_len > 0)
4478 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
4480 if (has_unaligned_cluster_map(ctx, pb->previous_block,
4484 for (i = 0; i < extent.e_len; i++) {
4485 pctx->blk = extent.e_lblk + i;
4486 pctx->blk2 = extent.e_pblk + i;
4487 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4488 mark_block_used(ctx, extent.e_pblk + i);
4489 mark_block_used(ctx, extent.e_pblk + i);
4494 * Check whether first cluster got marked in previous iteration.
4496 if (ctx->fs->cluster_ratio_bits &&
4497 pb->previous_block &&
4498 (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
4499 EXT2FS_B2C(ctx->fs, pb->previous_block)))
4500 /* Set blk to the beginning of next cluster. */
4503 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
4505 /* Set blk to the beginning of current cluster. */
4506 blk = EXT2FS_C2B(ctx->fs,
4507 EXT2FS_B2C(ctx->fs, extent.e_pblk));
4509 if (blk < extent.e_pblk + extent.e_len) {
4510 mark_blocks_used(ctx, blk,
4511 extent.e_pblk + extent.e_len - blk);
4512 n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
4513 EXT2FS_CLUSTER_RATIO(ctx->fs));
4514 pb->num_blocks += n;
4516 pb->last_block = extent.e_lblk + extent.e_len - 1;
4517 pb->previous_block = extent.e_pblk + extent.e_len - 1;
4518 start_block = pb->last_block = last_lblk;
4519 if (is_leaf && !is_dir &&
4520 !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
4521 pb->last_init_lblock = last_lblk;
4523 pctx->errcode = ext2fs_extent_get(ehandle,
4524 EXT2_EXTENT_NEXT_SIB,
4528 /* Failed csum but passes checks? Ask to fix checksum. */
4530 fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
4531 e2fsck_pass1_fix_lock(ctx);
4532 pb->inode_modified = 1;
4533 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
4534 e2fsck_pass1_fix_unlock(ctx);
4539 if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
4543 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
4544 struct process_block_struct *pb)
4546 struct ext2_extent_info info;
4547 struct ext2_inode *inode = pctx->inode;
4548 ext2_extent_handle_t ehandle;
4549 ext2_filsys fs = ctx->fs;
4550 ext2_ino_t ino = pctx->ino;
4553 struct ext3_extent_header *eh;
4555 /* Check for a proper extent header... */
4556 eh = (struct ext3_extent_header *) &inode->i_block[0];
4557 retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
4559 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
4560 e2fsck_clear_inode(ctx, ino, inode, 0,
4561 "check_blocks_extents");
4566 /* ...since this function doesn't fail if i_block is zeroed. */
4567 pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
4568 if (pctx->errcode) {
4569 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
4570 e2fsck_clear_inode(ctx, ino, inode, 0,
4571 "check_blocks_extents");
4576 retval = ext2fs_extent_get_info(ehandle, &info);
4578 int max_depth = info.max_depth;
4580 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
4581 max_depth = MAX_EXTENT_DEPTH_COUNT-1;
4582 ctx->extent_depth_count[max_depth]++;
4585 /* Check maximum extent depth */
4586 pctx->blk = info.max_depth;
4587 pctx->blk2 = ext2fs_max_extent_depth(ehandle);
4588 if (pctx->blk2 < pctx->blk &&
4589 fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
4590 pb->eti.force_rebuild = 1;
4592 /* Can we collect extent tree level stats? */
4593 pctx->blk = MAX_EXTENT_DEPTH_COUNT;
4594 if (pctx->blk2 > pctx->blk)
4595 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
4596 memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
4597 pb->eti.ino = pb->ino;
4599 pb->next_lblock = 0;
4601 eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
4602 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
4603 scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
4604 if (pctx->errcode &&
4605 fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
4607 inode->i_blocks = 0;
4608 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4609 "check_blocks_extents");
4612 ext2fs_extent_free(ehandle);
4614 /* Rebuild unless it's a dir and we're rehashing it */
4615 if (LINUX_S_ISDIR(inode->i_mode) &&
4616 e2fsck_dir_will_be_rehashed(ctx, ino))
4619 if (ctx->options & E2F_OPT_CONVERT_BMAP)
4620 e2fsck_rebuild_extents_later(ctx, ino);
4622 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
4626 * In fact we don't need to check blocks for an inode with inline data
4627 * because this inode doesn't have any blocks. In this function all
4628 * we need to do is add this inode into dblist when it is a directory.
4630 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
4631 struct process_block_struct *pb)
4634 size_t inline_data_size = 0;
4641 /* Process the dirents in i_block[] as the "first" block. */
4642 pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
4646 /* Process the dirents in the EA as a "second" block. */
4647 flags = ctx->fs->flags;
4648 ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4649 pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
4651 ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4652 (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4653 if (pctx->errcode) {
4658 if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
4661 pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
4669 fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4670 ctx->flags |= E2F_FLAG_ABORT;
4674 * This subroutine is called on each inode to account for all of the
4675 * blocks used by that inode.
4677 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
4678 char *block_buf, const struct ea_quota *ea_ibody_quota)
4680 ext2_filsys fs = ctx->fs;
4681 struct process_block_struct pb;
4682 ext2_ino_t ino = pctx->ino;
4683 struct ext2_inode *inode = pctx->inode;
4684 unsigned bad_size = 0;
4685 int dirty_inode = 0;
4689 struct ea_quota ea_block_quota;
4692 pb.num_blocks = EXT2FS_B2C(ctx->fs,
4693 ea_ibody_quota ? ea_ibody_quota->blocks : 0);
4695 pb.last_init_lblock = -1;
4696 pb.last_db_block = -1;
4697 pb.num_illegal_blocks = 0;
4698 pb.suppress = 0; pb.clear = 0;
4701 pb.previous_block = 0;
4702 pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
4703 pb.is_reg = LINUX_S_ISREG(inode->i_mode);
4704 pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
4708 pb.inode_modified = 0;
4709 pb.eti.force_rebuild = 0;
4713 extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
4714 inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
4716 if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
4717 if (e2fsck_should_abort(ctx))
4719 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
4722 if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
4723 check_blocks_inline_data(ctx, pctx, &pb);
4724 else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
4725 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
4726 check_blocks_extents(ctx, pctx, &pb);
4730 * If we've modified the inode, write it out before
4731 * iterate() tries to use it.
4734 e2fsck_write_inode(ctx, ino, inode,
4739 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4740 pctx->errcode = ext2fs_block_iterate3(fs, ino,
4741 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
4742 block_buf, process_block, &pb);
4744 * We do not have uninitialized extents in non extent
4747 pb.last_init_lblock = pb.last_block;
4749 * If iterate() changed a block mapping, we have to
4750 * re-read the inode. If we decide to clear the
4751 * inode after clearing some stuff, we'll re-write the
4752 * bad mappings into the inode!
4754 if (pb.inode_modified)
4755 e2fsck_read_inode(ctx, ino, inode,
4757 fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4758 (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4760 if (ctx->options & E2F_OPT_CONVERT_BMAP) {
4762 printf("bmap rebuild ino=%d\n", ino);
4764 if (!LINUX_S_ISDIR(inode->i_mode) ||
4765 !e2fsck_dir_will_be_rehashed(ctx, ino))
4766 e2fsck_rebuild_extents_later(ctx, ino);
4770 end_problem_latch(ctx, PR_LATCH_BLOCK);
4771 end_problem_latch(ctx, PR_LATCH_TOOBIG);
4772 if (e2fsck_should_abort(ctx))
4775 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
4777 if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
4778 if (LINUX_S_ISDIR(inode->i_mode))
4779 ctx->fs_fragmented_dir++;
4781 ctx->fs_fragmented++;
4785 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4790 if (inode->i_flags & EXT2_INDEX_FL) {
4791 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
4792 inode->i_flags &= ~EXT2_INDEX_FL;
4795 e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
4799 if (!pb.num_blocks && pb.is_dir &&
4800 !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
4801 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
4802 e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
4803 ctx->fs_directory_count--;
4808 if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
4809 (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
4810 !(inode->i_flags & EXT4_EA_INODE_FL)) {
4811 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
4813 pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
4814 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
4815 ino, (ea_ibody_quota ?
4816 ea_ibody_quota->inodes : 0) +
4817 ea_block_quota.inodes + 1);
4820 if (!ext2fs_has_feature_huge_file(fs->super) ||
4821 !(inode->i_flags & EXT4_HUGE_FILE_FL))
4822 pb.num_blocks *= (fs->blocksize / 512);
4823 pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
4825 printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
4826 ino, inode->i_size, (unsigned long long) pb.last_block,
4827 (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
4828 (unsigned long long) pb.num_blocks);
4830 size = EXT2_I_SIZE(inode);
4832 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
4833 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
4838 flags = ctx->fs->flags;
4839 ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4840 err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
4842 ctx->fs->flags = (flags &
4843 EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4845 ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4846 if (err || sz != size) {
4850 } else if (size & (fs->blocksize - 1))
4852 else if (nblock > (pb.last_block + 1))
4854 else if (nblock < (pb.last_block + 1)) {
4855 if (((pb.last_block + 1) - nblock) >
4856 fs->super->s_prealloc_dir_blocks)
4860 if ((pb.last_init_lblock >= 0) &&
4861 /* Do not allow initialized allocated blocks past i_size*/
4862 (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
4863 !(inode->i_flags & EXT4_VERITY_FL))
4865 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4866 size > ext2_max_sizes[fs->super->s_log_block_size])
4867 /* too big for a direct/indirect-mapped file */
4869 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4871 ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
4872 /* too big for an extent-based file - 32bit ee_block */
4875 /* i_size for symlinks is checked elsewhere */
4876 if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
4877 /* Did inline_data set pctx->num earlier? */
4879 pctx->num = (pb.last_block + 1) * fs->blocksize;
4880 pctx->group = bad_size;
4881 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
4882 ext2fs_inode_size_set(fs, inode, pctx->num);
4883 if (EXT2_I_SIZE(inode) == 0 &&
4884 (inode->i_flags & EXT4_INLINE_DATA_FL)) {
4885 memset(inode->i_block, 0,
4886 sizeof(inode->i_block));
4887 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
4893 if (LINUX_S_ISREG(inode->i_mode) &&
4894 ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4896 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
4897 ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
4898 (ext2fs_has_feature_huge_file(fs->super) &&
4899 (inode->i_flags & EXT4_HUGE_FILE_FL) &&
4900 (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
4901 pctx->num = pb.num_blocks;
4902 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
4903 inode->i_blocks = pb.num_blocks;
4904 inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
4911 * The kernel gets mad if we ask it to allocate bigalloc clusters to
4912 * a block mapped file, so rebuild it as an extent file. We can skip
4913 * symlinks because they're never rewritten.
4915 if (ext2fs_has_feature_bigalloc(fs->super) &&
4916 (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
4917 ext2fs_inode_data_blocks2(fs, inode) > 0 &&
4918 (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
4919 !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
4920 fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
4921 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
4926 if (ctx->dirs_to_hash && pb.is_dir &&
4927 !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
4928 !(inode->i_flags & EXT2_INDEX_FL) &&
4929 ((inode->i_size / fs->blocksize) >= 3))
4930 e2fsck_rehash_dir_later(ctx, ino);
4934 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
4939 * Helper function called by process block when an illegal block is
4940 * found. It returns a description about why the block is illegal
4942 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
4946 static char problem[80];
4948 super = fs->super->s_first_data_block;
4949 strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
4950 if (block < super) {
4951 sprintf(problem, "< FIRSTBLOCK (%u)", super);
4953 } else if (block >= ext2fs_blocks_count(fs->super)) {
4954 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
4957 for (i = 0; i < fs->group_desc_count; i++) {
4958 if (block == super) {
4959 sprintf(problem, "is the superblock in group %d", i);
4962 if (block > super &&
4963 block <= (super + fs->desc_blocks)) {
4964 sprintf(problem, "is in the group descriptors "
4968 if (block == ext2fs_block_bitmap_loc(fs, i)) {
4969 sprintf(problem, "is the block bitmap of group %d", i);
4972 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
4973 sprintf(problem, "is the inode bitmap of group %d", i);
4976 if (block >= ext2fs_inode_table_loc(fs, i) &&
4977 (block < ext2fs_inode_table_loc(fs, i)
4978 + fs->inode_blocks_per_group)) {
4979 sprintf(problem, "is in the inode table of group %d",
4983 super += fs->super->s_blocks_per_group;
4990 * This is a helper function for check_blocks().
4992 static int process_block(ext2_filsys fs,
4994 e2_blkcnt_t blockcnt,
4995 blk64_t ref_block EXT2FS_ATTR((unused)),
4996 int ref_offset EXT2FS_ATTR((unused)),
4999 struct process_block_struct *p;
5000 struct problem_context *pctx;
5001 blk64_t blk = *block_nr;
5003 problem_t problem = 0;
5006 p = (struct process_block_struct *) priv_data;
5011 * For a directory, add logical block zero for processing even if it's
5012 * not mapped or we'll be perennially stuck with broken "." and ".."
5015 if (p->is_dir && blockcnt == 0 && blk == 0) {
5016 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
5017 if (pctx->errcode) {
5019 pctx->num = blockcnt;
5020 goto failed_add_dir_block;
5029 printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
5034 * Simplistic fragmentation check. We merely require that the
5035 * file be contiguous. (Which can never be true for really
5036 * big files that are greater than a block group.)
5038 if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
5039 if (p->previous_block+1 != blk) {
5040 if (ctx->options & E2F_OPT_FRAGCHECK) {
5048 printf(_("%6lu(%c): expecting %6lu "
5049 "got phys %6lu (blkcnt %lld)\n"),
5050 (unsigned long) pctx->ino, type,
5051 (unsigned long) p->previous_block+1,
5052 (unsigned long) blk,
5053 (long long) blockcnt);
5059 if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
5060 !pctx->inode->i_size_high &&
5061 blockcnt > (1 << (21 - fs->super->s_log_block_size)))
5062 problem = PR_1_TOOBIG_DIR;
5063 if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
5064 problem = PR_1_TOOBIG_DIR;
5065 if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
5066 problem = PR_1_TOOBIG_REG;
5067 if (!p->is_dir && !p->is_reg && blockcnt > 0)
5068 problem = PR_1_TOOBIG_SYMLINK;
5070 if (blk < fs->super->s_first_data_block ||
5071 blk >= ext2fs_blocks_count(fs->super))
5072 problem = PR_1_ILLEGAL_BLOCK_NUM;
5075 * If this IND/DIND/TIND block is squatting atop some critical metadata
5076 * (group descriptors, superblock, bitmap, inode table), any write to
5077 * "fix" mapping problems will destroy the metadata. We'll let pass 1b
5078 * fix that and restart fsck.
5081 p->ino != EXT2_RESIZE_INO &&
5082 blk < ctx->fs->super->s_blocks_count &&
5083 ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
5085 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
5086 if ((ctx->options & E2F_OPT_NO) == 0)
5087 ctx->flags |= E2F_FLAG_RESTART_LATER;
5091 p->num_illegal_blocks++;
5093 * A bit of subterfuge here -- we're trying to fix a block
5094 * mapping, but the IND/DIND/TIND block could have collided
5095 * with some critical metadata. So, fix the in-core mapping so
5096 * iterate won't go insane, but return 0 instead of
5097 * BLOCK_CHANGED so that it won't write the remapping out to
5098 * our multiply linked block.
5100 * Even if we previously determined that an *IND block
5101 * conflicts with critical metadata, we must still try to
5102 * iterate the *IND block as if it is an *IND block to find and
5103 * mark the blocks it points to. Better to be overly cautious
5104 * with the used_blocks map so that we don't move the *IND
5105 * block to a block that's really in use!
5107 if (p->ino != EXT2_RESIZE_INO &&
5109 ext2fs_test_block_bitmap2(ctx->block_metadata_map,
5114 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
5115 if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
5119 if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
5121 set_latch_flags(PR_LATCH_BLOCK,
5126 pctx->blkcount = blockcnt;
5127 if (fix_problem(ctx, problem, pctx)) {
5128 blk = *block_nr = 0;
5129 ret_code = BLOCK_CHANGED;
5130 p->inode_modified = 1;
5132 * If the directory block is too big and is beyond the
5133 * end of the FS, don't bother trying to add it for
5134 * processing -- the kernel would never have created a
5135 * directory this large, and we risk an ENOMEM abort.
5136 * In any case, the toobig handler for extent-based
5137 * directories also doesn't feed toobig blocks to
5140 if (problem == PR_1_TOOBIG_DIR)
5147 if (p->ino == EXT2_RESIZE_INO) {
5149 * The resize inode has already be sanity checked
5150 * during pass #0 (the superblock checks). All we
5151 * have to do is mark the double indirect block as
5152 * being in use; all of the other blocks are handled
5153 * by mark_table_blocks()).
5155 if (blockcnt == BLOCK_COUNT_DIND)
5156 mark_block_used(ctx, blk);
5158 } else if (!(ctx->fs->cluster_ratio_bits &&
5159 p->previous_block &&
5160 (EXT2FS_B2C(ctx->fs, blk) ==
5161 EXT2FS_B2C(ctx->fs, p->previous_block)) &&
5162 (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
5163 ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
5164 mark_block_used(ctx, blk);
5166 } else if (has_unaligned_cluster_map(ctx, p->previous_block,
5167 p->last_block, blk, blockcnt)) {
5168 pctx->blk = blockcnt;
5170 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
5171 mark_block_used(ctx, blk);
5172 mark_block_used(ctx, blk);
5175 p->last_block = blockcnt;
5176 p->previous_block = blk;
5178 if (p->is_dir && (blockcnt >= 0)) {
5179 while (++p->last_db_block < blockcnt) {
5180 pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
5183 if (pctx->errcode) {
5185 pctx->num = p->last_db_block;
5186 goto failed_add_dir_block;
5189 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
5191 if (pctx->errcode) {
5193 pctx->num = blockcnt;
5194 failed_add_dir_block:
5195 fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
5196 /* Should never get here */
5197 ctx->flags |= E2F_FLAG_ABORT;
5204 static int process_bad_block(ext2_filsys fs,
5206 e2_blkcnt_t blockcnt,
5207 blk64_t ref_block EXT2FS_ATTR((unused)),
5208 int ref_offset EXT2FS_ATTR((unused)),
5211 struct process_block_struct *p;
5212 blk64_t blk = *block_nr;
5213 blk64_t first_block;
5215 struct problem_context *pctx;
5221 p = (struct process_block_struct *) priv_data;
5225 pctx->ino = EXT2_BAD_INO;
5227 pctx->blkcount = blockcnt;
5229 if ((blk < fs->super->s_first_data_block) ||
5230 (blk >= ext2fs_blocks_count(fs->super))) {
5231 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
5233 return BLOCK_CHANGED;
5239 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
5241 if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
5243 return BLOCK_CHANGED;
5245 } else if (is_blocks_used(ctx, blk, 1)) {
5247 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
5250 return BLOCK_CHANGED;
5252 if (e2fsck_should_abort(ctx))
5255 mark_block_used(ctx, blk);
5260 printf ("DEBUG: Marking %u as bad.\n", blk);
5262 ctx->fs_badblocks_count++;
5264 * If the block is not used, then mark it as used and return.
5265 * If it is already marked as found, this must mean that
5266 * there's an overlap between the filesystem table blocks
5267 * (bitmaps and inode table) and the bad block list.
5269 if (!is_blocks_used(ctx, blk, 1)) {
5270 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
5274 * Try to find the where the filesystem block was used...
5276 first_block = fs->super->s_first_data_block;
5278 for (i = 0; i < fs->group_desc_count; i++ ) {
5281 if (!ext2fs_bg_has_super(fs, i))
5283 if (blk == first_block) {
5285 if (fix_problem(ctx,
5286 PR_1_BAD_PRIMARY_SUPERBLOCK,
5289 return BLOCK_CHANGED;
5293 fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
5296 if ((blk > first_block) &&
5297 (blk <= first_block + fs->desc_blocks)) {
5299 pctx->blk = *block_nr;
5300 if (fix_problem(ctx,
5301 PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
5303 return BLOCK_CHANGED;
5307 fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
5311 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
5312 if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
5313 ctx->invalid_block_bitmap_flag[i]++;
5314 ctx->invalid_bitmaps++;
5318 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
5319 if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
5320 ctx->invalid_inode_bitmap_flag[i]++;
5321 ctx->invalid_bitmaps++;
5325 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
5326 (blk < (ext2fs_inode_table_loc(fs, i) +
5327 fs->inode_blocks_per_group))) {
5329 * If there are bad blocks in the inode table,
5330 * the inode scan code will try to do
5331 * something reasonable automatically.
5335 first_block += fs->super->s_blocks_per_group;
5338 * If we've gotten to this point, then the only
5339 * possibility is that the bad block inode meta data
5340 * is using a bad block.
5342 if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
5343 (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
5344 (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
5346 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
5348 return BLOCK_CHANGED;
5350 if (e2fsck_should_abort(ctx))
5357 /* Warn user that the block wasn't claimed */
5358 fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
5363 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
5364 const char *name, int num, blk64_t *new_block)
5366 ext2_filsys fs = ctx->fs;
5368 blk64_t old_block = *new_block;
5371 unsigned flexbg_size;
5374 struct problem_context pctx;
5376 clear_problem_context(&pctx);
5379 pctx.blk = old_block;
5383 * For flex_bg filesystems, first try to allocate the metadata
5384 * within the flex_bg, and if that fails then try finding the
5385 * space anywhere in the filesystem.
5387 is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
5389 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
5390 flexbg = group / flexbg_size;
5391 first_block = ext2fs_group_first_block2(fs,
5392 flexbg_size * flexbg);
5393 last_grp = group | (flexbg_size - 1);
5394 if (last_grp >= fs->group_desc_count)
5395 last_grp = fs->group_desc_count - 1;
5396 last_block = ext2fs_group_last_block2(fs, last_grp);
5398 last_block = ext2fs_group_last_block2(fs, group);
5399 pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
5400 num, ctx->block_found_map,
5402 if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
5403 pctx.errcode = ext2fs_get_free_blocks2(fs,
5404 fs->super->s_first_data_block,
5405 ext2fs_blocks_count(fs->super),
5406 num, ctx->block_found_map, new_block);
5409 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
5410 ext2fs_unmark_valid(fs);
5411 ctx->flags |= E2F_FLAG_ABORT;
5414 pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
5416 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
5417 ext2fs_unmark_valid(fs);
5418 ctx->flags |= E2F_FLAG_ABORT;
5421 ext2fs_mark_super_dirty(fs);
5422 fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
5423 pctx.blk2 = *new_block;
5424 fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
5425 PR_1_RELOC_TO), &pctx);
5427 for (i = 0; i < num; i++) {
5429 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
5431 pctx.errcode = io_channel_read_blk64(fs->io,
5432 old_block + i, 1, buf);
5434 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
5435 pctx.blk = (*new_block) + i;
5436 pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
5439 pctx.blk = (*new_block) + i;
5440 pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
5445 fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
5447 ext2fs_free_mem(&buf);
5451 * This routine gets called at the end of pass 1 if bad blocks are
5452 * detected in the superblock, group descriptors, inode_bitmaps, or
5453 * block bitmaps. At this point, all of the blocks have been mapped
5454 * out, so we can try to allocate new block(s) to replace the bad
5457 static void handle_fs_bad_blocks(e2fsck_t ctx)
5459 ext2_filsys fs = ctx->fs;
5461 blk64_t first_block;
5464 for (i = 0; i < fs->group_desc_count; i++) {
5465 first_block = ext2fs_group_first_block2(fs, i);
5467 if (ctx->invalid_block_bitmap_flag[i]) {
5468 new_blk = ext2fs_block_bitmap_loc(fs, i);
5469 new_table_block(ctx, first_block, i, _("block bitmap"),
5471 ext2fs_block_bitmap_loc_set(fs, i, new_blk);
5473 if (ctx->invalid_inode_bitmap_flag[i]) {
5474 new_blk = ext2fs_inode_bitmap_loc(fs, i);
5475 new_table_block(ctx, first_block, i, _("inode bitmap"),
5477 ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
5479 if (ctx->invalid_inode_table_flag[i]) {
5480 new_blk = ext2fs_inode_table_loc(fs, i);
5481 new_table_block(ctx, first_block, i, _("inode table"),
5482 fs->inode_blocks_per_group,
5484 ext2fs_inode_table_loc_set(fs, i, new_blk);
5485 ctx->flags |= E2F_FLAG_RESTART;
5488 ctx->invalid_bitmaps = 0;
5492 * This routine marks all blocks which are used by the superblock,
5493 * group descriptors, inode bitmaps, and block bitmaps.
5495 static void mark_table_blocks(e2fsck_t ctx)
5497 ext2_filsys fs = ctx->fs;
5501 struct problem_context pctx;
5503 clear_problem_context(&pctx);
5505 for (i = 0; i < fs->group_desc_count; i++) {
5508 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
5509 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
5512 * Mark the blocks used for the inode table
5514 if (ext2fs_inode_table_loc(fs, i)) {
5515 for (j = 0, b = ext2fs_inode_table_loc(fs, i);
5516 j < fs->inode_blocks_per_group;
5518 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5521 if (!ctx->invalid_inode_table_flag[i] &&
5523 PR_1_ITABLE_CONFLICT, &pctx)) {
5524 ctx->invalid_inode_table_flag[i]++;
5525 ctx->invalid_bitmaps++;
5528 ext2fs_mark_block_bitmap2(
5529 ctx->block_found_map, b);
5530 ext2fs_mark_block_bitmap2(
5531 ctx->block_metadata_map, b);
5537 * Mark block used for the block bitmap
5539 if (ext2fs_block_bitmap_loc(fs, i)) {
5540 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5541 ext2fs_block_bitmap_loc(fs, i))) {
5542 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
5543 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
5544 ctx->invalid_block_bitmap_flag[i]++;
5545 ctx->invalid_bitmaps++;
5548 ext2fs_mark_block_bitmap2(ctx->block_found_map,
5549 ext2fs_block_bitmap_loc(fs, i));
5550 ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5551 ext2fs_block_bitmap_loc(fs, i));
5555 * Mark block used for the inode bitmap
5557 if (ext2fs_inode_bitmap_loc(fs, i)) {
5558 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5559 ext2fs_inode_bitmap_loc(fs, i))) {
5560 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
5561 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
5562 ctx->invalid_inode_bitmap_flag[i]++;
5563 ctx->invalid_bitmaps++;
5566 ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5567 ext2fs_inode_bitmap_loc(fs, i));
5568 ext2fs_mark_block_bitmap2(ctx->block_found_map,
5569 ext2fs_inode_bitmap_loc(fs, i));
5576 * These subroutines short circuits ext2fs_get_blocks and
5577 * ext2fs_check_directory; we use them since we already have the inode
5578 * structure, so there's no point in letting the ext2fs library read
5581 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
5584 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5587 if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5588 return EXT2_ET_CALLBACK_NOTHANDLED;
5590 for (i=0; i < EXT2_N_BLOCKS; i++)
5591 blocks[i] = ctx->stashed_inode->i_block[i];
5595 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
5596 struct ext2_inode *inode)
5598 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5600 if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5601 return EXT2_ET_CALLBACK_NOTHANDLED;
5602 *inode = *ctx->stashed_inode;
5606 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
5607 struct ext2_inode *inode)
5609 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5611 if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
5612 (inode != ctx->stashed_inode))
5613 *ctx->stashed_inode = *inode;
5614 return EXT2_ET_CALLBACK_NOTHANDLED;
5617 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
5619 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5621 if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5622 return EXT2_ET_CALLBACK_NOTHANDLED;
5624 if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
5625 return EXT2_ET_NO_DIRECTORY;
5629 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
5632 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5636 if (ctx->block_found_map) {
5637 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
5641 if (fs->block_map) {
5642 ext2fs_mark_block_bitmap2(fs->block_map, new_block);
5643 ext2fs_mark_bb_dirty(fs);
5646 if (!fs->block_map) {
5647 retval = ext2fs_read_block_bitmap(fs);
5652 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
5661 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
5662 blk64_t len, blk64_t *pblk, blk64_t *plen)
5664 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5667 if (ctx->block_found_map)
5668 return ext2fs_new_range(fs, flags, goal, len,
5669 ctx->block_found_map, pblk, plen);
5671 if (!fs->block_map) {
5672 retval = ext2fs_read_block_bitmap(fs);
5677 return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
5681 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
5683 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5685 /* Never free a critical metadata block */
5686 if (ctx->block_found_map &&
5687 ctx->block_metadata_map &&
5689 ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
5692 if (ctx->block_found_map) {
5694 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
5696 ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
5700 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
5701 blk_t num, int inuse)
5703 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5705 /* Never free a critical metadata block */
5706 if (ctx->block_found_map &&
5707 ctx->block_metadata_map &&
5709 ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
5712 if (ctx->block_found_map) {
5714 ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
5717 ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
5722 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
5724 ext2_filsys fs = ctx->fs;
5726 if (use_shortcuts) {
5727 fs->get_blocks = pass1_get_blocks;
5728 fs->check_directory = pass1_check_directory;
5729 fs->read_inode = pass1_read_inode;
5730 fs->write_inode = pass1_write_inode;
5731 ctx->stashed_ino = 0;
5734 fs->check_directory = 0;
5736 fs->write_inode = 0;
5740 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
5742 ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
5743 ext2fs_set_block_alloc_stats_callback(ctx->fs,
5744 e2fsck_block_alloc_stats, 0);
5745 ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
5746 ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
5747 e2fsck_block_alloc_stats_range, NULL);