1 /*
2 * pass1.c -- pass #1 of e2fsck: sequential scan of the inode table
3 *
4 * Copyright (C) 1993, 1994, 1995, 1996, 1997 Theodore Ts'o.
5 *
6 * %Begin-Header%
7 * This file may be redistributed under the terms of the GNU Public
8 * License.
9 * %End-Header%
10 *
11 * Pass 1 of e2fsck iterates over all the inodes in the filesystems,
12 * and applies the following tests to each inode:
13 *
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
17 *
18 * Pass 1 also gathers the collects the following information:
19 *
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)
32 *
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
37 * inode to fix it.)
38 *
39 * Note that pass 1B will be invoked if there are any duplicate blocks
40 * found.
41 */
42
43 #define _GNU_SOURCE 1 /* get strnlen() */
44 #include "config.h"
45 #include <string.h>
46 #include <time.h>
47 #ifdef HAVE_ERRNO_H
48 #include <errno.h>
49 #endif
50
51 #include "e2fsck.h"
52 #include <ext2fs/ext2_ext_attr.h>
53 #include <e2p/e2p.h>
54
55 #include "problem.h"
56
57 #ifdef NO_INLINE_FUNCS
58 #define _INLINE_
59 #else
60 #define _INLINE_ inline
61 #endif
62
63 #undef DEBUG
64
65 struct ea_quota {
66 blk64_t blocks;
67 __u64 inodes;
68 };
69
70 static int process_block(ext2_filsys fs, blk64_t *blocknr,
71 e2_blkcnt_t blockcnt, blk64_t ref_blk,
72 int ref_offset, void *priv_data);
73 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
74 e2_blkcnt_t blockcnt, blk64_t ref_blk,
75 int ref_offset, void *priv_data);
76 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
77 char *block_buf,
78 const struct ea_quota *ea_ibody_quota);
79 static void mark_table_blocks(e2fsck_t ctx);
80 static void alloc_bb_map(e2fsck_t ctx);
81 static void alloc_imagic_map(e2fsck_t ctx);
82 static void mark_inode_bad(e2fsck_t ctx, ext2_ino_t ino);
83 static void add_casefolded_dir(e2fsck_t ctx, ext2_ino_t ino);
84 static void handle_fs_bad_blocks(e2fsck_t ctx);
85 static void process_inodes(e2fsck_t ctx, char *block_buf);
86 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
87 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
88 dgrp_t group, void * priv_data);
89 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
90 char *block_buf, int adjust_sign);
91 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
92
93 struct process_block_struct {
94 ext2_ino_t ino;
95 unsigned is_dir:1, is_reg:1, clear:1, suppress:1,
96 fragmented:1, compressed:1, bbcheck:1,
97 inode_modified:1;
98 blk64_t num_blocks;
99 blk64_t max_blocks;
100 blk64_t last_block;
101 e2_blkcnt_t last_init_lblock;
102 e2_blkcnt_t last_db_block;
103 int num_illegal_blocks;
104 blk64_t previous_block;
105 struct ext2_inode *inode;
106 struct problem_context *pctx;
107 ext2fs_block_bitmap fs_meta_blocks;
108 e2fsck_t ctx;
109 blk64_t next_lblock;
110 struct extent_tree_info eti;
111 };
112
113 struct process_inode_block {
114 ext2_ino_t ino;
115 struct ea_quota ea_ibody_quota;
116 struct ext2_inode_large inode;
117 };
118
119 struct scan_callback_struct {
120 e2fsck_t ctx;
121 char *block_buf;
122 };
123
124 /*
125 * For the inodes to process list.
126 */
127 static struct process_inode_block *inodes_to_process;
128 static int process_inode_count;
129
130 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
131 EXT2_MIN_BLOCK_LOG_SIZE + 1];
132
133 /*
134 * Check to make sure a device inode is real. Returns 1 if the device
135 * checks out, 0 if not.
136 *
137 * Note: this routine is now also used to check FIFO's and Sockets,
138 * since they have the same requirement; the i_block fields should be
139 * zero.
140 */
e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR ((unused)),struct ext2_inode * inode)141 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
142 struct ext2_inode *inode)
143 {
144 int i;
145
146 /*
147 * If the index or extents flag is set, then this is a bogus
148 * device/fifo/socket
149 */
150 if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
151 return 0;
152
153 /*
154 * We should be able to do the test below all the time, but
155 * because the kernel doesn't forcibly clear the device
156 * inode's additional i_block fields, there are some rare
157 * occasions when a legitimate device inode will have non-zero
158 * additional i_block fields. So for now, we only complain
159 * when the immutable flag is set, which should never happen
160 * for devices. (And that's when the problem is caused, since
161 * you can't set or clear immutable flags for devices.) Once
162 * the kernel has been fixed we can change this...
163 */
164 if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
165 for (i=4; i < EXT2_N_BLOCKS; i++)
166 if (inode->i_block[i])
167 return 0;
168 }
169 return 1;
170 }
171
172 /*
173 * Check to make sure a symlink inode is real. Returns 1 if the symlink
174 * checks out, 0 if not.
175 */
e2fsck_pass1_check_symlink(ext2_filsys fs,ext2_ino_t ino,struct ext2_inode * inode,char * buf)176 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
177 struct ext2_inode *inode, char *buf)
178 {
179 unsigned int buflen;
180 unsigned int len;
181
182 if ((inode->i_size_high || inode->i_size == 0) ||
183 (inode->i_flags & EXT2_INDEX_FL))
184 return 0;
185
186 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
187 size_t inline_size;
188
189 if (inode->i_flags & EXT4_EXTENTS_FL)
190 return 0;
191 if (ext2fs_inline_data_size(fs, ino, &inline_size))
192 return 0;
193 if (inode->i_size != inline_size)
194 return 0;
195
196 return 1;
197 }
198
199 if (ext2fs_is_fast_symlink(inode)) {
200 if (inode->i_flags & EXT4_EXTENTS_FL)
201 return 0;
202 buf = (char *)inode->i_block;
203 buflen = sizeof(inode->i_block);
204 } else {
205 ext2_extent_handle_t handle;
206 struct ext2_extent_info info;
207 struct ext2fs_extent extent;
208 blk64_t blk;
209 int i;
210
211 if (inode->i_flags & EXT4_EXTENTS_FL) {
212 if (ext2fs_extent_open2(fs, ino, inode, &handle))
213 return 0;
214 if (ext2fs_extent_get_info(handle, &info) ||
215 (info.num_entries != 1) ||
216 (info.max_depth != 0)) {
217 ext2fs_extent_free(handle);
218 return 0;
219 }
220 if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
221 &extent) ||
222 (extent.e_lblk != 0) ||
223 (extent.e_len != 1)) {
224 ext2fs_extent_free(handle);
225 return 0;
226 }
227 blk = extent.e_pblk;
228 ext2fs_extent_free(handle);
229 } else {
230 blk = inode->i_block[0];
231
232 for (i = 1; i < EXT2_N_BLOCKS; i++)
233 if (inode->i_block[i])
234 return 0;
235 }
236
237 if (blk < fs->super->s_first_data_block ||
238 blk >= ext2fs_blocks_count(fs->super))
239 return 0;
240
241 if (io_channel_read_blk64(fs->io, blk, 1, buf))
242 return 0;
243
244 buflen = fs->blocksize;
245 }
246
247 if (inode->i_flags & EXT4_ENCRYPT_FL)
248 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
249 else
250 len = strnlen(buf, buflen);
251
252 if (len >= buflen)
253 return 0;
254
255 if (len != inode->i_size)
256 return 0;
257 return 1;
258 }
259
260 /*
261 * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
262 */
263 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
check_extents_inlinedata(e2fsck_t ctx,struct problem_context * pctx)264 static void check_extents_inlinedata(e2fsck_t ctx,
265 struct problem_context *pctx)
266 {
267 if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
268 return;
269
270 if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
271 return;
272
273 pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
274 e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
275 }
276 #undef BAD_SPECIAL_FLAGS
277
278 /*
279 * If the immutable (or append-only) flag is set on the inode, offer
280 * to clear it.
281 */
282 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
check_immutable(e2fsck_t ctx,struct problem_context * pctx)283 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
284 {
285 if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
286 return;
287
288 if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
289 return;
290
291 pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
292 e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
293 }
294
295 /*
296 * If device, fifo or socket, check size is zero -- if not offer to
297 * clear it
298 */
check_size(e2fsck_t ctx,struct problem_context * pctx)299 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
300 {
301 struct ext2_inode *inode = pctx->inode;
302
303 if (EXT2_I_SIZE(inode) == 0)
304 return;
305
306 if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
307 return;
308
309 ext2fs_inode_size_set(ctx->fs, inode, 0);
310 e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
311 }
312
313 /*
314 * For a given size, calculate how many blocks would be charged towards quota.
315 */
size_to_quota_blocks(ext2_filsys fs,size_t size)316 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
317 {
318 blk64_t clusters;
319
320 clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
321 return EXT2FS_C2B(fs, clusters);
322 }
323
324 /*
325 * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
326 * the problem code.
327 */
check_large_ea_inode(e2fsck_t ctx,struct ext2_ext_attr_entry * entry,struct problem_context * pctx,blk64_t * quota_blocks)328 static problem_t check_large_ea_inode(e2fsck_t ctx,
329 struct ext2_ext_attr_entry *entry,
330 struct problem_context *pctx,
331 blk64_t *quota_blocks)
332 {
333 struct ext2_inode inode;
334 __u32 hash, signed_hash;
335 errcode_t retval;
336
337 /* Check if inode is within valid range */
338 if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
339 (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
340 pctx->num = entry->e_value_inum;
341 return PR_1_ATTR_VALUE_EA_INODE;
342 }
343
344 e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
345
346 retval = ext2fs_ext_attr_hash_entry3(ctx->fs, entry, NULL, &hash,
347 &signed_hash);
348 if (retval) {
349 com_err("check_large_ea_inode", retval,
350 _("while hashing entry with e_value_inum = %u"),
351 entry->e_value_inum);
352 fatal_error(ctx, 0);
353 }
354
355 if ((hash == entry->e_hash) || (signed_hash == entry->e_hash)) {
356 *quota_blocks = size_to_quota_blocks(ctx->fs,
357 entry->e_value_size);
358 } else {
359 /* This might be an old Lustre-style ea_inode reference. */
360 if (inode.i_mtime == pctx->ino &&
361 inode.i_generation == pctx->inode->i_generation) {
362 *quota_blocks = 0;
363 } else {
364 /* If target inode is also missing EA_INODE flag,
365 * this is likely to be a bad reference.
366 */
367 if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
368 pctx->num = entry->e_value_inum;
369 return PR_1_ATTR_VALUE_EA_INODE;
370 } else {
371 pctx->num = entry->e_hash;
372 return PR_1_ATTR_HASH;
373 }
374 }
375 }
376
377 if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
378 pctx->num = entry->e_value_inum;
379 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
380 inode.i_flags |= EXT4_EA_INODE_FL;
381 ext2fs_write_inode(ctx->fs, entry->e_value_inum,
382 &inode);
383 } else {
384 return PR_1_ATTR_NO_EA_INODE_FL;
385 }
386 }
387 return 0;
388 }
389
inc_ea_inode_refs(e2fsck_t ctx,struct problem_context * pctx,struct ext2_ext_attr_entry * first,void * end)390 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
391 struct ext2_ext_attr_entry *first, void *end)
392 {
393 struct ext2_ext_attr_entry *entry = first;
394 struct ext2_ext_attr_entry *np = EXT2_EXT_ATTR_NEXT(entry);
395
396 while ((void *) entry < end && (void *) np < end &&
397 !EXT2_EXT_IS_LAST_ENTRY(entry)) {
398 if (!entry->e_value_inum)
399 goto next;
400 if (!ctx->ea_inode_refs) {
401 pctx->errcode = ea_refcount_create(0,
402 &ctx->ea_inode_refs);
403 if (pctx->errcode) {
404 pctx->num = 4;
405 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
406 ctx->flags |= E2F_FLAG_ABORT;
407 return;
408 }
409 }
410 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
411 0);
412 next:
413 entry = np;
414 np = EXT2_EXT_ATTR_NEXT(entry);
415 }
416 }
417
check_ea_in_inode(e2fsck_t ctx,struct problem_context * pctx,struct ea_quota * ea_ibody_quota)418 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
419 struct ea_quota *ea_ibody_quota)
420 {
421 struct ext2_super_block *sb = ctx->fs->super;
422 struct ext2_inode_large *inode;
423 struct ext2_ext_attr_entry *entry;
424 char *start, *header, *end;
425 unsigned int storage_size, remain;
426 problem_t problem = 0;
427 region_t region = 0;
428
429 ea_ibody_quota->blocks = 0;
430 ea_ibody_quota->inodes = 0;
431
432 inode = (struct ext2_inode_large *) pctx->inode;
433 storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
434 inode->i_extra_isize;
435 header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
436 inode->i_extra_isize;
437 end = header + storage_size;
438 start = header + sizeof(__u32);
439 entry = (struct ext2_ext_attr_entry *) start;
440
441 /* scan all entry's headers first */
442
443 /* take finish entry 0UL into account */
444 remain = storage_size - sizeof(__u32);
445
446 region = region_create(0, storage_size);
447 if (!region) {
448 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
449 problem = 0;
450 ctx->flags |= E2F_FLAG_ABORT;
451 return;
452 }
453 if (region_allocate(region, 0, sizeof(__u32))) {
454 problem = PR_1_INODE_EA_ALLOC_COLLISION;
455 goto fix;
456 }
457
458 while (remain >= sizeof(struct ext2_ext_attr_entry) &&
459 !EXT2_EXT_IS_LAST_ENTRY(entry)) {
460 __u32 hash;
461
462 if (region_allocate(region, (char *)entry - (char *)header,
463 EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
464 problem = PR_1_INODE_EA_ALLOC_COLLISION;
465 goto fix;
466 }
467
468 /* header eats this space */
469 remain -= sizeof(struct ext2_ext_attr_entry);
470
471 /* is attribute name valid? */
472 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
473 pctx->num = entry->e_name_len;
474 problem = PR_1_ATTR_NAME_LEN;
475 goto fix;
476 }
477
478 /* attribute len eats this space */
479 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
480
481 if (entry->e_value_inum == 0) {
482 /* check value size */
483 if (entry->e_value_size > remain) {
484 pctx->num = entry->e_value_size;
485 problem = PR_1_ATTR_VALUE_SIZE;
486 goto fix;
487 }
488
489 if (entry->e_value_size &&
490 region_allocate(region,
491 sizeof(__u32) + entry->e_value_offs,
492 EXT2_EXT_ATTR_SIZE(
493 entry->e_value_size))) {
494 problem = PR_1_INODE_EA_ALLOC_COLLISION;
495 goto fix;
496 }
497
498 hash = ext2fs_ext_attr_hash_entry(entry,
499 start + entry->e_value_offs);
500 if (entry->e_hash != 0 && entry->e_hash != hash)
501 hash = ext2fs_ext_attr_hash_entry_signed(entry,
502 start + entry->e_value_offs);
503
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;
508 goto fix;
509 }
510 } else {
511 blk64_t quota_blocks;
512
513 problem = check_large_ea_inode(ctx, entry, pctx,
514 "a_blocks);
515 if (problem != 0)
516 goto fix;
517
518 ea_ibody_quota->blocks += quota_blocks;
519 ea_ibody_quota->inodes++;
520 }
521
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;
526
527 entry = EXT2_EXT_ATTR_NEXT(entry);
528 }
529
530 if (region_allocate(region, (char *)entry - (char *)header,
531 sizeof(__u32))) {
532 problem = PR_1_INODE_EA_ALLOC_COLLISION;
533 goto fix;
534 }
535 fix:
536 if (region)
537 region_free(region);
538 /*
539 * it seems like a corruption. it's very unlikely we could repair
540 * EA(s) in automatic fashion -bzzz
541 */
542 if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
543 inc_ea_inode_refs(ctx, pctx,
544 (struct ext2_ext_attr_entry *)start, end);
545 return;
546 }
547
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;
554 }
555
check_inode_extra_negative_epoch(__u32 xtime,__u32 extra)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;
559 }
560
561 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
562 check_inode_extra_negative_epoch(inode->i_##xtime, \
563 inode->i_##xtime##_extra)
564
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.
568 */
569 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
570
check_inode_extra_space(e2fsck_t ctx,struct problem_context * pctx,struct ea_quota * ea_ibody_quota)571 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
572 struct ea_quota *ea_ibody_quota)
573 {
574 struct ext2_super_block *sb = ctx->fs->super;
575 struct ext2_inode_large *inode;
576 __u32 *eamagic;
577 int min, max;
578
579 ea_ibody_quota->blocks = 0;
580 ea_ibody_quota->inodes = 0;
581
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 */
585 return;
586 }
587
588 #if 0
589 printf("inode #%u, i_extra_size %d\n", pctx->ino,
590 inode->i_extra_isize);
591 #endif
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;
595 /*
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
598 */
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))
603 return;
604 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
605 inode->i_extra_isize = sb->s_want_extra_isize;
606 else
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");
610 }
611
612 /* check if there is no place for an EA header */
613 if (inode->i_extra_isize >= max - sizeof(__u32))
614 return;
615
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);
621 }
622
623 /*
624 * If the inode's extended atime (ctime, crtime, mtime) is stored in
625 * the old, invalid format, repair it.
626 */
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))) {
634
635 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
636 return;
637
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");
648 }
649
650 }
651
652 /*
653 * Check to see if the inode might really be a directory, despite i_mode
654 *
655 * This is a lot of complexity for something for which I'm not really
656 * convinced happens frequently in the wild. If for any reason this
657 * causes any problems, take this code out.
658 * [tytso:20070331.0827EDT]
659 */
check_is_really_dir(e2fsck_t ctx,struct problem_context * pctx,char * buf)660 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
661 char *buf)
662 {
663 struct ext2_inode *inode = pctx->inode;
664 struct ext2_dir_entry *dirent;
665 errcode_t retval;
666 blk64_t blk;
667 unsigned int i, rec_len, not_device = 0;
668 int extent_fs;
669 int inlinedata_fs;
670
671 /*
672 * If the mode looks OK, we believe it. If the first block in
673 * the i_block array is 0, this cannot be a directory. If the
674 * inode is extent-mapped, it is still the case that the latter
675 * cannot be 0 - the magic number in the extent header would make
676 * it nonzero.
677 */
678 if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
679 LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
680 return;
681
682 /*
683 * Check the block numbers in the i_block array for validity:
684 * zero blocks are skipped (but the first one cannot be zero -
685 * see above), other blocks are checked against the first and
686 * max data blocks (from the the superblock) and against the
687 * block bitmap. Any invalid block found means this cannot be
688 * a directory.
689 *
690 * If there are non-zero blocks past the fourth entry, then
691 * this cannot be a device file: we remember that for the next
692 * check.
693 *
694 * For extent mapped files, we don't do any sanity checking:
695 * just try to get the phys block of logical block 0 and run
696 * with it.
697 *
698 * For inline data files, we just try to get the size of inline
699 * data. If it's true, we will treat it as a directory.
700 */
701
702 extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
703 inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
704 if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
705 size_t size;
706 __u32 dotdot;
707 unsigned int rec_len2;
708 struct ext2_dir_entry de;
709
710 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
711 return;
712 /*
713 * If the size isn't a multiple of 4, it's probably not a
714 * directory??
715 */
716 if (size & 3)
717 return;
718 /*
719 * If the first 10 bytes don't look like a directory entry,
720 * it's probably not a directory.
721 */
722 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
723 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
724 EXT2_DIR_REC_LEN(0));
725 dotdot = ext2fs_le32_to_cpu(dotdot);
726 de.inode = ext2fs_le32_to_cpu(de.inode);
727 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
728 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
729 if (dotdot >= ctx->fs->super->s_inodes_count ||
730 (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
731 dotdot != EXT2_ROOT_INO) ||
732 de.inode >= ctx->fs->super->s_inodes_count ||
733 (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
734 de.inode != 0) ||
735 rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
736 EXT4_INLINE_DATA_DOTDOT_SIZE)
737 return;
738 /* device files never have a "system.data" entry */
739 goto isdir;
740 } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
741 /* extent mapped */
742 if (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
743 &blk))
744 return;
745 /* device files are never extent mapped */
746 not_device++;
747 } else {
748 for (i=0; i < EXT2_N_BLOCKS; i++) {
749 blk = inode->i_block[i];
750 if (!blk)
751 continue;
752 if (i >= 4)
753 not_device++;
754
755 if (blk < ctx->fs->super->s_first_data_block ||
756 blk >= ext2fs_blocks_count(ctx->fs->super) ||
757 ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
758 blk))
759 return; /* Invalid block, can't be dir */
760 }
761 blk = inode->i_block[0];
762 }
763
764 /*
765 * If the mode says this is a device file and the i_links_count field
766 * is sane and we have not ruled it out as a device file previously,
767 * we declare it a device file, not a directory.
768 */
769 if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
770 (inode->i_links_count == 1) && !not_device)
771 return;
772
773 /* read the first block */
774 ehandler_operation(_("reading directory block"));
775 retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
776 ehandler_operation(0);
777 if (retval)
778 return;
779
780 dirent = (struct ext2_dir_entry *) buf;
781 retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
782 if (retval)
783 return;
784 if ((ext2fs_dirent_name_len(dirent) != 1) ||
785 (dirent->name[0] != '.') ||
786 (dirent->inode != pctx->ino) ||
787 (rec_len < 12) ||
788 (rec_len % 4) ||
789 (rec_len >= ctx->fs->blocksize - 12))
790 return;
791
792 dirent = (struct ext2_dir_entry *) (buf + rec_len);
793 retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
794 if (retval)
795 return;
796 if ((ext2fs_dirent_name_len(dirent) != 2) ||
797 (dirent->name[0] != '.') ||
798 (dirent->name[1] != '.') ||
799 (rec_len < 12) ||
800 (rec_len % 4))
801 return;
802
803 isdir:
804 if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
805 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
806 e2fsck_write_inode_full(ctx, pctx->ino, inode,
807 EXT2_INODE_SIZE(ctx->fs->super),
808 "check_is_really_dir");
809 }
810 }
811
e2fsck_setup_icount(e2fsck_t ctx,const char * icount_name,int flags,ext2_icount_t hint,ext2_icount_t * ret)812 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
813 int flags, ext2_icount_t hint,
814 ext2_icount_t *ret)
815 {
816 unsigned int threshold;
817 unsigned int save_type;
818 ext2_ino_t num_dirs;
819 errcode_t retval;
820 char *tdb_dir;
821 int enable;
822
823 *ret = 0;
824
825 profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
826 &tdb_dir);
827 profile_get_uint(ctx->profile, "scratch_files",
828 "numdirs_threshold", 0, 0, &threshold);
829 profile_get_boolean(ctx->profile, "scratch_files",
830 "icount", 0, 1, &enable);
831
832 retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
833 if (retval)
834 num_dirs = 1024; /* Guess */
835
836 if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
837 (!threshold || num_dirs > threshold)) {
838 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
839 flags, ret);
840 if (retval == 0)
841 return 0;
842 }
843 e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
844 &save_type);
845 if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
846 flags |= EXT2_ICOUNT_OPT_FULLMAP;
847 retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
848 ctx->fs->default_bitmap_type = save_type;
849 return retval;
850 }
851
recheck_bad_inode_checksum(ext2_filsys fs,ext2_ino_t ino,e2fsck_t ctx,struct problem_context * pctx)852 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
853 e2fsck_t ctx,
854 struct problem_context *pctx)
855 {
856 errcode_t retval;
857 struct ext2_inode_large inode;
858
859 /*
860 * Reread inode. If we don't see checksum error, then this inode
861 * has been fixed elsewhere.
862 */
863 ctx->stashed_ino = 0;
864 retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
865 sizeof(inode));
866 if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
867 return retval;
868 if (!retval)
869 return 0;
870
871 /*
872 * Checksum still doesn't match. That implies that the inode passes
873 * all the sanity checks, so maybe the checksum is simply corrupt.
874 * See if the user will go for fixing that.
875 */
876 if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
877 return 0;
878
879 retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
880 sizeof(inode));
881 return retval;
882 }
883
reserve_block_for_root_repair(e2fsck_t ctx)884 static void reserve_block_for_root_repair(e2fsck_t ctx)
885 {
886 blk64_t blk = 0;
887 errcode_t err;
888 ext2_filsys fs = ctx->fs;
889
890 ctx->root_repair_block = 0;
891 if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
892 return;
893
894 err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
895 if (err)
896 return;
897 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
898 ctx->root_repair_block = blk;
899 }
900
reserve_block_for_lnf_repair(e2fsck_t ctx)901 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
902 {
903 blk64_t blk = 0;
904 errcode_t err;
905 ext2_filsys fs = ctx->fs;
906 static const char name[] = "lost+found";
907 ext2_ino_t ino;
908
909 ctx->lnf_repair_block = 0;
910 if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
911 return;
912
913 err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
914 if (err)
915 return;
916 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
917 ctx->lnf_repair_block = blk;
918 }
919
get_inline_data_ea_size(ext2_filsys fs,ext2_ino_t ino,size_t * sz)920 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
921 size_t *sz)
922 {
923 void *p;
924 struct ext2_xattr_handle *handle;
925 errcode_t retval;
926
927 retval = ext2fs_xattrs_open(fs, ino, &handle);
928 if (retval)
929 return retval;
930
931 retval = ext2fs_xattrs_read(handle);
932 if (retval)
933 goto err;
934
935 retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
936 if (retval)
937 goto err;
938 ext2fs_free_mem(&p);
939 err:
940 (void) ext2fs_xattrs_close(&handle);
941 return retval;
942 }
943
finish_processing_inode(e2fsck_t ctx,ext2_ino_t ino,struct problem_context * pctx,int failed_csum)944 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
945 struct problem_context *pctx,
946 int failed_csum)
947 {
948 if (!failed_csum)
949 return;
950
951 /*
952 * If the inode failed the checksum and the user didn't
953 * clear the inode, test the checksum again -- if it still
954 * fails, ask the user if the checksum should be corrected.
955 */
956 pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
957 if (pctx->errcode)
958 ctx->flags |= E2F_FLAG_ABORT;
959 }
960 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
961 do { \
962 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
963 if ((ctx)->flags & E2F_FLAG_ABORT) \
964 return; \
965 } while (0)
966
could_be_block_map(ext2_filsys fs,struct ext2_inode * inode)967 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
968 {
969 __u32 x;
970 int i;
971
972 for (i = 0; i < EXT2_N_BLOCKS; i++) {
973 x = inode->i_block[i];
974 #ifdef WORDS_BIGENDIAN
975 x = ext2fs_swab32(x);
976 #endif
977 if (x >= ext2fs_blocks_count(fs->super))
978 return 0;
979 }
980
981 return 1;
982 }
983
984 /*
985 * Figure out what to do with an inode that has both extents and inline data
986 * inode flags set. Returns -1 if we decide to erase the inode, 0 otherwise.
987 */
fix_inline_data_extents_file(e2fsck_t ctx,ext2_ino_t ino,struct ext2_inode * inode,int inode_size,struct problem_context * pctx)988 static int fix_inline_data_extents_file(e2fsck_t ctx,
989 ext2_ino_t ino,
990 struct ext2_inode *inode,
991 int inode_size,
992 struct problem_context *pctx)
993 {
994 size_t max_inline_ea_size;
995 ext2_filsys fs = ctx->fs;
996 int dirty = 0;
997
998 /* Both feature flags not set? Just run the regular checks */
999 if (!ext2fs_has_feature_extents(fs->super) &&
1000 !ext2fs_has_feature_inline_data(fs->super))
1001 return 0;
1002
1003 /* Clear both flags if it's a special file */
1004 if (LINUX_S_ISCHR(inode->i_mode) ||
1005 LINUX_S_ISBLK(inode->i_mode) ||
1006 LINUX_S_ISFIFO(inode->i_mode) ||
1007 LINUX_S_ISSOCK(inode->i_mode)) {
1008 check_extents_inlinedata(ctx, pctx);
1009 return 0;
1010 }
1011
1012 /* If it looks like an extent tree, try to clear inlinedata */
1013 if (ext2fs_extent_header_verify(inode->i_block,
1014 sizeof(inode->i_block)) == 0 &&
1015 fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1016 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1017 dirty = 1;
1018 goto out;
1019 }
1020
1021 /* If it looks short enough to be inline data, try to clear extents */
1022 if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1023 max_inline_ea_size = inode_size -
1024 (EXT2_GOOD_OLD_INODE_SIZE +
1025 ((struct ext2_inode_large *)inode)->i_extra_isize);
1026 else
1027 max_inline_ea_size = 0;
1028 if (EXT2_I_SIZE(inode) <
1029 EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1030 fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1031 inode->i_flags &= ~EXT4_EXTENTS_FL;
1032 dirty = 1;
1033 goto out;
1034 }
1035
1036 /*
1037 * Too big for inline data, but no evidence of extent tree -
1038 * maybe it's a block map file? If the mappings all look valid?
1039 */
1040 if (could_be_block_map(fs, inode) &&
1041 fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1042 #ifdef WORDS_BIGENDIAN
1043 int i;
1044
1045 for (i = 0; i < EXT2_N_BLOCKS; i++)
1046 inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1047 #endif
1048
1049 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1050 dirty = 1;
1051 goto out;
1052 }
1053
1054 /* Oh well, just clear the busted inode. */
1055 if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1056 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1057 return -1;
1058 }
1059
1060 out:
1061 if (dirty)
1062 e2fsck_write_inode(ctx, ino, inode, "pass1");
1063
1064 return 0;
1065 }
1066
pass1_readahead(e2fsck_t ctx,dgrp_t * group,ext2_ino_t * next_ino)1067 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1068 {
1069 ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1070 dgrp_t start = *group, grp;
1071 blk64_t blocks_to_read = 0;
1072 errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1073
1074 if (ctx->readahead_kb == 0)
1075 goto out;
1076
1077 /* Keep iterating groups until we have enough to readahead */
1078 inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1079 for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1080 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1081 continue;
1082 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1083 ext2fs_bg_itable_unused(ctx->fs, grp);
1084 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1085 inodes_per_block;
1086 if (blocks_to_read * ctx->fs->blocksize >
1087 ctx->readahead_kb * 1024)
1088 break;
1089 }
1090
1091 err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1092 grp - start + 1);
1093 if (err == EAGAIN) {
1094 ctx->readahead_kb /= 2;
1095 err = 0;
1096 }
1097
1098 out:
1099 if (err) {
1100 /* Error; disable itable readahead */
1101 *group = ctx->fs->group_desc_count;
1102 *next_ino = ctx->fs->super->s_inodes_count;
1103 } else {
1104 /*
1105 * Don't do more readahead until we've reached the first inode
1106 * of the last inode scan buffer block for the last group.
1107 */
1108 *group = grp + 1;
1109 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1110 ctx->inode_buffer_blocks :
1111 EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1112 ctx->fs->blocksize /
1113 EXT2_INODE_SIZE(ctx->fs->super);
1114 inodes_in_group--;
1115 *next_ino = inodes_in_group -
1116 (inodes_in_group % inodes_per_buffer) + 1 +
1117 (grp * ctx->fs->super->s_inodes_per_group);
1118 }
1119 }
1120
1121 /*
1122 * Check if the passed ino is one of the used superblock quota inodes.
1123 *
1124 * Before the quota inodes were journaled, older superblock quota inodes
1125 * were just regular files in the filesystem and not reserved inodes. This
1126 * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1127 * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1128 */
quota_inum_is_super(struct ext2_super_block * sb,ext2_ino_t ino)1129 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1130 {
1131 enum quota_type qtype;
1132
1133 for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1134 if (*quota_sb_inump(sb, qtype) == ino)
1135 return 1;
1136
1137 return 0;
1138 }
1139
1140 /*
1141 * Check if the passed ino is one of the reserved quota inodes.
1142 * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1143 * inodes. These inodes may or may not be in use by the quota feature.
1144 */
quota_inum_is_reserved(ext2_filsys fs,ext2_ino_t ino)1145 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1146 {
1147 enum quota_type qtype;
1148
1149 for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1150 if (quota_type2inum(qtype, fs->super) == ino)
1151 return 1;
1152
1153 return 0;
1154 }
1155
e2fsck_pass1(e2fsck_t ctx)1156 void e2fsck_pass1(e2fsck_t ctx)
1157 {
1158 int i;
1159 __u64 max_sizes;
1160 ext2_filsys fs = ctx->fs;
1161 ext2_ino_t ino = 0;
1162 struct ext2_inode *inode = NULL;
1163 ext2_inode_scan scan = NULL;
1164 char *block_buf = NULL;
1165 #ifdef RESOURCE_TRACK
1166 struct resource_track rtrack;
1167 #endif
1168 unsigned char frag, fsize;
1169 struct problem_context pctx;
1170 struct scan_callback_struct scan_struct;
1171 struct ext2_super_block *sb = ctx->fs->super;
1172 const char *old_op;
1173 const char *eop_next_inode = _("getting next inode from scan");
1174 int imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1175 int low_dtime_check = 1;
1176 unsigned int inode_size = EXT2_INODE_SIZE(fs->super);
1177 unsigned int bufsize;
1178 int failed_csum = 0;
1179 ext2_ino_t ino_threshold = 0;
1180 dgrp_t ra_group = 0;
1181 struct ea_quota ea_ibody_quota;
1182
1183 init_resource_track(&rtrack, ctx->fs->io);
1184 clear_problem_context(&pctx);
1185
1186 /* If we can do readahead, figure out how many groups to pull in. */
1187 if (!e2fsck_can_readahead(ctx->fs))
1188 ctx->readahead_kb = 0;
1189 else if (ctx->readahead_kb == ~0ULL)
1190 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1191 pass1_readahead(ctx, &ra_group, &ino_threshold);
1192
1193 if (!(ctx->options & E2F_OPT_PREEN))
1194 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1195
1196 if (ext2fs_has_feature_dir_index(fs->super) &&
1197 !(ctx->options & E2F_OPT_NO)) {
1198 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1199 ctx->dirs_to_hash = 0;
1200 }
1201
1202 #ifdef MTRACE
1203 mtrace_print("Pass 1");
1204 #endif
1205
1206 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1207
1208 for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1209 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1210 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1211 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1212 max_sizes = (max_sizes * (1UL << i));
1213 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1214 }
1215 #undef EXT2_BPP
1216
1217 imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1218 extent_fs = ext2fs_has_feature_extents(sb);
1219 inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1220 casefold_fs = ext2fs_has_feature_casefold(sb);
1221
1222 /*
1223 * Allocate bitmaps structures
1224 */
1225 pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1226 EXT2FS_BMAP64_RBTREE,
1227 "inode_used_map",
1228 &ctx->inode_used_map);
1229 if (pctx.errcode) {
1230 pctx.num = 1;
1231 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1232 ctx->flags |= E2F_FLAG_ABORT;
1233 return;
1234 }
1235 pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1236 _("directory inode map"),
1237 EXT2FS_BMAP64_AUTODIR,
1238 "inode_dir_map", &ctx->inode_dir_map);
1239 if (pctx.errcode) {
1240 pctx.num = 2;
1241 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1242 ctx->flags |= E2F_FLAG_ABORT;
1243 return;
1244 }
1245 pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1246 _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1247 "inode_reg_map", &ctx->inode_reg_map);
1248 if (pctx.errcode) {
1249 pctx.num = 6;
1250 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1251 ctx->flags |= E2F_FLAG_ABORT;
1252 return;
1253 }
1254 pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1255 _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1256 "block_found_map", &ctx->block_found_map);
1257 if (pctx.errcode) {
1258 pctx.num = 1;
1259 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1260 ctx->flags |= E2F_FLAG_ABORT;
1261 return;
1262 }
1263 pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1264 _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1265 "block_metadata_map", &ctx->block_metadata_map);
1266 if (pctx.errcode) {
1267 pctx.num = 1;
1268 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1269 ctx->flags |= E2F_FLAG_ABORT;
1270 return;
1271 }
1272 if (casefold_fs) {
1273 pctx.errcode =
1274 e2fsck_allocate_inode_bitmap(fs,
1275 _("inode casefold map"),
1276 EXT2FS_BMAP64_RBTREE,
1277 "inode_casefold_map",
1278 &ctx->inode_casefold_map);
1279 if (pctx.errcode) {
1280 pctx.num = 1;
1281 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1282 ctx->flags |= E2F_FLAG_ABORT;
1283 return;
1284 }
1285 }
1286 pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1287 &ctx->inode_link_info);
1288 if (pctx.errcode) {
1289 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1290 ctx->flags |= E2F_FLAG_ABORT;
1291 return;
1292 }
1293 bufsize = inode_size;
1294 if (bufsize < sizeof(struct ext2_inode_large))
1295 bufsize = sizeof(struct ext2_inode_large);
1296 inode = (struct ext2_inode *)
1297 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1298
1299 inodes_to_process = (struct process_inode_block *)
1300 e2fsck_allocate_memory(ctx,
1301 (ctx->process_inode_size *
1302 sizeof(struct process_inode_block)),
1303 "array of inodes to process");
1304 process_inode_count = 0;
1305
1306 pctx.errcode = ext2fs_init_dblist(fs, 0);
1307 if (pctx.errcode) {
1308 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1309 ctx->flags |= E2F_FLAG_ABORT;
1310 goto endit;
1311 }
1312
1313 /*
1314 * If the last orphan field is set, clear it, since the pass1
1315 * processing will automatically find and clear the orphans.
1316 * In the future, we may want to try using the last_orphan
1317 * linked list ourselves, but for now, we clear it so that the
1318 * ext3 mount code won't get confused.
1319 */
1320 if (!(ctx->options & E2F_OPT_READONLY)) {
1321 if (fs->super->s_last_orphan) {
1322 fs->super->s_last_orphan = 0;
1323 ext2fs_mark_super_dirty(fs);
1324 }
1325 }
1326
1327 mark_table_blocks(ctx);
1328 pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1329 &ctx->block_found_map);
1330 if (pctx.errcode) {
1331 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1332 ctx->flags |= E2F_FLAG_ABORT;
1333 goto endit;
1334 }
1335 block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1336 "block iterate buffer");
1337 if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1338 e2fsck_use_inode_shortcuts(ctx, 1);
1339 e2fsck_intercept_block_allocations(ctx);
1340 old_op = ehandler_operation(_("opening inode scan"));
1341 pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1342 &scan);
1343 ehandler_operation(old_op);
1344 if (pctx.errcode) {
1345 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1346 ctx->flags |= E2F_FLAG_ABORT;
1347 goto endit;
1348 }
1349 ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1350 EXT2_SF_WARN_GARBAGE_INODES, 0);
1351 ctx->stashed_inode = inode;
1352 scan_struct.ctx = ctx;
1353 scan_struct.block_buf = block_buf;
1354 ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1355 if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1356 ctx->fs->group_desc_count)))
1357 goto endit;
1358 if ((fs->super->s_wtime &&
1359 fs->super->s_wtime < fs->super->s_inodes_count) ||
1360 (fs->super->s_mtime &&
1361 fs->super->s_mtime < fs->super->s_inodes_count) ||
1362 (fs->super->s_mkfs_time &&
1363 fs->super->s_mkfs_time < fs->super->s_inodes_count))
1364 low_dtime_check = 0;
1365
1366 if (ext2fs_has_feature_mmp(fs->super) &&
1367 fs->super->s_mmp_block > fs->super->s_first_data_block &&
1368 fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1369 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1370 fs->super->s_mmp_block);
1371
1372 /* Set up ctx->lost_and_found if possible */
1373 (void) e2fsck_get_lost_and_found(ctx, 0);
1374
1375 while (1) {
1376 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1377 if (e2fsck_mmp_update(fs))
1378 fatal_error(ctx, 0);
1379 }
1380 old_op = ehandler_operation(eop_next_inode);
1381 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1382 inode, inode_size);
1383 if (ino > ino_threshold)
1384 pass1_readahead(ctx, &ra_group, &ino_threshold);
1385 ehandler_operation(old_op);
1386 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1387 goto endit;
1388 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1389 /*
1390 * If badblocks says badblocks is bad, offer to clear
1391 * the list, update the in-core bb list, and restart
1392 * the inode scan.
1393 */
1394 if (ino == EXT2_BAD_INO &&
1395 fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1396 &pctx)) {
1397 errcode_t err;
1398
1399 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1400 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1401 ctx->fs->badblocks = NULL;
1402 err = ext2fs_read_bb_inode(ctx->fs,
1403 &ctx->fs->badblocks);
1404 if (err) {
1405 fix_problem(ctx, PR_1_ISCAN_ERROR,
1406 &pctx);
1407 ctx->flags |= E2F_FLAG_ABORT;
1408 } else
1409 ctx->flags |= E2F_FLAG_RESTART;
1410 goto endit;
1411 }
1412 if (!ctx->inode_bb_map)
1413 alloc_bb_map(ctx);
1414 ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1415 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1416 continue;
1417 }
1418 if (pctx.errcode &&
1419 pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1420 pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1421 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1422 ctx->flags |= E2F_FLAG_ABORT;
1423 goto endit;
1424 }
1425 if (!ino)
1426 break;
1427 pctx.ino = ino;
1428 pctx.inode = inode;
1429 ctx->stashed_ino = ino;
1430
1431 /* Clear trashed inode? */
1432 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1433 inode->i_links_count > 0 &&
1434 fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1435 pctx.errcode = 0;
1436 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1437 }
1438 failed_csum = pctx.errcode != 0;
1439
1440 /*
1441 * Check for inodes who might have been part of the
1442 * orphaned list linked list. They should have gotten
1443 * dealt with by now, unless the list had somehow been
1444 * corrupted.
1445 *
1446 * FIXME: In the future, inodes which are still in use
1447 * (and which are therefore) pending truncation should
1448 * be handled specially. Right now we just clear the
1449 * dtime field, and the normal e2fsck handling of
1450 * inodes where i_size and the inode blocks are
1451 * inconsistent is to fix i_size, instead of releasing
1452 * the extra blocks. This won't catch the inodes that
1453 * was at the end of the orphan list, but it's better
1454 * than nothing. The right answer is that there
1455 * shouldn't be any bugs in the orphan list handling. :-)
1456 */
1457 if (inode->i_dtime && low_dtime_check &&
1458 inode->i_dtime < ctx->fs->super->s_inodes_count) {
1459 if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1460 inode->i_dtime = inode->i_links_count ?
1461 0 : ctx->now;
1462 e2fsck_write_inode(ctx, ino, inode,
1463 "pass1");
1464 failed_csum = 0;
1465 }
1466 }
1467
1468 if (inode->i_links_count) {
1469 pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1470 ino, inode->i_links_count);
1471 if (pctx.errcode) {
1472 pctx.num = inode->i_links_count;
1473 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1474 ctx->flags |= E2F_FLAG_ABORT;
1475 goto endit;
1476 }
1477 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1478 !quota_inum_is_reserved(fs, ino)) {
1479 if (!inode->i_dtime && inode->i_mode) {
1480 if (fix_problem(ctx,
1481 PR_1_ZERO_DTIME, &pctx)) {
1482 inode->i_dtime = ctx->now;
1483 e2fsck_write_inode(ctx, ino, inode,
1484 "pass1");
1485 failed_csum = 0;
1486 }
1487 }
1488 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1489 continue;
1490 }
1491
1492 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1493 ((!LINUX_S_ISDIR(inode->i_mode) &&
1494 fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1495 (!casefold_fs &&
1496 fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1497 inode->i_flags &= ~EXT4_CASEFOLD_FL;
1498 e2fsck_write_inode(ctx, ino, inode, "pass1");
1499 }
1500
1501 /* Conflicting inlinedata/extents inode flags? */
1502 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1503 (inode->i_flags & EXT4_EXTENTS_FL)) {
1504 int res = fix_inline_data_extents_file(ctx, ino, inode,
1505 inode_size,
1506 &pctx);
1507 if (res < 0) {
1508 /* skip FINISH_INODE_LOOP */
1509 continue;
1510 }
1511 }
1512
1513 /* Test for incorrect inline_data flags settings. */
1514 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1515 (ino >= EXT2_FIRST_INODE(fs->super))) {
1516 size_t size = 0;
1517
1518 pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1519 if (!pctx.errcode &&
1520 fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1521 ext2fs_set_feature_inline_data(sb);
1522 ext2fs_mark_super_dirty(fs);
1523 inlinedata_fs = 1;
1524 } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1525 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1526 /* skip FINISH_INODE_LOOP */
1527 continue;
1528 }
1529 }
1530
1531 /* Test for inline data flag but no attr */
1532 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1533 (ino >= EXT2_FIRST_INODE(fs->super))) {
1534 size_t size = 0;
1535 errcode_t err;
1536 int flags;
1537
1538 flags = fs->flags;
1539 if (failed_csum)
1540 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1541 err = get_inline_data_ea_size(fs, ino, &size);
1542 fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1543 (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1544
1545 switch (err) {
1546 case 0:
1547 /* Everything is awesome... */
1548 break;
1549 case EXT2_ET_BAD_EA_BLOCK_NUM:
1550 case EXT2_ET_BAD_EA_HASH:
1551 case EXT2_ET_BAD_EA_HEADER:
1552 case EXT2_ET_EA_BAD_NAME_LEN:
1553 case EXT2_ET_EA_BAD_VALUE_SIZE:
1554 case EXT2_ET_EA_KEY_NOT_FOUND:
1555 case EXT2_ET_EA_NO_SPACE:
1556 case EXT2_ET_MISSING_EA_FEATURE:
1557 case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1558 case EXT2_ET_INLINE_DATA_NO_BLOCK:
1559 case EXT2_ET_INLINE_DATA_NO_SPACE:
1560 case EXT2_ET_NO_INLINE_DATA:
1561 case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1562 case EXT2_ET_EA_BAD_VALUE_OFFSET:
1563 case EXT2_ET_EA_INODE_CORRUPTED:
1564 /* broken EA or no system.data EA; truncate */
1565 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1566 &pctx)) {
1567 err = ext2fs_inode_size_set(fs, inode, 0);
1568 if (err) {
1569 pctx.errcode = err;
1570 ctx->flags |= E2F_FLAG_ABORT;
1571 goto endit;
1572 }
1573 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1574 memset(&inode->i_block, 0,
1575 sizeof(inode->i_block));
1576 e2fsck_write_inode(ctx, ino, inode,
1577 "pass1");
1578 failed_csum = 0;
1579 }
1580 break;
1581 default:
1582 /* Some other kind of non-xattr error? */
1583 pctx.errcode = err;
1584 ctx->flags |= E2F_FLAG_ABORT;
1585 goto endit;
1586 }
1587 }
1588
1589 /*
1590 * Test for incorrect extent flag settings.
1591 *
1592 * On big-endian machines we must be careful:
1593 * When the inode is read, the i_block array is not swapped
1594 * if the extent flag is set. Therefore if we are testing
1595 * for or fixing a wrongly-set flag, we must potentially
1596 * (un)swap before testing, or after fixing.
1597 */
1598
1599 /*
1600 * In this case the extents flag was set when read, so
1601 * extent_header_verify is ok. If the inode is cleared,
1602 * no need to swap... so no extra swapping here.
1603 */
1604 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1605 (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1606 (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1607 if ((ext2fs_extent_header_verify(inode->i_block,
1608 sizeof(inode->i_block)) == 0) &&
1609 fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1610 ext2fs_set_feature_extents(sb);
1611 ext2fs_mark_super_dirty(fs);
1612 extent_fs = 1;
1613 } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1614 clear_inode:
1615 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1616 if (ino == EXT2_BAD_INO)
1617 ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1618 ino);
1619 /* skip FINISH_INODE_LOOP */
1620 continue;
1621 }
1622 }
1623
1624 /*
1625 * For big-endian machines:
1626 * If the inode didn't have the extents flag set when it
1627 * was read, then the i_blocks array was swapped. To test
1628 * as an extents header, we must swap it back first.
1629 * IF we then set the extents flag, the entire i_block
1630 * array must be un/re-swapped to make it proper extents data.
1631 */
1632 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1633 (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1634 (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1635 (LINUX_S_ISREG(inode->i_mode) ||
1636 LINUX_S_ISDIR(inode->i_mode))) {
1637 void *ehp;
1638 #ifdef WORDS_BIGENDIAN
1639 __u32 tmp_block[EXT2_N_BLOCKS];
1640
1641 for (i = 0; i < EXT2_N_BLOCKS; i++)
1642 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1643 ehp = tmp_block;
1644 #else
1645 ehp = inode->i_block;
1646 #endif
1647 if ((ext2fs_extent_header_verify(ehp,
1648 sizeof(inode->i_block)) == 0) &&
1649 (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1650 inode->i_flags |= EXT4_EXTENTS_FL;
1651 #ifdef WORDS_BIGENDIAN
1652 memcpy(inode->i_block, tmp_block,
1653 sizeof(inode->i_block));
1654 #endif
1655 e2fsck_write_inode(ctx, ino, inode, "pass1");
1656 failed_csum = 0;
1657 }
1658 }
1659
1660 if (ino == EXT2_BAD_INO) {
1661 struct process_block_struct pb;
1662
1663 if ((failed_csum || inode->i_mode || inode->i_uid ||
1664 inode->i_gid || inode->i_links_count ||
1665 (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1666 inode->i_file_acl) &&
1667 fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1668 memset(inode, 0, sizeof(struct ext2_inode));
1669 e2fsck_write_inode(ctx, ino, inode,
1670 "clear bad inode");
1671 failed_csum = 0;
1672 }
1673
1674 pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1675 &pb.fs_meta_blocks);
1676 if (pctx.errcode) {
1677 pctx.num = 4;
1678 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1679 ctx->flags |= E2F_FLAG_ABORT;
1680 goto endit;
1681 }
1682 pb.ino = EXT2_BAD_INO;
1683 pb.num_blocks = pb.last_block = 0;
1684 pb.last_db_block = -1;
1685 pb.num_illegal_blocks = 0;
1686 pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1687 pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1688 pb.inode = inode;
1689 pb.pctx = &pctx;
1690 pb.ctx = ctx;
1691 pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1692 block_buf, process_bad_block, &pb);
1693 ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1694 if (pctx.errcode) {
1695 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1696 ctx->flags |= E2F_FLAG_ABORT;
1697 goto endit;
1698 }
1699 if (pb.bbcheck)
1700 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1701 ctx->flags |= E2F_FLAG_ABORT;
1702 goto endit;
1703 }
1704 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1705 clear_problem_context(&pctx);
1706 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1707 continue;
1708 } else if (ino == EXT2_ROOT_INO) {
1709 /*
1710 * Make sure the root inode is a directory; if
1711 * not, offer to clear it. It will be
1712 * regenerated in pass #3.
1713 */
1714 if (!LINUX_S_ISDIR(inode->i_mode)) {
1715 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1716 goto clear_inode;
1717 }
1718 /*
1719 * If dtime is set, offer to clear it. mke2fs
1720 * version 0.2b created filesystems with the
1721 * dtime field set for the root and lost+found
1722 * directories. We won't worry about
1723 * /lost+found, since that can be regenerated
1724 * easily. But we will fix the root directory
1725 * as a special case.
1726 */
1727 if (inode->i_dtime && inode->i_links_count) {
1728 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1729 inode->i_dtime = 0;
1730 e2fsck_write_inode(ctx, ino, inode,
1731 "pass1");
1732 failed_csum = 0;
1733 }
1734 }
1735 } else if (ino == EXT2_JOURNAL_INO) {
1736 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1737 if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1738 if (!LINUX_S_ISREG(inode->i_mode) &&
1739 fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1740 &pctx)) {
1741 inode->i_mode = LINUX_S_IFREG;
1742 e2fsck_write_inode(ctx, ino, inode,
1743 "pass1");
1744 failed_csum = 0;
1745 }
1746 check_blocks(ctx, &pctx, block_buf, NULL);
1747 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1748 continue;
1749 }
1750 if ((inode->i_links_count ||
1751 inode->i_blocks || inode->i_block[0]) &&
1752 fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1753 &pctx)) {
1754 memset(inode, 0, inode_size);
1755 ext2fs_icount_store(ctx->inode_link_info,
1756 ino, 0);
1757 e2fsck_write_inode_full(ctx, ino, inode,
1758 inode_size, "pass1");
1759 failed_csum = 0;
1760 }
1761 } else if (quota_inum_is_reserved(fs, ino)) {
1762 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1763 if (ext2fs_has_feature_quota(fs->super) &&
1764 quota_inum_is_super(fs->super, ino)) {
1765 if (!LINUX_S_ISREG(inode->i_mode) &&
1766 fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1767 &pctx)) {
1768 inode->i_mode = LINUX_S_IFREG;
1769 e2fsck_write_inode(ctx, ino, inode,
1770 "pass1");
1771 failed_csum = 0;
1772 }
1773 check_blocks(ctx, &pctx, block_buf, NULL);
1774 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1775 continue;
1776 }
1777 if ((inode->i_links_count ||
1778 inode->i_blocks || inode->i_block[0]) &&
1779 fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1780 &pctx)) {
1781 memset(inode, 0, inode_size);
1782 ext2fs_icount_store(ctx->inode_link_info,
1783 ino, 0);
1784 e2fsck_write_inode_full(ctx, ino, inode,
1785 inode_size, "pass1");
1786 failed_csum = 0;
1787 }
1788 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1789 problem_t problem = 0;
1790
1791 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1792 if (ino == EXT2_BOOT_LOADER_INO) {
1793 if (LINUX_S_ISDIR(inode->i_mode))
1794 problem = PR_1_RESERVED_BAD_MODE;
1795 } else if (ino == EXT2_RESIZE_INO) {
1796 if (inode->i_mode &&
1797 !LINUX_S_ISREG(inode->i_mode))
1798 problem = PR_1_RESERVED_BAD_MODE;
1799 } else {
1800 if (inode->i_mode != 0)
1801 problem = PR_1_RESERVED_BAD_MODE;
1802 }
1803 if (problem) {
1804 if (fix_problem(ctx, problem, &pctx)) {
1805 inode->i_mode = 0;
1806 e2fsck_write_inode(ctx, ino, inode,
1807 "pass1");
1808 failed_csum = 0;
1809 }
1810 }
1811 check_blocks(ctx, &pctx, block_buf, NULL);
1812 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1813 continue;
1814 }
1815
1816 if (!inode->i_links_count) {
1817 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1818 continue;
1819 }
1820 /*
1821 * n.b. 0.3c ext2fs code didn't clear i_links_count for
1822 * deleted files. Oops.
1823 *
1824 * Since all new ext2 implementations get this right,
1825 * we now assume that the case of non-zero
1826 * i_links_count and non-zero dtime means that we
1827 * should keep the file, not delete it.
1828 *
1829 */
1830 if (inode->i_dtime) {
1831 if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1832 inode->i_dtime = 0;
1833 e2fsck_write_inode(ctx, ino, inode, "pass1");
1834 failed_csum = 0;
1835 }
1836 }
1837
1838 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1839 switch (fs->super->s_creator_os) {
1840 case EXT2_OS_HURD:
1841 frag = inode->osd2.hurd2.h_i_frag;
1842 fsize = inode->osd2.hurd2.h_i_fsize;
1843 break;
1844 default:
1845 frag = fsize = 0;
1846 }
1847
1848 if (inode->i_faddr || frag || fsize ||
1849 (!ext2fs_has_feature_largedir(fs->super) &&
1850 (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1851 mark_inode_bad(ctx, ino);
1852 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1853 !ext2fs_has_feature_64bit(fs->super) &&
1854 inode->osd2.linux2.l_i_file_acl_high != 0)
1855 mark_inode_bad(ctx, ino);
1856 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1857 !ext2fs_has_feature_huge_file(fs->super) &&
1858 (inode->osd2.linux2.l_i_blocks_hi != 0))
1859 mark_inode_bad(ctx, ino);
1860 if (inode->i_flags & EXT2_IMAGIC_FL) {
1861 if (imagic_fs) {
1862 if (!ctx->inode_imagic_map)
1863 alloc_imagic_map(ctx);
1864 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1865 ino);
1866 } else {
1867 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1868 inode->i_flags &= ~EXT2_IMAGIC_FL;
1869 e2fsck_write_inode(ctx, ino,
1870 inode, "pass1");
1871 failed_csum = 0;
1872 }
1873 }
1874 }
1875
1876 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1877 check_is_really_dir(ctx, &pctx, block_buf);
1878
1879 /*
1880 * ext2fs_inode_has_valid_blocks2 does not actually look
1881 * at i_block[] values, so not endian-sensitive here.
1882 */
1883 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1884 LINUX_S_ISLNK(inode->i_mode) &&
1885 !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1886 fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1887 inode->i_flags &= ~EXT4_EXTENTS_FL;
1888 e2fsck_write_inode(ctx, ino, inode, "pass1");
1889 failed_csum = 0;
1890 }
1891
1892 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
1893 add_encrypted_file(ctx, &pctx) < 0)
1894 goto clear_inode;
1895
1896 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
1897 ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
1898
1899 if (LINUX_S_ISDIR(inode->i_mode)) {
1900 ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1901 e2fsck_add_dir_info(ctx, ino, 0);
1902 ctx->fs_directory_count++;
1903 if (inode->i_flags & EXT4_CASEFOLD_FL)
1904 add_casefolded_dir(ctx, ino);
1905 } else if (LINUX_S_ISREG (inode->i_mode)) {
1906 ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1907 ctx->fs_regular_count++;
1908 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1909 e2fsck_pass1_check_device_inode(fs, inode)) {
1910 check_extents_inlinedata(ctx, &pctx);
1911 check_immutable(ctx, &pctx);
1912 check_size(ctx, &pctx);
1913 ctx->fs_chardev_count++;
1914 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1915 e2fsck_pass1_check_device_inode(fs, inode)) {
1916 check_extents_inlinedata(ctx, &pctx);
1917 check_immutable(ctx, &pctx);
1918 check_size(ctx, &pctx);
1919 ctx->fs_blockdev_count++;
1920 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1921 e2fsck_pass1_check_symlink(fs, ino, inode,
1922 block_buf)) {
1923 check_immutable(ctx, &pctx);
1924 ctx->fs_symlinks_count++;
1925 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1926 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1927 continue;
1928 } else if (ext2fs_is_fast_symlink(inode)) {
1929 ctx->fs_fast_symlinks_count++;
1930 check_blocks(ctx, &pctx, block_buf,
1931 &ea_ibody_quota);
1932 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1933 continue;
1934 }
1935 }
1936 else if (LINUX_S_ISFIFO (inode->i_mode) &&
1937 e2fsck_pass1_check_device_inode(fs, inode)) {
1938 check_extents_inlinedata(ctx, &pctx);
1939 check_immutable(ctx, &pctx);
1940 check_size(ctx, &pctx);
1941 ctx->fs_fifo_count++;
1942 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
1943 e2fsck_pass1_check_device_inode(fs, inode)) {
1944 check_extents_inlinedata(ctx, &pctx);
1945 check_immutable(ctx, &pctx);
1946 check_size(ctx, &pctx);
1947 ctx->fs_sockets_count++;
1948 } else
1949 mark_inode_bad(ctx, ino);
1950 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1951 !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
1952 if (inode->i_block[EXT2_IND_BLOCK])
1953 ctx->fs_ind_count++;
1954 if (inode->i_block[EXT2_DIND_BLOCK])
1955 ctx->fs_dind_count++;
1956 if (inode->i_block[EXT2_TIND_BLOCK])
1957 ctx->fs_tind_count++;
1958 }
1959 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1960 !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
1961 (inode->i_block[EXT2_IND_BLOCK] ||
1962 inode->i_block[EXT2_DIND_BLOCK] ||
1963 inode->i_block[EXT2_TIND_BLOCK] ||
1964 ext2fs_file_acl_block(fs, inode))) {
1965 struct process_inode_block *itp;
1966
1967 itp = &inodes_to_process[process_inode_count];
1968 itp->ino = ino;
1969 itp->ea_ibody_quota = ea_ibody_quota;
1970 if (inode_size < sizeof(struct ext2_inode_large))
1971 memcpy(&itp->inode, inode, inode_size);
1972 else
1973 memcpy(&itp->inode, inode, sizeof(itp->inode));
1974 process_inode_count++;
1975 } else
1976 check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
1977
1978 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1979
1980 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1981 goto endit;
1982
1983 if (process_inode_count >= ctx->process_inode_size) {
1984 process_inodes(ctx, block_buf);
1985
1986 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1987 goto endit;
1988 }
1989 }
1990 process_inodes(ctx, block_buf);
1991 ext2fs_close_inode_scan(scan);
1992 scan = NULL;
1993
1994 reserve_block_for_root_repair(ctx);
1995 reserve_block_for_lnf_repair(ctx);
1996
1997 /*
1998 * If any extended attribute blocks' reference counts need to
1999 * be adjusted, either up (ctx->refcount_extra), or down
2000 * (ctx->refcount), then fix them.
2001 */
2002 if (ctx->refcount) {
2003 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
2004 ea_refcount_free(ctx->refcount);
2005 ctx->refcount = 0;
2006 }
2007 if (ctx->refcount_extra) {
2008 adjust_extattr_refcount(ctx, ctx->refcount_extra,
2009 block_buf, +1);
2010 ea_refcount_free(ctx->refcount_extra);
2011 ctx->refcount_extra = 0;
2012 }
2013
2014 if (ctx->ea_block_quota_blocks) {
2015 ea_refcount_free(ctx->ea_block_quota_blocks);
2016 ctx->ea_block_quota_blocks = 0;
2017 }
2018
2019 if (ctx->ea_block_quota_inodes) {
2020 ea_refcount_free(ctx->ea_block_quota_inodes);
2021 ctx->ea_block_quota_inodes = 0;
2022 }
2023
2024 if (ctx->invalid_bitmaps)
2025 handle_fs_bad_blocks(ctx);
2026
2027 /* We don't need the block_ea_map any more */
2028 if (ctx->block_ea_map) {
2029 ext2fs_free_block_bitmap(ctx->block_ea_map);
2030 ctx->block_ea_map = 0;
2031 }
2032
2033 /* We don't need the encryption policy => ID map any more */
2034 destroy_encryption_policy_map(ctx);
2035
2036 if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2037 clear_problem_context(&pctx);
2038 pctx.errcode = ext2fs_create_resize_inode(fs);
2039 if (pctx.errcode) {
2040 if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2041 &pctx)) {
2042 ctx->flags |= E2F_FLAG_ABORT;
2043 goto endit;
2044 }
2045 pctx.errcode = 0;
2046 }
2047 if (!pctx.errcode) {
2048 e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2049 "recreate inode");
2050 inode->i_mtime = ctx->now;
2051 e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2052 "recreate inode");
2053 }
2054 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2055 }
2056
2057 if (ctx->flags & E2F_FLAG_RESTART) {
2058 /*
2059 * Only the master copy of the superblock and block
2060 * group descriptors are going to be written during a
2061 * restart, so set the superblock to be used to be the
2062 * master superblock.
2063 */
2064 ctx->use_superblock = 0;
2065 goto endit;
2066 }
2067
2068 if (ctx->large_dirs && !ext2fs_has_feature_largedir(fs->super)) {
2069 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2070 ext2fs_set_feature_largedir(fs->super);
2071 fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2072 ext2fs_mark_super_dirty(fs);
2073 }
2074 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2075 fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2076 ext2fs_update_dynamic_rev(fs);
2077 ext2fs_mark_super_dirty(fs);
2078 }
2079 }
2080
2081 if (ctx->block_dup_map) {
2082 if (ctx->options & E2F_OPT_PREEN) {
2083 clear_problem_context(&pctx);
2084 fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2085 }
2086 e2fsck_pass1_dupblocks(ctx, block_buf);
2087 }
2088 ctx->flags |= E2F_FLAG_ALLOC_OK;
2089 endit:
2090 e2fsck_use_inode_shortcuts(ctx, 0);
2091 ext2fs_free_mem(&inodes_to_process);
2092 inodes_to_process = 0;
2093
2094 if (scan)
2095 ext2fs_close_inode_scan(scan);
2096 if (block_buf)
2097 ext2fs_free_mem(&block_buf);
2098 if (inode)
2099 ext2fs_free_mem(&inode);
2100
2101 /*
2102 * The l+f inode may have been cleared, so zap it now and
2103 * later passes will recalculate it if necessary
2104 */
2105 ctx->lost_and_found = 0;
2106
2107 if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2108 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2109 else
2110 ctx->invalid_bitmaps++;
2111 }
2112 #undef FINISH_INODE_LOOP
2113
2114 /*
2115 * When the inode_scan routines call this callback at the end of the
2116 * glock group, call process_inodes.
2117 */
scan_callback(ext2_filsys fs,ext2_inode_scan scan EXT2FS_ATTR ((unused)),dgrp_t group,void * priv_data)2118 static errcode_t scan_callback(ext2_filsys fs,
2119 ext2_inode_scan scan EXT2FS_ATTR((unused)),
2120 dgrp_t group, void * priv_data)
2121 {
2122 struct scan_callback_struct *scan_struct;
2123 e2fsck_t ctx;
2124
2125 scan_struct = (struct scan_callback_struct *) priv_data;
2126 ctx = scan_struct->ctx;
2127
2128 process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf);
2129
2130 if (ctx->progress)
2131 if ((ctx->progress)(ctx, 1, group+1,
2132 ctx->fs->group_desc_count))
2133 return EXT2_ET_CANCEL_REQUESTED;
2134
2135 return 0;
2136 }
2137
2138 /*
2139 * Process the inodes in the "inodes to process" list.
2140 */
process_inodes(e2fsck_t ctx,char * block_buf)2141 static void process_inodes(e2fsck_t ctx, char *block_buf)
2142 {
2143 int i;
2144 struct ext2_inode *old_stashed_inode;
2145 ext2_ino_t old_stashed_ino;
2146 const char *old_operation;
2147 char buf[80];
2148 struct problem_context pctx;
2149
2150 #if 0
2151 printf("begin process_inodes: ");
2152 #endif
2153 if (process_inode_count == 0)
2154 return;
2155 old_operation = ehandler_operation(0);
2156 old_stashed_inode = ctx->stashed_inode;
2157 old_stashed_ino = ctx->stashed_ino;
2158 qsort(inodes_to_process, process_inode_count,
2159 sizeof(struct process_inode_block), process_inode_cmp);
2160 clear_problem_context(&pctx);
2161 for (i=0; i < process_inode_count; i++) {
2162 pctx.inode = ctx->stashed_inode =
2163 (struct ext2_inode *) &inodes_to_process[i].inode;
2164 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2165
2166 #if 0
2167 printf("%u ", pctx.ino);
2168 #endif
2169 sprintf(buf, _("reading indirect blocks of inode %u"),
2170 pctx.ino);
2171 ehandler_operation(buf);
2172 check_blocks(ctx, &pctx, block_buf,
2173 &inodes_to_process[i].ea_ibody_quota);
2174 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2175 break;
2176 }
2177 ctx->stashed_inode = old_stashed_inode;
2178 ctx->stashed_ino = old_stashed_ino;
2179 process_inode_count = 0;
2180 #if 0
2181 printf("end process inodes\n");
2182 #endif
2183 ehandler_operation(old_operation);
2184 }
2185
process_inode_cmp(const void * a,const void * b)2186 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
2187 {
2188 const struct process_inode_block *ib_a =
2189 (const struct process_inode_block *) a;
2190 const struct process_inode_block *ib_b =
2191 (const struct process_inode_block *) b;
2192 int ret;
2193
2194 ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
2195 ib_b->inode.i_block[EXT2_IND_BLOCK]);
2196 if (ret == 0)
2197 /*
2198 * We only call process_inodes() for non-extent
2199 * inodes, so it's OK to pass NULL to
2200 * ext2fs_file_acl_block() here.
2201 */
2202 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
2203 ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
2204 if (ret == 0)
2205 ret = ib_a->ino - ib_b->ino;
2206 return ret;
2207 }
2208
2209 /*
2210 * Mark an inode as being bad in some what
2211 */
mark_inode_bad(e2fsck_t ctx,ext2_ino_t ino)2212 static void mark_inode_bad(e2fsck_t ctx, ext2_ino_t ino)
2213 {
2214 struct problem_context pctx;
2215
2216 if (!ctx->inode_bad_map) {
2217 clear_problem_context(&pctx);
2218
2219 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2220 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
2221 "inode_bad_map", &ctx->inode_bad_map);
2222 if (pctx.errcode) {
2223 pctx.num = 3;
2224 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2225 /* Should never get here */
2226 ctx->flags |= E2F_FLAG_ABORT;
2227 return;
2228 }
2229 }
2230 ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
2231 }
2232
add_casefolded_dir(e2fsck_t ctx,ext2_ino_t ino)2233 static void add_casefolded_dir(e2fsck_t ctx, ext2_ino_t ino)
2234 {
2235 struct problem_context pctx;
2236
2237 if (!ctx->casefolded_dirs) {
2238 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
2239 if (pctx.errcode)
2240 goto error;
2241 }
2242 pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
2243 if (pctx.errcode == 0)
2244 return;
2245 error:
2246 fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
2247 /* Should never get here */
2248 ctx->flags |= E2F_FLAG_ABORT;
2249 }
2250
2251 /*
2252 * This procedure will allocate the inode "bb" (badblock) map table
2253 */
alloc_bb_map(e2fsck_t ctx)2254 static void alloc_bb_map(e2fsck_t ctx)
2255 {
2256 struct problem_context pctx;
2257
2258 clear_problem_context(&pctx);
2259 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2260 _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
2261 "inode_bb_map", &ctx->inode_bb_map);
2262 if (pctx.errcode) {
2263 pctx.num = 4;
2264 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2265 /* Should never get here */
2266 ctx->flags |= E2F_FLAG_ABORT;
2267 return;
2268 }
2269 }
2270
2271 /*
2272 * This procedure will allocate the inode imagic table
2273 */
alloc_imagic_map(e2fsck_t ctx)2274 static void alloc_imagic_map(e2fsck_t ctx)
2275 {
2276 struct problem_context pctx;
2277
2278 clear_problem_context(&pctx);
2279 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2280 _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
2281 "inode_imagic_map", &ctx->inode_imagic_map);
2282 if (pctx.errcode) {
2283 pctx.num = 5;
2284 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2285 /* Should never get here */
2286 ctx->flags |= E2F_FLAG_ABORT;
2287 return;
2288 }
2289 }
2290
2291 /*
2292 * Marks a block as in use, setting the dup_map if it's been set
2293 * already. Called by process_block and process_bad_block.
2294 *
2295 * WARNING: Assumes checks have already been done to make sure block
2296 * is valid. This is true in both process_block and process_bad_block.
2297 */
mark_block_used(e2fsck_t ctx,blk64_t block)2298 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
2299 {
2300 struct problem_context pctx;
2301
2302 clear_problem_context(&pctx);
2303
2304 if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
2305 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
2306 !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
2307 return;
2308 }
2309 if (!ctx->block_dup_map) {
2310 pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
2311 _("multiply claimed block map"),
2312 EXT2FS_BMAP64_RBTREE, "block_dup_map",
2313 &ctx->block_dup_map);
2314 if (pctx.errcode) {
2315 pctx.num = 3;
2316 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
2317 &pctx);
2318 /* Should never get here */
2319 ctx->flags |= E2F_FLAG_ABORT;
2320 return;
2321 }
2322 }
2323 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
2324 } else {
2325 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
2326 }
2327 }
2328
2329 /*
2330 * When cluster size is greater than one block, it is caller's responsibility
2331 * to make sure block parameter starts at a cluster boundary.
2332 */
mark_blocks_used(e2fsck_t ctx,blk64_t block,unsigned int num)2333 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
2334 unsigned int num)
2335 {
2336 if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
2337 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
2338 else {
2339 unsigned int i;
2340
2341 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
2342 mark_block_used(ctx, block + i);
2343 }
2344 }
2345
2346 /*
2347 * Adjust the extended attribute block's reference counts at the end
2348 * of pass 1, either by subtracting out references for EA blocks that
2349 * are still referenced in ctx->refcount, or by adding references for
2350 * EA blocks that had extra references as accounted for in
2351 * ctx->refcount_extra.
2352 */
adjust_extattr_refcount(e2fsck_t ctx,ext2_refcount_t refcount,char * block_buf,int adjust_sign)2353 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
2354 char *block_buf, int adjust_sign)
2355 {
2356 struct ext2_ext_attr_header *header;
2357 struct problem_context pctx;
2358 ext2_filsys fs = ctx->fs;
2359 blk64_t blk;
2360 __u32 should_be;
2361 ea_value_t count;
2362
2363 clear_problem_context(&pctx);
2364
2365 ea_refcount_intr_begin(refcount);
2366 while (1) {
2367 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
2368 break;
2369 pctx.blk = blk;
2370 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
2371 pctx.ino);
2372 if (pctx.errcode) {
2373 fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
2374 return;
2375 }
2376 header = (struct ext2_ext_attr_header *) block_buf;
2377 pctx.blkcount = header->h_refcount;
2378 should_be = header->h_refcount + adjust_sign * (int)count;
2379 pctx.num = should_be;
2380 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
2381 header->h_refcount = should_be;
2382 pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
2383 block_buf,
2384 pctx.ino);
2385 if (pctx.errcode) {
2386 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
2387 &pctx);
2388 continue;
2389 }
2390 }
2391 }
2392 }
2393
2394 /*
2395 * Handle processing the extended attribute blocks
2396 */
check_ext_attr(e2fsck_t ctx,struct problem_context * pctx,char * block_buf,struct ea_quota * ea_block_quota)2397 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
2398 char *block_buf, struct ea_quota *ea_block_quota)
2399 {
2400 ext2_filsys fs = ctx->fs;
2401 ext2_ino_t ino = pctx->ino;
2402 struct ext2_inode *inode = pctx->inode;
2403 blk64_t blk;
2404 char * end;
2405 struct ext2_ext_attr_header *header;
2406 struct ext2_ext_attr_entry *first, *entry;
2407 blk64_t quota_blocks = EXT2FS_C2B(fs, 1);
2408 __u64 quota_inodes = 0;
2409 region_t region = 0;
2410 int failed_csum = 0;
2411
2412 ea_block_quota->blocks = 0;
2413 ea_block_quota->inodes = 0;
2414
2415 blk = ext2fs_file_acl_block(fs, inode);
2416 if (blk == 0)
2417 return 0;
2418
2419 /*
2420 * If the Extended attribute flag isn't set, then a non-zero
2421 * file acl means that the inode is corrupted.
2422 *
2423 * Or if the extended attribute block is an invalid block,
2424 * then the inode is also corrupted.
2425 */
2426 if (!ext2fs_has_feature_xattr(fs->super) ||
2427 (blk < fs->super->s_first_data_block) ||
2428 (blk >= ext2fs_blocks_count(fs->super))) {
2429 mark_inode_bad(ctx, ino);
2430 return 0;
2431 }
2432
2433 /* If ea bitmap hasn't been allocated, create it */
2434 if (!ctx->block_ea_map) {
2435 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
2436 _("ext attr block map"),
2437 EXT2FS_BMAP64_RBTREE, "block_ea_map",
2438 &ctx->block_ea_map);
2439 if (pctx->errcode) {
2440 pctx->num = 2;
2441 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
2442 ctx->flags |= E2F_FLAG_ABORT;
2443 return 0;
2444 }
2445 }
2446
2447 /* Create the EA refcount structure if necessary */
2448 if (!ctx->refcount) {
2449 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
2450 if (pctx->errcode) {
2451 pctx->num = 1;
2452 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2453 ctx->flags |= E2F_FLAG_ABORT;
2454 return 0;
2455 }
2456 }
2457
2458 #if 0
2459 /* Debugging text */
2460 printf("Inode %u has EA block %u\n", ino, blk);
2461 #endif
2462
2463 /* Have we seen this EA block before? */
2464 if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
2465 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
2466 ea_block_quota->inodes = 0;
2467
2468 if (ctx->ea_block_quota_blocks) {
2469 ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
2470 "a_blocks);
2471 if (quota_blocks)
2472 ea_block_quota->blocks = quota_blocks;
2473 }
2474
2475 if (ctx->ea_block_quota_inodes)
2476 ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
2477 &ea_block_quota->inodes);
2478
2479 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
2480 return 1;
2481 /* Ooops, this EA was referenced more than it stated */
2482 if (!ctx->refcount_extra) {
2483 pctx->errcode = ea_refcount_create(0,
2484 &ctx->refcount_extra);
2485 if (pctx->errcode) {
2486 pctx->num = 2;
2487 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2488 ctx->flags |= E2F_FLAG_ABORT;
2489 return 0;
2490 }
2491 }
2492 ea_refcount_increment(ctx->refcount_extra, blk, 0);
2493 return 1;
2494 }
2495
2496 /*
2497 * OK, we haven't seen this EA block yet. So we need to
2498 * validate it
2499 */
2500 pctx->blk = blk;
2501 pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
2502 if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
2503 pctx->errcode = 0;
2504 failed_csum = 1;
2505 } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
2506 pctx->errcode = 0;
2507
2508 if (pctx->errcode &&
2509 fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
2510 pctx->errcode = 0;
2511 goto clear_extattr;
2512 }
2513 header = (struct ext2_ext_attr_header *) block_buf;
2514 pctx->blk = ext2fs_file_acl_block(fs, inode);
2515 if (((ctx->ext_attr_ver == 1) &&
2516 (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
2517 ((ctx->ext_attr_ver == 2) &&
2518 (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
2519 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
2520 goto clear_extattr;
2521 }
2522
2523 if (header->h_blocks != 1) {
2524 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
2525 goto clear_extattr;
2526 }
2527
2528 if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
2529 goto clear_extattr;
2530
2531 region = region_create(0, fs->blocksize);
2532 if (!region) {
2533 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
2534 ctx->flags |= E2F_FLAG_ABORT;
2535 return 0;
2536 }
2537 if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
2538 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2539 goto clear_extattr;
2540 }
2541
2542 first = (struct ext2_ext_attr_entry *)(header+1);
2543 end = block_buf + fs->blocksize;
2544 entry = first;
2545 while ((char *)entry < end && *(__u32 *)entry) {
2546 __u32 hash;
2547
2548 if (region_allocate(region, (char *)entry - (char *)header,
2549 EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
2550 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2551 goto clear_extattr;
2552 break;
2553 }
2554 if ((ctx->ext_attr_ver == 1 &&
2555 (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
2556 (ctx->ext_attr_ver == 2 &&
2557 entry->e_name_index == 0)) {
2558 if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
2559 goto clear_extattr;
2560 break;
2561 }
2562 if (entry->e_value_inum == 0) {
2563 if (entry->e_value_size > EXT2_XATTR_SIZE_MAX ||
2564 (entry->e_value_offs + entry->e_value_size >
2565 fs->blocksize)) {
2566 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
2567 goto clear_extattr;
2568 break;
2569 }
2570 if (entry->e_value_size &&
2571 region_allocate(region, entry->e_value_offs,
2572 EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
2573 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
2574 pctx))
2575 goto clear_extattr;
2576 }
2577
2578 hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
2579 entry->e_value_offs);
2580 if (entry->e_hash != hash)
2581 hash = ext2fs_ext_attr_hash_entry_signed(entry,
2582 block_buf + entry->e_value_offs);
2583
2584 if (entry->e_hash != hash) {
2585 pctx->num = entry->e_hash;
2586 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
2587 goto clear_extattr;
2588 entry->e_hash = hash;
2589 }
2590 } else {
2591 problem_t problem;
2592 blk64_t entry_quota_blocks;
2593
2594 problem = check_large_ea_inode(ctx, entry, pctx,
2595 &entry_quota_blocks);
2596 if (problem && fix_problem(ctx, problem, pctx))
2597 goto clear_extattr;
2598
2599 quota_blocks += entry_quota_blocks;
2600 quota_inodes++;
2601 }
2602
2603 entry = EXT2_EXT_ATTR_NEXT(entry);
2604 }
2605 if (region_allocate(region, (char *)entry - (char *)header, 4)) {
2606 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2607 goto clear_extattr;
2608 }
2609 region_free(region);
2610
2611 /*
2612 * We only get here if there was no other errors that were fixed.
2613 * If there was a checksum fail, ask to correct it.
2614 */
2615 if (failed_csum &&
2616 fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
2617 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
2618 pctx->ino);
2619 if (pctx->errcode)
2620 return 0;
2621 }
2622
2623 if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
2624 if (!ctx->ea_block_quota_blocks) {
2625 pctx->errcode = ea_refcount_create(0,
2626 &ctx->ea_block_quota_blocks);
2627 if (pctx->errcode) {
2628 pctx->num = 3;
2629 goto refcount_fail;
2630 }
2631 }
2632 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
2633 quota_blocks);
2634 }
2635
2636 if (quota_inodes) {
2637 if (!ctx->ea_block_quota_inodes) {
2638 pctx->errcode = ea_refcount_create(0,
2639 &ctx->ea_block_quota_inodes);
2640 if (pctx->errcode) {
2641 pctx->num = 4;
2642 refcount_fail:
2643 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2644 ctx->flags |= E2F_FLAG_ABORT;
2645 return 0;
2646 }
2647 }
2648
2649 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
2650 quota_inodes);
2651 }
2652 ea_block_quota->blocks = quota_blocks;
2653 ea_block_quota->inodes = quota_inodes;
2654
2655 inc_ea_inode_refs(ctx, pctx, first, end);
2656 ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
2657 mark_block_used(ctx, blk);
2658 ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
2659 return 1;
2660
2661 clear_extattr:
2662 if (region)
2663 region_free(region);
2664 ext2fs_file_acl_block_set(fs, inode, 0);
2665 e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
2666 return 0;
2667 }
2668
2669 /* Returns 1 if bad htree, 0 if OK */
handle_htree(e2fsck_t ctx,struct problem_context * pctx,ext2_ino_t ino,struct ext2_inode * inode,char * block_buf)2670 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
2671 ext2_ino_t ino, struct ext2_inode *inode,
2672 char *block_buf)
2673 {
2674 struct ext2_dx_root_info *root;
2675 ext2_filsys fs = ctx->fs;
2676 errcode_t retval;
2677 blk64_t blk;
2678
2679 if ((!LINUX_S_ISDIR(inode->i_mode) &&
2680 fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
2681 (!ext2fs_has_feature_dir_index(fs->super) &&
2682 fix_problem(ctx, PR_1_HTREE_SET, pctx)))
2683 return 1;
2684
2685 pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
2686
2687 if ((pctx->errcode) ||
2688 (blk == 0) ||
2689 (blk < fs->super->s_first_data_block) ||
2690 (blk >= ext2fs_blocks_count(fs->super))) {
2691 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2692 return 1;
2693 else
2694 return 0;
2695 }
2696
2697 retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
2698 if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2699 return 1;
2700
2701 /* XXX should check that beginning matches a directory */
2702 root = (struct ext2_dx_root_info *) (block_buf + 24);
2703
2704 if ((root->reserved_zero || root->info_length < 8) &&
2705 fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2706 return 1;
2707
2708 pctx->num = root->hash_version;
2709 if ((root->hash_version != EXT2_HASH_LEGACY) &&
2710 (root->hash_version != EXT2_HASH_HALF_MD4) &&
2711 (root->hash_version != EXT2_HASH_TEA) &&
2712 (root->hash_version != EXT2_HASH_SIPHASH) &&
2713 fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
2714 return 1;
2715
2716 if (ext4_hash_in_dirent(inode)) {
2717 if (root->hash_version != EXT2_HASH_SIPHASH &&
2718 fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
2719 return 1;
2720 } else {
2721 if (root->hash_version == EXT2_HASH_SIPHASH &&
2722 fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
2723 return 1;
2724 }
2725
2726 if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
2727 fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
2728 return 1;
2729
2730 pctx->num = root->indirect_levels;
2731 /* if htree level is clearly too high, consider it to be broken */
2732 if (root->indirect_levels > EXT4_HTREE_LEVEL &&
2733 fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
2734 return 1;
2735
2736 /* if level is only maybe too high, LARGE_DIR feature could be unset */
2737 if (root->indirect_levels > ext2_dir_htree_level(fs) &&
2738 !ext2fs_has_feature_largedir(fs->super)) {
2739 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
2740 unsigned idx_pb = 1 << (blockbits - 3);
2741
2742 /* compare inode size/blocks vs. max-sized 2-level htree */
2743 if (EXT2_I_SIZE(pctx->inode) <
2744 (idx_pb - 1) * (idx_pb - 2) << blockbits &&
2745 pctx->inode->i_blocks <
2746 (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
2747 fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
2748 return 1;
2749 }
2750
2751 if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
2752 ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
2753 ctx->large_dirs++;
2754
2755 return 0;
2756 }
2757
e2fsck_clear_inode(e2fsck_t ctx,ext2_ino_t ino,struct ext2_inode * inode,int restart_flag,const char * source)2758 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
2759 struct ext2_inode *inode, int restart_flag,
2760 const char *source)
2761 {
2762 inode->i_flags = 0;
2763 inode->i_links_count = 0;
2764 ext2fs_icount_store(ctx->inode_link_info, ino, 0);
2765 inode->i_dtime = ctx->now;
2766
2767 /*
2768 * If a special inode has such rotten block mappings that we
2769 * want to clear the whole inode, be sure to actually zap
2770 * the block maps because i_links_count isn't checked for
2771 * special inodes, and we'll end up right back here the next
2772 * time we run fsck.
2773 */
2774 if (ino < EXT2_FIRST_INODE(ctx->fs->super))
2775 memset(inode->i_block, 0, sizeof(inode->i_block));
2776
2777 ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
2778 ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
2779 if (ctx->inode_reg_map)
2780 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
2781 if (ctx->inode_bad_map)
2782 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
2783
2784 /*
2785 * If the inode was partially accounted for before processing
2786 * was aborted, we need to restart the pass 1 scan.
2787 */
2788 ctx->flags |= restart_flag;
2789
2790 if (ino == EXT2_BAD_INO)
2791 memset(inode, 0, sizeof(struct ext2_inode));
2792
2793 e2fsck_write_inode(ctx, ino, inode, source);
2794 }
2795
2796 /*
2797 * Use the multiple-blocks reclamation code to fix alignment problems in
2798 * a bigalloc filesystem. We want a logical cluster to map to *only* one
2799 * physical cluster, and we want the block offsets within that cluster to
2800 * line up.
2801 */
has_unaligned_cluster_map(e2fsck_t ctx,blk64_t last_pblk,blk64_t last_lblk,blk64_t pblk,blk64_t lblk)2802 static int has_unaligned_cluster_map(e2fsck_t ctx,
2803 blk64_t last_pblk, blk64_t last_lblk,
2804 blk64_t pblk, blk64_t lblk)
2805 {
2806 blk64_t cluster_mask;
2807
2808 if (!ctx->fs->cluster_ratio_bits)
2809 return 0;
2810 cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
2811
2812 /*
2813 * If the block in the logical cluster doesn't align with the block in
2814 * the physical cluster...
2815 */
2816 if ((lblk & cluster_mask) != (pblk & cluster_mask))
2817 return 1;
2818
2819 /*
2820 * If we cross a physical cluster boundary within a logical cluster...
2821 */
2822 if (last_pblk && (lblk & cluster_mask) != 0 &&
2823 EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
2824 EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
2825 return 1;
2826
2827 return 0;
2828 }
2829
scan_extent_node(e2fsck_t ctx,struct problem_context * pctx,struct process_block_struct * pb,blk64_t start_block,blk64_t end_block,blk64_t eof_block,ext2_extent_handle_t ehandle,int try_repairs)2830 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
2831 struct process_block_struct *pb,
2832 blk64_t start_block, blk64_t end_block,
2833 blk64_t eof_block,
2834 ext2_extent_handle_t ehandle,
2835 int try_repairs)
2836 {
2837 struct ext2fs_extent extent;
2838 blk64_t blk, last_lblk;
2839 unsigned int i, n;
2840 int is_dir, is_leaf;
2841 problem_t problem;
2842 struct ext2_extent_info info;
2843 int failed_csum = 0;
2844
2845 if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
2846 failed_csum = 1;
2847
2848 pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
2849 if (pctx->errcode)
2850 return;
2851 if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
2852 !pb->eti.force_rebuild &&
2853 info.curr_level < MAX_EXTENT_DEPTH_COUNT) {
2854 struct extent_tree_level *etl;
2855
2856 etl = pb->eti.ext_info + info.curr_level;
2857 etl->num_extents += info.num_entries;
2858 etl->max_extents += info.max_entries;
2859 /*
2860 * Implementation wart: Splitting extent blocks when appending
2861 * will leave the old block with one free entry. Therefore
2862 * unless the node is totally full, pretend that a non-root
2863 * extent block can hold one fewer entry than it actually does,
2864 * so that we don't repeatedly rebuild the extent tree.
2865 */
2866 if (info.curr_level && info.num_entries < info.max_entries)
2867 etl->max_extents--;
2868 }
2869
2870 pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
2871 &extent);
2872 while ((pctx->errcode == 0 ||
2873 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
2874 info.num_entries-- > 0) {
2875 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
2876 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
2877 last_lblk = extent.e_lblk + extent.e_len - 1;
2878
2879 problem = 0;
2880 pctx->blk = extent.e_pblk;
2881 pctx->blk2 = extent.e_lblk;
2882 pctx->num = extent.e_len;
2883 pctx->blkcount = extent.e_lblk + extent.e_len;
2884
2885 if (extent.e_pblk == 0 ||
2886 extent.e_pblk < ctx->fs->super->s_first_data_block ||
2887 extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
2888 problem = PR_1_EXTENT_BAD_START_BLK;
2889 else if (extent.e_lblk < start_block)
2890 problem = PR_1_OUT_OF_ORDER_EXTENTS;
2891 else if ((end_block && last_lblk > end_block) &&
2892 !(last_lblk > eof_block &&
2893 ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
2894 (pctx->inode->i_flags & EXT4_VERITY_FL))))
2895 problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
2896 else if (is_leaf && extent.e_len == 0)
2897 problem = PR_1_EXTENT_LENGTH_ZERO;
2898 else if (is_leaf &&
2899 (extent.e_pblk + extent.e_len) >
2900 ext2fs_blocks_count(ctx->fs->super))
2901 problem = PR_1_EXTENT_ENDS_BEYOND;
2902 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
2903 !ext2fs_has_feature_largedir(ctx->fs->super) &&
2904 ((extent.e_lblk + extent.e_len) >
2905 (1U << (21 - ctx->fs->super->s_log_block_size))))
2906 problem = PR_1_TOOBIG_DIR;
2907
2908 if (is_leaf && problem == 0 && extent.e_len > 0) {
2909 #if 0
2910 printf("extent_region(ino=%u, expect=%llu, "
2911 "lblk=%llu, len=%u)\n", pb->ino,
2912 (unsigned long long) pb->next_lblock,
2913 (unsigned long long) extent.e_lblk,
2914 extent.e_len);
2915 #endif
2916 if (extent.e_lblk < pb->next_lblock)
2917 problem = PR_1_EXTENT_COLLISION;
2918 else if (extent.e_lblk + extent.e_len > pb->next_lblock)
2919 pb->next_lblock = extent.e_lblk + extent.e_len;
2920 }
2921
2922 /*
2923 * Uninitialized blocks in a directory? Clear the flag and
2924 * we'll interpret the blocks later.
2925 */
2926 if (try_repairs && is_dir && problem == 0 &&
2927 (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
2928 fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
2929 extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
2930 pb->inode_modified = 1;
2931 pctx->errcode = ext2fs_extent_replace(ehandle, 0,
2932 &extent);
2933 if (pctx->errcode)
2934 return;
2935 failed_csum = 0;
2936 }
2937 #ifdef CONFIG_DEVELOPER_FEATURES
2938 if (try_repairs && !is_dir && problem == 0 &&
2939 (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
2940 (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
2941 fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
2942 extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
2943 pb->inode_modified = 1;
2944 pctx->errcode = ext2fs_extent_replace(ehandle, 0,
2945 &extent);
2946 if (pctx->errcode)
2947 return;
2948 failed_csum = 0;
2949 }
2950 #endif
2951 if (try_repairs && problem) {
2952 report_problem:
2953 if (fix_problem(ctx, problem, pctx)) {
2954 if (ctx->invalid_bitmaps) {
2955 /*
2956 * If fsck knows the bitmaps are bad,
2957 * skip to the next extent and
2958 * try to clear this extent again
2959 * after fixing the bitmaps, by
2960 * restarting fsck.
2961 */
2962 pctx->errcode = ext2fs_extent_get(
2963 ehandle,
2964 EXT2_EXTENT_NEXT_SIB,
2965 &extent);
2966 ctx->flags |= E2F_FLAG_RESTART_LATER;
2967 if (pctx->errcode ==
2968 EXT2_ET_NO_CURRENT_NODE) {
2969 pctx->errcode = 0;
2970 break;
2971 }
2972 continue;
2973 }
2974 e2fsck_read_bitmaps(ctx);
2975 pb->inode_modified = 1;
2976 pctx->errcode =
2977 ext2fs_extent_delete(ehandle, 0);
2978 if (pctx->errcode) {
2979 pctx->str = "ext2fs_extent_delete";
2980 return;
2981 }
2982 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
2983 if (pctx->errcode &&
2984 pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
2985 pctx->str = "ext2fs_extent_fix_parents";
2986 return;
2987 }
2988 pctx->errcode = ext2fs_extent_get(ehandle,
2989 EXT2_EXTENT_CURRENT,
2990 &extent);
2991 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
2992 pctx->errcode = 0;
2993 break;
2994 }
2995 failed_csum = 0;
2996 continue;
2997 }
2998 goto next;
2999 }
3000
3001 if (!is_leaf) {
3002 blk64_t lblk = extent.e_lblk;
3003 int next_try_repairs = 1;
3004
3005 blk = extent.e_pblk;
3006
3007 /*
3008 * If this lower extent block collides with critical
3009 * metadata, don't try to repair the damage. Pass 1b
3010 * will reallocate the block; then we can try again.
3011 */
3012 if (pb->ino != EXT2_RESIZE_INO &&
3013 extent.e_pblk < ctx->fs->super->s_blocks_count &&
3014 ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3015 extent.e_pblk)) {
3016 next_try_repairs = 0;
3017 pctx->blk = blk;
3018 fix_problem(ctx,
3019 PR_1_CRITICAL_METADATA_COLLISION,
3020 pctx);
3021 if ((ctx->options & E2F_OPT_NO) == 0)
3022 ctx->flags |= E2F_FLAG_RESTART_LATER;
3023 }
3024 pctx->errcode = ext2fs_extent_get(ehandle,
3025 EXT2_EXTENT_DOWN, &extent);
3026 if (pctx->errcode &&
3027 pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
3028 pctx->str = "EXT2_EXTENT_DOWN";
3029 problem = PR_1_EXTENT_HEADER_INVALID;
3030 if (!next_try_repairs)
3031 return;
3032 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
3033 goto report_problem;
3034 return;
3035 }
3036 /* The next extent should match this index's logical start */
3037 if (extent.e_lblk != lblk) {
3038 struct ext2_extent_info e_info;
3039
3040 pctx->errcode = ext2fs_extent_get_info(ehandle,
3041 &e_info);
3042 if (pctx->errcode) {
3043 pctx->str = "ext2fs_extent_get_info";
3044 return;
3045 }
3046 pctx->blk = lblk;
3047 pctx->blk2 = extent.e_lblk;
3048 pctx->num = e_info.curr_level - 1;
3049 problem = PR_1_EXTENT_INDEX_START_INVALID;
3050 if (fix_problem(ctx, problem, pctx)) {
3051 pb->inode_modified = 1;
3052 pctx->errcode =
3053 ext2fs_extent_fix_parents(ehandle);
3054 if (pctx->errcode) {
3055 pctx->str = "ext2fs_extent_fix_parents";
3056 return;
3057 }
3058 }
3059 }
3060 scan_extent_node(ctx, pctx, pb, extent.e_lblk,
3061 last_lblk, eof_block, ehandle,
3062 next_try_repairs);
3063 if (pctx->errcode)
3064 return;
3065 pctx->errcode = ext2fs_extent_get(ehandle,
3066 EXT2_EXTENT_UP, &extent);
3067 if (pctx->errcode) {
3068 pctx->str = "EXT2_EXTENT_UP";
3069 return;
3070 }
3071 mark_block_used(ctx, blk);
3072 pb->num_blocks++;
3073 goto next;
3074 }
3075
3076 if ((pb->previous_block != 0) &&
3077 (pb->previous_block+1 != extent.e_pblk)) {
3078 if (ctx->options & E2F_OPT_FRAGCHECK) {
3079 char type = '?';
3080
3081 if (pb->is_dir)
3082 type = 'd';
3083 else if (pb->is_reg)
3084 type = 'f';
3085
3086 printf(("%6lu(%c): expecting %6lu "
3087 "actual extent "
3088 "phys %6lu log %lu len %lu\n"),
3089 (unsigned long) pctx->ino, type,
3090 (unsigned long) pb->previous_block+1,
3091 (unsigned long) extent.e_pblk,
3092 (unsigned long) extent.e_lblk,
3093 (unsigned long) extent.e_len);
3094 }
3095 pb->fragmented = 1;
3096 }
3097 /*
3098 * If we notice a gap in the logical block mappings of an
3099 * extent-mapped directory, offer to close the hole by
3100 * moving the logical block down, otherwise we'll go mad in
3101 * pass 3 allocating empty directory blocks to fill the hole.
3102 */
3103 if (try_repairs && is_dir &&
3104 pb->last_block + 1 < extent.e_lblk) {
3105 blk64_t new_lblk;
3106
3107 new_lblk = pb->last_block + 1;
3108 if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3109 new_lblk = ((new_lblk +
3110 EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3111 ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3112 (extent.e_pblk &
3113 EXT2FS_CLUSTER_MASK(ctx->fs));
3114 pctx->blk = extent.e_lblk;
3115 pctx->blk2 = new_lblk;
3116 if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3117 extent.e_lblk = new_lblk;
3118 pb->inode_modified = 1;
3119 pctx->errcode = ext2fs_extent_replace(ehandle,
3120 0, &extent);
3121 if (pctx->errcode) {
3122 pctx->errcode = 0;
3123 goto alloc_later;
3124 }
3125 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3126 if (pctx->errcode)
3127 goto failed_add_dir_block;
3128 pctx->errcode = ext2fs_extent_goto(ehandle,
3129 extent.e_lblk);
3130 if (pctx->errcode)
3131 goto failed_add_dir_block;
3132 last_lblk = extent.e_lblk + extent.e_len - 1;
3133 failed_csum = 0;
3134 }
3135 }
3136 alloc_later:
3137 if (is_dir) {
3138 while (++pb->last_db_block <
3139 (e2_blkcnt_t) extent.e_lblk) {
3140 pctx->errcode = ext2fs_add_dir_block2(
3141 ctx->fs->dblist,
3142 pb->ino, 0,
3143 pb->last_db_block);
3144 if (pctx->errcode) {
3145 pctx->blk = 0;
3146 pctx->num = pb->last_db_block;
3147 goto failed_add_dir_block;
3148 }
3149 }
3150
3151 for (i = 0; i < extent.e_len; i++) {
3152 pctx->errcode = ext2fs_add_dir_block2(
3153 ctx->fs->dblist,
3154 pctx->ino,
3155 extent.e_pblk + i,
3156 extent.e_lblk + i);
3157 if (pctx->errcode) {
3158 pctx->blk = extent.e_pblk + i;
3159 pctx->num = extent.e_lblk + i;
3160 failed_add_dir_block:
3161 fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3162 /* Should never get here */
3163 ctx->flags |= E2F_FLAG_ABORT;
3164 return;
3165 }
3166 }
3167 if (extent.e_len > 0)
3168 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3169 }
3170 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3171 pb->last_block,
3172 extent.e_pblk,
3173 extent.e_lblk)) {
3174 for (i = 0; i < extent.e_len; i++) {
3175 pctx->blk = extent.e_lblk + i;
3176 pctx->blk2 = extent.e_pblk + i;
3177 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3178 mark_block_used(ctx, extent.e_pblk + i);
3179 mark_block_used(ctx, extent.e_pblk + i);
3180 }
3181 }
3182
3183 /*
3184 * Check whether first cluster got marked in previous iteration.
3185 */
3186 if (ctx->fs->cluster_ratio_bits &&
3187 pb->previous_block &&
3188 (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
3189 EXT2FS_B2C(ctx->fs, pb->previous_block)))
3190 /* Set blk to the beginning of next cluster. */
3191 blk = EXT2FS_C2B(
3192 ctx->fs,
3193 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
3194 else
3195 /* Set blk to the beginning of current cluster. */
3196 blk = EXT2FS_C2B(ctx->fs,
3197 EXT2FS_B2C(ctx->fs, extent.e_pblk));
3198
3199 if (blk < extent.e_pblk + extent.e_len) {
3200 mark_blocks_used(ctx, blk,
3201 extent.e_pblk + extent.e_len - blk);
3202 n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
3203 EXT2FS_CLUSTER_RATIO(ctx->fs));
3204 pb->num_blocks += n;
3205 }
3206 pb->last_block = extent.e_lblk + extent.e_len - 1;
3207 pb->previous_block = extent.e_pblk + extent.e_len - 1;
3208 start_block = pb->last_block = last_lblk;
3209 if (is_leaf && !is_dir &&
3210 !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
3211 pb->last_init_lblock = last_lblk;
3212 next:
3213 pctx->errcode = ext2fs_extent_get(ehandle,
3214 EXT2_EXTENT_NEXT_SIB,
3215 &extent);
3216 }
3217
3218 /* Failed csum but passes checks? Ask to fix checksum. */
3219 if (failed_csum &&
3220 fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
3221 pb->inode_modified = 1;
3222 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
3223 if (pctx->errcode)
3224 return;
3225 }
3226
3227 if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
3228 pctx->errcode = 0;
3229 }
3230
check_blocks_extents(e2fsck_t ctx,struct problem_context * pctx,struct process_block_struct * pb)3231 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
3232 struct process_block_struct *pb)
3233 {
3234 struct ext2_extent_info info;
3235 struct ext2_inode *inode = pctx->inode;
3236 ext2_extent_handle_t ehandle;
3237 ext2_filsys fs = ctx->fs;
3238 ext2_ino_t ino = pctx->ino;
3239 errcode_t retval;
3240 blk64_t eof_lblk;
3241 struct ext3_extent_header *eh;
3242
3243 /* Check for a proper extent header... */
3244 eh = (struct ext3_extent_header *) &inode->i_block[0];
3245 retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
3246 if (retval) {
3247 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
3248 e2fsck_clear_inode(ctx, ino, inode, 0,
3249 "check_blocks_extents");
3250 pctx->errcode = 0;
3251 return;
3252 }
3253
3254 /* ...since this function doesn't fail if i_block is zeroed. */
3255 pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
3256 if (pctx->errcode) {
3257 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
3258 e2fsck_clear_inode(ctx, ino, inode, 0,
3259 "check_blocks_extents");
3260 pctx->errcode = 0;
3261 return;
3262 }
3263
3264 retval = ext2fs_extent_get_info(ehandle, &info);
3265 if (retval == 0) {
3266 int max_depth = info.max_depth;
3267
3268 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
3269 max_depth = MAX_EXTENT_DEPTH_COUNT-1;
3270 ctx->extent_depth_count[max_depth]++;
3271 }
3272
3273 /* Check maximum extent depth */
3274 pctx->blk = info.max_depth;
3275 pctx->blk2 = ext2fs_max_extent_depth(ehandle);
3276 if (pctx->blk2 < pctx->blk &&
3277 fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
3278 pb->eti.force_rebuild = 1;
3279
3280 /* Can we collect extent tree level stats? */
3281 pctx->blk = MAX_EXTENT_DEPTH_COUNT;
3282 if (pctx->blk2 > pctx->blk)
3283 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
3284 memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
3285 pb->eti.ino = pb->ino;
3286
3287 pb->next_lblock = 0;
3288
3289 eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
3290 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
3291 scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
3292 if (pctx->errcode &&
3293 fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
3294 pb->num_blocks = 0;
3295 inode->i_blocks = 0;
3296 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3297 "check_blocks_extents");
3298 pctx->errcode = 0;
3299 }
3300 ext2fs_extent_free(ehandle);
3301
3302 /* Rebuild unless it's a dir and we're rehashing it */
3303 if (LINUX_S_ISDIR(inode->i_mode) &&
3304 e2fsck_dir_will_be_rehashed(ctx, ino))
3305 return;
3306
3307 if (ctx->options & E2F_OPT_CONVERT_BMAP)
3308 e2fsck_rebuild_extents_later(ctx, ino);
3309 else
3310 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
3311 }
3312
3313 /*
3314 * In fact we don't need to check blocks for an inode with inline data
3315 * because this inode doesn't have any blocks. In this function all
3316 * we need to do is add this inode into dblist when it is a directory.
3317 */
check_blocks_inline_data(e2fsck_t ctx,struct problem_context * pctx,struct process_block_struct * pb)3318 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
3319 struct process_block_struct *pb)
3320 {
3321 int flags;
3322 size_t inline_data_size = 0;
3323
3324 if (!pb->is_dir) {
3325 pctx->errcode = 0;
3326 return;
3327 }
3328
3329 /* Process the dirents in i_block[] as the "first" block. */
3330 pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
3331 if (pctx->errcode)
3332 goto err;
3333
3334 /* Process the dirents in the EA as a "second" block. */
3335 flags = ctx->fs->flags;
3336 ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3337 pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
3338 &inline_data_size);
3339 ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3340 (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3341 if (pctx->errcode) {
3342 pctx->errcode = 0;
3343 return;
3344 }
3345
3346 if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
3347 return;
3348
3349 pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
3350 if (pctx->errcode)
3351 goto err;
3352
3353 return;
3354 err:
3355 pctx->blk = 0;
3356 pctx->num = 0;
3357 fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3358 ctx->flags |= E2F_FLAG_ABORT;
3359 }
3360
3361 /*
3362 * This subroutine is called on each inode to account for all of the
3363 * blocks used by that inode.
3364 */
check_blocks(e2fsck_t ctx,struct problem_context * pctx,char * block_buf,const struct ea_quota * ea_ibody_quota)3365 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
3366 char *block_buf, const struct ea_quota *ea_ibody_quota)
3367 {
3368 ext2_filsys fs = ctx->fs;
3369 struct process_block_struct pb;
3370 ext2_ino_t ino = pctx->ino;
3371 struct ext2_inode *inode = pctx->inode;
3372 unsigned bad_size = 0;
3373 int dirty_inode = 0;
3374 int extent_fs;
3375 int inlinedata_fs;
3376 __u64 size;
3377 struct ea_quota ea_block_quota;
3378
3379 pb.ino = ino;
3380 pb.num_blocks = EXT2FS_B2C(ctx->fs,
3381 ea_ibody_quota ? ea_ibody_quota->blocks : 0);
3382 pb.last_block = ~0;
3383 pb.last_init_lblock = -1;
3384 pb.last_db_block = -1;
3385 pb.num_illegal_blocks = 0;
3386 pb.suppress = 0; pb.clear = 0;
3387 pb.fragmented = 0;
3388 pb.compressed = 0;
3389 pb.previous_block = 0;
3390 pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
3391 pb.is_reg = LINUX_S_ISREG(inode->i_mode);
3392 pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
3393 pb.inode = inode;
3394 pb.pctx = pctx;
3395 pb.ctx = ctx;
3396 pb.inode_modified = 0;
3397 pb.eti.force_rebuild = 0;
3398 pctx->ino = ino;
3399 pctx->errcode = 0;
3400
3401 extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
3402 inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
3403
3404 if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
3405 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3406 goto out;
3407 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
3408 }
3409
3410 if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
3411 check_blocks_inline_data(ctx, pctx, &pb);
3412 else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
3413 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
3414 check_blocks_extents(ctx, pctx, &pb);
3415 else {
3416 int flags;
3417 /*
3418 * If we've modified the inode, write it out before
3419 * iterate() tries to use it.
3420 */
3421 if (dirty_inode) {
3422 e2fsck_write_inode(ctx, ino, inode,
3423 "check_blocks");
3424 dirty_inode = 0;
3425 }
3426 flags = fs->flags;
3427 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3428 pctx->errcode = ext2fs_block_iterate3(fs, ino,
3429 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
3430 block_buf, process_block, &pb);
3431 /*
3432 * We do not have uninitialized extents in non extent
3433 * files.
3434 */
3435 pb.last_init_lblock = pb.last_block;
3436 /*
3437 * If iterate() changed a block mapping, we have to
3438 * re-read the inode. If we decide to clear the
3439 * inode after clearing some stuff, we'll re-write the
3440 * bad mappings into the inode!
3441 */
3442 if (pb.inode_modified)
3443 e2fsck_read_inode(ctx, ino, inode,
3444 "check_blocks");
3445 fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3446 (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3447
3448 if (ctx->options & E2F_OPT_CONVERT_BMAP) {
3449 #ifdef DEBUG
3450 printf("bmap rebuild ino=%d\n", ino);
3451 #endif
3452 if (!LINUX_S_ISDIR(inode->i_mode) ||
3453 !e2fsck_dir_will_be_rehashed(ctx, ino))
3454 e2fsck_rebuild_extents_later(ctx, ino);
3455 }
3456 }
3457 }
3458 end_problem_latch(ctx, PR_LATCH_BLOCK);
3459 end_problem_latch(ctx, PR_LATCH_TOOBIG);
3460 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3461 goto out;
3462 if (pctx->errcode)
3463 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
3464
3465 if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
3466 if (LINUX_S_ISDIR(inode->i_mode))
3467 ctx->fs_fragmented_dir++;
3468 else
3469 ctx->fs_fragmented++;
3470 }
3471
3472 if (pb.clear) {
3473 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3474 "check_blocks");
3475 return;
3476 }
3477
3478 if (inode->i_flags & EXT2_INDEX_FL) {
3479 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
3480 inode->i_flags &= ~EXT2_INDEX_FL;
3481 dirty_inode++;
3482 } else {
3483 e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
3484 }
3485 }
3486
3487 if (!pb.num_blocks && pb.is_dir &&
3488 !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
3489 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
3490 e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
3491 ctx->fs_directory_count--;
3492 return;
3493 }
3494 }
3495
3496 if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
3497 (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
3498 !(inode->i_flags & EXT4_EA_INODE_FL)) {
3499 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
3500 ino,
3501 pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
3502 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
3503 ino, (ea_ibody_quota ?
3504 ea_ibody_quota->inodes : 0) +
3505 ea_block_quota.inodes + 1);
3506 }
3507
3508 if (!ext2fs_has_feature_huge_file(fs->super) ||
3509 !(inode->i_flags & EXT4_HUGE_FILE_FL))
3510 pb.num_blocks *= (fs->blocksize / 512);
3511 pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
3512 #if 0
3513 printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
3514 ino, inode->i_size, (unsigned long long) pb.last_block,
3515 (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
3516 (unsigned long long) pb.num_blocks);
3517 #endif
3518 size = EXT2_I_SIZE(inode);
3519 if (pb.is_dir) {
3520 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
3521 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
3522 int flags;
3523 size_t sz = 0;
3524 errcode_t err;
3525
3526 flags = ctx->fs->flags;
3527 ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3528 err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
3529 &sz);
3530 ctx->fs->flags = (flags &
3531 EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3532 (ctx->fs->flags &
3533 ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3534 if (err || sz != size) {
3535 bad_size = 7;
3536 pctx->num = sz;
3537 }
3538 } else if (size & (fs->blocksize - 1))
3539 bad_size = 5;
3540 else if (nblock > (pb.last_block + 1))
3541 bad_size = 1;
3542 else if (nblock < (pb.last_block + 1)) {
3543 if (((pb.last_block + 1) - nblock) >
3544 fs->super->s_prealloc_dir_blocks)
3545 bad_size = 2;
3546 }
3547 } else {
3548 if ((pb.last_init_lblock >= 0) &&
3549 /* Do not allow initialized allocated blocks past i_size*/
3550 (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
3551 !(inode->i_flags & EXT4_VERITY_FL))
3552 bad_size = 3;
3553 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3554 size > ext2_max_sizes[fs->super->s_log_block_size])
3555 /* too big for a direct/indirect-mapped file */
3556 bad_size = 4;
3557 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3558 size >
3559 ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
3560 /* too big for an extent-based file - 32bit ee_block */
3561 bad_size = 6;
3562 }
3563 /* i_size for symlinks is checked elsewhere */
3564 if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
3565 /* Did inline_data set pctx->num earlier? */
3566 if (bad_size != 7)
3567 pctx->num = (pb.last_block + 1) * fs->blocksize;
3568 pctx->group = bad_size;
3569 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
3570 ext2fs_inode_size_set(fs, inode, pctx->num);
3571 if (EXT2_I_SIZE(inode) == 0 &&
3572 (inode->i_flags & EXT4_INLINE_DATA_FL)) {
3573 memset(inode->i_block, 0,
3574 sizeof(inode->i_block));
3575 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
3576 }
3577 dirty_inode++;
3578 }
3579 pctx->num = 0;
3580 }
3581 if (LINUX_S_ISREG(inode->i_mode) &&
3582 ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3583 ctx->large_files++;
3584 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
3585 ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
3586 (ext2fs_has_feature_huge_file(fs->super) &&
3587 (inode->i_flags & EXT4_HUGE_FILE_FL) &&
3588 (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
3589 pctx->num = pb.num_blocks;
3590 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
3591 inode->i_blocks = pb.num_blocks;
3592 inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
3593 dirty_inode++;
3594 }
3595 pctx->num = 0;
3596 }
3597
3598 /*
3599 * The kernel gets mad if we ask it to allocate bigalloc clusters to
3600 * a block mapped file, so rebuild it as an extent file. We can skip
3601 * symlinks because they're never rewritten.
3602 */
3603 if (ext2fs_has_feature_bigalloc(fs->super) &&
3604 (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
3605 ext2fs_inode_data_blocks2(fs, inode) > 0 &&
3606 (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
3607 !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
3608 fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
3609 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
3610 if (pctx->errcode)
3611 goto out;
3612 }
3613
3614 if (ctx->dirs_to_hash && pb.is_dir &&
3615 !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
3616 !(inode->i_flags & EXT2_INDEX_FL) &&
3617 ((inode->i_size / fs->blocksize) >= 3))
3618 e2fsck_rehash_dir_later(ctx, ino);
3619
3620 out:
3621 if (dirty_inode)
3622 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
3623 }
3624
3625 #if 0
3626 /*
3627 * Helper function called by process block when an illegal block is
3628 * found. It returns a description about why the block is illegal
3629 */
3630 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
3631 {
3632 blk64_t super;
3633 int i;
3634 static char problem[80];
3635
3636 super = fs->super->s_first_data_block;
3637 strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
3638 if (block < super) {
3639 sprintf(problem, "< FIRSTBLOCK (%u)", super);
3640 return(problem);
3641 } else if (block >= ext2fs_blocks_count(fs->super)) {
3642 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
3643 return(problem);
3644 }
3645 for (i = 0; i < fs->group_desc_count; i++) {
3646 if (block == super) {
3647 sprintf(problem, "is the superblock in group %d", i);
3648 break;
3649 }
3650 if (block > super &&
3651 block <= (super + fs->desc_blocks)) {
3652 sprintf(problem, "is in the group descriptors "
3653 "of group %d", i);
3654 break;
3655 }
3656 if (block == ext2fs_block_bitmap_loc(fs, i)) {
3657 sprintf(problem, "is the block bitmap of group %d", i);
3658 break;
3659 }
3660 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
3661 sprintf(problem, "is the inode bitmap of group %d", i);
3662 break;
3663 }
3664 if (block >= ext2fs_inode_table_loc(fs, i) &&
3665 (block < ext2fs_inode_table_loc(fs, i)
3666 + fs->inode_blocks_per_group)) {
3667 sprintf(problem, "is in the inode table of group %d",
3668 i);
3669 break;
3670 }
3671 super += fs->super->s_blocks_per_group;
3672 }
3673 return(problem);
3674 }
3675 #endif
3676
3677 /*
3678 * This is a helper function for check_blocks().
3679 */
process_block(ext2_filsys fs,blk64_t * block_nr,e2_blkcnt_t blockcnt,blk64_t ref_block EXT2FS_ATTR ((unused)),int ref_offset EXT2FS_ATTR ((unused)),void * priv_data)3680 static int process_block(ext2_filsys fs,
3681 blk64_t *block_nr,
3682 e2_blkcnt_t blockcnt,
3683 blk64_t ref_block EXT2FS_ATTR((unused)),
3684 int ref_offset EXT2FS_ATTR((unused)),
3685 void *priv_data)
3686 {
3687 struct process_block_struct *p;
3688 struct problem_context *pctx;
3689 blk64_t blk = *block_nr;
3690 int ret_code = 0;
3691 problem_t problem = 0;
3692 e2fsck_t ctx;
3693
3694 p = (struct process_block_struct *) priv_data;
3695 pctx = p->pctx;
3696 ctx = p->ctx;
3697
3698 /*
3699 * For a directory, add logical block zero for processing even if it's
3700 * not mapped or we'll be perennially stuck with broken "." and ".."
3701 * entries.
3702 */
3703 if (p->is_dir && blockcnt == 0 && blk == 0) {
3704 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
3705 if (pctx->errcode) {
3706 pctx->blk = blk;
3707 pctx->num = blockcnt;
3708 goto failed_add_dir_block;
3709 }
3710 p->last_db_block++;
3711 }
3712
3713 if (blk == 0)
3714 return 0;
3715
3716 #if 0
3717 printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
3718 blockcnt);
3719 #endif
3720
3721 /*
3722 * Simplistic fragmentation check. We merely require that the
3723 * file be contiguous. (Which can never be true for really
3724 * big files that are greater than a block group.)
3725 */
3726 if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
3727 if (p->previous_block+1 != blk) {
3728 if (ctx->options & E2F_OPT_FRAGCHECK) {
3729 char type = '?';
3730
3731 if (p->is_dir)
3732 type = 'd';
3733 else if (p->is_reg)
3734 type = 'f';
3735
3736 printf(_("%6lu(%c): expecting %6lu "
3737 "got phys %6lu (blkcnt %lld)\n"),
3738 (unsigned long) pctx->ino, type,
3739 (unsigned long) p->previous_block+1,
3740 (unsigned long) blk,
3741 (long long) blockcnt);
3742 }
3743 p->fragmented = 1;
3744 }
3745 }
3746
3747 if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
3748 !pctx->inode->i_size_high &&
3749 blockcnt > (1 << (21 - fs->super->s_log_block_size)))
3750 problem = PR_1_TOOBIG_DIR;
3751 if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
3752 problem = PR_1_TOOBIG_DIR;
3753 if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
3754 problem = PR_1_TOOBIG_REG;
3755 if (!p->is_dir && !p->is_reg && blockcnt > 0)
3756 problem = PR_1_TOOBIG_SYMLINK;
3757
3758 if (blk < fs->super->s_first_data_block ||
3759 blk >= ext2fs_blocks_count(fs->super))
3760 problem = PR_1_ILLEGAL_BLOCK_NUM;
3761
3762 /*
3763 * If this IND/DIND/TIND block is squatting atop some critical metadata
3764 * (group descriptors, superblock, bitmap, inode table), any write to
3765 * "fix" mapping problems will destroy the metadata. We'll let pass 1b
3766 * fix that and restart fsck.
3767 */
3768 if (blockcnt < 0 &&
3769 p->ino != EXT2_RESIZE_INO &&
3770 blk < ctx->fs->super->s_blocks_count &&
3771 ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
3772 pctx->blk = blk;
3773 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
3774 if ((ctx->options & E2F_OPT_NO) == 0)
3775 ctx->flags |= E2F_FLAG_RESTART_LATER;
3776 }
3777
3778 if (problem) {
3779 p->num_illegal_blocks++;
3780 /*
3781 * A bit of subterfuge here -- we're trying to fix a block
3782 * mapping, but the IND/DIND/TIND block could have collided
3783 * with some critical metadata. So, fix the in-core mapping so
3784 * iterate won't go insane, but return 0 instead of
3785 * BLOCK_CHANGED so that it won't write the remapping out to
3786 * our multiply linked block.
3787 *
3788 * Even if we previously determined that an *IND block
3789 * conflicts with critical metadata, we must still try to
3790 * iterate the *IND block as if it is an *IND block to find and
3791 * mark the blocks it points to. Better to be overly cautious
3792 * with the used_blocks map so that we don't move the *IND
3793 * block to a block that's really in use!
3794 */
3795 if (p->ino != EXT2_RESIZE_INO &&
3796 ref_block != 0 &&
3797 ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3798 ref_block)) {
3799 *block_nr = 0;
3800 return 0;
3801 }
3802 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
3803 if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
3804 p->clear = 1;
3805 return BLOCK_ABORT;
3806 }
3807 if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
3808 p->suppress = 1;
3809 set_latch_flags(PR_LATCH_BLOCK,
3810 PRL_SUPPRESS, 0);
3811 }
3812 }
3813 pctx->blk = blk;
3814 pctx->blkcount = blockcnt;
3815 if (fix_problem(ctx, problem, pctx)) {
3816 blk = *block_nr = 0;
3817 ret_code = BLOCK_CHANGED;
3818 p->inode_modified = 1;
3819 /*
3820 * If the directory block is too big and is beyond the
3821 * end of the FS, don't bother trying to add it for
3822 * processing -- the kernel would never have created a
3823 * directory this large, and we risk an ENOMEM abort.
3824 * In any case, the toobig handler for extent-based
3825 * directories also doesn't feed toobig blocks to
3826 * pass 2.
3827 */
3828 if (problem == PR_1_TOOBIG_DIR)
3829 return ret_code;
3830 goto mark_dir;
3831 } else
3832 return 0;
3833 }
3834
3835 if (p->ino == EXT2_RESIZE_INO) {
3836 /*
3837 * The resize inode has already be sanity checked
3838 * during pass #0 (the superblock checks). All we
3839 * have to do is mark the double indirect block as
3840 * being in use; all of the other blocks are handled
3841 * by mark_table_blocks()).
3842 */
3843 if (blockcnt == BLOCK_COUNT_DIND)
3844 mark_block_used(ctx, blk);
3845 p->num_blocks++;
3846 } else if (!(ctx->fs->cluster_ratio_bits &&
3847 p->previous_block &&
3848 (EXT2FS_B2C(ctx->fs, blk) ==
3849 EXT2FS_B2C(ctx->fs, p->previous_block)) &&
3850 (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
3851 ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
3852 mark_block_used(ctx, blk);
3853 p->num_blocks++;
3854 } else if (has_unaligned_cluster_map(ctx, p->previous_block,
3855 p->last_block, blk, blockcnt)) {
3856 pctx->blk = blockcnt;
3857 pctx->blk2 = blk;
3858 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3859 mark_block_used(ctx, blk);
3860 mark_block_used(ctx, blk);
3861 }
3862 if (blockcnt >= 0)
3863 p->last_block = blockcnt;
3864 p->previous_block = blk;
3865 mark_dir:
3866 if (p->is_dir && (blockcnt >= 0)) {
3867 while (++p->last_db_block < blockcnt) {
3868 pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
3869 p->ino, 0,
3870 p->last_db_block);
3871 if (pctx->errcode) {
3872 pctx->blk = 0;
3873 pctx->num = p->last_db_block;
3874 goto failed_add_dir_block;
3875 }
3876 }
3877 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
3878 blk, blockcnt);
3879 if (pctx->errcode) {
3880 pctx->blk = blk;
3881 pctx->num = blockcnt;
3882 failed_add_dir_block:
3883 fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3884 /* Should never get here */
3885 ctx->flags |= E2F_FLAG_ABORT;
3886 return BLOCK_ABORT;
3887 }
3888 }
3889 return ret_code;
3890 }
3891
process_bad_block(ext2_filsys fs,blk64_t * block_nr,e2_blkcnt_t blockcnt,blk64_t ref_block EXT2FS_ATTR ((unused)),int ref_offset EXT2FS_ATTR ((unused)),void * priv_data)3892 static int process_bad_block(ext2_filsys fs,
3893 blk64_t *block_nr,
3894 e2_blkcnt_t blockcnt,
3895 blk64_t ref_block EXT2FS_ATTR((unused)),
3896 int ref_offset EXT2FS_ATTR((unused)),
3897 void *priv_data)
3898 {
3899 struct process_block_struct *p;
3900 blk64_t blk = *block_nr;
3901 blk64_t first_block;
3902 dgrp_t i;
3903 struct problem_context *pctx;
3904 e2fsck_t ctx;
3905
3906 if (!blk)
3907 return 0;
3908
3909 p = (struct process_block_struct *) priv_data;
3910 ctx = p->ctx;
3911 pctx = p->pctx;
3912
3913 pctx->ino = EXT2_BAD_INO;
3914 pctx->blk = blk;
3915 pctx->blkcount = blockcnt;
3916
3917 if ((blk < fs->super->s_first_data_block) ||
3918 (blk >= ext2fs_blocks_count(fs->super))) {
3919 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
3920 *block_nr = 0;
3921 return BLOCK_CHANGED;
3922 } else
3923 return 0;
3924 }
3925
3926 if (blockcnt < 0) {
3927 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
3928 p->bbcheck = 1;
3929 if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
3930 *block_nr = 0;
3931 return BLOCK_CHANGED;
3932 }
3933 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
3934 blk)) {
3935 p->bbcheck = 1;
3936 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
3937 pctx)) {
3938 *block_nr = 0;
3939 return BLOCK_CHANGED;
3940 }
3941 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3942 return BLOCK_ABORT;
3943 } else
3944 mark_block_used(ctx, blk);
3945 return 0;
3946 }
3947 #if 0
3948 printf ("DEBUG: Marking %u as bad.\n", blk);
3949 #endif
3950 ctx->fs_badblocks_count++;
3951 /*
3952 * If the block is not used, then mark it as used and return.
3953 * If it is already marked as found, this must mean that
3954 * there's an overlap between the filesystem table blocks
3955 * (bitmaps and inode table) and the bad block list.
3956 */
3957 if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
3958 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
3959 return 0;
3960 }
3961 /*
3962 * Try to find the where the filesystem block was used...
3963 */
3964 first_block = fs->super->s_first_data_block;
3965
3966 for (i = 0; i < fs->group_desc_count; i++ ) {
3967 pctx->group = i;
3968 pctx->blk = blk;
3969 if (!ext2fs_bg_has_super(fs, i))
3970 goto skip_super;
3971 if (blk == first_block) {
3972 if (i == 0) {
3973 if (fix_problem(ctx,
3974 PR_1_BAD_PRIMARY_SUPERBLOCK,
3975 pctx)) {
3976 *block_nr = 0;
3977 return BLOCK_CHANGED;
3978 }
3979 return 0;
3980 }
3981 fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
3982 return 0;
3983 }
3984 if ((blk > first_block) &&
3985 (blk <= first_block + fs->desc_blocks)) {
3986 if (i == 0) {
3987 pctx->blk = *block_nr;
3988 if (fix_problem(ctx,
3989 PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
3990 *block_nr = 0;
3991 return BLOCK_CHANGED;
3992 }
3993 return 0;
3994 }
3995 fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
3996 return 0;
3997 }
3998 skip_super:
3999 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
4000 if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
4001 ctx->invalid_block_bitmap_flag[i]++;
4002 ctx->invalid_bitmaps++;
4003 }
4004 return 0;
4005 }
4006 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
4007 if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
4008 ctx->invalid_inode_bitmap_flag[i]++;
4009 ctx->invalid_bitmaps++;
4010 }
4011 return 0;
4012 }
4013 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
4014 (blk < (ext2fs_inode_table_loc(fs, i) +
4015 fs->inode_blocks_per_group))) {
4016 /*
4017 * If there are bad blocks in the inode table,
4018 * the inode scan code will try to do
4019 * something reasonable automatically.
4020 */
4021 return 0;
4022 }
4023 first_block += fs->super->s_blocks_per_group;
4024 }
4025 /*
4026 * If we've gotten to this point, then the only
4027 * possibility is that the bad block inode meta data
4028 * is using a bad block.
4029 */
4030 if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
4031 (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
4032 (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
4033 p->bbcheck = 1;
4034 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
4035 *block_nr = 0;
4036 return BLOCK_CHANGED;
4037 }
4038 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
4039 return BLOCK_ABORT;
4040 return 0;
4041 }
4042
4043 pctx->group = -1;
4044
4045 /* Warn user that the block wasn't claimed */
4046 fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
4047
4048 return 0;
4049 }
4050
new_table_block(e2fsck_t ctx,blk64_t first_block,dgrp_t group,const char * name,int num,blk64_t * new_block)4051 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
4052 const char *name, int num, blk64_t *new_block)
4053 {
4054 ext2_filsys fs = ctx->fs;
4055 dgrp_t last_grp;
4056 blk64_t old_block = *new_block;
4057 blk64_t last_block;
4058 dgrp_t flexbg;
4059 unsigned flexbg_size;
4060 int i, is_flexbg;
4061 char *buf;
4062 struct problem_context pctx;
4063
4064 clear_problem_context(&pctx);
4065
4066 pctx.group = group;
4067 pctx.blk = old_block;
4068 pctx.str = name;
4069
4070 /*
4071 * For flex_bg filesystems, first try to allocate the metadata
4072 * within the flex_bg, and if that fails then try finding the
4073 * space anywhere in the filesystem.
4074 */
4075 is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
4076 if (is_flexbg) {
4077 flexbg_size = 1U << fs->super->s_log_groups_per_flex;
4078 flexbg = group / flexbg_size;
4079 first_block = ext2fs_group_first_block2(fs,
4080 flexbg_size * flexbg);
4081 last_grp = group | (flexbg_size - 1);
4082 if (last_grp >= fs->group_desc_count)
4083 last_grp = fs->group_desc_count - 1;
4084 last_block = ext2fs_group_last_block2(fs, last_grp);
4085 } else
4086 last_block = ext2fs_group_last_block2(fs, group);
4087 pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
4088 num, ctx->block_found_map,
4089 new_block);
4090 if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
4091 pctx.errcode = ext2fs_get_free_blocks2(fs,
4092 fs->super->s_first_data_block,
4093 ext2fs_blocks_count(fs->super),
4094 num, ctx->block_found_map, new_block);
4095 if (pctx.errcode) {
4096 pctx.num = num;
4097 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4098 ext2fs_unmark_valid(fs);
4099 ctx->flags |= E2F_FLAG_ABORT;
4100 return;
4101 }
4102 pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4103 if (pctx.errcode) {
4104 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4105 ext2fs_unmark_valid(fs);
4106 ctx->flags |= E2F_FLAG_ABORT;
4107 return;
4108 }
4109 ext2fs_mark_super_dirty(fs);
4110 fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4111 pctx.blk2 = *new_block;
4112 fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4113 PR_1_RELOC_TO), &pctx);
4114 pctx.blk2 = 0;
4115 for (i = 0; i < num; i++) {
4116 pctx.blk = i;
4117 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4118 if (old_block) {
4119 pctx.errcode = io_channel_read_blk64(fs->io,
4120 old_block + i, 1, buf);
4121 if (pctx.errcode)
4122 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4123 pctx.blk = (*new_block) + i;
4124 pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4125 1, buf);
4126 } else {
4127 pctx.blk = (*new_block) + i;
4128 pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4129 NULL, NULL);
4130 }
4131
4132 if (pctx.errcode)
4133 fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4134 }
4135 ext2fs_free_mem(&buf);
4136 }
4137
4138 /*
4139 * This routine gets called at the end of pass 1 if bad blocks are
4140 * detected in the superblock, group descriptors, inode_bitmaps, or
4141 * block bitmaps. At this point, all of the blocks have been mapped
4142 * out, so we can try to allocate new block(s) to replace the bad
4143 * blocks.
4144 */
handle_fs_bad_blocks(e2fsck_t ctx)4145 static void handle_fs_bad_blocks(e2fsck_t ctx)
4146 {
4147 ext2_filsys fs = ctx->fs;
4148 dgrp_t i;
4149 blk64_t first_block;
4150 blk64_t new_blk;
4151
4152 for (i = 0; i < fs->group_desc_count; i++) {
4153 first_block = ext2fs_group_first_block2(fs, i);
4154
4155 if (ctx->invalid_block_bitmap_flag[i]) {
4156 new_blk = ext2fs_block_bitmap_loc(fs, i);
4157 new_table_block(ctx, first_block, i, _("block bitmap"),
4158 1, &new_blk);
4159 ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4160 }
4161 if (ctx->invalid_inode_bitmap_flag[i]) {
4162 new_blk = ext2fs_inode_bitmap_loc(fs, i);
4163 new_table_block(ctx, first_block, i, _("inode bitmap"),
4164 1, &new_blk);
4165 ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4166 }
4167 if (ctx->invalid_inode_table_flag[i]) {
4168 new_blk = ext2fs_inode_table_loc(fs, i);
4169 new_table_block(ctx, first_block, i, _("inode table"),
4170 fs->inode_blocks_per_group,
4171 &new_blk);
4172 ext2fs_inode_table_loc_set(fs, i, new_blk);
4173 ctx->flags |= E2F_FLAG_RESTART;
4174 }
4175 }
4176 ctx->invalid_bitmaps = 0;
4177 }
4178
4179 /*
4180 * This routine marks all blocks which are used by the superblock,
4181 * group descriptors, inode bitmaps, and block bitmaps.
4182 */
mark_table_blocks(e2fsck_t ctx)4183 static void mark_table_blocks(e2fsck_t ctx)
4184 {
4185 ext2_filsys fs = ctx->fs;
4186 blk64_t b;
4187 dgrp_t i;
4188 unsigned int j;
4189 struct problem_context pctx;
4190
4191 clear_problem_context(&pctx);
4192
4193 for (i = 0; i < fs->group_desc_count; i++) {
4194 pctx.group = i;
4195
4196 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
4197 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
4198
4199 /*
4200 * Mark the blocks used for the inode table
4201 */
4202 if (ext2fs_inode_table_loc(fs, i)) {
4203 for (j = 0, b = ext2fs_inode_table_loc(fs, i);
4204 j < fs->inode_blocks_per_group;
4205 j++, b++) {
4206 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4207 b)) {
4208 pctx.blk = b;
4209 if (!ctx->invalid_inode_table_flag[i] &&
4210 fix_problem(ctx,
4211 PR_1_ITABLE_CONFLICT, &pctx)) {
4212 ctx->invalid_inode_table_flag[i]++;
4213 ctx->invalid_bitmaps++;
4214 }
4215 } else {
4216 ext2fs_mark_block_bitmap2(
4217 ctx->block_found_map, b);
4218 ext2fs_mark_block_bitmap2(
4219 ctx->block_metadata_map, b);
4220 }
4221 }
4222 }
4223
4224 /*
4225 * Mark block used for the block bitmap
4226 */
4227 if (ext2fs_block_bitmap_loc(fs, i)) {
4228 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4229 ext2fs_block_bitmap_loc(fs, i))) {
4230 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
4231 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
4232 ctx->invalid_block_bitmap_flag[i]++;
4233 ctx->invalid_bitmaps++;
4234 }
4235 } else {
4236 ext2fs_mark_block_bitmap2(ctx->block_found_map,
4237 ext2fs_block_bitmap_loc(fs, i));
4238 ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4239 ext2fs_block_bitmap_loc(fs, i));
4240 }
4241 }
4242 /*
4243 * Mark block used for the inode bitmap
4244 */
4245 if (ext2fs_inode_bitmap_loc(fs, i)) {
4246 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4247 ext2fs_inode_bitmap_loc(fs, i))) {
4248 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
4249 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
4250 ctx->invalid_inode_bitmap_flag[i]++;
4251 ctx->invalid_bitmaps++;
4252 }
4253 } else {
4254 ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4255 ext2fs_inode_bitmap_loc(fs, i));
4256 ext2fs_mark_block_bitmap2(ctx->block_found_map,
4257 ext2fs_inode_bitmap_loc(fs, i));
4258 }
4259 }
4260 }
4261 }
4262
4263 /*
4264 * These subroutines short circuits ext2fs_get_blocks and
4265 * ext2fs_check_directory; we use them since we already have the inode
4266 * structure, so there's no point in letting the ext2fs library read
4267 * the inode again.
4268 */
pass1_get_blocks(ext2_filsys fs,ext2_ino_t ino,blk_t * blocks)4269 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
4270 blk_t *blocks)
4271 {
4272 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4273 int i;
4274
4275 if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4276 return EXT2_ET_CALLBACK_NOTHANDLED;
4277
4278 for (i=0; i < EXT2_N_BLOCKS; i++)
4279 blocks[i] = ctx->stashed_inode->i_block[i];
4280 return 0;
4281 }
4282
pass1_read_inode(ext2_filsys fs,ext2_ino_t ino,struct ext2_inode * inode)4283 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
4284 struct ext2_inode *inode)
4285 {
4286 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4287
4288 if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4289 return EXT2_ET_CALLBACK_NOTHANDLED;
4290 *inode = *ctx->stashed_inode;
4291 return 0;
4292 }
4293
pass1_write_inode(ext2_filsys fs,ext2_ino_t ino,struct ext2_inode * inode)4294 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
4295 struct ext2_inode *inode)
4296 {
4297 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4298
4299 if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
4300 (inode != ctx->stashed_inode))
4301 *ctx->stashed_inode = *inode;
4302 return EXT2_ET_CALLBACK_NOTHANDLED;
4303 }
4304
pass1_check_directory(ext2_filsys fs,ext2_ino_t ino)4305 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
4306 {
4307 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4308
4309 if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4310 return EXT2_ET_CALLBACK_NOTHANDLED;
4311
4312 if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
4313 return EXT2_ET_NO_DIRECTORY;
4314 return 0;
4315 }
4316
e2fsck_get_alloc_block(ext2_filsys fs,blk64_t goal,blk64_t * ret)4317 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
4318 blk64_t *ret)
4319 {
4320 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4321 errcode_t retval;
4322 blk64_t new_block;
4323
4324 if (ctx->block_found_map) {
4325 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
4326 &new_block);
4327 if (retval)
4328 return retval;
4329 if (fs->block_map) {
4330 ext2fs_mark_block_bitmap2(fs->block_map, new_block);
4331 ext2fs_mark_bb_dirty(fs);
4332 }
4333 } else {
4334 if (!fs->block_map) {
4335 retval = ext2fs_read_block_bitmap(fs);
4336 if (retval)
4337 return retval;
4338 }
4339
4340 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
4341 if (retval)
4342 return retval;
4343 }
4344
4345 *ret = new_block;
4346 return (0);
4347 }
4348
e2fsck_new_range(ext2_filsys fs,int flags,blk64_t goal,blk64_t len,blk64_t * pblk,blk64_t * plen)4349 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
4350 blk64_t len, blk64_t *pblk, blk64_t *plen)
4351 {
4352 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4353 errcode_t retval;
4354
4355 if (ctx->block_found_map)
4356 return ext2fs_new_range(fs, flags, goal, len,
4357 ctx->block_found_map, pblk, plen);
4358
4359 if (!fs->block_map) {
4360 retval = ext2fs_read_block_bitmap(fs);
4361 if (retval)
4362 return retval;
4363 }
4364
4365 return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
4366 pblk, plen);
4367 }
4368
e2fsck_block_alloc_stats(ext2_filsys fs,blk64_t blk,int inuse)4369 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
4370 {
4371 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4372
4373 /* Never free a critical metadata block */
4374 if (ctx->block_found_map &&
4375 ctx->block_metadata_map &&
4376 inuse < 0 &&
4377 ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
4378 return;
4379
4380 if (ctx->block_found_map) {
4381 if (inuse > 0)
4382 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4383 else
4384 ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
4385 }
4386 }
4387
e2fsck_block_alloc_stats_range(ext2_filsys fs,blk64_t blk,blk_t num,int inuse)4388 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
4389 blk_t num, int inuse)
4390 {
4391 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4392
4393 /* Never free a critical metadata block */
4394 if (ctx->block_found_map &&
4395 ctx->block_metadata_map &&
4396 inuse < 0 &&
4397 ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
4398 return;
4399
4400 if (ctx->block_found_map) {
4401 if (inuse > 0)
4402 ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
4403 blk, num);
4404 else
4405 ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
4406 blk, num);
4407 }
4408 }
4409
e2fsck_use_inode_shortcuts(e2fsck_t ctx,int use_shortcuts)4410 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
4411 {
4412 ext2_filsys fs = ctx->fs;
4413
4414 if (use_shortcuts) {
4415 fs->get_blocks = pass1_get_blocks;
4416 fs->check_directory = pass1_check_directory;
4417 fs->read_inode = pass1_read_inode;
4418 fs->write_inode = pass1_write_inode;
4419 ctx->stashed_ino = 0;
4420 } else {
4421 fs->get_blocks = 0;
4422 fs->check_directory = 0;
4423 fs->read_inode = 0;
4424 fs->write_inode = 0;
4425 }
4426 }
4427
e2fsck_intercept_block_allocations(e2fsck_t ctx)4428 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
4429 {
4430 ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
4431 ext2fs_set_block_alloc_stats_callback(ctx->fs,
4432 e2fsck_block_alloc_stats, 0);
4433 ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
4434 ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
4435 e2fsck_block_alloc_stats_range, NULL);
4436 }
4437