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, ino_t ino);
83 static void add_casefolded_dir(e2fsck_t ctx, 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;
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_entry2(ctx->fs, entry, NULL, &hash);
347 if (retval) {
348 com_err("check_large_ea_inode", retval,
349 _("while hashing entry with e_value_inum = %u"),
350 entry->e_value_inum);
351 fatal_error(ctx, 0);
352 }
353
354 if (hash == entry->e_hash) {
355 *quota_blocks = size_to_quota_blocks(ctx->fs,
356 entry->e_value_size);
357 } else {
358 /* This might be an old Lustre-style ea_inode reference. */
359 if (inode.i_mtime == pctx->ino &&
360 inode.i_generation == pctx->inode->i_generation) {
361 *quota_blocks = 0;
362 } else {
363 /* If target inode is also missing EA_INODE flag,
364 * this is likely to be a bad reference.
365 */
366 if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
367 pctx->num = entry->e_value_inum;
368 return PR_1_ATTR_VALUE_EA_INODE;
369 } else {
370 pctx->num = entry->e_hash;
371 return PR_1_ATTR_HASH;
372 }
373 }
374 }
375
376 if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
377 pctx->num = entry->e_value_inum;
378 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
379 inode.i_flags |= EXT4_EA_INODE_FL;
380 ext2fs_write_inode(ctx->fs, entry->e_value_inum,
381 &inode);
382 } else {
383 return PR_1_ATTR_NO_EA_INODE_FL;
384 }
385 }
386 return 0;
387 }
388
inc_ea_inode_refs(e2fsck_t ctx,struct problem_context * pctx,struct ext2_ext_attr_entry * first,void * end)389 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
390 struct ext2_ext_attr_entry *first, void *end)
391 {
392 struct ext2_ext_attr_entry *entry;
393
394 for (entry = first;
395 (void *)entry < end && !EXT2_EXT_IS_LAST_ENTRY(entry);
396 entry = EXT2_EXT_ATTR_NEXT(entry)) {
397 if (!entry->e_value_inum)
398 continue;
399 if (!ctx->ea_inode_refs) {
400 pctx->errcode = ea_refcount_create(0,
401 &ctx->ea_inode_refs);
402 if (pctx->errcode) {
403 pctx->num = 4;
404 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
405 ctx->flags |= E2F_FLAG_ABORT;
406 return;
407 }
408 }
409 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
410 0);
411 }
412 }
413
check_ea_in_inode(e2fsck_t ctx,struct problem_context * pctx,struct ea_quota * ea_ibody_quota)414 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
415 struct ea_quota *ea_ibody_quota)
416 {
417 struct ext2_super_block *sb = ctx->fs->super;
418 struct ext2_inode_large *inode;
419 struct ext2_ext_attr_entry *entry;
420 char *start, *header, *end;
421 unsigned int storage_size, remain;
422 problem_t problem = 0;
423 region_t region = 0;
424
425 ea_ibody_quota->blocks = 0;
426 ea_ibody_quota->inodes = 0;
427
428 inode = (struct ext2_inode_large *) pctx->inode;
429 storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
430 inode->i_extra_isize;
431 header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
432 inode->i_extra_isize;
433 end = header + storage_size;
434 start = header + sizeof(__u32);
435 entry = (struct ext2_ext_attr_entry *) start;
436
437 /* scan all entry's headers first */
438
439 /* take finish entry 0UL into account */
440 remain = storage_size - sizeof(__u32);
441
442 region = region_create(0, storage_size);
443 if (!region) {
444 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
445 problem = 0;
446 ctx->flags |= E2F_FLAG_ABORT;
447 return;
448 }
449 if (region_allocate(region, 0, sizeof(__u32))) {
450 problem = PR_1_INODE_EA_ALLOC_COLLISION;
451 goto fix;
452 }
453
454 while (remain >= sizeof(struct ext2_ext_attr_entry) &&
455 !EXT2_EXT_IS_LAST_ENTRY(entry)) {
456 __u32 hash;
457
458 if (region_allocate(region, (char *)entry - (char *)header,
459 EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
460 problem = PR_1_INODE_EA_ALLOC_COLLISION;
461 goto fix;
462 }
463
464 /* header eats this space */
465 remain -= sizeof(struct ext2_ext_attr_entry);
466
467 /* is attribute name valid? */
468 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
469 pctx->num = entry->e_name_len;
470 problem = PR_1_ATTR_NAME_LEN;
471 goto fix;
472 }
473
474 /* attribute len eats this space */
475 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
476
477 if (entry->e_value_inum == 0) {
478 /* check value size */
479 if (entry->e_value_size > remain) {
480 pctx->num = entry->e_value_size;
481 problem = PR_1_ATTR_VALUE_SIZE;
482 goto fix;
483 }
484
485 if (entry->e_value_size &&
486 region_allocate(region,
487 sizeof(__u32) + entry->e_value_offs,
488 EXT2_EXT_ATTR_SIZE(
489 entry->e_value_size))) {
490 problem = PR_1_INODE_EA_ALLOC_COLLISION;
491 goto fix;
492 }
493
494 hash = ext2fs_ext_attr_hash_entry(entry,
495 start + entry->e_value_offs);
496
497 /* e_hash may be 0 in older inode's ea */
498 if (entry->e_hash != 0 && entry->e_hash != hash) {
499 pctx->num = entry->e_hash;
500 problem = PR_1_ATTR_HASH;
501 goto fix;
502 }
503 } else {
504 blk64_t quota_blocks;
505
506 problem = check_large_ea_inode(ctx, entry, pctx,
507 "a_blocks);
508 if (problem != 0)
509 goto fix;
510
511 ea_ibody_quota->blocks += quota_blocks;
512 ea_ibody_quota->inodes++;
513 }
514
515 /* If EA value is stored in external inode then it does not
516 * consume space here */
517 if (entry->e_value_inum == 0)
518 remain -= entry->e_value_size;
519
520 entry = EXT2_EXT_ATTR_NEXT(entry);
521 }
522
523 if (region_allocate(region, (char *)entry - (char *)header,
524 sizeof(__u32))) {
525 problem = PR_1_INODE_EA_ALLOC_COLLISION;
526 goto fix;
527 }
528 fix:
529 if (region)
530 region_free(region);
531 /*
532 * it seems like a corruption. it's very unlikely we could repair
533 * EA(s) in automatic fashion -bzzz
534 */
535 if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
536 inc_ea_inode_refs(ctx, pctx,
537 (struct ext2_ext_attr_entry *)start, end);
538 return;
539 }
540
541 /* simply remove all possible EA(s) */
542 *((__u32 *)header) = 0UL;
543 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
544 EXT2_INODE_SIZE(sb), "pass1");
545 ea_ibody_quota->blocks = 0;
546 ea_ibody_quota->inodes = 0;
547 }
548
check_inode_extra_negative_epoch(__u32 xtime,__u32 extra)549 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
550 return (xtime & (1U << 31)) != 0 &&
551 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
552 }
553
554 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
555 check_inode_extra_negative_epoch(inode->i_##xtime, \
556 inode->i_##xtime##_extra)
557
558 /* When today's date is earlier than 2242, we assume that atimes,
559 * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
560 * actually pre-1970 dates mis-encoded.
561 */
562 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
563
check_inode_extra_space(e2fsck_t ctx,struct problem_context * pctx,struct ea_quota * ea_ibody_quota)564 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
565 struct ea_quota *ea_ibody_quota)
566 {
567 struct ext2_super_block *sb = ctx->fs->super;
568 struct ext2_inode_large *inode;
569 __u32 *eamagic;
570 int min, max;
571
572 ea_ibody_quota->blocks = 0;
573 ea_ibody_quota->inodes = 0;
574
575 inode = (struct ext2_inode_large *) pctx->inode;
576 if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
577 /* this isn't large inode. so, nothing to check */
578 return;
579 }
580
581 #if 0
582 printf("inode #%u, i_extra_size %d\n", pctx->ino,
583 inode->i_extra_isize);
584 #endif
585 /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
586 min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
587 max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
588 /*
589 * For now we will allow i_extra_isize to be 0, but really
590 * implementations should never allow i_extra_isize to be 0
591 */
592 if (inode->i_extra_isize &&
593 (inode->i_extra_isize < min || inode->i_extra_isize > max ||
594 inode->i_extra_isize & 3)) {
595 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
596 return;
597 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
598 inode->i_extra_isize = sb->s_want_extra_isize;
599 else
600 inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
601 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
602 EXT2_INODE_SIZE(sb), "pass1");
603 }
604
605 /* check if there is no place for an EA header */
606 if (inode->i_extra_isize >= max - sizeof(__u32))
607 return;
608
609 eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
610 inode->i_extra_isize);
611 if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
612 /* it seems inode has an extended attribute(s) in body */
613 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
614 }
615
616 /*
617 * If the inode's extended atime (ctime, crtime, mtime) is stored in
618 * the old, invalid format, repair it.
619 */
620 if (((sizeof(time_t) <= 4) ||
621 (((sizeof(time_t) > 4) &&
622 ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
623 (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
624 CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
625 CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
626 CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
627
628 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
629 return;
630
631 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
632 inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
633 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
634 inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
635 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
636 inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
637 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
638 inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
639 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
640 EXT2_INODE_SIZE(sb), "pass1");
641 }
642
643 }
644
645 /*
646 * Check to see if the inode might really be a directory, despite i_mode
647 *
648 * This is a lot of complexity for something for which I'm not really
649 * convinced happens frequently in the wild. If for any reason this
650 * causes any problems, take this code out.
651 * [tytso:20070331.0827EDT]
652 */
check_is_really_dir(e2fsck_t ctx,struct problem_context * pctx,char * buf)653 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
654 char *buf)
655 {
656 struct ext2_inode *inode = pctx->inode;
657 struct ext2_dir_entry *dirent;
658 errcode_t retval;
659 blk64_t blk;
660 unsigned int i, rec_len, not_device = 0;
661 int extent_fs;
662 int inlinedata_fs;
663
664 /*
665 * If the mode looks OK, we believe it. If the first block in
666 * the i_block array is 0, this cannot be a directory. If the
667 * inode is extent-mapped, it is still the case that the latter
668 * cannot be 0 - the magic number in the extent header would make
669 * it nonzero.
670 */
671 if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
672 LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
673 return;
674
675 /*
676 * Check the block numbers in the i_block array for validity:
677 * zero blocks are skipped (but the first one cannot be zero -
678 * see above), other blocks are checked against the first and
679 * max data blocks (from the the superblock) and against the
680 * block bitmap. Any invalid block found means this cannot be
681 * a directory.
682 *
683 * If there are non-zero blocks past the fourth entry, then
684 * this cannot be a device file: we remember that for the next
685 * check.
686 *
687 * For extent mapped files, we don't do any sanity checking:
688 * just try to get the phys block of logical block 0 and run
689 * with it.
690 *
691 * For inline data files, we just try to get the size of inline
692 * data. If it's true, we will treat it as a directory.
693 */
694
695 extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
696 inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
697 if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
698 size_t size;
699 __u32 dotdot;
700 unsigned int rec_len2;
701 struct ext2_dir_entry de;
702
703 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
704 return;
705 /*
706 * If the size isn't a multiple of 4, it's probably not a
707 * directory??
708 */
709 if (size & 3)
710 return;
711 /*
712 * If the first 10 bytes don't look like a directory entry,
713 * it's probably not a directory.
714 */
715 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
716 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
717 EXT2_DIR_REC_LEN(0));
718 dotdot = ext2fs_le32_to_cpu(dotdot);
719 de.inode = ext2fs_le32_to_cpu(de.inode);
720 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
721 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
722 if (dotdot >= ctx->fs->super->s_inodes_count ||
723 (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
724 dotdot != EXT2_ROOT_INO) ||
725 de.inode >= ctx->fs->super->s_inodes_count ||
726 (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
727 de.inode != 0) ||
728 rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
729 EXT4_INLINE_DATA_DOTDOT_SIZE)
730 return;
731 /* device files never have a "system.data" entry */
732 goto isdir;
733 } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
734 /* extent mapped */
735 if (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
736 &blk))
737 return;
738 /* device files are never extent mapped */
739 not_device++;
740 } else {
741 for (i=0; i < EXT2_N_BLOCKS; i++) {
742 blk = inode->i_block[i];
743 if (!blk)
744 continue;
745 if (i >= 4)
746 not_device++;
747
748 if (blk < ctx->fs->super->s_first_data_block ||
749 blk >= ext2fs_blocks_count(ctx->fs->super) ||
750 ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
751 blk))
752 return; /* Invalid block, can't be dir */
753 }
754 blk = inode->i_block[0];
755 }
756
757 /*
758 * If the mode says this is a device file and the i_links_count field
759 * is sane and we have not ruled it out as a device file previously,
760 * we declare it a device file, not a directory.
761 */
762 if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
763 (inode->i_links_count == 1) && !not_device)
764 return;
765
766 /* read the first block */
767 ehandler_operation(_("reading directory block"));
768 retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
769 ehandler_operation(0);
770 if (retval)
771 return;
772
773 dirent = (struct ext2_dir_entry *) buf;
774 retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
775 if (retval)
776 return;
777 if ((ext2fs_dirent_name_len(dirent) != 1) ||
778 (dirent->name[0] != '.') ||
779 (dirent->inode != pctx->ino) ||
780 (rec_len < 12) ||
781 (rec_len % 4) ||
782 (rec_len >= ctx->fs->blocksize - 12))
783 return;
784
785 dirent = (struct ext2_dir_entry *) (buf + rec_len);
786 retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
787 if (retval)
788 return;
789 if ((ext2fs_dirent_name_len(dirent) != 2) ||
790 (dirent->name[0] != '.') ||
791 (dirent->name[1] != '.') ||
792 (rec_len < 12) ||
793 (rec_len % 4))
794 return;
795
796 isdir:
797 if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
798 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
799 e2fsck_write_inode_full(ctx, pctx->ino, inode,
800 EXT2_INODE_SIZE(ctx->fs->super),
801 "check_is_really_dir");
802 }
803 }
804
e2fsck_setup_icount(e2fsck_t ctx,const char * icount_name,int flags,ext2_icount_t hint,ext2_icount_t * ret)805 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
806 int flags, ext2_icount_t hint,
807 ext2_icount_t *ret)
808 {
809 unsigned int threshold;
810 unsigned int save_type;
811 ext2_ino_t num_dirs;
812 errcode_t retval;
813 char *tdb_dir;
814 int enable;
815
816 *ret = 0;
817
818 profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
819 &tdb_dir);
820 profile_get_uint(ctx->profile, "scratch_files",
821 "numdirs_threshold", 0, 0, &threshold);
822 profile_get_boolean(ctx->profile, "scratch_files",
823 "icount", 0, 1, &enable);
824
825 retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
826 if (retval)
827 num_dirs = 1024; /* Guess */
828
829 if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
830 (!threshold || num_dirs > threshold)) {
831 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
832 flags, ret);
833 if (retval == 0)
834 return 0;
835 }
836 e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
837 &save_type);
838 if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
839 flags |= EXT2_ICOUNT_OPT_FULLMAP;
840 retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
841 ctx->fs->default_bitmap_type = save_type;
842 return retval;
843 }
844
recheck_bad_inode_checksum(ext2_filsys fs,ext2_ino_t ino,e2fsck_t ctx,struct problem_context * pctx)845 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
846 e2fsck_t ctx,
847 struct problem_context *pctx)
848 {
849 errcode_t retval;
850 struct ext2_inode_large inode;
851
852 /*
853 * Reread inode. If we don't see checksum error, then this inode
854 * has been fixed elsewhere.
855 */
856 ctx->stashed_ino = 0;
857 retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
858 sizeof(inode));
859 if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
860 return retval;
861 if (!retval)
862 return 0;
863
864 /*
865 * Checksum still doesn't match. That implies that the inode passes
866 * all the sanity checks, so maybe the checksum is simply corrupt.
867 * See if the user will go for fixing that.
868 */
869 if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
870 return 0;
871
872 retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
873 sizeof(inode));
874 return retval;
875 }
876
reserve_block_for_root_repair(e2fsck_t ctx)877 static void reserve_block_for_root_repair(e2fsck_t ctx)
878 {
879 blk64_t blk = 0;
880 errcode_t err;
881 ext2_filsys fs = ctx->fs;
882
883 ctx->root_repair_block = 0;
884 if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
885 return;
886
887 err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
888 if (err)
889 return;
890 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
891 ctx->root_repair_block = blk;
892 }
893
reserve_block_for_lnf_repair(e2fsck_t ctx)894 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
895 {
896 blk64_t blk = 0;
897 errcode_t err;
898 ext2_filsys fs = ctx->fs;
899 static const char name[] = "lost+found";
900 ext2_ino_t ino;
901
902 ctx->lnf_repair_block = 0;
903 if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
904 return;
905
906 err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
907 if (err)
908 return;
909 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
910 ctx->lnf_repair_block = blk;
911 }
912
get_inline_data_ea_size(ext2_filsys fs,ext2_ino_t ino,size_t * sz)913 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
914 size_t *sz)
915 {
916 void *p;
917 struct ext2_xattr_handle *handle;
918 errcode_t retval;
919
920 retval = ext2fs_xattrs_open(fs, ino, &handle);
921 if (retval)
922 return retval;
923
924 retval = ext2fs_xattrs_read(handle);
925 if (retval)
926 goto err;
927
928 retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
929 if (retval)
930 goto err;
931 ext2fs_free_mem(&p);
932 err:
933 (void) ext2fs_xattrs_close(&handle);
934 return retval;
935 }
936
finish_processing_inode(e2fsck_t ctx,ext2_ino_t ino,struct problem_context * pctx,int failed_csum)937 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
938 struct problem_context *pctx,
939 int failed_csum)
940 {
941 if (!failed_csum)
942 return;
943
944 /*
945 * If the inode failed the checksum and the user didn't
946 * clear the inode, test the checksum again -- if it still
947 * fails, ask the user if the checksum should be corrected.
948 */
949 pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
950 if (pctx->errcode)
951 ctx->flags |= E2F_FLAG_ABORT;
952 }
953 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
954 do { \
955 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
956 if ((ctx)->flags & E2F_FLAG_ABORT) \
957 return; \
958 } while (0)
959
could_be_block_map(ext2_filsys fs,struct ext2_inode * inode)960 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
961 {
962 __u32 x;
963 int i;
964
965 for (i = 0; i < EXT2_N_BLOCKS; i++) {
966 x = inode->i_block[i];
967 #ifdef WORDS_BIGENDIAN
968 x = ext2fs_swab32(x);
969 #endif
970 if (x >= ext2fs_blocks_count(fs->super))
971 return 0;
972 }
973
974 return 1;
975 }
976
977 /*
978 * Figure out what to do with an inode that has both extents and inline data
979 * inode flags set. Returns -1 if we decide to erase the inode, 0 otherwise.
980 */
fix_inline_data_extents_file(e2fsck_t ctx,ext2_ino_t ino,struct ext2_inode * inode,int inode_size,struct problem_context * pctx)981 static int fix_inline_data_extents_file(e2fsck_t ctx,
982 ext2_ino_t ino,
983 struct ext2_inode *inode,
984 int inode_size,
985 struct problem_context *pctx)
986 {
987 size_t max_inline_ea_size;
988 ext2_filsys fs = ctx->fs;
989 int dirty = 0;
990
991 /* Both feature flags not set? Just run the regular checks */
992 if (!ext2fs_has_feature_extents(fs->super) &&
993 !ext2fs_has_feature_inline_data(fs->super))
994 return 0;
995
996 /* Clear both flags if it's a special file */
997 if (LINUX_S_ISCHR(inode->i_mode) ||
998 LINUX_S_ISBLK(inode->i_mode) ||
999 LINUX_S_ISFIFO(inode->i_mode) ||
1000 LINUX_S_ISSOCK(inode->i_mode)) {
1001 check_extents_inlinedata(ctx, pctx);
1002 return 0;
1003 }
1004
1005 /* If it looks like an extent tree, try to clear inlinedata */
1006 if (ext2fs_extent_header_verify(inode->i_block,
1007 sizeof(inode->i_block)) == 0 &&
1008 fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1009 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1010 dirty = 1;
1011 goto out;
1012 }
1013
1014 /* If it looks short enough to be inline data, try to clear extents */
1015 if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1016 max_inline_ea_size = inode_size -
1017 (EXT2_GOOD_OLD_INODE_SIZE +
1018 ((struct ext2_inode_large *)inode)->i_extra_isize);
1019 else
1020 max_inline_ea_size = 0;
1021 if (EXT2_I_SIZE(inode) <
1022 EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1023 fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1024 inode->i_flags &= ~EXT4_EXTENTS_FL;
1025 dirty = 1;
1026 goto out;
1027 }
1028
1029 /*
1030 * Too big for inline data, but no evidence of extent tree -
1031 * maybe it's a block map file? If the mappings all look valid?
1032 */
1033 if (could_be_block_map(fs, inode) &&
1034 fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1035 #ifdef WORDS_BIGENDIAN
1036 int i;
1037
1038 for (i = 0; i < EXT2_N_BLOCKS; i++)
1039 inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1040 #endif
1041
1042 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1043 dirty = 1;
1044 goto out;
1045 }
1046
1047 /* Oh well, just clear the busted inode. */
1048 if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1049 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1050 return -1;
1051 }
1052
1053 out:
1054 if (dirty)
1055 e2fsck_write_inode(ctx, ino, inode, "pass1");
1056
1057 return 0;
1058 }
1059
pass1_readahead(e2fsck_t ctx,dgrp_t * group,ext2_ino_t * next_ino)1060 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1061 {
1062 ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1063 dgrp_t start = *group, grp;
1064 blk64_t blocks_to_read = 0;
1065 errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1066
1067 if (ctx->readahead_kb == 0)
1068 goto out;
1069
1070 /* Keep iterating groups until we have enough to readahead */
1071 inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1072 for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1073 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1074 continue;
1075 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1076 ext2fs_bg_itable_unused(ctx->fs, grp);
1077 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1078 inodes_per_block;
1079 if (blocks_to_read * ctx->fs->blocksize >
1080 ctx->readahead_kb * 1024)
1081 break;
1082 }
1083
1084 err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1085 grp - start + 1);
1086 if (err == EAGAIN) {
1087 ctx->readahead_kb /= 2;
1088 err = 0;
1089 }
1090
1091 out:
1092 if (err) {
1093 /* Error; disable itable readahead */
1094 *group = ctx->fs->group_desc_count;
1095 *next_ino = ctx->fs->super->s_inodes_count;
1096 } else {
1097 /*
1098 * Don't do more readahead until we've reached the first inode
1099 * of the last inode scan buffer block for the last group.
1100 */
1101 *group = grp + 1;
1102 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1103 ctx->inode_buffer_blocks :
1104 EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1105 ctx->fs->blocksize /
1106 EXT2_INODE_SIZE(ctx->fs->super);
1107 inodes_in_group--;
1108 *next_ino = inodes_in_group -
1109 (inodes_in_group % inodes_per_buffer) + 1 +
1110 (grp * ctx->fs->super->s_inodes_per_group);
1111 }
1112 }
1113
1114 /*
1115 * Check if the passed ino is one of the used superblock quota inodes.
1116 *
1117 * Before the quota inodes were journaled, older superblock quota inodes
1118 * were just regular files in the filesystem and not reserved inodes. This
1119 * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1120 * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1121 */
quota_inum_is_super(struct ext2_super_block * sb,ext2_ino_t ino)1122 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1123 {
1124 enum quota_type qtype;
1125
1126 for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1127 if (*quota_sb_inump(sb, qtype) == ino)
1128 return 1;
1129
1130 return 0;
1131 }
1132
1133 /*
1134 * Check if the passed ino is one of the reserved quota inodes.
1135 * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1136 * inodes. These inodes may or may not be in use by the quota feature.
1137 */
quota_inum_is_reserved(ext2_filsys fs,ext2_ino_t ino)1138 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1139 {
1140 enum quota_type qtype;
1141
1142 for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1143 if (quota_type2inum(qtype, fs->super) == ino)
1144 return 1;
1145
1146 return 0;
1147 }
1148
e2fsck_pass1(e2fsck_t ctx)1149 void e2fsck_pass1(e2fsck_t ctx)
1150 {
1151 int i;
1152 __u64 max_sizes;
1153 ext2_filsys fs = ctx->fs;
1154 ext2_ino_t ino = 0;
1155 struct ext2_inode *inode = NULL;
1156 ext2_inode_scan scan = NULL;
1157 char *block_buf = NULL;
1158 #ifdef RESOURCE_TRACK
1159 struct resource_track rtrack;
1160 #endif
1161 unsigned char frag, fsize;
1162 struct problem_context pctx;
1163 struct scan_callback_struct scan_struct;
1164 struct ext2_super_block *sb = ctx->fs->super;
1165 const char *old_op;
1166 const char *eop_next_inode = _("getting next inode from scan");
1167 int imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1168 int low_dtime_check = 1;
1169 unsigned int inode_size = EXT2_INODE_SIZE(fs->super);
1170 unsigned int bufsize;
1171 int failed_csum = 0;
1172 ext2_ino_t ino_threshold = 0;
1173 dgrp_t ra_group = 0;
1174 struct ea_quota ea_ibody_quota;
1175
1176 init_resource_track(&rtrack, ctx->fs->io);
1177 clear_problem_context(&pctx);
1178
1179 /* If we can do readahead, figure out how many groups to pull in. */
1180 if (!e2fsck_can_readahead(ctx->fs))
1181 ctx->readahead_kb = 0;
1182 else if (ctx->readahead_kb == ~0ULL)
1183 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1184 pass1_readahead(ctx, &ra_group, &ino_threshold);
1185
1186 if (!(ctx->options & E2F_OPT_PREEN))
1187 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1188
1189 if (ext2fs_has_feature_dir_index(fs->super) &&
1190 !(ctx->options & E2F_OPT_NO)) {
1191 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1192 ctx->dirs_to_hash = 0;
1193 }
1194
1195 #ifdef MTRACE
1196 mtrace_print("Pass 1");
1197 #endif
1198
1199 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1200
1201 for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1202 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1203 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1204 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1205 max_sizes = (max_sizes * (1UL << i));
1206 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1207 }
1208 #undef EXT2_BPP
1209
1210 imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1211 extent_fs = ext2fs_has_feature_extents(sb);
1212 inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1213 casefold_fs = ext2fs_has_feature_casefold(sb);
1214
1215 /*
1216 * Allocate bitmaps structures
1217 */
1218 pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1219 EXT2FS_BMAP64_RBTREE,
1220 "inode_used_map",
1221 &ctx->inode_used_map);
1222 if (pctx.errcode) {
1223 pctx.num = 1;
1224 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1225 ctx->flags |= E2F_FLAG_ABORT;
1226 return;
1227 }
1228 pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1229 _("directory inode map"),
1230 EXT2FS_BMAP64_AUTODIR,
1231 "inode_dir_map", &ctx->inode_dir_map);
1232 if (pctx.errcode) {
1233 pctx.num = 2;
1234 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1235 ctx->flags |= E2F_FLAG_ABORT;
1236 return;
1237 }
1238 pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1239 _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1240 "inode_reg_map", &ctx->inode_reg_map);
1241 if (pctx.errcode) {
1242 pctx.num = 6;
1243 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1244 ctx->flags |= E2F_FLAG_ABORT;
1245 return;
1246 }
1247 pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1248 _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1249 "block_found_map", &ctx->block_found_map);
1250 if (pctx.errcode) {
1251 pctx.num = 1;
1252 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1253 ctx->flags |= E2F_FLAG_ABORT;
1254 return;
1255 }
1256 pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1257 _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1258 "block_metadata_map", &ctx->block_metadata_map);
1259 if (pctx.errcode) {
1260 pctx.num = 1;
1261 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1262 ctx->flags |= E2F_FLAG_ABORT;
1263 return;
1264 }
1265 if (casefold_fs) {
1266 pctx.errcode =
1267 e2fsck_allocate_inode_bitmap(fs,
1268 _("inode casefold map"),
1269 EXT2FS_BMAP64_RBTREE,
1270 "inode_casefold_map",
1271 &ctx->inode_casefold_map);
1272 if (pctx.errcode) {
1273 pctx.num = 1;
1274 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1275 ctx->flags |= E2F_FLAG_ABORT;
1276 return;
1277 }
1278 }
1279 pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1280 &ctx->inode_link_info);
1281 if (pctx.errcode) {
1282 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1283 ctx->flags |= E2F_FLAG_ABORT;
1284 return;
1285 }
1286 bufsize = inode_size;
1287 if (bufsize < sizeof(struct ext2_inode_large))
1288 bufsize = sizeof(struct ext2_inode_large);
1289 inode = (struct ext2_inode *)
1290 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1291
1292 inodes_to_process = (struct process_inode_block *)
1293 e2fsck_allocate_memory(ctx,
1294 (ctx->process_inode_size *
1295 sizeof(struct process_inode_block)),
1296 "array of inodes to process");
1297 process_inode_count = 0;
1298
1299 pctx.errcode = ext2fs_init_dblist(fs, 0);
1300 if (pctx.errcode) {
1301 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1302 ctx->flags |= E2F_FLAG_ABORT;
1303 goto endit;
1304 }
1305
1306 /*
1307 * If the last orphan field is set, clear it, since the pass1
1308 * processing will automatically find and clear the orphans.
1309 * In the future, we may want to try using the last_orphan
1310 * linked list ourselves, but for now, we clear it so that the
1311 * ext3 mount code won't get confused.
1312 */
1313 if (!(ctx->options & E2F_OPT_READONLY)) {
1314 if (fs->super->s_last_orphan) {
1315 fs->super->s_last_orphan = 0;
1316 ext2fs_mark_super_dirty(fs);
1317 }
1318 }
1319
1320 mark_table_blocks(ctx);
1321 pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1322 &ctx->block_found_map);
1323 if (pctx.errcode) {
1324 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1325 ctx->flags |= E2F_FLAG_ABORT;
1326 goto endit;
1327 }
1328 block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1329 "block interate buffer");
1330 if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1331 e2fsck_use_inode_shortcuts(ctx, 1);
1332 e2fsck_intercept_block_allocations(ctx);
1333 old_op = ehandler_operation(_("opening inode scan"));
1334 pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1335 &scan);
1336 ehandler_operation(old_op);
1337 if (pctx.errcode) {
1338 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1339 ctx->flags |= E2F_FLAG_ABORT;
1340 goto endit;
1341 }
1342 ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1343 EXT2_SF_WARN_GARBAGE_INODES, 0);
1344 ctx->stashed_inode = inode;
1345 scan_struct.ctx = ctx;
1346 scan_struct.block_buf = block_buf;
1347 ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1348 if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1349 ctx->fs->group_desc_count)))
1350 goto endit;
1351 if ((fs->super->s_wtime &&
1352 fs->super->s_wtime < fs->super->s_inodes_count) ||
1353 (fs->super->s_mtime &&
1354 fs->super->s_mtime < fs->super->s_inodes_count) ||
1355 (fs->super->s_mkfs_time &&
1356 fs->super->s_mkfs_time < fs->super->s_inodes_count))
1357 low_dtime_check = 0;
1358
1359 if (ext2fs_has_feature_mmp(fs->super) &&
1360 fs->super->s_mmp_block > fs->super->s_first_data_block &&
1361 fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1362 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1363 fs->super->s_mmp_block);
1364
1365 /* Set up ctx->lost_and_found if possible */
1366 (void) e2fsck_get_lost_and_found(ctx, 0);
1367
1368 while (1) {
1369 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1370 if (e2fsck_mmp_update(fs))
1371 fatal_error(ctx, 0);
1372 }
1373 old_op = ehandler_operation(eop_next_inode);
1374 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1375 inode, inode_size);
1376 if (ino > ino_threshold)
1377 pass1_readahead(ctx, &ra_group, &ino_threshold);
1378 ehandler_operation(old_op);
1379 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1380 goto endit;
1381 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1382 /*
1383 * If badblocks says badblocks is bad, offer to clear
1384 * the list, update the in-core bb list, and restart
1385 * the inode scan.
1386 */
1387 if (ino == EXT2_BAD_INO &&
1388 fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1389 &pctx)) {
1390 errcode_t err;
1391
1392 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1393 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1394 ctx->fs->badblocks = NULL;
1395 err = ext2fs_read_bb_inode(ctx->fs,
1396 &ctx->fs->badblocks);
1397 if (err) {
1398 fix_problem(ctx, PR_1_ISCAN_ERROR,
1399 &pctx);
1400 ctx->flags |= E2F_FLAG_ABORT;
1401 } else
1402 ctx->flags |= E2F_FLAG_RESTART;
1403 goto endit;
1404 }
1405 if (!ctx->inode_bb_map)
1406 alloc_bb_map(ctx);
1407 ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1408 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1409 continue;
1410 }
1411 if (pctx.errcode &&
1412 pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1413 pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1414 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1415 ctx->flags |= E2F_FLAG_ABORT;
1416 goto endit;
1417 }
1418 if (!ino)
1419 break;
1420 pctx.ino = ino;
1421 pctx.inode = inode;
1422 ctx->stashed_ino = ino;
1423
1424 /* Clear trashed inode? */
1425 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1426 inode->i_links_count > 0 &&
1427 fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1428 pctx.errcode = 0;
1429 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1430 }
1431 failed_csum = pctx.errcode != 0;
1432
1433 /*
1434 * Check for inodes who might have been part of the
1435 * orphaned list linked list. They should have gotten
1436 * dealt with by now, unless the list had somehow been
1437 * corrupted.
1438 *
1439 * FIXME: In the future, inodes which are still in use
1440 * (and which are therefore) pending truncation should
1441 * be handled specially. Right now we just clear the
1442 * dtime field, and the normal e2fsck handling of
1443 * inodes where i_size and the inode blocks are
1444 * inconsistent is to fix i_size, instead of releasing
1445 * the extra blocks. This won't catch the inodes that
1446 * was at the end of the orphan list, but it's better
1447 * than nothing. The right answer is that there
1448 * shouldn't be any bugs in the orphan list handling. :-)
1449 */
1450 if (inode->i_dtime && low_dtime_check &&
1451 inode->i_dtime < ctx->fs->super->s_inodes_count) {
1452 if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1453 inode->i_dtime = inode->i_links_count ?
1454 0 : ctx->now;
1455 e2fsck_write_inode(ctx, ino, inode,
1456 "pass1");
1457 failed_csum = 0;
1458 }
1459 }
1460
1461 if (inode->i_links_count) {
1462 pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1463 ino, inode->i_links_count);
1464 if (pctx.errcode) {
1465 pctx.num = inode->i_links_count;
1466 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1467 ctx->flags |= E2F_FLAG_ABORT;
1468 goto endit;
1469 }
1470 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1471 !quota_inum_is_reserved(fs, ino)) {
1472 if (!inode->i_dtime && inode->i_mode) {
1473 if (fix_problem(ctx,
1474 PR_1_ZERO_DTIME, &pctx)) {
1475 inode->i_dtime = ctx->now;
1476 e2fsck_write_inode(ctx, ino, inode,
1477 "pass1");
1478 failed_csum = 0;
1479 }
1480 }
1481 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1482 continue;
1483 }
1484
1485 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1486 ((!LINUX_S_ISDIR(inode->i_mode) &&
1487 fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1488 (!casefold_fs &&
1489 fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1490 inode->i_flags &= ~EXT4_CASEFOLD_FL;
1491 e2fsck_write_inode(ctx, ino, inode, "pass1");
1492 }
1493
1494 /* Conflicting inlinedata/extents inode flags? */
1495 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1496 (inode->i_flags & EXT4_EXTENTS_FL)) {
1497 int res = fix_inline_data_extents_file(ctx, ino, inode,
1498 inode_size,
1499 &pctx);
1500 if (res < 0) {
1501 /* skip FINISH_INODE_LOOP */
1502 continue;
1503 }
1504 }
1505
1506 /* Test for incorrect inline_data flags settings. */
1507 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1508 (ino >= EXT2_FIRST_INODE(fs->super))) {
1509 size_t size = 0;
1510
1511 pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1512 if (!pctx.errcode &&
1513 fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1514 ext2fs_set_feature_inline_data(sb);
1515 ext2fs_mark_super_dirty(fs);
1516 inlinedata_fs = 1;
1517 } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1518 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1519 /* skip FINISH_INODE_LOOP */
1520 continue;
1521 }
1522 }
1523
1524 /* Test for inline data flag but no attr */
1525 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1526 (ino >= EXT2_FIRST_INODE(fs->super))) {
1527 size_t size = 0;
1528 errcode_t err;
1529 int flags;
1530
1531 flags = fs->flags;
1532 if (failed_csum)
1533 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1534 err = get_inline_data_ea_size(fs, ino, &size);
1535 fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1536 (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1537
1538 switch (err) {
1539 case 0:
1540 /* Everything is awesome... */
1541 break;
1542 case EXT2_ET_BAD_EA_BLOCK_NUM:
1543 case EXT2_ET_BAD_EA_HASH:
1544 case EXT2_ET_BAD_EA_HEADER:
1545 case EXT2_ET_EA_BAD_NAME_LEN:
1546 case EXT2_ET_EA_BAD_VALUE_SIZE:
1547 case EXT2_ET_EA_KEY_NOT_FOUND:
1548 case EXT2_ET_EA_NO_SPACE:
1549 case EXT2_ET_MISSING_EA_FEATURE:
1550 case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1551 case EXT2_ET_INLINE_DATA_NO_BLOCK:
1552 case EXT2_ET_INLINE_DATA_NO_SPACE:
1553 case EXT2_ET_NO_INLINE_DATA:
1554 case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1555 case EXT2_ET_EA_BAD_VALUE_OFFSET:
1556 case EXT2_ET_EA_INODE_CORRUPTED:
1557 /* broken EA or no system.data EA; truncate */
1558 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1559 &pctx)) {
1560 err = ext2fs_inode_size_set(fs, inode, 0);
1561 if (err) {
1562 pctx.errcode = err;
1563 ctx->flags |= E2F_FLAG_ABORT;
1564 goto endit;
1565 }
1566 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1567 memset(&inode->i_block, 0,
1568 sizeof(inode->i_block));
1569 e2fsck_write_inode(ctx, ino, inode,
1570 "pass1");
1571 failed_csum = 0;
1572 }
1573 break;
1574 default:
1575 /* Some other kind of non-xattr error? */
1576 pctx.errcode = err;
1577 ctx->flags |= E2F_FLAG_ABORT;
1578 goto endit;
1579 }
1580 }
1581
1582 /*
1583 * Test for incorrect extent flag settings.
1584 *
1585 * On big-endian machines we must be careful:
1586 * When the inode is read, the i_block array is not swapped
1587 * if the extent flag is set. Therefore if we are testing
1588 * for or fixing a wrongly-set flag, we must potentially
1589 * (un)swap before testing, or after fixing.
1590 */
1591
1592 /*
1593 * In this case the extents flag was set when read, so
1594 * extent_header_verify is ok. If the inode is cleared,
1595 * no need to swap... so no extra swapping here.
1596 */
1597 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1598 (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1599 (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1600 if ((ext2fs_extent_header_verify(inode->i_block,
1601 sizeof(inode->i_block)) == 0) &&
1602 fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1603 ext2fs_set_feature_extents(sb);
1604 ext2fs_mark_super_dirty(fs);
1605 extent_fs = 1;
1606 } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1607 clear_inode:
1608 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1609 if (ino == EXT2_BAD_INO)
1610 ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1611 ino);
1612 /* skip FINISH_INODE_LOOP */
1613 continue;
1614 }
1615 }
1616
1617 /*
1618 * For big-endian machines:
1619 * If the inode didn't have the extents flag set when it
1620 * was read, then the i_blocks array was swapped. To test
1621 * as an extents header, we must swap it back first.
1622 * IF we then set the extents flag, the entire i_block
1623 * array must be un/re-swapped to make it proper extents data.
1624 */
1625 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1626 (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1627 (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1628 (LINUX_S_ISREG(inode->i_mode) ||
1629 LINUX_S_ISDIR(inode->i_mode))) {
1630 void *ehp;
1631 #ifdef WORDS_BIGENDIAN
1632 __u32 tmp_block[EXT2_N_BLOCKS];
1633
1634 for (i = 0; i < EXT2_N_BLOCKS; i++)
1635 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1636 ehp = tmp_block;
1637 #else
1638 ehp = inode->i_block;
1639 #endif
1640 if ((ext2fs_extent_header_verify(ehp,
1641 sizeof(inode->i_block)) == 0) &&
1642 (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1643 inode->i_flags |= EXT4_EXTENTS_FL;
1644 #ifdef WORDS_BIGENDIAN
1645 memcpy(inode->i_block, tmp_block,
1646 sizeof(inode->i_block));
1647 #endif
1648 e2fsck_write_inode(ctx, ino, inode, "pass1");
1649 failed_csum = 0;
1650 }
1651 }
1652
1653 if (ino == EXT2_BAD_INO) {
1654 struct process_block_struct pb;
1655
1656 if ((failed_csum || inode->i_mode || inode->i_uid ||
1657 inode->i_gid || inode->i_links_count ||
1658 (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1659 inode->i_file_acl) &&
1660 fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1661 memset(inode, 0, sizeof(struct ext2_inode));
1662 e2fsck_write_inode(ctx, ino, inode,
1663 "clear bad inode");
1664 failed_csum = 0;
1665 }
1666
1667 pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1668 &pb.fs_meta_blocks);
1669 if (pctx.errcode) {
1670 pctx.num = 4;
1671 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1672 ctx->flags |= E2F_FLAG_ABORT;
1673 goto endit;
1674 }
1675 pb.ino = EXT2_BAD_INO;
1676 pb.num_blocks = pb.last_block = 0;
1677 pb.last_db_block = -1;
1678 pb.num_illegal_blocks = 0;
1679 pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1680 pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1681 pb.inode = inode;
1682 pb.pctx = &pctx;
1683 pb.ctx = ctx;
1684 pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1685 block_buf, process_bad_block, &pb);
1686 ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1687 if (pctx.errcode) {
1688 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1689 ctx->flags |= E2F_FLAG_ABORT;
1690 goto endit;
1691 }
1692 if (pb.bbcheck)
1693 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1694 ctx->flags |= E2F_FLAG_ABORT;
1695 goto endit;
1696 }
1697 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1698 clear_problem_context(&pctx);
1699 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1700 continue;
1701 } else if (ino == EXT2_ROOT_INO) {
1702 /*
1703 * Make sure the root inode is a directory; if
1704 * not, offer to clear it. It will be
1705 * regenerated in pass #3.
1706 */
1707 if (!LINUX_S_ISDIR(inode->i_mode)) {
1708 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1709 goto clear_inode;
1710 }
1711 /*
1712 * If dtime is set, offer to clear it. mke2fs
1713 * version 0.2b created filesystems with the
1714 * dtime field set for the root and lost+found
1715 * directories. We won't worry about
1716 * /lost+found, since that can be regenerated
1717 * easily. But we will fix the root directory
1718 * as a special case.
1719 */
1720 if (inode->i_dtime && inode->i_links_count) {
1721 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1722 inode->i_dtime = 0;
1723 e2fsck_write_inode(ctx, ino, inode,
1724 "pass1");
1725 failed_csum = 0;
1726 }
1727 }
1728 } else if (ino == EXT2_JOURNAL_INO) {
1729 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1730 if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1731 if (!LINUX_S_ISREG(inode->i_mode) &&
1732 fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1733 &pctx)) {
1734 inode->i_mode = LINUX_S_IFREG;
1735 e2fsck_write_inode(ctx, ino, inode,
1736 "pass1");
1737 failed_csum = 0;
1738 }
1739 check_blocks(ctx, &pctx, block_buf, NULL);
1740 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1741 continue;
1742 }
1743 if ((inode->i_links_count ||
1744 inode->i_blocks || inode->i_block[0]) &&
1745 fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1746 &pctx)) {
1747 memset(inode, 0, inode_size);
1748 ext2fs_icount_store(ctx->inode_link_info,
1749 ino, 0);
1750 e2fsck_write_inode_full(ctx, ino, inode,
1751 inode_size, "pass1");
1752 failed_csum = 0;
1753 }
1754 } else if (quota_inum_is_reserved(fs, ino)) {
1755 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1756 if (ext2fs_has_feature_quota(fs->super) &&
1757 quota_inum_is_super(fs->super, ino)) {
1758 if (!LINUX_S_ISREG(inode->i_mode) &&
1759 fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1760 &pctx)) {
1761 inode->i_mode = LINUX_S_IFREG;
1762 e2fsck_write_inode(ctx, ino, inode,
1763 "pass1");
1764 failed_csum = 0;
1765 }
1766 check_blocks(ctx, &pctx, block_buf, NULL);
1767 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1768 continue;
1769 }
1770 if ((inode->i_links_count ||
1771 inode->i_blocks || inode->i_block[0]) &&
1772 fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1773 &pctx)) {
1774 memset(inode, 0, inode_size);
1775 ext2fs_icount_store(ctx->inode_link_info,
1776 ino, 0);
1777 e2fsck_write_inode_full(ctx, ino, inode,
1778 inode_size, "pass1");
1779 failed_csum = 0;
1780 }
1781 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1782 problem_t problem = 0;
1783
1784 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1785 if (ino == EXT2_BOOT_LOADER_INO) {
1786 if (LINUX_S_ISDIR(inode->i_mode))
1787 problem = PR_1_RESERVED_BAD_MODE;
1788 } else if (ino == EXT2_RESIZE_INO) {
1789 if (inode->i_mode &&
1790 !LINUX_S_ISREG(inode->i_mode))
1791 problem = PR_1_RESERVED_BAD_MODE;
1792 } else {
1793 if (inode->i_mode != 0)
1794 problem = PR_1_RESERVED_BAD_MODE;
1795 }
1796 if (problem) {
1797 if (fix_problem(ctx, problem, &pctx)) {
1798 inode->i_mode = 0;
1799 e2fsck_write_inode(ctx, ino, inode,
1800 "pass1");
1801 failed_csum = 0;
1802 }
1803 }
1804 check_blocks(ctx, &pctx, block_buf, NULL);
1805 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1806 continue;
1807 }
1808
1809 if (!inode->i_links_count) {
1810 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1811 continue;
1812 }
1813 /*
1814 * n.b. 0.3c ext2fs code didn't clear i_links_count for
1815 * deleted files. Oops.
1816 *
1817 * Since all new ext2 implementations get this right,
1818 * we now assume that the case of non-zero
1819 * i_links_count and non-zero dtime means that we
1820 * should keep the file, not delete it.
1821 *
1822 */
1823 if (inode->i_dtime) {
1824 if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1825 inode->i_dtime = 0;
1826 e2fsck_write_inode(ctx, ino, inode, "pass1");
1827 failed_csum = 0;
1828 }
1829 }
1830
1831 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1832 switch (fs->super->s_creator_os) {
1833 case EXT2_OS_HURD:
1834 frag = inode->osd2.hurd2.h_i_frag;
1835 fsize = inode->osd2.hurd2.h_i_fsize;
1836 break;
1837 default:
1838 frag = fsize = 0;
1839 }
1840
1841 if (inode->i_faddr || frag || fsize ||
1842 (!ext2fs_has_feature_largedir(fs->super) &&
1843 (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1844 mark_inode_bad(ctx, ino);
1845 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1846 !ext2fs_has_feature_64bit(fs->super) &&
1847 inode->osd2.linux2.l_i_file_acl_high != 0)
1848 mark_inode_bad(ctx, ino);
1849 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1850 !ext2fs_has_feature_huge_file(fs->super) &&
1851 (inode->osd2.linux2.l_i_blocks_hi != 0))
1852 mark_inode_bad(ctx, ino);
1853 if (inode->i_flags & EXT2_IMAGIC_FL) {
1854 if (imagic_fs) {
1855 if (!ctx->inode_imagic_map)
1856 alloc_imagic_map(ctx);
1857 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1858 ino);
1859 } else {
1860 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1861 inode->i_flags &= ~EXT2_IMAGIC_FL;
1862 e2fsck_write_inode(ctx, ino,
1863 inode, "pass1");
1864 failed_csum = 0;
1865 }
1866 }
1867 }
1868
1869 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1870 check_is_really_dir(ctx, &pctx, block_buf);
1871
1872 /*
1873 * ext2fs_inode_has_valid_blocks2 does not actually look
1874 * at i_block[] values, so not endian-sensitive here.
1875 */
1876 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1877 LINUX_S_ISLNK(inode->i_mode) &&
1878 !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1879 fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1880 inode->i_flags &= ~EXT4_EXTENTS_FL;
1881 e2fsck_write_inode(ctx, ino, inode, "pass1");
1882 failed_csum = 0;
1883 }
1884
1885 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
1886 add_encrypted_file(ctx, &pctx) < 0)
1887 goto clear_inode;
1888
1889 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
1890 ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
1891
1892 if (LINUX_S_ISDIR(inode->i_mode)) {
1893 ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1894 e2fsck_add_dir_info(ctx, ino, 0);
1895 ctx->fs_directory_count++;
1896 if (inode->i_flags & EXT4_CASEFOLD_FL)
1897 add_casefolded_dir(ctx, ino);
1898 } else if (LINUX_S_ISREG (inode->i_mode)) {
1899 ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1900 ctx->fs_regular_count++;
1901 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1902 e2fsck_pass1_check_device_inode(fs, inode)) {
1903 check_extents_inlinedata(ctx, &pctx);
1904 check_immutable(ctx, &pctx);
1905 check_size(ctx, &pctx);
1906 ctx->fs_chardev_count++;
1907 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1908 e2fsck_pass1_check_device_inode(fs, inode)) {
1909 check_extents_inlinedata(ctx, &pctx);
1910 check_immutable(ctx, &pctx);
1911 check_size(ctx, &pctx);
1912 ctx->fs_blockdev_count++;
1913 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1914 e2fsck_pass1_check_symlink(fs, ino, inode,
1915 block_buf)) {
1916 check_immutable(ctx, &pctx);
1917 ctx->fs_symlinks_count++;
1918 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1919 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1920 continue;
1921 } else if (ext2fs_is_fast_symlink(inode)) {
1922 ctx->fs_fast_symlinks_count++;
1923 check_blocks(ctx, &pctx, block_buf,
1924 &ea_ibody_quota);
1925 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1926 continue;
1927 }
1928 }
1929 else if (LINUX_S_ISFIFO (inode->i_mode) &&
1930 e2fsck_pass1_check_device_inode(fs, inode)) {
1931 check_extents_inlinedata(ctx, &pctx);
1932 check_immutable(ctx, &pctx);
1933 check_size(ctx, &pctx);
1934 ctx->fs_fifo_count++;
1935 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
1936 e2fsck_pass1_check_device_inode(fs, inode)) {
1937 check_extents_inlinedata(ctx, &pctx);
1938 check_immutable(ctx, &pctx);
1939 check_size(ctx, &pctx);
1940 ctx->fs_sockets_count++;
1941 } else
1942 mark_inode_bad(ctx, ino);
1943 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1944 !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
1945 if (inode->i_block[EXT2_IND_BLOCK])
1946 ctx->fs_ind_count++;
1947 if (inode->i_block[EXT2_DIND_BLOCK])
1948 ctx->fs_dind_count++;
1949 if (inode->i_block[EXT2_TIND_BLOCK])
1950 ctx->fs_tind_count++;
1951 }
1952 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1953 !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
1954 (inode->i_block[EXT2_IND_BLOCK] ||
1955 inode->i_block[EXT2_DIND_BLOCK] ||
1956 inode->i_block[EXT2_TIND_BLOCK] ||
1957 ext2fs_file_acl_block(fs, inode))) {
1958 struct process_inode_block *itp;
1959
1960 itp = &inodes_to_process[process_inode_count];
1961 itp->ino = ino;
1962 itp->ea_ibody_quota = ea_ibody_quota;
1963 if (inode_size < sizeof(struct ext2_inode_large))
1964 memcpy(&itp->inode, inode, inode_size);
1965 else
1966 memcpy(&itp->inode, inode, sizeof(itp->inode));
1967 process_inode_count++;
1968 } else
1969 check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
1970
1971 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1972
1973 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1974 goto endit;
1975
1976 if (process_inode_count >= ctx->process_inode_size) {
1977 process_inodes(ctx, block_buf);
1978
1979 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1980 goto endit;
1981 }
1982 }
1983 process_inodes(ctx, block_buf);
1984 ext2fs_close_inode_scan(scan);
1985 scan = NULL;
1986
1987 reserve_block_for_root_repair(ctx);
1988 reserve_block_for_lnf_repair(ctx);
1989
1990 /*
1991 * If any extended attribute blocks' reference counts need to
1992 * be adjusted, either up (ctx->refcount_extra), or down
1993 * (ctx->refcount), then fix them.
1994 */
1995 if (ctx->refcount) {
1996 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
1997 ea_refcount_free(ctx->refcount);
1998 ctx->refcount = 0;
1999 }
2000 if (ctx->refcount_extra) {
2001 adjust_extattr_refcount(ctx, ctx->refcount_extra,
2002 block_buf, +1);
2003 ea_refcount_free(ctx->refcount_extra);
2004 ctx->refcount_extra = 0;
2005 }
2006
2007 if (ctx->ea_block_quota_blocks) {
2008 ea_refcount_free(ctx->ea_block_quota_blocks);
2009 ctx->ea_block_quota_blocks = 0;
2010 }
2011
2012 if (ctx->ea_block_quota_inodes) {
2013 ea_refcount_free(ctx->ea_block_quota_inodes);
2014 ctx->ea_block_quota_inodes = 0;
2015 }
2016
2017 if (ctx->invalid_bitmaps)
2018 handle_fs_bad_blocks(ctx);
2019
2020 /* We don't need the block_ea_map any more */
2021 if (ctx->block_ea_map) {
2022 ext2fs_free_block_bitmap(ctx->block_ea_map);
2023 ctx->block_ea_map = 0;
2024 }
2025
2026 /* We don't need the encryption policy => ID map any more */
2027 destroy_encryption_policy_map(ctx);
2028
2029 if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2030 clear_problem_context(&pctx);
2031 pctx.errcode = ext2fs_create_resize_inode(fs);
2032 if (pctx.errcode) {
2033 if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2034 &pctx)) {
2035 ctx->flags |= E2F_FLAG_ABORT;
2036 goto endit;
2037 }
2038 pctx.errcode = 0;
2039 }
2040 if (!pctx.errcode) {
2041 e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2042 "recreate inode");
2043 inode->i_mtime = ctx->now;
2044 e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2045 "recreate inode");
2046 }
2047 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2048 }
2049
2050 if (ctx->flags & E2F_FLAG_RESTART) {
2051 /*
2052 * Only the master copy of the superblock and block
2053 * group descriptors are going to be written during a
2054 * restart, so set the superblock to be used to be the
2055 * master superblock.
2056 */
2057 ctx->use_superblock = 0;
2058 goto endit;
2059 }
2060
2061 if (ctx->large_dirs && !ext2fs_has_feature_largedir(fs->super)) {
2062 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2063 ext2fs_set_feature_largedir(fs->super);
2064 fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2065 ext2fs_mark_super_dirty(fs);
2066 }
2067 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2068 fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2069 ext2fs_update_dynamic_rev(fs);
2070 ext2fs_mark_super_dirty(fs);
2071 }
2072 }
2073
2074 if (ctx->block_dup_map) {
2075 if (ctx->options & E2F_OPT_PREEN) {
2076 clear_problem_context(&pctx);
2077 fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2078 }
2079 e2fsck_pass1_dupblocks(ctx, block_buf);
2080 }
2081 ctx->flags |= E2F_FLAG_ALLOC_OK;
2082 endit:
2083 e2fsck_use_inode_shortcuts(ctx, 0);
2084 ext2fs_free_mem(&inodes_to_process);
2085 inodes_to_process = 0;
2086
2087 if (scan)
2088 ext2fs_close_inode_scan(scan);
2089 if (block_buf)
2090 ext2fs_free_mem(&block_buf);
2091 if (inode)
2092 ext2fs_free_mem(&inode);
2093
2094 /*
2095 * The l+f inode may have been cleared, so zap it now and
2096 * later passes will recalculate it if necessary
2097 */
2098 ctx->lost_and_found = 0;
2099
2100 if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2101 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2102 else
2103 ctx->invalid_bitmaps++;
2104 }
2105 #undef FINISH_INODE_LOOP
2106
2107 /*
2108 * When the inode_scan routines call this callback at the end of the
2109 * glock group, call process_inodes.
2110 */
scan_callback(ext2_filsys fs,ext2_inode_scan scan EXT2FS_ATTR ((unused)),dgrp_t group,void * priv_data)2111 static errcode_t scan_callback(ext2_filsys fs,
2112 ext2_inode_scan scan EXT2FS_ATTR((unused)),
2113 dgrp_t group, void * priv_data)
2114 {
2115 struct scan_callback_struct *scan_struct;
2116 e2fsck_t ctx;
2117
2118 scan_struct = (struct scan_callback_struct *) priv_data;
2119 ctx = scan_struct->ctx;
2120
2121 process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf);
2122
2123 if (ctx->progress)
2124 if ((ctx->progress)(ctx, 1, group+1,
2125 ctx->fs->group_desc_count))
2126 return EXT2_ET_CANCEL_REQUESTED;
2127
2128 return 0;
2129 }
2130
2131 /*
2132 * Process the inodes in the "inodes to process" list.
2133 */
process_inodes(e2fsck_t ctx,char * block_buf)2134 static void process_inodes(e2fsck_t ctx, char *block_buf)
2135 {
2136 int i;
2137 struct ext2_inode *old_stashed_inode;
2138 ext2_ino_t old_stashed_ino;
2139 const char *old_operation;
2140 char buf[80];
2141 struct problem_context pctx;
2142
2143 #if 0
2144 printf("begin process_inodes: ");
2145 #endif
2146 if (process_inode_count == 0)
2147 return;
2148 old_operation = ehandler_operation(0);
2149 old_stashed_inode = ctx->stashed_inode;
2150 old_stashed_ino = ctx->stashed_ino;
2151 qsort(inodes_to_process, process_inode_count,
2152 sizeof(struct process_inode_block), process_inode_cmp);
2153 clear_problem_context(&pctx);
2154 for (i=0; i < process_inode_count; i++) {
2155 pctx.inode = ctx->stashed_inode =
2156 (struct ext2_inode *) &inodes_to_process[i].inode;
2157 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2158
2159 #if 0
2160 printf("%u ", pctx.ino);
2161 #endif
2162 sprintf(buf, _("reading indirect blocks of inode %u"),
2163 pctx.ino);
2164 ehandler_operation(buf);
2165 check_blocks(ctx, &pctx, block_buf,
2166 &inodes_to_process[i].ea_ibody_quota);
2167 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2168 break;
2169 }
2170 ctx->stashed_inode = old_stashed_inode;
2171 ctx->stashed_ino = old_stashed_ino;
2172 process_inode_count = 0;
2173 #if 0
2174 printf("end process inodes\n");
2175 #endif
2176 ehandler_operation(old_operation);
2177 }
2178
process_inode_cmp(const void * a,const void * b)2179 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
2180 {
2181 const struct process_inode_block *ib_a =
2182 (const struct process_inode_block *) a;
2183 const struct process_inode_block *ib_b =
2184 (const struct process_inode_block *) b;
2185 int ret;
2186
2187 ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
2188 ib_b->inode.i_block[EXT2_IND_BLOCK]);
2189 if (ret == 0)
2190 /*
2191 * We only call process_inodes() for non-extent
2192 * inodes, so it's OK to pass NULL to
2193 * ext2fs_file_acl_block() here.
2194 */
2195 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
2196 ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
2197 if (ret == 0)
2198 ret = ib_a->ino - ib_b->ino;
2199 return ret;
2200 }
2201
2202 /*
2203 * Mark an inode as being bad in some what
2204 */
mark_inode_bad(e2fsck_t ctx,ino_t ino)2205 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
2206 {
2207 struct problem_context pctx;
2208
2209 if (!ctx->inode_bad_map) {
2210 clear_problem_context(&pctx);
2211
2212 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2213 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
2214 "inode_bad_map", &ctx->inode_bad_map);
2215 if (pctx.errcode) {
2216 pctx.num = 3;
2217 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2218 /* Should never get here */
2219 ctx->flags |= E2F_FLAG_ABORT;
2220 return;
2221 }
2222 }
2223 ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
2224 }
2225
add_casefolded_dir(e2fsck_t ctx,ino_t ino)2226 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino)
2227 {
2228 struct problem_context pctx;
2229
2230 if (!ctx->casefolded_dirs) {
2231 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
2232 if (pctx.errcode)
2233 goto error;
2234 }
2235 pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
2236 if (pctx.errcode == 0)
2237 return;
2238 error:
2239 fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
2240 /* Should never get here */
2241 ctx->flags |= E2F_FLAG_ABORT;
2242 }
2243
2244 /*
2245 * This procedure will allocate the inode "bb" (badblock) map table
2246 */
alloc_bb_map(e2fsck_t ctx)2247 static void alloc_bb_map(e2fsck_t ctx)
2248 {
2249 struct problem_context pctx;
2250
2251 clear_problem_context(&pctx);
2252 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2253 _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
2254 "inode_bb_map", &ctx->inode_bb_map);
2255 if (pctx.errcode) {
2256 pctx.num = 4;
2257 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2258 /* Should never get here */
2259 ctx->flags |= E2F_FLAG_ABORT;
2260 return;
2261 }
2262 }
2263
2264 /*
2265 * This procedure will allocate the inode imagic table
2266 */
alloc_imagic_map(e2fsck_t ctx)2267 static void alloc_imagic_map(e2fsck_t ctx)
2268 {
2269 struct problem_context pctx;
2270
2271 clear_problem_context(&pctx);
2272 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2273 _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
2274 "inode_imagic_map", &ctx->inode_imagic_map);
2275 if (pctx.errcode) {
2276 pctx.num = 5;
2277 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2278 /* Should never get here */
2279 ctx->flags |= E2F_FLAG_ABORT;
2280 return;
2281 }
2282 }
2283
2284 /*
2285 * Marks a block as in use, setting the dup_map if it's been set
2286 * already. Called by process_block and process_bad_block.
2287 *
2288 * WARNING: Assumes checks have already been done to make sure block
2289 * is valid. This is true in both process_block and process_bad_block.
2290 */
mark_block_used(e2fsck_t ctx,blk64_t block)2291 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
2292 {
2293 struct problem_context pctx;
2294
2295 clear_problem_context(&pctx);
2296
2297 if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
2298 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
2299 !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
2300 return;
2301 }
2302 if (!ctx->block_dup_map) {
2303 pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
2304 _("multiply claimed block map"),
2305 EXT2FS_BMAP64_RBTREE, "block_dup_map",
2306 &ctx->block_dup_map);
2307 if (pctx.errcode) {
2308 pctx.num = 3;
2309 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
2310 &pctx);
2311 /* Should never get here */
2312 ctx->flags |= E2F_FLAG_ABORT;
2313 return;
2314 }
2315 }
2316 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
2317 } else {
2318 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
2319 }
2320 }
2321
2322 /*
2323 * When cluster size is greater than one block, it is caller's responsibility
2324 * to make sure block parameter starts at a cluster boundary.
2325 */
mark_blocks_used(e2fsck_t ctx,blk64_t block,unsigned int num)2326 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
2327 unsigned int num)
2328 {
2329 if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
2330 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
2331 else {
2332 unsigned int i;
2333
2334 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
2335 mark_block_used(ctx, block + i);
2336 }
2337 }
2338
2339 /*
2340 * Adjust the extended attribute block's reference counts at the end
2341 * of pass 1, either by subtracting out references for EA blocks that
2342 * are still referenced in ctx->refcount, or by adding references for
2343 * EA blocks that had extra references as accounted for in
2344 * ctx->refcount_extra.
2345 */
adjust_extattr_refcount(e2fsck_t ctx,ext2_refcount_t refcount,char * block_buf,int adjust_sign)2346 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
2347 char *block_buf, int adjust_sign)
2348 {
2349 struct ext2_ext_attr_header *header;
2350 struct problem_context pctx;
2351 ext2_filsys fs = ctx->fs;
2352 blk64_t blk;
2353 __u32 should_be;
2354 ea_value_t count;
2355
2356 clear_problem_context(&pctx);
2357
2358 ea_refcount_intr_begin(refcount);
2359 while (1) {
2360 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
2361 break;
2362 pctx.blk = blk;
2363 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
2364 pctx.ino);
2365 if (pctx.errcode) {
2366 fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
2367 return;
2368 }
2369 header = (struct ext2_ext_attr_header *) block_buf;
2370 pctx.blkcount = header->h_refcount;
2371 should_be = header->h_refcount + adjust_sign * (int)count;
2372 pctx.num = should_be;
2373 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
2374 header->h_refcount = should_be;
2375 pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
2376 block_buf,
2377 pctx.ino);
2378 if (pctx.errcode) {
2379 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
2380 &pctx);
2381 continue;
2382 }
2383 }
2384 }
2385 }
2386
2387 /*
2388 * Handle processing the extended attribute blocks
2389 */
check_ext_attr(e2fsck_t ctx,struct problem_context * pctx,char * block_buf,struct ea_quota * ea_block_quota)2390 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
2391 char *block_buf, struct ea_quota *ea_block_quota)
2392 {
2393 ext2_filsys fs = ctx->fs;
2394 ext2_ino_t ino = pctx->ino;
2395 struct ext2_inode *inode = pctx->inode;
2396 blk64_t blk;
2397 char * end;
2398 struct ext2_ext_attr_header *header;
2399 struct ext2_ext_attr_entry *first, *entry;
2400 blk64_t quota_blocks = EXT2FS_C2B(fs, 1);
2401 __u64 quota_inodes = 0;
2402 region_t region = 0;
2403 int failed_csum = 0;
2404
2405 ea_block_quota->blocks = 0;
2406 ea_block_quota->inodes = 0;
2407
2408 blk = ext2fs_file_acl_block(fs, inode);
2409 if (blk == 0)
2410 return 0;
2411
2412 /*
2413 * If the Extended attribute flag isn't set, then a non-zero
2414 * file acl means that the inode is corrupted.
2415 *
2416 * Or if the extended attribute block is an invalid block,
2417 * then the inode is also corrupted.
2418 */
2419 if (!ext2fs_has_feature_xattr(fs->super) ||
2420 (blk < fs->super->s_first_data_block) ||
2421 (blk >= ext2fs_blocks_count(fs->super))) {
2422 mark_inode_bad(ctx, ino);
2423 return 0;
2424 }
2425
2426 /* If ea bitmap hasn't been allocated, create it */
2427 if (!ctx->block_ea_map) {
2428 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
2429 _("ext attr block map"),
2430 EXT2FS_BMAP64_RBTREE, "block_ea_map",
2431 &ctx->block_ea_map);
2432 if (pctx->errcode) {
2433 pctx->num = 2;
2434 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
2435 ctx->flags |= E2F_FLAG_ABORT;
2436 return 0;
2437 }
2438 }
2439
2440 /* Create the EA refcount structure if necessary */
2441 if (!ctx->refcount) {
2442 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
2443 if (pctx->errcode) {
2444 pctx->num = 1;
2445 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2446 ctx->flags |= E2F_FLAG_ABORT;
2447 return 0;
2448 }
2449 }
2450
2451 #if 0
2452 /* Debugging text */
2453 printf("Inode %u has EA block %u\n", ino, blk);
2454 #endif
2455
2456 /* Have we seen this EA block before? */
2457 if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
2458 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
2459 ea_block_quota->inodes = 0;
2460
2461 if (ctx->ea_block_quota_blocks) {
2462 ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
2463 "a_blocks);
2464 if (quota_blocks)
2465 ea_block_quota->blocks = quota_blocks;
2466 }
2467
2468 if (ctx->ea_block_quota_inodes)
2469 ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
2470 &ea_block_quota->inodes);
2471
2472 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
2473 return 1;
2474 /* Ooops, this EA was referenced more than it stated */
2475 if (!ctx->refcount_extra) {
2476 pctx->errcode = ea_refcount_create(0,
2477 &ctx->refcount_extra);
2478 if (pctx->errcode) {
2479 pctx->num = 2;
2480 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2481 ctx->flags |= E2F_FLAG_ABORT;
2482 return 0;
2483 }
2484 }
2485 ea_refcount_increment(ctx->refcount_extra, blk, 0);
2486 return 1;
2487 }
2488
2489 /*
2490 * OK, we haven't seen this EA block yet. So we need to
2491 * validate it
2492 */
2493 pctx->blk = blk;
2494 pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
2495 if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
2496 pctx->errcode = 0;
2497 failed_csum = 1;
2498 } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
2499 pctx->errcode = 0;
2500
2501 if (pctx->errcode &&
2502 fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
2503 pctx->errcode = 0;
2504 goto clear_extattr;
2505 }
2506 header = (struct ext2_ext_attr_header *) block_buf;
2507 pctx->blk = ext2fs_file_acl_block(fs, inode);
2508 if (((ctx->ext_attr_ver == 1) &&
2509 (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
2510 ((ctx->ext_attr_ver == 2) &&
2511 (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
2512 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
2513 goto clear_extattr;
2514 }
2515
2516 if (header->h_blocks != 1) {
2517 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
2518 goto clear_extattr;
2519 }
2520
2521 if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
2522 goto clear_extattr;
2523
2524 region = region_create(0, fs->blocksize);
2525 if (!region) {
2526 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
2527 ctx->flags |= E2F_FLAG_ABORT;
2528 return 0;
2529 }
2530 if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
2531 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2532 goto clear_extattr;
2533 }
2534
2535 first = (struct ext2_ext_attr_entry *)(header+1);
2536 end = block_buf + fs->blocksize;
2537 entry = first;
2538 while ((char *)entry < end && *(__u32 *)entry) {
2539 __u32 hash;
2540
2541 if (region_allocate(region, (char *)entry - (char *)header,
2542 EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
2543 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2544 goto clear_extattr;
2545 break;
2546 }
2547 if ((ctx->ext_attr_ver == 1 &&
2548 (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
2549 (ctx->ext_attr_ver == 2 &&
2550 entry->e_name_index == 0)) {
2551 if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
2552 goto clear_extattr;
2553 break;
2554 }
2555 if (entry->e_value_inum == 0) {
2556 if (entry->e_value_offs + entry->e_value_size >
2557 fs->blocksize) {
2558 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
2559 goto clear_extattr;
2560 break;
2561 }
2562 if (entry->e_value_size &&
2563 region_allocate(region, entry->e_value_offs,
2564 EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
2565 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
2566 pctx))
2567 goto clear_extattr;
2568 }
2569
2570 hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
2571 entry->e_value_offs);
2572
2573 if (entry->e_hash != hash) {
2574 pctx->num = entry->e_hash;
2575 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
2576 goto clear_extattr;
2577 entry->e_hash = hash;
2578 }
2579 } else {
2580 problem_t problem;
2581 blk64_t entry_quota_blocks;
2582
2583 problem = check_large_ea_inode(ctx, entry, pctx,
2584 &entry_quota_blocks);
2585 if (problem && fix_problem(ctx, problem, pctx))
2586 goto clear_extattr;
2587
2588 quota_blocks += entry_quota_blocks;
2589 quota_inodes++;
2590 }
2591
2592 entry = EXT2_EXT_ATTR_NEXT(entry);
2593 }
2594 if (region_allocate(region, (char *)entry - (char *)header, 4)) {
2595 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2596 goto clear_extattr;
2597 }
2598 region_free(region);
2599
2600 /*
2601 * We only get here if there was no other errors that were fixed.
2602 * If there was a checksum fail, ask to correct it.
2603 */
2604 if (failed_csum &&
2605 fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
2606 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
2607 pctx->ino);
2608 if (pctx->errcode)
2609 return 0;
2610 }
2611
2612 if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
2613 if (!ctx->ea_block_quota_blocks) {
2614 pctx->errcode = ea_refcount_create(0,
2615 &ctx->ea_block_quota_blocks);
2616 if (pctx->errcode) {
2617 pctx->num = 3;
2618 goto refcount_fail;
2619 }
2620 }
2621 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
2622 quota_blocks);
2623 }
2624
2625 if (quota_inodes) {
2626 if (!ctx->ea_block_quota_inodes) {
2627 pctx->errcode = ea_refcount_create(0,
2628 &ctx->ea_block_quota_inodes);
2629 if (pctx->errcode) {
2630 pctx->num = 4;
2631 refcount_fail:
2632 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2633 ctx->flags |= E2F_FLAG_ABORT;
2634 return 0;
2635 }
2636 }
2637
2638 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
2639 quota_inodes);
2640 }
2641 ea_block_quota->blocks = quota_blocks;
2642 ea_block_quota->inodes = quota_inodes;
2643
2644 inc_ea_inode_refs(ctx, pctx, first, end);
2645 ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
2646 mark_block_used(ctx, blk);
2647 ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
2648 return 1;
2649
2650 clear_extattr:
2651 if (region)
2652 region_free(region);
2653 ext2fs_file_acl_block_set(fs, inode, 0);
2654 e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
2655 return 0;
2656 }
2657
2658 /* 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)2659 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
2660 ext2_ino_t ino, struct ext2_inode *inode,
2661 char *block_buf)
2662 {
2663 struct ext2_dx_root_info *root;
2664 ext2_filsys fs = ctx->fs;
2665 errcode_t retval;
2666 blk64_t blk;
2667
2668 if ((!LINUX_S_ISDIR(inode->i_mode) &&
2669 fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
2670 (!ext2fs_has_feature_dir_index(fs->super) &&
2671 fix_problem(ctx, PR_1_HTREE_SET, pctx)))
2672 return 1;
2673
2674 pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
2675
2676 if ((pctx->errcode) ||
2677 (blk == 0) ||
2678 (blk < fs->super->s_first_data_block) ||
2679 (blk >= ext2fs_blocks_count(fs->super))) {
2680 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2681 return 1;
2682 else
2683 return 0;
2684 }
2685
2686 retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
2687 if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2688 return 1;
2689
2690 /* XXX should check that beginning matches a directory */
2691 root = (struct ext2_dx_root_info *) (block_buf + 24);
2692
2693 if ((root->reserved_zero || root->info_length < 8) &&
2694 fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2695 return 1;
2696
2697 pctx->num = root->hash_version;
2698 if ((root->hash_version != EXT2_HASH_LEGACY) &&
2699 (root->hash_version != EXT2_HASH_HALF_MD4) &&
2700 (root->hash_version != EXT2_HASH_TEA) &&
2701 (root->hash_version != EXT2_HASH_SIPHASH) &&
2702 fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
2703 return 1;
2704
2705 if (ext4_hash_in_dirent(inode)) {
2706 if (root->hash_version != EXT2_HASH_SIPHASH &&
2707 fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
2708 return 1;
2709 } else {
2710 if (root->hash_version == EXT2_HASH_SIPHASH &&
2711 fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
2712 return 1;
2713 }
2714
2715 if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
2716 fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
2717 return 1;
2718
2719 pctx->num = root->indirect_levels;
2720 /* if htree level is clearly too high, consider it to be broken */
2721 if (root->indirect_levels > EXT4_HTREE_LEVEL &&
2722 fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
2723 return 1;
2724
2725 /* if level is only maybe too high, LARGE_DIR feature could be unset */
2726 if (root->indirect_levels > ext2_dir_htree_level(fs) &&
2727 !ext2fs_has_feature_largedir(fs->super)) {
2728 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
2729 unsigned idx_pb = 1 << (blockbits - 3);
2730
2731 /* compare inode size/blocks vs. max-sized 2-level htree */
2732 if (EXT2_I_SIZE(pctx->inode) <
2733 (idx_pb - 1) * (idx_pb - 2) << blockbits &&
2734 pctx->inode->i_blocks <
2735 (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
2736 fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
2737 return 1;
2738 }
2739
2740 if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
2741 ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
2742 ctx->large_dirs++;
2743
2744 return 0;
2745 }
2746
e2fsck_clear_inode(e2fsck_t ctx,ext2_ino_t ino,struct ext2_inode * inode,int restart_flag,const char * source)2747 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
2748 struct ext2_inode *inode, int restart_flag,
2749 const char *source)
2750 {
2751 inode->i_flags = 0;
2752 inode->i_links_count = 0;
2753 ext2fs_icount_store(ctx->inode_link_info, ino, 0);
2754 inode->i_dtime = ctx->now;
2755
2756 /*
2757 * If a special inode has such rotten block mappings that we
2758 * want to clear the whole inode, be sure to actually zap
2759 * the block maps because i_links_count isn't checked for
2760 * special inodes, and we'll end up right back here the next
2761 * time we run fsck.
2762 */
2763 if (ino < EXT2_FIRST_INODE(ctx->fs->super))
2764 memset(inode->i_block, 0, sizeof(inode->i_block));
2765
2766 ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
2767 ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
2768 if (ctx->inode_reg_map)
2769 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
2770 if (ctx->inode_bad_map)
2771 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
2772
2773 /*
2774 * If the inode was partially accounted for before processing
2775 * was aborted, we need to restart the pass 1 scan.
2776 */
2777 ctx->flags |= restart_flag;
2778
2779 if (ino == EXT2_BAD_INO)
2780 memset(inode, 0, sizeof(struct ext2_inode));
2781
2782 e2fsck_write_inode(ctx, ino, inode, source);
2783 }
2784
2785 /*
2786 * Use the multiple-blocks reclamation code to fix alignment problems in
2787 * a bigalloc filesystem. We want a logical cluster to map to *only* one
2788 * physical cluster, and we want the block offsets within that cluster to
2789 * line up.
2790 */
has_unaligned_cluster_map(e2fsck_t ctx,blk64_t last_pblk,blk64_t last_lblk,blk64_t pblk,blk64_t lblk)2791 static int has_unaligned_cluster_map(e2fsck_t ctx,
2792 blk64_t last_pblk, blk64_t last_lblk,
2793 blk64_t pblk, blk64_t lblk)
2794 {
2795 blk64_t cluster_mask;
2796
2797 if (!ctx->fs->cluster_ratio_bits)
2798 return 0;
2799 cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
2800
2801 /*
2802 * If the block in the logical cluster doesn't align with the block in
2803 * the physical cluster...
2804 */
2805 if ((lblk & cluster_mask) != (pblk & cluster_mask))
2806 return 1;
2807
2808 /*
2809 * If we cross a physical cluster boundary within a logical cluster...
2810 */
2811 if (last_pblk && (lblk & cluster_mask) != 0 &&
2812 EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
2813 EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
2814 return 1;
2815
2816 return 0;
2817 }
2818
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)2819 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
2820 struct process_block_struct *pb,
2821 blk64_t start_block, blk64_t end_block,
2822 blk64_t eof_block,
2823 ext2_extent_handle_t ehandle,
2824 int try_repairs)
2825 {
2826 struct ext2fs_extent extent;
2827 blk64_t blk, last_lblk;
2828 unsigned int i, n;
2829 int is_dir, is_leaf;
2830 problem_t problem;
2831 struct ext2_extent_info info;
2832 int failed_csum = 0;
2833
2834 if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
2835 failed_csum = 1;
2836
2837 pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
2838 if (pctx->errcode)
2839 return;
2840 if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
2841 !pb->eti.force_rebuild) {
2842 struct extent_tree_level *etl;
2843
2844 etl = pb->eti.ext_info + info.curr_level;
2845 etl->num_extents += info.num_entries;
2846 etl->max_extents += info.max_entries;
2847 /*
2848 * Implementation wart: Splitting extent blocks when appending
2849 * will leave the old block with one free entry. Therefore
2850 * unless the node is totally full, pretend that a non-root
2851 * extent block can hold one fewer entry than it actually does,
2852 * so that we don't repeatedly rebuild the extent tree.
2853 */
2854 if (info.curr_level && info.num_entries < info.max_entries)
2855 etl->max_extents--;
2856 }
2857
2858 pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
2859 &extent);
2860 while ((pctx->errcode == 0 ||
2861 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
2862 info.num_entries-- > 0) {
2863 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
2864 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
2865 last_lblk = extent.e_lblk + extent.e_len - 1;
2866
2867 problem = 0;
2868 pctx->blk = extent.e_pblk;
2869 pctx->blk2 = extent.e_lblk;
2870 pctx->num = extent.e_len;
2871 pctx->blkcount = extent.e_lblk + extent.e_len;
2872
2873 if (extent.e_pblk == 0 ||
2874 extent.e_pblk < ctx->fs->super->s_first_data_block ||
2875 extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
2876 problem = PR_1_EXTENT_BAD_START_BLK;
2877 else if (extent.e_lblk < start_block)
2878 problem = PR_1_OUT_OF_ORDER_EXTENTS;
2879 else if ((end_block && last_lblk > end_block) &&
2880 !(last_lblk > eof_block &&
2881 ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
2882 (pctx->inode->i_flags & EXT4_VERITY_FL))))
2883 problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
2884 else if (is_leaf && extent.e_len == 0)
2885 problem = PR_1_EXTENT_LENGTH_ZERO;
2886 else if (is_leaf &&
2887 (extent.e_pblk + extent.e_len) >
2888 ext2fs_blocks_count(ctx->fs->super))
2889 problem = PR_1_EXTENT_ENDS_BEYOND;
2890 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
2891 !ext2fs_has_feature_largedir(ctx->fs->super) &&
2892 ((extent.e_lblk + extent.e_len) >
2893 (1U << (21 - ctx->fs->super->s_log_block_size))))
2894 problem = PR_1_TOOBIG_DIR;
2895
2896 if (is_leaf && problem == 0 && extent.e_len > 0) {
2897 #if 0
2898 printf("extent_region(ino=%u, expect=%llu, "
2899 "lblk=%llu, len=%u)\n", pb->ino,
2900 (unsigned long long) pb->next_lblock,
2901 (unsigned long long) extent.e_lblk,
2902 extent.e_len);
2903 #endif
2904 if (extent.e_lblk < pb->next_lblock)
2905 problem = PR_1_EXTENT_COLLISION;
2906 else if (extent.e_lblk + extent.e_len > pb->next_lblock)
2907 pb->next_lblock = extent.e_lblk + extent.e_len;
2908 }
2909
2910 /*
2911 * Uninitialized blocks in a directory? Clear the flag and
2912 * we'll interpret the blocks later.
2913 */
2914 if (try_repairs && is_dir && problem == 0 &&
2915 (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
2916 fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
2917 extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
2918 pb->inode_modified = 1;
2919 pctx->errcode = ext2fs_extent_replace(ehandle, 0,
2920 &extent);
2921 if (pctx->errcode)
2922 return;
2923 failed_csum = 0;
2924 }
2925 #ifdef CONFIG_DEVELOPER_FEATURES
2926 if (try_repairs && !is_dir && problem == 0 &&
2927 (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
2928 (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
2929 fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
2930 extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
2931 pb->inode_modified = 1;
2932 pctx->errcode = ext2fs_extent_replace(ehandle, 0,
2933 &extent);
2934 if (pctx->errcode)
2935 return;
2936 failed_csum = 0;
2937 }
2938 #endif
2939 if (try_repairs && problem) {
2940 report_problem:
2941 if (fix_problem(ctx, problem, pctx)) {
2942 if (ctx->invalid_bitmaps) {
2943 /*
2944 * If fsck knows the bitmaps are bad,
2945 * skip to the next extent and
2946 * try to clear this extent again
2947 * after fixing the bitmaps, by
2948 * restarting fsck.
2949 */
2950 pctx->errcode = ext2fs_extent_get(
2951 ehandle,
2952 EXT2_EXTENT_NEXT_SIB,
2953 &extent);
2954 ctx->flags |= E2F_FLAG_RESTART_LATER;
2955 if (pctx->errcode ==
2956 EXT2_ET_NO_CURRENT_NODE) {
2957 pctx->errcode = 0;
2958 break;
2959 }
2960 continue;
2961 }
2962 e2fsck_read_bitmaps(ctx);
2963 pb->inode_modified = 1;
2964 pctx->errcode =
2965 ext2fs_extent_delete(ehandle, 0);
2966 if (pctx->errcode) {
2967 pctx->str = "ext2fs_extent_delete";
2968 return;
2969 }
2970 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
2971 if (pctx->errcode &&
2972 pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
2973 pctx->str = "ext2fs_extent_fix_parents";
2974 return;
2975 }
2976 pctx->errcode = ext2fs_extent_get(ehandle,
2977 EXT2_EXTENT_CURRENT,
2978 &extent);
2979 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
2980 pctx->errcode = 0;
2981 break;
2982 }
2983 failed_csum = 0;
2984 continue;
2985 }
2986 goto next;
2987 }
2988
2989 if (!is_leaf) {
2990 blk64_t lblk = extent.e_lblk;
2991 int next_try_repairs = 1;
2992
2993 blk = extent.e_pblk;
2994
2995 /*
2996 * If this lower extent block collides with critical
2997 * metadata, don't try to repair the damage. Pass 1b
2998 * will reallocate the block; then we can try again.
2999 */
3000 if (pb->ino != EXT2_RESIZE_INO &&
3001 extent.e_pblk < ctx->fs->super->s_blocks_count &&
3002 ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3003 extent.e_pblk)) {
3004 next_try_repairs = 0;
3005 pctx->blk = blk;
3006 fix_problem(ctx,
3007 PR_1_CRITICAL_METADATA_COLLISION,
3008 pctx);
3009 if ((ctx->options & E2F_OPT_NO) == 0)
3010 ctx->flags |= E2F_FLAG_RESTART_LATER;
3011 }
3012 pctx->errcode = ext2fs_extent_get(ehandle,
3013 EXT2_EXTENT_DOWN, &extent);
3014 if (pctx->errcode &&
3015 pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
3016 pctx->str = "EXT2_EXTENT_DOWN";
3017 problem = PR_1_EXTENT_HEADER_INVALID;
3018 if (!next_try_repairs)
3019 return;
3020 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
3021 goto report_problem;
3022 return;
3023 }
3024 /* The next extent should match this index's logical start */
3025 if (extent.e_lblk != lblk) {
3026 struct ext2_extent_info e_info;
3027
3028 pctx->errcode = ext2fs_extent_get_info(ehandle,
3029 &e_info);
3030 if (pctx->errcode) {
3031 pctx->str = "ext2fs_extent_get_info";
3032 return;
3033 }
3034 pctx->blk = lblk;
3035 pctx->blk2 = extent.e_lblk;
3036 pctx->num = e_info.curr_level - 1;
3037 problem = PR_1_EXTENT_INDEX_START_INVALID;
3038 if (fix_problem(ctx, problem, pctx)) {
3039 pb->inode_modified = 1;
3040 pctx->errcode =
3041 ext2fs_extent_fix_parents(ehandle);
3042 if (pctx->errcode) {
3043 pctx->str = "ext2fs_extent_fix_parents";
3044 return;
3045 }
3046 }
3047 }
3048 scan_extent_node(ctx, pctx, pb, extent.e_lblk,
3049 last_lblk, eof_block, ehandle,
3050 next_try_repairs);
3051 if (pctx->errcode)
3052 return;
3053 pctx->errcode = ext2fs_extent_get(ehandle,
3054 EXT2_EXTENT_UP, &extent);
3055 if (pctx->errcode) {
3056 pctx->str = "EXT2_EXTENT_UP";
3057 return;
3058 }
3059 mark_block_used(ctx, blk);
3060 pb->num_blocks++;
3061 goto next;
3062 }
3063
3064 if ((pb->previous_block != 0) &&
3065 (pb->previous_block+1 != extent.e_pblk)) {
3066 if (ctx->options & E2F_OPT_FRAGCHECK) {
3067 char type = '?';
3068
3069 if (pb->is_dir)
3070 type = 'd';
3071 else if (pb->is_reg)
3072 type = 'f';
3073
3074 printf(("%6lu(%c): expecting %6lu "
3075 "actual extent "
3076 "phys %6lu log %lu len %lu\n"),
3077 (unsigned long) pctx->ino, type,
3078 (unsigned long) pb->previous_block+1,
3079 (unsigned long) extent.e_pblk,
3080 (unsigned long) extent.e_lblk,
3081 (unsigned long) extent.e_len);
3082 }
3083 pb->fragmented = 1;
3084 }
3085 /*
3086 * If we notice a gap in the logical block mappings of an
3087 * extent-mapped directory, offer to close the hole by
3088 * moving the logical block down, otherwise we'll go mad in
3089 * pass 3 allocating empty directory blocks to fill the hole.
3090 */
3091 if (try_repairs && is_dir &&
3092 pb->last_block + 1 < extent.e_lblk) {
3093 blk64_t new_lblk;
3094
3095 new_lblk = pb->last_block + 1;
3096 if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3097 new_lblk = ((new_lblk +
3098 EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3099 ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3100 (extent.e_pblk &
3101 EXT2FS_CLUSTER_MASK(ctx->fs));
3102 pctx->blk = extent.e_lblk;
3103 pctx->blk2 = new_lblk;
3104 if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3105 extent.e_lblk = new_lblk;
3106 pb->inode_modified = 1;
3107 pctx->errcode = ext2fs_extent_replace(ehandle,
3108 0, &extent);
3109 if (pctx->errcode) {
3110 pctx->errcode = 0;
3111 goto alloc_later;
3112 }
3113 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3114 if (pctx->errcode)
3115 goto failed_add_dir_block;
3116 pctx->errcode = ext2fs_extent_goto(ehandle,
3117 extent.e_lblk);
3118 if (pctx->errcode)
3119 goto failed_add_dir_block;
3120 last_lblk = extent.e_lblk + extent.e_len - 1;
3121 failed_csum = 0;
3122 }
3123 }
3124 alloc_later:
3125 if (is_dir) {
3126 while (++pb->last_db_block <
3127 (e2_blkcnt_t) extent.e_lblk) {
3128 pctx->errcode = ext2fs_add_dir_block2(
3129 ctx->fs->dblist,
3130 pb->ino, 0,
3131 pb->last_db_block);
3132 if (pctx->errcode) {
3133 pctx->blk = 0;
3134 pctx->num = pb->last_db_block;
3135 goto failed_add_dir_block;
3136 }
3137 }
3138
3139 for (i = 0; i < extent.e_len; i++) {
3140 pctx->errcode = ext2fs_add_dir_block2(
3141 ctx->fs->dblist,
3142 pctx->ino,
3143 extent.e_pblk + i,
3144 extent.e_lblk + i);
3145 if (pctx->errcode) {
3146 pctx->blk = extent.e_pblk + i;
3147 pctx->num = extent.e_lblk + i;
3148 failed_add_dir_block:
3149 fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3150 /* Should never get here */
3151 ctx->flags |= E2F_FLAG_ABORT;
3152 return;
3153 }
3154 }
3155 if (extent.e_len > 0)
3156 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3157 }
3158 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3159 pb->last_block,
3160 extent.e_pblk,
3161 extent.e_lblk)) {
3162 for (i = 0; i < extent.e_len; i++) {
3163 pctx->blk = extent.e_lblk + i;
3164 pctx->blk2 = extent.e_pblk + i;
3165 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3166 mark_block_used(ctx, extent.e_pblk + i);
3167 mark_block_used(ctx, extent.e_pblk + i);
3168 }
3169 }
3170
3171 /*
3172 * Check whether first cluster got marked in previous iteration.
3173 */
3174 if (ctx->fs->cluster_ratio_bits &&
3175 pb->previous_block &&
3176 (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
3177 EXT2FS_B2C(ctx->fs, pb->previous_block)))
3178 /* Set blk to the beginning of next cluster. */
3179 blk = EXT2FS_C2B(
3180 ctx->fs,
3181 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
3182 else
3183 /* Set blk to the beginning of current cluster. */
3184 blk = EXT2FS_C2B(ctx->fs,
3185 EXT2FS_B2C(ctx->fs, extent.e_pblk));
3186
3187 if (blk < extent.e_pblk + extent.e_len) {
3188 mark_blocks_used(ctx, blk,
3189 extent.e_pblk + extent.e_len - blk);
3190 n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
3191 EXT2FS_CLUSTER_RATIO(ctx->fs));
3192 pb->num_blocks += n;
3193 }
3194 pb->last_block = extent.e_lblk + extent.e_len - 1;
3195 pb->previous_block = extent.e_pblk + extent.e_len - 1;
3196 start_block = pb->last_block = last_lblk;
3197 if (is_leaf && !is_dir &&
3198 !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
3199 pb->last_init_lblock = last_lblk;
3200 next:
3201 pctx->errcode = ext2fs_extent_get(ehandle,
3202 EXT2_EXTENT_NEXT_SIB,
3203 &extent);
3204 }
3205
3206 /* Failed csum but passes checks? Ask to fix checksum. */
3207 if (failed_csum &&
3208 fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
3209 pb->inode_modified = 1;
3210 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
3211 if (pctx->errcode)
3212 return;
3213 }
3214
3215 if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
3216 pctx->errcode = 0;
3217 }
3218
check_blocks_extents(e2fsck_t ctx,struct problem_context * pctx,struct process_block_struct * pb)3219 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
3220 struct process_block_struct *pb)
3221 {
3222 struct ext2_extent_info info;
3223 struct ext2_inode *inode = pctx->inode;
3224 ext2_extent_handle_t ehandle;
3225 ext2_filsys fs = ctx->fs;
3226 ext2_ino_t ino = pctx->ino;
3227 errcode_t retval;
3228 blk64_t eof_lblk;
3229 struct ext3_extent_header *eh;
3230
3231 /* Check for a proper extent header... */
3232 eh = (struct ext3_extent_header *) &inode->i_block[0];
3233 retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
3234 if (retval) {
3235 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
3236 e2fsck_clear_inode(ctx, ino, inode, 0,
3237 "check_blocks_extents");
3238 pctx->errcode = 0;
3239 return;
3240 }
3241
3242 /* ...since this function doesn't fail if i_block is zeroed. */
3243 pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
3244 if (pctx->errcode) {
3245 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
3246 e2fsck_clear_inode(ctx, ino, inode, 0,
3247 "check_blocks_extents");
3248 pctx->errcode = 0;
3249 return;
3250 }
3251
3252 retval = ext2fs_extent_get_info(ehandle, &info);
3253 if (retval == 0) {
3254 int max_depth = info.max_depth;
3255
3256 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
3257 max_depth = MAX_EXTENT_DEPTH_COUNT-1;
3258 ctx->extent_depth_count[max_depth]++;
3259 }
3260
3261 /* Check maximum extent depth */
3262 pctx->blk = info.max_depth;
3263 pctx->blk2 = ext2fs_max_extent_depth(ehandle);
3264 if (pctx->blk2 < pctx->blk &&
3265 fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
3266 pb->eti.force_rebuild = 1;
3267
3268 /* Can we collect extent tree level stats? */
3269 pctx->blk = MAX_EXTENT_DEPTH_COUNT;
3270 if (pctx->blk2 > pctx->blk)
3271 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
3272 memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
3273 pb->eti.ino = pb->ino;
3274
3275 pb->next_lblock = 0;
3276
3277 eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
3278 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
3279 scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
3280 if (pctx->errcode &&
3281 fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
3282 pb->num_blocks = 0;
3283 inode->i_blocks = 0;
3284 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3285 "check_blocks_extents");
3286 pctx->errcode = 0;
3287 }
3288 ext2fs_extent_free(ehandle);
3289
3290 /* Rebuild unless it's a dir and we're rehashing it */
3291 if (LINUX_S_ISDIR(inode->i_mode) &&
3292 e2fsck_dir_will_be_rehashed(ctx, ino))
3293 return;
3294
3295 if (ctx->options & E2F_OPT_CONVERT_BMAP)
3296 e2fsck_rebuild_extents_later(ctx, ino);
3297 else
3298 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
3299 }
3300
3301 /*
3302 * In fact we don't need to check blocks for an inode with inline data
3303 * because this inode doesn't have any blocks. In this function all
3304 * we need to do is add this inode into dblist when it is a directory.
3305 */
check_blocks_inline_data(e2fsck_t ctx,struct problem_context * pctx,struct process_block_struct * pb)3306 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
3307 struct process_block_struct *pb)
3308 {
3309 int flags;
3310 size_t inline_data_size = 0;
3311
3312 if (!pb->is_dir) {
3313 pctx->errcode = 0;
3314 return;
3315 }
3316
3317 /* Process the dirents in i_block[] as the "first" block. */
3318 pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
3319 if (pctx->errcode)
3320 goto err;
3321
3322 /* Process the dirents in the EA as a "second" block. */
3323 flags = ctx->fs->flags;
3324 ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3325 pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
3326 &inline_data_size);
3327 ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3328 (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3329 if (pctx->errcode) {
3330 pctx->errcode = 0;
3331 return;
3332 }
3333
3334 if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
3335 return;
3336
3337 pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
3338 if (pctx->errcode)
3339 goto err;
3340
3341 return;
3342 err:
3343 pctx->blk = 0;
3344 pctx->num = 0;
3345 fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3346 ctx->flags |= E2F_FLAG_ABORT;
3347 }
3348
3349 /*
3350 * This subroutine is called on each inode to account for all of the
3351 * blocks used by that inode.
3352 */
check_blocks(e2fsck_t ctx,struct problem_context * pctx,char * block_buf,const struct ea_quota * ea_ibody_quota)3353 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
3354 char *block_buf, const struct ea_quota *ea_ibody_quota)
3355 {
3356 ext2_filsys fs = ctx->fs;
3357 struct process_block_struct pb;
3358 ext2_ino_t ino = pctx->ino;
3359 struct ext2_inode *inode = pctx->inode;
3360 unsigned bad_size = 0;
3361 int dirty_inode = 0;
3362 int extent_fs;
3363 int inlinedata_fs;
3364 __u64 size;
3365 struct ea_quota ea_block_quota;
3366
3367 pb.ino = ino;
3368 pb.num_blocks = EXT2FS_B2C(ctx->fs,
3369 ea_ibody_quota ? ea_ibody_quota->blocks : 0);
3370 pb.last_block = ~0;
3371 pb.last_init_lblock = -1;
3372 pb.last_db_block = -1;
3373 pb.num_illegal_blocks = 0;
3374 pb.suppress = 0; pb.clear = 0;
3375 pb.fragmented = 0;
3376 pb.compressed = 0;
3377 pb.previous_block = 0;
3378 pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
3379 pb.is_reg = LINUX_S_ISREG(inode->i_mode);
3380 pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
3381 pb.inode = inode;
3382 pb.pctx = pctx;
3383 pb.ctx = ctx;
3384 pb.inode_modified = 0;
3385 pb.eti.force_rebuild = 0;
3386 pctx->ino = ino;
3387 pctx->errcode = 0;
3388
3389 extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
3390 inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
3391
3392 if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
3393 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3394 goto out;
3395 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
3396 }
3397
3398 if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
3399 check_blocks_inline_data(ctx, pctx, &pb);
3400 else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
3401 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
3402 check_blocks_extents(ctx, pctx, &pb);
3403 else {
3404 int flags;
3405 /*
3406 * If we've modified the inode, write it out before
3407 * iterate() tries to use it.
3408 */
3409 if (dirty_inode) {
3410 e2fsck_write_inode(ctx, ino, inode,
3411 "check_blocks");
3412 dirty_inode = 0;
3413 }
3414 flags = fs->flags;
3415 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3416 pctx->errcode = ext2fs_block_iterate3(fs, ino,
3417 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
3418 block_buf, process_block, &pb);
3419 /*
3420 * We do not have uninitialized extents in non extent
3421 * files.
3422 */
3423 pb.last_init_lblock = pb.last_block;
3424 /*
3425 * If iterate() changed a block mapping, we have to
3426 * re-read the inode. If we decide to clear the
3427 * inode after clearing some stuff, we'll re-write the
3428 * bad mappings into the inode!
3429 */
3430 if (pb.inode_modified)
3431 e2fsck_read_inode(ctx, ino, inode,
3432 "check_blocks");
3433 fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3434 (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3435
3436 if (ctx->options & E2F_OPT_CONVERT_BMAP) {
3437 #ifdef DEBUG
3438 printf("bmap rebuild ino=%d\n", ino);
3439 #endif
3440 if (!LINUX_S_ISDIR(inode->i_mode) ||
3441 !e2fsck_dir_will_be_rehashed(ctx, ino))
3442 e2fsck_rebuild_extents_later(ctx, ino);
3443 }
3444 }
3445 }
3446 end_problem_latch(ctx, PR_LATCH_BLOCK);
3447 end_problem_latch(ctx, PR_LATCH_TOOBIG);
3448 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3449 goto out;
3450 if (pctx->errcode)
3451 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
3452
3453 if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
3454 if (LINUX_S_ISDIR(inode->i_mode))
3455 ctx->fs_fragmented_dir++;
3456 else
3457 ctx->fs_fragmented++;
3458 }
3459
3460 if (pb.clear) {
3461 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3462 "check_blocks");
3463 return;
3464 }
3465
3466 if (inode->i_flags & EXT2_INDEX_FL) {
3467 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
3468 inode->i_flags &= ~EXT2_INDEX_FL;
3469 dirty_inode++;
3470 } else {
3471 e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
3472 }
3473 }
3474
3475 if (!pb.num_blocks && pb.is_dir &&
3476 !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
3477 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
3478 e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
3479 ctx->fs_directory_count--;
3480 return;
3481 }
3482 }
3483
3484 if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
3485 (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
3486 !(inode->i_flags & EXT4_EA_INODE_FL)) {
3487 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
3488 ino,
3489 pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
3490 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
3491 ino, (ea_ibody_quota ?
3492 ea_ibody_quota->inodes : 0) +
3493 ea_block_quota.inodes + 1);
3494 }
3495
3496 if (!ext2fs_has_feature_huge_file(fs->super) ||
3497 !(inode->i_flags & EXT4_HUGE_FILE_FL))
3498 pb.num_blocks *= (fs->blocksize / 512);
3499 pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
3500 #if 0
3501 printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
3502 ino, inode->i_size, (unsigned long long) pb.last_block,
3503 (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
3504 (unsigned long long) pb.num_blocks);
3505 #endif
3506 size = EXT2_I_SIZE(inode);
3507 if (pb.is_dir) {
3508 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
3509 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
3510 int flags;
3511 size_t sz = 0;
3512 errcode_t err;
3513
3514 flags = ctx->fs->flags;
3515 ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3516 err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
3517 &sz);
3518 ctx->fs->flags = (flags &
3519 EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3520 (ctx->fs->flags &
3521 ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3522 if (err || sz != size) {
3523 bad_size = 7;
3524 pctx->num = sz;
3525 }
3526 } else if (size & (fs->blocksize - 1))
3527 bad_size = 5;
3528 else if (nblock > (pb.last_block + 1))
3529 bad_size = 1;
3530 else if (nblock < (pb.last_block + 1)) {
3531 if (((pb.last_block + 1) - nblock) >
3532 fs->super->s_prealloc_dir_blocks)
3533 bad_size = 2;
3534 }
3535 } else {
3536 if ((pb.last_init_lblock >= 0) &&
3537 /* Do not allow initialized allocated blocks past i_size*/
3538 (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
3539 !(inode->i_flags & EXT4_VERITY_FL))
3540 bad_size = 3;
3541 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3542 size > ext2_max_sizes[fs->super->s_log_block_size])
3543 /* too big for a direct/indirect-mapped file */
3544 bad_size = 4;
3545 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3546 size >
3547 ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
3548 /* too big for an extent-based file - 32bit ee_block */
3549 bad_size = 6;
3550 }
3551 /* i_size for symlinks is checked elsewhere */
3552 if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
3553 /* Did inline_data set pctx->num earlier? */
3554 if (bad_size != 7)
3555 pctx->num = (pb.last_block + 1) * fs->blocksize;
3556 pctx->group = bad_size;
3557 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
3558 ext2fs_inode_size_set(fs, inode, pctx->num);
3559 if (EXT2_I_SIZE(inode) == 0 &&
3560 (inode->i_flags & EXT4_INLINE_DATA_FL)) {
3561 memset(inode->i_block, 0,
3562 sizeof(inode->i_block));
3563 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
3564 }
3565 dirty_inode++;
3566 }
3567 pctx->num = 0;
3568 }
3569 if (LINUX_S_ISREG(inode->i_mode) &&
3570 ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3571 ctx->large_files++;
3572 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
3573 ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
3574 (ext2fs_has_feature_huge_file(fs->super) &&
3575 (inode->i_flags & EXT4_HUGE_FILE_FL) &&
3576 (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
3577 pctx->num = pb.num_blocks;
3578 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
3579 inode->i_blocks = pb.num_blocks;
3580 inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
3581 dirty_inode++;
3582 }
3583 pctx->num = 0;
3584 }
3585
3586 /*
3587 * The kernel gets mad if we ask it to allocate bigalloc clusters to
3588 * a block mapped file, so rebuild it as an extent file. We can skip
3589 * symlinks because they're never rewritten.
3590 */
3591 if (ext2fs_has_feature_bigalloc(fs->super) &&
3592 (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
3593 ext2fs_inode_data_blocks2(fs, inode) > 0 &&
3594 (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
3595 !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
3596 fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
3597 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
3598 if (pctx->errcode)
3599 goto out;
3600 }
3601
3602 if (ctx->dirs_to_hash && pb.is_dir &&
3603 !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
3604 !(inode->i_flags & EXT2_INDEX_FL) &&
3605 ((inode->i_size / fs->blocksize) >= 3))
3606 e2fsck_rehash_dir_later(ctx, ino);
3607
3608 out:
3609 if (dirty_inode)
3610 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
3611 }
3612
3613 #if 0
3614 /*
3615 * Helper function called by process block when an illegal block is
3616 * found. It returns a description about why the block is illegal
3617 */
3618 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
3619 {
3620 blk64_t super;
3621 int i;
3622 static char problem[80];
3623
3624 super = fs->super->s_first_data_block;
3625 strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
3626 if (block < super) {
3627 sprintf(problem, "< FIRSTBLOCK (%u)", super);
3628 return(problem);
3629 } else if (block >= ext2fs_blocks_count(fs->super)) {
3630 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
3631 return(problem);
3632 }
3633 for (i = 0; i < fs->group_desc_count; i++) {
3634 if (block == super) {
3635 sprintf(problem, "is the superblock in group %d", i);
3636 break;
3637 }
3638 if (block > super &&
3639 block <= (super + fs->desc_blocks)) {
3640 sprintf(problem, "is in the group descriptors "
3641 "of group %d", i);
3642 break;
3643 }
3644 if (block == ext2fs_block_bitmap_loc(fs, i)) {
3645 sprintf(problem, "is the block bitmap of group %d", i);
3646 break;
3647 }
3648 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
3649 sprintf(problem, "is the inode bitmap of group %d", i);
3650 break;
3651 }
3652 if (block >= ext2fs_inode_table_loc(fs, i) &&
3653 (block < ext2fs_inode_table_loc(fs, i)
3654 + fs->inode_blocks_per_group)) {
3655 sprintf(problem, "is in the inode table of group %d",
3656 i);
3657 break;
3658 }
3659 super += fs->super->s_blocks_per_group;
3660 }
3661 return(problem);
3662 }
3663 #endif
3664
3665 /*
3666 * This is a helper function for check_blocks().
3667 */
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)3668 static int process_block(ext2_filsys fs,
3669 blk64_t *block_nr,
3670 e2_blkcnt_t blockcnt,
3671 blk64_t ref_block EXT2FS_ATTR((unused)),
3672 int ref_offset EXT2FS_ATTR((unused)),
3673 void *priv_data)
3674 {
3675 struct process_block_struct *p;
3676 struct problem_context *pctx;
3677 blk64_t blk = *block_nr;
3678 int ret_code = 0;
3679 problem_t problem = 0;
3680 e2fsck_t ctx;
3681
3682 p = (struct process_block_struct *) priv_data;
3683 pctx = p->pctx;
3684 ctx = p->ctx;
3685
3686 /*
3687 * For a directory, add logical block zero for processing even if it's
3688 * not mapped or we'll be perennially stuck with broken "." and ".."
3689 * entries.
3690 */
3691 if (p->is_dir && blockcnt == 0 && blk == 0) {
3692 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
3693 if (pctx->errcode) {
3694 pctx->blk = blk;
3695 pctx->num = blockcnt;
3696 goto failed_add_dir_block;
3697 }
3698 p->last_db_block++;
3699 }
3700
3701 if (blk == 0)
3702 return 0;
3703
3704 #if 0
3705 printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
3706 blockcnt);
3707 #endif
3708
3709 /*
3710 * Simplistic fragmentation check. We merely require that the
3711 * file be contiguous. (Which can never be true for really
3712 * big files that are greater than a block group.)
3713 */
3714 if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
3715 if (p->previous_block+1 != blk) {
3716 if (ctx->options & E2F_OPT_FRAGCHECK) {
3717 char type = '?';
3718
3719 if (p->is_dir)
3720 type = 'd';
3721 else if (p->is_reg)
3722 type = 'f';
3723
3724 printf(_("%6lu(%c): expecting %6lu "
3725 "got phys %6lu (blkcnt %lld)\n"),
3726 (unsigned long) pctx->ino, type,
3727 (unsigned long) p->previous_block+1,
3728 (unsigned long) blk,
3729 (long long) blockcnt);
3730 }
3731 p->fragmented = 1;
3732 }
3733 }
3734
3735 if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
3736 !pctx->inode->i_size_high &&
3737 blockcnt > (1 << (21 - fs->super->s_log_block_size)))
3738 problem = PR_1_TOOBIG_DIR;
3739 if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
3740 problem = PR_1_TOOBIG_DIR;
3741 if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
3742 problem = PR_1_TOOBIG_REG;
3743 if (!p->is_dir && !p->is_reg && blockcnt > 0)
3744 problem = PR_1_TOOBIG_SYMLINK;
3745
3746 if (blk < fs->super->s_first_data_block ||
3747 blk >= ext2fs_blocks_count(fs->super))
3748 problem = PR_1_ILLEGAL_BLOCK_NUM;
3749
3750 /*
3751 * If this IND/DIND/TIND block is squatting atop some critical metadata
3752 * (group descriptors, superblock, bitmap, inode table), any write to
3753 * "fix" mapping problems will destroy the metadata. We'll let pass 1b
3754 * fix that and restart fsck.
3755 */
3756 if (blockcnt < 0 &&
3757 p->ino != EXT2_RESIZE_INO &&
3758 blk < ctx->fs->super->s_blocks_count &&
3759 ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
3760 pctx->blk = blk;
3761 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
3762 if ((ctx->options & E2F_OPT_NO) == 0)
3763 ctx->flags |= E2F_FLAG_RESTART_LATER;
3764 }
3765
3766 if (problem) {
3767 p->num_illegal_blocks++;
3768 /*
3769 * A bit of subterfuge here -- we're trying to fix a block
3770 * mapping, but the IND/DIND/TIND block could have collided
3771 * with some critical metadata. So, fix the in-core mapping so
3772 * iterate won't go insane, but return 0 instead of
3773 * BLOCK_CHANGED so that it won't write the remapping out to
3774 * our multiply linked block.
3775 *
3776 * Even if we previously determined that an *IND block
3777 * conflicts with critical metadata, we must still try to
3778 * iterate the *IND block as if it is an *IND block to find and
3779 * mark the blocks it points to. Better to be overly cautious
3780 * with the used_blocks map so that we don't move the *IND
3781 * block to a block that's really in use!
3782 */
3783 if (p->ino != EXT2_RESIZE_INO &&
3784 ref_block != 0 &&
3785 ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3786 ref_block)) {
3787 *block_nr = 0;
3788 return 0;
3789 }
3790 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
3791 if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
3792 p->clear = 1;
3793 return BLOCK_ABORT;
3794 }
3795 if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
3796 p->suppress = 1;
3797 set_latch_flags(PR_LATCH_BLOCK,
3798 PRL_SUPPRESS, 0);
3799 }
3800 }
3801 pctx->blk = blk;
3802 pctx->blkcount = blockcnt;
3803 if (fix_problem(ctx, problem, pctx)) {
3804 blk = *block_nr = 0;
3805 ret_code = BLOCK_CHANGED;
3806 p->inode_modified = 1;
3807 /*
3808 * If the directory block is too big and is beyond the
3809 * end of the FS, don't bother trying to add it for
3810 * processing -- the kernel would never have created a
3811 * directory this large, and we risk an ENOMEM abort.
3812 * In any case, the toobig handler for extent-based
3813 * directories also doesn't feed toobig blocks to
3814 * pass 2.
3815 */
3816 if (problem == PR_1_TOOBIG_DIR)
3817 return ret_code;
3818 goto mark_dir;
3819 } else
3820 return 0;
3821 }
3822
3823 if (p->ino == EXT2_RESIZE_INO) {
3824 /*
3825 * The resize inode has already be sanity checked
3826 * during pass #0 (the superblock checks). All we
3827 * have to do is mark the double indirect block as
3828 * being in use; all of the other blocks are handled
3829 * by mark_table_blocks()).
3830 */
3831 if (blockcnt == BLOCK_COUNT_DIND)
3832 mark_block_used(ctx, blk);
3833 p->num_blocks++;
3834 } else if (!(ctx->fs->cluster_ratio_bits &&
3835 p->previous_block &&
3836 (EXT2FS_B2C(ctx->fs, blk) ==
3837 EXT2FS_B2C(ctx->fs, p->previous_block)) &&
3838 (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
3839 ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
3840 mark_block_used(ctx, blk);
3841 p->num_blocks++;
3842 } else if (has_unaligned_cluster_map(ctx, p->previous_block,
3843 p->last_block, blk, blockcnt)) {
3844 pctx->blk = blockcnt;
3845 pctx->blk2 = blk;
3846 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3847 mark_block_used(ctx, blk);
3848 mark_block_used(ctx, blk);
3849 }
3850 if (blockcnt >= 0)
3851 p->last_block = blockcnt;
3852 p->previous_block = blk;
3853 mark_dir:
3854 if (p->is_dir && (blockcnt >= 0)) {
3855 while (++p->last_db_block < blockcnt) {
3856 pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
3857 p->ino, 0,
3858 p->last_db_block);
3859 if (pctx->errcode) {
3860 pctx->blk = 0;
3861 pctx->num = p->last_db_block;
3862 goto failed_add_dir_block;
3863 }
3864 }
3865 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
3866 blk, blockcnt);
3867 if (pctx->errcode) {
3868 pctx->blk = blk;
3869 pctx->num = blockcnt;
3870 failed_add_dir_block:
3871 fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3872 /* Should never get here */
3873 ctx->flags |= E2F_FLAG_ABORT;
3874 return BLOCK_ABORT;
3875 }
3876 }
3877 return ret_code;
3878 }
3879
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)3880 static int process_bad_block(ext2_filsys fs,
3881 blk64_t *block_nr,
3882 e2_blkcnt_t blockcnt,
3883 blk64_t ref_block EXT2FS_ATTR((unused)),
3884 int ref_offset EXT2FS_ATTR((unused)),
3885 void *priv_data)
3886 {
3887 struct process_block_struct *p;
3888 blk64_t blk = *block_nr;
3889 blk64_t first_block;
3890 dgrp_t i;
3891 struct problem_context *pctx;
3892 e2fsck_t ctx;
3893
3894 if (!blk)
3895 return 0;
3896
3897 p = (struct process_block_struct *) priv_data;
3898 ctx = p->ctx;
3899 pctx = p->pctx;
3900
3901 pctx->ino = EXT2_BAD_INO;
3902 pctx->blk = blk;
3903 pctx->blkcount = blockcnt;
3904
3905 if ((blk < fs->super->s_first_data_block) ||
3906 (blk >= ext2fs_blocks_count(fs->super))) {
3907 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
3908 *block_nr = 0;
3909 return BLOCK_CHANGED;
3910 } else
3911 return 0;
3912 }
3913
3914 if (blockcnt < 0) {
3915 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
3916 p->bbcheck = 1;
3917 if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
3918 *block_nr = 0;
3919 return BLOCK_CHANGED;
3920 }
3921 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
3922 blk)) {
3923 p->bbcheck = 1;
3924 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
3925 pctx)) {
3926 *block_nr = 0;
3927 return BLOCK_CHANGED;
3928 }
3929 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3930 return BLOCK_ABORT;
3931 } else
3932 mark_block_used(ctx, blk);
3933 return 0;
3934 }
3935 #if 0
3936 printf ("DEBUG: Marking %u as bad.\n", blk);
3937 #endif
3938 ctx->fs_badblocks_count++;
3939 /*
3940 * If the block is not used, then mark it as used and return.
3941 * If it is already marked as found, this must mean that
3942 * there's an overlap between the filesystem table blocks
3943 * (bitmaps and inode table) and the bad block list.
3944 */
3945 if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
3946 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
3947 return 0;
3948 }
3949 /*
3950 * Try to find the where the filesystem block was used...
3951 */
3952 first_block = fs->super->s_first_data_block;
3953
3954 for (i = 0; i < fs->group_desc_count; i++ ) {
3955 pctx->group = i;
3956 pctx->blk = blk;
3957 if (!ext2fs_bg_has_super(fs, i))
3958 goto skip_super;
3959 if (blk == first_block) {
3960 if (i == 0) {
3961 if (fix_problem(ctx,
3962 PR_1_BAD_PRIMARY_SUPERBLOCK,
3963 pctx)) {
3964 *block_nr = 0;
3965 return BLOCK_CHANGED;
3966 }
3967 return 0;
3968 }
3969 fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
3970 return 0;
3971 }
3972 if ((blk > first_block) &&
3973 (blk <= first_block + fs->desc_blocks)) {
3974 if (i == 0) {
3975 pctx->blk = *block_nr;
3976 if (fix_problem(ctx,
3977 PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
3978 *block_nr = 0;
3979 return BLOCK_CHANGED;
3980 }
3981 return 0;
3982 }
3983 fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
3984 return 0;
3985 }
3986 skip_super:
3987 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
3988 if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
3989 ctx->invalid_block_bitmap_flag[i]++;
3990 ctx->invalid_bitmaps++;
3991 }
3992 return 0;
3993 }
3994 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
3995 if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
3996 ctx->invalid_inode_bitmap_flag[i]++;
3997 ctx->invalid_bitmaps++;
3998 }
3999 return 0;
4000 }
4001 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
4002 (blk < (ext2fs_inode_table_loc(fs, i) +
4003 fs->inode_blocks_per_group))) {
4004 /*
4005 * If there are bad blocks in the inode table,
4006 * the inode scan code will try to do
4007 * something reasonable automatically.
4008 */
4009 return 0;
4010 }
4011 first_block += fs->super->s_blocks_per_group;
4012 }
4013 /*
4014 * If we've gotten to this point, then the only
4015 * possibility is that the bad block inode meta data
4016 * is using a bad block.
4017 */
4018 if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
4019 (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
4020 (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
4021 p->bbcheck = 1;
4022 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
4023 *block_nr = 0;
4024 return BLOCK_CHANGED;
4025 }
4026 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
4027 return BLOCK_ABORT;
4028 return 0;
4029 }
4030
4031 pctx->group = -1;
4032
4033 /* Warn user that the block wasn't claimed */
4034 fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
4035
4036 return 0;
4037 }
4038
new_table_block(e2fsck_t ctx,blk64_t first_block,dgrp_t group,const char * name,int num,blk64_t * new_block)4039 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
4040 const char *name, int num, blk64_t *new_block)
4041 {
4042 ext2_filsys fs = ctx->fs;
4043 dgrp_t last_grp;
4044 blk64_t old_block = *new_block;
4045 blk64_t last_block;
4046 dgrp_t flexbg;
4047 unsigned flexbg_size;
4048 int i, is_flexbg;
4049 char *buf;
4050 struct problem_context pctx;
4051
4052 clear_problem_context(&pctx);
4053
4054 pctx.group = group;
4055 pctx.blk = old_block;
4056 pctx.str = name;
4057
4058 /*
4059 * For flex_bg filesystems, first try to allocate the metadata
4060 * within the flex_bg, and if that fails then try finding the
4061 * space anywhere in the filesystem.
4062 */
4063 is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
4064 if (is_flexbg) {
4065 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
4066 flexbg = group / flexbg_size;
4067 first_block = ext2fs_group_first_block2(fs,
4068 flexbg_size * flexbg);
4069 last_grp = group | (flexbg_size - 1);
4070 if (last_grp >= fs->group_desc_count)
4071 last_grp = fs->group_desc_count - 1;
4072 last_block = ext2fs_group_last_block2(fs, last_grp);
4073 } else
4074 last_block = ext2fs_group_last_block2(fs, group);
4075 pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
4076 num, ctx->block_found_map,
4077 new_block);
4078 if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
4079 pctx.errcode = ext2fs_get_free_blocks2(fs,
4080 fs->super->s_first_data_block,
4081 ext2fs_blocks_count(fs->super),
4082 num, ctx->block_found_map, new_block);
4083 if (pctx.errcode) {
4084 pctx.num = num;
4085 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4086 ext2fs_unmark_valid(fs);
4087 ctx->flags |= E2F_FLAG_ABORT;
4088 return;
4089 }
4090 pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4091 if (pctx.errcode) {
4092 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4093 ext2fs_unmark_valid(fs);
4094 ctx->flags |= E2F_FLAG_ABORT;
4095 return;
4096 }
4097 ext2fs_mark_super_dirty(fs);
4098 fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4099 pctx.blk2 = *new_block;
4100 fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4101 PR_1_RELOC_TO), &pctx);
4102 pctx.blk2 = 0;
4103 for (i = 0; i < num; i++) {
4104 pctx.blk = i;
4105 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4106 if (old_block) {
4107 pctx.errcode = io_channel_read_blk64(fs->io,
4108 old_block + i, 1, buf);
4109 if (pctx.errcode)
4110 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4111 pctx.blk = (*new_block) + i;
4112 pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4113 1, buf);
4114 } else {
4115 pctx.blk = (*new_block) + i;
4116 pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4117 NULL, NULL);
4118 }
4119
4120 if (pctx.errcode)
4121 fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4122 }
4123 ext2fs_free_mem(&buf);
4124 }
4125
4126 /*
4127 * This routine gets called at the end of pass 1 if bad blocks are
4128 * detected in the superblock, group descriptors, inode_bitmaps, or
4129 * block bitmaps. At this point, all of the blocks have been mapped
4130 * out, so we can try to allocate new block(s) to replace the bad
4131 * blocks.
4132 */
handle_fs_bad_blocks(e2fsck_t ctx)4133 static void handle_fs_bad_blocks(e2fsck_t ctx)
4134 {
4135 ext2_filsys fs = ctx->fs;
4136 dgrp_t i;
4137 blk64_t first_block;
4138 blk64_t new_blk;
4139
4140 for (i = 0; i < fs->group_desc_count; i++) {
4141 first_block = ext2fs_group_first_block2(fs, i);
4142
4143 if (ctx->invalid_block_bitmap_flag[i]) {
4144 new_blk = ext2fs_block_bitmap_loc(fs, i);
4145 new_table_block(ctx, first_block, i, _("block bitmap"),
4146 1, &new_blk);
4147 ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4148 }
4149 if (ctx->invalid_inode_bitmap_flag[i]) {
4150 new_blk = ext2fs_inode_bitmap_loc(fs, i);
4151 new_table_block(ctx, first_block, i, _("inode bitmap"),
4152 1, &new_blk);
4153 ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4154 }
4155 if (ctx->invalid_inode_table_flag[i]) {
4156 new_blk = ext2fs_inode_table_loc(fs, i);
4157 new_table_block(ctx, first_block, i, _("inode table"),
4158 fs->inode_blocks_per_group,
4159 &new_blk);
4160 ext2fs_inode_table_loc_set(fs, i, new_blk);
4161 ctx->flags |= E2F_FLAG_RESTART;
4162 }
4163 }
4164 ctx->invalid_bitmaps = 0;
4165 }
4166
4167 /*
4168 * This routine marks all blocks which are used by the superblock,
4169 * group descriptors, inode bitmaps, and block bitmaps.
4170 */
mark_table_blocks(e2fsck_t ctx)4171 static void mark_table_blocks(e2fsck_t ctx)
4172 {
4173 ext2_filsys fs = ctx->fs;
4174 blk64_t b;
4175 dgrp_t i;
4176 unsigned int j;
4177 struct problem_context pctx;
4178
4179 clear_problem_context(&pctx);
4180
4181 for (i = 0; i < fs->group_desc_count; i++) {
4182 pctx.group = i;
4183
4184 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
4185 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
4186
4187 /*
4188 * Mark the blocks used for the inode table
4189 */
4190 if (ext2fs_inode_table_loc(fs, i)) {
4191 for (j = 0, b = ext2fs_inode_table_loc(fs, i);
4192 j < fs->inode_blocks_per_group;
4193 j++, b++) {
4194 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4195 b)) {
4196 pctx.blk = b;
4197 if (!ctx->invalid_inode_table_flag[i] &&
4198 fix_problem(ctx,
4199 PR_1_ITABLE_CONFLICT, &pctx)) {
4200 ctx->invalid_inode_table_flag[i]++;
4201 ctx->invalid_bitmaps++;
4202 }
4203 } else {
4204 ext2fs_mark_block_bitmap2(
4205 ctx->block_found_map, b);
4206 ext2fs_mark_block_bitmap2(
4207 ctx->block_metadata_map, b);
4208 }
4209 }
4210 }
4211
4212 /*
4213 * Mark block used for the block bitmap
4214 */
4215 if (ext2fs_block_bitmap_loc(fs, i)) {
4216 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4217 ext2fs_block_bitmap_loc(fs, i))) {
4218 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
4219 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
4220 ctx->invalid_block_bitmap_flag[i]++;
4221 ctx->invalid_bitmaps++;
4222 }
4223 } else {
4224 ext2fs_mark_block_bitmap2(ctx->block_found_map,
4225 ext2fs_block_bitmap_loc(fs, i));
4226 ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4227 ext2fs_block_bitmap_loc(fs, i));
4228 }
4229 }
4230 /*
4231 * Mark block used for the inode bitmap
4232 */
4233 if (ext2fs_inode_bitmap_loc(fs, i)) {
4234 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4235 ext2fs_inode_bitmap_loc(fs, i))) {
4236 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
4237 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
4238 ctx->invalid_inode_bitmap_flag[i]++;
4239 ctx->invalid_bitmaps++;
4240 }
4241 } else {
4242 ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4243 ext2fs_inode_bitmap_loc(fs, i));
4244 ext2fs_mark_block_bitmap2(ctx->block_found_map,
4245 ext2fs_inode_bitmap_loc(fs, i));
4246 }
4247 }
4248 }
4249 }
4250
4251 /*
4252 * These subroutines short circuits ext2fs_get_blocks and
4253 * ext2fs_check_directory; we use them since we already have the inode
4254 * structure, so there's no point in letting the ext2fs library read
4255 * the inode again.
4256 */
pass1_get_blocks(ext2_filsys fs,ext2_ino_t ino,blk_t * blocks)4257 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
4258 blk_t *blocks)
4259 {
4260 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4261 int i;
4262
4263 if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4264 return EXT2_ET_CALLBACK_NOTHANDLED;
4265
4266 for (i=0; i < EXT2_N_BLOCKS; i++)
4267 blocks[i] = ctx->stashed_inode->i_block[i];
4268 return 0;
4269 }
4270
pass1_read_inode(ext2_filsys fs,ext2_ino_t ino,struct ext2_inode * inode)4271 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
4272 struct ext2_inode *inode)
4273 {
4274 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4275
4276 if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4277 return EXT2_ET_CALLBACK_NOTHANDLED;
4278 *inode = *ctx->stashed_inode;
4279 return 0;
4280 }
4281
pass1_write_inode(ext2_filsys fs,ext2_ino_t ino,struct ext2_inode * inode)4282 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
4283 struct ext2_inode *inode)
4284 {
4285 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4286
4287 if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
4288 (inode != ctx->stashed_inode))
4289 *ctx->stashed_inode = *inode;
4290 return EXT2_ET_CALLBACK_NOTHANDLED;
4291 }
4292
pass1_check_directory(ext2_filsys fs,ext2_ino_t ino)4293 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
4294 {
4295 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4296
4297 if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4298 return EXT2_ET_CALLBACK_NOTHANDLED;
4299
4300 if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
4301 return EXT2_ET_NO_DIRECTORY;
4302 return 0;
4303 }
4304
e2fsck_get_alloc_block(ext2_filsys fs,blk64_t goal,blk64_t * ret)4305 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
4306 blk64_t *ret)
4307 {
4308 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4309 errcode_t retval;
4310 blk64_t new_block;
4311
4312 if (ctx->block_found_map) {
4313 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
4314 &new_block);
4315 if (retval)
4316 return retval;
4317 if (fs->block_map) {
4318 ext2fs_mark_block_bitmap2(fs->block_map, new_block);
4319 ext2fs_mark_bb_dirty(fs);
4320 }
4321 } else {
4322 if (!fs->block_map) {
4323 retval = ext2fs_read_block_bitmap(fs);
4324 if (retval)
4325 return retval;
4326 }
4327
4328 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
4329 if (retval)
4330 return retval;
4331 }
4332
4333 *ret = new_block;
4334 return (0);
4335 }
4336
e2fsck_new_range(ext2_filsys fs,int flags,blk64_t goal,blk64_t len,blk64_t * pblk,blk64_t * plen)4337 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
4338 blk64_t len, blk64_t *pblk, blk64_t *plen)
4339 {
4340 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4341 errcode_t retval;
4342
4343 if (ctx->block_found_map)
4344 return ext2fs_new_range(fs, flags, goal, len,
4345 ctx->block_found_map, pblk, plen);
4346
4347 if (!fs->block_map) {
4348 retval = ext2fs_read_block_bitmap(fs);
4349 if (retval)
4350 return retval;
4351 }
4352
4353 return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
4354 pblk, plen);
4355 }
4356
e2fsck_block_alloc_stats(ext2_filsys fs,blk64_t blk,int inuse)4357 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
4358 {
4359 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4360
4361 /* Never free a critical metadata block */
4362 if (ctx->block_found_map &&
4363 ctx->block_metadata_map &&
4364 inuse < 0 &&
4365 ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
4366 return;
4367
4368 if (ctx->block_found_map) {
4369 if (inuse > 0)
4370 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4371 else
4372 ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
4373 }
4374 }
4375
e2fsck_block_alloc_stats_range(ext2_filsys fs,blk64_t blk,blk_t num,int inuse)4376 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
4377 blk_t num, int inuse)
4378 {
4379 e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4380
4381 /* Never free a critical metadata block */
4382 if (ctx->block_found_map &&
4383 ctx->block_metadata_map &&
4384 inuse < 0 &&
4385 ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
4386 return;
4387
4388 if (ctx->block_found_map) {
4389 if (inuse > 0)
4390 ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
4391 blk, num);
4392 else
4393 ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
4394 blk, num);
4395 }
4396 }
4397
e2fsck_use_inode_shortcuts(e2fsck_t ctx,int use_shortcuts)4398 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
4399 {
4400 ext2_filsys fs = ctx->fs;
4401
4402 if (use_shortcuts) {
4403 fs->get_blocks = pass1_get_blocks;
4404 fs->check_directory = pass1_check_directory;
4405 fs->read_inode = pass1_read_inode;
4406 fs->write_inode = pass1_write_inode;
4407 ctx->stashed_ino = 0;
4408 } else {
4409 fs->get_blocks = 0;
4410 fs->check_directory = 0;
4411 fs->read_inode = 0;
4412 fs->write_inode = 0;
4413 }
4414 }
4415
e2fsck_intercept_block_allocations(e2fsck_t ctx)4416 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
4417 {
4418 ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
4419 ext2fs_set_block_alloc_stats_callback(ctx->fs,
4420 e2fsck_block_alloc_stats, 0);
4421 ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
4422 ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
4423 e2fsck_block_alloc_stats_range, NULL);
4424 }
4425