Whamcloud - gitweb
e2fsck: merge bitmaps after thread completes
[tools/e2fsprogs.git] / e2fsck / pass1.c
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 #include <assert.h>
51 #ifdef HAVE_PTHREAD
52 #include <pthread.h>
53 #endif
54
55 #include "e2fsck.h"
56 #include <ext2fs/ext2_ext_attr.h>
57 /* todo remove this finally */
58 #include <ext2fs/ext2fsP.h>
59 #include <e2p/e2p.h>
60
61 #include "problem.h"
62
63 #ifdef NO_INLINE_FUNCS
64 #define _INLINE_
65 #else
66 #define _INLINE_ inline
67 #endif
68
69 #undef DEBUG
70
71 struct ea_quota {
72         blk64_t blocks;
73         __u64 inodes;
74 };
75
76 static int process_block(ext2_filsys fs, blk64_t        *blocknr,
77                          e2_blkcnt_t blockcnt, blk64_t ref_blk,
78                          int ref_offset, void *priv_data);
79 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
80                              e2_blkcnt_t blockcnt, blk64_t ref_blk,
81                              int ref_offset, void *priv_data);
82 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
83                          char *block_buf,
84                          const struct ea_quota *ea_ibody_quota);
85 static void mark_table_blocks(e2fsck_t ctx);
86 static void alloc_bb_map(e2fsck_t ctx);
87 static void alloc_imagic_map(e2fsck_t ctx);
88 static void mark_inode_bad(e2fsck_t ctx, ino_t ino);
89 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino);
90 static void handle_fs_bad_blocks(e2fsck_t ctx);
91 static void process_inodes(e2fsck_t ctx, char *block_buf);
92 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
93 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
94                                   dgrp_t group, void * priv_data);
95 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
96                                     char *block_buf, int adjust_sign);
97 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
98
99 struct process_block_struct {
100         ext2_ino_t      ino;
101         unsigned        is_dir:1, is_reg:1, clear:1, suppress:1,
102                                 fragmented:1, compressed:1, bbcheck:1,
103                                 inode_modified:1;
104         blk64_t         num_blocks;
105         blk64_t         max_blocks;
106         blk64_t         last_block;
107         e2_blkcnt_t     last_init_lblock;
108         e2_blkcnt_t     last_db_block;
109         int             num_illegal_blocks;
110         blk64_t         previous_block;
111         struct ext2_inode *inode;
112         struct problem_context *pctx;
113         ext2fs_block_bitmap fs_meta_blocks;
114         e2fsck_t        ctx;
115         blk64_t         next_lblock;
116         struct extent_tree_info eti;
117 };
118
119 struct process_inode_block {
120         ext2_ino_t ino;
121         struct ea_quota ea_ibody_quota;
122         struct ext2_inode_large inode;
123 };
124
125 struct scan_callback_struct {
126         e2fsck_t        ctx;
127         char            *block_buf;
128 };
129
130 /*
131  * For the inodes to process list.
132  */
133 static struct process_inode_block *inodes_to_process;
134 static int process_inode_count;
135
136 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
137                             EXT2_MIN_BLOCK_LOG_SIZE + 1];
138
139 /*
140  * Check to make sure a device inode is real.  Returns 1 if the device
141  * checks out, 0 if not.
142  *
143  * Note: this routine is now also used to check FIFO's and Sockets,
144  * since they have the same requirement; the i_block fields should be
145  * zero.
146  */
147 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
148                                     struct ext2_inode *inode)
149 {
150         int     i;
151
152         /*
153          * If the index or extents flag is set, then this is a bogus
154          * device/fifo/socket
155          */
156         if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
157                 return 0;
158
159         /*
160          * We should be able to do the test below all the time, but
161          * because the kernel doesn't forcibly clear the device
162          * inode's additional i_block fields, there are some rare
163          * occasions when a legitimate device inode will have non-zero
164          * additional i_block fields.  So for now, we only complain
165          * when the immutable flag is set, which should never happen
166          * for devices.  (And that's when the problem is caused, since
167          * you can't set or clear immutable flags for devices.)  Once
168          * the kernel has been fixed we can change this...
169          */
170         if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
171                 for (i=4; i < EXT2_N_BLOCKS; i++)
172                         if (inode->i_block[i])
173                                 return 0;
174         }
175         return 1;
176 }
177
178 /*
179  * Check to make sure a symlink inode is real.  Returns 1 if the symlink
180  * checks out, 0 if not.
181  */
182 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
183                                struct ext2_inode *inode, char *buf)
184 {
185         unsigned int buflen;
186         unsigned int len;
187
188         if ((inode->i_size_high || inode->i_size == 0) ||
189             (inode->i_flags & EXT2_INDEX_FL))
190                 return 0;
191
192         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
193                 size_t inline_size;
194
195                 if (inode->i_flags & EXT4_EXTENTS_FL)
196                         return 0;
197                 if (ext2fs_inline_data_size(fs, ino, &inline_size))
198                         return 0;
199                 if (inode->i_size != inline_size)
200                         return 0;
201
202                 return 1;
203         }
204
205         if (ext2fs_is_fast_symlink(inode)) {
206                 if (inode->i_flags & EXT4_EXTENTS_FL)
207                         return 0;
208                 buf = (char *)inode->i_block;
209                 buflen = sizeof(inode->i_block);
210         } else {
211                 ext2_extent_handle_t    handle;
212                 struct ext2_extent_info info;
213                 struct ext2fs_extent    extent;
214                 blk64_t blk;
215                 int i;
216
217                 if (inode->i_flags & EXT4_EXTENTS_FL) {
218                         if (ext2fs_extent_open2(fs, ino, inode, &handle))
219                                 return 0;
220                         if (ext2fs_extent_get_info(handle, &info) ||
221                             (info.num_entries != 1) ||
222                             (info.max_depth != 0)) {
223                                 ext2fs_extent_free(handle);
224                                 return 0;
225                         }
226                         if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
227                                               &extent) ||
228                             (extent.e_lblk != 0) ||
229                             (extent.e_len != 1)) {
230                                 ext2fs_extent_free(handle);
231                                 return 0;
232                         }
233                         blk = extent.e_pblk;
234                         ext2fs_extent_free(handle);
235                 } else {
236                         blk = inode->i_block[0];
237
238                         for (i = 1; i < EXT2_N_BLOCKS; i++)
239                                 if (inode->i_block[i])
240                                         return 0;
241                 }
242
243                 if (blk < fs->super->s_first_data_block ||
244                     blk >= ext2fs_blocks_count(fs->super))
245                         return 0;
246
247                 if (io_channel_read_blk64(fs->io, blk, 1, buf))
248                         return 0;
249
250                 buflen = fs->blocksize;
251         }
252
253         if (inode->i_flags & EXT4_ENCRYPT_FL)
254                 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
255         else
256                 len = strnlen(buf, buflen);
257
258         if (len >= buflen)
259                 return 0;
260
261         if (len != inode->i_size)
262                 return 0;
263         return 1;
264 }
265
266 /*
267  * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
268  */
269 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
270 static void check_extents_inlinedata(e2fsck_t ctx,
271                                      struct problem_context *pctx)
272 {
273         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
274                 return;
275
276         if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
277                 return;
278
279         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
280         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
281 }
282 #undef BAD_SPECIAL_FLAGS
283
284 /*
285  * If the immutable (or append-only) flag is set on the inode, offer
286  * to clear it.
287  */
288 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
289 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
290 {
291         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
292                 return;
293
294         if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
295                 return;
296
297         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
298         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
299 }
300
301 /*
302  * If device, fifo or socket, check size is zero -- if not offer to
303  * clear it
304  */
305 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
306 {
307         struct ext2_inode *inode = pctx->inode;
308
309         if (EXT2_I_SIZE(inode) == 0)
310                 return;
311
312         if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
313                 return;
314
315         ext2fs_inode_size_set(ctx->fs, inode, 0);
316         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
317 }
318
319 /*
320  * For a given size, calculate how many blocks would be charged towards quota.
321  */
322 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
323 {
324         blk64_t clusters;
325
326         clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
327         return EXT2FS_C2B(fs, clusters);
328 }
329
330 /*
331  * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
332  * the problem code.
333  */
334 static problem_t check_large_ea_inode(e2fsck_t ctx,
335                                       struct ext2_ext_attr_entry *entry,
336                                       struct problem_context *pctx,
337                                       blk64_t *quota_blocks)
338 {
339         struct ext2_inode inode;
340         __u32 hash;
341         errcode_t retval;
342
343         /* Check if inode is within valid range */
344         if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
345             (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
346                 pctx->num = entry->e_value_inum;
347                 return PR_1_ATTR_VALUE_EA_INODE;
348         }
349
350         e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
351
352         retval = ext2fs_ext_attr_hash_entry2(ctx->fs, entry, NULL, &hash);
353         if (retval) {
354                 com_err("check_large_ea_inode", retval,
355                         _("while hashing entry with e_value_inum = %u"),
356                         entry->e_value_inum);
357                 fatal_error(ctx, 0);
358         }
359
360         if (hash == entry->e_hash) {
361                 *quota_blocks = size_to_quota_blocks(ctx->fs,
362                                                      entry->e_value_size);
363         } else {
364                 /* This might be an old Lustre-style ea_inode reference. */
365                 if (inode.i_mtime == pctx->ino &&
366                     inode.i_generation == pctx->inode->i_generation) {
367                         *quota_blocks = 0;
368                 } else {
369                         /* If target inode is also missing EA_INODE flag,
370                          * this is likely to be a bad reference.
371                          */
372                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
373                                 pctx->num = entry->e_value_inum;
374                                 return PR_1_ATTR_VALUE_EA_INODE;
375                         } else {
376                                 pctx->num = entry->e_hash;
377                                 return PR_1_ATTR_HASH;
378                         }
379                 }
380         }
381
382         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
383                 pctx->num = entry->e_value_inum;
384                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
385                         inode.i_flags |= EXT4_EA_INODE_FL;
386                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
387                                            &inode);
388                 } else {
389                         return PR_1_ATTR_NO_EA_INODE_FL;
390                 }
391         }
392         return 0;
393 }
394
395 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
396                               struct ext2_ext_attr_entry *first, void *end)
397 {
398         struct ext2_ext_attr_entry *entry = first;
399         struct ext2_ext_attr_entry *np = EXT2_EXT_ATTR_NEXT(entry);
400
401         while ((void *) entry < end && (void *) np < end &&
402                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
403                 if (!entry->e_value_inum)
404                         goto next;
405                 if (!ctx->ea_inode_refs) {
406                         pctx->errcode = ea_refcount_create(0,
407                                                            &ctx->ea_inode_refs);
408                         if (pctx->errcode) {
409                                 pctx->num = 4;
410                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
411                                 ctx->flags |= E2F_FLAG_ABORT;
412                                 return;
413                         }
414                 }
415                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
416                                       0);
417         next:
418                 entry = np;
419                 np = EXT2_EXT_ATTR_NEXT(entry);
420         }
421 }
422
423 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
424                               struct ea_quota *ea_ibody_quota)
425 {
426         struct ext2_super_block *sb = ctx->fs->super;
427         struct ext2_inode_large *inode;
428         struct ext2_ext_attr_entry *entry;
429         char *start, *header, *end;
430         unsigned int storage_size, remain;
431         problem_t problem = 0;
432         region_t region = 0;
433
434         ea_ibody_quota->blocks = 0;
435         ea_ibody_quota->inodes = 0;
436
437         inode = (struct ext2_inode_large *) pctx->inode;
438         storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
439                 inode->i_extra_isize;
440         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
441                  inode->i_extra_isize;
442         end = header + storage_size;
443         start = header + sizeof(__u32);
444         entry = (struct ext2_ext_attr_entry *) start;
445
446         /* scan all entry's headers first */
447
448         /* take finish entry 0UL into account */
449         remain = storage_size - sizeof(__u32);
450
451         region = region_create(0, storage_size);
452         if (!region) {
453                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
454                 problem = 0;
455                 ctx->flags |= E2F_FLAG_ABORT;
456                 return;
457         }
458         if (region_allocate(region, 0, sizeof(__u32))) {
459                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
460                 goto fix;
461         }
462
463         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
464                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
465                 __u32 hash;
466
467                 if (region_allocate(region, (char *)entry - (char *)header,
468                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
469                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
470                         goto fix;
471                 }
472
473                 /* header eats this space */
474                 remain -= sizeof(struct ext2_ext_attr_entry);
475
476                 /* is attribute name valid? */
477                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
478                         pctx->num = entry->e_name_len;
479                         problem = PR_1_ATTR_NAME_LEN;
480                         goto fix;
481                 }
482
483                 /* attribute len eats this space */
484                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
485
486                 if (entry->e_value_inum == 0) {
487                         /* check value size */
488                         if (entry->e_value_size > remain) {
489                                 pctx->num = entry->e_value_size;
490                                 problem = PR_1_ATTR_VALUE_SIZE;
491                                 goto fix;
492                         }
493
494                         if (entry->e_value_size &&
495                             region_allocate(region,
496                                             sizeof(__u32) + entry->e_value_offs,
497                                             EXT2_EXT_ATTR_SIZE(
498                                                 entry->e_value_size))) {
499                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
500                                 goto fix;
501                         }
502
503                         hash = ext2fs_ext_attr_hash_entry(entry,
504                                                           start + entry->e_value_offs);
505
506                         /* e_hash may be 0 in older inode's ea */
507                         if (entry->e_hash != 0 && entry->e_hash != hash) {
508                                 pctx->num = entry->e_hash;
509                                 problem = PR_1_ATTR_HASH;
510                                 goto fix;
511                         }
512                 } else {
513                         blk64_t quota_blocks;
514
515                         problem = check_large_ea_inode(ctx, entry, pctx,
516                                                        &quota_blocks);
517                         if (problem != 0)
518                                 goto fix;
519
520                         ea_ibody_quota->blocks += quota_blocks;
521                         ea_ibody_quota->inodes++;
522                 }
523
524                 /* If EA value is stored in external inode then it does not
525                  * consume space here */
526                 if (entry->e_value_inum == 0)
527                         remain -= entry->e_value_size;
528
529                 entry = EXT2_EXT_ATTR_NEXT(entry);
530         }
531
532         if (region_allocate(region, (char *)entry - (char *)header,
533                             sizeof(__u32))) {
534                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
535                 goto fix;
536         }
537 fix:
538         if (region)
539                 region_free(region);
540         /*
541          * it seems like a corruption. it's very unlikely we could repair
542          * EA(s) in automatic fashion -bzzz
543          */
544         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
545                 inc_ea_inode_refs(ctx, pctx,
546                                   (struct ext2_ext_attr_entry *)start, end);
547                 return;
548         }
549
550         /* simply remove all possible EA(s) */
551         *((__u32 *)header) = 0UL;
552         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
553                                 EXT2_INODE_SIZE(sb), "pass1");
554         ea_ibody_quota->blocks = 0;
555         ea_ibody_quota->inodes = 0;
556 }
557
558 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
559         return (xtime & (1U << 31)) != 0 &&
560                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
561 }
562
563 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
564         check_inode_extra_negative_epoch(inode->i_##xtime, \
565                                          inode->i_##xtime##_extra)
566
567 /* When today's date is earlier than 2242, we assume that atimes,
568  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
569  * actually pre-1970 dates mis-encoded.
570  */
571 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
572
573 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
574                                     struct ea_quota *ea_ibody_quota)
575 {
576         struct ext2_super_block *sb = ctx->fs->super;
577         struct ext2_inode_large *inode;
578         __u32 *eamagic;
579         int min, max;
580
581         ea_ibody_quota->blocks = 0;
582         ea_ibody_quota->inodes = 0;
583
584         inode = (struct ext2_inode_large *) pctx->inode;
585         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
586                 /* this isn't large inode. so, nothing to check */
587                 return;
588         }
589
590 #if 0
591         printf("inode #%u, i_extra_size %d\n", pctx->ino,
592                         inode->i_extra_isize);
593 #endif
594         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
595         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
596         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
597         /*
598          * For now we will allow i_extra_isize to be 0, but really
599          * implementations should never allow i_extra_isize to be 0
600          */
601         if (inode->i_extra_isize &&
602             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
603              inode->i_extra_isize & 3)) {
604                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
605                         return;
606                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
607                         inode->i_extra_isize = sb->s_want_extra_isize;
608                 else
609                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
610                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
611                                         EXT2_INODE_SIZE(sb), "pass1");
612         }
613
614         /* check if there is no place for an EA header */
615         if (inode->i_extra_isize >= max - sizeof(__u32))
616                 return;
617
618         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
619                         inode->i_extra_isize);
620         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
621                 /* it seems inode has an extended attribute(s) in body */
622                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
623         }
624
625         /*
626          * If the inode's extended atime (ctime, crtime, mtime) is stored in
627          * the old, invalid format, repair it.
628          */
629         if (((sizeof(time_t) <= 4) ||
630              (((sizeof(time_t) > 4) &&
631                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
632             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
633              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
634              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
635              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
636
637                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
638                         return;
639
640                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
641                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
642                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
643                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
644                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
645                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
646                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
647                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
648                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
649                                         EXT2_INODE_SIZE(sb), "pass1");
650         }
651
652 }
653
654 /*
655  * Check to see if the inode might really be a directory, despite i_mode
656  *
657  * This is a lot of complexity for something for which I'm not really
658  * convinced happens frequently in the wild.  If for any reason this
659  * causes any problems, take this code out.
660  * [tytso:20070331.0827EDT]
661  */
662 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
663                                 char *buf)
664 {
665         struct ext2_inode *inode = pctx->inode;
666         struct ext2_dir_entry   *dirent;
667         errcode_t               retval;
668         blk64_t                 blk;
669         unsigned int            i, rec_len, not_device = 0;
670         int                     extent_fs;
671         int                     inlinedata_fs;
672
673         /*
674          * If the mode looks OK, we believe it.  If the first block in
675          * the i_block array is 0, this cannot be a directory. If the
676          * inode is extent-mapped, it is still the case that the latter
677          * cannot be 0 - the magic number in the extent header would make
678          * it nonzero.
679          */
680         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
681             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
682                 return;
683
684         /* 
685          * Check the block numbers in the i_block array for validity:
686          * zero blocks are skipped (but the first one cannot be zero -
687          * see above), other blocks are checked against the first and
688          * max data blocks (from the the superblock) and against the
689          * block bitmap. Any invalid block found means this cannot be
690          * a directory.
691          * 
692          * If there are non-zero blocks past the fourth entry, then
693          * this cannot be a device file: we remember that for the next
694          * check.
695          *
696          * For extent mapped files, we don't do any sanity checking:
697          * just try to get the phys block of logical block 0 and run
698          * with it.
699          *
700          * For inline data files, we just try to get the size of inline
701          * data.  If it's true, we will treat it as a directory.
702          */
703
704         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
705         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
706         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
707                 size_t size;
708                 __u32 dotdot;
709                 unsigned int rec_len2;
710                 struct ext2_dir_entry de;
711
712                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
713                         return;
714                 /*
715                  * If the size isn't a multiple of 4, it's probably not a
716                  * directory??
717                  */
718                 if (size & 3)
719                         return;
720                 /*
721                  * If the first 10 bytes don't look like a directory entry,
722                  * it's probably not a directory.
723                  */
724                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
725                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
726                        EXT2_DIR_REC_LEN(0));
727                 dotdot = ext2fs_le32_to_cpu(dotdot);
728                 de.inode = ext2fs_le32_to_cpu(de.inode);
729                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
730                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
731                 if (dotdot >= ctx->fs->super->s_inodes_count ||
732                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
733                      dotdot != EXT2_ROOT_INO) ||
734                     de.inode >= ctx->fs->super->s_inodes_count ||
735                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
736                      de.inode != 0) ||
737                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
738                               EXT4_INLINE_DATA_DOTDOT_SIZE)
739                         return;
740                 /* device files never have a "system.data" entry */
741                 goto isdir;
742         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
743                 /* extent mapped */
744                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
745                                  &blk))
746                         return;
747                 /* device files are never extent mapped */
748                 not_device++;
749         } else {
750                 for (i=0; i < EXT2_N_BLOCKS; i++) {
751                         blk = inode->i_block[i];
752                         if (!blk)
753                                 continue;
754                         if (i >= 4)
755                                 not_device++;
756
757                         if (blk < ctx->fs->super->s_first_data_block ||
758                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
759                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
760                                                            blk))
761                                 return; /* Invalid block, can't be dir */
762                 }
763                 blk = inode->i_block[0];
764         }
765
766         /*
767          * If the mode says this is a device file and the i_links_count field
768          * is sane and we have not ruled it out as a device file previously,
769          * we declare it a device file, not a directory.
770          */
771         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
772             (inode->i_links_count == 1) && !not_device)
773                 return;
774
775         /* read the first block */
776         ehandler_operation(_("reading directory block"));
777         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
778         ehandler_operation(0);
779         if (retval)
780                 return;
781
782         dirent = (struct ext2_dir_entry *) buf;
783         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
784         if (retval)
785                 return;
786         if ((ext2fs_dirent_name_len(dirent) != 1) ||
787             (dirent->name[0] != '.') ||
788             (dirent->inode != pctx->ino) ||
789             (rec_len < 12) ||
790             (rec_len % 4) ||
791             (rec_len >= ctx->fs->blocksize - 12))
792                 return;
793
794         dirent = (struct ext2_dir_entry *) (buf + rec_len);
795         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
796         if (retval)
797                 return;
798         if ((ext2fs_dirent_name_len(dirent) != 2) ||
799             (dirent->name[0] != '.') ||
800             (dirent->name[1] != '.') ||
801             (rec_len < 12) ||
802             (rec_len % 4))
803                 return;
804
805 isdir:
806         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
807                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
808                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
809                                         EXT2_INODE_SIZE(ctx->fs->super),
810                                         "check_is_really_dir");
811         }
812 }
813
814 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
815                                      int flags, ext2_icount_t hint,
816                                      ext2_icount_t *ret)
817 {
818         unsigned int            threshold;
819         unsigned int            save_type;
820         ext2_ino_t              num_dirs;
821         errcode_t               retval;
822         char                    *tdb_dir;
823         int                     enable;
824
825         *ret = 0;
826
827         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
828                            &tdb_dir);
829         profile_get_uint(ctx->profile, "scratch_files",
830                          "numdirs_threshold", 0, 0, &threshold);
831         profile_get_boolean(ctx->profile, "scratch_files",
832                             "icount", 0, 1, &enable);
833
834         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
835         if (retval)
836                 num_dirs = 1024;        /* Guess */
837
838         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
839             (!threshold || num_dirs > threshold)) {
840                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
841                                                   flags, ret);
842                 if (retval == 0)
843                         return 0;
844         }
845         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
846                                &save_type);
847         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
848                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
849         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
850         ctx->fs->default_bitmap_type = save_type;
851         return retval;
852 }
853
854 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
855                                             e2fsck_t ctx,
856                                             struct problem_context *pctx)
857 {
858         errcode_t retval;
859         struct ext2_inode_large inode;
860
861         /*
862          * Reread inode.  If we don't see checksum error, then this inode
863          * has been fixed elsewhere.
864          */
865         ctx->stashed_ino = 0;
866         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
867                                         sizeof(inode));
868         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
869                 return retval;
870         if (!retval)
871                 return 0;
872
873         /*
874          * Checksum still doesn't match.  That implies that the inode passes
875          * all the sanity checks, so maybe the checksum is simply corrupt.
876          * See if the user will go for fixing that.
877          */
878         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
879                 return 0;
880
881         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
882                                          sizeof(inode));
883         return retval;
884 }
885
886 static void reserve_block_for_root_repair(e2fsck_t ctx)
887 {
888         blk64_t         blk = 0;
889         errcode_t       err;
890         ext2_filsys     fs = ctx->fs;
891
892         ctx->root_repair_block = 0;
893         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
894                 return;
895
896         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
897         if (err)
898                 return;
899         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
900         ctx->root_repair_block = blk;
901 }
902
903 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
904 {
905         blk64_t         blk = 0;
906         errcode_t       err;
907         ext2_filsys     fs = ctx->fs;
908         static const char name[] = "lost+found";
909         ext2_ino_t      ino;
910
911         ctx->lnf_repair_block = 0;
912         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
913                 return;
914
915         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
916         if (err)
917                 return;
918         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
919         ctx->lnf_repair_block = blk;
920 }
921
922 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
923                                          size_t *sz)
924 {
925         void *p;
926         struct ext2_xattr_handle *handle;
927         errcode_t retval;
928
929         retval = ext2fs_xattrs_open(fs, ino, &handle);
930         if (retval)
931                 return retval;
932
933         retval = ext2fs_xattrs_read(handle);
934         if (retval)
935                 goto err;
936
937         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
938         if (retval)
939                 goto err;
940         ext2fs_free_mem(&p);
941 err:
942         (void) ext2fs_xattrs_close(&handle);
943         return retval;
944 }
945
946 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
947                                     struct problem_context *pctx,
948                                     int failed_csum)
949 {
950         if (!failed_csum)
951                 return;
952
953         /*
954          * If the inode failed the checksum and the user didn't
955          * clear the inode, test the checksum again -- if it still
956          * fails, ask the user if the checksum should be corrected.
957          */
958         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
959         if (pctx->errcode)
960                 ctx->flags |= E2F_FLAG_ABORT;
961 }
962 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
963         do { \
964                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
965                 if ((ctx)->flags & E2F_FLAG_ABORT) \
966                         return; \
967         } while (0)
968
969 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
970 {
971         __u32 x;
972         int i;
973
974         for (i = 0; i < EXT2_N_BLOCKS; i++) {
975                 x = inode->i_block[i];
976 #ifdef WORDS_BIGENDIAN
977                 x = ext2fs_swab32(x);
978 #endif
979                 if (x >= ext2fs_blocks_count(fs->super))
980                         return 0;
981         }
982
983         return 1;
984 }
985
986 /*
987  * Figure out what to do with an inode that has both extents and inline data
988  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
989  */
990 static int fix_inline_data_extents_file(e2fsck_t ctx,
991                                         ext2_ino_t ino,
992                                         struct ext2_inode *inode,
993                                         int inode_size,
994                                         struct problem_context *pctx)
995 {
996         size_t max_inline_ea_size;
997         ext2_filsys fs = ctx->fs;
998         int dirty = 0;
999
1000         /* Both feature flags not set?  Just run the regular checks */
1001         if (!ext2fs_has_feature_extents(fs->super) &&
1002             !ext2fs_has_feature_inline_data(fs->super))
1003                 return 0;
1004
1005         /* Clear both flags if it's a special file */
1006         if (LINUX_S_ISCHR(inode->i_mode) ||
1007             LINUX_S_ISBLK(inode->i_mode) ||
1008             LINUX_S_ISFIFO(inode->i_mode) ||
1009             LINUX_S_ISSOCK(inode->i_mode)) {
1010                 check_extents_inlinedata(ctx, pctx);
1011                 return 0;
1012         }
1013
1014         /* If it looks like an extent tree, try to clear inlinedata */
1015         if (ext2fs_extent_header_verify(inode->i_block,
1016                                  sizeof(inode->i_block)) == 0 &&
1017             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1018                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1019                 dirty = 1;
1020                 goto out;
1021         }
1022
1023         /* If it looks short enough to be inline data, try to clear extents */
1024         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1025                 max_inline_ea_size = inode_size -
1026                                      (EXT2_GOOD_OLD_INODE_SIZE +
1027                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1028         else
1029                 max_inline_ea_size = 0;
1030         if (EXT2_I_SIZE(inode) <
1031             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1032             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1033                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1034                 dirty = 1;
1035                 goto out;
1036         }
1037
1038         /*
1039          * Too big for inline data, but no evidence of extent tree -
1040          * maybe it's a block map file?  If the mappings all look valid?
1041          */
1042         if (could_be_block_map(fs, inode) &&
1043             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1044 #ifdef WORDS_BIGENDIAN
1045                 int i;
1046
1047                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1048                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1049 #endif
1050
1051                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1052                 dirty = 1;
1053                 goto out;
1054         }
1055
1056         /* Oh well, just clear the busted inode. */
1057         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1058                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1059                 return -1;
1060         }
1061
1062 out:
1063         if (dirty)
1064                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1065
1066         return 0;
1067 }
1068
1069 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1070 {
1071         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1072         dgrp_t start = *group, grp;
1073         blk64_t blocks_to_read = 0;
1074         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1075
1076         if (ctx->readahead_kb == 0)
1077                 goto out;
1078
1079         /* Keep iterating groups until we have enough to readahead */
1080         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1081         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1082                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1083                         continue;
1084                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1085                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1086                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1087                                         inodes_per_block;
1088                 if (blocks_to_read * ctx->fs->blocksize >
1089                     ctx->readahead_kb * 1024)
1090                         break;
1091         }
1092
1093         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1094                                grp - start + 1);
1095         if (err == EAGAIN) {
1096                 ctx->readahead_kb /= 2;
1097                 err = 0;
1098         }
1099
1100 out:
1101         if (err) {
1102                 /* Error; disable itable readahead */
1103                 *group = ctx->fs->group_desc_count;
1104                 *next_ino = ctx->fs->super->s_inodes_count;
1105         } else {
1106                 /*
1107                  * Don't do more readahead until we've reached the first inode
1108                  * of the last inode scan buffer block for the last group.
1109                  */
1110                 *group = grp + 1;
1111                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1112                                      ctx->inode_buffer_blocks :
1113                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1114                                     ctx->fs->blocksize /
1115                                     EXT2_INODE_SIZE(ctx->fs->super);
1116                 inodes_in_group--;
1117                 *next_ino = inodes_in_group -
1118                             (inodes_in_group % inodes_per_buffer) + 1 +
1119                             (grp * ctx->fs->super->s_inodes_per_group);
1120         }
1121 }
1122
1123 /*
1124  * Check if the passed ino is one of the used superblock quota inodes.
1125  *
1126  * Before the quota inodes were journaled, older superblock quota inodes
1127  * were just regular files in the filesystem and not reserved inodes.  This
1128  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1129  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1130  */
1131 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1132 {
1133         enum quota_type qtype;
1134
1135         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1136                 if (*quota_sb_inump(sb, qtype) == ino)
1137                         return 1;
1138
1139         return 0;
1140 }
1141
1142 /*
1143  * Check if the passed ino is one of the reserved quota inodes.
1144  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1145  * inodes.  These inodes may or may not be in use by the quota feature.
1146  */
1147 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1148 {
1149         enum quota_type qtype;
1150
1151         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1152                 if (quota_type2inum(qtype, fs->super) == ino)
1153                         return 1;
1154
1155         return 0;
1156 }
1157
1158 static int e2fsck_should_abort(e2fsck_t ctx)
1159 {
1160         e2fsck_t global_ctx;
1161
1162         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1163                 return 1;
1164
1165         if (ctx->global_ctx) {
1166                 global_ctx = ctx->global_ctx;
1167                 if (global_ctx->flags & E2F_FLAG_SIGNAL_MASK)
1168                         return 1;
1169         }
1170         return 0;
1171 }
1172
1173 void e2fsck_pass1_run(e2fsck_t ctx)
1174 {
1175         int     i;
1176         __u64   max_sizes;
1177         ext2_filsys fs = ctx->fs;
1178         ext2_ino_t      ino = 0;
1179         struct ext2_inode *inode = NULL;
1180         ext2_inode_scan scan = NULL;
1181         char            *block_buf = NULL;
1182 #ifdef RESOURCE_TRACK
1183         struct resource_track   rtrack;
1184 #endif
1185         unsigned char   frag, fsize;
1186         struct          problem_context pctx;
1187         struct          scan_callback_struct scan_struct;
1188         struct ext2_super_block *sb = ctx->fs->super;
1189         const char      *old_op;
1190         const char      *eop_next_inode = _("getting next inode from scan");
1191         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1192         int             low_dtime_check = 1;
1193         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1194         unsigned int    bufsize;
1195         int             failed_csum = 0;
1196         ext2_ino_t      ino_threshold = 0;
1197         dgrp_t          ra_group = 0;
1198         struct ea_quota ea_ibody_quota;
1199
1200         init_resource_track(&rtrack, ctx->fs->io);
1201         clear_problem_context(&pctx);
1202
1203         /* If we can do readahead, figure out how many groups to pull in. */
1204         if (!e2fsck_can_readahead(ctx->fs))
1205                 ctx->readahead_kb = 0;
1206         else if (ctx->readahead_kb == ~0ULL)
1207                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1208         pass1_readahead(ctx, &ra_group, &ino_threshold);
1209
1210         if (!(ctx->options & E2F_OPT_PREEN))
1211                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1212
1213         if (ext2fs_has_feature_dir_index(fs->super) &&
1214             !(ctx->options & E2F_OPT_NO)) {
1215                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1216                         ctx->dirs_to_hash = 0;
1217         }
1218
1219 #ifdef MTRACE
1220         mtrace_print("Pass 1");
1221 #endif
1222
1223 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1224
1225         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1226                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1227                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1228                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1229                 max_sizes = (max_sizes * (1UL << i));
1230                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1231         }
1232 #undef EXT2_BPP
1233
1234         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1235         extent_fs = ext2fs_has_feature_extents(sb);
1236         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1237         casefold_fs = ext2fs_has_feature_casefold(sb);
1238
1239         /*
1240          * Allocate bitmaps structures
1241          */
1242         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1243                                                     EXT2FS_BMAP64_RBTREE,
1244                                                     "inode_used_map",
1245                                                     &ctx->inode_used_map);
1246         if (pctx.errcode) {
1247                 pctx.num = 1;
1248                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1249                 ctx->flags |= E2F_FLAG_ABORT;
1250                 return;
1251         }
1252         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1253                         _("directory inode map"),
1254                         EXT2FS_BMAP64_AUTODIR,
1255                         "inode_dir_map", &ctx->inode_dir_map);
1256         if (pctx.errcode) {
1257                 pctx.num = 2;
1258                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1259                 ctx->flags |= E2F_FLAG_ABORT;
1260                 return;
1261         }
1262         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1263                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1264                         "inode_reg_map", &ctx->inode_reg_map);
1265         if (pctx.errcode) {
1266                 pctx.num = 6;
1267                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1268                 ctx->flags |= E2F_FLAG_ABORT;
1269                 return;
1270         }
1271         pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1272                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1273                         "block_found_map", &ctx->block_found_map);
1274         if (pctx.errcode) {
1275                 pctx.num = 1;
1276                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1277                 ctx->flags |= E2F_FLAG_ABORT;
1278                 return;
1279         }
1280         pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1281                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1282                         "block_metadata_map", &ctx->block_metadata_map);
1283         if (pctx.errcode) {
1284                 pctx.num = 1;
1285                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1286                 ctx->flags |= E2F_FLAG_ABORT;
1287                 return;
1288         }
1289         if (casefold_fs) {
1290                 pctx.errcode =
1291                         e2fsck_allocate_inode_bitmap(fs,
1292                                                      _("inode casefold map"),
1293                                                      EXT2FS_BMAP64_RBTREE,
1294                                                      "inode_casefold_map",
1295                                                      &ctx->inode_casefold_map);
1296                 if (pctx.errcode) {
1297                         pctx.num = 1;
1298                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1299                         ctx->flags |= E2F_FLAG_ABORT;
1300                         return;
1301                 }
1302         }
1303         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1304                                            &ctx->inode_link_info);
1305         if (pctx.errcode) {
1306                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1307                 ctx->flags |= E2F_FLAG_ABORT;
1308                 return;
1309         }
1310         bufsize = inode_size;
1311         if (bufsize < sizeof(struct ext2_inode_large))
1312                 bufsize = sizeof(struct ext2_inode_large);
1313         inode = (struct ext2_inode *)
1314                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1315
1316         inodes_to_process = (struct process_inode_block *)
1317                 e2fsck_allocate_memory(ctx,
1318                                        (ctx->process_inode_size *
1319                                         sizeof(struct process_inode_block)),
1320                                        "array of inodes to process");
1321         process_inode_count = 0;
1322
1323         pctx.errcode = ext2fs_init_dblist(fs, 0);
1324         if (pctx.errcode) {
1325                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1326                 ctx->flags |= E2F_FLAG_ABORT;
1327                 goto endit;
1328         }
1329
1330         /*
1331          * If the last orphan field is set, clear it, since the pass1
1332          * processing will automatically find and clear the orphans.
1333          * In the future, we may want to try using the last_orphan
1334          * linked list ourselves, but for now, we clear it so that the
1335          * ext3 mount code won't get confused.
1336          */
1337         if (!(ctx->options & E2F_OPT_READONLY)) {
1338                 if (fs->super->s_last_orphan) {
1339                         fs->super->s_last_orphan = 0;
1340                         ext2fs_mark_super_dirty(fs);
1341                 }
1342         }
1343
1344         mark_table_blocks(ctx);
1345         pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1346                                                 &ctx->block_found_map);
1347         if (pctx.errcode) {
1348                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1349                 ctx->flags |= E2F_FLAG_ABORT;
1350                 goto endit;
1351         }
1352         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1353                                                     "block interate buffer");
1354         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1355                 e2fsck_use_inode_shortcuts(ctx, 1);
1356         e2fsck_intercept_block_allocations(ctx);
1357         old_op = ehandler_operation(_("opening inode scan"));
1358         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1359                                               &scan);
1360         ehandler_operation(old_op);
1361         if (pctx.errcode) {
1362                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1363                 ctx->flags |= E2F_FLAG_ABORT;
1364                 goto endit;
1365         }
1366         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1367                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1368         ctx->stashed_inode = inode;
1369         scan_struct.ctx = ctx;
1370         scan_struct.block_buf = block_buf;
1371         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1372         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1373                                               ctx->fs->group_desc_count)))
1374                 goto endit;
1375         if ((fs->super->s_wtime &&
1376              fs->super->s_wtime < fs->super->s_inodes_count) ||
1377             (fs->super->s_mtime &&
1378              fs->super->s_mtime < fs->super->s_inodes_count) ||
1379             (fs->super->s_mkfs_time &&
1380              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1381                 low_dtime_check = 0;
1382
1383         if (ext2fs_has_feature_mmp(fs->super) &&
1384             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1385             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1386                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1387                                           fs->super->s_mmp_block);
1388
1389         /* Set up ctx->lost_and_found if possible */
1390         (void) e2fsck_get_lost_and_found(ctx, 0);
1391
1392 #ifdef HAVE_PTHREAD
1393         if (ctx->global_ctx) {
1394                 if (ctx->options & E2F_OPT_DEBUG &&
1395                     ctx->options & E2F_OPT_MULTITHREAD)
1396                         fprintf(stderr, "thread %d jumping to group %d\n",
1397                                         ctx->thread_info.et_thread_index,
1398                                         ctx->thread_info.et_group_start);
1399                 pctx.errcode = ext2fs_inode_scan_goto_blockgroup(scan,
1400                                         ctx->thread_info.et_group_start);
1401                 if (pctx.errcode) {
1402                         fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1403                         ctx->flags |= E2F_FLAG_ABORT;
1404                         goto endit;
1405                 }
1406         }
1407 #endif
1408
1409         while (1) {
1410                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1411                         if (e2fsck_mmp_update(fs))
1412                                 fatal_error(ctx, 0);
1413                 }
1414                 old_op = ehandler_operation(eop_next_inode);
1415                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1416                                                           inode, inode_size);
1417                 if (ino > ino_threshold)
1418                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1419                 ehandler_operation(old_op);
1420                 if (e2fsck_should_abort(ctx))
1421                         goto endit;
1422                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1423                         /*
1424                          * If badblocks says badblocks is bad, offer to clear
1425                          * the list, update the in-core bb list, and restart
1426                          * the inode scan.
1427                          */
1428                         if (ino == EXT2_BAD_INO &&
1429                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1430                                         &pctx)) {
1431                                 errcode_t err;
1432
1433                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1434                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1435                                 ctx->fs->badblocks = NULL;
1436                                 err = ext2fs_read_bb_inode(ctx->fs,
1437                                                         &ctx->fs->badblocks);
1438                                 if (err) {
1439                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1440                                                     &pctx);
1441                                         ctx->flags |= E2F_FLAG_ABORT;
1442                                 } else
1443                                         ctx->flags |= E2F_FLAG_RESTART;
1444                                 goto endit;
1445                         }
1446                         if (!ctx->inode_bb_map)
1447                                 alloc_bb_map(ctx);
1448                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1449                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1450                         continue;
1451                 }
1452                 if (pctx.errcode == EXT2_ET_SCAN_FINISHED)
1453                         break;
1454                 if (pctx.errcode &&
1455                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1456                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1457                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1458                         ctx->flags |= E2F_FLAG_ABORT;
1459                         goto endit;
1460                 }
1461                 if (!ino)
1462                         break;
1463 #ifdef HAVE_PTHREAD
1464                 if (ctx->global_ctx)
1465                         ctx->thread_info.et_inode_number++;
1466 #endif
1467                 pctx.ino = ino;
1468                 pctx.inode = inode;
1469                 ctx->stashed_ino = ino;
1470
1471                 /* Clear trashed inode? */
1472                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1473                     inode->i_links_count > 0 &&
1474                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1475                         pctx.errcode = 0;
1476                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1477                 }
1478                 failed_csum = pctx.errcode != 0;
1479
1480                 /*
1481                  * Check for inodes who might have been part of the
1482                  * orphaned list linked list.  They should have gotten
1483                  * dealt with by now, unless the list had somehow been
1484                  * corrupted.
1485                  *
1486                  * FIXME: In the future, inodes which are still in use
1487                  * (and which are therefore) pending truncation should
1488                  * be handled specially.  Right now we just clear the
1489                  * dtime field, and the normal e2fsck handling of
1490                  * inodes where i_size and the inode blocks are
1491                  * inconsistent is to fix i_size, instead of releasing
1492                  * the extra blocks.  This won't catch the inodes that
1493                  * was at the end of the orphan list, but it's better
1494                  * than nothing.  The right answer is that there
1495                  * shouldn't be any bugs in the orphan list handling.  :-)
1496                  */
1497                 if (inode->i_dtime && low_dtime_check &&
1498                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1499                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1500                                 inode->i_dtime = inode->i_links_count ?
1501                                         0 : ctx->now;
1502                                 e2fsck_write_inode(ctx, ino, inode,
1503                                                    "pass1");
1504                                 failed_csum = 0;
1505                         }
1506                 }
1507
1508                 if (inode->i_links_count) {
1509                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1510                                            ino, inode->i_links_count);
1511                         if (pctx.errcode) {
1512                                 pctx.num = inode->i_links_count;
1513                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1514                                 ctx->flags |= E2F_FLAG_ABORT;
1515                                 goto endit;
1516                         }
1517                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1518                            !quota_inum_is_reserved(fs, ino)) {
1519                         if (!inode->i_dtime && inode->i_mode) {
1520                                 if (fix_problem(ctx,
1521                                             PR_1_ZERO_DTIME, &pctx)) {
1522                                         inode->i_dtime = ctx->now;
1523                                         e2fsck_write_inode(ctx, ino, inode,
1524                                                            "pass1");
1525                                         failed_csum = 0;
1526                                 }
1527                         }
1528                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1529                         continue;
1530                 }
1531
1532                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1533                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1534                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1535                      (!casefold_fs &&
1536                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1537                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1538                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1539                 }
1540
1541                 /* Conflicting inlinedata/extents inode flags? */
1542                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1543                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1544                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1545                                                                inode_size,
1546                                                                &pctx);
1547                         if (res < 0) {
1548                                 /* skip FINISH_INODE_LOOP */
1549                                 continue;
1550                         }
1551                 }
1552
1553                 /* Test for incorrect inline_data flags settings. */
1554                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1555                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1556                         size_t size = 0;
1557
1558                         pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1559                         if (!pctx.errcode &&
1560                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1561                                 ext2fs_set_feature_inline_data(sb);
1562                                 ext2fs_mark_super_dirty(fs);
1563                                 inlinedata_fs = 1;
1564                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1565                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1566                                 /* skip FINISH_INODE_LOOP */
1567                                 continue;
1568                         }
1569                 }
1570
1571                 /* Test for inline data flag but no attr */
1572                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1573                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1574                         size_t size = 0;
1575                         errcode_t err;
1576                         int flags;
1577
1578                         flags = fs->flags;
1579                         if (failed_csum)
1580                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1581                         err = get_inline_data_ea_size(fs, ino, &size);
1582                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1583                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1584
1585                         switch (err) {
1586                         case 0:
1587                                 /* Everything is awesome... */
1588                                 break;
1589                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1590                         case EXT2_ET_BAD_EA_HASH:
1591                         case EXT2_ET_BAD_EA_HEADER:
1592                         case EXT2_ET_EA_BAD_NAME_LEN:
1593                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1594                         case EXT2_ET_EA_KEY_NOT_FOUND:
1595                         case EXT2_ET_EA_NO_SPACE:
1596                         case EXT2_ET_MISSING_EA_FEATURE:
1597                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1598                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1599                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1600                         case EXT2_ET_NO_INLINE_DATA:
1601                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1602                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1603                         case EXT2_ET_EA_INODE_CORRUPTED:
1604                                 /* broken EA or no system.data EA; truncate */
1605                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1606                                                 &pctx)) {
1607                                         err = ext2fs_inode_size_set(fs, inode, 0);
1608                                         if (err) {
1609                                                 pctx.errcode = err;
1610                                                 ctx->flags |= E2F_FLAG_ABORT;
1611                                                 goto endit;
1612                                         }
1613                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1614                                         memset(&inode->i_block, 0,
1615                                                sizeof(inode->i_block));
1616                                         e2fsck_write_inode(ctx, ino, inode,
1617                                                            "pass1");
1618                                         failed_csum = 0;
1619                                 }
1620                                 break;
1621                         default:
1622                                 /* Some other kind of non-xattr error? */
1623                                 pctx.errcode = err;
1624                                 ctx->flags |= E2F_FLAG_ABORT;
1625                                 goto endit;
1626                         }
1627                 }
1628
1629                 /*
1630                  * Test for incorrect extent flag settings.
1631                  *
1632                  * On big-endian machines we must be careful:
1633                  * When the inode is read, the i_block array is not swapped
1634                  * if the extent flag is set.  Therefore if we are testing
1635                  * for or fixing a wrongly-set flag, we must potentially
1636                  * (un)swap before testing, or after fixing.
1637                  */
1638
1639                 /*
1640                  * In this case the extents flag was set when read, so
1641                  * extent_header_verify is ok.  If the inode is cleared,
1642                  * no need to swap... so no extra swapping here.
1643                  */
1644                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1645                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1646                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1647                         if ((ext2fs_extent_header_verify(inode->i_block,
1648                                                  sizeof(inode->i_block)) == 0) &&
1649                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1650                                 ext2fs_set_feature_extents(sb);
1651                                 ext2fs_mark_super_dirty(fs);
1652                                 extent_fs = 1;
1653                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1654                         clear_inode:
1655                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1656                                 if (ino == EXT2_BAD_INO)
1657                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1658                                                                  ino);
1659                                 /* skip FINISH_INODE_LOOP */
1660                                 continue;
1661                         }
1662                 }
1663
1664                 /*
1665                  * For big-endian machines:
1666                  * If the inode didn't have the extents flag set when it
1667                  * was read, then the i_blocks array was swapped.  To test
1668                  * as an extents header, we must swap it back first.
1669                  * IF we then set the extents flag, the entire i_block
1670                  * array must be un/re-swapped to make it proper extents data.
1671                  */
1672                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1673                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1674                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1675                     (LINUX_S_ISREG(inode->i_mode) ||
1676                      LINUX_S_ISDIR(inode->i_mode))) {
1677                         void *ehp;
1678 #ifdef WORDS_BIGENDIAN
1679                         __u32 tmp_block[EXT2_N_BLOCKS];
1680
1681                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1682                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1683                         ehp = tmp_block;
1684 #else
1685                         ehp = inode->i_block;
1686 #endif
1687                         if ((ext2fs_extent_header_verify(ehp,
1688                                          sizeof(inode->i_block)) == 0) &&
1689                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1690                                 inode->i_flags |= EXT4_EXTENTS_FL;
1691 #ifdef WORDS_BIGENDIAN
1692                                 memcpy(inode->i_block, tmp_block,
1693                                        sizeof(inode->i_block));
1694 #endif
1695                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1696                                 failed_csum = 0;
1697                         }
1698                 }
1699
1700                 if (ino == EXT2_BAD_INO) {
1701                         struct process_block_struct pb;
1702
1703                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1704                              inode->i_gid || inode->i_links_count ||
1705                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1706                              inode->i_file_acl) &&
1707                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1708                                 memset(inode, 0, sizeof(struct ext2_inode));
1709                                 e2fsck_write_inode(ctx, ino, inode,
1710                                                    "clear bad inode");
1711                                 failed_csum = 0;
1712                         }
1713
1714                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1715                                                           &pb.fs_meta_blocks);
1716                         if (pctx.errcode) {
1717                                 pctx.num = 4;
1718                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1719                                 ctx->flags |= E2F_FLAG_ABORT;
1720                                 goto endit;
1721                         }
1722                         pb.ino = EXT2_BAD_INO;
1723                         pb.num_blocks = pb.last_block = 0;
1724                         pb.last_db_block = -1;
1725                         pb.num_illegal_blocks = 0;
1726                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1727                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1728                         pb.inode = inode;
1729                         pb.pctx = &pctx;
1730                         pb.ctx = ctx;
1731                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1732                                      block_buf, process_bad_block, &pb);
1733                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1734                         if (pctx.errcode) {
1735                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1736                                 ctx->flags |= E2F_FLAG_ABORT;
1737                                 goto endit;
1738                         }
1739                         if (pb.bbcheck)
1740                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1741                                 ctx->flags |= E2F_FLAG_ABORT;
1742                                 goto endit;
1743                         }
1744                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1745                         clear_problem_context(&pctx);
1746                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1747                         continue;
1748                 } else if (ino == EXT2_ROOT_INO) {
1749                         /*
1750                          * Make sure the root inode is a directory; if
1751                          * not, offer to clear it.  It will be
1752                          * regenerated in pass #3.
1753                          */
1754                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1755                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1756                                         goto clear_inode;
1757                         }
1758                         /*
1759                          * If dtime is set, offer to clear it.  mke2fs
1760                          * version 0.2b created filesystems with the
1761                          * dtime field set for the root and lost+found
1762                          * directories.  We won't worry about
1763                          * /lost+found, since that can be regenerated
1764                          * easily.  But we will fix the root directory
1765                          * as a special case.
1766                          */
1767                         if (inode->i_dtime && inode->i_links_count) {
1768                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1769                                         inode->i_dtime = 0;
1770                                         e2fsck_write_inode(ctx, ino, inode,
1771                                                            "pass1");
1772                                         failed_csum = 0;
1773                                 }
1774                         }
1775                 } else if (ino == EXT2_JOURNAL_INO) {
1776                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1777                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1778                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1779                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1780                                                 &pctx)) {
1781                                         inode->i_mode = LINUX_S_IFREG;
1782                                         e2fsck_write_inode(ctx, ino, inode,
1783                                                            "pass1");
1784                                         failed_csum = 0;
1785                                 }
1786                                 check_blocks(ctx, &pctx, block_buf, NULL);
1787                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1788                                 continue;
1789                         }
1790                         if ((inode->i_links_count ||
1791                              inode->i_blocks || inode->i_block[0]) &&
1792                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1793                                         &pctx)) {
1794                                 memset(inode, 0, inode_size);
1795                                 ext2fs_icount_store(ctx->inode_link_info,
1796                                                     ino, 0);
1797                                 e2fsck_write_inode_full(ctx, ino, inode,
1798                                                         inode_size, "pass1");
1799                                 failed_csum = 0;
1800                         }
1801                 } else if (quota_inum_is_reserved(fs, ino)) {
1802                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1803                         if (ext2fs_has_feature_quota(fs->super) &&
1804                             quota_inum_is_super(fs->super, ino)) {
1805                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1806                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1807                                                         &pctx)) {
1808                                         inode->i_mode = LINUX_S_IFREG;
1809                                         e2fsck_write_inode(ctx, ino, inode,
1810                                                         "pass1");
1811                                         failed_csum = 0;
1812                                 }
1813                                 check_blocks(ctx, &pctx, block_buf, NULL);
1814                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1815                                 continue;
1816                         }
1817                         if ((inode->i_links_count ||
1818                              inode->i_blocks || inode->i_block[0]) &&
1819                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1820                                         &pctx)) {
1821                                 memset(inode, 0, inode_size);
1822                                 ext2fs_icount_store(ctx->inode_link_info,
1823                                                     ino, 0);
1824                                 e2fsck_write_inode_full(ctx, ino, inode,
1825                                                         inode_size, "pass1");
1826                                 failed_csum = 0;
1827                         }
1828                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1829                         problem_t problem = 0;
1830
1831                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1832                         if (ino == EXT2_BOOT_LOADER_INO) {
1833                                 if (LINUX_S_ISDIR(inode->i_mode))
1834                                         problem = PR_1_RESERVED_BAD_MODE;
1835                         } else if (ino == EXT2_RESIZE_INO) {
1836                                 if (inode->i_mode &&
1837                                     !LINUX_S_ISREG(inode->i_mode))
1838                                         problem = PR_1_RESERVED_BAD_MODE;
1839                         } else {
1840                                 if (inode->i_mode != 0)
1841                                         problem = PR_1_RESERVED_BAD_MODE;
1842                         }
1843                         if (problem) {
1844                                 if (fix_problem(ctx, problem, &pctx)) {
1845                                         inode->i_mode = 0;
1846                                         e2fsck_write_inode(ctx, ino, inode,
1847                                                            "pass1");
1848                                         failed_csum = 0;
1849                                 }
1850                         }
1851                         check_blocks(ctx, &pctx, block_buf, NULL);
1852                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1853                         continue;
1854                 }
1855
1856                 if (!inode->i_links_count) {
1857                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1858                         continue;
1859                 }
1860                 /*
1861                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1862                  * deleted files.  Oops.
1863                  *
1864                  * Since all new ext2 implementations get this right,
1865                  * we now assume that the case of non-zero
1866                  * i_links_count and non-zero dtime means that we
1867                  * should keep the file, not delete it.
1868                  *
1869                  */
1870                 if (inode->i_dtime) {
1871                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1872                                 inode->i_dtime = 0;
1873                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1874                                 failed_csum = 0;
1875                         }
1876                 }
1877
1878                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1879                 switch (fs->super->s_creator_os) {
1880                     case EXT2_OS_HURD:
1881                         frag = inode->osd2.hurd2.h_i_frag;
1882                         fsize = inode->osd2.hurd2.h_i_fsize;
1883                         break;
1884                     default:
1885                         frag = fsize = 0;
1886                 }
1887
1888                 if (inode->i_faddr || frag || fsize ||
1889                     (!ext2fs_has_feature_largedir(fs->super) &&
1890                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1891                         mark_inode_bad(ctx, ino);
1892                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1893                     !ext2fs_has_feature_64bit(fs->super) &&
1894                     inode->osd2.linux2.l_i_file_acl_high != 0)
1895                         mark_inode_bad(ctx, ino);
1896                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1897                     !ext2fs_has_feature_huge_file(fs->super) &&
1898                     (inode->osd2.linux2.l_i_blocks_hi != 0))
1899                         mark_inode_bad(ctx, ino);
1900                 if (inode->i_flags & EXT2_IMAGIC_FL) {
1901                         if (imagic_fs) {
1902                                 if (!ctx->inode_imagic_map)
1903                                         alloc_imagic_map(ctx);
1904                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1905                                                          ino);
1906                         } else {
1907                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1908                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
1909                                         e2fsck_write_inode(ctx, ino,
1910                                                            inode, "pass1");
1911                                         failed_csum = 0;
1912                                 }
1913                         }
1914                 }
1915
1916                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1917                 check_is_really_dir(ctx, &pctx, block_buf);
1918
1919                 /*
1920                  * ext2fs_inode_has_valid_blocks2 does not actually look
1921                  * at i_block[] values, so not endian-sensitive here.
1922                  */
1923                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1924                     LINUX_S_ISLNK(inode->i_mode) &&
1925                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1926                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1927                         inode->i_flags &= ~EXT4_EXTENTS_FL;
1928                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1929                         failed_csum = 0;
1930                 }
1931
1932                 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
1933                     add_encrypted_file(ctx, &pctx) < 0)
1934                         goto clear_inode;
1935
1936                 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
1937                         ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
1938
1939                 if (LINUX_S_ISDIR(inode->i_mode)) {
1940                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1941                         e2fsck_add_dir_info(ctx, ino, 0);
1942                         ctx->fs_directory_count++;
1943                         if (inode->i_flags & EXT4_CASEFOLD_FL)
1944                                 add_casefolded_dir(ctx, ino);
1945                 } else if (LINUX_S_ISREG (inode->i_mode)) {
1946                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1947                         ctx->fs_regular_count++;
1948                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1949                            e2fsck_pass1_check_device_inode(fs, inode)) {
1950                         check_extents_inlinedata(ctx, &pctx);
1951                         check_immutable(ctx, &pctx);
1952                         check_size(ctx, &pctx);
1953                         ctx->fs_chardev_count++;
1954                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1955                            e2fsck_pass1_check_device_inode(fs, inode)) {
1956                         check_extents_inlinedata(ctx, &pctx);
1957                         check_immutable(ctx, &pctx);
1958                         check_size(ctx, &pctx);
1959                         ctx->fs_blockdev_count++;
1960                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1961                            e2fsck_pass1_check_symlink(fs, ino, inode,
1962                                                       block_buf)) {
1963                         check_immutable(ctx, &pctx);
1964                         ctx->fs_symlinks_count++;
1965                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1966                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1967                                 continue;
1968                         } else if (ext2fs_is_fast_symlink(inode)) {
1969                                 ctx->fs_fast_symlinks_count++;
1970                                 check_blocks(ctx, &pctx, block_buf,
1971                                              &ea_ibody_quota);
1972                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1973                                 continue;
1974                         }
1975                 }
1976                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
1977                          e2fsck_pass1_check_device_inode(fs, inode)) {
1978                         check_extents_inlinedata(ctx, &pctx);
1979                         check_immutable(ctx, &pctx);
1980                         check_size(ctx, &pctx);
1981                         ctx->fs_fifo_count++;
1982                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
1983                            e2fsck_pass1_check_device_inode(fs, inode)) {
1984                         check_extents_inlinedata(ctx, &pctx);
1985                         check_immutable(ctx, &pctx);
1986                         check_size(ctx, &pctx);
1987                         ctx->fs_sockets_count++;
1988                 } else
1989                         mark_inode_bad(ctx, ino);
1990                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1991                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
1992                         if (inode->i_block[EXT2_IND_BLOCK])
1993                                 ctx->fs_ind_count++;
1994                         if (inode->i_block[EXT2_DIND_BLOCK])
1995                                 ctx->fs_dind_count++;
1996                         if (inode->i_block[EXT2_TIND_BLOCK])
1997                                 ctx->fs_tind_count++;
1998                 }
1999                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2000                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
2001                     (inode->i_block[EXT2_IND_BLOCK] ||
2002                      inode->i_block[EXT2_DIND_BLOCK] ||
2003                      inode->i_block[EXT2_TIND_BLOCK] ||
2004                      ext2fs_file_acl_block(fs, inode))) {
2005                         struct process_inode_block *itp;
2006
2007                         itp = &inodes_to_process[process_inode_count];
2008                         itp->ino = ino;
2009                         itp->ea_ibody_quota = ea_ibody_quota;
2010                         if (inode_size < sizeof(struct ext2_inode_large))
2011                                 memcpy(&itp->inode, inode, inode_size);
2012                         else
2013                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
2014                         process_inode_count++;
2015                 } else
2016                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
2017
2018                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2019
2020                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2021                         goto endit;
2022
2023                 if (process_inode_count >= ctx->process_inode_size) {
2024                         process_inodes(ctx, block_buf);
2025
2026                         if (e2fsck_should_abort(ctx))
2027                                 goto endit;
2028                 }
2029         }
2030         process_inodes(ctx, block_buf);
2031         ext2fs_close_inode_scan(scan);
2032         scan = NULL;
2033
2034         reserve_block_for_root_repair(ctx);
2035         reserve_block_for_lnf_repair(ctx);
2036
2037         /*
2038          * If any extended attribute blocks' reference counts need to
2039          * be adjusted, either up (ctx->refcount_extra), or down
2040          * (ctx->refcount), then fix them.
2041          */
2042         if (ctx->refcount) {
2043                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
2044                 ea_refcount_free(ctx->refcount);
2045                 ctx->refcount = 0;
2046         }
2047         if (ctx->refcount_extra) {
2048                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
2049                                         block_buf, +1);
2050                 ea_refcount_free(ctx->refcount_extra);
2051                 ctx->refcount_extra = 0;
2052         }
2053
2054         if (ctx->ea_block_quota_blocks) {
2055                 ea_refcount_free(ctx->ea_block_quota_blocks);
2056                 ctx->ea_block_quota_blocks = 0;
2057         }
2058
2059         if (ctx->ea_block_quota_inodes) {
2060                 ea_refcount_free(ctx->ea_block_quota_inodes);
2061                 ctx->ea_block_quota_inodes = 0;
2062         }
2063
2064         if (ctx->invalid_bitmaps)
2065                 handle_fs_bad_blocks(ctx);
2066
2067         /* We don't need the block_ea_map any more */
2068         if (ctx->block_ea_map) {
2069                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2070                 ctx->block_ea_map = 0;
2071         }
2072
2073         /* We don't need the encryption policy => ID map any more */
2074         destroy_encryption_policy_map(ctx);
2075
2076         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2077                 clear_problem_context(&pctx);
2078                 pctx.errcode = ext2fs_create_resize_inode(fs);
2079                 if (pctx.errcode) {
2080                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2081                                          &pctx)) {
2082                                 ctx->flags |= E2F_FLAG_ABORT;
2083                                 goto endit;
2084                         }
2085                         pctx.errcode = 0;
2086                 }
2087                 if (!pctx.errcode) {
2088                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2089                                           "recreate inode");
2090                         inode->i_mtime = ctx->now;
2091                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2092                                            "recreate inode");
2093                 }
2094                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2095         }
2096
2097         if (ctx->flags & E2F_FLAG_RESTART) {
2098                 /*
2099                  * Only the master copy of the superblock and block
2100                  * group descriptors are going to be written during a
2101                  * restart, so set the superblock to be used to be the
2102                  * master superblock.
2103                  */
2104                 ctx->use_superblock = 0;
2105                 goto endit;
2106         }
2107
2108         if (ctx->large_dirs && !ext2fs_has_feature_largedir(fs->super)) {
2109                 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2110                         ext2fs_set_feature_largedir(fs->super);
2111                         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2112                         ext2fs_mark_super_dirty(fs);
2113                 }
2114                 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2115                     fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2116                         ext2fs_update_dynamic_rev(fs);
2117                         ext2fs_mark_super_dirty(fs);
2118                 }
2119         }
2120
2121         if (ctx->block_dup_map) {
2122                 if (ctx->options & E2F_OPT_PREEN) {
2123                         clear_problem_context(&pctx);
2124                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2125                 }
2126                 e2fsck_pass1_dupblocks(ctx, block_buf);
2127         }
2128         ctx->flags |= E2F_FLAG_ALLOC_OK;
2129 endit:
2130         e2fsck_use_inode_shortcuts(ctx, 0);
2131         ext2fs_free_mem(&inodes_to_process);
2132         inodes_to_process = 0;
2133
2134         if (scan)
2135                 ext2fs_close_inode_scan(scan);
2136         if (block_buf)
2137                 ext2fs_free_mem(&block_buf);
2138         if (inode)
2139                 ext2fs_free_mem(&inode);
2140
2141         /*
2142          * The l+f inode may have been cleared, so zap it now and
2143          * later passes will recalculate it if necessary
2144          */
2145         ctx->lost_and_found = 0;
2146
2147         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2148                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2149         else
2150                 ctx->invalid_bitmaps++;
2151 }
2152
2153 #ifdef HAVE_PTHREAD
2154 static errcode_t e2fsck_pass1_copy_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2155                                           ext2fs_generic_bitmap *dest)
2156 {
2157         errcode_t ret;
2158
2159         ret = ext2fs_copy_bitmap(*src, dest);
2160         if (ret)
2161                 return ret;
2162
2163         (*dest)->fs = fs;
2164
2165         return 0;
2166 }
2167
2168 static void e2fsck_pass1_free_bitmap(ext2fs_generic_bitmap *bitmap)
2169 {
2170         if (*bitmap) {
2171                 ext2fs_free_generic_bmap(*bitmap);
2172                 *bitmap = NULL;
2173         }
2174
2175 }
2176
2177 static errcode_t e2fsck_pass1_merge_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2178                                           ext2fs_generic_bitmap *dest)
2179 {
2180         errcode_t ret = 0;
2181
2182         if (*src) {
2183                 if (*dest == NULL) {
2184                         *dest = *src;
2185                         *src = NULL;
2186                 } else {
2187                         ret = ext2fs_merge_bitmap(*src, *dest, NULL, NULL);
2188                         if (ret)
2189                                 return ret;
2190                 }
2191                 (*dest)->fs = fs;
2192         }
2193
2194         return 0;
2195 }
2196
2197 static errcode_t e2fsck_pass1_copy_fs(ext2_filsys dest, e2fsck_t src_context,
2198                                       ext2_filsys src)
2199 {
2200         errcode_t       retval;
2201
2202         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2203         dest->inode_map = NULL;
2204         dest->block_map = NULL;
2205         if (dest->dblist)
2206                 dest->dblist->fs = dest;
2207         if (src->block_map) {
2208                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2209                                                   &dest->block_map);
2210                 if (retval)
2211                         return retval;
2212         }
2213         if (src->inode_map) {
2214                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2215                                                   &dest->inode_map);
2216                 if (retval)
2217                         return retval;
2218         }
2219
2220         if (src->badblocks) {
2221                 retval = ext2fs_badblocks_copy(src->badblocks, &dest->badblocks);
2222                 if (retval)
2223                         return retval;
2224         }
2225
2226         /* disable it for now */
2227         src_context->openfs_flags &= ~EXT2_FLAG_EXCLUSIVE;
2228         retval = ext2fs_open_channel(dest, src_context->io_options,
2229                                      src_context->io_manager,
2230                                      src_context->openfs_flags,
2231                                      src->io->block_size);
2232         if (retval)
2233                 return retval;
2234
2235         /* Block size might not be default */
2236         io_channel_set_blksize(dest->io, src->io->block_size);
2237         ehandler_init(dest->io);
2238
2239         assert(dest->io->magic == src->io->magic);
2240         assert(dest->io->manager == src->io->manager);
2241         assert(strcmp(dest->io->name, src->io->name) == 0);
2242         assert(dest->io->block_size == src->io->block_size);
2243         assert(dest->io->read_error == src->io->read_error);
2244         assert(dest->io->write_error == src->io->write_error);
2245         assert(dest->io->refcount == src->io->refcount);
2246         assert(dest->io->flags == src->io->flags);
2247         assert(dest->io->app_data == dest);
2248         assert(src->io->app_data == src);
2249         assert(dest->io->align == src->io->align);
2250
2251         /* The data should be written to disk immediately */
2252         dest->io->flags |= CHANNEL_FLAGS_WRITETHROUGH;
2253         /* icache will be rebuilt if needed, so do not copy from @src */
2254         src->icache = NULL;
2255         return 0;
2256 }
2257
2258 static int e2fsck_pass1_merge_fs(ext2_filsys dest, ext2_filsys src)
2259 {
2260         struct ext2_inode_cache *icache = dest->icache;
2261         errcode_t retval = 0;
2262         io_channel dest_io;
2263         io_channel dest_image_io;
2264         ext2fs_inode_bitmap inode_map;
2265         ext2fs_block_bitmap block_map;
2266
2267         dest_io = dest->io;
2268         dest_image_io = dest->image_io;
2269         inode_map = dest->inode_map;
2270         block_map = dest->block_map;
2271
2272         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2273         dest->io = dest_io;
2274         dest->image_io = dest_image_io;
2275         dest->icache = icache;
2276         dest->inode_map = inode_map;
2277         dest->block_map = block_map;
2278         if (dest->dblist)
2279                 dest->dblist->fs = dest;
2280
2281         if (src->icache) {
2282                 ext2fs_free_inode_cache(src->icache);
2283                 src->icache = NULL;
2284         }
2285
2286         retval = e2fsck_pass1_merge_bitmap(dest, &src->inode_map,
2287                                            &dest->inode_map);
2288         if (retval)
2289                 goto out;
2290
2291         retval = e2fsck_pass1_merge_bitmap(dest, &src->block_map,
2292                                           &dest->block_map);
2293         if (retval)
2294                 goto out;
2295
2296         if (src->badblocks) {
2297                 retval = ext2fs_badblocks_copy(src->badblocks, &dest->badblocks);
2298
2299                 ext2fs_badblocks_list_free(src->badblocks);
2300                 src->badblocks = NULL;
2301         }
2302 out:
2303         io_channel_close(src->io);
2304         if (src->inode_map)
2305                 ext2fs_free_generic_bmap(src->inode_map);
2306         if (src->block_map)
2307                 ext2fs_free_generic_bmap(src->block_map);
2308         return retval;
2309 }
2310
2311 static errcode_t e2fsck_pass1_thread_prepare(e2fsck_t global_ctx, e2fsck_t *thread_ctx,
2312                                              int thread_index, int num_threads)
2313 {
2314         errcode_t               retval;
2315         e2fsck_t                thread_context;
2316         ext2_filsys             thread_fs;
2317         ext2_filsys             global_fs = global_ctx->fs;
2318         struct e2fsck_thread    *tinfo;
2319         dgrp_t                  average_group;
2320
2321         assert(global_ctx->inode_used_map == NULL);
2322         assert(global_ctx->inode_dir_map == NULL);
2323         assert(global_ctx->inode_bb_map == NULL);
2324         assert(global_ctx->inode_imagic_map == NULL);
2325         assert(global_ctx->inode_reg_map == NULL);
2326         assert(global_ctx->inodes_to_rebuild == NULL);
2327
2328         assert(global_ctx->block_found_map == NULL);
2329         assert(global_ctx->block_dup_map == NULL);
2330         assert(global_ctx->block_ea_map == NULL);
2331         assert(global_ctx->block_metadata_map == NULL);
2332         assert(global_ctx->fs->dblist == NULL);
2333
2334         retval = ext2fs_get_mem(sizeof(struct e2fsck_struct), &thread_context);
2335         if (retval) {
2336                 com_err(global_ctx->program_name, retval, "while allocating memory");
2337                 return retval;
2338         }
2339         memcpy(thread_context, global_ctx, sizeof(struct e2fsck_struct));
2340         thread_context->global_ctx = global_ctx;
2341
2342         retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &thread_fs);
2343         if (retval) {
2344                 com_err(global_ctx->program_name, retval, "while allocating memory");
2345                 goto out_context;
2346         }
2347
2348         io_channel_flush_cleanup(global_fs->io);
2349         retval = e2fsck_pass1_copy_fs(thread_fs, global_ctx, global_fs);
2350         if (retval) {
2351                 com_err(global_ctx->program_name, retval, "while copying fs");
2352                 goto out_fs;
2353         }
2354         thread_fs->priv_data = thread_context;
2355
2356         thread_context->thread_info.et_thread_index = thread_index;
2357         set_up_logging(thread_context);
2358
2359         /*
2360          * Distribute work to multiple threads:
2361          * Each thread work on fs->group_desc_count / nthread groups.
2362          */
2363         tinfo = &thread_context->thread_info;
2364         average_group = thread_fs->group_desc_count / num_threads;
2365         if (average_group == 0)
2366                 average_group = 1;
2367         tinfo->et_group_start = average_group * thread_index;
2368         if (thread_index == num_threads - 1)
2369                 tinfo->et_group_end = thread_fs->group_desc_count;
2370         else
2371                 tinfo->et_group_end = average_group * (thread_index + 1);
2372         tinfo->et_group_next = tinfo->et_group_start;
2373         tinfo->et_inode_number = 0;
2374         tinfo->et_log_buf[0] = '\0';
2375         tinfo->et_log_length = 0;
2376         if (thread_context->options & E2F_OPT_MULTITHREAD)
2377                 log_out(thread_context, _("Scan group range [%d, %d)\n"),
2378                         tinfo->et_group_start, tinfo->et_group_end);
2379         thread_context->fs = thread_fs;
2380         *thread_ctx = thread_context;
2381         return 0;
2382 out_fs:
2383         ext2fs_free_mem(&thread_fs);
2384 out_context:
2385         ext2fs_free_mem(&thread_context);
2386         return retval;
2387 }
2388
2389 static int e2fsck_pass1_thread_join_one(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2390 {
2391         errcode_t        retval;
2392         int              flags = global_ctx->flags;
2393         ext2_filsys      thread_fs = thread_ctx->fs;
2394         ext2_filsys      global_fs = global_ctx->fs;
2395         FILE            *global_logf = global_ctx->logf;
2396         FILE            *global_problem_logf = global_ctx->problem_logf;
2397         ext2fs_inode_bitmap inode_bad_map = global_ctx->inode_bad_map;
2398         ext2fs_inode_bitmap inode_used_map = global_ctx->inode_used_map;
2399         ext2fs_inode_bitmap inode_dir_map = global_ctx->inode_dir_map;
2400         ext2fs_inode_bitmap inode_bb_map = global_ctx->inode_bb_map;
2401         ext2fs_inode_bitmap inode_imagic_map = global_ctx->inode_imagic_map;
2402         ext2fs_inode_bitmap inode_reg_map = global_ctx->inode_reg_map;
2403         ext2fs_block_bitmap block_found_map = global_ctx->block_found_map;
2404         ext2fs_block_bitmap block_dup_map = global_ctx->block_dup_map;
2405         ext2fs_block_bitmap block_ea_map = global_ctx->block_ea_map;
2406         ext2fs_block_bitmap block_metadata_map = global_ctx->block_metadata_map;
2407         ext2fs_block_bitmap inodes_to_rebuild = global_ctx->inodes_to_rebuild;
2408
2409 #ifdef HAVE_SETJMP_H
2410         jmp_buf          old_jmp;
2411
2412         memcpy(old_jmp, global_ctx->abort_loc, sizeof(jmp_buf));
2413 #endif
2414         memcpy(global_ctx, thread_ctx, sizeof(struct e2fsck_struct));
2415 #ifdef HAVE_SETJMP_H
2416         memcpy(global_ctx->abort_loc, old_jmp, sizeof(jmp_buf));
2417 #endif
2418
2419         global_ctx->inode_used_map = inode_used_map;
2420         global_ctx->inode_bad_map = inode_bad_map;
2421         global_ctx->inode_dir_map = inode_dir_map;
2422         global_ctx->inode_bb_map = inode_bb_map;
2423         global_ctx->inode_imagic_map = inode_imagic_map;
2424         global_ctx->inodes_to_rebuild = inodes_to_rebuild;
2425         global_ctx->inode_reg_map = inode_reg_map;
2426         global_ctx->block_found_map = block_found_map;
2427         global_ctx->block_dup_map = block_dup_map;
2428         global_ctx->block_ea_map = block_ea_map;
2429         global_ctx->block_metadata_map = block_metadata_map;
2430
2431         /* Keep the global singal flags*/
2432         global_ctx->flags |= (flags & E2F_FLAG_SIGNAL_MASK) |
2433                              (global_ctx->flags & E2F_FLAG_SIGNAL_MASK);
2434
2435         retval = e2fsck_pass1_merge_fs(global_fs, thread_fs);
2436         if (retval) {
2437                 com_err(global_ctx->program_name, 0, _("while merging fs\n"));
2438                 return retval;
2439         }
2440         global_fs->priv_data = global_ctx;
2441         global_ctx->fs = global_fs;
2442         global_ctx->logf = global_logf;
2443         global_ctx->problem_logf = global_problem_logf;
2444         global_ctx->global_ctx = NULL;
2445
2446         retval = e2fsck_pass1_merge_bitmap(global_fs,
2447                                 &thread_ctx->inode_used_map,
2448                                 &global_ctx->inode_used_map);
2449         if (retval)
2450                 return retval;
2451
2452         retval = e2fsck_pass1_merge_bitmap(global_fs,
2453                                 &thread_ctx->inode_bad_map,
2454                                 &global_ctx->inode_bad_map);
2455         if (retval)
2456                 return retval;
2457         retval = e2fsck_pass1_merge_bitmap(global_fs,
2458                                         &thread_ctx->inode_dir_map,
2459                                         &global_ctx->inode_dir_map);
2460         if (retval)
2461                 return retval;
2462         retval = e2fsck_pass1_merge_bitmap(global_fs,
2463                                 &thread_ctx->inode_bb_map,
2464                                 &global_ctx->inode_bb_map);
2465         if (retval)
2466                 return retval;
2467         retval = e2fsck_pass1_merge_bitmap(global_fs,
2468                                 &thread_ctx->inode_imagic_map,
2469                                 &global_ctx->inode_imagic_map);
2470         if (retval)
2471                 return retval;
2472         retval = e2fsck_pass1_merge_bitmap(global_fs,
2473                                 &thread_ctx->inode_reg_map,
2474                                 &global_ctx->inode_reg_map);
2475         if (retval)
2476                 return retval;
2477         retval = e2fsck_pass1_merge_bitmap(global_fs,
2478                                 &thread_ctx->inodes_to_rebuild,
2479                                 &global_ctx->inodes_to_rebuild);
2480         if (retval)
2481                 return retval;
2482         retval = e2fsck_pass1_merge_bitmap(global_fs,
2483                                 &thread_ctx->block_found_map,
2484                                 &global_ctx->block_found_map);
2485         if (retval)
2486                 return retval;
2487         retval = e2fsck_pass1_merge_bitmap(global_fs,
2488                                 &thread_ctx->block_dup_map,
2489                                 &global_ctx->block_dup_map);
2490         if (retval)
2491                 return retval;
2492         retval = e2fsck_pass1_merge_bitmap(global_fs,
2493                                 &thread_ctx->block_ea_map,
2494                                 &global_ctx->block_ea_map);
2495         if (retval)
2496                 return retval;
2497         retval = e2fsck_pass1_merge_bitmap(global_fs,
2498                                 &thread_ctx->block_metadata_map,
2499                                 &global_ctx->block_metadata_map);
2500         if (retval)
2501                 return retval;
2502
2503         return 0;
2504 }
2505
2506 static int e2fsck_pass1_thread_join(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2507 {
2508         errcode_t       retval;
2509
2510         retval = e2fsck_pass1_thread_join_one(global_ctx, thread_ctx);
2511         ext2fs_free_mem(&thread_ctx->fs);
2512         if (thread_ctx->logf)
2513                 fclose(thread_ctx->logf);
2514         if (thread_ctx->problem_logf) {
2515                 fputs("</problem_log>\n", thread_ctx->problem_logf);
2516                 fclose(thread_ctx->problem_logf);
2517         }
2518         e2fsck_pass1_free_bitmap(&thread_ctx->inode_used_map);
2519         e2fsck_pass1_free_bitmap(&thread_ctx->inode_bad_map);
2520         e2fsck_pass1_free_bitmap(&thread_ctx->inode_dir_map);
2521         e2fsck_pass1_free_bitmap(&thread_ctx->inode_bb_map);
2522         e2fsck_pass1_free_bitmap(&thread_ctx->inode_imagic_map);
2523         e2fsck_pass1_free_bitmap(&thread_ctx->inode_reg_map);
2524         e2fsck_pass1_free_bitmap(&thread_ctx->inodes_to_rebuild);
2525         e2fsck_pass1_free_bitmap(&thread_ctx->block_found_map);
2526         e2fsck_pass1_free_bitmap(&thread_ctx->block_dup_map);
2527         e2fsck_pass1_free_bitmap(&thread_ctx->block_ea_map);
2528         e2fsck_pass1_free_bitmap(&thread_ctx->block_metadata_map);
2529         ext2fs_free_mem(&thread_ctx);
2530
2531         return retval;
2532 }
2533
2534 static int e2fsck_pass1_threads_join(struct e2fsck_thread_info *infos,
2535                                       int num_threads, e2fsck_t global_ctx)
2536 {
2537         errcode_t                        rc;
2538         errcode_t                        ret = 0;
2539         int                              i;
2540         struct e2fsck_thread_info       *pinfo;
2541
2542         for (i = 0; i < num_threads; i++) {
2543                 pinfo = &infos[i];
2544
2545                 if (!pinfo->eti_started)
2546                         continue;
2547
2548                 rc = pthread_join(pinfo->eti_thread_id, NULL);
2549                 if (rc) {
2550                         com_err(global_ctx->program_name, rc,
2551                                 _("while joining thread\n"));
2552                         if (ret == 0)
2553                                 ret = rc;
2554                 }
2555                 rc = e2fsck_pass1_thread_join(global_ctx, infos[i].eti_thread_ctx);
2556                 if (rc) {
2557                         com_err(global_ctx->program_name, rc,
2558                                 _("while joining pass1 thread\n"));
2559                         if (ret == 0)
2560                                 ret = rc;
2561                 }
2562         }
2563         free(infos);
2564
2565         return ret;
2566 }
2567
2568 static void *e2fsck_pass1_thread(void *arg)
2569 {
2570         struct e2fsck_thread_info       *info = arg;
2571         e2fsck_t                         thread_ctx = info->eti_thread_ctx;
2572
2573 #ifdef HAVE_SETJMP_H
2574         /*
2575          * When fatal_error() happens, jump to here. The thread
2576          * context's flags will be saved, but its abort_loc will
2577          * be overwritten by original jump buffer for the later
2578          * tests.
2579          */
2580         if (setjmp(thread_ctx->abort_loc)) {
2581                 thread_ctx->flags &= ~E2F_FLAG_SETJMP_OK;
2582                 goto out;
2583         }
2584         thread_ctx->flags |= E2F_FLAG_SETJMP_OK;
2585 #endif
2586
2587         e2fsck_pass1_run(thread_ctx);
2588
2589 out:
2590         if (thread_ctx->options & E2F_OPT_MULTITHREAD)
2591                 log_out(thread_ctx,
2592                         _("Scanned group range [%lu, %lu), inodes %lu\n"),
2593                         thread_ctx->thread_info.et_group_start,
2594                         thread_ctx->thread_info.et_group_end,
2595                         thread_ctx->thread_info.et_inode_number);
2596         return NULL;
2597 }
2598
2599 static int e2fsck_pass1_threads_start(struct e2fsck_thread_info **pinfo,
2600                                       int num_threads, e2fsck_t global_ctx)
2601 {
2602         struct e2fsck_thread_info       *infos;
2603         pthread_attr_t                   attr;
2604         errcode_t                        retval;
2605         errcode_t                        ret;
2606         struct e2fsck_thread_info       *tmp_pinfo;
2607         int                              i;
2608         e2fsck_t                         thread_ctx;
2609
2610         retval = pthread_attr_init(&attr);
2611         if (retval) {
2612                 com_err(global_ctx->program_name, retval,
2613                         _("while setting pthread attribute\n"));
2614                 return retval;
2615         }
2616
2617         infos = calloc(num_threads, sizeof(struct e2fsck_thread_info));
2618         if (infos == NULL) {
2619                 retval = -ENOMEM;
2620                 com_err(global_ctx->program_name, retval,
2621                         _("while allocating memory for threads\n"));
2622                 pthread_attr_destroy(&attr);
2623                 return retval;
2624         }
2625
2626         for (i = 0; i < num_threads; i++) {
2627                 tmp_pinfo = &infos[i];
2628                 tmp_pinfo->eti_thread_index = i;
2629                 retval = e2fsck_pass1_thread_prepare(global_ctx, &thread_ctx,
2630                                                      i, num_threads);
2631                 if (retval) {
2632                         com_err(global_ctx->program_name, retval,
2633                                 _("while preparing pass1 thread\n"));
2634                         break;
2635                 }
2636                 tmp_pinfo->eti_thread_ctx = thread_ctx;
2637
2638                 retval = pthread_create(&tmp_pinfo->eti_thread_id, &attr,
2639                                         &e2fsck_pass1_thread, tmp_pinfo);
2640                 if (retval) {
2641                         com_err(global_ctx->program_name, retval,
2642                                 _("while creating thread\n"));
2643                         e2fsck_pass1_thread_join(global_ctx, thread_ctx);
2644                         break;
2645                 }
2646
2647                 tmp_pinfo->eti_started = 1;
2648         }
2649
2650         /* destroy the thread attribute object, since it is no longer needed */
2651         ret = pthread_attr_destroy(&attr);
2652         if (ret) {
2653                 com_err(global_ctx->program_name, ret,
2654                         _("while destroying thread attribute\n"));
2655                 if (retval == 0)
2656                         retval = ret;
2657         }
2658
2659         if (retval) {
2660                 e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
2661                 return retval;
2662         }
2663         *pinfo = infos;
2664         return 0;
2665 }
2666
2667 static void e2fsck_pass1_multithread(e2fsck_t global_ctx)
2668 {
2669         struct e2fsck_thread_info       *infos = NULL;
2670         int                              num_threads = 1;
2671         errcode_t                        retval;
2672
2673         retval = e2fsck_pass1_threads_start(&infos, num_threads, global_ctx);
2674         if (retval) {
2675                 com_err(global_ctx->program_name, retval,
2676                         _("while starting pass1 threads\n"));
2677                 goto out_abort;
2678         }
2679
2680         retval = e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
2681         if (retval) {
2682                 com_err(global_ctx->program_name, retval,
2683                         _("while joining pass1 threads\n"));
2684                 goto out_abort;
2685         }
2686         return;
2687 out_abort:
2688         global_ctx->flags |= E2F_FLAG_ABORT;
2689         return;
2690 }
2691 #endif
2692
2693 void e2fsck_pass1(e2fsck_t ctx)
2694 {
2695
2696 #ifdef HAVE_PTHREAD
2697         e2fsck_pass1_multithread(ctx);
2698 #else
2699         e2fsck_pass1_run(ctx);
2700 #endif
2701 }
2702
2703 #undef FINISH_INODE_LOOP
2704
2705 /*
2706  * When the inode_scan routines call this callback at the end of the
2707  * glock group, call process_inodes.
2708  */
2709 static errcode_t scan_callback(ext2_filsys fs,
2710                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
2711                                dgrp_t group, void * priv_data)
2712 {
2713         struct scan_callback_struct *scan_struct;
2714         e2fsck_t ctx;
2715         struct e2fsck_thread *tinfo;
2716
2717         scan_struct = (struct scan_callback_struct *) priv_data;
2718         ctx = scan_struct->ctx;
2719
2720         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf);
2721
2722         if (ctx->progress)
2723                 if ((ctx->progress)(ctx, 1, group+1,
2724                                     ctx->fs->group_desc_count))
2725                         return EXT2_ET_CANCEL_REQUESTED;
2726
2727 #ifdef HAVE_PTHREAD
2728         if (ctx->global_ctx) {
2729                 tinfo = &ctx->thread_info;
2730                 tinfo->et_group_next++;
2731                 if (ctx->options & E2F_OPT_DEBUG &&
2732                     ctx->options & E2F_OPT_MULTITHREAD)
2733                         log_out(ctx, _("group %d finished\n"),
2734                                 tinfo->et_group_next);
2735                 if (tinfo->et_group_next >= tinfo->et_group_end)
2736                         return EXT2_ET_SCAN_FINISHED;
2737         }
2738 #endif
2739
2740         return 0;
2741 }
2742
2743 /*
2744  * Process the inodes in the "inodes to process" list.
2745  */
2746 static void process_inodes(e2fsck_t ctx, char *block_buf)
2747 {
2748         int                     i;
2749         struct ext2_inode       *old_stashed_inode;
2750         ext2_ino_t              old_stashed_ino;
2751         const char              *old_operation;
2752         char                    buf[80];
2753         struct problem_context  pctx;
2754
2755 #if 0
2756         printf("begin process_inodes: ");
2757 #endif
2758         if (process_inode_count == 0)
2759                 return;
2760         old_operation = ehandler_operation(0);
2761         old_stashed_inode = ctx->stashed_inode;
2762         old_stashed_ino = ctx->stashed_ino;
2763         qsort(inodes_to_process, process_inode_count,
2764                       sizeof(struct process_inode_block), process_inode_cmp);
2765         clear_problem_context(&pctx);
2766         for (i=0; i < process_inode_count; i++) {
2767                 pctx.inode = ctx->stashed_inode =
2768                         (struct ext2_inode *) &inodes_to_process[i].inode;
2769                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2770
2771 #if 0
2772                 printf("%u ", pctx.ino);
2773 #endif
2774                 sprintf(buf, _("reading indirect blocks of inode %u"),
2775                         pctx.ino);
2776                 ehandler_operation(buf);
2777                 check_blocks(ctx, &pctx, block_buf,
2778                              &inodes_to_process[i].ea_ibody_quota);
2779                 if (e2fsck_should_abort(ctx))
2780                         break;
2781         }
2782         ctx->stashed_inode = old_stashed_inode;
2783         ctx->stashed_ino = old_stashed_ino;
2784         process_inode_count = 0;
2785 #if 0
2786         printf("end process inodes\n");
2787 #endif
2788         ehandler_operation(old_operation);
2789 }
2790
2791 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
2792 {
2793         const struct process_inode_block *ib_a =
2794                 (const struct process_inode_block *) a;
2795         const struct process_inode_block *ib_b =
2796                 (const struct process_inode_block *) b;
2797         int     ret;
2798
2799         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
2800                ib_b->inode.i_block[EXT2_IND_BLOCK]);
2801         if (ret == 0)
2802                 /*
2803                  * We only call process_inodes() for non-extent
2804                  * inodes, so it's OK to pass NULL to
2805                  * ext2fs_file_acl_block() here.
2806                  */
2807                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
2808                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
2809         if (ret == 0)
2810                 ret = ib_a->ino - ib_b->ino;
2811         return ret;
2812 }
2813
2814 /*
2815  * Mark an inode as being bad in some what
2816  */
2817 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
2818 {
2819         struct          problem_context pctx;
2820
2821         if (!ctx->inode_bad_map) {
2822                 clear_problem_context(&pctx);
2823
2824                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2825                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
2826                                 "inode_bad_map", &ctx->inode_bad_map);
2827                 if (pctx.errcode) {
2828                         pctx.num = 3;
2829                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2830                         /* Should never get here */
2831                         ctx->flags |= E2F_FLAG_ABORT;
2832                         return;
2833                 }
2834         }
2835         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
2836 }
2837
2838 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino)
2839 {
2840         struct          problem_context pctx;
2841
2842         if (!ctx->casefolded_dirs) {
2843                 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
2844                 if (pctx.errcode)
2845                         goto error;
2846         }
2847         pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
2848         if (pctx.errcode == 0)
2849                 return;
2850 error:
2851         fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
2852         /* Should never get here */
2853         ctx->flags |= E2F_FLAG_ABORT;
2854 }
2855
2856 /*
2857  * This procedure will allocate the inode "bb" (badblock) map table
2858  */
2859 static void alloc_bb_map(e2fsck_t ctx)
2860 {
2861         struct          problem_context pctx;
2862
2863         clear_problem_context(&pctx);
2864         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2865                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
2866                         "inode_bb_map", &ctx->inode_bb_map);
2867         if (pctx.errcode) {
2868                 pctx.num = 4;
2869                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2870                 /* Should never get here */
2871                 ctx->flags |= E2F_FLAG_ABORT;
2872                 return;
2873         }
2874 }
2875
2876 /*
2877  * This procedure will allocate the inode imagic table
2878  */
2879 static void alloc_imagic_map(e2fsck_t ctx)
2880 {
2881         struct          problem_context pctx;
2882
2883         clear_problem_context(&pctx);
2884         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2885                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
2886                         "inode_imagic_map", &ctx->inode_imagic_map);
2887         if (pctx.errcode) {
2888                 pctx.num = 5;
2889                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2890                 /* Should never get here */
2891                 ctx->flags |= E2F_FLAG_ABORT;
2892                 return;
2893         }
2894 }
2895
2896 /*
2897  * Marks a block as in use, setting the dup_map if it's been set
2898  * already.  Called by process_block and process_bad_block.
2899  *
2900  * WARNING: Assumes checks have already been done to make sure block
2901  * is valid.  This is true in both process_block and process_bad_block.
2902  */
2903 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
2904 {
2905         struct          problem_context pctx;
2906
2907         clear_problem_context(&pctx);
2908
2909         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
2910                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
2911                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
2912                         return;
2913                 }
2914                 if (!ctx->block_dup_map) {
2915                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
2916                                         _("multiply claimed block map"),
2917                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
2918                                         &ctx->block_dup_map);
2919                         if (pctx.errcode) {
2920                                 pctx.num = 3;
2921                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
2922                                             &pctx);
2923                                 /* Should never get here */
2924                                 ctx->flags |= E2F_FLAG_ABORT;
2925                                 return;
2926                         }
2927                 }
2928                 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
2929         } else {
2930                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
2931         }
2932 }
2933
2934 /*
2935  * When cluster size is greater than one block, it is caller's responsibility
2936  * to make sure block parameter starts at a cluster boundary.
2937  */
2938 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
2939                                       unsigned int num)
2940 {
2941         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
2942                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
2943         else {
2944                 unsigned int i;
2945
2946                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
2947                         mark_block_used(ctx, block + i);
2948         }
2949 }
2950
2951 /*
2952  * Adjust the extended attribute block's reference counts at the end
2953  * of pass 1, either by subtracting out references for EA blocks that
2954  * are still referenced in ctx->refcount, or by adding references for
2955  * EA blocks that had extra references as accounted for in
2956  * ctx->refcount_extra.
2957  */
2958 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
2959                                     char *block_buf, int adjust_sign)
2960 {
2961         struct ext2_ext_attr_header     *header;
2962         struct problem_context          pctx;
2963         ext2_filsys                     fs = ctx->fs;
2964         blk64_t                         blk;
2965         __u32                           should_be;
2966         ea_value_t                      count;
2967
2968         clear_problem_context(&pctx);
2969
2970         ea_refcount_intr_begin(refcount);
2971         while (1) {
2972                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
2973                         break;
2974                 pctx.blk = blk;
2975                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
2976                                                      pctx.ino);
2977                 if (pctx.errcode) {
2978                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
2979                         return;
2980                 }
2981                 header = (struct ext2_ext_attr_header *) block_buf;
2982                 pctx.blkcount = header->h_refcount;
2983                 should_be = header->h_refcount + adjust_sign * (int)count;
2984                 pctx.num = should_be;
2985                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
2986                         header->h_refcount = should_be;
2987                         pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
2988                                                              block_buf,
2989                                                              pctx.ino);
2990                         if (pctx.errcode) {
2991                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
2992                                             &pctx);
2993                                 continue;
2994                         }
2995                 }
2996         }
2997 }
2998
2999 /*
3000  * Handle processing the extended attribute blocks
3001  */
3002 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
3003                            char *block_buf, struct ea_quota *ea_block_quota)
3004 {
3005         ext2_filsys fs = ctx->fs;
3006         ext2_ino_t      ino = pctx->ino;
3007         struct ext2_inode *inode = pctx->inode;
3008         blk64_t         blk;
3009         char *          end;
3010         struct ext2_ext_attr_header *header;
3011         struct ext2_ext_attr_entry *first, *entry;
3012         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
3013         __u64           quota_inodes = 0;
3014         region_t        region = 0;
3015         int             failed_csum = 0;
3016
3017         ea_block_quota->blocks = 0;
3018         ea_block_quota->inodes = 0;
3019
3020         blk = ext2fs_file_acl_block(fs, inode);
3021         if (blk == 0)
3022                 return 0;
3023
3024         /*
3025          * If the Extended attribute flag isn't set, then a non-zero
3026          * file acl means that the inode is corrupted.
3027          *
3028          * Or if the extended attribute block is an invalid block,
3029          * then the inode is also corrupted.
3030          */
3031         if (!ext2fs_has_feature_xattr(fs->super) ||
3032             (blk < fs->super->s_first_data_block) ||
3033             (blk >= ext2fs_blocks_count(fs->super))) {
3034                 mark_inode_bad(ctx, ino);
3035                 return 0;
3036         }
3037
3038         /* If ea bitmap hasn't been allocated, create it */
3039         if (!ctx->block_ea_map) {
3040                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
3041                                         _("ext attr block map"),
3042                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
3043                                         &ctx->block_ea_map);
3044                 if (pctx->errcode) {
3045                         pctx->num = 2;
3046                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
3047                         ctx->flags |= E2F_FLAG_ABORT;
3048                         return 0;
3049                 }
3050         }
3051
3052         /* Create the EA refcount structure if necessary */
3053         if (!ctx->refcount) {
3054                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
3055                 if (pctx->errcode) {
3056                         pctx->num = 1;
3057                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3058                         ctx->flags |= E2F_FLAG_ABORT;
3059                         return 0;
3060                 }
3061         }
3062
3063 #if 0
3064         /* Debugging text */
3065         printf("Inode %u has EA block %u\n", ino, blk);
3066 #endif
3067
3068         /* Have we seen this EA block before? */
3069         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
3070                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
3071                 ea_block_quota->inodes = 0;
3072
3073                 if (ctx->ea_block_quota_blocks) {
3074                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
3075                                           &quota_blocks);
3076                         if (quota_blocks)
3077                                 ea_block_quota->blocks = quota_blocks;
3078                 }
3079
3080                 if (ctx->ea_block_quota_inodes)
3081                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
3082                                           &ea_block_quota->inodes);
3083
3084                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
3085                         return 1;
3086                 /* Ooops, this EA was referenced more than it stated */
3087                 if (!ctx->refcount_extra) {
3088                         pctx->errcode = ea_refcount_create(0,
3089                                            &ctx->refcount_extra);
3090                         if (pctx->errcode) {
3091                                 pctx->num = 2;
3092                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3093                                 ctx->flags |= E2F_FLAG_ABORT;
3094                                 return 0;
3095                         }
3096                 }
3097                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
3098                 return 1;
3099         }
3100
3101         /*
3102          * OK, we haven't seen this EA block yet.  So we need to
3103          * validate it
3104          */
3105         pctx->blk = blk;
3106         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
3107         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
3108                 pctx->errcode = 0;
3109                 failed_csum = 1;
3110         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
3111                 pctx->errcode = 0;
3112
3113         if (pctx->errcode &&
3114             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
3115                 pctx->errcode = 0;
3116                 goto clear_extattr;
3117         }
3118         header = (struct ext2_ext_attr_header *) block_buf;
3119         pctx->blk = ext2fs_file_acl_block(fs, inode);
3120         if (((ctx->ext_attr_ver == 1) &&
3121              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
3122             ((ctx->ext_attr_ver == 2) &&
3123              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
3124                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
3125                         goto clear_extattr;
3126         }
3127
3128         if (header->h_blocks != 1) {
3129                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
3130                         goto clear_extattr;
3131         }
3132
3133         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
3134                 goto clear_extattr;
3135
3136         region = region_create(0, fs->blocksize);
3137         if (!region) {
3138                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
3139                 ctx->flags |= E2F_FLAG_ABORT;
3140                 return 0;
3141         }
3142         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
3143                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3144                         goto clear_extattr;
3145         }
3146
3147         first = (struct ext2_ext_attr_entry *)(header+1);
3148         end = block_buf + fs->blocksize;
3149         entry = first;
3150         while ((char *)entry < end && *(__u32 *)entry) {
3151                 __u32 hash;
3152
3153                 if (region_allocate(region, (char *)entry - (char *)header,
3154                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
3155                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3156                                 goto clear_extattr;
3157                         break;
3158                 }
3159                 if ((ctx->ext_attr_ver == 1 &&
3160                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
3161                     (ctx->ext_attr_ver == 2 &&
3162                      entry->e_name_index == 0)) {
3163                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
3164                                 goto clear_extattr;
3165                         break;
3166                 }
3167                 if (entry->e_value_inum == 0) {
3168                         if (entry->e_value_size > EXT2_XATTR_SIZE_MAX ||
3169                             (entry->e_value_offs + entry->e_value_size >
3170                              fs->blocksize)) {
3171                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
3172                                         goto clear_extattr;
3173                                 break;
3174                         }
3175                         if (entry->e_value_size &&
3176                             region_allocate(region, entry->e_value_offs,
3177                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
3178                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
3179                                                 pctx))
3180                                         goto clear_extattr;
3181                         }
3182
3183                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
3184                                                           entry->e_value_offs);
3185
3186                         if (entry->e_hash != hash) {
3187                                 pctx->num = entry->e_hash;
3188                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
3189                                         goto clear_extattr;
3190                                 entry->e_hash = hash;
3191                         }
3192                 } else {
3193                         problem_t problem;
3194                         blk64_t entry_quota_blocks;
3195
3196                         problem = check_large_ea_inode(ctx, entry, pctx,
3197                                                        &entry_quota_blocks);
3198                         if (problem && fix_problem(ctx, problem, pctx))
3199                                 goto clear_extattr;
3200
3201                         quota_blocks += entry_quota_blocks;
3202                         quota_inodes++;
3203                 }
3204
3205                 entry = EXT2_EXT_ATTR_NEXT(entry);
3206         }
3207         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
3208                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3209                         goto clear_extattr;
3210         }
3211         region_free(region);
3212
3213         /*
3214          * We only get here if there was no other errors that were fixed.
3215          * If there was a checksum fail, ask to correct it.
3216          */
3217         if (failed_csum &&
3218             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
3219                 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
3220                                                        pctx->ino);
3221                 if (pctx->errcode)
3222                         return 0;
3223         }
3224
3225         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
3226                 if (!ctx->ea_block_quota_blocks) {
3227                         pctx->errcode = ea_refcount_create(0,
3228                                                 &ctx->ea_block_quota_blocks);
3229                         if (pctx->errcode) {
3230                                 pctx->num = 3;
3231                                 goto refcount_fail;
3232                         }
3233                 }
3234                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
3235                                   quota_blocks);
3236         }
3237
3238         if (quota_inodes) {
3239                 if (!ctx->ea_block_quota_inodes) {
3240                         pctx->errcode = ea_refcount_create(0,
3241                                                 &ctx->ea_block_quota_inodes);
3242                         if (pctx->errcode) {
3243                                 pctx->num = 4;
3244 refcount_fail:
3245                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3246                                 ctx->flags |= E2F_FLAG_ABORT;
3247                                 return 0;
3248                         }
3249                 }
3250
3251                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
3252                                   quota_inodes);
3253         }
3254         ea_block_quota->blocks = quota_blocks;
3255         ea_block_quota->inodes = quota_inodes;
3256
3257         inc_ea_inode_refs(ctx, pctx, first, end);
3258         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
3259         mark_block_used(ctx, blk);
3260         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
3261         return 1;
3262
3263 clear_extattr:
3264         if (region)
3265                 region_free(region);
3266         ext2fs_file_acl_block_set(fs, inode, 0);
3267         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
3268         return 0;
3269 }
3270
3271 /* Returns 1 if bad htree, 0 if OK */
3272 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
3273                         ext2_ino_t ino, struct ext2_inode *inode,
3274                         char *block_buf)
3275 {
3276         struct ext2_dx_root_info        *root;
3277         ext2_filsys                     fs = ctx->fs;
3278         errcode_t                       retval;
3279         blk64_t                         blk;
3280
3281         if ((!LINUX_S_ISDIR(inode->i_mode) &&
3282              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
3283             (!ext2fs_has_feature_dir_index(fs->super) &&
3284              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
3285                 return 1;
3286
3287         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
3288
3289         if ((pctx->errcode) ||
3290             (blk == 0) ||
3291             (blk < fs->super->s_first_data_block) ||
3292             (blk >= ext2fs_blocks_count(fs->super))) {
3293                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3294                         return 1;
3295                 else
3296                         return 0;
3297         }
3298
3299         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
3300         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3301                 return 1;
3302
3303         /* XXX should check that beginning matches a directory */
3304         root = (struct ext2_dx_root_info *) (block_buf + 24);
3305
3306         if ((root->reserved_zero || root->info_length < 8) &&
3307             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3308                 return 1;
3309
3310         pctx->num = root->hash_version;
3311         if ((root->hash_version != EXT2_HASH_LEGACY) &&
3312             (root->hash_version != EXT2_HASH_HALF_MD4) &&
3313             (root->hash_version != EXT2_HASH_TEA) &&
3314             (root->hash_version != EXT2_HASH_SIPHASH) &&
3315             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
3316                 return 1;
3317
3318         if (ext4_hash_in_dirent(inode)) {
3319                 if (root->hash_version != EXT2_HASH_SIPHASH &&
3320                     fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
3321                         return 1;
3322         } else {
3323                 if (root->hash_version == EXT2_HASH_SIPHASH &&
3324                    fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
3325                         return 1;
3326         }
3327
3328         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
3329             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
3330                 return 1;
3331
3332         pctx->num = root->indirect_levels;
3333         /* if htree level is clearly too high, consider it to be broken */
3334         if (root->indirect_levels > EXT4_HTREE_LEVEL &&
3335             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3336                 return 1;
3337
3338         /* if level is only maybe too high, LARGE_DIR feature could be unset */
3339         if (root->indirect_levels > ext2_dir_htree_level(fs) &&
3340             !ext2fs_has_feature_largedir(fs->super)) {
3341                 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
3342                 unsigned idx_pb = 1 << (blockbits - 3);
3343
3344                 /* compare inode size/blocks vs. max-sized 2-level htree */
3345                 if (EXT2_I_SIZE(pctx->inode) <
3346                     (idx_pb - 1) * (idx_pb - 2) << blockbits &&
3347                     pctx->inode->i_blocks <
3348                     (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
3349                     fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3350                         return 1;
3351         }
3352
3353         if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
3354             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3355                 ctx->large_dirs++;
3356
3357         return 0;
3358 }
3359
3360 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
3361                         struct ext2_inode *inode, int restart_flag,
3362                         const char *source)
3363 {
3364         inode->i_flags = 0;
3365         inode->i_links_count = 0;
3366         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
3367         inode->i_dtime = ctx->now;
3368
3369         /*
3370          * If a special inode has such rotten block mappings that we
3371          * want to clear the whole inode, be sure to actually zap
3372          * the block maps because i_links_count isn't checked for
3373          * special inodes, and we'll end up right back here the next
3374          * time we run fsck.
3375          */
3376         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
3377                 memset(inode->i_block, 0, sizeof(inode->i_block));
3378
3379         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
3380         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
3381         if (ctx->inode_reg_map)
3382                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
3383         if (ctx->inode_bad_map)
3384                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
3385
3386         /*
3387          * If the inode was partially accounted for before processing
3388          * was aborted, we need to restart the pass 1 scan.
3389          */
3390         ctx->flags |= restart_flag;
3391
3392         if (ino == EXT2_BAD_INO)
3393                 memset(inode, 0, sizeof(struct ext2_inode));
3394
3395         e2fsck_write_inode(ctx, ino, inode, source);
3396 }
3397
3398 /*
3399  * Use the multiple-blocks reclamation code to fix alignment problems in
3400  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
3401  * physical cluster, and we want the block offsets within that cluster to
3402  * line up.
3403  */
3404 static int has_unaligned_cluster_map(e2fsck_t ctx,
3405                                      blk64_t last_pblk, blk64_t last_lblk,
3406                                      blk64_t pblk, blk64_t lblk)
3407 {
3408         blk64_t cluster_mask;
3409
3410         if (!ctx->fs->cluster_ratio_bits)
3411                 return 0;
3412         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
3413
3414         /*
3415          * If the block in the logical cluster doesn't align with the block in
3416          * the physical cluster...
3417          */
3418         if ((lblk & cluster_mask) != (pblk & cluster_mask))
3419                 return 1;
3420
3421         /*
3422          * If we cross a physical cluster boundary within a logical cluster...
3423          */
3424         if (last_pblk && (lblk & cluster_mask) != 0 &&
3425             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
3426             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
3427                 return 1;
3428
3429         return 0;
3430 }
3431
3432 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
3433                              struct process_block_struct *pb,
3434                              blk64_t start_block, blk64_t end_block,
3435                              blk64_t eof_block,
3436                              ext2_extent_handle_t ehandle,
3437                              int try_repairs)
3438 {
3439         struct ext2fs_extent    extent;
3440         blk64_t                 blk, last_lblk;
3441         unsigned int            i, n;
3442         int                     is_dir, is_leaf;
3443         problem_t               problem;
3444         struct ext2_extent_info info;
3445         int                     failed_csum = 0;
3446
3447         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
3448                 failed_csum = 1;
3449
3450         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
3451         if (pctx->errcode)
3452                 return;
3453         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
3454             !pb->eti.force_rebuild &&
3455             info.curr_level < MAX_EXTENT_DEPTH_COUNT) {
3456                 struct extent_tree_level *etl;
3457
3458                 etl = pb->eti.ext_info + info.curr_level;
3459                 etl->num_extents += info.num_entries;
3460                 etl->max_extents += info.max_entries;
3461                 /*
3462                  * Implementation wart: Splitting extent blocks when appending
3463                  * will leave the old block with one free entry.  Therefore
3464                  * unless the node is totally full, pretend that a non-root
3465                  * extent block can hold one fewer entry than it actually does,
3466                  * so that we don't repeatedly rebuild the extent tree.
3467                  */
3468                 if (info.curr_level && info.num_entries < info.max_entries)
3469                         etl->max_extents--;
3470         }
3471
3472         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
3473                                           &extent);
3474         while ((pctx->errcode == 0 ||
3475                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
3476                info.num_entries-- > 0) {
3477                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
3478                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
3479                 last_lblk = extent.e_lblk + extent.e_len - 1;
3480
3481                 problem = 0;
3482                 pctx->blk = extent.e_pblk;
3483                 pctx->blk2 = extent.e_lblk;
3484                 pctx->num = extent.e_len;
3485                 pctx->blkcount = extent.e_lblk + extent.e_len;
3486
3487                 if (extent.e_pblk == 0 ||
3488                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
3489                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
3490                         problem = PR_1_EXTENT_BAD_START_BLK;
3491                 else if (extent.e_lblk < start_block)
3492                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
3493                 else if ((end_block && last_lblk > end_block) &&
3494                          !(last_lblk > eof_block &&
3495                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
3496                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
3497                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
3498                 else if (is_leaf && extent.e_len == 0)
3499                         problem = PR_1_EXTENT_LENGTH_ZERO;
3500                 else if (is_leaf &&
3501                          (extent.e_pblk + extent.e_len) >
3502                          ext2fs_blocks_count(ctx->fs->super))
3503                         problem = PR_1_EXTENT_ENDS_BEYOND;
3504                 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
3505                          !ext2fs_has_feature_largedir(ctx->fs->super) &&
3506                          ((extent.e_lblk + extent.e_len) >
3507                           (1U << (21 - ctx->fs->super->s_log_block_size))))
3508                         problem = PR_1_TOOBIG_DIR;
3509
3510                 if (is_leaf && problem == 0 && extent.e_len > 0) {
3511 #if 0
3512                         printf("extent_region(ino=%u, expect=%llu, "
3513                                "lblk=%llu, len=%u)\n", pb->ino,
3514                                (unsigned long long) pb->next_lblock,
3515                                (unsigned long long) extent.e_lblk,
3516                                extent.e_len);
3517 #endif
3518                         if (extent.e_lblk < pb->next_lblock)
3519                                 problem = PR_1_EXTENT_COLLISION;
3520                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
3521                                 pb->next_lblock = extent.e_lblk + extent.e_len;
3522                 }
3523
3524                 /*
3525                  * Uninitialized blocks in a directory?  Clear the flag and
3526                  * we'll interpret the blocks later.
3527                  */
3528                 if (try_repairs && is_dir && problem == 0 &&
3529                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3530                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
3531                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3532                         pb->inode_modified = 1;
3533                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3534                                                               &extent);
3535                         if (pctx->errcode)
3536                                 return;
3537                         failed_csum = 0;
3538                 }
3539 #ifdef CONFIG_DEVELOPER_FEATURES
3540                 if (try_repairs && !is_dir && problem == 0 &&
3541                     (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
3542                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3543                     fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
3544                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3545                         pb->inode_modified = 1;
3546                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3547                                                               &extent);
3548                         if (pctx->errcode)
3549                                 return;
3550                         failed_csum = 0;
3551                 }
3552 #endif
3553                 if (try_repairs && problem) {
3554 report_problem:
3555                         if (fix_problem(ctx, problem, pctx)) {
3556                                 if (ctx->invalid_bitmaps) {
3557                                         /*
3558                                          * If fsck knows the bitmaps are bad,
3559                                          * skip to the next extent and
3560                                          * try to clear this extent again
3561                                          * after fixing the bitmaps, by
3562                                          * restarting fsck.
3563                                          */
3564                                         pctx->errcode = ext2fs_extent_get(
3565                                                           ehandle,
3566                                                           EXT2_EXTENT_NEXT_SIB,
3567                                                           &extent);
3568                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3569                                         if (pctx->errcode ==
3570                                                     EXT2_ET_NO_CURRENT_NODE) {
3571                                                 pctx->errcode = 0;
3572                                                 break;
3573                                         }
3574                                         continue;
3575                                 }
3576                                 e2fsck_read_bitmaps(ctx);
3577                                 pb->inode_modified = 1;
3578                                 pctx->errcode =
3579                                         ext2fs_extent_delete(ehandle, 0);
3580                                 if (pctx->errcode) {
3581                                         pctx->str = "ext2fs_extent_delete";
3582                                         return;
3583                                 }
3584                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3585                                 if (pctx->errcode &&
3586                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
3587                                         pctx->str = "ext2fs_extent_fix_parents";
3588                                         return;
3589                                 }
3590                                 pctx->errcode = ext2fs_extent_get(ehandle,
3591                                                                   EXT2_EXTENT_CURRENT,
3592                                                                   &extent);
3593                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
3594                                         pctx->errcode = 0;
3595                                         break;
3596                                 }
3597                                 failed_csum = 0;
3598                                 continue;
3599                         }
3600                         goto next;
3601                 }
3602
3603                 if (!is_leaf) {
3604                         blk64_t lblk = extent.e_lblk;
3605                         int next_try_repairs = 1;
3606
3607                         blk = extent.e_pblk;
3608
3609                         /*
3610                          * If this lower extent block collides with critical
3611                          * metadata, don't try to repair the damage.  Pass 1b
3612                          * will reallocate the block; then we can try again.
3613                          */
3614                         if (pb->ino != EXT2_RESIZE_INO &&
3615                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
3616                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3617                                                       extent.e_pblk)) {
3618                                 next_try_repairs = 0;
3619                                 pctx->blk = blk;
3620                                 fix_problem(ctx,
3621                                             PR_1_CRITICAL_METADATA_COLLISION,
3622                                             pctx);
3623                                 if ((ctx->options & E2F_OPT_NO) == 0)
3624                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3625                         }
3626                         pctx->errcode = ext2fs_extent_get(ehandle,
3627                                                   EXT2_EXTENT_DOWN, &extent);
3628                         if (pctx->errcode &&
3629                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
3630                                 pctx->str = "EXT2_EXTENT_DOWN";
3631                                 problem = PR_1_EXTENT_HEADER_INVALID;
3632                                 if (!next_try_repairs)
3633                                         return;
3634                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
3635                                         goto report_problem;
3636                                 return;
3637                         }
3638                         /* The next extent should match this index's logical start */
3639                         if (extent.e_lblk != lblk) {
3640                                 struct ext2_extent_info e_info;
3641
3642                                 pctx->errcode = ext2fs_extent_get_info(ehandle,
3643                                                                        &e_info);
3644                                 if (pctx->errcode) {
3645                                         pctx->str = "ext2fs_extent_get_info";
3646                                         return;
3647                                 }
3648                                 pctx->blk = lblk;
3649                                 pctx->blk2 = extent.e_lblk;
3650                                 pctx->num = e_info.curr_level - 1;
3651                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
3652                                 if (fix_problem(ctx, problem, pctx)) {
3653                                         pb->inode_modified = 1;
3654                                         pctx->errcode =
3655                                                 ext2fs_extent_fix_parents(ehandle);
3656                                         if (pctx->errcode) {
3657                                                 pctx->str = "ext2fs_extent_fix_parents";
3658                                                 return;
3659                                         }
3660                                 }
3661                         }
3662                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
3663                                          last_lblk, eof_block, ehandle,
3664                                          next_try_repairs);
3665                         if (pctx->errcode)
3666                                 return;
3667                         pctx->errcode = ext2fs_extent_get(ehandle,
3668                                                   EXT2_EXTENT_UP, &extent);
3669                         if (pctx->errcode) {
3670                                 pctx->str = "EXT2_EXTENT_UP";
3671                                 return;
3672                         }
3673                         mark_block_used(ctx, blk);
3674                         pb->num_blocks++;
3675                         goto next;
3676                 }
3677
3678                 if ((pb->previous_block != 0) &&
3679                     (pb->previous_block+1 != extent.e_pblk)) {
3680                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3681                                 char type = '?';
3682
3683                                 if (pb->is_dir)
3684                                         type = 'd';
3685                                 else if (pb->is_reg)
3686                                         type = 'f';
3687
3688                                 printf(("%6lu(%c): expecting %6lu "
3689                                         "actual extent "
3690                                         "phys %6lu log %lu len %lu\n"),
3691                                        (unsigned long) pctx->ino, type,
3692                                        (unsigned long) pb->previous_block+1,
3693                                        (unsigned long) extent.e_pblk,
3694                                        (unsigned long) extent.e_lblk,
3695                                        (unsigned long) extent.e_len);
3696                         }
3697                         pb->fragmented = 1;
3698                 }
3699                 /*
3700                  * If we notice a gap in the logical block mappings of an
3701                  * extent-mapped directory, offer to close the hole by
3702                  * moving the logical block down, otherwise we'll go mad in
3703                  * pass 3 allocating empty directory blocks to fill the hole.
3704                  */
3705                 if (try_repairs && is_dir &&
3706                     pb->last_block + 1 < extent.e_lblk) {
3707                         blk64_t new_lblk;
3708
3709                         new_lblk = pb->last_block + 1;
3710                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3711                                 new_lblk = ((new_lblk +
3712                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3713                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3714                                            (extent.e_pblk &
3715                                             EXT2FS_CLUSTER_MASK(ctx->fs));
3716                         pctx->blk = extent.e_lblk;
3717                         pctx->blk2 = new_lblk;
3718                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3719                                 extent.e_lblk = new_lblk;
3720                                 pb->inode_modified = 1;
3721                                 pctx->errcode = ext2fs_extent_replace(ehandle,
3722                                                                 0, &extent);
3723                                 if (pctx->errcode) {
3724                                         pctx->errcode = 0;
3725                                         goto alloc_later;
3726                                 }
3727                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3728                                 if (pctx->errcode)
3729                                         goto failed_add_dir_block;
3730                                 pctx->errcode = ext2fs_extent_goto(ehandle,
3731                                                                 extent.e_lblk);
3732                                 if (pctx->errcode)
3733                                         goto failed_add_dir_block;
3734                                 last_lblk = extent.e_lblk + extent.e_len - 1;
3735                                 failed_csum = 0;
3736                         }
3737                 }
3738 alloc_later:
3739                 if (is_dir) {
3740                         while (++pb->last_db_block <
3741                                (e2_blkcnt_t) extent.e_lblk) {
3742                                 pctx->errcode = ext2fs_add_dir_block2(
3743                                                         ctx->fs->dblist,
3744                                                         pb->ino, 0,
3745                                                         pb->last_db_block);
3746                                 if (pctx->errcode) {
3747                                         pctx->blk = 0;
3748                                         pctx->num = pb->last_db_block;
3749                                         goto failed_add_dir_block;
3750                                 }
3751                         }
3752
3753                         for (i = 0; i < extent.e_len; i++) {
3754                                 pctx->errcode = ext2fs_add_dir_block2(
3755                                                         ctx->fs->dblist,
3756                                                         pctx->ino,
3757                                                         extent.e_pblk + i,
3758                                                         extent.e_lblk + i);
3759                                 if (pctx->errcode) {
3760                                         pctx->blk = extent.e_pblk + i;
3761                                         pctx->num = extent.e_lblk + i;
3762                                 failed_add_dir_block:
3763                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3764                                         /* Should never get here */
3765                                         ctx->flags |= E2F_FLAG_ABORT;
3766                                         return;
3767                                 }
3768                         }
3769                         if (extent.e_len > 0)
3770                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3771                 }
3772                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3773                                               pb->last_block,
3774                                               extent.e_pblk,
3775                                               extent.e_lblk)) {
3776                         for (i = 0; i < extent.e_len; i++) {
3777                                 pctx->blk = extent.e_lblk + i;
3778                                 pctx->blk2 = extent.e_pblk + i;
3779                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3780                                 mark_block_used(ctx, extent.e_pblk + i);
3781                                 mark_block_used(ctx, extent.e_pblk + i);
3782                         }
3783                 }
3784
3785                 /*
3786                  * Check whether first cluster got marked in previous iteration.
3787                  */
3788                 if (ctx->fs->cluster_ratio_bits &&
3789                     pb->previous_block &&
3790                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
3791                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
3792                         /* Set blk to the beginning of next cluster. */
3793                         blk = EXT2FS_C2B(
3794                                 ctx->fs,
3795                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
3796                 else
3797                         /* Set blk to the beginning of current cluster. */
3798                         blk = EXT2FS_C2B(ctx->fs,
3799                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
3800
3801                 if (blk < extent.e_pblk + extent.e_len) {
3802                         mark_blocks_used(ctx, blk,
3803                                          extent.e_pblk + extent.e_len - blk);
3804                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
3805                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
3806                         pb->num_blocks += n;
3807                 }
3808                 pb->last_block = extent.e_lblk + extent.e_len - 1;
3809                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
3810                 start_block = pb->last_block = last_lblk;
3811                 if (is_leaf && !is_dir &&
3812                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
3813                         pb->last_init_lblock = last_lblk;
3814         next:
3815                 pctx->errcode = ext2fs_extent_get(ehandle,
3816                                                   EXT2_EXTENT_NEXT_SIB,
3817                                                   &extent);
3818         }
3819
3820         /* Failed csum but passes checks?  Ask to fix checksum. */
3821         if (failed_csum &&
3822             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
3823                 pb->inode_modified = 1;
3824                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
3825                 if (pctx->errcode)
3826                         return;
3827         }
3828
3829         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
3830                 pctx->errcode = 0;
3831 }
3832
3833 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
3834                                  struct process_block_struct *pb)
3835 {
3836         struct ext2_extent_info info;
3837         struct ext2_inode       *inode = pctx->inode;
3838         ext2_extent_handle_t    ehandle;
3839         ext2_filsys             fs = ctx->fs;
3840         ext2_ino_t              ino = pctx->ino;
3841         errcode_t               retval;
3842         blk64_t                 eof_lblk;
3843         struct ext3_extent_header       *eh;
3844
3845         /* Check for a proper extent header... */
3846         eh = (struct ext3_extent_header *) &inode->i_block[0];
3847         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
3848         if (retval) {
3849                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
3850                         e2fsck_clear_inode(ctx, ino, inode, 0,
3851                                            "check_blocks_extents");
3852                 pctx->errcode = 0;
3853                 return;
3854         }
3855
3856         /* ...since this function doesn't fail if i_block is zeroed. */
3857         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
3858         if (pctx->errcode) {
3859                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
3860                         e2fsck_clear_inode(ctx, ino, inode, 0,
3861                                            "check_blocks_extents");
3862                 pctx->errcode = 0;
3863                 return;
3864         }
3865
3866         retval = ext2fs_extent_get_info(ehandle, &info);
3867         if (retval == 0) {
3868                 int max_depth = info.max_depth;
3869
3870                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
3871                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
3872                 ctx->extent_depth_count[max_depth]++;
3873         }
3874
3875         /* Check maximum extent depth */
3876         pctx->blk = info.max_depth;
3877         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
3878         if (pctx->blk2 < pctx->blk &&
3879             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
3880                 pb->eti.force_rebuild = 1;
3881
3882         /* Can we collect extent tree level stats? */
3883         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
3884         if (pctx->blk2 > pctx->blk)
3885                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
3886         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
3887         pb->eti.ino = pb->ino;
3888
3889         pb->next_lblock = 0;
3890
3891         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
3892                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
3893         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
3894         if (pctx->errcode &&
3895             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
3896                 pb->num_blocks = 0;
3897                 inode->i_blocks = 0;
3898                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3899                                    "check_blocks_extents");
3900                 pctx->errcode = 0;
3901         }
3902         ext2fs_extent_free(ehandle);
3903
3904         /* Rebuild unless it's a dir and we're rehashing it */
3905         if (LINUX_S_ISDIR(inode->i_mode) &&
3906             e2fsck_dir_will_be_rehashed(ctx, ino))
3907                 return;
3908
3909         if (ctx->options & E2F_OPT_CONVERT_BMAP)
3910                 e2fsck_rebuild_extents_later(ctx, ino);
3911         else
3912                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
3913 }
3914
3915 /*
3916  * In fact we don't need to check blocks for an inode with inline data
3917  * because this inode doesn't have any blocks.  In this function all
3918  * we need to do is add this inode into dblist when it is a directory.
3919  */
3920 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
3921                                      struct process_block_struct *pb)
3922 {
3923         int     flags;
3924         size_t  inline_data_size = 0;
3925
3926         if (!pb->is_dir) {
3927                 pctx->errcode = 0;
3928                 return;
3929         }
3930
3931         /* Process the dirents in i_block[] as the "first" block. */
3932         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
3933         if (pctx->errcode)
3934                 goto err;
3935
3936         /* Process the dirents in the EA as a "second" block. */
3937         flags = ctx->fs->flags;
3938         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3939         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
3940                                                 &inline_data_size);
3941         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3942                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3943         if (pctx->errcode) {
3944                 pctx->errcode = 0;
3945                 return;
3946         }
3947
3948         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
3949                 return;
3950
3951         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
3952         if (pctx->errcode)
3953                 goto err;
3954
3955         return;
3956 err:
3957         pctx->blk = 0;
3958         pctx->num = 0;
3959         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3960         ctx->flags |= E2F_FLAG_ABORT;
3961 }
3962
3963 /*
3964  * This subroutine is called on each inode to account for all of the
3965  * blocks used by that inode.
3966  */
3967 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
3968                          char *block_buf, const struct ea_quota *ea_ibody_quota)
3969 {
3970         ext2_filsys fs = ctx->fs;
3971         struct process_block_struct pb;
3972         ext2_ino_t      ino = pctx->ino;
3973         struct ext2_inode *inode = pctx->inode;
3974         unsigned        bad_size = 0;
3975         int             dirty_inode = 0;
3976         int             extent_fs;
3977         int             inlinedata_fs;
3978         __u64           size;
3979         struct ea_quota ea_block_quota;
3980
3981         pb.ino = ino;
3982         pb.num_blocks = EXT2FS_B2C(ctx->fs,
3983                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
3984         pb.last_block = ~0;
3985         pb.last_init_lblock = -1;
3986         pb.last_db_block = -1;
3987         pb.num_illegal_blocks = 0;
3988         pb.suppress = 0; pb.clear = 0;
3989         pb.fragmented = 0;
3990         pb.compressed = 0;
3991         pb.previous_block = 0;
3992         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
3993         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
3994         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
3995         pb.inode = inode;
3996         pb.pctx = pctx;
3997         pb.ctx = ctx;
3998         pb.inode_modified = 0;
3999         pb.eti.force_rebuild = 0;
4000         pctx->ino = ino;
4001         pctx->errcode = 0;
4002
4003         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
4004         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
4005
4006         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
4007                 if (e2fsck_should_abort(ctx))
4008                         goto out;
4009                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
4010         }
4011
4012         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
4013                 check_blocks_inline_data(ctx, pctx, &pb);
4014         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
4015                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
4016                         check_blocks_extents(ctx, pctx, &pb);
4017                 else {
4018                         int flags;
4019                         /*
4020                          * If we've modified the inode, write it out before
4021                          * iterate() tries to use it.
4022                          */
4023                         if (dirty_inode) {
4024                                 e2fsck_write_inode(ctx, ino, inode,
4025                                                    "check_blocks");
4026                                 dirty_inode = 0;
4027                         }
4028                         flags = fs->flags;
4029                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4030                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
4031                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
4032                                                 block_buf, process_block, &pb);
4033                         /*
4034                          * We do not have uninitialized extents in non extent
4035                          * files.
4036                          */
4037                         pb.last_init_lblock = pb.last_block;
4038                         /*
4039                          * If iterate() changed a block mapping, we have to
4040                          * re-read the inode.  If we decide to clear the
4041                          * inode after clearing some stuff, we'll re-write the
4042                          * bad mappings into the inode!
4043                          */
4044                         if (pb.inode_modified)
4045                                 e2fsck_read_inode(ctx, ino, inode,
4046                                                   "check_blocks");
4047                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4048                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4049
4050                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
4051 #ifdef DEBUG
4052                                 printf("bmap rebuild ino=%d\n", ino);
4053 #endif
4054                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
4055                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
4056                                         e2fsck_rebuild_extents_later(ctx, ino);
4057                         }
4058                 }
4059         }
4060         end_problem_latch(ctx, PR_LATCH_BLOCK);
4061         end_problem_latch(ctx, PR_LATCH_TOOBIG);
4062         if (e2fsck_should_abort(ctx))
4063                 goto out;
4064         if (pctx->errcode)
4065                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
4066
4067         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
4068                 if (LINUX_S_ISDIR(inode->i_mode))
4069                         ctx->fs_fragmented_dir++;
4070                 else
4071                         ctx->fs_fragmented++;
4072         }
4073
4074         if (pb.clear) {
4075                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4076                                    "check_blocks");
4077                 return;
4078         }
4079
4080         if (inode->i_flags & EXT2_INDEX_FL) {
4081                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
4082                         inode->i_flags &= ~EXT2_INDEX_FL;
4083                         dirty_inode++;
4084                 } else {
4085                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
4086                 }
4087         }
4088
4089         if (!pb.num_blocks && pb.is_dir &&
4090             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
4091                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
4092                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
4093                         ctx->fs_directory_count--;
4094                         return;
4095                 }
4096         }
4097
4098         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
4099             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
4100             !(inode->i_flags & EXT4_EA_INODE_FL)) {
4101                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
4102                                ino,
4103                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
4104                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
4105                                   ino, (ea_ibody_quota ?
4106                                         ea_ibody_quota->inodes : 0) +
4107                                                 ea_block_quota.inodes + 1);
4108         }
4109
4110         if (!ext2fs_has_feature_huge_file(fs->super) ||
4111             !(inode->i_flags & EXT4_HUGE_FILE_FL))
4112                 pb.num_blocks *= (fs->blocksize / 512);
4113         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
4114 #if 0
4115         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
4116                ino, inode->i_size, (unsigned long long) pb.last_block,
4117                (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
4118                (unsigned long long) pb.num_blocks);
4119 #endif
4120         size = EXT2_I_SIZE(inode);
4121         if (pb.is_dir) {
4122                 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
4123                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
4124                         int flags;
4125                         size_t sz = 0;
4126                         errcode_t err;
4127
4128                         flags = ctx->fs->flags;
4129                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4130                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
4131                                                       &sz);
4132                         ctx->fs->flags = (flags &
4133                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4134                                          (ctx->fs->flags &
4135                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4136                         if (err || sz != size) {
4137                                 bad_size = 7;
4138                                 pctx->num = sz;
4139                         }
4140                 } else if (size & (fs->blocksize - 1))
4141                         bad_size = 5;
4142                 else if (nblock > (pb.last_block + 1))
4143                         bad_size = 1;
4144                 else if (nblock < (pb.last_block + 1)) {
4145                         if (((pb.last_block + 1) - nblock) >
4146                             fs->super->s_prealloc_dir_blocks)
4147                                 bad_size = 2;
4148                 }
4149         } else {
4150                 if ((pb.last_init_lblock >= 0) &&
4151                     /* Do not allow initialized allocated blocks past i_size*/
4152                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
4153                     !(inode->i_flags & EXT4_VERITY_FL))
4154                         bad_size = 3;
4155                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4156                          size > ext2_max_sizes[fs->super->s_log_block_size])
4157                         /* too big for a direct/indirect-mapped file */
4158                         bad_size = 4;
4159                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4160                          size >
4161                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
4162                         /* too big for an extent-based file - 32bit ee_block */
4163                         bad_size = 6;
4164         }
4165         /* i_size for symlinks is checked elsewhere */
4166         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
4167                 /* Did inline_data set pctx->num earlier? */
4168                 if (bad_size != 7)
4169                         pctx->num = (pb.last_block + 1) * fs->blocksize;
4170                 pctx->group = bad_size;
4171                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
4172                         ext2fs_inode_size_set(fs, inode, pctx->num);
4173                         if (EXT2_I_SIZE(inode) == 0 &&
4174                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
4175                                 memset(inode->i_block, 0,
4176                                        sizeof(inode->i_block));
4177                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
4178                         }
4179                         dirty_inode++;
4180                 }
4181                 pctx->num = 0;
4182         }
4183         if (LINUX_S_ISREG(inode->i_mode) &&
4184             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4185                 ctx->large_files++;
4186         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
4187             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
4188              (ext2fs_has_feature_huge_file(fs->super) &&
4189               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
4190               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
4191                 pctx->num = pb.num_blocks;
4192                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
4193                         inode->i_blocks = pb.num_blocks;
4194                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
4195                         dirty_inode++;
4196                 }
4197                 pctx->num = 0;
4198         }
4199
4200         /*
4201          * The kernel gets mad if we ask it to allocate bigalloc clusters to
4202          * a block mapped file, so rebuild it as an extent file.  We can skip
4203          * symlinks because they're never rewritten.
4204          */
4205         if (ext2fs_has_feature_bigalloc(fs->super) &&
4206             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
4207             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
4208             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
4209             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
4210             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
4211                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
4212                 if (pctx->errcode)
4213                         goto out;
4214         }
4215
4216         if (ctx->dirs_to_hash && pb.is_dir &&
4217             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
4218             !(inode->i_flags & EXT2_INDEX_FL) &&
4219             ((inode->i_size / fs->blocksize) >= 3))
4220                 e2fsck_rehash_dir_later(ctx, ino);
4221
4222 out:
4223         if (dirty_inode)
4224                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
4225 }
4226
4227 #if 0
4228 /*
4229  * Helper function called by process block when an illegal block is
4230  * found.  It returns a description about why the block is illegal
4231  */
4232 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
4233 {
4234         blk64_t super;
4235         int     i;
4236         static char     problem[80];
4237
4238         super = fs->super->s_first_data_block;
4239         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
4240         if (block < super) {
4241                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
4242                 return(problem);
4243         } else if (block >= ext2fs_blocks_count(fs->super)) {
4244                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
4245                 return(problem);
4246         }
4247         for (i = 0; i < fs->group_desc_count; i++) {
4248                 if (block == super) {
4249                         sprintf(problem, "is the superblock in group %d", i);
4250                         break;
4251                 }
4252                 if (block > super &&
4253                     block <= (super + fs->desc_blocks)) {
4254                         sprintf(problem, "is in the group descriptors "
4255                                 "of group %d", i);
4256                         break;
4257                 }
4258                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
4259                         sprintf(problem, "is the block bitmap of group %d", i);
4260                         break;
4261                 }
4262                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
4263                         sprintf(problem, "is the inode bitmap of group %d", i);
4264                         break;
4265                 }
4266                 if (block >= ext2fs_inode_table_loc(fs, i) &&
4267                     (block < ext2fs_inode_table_loc(fs, i)
4268                      + fs->inode_blocks_per_group)) {
4269                         sprintf(problem, "is in the inode table of group %d",
4270                                 i);
4271                         break;
4272                 }
4273                 super += fs->super->s_blocks_per_group;
4274         }
4275         return(problem);
4276 }
4277 #endif
4278
4279 /*
4280  * This is a helper function for check_blocks().
4281  */
4282 static int process_block(ext2_filsys fs,
4283                   blk64_t       *block_nr,
4284                   e2_blkcnt_t blockcnt,
4285                   blk64_t ref_block EXT2FS_ATTR((unused)),
4286                   int ref_offset EXT2FS_ATTR((unused)),
4287                   void *priv_data)
4288 {
4289         struct process_block_struct *p;
4290         struct problem_context *pctx;
4291         blk64_t blk = *block_nr;
4292         int     ret_code = 0;
4293         problem_t       problem = 0;
4294         e2fsck_t        ctx;
4295
4296         p = (struct process_block_struct *) priv_data;
4297         pctx = p->pctx;
4298         ctx = p->ctx;
4299
4300         /*
4301          * For a directory, add logical block zero for processing even if it's
4302          * not mapped or we'll be perennially stuck with broken "." and ".."
4303          * entries.
4304          */
4305         if (p->is_dir && blockcnt == 0 && blk == 0) {
4306                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
4307                 if (pctx->errcode) {
4308                         pctx->blk = blk;
4309                         pctx->num = blockcnt;
4310                         goto failed_add_dir_block;
4311                 }
4312                 p->last_db_block++;
4313         }
4314
4315         if (blk == 0)
4316                 return 0;
4317
4318 #if 0
4319         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
4320                blockcnt);
4321 #endif
4322
4323         /*
4324          * Simplistic fragmentation check.  We merely require that the
4325          * file be contiguous.  (Which can never be true for really
4326          * big files that are greater than a block group.)
4327          */
4328         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
4329                 if (p->previous_block+1 != blk) {
4330                         if (ctx->options & E2F_OPT_FRAGCHECK) {
4331                                 char type = '?';
4332
4333                                 if (p->is_dir)
4334                                         type = 'd';
4335                                 else if (p->is_reg)
4336                                         type = 'f';
4337
4338                                 printf(_("%6lu(%c): expecting %6lu "
4339                                          "got phys %6lu (blkcnt %lld)\n"),
4340                                        (unsigned long) pctx->ino, type,
4341                                        (unsigned long) p->previous_block+1,
4342                                        (unsigned long) blk,
4343                                        (long long) blockcnt);
4344                         }
4345                         p->fragmented = 1;
4346                 }
4347         }
4348
4349         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
4350             !pctx->inode->i_size_high &&
4351             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
4352                 problem = PR_1_TOOBIG_DIR;
4353         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
4354                 problem = PR_1_TOOBIG_DIR;
4355         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
4356                 problem = PR_1_TOOBIG_REG;
4357         if (!p->is_dir && !p->is_reg && blockcnt > 0)
4358                 problem = PR_1_TOOBIG_SYMLINK;
4359
4360         if (blk < fs->super->s_first_data_block ||
4361             blk >= ext2fs_blocks_count(fs->super))
4362                 problem = PR_1_ILLEGAL_BLOCK_NUM;
4363
4364         /*
4365          * If this IND/DIND/TIND block is squatting atop some critical metadata
4366          * (group descriptors, superblock, bitmap, inode table), any write to
4367          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
4368          * fix that and restart fsck.
4369          */
4370         if (blockcnt < 0 &&
4371             p->ino != EXT2_RESIZE_INO &&
4372             blk < ctx->fs->super->s_blocks_count &&
4373             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
4374                 pctx->blk = blk;
4375                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
4376                 if ((ctx->options & E2F_OPT_NO) == 0)
4377                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4378         }
4379
4380         if (problem) {
4381                 p->num_illegal_blocks++;
4382                 /*
4383                  * A bit of subterfuge here -- we're trying to fix a block
4384                  * mapping, but the IND/DIND/TIND block could have collided
4385                  * with some critical metadata.  So, fix the in-core mapping so
4386                  * iterate won't go insane, but return 0 instead of
4387                  * BLOCK_CHANGED so that it won't write the remapping out to
4388                  * our multiply linked block.
4389                  *
4390                  * Even if we previously determined that an *IND block
4391                  * conflicts with critical metadata, we must still try to
4392                  * iterate the *IND block as if it is an *IND block to find and
4393                  * mark the blocks it points to.  Better to be overly cautious
4394                  * with the used_blocks map so that we don't move the *IND
4395                  * block to a block that's really in use!
4396                  */
4397                 if (p->ino != EXT2_RESIZE_INO &&
4398                     ref_block != 0 &&
4399                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4400                                               ref_block)) {
4401                         *block_nr = 0;
4402                         return 0;
4403                 }
4404                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
4405                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
4406                                 p->clear = 1;
4407                                 return BLOCK_ABORT;
4408                         }
4409                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
4410                                 p->suppress = 1;
4411                                 set_latch_flags(PR_LATCH_BLOCK,
4412                                                 PRL_SUPPRESS, 0);
4413                         }
4414                 }
4415                 pctx->blk = blk;
4416                 pctx->blkcount = blockcnt;
4417                 if (fix_problem(ctx, problem, pctx)) {
4418                         blk = *block_nr = 0;
4419                         ret_code = BLOCK_CHANGED;
4420                         p->inode_modified = 1;
4421                         /*
4422                          * If the directory block is too big and is beyond the
4423                          * end of the FS, don't bother trying to add it for
4424                          * processing -- the kernel would never have created a
4425                          * directory this large, and we risk an ENOMEM abort.
4426                          * In any case, the toobig handler for extent-based
4427                          * directories also doesn't feed toobig blocks to
4428                          * pass 2.
4429                          */
4430                         if (problem == PR_1_TOOBIG_DIR)
4431                                 return ret_code;
4432                         goto mark_dir;
4433                 } else
4434                         return 0;
4435         }
4436
4437         if (p->ino == EXT2_RESIZE_INO) {
4438                 /*
4439                  * The resize inode has already be sanity checked
4440                  * during pass #0 (the superblock checks).  All we
4441                  * have to do is mark the double indirect block as
4442                  * being in use; all of the other blocks are handled
4443                  * by mark_table_blocks()).
4444                  */
4445                 if (blockcnt == BLOCK_COUNT_DIND)
4446                         mark_block_used(ctx, blk);
4447                 p->num_blocks++;
4448         } else if (!(ctx->fs->cluster_ratio_bits &&
4449                      p->previous_block &&
4450                      (EXT2FS_B2C(ctx->fs, blk) ==
4451                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
4452                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
4453                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
4454                 mark_block_used(ctx, blk);
4455                 p->num_blocks++;
4456         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
4457                                              p->last_block, blk, blockcnt)) {
4458                 pctx->blk = blockcnt;
4459                 pctx->blk2 = blk;
4460                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4461                 mark_block_used(ctx, blk);
4462                 mark_block_used(ctx, blk);
4463         }
4464         if (blockcnt >= 0)
4465                 p->last_block = blockcnt;
4466         p->previous_block = blk;
4467 mark_dir:
4468         if (p->is_dir && (blockcnt >= 0)) {
4469                 while (++p->last_db_block < blockcnt) {
4470                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
4471                                                               p->ino, 0,
4472                                                               p->last_db_block);
4473                         if (pctx->errcode) {
4474                                 pctx->blk = 0;
4475                                 pctx->num = p->last_db_block;
4476                                 goto failed_add_dir_block;
4477                         }
4478                 }
4479                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
4480                                                       blk, blockcnt);
4481                 if (pctx->errcode) {
4482                         pctx->blk = blk;
4483                         pctx->num = blockcnt;
4484                 failed_add_dir_block:
4485                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4486                         /* Should never get here */
4487                         ctx->flags |= E2F_FLAG_ABORT;
4488                         return BLOCK_ABORT;
4489                 }
4490         }
4491         return ret_code;
4492 }
4493
4494 static int process_bad_block(ext2_filsys fs,
4495                       blk64_t *block_nr,
4496                       e2_blkcnt_t blockcnt,
4497                       blk64_t ref_block EXT2FS_ATTR((unused)),
4498                       int ref_offset EXT2FS_ATTR((unused)),
4499                       void *priv_data)
4500 {
4501         struct process_block_struct *p;
4502         blk64_t         blk = *block_nr;
4503         blk64_t         first_block;
4504         dgrp_t          i;
4505         struct problem_context *pctx;
4506         e2fsck_t        ctx;
4507
4508         if (!blk)
4509                 return 0;
4510
4511         p = (struct process_block_struct *) priv_data;
4512         ctx = p->ctx;
4513         pctx = p->pctx;
4514
4515         pctx->ino = EXT2_BAD_INO;
4516         pctx->blk = blk;
4517         pctx->blkcount = blockcnt;
4518
4519         if ((blk < fs->super->s_first_data_block) ||
4520             (blk >= ext2fs_blocks_count(fs->super))) {
4521                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
4522                         *block_nr = 0;
4523                         return BLOCK_CHANGED;
4524                 } else
4525                         return 0;
4526         }
4527
4528         if (blockcnt < 0) {
4529                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
4530                         p->bbcheck = 1;
4531                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
4532                                 *block_nr = 0;
4533                                 return BLOCK_CHANGED;
4534                         }
4535                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4536                                                     blk)) {
4537                         p->bbcheck = 1;
4538                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
4539                                         pctx)) {
4540                                 *block_nr = 0;
4541                                 return BLOCK_CHANGED;
4542                         }
4543                         if (e2fsck_should_abort(ctx))
4544                                 return BLOCK_ABORT;
4545                 } else
4546                         mark_block_used(ctx, blk);
4547                 return 0;
4548         }
4549 #if 0
4550         printf ("DEBUG: Marking %u as bad.\n", blk);
4551 #endif
4552         ctx->fs_badblocks_count++;
4553         /*
4554          * If the block is not used, then mark it as used and return.
4555          * If it is already marked as found, this must mean that
4556          * there's an overlap between the filesystem table blocks
4557          * (bitmaps and inode table) and the bad block list.
4558          */
4559         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
4560                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4561                 return 0;
4562         }
4563         /*
4564          * Try to find the where the filesystem block was used...
4565          */
4566         first_block = fs->super->s_first_data_block;
4567
4568         for (i = 0; i < fs->group_desc_count; i++ ) {
4569                 pctx->group = i;
4570                 pctx->blk = blk;
4571                 if (!ext2fs_bg_has_super(fs, i))
4572                         goto skip_super;
4573                 if (blk == first_block) {
4574                         if (i == 0) {
4575                                 if (fix_problem(ctx,
4576                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
4577                                                 pctx)) {
4578                                         *block_nr = 0;
4579                                         return BLOCK_CHANGED;
4580                                 }
4581                                 return 0;
4582                         }
4583                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
4584                         return 0;
4585                 }
4586                 if ((blk > first_block) &&
4587                     (blk <= first_block + fs->desc_blocks)) {
4588                         if (i == 0) {
4589                                 pctx->blk = *block_nr;
4590                                 if (fix_problem(ctx,
4591                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
4592                                         *block_nr = 0;
4593                                         return BLOCK_CHANGED;
4594                                 }
4595                                 return 0;
4596                         }
4597                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
4598                         return 0;
4599                 }
4600         skip_super:
4601                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
4602                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
4603                                 ctx->invalid_block_bitmap_flag[i]++;
4604                                 ctx->invalid_bitmaps++;
4605                         }
4606                         return 0;
4607                 }
4608                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
4609                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
4610                                 ctx->invalid_inode_bitmap_flag[i]++;
4611                                 ctx->invalid_bitmaps++;
4612                         }
4613                         return 0;
4614                 }
4615                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
4616                     (blk < (ext2fs_inode_table_loc(fs, i) +
4617                             fs->inode_blocks_per_group))) {
4618                         /*
4619                          * If there are bad blocks in the inode table,
4620                          * the inode scan code will try to do
4621                          * something reasonable automatically.
4622                          */
4623                         return 0;
4624                 }
4625                 first_block += fs->super->s_blocks_per_group;
4626         }
4627         /*
4628          * If we've gotten to this point, then the only
4629          * possibility is that the bad block inode meta data
4630          * is using a bad block.
4631          */
4632         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
4633             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
4634             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
4635                 p->bbcheck = 1;
4636                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
4637                         *block_nr = 0;
4638                         return BLOCK_CHANGED;
4639                 }
4640                 if (e2fsck_should_abort(ctx))
4641                         return BLOCK_ABORT;
4642                 return 0;
4643         }
4644
4645         pctx->group = -1;
4646
4647         /* Warn user that the block wasn't claimed */
4648         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
4649
4650         return 0;
4651 }
4652
4653 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
4654                             const char *name, int num, blk64_t *new_block)
4655 {
4656         ext2_filsys fs = ctx->fs;
4657         dgrp_t          last_grp;
4658         blk64_t         old_block = *new_block;
4659         blk64_t         last_block;
4660         dgrp_t          flexbg;
4661         unsigned        flexbg_size;
4662         int             i, is_flexbg;
4663         char            *buf;
4664         struct problem_context  pctx;
4665
4666         clear_problem_context(&pctx);
4667
4668         pctx.group = group;
4669         pctx.blk = old_block;
4670         pctx.str = name;
4671
4672         /*
4673          * For flex_bg filesystems, first try to allocate the metadata
4674          * within the flex_bg, and if that fails then try finding the
4675          * space anywhere in the filesystem.
4676          */
4677         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
4678         if (is_flexbg) {
4679                 flexbg_size = 1U << fs->super->s_log_groups_per_flex;
4680                 flexbg = group / flexbg_size;
4681                 first_block = ext2fs_group_first_block2(fs,
4682                                                         flexbg_size * flexbg);
4683                 last_grp = group | (flexbg_size - 1);
4684                 if (last_grp >= fs->group_desc_count)
4685                         last_grp = fs->group_desc_count - 1;
4686                 last_block = ext2fs_group_last_block2(fs, last_grp);
4687         } else
4688                 last_block = ext2fs_group_last_block2(fs, group);
4689         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
4690                                                num, ctx->block_found_map,
4691                                                new_block);
4692         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
4693                 pctx.errcode = ext2fs_get_free_blocks2(fs,
4694                                 fs->super->s_first_data_block,
4695                                 ext2fs_blocks_count(fs->super),
4696                                 num, ctx->block_found_map, new_block);
4697         if (pctx.errcode) {
4698                 pctx.num = num;
4699                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4700                 ext2fs_unmark_valid(fs);
4701                 ctx->flags |= E2F_FLAG_ABORT;
4702                 return;
4703         }
4704         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4705         if (pctx.errcode) {
4706                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4707                 ext2fs_unmark_valid(fs);
4708                 ctx->flags |= E2F_FLAG_ABORT;
4709                 return;
4710         }
4711         ext2fs_mark_super_dirty(fs);
4712         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4713         pctx.blk2 = *new_block;
4714         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4715                           PR_1_RELOC_TO), &pctx);
4716         pctx.blk2 = 0;
4717         for (i = 0; i < num; i++) {
4718                 pctx.blk = i;
4719                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4720                 if (old_block) {
4721                         pctx.errcode = io_channel_read_blk64(fs->io,
4722                                    old_block + i, 1, buf);
4723                         if (pctx.errcode)
4724                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4725                         pctx.blk = (*new_block) + i;
4726                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4727                                                               1, buf);
4728                 } else {
4729                         pctx.blk = (*new_block) + i;
4730                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4731                                                            NULL, NULL);
4732                 }
4733
4734                 if (pctx.errcode)
4735                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4736         }
4737         ext2fs_free_mem(&buf);
4738 }
4739
4740 /*
4741  * This routine gets called at the end of pass 1 if bad blocks are
4742  * detected in the superblock, group descriptors, inode_bitmaps, or
4743  * block bitmaps.  At this point, all of the blocks have been mapped
4744  * out, so we can try to allocate new block(s) to replace the bad
4745  * blocks.
4746  */
4747 static void handle_fs_bad_blocks(e2fsck_t ctx)
4748 {
4749         ext2_filsys fs = ctx->fs;
4750         dgrp_t          i;
4751         blk64_t         first_block;
4752         blk64_t         new_blk;
4753
4754         for (i = 0; i < fs->group_desc_count; i++) {
4755                 first_block = ext2fs_group_first_block2(fs, i);
4756
4757                 if (ctx->invalid_block_bitmap_flag[i]) {
4758                         new_blk = ext2fs_block_bitmap_loc(fs, i);
4759                         new_table_block(ctx, first_block, i, _("block bitmap"),
4760                                         1, &new_blk);
4761                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4762                 }
4763                 if (ctx->invalid_inode_bitmap_flag[i]) {
4764                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
4765                         new_table_block(ctx, first_block, i, _("inode bitmap"),
4766                                         1, &new_blk);
4767                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4768                 }
4769                 if (ctx->invalid_inode_table_flag[i]) {
4770                         new_blk = ext2fs_inode_table_loc(fs, i);
4771                         new_table_block(ctx, first_block, i, _("inode table"),
4772                                         fs->inode_blocks_per_group,
4773                                         &new_blk);
4774                         ext2fs_inode_table_loc_set(fs, i, new_blk);
4775                         ctx->flags |= E2F_FLAG_RESTART;
4776                 }
4777         }
4778         ctx->invalid_bitmaps = 0;
4779 }
4780
4781 /*
4782  * This routine marks all blocks which are used by the superblock,
4783  * group descriptors, inode bitmaps, and block bitmaps.
4784  */
4785 static void mark_table_blocks(e2fsck_t ctx)
4786 {
4787         ext2_filsys fs = ctx->fs;
4788         blk64_t b;
4789         dgrp_t  i;
4790         unsigned int    j;
4791         struct problem_context pctx;
4792
4793         clear_problem_context(&pctx);
4794
4795         for (i = 0; i < fs->group_desc_count; i++) {
4796                 pctx.group = i;
4797
4798                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
4799                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
4800
4801                 /*
4802                  * Mark the blocks used for the inode table
4803                  */
4804                 if (ext2fs_inode_table_loc(fs, i)) {
4805                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
4806                              j < fs->inode_blocks_per_group;
4807                              j++, b++) {
4808                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4809                                                              b)) {
4810                                         pctx.blk = b;
4811                                         if (!ctx->invalid_inode_table_flag[i] &&
4812                                             fix_problem(ctx,
4813                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
4814                                                 ctx->invalid_inode_table_flag[i]++;
4815                                                 ctx->invalid_bitmaps++;
4816                                         }
4817                                 } else {
4818                                     ext2fs_mark_block_bitmap2(
4819                                                 ctx->block_found_map, b);
4820                                     ext2fs_mark_block_bitmap2(
4821                                                 ctx->block_metadata_map, b);
4822                                 }
4823                         }
4824                 }
4825
4826                 /*
4827                  * Mark block used for the block bitmap
4828                  */
4829                 if (ext2fs_block_bitmap_loc(fs, i)) {
4830                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4831                                      ext2fs_block_bitmap_loc(fs, i))) {
4832                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
4833                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
4834                                         ctx->invalid_block_bitmap_flag[i]++;
4835                                         ctx->invalid_bitmaps++;
4836                                 }
4837                         } else {
4838                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4839                                      ext2fs_block_bitmap_loc(fs, i));
4840                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4841                                      ext2fs_block_bitmap_loc(fs, i));
4842                         }
4843                 }
4844                 /*
4845                  * Mark block used for the inode bitmap
4846                  */
4847                 if (ext2fs_inode_bitmap_loc(fs, i)) {
4848                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4849                                      ext2fs_inode_bitmap_loc(fs, i))) {
4850                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
4851                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
4852                                         ctx->invalid_inode_bitmap_flag[i]++;
4853                                         ctx->invalid_bitmaps++;
4854                                 }
4855                         } else {
4856                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4857                                      ext2fs_inode_bitmap_loc(fs, i));
4858                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4859                                      ext2fs_inode_bitmap_loc(fs, i));
4860                         }
4861                 }
4862         }
4863 }
4864
4865 /*
4866  * These subroutines short circuits ext2fs_get_blocks and
4867  * ext2fs_check_directory; we use them since we already have the inode
4868  * structure, so there's no point in letting the ext2fs library read
4869  * the inode again.
4870  */
4871 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
4872                                   blk_t *blocks)
4873 {
4874         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4875         int     i;
4876
4877         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4878                 return EXT2_ET_CALLBACK_NOTHANDLED;
4879
4880         for (i=0; i < EXT2_N_BLOCKS; i++)
4881                 blocks[i] = ctx->stashed_inode->i_block[i];
4882         return 0;
4883 }
4884
4885 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
4886                                   struct ext2_inode *inode)
4887 {
4888         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4889
4890         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4891                 return EXT2_ET_CALLBACK_NOTHANDLED;
4892         *inode = *ctx->stashed_inode;
4893         return 0;
4894 }
4895
4896 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
4897                             struct ext2_inode *inode)
4898 {
4899         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4900
4901         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
4902                 (inode != ctx->stashed_inode))
4903                 *ctx->stashed_inode = *inode;
4904         return EXT2_ET_CALLBACK_NOTHANDLED;
4905 }
4906
4907 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
4908 {
4909         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4910
4911         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4912                 return EXT2_ET_CALLBACK_NOTHANDLED;
4913
4914         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
4915                 return EXT2_ET_NO_DIRECTORY;
4916         return 0;
4917 }
4918
4919 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
4920                                         blk64_t *ret)
4921 {
4922         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4923         errcode_t       retval;
4924         blk64_t         new_block;
4925
4926         if (ctx->block_found_map) {
4927                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
4928                                            &new_block);
4929                 if (retval)
4930                         return retval;
4931                 if (fs->block_map) {
4932                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
4933                         ext2fs_mark_bb_dirty(fs);
4934                 }
4935         } else {
4936                 if (!fs->block_map) {
4937                         retval = ext2fs_read_block_bitmap(fs);
4938                         if (retval)
4939                                 return retval;
4940                 }
4941
4942                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
4943                 if (retval)
4944                         return retval;
4945         }
4946
4947         *ret = new_block;
4948         return (0);
4949 }
4950
4951 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
4952                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
4953 {
4954         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4955         errcode_t       retval;
4956
4957         if (ctx->block_found_map)
4958                 return ext2fs_new_range(fs, flags, goal, len,
4959                                         ctx->block_found_map, pblk, plen);
4960
4961         if (!fs->block_map) {
4962                 retval = ext2fs_read_block_bitmap(fs);
4963                 if (retval)
4964                         return retval;
4965         }
4966
4967         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
4968                                 pblk, plen);
4969 }
4970
4971 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
4972 {
4973         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4974
4975         /* Never free a critical metadata block */
4976         if (ctx->block_found_map &&
4977             ctx->block_metadata_map &&
4978             inuse < 0 &&
4979             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
4980                 return;
4981
4982         if (ctx->block_found_map) {
4983                 if (inuse > 0)
4984                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4985                 else
4986                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
4987         }
4988 }
4989
4990 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
4991                                            blk_t num, int inuse)
4992 {
4993         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4994
4995         /* Never free a critical metadata block */
4996         if (ctx->block_found_map &&
4997             ctx->block_metadata_map &&
4998             inuse < 0 &&
4999             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
5000                 return;
5001
5002         if (ctx->block_found_map) {
5003                 if (inuse > 0)
5004                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
5005                                                         blk, num);
5006                 else
5007                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
5008                                                         blk, num);
5009         }
5010 }
5011
5012 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
5013 {
5014         ext2_filsys fs = ctx->fs;
5015
5016         if (use_shortcuts) {
5017                 fs->get_blocks = pass1_get_blocks;
5018                 fs->check_directory = pass1_check_directory;
5019                 fs->read_inode = pass1_read_inode;
5020                 fs->write_inode = pass1_write_inode;
5021                 ctx->stashed_ino = 0;
5022         } else {
5023                 fs->get_blocks = 0;
5024                 fs->check_directory = 0;
5025                 fs->read_inode = 0;
5026                 fs->write_inode = 0;
5027         }
5028 }
5029
5030 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
5031 {
5032         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
5033         ext2fs_set_block_alloc_stats_callback(ctx->fs,
5034                                                 e2fsck_block_alloc_stats, 0);
5035         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
5036         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
5037                                         e2fsck_block_alloc_stats_range, NULL);
5038 }