Whamcloud - gitweb
e2fsck: merge counts after threads finish
[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, ext2_ino_t ino);
89 static void add_casefolded_dir(e2fsck_t ctx, ext2_ino_t ino);
90 static void handle_fs_bad_blocks(e2fsck_t ctx);
91 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
92 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
93                                   dgrp_t group, void * priv_data);
94 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
95                                     char *block_buf, int adjust_sign);
96 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
97
98 struct process_block_struct {
99         ext2_ino_t      ino;
100         unsigned        is_dir:1, is_reg:1, clear:1, suppress:1,
101                                 fragmented:1, compressed:1, bbcheck:1,
102                                 inode_modified:1;
103         blk64_t         num_blocks;
104         blk64_t         max_blocks;
105         blk64_t         last_block;
106         e2_blkcnt_t     last_init_lblock;
107         e2_blkcnt_t     last_db_block;
108         int             num_illegal_blocks;
109         blk64_t         previous_block;
110         struct ext2_inode *inode;
111         struct problem_context *pctx;
112         ext2fs_block_bitmap fs_meta_blocks;
113         e2fsck_t        ctx;
114         blk64_t         next_lblock;
115         struct extent_tree_info eti;
116 };
117
118 struct process_inode_block {
119         ext2_ino_t ino;
120         struct ea_quota ea_ibody_quota;
121         struct ext2_inode_large inode;
122 };
123
124 struct scan_callback_struct {
125         e2fsck_t                         ctx;
126         char                            *block_buf;
127         struct process_inode_block      *inodes_to_process;
128         int                             *process_inode_count;
129 };
130
131 static void process_inodes(e2fsck_t ctx, char *block_buf,
132                            struct process_inode_block *inodes_to_process,
133                            int *process_inode_count);
134
135 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
136                             EXT2_MIN_BLOCK_LOG_SIZE + 1];
137
138 /*
139  * Check to make sure a device inode is real.  Returns 1 if the device
140  * checks out, 0 if not.
141  *
142  * Note: this routine is now also used to check FIFO's and Sockets,
143  * since they have the same requirement; the i_block fields should be
144  * zero.
145  */
146 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
147                                     struct ext2_inode *inode)
148 {
149         int     i;
150
151         /*
152          * If the index or extents flag is set, then this is a bogus
153          * device/fifo/socket
154          */
155         if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
156                 return 0;
157
158         /*
159          * We should be able to do the test below all the time, but
160          * because the kernel doesn't forcibly clear the device
161          * inode's additional i_block fields, there are some rare
162          * occasions when a legitimate device inode will have non-zero
163          * additional i_block fields.  So for now, we only complain
164          * when the immutable flag is set, which should never happen
165          * for devices.  (And that's when the problem is caused, since
166          * you can't set or clear immutable flags for devices.)  Once
167          * the kernel has been fixed we can change this...
168          */
169         if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
170                 for (i=4; i < EXT2_N_BLOCKS; i++)
171                         if (inode->i_block[i])
172                                 return 0;
173         }
174         return 1;
175 }
176
177 /*
178  * Check to make sure a symlink inode is real.  Returns 1 if the symlink
179  * checks out, 0 if not.
180  */
181 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
182                                struct ext2_inode *inode, char *buf)
183 {
184         unsigned int buflen;
185         unsigned int len;
186
187         if ((inode->i_size_high || inode->i_size == 0) ||
188             (inode->i_flags & EXT2_INDEX_FL))
189                 return 0;
190
191         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
192                 size_t inline_size;
193
194                 if (inode->i_flags & EXT4_EXTENTS_FL)
195                         return 0;
196                 if (ext2fs_inline_data_size(fs, ino, &inline_size))
197                         return 0;
198                 if (inode->i_size != inline_size)
199                         return 0;
200
201                 return 1;
202         }
203
204         if (ext2fs_is_fast_symlink(inode)) {
205                 if (inode->i_flags & EXT4_EXTENTS_FL)
206                         return 0;
207                 buf = (char *)inode->i_block;
208                 buflen = sizeof(inode->i_block);
209         } else {
210                 ext2_extent_handle_t    handle;
211                 struct ext2_extent_info info;
212                 struct ext2fs_extent    extent;
213                 blk64_t blk;
214                 int i;
215
216                 if (inode->i_flags & EXT4_EXTENTS_FL) {
217                         if (ext2fs_extent_open2(fs, ino, inode, &handle))
218                                 return 0;
219                         if (ext2fs_extent_get_info(handle, &info) ||
220                             (info.num_entries != 1) ||
221                             (info.max_depth != 0)) {
222                                 ext2fs_extent_free(handle);
223                                 return 0;
224                         }
225                         if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
226                                               &extent) ||
227                             (extent.e_lblk != 0) ||
228                             (extent.e_len != 1)) {
229                                 ext2fs_extent_free(handle);
230                                 return 0;
231                         }
232                         blk = extent.e_pblk;
233                         ext2fs_extent_free(handle);
234                 } else {
235                         blk = inode->i_block[0];
236
237                         for (i = 1; i < EXT2_N_BLOCKS; i++)
238                                 if (inode->i_block[i])
239                                         return 0;
240                 }
241
242                 if (blk < fs->super->s_first_data_block ||
243                     blk >= ext2fs_blocks_count(fs->super))
244                         return 0;
245
246                 if (io_channel_read_blk64(fs->io, blk, 1, buf))
247                         return 0;
248
249                 buflen = fs->blocksize;
250         }
251
252         if (inode->i_flags & EXT4_ENCRYPT_FL)
253                 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
254         else
255                 len = strnlen(buf, buflen);
256
257         if (len >= buflen)
258                 return 0;
259
260         if (len != inode->i_size)
261                 return 0;
262         return 1;
263 }
264
265 /*
266  * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
267  */
268 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
269 static void check_extents_inlinedata(e2fsck_t ctx,
270                                      struct problem_context *pctx)
271 {
272         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
273                 return;
274
275         if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
276                 return;
277
278         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
279         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
280 }
281 #undef BAD_SPECIAL_FLAGS
282
283 /*
284  * If the immutable (or append-only) flag is set on the inode, offer
285  * to clear it.
286  */
287 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
288 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
289 {
290         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
291                 return;
292
293         if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
294                 return;
295
296         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
297         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
298 }
299
300 /*
301  * If device, fifo or socket, check size is zero -- if not offer to
302  * clear it
303  */
304 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
305 {
306         struct ext2_inode *inode = pctx->inode;
307
308         if (EXT2_I_SIZE(inode) == 0)
309                 return;
310
311         if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
312                 return;
313
314         ext2fs_inode_size_set(ctx->fs, inode, 0);
315         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
316 }
317
318 /*
319  * For a given size, calculate how many blocks would be charged towards quota.
320  */
321 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
322 {
323         blk64_t clusters;
324
325         clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
326         return EXT2FS_C2B(fs, clusters);
327 }
328
329 /*
330  * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
331  * the problem code.
332  */
333 static problem_t check_large_ea_inode(e2fsck_t ctx,
334                                       struct ext2_ext_attr_entry *entry,
335                                       struct problem_context *pctx,
336                                       blk64_t *quota_blocks)
337 {
338         struct ext2_inode inode;
339         __u32 hash, signed_hash;
340         errcode_t retval;
341
342         /* Check if inode is within valid range */
343         if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
344             (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
345                 pctx->num = entry->e_value_inum;
346                 return PR_1_ATTR_VALUE_EA_INODE;
347         }
348
349         e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
350
351         retval = ext2fs_ext_attr_hash_entry3(ctx->fs, entry, NULL, &hash,
352                                              &signed_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) || (signed_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                         if (entry->e_hash != 0 && entry->e_hash != hash)
506                                 hash = ext2fs_ext_attr_hash_entry_signed(entry,
507                                                 start + entry->e_value_offs);
508
509                         /* e_hash may be 0 in older inode's ea */
510                         if (entry->e_hash != 0 && entry->e_hash != hash) {
511                                 pctx->num = entry->e_hash;
512                                 problem = PR_1_ATTR_HASH;
513                                 goto fix;
514                         }
515                 } else {
516                         blk64_t quota_blocks;
517
518                         problem = check_large_ea_inode(ctx, entry, pctx,
519                                                        &quota_blocks);
520                         if (problem != 0)
521                                 goto fix;
522
523                         ea_ibody_quota->blocks += quota_blocks;
524                         ea_ibody_quota->inodes++;
525                 }
526
527                 /* If EA value is stored in external inode then it does not
528                  * consume space here */
529                 if (entry->e_value_inum == 0)
530                         remain -= entry->e_value_size;
531
532                 entry = EXT2_EXT_ATTR_NEXT(entry);
533         }
534
535         if (region_allocate(region, (char *)entry - (char *)header,
536                             sizeof(__u32))) {
537                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
538                 goto fix;
539         }
540 fix:
541         if (region)
542                 region_free(region);
543         /*
544          * it seems like a corruption. it's very unlikely we could repair
545          * EA(s) in automatic fashion -bzzz
546          */
547         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
548                 inc_ea_inode_refs(ctx, pctx,
549                                   (struct ext2_ext_attr_entry *)start, end);
550                 return;
551         }
552
553         /* simply remove all possible EA(s) */
554         *((__u32 *)header) = 0UL;
555         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
556                                 EXT2_INODE_SIZE(sb), "pass1");
557         ea_ibody_quota->blocks = 0;
558         ea_ibody_quota->inodes = 0;
559 }
560
561 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
562         return (xtime & (1U << 31)) != 0 &&
563                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
564 }
565
566 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
567         check_inode_extra_negative_epoch(inode->i_##xtime, \
568                                          inode->i_##xtime##_extra)
569
570 /* When today's date is earlier than 2242, we assume that atimes,
571  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
572  * actually pre-1970 dates mis-encoded.
573  */
574 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
575
576 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
577                                     struct ea_quota *ea_ibody_quota)
578 {
579         struct ext2_super_block *sb = ctx->fs->super;
580         struct ext2_inode_large *inode;
581         __u32 *eamagic;
582         int min, max;
583
584         ea_ibody_quota->blocks = 0;
585         ea_ibody_quota->inodes = 0;
586
587         inode = (struct ext2_inode_large *) pctx->inode;
588         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
589                 /* this isn't large inode. so, nothing to check */
590                 return;
591         }
592
593 #if 0
594         printf("inode #%u, i_extra_size %d\n", pctx->ino,
595                         inode->i_extra_isize);
596 #endif
597         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
598         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
599         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
600         /*
601          * For now we will allow i_extra_isize to be 0, but really
602          * implementations should never allow i_extra_isize to be 0
603          */
604         if (inode->i_extra_isize &&
605             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
606              inode->i_extra_isize & 3)) {
607                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
608                         return;
609                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
610                         inode->i_extra_isize = sb->s_want_extra_isize;
611                 else
612                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
613                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
614                                         EXT2_INODE_SIZE(sb), "pass1");
615         }
616
617         /* check if there is no place for an EA header */
618         if (inode->i_extra_isize >= max - sizeof(__u32))
619                 return;
620
621         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
622                         inode->i_extra_isize);
623         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
624                 /* it seems inode has an extended attribute(s) in body */
625                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
626         }
627
628         /*
629          * If the inode's extended atime (ctime, crtime, mtime) is stored in
630          * the old, invalid format, repair it.
631          */
632         if (((sizeof(time_t) <= 4) ||
633              (((sizeof(time_t) > 4) &&
634                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
635             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
636              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
637              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
638              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
639
640                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
641                         return;
642
643                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
644                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
645                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
646                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
647                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
648                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
649                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
650                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
651                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
652                                         EXT2_INODE_SIZE(sb), "pass1");
653         }
654
655 }
656
657 /*
658  * Check to see if the inode might really be a directory, despite i_mode
659  *
660  * This is a lot of complexity for something for which I'm not really
661  * convinced happens frequently in the wild.  If for any reason this
662  * causes any problems, take this code out.
663  * [tytso:20070331.0827EDT]
664  */
665 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
666                                 char *buf)
667 {
668         struct ext2_inode *inode = pctx->inode;
669         struct ext2_dir_entry   *dirent;
670         errcode_t               retval;
671         blk64_t                 blk;
672         unsigned int            i, rec_len, not_device = 0;
673         int                     extent_fs;
674         int                     inlinedata_fs;
675
676         /*
677          * If the mode looks OK, we believe it.  If the first block in
678          * the i_block array is 0, this cannot be a directory. If the
679          * inode is extent-mapped, it is still the case that the latter
680          * cannot be 0 - the magic number in the extent header would make
681          * it nonzero.
682          */
683         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
684             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
685                 return;
686
687         /* 
688          * Check the block numbers in the i_block array for validity:
689          * zero blocks are skipped (but the first one cannot be zero -
690          * see above), other blocks are checked against the first and
691          * max data blocks (from the the superblock) and against the
692          * block bitmap. Any invalid block found means this cannot be
693          * a directory.
694          * 
695          * If there are non-zero blocks past the fourth entry, then
696          * this cannot be a device file: we remember that for the next
697          * check.
698          *
699          * For extent mapped files, we don't do any sanity checking:
700          * just try to get the phys block of logical block 0 and run
701          * with it.
702          *
703          * For inline data files, we just try to get the size of inline
704          * data.  If it's true, we will treat it as a directory.
705          */
706
707         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
708         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
709         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
710                 size_t size;
711                 __u32 dotdot;
712                 unsigned int rec_len2;
713                 struct ext2_dir_entry de;
714
715                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
716                         return;
717                 /*
718                  * If the size isn't a multiple of 4, it's probably not a
719                  * directory??
720                  */
721                 if (size & 3)
722                         return;
723                 /*
724                  * If the first 10 bytes don't look like a directory entry,
725                  * it's probably not a directory.
726                  */
727                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
728                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
729                        EXT2_DIR_REC_LEN(0));
730                 dotdot = ext2fs_le32_to_cpu(dotdot);
731                 de.inode = ext2fs_le32_to_cpu(de.inode);
732                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
733                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
734                 if (dotdot >= ctx->fs->super->s_inodes_count ||
735                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
736                      dotdot != EXT2_ROOT_INO) ||
737                     de.inode >= ctx->fs->super->s_inodes_count ||
738                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
739                      de.inode != 0) ||
740                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
741                               EXT4_INLINE_DATA_DOTDOT_SIZE)
742                         return;
743                 /* device files never have a "system.data" entry */
744                 goto isdir;
745         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
746                 /* extent mapped */
747                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
748                                  &blk))
749                         return;
750                 /* device files are never extent mapped */
751                 not_device++;
752         } else {
753                 for (i=0; i < EXT2_N_BLOCKS; i++) {
754                         blk = inode->i_block[i];
755                         if (!blk)
756                                 continue;
757                         if (i >= 4)
758                                 not_device++;
759
760                         if (blk < ctx->fs->super->s_first_data_block ||
761                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
762                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
763                                                            blk))
764                                 return; /* Invalid block, can't be dir */
765                 }
766                 blk = inode->i_block[0];
767         }
768
769         /*
770          * If the mode says this is a device file and the i_links_count field
771          * is sane and we have not ruled it out as a device file previously,
772          * we declare it a device file, not a directory.
773          */
774         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
775             (inode->i_links_count == 1) && !not_device)
776                 return;
777
778         /* read the first block */
779         ehandler_operation(_("reading directory block"));
780         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
781         ehandler_operation(0);
782         if (retval)
783                 return;
784
785         dirent = (struct ext2_dir_entry *) buf;
786         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
787         if (retval)
788                 return;
789         if ((ext2fs_dirent_name_len(dirent) != 1) ||
790             (dirent->name[0] != '.') ||
791             (dirent->inode != pctx->ino) ||
792             (rec_len < 12) ||
793             (rec_len % 4) ||
794             (rec_len >= ctx->fs->blocksize - 12))
795                 return;
796
797         dirent = (struct ext2_dir_entry *) (buf + rec_len);
798         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
799         if (retval)
800                 return;
801         if ((ext2fs_dirent_name_len(dirent) != 2) ||
802             (dirent->name[0] != '.') ||
803             (dirent->name[1] != '.') ||
804             (rec_len < 12) ||
805             (rec_len % 4))
806                 return;
807
808 isdir:
809         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
810                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
811                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
812                                         EXT2_INODE_SIZE(ctx->fs->super),
813                                         "check_is_really_dir");
814         }
815 }
816
817 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
818                                      int flags, ext2_icount_t hint,
819                                      ext2_icount_t *ret)
820 {
821         unsigned int            threshold;
822         unsigned int            save_type;
823         ext2_ino_t              num_dirs;
824         errcode_t               retval;
825         char                    *tdb_dir;
826         int                     enable;
827
828         *ret = 0;
829
830         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
831                            &tdb_dir);
832         profile_get_uint(ctx->profile, "scratch_files",
833                          "numdirs_threshold", 0, 0, &threshold);
834         profile_get_boolean(ctx->profile, "scratch_files",
835                             "icount", 0, 1, &enable);
836
837         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
838         if (retval)
839                 num_dirs = 1024;        /* Guess */
840
841         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
842             (!threshold || num_dirs > threshold)) {
843                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
844                                                   flags, ret);
845                 if (retval == 0)
846                         return 0;
847         }
848         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
849                                &save_type);
850         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
851                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
852         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
853         ctx->fs->default_bitmap_type = save_type;
854         return retval;
855 }
856
857 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
858                                             e2fsck_t ctx,
859                                             struct problem_context *pctx)
860 {
861         errcode_t retval;
862         struct ext2_inode_large inode;
863
864         /*
865          * Reread inode.  If we don't see checksum error, then this inode
866          * has been fixed elsewhere.
867          */
868         ctx->stashed_ino = 0;
869         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
870                                         sizeof(inode));
871         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
872                 return retval;
873         if (!retval)
874                 return 0;
875
876         /*
877          * Checksum still doesn't match.  That implies that the inode passes
878          * all the sanity checks, so maybe the checksum is simply corrupt.
879          * See if the user will go for fixing that.
880          */
881         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
882                 return 0;
883
884         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
885                                          sizeof(inode));
886         return retval;
887 }
888
889 static void reserve_block_for_root_repair(e2fsck_t ctx)
890 {
891         blk64_t         blk = 0;
892         errcode_t       err;
893         ext2_filsys     fs = ctx->fs;
894
895         ctx->root_repair_block = 0;
896         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
897                 return;
898
899         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
900         if (err)
901                 return;
902         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
903         ctx->root_repair_block = blk;
904 }
905
906 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
907 {
908         blk64_t         blk = 0;
909         errcode_t       err;
910         ext2_filsys     fs = ctx->fs;
911         static const char name[] = "lost+found";
912         ext2_ino_t      ino;
913
914         ctx->lnf_repair_block = 0;
915         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
916                 return;
917
918         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
919         if (err)
920                 return;
921         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
922         ctx->lnf_repair_block = blk;
923 }
924
925 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
926                                          struct ext2_inode *inode,
927                                          size_t *sz)
928 {
929         void *p;
930         struct ext2_xattr_handle *handle;
931         errcode_t retval;
932
933         retval = ext2fs_xattrs_open(fs, ino, &handle);
934         if (retval)
935                 return retval;
936
937         retval = ext2fs_xattrs_read_inode(handle,
938                                           (struct ext2_inode_large *)inode);
939         if (retval)
940                 goto err;
941
942         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
943         if (retval)
944                 goto err;
945         ext2fs_free_mem(&p);
946 err:
947         (void) ext2fs_xattrs_close(&handle);
948         return retval;
949 }
950
951 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
952                                     struct problem_context *pctx,
953                                     int failed_csum)
954 {
955         if (!failed_csum)
956                 return;
957
958         /*
959          * If the inode failed the checksum and the user didn't
960          * clear the inode, test the checksum again -- if it still
961          * fails, ask the user if the checksum should be corrected.
962          */
963         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
964         if (pctx->errcode)
965                 ctx->flags |= E2F_FLAG_ABORT;
966 }
967 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
968         do { \
969                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
970                 if ((ctx)->flags & E2F_FLAG_ABORT) \
971                         return; \
972         } while (0)
973
974 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
975 {
976         __u32 x;
977         int i;
978
979         for (i = 0; i < EXT2_N_BLOCKS; i++) {
980                 x = inode->i_block[i];
981 #ifdef WORDS_BIGENDIAN
982                 x = ext2fs_swab32(x);
983 #endif
984                 if (x >= ext2fs_blocks_count(fs->super))
985                         return 0;
986         }
987
988         return 1;
989 }
990
991 /*
992  * Figure out what to do with an inode that has both extents and inline data
993  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
994  */
995 static int fix_inline_data_extents_file(e2fsck_t ctx,
996                                         ext2_ino_t ino,
997                                         struct ext2_inode *inode,
998                                         int inode_size,
999                                         struct problem_context *pctx)
1000 {
1001         size_t max_inline_ea_size;
1002         ext2_filsys fs = ctx->fs;
1003         int dirty = 0;
1004
1005         /* Both feature flags not set?  Just run the regular checks */
1006         if (!ext2fs_has_feature_extents(fs->super) &&
1007             !ext2fs_has_feature_inline_data(fs->super))
1008                 return 0;
1009
1010         /* Clear both flags if it's a special file */
1011         if (LINUX_S_ISCHR(inode->i_mode) ||
1012             LINUX_S_ISBLK(inode->i_mode) ||
1013             LINUX_S_ISFIFO(inode->i_mode) ||
1014             LINUX_S_ISSOCK(inode->i_mode)) {
1015                 check_extents_inlinedata(ctx, pctx);
1016                 return 0;
1017         }
1018
1019         /* If it looks like an extent tree, try to clear inlinedata */
1020         if (ext2fs_extent_header_verify(inode->i_block,
1021                                  sizeof(inode->i_block)) == 0 &&
1022             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1023                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1024                 dirty = 1;
1025                 goto out;
1026         }
1027
1028         /* If it looks short enough to be inline data, try to clear extents */
1029         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1030                 max_inline_ea_size = inode_size -
1031                                      (EXT2_GOOD_OLD_INODE_SIZE +
1032                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1033         else
1034                 max_inline_ea_size = 0;
1035         if (EXT2_I_SIZE(inode) <
1036             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1037             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1038                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1039                 dirty = 1;
1040                 goto out;
1041         }
1042
1043         /*
1044          * Too big for inline data, but no evidence of extent tree -
1045          * maybe it's a block map file?  If the mappings all look valid?
1046          */
1047         if (could_be_block_map(fs, inode) &&
1048             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1049 #ifdef WORDS_BIGENDIAN
1050                 int i;
1051
1052                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1053                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1054 #endif
1055
1056                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1057                 dirty = 1;
1058                 goto out;
1059         }
1060
1061         /* Oh well, just clear the busted inode. */
1062         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1063                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1064                 return -1;
1065         }
1066
1067 out:
1068         if (dirty)
1069                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1070
1071         return 0;
1072 }
1073
1074 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1075 {
1076         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1077         dgrp_t start = *group, grp;
1078         blk64_t blocks_to_read = 0;
1079         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1080
1081         if (ctx->readahead_kb == 0)
1082                 goto out;
1083
1084         /* Keep iterating groups until we have enough to readahead */
1085         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1086         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1087                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1088                         continue;
1089                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1090                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1091                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1092                                         inodes_per_block;
1093                 if (blocks_to_read * ctx->fs->blocksize >
1094                     ctx->readahead_kb * 1024)
1095                         break;
1096         }
1097
1098         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1099                                grp - start + 1);
1100         if (err == EAGAIN) {
1101                 ctx->readahead_kb /= 2;
1102                 err = 0;
1103         }
1104
1105 out:
1106         if (err) {
1107                 /* Error; disable itable readahead */
1108                 *group = ctx->fs->group_desc_count;
1109                 *next_ino = ctx->fs->super->s_inodes_count;
1110         } else {
1111                 /*
1112                  * Don't do more readahead until we've reached the first inode
1113                  * of the last inode scan buffer block for the last group.
1114                  */
1115                 *group = grp + 1;
1116                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1117                                      ctx->inode_buffer_blocks :
1118                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1119                                     ctx->fs->blocksize /
1120                                     EXT2_INODE_SIZE(ctx->fs->super);
1121                 inodes_in_group--;
1122                 *next_ino = inodes_in_group -
1123                             (inodes_in_group % inodes_per_buffer) + 1 +
1124                             (grp * ctx->fs->super->s_inodes_per_group);
1125         }
1126 }
1127
1128 /*
1129  * Check if the passed ino is one of the used superblock quota inodes.
1130  *
1131  * Before the quota inodes were journaled, older superblock quota inodes
1132  * were just regular files in the filesystem and not reserved inodes.  This
1133  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1134  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1135  */
1136 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1137 {
1138         enum quota_type qtype;
1139
1140         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1141                 if (*quota_sb_inump(sb, qtype) == ino)
1142                         return 1;
1143
1144         return 0;
1145 }
1146
1147 /*
1148  * Check if the passed ino is one of the reserved quota inodes.
1149  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1150  * inodes.  These inodes may or may not be in use by the quota feature.
1151  */
1152 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1153 {
1154         enum quota_type qtype;
1155
1156         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1157                 if (quota_type2inum(qtype, fs->super) == ino)
1158                         return 1;
1159
1160         return 0;
1161 }
1162
1163 static int e2fsck_should_abort(e2fsck_t ctx)
1164 {
1165         e2fsck_t global_ctx;
1166
1167         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1168                 return 1;
1169
1170         if (ctx->global_ctx) {
1171                 global_ctx = ctx->global_ctx;
1172                 if (global_ctx->flags & E2F_FLAG_SIGNAL_MASK)
1173                         return 1;
1174         }
1175         return 0;
1176 }
1177
1178 void e2fsck_pass1_run(e2fsck_t ctx)
1179 {
1180         int     i;
1181         ext2_filsys fs = ctx->fs;
1182         ext2_ino_t      ino = 0;
1183         struct ext2_inode *inode = NULL;
1184         ext2_inode_scan scan = NULL;
1185         char            *block_buf = NULL;
1186 #ifdef RESOURCE_TRACK
1187         struct resource_track   rtrack;
1188 #endif
1189         unsigned char   frag, fsize;
1190         struct          problem_context pctx;
1191         struct          scan_callback_struct scan_struct;
1192         struct ext2_super_block *sb = ctx->fs->super;
1193         const char      *old_op;
1194         const char      *eop_next_inode = _("getting next inode from scan");
1195         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1196         int             low_dtime_check = 1;
1197         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1198         unsigned int    bufsize;
1199         int             failed_csum = 0;
1200         ext2_ino_t      ino_threshold = 0;
1201         dgrp_t          ra_group = 0;
1202         struct ea_quota ea_ibody_quota;
1203         struct process_inode_block *inodes_to_process;
1204         int             process_inode_count;
1205
1206         init_resource_track(&rtrack, ctx->fs->io);
1207         clear_problem_context(&pctx);
1208
1209         /* If we can do readahead, figure out how many groups to pull in. */
1210         if (!e2fsck_can_readahead(ctx->fs))
1211                 ctx->readahead_kb = 0;
1212         else if (ctx->readahead_kb == ~0ULL)
1213                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1214         pass1_readahead(ctx, &ra_group, &ino_threshold);
1215
1216         if (!(ctx->options & E2F_OPT_PREEN))
1217                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1218
1219         if (ext2fs_has_feature_dir_index(fs->super) &&
1220             !(ctx->options & E2F_OPT_NO)) {
1221                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1222                         ctx->dirs_to_hash = 0;
1223         }
1224
1225 #ifdef MTRACE
1226         mtrace_print("Pass 1");
1227 #endif
1228
1229         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1230         extent_fs = ext2fs_has_feature_extents(sb);
1231         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1232         casefold_fs = ext2fs_has_feature_casefold(sb);
1233
1234         /*
1235          * Allocate bitmaps structures
1236          */
1237         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1238                                                     EXT2FS_BMAP64_RBTREE,
1239                                                     "inode_used_map",
1240                                                     &ctx->inode_used_map);
1241         if (pctx.errcode) {
1242                 pctx.num = 1;
1243                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1244                 ctx->flags |= E2F_FLAG_ABORT;
1245                 return;
1246         }
1247         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1248                         _("directory inode map"),
1249                         ctx->global_ctx ? EXT2FS_BMAP64_RBTREE :
1250                         EXT2FS_BMAP64_AUTODIR,
1251                         "inode_dir_map", &ctx->inode_dir_map);
1252         if (pctx.errcode) {
1253                 pctx.num = 2;
1254                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1255                 ctx->flags |= E2F_FLAG_ABORT;
1256                 return;
1257         }
1258         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1259                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1260                         "inode_reg_map", &ctx->inode_reg_map);
1261         if (pctx.errcode) {
1262                 pctx.num = 6;
1263                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1264                 ctx->flags |= E2F_FLAG_ABORT;
1265                 return;
1266         }
1267         pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1268                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1269                         "block_found_map", &ctx->block_found_map);
1270         if (pctx.errcode) {
1271                 pctx.num = 1;
1272                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1273                 ctx->flags |= E2F_FLAG_ABORT;
1274                 return;
1275         }
1276         pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1277                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1278                         "block_metadata_map", &ctx->block_metadata_map);
1279         if (pctx.errcode) {
1280                 pctx.num = 1;
1281                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1282                 ctx->flags |= E2F_FLAG_ABORT;
1283                 return;
1284         }
1285         if (casefold_fs) {
1286                 pctx.errcode =
1287                         e2fsck_allocate_inode_bitmap(fs,
1288                                                      _("inode casefold map"),
1289                                                      EXT2FS_BMAP64_RBTREE,
1290                                                      "inode_casefold_map",
1291                                                      &ctx->inode_casefold_map);
1292                 if (pctx.errcode) {
1293                         pctx.num = 1;
1294                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1295                         ctx->flags |= E2F_FLAG_ABORT;
1296                         return;
1297                 }
1298         }
1299         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1300                                            &ctx->inode_link_info);
1301         if (pctx.errcode) {
1302                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1303                 ctx->flags |= E2F_FLAG_ABORT;
1304                 return;
1305         }
1306         bufsize = inode_size;
1307         if (bufsize < sizeof(struct ext2_inode_large))
1308                 bufsize = sizeof(struct ext2_inode_large);
1309         inode = (struct ext2_inode *)
1310                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1311
1312         inodes_to_process = (struct process_inode_block *)
1313                 e2fsck_allocate_memory(ctx,
1314                                        (ctx->process_inode_size *
1315                                         sizeof(struct process_inode_block)),
1316                                        "array of inodes to process");
1317         process_inode_count = 0;
1318
1319         pctx.errcode = ext2fs_init_dblist(fs, 0);
1320         if (pctx.errcode) {
1321                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1322                 ctx->flags |= E2F_FLAG_ABORT;
1323                 goto endit;
1324         }
1325
1326         /*
1327          * If the last orphan field is set, clear it, since the pass1
1328          * processing will automatically find and clear the orphans.
1329          * In the future, we may want to try using the last_orphan
1330          * linked list ourselves, but for now, we clear it so that the
1331          * ext3 mount code won't get confused.
1332          */
1333         if (!(ctx->options & E2F_OPT_READONLY)) {
1334                 if (fs->super->s_last_orphan) {
1335                         fs->super->s_last_orphan = 0;
1336                         ext2fs_mark_super_dirty(fs);
1337                 }
1338         }
1339
1340         mark_table_blocks(ctx);
1341         pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1342                                                 &ctx->block_found_map);
1343         if (pctx.errcode) {
1344                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1345                 ctx->flags |= E2F_FLAG_ABORT;
1346                 goto endit;
1347         }
1348         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1349                                                     "block iterate buffer");
1350         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1351                 e2fsck_use_inode_shortcuts(ctx, 1);
1352         e2fsck_intercept_block_allocations(ctx);
1353         old_op = ehandler_operation(_("opening inode scan"));
1354         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1355                                               &scan);
1356         ehandler_operation(old_op);
1357         if (pctx.errcode) {
1358                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1359                 ctx->flags |= E2F_FLAG_ABORT;
1360                 goto endit;
1361         }
1362         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1363                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1364         ctx->stashed_inode = inode;
1365         scan_struct.ctx = ctx;
1366         scan_struct.block_buf = block_buf;
1367         scan_struct.inodes_to_process = inodes_to_process;
1368         scan_struct.process_inode_count = &process_inode_count;
1369         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1370         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1371                                               ctx->fs->group_desc_count)))
1372                 goto endit;
1373         if ((fs->super->s_wtime &&
1374              fs->super->s_wtime < fs->super->s_inodes_count) ||
1375             (fs->super->s_mtime &&
1376              fs->super->s_mtime < fs->super->s_inodes_count) ||
1377             (fs->super->s_mkfs_time &&
1378              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1379                 low_dtime_check = 0;
1380
1381         if (ext2fs_has_feature_mmp(fs->super) &&
1382             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1383             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1384                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1385                                           fs->super->s_mmp_block);
1386
1387         /* Set up ctx->lost_and_found if possible */
1388         (void) e2fsck_get_lost_and_found(ctx, 0);
1389
1390 #ifdef HAVE_PTHREAD
1391         if (ctx->global_ctx) {
1392                 if (ctx->options & E2F_OPT_DEBUG &&
1393                     ctx->options & E2F_OPT_MULTITHREAD)
1394                         fprintf(stderr, "thread %d jumping to group %d\n",
1395                                         ctx->thread_info.et_thread_index,
1396                                         ctx->thread_info.et_group_start);
1397                 pctx.errcode = ext2fs_inode_scan_goto_blockgroup(scan,
1398                                         ctx->thread_info.et_group_start);
1399                 if (pctx.errcode) {
1400                         fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1401                         ctx->flags |= E2F_FLAG_ABORT;
1402                         goto endit;
1403                 }
1404         }
1405 #endif
1406
1407         while (1) {
1408                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1409                         if (e2fsck_mmp_update(fs))
1410                                 fatal_error(ctx, 0);
1411                 }
1412                 old_op = ehandler_operation(eop_next_inode);
1413                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1414                                                           inode, inode_size);
1415                 if (ino > ino_threshold)
1416                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1417                 ehandler_operation(old_op);
1418                 if (e2fsck_should_abort(ctx))
1419                         goto endit;
1420                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1421                         /*
1422                          * If badblocks says badblocks is bad, offer to clear
1423                          * the list, update the in-core bb list, and restart
1424                          * the inode scan.
1425                          */
1426                         if (ino == EXT2_BAD_INO &&
1427                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1428                                         &pctx)) {
1429                                 errcode_t err;
1430
1431                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1432                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1433                                 ctx->fs->badblocks = NULL;
1434                                 err = ext2fs_read_bb_inode(ctx->fs,
1435                                                         &ctx->fs->badblocks);
1436                                 if (err) {
1437                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1438                                                     &pctx);
1439                                         ctx->flags |= E2F_FLAG_ABORT;
1440                                 } else
1441                                         ctx->flags |= E2F_FLAG_RESTART;
1442                                 goto endit;
1443                         }
1444                         if (!ctx->inode_bb_map)
1445                                 alloc_bb_map(ctx);
1446                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1447                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1448                         continue;
1449                 }
1450                 if (pctx.errcode == EXT2_ET_SCAN_FINISHED)
1451                         break;
1452                 if (pctx.errcode &&
1453                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1454                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1455                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1456                         ctx->flags |= E2F_FLAG_ABORT;
1457                         goto endit;
1458                 }
1459                 if (!ino)
1460                         break;
1461 #ifdef HAVE_PTHREAD
1462                 if (ctx->global_ctx)
1463                         ctx->thread_info.et_inode_number++;
1464 #endif
1465                 pctx.ino = ino;
1466                 pctx.inode = inode;
1467                 ctx->stashed_ino = ino;
1468
1469                 /* Clear trashed inode? */
1470                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1471                     inode->i_links_count > 0 &&
1472                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1473                         pctx.errcode = 0;
1474                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1475                 }
1476                 failed_csum = pctx.errcode != 0;
1477
1478                 /*
1479                  * Check for inodes who might have been part of the
1480                  * orphaned list linked list.  They should have gotten
1481                  * dealt with by now, unless the list had somehow been
1482                  * corrupted.
1483                  *
1484                  * FIXME: In the future, inodes which are still in use
1485                  * (and which are therefore) pending truncation should
1486                  * be handled specially.  Right now we just clear the
1487                  * dtime field, and the normal e2fsck handling of
1488                  * inodes where i_size and the inode blocks are
1489                  * inconsistent is to fix i_size, instead of releasing
1490                  * the extra blocks.  This won't catch the inodes that
1491                  * was at the end of the orphan list, but it's better
1492                  * than nothing.  The right answer is that there
1493                  * shouldn't be any bugs in the orphan list handling.  :-)
1494                  */
1495                 if (inode->i_dtime && low_dtime_check &&
1496                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1497                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1498                                 inode->i_dtime = inode->i_links_count ?
1499                                         0 : ctx->now;
1500                                 e2fsck_write_inode(ctx, ino, inode,
1501                                                    "pass1");
1502                                 failed_csum = 0;
1503                         }
1504                 }
1505
1506                 if (inode->i_links_count) {
1507                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1508                                            ino, inode->i_links_count);
1509                         if (pctx.errcode) {
1510                                 pctx.num = inode->i_links_count;
1511                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1512                                 ctx->flags |= E2F_FLAG_ABORT;
1513                                 goto endit;
1514                         }
1515                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1516                            !quota_inum_is_reserved(fs, ino)) {
1517                         if (!inode->i_dtime && inode->i_mode) {
1518                                 if (fix_problem(ctx,
1519                                             PR_1_ZERO_DTIME, &pctx)) {
1520                                         inode->i_dtime = ctx->now;
1521                                         e2fsck_write_inode(ctx, ino, inode,
1522                                                            "pass1");
1523                                         failed_csum = 0;
1524                                 }
1525                         }
1526                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1527                         continue;
1528                 }
1529
1530                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1531                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1532                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1533                      (!casefold_fs &&
1534                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1535                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1536                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1537                 }
1538
1539                 /* Conflicting inlinedata/extents inode flags? */
1540                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1541                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1542                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1543                                                                inode_size,
1544                                                                &pctx);
1545                         if (res < 0) {
1546                                 /* skip FINISH_INODE_LOOP */
1547                                 continue;
1548                         }
1549                 }
1550
1551                 /* Test for incorrect inline_data flags settings. */
1552                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1553                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1554                         size_t size = 0;
1555
1556                         pctx.errcode = get_inline_data_ea_size(fs, ino, inode,
1557                                                                &size);
1558                         if (!pctx.errcode &&
1559                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1560                                 ext2fs_set_feature_inline_data(sb);
1561                                 ext2fs_mark_super_dirty(fs);
1562                                 inlinedata_fs = 1;
1563                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1564                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1565                                 /* skip FINISH_INODE_LOOP */
1566                                 continue;
1567                         }
1568                 }
1569
1570                 /* Test for inline data flag but no attr */
1571                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1572                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1573                         size_t size = 0;
1574                         errcode_t err;
1575                         int flags;
1576
1577                         flags = fs->flags;
1578                         if (failed_csum)
1579                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1580                         err = get_inline_data_ea_size(fs, ino, inode, &size);
1581                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1582                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1583
1584                         switch (err) {
1585                         case 0:
1586                                 /* Everything is awesome... */
1587                                 break;
1588                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1589                         case EXT2_ET_BAD_EA_HASH:
1590                         case EXT2_ET_BAD_EA_HEADER:
1591                         case EXT2_ET_EA_BAD_NAME_LEN:
1592                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1593                         case EXT2_ET_EA_KEY_NOT_FOUND:
1594                         case EXT2_ET_EA_NO_SPACE:
1595                         case EXT2_ET_MISSING_EA_FEATURE:
1596                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1597                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1598                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1599                         case EXT2_ET_NO_INLINE_DATA:
1600                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1601                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1602                         case EXT2_ET_EA_INODE_CORRUPTED:
1603                                 /* broken EA or no system.data EA; truncate */
1604                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1605                                                 &pctx)) {
1606                                         err = ext2fs_inode_size_set(fs, inode, 0);
1607                                         if (err) {
1608                                                 pctx.errcode = err;
1609                                                 ctx->flags |= E2F_FLAG_ABORT;
1610                                                 goto endit;
1611                                         }
1612                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1613                                         memset(&inode->i_block, 0,
1614                                                sizeof(inode->i_block));
1615                                         e2fsck_write_inode(ctx, ino, inode,
1616                                                            "pass1");
1617                                         failed_csum = 0;
1618                                 }
1619                                 break;
1620                         default:
1621                                 /* Some other kind of non-xattr error? */
1622                                 pctx.errcode = err;
1623                                 ctx->flags |= E2F_FLAG_ABORT;
1624                                 goto endit;
1625                         }
1626                 }
1627
1628                 /*
1629                  * Test for incorrect extent flag settings.
1630                  *
1631                  * On big-endian machines we must be careful:
1632                  * When the inode is read, the i_block array is not swapped
1633                  * if the extent flag is set.  Therefore if we are testing
1634                  * for or fixing a wrongly-set flag, we must potentially
1635                  * (un)swap before testing, or after fixing.
1636                  */
1637
1638                 /*
1639                  * In this case the extents flag was set when read, so
1640                  * extent_header_verify is ok.  If the inode is cleared,
1641                  * no need to swap... so no extra swapping here.
1642                  */
1643                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1644                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1645                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1646                         if ((ext2fs_extent_header_verify(inode->i_block,
1647                                                  sizeof(inode->i_block)) == 0) &&
1648                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1649                                 ext2fs_set_feature_extents(sb);
1650                                 ext2fs_mark_super_dirty(fs);
1651                                 extent_fs = 1;
1652                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1653                         clear_inode:
1654                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1655                                 if (ino == EXT2_BAD_INO)
1656                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1657                                                                  ino);
1658                                 /* skip FINISH_INODE_LOOP */
1659                                 continue;
1660                         }
1661                 }
1662
1663                 /*
1664                  * For big-endian machines:
1665                  * If the inode didn't have the extents flag set when it
1666                  * was read, then the i_blocks array was swapped.  To test
1667                  * as an extents header, we must swap it back first.
1668                  * IF we then set the extents flag, the entire i_block
1669                  * array must be un/re-swapped to make it proper extents data.
1670                  */
1671                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1672                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1673                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1674                     (LINUX_S_ISREG(inode->i_mode) ||
1675                      LINUX_S_ISDIR(inode->i_mode))) {
1676                         void *ehp;
1677 #ifdef WORDS_BIGENDIAN
1678                         __u32 tmp_block[EXT2_N_BLOCKS];
1679
1680                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1681                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1682                         ehp = tmp_block;
1683 #else
1684                         ehp = inode->i_block;
1685 #endif
1686                         if ((ext2fs_extent_header_verify(ehp,
1687                                          sizeof(inode->i_block)) == 0) &&
1688                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1689                                 inode->i_flags |= EXT4_EXTENTS_FL;
1690 #ifdef WORDS_BIGENDIAN
1691                                 memcpy(inode->i_block, tmp_block,
1692                                        sizeof(inode->i_block));
1693 #endif
1694                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1695                                 failed_csum = 0;
1696                         }
1697                 }
1698
1699                 if (ino == EXT2_BAD_INO) {
1700                         struct process_block_struct pb;
1701
1702                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1703                              inode->i_gid || inode->i_links_count ||
1704                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1705                              inode->i_file_acl) &&
1706                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1707                                 memset(inode, 0, sizeof(struct ext2_inode));
1708                                 e2fsck_write_inode(ctx, ino, inode,
1709                                                    "clear bad inode");
1710                                 failed_csum = 0;
1711                         }
1712
1713                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1714                                                           &pb.fs_meta_blocks);
1715                         if (pctx.errcode) {
1716                                 pctx.num = 4;
1717                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1718                                 ctx->flags |= E2F_FLAG_ABORT;
1719                                 goto endit;
1720                         }
1721                         pb.ino = EXT2_BAD_INO;
1722                         pb.num_blocks = pb.last_block = 0;
1723                         pb.last_db_block = -1;
1724                         pb.num_illegal_blocks = 0;
1725                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1726                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1727                         pb.inode = inode;
1728                         pb.pctx = &pctx;
1729                         pb.ctx = ctx;
1730                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1731                                      block_buf, process_bad_block, &pb);
1732                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1733                         if (pctx.errcode) {
1734                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1735                                 ctx->flags |= E2F_FLAG_ABORT;
1736                                 goto endit;
1737                         }
1738                         if (pb.bbcheck)
1739                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1740                                 ctx->flags |= E2F_FLAG_ABORT;
1741                                 goto endit;
1742                         }
1743                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1744                         clear_problem_context(&pctx);
1745                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1746                         continue;
1747                 } else if (ino == EXT2_ROOT_INO) {
1748                         /*
1749                          * Make sure the root inode is a directory; if
1750                          * not, offer to clear it.  It will be
1751                          * regenerated in pass #3.
1752                          */
1753                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1754                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1755                                         goto clear_inode;
1756                         }
1757                         /*
1758                          * If dtime is set, offer to clear it.  mke2fs
1759                          * version 0.2b created filesystems with the
1760                          * dtime field set for the root and lost+found
1761                          * directories.  We won't worry about
1762                          * /lost+found, since that can be regenerated
1763                          * easily.  But we will fix the root directory
1764                          * as a special case.
1765                          */
1766                         if (inode->i_dtime && inode->i_links_count) {
1767                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1768                                         inode->i_dtime = 0;
1769                                         e2fsck_write_inode(ctx, ino, inode,
1770                                                            "pass1");
1771                                         failed_csum = 0;
1772                                 }
1773                         }
1774                 } else if (ino == EXT2_JOURNAL_INO) {
1775                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1776                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1777                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1778                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1779                                                 &pctx)) {
1780                                         inode->i_mode = LINUX_S_IFREG;
1781                                         e2fsck_write_inode(ctx, ino, inode,
1782                                                            "pass1");
1783                                         failed_csum = 0;
1784                                 }
1785                                 check_blocks(ctx, &pctx, block_buf, NULL);
1786                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1787                                 continue;
1788                         }
1789                         if ((inode->i_links_count ||
1790                              inode->i_blocks || inode->i_block[0]) &&
1791                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1792                                         &pctx)) {
1793                                 memset(inode, 0, inode_size);
1794                                 ext2fs_icount_store(ctx->inode_link_info,
1795                                                     ino, 0);
1796                                 e2fsck_write_inode_full(ctx, ino, inode,
1797                                                         inode_size, "pass1");
1798                                 failed_csum = 0;
1799                         }
1800                 } else if (quota_inum_is_reserved(fs, ino)) {
1801                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1802                         if (ext2fs_has_feature_quota(fs->super) &&
1803                             quota_inum_is_super(fs->super, ino)) {
1804                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1805                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1806                                                         &pctx)) {
1807                                         inode->i_mode = LINUX_S_IFREG;
1808                                         e2fsck_write_inode(ctx, ino, inode,
1809                                                         "pass1");
1810                                         failed_csum = 0;
1811                                 }
1812                                 check_blocks(ctx, &pctx, block_buf, NULL);
1813                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1814                                 continue;
1815                         }
1816                         if ((inode->i_links_count ||
1817                              inode->i_blocks || inode->i_block[0]) &&
1818                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1819                                         &pctx)) {
1820                                 memset(inode, 0, inode_size);
1821                                 ext2fs_icount_store(ctx->inode_link_info,
1822                                                     ino, 0);
1823                                 e2fsck_write_inode_full(ctx, ino, inode,
1824                                                         inode_size, "pass1");
1825                                 failed_csum = 0;
1826                         }
1827                 } else if (ino == fs->super->s_orphan_file_inum) {
1828                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1829                         if (ext2fs_has_feature_orphan_file(fs->super)) {
1830                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1831                                     fix_problem(ctx, PR_1_ORPHAN_FILE_BAD_MODE,
1832                                                 &pctx)) {
1833                                         inode->i_mode = LINUX_S_IFREG;
1834                                         e2fsck_write_inode(ctx, ino, inode,
1835                                                            "pass1");
1836                                         failed_csum = 0;
1837                                 }
1838                                 check_blocks(ctx, &pctx, block_buf, NULL);
1839                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1840                                 continue;
1841                         }
1842                         if ((inode->i_links_count ||
1843                              inode->i_blocks || inode->i_block[0]) &&
1844                             fix_problem(ctx, PR_1_ORPHAN_FILE_NOT_CLEAR,
1845                                         &pctx)) {
1846                                 memset(inode, 0, inode_size);
1847                                 ext2fs_icount_store(ctx->inode_link_info, ino,
1848                                                     0);
1849                                 e2fsck_write_inode_full(ctx, ino, inode,
1850                                                         inode_size, "pass1");
1851                                 failed_csum = 0;
1852                         }
1853                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1854                         problem_t problem = 0;
1855
1856                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1857                         if (ino == EXT2_BOOT_LOADER_INO) {
1858                                 if (LINUX_S_ISDIR(inode->i_mode))
1859                                         problem = PR_1_RESERVED_BAD_MODE;
1860                         } else if (ino == EXT2_RESIZE_INO) {
1861                                 if (inode->i_mode &&
1862                                     !LINUX_S_ISREG(inode->i_mode))
1863                                         problem = PR_1_RESERVED_BAD_MODE;
1864                         } else {
1865                                 if (inode->i_mode != 0)
1866                                         problem = PR_1_RESERVED_BAD_MODE;
1867                         }
1868                         if (problem) {
1869                                 if (fix_problem(ctx, problem, &pctx)) {
1870                                         inode->i_mode = 0;
1871                                         e2fsck_write_inode(ctx, ino, inode,
1872                                                            "pass1");
1873                                         failed_csum = 0;
1874                                 }
1875                         }
1876                         check_blocks(ctx, &pctx, block_buf, NULL);
1877                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1878                         continue;
1879                 }
1880
1881                 if (!inode->i_links_count) {
1882                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1883                         continue;
1884                 }
1885                 /*
1886                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1887                  * deleted files.  Oops.
1888                  *
1889                  * Since all new ext2 implementations get this right,
1890                  * we now assume that the case of non-zero
1891                  * i_links_count and non-zero dtime means that we
1892                  * should keep the file, not delete it.
1893                  *
1894                  */
1895                 if (inode->i_dtime) {
1896                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1897                                 inode->i_dtime = 0;
1898                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1899                                 failed_csum = 0;
1900                         }
1901                 }
1902
1903                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1904                 switch (fs->super->s_creator_os) {
1905                     case EXT2_OS_HURD:
1906                         frag = inode->osd2.hurd2.h_i_frag;
1907                         fsize = inode->osd2.hurd2.h_i_fsize;
1908                         break;
1909                     default:
1910                         frag = fsize = 0;
1911                 }
1912
1913                 if (inode->i_faddr || frag || fsize ||
1914                     (!ext2fs_has_feature_largedir(fs->super) &&
1915                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1916                         mark_inode_bad(ctx, ino);
1917                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1918                     !ext2fs_has_feature_64bit(fs->super) &&
1919                     inode->osd2.linux2.l_i_file_acl_high != 0)
1920                         mark_inode_bad(ctx, ino);
1921                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1922                     !ext2fs_has_feature_huge_file(fs->super) &&
1923                     (inode->osd2.linux2.l_i_blocks_hi != 0))
1924                         mark_inode_bad(ctx, ino);
1925                 if (inode->i_flags & EXT2_IMAGIC_FL) {
1926                         if (imagic_fs) {
1927                                 if (!ctx->inode_imagic_map)
1928                                         alloc_imagic_map(ctx);
1929                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1930                                                          ino);
1931                         } else {
1932                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1933                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
1934                                         e2fsck_write_inode(ctx, ino,
1935                                                            inode, "pass1");
1936                                         failed_csum = 0;
1937                                 }
1938                         }
1939                 }
1940
1941                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1942                 check_is_really_dir(ctx, &pctx, block_buf);
1943
1944                 /*
1945                  * ext2fs_inode_has_valid_blocks2 does not actually look
1946                  * at i_block[] values, so not endian-sensitive here.
1947                  */
1948                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1949                     LINUX_S_ISLNK(inode->i_mode) &&
1950                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1951                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1952                         inode->i_flags &= ~EXT4_EXTENTS_FL;
1953                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1954                         failed_csum = 0;
1955                 }
1956
1957                 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
1958                     add_encrypted_file(ctx, &pctx) < 0)
1959                         goto clear_inode;
1960
1961                 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
1962                         ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
1963
1964                 if (LINUX_S_ISDIR(inode->i_mode)) {
1965                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1966                         e2fsck_add_dir_info(ctx, ino, 0);
1967                         ctx->fs_directory_count++;
1968                         if (inode->i_flags & EXT4_CASEFOLD_FL)
1969                                 add_casefolded_dir(ctx, ino);
1970                 } else if (LINUX_S_ISREG (inode->i_mode)) {
1971                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1972                         ctx->fs_regular_count++;
1973                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1974                            e2fsck_pass1_check_device_inode(fs, inode)) {
1975                         check_extents_inlinedata(ctx, &pctx);
1976                         check_immutable(ctx, &pctx);
1977                         check_size(ctx, &pctx);
1978                         ctx->fs_chardev_count++;
1979                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1980                            e2fsck_pass1_check_device_inode(fs, inode)) {
1981                         check_extents_inlinedata(ctx, &pctx);
1982                         check_immutable(ctx, &pctx);
1983                         check_size(ctx, &pctx);
1984                         ctx->fs_blockdev_count++;
1985                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1986                            e2fsck_pass1_check_symlink(fs, ino, inode,
1987                                                       block_buf)) {
1988                         check_immutable(ctx, &pctx);
1989                         ctx->fs_symlinks_count++;
1990                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1991                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1992                                 continue;
1993                         } else if (ext2fs_is_fast_symlink(inode)) {
1994                                 ctx->fs_fast_symlinks_count++;
1995                                 check_blocks(ctx, &pctx, block_buf,
1996                                              &ea_ibody_quota);
1997                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1998                                 continue;
1999                         }
2000                 }
2001                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
2002                          e2fsck_pass1_check_device_inode(fs, inode)) {
2003                         check_extents_inlinedata(ctx, &pctx);
2004                         check_immutable(ctx, &pctx);
2005                         check_size(ctx, &pctx);
2006                         ctx->fs_fifo_count++;
2007                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
2008                            e2fsck_pass1_check_device_inode(fs, inode)) {
2009                         check_extents_inlinedata(ctx, &pctx);
2010                         check_immutable(ctx, &pctx);
2011                         check_size(ctx, &pctx);
2012                         ctx->fs_sockets_count++;
2013                 } else
2014                         mark_inode_bad(ctx, ino);
2015                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2016                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
2017                         if (inode->i_block[EXT2_IND_BLOCK])
2018                                 ctx->fs_ind_count++;
2019                         if (inode->i_block[EXT2_DIND_BLOCK])
2020                                 ctx->fs_dind_count++;
2021                         if (inode->i_block[EXT2_TIND_BLOCK])
2022                                 ctx->fs_tind_count++;
2023                 }
2024                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2025                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
2026                     (inode->i_block[EXT2_IND_BLOCK] ||
2027                      inode->i_block[EXT2_DIND_BLOCK] ||
2028                      inode->i_block[EXT2_TIND_BLOCK] ||
2029                      ext2fs_file_acl_block(fs, inode))) {
2030                         struct process_inode_block *itp;
2031
2032                         itp = &inodes_to_process[process_inode_count];
2033                         itp->ino = ino;
2034                         itp->ea_ibody_quota = ea_ibody_quota;
2035                         if (inode_size < sizeof(struct ext2_inode_large))
2036                                 memcpy(&itp->inode, inode, inode_size);
2037                         else
2038                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
2039                         process_inode_count++;
2040                 } else
2041                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
2042
2043                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2044
2045                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2046                         goto endit;
2047
2048                 if (process_inode_count >= ctx->process_inode_size) {
2049                         process_inodes(ctx, block_buf, inodes_to_process,
2050                                        &process_inode_count);
2051
2052                         if (e2fsck_should_abort(ctx))
2053                                 goto endit;
2054                 }
2055         }
2056         process_inodes(ctx, block_buf, inodes_to_process,
2057                        &process_inode_count);
2058         ext2fs_close_inode_scan(scan);
2059         scan = NULL;
2060
2061         reserve_block_for_root_repair(ctx);
2062         reserve_block_for_lnf_repair(ctx);
2063
2064         /*
2065          * If any extended attribute blocks' reference counts need to
2066          * be adjusted, either up (ctx->refcount_extra), or down
2067          * (ctx->refcount), then fix them.
2068          */
2069         if (ctx->refcount) {
2070                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
2071                 ea_refcount_free(ctx->refcount);
2072                 ctx->refcount = 0;
2073         }
2074         if (ctx->refcount_extra) {
2075                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
2076                                         block_buf, +1);
2077                 ea_refcount_free(ctx->refcount_extra);
2078                 ctx->refcount_extra = 0;
2079         }
2080
2081         if (ctx->ea_block_quota_blocks) {
2082                 ea_refcount_free(ctx->ea_block_quota_blocks);
2083                 ctx->ea_block_quota_blocks = 0;
2084         }
2085
2086         if (ctx->ea_block_quota_inodes) {
2087                 ea_refcount_free(ctx->ea_block_quota_inodes);
2088                 ctx->ea_block_quota_inodes = 0;
2089         }
2090
2091         if (ctx->invalid_bitmaps)
2092                 handle_fs_bad_blocks(ctx);
2093
2094         /* We don't need the block_ea_map any more */
2095         if (ctx->block_ea_map) {
2096                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2097                 ctx->block_ea_map = 0;
2098         }
2099
2100         /* We don't need the encryption policy => ID map any more */
2101         destroy_encryption_policy_map(ctx);
2102
2103         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2104                 clear_problem_context(&pctx);
2105                 pctx.errcode = ext2fs_create_resize_inode(fs);
2106                 if (pctx.errcode) {
2107                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2108                                          &pctx)) {
2109                                 ctx->flags |= E2F_FLAG_ABORT;
2110                                 goto endit;
2111                         }
2112                         pctx.errcode = 0;
2113                 }
2114                 if (!pctx.errcode) {
2115                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2116                                           "recreate inode");
2117                         inode->i_mtime = ctx->now;
2118                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2119                                            "recreate inode");
2120                 }
2121                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2122         }
2123
2124         if (ctx->flags & E2F_FLAG_RESTART) {
2125                 /*
2126                  * Only the master copy of the superblock and block
2127                  * group descriptors are going to be written during a
2128                  * restart, so set the superblock to be used to be the
2129                  * master superblock.
2130                  */
2131                 ctx->use_superblock = 0;
2132                 goto endit;
2133         }
2134
2135         if (ctx->large_dirs && !ext2fs_has_feature_largedir(fs->super)) {
2136                 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2137                         ext2fs_set_feature_largedir(fs->super);
2138                         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2139                         ext2fs_mark_super_dirty(fs);
2140                 }
2141                 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2142                     fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2143                         ext2fs_update_dynamic_rev(fs);
2144                         ext2fs_mark_super_dirty(fs);
2145                 }
2146         }
2147
2148         if (ctx->block_dup_map) {
2149                 if (ctx->options & E2F_OPT_PREEN) {
2150                         clear_problem_context(&pctx);
2151                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2152                 }
2153                 e2fsck_pass1_dupblocks(ctx, block_buf);
2154         }
2155         ctx->flags |= E2F_FLAG_ALLOC_OK;
2156 endit:
2157         e2fsck_use_inode_shortcuts(ctx, 0);
2158         ext2fs_free_mem(&inodes_to_process);
2159         inodes_to_process = 0;
2160
2161         if (scan)
2162                 ext2fs_close_inode_scan(scan);
2163         if (block_buf)
2164                 ext2fs_free_mem(&block_buf);
2165         if (inode)
2166                 ext2fs_free_mem(&inode);
2167
2168         /*
2169          * The l+f inode may have been cleared, so zap it now and
2170          * later passes will recalculate it if necessary
2171          */
2172         ctx->lost_and_found = 0;
2173
2174         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2175                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2176         else
2177                 ctx->invalid_bitmaps++;
2178 }
2179
2180 static void init_ext2_max_sizes()
2181 {
2182         int     i;
2183         __u64   max_sizes;
2184
2185         /*
2186          * Init ext2_max_sizes which will be immutable and shared between
2187          * threads
2188          */
2189 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
2190
2191         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
2192                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
2193                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
2194                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
2195                 max_sizes = (max_sizes * (1UL << i));
2196                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
2197         }
2198 #undef EXT2_BPP
2199 }
2200
2201 #ifdef HAVE_PTHREAD
2202 static errcode_t e2fsck_pass1_copy_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2203                                           ext2fs_generic_bitmap *dest)
2204 {
2205         errcode_t ret;
2206
2207         ret = ext2fs_copy_bitmap(*src, dest);
2208         if (ret)
2209                 return ret;
2210
2211         (*dest)->fs = fs;
2212
2213         return 0;
2214 }
2215
2216 static void e2fsck_pass1_free_bitmap(ext2fs_generic_bitmap *bitmap)
2217 {
2218         if (*bitmap) {
2219                 ext2fs_free_generic_bmap(*bitmap);
2220                 *bitmap = NULL;
2221         }
2222
2223 }
2224
2225 static errcode_t e2fsck_pass1_merge_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2226                                           ext2fs_generic_bitmap *dest)
2227 {
2228         errcode_t ret = 0;
2229
2230         if (*src) {
2231                 if (*dest == NULL) {
2232                         *dest = *src;
2233                         *src = NULL;
2234                 } else {
2235                         ret = ext2fs_merge_bitmap(*src, *dest, NULL, NULL);
2236                         if (ret)
2237                                 return ret;
2238                 }
2239                 (*dest)->fs = fs;
2240         }
2241
2242         return 0;
2243 }
2244
2245 static errcode_t e2fsck_pass1_copy_fs(ext2_filsys dest, e2fsck_t src_context,
2246                                       ext2_filsys src)
2247 {
2248         errcode_t       retval;
2249
2250         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2251         dest->inode_map = NULL;
2252         dest->block_map = NULL;
2253         dest->badblocks = NULL;
2254         if (dest->dblist)
2255                 dest->dblist->fs = dest;
2256         if (src->block_map) {
2257                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2258                                                   &dest->block_map);
2259                 if (retval)
2260                         return retval;
2261         }
2262         if (src->inode_map) {
2263                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2264                                                   &dest->inode_map);
2265                 if (retval)
2266                         return retval;
2267         }
2268
2269         if (src->badblocks) {
2270                 retval = ext2fs_badblocks_copy(src->badblocks,
2271                                                &dest->badblocks);
2272                 if (retval)
2273                         return retval;
2274         }
2275
2276         /* disable it for now */
2277         src_context->openfs_flags &= ~EXT2_FLAG_EXCLUSIVE;
2278         retval = ext2fs_open_channel(dest, src_context->io_options,
2279                                      src_context->io_manager,
2280                                      src_context->openfs_flags,
2281                                      src->io->block_size);
2282         if (retval)
2283                 return retval;
2284
2285         /* Block size might not be default */
2286         io_channel_set_blksize(dest->io, src->io->block_size);
2287         ehandler_init(dest->io);
2288
2289         assert(dest->io->magic == src->io->magic);
2290         assert(dest->io->manager == src->io->manager);
2291         assert(strcmp(dest->io->name, src->io->name) == 0);
2292         assert(dest->io->block_size == src->io->block_size);
2293         assert(dest->io->read_error == src->io->read_error);
2294         assert(dest->io->write_error == src->io->write_error);
2295         assert(dest->io->refcount == src->io->refcount);
2296         assert(dest->io->flags == src->io->flags);
2297         assert(dest->io->app_data == dest);
2298         assert(src->io->app_data == src);
2299         assert(dest->io->align == src->io->align);
2300
2301         /* The data should be written to disk immediately */
2302         dest->io->flags |= CHANNEL_FLAGS_WRITETHROUGH;
2303         /* icache will be rebuilt if needed, so do not copy from @src */
2304         src->icache = NULL;
2305         return 0;
2306 }
2307
2308 static int e2fsck_pass1_merge_fs(ext2_filsys dest, ext2_filsys src)
2309 {
2310         struct ext2_inode_cache *icache = dest->icache;
2311         errcode_t retval = 0;
2312         io_channel dest_io;
2313         io_channel dest_image_io;
2314         ext2fs_inode_bitmap inode_map;
2315         ext2fs_block_bitmap block_map;
2316         ext2_badblocks_list badblocks;
2317         ext2_dblist dblist;
2318
2319         dest_io = dest->io;
2320         dest_image_io = dest->image_io;
2321         inode_map = dest->inode_map;
2322         block_map = dest->block_map;
2323         badblocks = dest->badblocks;
2324         dblist = dest->dblist;
2325
2326         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2327         dest->io = dest_io;
2328         dest->image_io = dest_image_io;
2329         dest->icache = icache;
2330         dest->inode_map = inode_map;
2331         dest->block_map = block_map;
2332         dest->badblocks = badblocks;
2333         dest->dblist = dblist;
2334         if (dest->dblist)
2335                 dest->dblist->fs = dest;
2336
2337         if (src->icache) {
2338                 ext2fs_free_inode_cache(src->icache);
2339                 src->icache = NULL;
2340         }
2341
2342         retval = e2fsck_pass1_merge_bitmap(dest, &src->inode_map,
2343                                            &dest->inode_map);
2344         if (retval)
2345                 goto out;
2346
2347         retval = e2fsck_pass1_merge_bitmap(dest, &src->block_map,
2348                                           &dest->block_map);
2349         if (retval)
2350                 goto out;
2351
2352         if (src->dblist) {
2353                 if (dest->dblist) {
2354                         retval = ext2fs_merge_dblist(src->dblist,
2355                                                      dest->dblist);
2356                         if (retval)
2357                                 goto out;
2358                 } else {
2359                         dest->dblist = src->dblist;
2360                         dest->dblist->fs = dest;
2361                         src->dblist = NULL;
2362                 }
2363         }
2364
2365         if (src->badblocks) {
2366                 if (dest->badblocks == NULL)
2367                         retval = ext2fs_badblocks_copy(src->badblocks,
2368                                                        &dest->badblocks);
2369                 else
2370                         retval = ext2fs_badblocks_merge(src->badblocks,
2371                                                         dest->badblocks);
2372         }
2373 out:
2374         io_channel_close(src->io);
2375         if (src->inode_map)
2376                 ext2fs_free_generic_bmap(src->inode_map);
2377         if (src->block_map)
2378                 ext2fs_free_generic_bmap(src->block_map);
2379         if (src->badblocks)
2380                 ext2fs_badblocks_list_free(src->badblocks);
2381         if (src->dblist)
2382                 ext2fs_free_dblist(src->dblist);
2383
2384         return retval;
2385 }
2386
2387 static errcode_t e2fsck_pass1_thread_prepare(e2fsck_t global_ctx, e2fsck_t *thread_ctx,
2388                                              int thread_index, int num_threads)
2389 {
2390         errcode_t               retval;
2391         e2fsck_t                thread_context;
2392         ext2_filsys             thread_fs;
2393         ext2_filsys             global_fs = global_ctx->fs;
2394         struct e2fsck_thread    *tinfo;
2395         dgrp_t                  average_group;
2396
2397         assert(global_ctx->inode_used_map == NULL);
2398         assert(global_ctx->inode_dir_map == NULL);
2399         assert(global_ctx->inode_bb_map == NULL);
2400         assert(global_ctx->inode_imagic_map == NULL);
2401         assert(global_ctx->inode_reg_map == NULL);
2402         assert(global_ctx->inodes_to_rebuild == NULL);
2403
2404         assert(global_ctx->block_found_map == NULL);
2405         assert(global_ctx->block_dup_map == NULL);
2406         assert(global_ctx->block_ea_map == NULL);
2407         assert(global_ctx->block_metadata_map == NULL);
2408         assert(global_ctx->fs->dblist == NULL);
2409
2410         retval = ext2fs_get_mem(sizeof(struct e2fsck_struct), &thread_context);
2411         if (retval) {
2412                 com_err(global_ctx->program_name, retval, "while allocating memory");
2413                 return retval;
2414         }
2415         memcpy(thread_context, global_ctx, sizeof(struct e2fsck_struct));
2416         thread_context->global_ctx = global_ctx;
2417
2418         retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &thread_fs);
2419         if (retval) {
2420                 com_err(global_ctx->program_name, retval, "while allocating memory");
2421                 goto out_context;
2422         }
2423
2424         io_channel_flush_cleanup(global_fs->io);
2425         retval = e2fsck_pass1_copy_fs(thread_fs, global_ctx, global_fs);
2426         if (retval) {
2427                 com_err(global_ctx->program_name, retval, "while copying fs");
2428                 goto out_fs;
2429         }
2430         thread_fs->priv_data = thread_context;
2431
2432         thread_context->thread_info.et_thread_index = thread_index;
2433         set_up_logging(thread_context);
2434
2435         /*
2436          * Distribute work to multiple threads:
2437          * Each thread work on fs->group_desc_count / nthread groups.
2438          */
2439         tinfo = &thread_context->thread_info;
2440         average_group = thread_fs->group_desc_count / num_threads;
2441         if (average_group == 0)
2442                 average_group = 1;
2443         tinfo->et_group_start = average_group * thread_index;
2444         if (thread_index == num_threads - 1)
2445                 tinfo->et_group_end = thread_fs->group_desc_count;
2446         else
2447                 tinfo->et_group_end = average_group * (thread_index + 1);
2448         tinfo->et_group_next = tinfo->et_group_start;
2449         tinfo->et_inode_number = 0;
2450         tinfo->et_log_buf[0] = '\0';
2451         tinfo->et_log_length = 0;
2452         if (thread_context->options & E2F_OPT_MULTITHREAD)
2453                 log_out(thread_context, _("Scan group range [%d, %d)\n"),
2454                         tinfo->et_group_start, tinfo->et_group_end);
2455         thread_context->fs = thread_fs;
2456         *thread_ctx = thread_context;
2457         return 0;
2458 out_fs:
2459         ext2fs_free_mem(&thread_fs);
2460 out_context:
2461         ext2fs_free_mem(&thread_context);
2462         return retval;
2463 }
2464
2465 static void e2fsck_pass1_merge_dir_info(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2466 {
2467         if (thread_ctx->dir_info == NULL)
2468                 return;
2469
2470         if (global_ctx->dir_info == NULL) {
2471                 global_ctx->dir_info = thread_ctx->dir_info;
2472                 thread_ctx->dir_info = NULL;
2473                 return;
2474         }
2475
2476         e2fsck_merge_dir_info(global_ctx, thread_ctx->dir_info,
2477                               global_ctx->dir_info);
2478 }
2479
2480 static inline errcode_t
2481 e2fsck_pass1_merge_icount(ext2_icount_t *dest_icount,
2482                           ext2_icount_t *src_icount)
2483 {
2484         if (*src_icount) {
2485                 if (*dest_icount == NULL) {
2486                         *dest_icount = *src_icount;
2487                         *src_icount = NULL;
2488                 } else {
2489                         errcode_t ret;
2490
2491                         ret = ext2fs_icount_merge(*src_icount,
2492                                                   *dest_icount);
2493                         if (ret)
2494                                 return ret;
2495                 }
2496         }
2497
2498         return 0;
2499 }
2500
2501 static errcode_t e2fsck_pass1_merge_icounts(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2502 {
2503         errcode_t ret;
2504
2505         ret = e2fsck_pass1_merge_icount(&global_ctx->inode_count,
2506                                         &thread_ctx->inode_count);
2507         if (ret)
2508                 return ret;
2509         ret = e2fsck_pass1_merge_icount(&global_ctx->inode_link_info,
2510                                         &thread_ctx->inode_link_info);
2511
2512         return ret;
2513 }
2514
2515 static int e2fsck_pass1_thread_join_one(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2516 {
2517         errcode_t        retval;
2518         int              flags = global_ctx->flags;
2519         ext2_filsys      thread_fs = thread_ctx->fs;
2520         ext2_filsys      global_fs = global_ctx->fs;
2521         FILE            *global_logf = global_ctx->logf;
2522         FILE            *global_problem_logf = global_ctx->problem_logf;
2523         ext2fs_inode_bitmap inode_bad_map = global_ctx->inode_bad_map;
2524         struct dir_info_db *dir_info = global_ctx->dir_info;
2525         ext2fs_inode_bitmap inode_used_map = global_ctx->inode_used_map;
2526         ext2fs_inode_bitmap inode_dir_map = global_ctx->inode_dir_map;
2527         ext2fs_inode_bitmap inode_bb_map = global_ctx->inode_bb_map;
2528         ext2fs_inode_bitmap inode_imagic_map = global_ctx->inode_imagic_map;
2529         ext2fs_inode_bitmap inode_reg_map = global_ctx->inode_reg_map;
2530         ext2fs_block_bitmap block_found_map = global_ctx->block_found_map;
2531         ext2fs_block_bitmap block_dup_map = global_ctx->block_dup_map;
2532         ext2fs_block_bitmap block_ea_map = global_ctx->block_ea_map;
2533         ext2fs_block_bitmap block_metadata_map = global_ctx->block_metadata_map;
2534         ext2fs_block_bitmap inodes_to_rebuild = global_ctx->inodes_to_rebuild;
2535         ext2_icount_t inode_count = global_ctx->inode_count;
2536         ext2_icount_t inode_link_info = global_ctx->inode_link_info;
2537         __u32 fs_directory_count = global_ctx->fs_directory_count;
2538         __u32 fs_regular_count = global_ctx->fs_regular_count;
2539         __u32 fs_blockdev_count = global_ctx->fs_blockdev_count;
2540         __u32 fs_chardev_count = global_ctx->fs_chardev_count;
2541         __u32 fs_links_count = global_ctx->fs_links_count;
2542         __u32 fs_symlinks_count = global_ctx->fs_symlinks_count;
2543         __u32 fs_fast_symlinks_count = global_ctx->fs_fast_symlinks_count;
2544         __u32 fs_fifo_count = global_ctx->fs_fifo_count;
2545         __u32 fs_total_count = global_ctx->fs_total_count;
2546         __u32 fs_badblocks_count = global_ctx->fs_badblocks_count;
2547         __u32 fs_sockets_count = global_ctx->fs_sockets_count;
2548         __u32 fs_ind_count = global_ctx->fs_ind_count;
2549         __u32 fs_dind_count = global_ctx->fs_dind_count;
2550         __u32 fs_tind_count = global_ctx->fs_tind_count;
2551         __u32 fs_fragmented = global_ctx->fs_fragmented;
2552         __u32 fs_fragmented_dir = global_ctx->fs_fragmented_dir;
2553         __u32 large_files = global_ctx->large_files;
2554
2555 #ifdef HAVE_SETJMP_H
2556         jmp_buf          old_jmp;
2557
2558         memcpy(old_jmp, global_ctx->abort_loc, sizeof(jmp_buf));
2559 #endif
2560         memcpy(global_ctx, thread_ctx, sizeof(struct e2fsck_struct));
2561 #ifdef HAVE_SETJMP_H
2562         memcpy(global_ctx->abort_loc, old_jmp, sizeof(jmp_buf));
2563 #endif
2564
2565         global_ctx->inode_used_map = inode_used_map;
2566         global_ctx->inode_bad_map = inode_bad_map;
2567         global_ctx->inode_dir_map = inode_dir_map;
2568         global_ctx->inode_bb_map = inode_bb_map;
2569         global_ctx->inode_imagic_map = inode_imagic_map;
2570         global_ctx->inodes_to_rebuild = inodes_to_rebuild;
2571         global_ctx->inode_reg_map = inode_reg_map;
2572         global_ctx->block_found_map = block_found_map;
2573         global_ctx->block_dup_map = block_dup_map;
2574         global_ctx->block_ea_map = block_ea_map;
2575         global_ctx->block_metadata_map = block_metadata_map;
2576         global_ctx->dir_info = dir_info;
2577         e2fsck_pass1_merge_dir_info(global_ctx, thread_ctx);
2578         global_ctx->inode_count = inode_count;
2579         global_ctx->inode_link_info = inode_link_info;
2580         global_ctx->fs_directory_count += fs_directory_count;
2581         global_ctx->fs_regular_count += fs_regular_count;
2582         global_ctx->fs_blockdev_count += fs_blockdev_count;
2583         global_ctx->fs_chardev_count += fs_chardev_count;
2584         global_ctx->fs_links_count += fs_links_count;
2585         global_ctx->fs_symlinks_count += fs_symlinks_count;
2586         global_ctx->fs_fast_symlinks_count += fs_fast_symlinks_count;
2587         global_ctx->fs_fifo_count += fs_fifo_count;
2588         global_ctx->fs_total_count += fs_total_count;
2589         global_ctx->fs_badblocks_count += fs_badblocks_count;
2590         global_ctx->fs_sockets_count += fs_sockets_count;
2591         global_ctx->fs_ind_count += fs_ind_count;
2592         global_ctx->fs_dind_count += fs_dind_count;
2593         global_ctx->fs_tind_count += fs_tind_count;
2594         global_ctx->fs_fragmented += fs_fragmented;
2595         global_ctx->fs_fragmented_dir += fs_fragmented_dir;
2596         global_ctx->large_files += large_files;
2597
2598         /* Keep the global singal flags*/
2599         global_ctx->flags |= (flags & E2F_FLAG_SIGNAL_MASK) |
2600                              (global_ctx->flags & E2F_FLAG_SIGNAL_MASK);
2601
2602         retval = e2fsck_pass1_merge_fs(global_fs, thread_fs);
2603         if (retval) {
2604                 com_err(global_ctx->program_name, 0, _("while merging fs\n"));
2605                 return retval;
2606         }
2607         global_fs->priv_data = global_ctx;
2608         global_ctx->fs = global_fs;
2609         global_ctx->logf = global_logf;
2610         global_ctx->problem_logf = global_problem_logf;
2611         global_ctx->global_ctx = NULL;
2612         retval = e2fsck_pass1_merge_icounts(global_ctx, thread_ctx);
2613         if (retval) {
2614                 com_err(global_ctx->program_name, 0,
2615                         _("while merging icounts\n"));
2616                 return retval;
2617         }
2618
2619         retval = e2fsck_pass1_merge_bitmap(global_fs,
2620                                 &thread_ctx->inode_used_map,
2621                                 &global_ctx->inode_used_map);
2622         if (retval)
2623                 return retval;
2624
2625         retval = e2fsck_pass1_merge_bitmap(global_fs,
2626                                 &thread_ctx->inode_bad_map,
2627                                 &global_ctx->inode_bad_map);
2628         if (retval)
2629                 return retval;
2630         retval = e2fsck_pass1_merge_bitmap(global_fs,
2631                                         &thread_ctx->inode_dir_map,
2632                                         &global_ctx->inode_dir_map);
2633         if (retval)
2634                 return retval;
2635         retval = e2fsck_pass1_merge_bitmap(global_fs,
2636                                 &thread_ctx->inode_bb_map,
2637                                 &global_ctx->inode_bb_map);
2638         if (retval)
2639                 return retval;
2640         retval = e2fsck_pass1_merge_bitmap(global_fs,
2641                                 &thread_ctx->inode_imagic_map,
2642                                 &global_ctx->inode_imagic_map);
2643         if (retval)
2644                 return retval;
2645         retval = e2fsck_pass1_merge_bitmap(global_fs,
2646                                 &thread_ctx->inode_reg_map,
2647                                 &global_ctx->inode_reg_map);
2648         if (retval)
2649                 return retval;
2650         retval = e2fsck_pass1_merge_bitmap(global_fs,
2651                                 &thread_ctx->inodes_to_rebuild,
2652                                 &global_ctx->inodes_to_rebuild);
2653         if (retval)
2654                 return retval;
2655         retval = e2fsck_pass1_merge_bitmap(global_fs,
2656                                 &thread_ctx->block_found_map,
2657                                 &global_ctx->block_found_map);
2658         if (retval)
2659                 return retval;
2660         retval = e2fsck_pass1_merge_bitmap(global_fs,
2661                                 &thread_ctx->block_dup_map,
2662                                 &global_ctx->block_dup_map);
2663         if (retval)
2664                 return retval;
2665         retval = e2fsck_pass1_merge_bitmap(global_fs,
2666                                 &thread_ctx->block_ea_map,
2667                                 &global_ctx->block_ea_map);
2668         if (retval)
2669                 return retval;
2670         retval = e2fsck_pass1_merge_bitmap(global_fs,
2671                                 &thread_ctx->block_metadata_map,
2672                                 &global_ctx->block_metadata_map);
2673         if (retval)
2674                 return retval;
2675
2676         return 0;
2677 }
2678
2679 static int e2fsck_pass1_thread_join(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2680 {
2681         errcode_t       retval;
2682
2683         retval = e2fsck_pass1_thread_join_one(global_ctx, thread_ctx);
2684         ext2fs_free_mem(&thread_ctx->fs);
2685         if (thread_ctx->logf)
2686                 fclose(thread_ctx->logf);
2687         if (thread_ctx->problem_logf) {
2688                 fputs("</problem_log>\n", thread_ctx->problem_logf);
2689                 fclose(thread_ctx->problem_logf);
2690         }
2691         e2fsck_pass1_free_bitmap(&thread_ctx->inode_used_map);
2692         e2fsck_pass1_free_bitmap(&thread_ctx->inode_bad_map);
2693         e2fsck_pass1_free_bitmap(&thread_ctx->inode_dir_map);
2694         e2fsck_pass1_free_bitmap(&thread_ctx->inode_bb_map);
2695         e2fsck_pass1_free_bitmap(&thread_ctx->inode_imagic_map);
2696         e2fsck_pass1_free_bitmap(&thread_ctx->inode_reg_map);
2697         e2fsck_pass1_free_bitmap(&thread_ctx->inodes_to_rebuild);
2698         e2fsck_pass1_free_bitmap(&thread_ctx->block_found_map);
2699         e2fsck_pass1_free_bitmap(&thread_ctx->block_dup_map);
2700         e2fsck_pass1_free_bitmap(&thread_ctx->block_ea_map);
2701         e2fsck_pass1_free_bitmap(&thread_ctx->block_metadata_map);
2702         e2fsck_free_dir_info(thread_ctx);
2703         ext2fs_free_icount(thread_ctx->inode_count);
2704         ext2fs_free_icount(thread_ctx->inode_link_info);
2705         ext2fs_free_mem(&thread_ctx);
2706
2707         return retval;
2708 }
2709
2710 static int e2fsck_pass1_threads_join(struct e2fsck_thread_info *infos,
2711                                       int num_threads, e2fsck_t global_ctx)
2712 {
2713         errcode_t                        rc;
2714         errcode_t                        ret = 0;
2715         int                              i;
2716         struct e2fsck_thread_info       *pinfo;
2717
2718         for (i = 0; i < num_threads; i++) {
2719                 pinfo = &infos[i];
2720
2721                 if (!pinfo->eti_started)
2722                         continue;
2723
2724                 rc = pthread_join(pinfo->eti_thread_id, NULL);
2725                 if (rc) {
2726                         com_err(global_ctx->program_name, rc,
2727                                 _("while joining thread\n"));
2728                         if (ret == 0)
2729                                 ret = rc;
2730                 }
2731                 rc = e2fsck_pass1_thread_join(global_ctx, infos[i].eti_thread_ctx);
2732                 if (rc) {
2733                         com_err(global_ctx->program_name, rc,
2734                                 _("while joining pass1 thread\n"));
2735                         if (ret == 0)
2736                                 ret = rc;
2737                 }
2738         }
2739         free(infos);
2740
2741         return ret;
2742 }
2743
2744 static void *e2fsck_pass1_thread(void *arg)
2745 {
2746         struct e2fsck_thread_info       *info = arg;
2747         e2fsck_t                         thread_ctx = info->eti_thread_ctx;
2748 #ifdef DEBUG_THREADS
2749         struct e2fsck_thread_debug      *thread_debug = info->eti_debug;
2750 #endif
2751
2752 #ifdef DEBUG_THREADS
2753         pthread_mutex_lock(&thread_debug->etd_mutex);
2754         while (info->eti_thread_index > thread_debug->etd_finished_threads) {
2755                 pthread_cond_wait(&thread_debug->etd_cond,
2756                                   &thread_debug->etd_mutex);
2757         }
2758         pthread_mutex_unlock(&thread_debug->etd_mutex);
2759 #endif
2760
2761 #ifdef HAVE_SETJMP_H
2762         /*
2763          * When fatal_error() happens, jump to here. The thread
2764          * context's flags will be saved, but its abort_loc will
2765          * be overwritten by original jump buffer for the later
2766          * tests.
2767          */
2768         if (setjmp(thread_ctx->abort_loc)) {
2769                 thread_ctx->flags &= ~E2F_FLAG_SETJMP_OK;
2770                 goto out;
2771         }
2772         thread_ctx->flags |= E2F_FLAG_SETJMP_OK;
2773 #endif
2774
2775         e2fsck_pass1_run(thread_ctx);
2776
2777 out:
2778         if (thread_ctx->options & E2F_OPT_MULTITHREAD)
2779                 log_out(thread_ctx,
2780                         _("Scanned group range [%lu, %lu), inodes %lu\n"),
2781                         thread_ctx->thread_info.et_group_start,
2782                         thread_ctx->thread_info.et_group_end,
2783                         thread_ctx->thread_info.et_inode_number);
2784
2785 #ifdef DEBUG_THREADS
2786         pthread_mutex_lock(&thread_debug->etd_mutex);
2787         thread_debug->etd_finished_threads++;
2788         pthread_cond_broadcast(&thread_debug->etd_cond);
2789         pthread_mutex_unlock(&thread_debug->etd_mutex);
2790 #endif
2791
2792         return NULL;
2793 }
2794
2795 static int e2fsck_pass1_threads_start(struct e2fsck_thread_info **pinfo,
2796                                       int num_threads, e2fsck_t global_ctx)
2797 {
2798         struct e2fsck_thread_info       *infos;
2799         pthread_attr_t                   attr;
2800         errcode_t                        retval;
2801         errcode_t                        ret;
2802         struct e2fsck_thread_info       *tmp_pinfo;
2803         int                              i;
2804         e2fsck_t                         thread_ctx;
2805 #ifdef DEBUG_THREADS
2806         struct e2fsck_thread_debug       thread_debug =
2807                 {PTHREAD_MUTEX_INITIALIZER, PTHREAD_COND_INITIALIZER, 0};
2808
2809         thread_debug.etd_finished_threads = 0;
2810 #endif
2811
2812         retval = pthread_attr_init(&attr);
2813         if (retval) {
2814                 com_err(global_ctx->program_name, retval,
2815                         _("while setting pthread attribute\n"));
2816                 return retval;
2817         }
2818
2819         infos = calloc(num_threads, sizeof(struct e2fsck_thread_info));
2820         if (infos == NULL) {
2821                 retval = -ENOMEM;
2822                 com_err(global_ctx->program_name, retval,
2823                         _("while allocating memory for threads\n"));
2824                 pthread_attr_destroy(&attr);
2825                 return retval;
2826         }
2827
2828         for (i = 0; i < num_threads; i++) {
2829                 tmp_pinfo = &infos[i];
2830                 tmp_pinfo->eti_thread_index = i;
2831 #ifdef DEBUG_THREADS
2832                 tmp_pinfo->eti_debug = &thread_debug;
2833 #endif
2834                 retval = e2fsck_pass1_thread_prepare(global_ctx, &thread_ctx,
2835                                                      i, num_threads);
2836                 if (retval) {
2837                         com_err(global_ctx->program_name, retval,
2838                                 _("while preparing pass1 thread\n"));
2839                         break;
2840                 }
2841                 tmp_pinfo->eti_thread_ctx = thread_ctx;
2842
2843                 retval = pthread_create(&tmp_pinfo->eti_thread_id, &attr,
2844                                         &e2fsck_pass1_thread, tmp_pinfo);
2845                 if (retval) {
2846                         com_err(global_ctx->program_name, retval,
2847                                 _("while creating thread\n"));
2848                         e2fsck_pass1_thread_join(global_ctx, thread_ctx);
2849                         break;
2850                 }
2851
2852                 tmp_pinfo->eti_started = 1;
2853         }
2854
2855         /* destroy the thread attribute object, since it is no longer needed */
2856         ret = pthread_attr_destroy(&attr);
2857         if (ret) {
2858                 com_err(global_ctx->program_name, ret,
2859                         _("while destroying thread attribute\n"));
2860                 if (retval == 0)
2861                         retval = ret;
2862         }
2863
2864         if (retval) {
2865                 e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
2866                 return retval;
2867         }
2868         *pinfo = infos;
2869         return 0;
2870 }
2871
2872 static void e2fsck_pass1_multithread(e2fsck_t global_ctx)
2873 {
2874         struct e2fsck_thread_info       *infos = NULL;
2875         int                              num_threads = 1;
2876         errcode_t                        retval;
2877
2878         retval = e2fsck_pass1_threads_start(&infos, num_threads, global_ctx);
2879         if (retval) {
2880                 com_err(global_ctx->program_name, retval,
2881                         _("while starting pass1 threads\n"));
2882                 goto out_abort;
2883         }
2884
2885         retval = e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
2886         if (retval) {
2887                 com_err(global_ctx->program_name, retval,
2888                         _("while joining pass1 threads\n"));
2889                 goto out_abort;
2890         }
2891         return;
2892 out_abort:
2893         global_ctx->flags |= E2F_FLAG_ABORT;
2894         return;
2895 }
2896 #endif
2897
2898 /* TODO: tdb needs to be handled properly for multiple threads*/
2899 static int multiple_threads_supported(e2fsck_t ctx)
2900 {
2901 #ifdef  CONFIG_TDB
2902         unsigned int            threshold;
2903         ext2_ino_t              num_dirs;
2904         errcode_t               retval;
2905         char                    *tdb_dir;
2906         int                     enable;
2907
2908         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
2909                            &tdb_dir);
2910         profile_get_uint(ctx->profile, "scratch_files",
2911                          "numdirs_threshold", 0, 0, &threshold);
2912         profile_get_boolean(ctx->profile, "scratch_files",
2913                             "icount", 0, 1, &enable);
2914
2915         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
2916         if (retval)
2917                 num_dirs = 1024;        /* Guess */
2918
2919         /* tdb is unsupported now */
2920         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
2921             (!threshold || num_dirs > threshold))
2922                 return 0;
2923  #endif
2924         return 1;
2925 }
2926
2927 void e2fsck_pass1(e2fsck_t ctx)
2928 {
2929
2930         init_ext2_max_sizes();
2931 #ifdef HAVE_PTHREAD
2932         if (multiple_threads_supported(ctx)) {
2933                 e2fsck_pass1_multithread(ctx);
2934                 return;
2935         }
2936         fprintf(stderr, "Fall through single thread for pass1 "
2937                         "because tdb could not handle properly\n");
2938 #endif
2939         e2fsck_pass1_run(ctx);
2940 }
2941
2942 #undef FINISH_INODE_LOOP
2943
2944 /*
2945  * When the inode_scan routines call this callback at the end of the
2946  * glock group, call process_inodes.
2947  */
2948 static errcode_t scan_callback(ext2_filsys fs,
2949                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
2950                                dgrp_t group, void * priv_data)
2951 {
2952         struct scan_callback_struct *scan_struct;
2953         e2fsck_t ctx;
2954         struct e2fsck_thread *tinfo;
2955
2956         scan_struct = (struct scan_callback_struct *) priv_data;
2957         ctx = scan_struct->ctx;
2958
2959         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf,
2960                        scan_struct->inodes_to_process,
2961                        scan_struct->process_inode_count);
2962
2963         if (ctx->progress)
2964                 if ((ctx->progress)(ctx, 1, group+1,
2965                                     ctx->fs->group_desc_count))
2966                         return EXT2_ET_CANCEL_REQUESTED;
2967
2968 #ifdef HAVE_PTHREAD
2969         if (ctx->global_ctx) {
2970                 tinfo = &ctx->thread_info;
2971                 tinfo->et_group_next++;
2972                 if (ctx->options & E2F_OPT_DEBUG &&
2973                     ctx->options & E2F_OPT_MULTITHREAD)
2974                         log_out(ctx, _("group %d finished\n"),
2975                                 tinfo->et_group_next);
2976                 if (tinfo->et_group_next >= tinfo->et_group_end)
2977                         return EXT2_ET_SCAN_FINISHED;
2978         }
2979 #endif
2980
2981         return 0;
2982 }
2983
2984 /*
2985  * Process the inodes in the "inodes to process" list.
2986  */
2987 static void process_inodes(e2fsck_t ctx, char *block_buf,
2988                            struct process_inode_block *inodes_to_process,
2989                            int *process_inode_count)
2990 {
2991         int                     i;
2992         struct ext2_inode       *old_stashed_inode;
2993         ext2_ino_t              old_stashed_ino;
2994         const char              *old_operation;
2995         char                    buf[80];
2996         struct problem_context  pctx;
2997
2998 #if 0
2999         printf("begin process_inodes: ");
3000 #endif
3001         if (*process_inode_count == 0)
3002                 return;
3003         old_operation = ehandler_operation(0);
3004         old_stashed_inode = ctx->stashed_inode;
3005         old_stashed_ino = ctx->stashed_ino;
3006         qsort(inodes_to_process, *process_inode_count,
3007                       sizeof(struct process_inode_block), process_inode_cmp);
3008         clear_problem_context(&pctx);
3009         for (i=0; i < *process_inode_count; i++) {
3010                 pctx.inode = ctx->stashed_inode =
3011                         (struct ext2_inode *) &inodes_to_process[i].inode;
3012                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
3013
3014 #if 0
3015                 printf("%u ", pctx.ino);
3016 #endif
3017                 sprintf(buf, _("reading indirect blocks of inode %u"),
3018                         pctx.ino);
3019                 ehandler_operation(buf);
3020                 check_blocks(ctx, &pctx, block_buf,
3021                              &inodes_to_process[i].ea_ibody_quota);
3022                 if (e2fsck_should_abort(ctx))
3023                         break;
3024         }
3025         ctx->stashed_inode = old_stashed_inode;
3026         ctx->stashed_ino = old_stashed_ino;
3027         *process_inode_count = 0;
3028 #if 0
3029         printf("end process inodes\n");
3030 #endif
3031         ehandler_operation(old_operation);
3032 }
3033
3034 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
3035 {
3036         const struct process_inode_block *ib_a =
3037                 (const struct process_inode_block *) a;
3038         const struct process_inode_block *ib_b =
3039                 (const struct process_inode_block *) b;
3040         int     ret;
3041
3042         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
3043                ib_b->inode.i_block[EXT2_IND_BLOCK]);
3044         if (ret == 0)
3045                 /*
3046                  * We only call process_inodes() for non-extent
3047                  * inodes, so it's OK to pass NULL to
3048                  * ext2fs_file_acl_block() here.
3049                  */
3050                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
3051                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
3052         if (ret == 0)
3053                 ret = ib_a->ino - ib_b->ino;
3054         return ret;
3055 }
3056
3057 /*
3058  * Mark an inode as being bad in some what
3059  */
3060 static void mark_inode_bad(e2fsck_t ctx, ext2_ino_t ino)
3061 {
3062         struct          problem_context pctx;
3063
3064         if (!ctx->inode_bad_map) {
3065                 clear_problem_context(&pctx);
3066
3067                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3068                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
3069                                 "inode_bad_map", &ctx->inode_bad_map);
3070                 if (pctx.errcode) {
3071                         pctx.num = 3;
3072                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3073                         /* Should never get here */
3074                         ctx->flags |= E2F_FLAG_ABORT;
3075                         return;
3076                 }
3077         }
3078         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
3079 }
3080
3081 static void add_casefolded_dir(e2fsck_t ctx, ext2_ino_t ino)
3082 {
3083         struct          problem_context pctx;
3084
3085         if (!ctx->casefolded_dirs) {
3086                 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
3087                 if (pctx.errcode)
3088                         goto error;
3089         }
3090         pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
3091         if (pctx.errcode == 0)
3092                 return;
3093 error:
3094         fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
3095         /* Should never get here */
3096         ctx->flags |= E2F_FLAG_ABORT;
3097 }
3098
3099 /*
3100  * This procedure will allocate the inode "bb" (badblock) map table
3101  */
3102 static void alloc_bb_map(e2fsck_t ctx)
3103 {
3104         struct          problem_context pctx;
3105
3106         clear_problem_context(&pctx);
3107         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3108                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
3109                         "inode_bb_map", &ctx->inode_bb_map);
3110         if (pctx.errcode) {
3111                 pctx.num = 4;
3112                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3113                 /* Should never get here */
3114                 ctx->flags |= E2F_FLAG_ABORT;
3115                 return;
3116         }
3117 }
3118
3119 /*
3120  * This procedure will allocate the inode imagic table
3121  */
3122 static void alloc_imagic_map(e2fsck_t ctx)
3123 {
3124         struct          problem_context pctx;
3125
3126         clear_problem_context(&pctx);
3127         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3128                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
3129                         "inode_imagic_map", &ctx->inode_imagic_map);
3130         if (pctx.errcode) {
3131                 pctx.num = 5;
3132                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3133                 /* Should never get here */
3134                 ctx->flags |= E2F_FLAG_ABORT;
3135                 return;
3136         }
3137 }
3138
3139 /*
3140  * Marks a block as in use, setting the dup_map if it's been set
3141  * already.  Called by process_block and process_bad_block.
3142  *
3143  * WARNING: Assumes checks have already been done to make sure block
3144  * is valid.  This is true in both process_block and process_bad_block.
3145  */
3146 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
3147 {
3148         struct          problem_context pctx;
3149
3150         clear_problem_context(&pctx);
3151
3152         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
3153                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
3154                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
3155                         return;
3156                 }
3157                 if (!ctx->block_dup_map) {
3158                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
3159                                         _("multiply claimed block map"),
3160                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
3161                                         &ctx->block_dup_map);
3162                         if (pctx.errcode) {
3163                                 pctx.num = 3;
3164                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
3165                                             &pctx);
3166                                 /* Should never get here */
3167                                 ctx->flags |= E2F_FLAG_ABORT;
3168                                 return;
3169                         }
3170                 }
3171                 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
3172         } else {
3173                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
3174         }
3175 }
3176
3177 /*
3178  * When cluster size is greater than one block, it is caller's responsibility
3179  * to make sure block parameter starts at a cluster boundary.
3180  */
3181 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
3182                                       unsigned int num)
3183 {
3184         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
3185                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
3186         else {
3187                 unsigned int i;
3188
3189                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
3190                         mark_block_used(ctx, block + i);
3191         }
3192 }
3193
3194 /*
3195  * Adjust the extended attribute block's reference counts at the end
3196  * of pass 1, either by subtracting out references for EA blocks that
3197  * are still referenced in ctx->refcount, or by adding references for
3198  * EA blocks that had extra references as accounted for in
3199  * ctx->refcount_extra.
3200  */
3201 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
3202                                     char *block_buf, int adjust_sign)
3203 {
3204         struct ext2_ext_attr_header     *header;
3205         struct problem_context          pctx;
3206         ext2_filsys                     fs = ctx->fs;
3207         blk64_t                         blk;
3208         __u32                           should_be;
3209         ea_value_t                      count;
3210
3211         clear_problem_context(&pctx);
3212
3213         ea_refcount_intr_begin(refcount);
3214         while (1) {
3215                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
3216                         break;
3217                 pctx.blk = blk;
3218                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
3219                                                      pctx.ino);
3220                 if (pctx.errcode) {
3221                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
3222                         return;
3223                 }
3224                 header = (struct ext2_ext_attr_header *) block_buf;
3225                 pctx.blkcount = header->h_refcount;
3226                 should_be = header->h_refcount + adjust_sign * (int)count;
3227                 pctx.num = should_be;
3228                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
3229                         header->h_refcount = should_be;
3230                         pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
3231                                                              block_buf,
3232                                                              pctx.ino);
3233                         if (pctx.errcode) {
3234                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
3235                                             &pctx);
3236                                 continue;
3237                         }
3238                 }
3239         }
3240 }
3241
3242 /*
3243  * Handle processing the extended attribute blocks
3244  */
3245 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
3246                            char *block_buf, struct ea_quota *ea_block_quota)
3247 {
3248         ext2_filsys fs = ctx->fs;
3249         ext2_ino_t      ino = pctx->ino;
3250         struct ext2_inode *inode = pctx->inode;
3251         blk64_t         blk;
3252         char *          end;
3253         struct ext2_ext_attr_header *header;
3254         struct ext2_ext_attr_entry *first, *entry;
3255         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
3256         __u64           quota_inodes = 0;
3257         region_t        region = 0;
3258         int             failed_csum = 0;
3259
3260         ea_block_quota->blocks = 0;
3261         ea_block_quota->inodes = 0;
3262
3263         blk = ext2fs_file_acl_block(fs, inode);
3264         if (blk == 0)
3265                 return 0;
3266
3267         /*
3268          * If the Extended attribute flag isn't set, then a non-zero
3269          * file acl means that the inode is corrupted.
3270          *
3271          * Or if the extended attribute block is an invalid block,
3272          * then the inode is also corrupted.
3273          */
3274         if (!ext2fs_has_feature_xattr(fs->super) ||
3275             (blk < fs->super->s_first_data_block) ||
3276             (blk >= ext2fs_blocks_count(fs->super))) {
3277                 mark_inode_bad(ctx, ino);
3278                 return 0;
3279         }
3280
3281         /* If ea bitmap hasn't been allocated, create it */
3282         if (!ctx->block_ea_map) {
3283                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
3284                                         _("ext attr block map"),
3285                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
3286                                         &ctx->block_ea_map);
3287                 if (pctx->errcode) {
3288                         pctx->num = 2;
3289                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
3290                         ctx->flags |= E2F_FLAG_ABORT;
3291                         return 0;
3292                 }
3293         }
3294
3295         /* Create the EA refcount structure if necessary */
3296         if (!ctx->refcount) {
3297                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
3298                 if (pctx->errcode) {
3299                         pctx->num = 1;
3300                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3301                         ctx->flags |= E2F_FLAG_ABORT;
3302                         return 0;
3303                 }
3304         }
3305
3306 #if 0
3307         /* Debugging text */
3308         printf("Inode %u has EA block %u\n", ino, blk);
3309 #endif
3310
3311         /* Have we seen this EA block before? */
3312         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
3313                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
3314                 ea_block_quota->inodes = 0;
3315
3316                 if (ctx->ea_block_quota_blocks) {
3317                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
3318                                           &quota_blocks);
3319                         if (quota_blocks)
3320                                 ea_block_quota->blocks = quota_blocks;
3321                 }
3322
3323                 if (ctx->ea_block_quota_inodes)
3324                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
3325                                           &ea_block_quota->inodes);
3326
3327                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
3328                         return 1;
3329                 /* Ooops, this EA was referenced more than it stated */
3330                 if (!ctx->refcount_extra) {
3331                         pctx->errcode = ea_refcount_create(0,
3332                                            &ctx->refcount_extra);
3333                         if (pctx->errcode) {
3334                                 pctx->num = 2;
3335                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3336                                 ctx->flags |= E2F_FLAG_ABORT;
3337                                 return 0;
3338                         }
3339                 }
3340                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
3341                 return 1;
3342         }
3343
3344         /*
3345          * OK, we haven't seen this EA block yet.  So we need to
3346          * validate it
3347          */
3348         pctx->blk = blk;
3349         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
3350         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
3351                 pctx->errcode = 0;
3352                 failed_csum = 1;
3353         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
3354                 pctx->errcode = 0;
3355
3356         if (pctx->errcode &&
3357             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
3358                 pctx->errcode = 0;
3359                 goto clear_extattr;
3360         }
3361         header = (struct ext2_ext_attr_header *) block_buf;
3362         pctx->blk = ext2fs_file_acl_block(fs, inode);
3363         if (((ctx->ext_attr_ver == 1) &&
3364              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
3365             ((ctx->ext_attr_ver == 2) &&
3366              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
3367                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
3368                         goto clear_extattr;
3369         }
3370
3371         if (header->h_blocks != 1) {
3372                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
3373                         goto clear_extattr;
3374         }
3375
3376         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
3377                 goto clear_extattr;
3378
3379         region = region_create(0, fs->blocksize);
3380         if (!region) {
3381                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
3382                 ctx->flags |= E2F_FLAG_ABORT;
3383                 return 0;
3384         }
3385         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
3386                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3387                         goto clear_extattr;
3388         }
3389
3390         first = (struct ext2_ext_attr_entry *)(header+1);
3391         end = block_buf + fs->blocksize;
3392         entry = first;
3393         while ((char *)entry < end && *(__u32 *)entry) {
3394                 __u32 hash;
3395
3396                 if (region_allocate(region, (char *)entry - (char *)header,
3397                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
3398                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3399                                 goto clear_extattr;
3400                         break;
3401                 }
3402                 if ((ctx->ext_attr_ver == 1 &&
3403                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
3404                     (ctx->ext_attr_ver == 2 &&
3405                      entry->e_name_index == 0)) {
3406                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
3407                                 goto clear_extattr;
3408                         break;
3409                 }
3410                 if (entry->e_value_inum == 0) {
3411                         if (entry->e_value_size > EXT2_XATTR_SIZE_MAX ||
3412                             (entry->e_value_offs + entry->e_value_size >
3413                              fs->blocksize)) {
3414                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
3415                                         goto clear_extattr;
3416                                 break;
3417                         }
3418                         if (entry->e_value_size &&
3419                             region_allocate(region, entry->e_value_offs,
3420                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
3421                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
3422                                                 pctx))
3423                                         goto clear_extattr;
3424                         }
3425
3426                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
3427                                                           entry->e_value_offs);
3428                         if (entry->e_hash != hash)
3429                                 hash = ext2fs_ext_attr_hash_entry_signed(entry,
3430                                         block_buf + entry->e_value_offs);
3431
3432                         if (entry->e_hash != hash) {
3433                                 pctx->num = entry->e_hash;
3434                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
3435                                         goto clear_extattr;
3436                                 entry->e_hash = hash;
3437                         }
3438                 } else {
3439                         problem_t problem;
3440                         blk64_t entry_quota_blocks;
3441
3442                         problem = check_large_ea_inode(ctx, entry, pctx,
3443                                                        &entry_quota_blocks);
3444                         if (problem && fix_problem(ctx, problem, pctx))
3445                                 goto clear_extattr;
3446
3447                         quota_blocks += entry_quota_blocks;
3448                         quota_inodes++;
3449                 }
3450
3451                 entry = EXT2_EXT_ATTR_NEXT(entry);
3452         }
3453         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
3454                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3455                         goto clear_extattr;
3456         }
3457         region_free(region);
3458
3459         /*
3460          * We only get here if there was no other errors that were fixed.
3461          * If there was a checksum fail, ask to correct it.
3462          */
3463         if (failed_csum &&
3464             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
3465                 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
3466                                                        pctx->ino);
3467                 if (pctx->errcode)
3468                         return 0;
3469         }
3470
3471         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
3472                 if (!ctx->ea_block_quota_blocks) {
3473                         pctx->errcode = ea_refcount_create(0,
3474                                                 &ctx->ea_block_quota_blocks);
3475                         if (pctx->errcode) {
3476                                 pctx->num = 3;
3477                                 goto refcount_fail;
3478                         }
3479                 }
3480                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
3481                                   quota_blocks);
3482         }
3483
3484         if (quota_inodes) {
3485                 if (!ctx->ea_block_quota_inodes) {
3486                         pctx->errcode = ea_refcount_create(0,
3487                                                 &ctx->ea_block_quota_inodes);
3488                         if (pctx->errcode) {
3489                                 pctx->num = 4;
3490 refcount_fail:
3491                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3492                                 ctx->flags |= E2F_FLAG_ABORT;
3493                                 return 0;
3494                         }
3495                 }
3496
3497                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
3498                                   quota_inodes);
3499         }
3500         ea_block_quota->blocks = quota_blocks;
3501         ea_block_quota->inodes = quota_inodes;
3502
3503         inc_ea_inode_refs(ctx, pctx, first, end);
3504         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
3505         mark_block_used(ctx, blk);
3506         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
3507         return 1;
3508
3509 clear_extattr:
3510         if (region)
3511                 region_free(region);
3512         ext2fs_file_acl_block_set(fs, inode, 0);
3513         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
3514         return 0;
3515 }
3516
3517 /* Returns 1 if bad htree, 0 if OK */
3518 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
3519                         ext2_ino_t ino, struct ext2_inode *inode,
3520                         char *block_buf)
3521 {
3522         struct ext2_dx_root_info        *root;
3523         ext2_filsys                     fs = ctx->fs;
3524         errcode_t                       retval;
3525         blk64_t                         blk;
3526
3527         if ((!LINUX_S_ISDIR(inode->i_mode) &&
3528              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
3529             (!ext2fs_has_feature_dir_index(fs->super) &&
3530              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
3531                 return 1;
3532
3533         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
3534
3535         if ((pctx->errcode) ||
3536             (blk == 0) ||
3537             (blk < fs->super->s_first_data_block) ||
3538             (blk >= ext2fs_blocks_count(fs->super))) {
3539                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3540                         return 1;
3541                 else
3542                         return 0;
3543         }
3544
3545         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
3546         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3547                 return 1;
3548
3549         /* XXX should check that beginning matches a directory */
3550         root = (struct ext2_dx_root_info *) (block_buf + 24);
3551
3552         if ((root->reserved_zero || root->info_length < 8) &&
3553             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3554                 return 1;
3555
3556         pctx->num = root->hash_version;
3557         if ((root->hash_version != EXT2_HASH_LEGACY) &&
3558             (root->hash_version != EXT2_HASH_HALF_MD4) &&
3559             (root->hash_version != EXT2_HASH_TEA) &&
3560             (root->hash_version != EXT2_HASH_SIPHASH) &&
3561             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
3562                 return 1;
3563
3564         if (ext4_hash_in_dirent(inode)) {
3565                 if (root->hash_version != EXT2_HASH_SIPHASH &&
3566                     fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
3567                         return 1;
3568         } else {
3569                 if (root->hash_version == EXT2_HASH_SIPHASH &&
3570                    fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
3571                         return 1;
3572         }
3573
3574         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
3575             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
3576                 return 1;
3577
3578         pctx->num = root->indirect_levels;
3579         /* if htree level is clearly too high, consider it to be broken */
3580         if (root->indirect_levels > EXT4_HTREE_LEVEL &&
3581             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3582                 return 1;
3583
3584         /* if level is only maybe too high, LARGE_DIR feature could be unset */
3585         if (root->indirect_levels > ext2_dir_htree_level(fs) &&
3586             !ext2fs_has_feature_largedir(fs->super)) {
3587                 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
3588                 unsigned idx_pb = 1 << (blockbits - 3);
3589
3590                 /* compare inode size/blocks vs. max-sized 2-level htree */
3591                 if (EXT2_I_SIZE(pctx->inode) <
3592                     (idx_pb - 1) * (idx_pb - 2) << blockbits &&
3593                     pctx->inode->i_blocks <
3594                     (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
3595                     fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3596                         return 1;
3597         }
3598
3599         if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
3600             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3601                 ctx->large_dirs++;
3602
3603         return 0;
3604 }
3605
3606 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
3607                         struct ext2_inode *inode, int restart_flag,
3608                         const char *source)
3609 {
3610         inode->i_flags = 0;
3611         inode->i_links_count = 0;
3612         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
3613         inode->i_dtime = ctx->now;
3614
3615         /*
3616          * If a special inode has such rotten block mappings that we
3617          * want to clear the whole inode, be sure to actually zap
3618          * the block maps because i_links_count isn't checked for
3619          * special inodes, and we'll end up right back here the next
3620          * time we run fsck.
3621          */
3622         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
3623                 memset(inode->i_block, 0, sizeof(inode->i_block));
3624
3625         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
3626         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
3627         if (ctx->inode_reg_map)
3628                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
3629         if (ctx->inode_bad_map)
3630                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
3631
3632         /*
3633          * If the inode was partially accounted for before processing
3634          * was aborted, we need to restart the pass 1 scan.
3635          */
3636         ctx->flags |= restart_flag;
3637
3638         if (ino == EXT2_BAD_INO)
3639                 memset(inode, 0, sizeof(struct ext2_inode));
3640
3641         e2fsck_write_inode(ctx, ino, inode, source);
3642 }
3643
3644 /*
3645  * Use the multiple-blocks reclamation code to fix alignment problems in
3646  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
3647  * physical cluster, and we want the block offsets within that cluster to
3648  * line up.
3649  */
3650 static int has_unaligned_cluster_map(e2fsck_t ctx,
3651                                      blk64_t last_pblk, blk64_t last_lblk,
3652                                      blk64_t pblk, blk64_t lblk)
3653 {
3654         blk64_t cluster_mask;
3655
3656         if (!ctx->fs->cluster_ratio_bits)
3657                 return 0;
3658         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
3659
3660         /*
3661          * If the block in the logical cluster doesn't align with the block in
3662          * the physical cluster...
3663          */
3664         if ((lblk & cluster_mask) != (pblk & cluster_mask))
3665                 return 1;
3666
3667         /*
3668          * If we cross a physical cluster boundary within a logical cluster...
3669          */
3670         if (last_pblk && (lblk & cluster_mask) != 0 &&
3671             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
3672             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
3673                 return 1;
3674
3675         return 0;
3676 }
3677
3678 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
3679                              struct process_block_struct *pb,
3680                              blk64_t start_block, blk64_t end_block,
3681                              blk64_t eof_block,
3682                              ext2_extent_handle_t ehandle,
3683                              int try_repairs)
3684 {
3685         struct ext2fs_extent    extent;
3686         blk64_t                 blk, last_lblk;
3687         unsigned int            i, n;
3688         int                     is_dir, is_leaf;
3689         problem_t               problem;
3690         struct ext2_extent_info info;
3691         int                     failed_csum = 0;
3692
3693         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
3694                 failed_csum = 1;
3695
3696         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
3697         if (pctx->errcode)
3698                 return;
3699         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
3700             !pb->eti.force_rebuild &&
3701             info.curr_level < MAX_EXTENT_DEPTH_COUNT) {
3702                 struct extent_tree_level *etl;
3703
3704                 etl = pb->eti.ext_info + info.curr_level;
3705                 etl->num_extents += info.num_entries;
3706                 etl->max_extents += info.max_entries;
3707                 /*
3708                  * Implementation wart: Splitting extent blocks when appending
3709                  * will leave the old block with one free entry.  Therefore
3710                  * unless the node is totally full, pretend that a non-root
3711                  * extent block can hold one fewer entry than it actually does,
3712                  * so that we don't repeatedly rebuild the extent tree.
3713                  */
3714                 if (info.curr_level && info.num_entries < info.max_entries)
3715                         etl->max_extents--;
3716         }
3717
3718         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
3719                                           &extent);
3720         while ((pctx->errcode == 0 ||
3721                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
3722                info.num_entries-- > 0) {
3723                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
3724                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
3725                 last_lblk = extent.e_lblk + extent.e_len - 1;
3726
3727                 problem = 0;
3728                 pctx->blk = extent.e_pblk;
3729                 pctx->blk2 = extent.e_lblk;
3730                 pctx->num = extent.e_len;
3731                 pctx->blkcount = extent.e_lblk + extent.e_len;
3732
3733                 if (extent.e_pblk == 0 ||
3734                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
3735                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
3736                         problem = PR_1_EXTENT_BAD_START_BLK;
3737                 else if (extent.e_lblk < start_block)
3738                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
3739                 else if ((end_block && last_lblk > end_block) &&
3740                          !(last_lblk > eof_block &&
3741                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
3742                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
3743                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
3744                 else if (is_leaf && extent.e_len == 0)
3745                         problem = PR_1_EXTENT_LENGTH_ZERO;
3746                 else if (is_leaf &&
3747                          (extent.e_pblk + extent.e_len) >
3748                          ext2fs_blocks_count(ctx->fs->super))
3749                         problem = PR_1_EXTENT_ENDS_BEYOND;
3750                 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
3751                          !ext2fs_has_feature_largedir(ctx->fs->super) &&
3752                          ((extent.e_lblk + extent.e_len) >
3753                           (1U << (21 - ctx->fs->super->s_log_block_size))))
3754                         problem = PR_1_TOOBIG_DIR;
3755
3756                 if (is_leaf && problem == 0 && extent.e_len > 0) {
3757 #if 0
3758                         printf("extent_region(ino=%u, expect=%llu, "
3759                                "lblk=%llu, len=%u)\n", pb->ino,
3760                                (unsigned long long) pb->next_lblock,
3761                                (unsigned long long) extent.e_lblk,
3762                                extent.e_len);
3763 #endif
3764                         if (extent.e_lblk < pb->next_lblock)
3765                                 problem = PR_1_EXTENT_COLLISION;
3766                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
3767                                 pb->next_lblock = extent.e_lblk + extent.e_len;
3768                 }
3769
3770                 /*
3771                  * Uninitialized blocks in a directory?  Clear the flag and
3772                  * we'll interpret the blocks later.
3773                  */
3774                 if (try_repairs && is_dir && problem == 0 &&
3775                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3776                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
3777                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3778                         pb->inode_modified = 1;
3779                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3780                                                               &extent);
3781                         if (pctx->errcode)
3782                                 return;
3783                         failed_csum = 0;
3784                 }
3785 #ifdef CONFIG_DEVELOPER_FEATURES
3786                 if (try_repairs && !is_dir && problem == 0 &&
3787                     (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
3788                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3789                     fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
3790                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3791                         pb->inode_modified = 1;
3792                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3793                                                               &extent);
3794                         if (pctx->errcode)
3795                                 return;
3796                         failed_csum = 0;
3797                 }
3798 #endif
3799                 if (try_repairs && problem) {
3800 report_problem:
3801                         if (fix_problem(ctx, problem, pctx)) {
3802                                 if (ctx->invalid_bitmaps) {
3803                                         /*
3804                                          * If fsck knows the bitmaps are bad,
3805                                          * skip to the next extent and
3806                                          * try to clear this extent again
3807                                          * after fixing the bitmaps, by
3808                                          * restarting fsck.
3809                                          */
3810                                         pctx->errcode = ext2fs_extent_get(
3811                                                           ehandle,
3812                                                           EXT2_EXTENT_NEXT_SIB,
3813                                                           &extent);
3814                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3815                                         if (pctx->errcode ==
3816                                                     EXT2_ET_NO_CURRENT_NODE) {
3817                                                 pctx->errcode = 0;
3818                                                 break;
3819                                         }
3820                                         continue;
3821                                 }
3822                                 e2fsck_read_bitmaps(ctx);
3823                                 pb->inode_modified = 1;
3824                                 pctx->errcode =
3825                                         ext2fs_extent_delete(ehandle, 0);
3826                                 if (pctx->errcode) {
3827                                         pctx->str = "ext2fs_extent_delete";
3828                                         return;
3829                                 }
3830                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3831                                 if (pctx->errcode &&
3832                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
3833                                         pctx->str = "ext2fs_extent_fix_parents";
3834                                         return;
3835                                 }
3836                                 pctx->errcode = ext2fs_extent_get(ehandle,
3837                                                                   EXT2_EXTENT_CURRENT,
3838                                                                   &extent);
3839                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
3840                                         pctx->errcode = 0;
3841                                         break;
3842                                 }
3843                                 failed_csum = 0;
3844                                 continue;
3845                         }
3846                         goto next;
3847                 }
3848
3849                 if (!is_leaf) {
3850                         blk64_t lblk = extent.e_lblk;
3851                         int next_try_repairs = 1;
3852
3853                         blk = extent.e_pblk;
3854
3855                         /*
3856                          * If this lower extent block collides with critical
3857                          * metadata, don't try to repair the damage.  Pass 1b
3858                          * will reallocate the block; then we can try again.
3859                          */
3860                         if (pb->ino != EXT2_RESIZE_INO &&
3861                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
3862                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3863                                                       extent.e_pblk)) {
3864                                 next_try_repairs = 0;
3865                                 pctx->blk = blk;
3866                                 fix_problem(ctx,
3867                                             PR_1_CRITICAL_METADATA_COLLISION,
3868                                             pctx);
3869                                 if ((ctx->options & E2F_OPT_NO) == 0)
3870                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3871                         }
3872                         pctx->errcode = ext2fs_extent_get(ehandle,
3873                                                   EXT2_EXTENT_DOWN, &extent);
3874                         if (pctx->errcode &&
3875                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
3876                                 pctx->str = "EXT2_EXTENT_DOWN";
3877                                 problem = PR_1_EXTENT_HEADER_INVALID;
3878                                 if (!next_try_repairs)
3879                                         return;
3880                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
3881                                         goto report_problem;
3882                                 return;
3883                         }
3884                         /* The next extent should match this index's logical start */
3885                         if (extent.e_lblk != lblk) {
3886                                 struct ext2_extent_info e_info;
3887
3888                                 pctx->errcode = ext2fs_extent_get_info(ehandle,
3889                                                                        &e_info);
3890                                 if (pctx->errcode) {
3891                                         pctx->str = "ext2fs_extent_get_info";
3892                                         return;
3893                                 }
3894                                 pctx->blk = lblk;
3895                                 pctx->blk2 = extent.e_lblk;
3896                                 pctx->num = e_info.curr_level - 1;
3897                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
3898                                 if (fix_problem(ctx, problem, pctx)) {
3899                                         pb->inode_modified = 1;
3900                                         pctx->errcode =
3901                                                 ext2fs_extent_fix_parents(ehandle);
3902                                         if (pctx->errcode) {
3903                                                 pctx->str = "ext2fs_extent_fix_parents";
3904                                                 return;
3905                                         }
3906                                 }
3907                         }
3908                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
3909                                          last_lblk, eof_block, ehandle,
3910                                          next_try_repairs);
3911                         if (pctx->errcode)
3912                                 return;
3913                         pctx->errcode = ext2fs_extent_get(ehandle,
3914                                                   EXT2_EXTENT_UP, &extent);
3915                         if (pctx->errcode) {
3916                                 pctx->str = "EXT2_EXTENT_UP";
3917                                 return;
3918                         }
3919                         mark_block_used(ctx, blk);
3920                         pb->num_blocks++;
3921                         goto next;
3922                 }
3923
3924                 if ((pb->previous_block != 0) &&
3925                     (pb->previous_block+1 != extent.e_pblk)) {
3926                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3927                                 char type = '?';
3928
3929                                 if (pb->is_dir)
3930                                         type = 'd';
3931                                 else if (pb->is_reg)
3932                                         type = 'f';
3933
3934                                 printf(("%6lu(%c): expecting %6lu "
3935                                         "actual extent "
3936                                         "phys %6lu log %lu len %lu\n"),
3937                                        (unsigned long) pctx->ino, type,
3938                                        (unsigned long) pb->previous_block+1,
3939                                        (unsigned long) extent.e_pblk,
3940                                        (unsigned long) extent.e_lblk,
3941                                        (unsigned long) extent.e_len);
3942                         }
3943                         pb->fragmented = 1;
3944                 }
3945                 /*
3946                  * If we notice a gap in the logical block mappings of an
3947                  * extent-mapped directory, offer to close the hole by
3948                  * moving the logical block down, otherwise we'll go mad in
3949                  * pass 3 allocating empty directory blocks to fill the hole.
3950                  */
3951                 if (try_repairs && is_dir &&
3952                     pb->last_block + 1 < extent.e_lblk) {
3953                         blk64_t new_lblk;
3954
3955                         new_lblk = pb->last_block + 1;
3956                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3957                                 new_lblk = ((new_lblk +
3958                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3959                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3960                                            (extent.e_pblk &
3961                                             EXT2FS_CLUSTER_MASK(ctx->fs));
3962                         pctx->blk = extent.e_lblk;
3963                         pctx->blk2 = new_lblk;
3964                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3965                                 extent.e_lblk = new_lblk;
3966                                 pb->inode_modified = 1;
3967                                 pctx->errcode = ext2fs_extent_replace(ehandle,
3968                                                                 0, &extent);
3969                                 if (pctx->errcode) {
3970                                         pctx->errcode = 0;
3971                                         goto alloc_later;
3972                                 }
3973                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3974                                 if (pctx->errcode)
3975                                         goto failed_add_dir_block;
3976                                 pctx->errcode = ext2fs_extent_goto(ehandle,
3977                                                                 extent.e_lblk);
3978                                 if (pctx->errcode)
3979                                         goto failed_add_dir_block;
3980                                 last_lblk = extent.e_lblk + extent.e_len - 1;
3981                                 failed_csum = 0;
3982                         }
3983                 }
3984 alloc_later:
3985                 if (is_dir) {
3986                         while (++pb->last_db_block <
3987                                (e2_blkcnt_t) extent.e_lblk) {
3988                                 pctx->errcode = ext2fs_add_dir_block2(
3989                                                         ctx->fs->dblist,
3990                                                         pb->ino, 0,
3991                                                         pb->last_db_block);
3992                                 if (pctx->errcode) {
3993                                         pctx->blk = 0;
3994                                         pctx->num = pb->last_db_block;
3995                                         goto failed_add_dir_block;
3996                                 }
3997                         }
3998
3999                         for (i = 0; i < extent.e_len; i++) {
4000                                 pctx->errcode = ext2fs_add_dir_block2(
4001                                                         ctx->fs->dblist,
4002                                                         pctx->ino,
4003                                                         extent.e_pblk + i,
4004                                                         extent.e_lblk + i);
4005                                 if (pctx->errcode) {
4006                                         pctx->blk = extent.e_pblk + i;
4007                                         pctx->num = extent.e_lblk + i;
4008                                 failed_add_dir_block:
4009                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4010                                         /* Should never get here */
4011                                         ctx->flags |= E2F_FLAG_ABORT;
4012                                         return;
4013                                 }
4014                         }
4015                         if (extent.e_len > 0)
4016                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
4017                 }
4018                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
4019                                               pb->last_block,
4020                                               extent.e_pblk,
4021                                               extent.e_lblk)) {
4022                         for (i = 0; i < extent.e_len; i++) {
4023                                 pctx->blk = extent.e_lblk + i;
4024                                 pctx->blk2 = extent.e_pblk + i;
4025                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4026                                 mark_block_used(ctx, extent.e_pblk + i);
4027                                 mark_block_used(ctx, extent.e_pblk + i);
4028                         }
4029                 }
4030
4031                 /*
4032                  * Check whether first cluster got marked in previous iteration.
4033                  */
4034                 if (ctx->fs->cluster_ratio_bits &&
4035                     pb->previous_block &&
4036                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
4037                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
4038                         /* Set blk to the beginning of next cluster. */
4039                         blk = EXT2FS_C2B(
4040                                 ctx->fs,
4041                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
4042                 else
4043                         /* Set blk to the beginning of current cluster. */
4044                         blk = EXT2FS_C2B(ctx->fs,
4045                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
4046
4047                 if (blk < extent.e_pblk + extent.e_len) {
4048                         mark_blocks_used(ctx, blk,
4049                                          extent.e_pblk + extent.e_len - blk);
4050                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
4051                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
4052                         pb->num_blocks += n;
4053                 }
4054                 pb->last_block = extent.e_lblk + extent.e_len - 1;
4055                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
4056                 start_block = pb->last_block = last_lblk;
4057                 if (is_leaf && !is_dir &&
4058                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
4059                         pb->last_init_lblock = last_lblk;
4060         next:
4061                 pctx->errcode = ext2fs_extent_get(ehandle,
4062                                                   EXT2_EXTENT_NEXT_SIB,
4063                                                   &extent);
4064         }
4065
4066         /* Failed csum but passes checks?  Ask to fix checksum. */
4067         if (failed_csum &&
4068             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
4069                 pb->inode_modified = 1;
4070                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
4071                 if (pctx->errcode)
4072                         return;
4073         }
4074
4075         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
4076                 pctx->errcode = 0;
4077 }
4078
4079 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
4080                                  struct process_block_struct *pb)
4081 {
4082         struct ext2_extent_info info;
4083         struct ext2_inode       *inode = pctx->inode;
4084         ext2_extent_handle_t    ehandle;
4085         ext2_filsys             fs = ctx->fs;
4086         ext2_ino_t              ino = pctx->ino;
4087         errcode_t               retval;
4088         blk64_t                 eof_lblk;
4089         struct ext3_extent_header       *eh;
4090
4091         /* Check for a proper extent header... */
4092         eh = (struct ext3_extent_header *) &inode->i_block[0];
4093         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
4094         if (retval) {
4095                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
4096                         e2fsck_clear_inode(ctx, ino, inode, 0,
4097                                            "check_blocks_extents");
4098                 pctx->errcode = 0;
4099                 return;
4100         }
4101
4102         /* ...since this function doesn't fail if i_block is zeroed. */
4103         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
4104         if (pctx->errcode) {
4105                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
4106                         e2fsck_clear_inode(ctx, ino, inode, 0,
4107                                            "check_blocks_extents");
4108                 pctx->errcode = 0;
4109                 return;
4110         }
4111
4112         retval = ext2fs_extent_get_info(ehandle, &info);
4113         if (retval == 0) {
4114                 int max_depth = info.max_depth;
4115
4116                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
4117                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
4118                 ctx->extent_depth_count[max_depth]++;
4119         }
4120
4121         /* Check maximum extent depth */
4122         pctx->blk = info.max_depth;
4123         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
4124         if (pctx->blk2 < pctx->blk &&
4125             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
4126                 pb->eti.force_rebuild = 1;
4127
4128         /* Can we collect extent tree level stats? */
4129         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
4130         if (pctx->blk2 > pctx->blk)
4131                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
4132         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
4133         pb->eti.ino = pb->ino;
4134
4135         pb->next_lblock = 0;
4136
4137         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
4138                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
4139         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
4140         if (pctx->errcode &&
4141             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
4142                 pb->num_blocks = 0;
4143                 inode->i_blocks = 0;
4144                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4145                                    "check_blocks_extents");
4146                 pctx->errcode = 0;
4147         }
4148         ext2fs_extent_free(ehandle);
4149
4150         /* Rebuild unless it's a dir and we're rehashing it */
4151         if (LINUX_S_ISDIR(inode->i_mode) &&
4152             e2fsck_dir_will_be_rehashed(ctx, ino))
4153                 return;
4154
4155         if (ctx->options & E2F_OPT_CONVERT_BMAP)
4156                 e2fsck_rebuild_extents_later(ctx, ino);
4157         else
4158                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
4159 }
4160
4161 /*
4162  * In fact we don't need to check blocks for an inode with inline data
4163  * because this inode doesn't have any blocks.  In this function all
4164  * we need to do is add this inode into dblist when it is a directory.
4165  */
4166 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
4167                                      struct process_block_struct *pb)
4168 {
4169         int     flags;
4170         size_t  inline_data_size = 0;
4171
4172         if (!pb->is_dir) {
4173                 pctx->errcode = 0;
4174                 return;
4175         }
4176
4177         /* Process the dirents in i_block[] as the "first" block. */
4178         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
4179         if (pctx->errcode)
4180                 goto err;
4181
4182         /* Process the dirents in the EA as a "second" block. */
4183         flags = ctx->fs->flags;
4184         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4185         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
4186                                                 &inline_data_size);
4187         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4188                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4189         if (pctx->errcode) {
4190                 pctx->errcode = 0;
4191                 return;
4192         }
4193
4194         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
4195                 return;
4196
4197         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
4198         if (pctx->errcode)
4199                 goto err;
4200
4201         return;
4202 err:
4203         pctx->blk = 0;
4204         pctx->num = 0;
4205         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4206         ctx->flags |= E2F_FLAG_ABORT;
4207 }
4208
4209 /*
4210  * This subroutine is called on each inode to account for all of the
4211  * blocks used by that inode.
4212  */
4213 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
4214                          char *block_buf, const struct ea_quota *ea_ibody_quota)
4215 {
4216         ext2_filsys fs = ctx->fs;
4217         struct process_block_struct pb;
4218         ext2_ino_t      ino = pctx->ino;
4219         struct ext2_inode *inode = pctx->inode;
4220         unsigned        bad_size = 0;
4221         int             dirty_inode = 0;
4222         int             extent_fs;
4223         int             inlinedata_fs;
4224         __u64           size;
4225         struct ea_quota ea_block_quota;
4226
4227         pb.ino = ino;
4228         pb.num_blocks = EXT2FS_B2C(ctx->fs,
4229                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
4230         pb.last_block = ~0;
4231         pb.last_init_lblock = -1;
4232         pb.last_db_block = -1;
4233         pb.num_illegal_blocks = 0;
4234         pb.suppress = 0; pb.clear = 0;
4235         pb.fragmented = 0;
4236         pb.compressed = 0;
4237         pb.previous_block = 0;
4238         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
4239         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
4240         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
4241         pb.inode = inode;
4242         pb.pctx = pctx;
4243         pb.ctx = ctx;
4244         pb.inode_modified = 0;
4245         pb.eti.force_rebuild = 0;
4246         pctx->ino = ino;
4247         pctx->errcode = 0;
4248
4249         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
4250         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
4251
4252         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
4253                 if (e2fsck_should_abort(ctx))
4254                         goto out;
4255                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
4256         }
4257
4258         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
4259                 check_blocks_inline_data(ctx, pctx, &pb);
4260         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
4261                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
4262                         check_blocks_extents(ctx, pctx, &pb);
4263                 else {
4264                         int flags;
4265                         /*
4266                          * If we've modified the inode, write it out before
4267                          * iterate() tries to use it.
4268                          */
4269                         if (dirty_inode) {
4270                                 e2fsck_write_inode(ctx, ino, inode,
4271                                                    "check_blocks");
4272                                 dirty_inode = 0;
4273                         }
4274                         flags = fs->flags;
4275                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4276                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
4277                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
4278                                                 block_buf, process_block, &pb);
4279                         /*
4280                          * We do not have uninitialized extents in non extent
4281                          * files.
4282                          */
4283                         pb.last_init_lblock = pb.last_block;
4284                         /*
4285                          * If iterate() changed a block mapping, we have to
4286                          * re-read the inode.  If we decide to clear the
4287                          * inode after clearing some stuff, we'll re-write the
4288                          * bad mappings into the inode!
4289                          */
4290                         if (pb.inode_modified)
4291                                 e2fsck_read_inode(ctx, ino, inode,
4292                                                   "check_blocks");
4293                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4294                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4295
4296                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
4297 #ifdef DEBUG
4298                                 printf("bmap rebuild ino=%d\n", ino);
4299 #endif
4300                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
4301                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
4302                                         e2fsck_rebuild_extents_later(ctx, ino);
4303                         }
4304                 }
4305         }
4306         end_problem_latch(ctx, PR_LATCH_BLOCK);
4307         end_problem_latch(ctx, PR_LATCH_TOOBIG);
4308         if (e2fsck_should_abort(ctx))
4309                 goto out;
4310         if (pctx->errcode)
4311                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
4312
4313         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
4314                 if (LINUX_S_ISDIR(inode->i_mode))
4315                         ctx->fs_fragmented_dir++;
4316                 else
4317                         ctx->fs_fragmented++;
4318         }
4319
4320         if (pb.clear) {
4321                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4322                                    "check_blocks");
4323                 return;
4324         }
4325
4326         if (inode->i_flags & EXT2_INDEX_FL) {
4327                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
4328                         inode->i_flags &= ~EXT2_INDEX_FL;
4329                         dirty_inode++;
4330                 } else {
4331                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
4332                 }
4333         }
4334
4335         if (!pb.num_blocks && pb.is_dir &&
4336             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
4337                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
4338                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
4339                         ctx->fs_directory_count--;
4340                         return;
4341                 }
4342         }
4343
4344         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
4345             ino != fs->super->s_orphan_file_inum &&
4346             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
4347             !(inode->i_flags & EXT4_EA_INODE_FL)) {
4348                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
4349                                ino,
4350                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
4351                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
4352                                   ino, (ea_ibody_quota ?
4353                                         ea_ibody_quota->inodes : 0) +
4354                                                 ea_block_quota.inodes + 1);
4355         }
4356
4357         if (!ext2fs_has_feature_huge_file(fs->super) ||
4358             !(inode->i_flags & EXT4_HUGE_FILE_FL))
4359                 pb.num_blocks *= (fs->blocksize / 512);
4360         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
4361 #if 0
4362         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
4363                ino, inode->i_size, (unsigned long long) pb.last_block,
4364                (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
4365                (unsigned long long) pb.num_blocks);
4366 #endif
4367         size = EXT2_I_SIZE(inode);
4368         if (pb.is_dir) {
4369                 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
4370                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
4371                         int flags;
4372                         size_t sz = 0;
4373                         errcode_t err;
4374
4375                         flags = ctx->fs->flags;
4376                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4377                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
4378                                                       &sz);
4379                         ctx->fs->flags = (flags &
4380                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4381                                          (ctx->fs->flags &
4382                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4383                         if (err || sz != size) {
4384                                 bad_size = 7;
4385                                 pctx->num = sz;
4386                         }
4387                 } else if (size & (fs->blocksize - 1))
4388                         bad_size = 5;
4389                 else if (nblock > (pb.last_block + 1))
4390                         bad_size = 1;
4391                 else if (nblock < (pb.last_block + 1)) {
4392                         if (((pb.last_block + 1) - nblock) >
4393                             fs->super->s_prealloc_dir_blocks)
4394                                 bad_size = 2;
4395                 }
4396         } else {
4397                 if ((pb.last_init_lblock >= 0) &&
4398                     /* Do not allow initialized allocated blocks past i_size*/
4399                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
4400                     !(inode->i_flags & EXT4_VERITY_FL))
4401                         bad_size = 3;
4402                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4403                          size > ext2_max_sizes[fs->super->s_log_block_size])
4404                         /* too big for a direct/indirect-mapped file */
4405                         bad_size = 4;
4406                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4407                          size >
4408                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
4409                         /* too big for an extent-based file - 32bit ee_block */
4410                         bad_size = 6;
4411         }
4412         /* i_size for symlinks is checked elsewhere */
4413         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
4414                 /* Did inline_data set pctx->num earlier? */
4415                 if (bad_size != 7)
4416                         pctx->num = (pb.last_block + 1) * fs->blocksize;
4417                 pctx->group = bad_size;
4418                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
4419                         ext2fs_inode_size_set(fs, inode, pctx->num);
4420                         if (EXT2_I_SIZE(inode) == 0 &&
4421                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
4422                                 memset(inode->i_block, 0,
4423                                        sizeof(inode->i_block));
4424                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
4425                         }
4426                         dirty_inode++;
4427                 }
4428                 pctx->num = 0;
4429         }
4430         if (LINUX_S_ISREG(inode->i_mode) &&
4431             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4432                 ctx->large_files++;
4433         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
4434             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
4435              (ext2fs_has_feature_huge_file(fs->super) &&
4436               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
4437               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
4438                 pctx->num = pb.num_blocks;
4439                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
4440                         inode->i_blocks = pb.num_blocks;
4441                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
4442                         dirty_inode++;
4443                 }
4444                 pctx->num = 0;
4445         }
4446
4447         /*
4448          * The kernel gets mad if we ask it to allocate bigalloc clusters to
4449          * a block mapped file, so rebuild it as an extent file.  We can skip
4450          * symlinks because they're never rewritten.
4451          */
4452         if (ext2fs_has_feature_bigalloc(fs->super) &&
4453             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
4454             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
4455             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
4456             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
4457             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
4458                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
4459                 if (pctx->errcode)
4460                         goto out;
4461         }
4462
4463         if (ctx->dirs_to_hash && pb.is_dir &&
4464             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
4465             !(inode->i_flags & EXT2_INDEX_FL) &&
4466             ((inode->i_size / fs->blocksize) >= 3))
4467                 e2fsck_rehash_dir_later(ctx, ino);
4468
4469 out:
4470         if (dirty_inode)
4471                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
4472 }
4473
4474 #if 0
4475 /*
4476  * Helper function called by process block when an illegal block is
4477  * found.  It returns a description about why the block is illegal
4478  */
4479 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
4480 {
4481         blk64_t super;
4482         int     i;
4483         static char     problem[80];
4484
4485         super = fs->super->s_first_data_block;
4486         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
4487         if (block < super) {
4488                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
4489                 return(problem);
4490         } else if (block >= ext2fs_blocks_count(fs->super)) {
4491                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
4492                 return(problem);
4493         }
4494         for (i = 0; i < fs->group_desc_count; i++) {
4495                 if (block == super) {
4496                         sprintf(problem, "is the superblock in group %d", i);
4497                         break;
4498                 }
4499                 if (block > super &&
4500                     block <= (super + fs->desc_blocks)) {
4501                         sprintf(problem, "is in the group descriptors "
4502                                 "of group %d", i);
4503                         break;
4504                 }
4505                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
4506                         sprintf(problem, "is the block bitmap of group %d", i);
4507                         break;
4508                 }
4509                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
4510                         sprintf(problem, "is the inode bitmap of group %d", i);
4511                         break;
4512                 }
4513                 if (block >= ext2fs_inode_table_loc(fs, i) &&
4514                     (block < ext2fs_inode_table_loc(fs, i)
4515                      + fs->inode_blocks_per_group)) {
4516                         sprintf(problem, "is in the inode table of group %d",
4517                                 i);
4518                         break;
4519                 }
4520                 super += fs->super->s_blocks_per_group;
4521         }
4522         return(problem);
4523 }
4524 #endif
4525
4526 /*
4527  * This is a helper function for check_blocks().
4528  */
4529 static int process_block(ext2_filsys fs,
4530                   blk64_t       *block_nr,
4531                   e2_blkcnt_t blockcnt,
4532                   blk64_t ref_block EXT2FS_ATTR((unused)),
4533                   int ref_offset EXT2FS_ATTR((unused)),
4534                   void *priv_data)
4535 {
4536         struct process_block_struct *p;
4537         struct problem_context *pctx;
4538         blk64_t blk = *block_nr;
4539         int     ret_code = 0;
4540         problem_t       problem = 0;
4541         e2fsck_t        ctx;
4542
4543         p = (struct process_block_struct *) priv_data;
4544         pctx = p->pctx;
4545         ctx = p->ctx;
4546
4547         /*
4548          * For a directory, add logical block zero for processing even if it's
4549          * not mapped or we'll be perennially stuck with broken "." and ".."
4550          * entries.
4551          */
4552         if (p->is_dir && blockcnt == 0 && blk == 0) {
4553                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
4554                 if (pctx->errcode) {
4555                         pctx->blk = blk;
4556                         pctx->num = blockcnt;
4557                         goto failed_add_dir_block;
4558                 }
4559                 p->last_db_block++;
4560         }
4561
4562         if (blk == 0)
4563                 return 0;
4564
4565 #if 0
4566         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
4567                blockcnt);
4568 #endif
4569
4570         /*
4571          * Simplistic fragmentation check.  We merely require that the
4572          * file be contiguous.  (Which can never be true for really
4573          * big files that are greater than a block group.)
4574          */
4575         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
4576                 if (p->previous_block+1 != blk) {
4577                         if (ctx->options & E2F_OPT_FRAGCHECK) {
4578                                 char type = '?';
4579
4580                                 if (p->is_dir)
4581                                         type = 'd';
4582                                 else if (p->is_reg)
4583                                         type = 'f';
4584
4585                                 printf(_("%6lu(%c): expecting %6lu "
4586                                          "got phys %6lu (blkcnt %lld)\n"),
4587                                        (unsigned long) pctx->ino, type,
4588                                        (unsigned long) p->previous_block+1,
4589                                        (unsigned long) blk,
4590                                        (long long) blockcnt);
4591                         }
4592                         p->fragmented = 1;
4593                 }
4594         }
4595
4596         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
4597             !pctx->inode->i_size_high &&
4598             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
4599                 problem = PR_1_TOOBIG_DIR;
4600         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
4601                 problem = PR_1_TOOBIG_DIR;
4602         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
4603                 problem = PR_1_TOOBIG_REG;
4604         if (!p->is_dir && !p->is_reg && blockcnt > 0)
4605                 problem = PR_1_TOOBIG_SYMLINK;
4606
4607         if (blk < fs->super->s_first_data_block ||
4608             blk >= ext2fs_blocks_count(fs->super))
4609                 problem = PR_1_ILLEGAL_BLOCK_NUM;
4610
4611         /*
4612          * If this IND/DIND/TIND block is squatting atop some critical metadata
4613          * (group descriptors, superblock, bitmap, inode table), any write to
4614          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
4615          * fix that and restart fsck.
4616          */
4617         if (blockcnt < 0 &&
4618             p->ino != EXT2_RESIZE_INO &&
4619             blk < ctx->fs->super->s_blocks_count &&
4620             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
4621                 pctx->blk = blk;
4622                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
4623                 if ((ctx->options & E2F_OPT_NO) == 0)
4624                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4625         }
4626
4627         if (problem) {
4628                 p->num_illegal_blocks++;
4629                 /*
4630                  * A bit of subterfuge here -- we're trying to fix a block
4631                  * mapping, but the IND/DIND/TIND block could have collided
4632                  * with some critical metadata.  So, fix the in-core mapping so
4633                  * iterate won't go insane, but return 0 instead of
4634                  * BLOCK_CHANGED so that it won't write the remapping out to
4635                  * our multiply linked block.
4636                  *
4637                  * Even if we previously determined that an *IND block
4638                  * conflicts with critical metadata, we must still try to
4639                  * iterate the *IND block as if it is an *IND block to find and
4640                  * mark the blocks it points to.  Better to be overly cautious
4641                  * with the used_blocks map so that we don't move the *IND
4642                  * block to a block that's really in use!
4643                  */
4644                 if (p->ino != EXT2_RESIZE_INO &&
4645                     ref_block != 0 &&
4646                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4647                                               ref_block)) {
4648                         *block_nr = 0;
4649                         return 0;
4650                 }
4651                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
4652                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
4653                                 p->clear = 1;
4654                                 return BLOCK_ABORT;
4655                         }
4656                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
4657                                 p->suppress = 1;
4658                                 set_latch_flags(PR_LATCH_BLOCK,
4659                                                 PRL_SUPPRESS, 0);
4660                         }
4661                 }
4662                 pctx->blk = blk;
4663                 pctx->blkcount = blockcnt;
4664                 if (fix_problem(ctx, problem, pctx)) {
4665                         blk = *block_nr = 0;
4666                         ret_code = BLOCK_CHANGED;
4667                         p->inode_modified = 1;
4668                         /*
4669                          * If the directory block is too big and is beyond the
4670                          * end of the FS, don't bother trying to add it for
4671                          * processing -- the kernel would never have created a
4672                          * directory this large, and we risk an ENOMEM abort.
4673                          * In any case, the toobig handler for extent-based
4674                          * directories also doesn't feed toobig blocks to
4675                          * pass 2.
4676                          */
4677                         if (problem == PR_1_TOOBIG_DIR)
4678                                 return ret_code;
4679                         goto mark_dir;
4680                 } else
4681                         return 0;
4682         }
4683
4684         if (p->ino == EXT2_RESIZE_INO) {
4685                 /*
4686                  * The resize inode has already be sanity checked
4687                  * during pass #0 (the superblock checks).  All we
4688                  * have to do is mark the double indirect block as
4689                  * being in use; all of the other blocks are handled
4690                  * by mark_table_blocks()).
4691                  */
4692                 if (blockcnt == BLOCK_COUNT_DIND)
4693                         mark_block_used(ctx, blk);
4694                 p->num_blocks++;
4695         } else if (!(ctx->fs->cluster_ratio_bits &&
4696                      p->previous_block &&
4697                      (EXT2FS_B2C(ctx->fs, blk) ==
4698                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
4699                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
4700                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
4701                 mark_block_used(ctx, blk);
4702                 p->num_blocks++;
4703         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
4704                                              p->last_block, blk, blockcnt)) {
4705                 pctx->blk = blockcnt;
4706                 pctx->blk2 = blk;
4707                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4708                 mark_block_used(ctx, blk);
4709                 mark_block_used(ctx, blk);
4710         }
4711         if (blockcnt >= 0)
4712                 p->last_block = blockcnt;
4713         p->previous_block = blk;
4714 mark_dir:
4715         if (p->is_dir && (blockcnt >= 0)) {
4716                 while (++p->last_db_block < blockcnt) {
4717                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
4718                                                               p->ino, 0,
4719                                                               p->last_db_block);
4720                         if (pctx->errcode) {
4721                                 pctx->blk = 0;
4722                                 pctx->num = p->last_db_block;
4723                                 goto failed_add_dir_block;
4724                         }
4725                 }
4726                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
4727                                                       blk, blockcnt);
4728                 if (pctx->errcode) {
4729                         pctx->blk = blk;
4730                         pctx->num = blockcnt;
4731                 failed_add_dir_block:
4732                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4733                         /* Should never get here */
4734                         ctx->flags |= E2F_FLAG_ABORT;
4735                         return BLOCK_ABORT;
4736                 }
4737         }
4738         return ret_code;
4739 }
4740
4741 static int process_bad_block(ext2_filsys fs,
4742                       blk64_t *block_nr,
4743                       e2_blkcnt_t blockcnt,
4744                       blk64_t ref_block EXT2FS_ATTR((unused)),
4745                       int ref_offset EXT2FS_ATTR((unused)),
4746                       void *priv_data)
4747 {
4748         struct process_block_struct *p;
4749         blk64_t         blk = *block_nr;
4750         blk64_t         first_block;
4751         dgrp_t          i;
4752         struct problem_context *pctx;
4753         e2fsck_t        ctx;
4754
4755         if (!blk)
4756                 return 0;
4757
4758         p = (struct process_block_struct *) priv_data;
4759         ctx = p->ctx;
4760         pctx = p->pctx;
4761
4762         pctx->ino = EXT2_BAD_INO;
4763         pctx->blk = blk;
4764         pctx->blkcount = blockcnt;
4765
4766         if ((blk < fs->super->s_first_data_block) ||
4767             (blk >= ext2fs_blocks_count(fs->super))) {
4768                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
4769                         *block_nr = 0;
4770                         return BLOCK_CHANGED;
4771                 } else
4772                         return 0;
4773         }
4774
4775         if (blockcnt < 0) {
4776                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
4777                         p->bbcheck = 1;
4778                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
4779                                 *block_nr = 0;
4780                                 return BLOCK_CHANGED;
4781                         }
4782                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4783                                                     blk)) {
4784                         p->bbcheck = 1;
4785                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
4786                                         pctx)) {
4787                                 *block_nr = 0;
4788                                 return BLOCK_CHANGED;
4789                         }
4790                         if (e2fsck_should_abort(ctx))
4791                                 return BLOCK_ABORT;
4792                 } else
4793                         mark_block_used(ctx, blk);
4794                 return 0;
4795         }
4796 #if 0
4797         printf ("DEBUG: Marking %u as bad.\n", blk);
4798 #endif
4799         ctx->fs_badblocks_count++;
4800         /*
4801          * If the block is not used, then mark it as used and return.
4802          * If it is already marked as found, this must mean that
4803          * there's an overlap between the filesystem table blocks
4804          * (bitmaps and inode table) and the bad block list.
4805          */
4806         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
4807                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4808                 return 0;
4809         }
4810         /*
4811          * Try to find the where the filesystem block was used...
4812          */
4813         first_block = fs->super->s_first_data_block;
4814
4815         for (i = 0; i < fs->group_desc_count; i++ ) {
4816                 pctx->group = i;
4817                 pctx->blk = blk;
4818                 if (!ext2fs_bg_has_super(fs, i))
4819                         goto skip_super;
4820                 if (blk == first_block) {
4821                         if (i == 0) {
4822                                 if (fix_problem(ctx,
4823                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
4824                                                 pctx)) {
4825                                         *block_nr = 0;
4826                                         return BLOCK_CHANGED;
4827                                 }
4828                                 return 0;
4829                         }
4830                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
4831                         return 0;
4832                 }
4833                 if ((blk > first_block) &&
4834                     (blk <= first_block + fs->desc_blocks)) {
4835                         if (i == 0) {
4836                                 pctx->blk = *block_nr;
4837                                 if (fix_problem(ctx,
4838                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
4839                                         *block_nr = 0;
4840                                         return BLOCK_CHANGED;
4841                                 }
4842                                 return 0;
4843                         }
4844                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
4845                         return 0;
4846                 }
4847         skip_super:
4848                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
4849                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
4850                                 ctx->invalid_block_bitmap_flag[i]++;
4851                                 ctx->invalid_bitmaps++;
4852                         }
4853                         return 0;
4854                 }
4855                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
4856                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
4857                                 ctx->invalid_inode_bitmap_flag[i]++;
4858                                 ctx->invalid_bitmaps++;
4859                         }
4860                         return 0;
4861                 }
4862                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
4863                     (blk < (ext2fs_inode_table_loc(fs, i) +
4864                             fs->inode_blocks_per_group))) {
4865                         /*
4866                          * If there are bad blocks in the inode table,
4867                          * the inode scan code will try to do
4868                          * something reasonable automatically.
4869                          */
4870                         return 0;
4871                 }
4872                 first_block += fs->super->s_blocks_per_group;
4873         }
4874         /*
4875          * If we've gotten to this point, then the only
4876          * possibility is that the bad block inode meta data
4877          * is using a bad block.
4878          */
4879         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
4880             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
4881             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
4882                 p->bbcheck = 1;
4883                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
4884                         *block_nr = 0;
4885                         return BLOCK_CHANGED;
4886                 }
4887                 if (e2fsck_should_abort(ctx))
4888                         return BLOCK_ABORT;
4889                 return 0;
4890         }
4891
4892         pctx->group = -1;
4893
4894         /* Warn user that the block wasn't claimed */
4895         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
4896
4897         return 0;
4898 }
4899
4900 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
4901                             const char *name, int num, blk64_t *new_block)
4902 {
4903         ext2_filsys fs = ctx->fs;
4904         dgrp_t          last_grp;
4905         blk64_t         old_block = *new_block;
4906         blk64_t         last_block;
4907         dgrp_t          flexbg;
4908         unsigned        flexbg_size;
4909         int             i, is_flexbg;
4910         char            *buf;
4911         struct problem_context  pctx;
4912
4913         clear_problem_context(&pctx);
4914
4915         pctx.group = group;
4916         pctx.blk = old_block;
4917         pctx.str = name;
4918
4919         /*
4920          * For flex_bg filesystems, first try to allocate the metadata
4921          * within the flex_bg, and if that fails then try finding the
4922          * space anywhere in the filesystem.
4923          */
4924         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
4925         if (is_flexbg) {
4926                 flexbg_size = 1U << fs->super->s_log_groups_per_flex;
4927                 flexbg = group / flexbg_size;
4928                 first_block = ext2fs_group_first_block2(fs,
4929                                                         flexbg_size * flexbg);
4930                 last_grp = group | (flexbg_size - 1);
4931                 if (last_grp >= fs->group_desc_count)
4932                         last_grp = fs->group_desc_count - 1;
4933                 last_block = ext2fs_group_last_block2(fs, last_grp);
4934         } else
4935                 last_block = ext2fs_group_last_block2(fs, group);
4936         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
4937                                                num, ctx->block_found_map,
4938                                                new_block);
4939         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
4940                 pctx.errcode = ext2fs_get_free_blocks2(fs,
4941                                 fs->super->s_first_data_block,
4942                                 ext2fs_blocks_count(fs->super),
4943                                 num, ctx->block_found_map, new_block);
4944         if (pctx.errcode) {
4945                 pctx.num = num;
4946                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4947                 ext2fs_unmark_valid(fs);
4948                 ctx->flags |= E2F_FLAG_ABORT;
4949                 return;
4950         }
4951         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4952         if (pctx.errcode) {
4953                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4954                 ext2fs_unmark_valid(fs);
4955                 ctx->flags |= E2F_FLAG_ABORT;
4956                 return;
4957         }
4958         ext2fs_mark_super_dirty(fs);
4959         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4960         pctx.blk2 = *new_block;
4961         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4962                           PR_1_RELOC_TO), &pctx);
4963         pctx.blk2 = 0;
4964         for (i = 0; i < num; i++) {
4965                 pctx.blk = i;
4966                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4967                 if (old_block) {
4968                         pctx.errcode = io_channel_read_blk64(fs->io,
4969                                    old_block + i, 1, buf);
4970                         if (pctx.errcode)
4971                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4972                         pctx.blk = (*new_block) + i;
4973                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4974                                                               1, buf);
4975                 } else {
4976                         pctx.blk = (*new_block) + i;
4977                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4978                                                            NULL, NULL);
4979                 }
4980
4981                 if (pctx.errcode)
4982                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4983         }
4984         ext2fs_free_mem(&buf);
4985 }
4986
4987 /*
4988  * This routine gets called at the end of pass 1 if bad blocks are
4989  * detected in the superblock, group descriptors, inode_bitmaps, or
4990  * block bitmaps.  At this point, all of the blocks have been mapped
4991  * out, so we can try to allocate new block(s) to replace the bad
4992  * blocks.
4993  */
4994 static void handle_fs_bad_blocks(e2fsck_t ctx)
4995 {
4996         ext2_filsys fs = ctx->fs;
4997         dgrp_t          i;
4998         blk64_t         first_block;
4999         blk64_t         new_blk;
5000
5001         for (i = 0; i < fs->group_desc_count; i++) {
5002                 first_block = ext2fs_group_first_block2(fs, i);
5003
5004                 if (ctx->invalid_block_bitmap_flag[i]) {
5005                         new_blk = ext2fs_block_bitmap_loc(fs, i);
5006                         new_table_block(ctx, first_block, i, _("block bitmap"),
5007                                         1, &new_blk);
5008                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
5009                 }
5010                 if (ctx->invalid_inode_bitmap_flag[i]) {
5011                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
5012                         new_table_block(ctx, first_block, i, _("inode bitmap"),
5013                                         1, &new_blk);
5014                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
5015                 }
5016                 if (ctx->invalid_inode_table_flag[i]) {
5017                         new_blk = ext2fs_inode_table_loc(fs, i);
5018                         new_table_block(ctx, first_block, i, _("inode table"),
5019                                         fs->inode_blocks_per_group,
5020                                         &new_blk);
5021                         ext2fs_inode_table_loc_set(fs, i, new_blk);
5022                         ctx->flags |= E2F_FLAG_RESTART;
5023                 }
5024         }
5025         ctx->invalid_bitmaps = 0;
5026 }
5027
5028 /*
5029  * This routine marks all blocks which are used by the superblock,
5030  * group descriptors, inode bitmaps, and block bitmaps.
5031  */
5032 static void mark_table_blocks(e2fsck_t ctx)
5033 {
5034         ext2_filsys fs = ctx->fs;
5035         blk64_t b;
5036         dgrp_t  i;
5037         unsigned int    j;
5038         struct problem_context pctx;
5039
5040         clear_problem_context(&pctx);
5041
5042         for (i = 0; i < fs->group_desc_count; i++) {
5043                 pctx.group = i;
5044
5045                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
5046                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
5047
5048                 /*
5049                  * Mark the blocks used for the inode table
5050                  */
5051                 if (ext2fs_inode_table_loc(fs, i)) {
5052                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
5053                              j < fs->inode_blocks_per_group;
5054                              j++, b++) {
5055                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5056                                                              b)) {
5057                                         pctx.blk = b;
5058                                         if (!ctx->invalid_inode_table_flag[i] &&
5059                                             fix_problem(ctx,
5060                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
5061                                                 ctx->invalid_inode_table_flag[i]++;
5062                                                 ctx->invalid_bitmaps++;
5063                                         }
5064                                 } else {
5065                                     ext2fs_mark_block_bitmap2(
5066                                                 ctx->block_found_map, b);
5067                                     ext2fs_mark_block_bitmap2(
5068                                                 ctx->block_metadata_map, b);
5069                                 }
5070                         }
5071                 }
5072
5073                 /*
5074                  * Mark block used for the block bitmap
5075                  */
5076                 if (ext2fs_block_bitmap_loc(fs, i)) {
5077                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5078                                      ext2fs_block_bitmap_loc(fs, i))) {
5079                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
5080                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
5081                                         ctx->invalid_block_bitmap_flag[i]++;
5082                                         ctx->invalid_bitmaps++;
5083                                 }
5084                         } else {
5085                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
5086                                      ext2fs_block_bitmap_loc(fs, i));
5087                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5088                                      ext2fs_block_bitmap_loc(fs, i));
5089                         }
5090                 }
5091                 /*
5092                  * Mark block used for the inode bitmap
5093                  */
5094                 if (ext2fs_inode_bitmap_loc(fs, i)) {
5095                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5096                                      ext2fs_inode_bitmap_loc(fs, i))) {
5097                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
5098                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
5099                                         ctx->invalid_inode_bitmap_flag[i]++;
5100                                         ctx->invalid_bitmaps++;
5101                                 }
5102                         } else {
5103                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5104                                      ext2fs_inode_bitmap_loc(fs, i));
5105                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
5106                                      ext2fs_inode_bitmap_loc(fs, i));
5107                         }
5108                 }
5109         }
5110 }
5111
5112 /*
5113  * These subroutines short circuits ext2fs_get_blocks and
5114  * ext2fs_check_directory; we use them since we already have the inode
5115  * structure, so there's no point in letting the ext2fs library read
5116  * the inode again.
5117  */
5118 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
5119                                   blk_t *blocks)
5120 {
5121         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5122         int     i;
5123
5124         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5125                 return EXT2_ET_CALLBACK_NOTHANDLED;
5126
5127         for (i=0; i < EXT2_N_BLOCKS; i++)
5128                 blocks[i] = ctx->stashed_inode->i_block[i];
5129         return 0;
5130 }
5131
5132 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
5133                                   struct ext2_inode *inode)
5134 {
5135         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5136
5137         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5138                 return EXT2_ET_CALLBACK_NOTHANDLED;
5139         *inode = *ctx->stashed_inode;
5140         return 0;
5141 }
5142
5143 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
5144                             struct ext2_inode *inode)
5145 {
5146         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5147
5148         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
5149                 (inode != ctx->stashed_inode))
5150                 *ctx->stashed_inode = *inode;
5151         return EXT2_ET_CALLBACK_NOTHANDLED;
5152 }
5153
5154 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
5155 {
5156         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5157
5158         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5159                 return EXT2_ET_CALLBACK_NOTHANDLED;
5160
5161         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
5162                 return EXT2_ET_NO_DIRECTORY;
5163         return 0;
5164 }
5165
5166 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
5167                                         blk64_t *ret)
5168 {
5169         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5170         errcode_t       retval;
5171         blk64_t         new_block;
5172
5173         if (ctx->block_found_map) {
5174                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
5175                                            &new_block);
5176                 if (retval)
5177                         return retval;
5178                 if (fs->block_map) {
5179                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
5180                         ext2fs_mark_bb_dirty(fs);
5181                 }
5182         } else {
5183                 if (!fs->block_map) {
5184                         retval = ext2fs_read_block_bitmap(fs);
5185                         if (retval)
5186                                 return retval;
5187                 }
5188
5189                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
5190                 if (retval)
5191                         return retval;
5192         }
5193
5194         *ret = new_block;
5195         return (0);
5196 }
5197
5198 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
5199                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
5200 {
5201         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5202         errcode_t       retval;
5203
5204         if (ctx->block_found_map)
5205                 return ext2fs_new_range(fs, flags, goal, len,
5206                                         ctx->block_found_map, pblk, plen);
5207
5208         if (!fs->block_map) {
5209                 retval = ext2fs_read_block_bitmap(fs);
5210                 if (retval)
5211                         return retval;
5212         }
5213
5214         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
5215                                 pblk, plen);
5216 }
5217
5218 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
5219 {
5220         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5221
5222         /* Never free a critical metadata block */
5223         if (ctx->block_found_map &&
5224             ctx->block_metadata_map &&
5225             inuse < 0 &&
5226             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
5227                 return;
5228
5229         if (ctx->block_found_map) {
5230                 if (inuse > 0)
5231                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
5232                 else
5233                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
5234         }
5235 }
5236
5237 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
5238                                            blk_t num, int inuse)
5239 {
5240         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5241
5242         /* Never free a critical metadata block */
5243         if (ctx->block_found_map &&
5244             ctx->block_metadata_map &&
5245             inuse < 0 &&
5246             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
5247                 return;
5248
5249         if (ctx->block_found_map) {
5250                 if (inuse > 0)
5251                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
5252                                                         blk, num);
5253                 else
5254                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
5255                                                         blk, num);
5256         }
5257 }
5258
5259 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
5260 {
5261         ext2_filsys fs = ctx->fs;
5262
5263         if (use_shortcuts) {
5264                 fs->get_blocks = pass1_get_blocks;
5265                 fs->check_directory = pass1_check_directory;
5266                 fs->read_inode = pass1_read_inode;
5267                 fs->write_inode = pass1_write_inode;
5268                 ctx->stashed_ino = 0;
5269         } else {
5270                 fs->get_blocks = 0;
5271                 fs->check_directory = 0;
5272                 fs->read_inode = 0;
5273                 fs->write_inode = 0;
5274         }
5275 }
5276
5277 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
5278 {
5279         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
5280         ext2fs_set_block_alloc_stats_callback(ctx->fs,
5281                                                 e2fsck_block_alloc_stats, 0);
5282         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
5283         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
5284                                         e2fsck_block_alloc_stats_range, NULL);
5285 }