Whamcloud - gitweb
e2fsck: merge EA blocks properly
[tools/e2fsprogs.git] / e2fsck / pass1.c
1 /*
2  * pass1.c -- pass #1 of e2fsck: sequential scan of the inode table
3  *
4  * Copyright (C) 1993, 1994, 1995, 1996, 1997 Theodore Ts'o.
5  *
6  * %Begin-Header%
7  * This file may be redistributed under the terms of the GNU Public
8  * License.
9  * %End-Header%
10  *
11  * Pass 1 of e2fsck iterates over all the inodes in the filesystems,
12  * and applies the following tests to each inode:
13  *
14  *      - The mode field of the inode must be legal.
15  *      - The size and block count fields of the inode are correct.
16  *      - A data block must not be used by another inode
17  *
18  * Pass 1 also gathers the collects the following information:
19  *
20  *      - A bitmap of which inodes are in use.          (inode_used_map)
21  *      - A bitmap of which inodes are directories.     (inode_dir_map)
22  *      - A bitmap of which inodes are regular files.   (inode_reg_map)
23  *      - A bitmap of which inodes have bad fields.     (inode_bad_map)
24  *      - A bitmap of which inodes are in bad blocks.   (inode_bb_map)
25  *      - A bitmap of which inodes are imagic inodes.   (inode_imagic_map)
26  *      - A bitmap of which inodes are casefolded.      (inode_casefold_map)
27  *      - A bitmap of which blocks are in use.          (block_found_map)
28  *      - A bitmap of which blocks are in use by two inodes     (block_dup_map)
29  *      - The data blocks of the directory inodes.      (dir_map)
30  *      - Ref counts for ea_inodes.                     (ea_inode_refs)
31  *      - The encryption policy ID of each encrypted inode. (encrypted_files)
32  *
33  * Pass 1 is designed to stash away enough information so that the
34  * other passes should not need to read in the inode information
35  * during the normal course of a filesystem check.  (Although if an
36  * inconsistency is detected, other passes may need to read in an
37  * inode to fix it.)
38  *
39  * Note that pass 1B will be invoked if there are any duplicate blocks
40  * found.
41  */
42
43 #define _GNU_SOURCE 1 /* get strnlen() */
44 #include "config.h"
45 #include <string.h>
46 #include <time.h>
47 #ifdef HAVE_ERRNO_H
48 #include <errno.h>
49 #endif
50 #include <assert.h>
51 #ifdef HAVE_PTHREAD
52 #include <pthread.h>
53 #endif
54
55 #include "e2fsck.h"
56 #include <ext2fs/ext2_ext_attr.h>
57 /* todo remove this finally */
58 #include <ext2fs/ext2fsP.h>
59 #include <e2p/e2p.h>
60
61 #include "problem.h"
62
63 #ifdef NO_INLINE_FUNCS
64 #define _INLINE_
65 #else
66 #define _INLINE_ inline
67 #endif
68
69 #undef DEBUG
70
71 struct ea_quota {
72         blk64_t blocks;
73         __u64 inodes;
74 };
75
76 static int process_block(ext2_filsys fs, blk64_t        *blocknr,
77                          e2_blkcnt_t blockcnt, blk64_t ref_blk,
78                          int ref_offset, void *priv_data);
79 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
80                              e2_blkcnt_t blockcnt, blk64_t ref_blk,
81                              int ref_offset, void *priv_data);
82 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
83                          char *block_buf,
84                          const struct ea_quota *ea_ibody_quota);
85 static void mark_table_blocks(e2fsck_t ctx);
86 static void alloc_bb_map(e2fsck_t ctx);
87 static void alloc_imagic_map(e2fsck_t ctx);
88 static void mark_inode_bad(e2fsck_t ctx, ino_t ino);
89 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino);
90 static void handle_fs_bad_blocks(e2fsck_t ctx);
91 static 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;
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_entry2(ctx->fs, entry, NULL, &hash);
352         if (retval) {
353                 com_err("check_large_ea_inode", retval,
354                         _("while hashing entry with e_value_inum = %u"),
355                         entry->e_value_inum);
356                 fatal_error(ctx, 0);
357         }
358
359         if (hash == entry->e_hash) {
360                 *quota_blocks = size_to_quota_blocks(ctx->fs,
361                                                      entry->e_value_size);
362         } else {
363                 /* This might be an old Lustre-style ea_inode reference. */
364                 if (inode.i_mtime == pctx->ino &&
365                     inode.i_generation == pctx->inode->i_generation) {
366                         *quota_blocks = 0;
367                 } else {
368                         /* If target inode is also missing EA_INODE flag,
369                          * this is likely to be a bad reference.
370                          */
371                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
372                                 pctx->num = entry->e_value_inum;
373                                 return PR_1_ATTR_VALUE_EA_INODE;
374                         } else {
375                                 pctx->num = entry->e_hash;
376                                 return PR_1_ATTR_HASH;
377                         }
378                 }
379         }
380
381         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
382                 pctx->num = entry->e_value_inum;
383                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
384                         inode.i_flags |= EXT4_EA_INODE_FL;
385                         e2fsck_pass1_fix_lock(ctx);
386                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
387                                            &inode);
388                         e2fsck_pass1_fix_unlock(ctx);
389                 } else {
390                         return PR_1_ATTR_NO_EA_INODE_FL;
391                 }
392         }
393         return 0;
394 }
395
396 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
397                               struct ext2_ext_attr_entry *first, void *end)
398 {
399         struct ext2_ext_attr_entry *entry;
400
401         for (entry = first;
402              (void *)entry < end && !EXT2_EXT_IS_LAST_ENTRY(entry);
403              entry = EXT2_EXT_ATTR_NEXT(entry)) {
404                 if (!entry->e_value_inum)
405                         continue;
406                 if (!ctx->ea_inode_refs) {
407                         pctx->errcode = ea_refcount_create(0,
408                                                            &ctx->ea_inode_refs);
409                         if (pctx->errcode) {
410                                 pctx->num = 4;
411                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
412                                 ctx->flags |= E2F_FLAG_ABORT;
413                                 return;
414                         }
415                 }
416                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
417                                       0);
418         }
419 }
420
421 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
422                               struct ea_quota *ea_ibody_quota)
423 {
424         struct ext2_super_block *sb = ctx->fs->super;
425         struct ext2_inode_large *inode;
426         struct ext2_ext_attr_entry *entry;
427         char *start, *header, *end;
428         unsigned int storage_size, remain;
429         problem_t problem = 0;
430         region_t region = 0;
431
432         ea_ibody_quota->blocks = 0;
433         ea_ibody_quota->inodes = 0;
434
435         inode = (struct ext2_inode_large *) pctx->inode;
436         storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
437                 inode->i_extra_isize;
438         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
439                  inode->i_extra_isize;
440         end = header + storage_size;
441         start = header + sizeof(__u32);
442         entry = (struct ext2_ext_attr_entry *) start;
443
444         /* scan all entry's headers first */
445
446         /* take finish entry 0UL into account */
447         remain = storage_size - sizeof(__u32);
448
449         region = region_create(0, storage_size);
450         if (!region) {
451                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
452                 problem = 0;
453                 ctx->flags |= E2F_FLAG_ABORT;
454                 return;
455         }
456         if (region_allocate(region, 0, sizeof(__u32))) {
457                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
458                 goto fix;
459         }
460
461         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
462                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
463                 __u32 hash;
464
465                 if (region_allocate(region, (char *)entry - (char *)header,
466                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
467                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
468                         goto fix;
469                 }
470
471                 /* header eats this space */
472                 remain -= sizeof(struct ext2_ext_attr_entry);
473
474                 /* is attribute name valid? */
475                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
476                         pctx->num = entry->e_name_len;
477                         problem = PR_1_ATTR_NAME_LEN;
478                         goto fix;
479                 }
480
481                 /* attribute len eats this space */
482                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
483
484                 if (entry->e_value_inum == 0) {
485                         /* check value size */
486                         if (entry->e_value_size > remain) {
487                                 pctx->num = entry->e_value_size;
488                                 problem = PR_1_ATTR_VALUE_SIZE;
489                                 goto fix;
490                         }
491
492                         if (entry->e_value_size &&
493                             region_allocate(region,
494                                             sizeof(__u32) + entry->e_value_offs,
495                                             EXT2_EXT_ATTR_SIZE(
496                                                 entry->e_value_size))) {
497                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
498                                 goto fix;
499                         }
500
501                         hash = ext2fs_ext_attr_hash_entry(entry,
502                                                           start + entry->e_value_offs);
503
504                         /* e_hash may be 0 in older inode's ea */
505                         if (entry->e_hash != 0 && entry->e_hash != hash) {
506                                 pctx->num = entry->e_hash;
507                                 problem = PR_1_ATTR_HASH;
508                                 goto fix;
509                         }
510                 } else {
511                         blk64_t quota_blocks;
512
513                         problem = check_large_ea_inode(ctx, entry, pctx,
514                                                        &quota_blocks);
515                         if (problem != 0)
516                                 goto fix;
517
518                         ea_ibody_quota->blocks += quota_blocks;
519                         ea_ibody_quota->inodes++;
520                 }
521
522                 /* If EA value is stored in external inode then it does not
523                  * consume space here */
524                 if (entry->e_value_inum == 0)
525                         remain -= entry->e_value_size;
526
527                 entry = EXT2_EXT_ATTR_NEXT(entry);
528         }
529
530         if (region_allocate(region, (char *)entry - (char *)header,
531                             sizeof(__u32))) {
532                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
533                 goto fix;
534         }
535 fix:
536         if (region)
537                 region_free(region);
538         /*
539          * it seems like a corruption. it's very unlikely we could repair
540          * EA(s) in automatic fashion -bzzz
541          */
542         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
543                 inc_ea_inode_refs(ctx, pctx,
544                                   (struct ext2_ext_attr_entry *)start, end);
545                 return;
546         }
547
548         /* simply remove all possible EA(s) */
549         *((__u32 *)header) = 0UL;
550         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
551                                 EXT2_INODE_SIZE(sb), "pass1");
552         ea_ibody_quota->blocks = 0;
553         ea_ibody_quota->inodes = 0;
554 }
555
556 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
557         return (xtime & (1U << 31)) != 0 &&
558                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
559 }
560
561 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
562         check_inode_extra_negative_epoch(inode->i_##xtime, \
563                                          inode->i_##xtime##_extra)
564
565 /* When today's date is earlier than 2242, we assume that atimes,
566  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
567  * actually pre-1970 dates mis-encoded.
568  */
569 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
570
571 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
572                                     struct ea_quota *ea_ibody_quota)
573 {
574         struct ext2_super_block *sb = ctx->fs->super;
575         struct ext2_inode_large *inode;
576         __u32 *eamagic;
577         int min, max;
578
579         ea_ibody_quota->blocks = 0;
580         ea_ibody_quota->inodes = 0;
581
582         inode = (struct ext2_inode_large *) pctx->inode;
583         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
584                 /* this isn't large inode. so, nothing to check */
585                 return;
586         }
587
588 #if 0
589         printf("inode #%u, i_extra_size %d\n", pctx->ino,
590                         inode->i_extra_isize);
591 #endif
592         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
593         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
594         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
595         /*
596          * For now we will allow i_extra_isize to be 0, but really
597          * implementations should never allow i_extra_isize to be 0
598          */
599         if (inode->i_extra_isize &&
600             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
601              inode->i_extra_isize & 3)) {
602                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
603                         return;
604                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
605                         inode->i_extra_isize = sb->s_want_extra_isize;
606                 else
607                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
608                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
609                                         EXT2_INODE_SIZE(sb), "pass1");
610         }
611
612         /* check if there is no place for an EA header */
613         if (inode->i_extra_isize >= max - sizeof(__u32))
614                 return;
615
616         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
617                         inode->i_extra_isize);
618         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
619                 /* it seems inode has an extended attribute(s) in body */
620                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
621         }
622
623         /*
624          * If the inode's extended atime (ctime, crtime, mtime) is stored in
625          * the old, invalid format, repair it.
626          */
627         if (((sizeof(time_t) <= 4) ||
628              (((sizeof(time_t) > 4) &&
629                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
630             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
631              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
632              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
633              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
634
635                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
636                         return;
637
638                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
639                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
640                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
641                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
642                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
643                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
644                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
645                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
646                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
647                                         EXT2_INODE_SIZE(sb), "pass1");
648         }
649
650 }
651
652 /*
653  * Check to see if the inode might really be a directory, despite i_mode
654  *
655  * This is a lot of complexity for something for which I'm not really
656  * convinced happens frequently in the wild.  If for any reason this
657  * causes any problems, take this code out.
658  * [tytso:20070331.0827EDT]
659  */
660 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
661                                 char *buf)
662 {
663         struct ext2_inode *inode = pctx->inode;
664         struct ext2_dir_entry   *dirent;
665         errcode_t               retval;
666         blk64_t                 blk;
667         unsigned int            i, rec_len, not_device = 0;
668         int                     extent_fs;
669         int                     inlinedata_fs;
670
671         /*
672          * If the mode looks OK, we believe it.  If the first block in
673          * the i_block array is 0, this cannot be a directory. If the
674          * inode is extent-mapped, it is still the case that the latter
675          * cannot be 0 - the magic number in the extent header would make
676          * it nonzero.
677          */
678         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
679             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
680                 return;
681
682         /*
683          * Check the block numbers in the i_block array for validity:
684          * zero blocks are skipped (but the first one cannot be zero -
685          * see above), other blocks are checked against the first and
686          * max data blocks (from the the superblock) and against the
687          * block bitmap. Any invalid block found means this cannot be
688          * a directory.
689          *
690          * If there are non-zero blocks past the fourth entry, then
691          * this cannot be a device file: we remember that for the next
692          * check.
693          *
694          * For extent mapped files, we don't do any sanity checking:
695          * just try to get the phys block of logical block 0 and run
696          * with it.
697          *
698          * For inline data files, we just try to get the size of inline
699          * data.  If it's true, we will treat it as a directory.
700          */
701
702         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
703         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
704         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
705                 size_t size;
706                 __u32 dotdot;
707                 unsigned int rec_len2;
708                 struct ext2_dir_entry de;
709
710                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
711                         return;
712                 /*
713                  * If the size isn't a multiple of 4, it's probably not a
714                  * directory??
715                  */
716                 if (size & 3)
717                         return;
718                 /*
719                  * If the first 10 bytes don't look like a directory entry,
720                  * it's probably not a directory.
721                  */
722                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
723                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
724                        EXT2_DIR_REC_LEN(0));
725                 dotdot = ext2fs_le32_to_cpu(dotdot);
726                 de.inode = ext2fs_le32_to_cpu(de.inode);
727                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
728                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
729                 if (dotdot >= ctx->fs->super->s_inodes_count ||
730                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
731                      dotdot != EXT2_ROOT_INO) ||
732                     de.inode >= ctx->fs->super->s_inodes_count ||
733                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
734                      de.inode != 0) ||
735                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
736                               EXT4_INLINE_DATA_DOTDOT_SIZE)
737                         return;
738                 /* device files never have a "system.data" entry */
739                 goto isdir;
740         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
741                 /* extent mapped */
742                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
743                                  &blk))
744                         return;
745                 /* device files are never extent mapped */
746                 not_device++;
747         } else {
748                 for (i=0; i < EXT2_N_BLOCKS; i++) {
749                         blk = inode->i_block[i];
750                         if (!blk)
751                                 continue;
752                         if (i >= 4)
753                                 not_device++;
754
755                         e2fsck_pass1_block_map_lock(ctx);
756                         if (blk < ctx->fs->super->s_first_data_block ||
757                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
758                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
759                                                            blk)) {
760                                 e2fsck_pass1_block_map_unlock(ctx);
761                                 return; /* Invalid block, can't be dir */
762                         }
763                         e2fsck_pass1_block_map_unlock(ctx);
764                 }
765                 blk = inode->i_block[0];
766         }
767
768         /*
769          * If the mode says this is a device file and the i_links_count field
770          * is sane and we have not ruled it out as a device file previously,
771          * we declare it a device file, not a directory.
772          */
773         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
774             (inode->i_links_count == 1) && !not_device)
775                 return;
776
777         /* read the first block */
778         ehandler_operation(_("reading directory block"));
779         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
780         ehandler_operation(0);
781         if (retval)
782                 return;
783
784         dirent = (struct ext2_dir_entry *) buf;
785         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
786         if (retval)
787                 return;
788         if ((ext2fs_dirent_name_len(dirent) != 1) ||
789             (dirent->name[0] != '.') ||
790             (dirent->inode != pctx->ino) ||
791             (rec_len < 12) ||
792             (rec_len % 4) ||
793             (rec_len >= ctx->fs->blocksize - 12))
794                 return;
795
796         dirent = (struct ext2_dir_entry *) (buf + rec_len);
797         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
798         if (retval)
799                 return;
800         if ((ext2fs_dirent_name_len(dirent) != 2) ||
801             (dirent->name[0] != '.') ||
802             (dirent->name[1] != '.') ||
803             (rec_len < 12) ||
804             (rec_len % 4))
805                 return;
806
807 isdir:
808         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
809                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
810                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
811                                         EXT2_INODE_SIZE(ctx->fs->super),
812                                         "check_is_really_dir");
813         }
814 }
815
816 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
817                                      int flags, ext2_icount_t hint,
818                                      ext2_icount_t *ret)
819 {
820         unsigned int            threshold;
821         unsigned int            save_type;
822         ext2_ino_t              num_dirs;
823         errcode_t               retval;
824         char                    *tdb_dir;
825         int                     enable;
826
827         *ret = 0;
828
829         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
830                            &tdb_dir);
831         profile_get_uint(ctx->profile, "scratch_files",
832                          "numdirs_threshold", 0, 0, &threshold);
833         profile_get_boolean(ctx->profile, "scratch_files",
834                             "icount", 0, 1, &enable);
835
836         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
837         if (retval)
838                 num_dirs = 1024;        /* Guess */
839
840         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
841             (!threshold || num_dirs > threshold)) {
842                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
843                                                   flags, ret);
844                 if (retval == 0)
845                         return 0;
846         }
847         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
848                                &save_type);
849         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
850                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
851         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
852         ctx->fs->default_bitmap_type = save_type;
853         return retval;
854 }
855
856 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
857                                             e2fsck_t ctx,
858                                             struct problem_context *pctx)
859 {
860         errcode_t retval;
861         struct ext2_inode_large inode;
862
863         /*
864          * Reread inode.  If we don't see checksum error, then this inode
865          * has been fixed elsewhere.
866          */
867         ctx->stashed_ino = 0;
868         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
869                                         sizeof(inode));
870         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
871                 return retval;
872         if (!retval)
873                 return 0;
874
875         /*
876          * Checksum still doesn't match.  That implies that the inode passes
877          * all the sanity checks, so maybe the checksum is simply corrupt.
878          * See if the user will go for fixing that.
879          */
880         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
881                 return 0;
882
883
884         e2fsck_pass1_fix_lock(ctx);
885         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
886                                          sizeof(inode));
887         e2fsck_pass1_fix_unlock(ctx);
888         return retval;
889 }
890
891 static void reserve_block_for_root_repair(e2fsck_t ctx)
892 {
893         blk64_t         blk = 0;
894         errcode_t       err;
895         ext2_filsys     fs = ctx->fs;
896
897         ctx->root_repair_block = 0;
898         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
899                 return;
900
901         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
902         if (err)
903                 return;
904         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
905         ctx->root_repair_block = blk;
906 }
907
908 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
909 {
910         blk64_t         blk = 0;
911         errcode_t       err;
912         ext2_filsys     fs = ctx->fs;
913         static const char name[] = "lost+found";
914         ext2_ino_t      ino;
915
916         ctx->lnf_repair_block = 0;
917         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
918                 return;
919
920         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
921         if (err)
922                 return;
923         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
924         ctx->lnf_repair_block = blk;
925         return;
926 }
927
928 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
929                                          size_t *sz)
930 {
931         void *p;
932         struct ext2_xattr_handle *handle;
933         errcode_t retval;
934
935         retval = ext2fs_xattrs_open(fs, ino, &handle);
936         if (retval)
937                 return retval;
938
939         retval = ext2fs_xattrs_read(handle);
940         if (retval)
941                 goto err;
942
943         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
944         if (retval)
945                 goto err;
946         ext2fs_free_mem(&p);
947 err:
948         (void) ext2fs_xattrs_close(&handle);
949         return retval;
950 }
951
952 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
953                                     struct problem_context *pctx,
954                                     int failed_csum)
955 {
956         if (!failed_csum)
957                 return;
958
959         /*
960          * If the inode failed the checksum and the user didn't
961          * clear the inode, test the checksum again -- if it still
962          * fails, ask the user if the checksum should be corrected.
963          */
964         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
965         if (pctx->errcode)
966                 ctx->flags |= E2F_FLAG_ABORT;
967 }
968 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
969         do { \
970                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
971                 if ((ctx)->flags & E2F_FLAG_ABORT) \
972                         return; \
973         } while (0)
974
975 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
976 {
977         __u32 x;
978         int i;
979
980         for (i = 0; i < EXT2_N_BLOCKS; i++) {
981                 x = inode->i_block[i];
982 #ifdef WORDS_BIGENDIAN
983                 x = ext2fs_swab32(x);
984 #endif
985                 if (x >= ext2fs_blocks_count(fs->super))
986                         return 0;
987         }
988
989         return 1;
990 }
991
992 /*
993  * Figure out what to do with an inode that has both extents and inline data
994  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
995  */
996 static int fix_inline_data_extents_file(e2fsck_t ctx,
997                                         ext2_ino_t ino,
998                                         struct ext2_inode *inode,
999                                         int inode_size,
1000                                         struct problem_context *pctx)
1001 {
1002         size_t max_inline_ea_size;
1003         ext2_filsys fs = ctx->fs;
1004         int dirty = 0;
1005
1006         /* Both feature flags not set?  Just run the regular checks */
1007         if (!ext2fs_has_feature_extents(fs->super) &&
1008             !ext2fs_has_feature_inline_data(fs->super))
1009                 return 0;
1010
1011         /* Clear both flags if it's a special file */
1012         if (LINUX_S_ISCHR(inode->i_mode) ||
1013             LINUX_S_ISBLK(inode->i_mode) ||
1014             LINUX_S_ISFIFO(inode->i_mode) ||
1015             LINUX_S_ISSOCK(inode->i_mode)) {
1016                 check_extents_inlinedata(ctx, pctx);
1017                 return 0;
1018         }
1019
1020         /* If it looks like an extent tree, try to clear inlinedata */
1021         if (ext2fs_extent_header_verify(inode->i_block,
1022                                  sizeof(inode->i_block)) == 0 &&
1023             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1024                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1025                 dirty = 1;
1026                 goto out;
1027         }
1028
1029         /* If it looks short enough to be inline data, try to clear extents */
1030         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1031                 max_inline_ea_size = inode_size -
1032                                      (EXT2_GOOD_OLD_INODE_SIZE +
1033                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1034         else
1035                 max_inline_ea_size = 0;
1036         if (EXT2_I_SIZE(inode) <
1037             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1038             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1039                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1040                 dirty = 1;
1041                 goto out;
1042         }
1043
1044         /*
1045          * Too big for inline data, but no evidence of extent tree -
1046          * maybe it's a block map file?  If the mappings all look valid?
1047          */
1048         if (could_be_block_map(fs, inode) &&
1049             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1050 #ifdef WORDS_BIGENDIAN
1051                 int i;
1052
1053                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1054                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1055 #endif
1056
1057                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1058                 dirty = 1;
1059                 goto out;
1060         }
1061
1062         /* Oh well, just clear the busted inode. */
1063         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1064                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1065                 return -1;
1066         }
1067
1068 out:
1069         if (dirty)
1070                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1071
1072         return 0;
1073 }
1074
1075 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1076 {
1077         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1078         dgrp_t start = *group, grp;
1079         blk64_t blocks_to_read = 0;
1080         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1081
1082         if (ctx->readahead_kb == 0)
1083                 goto out;
1084
1085         /* Keep iterating groups until we have enough to readahead */
1086         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1087         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1088                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1089                         continue;
1090                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1091                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1092                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1093                                         inodes_per_block;
1094                 if (blocks_to_read * ctx->fs->blocksize >
1095                     ctx->readahead_kb * 1024)
1096                         break;
1097         }
1098
1099         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1100                                grp - start + 1);
1101         if (err == EAGAIN) {
1102                 ctx->readahead_kb /= 2;
1103                 err = 0;
1104         }
1105
1106 out:
1107         if (err) {
1108                 /* Error; disable itable readahead */
1109                 *group = ctx->fs->group_desc_count;
1110                 *next_ino = ctx->fs->super->s_inodes_count;
1111         } else {
1112                 /*
1113                  * Don't do more readahead until we've reached the first inode
1114                  * of the last inode scan buffer block for the last group.
1115                  */
1116                 *group = grp + 1;
1117                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1118                                      ctx->inode_buffer_blocks :
1119                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1120                                     ctx->fs->blocksize /
1121                                     EXT2_INODE_SIZE(ctx->fs->super);
1122                 inodes_in_group--;
1123                 *next_ino = inodes_in_group -
1124                             (inodes_in_group % inodes_per_buffer) + 1 +
1125                             (grp * ctx->fs->super->s_inodes_per_group);
1126         }
1127 }
1128
1129 /*
1130  * Check if the passed ino is one of the used superblock quota inodes.
1131  *
1132  * Before the quota inodes were journaled, older superblock quota inodes
1133  * were just regular files in the filesystem and not reserved inodes.  This
1134  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1135  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1136  */
1137 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1138 {
1139         enum quota_type qtype;
1140
1141         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1142                 if (*quota_sb_inump(sb, qtype) == ino)
1143                         return 1;
1144
1145         return 0;
1146 }
1147
1148 /*
1149  * Check if the passed ino is one of the reserved quota inodes.
1150  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1151  * inodes.  These inodes may or may not be in use by the quota feature.
1152  */
1153 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1154 {
1155         enum quota_type qtype;
1156
1157         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1158                 if (quota_type2inum(qtype, fs->super) == ino)
1159                         return 1;
1160
1161         return 0;
1162 }
1163
1164 static int e2fsck_should_abort(e2fsck_t ctx)
1165 {
1166         e2fsck_t global_ctx;
1167
1168         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1169                 return 1;
1170
1171         if (ctx->global_ctx) {
1172                 global_ctx = ctx->global_ctx;
1173                 if (global_ctx->flags & E2F_FLAG_SIGNAL_MASK)
1174                         return 1;
1175         }
1176         return 0;
1177 }
1178
1179 /*
1180  * We need call mark_table_blocks() before multiple
1181  * thread start, since all known system blocks should be
1182  * marked and checked later.
1183  */
1184 static errcode_t e2fsck_pass1_prepare(e2fsck_t ctx)
1185 {
1186         struct problem_context pctx;
1187         ext2_filsys fs = ctx->fs;
1188
1189         clear_problem_context(&pctx);
1190         if (!(ctx->options & E2F_OPT_PREEN))
1191                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1192
1193         pctx.errcode = e2fsck_allocate_subcluster_bitmap(ctx->fs,
1194                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1195                         "block_found_map", &ctx->block_found_map);
1196         if (pctx.errcode) {
1197                 pctx.num = 1;
1198                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1199                 ctx->flags |= E2F_FLAG_ABORT;
1200                 return pctx.errcode;
1201         }
1202         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
1203                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1204                         "block_metadata_map", &ctx->block_metadata_map);
1205         if (pctx.errcode) {
1206                 pctx.num = 1;
1207                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1208                 ctx->flags |= E2F_FLAG_ABORT;
1209                 return pctx.errcode;
1210         }
1211
1212         mark_table_blocks(ctx);
1213         pctx.errcode = ext2fs_convert_subcluster_bitmap(ctx->fs,
1214                                                 &ctx->block_found_map);
1215         if (pctx.errcode) {
1216                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1217                 ctx->flags |= E2F_FLAG_ABORT;
1218                 return pctx.errcode;
1219         }
1220
1221         if (ext2fs_has_feature_mmp(fs->super) &&
1222             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1223             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1224                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1225                                           fs->super->s_mmp_block);
1226
1227         return 0;
1228 }
1229
1230 static void e2fsck_pass1_post(e2fsck_t ctx)
1231 {
1232         struct problem_context pctx;
1233         ext2_filsys fs = ctx->fs;
1234
1235         char *block_buf =
1236                 (char *)e2fsck_allocate_memory(ctx, ctx->fs->blocksize * 3,
1237                                               "block interate buffer");
1238         reserve_block_for_root_repair(ctx);
1239         reserve_block_for_lnf_repair(ctx);
1240
1241         /*
1242          * If any extended attribute blocks' reference counts need to
1243          * be adjusted, either up (ctx->refcount_extra), or down
1244          * (ctx->refcount), then fix them.
1245          */
1246         if (ctx->refcount) {
1247                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
1248                 ea_refcount_free(ctx->refcount);
1249                 ctx->refcount = 0;
1250         }
1251         if (ctx->refcount_extra) {
1252                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
1253                                         block_buf, +1);
1254                 ea_refcount_free(ctx->refcount_extra);
1255                 ctx->refcount_extra = 0;
1256         }
1257
1258         if (ctx->invalid_bitmaps)
1259                 handle_fs_bad_blocks(ctx);
1260
1261         /* We don't need the block_ea_map any more */
1262         if (ctx->block_ea_map) {
1263                 ext2fs_free_block_bitmap(ctx->block_ea_map);
1264                 ctx->block_ea_map = 0;
1265         }
1266
1267         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
1268                 struct ext2_inode *inode;
1269                 int inode_size = EXT2_INODE_SIZE(fs->super);
1270                 inode = e2fsck_allocate_memory(ctx, inode_size,
1271                                                "scratch inode");
1272
1273                 clear_problem_context(&pctx);
1274                 pctx.errcode = ext2fs_create_resize_inode(fs);
1275                 if (pctx.errcode) {
1276                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
1277                                          &pctx)) {
1278                                 ctx->flags |= E2F_FLAG_ABORT;
1279                                 ext2fs_free_mem(&inode);
1280                                 ext2fs_free_mem(&block_buf);
1281                                 return;
1282                         }
1283                         pctx.errcode = 0;
1284                 }
1285                 if (!pctx.errcode) {
1286                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
1287                                           "recreate inode");
1288                         inode->i_mtime = ctx->now;
1289                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
1290                                            "recreate inode");
1291                 }
1292                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
1293                 ext2fs_free_mem(&inode);
1294         }
1295
1296         if (ctx->flags & E2F_FLAG_RESTART) {
1297                 ext2fs_free_mem(&block_buf);
1298                 return;
1299         }
1300
1301         if (ctx->block_dup_map) {
1302                 if (ctx->options & E2F_OPT_PREEN) {
1303                         clear_problem_context(&pctx);
1304                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
1305                 }
1306                 e2fsck_pass1_dupblocks(ctx, block_buf);
1307                 ext2fs_free_mem(&block_buf);
1308         }
1309 }
1310
1311
1312 void e2fsck_pass1_run(e2fsck_t ctx)
1313 {
1314         int     i;
1315         ext2_filsys fs = ctx->fs;
1316         ext2_ino_t      ino = 0;
1317         struct ext2_inode *inode = NULL;
1318         ext2_inode_scan scan = NULL;
1319         char            *block_buf = NULL;
1320 #ifdef RESOURCE_TRACK
1321         struct resource_track   rtrack;
1322 #endif
1323         unsigned char   frag, fsize;
1324         struct          problem_context pctx;
1325         struct          scan_callback_struct scan_struct;
1326         struct ext2_super_block *sb = ctx->fs->super;
1327         const char      *old_op;
1328         const char      *eop_next_inode = _("getting next inode from scan");
1329         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1330         int             low_dtime_check = 1;
1331         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1332         unsigned int    bufsize;
1333         int             failed_csum = 0;
1334         ext2_ino_t      ino_threshold = 0;
1335         dgrp_t          ra_group = 0;
1336         struct ea_quota ea_ibody_quota;
1337         struct process_inode_block *inodes_to_process;
1338         int             process_inode_count;
1339
1340         init_resource_track(&rtrack, ctx->fs->io);
1341         clear_problem_context(&pctx);
1342
1343         /* If we can do readahead, figure out how many groups to pull in. */
1344         if (!e2fsck_can_readahead(ctx->fs))
1345                 ctx->readahead_kb = 0;
1346         else if (ctx->readahead_kb == ~0ULL)
1347                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1348         pass1_readahead(ctx, &ra_group, &ino_threshold);
1349
1350         if (ext2fs_has_feature_dir_index(fs->super) &&
1351             !(ctx->options & E2F_OPT_NO)) {
1352                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1353                         ctx->dirs_to_hash = 0;
1354         }
1355
1356 #ifdef MTRACE
1357         mtrace_print("Pass 1");
1358 #endif
1359
1360         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1361         extent_fs = ext2fs_has_feature_extents(sb);
1362         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1363         casefold_fs = ext2fs_has_feature_casefold(sb);
1364
1365         /*
1366          * Allocate bitmaps structures
1367          */
1368         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1369                                                     EXT2FS_BMAP64_RBTREE,
1370                                                     "inode_used_map",
1371                                                     &ctx->inode_used_map);
1372         if (pctx.errcode) {
1373                 pctx.num = 1;
1374                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1375                 ctx->flags |= E2F_FLAG_ABORT;
1376                 return;
1377         }
1378         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1379                         _("directory inode map"),
1380                         ctx->global_ctx ? EXT2FS_BMAP64_RBTREE :
1381                         EXT2FS_BMAP64_AUTODIR,
1382                         "inode_dir_map", &ctx->inode_dir_map);
1383         if (pctx.errcode) {
1384                 pctx.num = 2;
1385                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1386                 ctx->flags |= E2F_FLAG_ABORT;
1387                 return;
1388         }
1389         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1390                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1391                         "inode_reg_map", &ctx->inode_reg_map);
1392         if (pctx.errcode) {
1393                 pctx.num = 6;
1394                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1395                 ctx->flags |= E2F_FLAG_ABORT;
1396                 return;
1397         }
1398         if (casefold_fs) {
1399                 pctx.errcode =
1400                         e2fsck_allocate_inode_bitmap(fs,
1401                                                      _("inode casefold map"),
1402                                                      EXT2FS_BMAP64_RBTREE,
1403                                                      "inode_casefold_map",
1404                                                      &ctx->inode_casefold_map);
1405                 if (pctx.errcode) {
1406                         pctx.num = 1;
1407                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1408                         ctx->flags |= E2F_FLAG_ABORT;
1409                         return;
1410                 }
1411         }
1412         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1413                                            &ctx->inode_link_info);
1414         if (pctx.errcode) {
1415                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1416                 ctx->flags |= E2F_FLAG_ABORT;
1417                 return;
1418         }
1419         bufsize = inode_size;
1420         if (bufsize < sizeof(struct ext2_inode_large))
1421                 bufsize = sizeof(struct ext2_inode_large);
1422         inode = (struct ext2_inode *)
1423                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1424
1425         inodes_to_process = (struct process_inode_block *)
1426                 e2fsck_allocate_memory(ctx,
1427                                        (ctx->process_inode_size *
1428                                         sizeof(struct process_inode_block)),
1429                                        "array of inodes to process");
1430         process_inode_count = 0;
1431
1432         pctx.errcode = ext2fs_init_dblist(fs, 0);
1433         if (pctx.errcode) {
1434                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1435                 ctx->flags |= E2F_FLAG_ABORT;
1436                 goto endit;
1437         }
1438
1439         /*
1440          * If the last orphan field is set, clear it, since the pass1
1441          * processing will automatically find and clear the orphans.
1442          * In the future, we may want to try using the last_orphan
1443          * linked list ourselves, but for now, we clear it so that the
1444          * ext3 mount code won't get confused.
1445          */
1446         if (!(ctx->options & E2F_OPT_READONLY)) {
1447                 if (fs->super->s_last_orphan) {
1448                         fs->super->s_last_orphan = 0;
1449                         ext2fs_mark_super_dirty(fs);
1450                 }
1451         }
1452
1453         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1454                                                     "block interate buffer");
1455         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1456                 e2fsck_use_inode_shortcuts(ctx, 1);
1457         e2fsck_intercept_block_allocations(ctx);
1458         old_op = ehandler_operation(_("opening inode scan"));
1459         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1460                                               &scan);
1461         ehandler_operation(old_op);
1462         if (pctx.errcode) {
1463                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1464                 ctx->flags |= E2F_FLAG_ABORT;
1465                 goto endit;
1466         }
1467         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1468                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1469         ctx->stashed_inode = inode;
1470         scan_struct.ctx = ctx;
1471         scan_struct.block_buf = block_buf;
1472         scan_struct.inodes_to_process = inodes_to_process;
1473         scan_struct.process_inode_count = &process_inode_count;
1474         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1475         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1476                                               ctx->fs->group_desc_count)))
1477                 goto endit;
1478         if ((fs->super->s_wtime &&
1479              fs->super->s_wtime < fs->super->s_inodes_count) ||
1480             (fs->super->s_mtime &&
1481              fs->super->s_mtime < fs->super->s_inodes_count) ||
1482             (fs->super->s_mkfs_time &&
1483              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1484                 low_dtime_check = 0;
1485
1486         /* Set up ctx->lost_and_found if possible */
1487         (void) e2fsck_get_lost_and_found(ctx, 0);
1488
1489 #ifdef HAVE_PTHREAD
1490         if (ctx->global_ctx) {
1491                 if (ctx->options & E2F_OPT_DEBUG &&
1492                     ctx->options & E2F_OPT_MULTITHREAD)
1493                         fprintf(stderr, "thread %d jumping to group %d\n",
1494                                         ctx->thread_info.et_thread_index,
1495                                         ctx->thread_info.et_group_start);
1496                 pctx.errcode = ext2fs_inode_scan_goto_blockgroup(scan,
1497                                         ctx->thread_info.et_group_start);
1498                 if (pctx.errcode) {
1499                         fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1500                         ctx->flags |= E2F_FLAG_ABORT;
1501                         goto endit;
1502                 }
1503         }
1504 #endif
1505
1506         while (1) {
1507                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1508                         if (e2fsck_mmp_update(fs))
1509                                 fatal_error(ctx, 0);
1510                 }
1511                 old_op = ehandler_operation(eop_next_inode);
1512                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1513                                                           inode, inode_size);
1514                 if (ino > ino_threshold)
1515                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1516                 ehandler_operation(old_op);
1517                 if (e2fsck_should_abort(ctx))
1518                         goto endit;
1519                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1520                         /*
1521                          * If badblocks says badblocks is bad, offer to clear
1522                          * the list, update the in-core bb list, and restart
1523                          * the inode scan.
1524                          */
1525                         if (ino == EXT2_BAD_INO &&
1526                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1527                                         &pctx)) {
1528                                 errcode_t err;
1529
1530                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1531                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1532                                 ctx->fs->badblocks = NULL;
1533                                 err = ext2fs_read_bb_inode(ctx->fs,
1534                                                         &ctx->fs->badblocks);
1535                                 if (err) {
1536                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1537                                                     &pctx);
1538                                         ctx->flags |= E2F_FLAG_ABORT;
1539                                 } else
1540                                         ctx->flags |= E2F_FLAG_RESTART;
1541                                 goto endit;
1542                         }
1543                         if (!ctx->inode_bb_map)
1544                                 alloc_bb_map(ctx);
1545                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1546                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1547                         continue;
1548                 }
1549                 if (pctx.errcode == EXT2_ET_SCAN_FINISHED)
1550                         break;
1551                 if (pctx.errcode &&
1552                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1553                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1554                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1555                         ctx->flags |= E2F_FLAG_ABORT;
1556                         goto endit;
1557                 }
1558                 if (!ino)
1559                         break;
1560 #ifdef HAVE_PTHREAD
1561                 if (ctx->global_ctx)
1562                         ctx->thread_info.et_inode_number++;
1563 #endif
1564                 pctx.ino = ino;
1565                 pctx.inode = inode;
1566                 ctx->stashed_ino = ino;
1567
1568                 /* Clear trashed inode? */
1569                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1570                     inode->i_links_count > 0 &&
1571                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1572                         pctx.errcode = 0;
1573                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1574                 }
1575                 failed_csum = pctx.errcode != 0;
1576
1577                 /*
1578                  * Check for inodes who might have been part of the
1579                  * orphaned list linked list.  They should have gotten
1580                  * dealt with by now, unless the list had somehow been
1581                  * corrupted.
1582                  *
1583                  * FIXME: In the future, inodes which are still in use
1584                  * (and which are therefore) pending truncation should
1585                  * be handled specially.  Right now we just clear the
1586                  * dtime field, and the normal e2fsck handling of
1587                  * inodes where i_size and the inode blocks are
1588                  * inconsistent is to fix i_size, instead of releasing
1589                  * the extra blocks.  This won't catch the inodes that
1590                  * was at the end of the orphan list, but it's better
1591                  * than nothing.  The right answer is that there
1592                  * shouldn't be any bugs in the orphan list handling.  :-)
1593                  */
1594                 if (inode->i_dtime && low_dtime_check &&
1595                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1596                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1597                                 inode->i_dtime = inode->i_links_count ?
1598                                         0 : ctx->now;
1599                                 e2fsck_write_inode(ctx, ino, inode,
1600                                                    "pass1");
1601                                 failed_csum = 0;
1602                         }
1603                 }
1604
1605                 if (inode->i_links_count) {
1606                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1607                                            ino, inode->i_links_count);
1608                         if (pctx.errcode) {
1609                                 pctx.num = inode->i_links_count;
1610                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1611                                 ctx->flags |= E2F_FLAG_ABORT;
1612                                 goto endit;
1613                         }
1614                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1615                            !quota_inum_is_reserved(fs, ino)) {
1616                         if (!inode->i_dtime && inode->i_mode) {
1617                                 if (fix_problem(ctx,
1618                                             PR_1_ZERO_DTIME, &pctx)) {
1619                                         inode->i_dtime = ctx->now;
1620                                         e2fsck_write_inode(ctx, ino, inode,
1621                                                            "pass1");
1622                                         failed_csum = 0;
1623                                 }
1624                         }
1625                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1626                         continue;
1627                 }
1628
1629                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1630                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1631                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1632                      (!casefold_fs &&
1633                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1634                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1635                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1636                 }
1637
1638                 /* Conflicting inlinedata/extents inode flags? */
1639                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1640                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1641                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1642                                                                inode_size,
1643                                                                &pctx);
1644                         if (res < 0) {
1645                                 /* skip FINISH_INODE_LOOP */
1646                                 continue;
1647                         }
1648                 }
1649
1650                 /* Test for incorrect inline_data flags settings. */
1651                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1652                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1653                         size_t size = 0;
1654
1655                         pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1656                         if (!pctx.errcode &&
1657                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1658                                 e2fsck_pass1_fix_lock(ctx);
1659                                 ext2fs_set_feature_inline_data(sb);
1660                                 ext2fs_mark_super_dirty(fs);
1661                                 e2fsck_pass1_fix_unlock(ctx);
1662                                 inlinedata_fs = 1;
1663                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1664                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1665                                 /* skip FINISH_INODE_LOOP */
1666                                 continue;
1667                         }
1668                 }
1669
1670                 /* Test for inline data flag but no attr */
1671                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1672                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1673                         size_t size = 0;
1674                         errcode_t err;
1675                         int flags;
1676
1677                         flags = fs->flags;
1678                         if (failed_csum)
1679                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1680                         err = get_inline_data_ea_size(fs, ino, &size);
1681                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1682                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1683
1684                         switch (err) {
1685                         case 0:
1686                                 /* Everything is awesome... */
1687                                 break;
1688                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1689                         case EXT2_ET_BAD_EA_HASH:
1690                         case EXT2_ET_BAD_EA_HEADER:
1691                         case EXT2_ET_EA_BAD_NAME_LEN:
1692                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1693                         case EXT2_ET_EA_KEY_NOT_FOUND:
1694                         case EXT2_ET_EA_NO_SPACE:
1695                         case EXT2_ET_MISSING_EA_FEATURE:
1696                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1697                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1698                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1699                         case EXT2_ET_NO_INLINE_DATA:
1700                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1701                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1702                         case EXT2_ET_EA_INODE_CORRUPTED:
1703                                 /* broken EA or no system.data EA; truncate */
1704                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1705                                                 &pctx)) {
1706                                         err = ext2fs_inode_size_set(fs, inode, 0);
1707                                         if (err) {
1708                                                 pctx.errcode = err;
1709                                                 ctx->flags |= E2F_FLAG_ABORT;
1710                                                 goto endit;
1711                                         }
1712                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1713                                         memset(&inode->i_block, 0,
1714                                                sizeof(inode->i_block));
1715                                         e2fsck_write_inode(ctx, ino, inode,
1716                                                            "pass1");
1717                                         failed_csum = 0;
1718                                 }
1719                                 break;
1720                         default:
1721                                 /* Some other kind of non-xattr error? */
1722                                 pctx.errcode = err;
1723                                 ctx->flags |= E2F_FLAG_ABORT;
1724                                 goto endit;
1725                         }
1726                 }
1727
1728                 /*
1729                  * Test for incorrect extent flag settings.
1730                  *
1731                  * On big-endian machines we must be careful:
1732                  * When the inode is read, the i_block array is not swapped
1733                  * if the extent flag is set.  Therefore if we are testing
1734                  * for or fixing a wrongly-set flag, we must potentially
1735                  * (un)swap before testing, or after fixing.
1736                  */
1737
1738                 /*
1739                  * In this case the extents flag was set when read, so
1740                  * extent_header_verify is ok.  If the inode is cleared,
1741                  * no need to swap... so no extra swapping here.
1742                  */
1743                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1744                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1745                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1746                         if ((ext2fs_extent_header_verify(inode->i_block,
1747                                                  sizeof(inode->i_block)) == 0) &&
1748                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1749                                 e2fsck_pass1_fix_lock(ctx);
1750                                 ext2fs_set_feature_extents(sb);
1751                                 ext2fs_mark_super_dirty(fs);
1752                                 extent_fs = 1;
1753                                 e2fsck_pass1_fix_unlock(ctx);
1754                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1755                         clear_inode:
1756                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1757                                 if (ino == EXT2_BAD_INO)
1758                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1759                                                                  ino);
1760                                 /* skip FINISH_INODE_LOOP */
1761                                 continue;
1762                         }
1763                 }
1764
1765                 /*
1766                  * For big-endian machines:
1767                  * If the inode didn't have the extents flag set when it
1768                  * was read, then the i_blocks array was swapped.  To test
1769                  * as an extents header, we must swap it back first.
1770                  * IF we then set the extents flag, the entire i_block
1771                  * array must be un/re-swapped to make it proper extents data.
1772                  */
1773                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1774                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1775                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1776                     (LINUX_S_ISREG(inode->i_mode) ||
1777                      LINUX_S_ISDIR(inode->i_mode))) {
1778                         void *ehp;
1779 #ifdef WORDS_BIGENDIAN
1780                         __u32 tmp_block[EXT2_N_BLOCKS];
1781
1782                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1783                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1784                         ehp = tmp_block;
1785 #else
1786                         ehp = inode->i_block;
1787 #endif
1788                         if ((ext2fs_extent_header_verify(ehp,
1789                                          sizeof(inode->i_block)) == 0) &&
1790                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1791                                 inode->i_flags |= EXT4_EXTENTS_FL;
1792 #ifdef WORDS_BIGENDIAN
1793                                 memcpy(inode->i_block, tmp_block,
1794                                        sizeof(inode->i_block));
1795 #endif
1796                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1797                                 failed_csum = 0;
1798                         }
1799                 }
1800
1801                 if (ino == EXT2_BAD_INO) {
1802                         struct process_block_struct pb;
1803
1804                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1805                              inode->i_gid || inode->i_links_count ||
1806                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1807                              inode->i_file_acl) &&
1808                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1809                                 memset(inode, 0, sizeof(struct ext2_inode));
1810                                 e2fsck_write_inode(ctx, ino, inode,
1811                                                    "clear bad inode");
1812                                 failed_csum = 0;
1813                         }
1814
1815                         e2fsck_pass1_block_map_lock(ctx);
1816                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1817                                                           &pb.fs_meta_blocks);
1818                         e2fsck_pass1_block_map_unlock(ctx);
1819                         if (pctx.errcode) {
1820                                 pctx.num = 4;
1821                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1822                                 ctx->flags |= E2F_FLAG_ABORT;
1823                                 goto endit;
1824                         }
1825                         pb.ino = EXT2_BAD_INO;
1826                         pb.num_blocks = pb.last_block = 0;
1827                         pb.last_db_block = -1;
1828                         pb.num_illegal_blocks = 0;
1829                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1830                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1831                         pb.inode = inode;
1832                         pb.pctx = &pctx;
1833                         pb.ctx = ctx;
1834                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1835                                      block_buf, process_bad_block, &pb);
1836                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1837                         if (pctx.errcode) {
1838                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1839                                 ctx->flags |= E2F_FLAG_ABORT;
1840                                 goto endit;
1841                         }
1842                         if (pb.bbcheck)
1843                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1844                                 ctx->flags |= E2F_FLAG_ABORT;
1845                                 goto endit;
1846                         }
1847                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1848                         clear_problem_context(&pctx);
1849                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1850                         continue;
1851                 } else if (ino == EXT2_ROOT_INO) {
1852                         /*
1853                          * Make sure the root inode is a directory; if
1854                          * not, offer to clear it.  It will be
1855                          * regenerated in pass #3.
1856                          */
1857                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1858                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1859                                         goto clear_inode;
1860                         }
1861                         /*
1862                          * If dtime is set, offer to clear it.  mke2fs
1863                          * version 0.2b created filesystems with the
1864                          * dtime field set for the root and lost+found
1865                          * directories.  We won't worry about
1866                          * /lost+found, since that can be regenerated
1867                          * easily.  But we will fix the root directory
1868                          * as a special case.
1869                          */
1870                         if (inode->i_dtime && inode->i_links_count) {
1871                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1872                                         inode->i_dtime = 0;
1873                                         e2fsck_write_inode(ctx, ino, inode,
1874                                                            "pass1");
1875                                         failed_csum = 0;
1876                                 }
1877                         }
1878                 } else if (ino == EXT2_JOURNAL_INO) {
1879                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1880                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1881                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1882                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1883                                                 &pctx)) {
1884                                         inode->i_mode = LINUX_S_IFREG;
1885                                         e2fsck_write_inode(ctx, ino, inode,
1886                                                            "pass1");
1887                                         failed_csum = 0;
1888                                 }
1889                                 check_blocks(ctx, &pctx, block_buf, NULL);
1890                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1891                                 continue;
1892                         }
1893                         if ((inode->i_links_count ||
1894                              inode->i_blocks || inode->i_block[0]) &&
1895                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1896                                         &pctx)) {
1897                                 memset(inode, 0, inode_size);
1898                                 ext2fs_icount_store(ctx->inode_link_info,
1899                                                     ino, 0);
1900                                 e2fsck_write_inode_full(ctx, ino, inode,
1901                                                         inode_size, "pass1");
1902                                 failed_csum = 0;
1903                         }
1904                 } else if (quota_inum_is_reserved(fs, ino)) {
1905                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1906                         if (ext2fs_has_feature_quota(fs->super) &&
1907                             quota_inum_is_super(fs->super, ino)) {
1908                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1909                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1910                                                         &pctx)) {
1911                                         inode->i_mode = LINUX_S_IFREG;
1912                                         e2fsck_write_inode(ctx, ino, inode,
1913                                                         "pass1");
1914                                         failed_csum = 0;
1915                                 }
1916                                 check_blocks(ctx, &pctx, block_buf, NULL);
1917                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1918                                 continue;
1919                         }
1920                         if ((inode->i_links_count ||
1921                              inode->i_blocks || inode->i_block[0]) &&
1922                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1923                                         &pctx)) {
1924                                 memset(inode, 0, inode_size);
1925                                 ext2fs_icount_store(ctx->inode_link_info,
1926                                                     ino, 0);
1927                                 e2fsck_write_inode_full(ctx, ino, inode,
1928                                                         inode_size, "pass1");
1929                                 failed_csum = 0;
1930                         }
1931                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1932                         problem_t problem = 0;
1933
1934                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1935                         if (ino == EXT2_BOOT_LOADER_INO) {
1936                                 if (LINUX_S_ISDIR(inode->i_mode))
1937                                         problem = PR_1_RESERVED_BAD_MODE;
1938                         } else if (ino == EXT2_RESIZE_INO) {
1939                                 if (inode->i_mode &&
1940                                     !LINUX_S_ISREG(inode->i_mode))
1941                                         problem = PR_1_RESERVED_BAD_MODE;
1942                         } else {
1943                                 if (inode->i_mode != 0)
1944                                         problem = PR_1_RESERVED_BAD_MODE;
1945                         }
1946                         if (problem) {
1947                                 if (fix_problem(ctx, problem, &pctx)) {
1948                                         inode->i_mode = 0;
1949                                         e2fsck_write_inode(ctx, ino, inode,
1950                                                            "pass1");
1951                                         failed_csum = 0;
1952                                 }
1953                         }
1954                         check_blocks(ctx, &pctx, block_buf, NULL);
1955                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1956                         continue;
1957                 }
1958
1959                 if (!inode->i_links_count) {
1960                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1961                         continue;
1962                 }
1963                 /*
1964                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1965                  * deleted files.  Oops.
1966                  *
1967                  * Since all new ext2 implementations get this right,
1968                  * we now assume that the case of non-zero
1969                  * i_links_count and non-zero dtime means that we
1970                  * should keep the file, not delete it.
1971                  *
1972                  */
1973                 if (inode->i_dtime) {
1974                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1975                                 inode->i_dtime = 0;
1976                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1977                                 failed_csum = 0;
1978                         }
1979                 }
1980
1981                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1982                 switch (fs->super->s_creator_os) {
1983                     case EXT2_OS_HURD:
1984                         frag = inode->osd2.hurd2.h_i_frag;
1985                         fsize = inode->osd2.hurd2.h_i_fsize;
1986                         break;
1987                     default:
1988                         frag = fsize = 0;
1989                 }
1990
1991                 if (inode->i_faddr || frag || fsize ||
1992                     (!ext2fs_has_feature_largedir(fs->super) &&
1993                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1994                         mark_inode_bad(ctx, ino);
1995                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1996                     !ext2fs_has_feature_64bit(fs->super) &&
1997                     inode->osd2.linux2.l_i_file_acl_high != 0)
1998                         mark_inode_bad(ctx, ino);
1999                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
2000                     !ext2fs_has_feature_huge_file(fs->super) &&
2001                     (inode->osd2.linux2.l_i_blocks_hi != 0))
2002                         mark_inode_bad(ctx, ino);
2003                 if (inode->i_flags & EXT2_IMAGIC_FL) {
2004                         if (imagic_fs) {
2005                                 if (!ctx->inode_imagic_map)
2006                                         alloc_imagic_map(ctx);
2007                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
2008                                                          ino);
2009                         } else {
2010                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
2011                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
2012                                         e2fsck_write_inode(ctx, ino,
2013                                                            inode, "pass1");
2014                                         failed_csum = 0;
2015                                 }
2016                         }
2017                 }
2018
2019                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
2020                 check_is_really_dir(ctx, &pctx, block_buf);
2021
2022                 /*
2023                  * ext2fs_inode_has_valid_blocks2 does not actually look
2024                  * at i_block[] values, so not endian-sensitive here.
2025                  */
2026                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
2027                     LINUX_S_ISLNK(inode->i_mode) &&
2028                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
2029                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
2030                         inode->i_flags &= ~EXT4_EXTENTS_FL;
2031                         e2fsck_write_inode(ctx, ino, inode, "pass1");
2032                         failed_csum = 0;
2033                 }
2034
2035                 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
2036                     add_encrypted_file(ctx, &pctx) < 0)
2037                         goto clear_inode;
2038
2039                 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
2040                         ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
2041
2042                 if (LINUX_S_ISDIR(inode->i_mode)) {
2043                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
2044                         e2fsck_add_dir_info(ctx, ino, 0);
2045                         ctx->fs_directory_count++;
2046                         if (inode->i_flags & EXT4_CASEFOLD_FL)
2047                                 add_casefolded_dir(ctx, ino);
2048                 } else if (LINUX_S_ISREG (inode->i_mode)) {
2049                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
2050                         ctx->fs_regular_count++;
2051                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
2052                            e2fsck_pass1_check_device_inode(fs, inode)) {
2053                         check_extents_inlinedata(ctx, &pctx);
2054                         check_immutable(ctx, &pctx);
2055                         check_size(ctx, &pctx);
2056                         ctx->fs_chardev_count++;
2057                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
2058                            e2fsck_pass1_check_device_inode(fs, inode)) {
2059                         check_extents_inlinedata(ctx, &pctx);
2060                         check_immutable(ctx, &pctx);
2061                         check_size(ctx, &pctx);
2062                         ctx->fs_blockdev_count++;
2063                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
2064                            e2fsck_pass1_check_symlink(fs, ino, inode,
2065                                                       block_buf)) {
2066                         check_immutable(ctx, &pctx);
2067                         ctx->fs_symlinks_count++;
2068                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
2069                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2070                                 continue;
2071                         } else if (ext2fs_is_fast_symlink(inode)) {
2072                                 ctx->fs_fast_symlinks_count++;
2073                                 check_blocks(ctx, &pctx, block_buf,
2074                                              &ea_ibody_quota);
2075                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2076                                 continue;
2077                         }
2078                 }
2079                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
2080                          e2fsck_pass1_check_device_inode(fs, inode)) {
2081                         check_extents_inlinedata(ctx, &pctx);
2082                         check_immutable(ctx, &pctx);
2083                         check_size(ctx, &pctx);
2084                         ctx->fs_fifo_count++;
2085                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
2086                            e2fsck_pass1_check_device_inode(fs, inode)) {
2087                         check_extents_inlinedata(ctx, &pctx);
2088                         check_immutable(ctx, &pctx);
2089                         check_size(ctx, &pctx);
2090                         ctx->fs_sockets_count++;
2091                 } else
2092                         mark_inode_bad(ctx, ino);
2093                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2094                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
2095                         if (inode->i_block[EXT2_IND_BLOCK])
2096                                 ctx->fs_ind_count++;
2097                         if (inode->i_block[EXT2_DIND_BLOCK])
2098                                 ctx->fs_dind_count++;
2099                         if (inode->i_block[EXT2_TIND_BLOCK])
2100                                 ctx->fs_tind_count++;
2101                 }
2102                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2103                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
2104                     (inode->i_block[EXT2_IND_BLOCK] ||
2105                      inode->i_block[EXT2_DIND_BLOCK] ||
2106                      inode->i_block[EXT2_TIND_BLOCK] ||
2107                      ext2fs_file_acl_block(fs, inode))) {
2108                         struct process_inode_block *itp;
2109
2110                         itp = &inodes_to_process[process_inode_count];
2111                         itp->ino = ino;
2112                         itp->ea_ibody_quota = ea_ibody_quota;
2113                         if (inode_size < sizeof(struct ext2_inode_large))
2114                                 memcpy(&itp->inode, inode, inode_size);
2115                         else
2116                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
2117                         process_inode_count++;
2118                 } else
2119                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
2120
2121                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2122
2123                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2124                         goto endit;
2125
2126                 if (process_inode_count >= ctx->process_inode_size) {
2127                         process_inodes(ctx, block_buf, inodes_to_process,
2128                                        &process_inode_count);
2129
2130                         if (e2fsck_should_abort(ctx))
2131                                 goto endit;
2132                 }
2133         }
2134         process_inodes(ctx, block_buf, inodes_to_process,
2135                        &process_inode_count);
2136         ext2fs_close_inode_scan(scan);
2137         scan = NULL;
2138
2139         if (ctx->ea_block_quota_blocks) {
2140                 ea_refcount_free(ctx->ea_block_quota_blocks);
2141                 ctx->ea_block_quota_blocks = 0;
2142         }
2143
2144         if (ctx->ea_block_quota_inodes) {
2145                 ea_refcount_free(ctx->ea_block_quota_inodes);
2146                 ctx->ea_block_quota_inodes = 0;
2147         }
2148
2149         /* We don't need the encryption policy => ID map any more */
2150         destroy_encryption_policy_map(ctx);
2151
2152         if (ctx->flags & E2F_FLAG_RESTART) {
2153                 /*
2154                  * Only the master copy of the superblock and block
2155                  * group descriptors are going to be written during a
2156                  * restart, so set the superblock to be used to be the
2157                  * master superblock.
2158                  */
2159                 ctx->use_superblock = 0;
2160                 goto endit;
2161         }
2162
2163         if (ctx->large_dirs && !ext2fs_has_feature_largedir(ctx->fs->super)) {
2164                 ext2_filsys fs = ctx->fs;
2165
2166                 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2167                         ext2fs_set_feature_largedir(fs->super);
2168                         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2169                         ext2fs_mark_super_dirty(fs);
2170                 }
2171                 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2172                     fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2173                         ext2fs_update_dynamic_rev(fs);
2174                         ext2fs_mark_super_dirty(fs);
2175                 }
2176         }
2177
2178         ctx->flags |= E2F_FLAG_ALLOC_OK;
2179 endit:
2180         e2fsck_use_inode_shortcuts(ctx, 0);
2181         ext2fs_free_mem(&inodes_to_process);
2182         inodes_to_process = 0;
2183
2184         if (scan)
2185                 ext2fs_close_inode_scan(scan);
2186         if (block_buf)
2187                 ext2fs_free_mem(&block_buf);
2188         if (inode)
2189                 ext2fs_free_mem(&inode);
2190
2191         /*
2192          * The l+f inode may have been cleared, so zap it now and
2193          * later passes will recalculate it if necessary
2194          */
2195         ctx->lost_and_found = 0;
2196
2197         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2198                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2199         else
2200                 ctx->invalid_bitmaps++;
2201 }
2202
2203 static void init_ext2_max_sizes()
2204 {
2205         int     i;
2206         __u64   max_sizes;
2207
2208         /*
2209          * Init ext2_max_sizes which will be immutable and shared between
2210          * threads
2211          */
2212 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
2213
2214         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
2215                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
2216                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
2217                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
2218                 max_sizes = (max_sizes * (1UL << i));
2219                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
2220         }
2221 #undef EXT2_BPP
2222 }
2223
2224 #ifdef HAVE_PTHREAD
2225 static errcode_t e2fsck_pass1_copy_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2226                                           ext2fs_generic_bitmap *dest)
2227 {
2228         errcode_t ret;
2229
2230         ret = ext2fs_copy_bitmap(*src, dest);
2231         if (ret)
2232                 return ret;
2233
2234         (*dest)->fs = fs;
2235
2236         return 0;
2237 }
2238
2239 static void e2fsck_pass1_free_bitmap(ext2fs_generic_bitmap *bitmap)
2240 {
2241         if (*bitmap) {
2242                 ext2fs_free_generic_bmap(*bitmap);
2243                 *bitmap = NULL;
2244         }
2245
2246 }
2247
2248 static errcode_t e2fsck_pass1_merge_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2249                                           ext2fs_generic_bitmap *dest)
2250 {
2251         errcode_t ret = 0;
2252
2253         if (*src) {
2254                 if (*dest == NULL) {
2255                         *dest = *src;
2256                         *src = NULL;
2257                 } else {
2258                         ret = ext2fs_merge_bitmap(*src, *dest, NULL, NULL);
2259                         if (ret)
2260                                 return ret;
2261                 }
2262                 (*dest)->fs = fs;
2263         }
2264
2265         return 0;
2266 }
2267
2268 static errcode_t e2fsck_pass1_copy_fs(ext2_filsys dest, e2fsck_t src_context,
2269                                       ext2_filsys src)
2270 {
2271         errcode_t       retval;
2272
2273         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2274         dest->inode_map = NULL;
2275         dest->block_map = NULL;
2276         dest->badblocks = NULL;
2277         if (dest->dblist)
2278                 dest->dblist->fs = dest;
2279         if (src->block_map) {
2280                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2281                                                   &dest->block_map);
2282                 if (retval)
2283                         return retval;
2284         }
2285         if (src->inode_map) {
2286                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2287                                                   &dest->inode_map);
2288                 if (retval)
2289                         return retval;
2290         }
2291
2292         if (src->badblocks) {
2293                 retval = ext2fs_badblocks_copy(src->badblocks,
2294                                                &dest->badblocks);
2295                 if (retval)
2296                         return retval;
2297         }
2298
2299         /* disable it for now */
2300         src_context->openfs_flags &= ~EXT2_FLAG_EXCLUSIVE;
2301         retval = ext2fs_open_channel(dest, src_context->io_options,
2302                                      src_context->io_manager,
2303                                      src_context->openfs_flags,
2304                                      src->io->block_size);
2305         if (retval)
2306                 return retval;
2307
2308         /* Block size might not be default */
2309         io_channel_set_blksize(dest->io, src->io->block_size);
2310         ehandler_init(dest->io);
2311
2312         assert(dest->io->magic == src->io->magic);
2313         assert(dest->io->manager == src->io->manager);
2314         assert(strcmp(dest->io->name, src->io->name) == 0);
2315         assert(dest->io->block_size == src->io->block_size);
2316         assert(dest->io->read_error == src->io->read_error);
2317         assert(dest->io->write_error == src->io->write_error);
2318         assert(dest->io->refcount == src->io->refcount);
2319         assert(dest->io->flags == src->io->flags);
2320         assert(dest->io->app_data == dest);
2321         assert(src->io->app_data == src);
2322         assert(dest->io->align == src->io->align);
2323
2324         /* The data should be written to disk immediately */
2325         dest->io->flags |= CHANNEL_FLAGS_WRITETHROUGH;
2326         /* icache will be rebuilt if needed, so do not copy from @src */
2327         src->icache = NULL;
2328         return 0;
2329 }
2330
2331 static int e2fsck_pass1_merge_fs(ext2_filsys dest, ext2_filsys src)
2332 {
2333         struct ext2_inode_cache *icache = dest->icache;
2334         errcode_t retval = 0;
2335         io_channel dest_io;
2336         io_channel dest_image_io;
2337         ext2fs_inode_bitmap inode_map;
2338         ext2fs_block_bitmap block_map;
2339         ext2_badblocks_list badblocks;
2340         ext2_dblist dblist;
2341         int flags;
2342
2343         dest_io = dest->io;
2344         dest_image_io = dest->image_io;
2345         inode_map = dest->inode_map;
2346         block_map = dest->block_map;
2347         badblocks = dest->badblocks;
2348         dblist = dest->dblist;
2349         flags = dest->flags;
2350
2351         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2352         dest->io = dest_io;
2353         dest->image_io = dest_image_io;
2354         dest->icache = icache;
2355         dest->inode_map = inode_map;
2356         dest->block_map = block_map;
2357         dest->badblocks = badblocks;
2358         dest->dblist = dblist;
2359         if (dest->dblist)
2360                 dest->dblist->fs = dest;
2361         dest->flags = src->flags | flags;
2362         if (!(src->flags & EXT2_FLAG_VALID) || !(flags & EXT2_FLAG_VALID))
2363                 ext2fs_unmark_valid(dest);
2364
2365         if (src->icache) {
2366                 ext2fs_free_inode_cache(src->icache);
2367                 src->icache = NULL;
2368         }
2369
2370         retval = e2fsck_pass1_merge_bitmap(dest, &src->inode_map,
2371                                            &dest->inode_map);
2372         if (retval)
2373                 goto out;
2374
2375         retval = e2fsck_pass1_merge_bitmap(dest, &src->block_map,
2376                                           &dest->block_map);
2377         if (retval)
2378                 goto out;
2379
2380         if (src->dblist) {
2381                 if (dest->dblist) {
2382                         retval = ext2fs_merge_dblist(src->dblist,
2383                                                      dest->dblist);
2384                         if (retval)
2385                                 goto out;
2386                 } else {
2387                         dest->dblist = src->dblist;
2388                         dest->dblist->fs = dest;
2389                         src->dblist = NULL;
2390                 }
2391         }
2392
2393         if (src->badblocks) {
2394                 if (dest->badblocks == NULL)
2395                         retval = ext2fs_badblocks_copy(src->badblocks,
2396                                                        &dest->badblocks);
2397                 else
2398                         retval = ext2fs_badblocks_merge(src->badblocks,
2399                                                         dest->badblocks);
2400         }
2401 out:
2402         io_channel_close(src->io);
2403         if (src->inode_map)
2404                 ext2fs_free_generic_bmap(src->inode_map);
2405         if (src->block_map)
2406                 ext2fs_free_generic_bmap(src->block_map);
2407         if (src->badblocks)
2408                 ext2fs_badblocks_list_free(src->badblocks);
2409         if (src->dblist)
2410                 ext2fs_free_dblist(src->dblist);
2411
2412         return retval;
2413 }
2414
2415 static void e2fsck_pass1_copy_invalid_bitmaps(e2fsck_t global_ctx,
2416                                               e2fsck_t thread_ctx)
2417 {
2418         dgrp_t i, j;
2419         dgrp_t grp_start = thread_ctx->thread_info.et_group_start;
2420         dgrp_t grp_end = thread_ctx->thread_info.et_group_end;
2421         dgrp_t total = grp_end - grp_start;
2422
2423         thread_ctx->invalid_inode_bitmap_flag =
2424                         e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2425                                                 "invalid_inode_bitmap");
2426         thread_ctx->invalid_block_bitmap_flag =
2427                         e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2428                                                "invalid_block_bitmap");
2429         thread_ctx->invalid_inode_table_flag =
2430                         e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2431                                                "invalid_inode_table");
2432
2433         memcpy(thread_ctx->invalid_block_bitmap_flag,
2434                &global_ctx->invalid_block_bitmap_flag[grp_start],
2435                total * sizeof(int));
2436         memcpy(thread_ctx->invalid_inode_bitmap_flag,
2437                &global_ctx->invalid_inode_bitmap_flag[grp_start],
2438                total * sizeof(int));
2439         memcpy(thread_ctx->invalid_inode_table_flag,
2440                &global_ctx->invalid_inode_table_flag[grp_start],
2441                total * sizeof(int));
2442
2443         thread_ctx->invalid_bitmaps = 0;
2444         for (i = grp_start, j = 0; i < grp_end; i++, j++) {
2445                 if (thread_ctx->invalid_block_bitmap_flag[j])
2446                         thread_ctx->invalid_bitmaps++;
2447                 if (thread_ctx->invalid_inode_bitmap_flag[j])
2448                         thread_ctx->invalid_bitmaps++;
2449                 if (thread_ctx->invalid_inode_table_flag[j])
2450                         thread_ctx->invalid_bitmaps++;
2451         }
2452 }
2453
2454 static void e2fsck_pass1_merge_invalid_bitmaps(e2fsck_t global_ctx,
2455                                                e2fsck_t thread_ctx)
2456 {
2457         dgrp_t i, j;
2458         dgrp_t grp_start = thread_ctx->thread_info.et_group_start;
2459         dgrp_t grp_end = thread_ctx->thread_info.et_group_end;
2460         dgrp_t total = grp_end - grp_start;
2461
2462         memcpy(&global_ctx->invalid_block_bitmap_flag[grp_start],
2463                thread_ctx->invalid_block_bitmap_flag, total * sizeof(int));
2464         memcpy(&global_ctx->invalid_inode_bitmap_flag[grp_start],
2465                thread_ctx->invalid_inode_bitmap_flag, total * sizeof(int));
2466         memcpy(&global_ctx->invalid_inode_table_flag[grp_start],
2467                thread_ctx->invalid_inode_table_flag, total * sizeof(int));
2468         global_ctx->invalid_bitmaps += thread_ctx->invalid_bitmaps;
2469 }
2470
2471 static errcode_t e2fsck_pass1_thread_prepare(e2fsck_t global_ctx, e2fsck_t *thread_ctx,
2472                                              int thread_index, int num_threads)
2473 {
2474         errcode_t               retval;
2475         e2fsck_t                thread_context;
2476         ext2_filsys             thread_fs;
2477         ext2_filsys             global_fs = global_ctx->fs;
2478         struct e2fsck_thread    *tinfo;
2479         dgrp_t                  average_group;
2480
2481         assert(global_ctx->inode_used_map == NULL);
2482         assert(global_ctx->inode_dir_map == NULL);
2483         assert(global_ctx->inode_bb_map == NULL);
2484         assert(global_ctx->inode_imagic_map == NULL);
2485         assert(global_ctx->inode_reg_map == NULL);
2486         assert(global_ctx->inodes_to_rebuild == NULL);
2487
2488         assert(global_ctx->block_found_map != NULL);
2489         assert(global_ctx->block_metadata_map != NULL);
2490         assert(global_ctx->block_dup_map == NULL);
2491         assert(global_ctx->block_ea_map == NULL);
2492         assert(global_ctx->fs->dblist == NULL);
2493
2494         retval = ext2fs_get_mem(sizeof(struct e2fsck_struct), &thread_context);
2495         if (retval) {
2496                 com_err(global_ctx->program_name, retval, "while allocating memory");
2497                 return retval;
2498         }
2499         memcpy(thread_context, global_ctx, sizeof(struct e2fsck_struct));
2500         thread_context->global_ctx = global_ctx;
2501
2502         retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &thread_fs);
2503         if (retval) {
2504                 com_err(global_ctx->program_name, retval, "while allocating memory");
2505                 goto out_context;
2506         }
2507
2508         io_channel_flush_cleanup(global_fs->io);
2509         retval = e2fsck_pass1_copy_fs(thread_fs, global_ctx, global_fs);
2510         if (retval) {
2511                 com_err(global_ctx->program_name, retval, "while copying fs");
2512                 goto out_fs;
2513         }
2514         thread_fs->priv_data = thread_context;
2515
2516         thread_context->thread_info.et_thread_index = thread_index;
2517         set_up_logging(thread_context);
2518
2519         /*
2520          * Distribute work to multiple threads:
2521          * Each thread work on fs->group_desc_count / nthread groups.
2522          */
2523         tinfo = &thread_context->thread_info;
2524         average_group = thread_fs->group_desc_count / num_threads;
2525         if (average_group == 0)
2526                 average_group = 1;
2527         tinfo->et_group_start = average_group * thread_index;
2528         if (thread_index == num_threads - 1)
2529                 tinfo->et_group_end = thread_fs->group_desc_count;
2530         else
2531                 tinfo->et_group_end = average_group * (thread_index + 1);
2532         tinfo->et_group_next = tinfo->et_group_start;
2533         tinfo->et_inode_number = 0;
2534         tinfo->et_log_buf[0] = '\0';
2535         tinfo->et_log_length = 0;
2536         if (thread_context->options & E2F_OPT_MULTITHREAD)
2537                 log_out(thread_context, _("Scan group range [%d, %d)\n"),
2538                         tinfo->et_group_start, tinfo->et_group_end);
2539         thread_context->fs = thread_fs;
2540         retval = quota_init_context(&thread_context->qctx, thread_fs, 0);
2541         if (retval) {
2542                 com_err(global_ctx->program_name, retval,
2543                         "while init quota context");
2544                 goto out_fs;
2545         }
2546         *thread_ctx = thread_context;
2547         e2fsck_pass1_copy_invalid_bitmaps(global_ctx, thread_context);
2548         return 0;
2549 out_fs:
2550         ext2fs_free_mem(&thread_fs);
2551 out_context:
2552         ext2fs_free_mem(&thread_context);
2553         return retval;
2554 }
2555
2556 static void e2fsck_pass1_merge_dir_info(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2557 {
2558         if (thread_ctx->dir_info == NULL)
2559                 return;
2560
2561         if (global_ctx->dir_info == NULL) {
2562                 global_ctx->dir_info = thread_ctx->dir_info;
2563                 thread_ctx->dir_info = NULL;
2564                 return;
2565         }
2566
2567         e2fsck_merge_dir_info(global_ctx, thread_ctx->dir_info,
2568                               global_ctx->dir_info);
2569 }
2570
2571 static void e2fsck_pass1_merge_dx_dir(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2572 {
2573         if (thread_ctx->dx_dir_info == NULL)
2574                 return;
2575
2576         if (global_ctx->dx_dir_info == NULL) {
2577                 global_ctx->dx_dir_info = thread_ctx->dx_dir_info;
2578                 global_ctx->dx_dir_info_size = thread_ctx->dx_dir_info_size;
2579                 global_ctx->dx_dir_info_count = thread_ctx->dx_dir_info_count;
2580                 thread_ctx->dx_dir_info = NULL;
2581                 return;
2582         }
2583
2584         e2fsck_merge_dx_dir(global_ctx, thread_ctx);
2585 }
2586
2587 static inline errcode_t
2588 e2fsck_pass1_merge_icount(ext2_icount_t *dest_icount,
2589                           ext2_icount_t *src_icount)
2590 {
2591         if (*src_icount) {
2592                 if (*dest_icount == NULL) {
2593                         *dest_icount = *src_icount;
2594                         *src_icount = NULL;
2595                 } else {
2596                         errcode_t ret;
2597
2598                         ret = ext2fs_icount_merge(*src_icount,
2599                                                   *dest_icount);
2600                         if (ret)
2601                                 return ret;
2602                 }
2603         }
2604
2605         return 0;
2606 }
2607
2608 static errcode_t e2fsck_pass1_merge_icounts(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2609 {
2610         errcode_t ret;
2611
2612         ret = e2fsck_pass1_merge_icount(&global_ctx->inode_count,
2613                                         &thread_ctx->inode_count);
2614         if (ret)
2615                 return ret;
2616         ret = e2fsck_pass1_merge_icount(&global_ctx->inode_link_info,
2617                                         &thread_ctx->inode_link_info);
2618
2619         return ret;
2620 }
2621
2622 static errcode_t e2fsck_pass1_merge_dirs_to_hash(e2fsck_t global_ctx,
2623                                                  e2fsck_t thread_ctx)
2624 {
2625         errcode_t retval = 0;
2626
2627         if (!thread_ctx->dirs_to_hash)
2628                 return 0;
2629
2630         if (!global_ctx->dirs_to_hash)
2631                 retval = ext2fs_badblocks_copy(thread_ctx->dirs_to_hash,
2632                                                &global_ctx->dirs_to_hash);
2633         else
2634                 retval = ext2fs_badblocks_merge(thread_ctx->dirs_to_hash,
2635                                                 global_ctx->dirs_to_hash);
2636
2637         return retval;
2638 }
2639
2640 static errcode_t e2fsck_pass1_merge_ea_inode_refs(e2fsck_t global_ctx,
2641                                                   e2fsck_t thread_ctx)
2642 {
2643         ea_value_t count;
2644         blk64_t blk;
2645         errcode_t retval;
2646
2647         if (!thread_ctx->ea_inode_refs)
2648                 return 0;
2649
2650         if (!global_ctx->ea_inode_refs) {
2651                 global_ctx->ea_inode_refs = thread_ctx->ea_inode_refs;
2652                 thread_ctx->ea_inode_refs = NULL;
2653                 return 0;
2654         }
2655
2656         ea_refcount_intr_begin(thread_ctx->ea_inode_refs);
2657         while (1) {
2658                 if ((blk = ea_refcount_intr_next(thread_ctx->ea_inode_refs,
2659                                                  &count)) == 0)
2660                         break;
2661                 if (!global_ctx->block_ea_map ||
2662                     !ext2fs_fast_test_block_bitmap2(global_ctx->block_ea_map,
2663                                                     blk)) {
2664                         retval = ea_refcount_store(global_ctx->ea_inode_refs,
2665                                                    blk, count);
2666                         if (retval)
2667                                 return retval;
2668                 }
2669         }
2670
2671         return retval;
2672 }
2673
2674 static ea_value_t ea_refcount_usage(e2fsck_t ctx, blk64_t blk,
2675                                     ea_value_t *orig)
2676 {
2677         ea_value_t count_cur;
2678         ea_value_t count_extra = 0;
2679         ea_value_t count_orig;
2680
2681         ea_refcount_fetch(ctx->refcount_orig, blk, &count_orig);
2682         ea_refcount_fetch(ctx->refcount, blk, &count_cur);
2683         /* most of time this is not needed */
2684         if (ctx->refcount_extra && count_cur == 0)
2685                 ea_refcount_fetch(ctx->refcount_extra, blk, &count_extra);
2686
2687         if (!count_orig)
2688                 count_orig = *orig;
2689         else if (orig)
2690                 *orig = count_orig;
2691
2692         return count_orig + count_extra - count_cur;
2693 }
2694
2695 static errcode_t e2fsck_pass1_merge_ea_refcount(e2fsck_t global_ctx,
2696                                                 e2fsck_t thread_ctx)
2697 {
2698         ea_value_t count;
2699         blk64_t blk;
2700         errcode_t retval = 0;
2701
2702         if (!thread_ctx->refcount)
2703                 return 0;
2704
2705         if (!global_ctx->refcount) {
2706                 global_ctx->refcount = thread_ctx->refcount;
2707                 thread_ctx->refcount = NULL;
2708                 global_ctx->refcount_extra = thread_ctx->refcount;
2709                 thread_ctx->refcount_extra = NULL;
2710                 return 0;
2711         }
2712
2713         ea_refcount_intr_begin(thread_ctx->refcount);
2714         while (1) {
2715                 if ((blk = ea_refcount_intr_next(thread_ctx->refcount,
2716                                                  &count)) == 0)
2717                         break;
2718                 /**
2719                  * this EA has never seen before, so just store its
2720                  * refcount and refcount_extra into global_ctx if needed.
2721                  */
2722                 if (!global_ctx->block_ea_map ||
2723                     !ext2fs_fast_test_block_bitmap2(global_ctx->block_ea_map,
2724                                                     blk)) {
2725                         ea_value_t extra;
2726
2727                         retval = ea_refcount_store(global_ctx->refcount,
2728                                                    blk, count);
2729                         if (retval)
2730                                 return retval;
2731
2732                         if (count > 0 || !thread_ctx->refcount_extra)
2733                                 continue;
2734                         ea_refcount_fetch(thread_ctx->refcount_extra, blk,
2735                                           &extra);
2736                         if (extra == 0)
2737                                 continue;
2738
2739                         if (!global_ctx->refcount_extra) {
2740                                 retval = ea_refcount_create(0,
2741                                                 &global_ctx->refcount_extra);
2742                                 if (retval)
2743                                         return retval;
2744                         }
2745                         retval = ea_refcount_store(global_ctx->refcount_extra,
2746                                                    blk, extra);
2747                         if (retval)
2748                                 return retval;
2749                 } else {
2750                         ea_value_t orig;
2751                         ea_value_t thread_usage;
2752                         ea_value_t global_usage;
2753                         ea_value_t new;
2754
2755                         thread_usage = ea_refcount_usage(thread_ctx,
2756                                                          blk, &orig);
2757                         global_usage = ea_refcount_usage(global_ctx,
2758                                                          blk, &orig);
2759                         if (thread_usage + global_usage <= orig) {
2760                                 new = orig - thread_usage - global_usage;
2761                                 retval = ea_refcount_store(global_ctx->refcount,
2762                                                            blk, new);
2763                                 if (retval)
2764                                         return retval;
2765                                 continue;
2766                         }
2767                         /* update it is as zero */
2768                         retval = ea_refcount_store(global_ctx->refcount,
2769                                                    blk, 0);
2770                         if (retval)
2771                                 return retval;
2772                         /* Ooops, this EA was referenced more than it stated */
2773                         if (!global_ctx->refcount_extra) {
2774                                 retval = ea_refcount_create(0,
2775                                                 &global_ctx->refcount_extra);
2776                                 if (retval)
2777                                         return retval;
2778                         }
2779                         new = global_usage + thread_usage - orig;
2780                         retval = ea_refcount_store(global_ctx->refcount_extra,
2781                                                    blk, new);
2782                         if (retval)
2783                                 return retval;
2784                 }
2785         }
2786
2787         return retval;
2788 }
2789
2790 static int e2fsck_pass1_thread_join_one(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2791 {
2792         errcode_t        retval;
2793         int              flags = global_ctx->flags;
2794         ext2_filsys      thread_fs = thread_ctx->fs;
2795         ext2_filsys      global_fs = global_ctx->fs;
2796         FILE            *global_logf = global_ctx->logf;
2797         FILE            *global_problem_logf = global_ctx->problem_logf;
2798         ext2fs_inode_bitmap inode_bad_map = global_ctx->inode_bad_map;
2799         struct dir_info_db *dir_info = global_ctx->dir_info;
2800         struct dx_dir_info *dx_dir_info = global_ctx->dx_dir_info;
2801         ext2fs_inode_bitmap inode_used_map = global_ctx->inode_used_map;
2802         ext2fs_inode_bitmap inode_dir_map = global_ctx->inode_dir_map;
2803         ext2fs_inode_bitmap inode_bb_map = global_ctx->inode_bb_map;
2804         ext2fs_inode_bitmap inode_imagic_map = global_ctx->inode_imagic_map;
2805         ext2fs_inode_bitmap inode_reg_map = global_ctx->inode_reg_map;
2806         ext2fs_block_bitmap block_dup_map = global_ctx->block_dup_map;
2807         ext2fs_block_bitmap inodes_to_rebuild = global_ctx->inodes_to_rebuild;
2808         ext2_icount_t inode_count = global_ctx->inode_count;
2809         ext2_icount_t inode_link_info = global_ctx->inode_link_info;
2810         __u32 fs_directory_count = global_ctx->fs_directory_count;
2811         __u32 fs_regular_count = global_ctx->fs_regular_count;
2812         __u32 fs_blockdev_count = global_ctx->fs_blockdev_count;
2813         __u32 fs_chardev_count = global_ctx->fs_chardev_count;
2814         __u32 fs_links_count = global_ctx->fs_links_count;
2815         __u32 fs_symlinks_count = global_ctx->fs_symlinks_count;
2816         __u32 fs_fast_symlinks_count = global_ctx->fs_fast_symlinks_count;
2817         __u32 fs_fifo_count = global_ctx->fs_fifo_count;
2818         __u32 fs_total_count = global_ctx->fs_total_count;
2819         __u32 fs_badblocks_count = global_ctx->fs_badblocks_count;
2820         __u32 fs_sockets_count = global_ctx->fs_sockets_count;
2821         __u32 fs_ind_count = global_ctx->fs_ind_count;
2822         __u32 fs_dind_count = global_ctx->fs_dind_count;
2823         __u32 fs_tind_count = global_ctx->fs_tind_count;
2824         __u32 fs_fragmented = global_ctx->fs_fragmented;
2825         __u32 fs_fragmented_dir = global_ctx->fs_fragmented_dir;
2826         __u32 large_files = global_ctx->large_files;
2827         ext2_ino_t dx_dir_info_size = global_ctx->dx_dir_info_size;
2828         ext2_ino_t dx_dir_info_count = global_ctx->dx_dir_info_count;
2829         ext2_u32_list dirs_to_hash = global_ctx->dirs_to_hash;
2830         quota_ctx_t qctx = global_ctx->qctx;
2831         int *invalid_block_bitmap_flag = global_ctx->invalid_block_bitmap_flag;
2832         int *invalid_inode_bitmap_flag = global_ctx->invalid_inode_bitmap_flag;
2833         int *invalid_inode_table_flag  = global_ctx->invalid_inode_table_flag;
2834         int invalid_bitmaps = global_ctx->invalid_bitmaps;
2835         ext2_refcount_t refcount = global_ctx->refcount;
2836         ext2_refcount_t refcount_extra = global_ctx->refcount_extra;
2837         ext2_refcount_t refcount_orig = global_ctx->refcount_orig;
2838         ext2_refcount_t ea_block_quota_blocks = global_ctx->ea_block_quota_blocks;
2839         ext2_refcount_t ea_block_quota_inodes = global_ctx->ea_block_quota_inodes;
2840         ext2fs_block_bitmap block_ea_map = global_ctx->block_ea_map;
2841         ext2_refcount_t ea_inode_refs = global_ctx->ea_inode_refs;
2842
2843 #ifdef HAVE_SETJMP_H
2844         jmp_buf          old_jmp;
2845
2846         memcpy(old_jmp, global_ctx->abort_loc, sizeof(jmp_buf));
2847 #endif
2848         memcpy(global_ctx, thread_ctx, sizeof(struct e2fsck_struct));
2849 #ifdef HAVE_SETJMP_H
2850         memcpy(global_ctx->abort_loc, old_jmp, sizeof(jmp_buf));
2851 #endif
2852
2853         global_ctx->inode_used_map = inode_used_map;
2854         global_ctx->inode_bad_map = inode_bad_map;
2855         global_ctx->inode_dir_map = inode_dir_map;
2856         global_ctx->inode_bb_map = inode_bb_map;
2857         global_ctx->inode_imagic_map = inode_imagic_map;
2858         global_ctx->inodes_to_rebuild = inodes_to_rebuild;
2859         global_ctx->inode_reg_map = inode_reg_map;
2860         global_ctx->block_dup_map = block_dup_map;
2861         global_ctx->dir_info = dir_info;
2862         e2fsck_pass1_merge_dir_info(global_ctx, thread_ctx);
2863         global_ctx->dx_dir_info = dx_dir_info;
2864         global_ctx->dx_dir_info_count = dx_dir_info_count;
2865         global_ctx->dx_dir_info_size = dx_dir_info_size;
2866         e2fsck_pass1_merge_dx_dir(global_ctx, thread_ctx);
2867         global_ctx->inode_count = inode_count;
2868         global_ctx->inode_link_info = inode_link_info;
2869         global_ctx->refcount = refcount;
2870         global_ctx->refcount_extra = refcount_extra;
2871         global_ctx->refcount_orig = refcount_orig;
2872         global_ctx->ea_block_quota_blocks = ea_block_quota_blocks;
2873         global_ctx->ea_block_quota_inodes = ea_block_quota_inodes;
2874         global_ctx->block_ea_map = block_ea_map;
2875         global_ctx->ea_inode_refs = ea_inode_refs;
2876         global_ctx->fs_directory_count += fs_directory_count;
2877         global_ctx->fs_regular_count += fs_regular_count;
2878         global_ctx->fs_blockdev_count += fs_blockdev_count;
2879         global_ctx->fs_chardev_count += fs_chardev_count;
2880         global_ctx->fs_links_count += fs_links_count;
2881         global_ctx->fs_symlinks_count += fs_symlinks_count;
2882         global_ctx->fs_fast_symlinks_count += fs_fast_symlinks_count;
2883         global_ctx->fs_fifo_count += fs_fifo_count;
2884         global_ctx->fs_total_count += fs_total_count;
2885         global_ctx->fs_badblocks_count += fs_badblocks_count;
2886         global_ctx->fs_sockets_count += fs_sockets_count;
2887         global_ctx->fs_ind_count += fs_ind_count;
2888         global_ctx->fs_dind_count += fs_dind_count;
2889         global_ctx->fs_tind_count += fs_tind_count;
2890         global_ctx->fs_fragmented += fs_fragmented;
2891         global_ctx->fs_fragmented_dir += fs_fragmented_dir;
2892         global_ctx->large_files += large_files;
2893
2894         global_ctx->flags |= flags;
2895
2896         retval = e2fsck_pass1_merge_fs(global_fs, thread_fs);
2897         if (retval) {
2898                 com_err(global_ctx->program_name, 0, _("while merging fs\n"));
2899                 return retval;
2900         }
2901         global_fs->priv_data = global_ctx;
2902         global_ctx->fs = global_fs;
2903         global_ctx->logf = global_logf;
2904         global_ctx->problem_logf = global_problem_logf;
2905         global_ctx->global_ctx = NULL;
2906         retval = e2fsck_pass1_merge_icounts(global_ctx, thread_ctx);
2907         if (retval) {
2908                 com_err(global_ctx->program_name, 0,
2909                         _("while merging icounts\n"));
2910                 return retval;
2911         }
2912
2913         global_ctx->dirs_to_hash = dirs_to_hash;
2914         retval = e2fsck_pass1_merge_dirs_to_hash(global_ctx, thread_ctx);
2915         if (retval) {
2916                 com_err(global_ctx->program_name, 0,
2917                         _("while merging dirs to hash\n"));
2918                 return retval;
2919         }
2920
2921         e2fsck_pass1_merge_ea_inode_refs(global_ctx, thread_ctx);
2922         e2fsck_pass1_merge_ea_refcount(global_ctx, thread_ctx);
2923         global_ctx->qctx = qctx;
2924         retval = quota_merge_and_update_usage(global_ctx->qctx,
2925                                               thread_ctx->qctx);
2926         if (retval)
2927                 return retval;
2928         global_ctx->invalid_block_bitmap_flag = invalid_block_bitmap_flag;
2929         global_ctx->invalid_inode_bitmap_flag = invalid_inode_bitmap_flag;
2930         global_ctx->invalid_inode_table_flag = invalid_inode_table_flag;
2931         global_ctx->invalid_bitmaps = invalid_bitmaps;
2932         e2fsck_pass1_merge_invalid_bitmaps(global_ctx, thread_ctx);
2933
2934         retval = e2fsck_pass1_merge_bitmap(global_fs,
2935                                 &thread_ctx->inode_used_map,
2936                                 &global_ctx->inode_used_map);
2937         if (retval)
2938                 return retval;
2939
2940         retval = e2fsck_pass1_merge_bitmap(global_fs,
2941                                 &thread_ctx->inode_bad_map,
2942                                 &global_ctx->inode_bad_map);
2943         if (retval)
2944                 return retval;
2945         retval = e2fsck_pass1_merge_bitmap(global_fs,
2946                                         &thread_ctx->inode_dir_map,
2947                                         &global_ctx->inode_dir_map);
2948         if (retval)
2949                 return retval;
2950         retval = e2fsck_pass1_merge_bitmap(global_fs,
2951                                 &thread_ctx->inode_bb_map,
2952                                 &global_ctx->inode_bb_map);
2953         if (retval)
2954                 return retval;
2955         retval = e2fsck_pass1_merge_bitmap(global_fs,
2956                                 &thread_ctx->inode_imagic_map,
2957                                 &global_ctx->inode_imagic_map);
2958         if (retval)
2959                 return retval;
2960         retval = e2fsck_pass1_merge_bitmap(global_fs,
2961                                 &thread_ctx->inode_reg_map,
2962                                 &global_ctx->inode_reg_map);
2963         if (retval)
2964                 return retval;
2965         retval = e2fsck_pass1_merge_bitmap(global_fs,
2966                                 &thread_ctx->inodes_to_rebuild,
2967                                 &global_ctx->inodes_to_rebuild);
2968         if (retval)
2969                 return retval;
2970         retval = e2fsck_pass1_merge_bitmap(global_fs,
2971                                 &thread_ctx->block_ea_map,
2972                                 &global_ctx->block_ea_map);
2973         if (retval)
2974                 return retval;
2975
2976         return 0;
2977 }
2978
2979 static int e2fsck_pass1_thread_join(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2980 {
2981         errcode_t       retval;
2982
2983         retval = e2fsck_pass1_thread_join_one(global_ctx, thread_ctx);
2984         ext2fs_free_mem(&thread_ctx->fs);
2985         if (thread_ctx->logf)
2986                 fclose(thread_ctx->logf);
2987         if (thread_ctx->problem_logf) {
2988                 fputs("</problem_log>\n", thread_ctx->problem_logf);
2989                 fclose(thread_ctx->problem_logf);
2990         }
2991         e2fsck_pass1_free_bitmap(&thread_ctx->inode_used_map);
2992         e2fsck_pass1_free_bitmap(&thread_ctx->inode_bad_map);
2993         e2fsck_pass1_free_bitmap(&thread_ctx->inode_dir_map);
2994         e2fsck_pass1_free_bitmap(&thread_ctx->inode_bb_map);
2995         e2fsck_pass1_free_bitmap(&thread_ctx->inode_imagic_map);
2996         e2fsck_pass1_free_bitmap(&thread_ctx->inode_reg_map);
2997         e2fsck_pass1_free_bitmap(&thread_ctx->inodes_to_rebuild);
2998         e2fsck_pass1_free_bitmap(&thread_ctx->block_ea_map);
2999         if (thread_ctx->refcount)
3000                 ea_refcount_free(thread_ctx->refcount);
3001         if (thread_ctx->refcount_extra)
3002                 ea_refcount_free(thread_ctx->refcount_extra);
3003         if (thread_ctx->ea_inode_refs)
3004                 ea_refcount_free(thread_ctx->ea_inode_refs);
3005         if (thread_ctx->refcount_orig)
3006                 ea_refcount_free(thread_ctx->refcount_orig);
3007         e2fsck_free_dir_info(thread_ctx);
3008         ext2fs_free_icount(thread_ctx->inode_count);
3009         ext2fs_free_icount(thread_ctx->inode_link_info);
3010         if (thread_ctx->dirs_to_hash)
3011                 ext2fs_badblocks_list_free(thread_ctx->dirs_to_hash);
3012         quota_release_context(&thread_ctx->qctx);
3013         ext2fs_free_mem(&thread_ctx->invalid_block_bitmap_flag);
3014         ext2fs_free_mem(&thread_ctx->invalid_inode_bitmap_flag);
3015         ext2fs_free_mem(&thread_ctx->invalid_inode_table_flag);
3016         ext2fs_free_mem(&thread_ctx);
3017
3018         return retval;
3019 }
3020
3021 static int e2fsck_pass1_threads_join(struct e2fsck_thread_info *infos,
3022                                       int num_threads, e2fsck_t global_ctx)
3023 {
3024         errcode_t                        rc;
3025         errcode_t                        ret = 0;
3026         int                              i;
3027         struct e2fsck_thread_info       *pinfo;
3028
3029         /* merge invalid bitmaps will recalculate it */
3030         global_ctx->invalid_bitmaps = 0;
3031         for (i = 0; i < num_threads; i++) {
3032                 pinfo = &infos[i];
3033
3034                 if (!pinfo->eti_started)
3035                         continue;
3036
3037                 rc = pthread_join(pinfo->eti_thread_id, NULL);
3038                 if (rc) {
3039                         com_err(global_ctx->program_name, rc,
3040                                 _("while joining thread\n"));
3041                         if (ret == 0)
3042                                 ret = rc;
3043                 }
3044                 rc = e2fsck_pass1_thread_join(global_ctx, infos[i].eti_thread_ctx);
3045                 if (rc) {
3046                         com_err(global_ctx->program_name, rc,
3047                                 _("while joining pass1 thread\n"));
3048                         if (ret == 0)
3049                                 ret = rc;
3050                 }
3051         }
3052         free(infos);
3053
3054         return ret;
3055 }
3056
3057 static void *e2fsck_pass1_thread(void *arg)
3058 {
3059         struct e2fsck_thread_info       *info = arg;
3060         e2fsck_t                         thread_ctx = info->eti_thread_ctx;
3061 #ifdef DEBUG_THREADS
3062         struct e2fsck_thread_debug      *thread_debug = info->eti_debug;
3063 #endif
3064
3065 #ifdef DEBUG_THREADS
3066         pthread_mutex_lock(&thread_debug->etd_mutex);
3067         while (info->eti_thread_index > thread_debug->etd_finished_threads) {
3068                 pthread_cond_wait(&thread_debug->etd_cond,
3069                                   &thread_debug->etd_mutex);
3070         }
3071         pthread_mutex_unlock(&thread_debug->etd_mutex);
3072 #endif
3073
3074 #ifdef HAVE_SETJMP_H
3075         /*
3076          * When fatal_error() happens, jump to here. The thread
3077          * context's flags will be saved, but its abort_loc will
3078          * be overwritten by original jump buffer for the later
3079          * tests.
3080          */
3081         if (setjmp(thread_ctx->abort_loc)) {
3082                 thread_ctx->flags &= ~E2F_FLAG_SETJMP_OK;
3083                 goto out;
3084         }
3085         thread_ctx->flags |= E2F_FLAG_SETJMP_OK;
3086 #endif
3087
3088         e2fsck_pass1_run(thread_ctx);
3089
3090 out:
3091         if (thread_ctx->options & E2F_OPT_MULTITHREAD)
3092                 log_out(thread_ctx,
3093                         _("Scanned group range [%lu, %lu), inodes %lu\n"),
3094                         thread_ctx->thread_info.et_group_start,
3095                         thread_ctx->thread_info.et_group_end,
3096                         thread_ctx->thread_info.et_inode_number);
3097
3098 #ifdef DEBUG_THREADS
3099         pthread_mutex_lock(&thread_debug->etd_mutex);
3100         thread_debug->etd_finished_threads++;
3101         pthread_cond_broadcast(&thread_debug->etd_cond);
3102         pthread_mutex_unlock(&thread_debug->etd_mutex);
3103 #endif
3104
3105         return NULL;
3106 }
3107
3108 static int e2fsck_pass1_threads_start(struct e2fsck_thread_info **pinfo,
3109                                       int num_threads, e2fsck_t global_ctx)
3110 {
3111         struct e2fsck_thread_info       *infos;
3112         pthread_attr_t                   attr;
3113         errcode_t                        retval;
3114         errcode_t                        ret;
3115         struct e2fsck_thread_info       *tmp_pinfo;
3116         int                              i;
3117         e2fsck_t                         thread_ctx;
3118 #ifdef DEBUG_THREADS
3119         struct e2fsck_thread_debug       thread_debug =
3120                 {PTHREAD_MUTEX_INITIALIZER, PTHREAD_COND_INITIALIZER, 0};
3121
3122         thread_debug.etd_finished_threads = 0;
3123 #endif
3124
3125         retval = pthread_attr_init(&attr);
3126         if (retval) {
3127                 com_err(global_ctx->program_name, retval,
3128                         _("while setting pthread attribute\n"));
3129                 return retval;
3130         }
3131
3132         infos = calloc(num_threads, sizeof(struct e2fsck_thread_info));
3133         if (infos == NULL) {
3134                 retval = -ENOMEM;
3135                 com_err(global_ctx->program_name, retval,
3136                         _("while allocating memory for threads\n"));
3137                 pthread_attr_destroy(&attr);
3138                 return retval;
3139         }
3140
3141         for (i = 0; i < num_threads; i++) {
3142                 tmp_pinfo = &infos[i];
3143                 tmp_pinfo->eti_thread_index = i;
3144 #ifdef DEBUG_THREADS
3145                 tmp_pinfo->eti_debug = &thread_debug;
3146 #endif
3147                 retval = e2fsck_pass1_thread_prepare(global_ctx, &thread_ctx,
3148                                                      i, num_threads);
3149                 if (retval) {
3150                         com_err(global_ctx->program_name, retval,
3151                                 _("while preparing pass1 thread\n"));
3152                         break;
3153                 }
3154                 tmp_pinfo->eti_thread_ctx = thread_ctx;
3155
3156                 retval = pthread_create(&tmp_pinfo->eti_thread_id, &attr,
3157                                         &e2fsck_pass1_thread, tmp_pinfo);
3158                 if (retval) {
3159                         com_err(global_ctx->program_name, retval,
3160                                 _("while creating thread\n"));
3161                         e2fsck_pass1_thread_join(global_ctx, thread_ctx);
3162                         break;
3163                 }
3164
3165                 tmp_pinfo->eti_started = 1;
3166         }
3167
3168         /* destroy the thread attribute object, since it is no longer needed */
3169         ret = pthread_attr_destroy(&attr);
3170         if (ret) {
3171                 com_err(global_ctx->program_name, ret,
3172                         _("while destroying thread attribute\n"));
3173                 if (retval == 0)
3174                         retval = ret;
3175         }
3176
3177         if (retval) {
3178                 e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
3179                 return retval;
3180         }
3181         *pinfo = infos;
3182         return 0;
3183 }
3184
3185 static void e2fsck_pass1_multithread(e2fsck_t global_ctx)
3186 {
3187         struct e2fsck_thread_info       *infos = NULL;
3188         int                              num_threads = 1;
3189         errcode_t                        retval;
3190
3191         pthread_mutex_init(&global_ctx->fs_fix_mutex, NULL);
3192         pthread_mutex_init(&global_ctx->fs_block_map_mutex, NULL);
3193         retval = e2fsck_pass1_threads_start(&infos, num_threads, global_ctx);
3194         if (retval) {
3195                 com_err(global_ctx->program_name, retval,
3196                         _("while starting pass1 threads\n"));
3197                 goto out_abort;
3198         }
3199
3200         retval = e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
3201         if (retval) {
3202                 com_err(global_ctx->program_name, retval,
3203                         _("while joining pass1 threads\n"));
3204                 goto out_abort;
3205         }
3206         return;
3207 out_abort:
3208         global_ctx->flags |= E2F_FLAG_ABORT;
3209         return;
3210 }
3211 #endif
3212
3213 /* TODO: tdb needs to be handled properly for multiple threads*/
3214 static int multiple_threads_supported(e2fsck_t ctx)
3215 {
3216 #ifdef  CONFIG_TDB
3217         unsigned int            threshold;
3218         ext2_ino_t              num_dirs;
3219         errcode_t               retval;
3220         char                    *tdb_dir;
3221         int                     enable;
3222
3223         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
3224                            &tdb_dir);
3225         profile_get_uint(ctx->profile, "scratch_files",
3226                          "numdirs_threshold", 0, 0, &threshold);
3227         profile_get_boolean(ctx->profile, "scratch_files",
3228                             "icount", 0, 1, &enable);
3229
3230         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
3231         if (retval)
3232                 num_dirs = 1024;        /* Guess */
3233
3234         /* tdb is unsupported now */
3235         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
3236             (!threshold || num_dirs > threshold))
3237                 return 0;
3238  #endif
3239         return 1;
3240 }
3241
3242 void e2fsck_pass1(e2fsck_t ctx)
3243 {
3244         errcode_t retval;
3245         int multiple = 0;
3246
3247         init_ext2_max_sizes();
3248         retval = e2fsck_pass1_prepare(ctx);
3249         if (retval)
3250                 return;
3251 #ifdef HAVE_PTHREAD
3252         if (multiple_threads_supported(ctx)) {
3253                 multiple = 1;
3254                 e2fsck_pass1_multithread(ctx);
3255         } else {
3256                 fprintf(stderr, "Fall through single thread for pass1 "
3257                                 "because tdb could not handle properly\n");
3258         }
3259 #endif
3260         if (!multiple)
3261                 e2fsck_pass1_run(ctx);
3262         e2fsck_pass1_post(ctx);
3263 }
3264
3265 #undef FINISH_INODE_LOOP
3266
3267 /*
3268  * When the inode_scan routines call this callback at the end of the
3269  * glock group, call process_inodes.
3270  */
3271 static errcode_t scan_callback(ext2_filsys fs,
3272                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
3273                                dgrp_t group, void * priv_data)
3274 {
3275         struct scan_callback_struct *scan_struct;
3276         e2fsck_t ctx;
3277         struct e2fsck_thread *tinfo;
3278
3279         scan_struct = (struct scan_callback_struct *) priv_data;
3280         ctx = scan_struct->ctx;
3281
3282         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf,
3283                        scan_struct->inodes_to_process,
3284                        scan_struct->process_inode_count);
3285
3286         if (ctx->progress)
3287                 if ((ctx->progress)(ctx, 1, group+1,
3288                                     ctx->fs->group_desc_count))
3289                         return EXT2_ET_CANCEL_REQUESTED;
3290
3291 #ifdef HAVE_PTHREAD
3292         if (ctx->global_ctx) {
3293                 tinfo = &ctx->thread_info;
3294                 tinfo->et_group_next++;
3295                 if (ctx->options & E2F_OPT_DEBUG &&
3296                     ctx->options & E2F_OPT_MULTITHREAD)
3297                         log_out(ctx, _("group %d finished\n"),
3298                                 tinfo->et_group_next);
3299                 if (tinfo->et_group_next >= tinfo->et_group_end)
3300                         return EXT2_ET_SCAN_FINISHED;
3301         }
3302 #endif
3303
3304         return 0;
3305 }
3306
3307 /*
3308  * Process the inodes in the "inodes to process" list.
3309  */
3310 static void process_inodes(e2fsck_t ctx, char *block_buf,
3311                            struct process_inode_block *inodes_to_process,
3312                            int *process_inode_count)
3313 {
3314         int                     i;
3315         struct ext2_inode       *old_stashed_inode;
3316         ext2_ino_t              old_stashed_ino;
3317         const char              *old_operation;
3318         char                    buf[80];
3319         struct problem_context  pctx;
3320
3321 #if 0
3322         printf("begin process_inodes: ");
3323 #endif
3324         if (*process_inode_count == 0)
3325                 return;
3326         old_operation = ehandler_operation(0);
3327         old_stashed_inode = ctx->stashed_inode;
3328         old_stashed_ino = ctx->stashed_ino;
3329         qsort(inodes_to_process, *process_inode_count,
3330                       sizeof(struct process_inode_block), process_inode_cmp);
3331         clear_problem_context(&pctx);
3332         for (i=0; i < *process_inode_count; i++) {
3333                 pctx.inode = ctx->stashed_inode =
3334                         (struct ext2_inode *) &inodes_to_process[i].inode;
3335                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
3336
3337 #if 0
3338                 printf("%u ", pctx.ino);
3339 #endif
3340                 sprintf(buf, _("reading indirect blocks of inode %u"),
3341                         pctx.ino);
3342                 ehandler_operation(buf);
3343                 check_blocks(ctx, &pctx, block_buf,
3344                              &inodes_to_process[i].ea_ibody_quota);
3345                 if (e2fsck_should_abort(ctx))
3346                         break;
3347         }
3348         ctx->stashed_inode = old_stashed_inode;
3349         ctx->stashed_ino = old_stashed_ino;
3350         *process_inode_count = 0;
3351 #if 0
3352         printf("end process inodes\n");
3353 #endif
3354         ehandler_operation(old_operation);
3355 }
3356
3357 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
3358 {
3359         const struct process_inode_block *ib_a =
3360                 (const struct process_inode_block *) a;
3361         const struct process_inode_block *ib_b =
3362                 (const struct process_inode_block *) b;
3363         int     ret;
3364
3365         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
3366                ib_b->inode.i_block[EXT2_IND_BLOCK]);
3367         if (ret == 0)
3368                 /*
3369                  * We only call process_inodes() for non-extent
3370                  * inodes, so it's OK to pass NULL to
3371                  * ext2fs_file_acl_block() here.
3372                  */
3373                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
3374                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
3375         if (ret == 0)
3376                 ret = ib_a->ino - ib_b->ino;
3377         return ret;
3378 }
3379
3380 /*
3381  * Mark an inode as being bad in some what
3382  */
3383 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
3384 {
3385         struct          problem_context pctx;
3386
3387         if (!ctx->inode_bad_map) {
3388                 clear_problem_context(&pctx);
3389
3390                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3391                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
3392                                 "inode_bad_map", &ctx->inode_bad_map);
3393                 if (pctx.errcode) {
3394                         pctx.num = 3;
3395                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3396                         /* Should never get here */
3397                         ctx->flags |= E2F_FLAG_ABORT;
3398                         return;
3399                 }
3400         }
3401         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
3402 }
3403
3404 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino)
3405 {
3406         struct          problem_context pctx;
3407
3408         if (!ctx->casefolded_dirs) {
3409                 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
3410                 if (pctx.errcode)
3411                         goto error;
3412         }
3413         pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
3414         if (pctx.errcode == 0)
3415                 return;
3416 error:
3417         fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
3418         /* Should never get here */
3419         ctx->flags |= E2F_FLAG_ABORT;
3420 }
3421
3422 /*
3423  * This procedure will allocate the inode "bb" (badblock) map table
3424  */
3425 static void alloc_bb_map(e2fsck_t ctx)
3426 {
3427         struct          problem_context pctx;
3428
3429         clear_problem_context(&pctx);
3430         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3431                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
3432                         "inode_bb_map", &ctx->inode_bb_map);
3433         if (pctx.errcode) {
3434                 pctx.num = 4;
3435                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3436                 /* Should never get here */
3437                 ctx->flags |= E2F_FLAG_ABORT;
3438                 return;
3439         }
3440 }
3441
3442 /*
3443  * This procedure will allocate the inode imagic table
3444  */
3445 static void alloc_imagic_map(e2fsck_t ctx)
3446 {
3447         struct          problem_context pctx;
3448
3449         clear_problem_context(&pctx);
3450         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3451                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
3452                         "inode_imagic_map", &ctx->inode_imagic_map);
3453         if (pctx.errcode) {
3454                 pctx.num = 5;
3455                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3456                 /* Should never get here */
3457                 ctx->flags |= E2F_FLAG_ABORT;
3458                 return;
3459         }
3460 }
3461
3462 /*
3463  * Marks a block as in use, setting the dup_map if it's been set
3464  * already.  Called by process_block and process_bad_block.
3465  *
3466  * WARNING: Assumes checks have already been done to make sure block
3467  * is valid.  This is true in both process_block and process_bad_block.
3468  */
3469 static _INLINE_ void mark_block_used_unlocked(e2fsck_t ctx, blk64_t block)
3470 {
3471         struct problem_context pctx;
3472         e2fsck_t global_ctx;
3473
3474         global_ctx = ctx->global_ctx;
3475         if (!global_ctx)
3476                 global_ctx = ctx;
3477
3478         clear_problem_context(&pctx);
3479
3480         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
3481                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
3482                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
3483                         return;
3484                 }
3485                 /**
3486                  * this should be safe because this operation has
3487                  * been serialized by mutex.
3488                  */
3489                 if (!global_ctx->block_dup_map) {
3490                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
3491                                         _("multiply claimed block map"),
3492                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
3493                                         &global_ctx->block_dup_map);
3494                         if (pctx.errcode) {
3495                                 pctx.num = 3;
3496                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
3497                                             &pctx);
3498                                 /* Should never get here */
3499                                 ctx->flags |= E2F_FLAG_ABORT;
3500                                 return;
3501                         }
3502                 }
3503                 ext2fs_fast_mark_block_bitmap2(global_ctx->block_dup_map, block);
3504         } else {
3505                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
3506         }
3507 }
3508
3509 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
3510 {
3511         e2fsck_pass1_block_map_lock(ctx);
3512         mark_block_used_unlocked(ctx, block);
3513         e2fsck_pass1_block_map_unlock(ctx);
3514
3515 }
3516
3517 /*
3518  * When cluster size is greater than one block, it is caller's responsibility
3519  * to make sure block parameter starts at a cluster boundary.
3520  */
3521 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
3522                                       unsigned int num)
3523 {
3524         e2fsck_pass1_block_map_lock(ctx);
3525         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num)) {
3526                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
3527         } else {
3528                 unsigned int i;
3529
3530                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
3531                         mark_block_used_unlocked(ctx, block + i);
3532         }
3533         e2fsck_pass1_block_map_unlock(ctx);
3534 }
3535
3536 static errcode_t _INLINE_ e2fsck_write_ext_attr3(e2fsck_t ctx, blk64_t block,
3537                                                  void *inbuf, ext2_ino_t inum)
3538 {
3539         errcode_t retval;
3540         ext2_filsys fs = ctx->fs;
3541
3542         e2fsck_pass1_fix_lock(ctx);
3543         retval = ext2fs_write_ext_attr3(fs, block, inbuf, inum);
3544         e2fsck_pass1_fix_unlock(ctx);
3545
3546         return retval;
3547 }
3548 /*
3549  * Adjust the extended attribute block's reference counts at the end
3550  * of pass 1, either by subtracting out references for EA blocks that
3551  * are still referenced in ctx->refcount, or by adding references for
3552  * EA blocks that had extra references as accounted for in
3553  * ctx->refcount_extra.
3554  */
3555 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
3556                                     char *block_buf, int adjust_sign)
3557 {
3558         struct ext2_ext_attr_header     *header;
3559         struct problem_context          pctx;
3560         ext2_filsys                     fs = ctx->fs;
3561         blk64_t                         blk;
3562         __u32                           should_be;
3563         ea_value_t                      count;
3564
3565         clear_problem_context(&pctx);
3566
3567         ea_refcount_intr_begin(refcount);
3568         while (1) {
3569                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
3570                         break;
3571                 pctx.blk = blk;
3572                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
3573                                                      pctx.ino);
3574                 if (pctx.errcode) {
3575                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
3576                         return;
3577                 }
3578                 header = (struct ext2_ext_attr_header *) block_buf;
3579                 pctx.blkcount = header->h_refcount;
3580                 should_be = header->h_refcount + adjust_sign * (int)count;
3581                 pctx.num = should_be;
3582                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
3583                         header->h_refcount = should_be;
3584                         pctx.errcode = e2fsck_write_ext_attr3(ctx, blk,
3585                                                              block_buf,
3586                                                              pctx.ino);
3587                         if (pctx.errcode) {
3588                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
3589                                             &pctx);
3590                                 continue;
3591                         }
3592                 }
3593         }
3594 }
3595
3596 /*
3597  * Handle processing the extended attribute blocks
3598  */
3599 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
3600                            char *block_buf, struct ea_quota *ea_block_quota)
3601 {
3602         ext2_filsys fs = ctx->fs;
3603         ext2_ino_t      ino = pctx->ino;
3604         struct ext2_inode *inode = pctx->inode;
3605         blk64_t         blk;
3606         char *          end;
3607         struct ext2_ext_attr_header *header;
3608         struct ext2_ext_attr_entry *first, *entry;
3609         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
3610         __u64           quota_inodes = 0;
3611         region_t        region = 0;
3612         int             failed_csum = 0;
3613
3614         ea_block_quota->blocks = 0;
3615         ea_block_quota->inodes = 0;
3616
3617         blk = ext2fs_file_acl_block(fs, inode);
3618         if (blk == 0)
3619                 return 0;
3620
3621         /*
3622          * If the Extended attribute flag isn't set, then a non-zero
3623          * file acl means that the inode is corrupted.
3624          *
3625          * Or if the extended attribute block is an invalid block,
3626          * then the inode is also corrupted.
3627          */
3628         if (!ext2fs_has_feature_xattr(fs->super) ||
3629             (blk < fs->super->s_first_data_block) ||
3630             (blk >= ext2fs_blocks_count(fs->super))) {
3631                 mark_inode_bad(ctx, ino);
3632                 return 0;
3633         }
3634
3635         /* If ea bitmap hasn't been allocated, create it */
3636         if (!ctx->block_ea_map) {
3637                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
3638                                         _("ext attr block map"),
3639                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
3640                                         &ctx->block_ea_map);
3641                 if (pctx->errcode) {
3642                         pctx->num = 2;
3643                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
3644                         ctx->flags |= E2F_FLAG_ABORT;
3645                         return 0;
3646                 }
3647         }
3648
3649         /* Create the EA refcount structure if necessary */
3650         if (!ctx->refcount) {
3651                 pctx->errcode = ea_refcount_create(0,
3652                                         &ctx->refcount_orig);
3653                 if (pctx->errcode) {
3654                         pctx->num = 1;
3655                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3656                         ctx->flags |= E2F_FLAG_ABORT;
3657                         return 0;
3658                 }
3659
3660                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
3661                 if (pctx->errcode) {
3662                         pctx->num = 1;
3663                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3664                         ctx->flags |= E2F_FLAG_ABORT;
3665                         return 0;
3666                 }
3667         }
3668
3669 #if 0
3670         /* Debugging text */
3671         printf("Inode %u has EA block %u\n", ino, blk);
3672 #endif
3673
3674         /* Have we seen this EA block before? */
3675         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
3676                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
3677                 ea_block_quota->inodes = 0;
3678
3679                 if (ctx->ea_block_quota_blocks) {
3680                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
3681                                           &quota_blocks);
3682                         if (quota_blocks)
3683                                 ea_block_quota->blocks = quota_blocks;
3684                 }
3685
3686                 if (ctx->ea_block_quota_inodes)
3687                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
3688                                           &ea_block_quota->inodes);
3689
3690                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
3691                         return 1;
3692                 /* Ooops, this EA was referenced more than it stated */
3693                 if (!ctx->refcount_extra) {
3694                         pctx->errcode = ea_refcount_create(0,
3695                                            &ctx->refcount_extra);
3696                         if (pctx->errcode) {
3697                                 pctx->num = 2;
3698                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3699                                 ctx->flags |= E2F_FLAG_ABORT;
3700                                 return 0;
3701                         }
3702                 }
3703                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
3704                 return 1;
3705         }
3706
3707         /*
3708          * OK, we haven't seen this EA block yet.  So we need to
3709          * validate it
3710          */
3711         pctx->blk = blk;
3712         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
3713         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
3714                 pctx->errcode = 0;
3715                 failed_csum = 1;
3716         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
3717                 pctx->errcode = 0;
3718
3719         if (pctx->errcode &&
3720             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
3721                 pctx->errcode = 0;
3722                 goto clear_extattr;
3723         }
3724         header = (struct ext2_ext_attr_header *) block_buf;
3725         pctx->blk = ext2fs_file_acl_block(fs, inode);
3726         if (((ctx->ext_attr_ver == 1) &&
3727              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
3728             ((ctx->ext_attr_ver == 2) &&
3729              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
3730                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
3731                         goto clear_extattr;
3732         }
3733
3734         if (header->h_blocks != 1) {
3735                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
3736                         goto clear_extattr;
3737         }
3738
3739         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
3740                 goto clear_extattr;
3741
3742         region = region_create(0, fs->blocksize);
3743         if (!region) {
3744                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
3745                 ctx->flags |= E2F_FLAG_ABORT;
3746                 return 0;
3747         }
3748         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
3749                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3750                         goto clear_extattr;
3751         }
3752
3753         first = (struct ext2_ext_attr_entry *)(header+1);
3754         end = block_buf + fs->blocksize;
3755         entry = first;
3756         while ((char *)entry < end && *(__u32 *)entry) {
3757                 __u32 hash;
3758
3759                 if (region_allocate(region, (char *)entry - (char *)header,
3760                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
3761                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3762                                 goto clear_extattr;
3763                         break;
3764                 }
3765                 if ((ctx->ext_attr_ver == 1 &&
3766                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
3767                     (ctx->ext_attr_ver == 2 &&
3768                      entry->e_name_index == 0)) {
3769                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
3770                                 goto clear_extattr;
3771                         break;
3772                 }
3773                 if (entry->e_value_inum == 0) {
3774                         if (entry->e_value_offs + entry->e_value_size >
3775                             fs->blocksize) {
3776                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
3777                                         goto clear_extattr;
3778                                 break;
3779                         }
3780                         if (entry->e_value_size &&
3781                             region_allocate(region, entry->e_value_offs,
3782                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
3783                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
3784                                                 pctx))
3785                                         goto clear_extattr;
3786                         }
3787
3788                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
3789                                                           entry->e_value_offs);
3790
3791                         if (entry->e_hash != hash) {
3792                                 pctx->num = entry->e_hash;
3793                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
3794                                         goto clear_extattr;
3795                                 entry->e_hash = hash;
3796                         }
3797                 } else {
3798                         problem_t problem;
3799                         blk64_t entry_quota_blocks;
3800
3801                         problem = check_large_ea_inode(ctx, entry, pctx,
3802                                                        &entry_quota_blocks);
3803                         if (problem && fix_problem(ctx, problem, pctx))
3804                                 goto clear_extattr;
3805
3806                         quota_blocks += entry_quota_blocks;
3807                         quota_inodes++;
3808                 }
3809
3810                 entry = EXT2_EXT_ATTR_NEXT(entry);
3811         }
3812         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
3813                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3814                         goto clear_extattr;
3815         }
3816         region_free(region);
3817
3818         /*
3819          * We only get here if there was no other errors that were fixed.
3820          * If there was a checksum fail, ask to correct it.
3821          */
3822         if (failed_csum &&
3823             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
3824                 pctx->errcode = e2fsck_write_ext_attr3(ctx, blk, block_buf,
3825                                                        pctx->ino);
3826                 if (pctx->errcode)
3827                         return 0;
3828         }
3829
3830         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
3831                 if (!ctx->ea_block_quota_blocks) {
3832                         pctx->errcode = ea_refcount_create(0,
3833                                                 &ctx->ea_block_quota_blocks);
3834                         if (pctx->errcode) {
3835                                 pctx->num = 3;
3836                                 goto refcount_fail;
3837                         }
3838                 }
3839                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
3840                                   quota_blocks);
3841         }
3842
3843         if (quota_inodes) {
3844                 if (!ctx->ea_block_quota_inodes) {
3845                         pctx->errcode = ea_refcount_create(0,
3846                                                 &ctx->ea_block_quota_inodes);
3847                         if (pctx->errcode) {
3848                                 pctx->num = 4;
3849 refcount_fail:
3850                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3851                                 ctx->flags |= E2F_FLAG_ABORT;
3852                                 return 0;
3853                         }
3854                 }
3855
3856                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
3857                                   quota_inodes);
3858         }
3859         ea_block_quota->blocks = quota_blocks;
3860         ea_block_quota->inodes = quota_inodes;
3861
3862         inc_ea_inode_refs(ctx, pctx, first, end);
3863         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
3864         ea_refcount_store(ctx->refcount_orig, blk, header->h_refcount);
3865         mark_block_used(ctx, blk);
3866         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
3867         return 1;
3868
3869 clear_extattr:
3870         if (region)
3871                 region_free(region);
3872         ext2fs_file_acl_block_set(fs, inode, 0);
3873         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
3874         return 0;
3875 }
3876
3877 /* Returns 1 if bad htree, 0 if OK */
3878 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
3879                         ext2_ino_t ino, struct ext2_inode *inode,
3880                         char *block_buf)
3881 {
3882         struct ext2_dx_root_info        *root;
3883         ext2_filsys                     fs = ctx->fs;
3884         errcode_t                       retval;
3885         blk64_t                         blk;
3886
3887         if ((!LINUX_S_ISDIR(inode->i_mode) &&
3888              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
3889             (!ext2fs_has_feature_dir_index(fs->super) &&
3890              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
3891                 return 1;
3892
3893         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
3894
3895         if ((pctx->errcode) ||
3896             (blk == 0) ||
3897             (blk < fs->super->s_first_data_block) ||
3898             (blk >= ext2fs_blocks_count(fs->super))) {
3899                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3900                         return 1;
3901                 else
3902                         return 0;
3903         }
3904
3905         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
3906         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3907                 return 1;
3908
3909         /* XXX should check that beginning matches a directory */
3910         root = (struct ext2_dx_root_info *) (block_buf + 24);
3911
3912         if ((root->reserved_zero || root->info_length < 8) &&
3913             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3914                 return 1;
3915
3916         pctx->num = root->hash_version;
3917         if ((root->hash_version != EXT2_HASH_LEGACY) &&
3918             (root->hash_version != EXT2_HASH_HALF_MD4) &&
3919             (root->hash_version != EXT2_HASH_TEA) &&
3920             (root->hash_version != EXT2_HASH_SIPHASH) &&
3921             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
3922                 return 1;
3923
3924         if (ext4_hash_in_dirent(inode)) {
3925                 if (root->hash_version != EXT2_HASH_SIPHASH &&
3926                     fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
3927                         return 1;
3928         } else {
3929                 if (root->hash_version == EXT2_HASH_SIPHASH &&
3930                    fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
3931                         return 1;
3932         }
3933
3934         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
3935             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
3936                 return 1;
3937
3938         pctx->num = root->indirect_levels;
3939         /* if htree level is clearly too high, consider it to be broken */
3940         if (root->indirect_levels > EXT4_HTREE_LEVEL &&
3941             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3942                 return 1;
3943
3944         /* if level is only maybe too high, LARGE_DIR feature could be unset */
3945         if (root->indirect_levels > ext2_dir_htree_level(fs) &&
3946             !ext2fs_has_feature_largedir(fs->super)) {
3947                 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
3948                 int idx_pb = 1 << (blockbits - 3);
3949
3950                 /* compare inode size/blocks vs. max-sized 2-level htree */
3951                 if (EXT2_I_SIZE(pctx->inode) <
3952                     (idx_pb - 1) * (idx_pb - 2) << blockbits &&
3953                     pctx->inode->i_blocks <
3954                     (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
3955                     fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3956                         return 1;
3957         }
3958
3959         if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
3960             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3961                 ctx->large_dirs++;
3962
3963         return 0;
3964 }
3965
3966 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
3967                         struct ext2_inode *inode, int restart_flag,
3968                         const char *source)
3969 {
3970         inode->i_flags = 0;
3971         inode->i_links_count = 0;
3972         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
3973         inode->i_dtime = ctx->now;
3974
3975         /*
3976          * If a special inode has such rotten block mappings that we
3977          * want to clear the whole inode, be sure to actually zap
3978          * the block maps because i_links_count isn't checked for
3979          * special inodes, and we'll end up right back here the next
3980          * time we run fsck.
3981          */
3982         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
3983                 memset(inode->i_block, 0, sizeof(inode->i_block));
3984
3985         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
3986         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
3987         if (ctx->inode_reg_map)
3988                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
3989         if (ctx->inode_bad_map)
3990                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
3991
3992         /*
3993          * If the inode was partially accounted for before processing
3994          * was aborted, we need to restart the pass 1 scan.
3995          */
3996         ctx->flags |= restart_flag;
3997
3998         if (ino == EXT2_BAD_INO)
3999                 memset(inode, 0, sizeof(struct ext2_inode));
4000
4001         e2fsck_write_inode(ctx, ino, inode, source);
4002 }
4003
4004 /*
4005  * Use the multiple-blocks reclamation code to fix alignment problems in
4006  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
4007  * physical cluster, and we want the block offsets within that cluster to
4008  * line up.
4009  */
4010 static int has_unaligned_cluster_map(e2fsck_t ctx,
4011                                      blk64_t last_pblk, blk64_t last_lblk,
4012                                      blk64_t pblk, blk64_t lblk)
4013 {
4014         blk64_t cluster_mask;
4015
4016         if (!ctx->fs->cluster_ratio_bits)
4017                 return 0;
4018         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
4019
4020         /*
4021          * If the block in the logical cluster doesn't align with the block in
4022          * the physical cluster...
4023          */
4024         if ((lblk & cluster_mask) != (pblk & cluster_mask))
4025                 return 1;
4026
4027         /*
4028          * If we cross a physical cluster boundary within a logical cluster...
4029          */
4030         if (last_pblk && (lblk & cluster_mask) != 0 &&
4031             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
4032             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
4033                 return 1;
4034
4035         return 0;
4036 }
4037
4038 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
4039                              struct process_block_struct *pb,
4040                              blk64_t start_block, blk64_t end_block,
4041                              blk64_t eof_block,
4042                              ext2_extent_handle_t ehandle,
4043                              int try_repairs)
4044 {
4045         struct ext2fs_extent    extent;
4046         blk64_t                 blk, last_lblk;
4047         unsigned int            i, n;
4048         int                     is_dir, is_leaf;
4049         problem_t               problem;
4050         struct ext2_extent_info info;
4051         int                     failed_csum = 0;
4052
4053         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
4054                 failed_csum = 1;
4055
4056         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
4057         if (pctx->errcode)
4058                 return;
4059         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
4060             !pb->eti.force_rebuild) {
4061                 struct extent_tree_level *etl;
4062
4063                 etl = pb->eti.ext_info + info.curr_level;
4064                 etl->num_extents += info.num_entries;
4065                 etl->max_extents += info.max_entries;
4066                 /*
4067                  * Implementation wart: Splitting extent blocks when appending
4068                  * will leave the old block with one free entry.  Therefore
4069                  * unless the node is totally full, pretend that a non-root
4070                  * extent block can hold one fewer entry than it actually does,
4071                  * so that we don't repeatedly rebuild the extent tree.
4072                  */
4073                 if (info.curr_level && info.num_entries < info.max_entries)
4074                         etl->max_extents--;
4075         }
4076
4077         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
4078                                           &extent);
4079         while ((pctx->errcode == 0 ||
4080                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
4081                info.num_entries-- > 0) {
4082                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
4083                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
4084                 last_lblk = extent.e_lblk + extent.e_len - 1;
4085
4086                 problem = 0;
4087                 pctx->blk = extent.e_pblk;
4088                 pctx->blk2 = extent.e_lblk;
4089                 pctx->num = extent.e_len;
4090                 pctx->blkcount = extent.e_lblk + extent.e_len;
4091
4092                 if (extent.e_pblk == 0 ||
4093                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
4094                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
4095                         problem = PR_1_EXTENT_BAD_START_BLK;
4096                 else if (extent.e_lblk < start_block)
4097                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
4098                 else if ((end_block && last_lblk > end_block) &&
4099                          !(last_lblk > eof_block &&
4100                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
4101                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
4102                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
4103                 else if (is_leaf && extent.e_len == 0)
4104                         problem = PR_1_EXTENT_LENGTH_ZERO;
4105                 else if (is_leaf &&
4106                          (extent.e_pblk + extent.e_len) >
4107                          ext2fs_blocks_count(ctx->fs->super))
4108                         problem = PR_1_EXTENT_ENDS_BEYOND;
4109                 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
4110                          !ext2fs_has_feature_largedir(ctx->fs->super) &&
4111                          ((extent.e_lblk + extent.e_len) >
4112                           (1U << (21 - ctx->fs->super->s_log_block_size))))
4113                         problem = PR_1_TOOBIG_DIR;
4114
4115                 if (is_leaf && problem == 0 && extent.e_len > 0) {
4116 #if 0
4117                         printf("extent_region(ino=%u, expect=%llu, "
4118                                "lblk=%llu, len=%u)\n", pb->ino,
4119                                (unsigned long long) pb->next_lblock,
4120                                (unsigned long long) extent.e_lblk,
4121                                extent.e_len);
4122 #endif
4123                         if (extent.e_lblk < pb->next_lblock)
4124                                 problem = PR_1_EXTENT_COLLISION;
4125                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
4126                                 pb->next_lblock = extent.e_lblk + extent.e_len;
4127                 }
4128
4129                 /*
4130                  * Uninitialized blocks in a directory?  Clear the flag and
4131                  * we'll interpret the blocks later.
4132                  */
4133                 if (try_repairs && is_dir && problem == 0 &&
4134                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
4135                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
4136                         e2fsck_pass1_fix_lock(ctx);
4137                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
4138                         pb->inode_modified = 1;
4139                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
4140                                                               &extent);
4141                         e2fsck_pass1_fix_unlock(ctx);
4142                         if (pctx->errcode)
4143                                 return;
4144                         failed_csum = 0;
4145                 }
4146 #ifdef CONFIG_DEVELOPER_FEATURES
4147                 if (try_repairs && !is_dir && problem == 0 &&
4148                     (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
4149                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
4150                     fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
4151                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
4152                         pb->inode_modified = 1;
4153                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
4154                                                               &extent);
4155                         if (pctx->errcode)
4156                                 return;
4157                         failed_csum = 0;
4158                 }
4159 #endif
4160                 if (try_repairs && problem) {
4161 report_problem:
4162                         if (fix_problem(ctx, problem, pctx)) {
4163                                 if (ctx->invalid_bitmaps) {
4164                                         /*
4165                                          * If fsck knows the bitmaps are bad,
4166                                          * skip to the next extent and
4167                                          * try to clear this extent again
4168                                          * after fixing the bitmaps, by
4169                                          * restarting fsck.
4170                                          */
4171                                         pctx->errcode = ext2fs_extent_get(
4172                                                           ehandle,
4173                                                           EXT2_EXTENT_NEXT_SIB,
4174                                                           &extent);
4175                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4176                                         if (pctx->errcode ==
4177                                                     EXT2_ET_NO_CURRENT_NODE) {
4178                                                 pctx->errcode = 0;
4179                                                 break;
4180                                         }
4181                                         continue;
4182                                 }
4183                                 e2fsck_pass1_fix_lock(ctx);
4184                                 e2fsck_read_bitmaps(ctx);
4185                                 pb->inode_modified = 1;
4186                                 pctx->errcode =
4187                                         ext2fs_extent_delete(ehandle, 0);
4188                                 e2fsck_pass1_fix_unlock(ctx);
4189                                 if (pctx->errcode) {
4190                                         pctx->str = "ext2fs_extent_delete";
4191                                         return;
4192                                 }
4193                                 e2fsck_pass1_fix_lock(ctx);
4194                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
4195                                 e2fsck_pass1_fix_unlock(ctx);
4196                                 if (pctx->errcode &&
4197                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
4198                                         pctx->str = "ext2fs_extent_fix_parents";
4199                                         return;
4200                                 }
4201                                 pctx->errcode = ext2fs_extent_get(ehandle,
4202                                                                   EXT2_EXTENT_CURRENT,
4203                                                                   &extent);
4204                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
4205                                         pctx->errcode = 0;
4206                                         break;
4207                                 }
4208                                 failed_csum = 0;
4209                                 continue;
4210                         }
4211                         goto next;
4212                 }
4213
4214                 if (!is_leaf) {
4215                         blk64_t lblk = extent.e_lblk;
4216                         int next_try_repairs = 1;
4217
4218                         blk = extent.e_pblk;
4219
4220                         /*
4221                          * If this lower extent block collides with critical
4222                          * metadata, don't try to repair the damage.  Pass 1b
4223                          * will reallocate the block; then we can try again.
4224                          */
4225                         if (pb->ino != EXT2_RESIZE_INO &&
4226                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
4227                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4228                                                       extent.e_pblk)) {
4229                                 next_try_repairs = 0;
4230                                 pctx->blk = blk;
4231                                 fix_problem(ctx,
4232                                             PR_1_CRITICAL_METADATA_COLLISION,
4233                                             pctx);
4234                                 if ((ctx->options & E2F_OPT_NO) == 0)
4235                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4236                         }
4237                         pctx->errcode = ext2fs_extent_get(ehandle,
4238                                                   EXT2_EXTENT_DOWN, &extent);
4239                         if (pctx->errcode &&
4240                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
4241                                 pctx->str = "EXT2_EXTENT_DOWN";
4242                                 problem = PR_1_EXTENT_HEADER_INVALID;
4243                                 if (!next_try_repairs)
4244                                         return;
4245                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
4246                                         goto report_problem;
4247                                 return;
4248                         }
4249                         /* The next extent should match this index's logical start */
4250                         if (extent.e_lblk != lblk) {
4251                                 struct ext2_extent_info e_info;
4252
4253                                 pctx->errcode = ext2fs_extent_get_info(ehandle,
4254                                                                        &e_info);
4255                                 if (pctx->errcode) {
4256                                         pctx->str = "ext2fs_extent_get_info";
4257                                         return;
4258                                 }
4259                                 pctx->blk = lblk;
4260                                 pctx->blk2 = extent.e_lblk;
4261                                 pctx->num = e_info.curr_level - 1;
4262                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
4263                                 if (fix_problem(ctx, problem, pctx)) {
4264                                         e2fsck_pass1_fix_lock(ctx);
4265                                         pb->inode_modified = 1;
4266                                         pctx->errcode =
4267                                                 ext2fs_extent_fix_parents(ehandle);
4268                                         e2fsck_pass1_fix_unlock(ctx);
4269                                         if (pctx->errcode) {
4270                                                 pctx->str = "ext2fs_extent_fix_parents";
4271                                                 return;
4272                                         }
4273                                 }
4274                         }
4275                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
4276                                          last_lblk, eof_block, ehandle,
4277                                          next_try_repairs);
4278                         if (pctx->errcode)
4279                                 return;
4280                         pctx->errcode = ext2fs_extent_get(ehandle,
4281                                                   EXT2_EXTENT_UP, &extent);
4282                         if (pctx->errcode) {
4283                                 pctx->str = "EXT2_EXTENT_UP";
4284                                 return;
4285                         }
4286                         mark_block_used(ctx, blk);
4287                         pb->num_blocks++;
4288                         goto next;
4289                 }
4290
4291                 if ((pb->previous_block != 0) &&
4292                     (pb->previous_block+1 != extent.e_pblk)) {
4293                         if (ctx->options & E2F_OPT_FRAGCHECK) {
4294                                 char type = '?';
4295
4296                                 if (pb->is_dir)
4297                                         type = 'd';
4298                                 else if (pb->is_reg)
4299                                         type = 'f';
4300
4301                                 printf(("%6lu(%c): expecting %6lu "
4302                                         "actual extent "
4303                                         "phys %6lu log %lu len %lu\n"),
4304                                        (unsigned long) pctx->ino, type,
4305                                        (unsigned long) pb->previous_block+1,
4306                                        (unsigned long) extent.e_pblk,
4307                                        (unsigned long) extent.e_lblk,
4308                                        (unsigned long) extent.e_len);
4309                         }
4310                         pb->fragmented = 1;
4311                 }
4312                 /*
4313                  * If we notice a gap in the logical block mappings of an
4314                  * extent-mapped directory, offer to close the hole by
4315                  * moving the logical block down, otherwise we'll go mad in
4316                  * pass 3 allocating empty directory blocks to fill the hole.
4317                  */
4318                 if (try_repairs && is_dir &&
4319                     pb->last_block + 1 < extent.e_lblk) {
4320                         blk64_t new_lblk;
4321
4322                         new_lblk = pb->last_block + 1;
4323                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
4324                                 new_lblk = ((new_lblk +
4325                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
4326                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
4327                                            (extent.e_pblk &
4328                                             EXT2FS_CLUSTER_MASK(ctx->fs));
4329                         pctx->blk = extent.e_lblk;
4330                         pctx->blk2 = new_lblk;
4331                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
4332                                 e2fsck_pass1_fix_lock(ctx);
4333                                 extent.e_lblk = new_lblk;
4334                                 pb->inode_modified = 1;
4335                                 pctx->errcode = ext2fs_extent_replace(ehandle,
4336                                                                 0, &extent);
4337                                 e2fsck_pass1_fix_unlock(ctx);
4338                                 if (pctx->errcode) {
4339                                         pctx->errcode = 0;
4340                                         goto alloc_later;
4341                                 }
4342                                 e2fsck_pass1_fix_lock(ctx);
4343                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
4344                                 e2fsck_pass1_fix_unlock(ctx);
4345                                 if (pctx->errcode)
4346                                         goto failed_add_dir_block;
4347                                 pctx->errcode = ext2fs_extent_goto(ehandle,
4348                                                                 extent.e_lblk);
4349                                 if (pctx->errcode)
4350                                         goto failed_add_dir_block;
4351                                 last_lblk = extent.e_lblk + extent.e_len - 1;
4352                                 failed_csum = 0;
4353                         }
4354                 }
4355 alloc_later:
4356                 if (is_dir) {
4357                         while (++pb->last_db_block <
4358                                (e2_blkcnt_t) extent.e_lblk) {
4359                                 pctx->errcode = ext2fs_add_dir_block2(
4360                                                         ctx->fs->dblist,
4361                                                         pb->ino, 0,
4362                                                         pb->last_db_block);
4363                                 if (pctx->errcode) {
4364                                         pctx->blk = 0;
4365                                         pctx->num = pb->last_db_block;
4366                                         goto failed_add_dir_block;
4367                                 }
4368                         }
4369
4370                         for (i = 0; i < extent.e_len; i++) {
4371                                 pctx->errcode = ext2fs_add_dir_block2(
4372                                                         ctx->fs->dblist,
4373                                                         pctx->ino,
4374                                                         extent.e_pblk + i,
4375                                                         extent.e_lblk + i);
4376                                 if (pctx->errcode) {
4377                                         pctx->blk = extent.e_pblk + i;
4378                                         pctx->num = extent.e_lblk + i;
4379                                 failed_add_dir_block:
4380                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4381                                         /* Should never get here */
4382                                         ctx->flags |= E2F_FLAG_ABORT;
4383                                         return;
4384                                 }
4385                         }
4386                         if (extent.e_len > 0)
4387                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
4388                 }
4389                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
4390                                               pb->last_block,
4391                                               extent.e_pblk,
4392                                               extent.e_lblk)) {
4393                         for (i = 0; i < extent.e_len; i++) {
4394                                 pctx->blk = extent.e_lblk + i;
4395                                 pctx->blk2 = extent.e_pblk + i;
4396                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4397                                 mark_block_used(ctx, extent.e_pblk + i);
4398                                 mark_block_used(ctx, extent.e_pblk + i);
4399                         }
4400                 }
4401
4402                 /*
4403                  * Check whether first cluster got marked in previous iteration.
4404                  */
4405                 if (ctx->fs->cluster_ratio_bits &&
4406                     pb->previous_block &&
4407                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
4408                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
4409                         /* Set blk to the beginning of next cluster. */
4410                         blk = EXT2FS_C2B(
4411                                 ctx->fs,
4412                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
4413                 else
4414                         /* Set blk to the beginning of current cluster. */
4415                         blk = EXT2FS_C2B(ctx->fs,
4416                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
4417
4418                 if (blk < extent.e_pblk + extent.e_len) {
4419                         mark_blocks_used(ctx, blk,
4420                                          extent.e_pblk + extent.e_len - blk);
4421                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
4422                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
4423                         pb->num_blocks += n;
4424                 }
4425                 pb->last_block = extent.e_lblk + extent.e_len - 1;
4426                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
4427                 start_block = pb->last_block = last_lblk;
4428                 if (is_leaf && !is_dir &&
4429                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
4430                         pb->last_init_lblock = last_lblk;
4431         next:
4432                 pctx->errcode = ext2fs_extent_get(ehandle,
4433                                                   EXT2_EXTENT_NEXT_SIB,
4434                                                   &extent);
4435         }
4436
4437         /* Failed csum but passes checks?  Ask to fix checksum. */
4438         if (failed_csum &&
4439             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
4440                 e2fsck_pass1_fix_lock(ctx);
4441                 pb->inode_modified = 1;
4442                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
4443                 e2fsck_pass1_fix_unlock(ctx);
4444                 if (pctx->errcode)
4445                         return;
4446         }
4447
4448         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
4449                 pctx->errcode = 0;
4450 }
4451
4452 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
4453                                  struct process_block_struct *pb)
4454 {
4455         struct ext2_extent_info info;
4456         struct ext2_inode       *inode = pctx->inode;
4457         ext2_extent_handle_t    ehandle;
4458         ext2_filsys             fs = ctx->fs;
4459         ext2_ino_t              ino = pctx->ino;
4460         errcode_t               retval;
4461         blk64_t                 eof_lblk;
4462         struct ext3_extent_header       *eh;
4463
4464         /* Check for a proper extent header... */
4465         eh = (struct ext3_extent_header *) &inode->i_block[0];
4466         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
4467         if (retval) {
4468                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
4469                         e2fsck_clear_inode(ctx, ino, inode, 0,
4470                                            "check_blocks_extents");
4471                 pctx->errcode = 0;
4472                 return;
4473         }
4474
4475         /* ...since this function doesn't fail if i_block is zeroed. */
4476         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
4477         if (pctx->errcode) {
4478                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
4479                         e2fsck_clear_inode(ctx, ino, inode, 0,
4480                                            "check_blocks_extents");
4481                 pctx->errcode = 0;
4482                 return;
4483         }
4484
4485         retval = ext2fs_extent_get_info(ehandle, &info);
4486         if (retval == 0) {
4487                 int max_depth = info.max_depth;
4488
4489                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
4490                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
4491                 ctx->extent_depth_count[max_depth]++;
4492         }
4493
4494         /* Check maximum extent depth */
4495         pctx->blk = info.max_depth;
4496         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
4497         if (pctx->blk2 < pctx->blk &&
4498             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
4499                 pb->eti.force_rebuild = 1;
4500
4501         /* Can we collect extent tree level stats? */
4502         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
4503         if (pctx->blk2 > pctx->blk)
4504                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
4505         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
4506         pb->eti.ino = pb->ino;
4507
4508         pb->next_lblock = 0;
4509
4510         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
4511                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
4512         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
4513         if (pctx->errcode &&
4514             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
4515                 pb->num_blocks = 0;
4516                 inode->i_blocks = 0;
4517                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4518                                    "check_blocks_extents");
4519                 pctx->errcode = 0;
4520         }
4521         ext2fs_extent_free(ehandle);
4522
4523         /* Rebuild unless it's a dir and we're rehashing it */
4524         if (LINUX_S_ISDIR(inode->i_mode) &&
4525             e2fsck_dir_will_be_rehashed(ctx, ino))
4526                 return;
4527
4528         if (ctx->options & E2F_OPT_CONVERT_BMAP)
4529                 e2fsck_rebuild_extents_later(ctx, ino);
4530         else
4531                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
4532 }
4533
4534 /*
4535  * In fact we don't need to check blocks for an inode with inline data
4536  * because this inode doesn't have any blocks.  In this function all
4537  * we need to do is add this inode into dblist when it is a directory.
4538  */
4539 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
4540                                      struct process_block_struct *pb)
4541 {
4542         int     flags;
4543         size_t  inline_data_size = 0;
4544
4545         if (!pb->is_dir) {
4546                 pctx->errcode = 0;
4547                 return;
4548         }
4549
4550         /* Process the dirents in i_block[] as the "first" block. */
4551         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
4552         if (pctx->errcode)
4553                 goto err;
4554
4555         /* Process the dirents in the EA as a "second" block. */
4556         flags = ctx->fs->flags;
4557         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4558         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
4559                                                 &inline_data_size);
4560         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4561                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4562         if (pctx->errcode) {
4563                 pctx->errcode = 0;
4564                 return;
4565         }
4566
4567         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
4568                 return;
4569
4570         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
4571         if (pctx->errcode)
4572                 goto err;
4573
4574         return;
4575 err:
4576         pctx->blk = 0;
4577         pctx->num = 0;
4578         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4579         ctx->flags |= E2F_FLAG_ABORT;
4580 }
4581
4582 /*
4583  * This subroutine is called on each inode to account for all of the
4584  * blocks used by that inode.
4585  */
4586 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
4587                          char *block_buf, const struct ea_quota *ea_ibody_quota)
4588 {
4589         ext2_filsys fs = ctx->fs;
4590         struct process_block_struct pb;
4591         ext2_ino_t      ino = pctx->ino;
4592         struct ext2_inode *inode = pctx->inode;
4593         unsigned        bad_size = 0;
4594         int             dirty_inode = 0;
4595         int             extent_fs;
4596         int             inlinedata_fs;
4597         __u64           size;
4598         struct ea_quota ea_block_quota;
4599
4600         pb.ino = ino;
4601         pb.num_blocks = EXT2FS_B2C(ctx->fs,
4602                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
4603         pb.last_block = ~0;
4604         pb.last_init_lblock = -1;
4605         pb.last_db_block = -1;
4606         pb.num_illegal_blocks = 0;
4607         pb.suppress = 0; pb.clear = 0;
4608         pb.fragmented = 0;
4609         pb.compressed = 0;
4610         pb.previous_block = 0;
4611         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
4612         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
4613         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
4614         pb.inode = inode;
4615         pb.pctx = pctx;
4616         pb.ctx = ctx;
4617         pb.inode_modified = 0;
4618         pb.eti.force_rebuild = 0;
4619         pctx->ino = ino;
4620         pctx->errcode = 0;
4621
4622         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
4623         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
4624
4625         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
4626                 if (e2fsck_should_abort(ctx))
4627                         goto out;
4628                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
4629         }
4630
4631         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
4632                 check_blocks_inline_data(ctx, pctx, &pb);
4633         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
4634                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
4635                         check_blocks_extents(ctx, pctx, &pb);
4636                 else {
4637                         int flags;
4638                         /*
4639                          * If we've modified the inode, write it out before
4640                          * iterate() tries to use it.
4641                          */
4642                         if (dirty_inode) {
4643                                 e2fsck_write_inode(ctx, ino, inode,
4644                                                    "check_blocks");
4645                                 dirty_inode = 0;
4646                         }
4647                         flags = fs->flags;
4648                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4649                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
4650                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
4651                                                 block_buf, process_block, &pb);
4652                         /*
4653                          * We do not have uninitialized extents in non extent
4654                          * files.
4655                          */
4656                         pb.last_init_lblock = pb.last_block;
4657                         /*
4658                          * If iterate() changed a block mapping, we have to
4659                          * re-read the inode.  If we decide to clear the
4660                          * inode after clearing some stuff, we'll re-write the
4661                          * bad mappings into the inode!
4662                          */
4663                         if (pb.inode_modified)
4664                                 e2fsck_read_inode(ctx, ino, inode,
4665                                                   "check_blocks");
4666                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4667                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4668
4669                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
4670 #ifdef DEBUG
4671                                 printf("bmap rebuild ino=%d\n", ino);
4672 #endif
4673                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
4674                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
4675                                         e2fsck_rebuild_extents_later(ctx, ino);
4676                         }
4677                 }
4678         }
4679         end_problem_latch(ctx, PR_LATCH_BLOCK);
4680         end_problem_latch(ctx, PR_LATCH_TOOBIG);
4681         if (e2fsck_should_abort(ctx))
4682                 goto out;
4683         if (pctx->errcode)
4684                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
4685
4686         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
4687                 if (LINUX_S_ISDIR(inode->i_mode))
4688                         ctx->fs_fragmented_dir++;
4689                 else
4690                         ctx->fs_fragmented++;
4691         }
4692
4693         if (pb.clear) {
4694                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4695                                    "check_blocks");
4696                 return;
4697         }
4698
4699         if (inode->i_flags & EXT2_INDEX_FL) {
4700                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
4701                         inode->i_flags &= ~EXT2_INDEX_FL;
4702                         dirty_inode++;
4703                 } else {
4704                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
4705                 }
4706         }
4707
4708         if (!pb.num_blocks && pb.is_dir &&
4709             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
4710                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
4711                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
4712                         ctx->fs_directory_count--;
4713                         return;
4714                 }
4715         }
4716
4717         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
4718             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
4719             !(inode->i_flags & EXT4_EA_INODE_FL)) {
4720                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
4721                                ino,
4722                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
4723                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
4724                                   ino, (ea_ibody_quota ?
4725                                         ea_ibody_quota->inodes : 0) +
4726                                                 ea_block_quota.inodes + 1);
4727         }
4728
4729         if (!ext2fs_has_feature_huge_file(fs->super) ||
4730             !(inode->i_flags & EXT4_HUGE_FILE_FL))
4731                 pb.num_blocks *= (fs->blocksize / 512);
4732         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
4733 #if 0
4734         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
4735                ino, inode->i_size, (unsigned long long) pb.last_block,
4736                (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
4737                (unsigned long long) pb.num_blocks);
4738 #endif
4739         size = EXT2_I_SIZE(inode);
4740         if (pb.is_dir) {
4741                 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
4742                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
4743                         int flags;
4744                         size_t sz = 0;
4745                         errcode_t err;
4746
4747                         flags = ctx->fs->flags;
4748                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4749                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
4750                                                       &sz);
4751                         ctx->fs->flags = (flags &
4752                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4753                                          (ctx->fs->flags &
4754                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4755                         if (err || sz != size) {
4756                                 bad_size = 7;
4757                                 pctx->num = sz;
4758                         }
4759                 } else if (size & (fs->blocksize - 1))
4760                         bad_size = 5;
4761                 else if (nblock > (pb.last_block + 1))
4762                         bad_size = 1;
4763                 else if (nblock < (pb.last_block + 1)) {
4764                         if (((pb.last_block + 1) - nblock) >
4765                             fs->super->s_prealloc_dir_blocks)
4766                                 bad_size = 2;
4767                 }
4768         } else {
4769                 if ((pb.last_init_lblock >= 0) &&
4770                     /* Do not allow initialized allocated blocks past i_size*/
4771                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
4772                     !(inode->i_flags & EXT4_VERITY_FL))
4773                         bad_size = 3;
4774                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4775                          size > ext2_max_sizes[fs->super->s_log_block_size])
4776                         /* too big for a direct/indirect-mapped file */
4777                         bad_size = 4;
4778                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4779                          size >
4780                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
4781                         /* too big for an extent-based file - 32bit ee_block */
4782                         bad_size = 6;
4783         }
4784         /* i_size for symlinks is checked elsewhere */
4785         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
4786                 /* Did inline_data set pctx->num earlier? */
4787                 if (bad_size != 7)
4788                         pctx->num = (pb.last_block + 1) * fs->blocksize;
4789                 pctx->group = bad_size;
4790                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
4791                         ext2fs_inode_size_set(fs, inode, pctx->num);
4792                         if (EXT2_I_SIZE(inode) == 0 &&
4793                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
4794                                 memset(inode->i_block, 0,
4795                                        sizeof(inode->i_block));
4796                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
4797                         }
4798                         dirty_inode++;
4799                 }
4800                 pctx->num = 0;
4801         }
4802         if (LINUX_S_ISREG(inode->i_mode) &&
4803             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4804                 ctx->large_files++;
4805         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
4806             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
4807              (ext2fs_has_feature_huge_file(fs->super) &&
4808               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
4809               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
4810                 pctx->num = pb.num_blocks;
4811                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
4812                         inode->i_blocks = pb.num_blocks;
4813                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
4814                         dirty_inode++;
4815                 }
4816                 pctx->num = 0;
4817         }
4818
4819         /*
4820          * The kernel gets mad if we ask it to allocate bigalloc clusters to
4821          * a block mapped file, so rebuild it as an extent file.  We can skip
4822          * symlinks because they're never rewritten.
4823          */
4824         if (ext2fs_has_feature_bigalloc(fs->super) &&
4825             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
4826             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
4827             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
4828             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
4829             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
4830                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
4831                 if (pctx->errcode)
4832                         goto out;
4833         }
4834
4835         if (ctx->dirs_to_hash && pb.is_dir &&
4836             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
4837             !(inode->i_flags & EXT2_INDEX_FL) &&
4838             ((inode->i_size / fs->blocksize) >= 3))
4839                 e2fsck_rehash_dir_later(ctx, ino);
4840
4841 out:
4842         if (dirty_inode)
4843                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
4844 }
4845
4846 #if 0
4847 /*
4848  * Helper function called by process block when an illegal block is
4849  * found.  It returns a description about why the block is illegal
4850  */
4851 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
4852 {
4853         blk64_t super;
4854         int     i;
4855         static char     problem[80];
4856
4857         super = fs->super->s_first_data_block;
4858         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
4859         if (block < super) {
4860                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
4861                 return(problem);
4862         } else if (block >= ext2fs_blocks_count(fs->super)) {
4863                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
4864                 return(problem);
4865         }
4866         for (i = 0; i < fs->group_desc_count; i++) {
4867                 if (block == super) {
4868                         sprintf(problem, "is the superblock in group %d", i);
4869                         break;
4870                 }
4871                 if (block > super &&
4872                     block <= (super + fs->desc_blocks)) {
4873                         sprintf(problem, "is in the group descriptors "
4874                                 "of group %d", i);
4875                         break;
4876                 }
4877                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
4878                         sprintf(problem, "is the block bitmap of group %d", i);
4879                         break;
4880                 }
4881                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
4882                         sprintf(problem, "is the inode bitmap of group %d", i);
4883                         break;
4884                 }
4885                 if (block >= ext2fs_inode_table_loc(fs, i) &&
4886                     (block < ext2fs_inode_table_loc(fs, i)
4887                      + fs->inode_blocks_per_group)) {
4888                         sprintf(problem, "is in the inode table of group %d",
4889                                 i);
4890                         break;
4891                 }
4892                 super += fs->super->s_blocks_per_group;
4893         }
4894         return(problem);
4895 }
4896 #endif
4897
4898 /*
4899  * This is a helper function for check_blocks().
4900  */
4901 static int process_block(ext2_filsys fs,
4902                   blk64_t       *block_nr,
4903                   e2_blkcnt_t blockcnt,
4904                   blk64_t ref_block EXT2FS_ATTR((unused)),
4905                   int ref_offset EXT2FS_ATTR((unused)),
4906                   void *priv_data)
4907 {
4908         struct process_block_struct *p;
4909         struct problem_context *pctx;
4910         blk64_t blk = *block_nr;
4911         int     ret_code = 0;
4912         problem_t       problem = 0;
4913         e2fsck_t        ctx;
4914
4915         p = (struct process_block_struct *) priv_data;
4916         pctx = p->pctx;
4917         ctx = p->ctx;
4918
4919         /*
4920          * For a directory, add logical block zero for processing even if it's
4921          * not mapped or we'll be perennially stuck with broken "." and ".."
4922          * entries.
4923          */
4924         if (p->is_dir && blockcnt == 0 && blk == 0) {
4925                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
4926                 if (pctx->errcode) {
4927                         pctx->blk = blk;
4928                         pctx->num = blockcnt;
4929                         goto failed_add_dir_block;
4930                 }
4931                 p->last_db_block++;
4932         }
4933
4934         if (blk == 0)
4935                 return 0;
4936
4937 #if 0
4938         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
4939                blockcnt);
4940 #endif
4941
4942         /*
4943          * Simplistic fragmentation check.  We merely require that the
4944          * file be contiguous.  (Which can never be true for really
4945          * big files that are greater than a block group.)
4946          */
4947         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
4948                 if (p->previous_block+1 != blk) {
4949                         if (ctx->options & E2F_OPT_FRAGCHECK) {
4950                                 char type = '?';
4951
4952                                 if (p->is_dir)
4953                                         type = 'd';
4954                                 else if (p->is_reg)
4955                                         type = 'f';
4956
4957                                 printf(_("%6lu(%c): expecting %6lu "
4958                                          "got phys %6lu (blkcnt %lld)\n"),
4959                                        (unsigned long) pctx->ino, type,
4960                                        (unsigned long) p->previous_block+1,
4961                                        (unsigned long) blk,
4962                                        (long long) blockcnt);
4963                         }
4964                         p->fragmented = 1;
4965                 }
4966         }
4967
4968         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
4969             !pctx->inode->i_size_high &&
4970             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
4971                 problem = PR_1_TOOBIG_DIR;
4972         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
4973                 problem = PR_1_TOOBIG_DIR;
4974         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
4975                 problem = PR_1_TOOBIG_REG;
4976         if (!p->is_dir && !p->is_reg && blockcnt > 0)
4977                 problem = PR_1_TOOBIG_SYMLINK;
4978
4979         if (blk < fs->super->s_first_data_block ||
4980             blk >= ext2fs_blocks_count(fs->super))
4981                 problem = PR_1_ILLEGAL_BLOCK_NUM;
4982
4983         /*
4984          * If this IND/DIND/TIND block is squatting atop some critical metadata
4985          * (group descriptors, superblock, bitmap, inode table), any write to
4986          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
4987          * fix that and restart fsck.
4988          */
4989         if (blockcnt < 0 &&
4990             p->ino != EXT2_RESIZE_INO &&
4991             blk < ctx->fs->super->s_blocks_count &&
4992             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
4993                 pctx->blk = blk;
4994                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
4995                 if ((ctx->options & E2F_OPT_NO) == 0)
4996                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4997         }
4998
4999         if (problem) {
5000                 p->num_illegal_blocks++;
5001                 /*
5002                  * A bit of subterfuge here -- we're trying to fix a block
5003                  * mapping, but the IND/DIND/TIND block could have collided
5004                  * with some critical metadata.  So, fix the in-core mapping so
5005                  * iterate won't go insane, but return 0 instead of
5006                  * BLOCK_CHANGED so that it won't write the remapping out to
5007                  * our multiply linked block.
5008                  *
5009                  * Even if we previously determined that an *IND block
5010                  * conflicts with critical metadata, we must still try to
5011                  * iterate the *IND block as if it is an *IND block to find and
5012                  * mark the blocks it points to.  Better to be overly cautious
5013                  * with the used_blocks map so that we don't move the *IND
5014                  * block to a block that's really in use!
5015                  */
5016                 if (p->ino != EXT2_RESIZE_INO &&
5017                     ref_block != 0 &&
5018                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
5019                                               ref_block)) {
5020                         *block_nr = 0;
5021                         return 0;
5022                 }
5023                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
5024                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
5025                                 p->clear = 1;
5026                                 return BLOCK_ABORT;
5027                         }
5028                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
5029                                 p->suppress = 1;
5030                                 set_latch_flags(PR_LATCH_BLOCK,
5031                                                 PRL_SUPPRESS, 0);
5032                         }
5033                 }
5034                 pctx->blk = blk;
5035                 pctx->blkcount = blockcnt;
5036                 if (fix_problem(ctx, problem, pctx)) {
5037                         blk = *block_nr = 0;
5038                         ret_code = BLOCK_CHANGED;
5039                         p->inode_modified = 1;
5040                         /*
5041                          * If the directory block is too big and is beyond the
5042                          * end of the FS, don't bother trying to add it for
5043                          * processing -- the kernel would never have created a
5044                          * directory this large, and we risk an ENOMEM abort.
5045                          * In any case, the toobig handler for extent-based
5046                          * directories also doesn't feed toobig blocks to
5047                          * pass 2.
5048                          */
5049                         if (problem == PR_1_TOOBIG_DIR)
5050                                 return ret_code;
5051                         goto mark_dir;
5052                 } else
5053                         return 0;
5054         }
5055
5056         if (p->ino == EXT2_RESIZE_INO) {
5057                 /*
5058                  * The resize inode has already be sanity checked
5059                  * during pass #0 (the superblock checks).  All we
5060                  * have to do is mark the double indirect block as
5061                  * being in use; all of the other blocks are handled
5062                  * by mark_table_blocks()).
5063                  */
5064                 if (blockcnt == BLOCK_COUNT_DIND)
5065                         mark_block_used(ctx, blk);
5066                 p->num_blocks++;
5067         } else if (!(ctx->fs->cluster_ratio_bits &&
5068                      p->previous_block &&
5069                      (EXT2FS_B2C(ctx->fs, blk) ==
5070                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
5071                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
5072                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
5073                 mark_block_used(ctx, blk);
5074                 p->num_blocks++;
5075         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
5076                                              p->last_block, blk, blockcnt)) {
5077                 pctx->blk = blockcnt;
5078                 pctx->blk2 = blk;
5079                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
5080                 mark_block_used(ctx, blk);
5081                 mark_block_used(ctx, blk);
5082         }
5083         if (blockcnt >= 0)
5084                 p->last_block = blockcnt;
5085         p->previous_block = blk;
5086 mark_dir:
5087         if (p->is_dir && (blockcnt >= 0)) {
5088                 while (++p->last_db_block < blockcnt) {
5089                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
5090                                                               p->ino, 0,
5091                                                               p->last_db_block);
5092                         if (pctx->errcode) {
5093                                 pctx->blk = 0;
5094                                 pctx->num = p->last_db_block;
5095                                 goto failed_add_dir_block;
5096                         }
5097                 }
5098                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
5099                                                       blk, blockcnt);
5100                 if (pctx->errcode) {
5101                         pctx->blk = blk;
5102                         pctx->num = blockcnt;
5103                 failed_add_dir_block:
5104                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
5105                         /* Should never get here */
5106                         ctx->flags |= E2F_FLAG_ABORT;
5107                         return BLOCK_ABORT;
5108                 }
5109         }
5110         return ret_code;
5111 }
5112
5113 static int process_bad_block(ext2_filsys fs,
5114                       blk64_t *block_nr,
5115                       e2_blkcnt_t blockcnt,
5116                       blk64_t ref_block EXT2FS_ATTR((unused)),
5117                       int ref_offset EXT2FS_ATTR((unused)),
5118                       void *priv_data)
5119 {
5120         struct process_block_struct *p;
5121         blk64_t         blk = *block_nr;
5122         blk64_t         first_block;
5123         dgrp_t          i;
5124         struct problem_context *pctx;
5125         e2fsck_t        ctx;
5126
5127         if (!blk)
5128                 return 0;
5129
5130         p = (struct process_block_struct *) priv_data;
5131         ctx = p->ctx;
5132         pctx = p->pctx;
5133
5134         pctx->ino = EXT2_BAD_INO;
5135         pctx->blk = blk;
5136         pctx->blkcount = blockcnt;
5137
5138         if ((blk < fs->super->s_first_data_block) ||
5139             (blk >= ext2fs_blocks_count(fs->super))) {
5140                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
5141                         *block_nr = 0;
5142                         return BLOCK_CHANGED;
5143                 } else
5144                         return 0;
5145         }
5146
5147         if (blockcnt < 0) {
5148                 e2fsck_pass1_block_map_lock(ctx);
5149                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
5150                         p->bbcheck = 1;
5151                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
5152                                 *block_nr = 0;
5153                                 e2fsck_pass1_block_map_unlock(ctx);
5154                                 return BLOCK_CHANGED;
5155                         }
5156                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5157                                                     blk)) {
5158                         p->bbcheck = 1;
5159                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
5160                                         pctx)) {
5161                                 *block_nr = 0;
5162                                 e2fsck_pass1_block_map_unlock(ctx);
5163                                 return BLOCK_CHANGED;
5164                         }
5165                         if (e2fsck_should_abort(ctx)) {
5166                                 e2fsck_pass1_block_map_unlock(ctx);
5167                                 return BLOCK_ABORT;
5168                         }
5169                 } else {
5170                         mark_block_used_unlocked(ctx, blk);
5171                 }
5172                 e2fsck_pass1_block_map_unlock(ctx);
5173                 return 0;
5174         }
5175 #if 0
5176         printf ("DEBUG: Marking %u as bad.\n", blk);
5177 #endif
5178         ctx->fs_badblocks_count++;
5179         /*
5180          * If the block is not used, then mark it as used and return.
5181          * If it is already marked as found, this must mean that
5182          * there's an overlap between the filesystem table blocks
5183          * (bitmaps and inode table) and the bad block list.
5184          */
5185         e2fsck_pass1_block_map_lock(ctx);
5186         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
5187                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
5188                 e2fsck_pass1_block_map_unlock(ctx);
5189                 return 0;
5190         }
5191         e2fsck_pass1_block_map_unlock(ctx);
5192         /*
5193          * Try to find the where the filesystem block was used...
5194          */
5195         first_block = fs->super->s_first_data_block;
5196
5197         for (i = 0; i < fs->group_desc_count; i++ ) {
5198                 pctx->group = i;
5199                 pctx->blk = blk;
5200                 if (!ext2fs_bg_has_super(fs, i))
5201                         goto skip_super;
5202                 if (blk == first_block) {
5203                         if (i == 0) {
5204                                 if (fix_problem(ctx,
5205                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
5206                                                 pctx)) {
5207                                         *block_nr = 0;
5208                                         return BLOCK_CHANGED;
5209                                 }
5210                                 return 0;
5211                         }
5212                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
5213                         return 0;
5214                 }
5215                 if ((blk > first_block) &&
5216                     (blk <= first_block + fs->desc_blocks)) {
5217                         if (i == 0) {
5218                                 pctx->blk = *block_nr;
5219                                 if (fix_problem(ctx,
5220                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
5221                                         *block_nr = 0;
5222                                         return BLOCK_CHANGED;
5223                                 }
5224                                 return 0;
5225                         }
5226                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
5227                         return 0;
5228                 }
5229         skip_super:
5230                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
5231                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
5232                                 ctx->invalid_block_bitmap_flag[i]++;
5233                                 ctx->invalid_bitmaps++;
5234                         }
5235                         return 0;
5236                 }
5237                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
5238                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
5239                                 ctx->invalid_inode_bitmap_flag[i]++;
5240                                 ctx->invalid_bitmaps++;
5241                         }
5242                         return 0;
5243                 }
5244                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
5245                     (blk < (ext2fs_inode_table_loc(fs, i) +
5246                             fs->inode_blocks_per_group))) {
5247                         /*
5248                          * If there are bad blocks in the inode table,
5249                          * the inode scan code will try to do
5250                          * something reasonable automatically.
5251                          */
5252                         return 0;
5253                 }
5254                 first_block += fs->super->s_blocks_per_group;
5255         }
5256         /*
5257          * If we've gotten to this point, then the only
5258          * possibility is that the bad block inode meta data
5259          * is using a bad block.
5260          */
5261         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
5262             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
5263             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
5264                 p->bbcheck = 1;
5265                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
5266                         *block_nr = 0;
5267                         return BLOCK_CHANGED;
5268                 }
5269                 if (e2fsck_should_abort(ctx))
5270                         return BLOCK_ABORT;
5271                 return 0;
5272         }
5273
5274         pctx->group = -1;
5275
5276         /* Warn user that the block wasn't claimed */
5277         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
5278
5279         return 0;
5280 }
5281
5282 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
5283                             const char *name, int num, blk64_t *new_block)
5284 {
5285         ext2_filsys fs = ctx->fs;
5286         dgrp_t          last_grp;
5287         blk64_t         old_block = *new_block;
5288         blk64_t         last_block;
5289         dgrp_t          flexbg;
5290         unsigned        flexbg_size;
5291         int             i, is_flexbg;
5292         char            *buf;
5293         struct problem_context  pctx;
5294
5295         clear_problem_context(&pctx);
5296
5297         pctx.group = group;
5298         pctx.blk = old_block;
5299         pctx.str = name;
5300
5301         /*
5302          * For flex_bg filesystems, first try to allocate the metadata
5303          * within the flex_bg, and if that fails then try finding the
5304          * space anywhere in the filesystem.
5305          */
5306         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
5307         if (is_flexbg) {
5308                 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
5309                 flexbg = group / flexbg_size;
5310                 first_block = ext2fs_group_first_block2(fs,
5311                                                         flexbg_size * flexbg);
5312                 last_grp = group | (flexbg_size - 1);
5313                 if (last_grp >= fs->group_desc_count)
5314                         last_grp = fs->group_desc_count - 1;
5315                 last_block = ext2fs_group_last_block2(fs, last_grp);
5316         } else
5317                 last_block = ext2fs_group_last_block2(fs, group);
5318         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
5319                                                num, ctx->block_found_map,
5320                                                new_block);
5321         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
5322                 pctx.errcode = ext2fs_get_free_blocks2(fs,
5323                                 fs->super->s_first_data_block,
5324                                 ext2fs_blocks_count(fs->super),
5325                                 num, ctx->block_found_map, new_block);
5326         if (pctx.errcode) {
5327                 pctx.num = num;
5328                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
5329                 ext2fs_unmark_valid(fs);
5330                 ctx->flags |= E2F_FLAG_ABORT;
5331                 return;
5332         }
5333         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
5334         if (pctx.errcode) {
5335                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
5336                 ext2fs_unmark_valid(fs);
5337                 ctx->flags |= E2F_FLAG_ABORT;
5338                 return;
5339         }
5340         ext2fs_mark_super_dirty(fs);
5341         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
5342         pctx.blk2 = *new_block;
5343         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
5344                           PR_1_RELOC_TO), &pctx);
5345         pctx.blk2 = 0;
5346         e2fsck_pass1_block_map_lock(ctx);
5347         for (i = 0; i < num; i++) {
5348                 pctx.blk = i;
5349                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
5350                 if (old_block) {
5351                         pctx.errcode = io_channel_read_blk64(fs->io,
5352                                    old_block + i, 1, buf);
5353                         if (pctx.errcode)
5354                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
5355                         pctx.blk = (*new_block) + i;
5356                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
5357                                                               1, buf);
5358                 } else {
5359                         pctx.blk = (*new_block) + i;
5360                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
5361                                                            NULL, NULL);
5362                 }
5363
5364                 if (pctx.errcode)
5365                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
5366         }
5367         e2fsck_pass1_block_map_unlock(ctx);
5368         ext2fs_free_mem(&buf);
5369 }
5370
5371 /*
5372  * This routine gets called at the end of pass 1 if bad blocks are
5373  * detected in the superblock, group descriptors, inode_bitmaps, or
5374  * block bitmaps.  At this point, all of the blocks have been mapped
5375  * out, so we can try to allocate new block(s) to replace the bad
5376  * blocks.
5377  */
5378 static void handle_fs_bad_blocks(e2fsck_t ctx)
5379 {
5380         ext2_filsys fs = ctx->fs;
5381         dgrp_t          i;
5382         blk64_t         first_block;
5383         blk64_t         new_blk;
5384
5385         for (i = 0; i < fs->group_desc_count; i++) {
5386                 first_block = ext2fs_group_first_block2(fs, i);
5387
5388                 if (ctx->invalid_block_bitmap_flag[i]) {
5389                         new_blk = ext2fs_block_bitmap_loc(fs, i);
5390                         new_table_block(ctx, first_block, i, _("block bitmap"),
5391                                         1, &new_blk);
5392                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
5393                 }
5394                 if (ctx->invalid_inode_bitmap_flag[i]) {
5395                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
5396                         new_table_block(ctx, first_block, i, _("inode bitmap"),
5397                                         1, &new_blk);
5398                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
5399                 }
5400                 if (ctx->invalid_inode_table_flag[i]) {
5401                         new_blk = ext2fs_inode_table_loc(fs, i);
5402                         new_table_block(ctx, first_block, i, _("inode table"),
5403                                         fs->inode_blocks_per_group,
5404                                         &new_blk);
5405                         ext2fs_inode_table_loc_set(fs, i, new_blk);
5406                         ctx->flags |= E2F_FLAG_RESTART;
5407                 }
5408         }
5409         ctx->invalid_bitmaps = 0;
5410 }
5411
5412 /*
5413  * This routine marks all blocks which are used by the superblock,
5414  * group descriptors, inode bitmaps, and block bitmaps.
5415  */
5416 static void mark_table_blocks(e2fsck_t ctx)
5417 {
5418         ext2_filsys fs = ctx->fs;
5419         blk64_t b;
5420         dgrp_t  i;
5421         unsigned int    j;
5422         struct problem_context pctx;
5423
5424         clear_problem_context(&pctx);
5425
5426         for (i = 0; i < fs->group_desc_count; i++) {
5427                 pctx.group = i;
5428
5429                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
5430                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
5431
5432                 /*
5433                  * Mark the blocks used for the inode table
5434                  */
5435                 if (ext2fs_inode_table_loc(fs, i)) {
5436                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
5437                              j < fs->inode_blocks_per_group;
5438                              j++, b++) {
5439                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5440                                                              b)) {
5441                                         pctx.blk = b;
5442                                         if (!ctx->invalid_inode_table_flag[i] &&
5443                                             fix_problem(ctx,
5444                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
5445                                                 ctx->invalid_inode_table_flag[i]++;
5446                                                 ctx->invalid_bitmaps++;
5447                                         }
5448                                 } else {
5449                                     ext2fs_mark_block_bitmap2(
5450                                                 ctx->block_found_map, b);
5451                                     ext2fs_mark_block_bitmap2(
5452                                                 ctx->block_metadata_map, b);
5453                                 }
5454                         }
5455                 }
5456
5457                 /*
5458                  * Mark block used for the block bitmap
5459                  */
5460                 if (ext2fs_block_bitmap_loc(fs, i)) {
5461                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5462                                      ext2fs_block_bitmap_loc(fs, i))) {
5463                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
5464                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
5465                                         ctx->invalid_block_bitmap_flag[i]++;
5466                                         ctx->invalid_bitmaps++;
5467                                 }
5468                         } else {
5469                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
5470                                      ext2fs_block_bitmap_loc(fs, i));
5471                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5472                                      ext2fs_block_bitmap_loc(fs, i));
5473                         }
5474                 }
5475                 /*
5476                  * Mark block used for the inode bitmap
5477                  */
5478                 if (ext2fs_inode_bitmap_loc(fs, i)) {
5479                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5480                                      ext2fs_inode_bitmap_loc(fs, i))) {
5481                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
5482                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
5483                                         ctx->invalid_inode_bitmap_flag[i]++;
5484                                         ctx->invalid_bitmaps++;
5485                                 }
5486                         } else {
5487                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5488                                      ext2fs_inode_bitmap_loc(fs, i));
5489                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
5490                                      ext2fs_inode_bitmap_loc(fs, i));
5491                         }
5492                 }
5493         }
5494 }
5495
5496 /*
5497  * These subroutines short circuits ext2fs_get_blocks and
5498  * ext2fs_check_directory; we use them since we already have the inode
5499  * structure, so there's no point in letting the ext2fs library read
5500  * the inode again.
5501  */
5502 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
5503                                   blk_t *blocks)
5504 {
5505         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5506         int     i;
5507
5508         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5509                 return EXT2_ET_CALLBACK_NOTHANDLED;
5510
5511         for (i=0; i < EXT2_N_BLOCKS; i++)
5512                 blocks[i] = ctx->stashed_inode->i_block[i];
5513         return 0;
5514 }
5515
5516 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
5517                                   struct ext2_inode *inode)
5518 {
5519         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5520
5521         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5522                 return EXT2_ET_CALLBACK_NOTHANDLED;
5523         *inode = *ctx->stashed_inode;
5524         return 0;
5525 }
5526
5527 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
5528                             struct ext2_inode *inode)
5529 {
5530         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5531
5532         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
5533                 (inode != ctx->stashed_inode))
5534                 *ctx->stashed_inode = *inode;
5535         return EXT2_ET_CALLBACK_NOTHANDLED;
5536 }
5537
5538 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
5539 {
5540         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5541
5542         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5543                 return EXT2_ET_CALLBACK_NOTHANDLED;
5544
5545         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
5546                 return EXT2_ET_NO_DIRECTORY;
5547         return 0;
5548 }
5549
5550 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
5551                                         blk64_t *ret)
5552 {
5553         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5554         errcode_t       retval;
5555         blk64_t         new_block;
5556
5557         if (ctx->block_found_map) {
5558                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
5559                                            &new_block);
5560                 if (retval)
5561                         return retval;
5562                 if (fs->block_map) {
5563                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
5564                         ext2fs_mark_bb_dirty(fs);
5565                 }
5566         } else {
5567                 if (!fs->block_map) {
5568                         retval = ext2fs_read_block_bitmap(fs);
5569                         if (retval)
5570                                 return retval;
5571                 }
5572
5573                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
5574                 if (retval)
5575                         return retval;
5576         }
5577
5578         *ret = new_block;
5579         return (0);
5580 }
5581
5582 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
5583                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
5584 {
5585         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5586         errcode_t       retval;
5587
5588         if (ctx->block_found_map)
5589                 return ext2fs_new_range(fs, flags, goal, len,
5590                                         ctx->block_found_map, pblk, plen);
5591
5592         if (!fs->block_map) {
5593                 retval = ext2fs_read_block_bitmap(fs);
5594                 if (retval)
5595                         return retval;
5596         }
5597
5598         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
5599                                 pblk, plen);
5600 }
5601
5602 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
5603 {
5604         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5605
5606         /* Never free a critical metadata block */
5607         if (ctx->block_found_map &&
5608             ctx->block_metadata_map &&
5609             inuse < 0 &&
5610             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
5611                 return;
5612
5613         if (ctx->block_found_map) {
5614                 if (inuse > 0)
5615                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
5616                 else
5617                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
5618         }
5619 }
5620
5621 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
5622                                            blk_t num, int inuse)
5623 {
5624         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5625
5626         /* Never free a critical metadata block */
5627         if (ctx->block_found_map &&
5628             ctx->block_metadata_map &&
5629             inuse < 0 &&
5630             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
5631                 return;
5632
5633         if (ctx->block_found_map) {
5634                 if (inuse > 0)
5635                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
5636                                                         blk, num);
5637                 else
5638                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
5639                                                         blk, num);
5640         }
5641 }
5642
5643 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
5644 {
5645         ext2_filsys fs = ctx->fs;
5646
5647         if (use_shortcuts) {
5648                 fs->get_blocks = pass1_get_blocks;
5649                 fs->check_directory = pass1_check_directory;
5650                 fs->read_inode = pass1_read_inode;
5651                 fs->write_inode = pass1_write_inode;
5652                 ctx->stashed_ino = 0;
5653         } else {
5654                 fs->get_blocks = 0;
5655                 fs->check_directory = 0;
5656                 fs->read_inode = 0;
5657                 fs->write_inode = 0;
5658         }
5659 }
5660
5661 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
5662 {
5663         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
5664         ext2fs_set_block_alloc_stats_callback(ctx->fs,
5665                                                 e2fsck_block_alloc_stats, 0);
5666         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
5667         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
5668                                         e2fsck_block_alloc_stats_range, NULL);
5669 }