Whamcloud - gitweb
e2fsck: split and merge invalid bitmaps
[tools/e2fsprogs.git] / e2fsck / pass1.c
1 /*
2  * pass1.c -- pass #1 of e2fsck: sequential scan of the inode table
3  *
4  * Copyright (C) 1993, 1994, 1995, 1996, 1997 Theodore Ts'o.
5  *
6  * %Begin-Header%
7  * This file may be redistributed under the terms of the GNU Public
8  * License.
9  * %End-Header%
10  *
11  * Pass 1 of e2fsck iterates over all the inodes in the filesystems,
12  * and applies the following tests to each inode:
13  *
14  *      - The mode field of the inode must be legal.
15  *      - The size and block count fields of the inode are correct.
16  *      - A data block must not be used by another inode
17  *
18  * Pass 1 also gathers the collects the following information:
19  *
20  *      - A bitmap of which inodes are in use.          (inode_used_map)
21  *      - A bitmap of which inodes are directories.     (inode_dir_map)
22  *      - A bitmap of which inodes are regular files.   (inode_reg_map)
23  *      - A bitmap of which inodes have bad fields.     (inode_bad_map)
24  *      - A bitmap of which inodes are in bad blocks.   (inode_bb_map)
25  *      - A bitmap of which inodes are imagic inodes.   (inode_imagic_map)
26  *      - A bitmap of which inodes are casefolded.      (inode_casefold_map)
27  *      - A bitmap of which blocks are in use.          (block_found_map)
28  *      - A bitmap of which blocks are in use by two inodes     (block_dup_map)
29  *      - The data blocks of the directory inodes.      (dir_map)
30  *      - Ref counts for ea_inodes.                     (ea_inode_refs)
31  *      - The encryption policy ID of each encrypted inode. (encrypted_files)
32  *
33  * Pass 1 is designed to stash away enough information so that the
34  * other passes should not need to read in the inode information
35  * during the normal course of a filesystem check.  (Although if an
36  * inconsistency is detected, other passes may need to read in an
37  * inode to fix it.)
38  *
39  * Note that pass 1B will be invoked if there are any duplicate blocks
40  * found.
41  */
42
43 #define _GNU_SOURCE 1 /* get strnlen() */
44 #include "config.h"
45 #include <string.h>
46 #include <time.h>
47 #ifdef HAVE_ERRNO_H
48 #include <errno.h>
49 #endif
50 #include <assert.h>
51 #ifdef HAVE_PTHREAD
52 #include <pthread.h>
53 #endif
54
55 #include "e2fsck.h"
56 #include <ext2fs/ext2_ext_attr.h>
57 /* todo remove this finally */
58 #include <ext2fs/ext2fsP.h>
59 #include <e2p/e2p.h>
60
61 #include "problem.h"
62
63 #ifdef NO_INLINE_FUNCS
64 #define _INLINE_
65 #else
66 #define _INLINE_ inline
67 #endif
68
69 #undef DEBUG
70
71 struct ea_quota {
72         blk64_t blocks;
73         __u64 inodes;
74 };
75
76 static int process_block(ext2_filsys fs, blk64_t        *blocknr,
77                          e2_blkcnt_t blockcnt, blk64_t ref_blk,
78                          int ref_offset, void *priv_data);
79 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
80                              e2_blkcnt_t blockcnt, blk64_t ref_blk,
81                              int ref_offset, void *priv_data);
82 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
83                          char *block_buf,
84                          const struct ea_quota *ea_ibody_quota);
85 static void mark_table_blocks(e2fsck_t ctx);
86 static void alloc_bb_map(e2fsck_t ctx);
87 static void alloc_imagic_map(e2fsck_t ctx);
88 static void mark_inode_bad(e2fsck_t ctx, ext2_ino_t ino);
89 static void add_casefolded_dir(e2fsck_t ctx, ext2_ino_t ino);
90 static void handle_fs_bad_blocks(e2fsck_t ctx);
91 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
92 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
93                                   dgrp_t group, void * priv_data);
94 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
95                                     char *block_buf, int adjust_sign);
96 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
97
98 struct process_block_struct {
99         ext2_ino_t      ino;
100         unsigned        is_dir:1, is_reg:1, clear:1, suppress:1,
101                                 fragmented:1, compressed:1, bbcheck:1,
102                                 inode_modified:1;
103         blk64_t         num_blocks;
104         blk64_t         max_blocks;
105         blk64_t         last_block;
106         e2_blkcnt_t     last_init_lblock;
107         e2_blkcnt_t     last_db_block;
108         int             num_illegal_blocks;
109         blk64_t         previous_block;
110         struct ext2_inode *inode;
111         struct problem_context *pctx;
112         ext2fs_block_bitmap fs_meta_blocks;
113         e2fsck_t        ctx;
114         blk64_t         next_lblock;
115         struct extent_tree_info eti;
116 };
117
118 struct process_inode_block {
119         ext2_ino_t ino;
120         struct ea_quota ea_ibody_quota;
121         struct ext2_inode_large inode;
122 };
123
124 struct scan_callback_struct {
125         e2fsck_t                         ctx;
126         char                            *block_buf;
127         struct process_inode_block      *inodes_to_process;
128         int                             *process_inode_count;
129 };
130
131 static void process_inodes(e2fsck_t ctx, char *block_buf,
132                            struct process_inode_block *inodes_to_process,
133                            int *process_inode_count);
134
135 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
136                             EXT2_MIN_BLOCK_LOG_SIZE + 1];
137
138 /*
139  * Check to make sure a device inode is real.  Returns 1 if the device
140  * checks out, 0 if not.
141  *
142  * Note: this routine is now also used to check FIFO's and Sockets,
143  * since they have the same requirement; the i_block fields should be
144  * zero.
145  */
146 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
147                                     struct ext2_inode *inode)
148 {
149         int     i;
150
151         /*
152          * If the index or extents flag is set, then this is a bogus
153          * device/fifo/socket
154          */
155         if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
156                 return 0;
157
158         /*
159          * We should be able to do the test below all the time, but
160          * because the kernel doesn't forcibly clear the device
161          * inode's additional i_block fields, there are some rare
162          * occasions when a legitimate device inode will have non-zero
163          * additional i_block fields.  So for now, we only complain
164          * when the immutable flag is set, which should never happen
165          * for devices.  (And that's when the problem is caused, since
166          * you can't set or clear immutable flags for devices.)  Once
167          * the kernel has been fixed we can change this...
168          */
169         if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
170                 for (i=4; i < EXT2_N_BLOCKS; i++)
171                         if (inode->i_block[i])
172                                 return 0;
173         }
174         return 1;
175 }
176
177 /*
178  * Check to make sure a symlink inode is real.  Returns 1 if the symlink
179  * checks out, 0 if not.
180  */
181 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
182                                struct ext2_inode *inode, char *buf)
183 {
184         unsigned int buflen;
185         unsigned int len;
186
187         if ((inode->i_size_high || inode->i_size == 0) ||
188             (inode->i_flags & EXT2_INDEX_FL))
189                 return 0;
190
191         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
192                 size_t inline_size;
193
194                 if (inode->i_flags & EXT4_EXTENTS_FL)
195                         return 0;
196                 if (ext2fs_inline_data_size(fs, ino, &inline_size))
197                         return 0;
198                 if (inode->i_size != inline_size)
199                         return 0;
200
201                 return 1;
202         }
203
204         if (ext2fs_is_fast_symlink(inode)) {
205                 if (inode->i_flags & EXT4_EXTENTS_FL)
206                         return 0;
207                 buf = (char *)inode->i_block;
208                 buflen = sizeof(inode->i_block);
209         } else {
210                 ext2_extent_handle_t    handle;
211                 struct ext2_extent_info info;
212                 struct ext2fs_extent    extent;
213                 blk64_t blk;
214                 int i;
215
216                 if (inode->i_flags & EXT4_EXTENTS_FL) {
217                         if (ext2fs_extent_open2(fs, ino, inode, &handle))
218                                 return 0;
219                         if (ext2fs_extent_get_info(handle, &info) ||
220                             (info.num_entries != 1) ||
221                             (info.max_depth != 0)) {
222                                 ext2fs_extent_free(handle);
223                                 return 0;
224                         }
225                         if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
226                                               &extent) ||
227                             (extent.e_lblk != 0) ||
228                             (extent.e_len != 1)) {
229                                 ext2fs_extent_free(handle);
230                                 return 0;
231                         }
232                         blk = extent.e_pblk;
233                         ext2fs_extent_free(handle);
234                 } else {
235                         blk = inode->i_block[0];
236
237                         for (i = 1; i < EXT2_N_BLOCKS; i++)
238                                 if (inode->i_block[i])
239                                         return 0;
240                 }
241
242                 if (blk < fs->super->s_first_data_block ||
243                     blk >= ext2fs_blocks_count(fs->super))
244                         return 0;
245
246                 if (io_channel_read_blk64(fs->io, blk, 1, buf))
247                         return 0;
248
249                 buflen = fs->blocksize;
250         }
251
252         if (inode->i_flags & EXT4_ENCRYPT_FL)
253                 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
254         else
255                 len = strnlen(buf, buflen);
256
257         if (len >= buflen)
258                 return 0;
259
260         if (len != inode->i_size)
261                 return 0;
262         return 1;
263 }
264
265 /*
266  * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
267  */
268 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
269 static void check_extents_inlinedata(e2fsck_t ctx,
270                                      struct problem_context *pctx)
271 {
272         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
273                 return;
274
275         if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
276                 return;
277
278         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
279         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
280 }
281 #undef BAD_SPECIAL_FLAGS
282
283 /*
284  * If the immutable (or append-only) flag is set on the inode, offer
285  * to clear it.
286  */
287 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
288 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
289 {
290         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
291                 return;
292
293         if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
294                 return;
295
296         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
297         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
298 }
299
300 /*
301  * If device, fifo or socket, check size is zero -- if not offer to
302  * clear it
303  */
304 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
305 {
306         struct ext2_inode *inode = pctx->inode;
307
308         if (EXT2_I_SIZE(inode) == 0)
309                 return;
310
311         if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
312                 return;
313
314         ext2fs_inode_size_set(ctx->fs, inode, 0);
315         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
316 }
317
318 /*
319  * For a given size, calculate how many blocks would be charged towards quota.
320  */
321 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
322 {
323         blk64_t clusters;
324
325         clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
326         return EXT2FS_C2B(fs, clusters);
327 }
328
329 /*
330  * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
331  * the problem code.
332  */
333 static problem_t check_large_ea_inode(e2fsck_t ctx,
334                                       struct ext2_ext_attr_entry *entry,
335                                       struct problem_context *pctx,
336                                       blk64_t *quota_blocks)
337 {
338         struct ext2_inode inode;
339         __u32 hash, signed_hash;
340         errcode_t retval;
341
342         /* Check if inode is within valid range */
343         if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
344             (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
345                 pctx->num = entry->e_value_inum;
346                 return PR_1_ATTR_VALUE_EA_INODE;
347         }
348
349         e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
350
351         retval = ext2fs_ext_attr_hash_entry3(ctx->fs, entry, NULL, &hash,
352                                              &signed_hash);
353         if (retval) {
354                 com_err("check_large_ea_inode", retval,
355                         _("while hashing entry with e_value_inum = %u"),
356                         entry->e_value_inum);
357                 fatal_error(ctx, 0);
358         }
359
360         if ((hash == entry->e_hash) || (signed_hash == entry->e_hash)) {
361                 *quota_blocks = size_to_quota_blocks(ctx->fs,
362                                                      entry->e_value_size);
363         } else {
364                 /* This might be an old Lustre-style ea_inode reference. */
365                 if (inode.i_mtime == pctx->ino &&
366                     inode.i_generation == pctx->inode->i_generation) {
367                         *quota_blocks = 0;
368                 } else {
369                         /* If target inode is also missing EA_INODE flag,
370                          * this is likely to be a bad reference.
371                          */
372                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
373                                 pctx->num = entry->e_value_inum;
374                                 return PR_1_ATTR_VALUE_EA_INODE;
375                         } else {
376                                 pctx->num = entry->e_hash;
377                                 return PR_1_ATTR_HASH;
378                         }
379                 }
380         }
381
382         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
383                 pctx->num = entry->e_value_inum;
384                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
385                         inode.i_flags |= EXT4_EA_INODE_FL;
386                         e2fsck_pass1_fix_lock(ctx);
387                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
388                                            &inode);
389                         e2fsck_pass1_fix_unlock(ctx);
390                 } else {
391                         return PR_1_ATTR_NO_EA_INODE_FL;
392                 }
393         }
394         return 0;
395 }
396
397 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
398                               struct ext2_ext_attr_entry *first, void *end)
399 {
400         struct ext2_ext_attr_entry *entry = first;
401         struct ext2_ext_attr_entry *np = EXT2_EXT_ATTR_NEXT(entry);
402
403         while ((void *) entry < end && (void *) np < end &&
404                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
405                 if (!entry->e_value_inum)
406                         goto next;
407                 if (!ctx->ea_inode_refs) {
408                         pctx->errcode = ea_refcount_create(0,
409                                                            &ctx->ea_inode_refs);
410                         if (pctx->errcode) {
411                                 pctx->num = 4;
412                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
413                                 ctx->flags |= E2F_FLAG_ABORT;
414                                 return;
415                         }
416                 }
417                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
418                                       0);
419         next:
420                 entry = np;
421                 np = EXT2_EXT_ATTR_NEXT(entry);
422         }
423 }
424
425 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
426                               struct ea_quota *ea_ibody_quota)
427 {
428         struct ext2_super_block *sb = ctx->fs->super;
429         struct ext2_inode_large *inode;
430         struct ext2_ext_attr_entry *entry;
431         char *start, *header, *end;
432         unsigned int storage_size, remain;
433         problem_t problem = 0;
434         region_t region = 0;
435
436         ea_ibody_quota->blocks = 0;
437         ea_ibody_quota->inodes = 0;
438
439         inode = (struct ext2_inode_large *) pctx->inode;
440         storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
441                 inode->i_extra_isize;
442         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
443                  inode->i_extra_isize;
444         end = header + storage_size;
445         start = header + sizeof(__u32);
446         entry = (struct ext2_ext_attr_entry *) start;
447
448         /* scan all entry's headers first */
449
450         /* take finish entry 0UL into account */
451         remain = storage_size - sizeof(__u32);
452
453         region = region_create(0, storage_size);
454         if (!region) {
455                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
456                 problem = 0;
457                 ctx->flags |= E2F_FLAG_ABORT;
458                 return;
459         }
460         if (region_allocate(region, 0, sizeof(__u32))) {
461                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
462                 goto fix;
463         }
464
465         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
466                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
467                 __u32 hash;
468
469                 if (region_allocate(region, (char *)entry - (char *)header,
470                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
471                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
472                         goto fix;
473                 }
474
475                 /* header eats this space */
476                 remain -= sizeof(struct ext2_ext_attr_entry);
477
478                 /* is attribute name valid? */
479                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
480                         pctx->num = entry->e_name_len;
481                         problem = PR_1_ATTR_NAME_LEN;
482                         goto fix;
483                 }
484
485                 /* attribute len eats this space */
486                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
487
488                 if (entry->e_value_inum == 0) {
489                         /* check value size */
490                         if (entry->e_value_size > remain) {
491                                 pctx->num = entry->e_value_size;
492                                 problem = PR_1_ATTR_VALUE_SIZE;
493                                 goto fix;
494                         }
495
496                         if (entry->e_value_size &&
497                             region_allocate(region,
498                                             sizeof(__u32) + entry->e_value_offs,
499                                             EXT2_EXT_ATTR_SIZE(
500                                                 entry->e_value_size))) {
501                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
502                                 goto fix;
503                         }
504
505                         hash = ext2fs_ext_attr_hash_entry(entry,
506                                                 start + entry->e_value_offs);
507                         if (entry->e_hash != 0 && entry->e_hash != hash)
508                                 hash = ext2fs_ext_attr_hash_entry_signed(entry,
509                                                 start + entry->e_value_offs);
510
511                         /* e_hash may be 0 in older inode's ea */
512                         if (entry->e_hash != 0 && entry->e_hash != hash) {
513                                 pctx->num = entry->e_hash;
514                                 problem = PR_1_ATTR_HASH;
515                                 goto fix;
516                         }
517                 } else {
518                         blk64_t quota_blocks;
519
520                         problem = check_large_ea_inode(ctx, entry, pctx,
521                                                        &quota_blocks);
522                         if (problem != 0)
523                                 goto fix;
524
525                         ea_ibody_quota->blocks += quota_blocks;
526                         ea_ibody_quota->inodes++;
527                 }
528
529                 /* If EA value is stored in external inode then it does not
530                  * consume space here */
531                 if (entry->e_value_inum == 0)
532                         remain -= entry->e_value_size;
533
534                 entry = EXT2_EXT_ATTR_NEXT(entry);
535         }
536
537         if (region_allocate(region, (char *)entry - (char *)header,
538                             sizeof(__u32))) {
539                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
540                 goto fix;
541         }
542 fix:
543         if (region)
544                 region_free(region);
545         /*
546          * it seems like a corruption. it's very unlikely we could repair
547          * EA(s) in automatic fashion -bzzz
548          */
549         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
550                 inc_ea_inode_refs(ctx, pctx,
551                                   (struct ext2_ext_attr_entry *)start, end);
552                 return;
553         }
554
555         /* simply remove all possible EA(s) */
556         *((__u32 *)header) = 0UL;
557         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
558                                 EXT2_INODE_SIZE(sb), "pass1");
559         ea_ibody_quota->blocks = 0;
560         ea_ibody_quota->inodes = 0;
561 }
562
563 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
564         return (xtime & (1U << 31)) != 0 &&
565                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
566 }
567
568 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
569         check_inode_extra_negative_epoch(inode->i_##xtime, \
570                                          inode->i_##xtime##_extra)
571
572 /* When today's date is earlier than 2242, we assume that atimes,
573  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
574  * actually pre-1970 dates mis-encoded.
575  */
576 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
577
578 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
579                                     struct ea_quota *ea_ibody_quota)
580 {
581         struct ext2_super_block *sb = ctx->fs->super;
582         struct ext2_inode_large *inode;
583         __u32 *eamagic;
584         int min, max;
585
586         ea_ibody_quota->blocks = 0;
587         ea_ibody_quota->inodes = 0;
588
589         inode = (struct ext2_inode_large *) pctx->inode;
590         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
591                 /* this isn't large inode. so, nothing to check */
592                 return;
593         }
594
595 #if 0
596         printf("inode #%u, i_extra_size %d\n", pctx->ino,
597                         inode->i_extra_isize);
598 #endif
599         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
600         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
601         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
602         /*
603          * For now we will allow i_extra_isize to be 0, but really
604          * implementations should never allow i_extra_isize to be 0
605          */
606         if (inode->i_extra_isize &&
607             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
608              inode->i_extra_isize & 3)) {
609                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
610                         return;
611                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
612                         inode->i_extra_isize = sb->s_want_extra_isize;
613                 else
614                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
615                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
616                                         EXT2_INODE_SIZE(sb), "pass1");
617         }
618
619         /* check if there is no place for an EA header */
620         if (inode->i_extra_isize >= max - sizeof(__u32))
621                 return;
622
623         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
624                         inode->i_extra_isize);
625         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
626                 /* it seems inode has an extended attribute(s) in body */
627                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
628         }
629
630         /*
631          * If the inode's extended atime (ctime, crtime, mtime) is stored in
632          * the old, invalid format, repair it.
633          */
634         if (((sizeof(time_t) <= 4) ||
635              (((sizeof(time_t) > 4) &&
636                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
637             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
638              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
639              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
640              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
641
642                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
643                         return;
644
645                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
646                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
647                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
648                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
649                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
650                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
651                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
652                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
653                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
654                                         EXT2_INODE_SIZE(sb), "pass1");
655         }
656
657 }
658
659 /*
660  * Check to see if the inode might really be a directory, despite i_mode
661  *
662  * This is a lot of complexity for something for which I'm not really
663  * convinced happens frequently in the wild.  If for any reason this
664  * causes any problems, take this code out.
665  * [tytso:20070331.0827EDT]
666  */
667 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
668                                 char *buf)
669 {
670         struct ext2_inode *inode = pctx->inode;
671         struct ext2_dir_entry   *dirent;
672         errcode_t               retval;
673         blk64_t                 blk;
674         unsigned int            i, rec_len, not_device = 0;
675         int                     extent_fs;
676         int                     inlinedata_fs;
677
678         /*
679          * If the mode looks OK, we believe it.  If the first block in
680          * the i_block array is 0, this cannot be a directory. If the
681          * inode is extent-mapped, it is still the case that the latter
682          * cannot be 0 - the magic number in the extent header would make
683          * it nonzero.
684          */
685         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
686             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
687                 return;
688
689         /* 
690          * Check the block numbers in the i_block array for validity:
691          * zero blocks are skipped (but the first one cannot be zero -
692          * see above), other blocks are checked against the first and
693          * max data blocks (from the the superblock) and against the
694          * block bitmap. Any invalid block found means this cannot be
695          * a directory.
696          * 
697          * If there are non-zero blocks past the fourth entry, then
698          * this cannot be a device file: we remember that for the next
699          * check.
700          *
701          * For extent mapped files, we don't do any sanity checking:
702          * just try to get the phys block of logical block 0 and run
703          * with it.
704          *
705          * For inline data files, we just try to get the size of inline
706          * data.  If it's true, we will treat it as a directory.
707          */
708
709         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
710         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
711         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
712                 size_t size;
713                 __u32 dotdot;
714                 unsigned int rec_len2;
715                 struct ext2_dir_entry de;
716
717                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
718                         return;
719                 /*
720                  * If the size isn't a multiple of 4, it's probably not a
721                  * directory??
722                  */
723                 if (size & 3)
724                         return;
725                 /*
726                  * If the first 10 bytes don't look like a directory entry,
727                  * it's probably not a directory.
728                  */
729                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
730                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
731                        EXT2_DIR_REC_LEN(0));
732                 dotdot = ext2fs_le32_to_cpu(dotdot);
733                 de.inode = ext2fs_le32_to_cpu(de.inode);
734                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
735                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
736                 if (dotdot >= ctx->fs->super->s_inodes_count ||
737                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
738                      dotdot != EXT2_ROOT_INO) ||
739                     de.inode >= ctx->fs->super->s_inodes_count ||
740                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
741                      de.inode != 0) ||
742                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
743                               EXT4_INLINE_DATA_DOTDOT_SIZE)
744                         return;
745                 /* device files never have a "system.data" entry */
746                 goto isdir;
747         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
748                 /* extent mapped */
749                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
750                                  &blk))
751                         return;
752                 /* device files are never extent mapped */
753                 not_device++;
754         } else {
755                 for (i=0; i < EXT2_N_BLOCKS; i++) {
756                         blk = inode->i_block[i];
757                         if (!blk)
758                                 continue;
759                         if (i >= 4)
760                                 not_device++;
761
762                         e2fsck_pass1_block_map_lock(ctx);
763                         if (blk < ctx->fs->super->s_first_data_block ||
764                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
765                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
766                                                            blk)) {
767                                 e2fsck_pass1_block_map_unlock(ctx);
768                                 return; /* Invalid block, can't be dir */
769                         }
770                         e2fsck_pass1_block_map_unlock(ctx);
771                 }
772                 blk = inode->i_block[0];
773         }
774
775         /*
776          * If the mode says this is a device file and the i_links_count field
777          * is sane and we have not ruled it out as a device file previously,
778          * we declare it a device file, not a directory.
779          */
780         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
781             (inode->i_links_count == 1) && !not_device)
782                 return;
783
784         /* read the first block */
785         ehandler_operation(_("reading directory block"));
786         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
787         ehandler_operation(0);
788         if (retval)
789                 return;
790
791         dirent = (struct ext2_dir_entry *) buf;
792         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
793         if (retval)
794                 return;
795         if ((ext2fs_dirent_name_len(dirent) != 1) ||
796             (dirent->name[0] != '.') ||
797             (dirent->inode != pctx->ino) ||
798             (rec_len < 12) ||
799             (rec_len % 4) ||
800             (rec_len >= ctx->fs->blocksize - 12))
801                 return;
802
803         dirent = (struct ext2_dir_entry *) (buf + rec_len);
804         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
805         if (retval)
806                 return;
807         if ((ext2fs_dirent_name_len(dirent) != 2) ||
808             (dirent->name[0] != '.') ||
809             (dirent->name[1] != '.') ||
810             (rec_len < 12) ||
811             (rec_len % 4))
812                 return;
813
814 isdir:
815         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
816                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
817                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
818                                         EXT2_INODE_SIZE(ctx->fs->super),
819                                         "check_is_really_dir");
820         }
821 }
822
823 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
824                                      int flags, ext2_icount_t hint,
825                                      ext2_icount_t *ret)
826 {
827         unsigned int            threshold;
828         unsigned int            save_type;
829         ext2_ino_t              num_dirs;
830         errcode_t               retval;
831         char                    *tdb_dir;
832         int                     enable;
833
834         *ret = 0;
835
836         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
837                            &tdb_dir);
838         profile_get_uint(ctx->profile, "scratch_files",
839                          "numdirs_threshold", 0, 0, &threshold);
840         profile_get_boolean(ctx->profile, "scratch_files",
841                             "icount", 0, 1, &enable);
842
843         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
844         if (retval)
845                 num_dirs = 1024;        /* Guess */
846
847         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
848             (!threshold || num_dirs > threshold)) {
849                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
850                                                   flags, ret);
851                 if (retval == 0)
852                         return 0;
853         }
854         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
855                                &save_type);
856         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
857                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
858         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
859         ctx->fs->default_bitmap_type = save_type;
860         return retval;
861 }
862
863 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
864                                             e2fsck_t ctx,
865                                             struct problem_context *pctx)
866 {
867         errcode_t retval;
868         struct ext2_inode_large inode;
869
870         /*
871          * Reread inode.  If we don't see checksum error, then this inode
872          * has been fixed elsewhere.
873          */
874         ctx->stashed_ino = 0;
875         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
876                                         sizeof(inode));
877         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
878                 return retval;
879         if (!retval)
880                 return 0;
881
882         /*
883          * Checksum still doesn't match.  That implies that the inode passes
884          * all the sanity checks, so maybe the checksum is simply corrupt.
885          * See if the user will go for fixing that.
886          */
887         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
888                 return 0;
889
890
891         e2fsck_pass1_fix_lock(ctx);
892         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
893                                          sizeof(inode));
894         e2fsck_pass1_fix_unlock(ctx);
895         return retval;
896 }
897
898 static void reserve_block_for_root_repair(e2fsck_t ctx)
899 {
900         blk64_t         blk = 0;
901         errcode_t       err;
902         ext2_filsys     fs = ctx->fs;
903
904         ctx->root_repair_block = 0;
905         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
906                 return;
907
908         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
909         if (err)
910                 return;
911         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
912         ctx->root_repair_block = blk;
913 }
914
915 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
916 {
917         blk64_t         blk = 0;
918         errcode_t       err;
919         ext2_filsys     fs = ctx->fs;
920         static const char name[] = "lost+found";
921         ext2_ino_t      ino;
922
923         ctx->lnf_repair_block = 0;
924         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
925                 return;
926
927         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
928         if (err)
929                 return;
930         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
931         ctx->lnf_repair_block = blk;
932         return;
933 }
934
935 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
936                                          struct ext2_inode *inode,
937                                          size_t *sz)
938 {
939         void *p;
940         struct ext2_xattr_handle *handle;
941         errcode_t retval;
942
943         retval = ext2fs_xattrs_open(fs, ino, &handle);
944         if (retval)
945                 return retval;
946
947         retval = ext2fs_xattrs_read_inode(handle,
948                                           (struct ext2_inode_large *)inode);
949         if (retval)
950                 goto err;
951
952         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
953         if (retval)
954                 goto err;
955         ext2fs_free_mem(&p);
956 err:
957         (void) ext2fs_xattrs_close(&handle);
958         return retval;
959 }
960
961 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
962                                     struct problem_context *pctx,
963                                     int failed_csum)
964 {
965         if (!failed_csum)
966                 return;
967
968         /*
969          * If the inode failed the checksum and the user didn't
970          * clear the inode, test the checksum again -- if it still
971          * fails, ask the user if the checksum should be corrected.
972          */
973         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
974         if (pctx->errcode)
975                 ctx->flags |= E2F_FLAG_ABORT;
976 }
977 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
978         do { \
979                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
980                 if ((ctx)->flags & E2F_FLAG_ABORT) \
981                         return; \
982         } while (0)
983
984 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
985 {
986         __u32 x;
987         int i;
988
989         for (i = 0; i < EXT2_N_BLOCKS; i++) {
990                 x = inode->i_block[i];
991 #ifdef WORDS_BIGENDIAN
992                 x = ext2fs_swab32(x);
993 #endif
994                 if (x >= ext2fs_blocks_count(fs->super))
995                         return 0;
996         }
997
998         return 1;
999 }
1000
1001 /*
1002  * Figure out what to do with an inode that has both extents and inline data
1003  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
1004  */
1005 static int fix_inline_data_extents_file(e2fsck_t ctx,
1006                                         ext2_ino_t ino,
1007                                         struct ext2_inode *inode,
1008                                         int inode_size,
1009                                         struct problem_context *pctx)
1010 {
1011         size_t max_inline_ea_size;
1012         ext2_filsys fs = ctx->fs;
1013         int dirty = 0;
1014
1015         /* Both feature flags not set?  Just run the regular checks */
1016         if (!ext2fs_has_feature_extents(fs->super) &&
1017             !ext2fs_has_feature_inline_data(fs->super))
1018                 return 0;
1019
1020         /* Clear both flags if it's a special file */
1021         if (LINUX_S_ISCHR(inode->i_mode) ||
1022             LINUX_S_ISBLK(inode->i_mode) ||
1023             LINUX_S_ISFIFO(inode->i_mode) ||
1024             LINUX_S_ISSOCK(inode->i_mode)) {
1025                 check_extents_inlinedata(ctx, pctx);
1026                 return 0;
1027         }
1028
1029         /* If it looks like an extent tree, try to clear inlinedata */
1030         if (ext2fs_extent_header_verify(inode->i_block,
1031                                  sizeof(inode->i_block)) == 0 &&
1032             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1033                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1034                 dirty = 1;
1035                 goto out;
1036         }
1037
1038         /* If it looks short enough to be inline data, try to clear extents */
1039         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1040                 max_inline_ea_size = inode_size -
1041                                      (EXT2_GOOD_OLD_INODE_SIZE +
1042                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1043         else
1044                 max_inline_ea_size = 0;
1045         if (EXT2_I_SIZE(inode) <
1046             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1047             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1048                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1049                 dirty = 1;
1050                 goto out;
1051         }
1052
1053         /*
1054          * Too big for inline data, but no evidence of extent tree -
1055          * maybe it's a block map file?  If the mappings all look valid?
1056          */
1057         if (could_be_block_map(fs, inode) &&
1058             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1059 #ifdef WORDS_BIGENDIAN
1060                 int i;
1061
1062                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1063                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1064 #endif
1065
1066                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1067                 dirty = 1;
1068                 goto out;
1069         }
1070
1071         /* Oh well, just clear the busted inode. */
1072         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1073                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1074                 return -1;
1075         }
1076
1077 out:
1078         if (dirty)
1079                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1080
1081         return 0;
1082 }
1083
1084 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1085 {
1086         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1087         dgrp_t start = *group, grp;
1088         blk64_t blocks_to_read = 0;
1089         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1090
1091         if (ctx->readahead_kb == 0)
1092                 goto out;
1093
1094         /* Keep iterating groups until we have enough to readahead */
1095         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1096         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1097                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1098                         continue;
1099                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1100                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1101                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1102                                         inodes_per_block;
1103                 if (blocks_to_read * ctx->fs->blocksize >
1104                     ctx->readahead_kb * 1024)
1105                         break;
1106         }
1107
1108         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1109                                grp - start + 1);
1110         if (err == EAGAIN) {
1111                 ctx->readahead_kb /= 2;
1112                 err = 0;
1113         }
1114
1115 out:
1116         if (err) {
1117                 /* Error; disable itable readahead */
1118                 *group = ctx->fs->group_desc_count;
1119                 *next_ino = ctx->fs->super->s_inodes_count;
1120         } else {
1121                 /*
1122                  * Don't do more readahead until we've reached the first inode
1123                  * of the last inode scan buffer block for the last group.
1124                  */
1125                 *group = grp + 1;
1126                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1127                                      ctx->inode_buffer_blocks :
1128                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1129                                     ctx->fs->blocksize /
1130                                     EXT2_INODE_SIZE(ctx->fs->super);
1131                 inodes_in_group--;
1132                 *next_ino = inodes_in_group -
1133                             (inodes_in_group % inodes_per_buffer) + 1 +
1134                             (grp * ctx->fs->super->s_inodes_per_group);
1135         }
1136 }
1137
1138 /*
1139  * Check if the passed ino is one of the used superblock quota inodes.
1140  *
1141  * Before the quota inodes were journaled, older superblock quota inodes
1142  * were just regular files in the filesystem and not reserved inodes.  This
1143  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1144  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1145  */
1146 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1147 {
1148         enum quota_type qtype;
1149
1150         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1151                 if (*quota_sb_inump(sb, qtype) == ino)
1152                         return 1;
1153
1154         return 0;
1155 }
1156
1157 /*
1158  * Check if the passed ino is one of the reserved quota inodes.
1159  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1160  * inodes.  These inodes may or may not be in use by the quota feature.
1161  */
1162 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1163 {
1164         enum quota_type qtype;
1165
1166         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1167                 if (quota_type2inum(qtype, fs->super) == ino)
1168                         return 1;
1169
1170         return 0;
1171 }
1172
1173 static int e2fsck_should_abort(e2fsck_t ctx)
1174 {
1175         e2fsck_t global_ctx;
1176
1177         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1178                 return 1;
1179
1180         if (ctx->global_ctx) {
1181                 global_ctx = ctx->global_ctx;
1182                 if (global_ctx->flags & E2F_FLAG_SIGNAL_MASK)
1183                         return 1;
1184         }
1185         return 0;
1186 }
1187
1188 /*
1189  * We need call mark_table_blocks() before multiple
1190  * thread start, since all known system blocks should be
1191  * marked and checked later.
1192  */
1193 static errcode_t e2fsck_pass1_prepare(e2fsck_t ctx)
1194 {
1195         struct problem_context pctx;
1196         ext2_filsys fs = ctx->fs;
1197
1198         clear_problem_context(&pctx);
1199         if (!(ctx->options & E2F_OPT_PREEN))
1200                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1201
1202         pctx.errcode = e2fsck_allocate_subcluster_bitmap(ctx->fs,
1203                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1204                         "block_found_map", &ctx->block_found_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         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
1212                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1213                         "block_metadata_map", &ctx->block_metadata_map);
1214         if (pctx.errcode) {
1215                 pctx.num = 1;
1216                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1217                 ctx->flags |= E2F_FLAG_ABORT;
1218                 return pctx.errcode;
1219         }
1220
1221         mark_table_blocks(ctx);
1222         pctx.errcode = ext2fs_convert_subcluster_bitmap(ctx->fs,
1223                                                 &ctx->block_found_map);
1224         if (pctx.errcode) {
1225                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1226                 ctx->flags |= E2F_FLAG_ABORT;
1227                 return pctx.errcode;
1228         }
1229
1230         if (ext2fs_has_feature_mmp(fs->super) &&
1231             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1232             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1233                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1234                                           fs->super->s_mmp_block);
1235
1236         return 0;
1237 }
1238
1239 static void e2fsck_pass1_post(e2fsck_t ctx)
1240 {
1241         struct problem_context pctx;
1242         ext2_filsys fs = ctx->fs;
1243         char *block_buf;
1244
1245         reserve_block_for_root_repair(ctx);
1246         reserve_block_for_lnf_repair(ctx);
1247
1248         if (ctx->invalid_bitmaps)
1249                 handle_fs_bad_blocks(ctx);
1250
1251         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
1252                 struct ext2_inode *inode;
1253                 int inode_size = EXT2_INODE_SIZE(fs->super);
1254                 inode = e2fsck_allocate_memory(ctx, inode_size,
1255                                                "scratch inode");
1256
1257                 clear_problem_context(&pctx);
1258                 pctx.errcode = ext2fs_create_resize_inode(fs);
1259                 if (pctx.errcode) {
1260                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
1261                                          &pctx)) {
1262                                 ctx->flags |= E2F_FLAG_ABORT;
1263                                 ext2fs_free_mem(&inode);
1264                                 ext2fs_free_mem(&block_buf);
1265                                 return;
1266                         }
1267                         pctx.errcode = 0;
1268                 }
1269                 if (!pctx.errcode) {
1270                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
1271                                           "recreate inode");
1272                         inode->i_mtime = ctx->now;
1273                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
1274                                            "recreate inode");
1275                 }
1276                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
1277                 ext2fs_free_mem(&inode);
1278         }
1279
1280         if (ctx->flags & E2F_FLAG_RESTART) {
1281                 ext2fs_free_mem(&block_buf);
1282                 return;
1283         }
1284
1285         if (ctx->block_dup_map) {
1286                 if (ctx->options & E2F_OPT_PREEN) {
1287                         clear_problem_context(&pctx);
1288                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
1289                 }
1290                 block_buf =
1291                         (char *)e2fsck_allocate_memory(ctx,
1292                                         ctx->fs->blocksize * 3,
1293                                         "block interate buffer");
1294                 e2fsck_pass1_dupblocks(ctx, block_buf);
1295                 ext2fs_free_mem(&block_buf);
1296         }
1297 }
1298
1299
1300 void e2fsck_pass1_run(e2fsck_t ctx)
1301 {
1302         int     i;
1303         ext2_filsys fs = ctx->fs;
1304         ext2_ino_t      ino = 0;
1305         struct ext2_inode *inode = NULL;
1306         ext2_inode_scan scan = NULL;
1307         char            *block_buf = NULL;
1308 #ifdef RESOURCE_TRACK
1309         struct resource_track   rtrack;
1310 #endif
1311         unsigned char   frag, fsize;
1312         struct          problem_context pctx;
1313         struct          scan_callback_struct scan_struct;
1314         struct ext2_super_block *sb = ctx->fs->super;
1315         const char      *old_op;
1316         const char      *eop_next_inode = _("getting next inode from scan");
1317         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1318         int             low_dtime_check = 1;
1319         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1320         unsigned int    bufsize;
1321         int             failed_csum = 0;
1322         ext2_ino_t      ino_threshold = 0;
1323         dgrp_t          ra_group = 0;
1324         struct ea_quota ea_ibody_quota;
1325         struct process_inode_block *inodes_to_process;
1326         int             process_inode_count;
1327
1328         init_resource_track(&rtrack, ctx->fs->io);
1329         clear_problem_context(&pctx);
1330
1331         /* If we can do readahead, figure out how many groups to pull in. */
1332         if (!e2fsck_can_readahead(ctx->fs))
1333                 ctx->readahead_kb = 0;
1334         else if (ctx->readahead_kb == ~0ULL)
1335                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1336         pass1_readahead(ctx, &ra_group, &ino_threshold);
1337
1338         if (ext2fs_has_feature_dir_index(fs->super) &&
1339             !(ctx->options & E2F_OPT_NO)) {
1340                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1341                         ctx->dirs_to_hash = 0;
1342         }
1343
1344 #ifdef MTRACE
1345         mtrace_print("Pass 1");
1346 #endif
1347
1348         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1349         extent_fs = ext2fs_has_feature_extents(sb);
1350         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1351         casefold_fs = ext2fs_has_feature_casefold(sb);
1352
1353         /*
1354          * Allocate bitmaps structures
1355          */
1356         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1357                                                     EXT2FS_BMAP64_RBTREE,
1358                                                     "inode_used_map",
1359                                                     &ctx->inode_used_map);
1360         if (pctx.errcode) {
1361                 pctx.num = 1;
1362                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1363                 ctx->flags |= E2F_FLAG_ABORT;
1364                 return;
1365         }
1366         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1367                         _("directory inode map"),
1368                         ctx->global_ctx ? EXT2FS_BMAP64_RBTREE :
1369                         EXT2FS_BMAP64_AUTODIR,
1370                         "inode_dir_map", &ctx->inode_dir_map);
1371         if (pctx.errcode) {
1372                 pctx.num = 2;
1373                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1374                 ctx->flags |= E2F_FLAG_ABORT;
1375                 return;
1376         }
1377         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1378                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1379                         "inode_reg_map", &ctx->inode_reg_map);
1380         if (pctx.errcode) {
1381                 pctx.num = 6;
1382                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1383                 ctx->flags |= E2F_FLAG_ABORT;
1384                 return;
1385         }
1386         if (casefold_fs) {
1387                 pctx.errcode =
1388                         e2fsck_allocate_inode_bitmap(fs,
1389                                                      _("inode casefold map"),
1390                                                      EXT2FS_BMAP64_RBTREE,
1391                                                      "inode_casefold_map",
1392                                                      &ctx->inode_casefold_map);
1393                 if (pctx.errcode) {
1394                         pctx.num = 1;
1395                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1396                         ctx->flags |= E2F_FLAG_ABORT;
1397                         return;
1398                 }
1399         }
1400         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1401                                            &ctx->inode_link_info);
1402         if (pctx.errcode) {
1403                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1404                 ctx->flags |= E2F_FLAG_ABORT;
1405                 return;
1406         }
1407         bufsize = inode_size;
1408         if (bufsize < sizeof(struct ext2_inode_large))
1409                 bufsize = sizeof(struct ext2_inode_large);
1410         inode = (struct ext2_inode *)
1411                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1412
1413         inodes_to_process = (struct process_inode_block *)
1414                 e2fsck_allocate_memory(ctx,
1415                                        (ctx->process_inode_size *
1416                                         sizeof(struct process_inode_block)),
1417                                        "array of inodes to process");
1418         process_inode_count = 0;
1419
1420         pctx.errcode = ext2fs_init_dblist(fs, 0);
1421         if (pctx.errcode) {
1422                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1423                 ctx->flags |= E2F_FLAG_ABORT;
1424                 goto endit;
1425         }
1426
1427         /*
1428          * If the last orphan field is set, clear it, since the pass1
1429          * processing will automatically find and clear the orphans.
1430          * In the future, we may want to try using the last_orphan
1431          * linked list ourselves, but for now, we clear it so that the
1432          * ext3 mount code won't get confused.
1433          */
1434         if (!(ctx->options & E2F_OPT_READONLY)) {
1435                 if (fs->super->s_last_orphan) {
1436                         fs->super->s_last_orphan = 0;
1437                         ext2fs_mark_super_dirty(fs);
1438                 }
1439         }
1440
1441         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1442                                                     "block iterate buffer");
1443         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1444                 e2fsck_use_inode_shortcuts(ctx, 1);
1445         e2fsck_intercept_block_allocations(ctx);
1446         old_op = ehandler_operation(_("opening inode scan"));
1447         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1448                                               &scan);
1449         ehandler_operation(old_op);
1450         if (pctx.errcode) {
1451                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1452                 ctx->flags |= E2F_FLAG_ABORT;
1453                 goto endit;
1454         }
1455         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1456                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1457         ctx->stashed_inode = inode;
1458         scan_struct.ctx = ctx;
1459         scan_struct.block_buf = block_buf;
1460         scan_struct.inodes_to_process = inodes_to_process;
1461         scan_struct.process_inode_count = &process_inode_count;
1462         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1463         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1464                                               ctx->fs->group_desc_count)))
1465                 goto endit;
1466         if ((fs->super->s_wtime &&
1467              fs->super->s_wtime < fs->super->s_inodes_count) ||
1468             (fs->super->s_mtime &&
1469              fs->super->s_mtime < fs->super->s_inodes_count) ||
1470             (fs->super->s_mkfs_time &&
1471              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1472                 low_dtime_check = 0;
1473
1474         /* Set up ctx->lost_and_found if possible */
1475         (void) e2fsck_get_lost_and_found(ctx, 0);
1476
1477 #ifdef HAVE_PTHREAD
1478         if (ctx->global_ctx) {
1479                 if (ctx->options & E2F_OPT_DEBUG &&
1480                     ctx->options & E2F_OPT_MULTITHREAD)
1481                         fprintf(stderr, "thread %d jumping to group %d\n",
1482                                         ctx->thread_info.et_thread_index,
1483                                         ctx->thread_info.et_group_start);
1484                 pctx.errcode = ext2fs_inode_scan_goto_blockgroup(scan,
1485                                         ctx->thread_info.et_group_start);
1486                 if (pctx.errcode) {
1487                         fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1488                         ctx->flags |= E2F_FLAG_ABORT;
1489                         goto endit;
1490                 }
1491         }
1492 #endif
1493
1494         while (1) {
1495                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1496                         if (e2fsck_mmp_update(fs))
1497                                 fatal_error(ctx, 0);
1498                 }
1499                 old_op = ehandler_operation(eop_next_inode);
1500                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1501                                                           inode, inode_size);
1502                 if (ino > ino_threshold)
1503                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1504                 ehandler_operation(old_op);
1505                 if (e2fsck_should_abort(ctx))
1506                         goto endit;
1507                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1508                         /*
1509                          * If badblocks says badblocks is bad, offer to clear
1510                          * the list, update the in-core bb list, and restart
1511                          * the inode scan.
1512                          */
1513                         if (ino == EXT2_BAD_INO &&
1514                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1515                                         &pctx)) {
1516                                 errcode_t err;
1517
1518                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1519                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1520                                 ctx->fs->badblocks = NULL;
1521                                 err = ext2fs_read_bb_inode(ctx->fs,
1522                                                         &ctx->fs->badblocks);
1523                                 if (err) {
1524                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1525                                                     &pctx);
1526                                         ctx->flags |= E2F_FLAG_ABORT;
1527                                 } else
1528                                         ctx->flags |= E2F_FLAG_RESTART;
1529                                 goto endit;
1530                         }
1531                         if (!ctx->inode_bb_map)
1532                                 alloc_bb_map(ctx);
1533                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1534                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1535                         continue;
1536                 }
1537                 if (pctx.errcode == EXT2_ET_SCAN_FINISHED)
1538                         break;
1539                 if (pctx.errcode &&
1540                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1541                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1542                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1543                         ctx->flags |= E2F_FLAG_ABORT;
1544                         goto endit;
1545                 }
1546                 if (!ino)
1547                         break;
1548 #ifdef HAVE_PTHREAD
1549                 if (ctx->global_ctx)
1550                         ctx->thread_info.et_inode_number++;
1551 #endif
1552                 pctx.ino = ino;
1553                 pctx.inode = inode;
1554                 ctx->stashed_ino = ino;
1555
1556                 /* Clear trashed inode? */
1557                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1558                     inode->i_links_count > 0 &&
1559                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1560                         pctx.errcode = 0;
1561                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1562                 }
1563                 failed_csum = pctx.errcode != 0;
1564
1565                 /*
1566                  * Check for inodes who might have been part of the
1567                  * orphaned list linked list.  They should have gotten
1568                  * dealt with by now, unless the list had somehow been
1569                  * corrupted.
1570                  *
1571                  * FIXME: In the future, inodes which are still in use
1572                  * (and which are therefore) pending truncation should
1573                  * be handled specially.  Right now we just clear the
1574                  * dtime field, and the normal e2fsck handling of
1575                  * inodes where i_size and the inode blocks are
1576                  * inconsistent is to fix i_size, instead of releasing
1577                  * the extra blocks.  This won't catch the inodes that
1578                  * was at the end of the orphan list, but it's better
1579                  * than nothing.  The right answer is that there
1580                  * shouldn't be any bugs in the orphan list handling.  :-)
1581                  */
1582                 if (inode->i_dtime && low_dtime_check &&
1583                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1584                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1585                                 inode->i_dtime = inode->i_links_count ?
1586                                         0 : ctx->now;
1587                                 e2fsck_write_inode(ctx, ino, inode,
1588                                                    "pass1");
1589                                 failed_csum = 0;
1590                         }
1591                 }
1592
1593                 if (inode->i_links_count) {
1594                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1595                                            ino, inode->i_links_count);
1596                         if (pctx.errcode) {
1597                                 pctx.num = inode->i_links_count;
1598                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1599                                 ctx->flags |= E2F_FLAG_ABORT;
1600                                 goto endit;
1601                         }
1602                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1603                            !quota_inum_is_reserved(fs, ino)) {
1604                         if (!inode->i_dtime && inode->i_mode) {
1605                                 if (fix_problem(ctx,
1606                                             PR_1_ZERO_DTIME, &pctx)) {
1607                                         inode->i_dtime = ctx->now;
1608                                         e2fsck_write_inode(ctx, ino, inode,
1609                                                            "pass1");
1610                                         failed_csum = 0;
1611                                 }
1612                         }
1613                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1614                         continue;
1615                 }
1616
1617                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1618                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1619                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1620                      (!casefold_fs &&
1621                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1622                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1623                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1624                 }
1625
1626                 /* Conflicting inlinedata/extents inode flags? */
1627                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1628                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1629                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1630                                                                inode_size,
1631                                                                &pctx);
1632                         if (res < 0) {
1633                                 /* skip FINISH_INODE_LOOP */
1634                                 continue;
1635                         }
1636                 }
1637
1638                 /* Test for incorrect inline_data flags settings. */
1639                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1640                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1641                         size_t size = 0;
1642
1643                         pctx.errcode = get_inline_data_ea_size(fs, ino, inode,
1644                                                                &size);
1645                         if (!pctx.errcode &&
1646                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1647                                 e2fsck_pass1_fix_lock(ctx);
1648                                 ext2fs_set_feature_inline_data(sb);
1649                                 ext2fs_mark_super_dirty(fs);
1650                                 e2fsck_pass1_fix_unlock(ctx);
1651                                 inlinedata_fs = 1;
1652                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1653                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1654                                 /* skip FINISH_INODE_LOOP */
1655                                 continue;
1656                         }
1657                 }
1658
1659                 /* Test for inline data flag but no attr */
1660                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1661                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1662                         size_t size = 0;
1663                         errcode_t err;
1664                         int flags;
1665
1666                         flags = fs->flags;
1667                         if (failed_csum)
1668                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1669                         err = get_inline_data_ea_size(fs, ino, inode, &size);
1670                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1671                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1672
1673                         switch (err) {
1674                         case 0:
1675                                 /* Everything is awesome... */
1676                                 break;
1677                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1678                         case EXT2_ET_BAD_EA_HASH:
1679                         case EXT2_ET_BAD_EA_HEADER:
1680                         case EXT2_ET_EA_BAD_NAME_LEN:
1681                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1682                         case EXT2_ET_EA_KEY_NOT_FOUND:
1683                         case EXT2_ET_EA_NO_SPACE:
1684                         case EXT2_ET_MISSING_EA_FEATURE:
1685                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1686                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1687                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1688                         case EXT2_ET_NO_INLINE_DATA:
1689                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1690                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1691                         case EXT2_ET_EA_INODE_CORRUPTED:
1692                                 /* broken EA or no system.data EA; truncate */
1693                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1694                                                 &pctx)) {
1695                                         err = ext2fs_inode_size_set(fs, inode, 0);
1696                                         if (err) {
1697                                                 pctx.errcode = err;
1698                                                 ctx->flags |= E2F_FLAG_ABORT;
1699                                                 goto endit;
1700                                         }
1701                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1702                                         memset(&inode->i_block, 0,
1703                                                sizeof(inode->i_block));
1704                                         e2fsck_write_inode(ctx, ino, inode,
1705                                                            "pass1");
1706                                         failed_csum = 0;
1707                                 }
1708                                 break;
1709                         default:
1710                                 /* Some other kind of non-xattr error? */
1711                                 pctx.errcode = err;
1712                                 ctx->flags |= E2F_FLAG_ABORT;
1713                                 goto endit;
1714                         }
1715                 }
1716
1717                 /*
1718                  * Test for incorrect extent flag settings.
1719                  *
1720                  * On big-endian machines we must be careful:
1721                  * When the inode is read, the i_block array is not swapped
1722                  * if the extent flag is set.  Therefore if we are testing
1723                  * for or fixing a wrongly-set flag, we must potentially
1724                  * (un)swap before testing, or after fixing.
1725                  */
1726
1727                 /*
1728                  * In this case the extents flag was set when read, so
1729                  * extent_header_verify is ok.  If the inode is cleared,
1730                  * no need to swap... so no extra swapping here.
1731                  */
1732                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1733                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1734                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1735                         if ((ext2fs_extent_header_verify(inode->i_block,
1736                                                  sizeof(inode->i_block)) == 0) &&
1737                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1738                                 e2fsck_pass1_fix_lock(ctx);
1739                                 ext2fs_set_feature_extents(sb);
1740                                 ext2fs_mark_super_dirty(fs);
1741                                 extent_fs = 1;
1742                                 e2fsck_pass1_fix_unlock(ctx);
1743                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1744                         clear_inode:
1745                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1746                                 if (ino == EXT2_BAD_INO)
1747                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1748                                                                  ino);
1749                                 /* skip FINISH_INODE_LOOP */
1750                                 continue;
1751                         }
1752                 }
1753
1754                 /*
1755                  * For big-endian machines:
1756                  * If the inode didn't have the extents flag set when it
1757                  * was read, then the i_blocks array was swapped.  To test
1758                  * as an extents header, we must swap it back first.
1759                  * IF we then set the extents flag, the entire i_block
1760                  * array must be un/re-swapped to make it proper extents data.
1761                  */
1762                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1763                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1764                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1765                     (LINUX_S_ISREG(inode->i_mode) ||
1766                      LINUX_S_ISDIR(inode->i_mode))) {
1767                         void *ehp;
1768 #ifdef WORDS_BIGENDIAN
1769                         __u32 tmp_block[EXT2_N_BLOCKS];
1770
1771                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1772                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1773                         ehp = tmp_block;
1774 #else
1775                         ehp = inode->i_block;
1776 #endif
1777                         if ((ext2fs_extent_header_verify(ehp,
1778                                          sizeof(inode->i_block)) == 0) &&
1779                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1780                                 inode->i_flags |= EXT4_EXTENTS_FL;
1781 #ifdef WORDS_BIGENDIAN
1782                                 memcpy(inode->i_block, tmp_block,
1783                                        sizeof(inode->i_block));
1784 #endif
1785                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1786                                 failed_csum = 0;
1787                         }
1788                 }
1789
1790                 if (ino == EXT2_BAD_INO) {
1791                         struct process_block_struct pb;
1792
1793                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1794                              inode->i_gid || inode->i_links_count ||
1795                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1796                              inode->i_file_acl) &&
1797                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1798                                 memset(inode, 0, sizeof(struct ext2_inode));
1799                                 e2fsck_write_inode(ctx, ino, inode,
1800                                                    "clear bad inode");
1801                                 failed_csum = 0;
1802                         }
1803
1804                         e2fsck_pass1_block_map_lock(ctx);
1805                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1806                                                           &pb.fs_meta_blocks);
1807                         e2fsck_pass1_block_map_unlock(ctx);
1808                         if (pctx.errcode) {
1809                                 pctx.num = 4;
1810                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1811                                 ctx->flags |= E2F_FLAG_ABORT;
1812                                 goto endit;
1813                         }
1814                         pb.ino = EXT2_BAD_INO;
1815                         pb.num_blocks = pb.last_block = 0;
1816                         pb.last_db_block = -1;
1817                         pb.num_illegal_blocks = 0;
1818                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1819                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1820                         pb.inode = inode;
1821                         pb.pctx = &pctx;
1822                         pb.ctx = ctx;
1823                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1824                                      block_buf, process_bad_block, &pb);
1825                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1826                         if (pctx.errcode) {
1827                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1828                                 ctx->flags |= E2F_FLAG_ABORT;
1829                                 goto endit;
1830                         }
1831                         if (pb.bbcheck)
1832                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1833                                 ctx->flags |= E2F_FLAG_ABORT;
1834                                 goto endit;
1835                         }
1836                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1837                         clear_problem_context(&pctx);
1838                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1839                         continue;
1840                 } else if (ino == EXT2_ROOT_INO) {
1841                         /*
1842                          * Make sure the root inode is a directory; if
1843                          * not, offer to clear it.  It will be
1844                          * regenerated in pass #3.
1845                          */
1846                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1847                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1848                                         goto clear_inode;
1849                         }
1850                         /*
1851                          * If dtime is set, offer to clear it.  mke2fs
1852                          * version 0.2b created filesystems with the
1853                          * dtime field set for the root and lost+found
1854                          * directories.  We won't worry about
1855                          * /lost+found, since that can be regenerated
1856                          * easily.  But we will fix the root directory
1857                          * as a special case.
1858                          */
1859                         if (inode->i_dtime && inode->i_links_count) {
1860                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1861                                         inode->i_dtime = 0;
1862                                         e2fsck_write_inode(ctx, ino, inode,
1863                                                            "pass1");
1864                                         failed_csum = 0;
1865                                 }
1866                         }
1867                 } else if (ino == EXT2_JOURNAL_INO) {
1868                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1869                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1870                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1871                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1872                                                 &pctx)) {
1873                                         inode->i_mode = LINUX_S_IFREG;
1874                                         e2fsck_write_inode(ctx, ino, inode,
1875                                                            "pass1");
1876                                         failed_csum = 0;
1877                                 }
1878                                 check_blocks(ctx, &pctx, block_buf, NULL);
1879                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1880                                 continue;
1881                         }
1882                         if ((inode->i_links_count ||
1883                              inode->i_blocks || inode->i_block[0]) &&
1884                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1885                                         &pctx)) {
1886                                 memset(inode, 0, inode_size);
1887                                 ext2fs_icount_store(ctx->inode_link_info,
1888                                                     ino, 0);
1889                                 e2fsck_write_inode_full(ctx, ino, inode,
1890                                                         inode_size, "pass1");
1891                                 failed_csum = 0;
1892                         }
1893                 } else if (quota_inum_is_reserved(fs, ino)) {
1894                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1895                         if (ext2fs_has_feature_quota(fs->super) &&
1896                             quota_inum_is_super(fs->super, ino)) {
1897                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1898                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1899                                                         &pctx)) {
1900                                         inode->i_mode = LINUX_S_IFREG;
1901                                         e2fsck_write_inode(ctx, ino, inode,
1902                                                         "pass1");
1903                                         failed_csum = 0;
1904                                 }
1905                                 check_blocks(ctx, &pctx, block_buf, NULL);
1906                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1907                                 continue;
1908                         }
1909                         if ((inode->i_links_count ||
1910                              inode->i_blocks || inode->i_block[0]) &&
1911                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1912                                         &pctx)) {
1913                                 memset(inode, 0, inode_size);
1914                                 ext2fs_icount_store(ctx->inode_link_info,
1915                                                     ino, 0);
1916                                 e2fsck_write_inode_full(ctx, ino, inode,
1917                                                         inode_size, "pass1");
1918                                 failed_csum = 0;
1919                         }
1920                 } else if (ino == fs->super->s_orphan_file_inum) {
1921                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1922                         if (ext2fs_has_feature_orphan_file(fs->super)) {
1923                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1924                                     fix_problem(ctx, PR_1_ORPHAN_FILE_BAD_MODE,
1925                                                 &pctx)) {
1926                                         inode->i_mode = LINUX_S_IFREG;
1927                                         e2fsck_write_inode(ctx, ino, inode,
1928                                                            "pass1");
1929                                         failed_csum = 0;
1930                                 }
1931                                 check_blocks(ctx, &pctx, block_buf, NULL);
1932                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1933                                 continue;
1934                         }
1935                         if ((inode->i_links_count ||
1936                              inode->i_blocks || inode->i_block[0]) &&
1937                             fix_problem(ctx, PR_1_ORPHAN_FILE_NOT_CLEAR,
1938                                         &pctx)) {
1939                                 memset(inode, 0, inode_size);
1940                                 ext2fs_icount_store(ctx->inode_link_info, ino,
1941                                                     0);
1942                                 e2fsck_write_inode_full(ctx, ino, inode,
1943                                                         inode_size, "pass1");
1944                                 failed_csum = 0;
1945                         }
1946                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1947                         problem_t problem = 0;
1948
1949                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1950                         if (ino == EXT2_BOOT_LOADER_INO) {
1951                                 if (LINUX_S_ISDIR(inode->i_mode))
1952                                         problem = PR_1_RESERVED_BAD_MODE;
1953                         } else if (ino == EXT2_RESIZE_INO) {
1954                                 if (inode->i_mode &&
1955                                     !LINUX_S_ISREG(inode->i_mode))
1956                                         problem = PR_1_RESERVED_BAD_MODE;
1957                         } else {
1958                                 if (inode->i_mode != 0)
1959                                         problem = PR_1_RESERVED_BAD_MODE;
1960                         }
1961                         if (problem) {
1962                                 if (fix_problem(ctx, problem, &pctx)) {
1963                                         inode->i_mode = 0;
1964                                         e2fsck_write_inode(ctx, ino, inode,
1965                                                            "pass1");
1966                                         failed_csum = 0;
1967                                 }
1968                         }
1969                         check_blocks(ctx, &pctx, block_buf, NULL);
1970                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1971                         continue;
1972                 }
1973
1974                 if (!inode->i_links_count) {
1975                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1976                         continue;
1977                 }
1978                 /*
1979                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1980                  * deleted files.  Oops.
1981                  *
1982                  * Since all new ext2 implementations get this right,
1983                  * we now assume that the case of non-zero
1984                  * i_links_count and non-zero dtime means that we
1985                  * should keep the file, not delete it.
1986                  *
1987                  */
1988                 if (inode->i_dtime) {
1989                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1990                                 inode->i_dtime = 0;
1991                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1992                                 failed_csum = 0;
1993                         }
1994                 }
1995
1996                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1997                 switch (fs->super->s_creator_os) {
1998                     case EXT2_OS_HURD:
1999                         frag = inode->osd2.hurd2.h_i_frag;
2000                         fsize = inode->osd2.hurd2.h_i_fsize;
2001                         break;
2002                     default:
2003                         frag = fsize = 0;
2004                 }
2005
2006                 if (inode->i_faddr || frag || fsize ||
2007                     (!ext2fs_has_feature_largedir(fs->super) &&
2008                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
2009                         mark_inode_bad(ctx, ino);
2010                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
2011                     !ext2fs_has_feature_64bit(fs->super) &&
2012                     inode->osd2.linux2.l_i_file_acl_high != 0)
2013                         mark_inode_bad(ctx, ino);
2014                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
2015                     !ext2fs_has_feature_huge_file(fs->super) &&
2016                     (inode->osd2.linux2.l_i_blocks_hi != 0))
2017                         mark_inode_bad(ctx, ino);
2018                 if (inode->i_flags & EXT2_IMAGIC_FL) {
2019                         if (imagic_fs) {
2020                                 if (!ctx->inode_imagic_map)
2021                                         alloc_imagic_map(ctx);
2022                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
2023                                                          ino);
2024                         } else {
2025                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
2026                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
2027                                         e2fsck_write_inode(ctx, ino,
2028                                                            inode, "pass1");
2029                                         failed_csum = 0;
2030                                 }
2031                         }
2032                 }
2033
2034                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
2035                 check_is_really_dir(ctx, &pctx, block_buf);
2036
2037                 /*
2038                  * ext2fs_inode_has_valid_blocks2 does not actually look
2039                  * at i_block[] values, so not endian-sensitive here.
2040                  */
2041                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
2042                     LINUX_S_ISLNK(inode->i_mode) &&
2043                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
2044                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
2045                         inode->i_flags &= ~EXT4_EXTENTS_FL;
2046                         e2fsck_write_inode(ctx, ino, inode, "pass1");
2047                         failed_csum = 0;
2048                 }
2049
2050                 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
2051                     add_encrypted_file(ctx, &pctx) < 0)
2052                         goto clear_inode;
2053
2054                 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
2055                         ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
2056
2057                 if (LINUX_S_ISDIR(inode->i_mode)) {
2058                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
2059                         e2fsck_add_dir_info(ctx, ino, 0);
2060                         ctx->fs_directory_count++;
2061                         if (inode->i_flags & EXT4_CASEFOLD_FL)
2062                                 add_casefolded_dir(ctx, ino);
2063                 } else if (LINUX_S_ISREG (inode->i_mode)) {
2064                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
2065                         ctx->fs_regular_count++;
2066                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
2067                            e2fsck_pass1_check_device_inode(fs, inode)) {
2068                         check_extents_inlinedata(ctx, &pctx);
2069                         check_immutable(ctx, &pctx);
2070                         check_size(ctx, &pctx);
2071                         ctx->fs_chardev_count++;
2072                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
2073                            e2fsck_pass1_check_device_inode(fs, inode)) {
2074                         check_extents_inlinedata(ctx, &pctx);
2075                         check_immutable(ctx, &pctx);
2076                         check_size(ctx, &pctx);
2077                         ctx->fs_blockdev_count++;
2078                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
2079                            e2fsck_pass1_check_symlink(fs, ino, inode,
2080                                                       block_buf)) {
2081                         check_immutable(ctx, &pctx);
2082                         ctx->fs_symlinks_count++;
2083                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
2084                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2085                                 continue;
2086                         } else if (ext2fs_is_fast_symlink(inode)) {
2087                                 ctx->fs_fast_symlinks_count++;
2088                                 check_blocks(ctx, &pctx, block_buf,
2089                                              &ea_ibody_quota);
2090                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2091                                 continue;
2092                         }
2093                 }
2094                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
2095                          e2fsck_pass1_check_device_inode(fs, inode)) {
2096                         check_extents_inlinedata(ctx, &pctx);
2097                         check_immutable(ctx, &pctx);
2098                         check_size(ctx, &pctx);
2099                         ctx->fs_fifo_count++;
2100                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
2101                            e2fsck_pass1_check_device_inode(fs, inode)) {
2102                         check_extents_inlinedata(ctx, &pctx);
2103                         check_immutable(ctx, &pctx);
2104                         check_size(ctx, &pctx);
2105                         ctx->fs_sockets_count++;
2106                 } else
2107                         mark_inode_bad(ctx, ino);
2108                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2109                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
2110                         if (inode->i_block[EXT2_IND_BLOCK])
2111                                 ctx->fs_ind_count++;
2112                         if (inode->i_block[EXT2_DIND_BLOCK])
2113                                 ctx->fs_dind_count++;
2114                         if (inode->i_block[EXT2_TIND_BLOCK])
2115                                 ctx->fs_tind_count++;
2116                 }
2117                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2118                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
2119                     (inode->i_block[EXT2_IND_BLOCK] ||
2120                      inode->i_block[EXT2_DIND_BLOCK] ||
2121                      inode->i_block[EXT2_TIND_BLOCK] ||
2122                      ext2fs_file_acl_block(fs, inode))) {
2123                         struct process_inode_block *itp;
2124
2125                         itp = &inodes_to_process[process_inode_count];
2126                         itp->ino = ino;
2127                         itp->ea_ibody_quota = ea_ibody_quota;
2128                         if (inode_size < sizeof(struct ext2_inode_large))
2129                                 memcpy(&itp->inode, inode, inode_size);
2130                         else
2131                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
2132                         process_inode_count++;
2133                 } else
2134                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
2135
2136                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2137
2138                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2139                         goto endit;
2140
2141                 if (process_inode_count >= ctx->process_inode_size) {
2142                         process_inodes(ctx, block_buf, inodes_to_process,
2143                                        &process_inode_count);
2144
2145                         if (e2fsck_should_abort(ctx))
2146                                 goto endit;
2147                 }
2148         }
2149         process_inodes(ctx, block_buf, inodes_to_process,
2150                        &process_inode_count);
2151         ext2fs_close_inode_scan(scan);
2152         scan = NULL;
2153
2154         /*
2155          * If any extended attribute blocks' reference counts need to
2156          * be adjusted, either up (ctx->refcount_extra), or down
2157          * (ctx->refcount), then fix them.
2158          */
2159         if (ctx->refcount) {
2160                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
2161                 ea_refcount_free(ctx->refcount);
2162                 ctx->refcount = 0;
2163         }
2164         if (ctx->refcount_extra) {
2165                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
2166                                         block_buf, +1);
2167                 ea_refcount_free(ctx->refcount_extra);
2168                 ctx->refcount_extra = 0;
2169         }
2170
2171         if (ctx->ea_block_quota_blocks) {
2172                 ea_refcount_free(ctx->ea_block_quota_blocks);
2173                 ctx->ea_block_quota_blocks = 0;
2174         }
2175
2176         if (ctx->ea_block_quota_inodes) {
2177                 ea_refcount_free(ctx->ea_block_quota_inodes);
2178                 ctx->ea_block_quota_inodes = 0;
2179         }
2180
2181
2182         /* We don't need the block_ea_map any more */
2183         if (ctx->block_ea_map) {
2184                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2185                 ctx->block_ea_map = 0;
2186         }
2187
2188         /* We don't need the encryption policy => ID map any more */
2189         destroy_encryption_policy_map(ctx);
2190
2191         if (ctx->flags & E2F_FLAG_RESTART) {
2192                 /*
2193                  * Only the master copy of the superblock and block
2194                  * group descriptors are going to be written during a
2195                  * restart, so set the superblock to be used to be the
2196                  * master superblock.
2197                  */
2198                 ctx->use_superblock = 0;
2199                 goto endit;
2200         }
2201
2202         if (ctx->large_dirs && !ext2fs_has_feature_largedir(fs->super)) {
2203                 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2204                         ext2fs_set_feature_largedir(fs->super);
2205                         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2206                         ext2fs_mark_super_dirty(fs);
2207                 }
2208                 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2209                     fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2210                         ext2fs_update_dynamic_rev(fs);
2211                         ext2fs_mark_super_dirty(fs);
2212                 }
2213         }
2214
2215         ctx->flags |= E2F_FLAG_ALLOC_OK;
2216 endit:
2217         e2fsck_use_inode_shortcuts(ctx, 0);
2218         ext2fs_free_mem(&inodes_to_process);
2219         inodes_to_process = 0;
2220
2221         if (scan)
2222                 ext2fs_close_inode_scan(scan);
2223         if (block_buf)
2224                 ext2fs_free_mem(&block_buf);
2225         if (inode)
2226                 ext2fs_free_mem(&inode);
2227
2228         /*
2229          * The l+f inode may have been cleared, so zap it now and
2230          * later passes will recalculate it if necessary
2231          */
2232         ctx->lost_and_found = 0;
2233
2234         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2235                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2236         else
2237                 ctx->invalid_bitmaps++;
2238 }
2239
2240 static void init_ext2_max_sizes()
2241 {
2242         int     i;
2243         __u64   max_sizes;
2244
2245         /*
2246          * Init ext2_max_sizes which will be immutable and shared between
2247          * threads
2248          */
2249 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
2250
2251         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
2252                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
2253                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
2254                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
2255                 max_sizes = (max_sizes * (1UL << i));
2256                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
2257         }
2258 #undef EXT2_BPP
2259 }
2260
2261 #ifdef HAVE_PTHREAD
2262 static errcode_t e2fsck_pass1_copy_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2263                                           ext2fs_generic_bitmap *dest)
2264 {
2265         errcode_t ret;
2266
2267         ret = ext2fs_copy_bitmap(*src, dest);
2268         if (ret)
2269                 return ret;
2270
2271         (*dest)->fs = fs;
2272
2273         return 0;
2274 }
2275
2276 static void e2fsck_pass1_free_bitmap(ext2fs_generic_bitmap *bitmap)
2277 {
2278         if (*bitmap) {
2279                 ext2fs_free_generic_bmap(*bitmap);
2280                 *bitmap = NULL;
2281         }
2282
2283 }
2284
2285 static errcode_t e2fsck_pass1_merge_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2286                                           ext2fs_generic_bitmap *dest)
2287 {
2288         errcode_t ret = 0;
2289
2290         if (*src) {
2291                 if (*dest == NULL) {
2292                         *dest = *src;
2293                         *src = NULL;
2294                 } else {
2295                         ret = ext2fs_merge_bitmap(*src, *dest, NULL, NULL);
2296                         if (ret)
2297                                 return ret;
2298                 }
2299                 (*dest)->fs = fs;
2300         }
2301
2302         return 0;
2303 }
2304
2305 static errcode_t e2fsck_pass1_copy_fs(ext2_filsys dest, e2fsck_t src_context,
2306                                       ext2_filsys src)
2307 {
2308         errcode_t       retval;
2309
2310         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2311         dest->inode_map = NULL;
2312         dest->block_map = NULL;
2313         dest->badblocks = NULL;
2314         if (dest->dblist)
2315                 dest->dblist->fs = dest;
2316         if (src->block_map) {
2317                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2318                                                   &dest->block_map);
2319                 if (retval)
2320                         return retval;
2321         }
2322         if (src->inode_map) {
2323                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2324                                                   &dest->inode_map);
2325                 if (retval)
2326                         return retval;
2327         }
2328
2329         if (src->badblocks) {
2330                 retval = ext2fs_badblocks_copy(src->badblocks,
2331                                                &dest->badblocks);
2332                 if (retval)
2333                         return retval;
2334         }
2335
2336         /* disable it for now */
2337         src_context->openfs_flags &= ~EXT2_FLAG_EXCLUSIVE;
2338         retval = ext2fs_open_channel(dest, src_context->io_options,
2339                                      src_context->io_manager,
2340                                      src_context->openfs_flags,
2341                                      src->io->block_size);
2342         if (retval)
2343                 return retval;
2344
2345         /* Block size might not be default */
2346         io_channel_set_blksize(dest->io, src->io->block_size);
2347         ehandler_init(dest->io);
2348
2349         assert(dest->io->magic == src->io->magic);
2350         assert(dest->io->manager == src->io->manager);
2351         assert(strcmp(dest->io->name, src->io->name) == 0);
2352         assert(dest->io->block_size == src->io->block_size);
2353         assert(dest->io->read_error == src->io->read_error);
2354         assert(dest->io->write_error == src->io->write_error);
2355         assert(dest->io->refcount == src->io->refcount);
2356         assert(dest->io->flags == src->io->flags);
2357         assert(dest->io->app_data == dest);
2358         assert(src->io->app_data == src);
2359         assert(dest->io->align == src->io->align);
2360
2361         /* The data should be written to disk immediately */
2362         dest->io->flags |= CHANNEL_FLAGS_WRITETHROUGH;
2363         /* icache will be rebuilt if needed, so do not copy from @src */
2364         src->icache = NULL;
2365         return 0;
2366 }
2367
2368 static int e2fsck_pass1_merge_fs(ext2_filsys dest, ext2_filsys src)
2369 {
2370         struct ext2_inode_cache *icache = dest->icache;
2371         errcode_t retval = 0;
2372         io_channel dest_io;
2373         io_channel dest_image_io;
2374         ext2fs_inode_bitmap inode_map;
2375         ext2fs_block_bitmap block_map;
2376         ext2_badblocks_list badblocks;
2377         ext2_dblist dblist;
2378         int flags;
2379
2380         dest_io = dest->io;
2381         dest_image_io = dest->image_io;
2382         inode_map = dest->inode_map;
2383         block_map = dest->block_map;
2384         badblocks = dest->badblocks;
2385         dblist = dest->dblist;
2386         flags = dest->flags;
2387
2388         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2389         dest->io = dest_io;
2390         dest->image_io = dest_image_io;
2391         dest->icache = icache;
2392         dest->inode_map = inode_map;
2393         dest->block_map = block_map;
2394         dest->badblocks = badblocks;
2395         dest->dblist = dblist;
2396         if (dest->dblist)
2397                 dest->dblist->fs = dest;
2398         dest->flags = src->flags | flags;
2399         if (!(src->flags & EXT2_FLAG_VALID) || !(flags & EXT2_FLAG_VALID))
2400                 ext2fs_unmark_valid(dest);
2401
2402         if (src->icache) {
2403                 ext2fs_free_inode_cache(src->icache);
2404                 src->icache = NULL;
2405         }
2406
2407         retval = e2fsck_pass1_merge_bitmap(dest, &src->inode_map,
2408                                            &dest->inode_map);
2409         if (retval)
2410                 goto out;
2411
2412         retval = e2fsck_pass1_merge_bitmap(dest, &src->block_map,
2413                                           &dest->block_map);
2414         if (retval)
2415                 goto out;
2416
2417         if (src->dblist) {
2418                 if (dest->dblist) {
2419                         retval = ext2fs_merge_dblist(src->dblist,
2420                                                      dest->dblist);
2421                         if (retval)
2422                                 goto out;
2423                 } else {
2424                         dest->dblist = src->dblist;
2425                         dest->dblist->fs = dest;
2426                         src->dblist = NULL;
2427                 }
2428         }
2429
2430         if (src->badblocks) {
2431                 if (dest->badblocks == NULL)
2432                         retval = ext2fs_badblocks_copy(src->badblocks,
2433                                                        &dest->badblocks);
2434                 else
2435                         retval = ext2fs_badblocks_merge(src->badblocks,
2436                                                         dest->badblocks);
2437         }
2438 out:
2439         io_channel_close(src->io);
2440         if (src->inode_map)
2441                 ext2fs_free_generic_bmap(src->inode_map);
2442         if (src->block_map)
2443                 ext2fs_free_generic_bmap(src->block_map);
2444         if (src->badblocks)
2445                 ext2fs_badblocks_list_free(src->badblocks);
2446         if (src->dblist)
2447                 ext2fs_free_dblist(src->dblist);
2448
2449         return retval;
2450 }
2451
2452 static void e2fsck_pass1_copy_invalid_bitmaps(e2fsck_t global_ctx,
2453                                               e2fsck_t thread_ctx)
2454 {
2455         dgrp_t i, j;
2456         dgrp_t grp_start = thread_ctx->thread_info.et_group_start;
2457         dgrp_t grp_end = thread_ctx->thread_info.et_group_end;
2458         dgrp_t total = grp_end - grp_start;
2459
2460         thread_ctx->invalid_inode_bitmap_flag =
2461                         e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2462                                                 "invalid_inode_bitmap");
2463         thread_ctx->invalid_block_bitmap_flag =
2464                         e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2465                                                "invalid_block_bitmap");
2466         thread_ctx->invalid_inode_table_flag =
2467                         e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2468                                                "invalid_inode_table");
2469
2470         memcpy(thread_ctx->invalid_block_bitmap_flag,
2471                &global_ctx->invalid_block_bitmap_flag[grp_start],
2472                total * sizeof(int));
2473         memcpy(thread_ctx->invalid_inode_bitmap_flag,
2474                &global_ctx->invalid_inode_bitmap_flag[grp_start],
2475                total * sizeof(int));
2476         memcpy(thread_ctx->invalid_inode_table_flag,
2477                &global_ctx->invalid_inode_table_flag[grp_start],
2478                total * sizeof(int));
2479
2480         thread_ctx->invalid_bitmaps = 0;
2481         for (i = grp_start, j = 0; i < grp_end; i++, j++) {
2482                 if (thread_ctx->invalid_block_bitmap_flag[j])
2483                         thread_ctx->invalid_bitmaps++;
2484                 if (thread_ctx->invalid_inode_bitmap_flag[j])
2485                         thread_ctx->invalid_bitmaps++;
2486                 if (thread_ctx->invalid_inode_table_flag[j])
2487                         thread_ctx->invalid_bitmaps++;
2488         }
2489 }
2490
2491 static void e2fsck_pass1_merge_invalid_bitmaps(e2fsck_t global_ctx,
2492                                                e2fsck_t thread_ctx)
2493 {
2494         dgrp_t grp_start = thread_ctx->thread_info.et_group_start;
2495         dgrp_t grp_end = thread_ctx->thread_info.et_group_end;
2496         dgrp_t total = grp_end - grp_start;
2497
2498         memcpy(&global_ctx->invalid_block_bitmap_flag[grp_start],
2499                thread_ctx->invalid_block_bitmap_flag, total * sizeof(int));
2500         memcpy(&global_ctx->invalid_inode_bitmap_flag[grp_start],
2501                thread_ctx->invalid_inode_bitmap_flag, total * sizeof(int));
2502         memcpy(&global_ctx->invalid_inode_table_flag[grp_start],
2503                thread_ctx->invalid_inode_table_flag, total * sizeof(int));
2504         global_ctx->invalid_bitmaps += thread_ctx->invalid_bitmaps;
2505 }
2506
2507 static errcode_t e2fsck_pass1_thread_prepare(e2fsck_t global_ctx, e2fsck_t *thread_ctx,
2508                                              int thread_index, int num_threads)
2509 {
2510         errcode_t               retval;
2511         e2fsck_t                thread_context;
2512         ext2_filsys             thread_fs;
2513         ext2_filsys             global_fs = global_ctx->fs;
2514         struct e2fsck_thread    *tinfo;
2515         dgrp_t                  average_group;
2516
2517         assert(global_ctx->inode_used_map == NULL);
2518         assert(global_ctx->inode_dir_map == NULL);
2519         assert(global_ctx->inode_bb_map == NULL);
2520         assert(global_ctx->inode_imagic_map == NULL);
2521         assert(global_ctx->inode_reg_map == NULL);
2522         assert(global_ctx->inodes_to_rebuild == NULL);
2523
2524         assert(global_ctx->block_found_map != NULL);
2525         assert(global_ctx->block_metadata_map != NULL);
2526         assert(global_ctx->block_dup_map == NULL);
2527         assert(global_ctx->block_ea_map == NULL);
2528         assert(global_ctx->fs->dblist == NULL);
2529
2530         retval = ext2fs_get_mem(sizeof(struct e2fsck_struct), &thread_context);
2531         if (retval) {
2532                 com_err(global_ctx->program_name, retval, "while allocating memory");
2533                 return retval;
2534         }
2535         memcpy(thread_context, global_ctx, sizeof(struct e2fsck_struct));
2536         thread_context->global_ctx = global_ctx;
2537
2538         retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &thread_fs);
2539         if (retval) {
2540                 com_err(global_ctx->program_name, retval, "while allocating memory");
2541                 goto out_context;
2542         }
2543
2544         io_channel_flush_cleanup(global_fs->io);
2545         retval = e2fsck_pass1_copy_fs(thread_fs, global_ctx, global_fs);
2546         if (retval) {
2547                 com_err(global_ctx->program_name, retval, "while copying fs");
2548                 goto out_fs;
2549         }
2550         thread_fs->priv_data = thread_context;
2551
2552         thread_context->thread_info.et_thread_index = thread_index;
2553         set_up_logging(thread_context);
2554
2555         /*
2556          * Distribute work to multiple threads:
2557          * Each thread work on fs->group_desc_count / nthread groups.
2558          */
2559         tinfo = &thread_context->thread_info;
2560         average_group = thread_fs->group_desc_count / num_threads;
2561         if (average_group == 0)
2562                 average_group = 1;
2563         tinfo->et_group_start = average_group * thread_index;
2564         if (thread_index == num_threads - 1)
2565                 tinfo->et_group_end = thread_fs->group_desc_count;
2566         else
2567                 tinfo->et_group_end = average_group * (thread_index + 1);
2568         tinfo->et_group_next = tinfo->et_group_start;
2569         tinfo->et_inode_number = 0;
2570         tinfo->et_log_buf[0] = '\0';
2571         tinfo->et_log_length = 0;
2572         if (thread_context->options & E2F_OPT_MULTITHREAD)
2573                 log_out(thread_context, _("Scan group range [%d, %d)\n"),
2574                         tinfo->et_group_start, tinfo->et_group_end);
2575         thread_context->fs = thread_fs;
2576         retval = quota_init_context(&thread_context->qctx, thread_fs, 0);
2577         if (retval) {
2578                 com_err(global_ctx->program_name, retval,
2579                         "while init quota context");
2580                 goto out_fs;
2581         }
2582         *thread_ctx = thread_context;
2583         e2fsck_pass1_copy_invalid_bitmaps(global_ctx, thread_context);
2584         return 0;
2585 out_fs:
2586         ext2fs_free_mem(&thread_fs);
2587 out_context:
2588         ext2fs_free_mem(&thread_context);
2589         return retval;
2590 }
2591
2592 static void e2fsck_pass1_merge_dir_info(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2593 {
2594         if (thread_ctx->dir_info == NULL)
2595                 return;
2596
2597         if (global_ctx->dir_info == NULL) {
2598                 global_ctx->dir_info = thread_ctx->dir_info;
2599                 thread_ctx->dir_info = NULL;
2600                 return;
2601         }
2602
2603         e2fsck_merge_dir_info(global_ctx, thread_ctx->dir_info,
2604                               global_ctx->dir_info);
2605 }
2606
2607 static void e2fsck_pass1_merge_dx_dir(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2608 {
2609         if (thread_ctx->dx_dir_info == NULL)
2610                 return;
2611
2612         if (global_ctx->dx_dir_info == NULL) {
2613                 global_ctx->dx_dir_info = thread_ctx->dx_dir_info;
2614                 global_ctx->dx_dir_info_size = thread_ctx->dx_dir_info_size;
2615                 global_ctx->dx_dir_info_count = thread_ctx->dx_dir_info_count;
2616                 thread_ctx->dx_dir_info = NULL;
2617                 return;
2618         }
2619
2620         e2fsck_merge_dx_dir(global_ctx, thread_ctx);
2621 }
2622
2623 static inline errcode_t
2624 e2fsck_pass1_merge_icount(ext2_icount_t *dest_icount,
2625                           ext2_icount_t *src_icount)
2626 {
2627         if (*src_icount) {
2628                 if (*dest_icount == NULL) {
2629                         *dest_icount = *src_icount;
2630                         *src_icount = NULL;
2631                 } else {
2632                         errcode_t ret;
2633
2634                         ret = ext2fs_icount_merge(*src_icount,
2635                                                   *dest_icount);
2636                         if (ret)
2637                                 return ret;
2638                 }
2639         }
2640
2641         return 0;
2642 }
2643
2644 static errcode_t e2fsck_pass1_merge_icounts(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2645 {
2646         errcode_t ret;
2647
2648         ret = e2fsck_pass1_merge_icount(&global_ctx->inode_count,
2649                                         &thread_ctx->inode_count);
2650         if (ret)
2651                 return ret;
2652         ret = e2fsck_pass1_merge_icount(&global_ctx->inode_link_info,
2653                                         &thread_ctx->inode_link_info);
2654
2655         return ret;
2656 }
2657
2658 static errcode_t e2fsck_pass1_merge_dirs_to_hash(e2fsck_t global_ctx,
2659                                                  e2fsck_t thread_ctx)
2660 {
2661         errcode_t retval = 0;
2662
2663         if (!thread_ctx->dirs_to_hash)
2664                 return 0;
2665
2666         if (!global_ctx->dirs_to_hash)
2667                 retval = ext2fs_badblocks_copy(thread_ctx->dirs_to_hash,
2668                                                &global_ctx->dirs_to_hash);
2669         else
2670                 retval = ext2fs_badblocks_merge(thread_ctx->dirs_to_hash,
2671                                                 global_ctx->dirs_to_hash);
2672
2673         return retval;
2674 }
2675
2676 static int e2fsck_pass1_thread_join_one(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2677 {
2678         errcode_t        retval;
2679         int              flags = global_ctx->flags;
2680         ext2_filsys      thread_fs = thread_ctx->fs;
2681         ext2_filsys      global_fs = global_ctx->fs;
2682         FILE            *global_logf = global_ctx->logf;
2683         FILE            *global_problem_logf = global_ctx->problem_logf;
2684         ext2fs_inode_bitmap inode_bad_map = global_ctx->inode_bad_map;
2685         struct dir_info_db *dir_info = global_ctx->dir_info;
2686         struct dx_dir_info *dx_dir_info = global_ctx->dx_dir_info;
2687         ext2fs_inode_bitmap inode_used_map = global_ctx->inode_used_map;
2688         ext2fs_inode_bitmap inode_dir_map = global_ctx->inode_dir_map;
2689         ext2fs_inode_bitmap inode_bb_map = global_ctx->inode_bb_map;
2690         ext2fs_inode_bitmap inode_imagic_map = global_ctx->inode_imagic_map;
2691         ext2fs_inode_bitmap inode_reg_map = global_ctx->inode_reg_map;
2692         ext2fs_block_bitmap block_dup_map = global_ctx->block_dup_map;
2693         ext2fs_block_bitmap block_ea_map = global_ctx->block_ea_map;
2694         ext2fs_block_bitmap inodes_to_rebuild = global_ctx->inodes_to_rebuild;
2695         ext2_icount_t inode_count = global_ctx->inode_count;
2696         ext2_icount_t inode_link_info = global_ctx->inode_link_info;
2697         __u32 fs_directory_count = global_ctx->fs_directory_count;
2698         __u32 fs_regular_count = global_ctx->fs_regular_count;
2699         __u32 fs_blockdev_count = global_ctx->fs_blockdev_count;
2700         __u32 fs_chardev_count = global_ctx->fs_chardev_count;
2701         __u32 fs_links_count = global_ctx->fs_links_count;
2702         __u32 fs_symlinks_count = global_ctx->fs_symlinks_count;
2703         __u32 fs_fast_symlinks_count = global_ctx->fs_fast_symlinks_count;
2704         __u32 fs_fifo_count = global_ctx->fs_fifo_count;
2705         __u32 fs_total_count = global_ctx->fs_total_count;
2706         __u32 fs_badblocks_count = global_ctx->fs_badblocks_count;
2707         __u32 fs_sockets_count = global_ctx->fs_sockets_count;
2708         __u32 fs_ind_count = global_ctx->fs_ind_count;
2709         __u32 fs_dind_count = global_ctx->fs_dind_count;
2710         __u32 fs_tind_count = global_ctx->fs_tind_count;
2711         __u32 fs_fragmented = global_ctx->fs_fragmented;
2712         __u32 fs_fragmented_dir = global_ctx->fs_fragmented_dir;
2713         __u32 large_files = global_ctx->large_files;
2714         ext2_ino_t dx_dir_info_size = global_ctx->dx_dir_info_size;
2715         ext2_ino_t dx_dir_info_count = global_ctx->dx_dir_info_count;
2716         ext2_u32_list dirs_to_hash = global_ctx->dirs_to_hash;
2717         quota_ctx_t qctx = global_ctx->qctx;
2718         int *invalid_block_bitmap_flag = global_ctx->invalid_block_bitmap_flag;
2719         int *invalid_inode_bitmap_flag = global_ctx->invalid_inode_bitmap_flag;
2720         int *invalid_inode_table_flag  = global_ctx->invalid_inode_table_flag;
2721         int invalid_bitmaps = global_ctx->invalid_bitmaps;
2722
2723 #ifdef HAVE_SETJMP_H
2724         jmp_buf          old_jmp;
2725
2726         memcpy(old_jmp, global_ctx->abort_loc, sizeof(jmp_buf));
2727 #endif
2728         memcpy(global_ctx, thread_ctx, sizeof(struct e2fsck_struct));
2729 #ifdef HAVE_SETJMP_H
2730         memcpy(global_ctx->abort_loc, old_jmp, sizeof(jmp_buf));
2731 #endif
2732
2733         global_ctx->inode_used_map = inode_used_map;
2734         global_ctx->inode_bad_map = inode_bad_map;
2735         global_ctx->inode_dir_map = inode_dir_map;
2736         global_ctx->inode_bb_map = inode_bb_map;
2737         global_ctx->inode_imagic_map = inode_imagic_map;
2738         global_ctx->inodes_to_rebuild = inodes_to_rebuild;
2739         global_ctx->inode_reg_map = inode_reg_map;
2740         global_ctx->block_ea_map = block_ea_map;
2741         global_ctx->block_dup_map = block_dup_map;
2742         global_ctx->dir_info = dir_info;
2743         e2fsck_pass1_merge_dir_info(global_ctx, thread_ctx);
2744         global_ctx->dx_dir_info = dx_dir_info;
2745         global_ctx->dx_dir_info_count = dx_dir_info_count;
2746         global_ctx->dx_dir_info_size = dx_dir_info_size;
2747         e2fsck_pass1_merge_dx_dir(global_ctx, thread_ctx);
2748         global_ctx->inode_count = inode_count;
2749         global_ctx->inode_link_info = inode_link_info;
2750         global_ctx->fs_directory_count += fs_directory_count;
2751         global_ctx->fs_regular_count += fs_regular_count;
2752         global_ctx->fs_blockdev_count += fs_blockdev_count;
2753         global_ctx->fs_chardev_count += fs_chardev_count;
2754         global_ctx->fs_links_count += fs_links_count;
2755         global_ctx->fs_symlinks_count += fs_symlinks_count;
2756         global_ctx->fs_fast_symlinks_count += fs_fast_symlinks_count;
2757         global_ctx->fs_fifo_count += fs_fifo_count;
2758         global_ctx->fs_total_count += fs_total_count;
2759         global_ctx->fs_badblocks_count += fs_badblocks_count;
2760         global_ctx->fs_sockets_count += fs_sockets_count;
2761         global_ctx->fs_ind_count += fs_ind_count;
2762         global_ctx->fs_dind_count += fs_dind_count;
2763         global_ctx->fs_tind_count += fs_tind_count;
2764         global_ctx->fs_fragmented += fs_fragmented;
2765         global_ctx->fs_fragmented_dir += fs_fragmented_dir;
2766         global_ctx->large_files += large_files;
2767
2768         global_ctx->flags |= flags;
2769
2770         retval = e2fsck_pass1_merge_fs(global_fs, thread_fs);
2771         if (retval) {
2772                 com_err(global_ctx->program_name, 0, _("while merging fs\n"));
2773                 return retval;
2774         }
2775         global_fs->priv_data = global_ctx;
2776         global_ctx->fs = global_fs;
2777         global_ctx->logf = global_logf;
2778         global_ctx->problem_logf = global_problem_logf;
2779         global_ctx->global_ctx = NULL;
2780         retval = e2fsck_pass1_merge_icounts(global_ctx, thread_ctx);
2781         if (retval) {
2782                 com_err(global_ctx->program_name, 0,
2783                         _("while merging icounts\n"));
2784                 return retval;
2785         }
2786
2787         global_ctx->dirs_to_hash = dirs_to_hash;
2788         retval = e2fsck_pass1_merge_dirs_to_hash(global_ctx, thread_ctx);
2789         if (retval) {
2790                 com_err(global_ctx->program_name, 0,
2791                         _("while merging dirs to hash\n"));
2792                 return retval;
2793         }
2794
2795         global_ctx->qctx = qctx;
2796         retval = quota_merge_and_update_usage(global_ctx->qctx,
2797                                               thread_ctx->qctx);
2798         if (retval)
2799                 return retval;
2800         global_ctx->invalid_block_bitmap_flag = invalid_block_bitmap_flag;
2801         global_ctx->invalid_inode_bitmap_flag = invalid_inode_bitmap_flag;
2802         global_ctx->invalid_inode_table_flag = invalid_inode_table_flag;
2803         global_ctx->invalid_bitmaps = invalid_bitmaps;
2804         e2fsck_pass1_merge_invalid_bitmaps(global_ctx, thread_ctx);
2805
2806         retval = e2fsck_pass1_merge_bitmap(global_fs,
2807                                 &thread_ctx->inode_used_map,
2808                                 &global_ctx->inode_used_map);
2809         if (retval)
2810                 return retval;
2811
2812         retval = e2fsck_pass1_merge_bitmap(global_fs,
2813                                 &thread_ctx->inode_bad_map,
2814                                 &global_ctx->inode_bad_map);
2815         if (retval)
2816                 return retval;
2817         retval = e2fsck_pass1_merge_bitmap(global_fs,
2818                                         &thread_ctx->inode_dir_map,
2819                                         &global_ctx->inode_dir_map);
2820         if (retval)
2821                 return retval;
2822         retval = e2fsck_pass1_merge_bitmap(global_fs,
2823                                 &thread_ctx->inode_bb_map,
2824                                 &global_ctx->inode_bb_map);
2825         if (retval)
2826                 return retval;
2827         retval = e2fsck_pass1_merge_bitmap(global_fs,
2828                                 &thread_ctx->inode_imagic_map,
2829                                 &global_ctx->inode_imagic_map);
2830         if (retval)
2831                 return retval;
2832         retval = e2fsck_pass1_merge_bitmap(global_fs,
2833                                 &thread_ctx->inode_reg_map,
2834                                 &global_ctx->inode_reg_map);
2835         if (retval)
2836                 return retval;
2837         retval = e2fsck_pass1_merge_bitmap(global_fs,
2838                                 &thread_ctx->inodes_to_rebuild,
2839                                 &global_ctx->inodes_to_rebuild);
2840         if (retval)
2841                 return retval;
2842         retval = e2fsck_pass1_merge_bitmap(global_fs,
2843                                 &thread_ctx->block_ea_map,
2844                                 &global_ctx->block_ea_map);
2845         if (retval)
2846                 return retval;
2847
2848         return 0;
2849 }
2850
2851 static int e2fsck_pass1_thread_join(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2852 {
2853         errcode_t       retval;
2854
2855         retval = e2fsck_pass1_thread_join_one(global_ctx, thread_ctx);
2856         ext2fs_free_mem(&thread_ctx->fs);
2857         if (thread_ctx->logf)
2858                 fclose(thread_ctx->logf);
2859         if (thread_ctx->problem_logf) {
2860                 fputs("</problem_log>\n", thread_ctx->problem_logf);
2861                 fclose(thread_ctx->problem_logf);
2862         }
2863         e2fsck_pass1_free_bitmap(&thread_ctx->inode_used_map);
2864         e2fsck_pass1_free_bitmap(&thread_ctx->inode_bad_map);
2865         e2fsck_pass1_free_bitmap(&thread_ctx->inode_dir_map);
2866         e2fsck_pass1_free_bitmap(&thread_ctx->inode_bb_map);
2867         e2fsck_pass1_free_bitmap(&thread_ctx->inode_imagic_map);
2868         e2fsck_pass1_free_bitmap(&thread_ctx->inode_reg_map);
2869         e2fsck_pass1_free_bitmap(&thread_ctx->inodes_to_rebuild);
2870         e2fsck_pass1_free_bitmap(&thread_ctx->block_ea_map);
2871         e2fsck_free_dir_info(thread_ctx);
2872         ext2fs_free_icount(thread_ctx->inode_count);
2873         ext2fs_free_icount(thread_ctx->inode_link_info);
2874         if (thread_ctx->dirs_to_hash)
2875                 ext2fs_badblocks_list_free(thread_ctx->dirs_to_hash);
2876         quota_release_context(&thread_ctx->qctx);
2877         ext2fs_free_mem(&thread_ctx->invalid_block_bitmap_flag);
2878         ext2fs_free_mem(&thread_ctx->invalid_inode_bitmap_flag);
2879         ext2fs_free_mem(&thread_ctx->invalid_inode_table_flag);
2880         ext2fs_free_mem(&thread_ctx);
2881
2882         return retval;
2883 }
2884
2885 static int e2fsck_pass1_threads_join(struct e2fsck_thread_info *infos,
2886                                       int num_threads, e2fsck_t global_ctx)
2887 {
2888         errcode_t                        rc;
2889         errcode_t                        ret = 0;
2890         int                              i;
2891         struct e2fsck_thread_info       *pinfo;
2892
2893         /* merge invalid bitmaps will recalculate it */
2894         global_ctx->invalid_bitmaps = 0;
2895         for (i = 0; i < num_threads; i++) {
2896                 pinfo = &infos[i];
2897
2898                 if (!pinfo->eti_started)
2899                         continue;
2900
2901                 rc = pthread_join(pinfo->eti_thread_id, NULL);
2902                 if (rc) {
2903                         com_err(global_ctx->program_name, rc,
2904                                 _("while joining thread\n"));
2905                         if (ret == 0)
2906                                 ret = rc;
2907                 }
2908                 rc = e2fsck_pass1_thread_join(global_ctx, infos[i].eti_thread_ctx);
2909                 if (rc) {
2910                         com_err(global_ctx->program_name, rc,
2911                                 _("while joining pass1 thread\n"));
2912                         if (ret == 0)
2913                                 ret = rc;
2914                 }
2915         }
2916         free(infos);
2917
2918         return ret;
2919 }
2920
2921 static void *e2fsck_pass1_thread(void *arg)
2922 {
2923         struct e2fsck_thread_info       *info = arg;
2924         e2fsck_t                         thread_ctx = info->eti_thread_ctx;
2925 #ifdef DEBUG_THREADS
2926         struct e2fsck_thread_debug      *thread_debug = info->eti_debug;
2927 #endif
2928
2929 #ifdef DEBUG_THREADS
2930         pthread_mutex_lock(&thread_debug->etd_mutex);
2931         while (info->eti_thread_index > thread_debug->etd_finished_threads) {
2932                 pthread_cond_wait(&thread_debug->etd_cond,
2933                                   &thread_debug->etd_mutex);
2934         }
2935         pthread_mutex_unlock(&thread_debug->etd_mutex);
2936 #endif
2937
2938 #ifdef HAVE_SETJMP_H
2939         /*
2940          * When fatal_error() happens, jump to here. The thread
2941          * context's flags will be saved, but its abort_loc will
2942          * be overwritten by original jump buffer for the later
2943          * tests.
2944          */
2945         if (setjmp(thread_ctx->abort_loc)) {
2946                 thread_ctx->flags &= ~E2F_FLAG_SETJMP_OK;
2947                 goto out;
2948         }
2949         thread_ctx->flags |= E2F_FLAG_SETJMP_OK;
2950 #endif
2951
2952         e2fsck_pass1_run(thread_ctx);
2953
2954 out:
2955         if (thread_ctx->options & E2F_OPT_MULTITHREAD)
2956                 log_out(thread_ctx,
2957                         _("Scanned group range [%lu, %lu), inodes %lu\n"),
2958                         thread_ctx->thread_info.et_group_start,
2959                         thread_ctx->thread_info.et_group_end,
2960                         thread_ctx->thread_info.et_inode_number);
2961
2962 #ifdef DEBUG_THREADS
2963         pthread_mutex_lock(&thread_debug->etd_mutex);
2964         thread_debug->etd_finished_threads++;
2965         pthread_cond_broadcast(&thread_debug->etd_cond);
2966         pthread_mutex_unlock(&thread_debug->etd_mutex);
2967 #endif
2968
2969         return NULL;
2970 }
2971
2972 static int e2fsck_pass1_threads_start(struct e2fsck_thread_info **pinfo,
2973                                       int num_threads, e2fsck_t global_ctx)
2974 {
2975         struct e2fsck_thread_info       *infos;
2976         pthread_attr_t                   attr;
2977         errcode_t                        retval;
2978         errcode_t                        ret;
2979         struct e2fsck_thread_info       *tmp_pinfo;
2980         int                              i;
2981         e2fsck_t                         thread_ctx;
2982 #ifdef DEBUG_THREADS
2983         struct e2fsck_thread_debug       thread_debug =
2984                 {PTHREAD_MUTEX_INITIALIZER, PTHREAD_COND_INITIALIZER, 0};
2985
2986         thread_debug.etd_finished_threads = 0;
2987 #endif
2988
2989         retval = pthread_attr_init(&attr);
2990         if (retval) {
2991                 com_err(global_ctx->program_name, retval,
2992                         _("while setting pthread attribute\n"));
2993                 return retval;
2994         }
2995
2996         infos = calloc(num_threads, sizeof(struct e2fsck_thread_info));
2997         if (infos == NULL) {
2998                 retval = -ENOMEM;
2999                 com_err(global_ctx->program_name, retval,
3000                         _("while allocating memory for threads\n"));
3001                 pthread_attr_destroy(&attr);
3002                 return retval;
3003         }
3004
3005         for (i = 0; i < num_threads; i++) {
3006                 tmp_pinfo = &infos[i];
3007                 tmp_pinfo->eti_thread_index = i;
3008 #ifdef DEBUG_THREADS
3009                 tmp_pinfo->eti_debug = &thread_debug;
3010 #endif
3011                 retval = e2fsck_pass1_thread_prepare(global_ctx, &thread_ctx,
3012                                                      i, num_threads);
3013                 if (retval) {
3014                         com_err(global_ctx->program_name, retval,
3015                                 _("while preparing pass1 thread\n"));
3016                         break;
3017                 }
3018                 tmp_pinfo->eti_thread_ctx = thread_ctx;
3019
3020                 retval = pthread_create(&tmp_pinfo->eti_thread_id, &attr,
3021                                         &e2fsck_pass1_thread, tmp_pinfo);
3022                 if (retval) {
3023                         com_err(global_ctx->program_name, retval,
3024                                 _("while creating thread\n"));
3025                         e2fsck_pass1_thread_join(global_ctx, thread_ctx);
3026                         break;
3027                 }
3028
3029                 tmp_pinfo->eti_started = 1;
3030         }
3031
3032         /* destroy the thread attribute object, since it is no longer needed */
3033         ret = pthread_attr_destroy(&attr);
3034         if (ret) {
3035                 com_err(global_ctx->program_name, ret,
3036                         _("while destroying thread attribute\n"));
3037                 if (retval == 0)
3038                         retval = ret;
3039         }
3040
3041         if (retval) {
3042                 e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
3043                 return retval;
3044         }
3045         *pinfo = infos;
3046         return 0;
3047 }
3048
3049 static void e2fsck_pass1_multithread(e2fsck_t global_ctx)
3050 {
3051         struct e2fsck_thread_info       *infos = NULL;
3052         int                              num_threads = 1;
3053         errcode_t                        retval;
3054
3055         pthread_mutex_init(&global_ctx->fs_fix_mutex, NULL);
3056         pthread_mutex_init(&global_ctx->fs_block_map_mutex, NULL);
3057         retval = e2fsck_pass1_threads_start(&infos, num_threads, global_ctx);
3058         if (retval) {
3059                 com_err(global_ctx->program_name, retval,
3060                         _("while starting pass1 threads\n"));
3061                 goto out_abort;
3062         }
3063
3064         retval = e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
3065         if (retval) {
3066                 com_err(global_ctx->program_name, retval,
3067                         _("while joining pass1 threads\n"));
3068                 goto out_abort;
3069         }
3070         return;
3071 out_abort:
3072         global_ctx->flags |= E2F_FLAG_ABORT;
3073         return;
3074 }
3075 #endif
3076
3077 /* TODO: tdb needs to be handled properly for multiple threads*/
3078 static int multiple_threads_supported(e2fsck_t ctx)
3079 {
3080 #ifdef  CONFIG_TDB
3081         unsigned int            threshold;
3082         ext2_ino_t              num_dirs;
3083         errcode_t               retval;
3084         char                    *tdb_dir;
3085         int                     enable;
3086
3087         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
3088                            &tdb_dir);
3089         profile_get_uint(ctx->profile, "scratch_files",
3090                          "numdirs_threshold", 0, 0, &threshold);
3091         profile_get_boolean(ctx->profile, "scratch_files",
3092                             "icount", 0, 1, &enable);
3093
3094         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
3095         if (retval)
3096                 num_dirs = 1024;        /* Guess */
3097
3098         /* tdb is unsupported now */
3099         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
3100             (!threshold || num_dirs > threshold))
3101                 return 0;
3102  #endif
3103         return 1;
3104 }
3105
3106 void e2fsck_pass1(e2fsck_t ctx)
3107 {
3108         errcode_t retval;
3109         int multiple = 0;
3110
3111         init_ext2_max_sizes();
3112         retval = e2fsck_pass1_prepare(ctx);
3113         if (retval)
3114                 return;
3115 #ifdef HAVE_PTHREAD
3116         if (multiple_threads_supported(ctx)) {
3117                 multiple = 1;
3118                 e2fsck_pass1_multithread(ctx);
3119         } else {
3120                 fprintf(stderr, "Fall through single thread for pass1 "
3121                                 "because tdb could not handle properly\n");
3122         }
3123 #endif
3124         if (!multiple)
3125                 e2fsck_pass1_run(ctx);
3126         e2fsck_pass1_post(ctx);
3127 }
3128
3129 #undef FINISH_INODE_LOOP
3130
3131 /*
3132  * When the inode_scan routines call this callback at the end of the
3133  * glock group, call process_inodes.
3134  */
3135 static errcode_t scan_callback(ext2_filsys fs,
3136                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
3137                                dgrp_t group, void * priv_data)
3138 {
3139         struct scan_callback_struct *scan_struct;
3140         e2fsck_t ctx;
3141         struct e2fsck_thread *tinfo;
3142
3143         scan_struct = (struct scan_callback_struct *) priv_data;
3144         ctx = scan_struct->ctx;
3145
3146         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf,
3147                        scan_struct->inodes_to_process,
3148                        scan_struct->process_inode_count);
3149
3150         if (ctx->progress)
3151                 if ((ctx->progress)(ctx, 1, group+1,
3152                                     ctx->fs->group_desc_count))
3153                         return EXT2_ET_CANCEL_REQUESTED;
3154
3155 #ifdef HAVE_PTHREAD
3156         if (ctx->global_ctx) {
3157                 tinfo = &ctx->thread_info;
3158                 tinfo->et_group_next++;
3159                 if (ctx->options & E2F_OPT_DEBUG &&
3160                     ctx->options & E2F_OPT_MULTITHREAD)
3161                         log_out(ctx, _("group %d finished\n"),
3162                                 tinfo->et_group_next);
3163                 if (tinfo->et_group_next >= tinfo->et_group_end)
3164                         return EXT2_ET_SCAN_FINISHED;
3165         }
3166 #endif
3167
3168         return 0;
3169 }
3170
3171 /*
3172  * Process the inodes in the "inodes to process" list.
3173  */
3174 static void process_inodes(e2fsck_t ctx, char *block_buf,
3175                            struct process_inode_block *inodes_to_process,
3176                            int *process_inode_count)
3177 {
3178         int                     i;
3179         struct ext2_inode       *old_stashed_inode;
3180         ext2_ino_t              old_stashed_ino;
3181         const char              *old_operation;
3182         char                    buf[80];
3183         struct problem_context  pctx;
3184
3185 #if 0
3186         printf("begin process_inodes: ");
3187 #endif
3188         if (*process_inode_count == 0)
3189                 return;
3190         old_operation = ehandler_operation(0);
3191         old_stashed_inode = ctx->stashed_inode;
3192         old_stashed_ino = ctx->stashed_ino;
3193         qsort(inodes_to_process, *process_inode_count,
3194                       sizeof(struct process_inode_block), process_inode_cmp);
3195         clear_problem_context(&pctx);
3196         for (i=0; i < *process_inode_count; i++) {
3197                 pctx.inode = ctx->stashed_inode =
3198                         (struct ext2_inode *) &inodes_to_process[i].inode;
3199                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
3200
3201 #if 0
3202                 printf("%u ", pctx.ino);
3203 #endif
3204                 sprintf(buf, _("reading indirect blocks of inode %u"),
3205                         pctx.ino);
3206                 ehandler_operation(buf);
3207                 check_blocks(ctx, &pctx, block_buf,
3208                              &inodes_to_process[i].ea_ibody_quota);
3209                 if (e2fsck_should_abort(ctx))
3210                         break;
3211         }
3212         ctx->stashed_inode = old_stashed_inode;
3213         ctx->stashed_ino = old_stashed_ino;
3214         *process_inode_count = 0;
3215 #if 0
3216         printf("end process inodes\n");
3217 #endif
3218         ehandler_operation(old_operation);
3219 }
3220
3221 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
3222 {
3223         const struct process_inode_block *ib_a =
3224                 (const struct process_inode_block *) a;
3225         const struct process_inode_block *ib_b =
3226                 (const struct process_inode_block *) b;
3227         int     ret;
3228
3229         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
3230                ib_b->inode.i_block[EXT2_IND_BLOCK]);
3231         if (ret == 0)
3232                 /*
3233                  * We only call process_inodes() for non-extent
3234                  * inodes, so it's OK to pass NULL to
3235                  * ext2fs_file_acl_block() here.
3236                  */
3237                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
3238                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
3239         if (ret == 0)
3240                 ret = ib_a->ino - ib_b->ino;
3241         return ret;
3242 }
3243
3244 /*
3245  * Mark an inode as being bad in some what
3246  */
3247 static void mark_inode_bad(e2fsck_t ctx, ext2_ino_t ino)
3248 {
3249         struct          problem_context pctx;
3250
3251         if (!ctx->inode_bad_map) {
3252                 clear_problem_context(&pctx);
3253
3254                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3255                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
3256                                 "inode_bad_map", &ctx->inode_bad_map);
3257                 if (pctx.errcode) {
3258                         pctx.num = 3;
3259                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3260                         /* Should never get here */
3261                         ctx->flags |= E2F_FLAG_ABORT;
3262                         return;
3263                 }
3264         }
3265         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
3266 }
3267
3268 static void add_casefolded_dir(e2fsck_t ctx, ext2_ino_t ino)
3269 {
3270         struct          problem_context pctx;
3271
3272         if (!ctx->casefolded_dirs) {
3273                 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
3274                 if (pctx.errcode)
3275                         goto error;
3276         }
3277         pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
3278         if (pctx.errcode == 0)
3279                 return;
3280 error:
3281         fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
3282         /* Should never get here */
3283         ctx->flags |= E2F_FLAG_ABORT;
3284 }
3285
3286 /*
3287  * This procedure will allocate the inode "bb" (badblock) map table
3288  */
3289 static void alloc_bb_map(e2fsck_t ctx)
3290 {
3291         struct          problem_context pctx;
3292
3293         clear_problem_context(&pctx);
3294         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3295                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
3296                         "inode_bb_map", &ctx->inode_bb_map);
3297         if (pctx.errcode) {
3298                 pctx.num = 4;
3299                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3300                 /* Should never get here */
3301                 ctx->flags |= E2F_FLAG_ABORT;
3302                 return;
3303         }
3304 }
3305
3306 /*
3307  * This procedure will allocate the inode imagic table
3308  */
3309 static void alloc_imagic_map(e2fsck_t ctx)
3310 {
3311         struct          problem_context pctx;
3312
3313         clear_problem_context(&pctx);
3314         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3315                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
3316                         "inode_imagic_map", &ctx->inode_imagic_map);
3317         if (pctx.errcode) {
3318                 pctx.num = 5;
3319                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3320                 /* Should never get here */
3321                 ctx->flags |= E2F_FLAG_ABORT;
3322                 return;
3323         }
3324 }
3325
3326 /*
3327  * Marks a block as in use, setting the dup_map if it's been set
3328  * already.  Called by process_block and process_bad_block.
3329  *
3330  * WARNING: Assumes checks have already been done to make sure block
3331  * is valid.  This is true in both process_block and process_bad_block.
3332  */
3333 static _INLINE_ void mark_block_used_unlocked(e2fsck_t ctx, blk64_t block)
3334 {
3335         struct problem_context pctx;
3336         e2fsck_t global_ctx;
3337
3338         global_ctx = ctx->global_ctx;
3339         if (!global_ctx)
3340                 global_ctx = ctx;
3341
3342         clear_problem_context(&pctx);
3343
3344         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
3345                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
3346                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
3347                         return;
3348                 }
3349                 /**
3350                  * this should be safe because this operation has
3351                  * been serialized by mutex.
3352                  */
3353                 if (!global_ctx->block_dup_map) {
3354                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
3355                                         _("multiply claimed block map"),
3356                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
3357                                         &global_ctx->block_dup_map);
3358                         if (pctx.errcode) {
3359                                 pctx.num = 3;
3360                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
3361                                             &pctx);
3362                                 /* Should never get here */
3363                                 ctx->flags |= E2F_FLAG_ABORT;
3364                                 return;
3365                         }
3366                 }
3367                 ext2fs_fast_mark_block_bitmap2(global_ctx->block_dup_map, block);
3368         } else {
3369                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
3370         }
3371 }
3372
3373 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
3374 {
3375         e2fsck_pass1_block_map_lock(ctx);
3376         mark_block_used_unlocked(ctx, block);
3377         e2fsck_pass1_block_map_unlock(ctx);
3378
3379 }
3380
3381 /*
3382  * When cluster size is greater than one block, it is caller's responsibility
3383  * to make sure block parameter starts at a cluster boundary.
3384  */
3385 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
3386                                       unsigned int num)
3387 {
3388         e2fsck_pass1_block_map_lock(ctx);
3389         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num)) {
3390                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
3391         } else {
3392                 unsigned int i;
3393
3394                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
3395                         mark_block_used_unlocked(ctx, block + i);
3396         }
3397         e2fsck_pass1_block_map_unlock(ctx);
3398 }
3399
3400 static errcode_t _INLINE_ e2fsck_write_ext_attr3(e2fsck_t ctx, blk64_t block,
3401                                                  void *inbuf, ext2_ino_t inum)
3402 {
3403         errcode_t retval;
3404         ext2_filsys fs = ctx->fs;
3405
3406         e2fsck_pass1_fix_lock(ctx);
3407         retval = ext2fs_write_ext_attr3(fs, block, inbuf, inum);
3408         e2fsck_pass1_fix_unlock(ctx);
3409
3410         return retval;
3411 }
3412 /*
3413  * Adjust the extended attribute block's reference counts at the end
3414  * of pass 1, either by subtracting out references for EA blocks that
3415  * are still referenced in ctx->refcount, or by adding references for
3416  * EA blocks that had extra references as accounted for in
3417  * ctx->refcount_extra.
3418  */
3419 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
3420                                     char *block_buf, int adjust_sign)
3421 {
3422         struct ext2_ext_attr_header     *header;
3423         struct problem_context          pctx;
3424         ext2_filsys                     fs = ctx->fs;
3425         blk64_t                         blk;
3426         __u32                           should_be;
3427         ea_value_t                      count;
3428
3429         clear_problem_context(&pctx);
3430
3431         ea_refcount_intr_begin(refcount);
3432         while (1) {
3433                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
3434                         break;
3435                 pctx.blk = blk;
3436                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
3437                                                      pctx.ino);
3438                 if (pctx.errcode) {
3439                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
3440                         return;
3441                 }
3442                 header = (struct ext2_ext_attr_header *) block_buf;
3443                 pctx.blkcount = header->h_refcount;
3444                 should_be = header->h_refcount + adjust_sign * (int)count;
3445                 pctx.num = should_be;
3446                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
3447                         header->h_refcount = should_be;
3448                         pctx.errcode = e2fsck_write_ext_attr3(ctx, blk,
3449                                                              block_buf,
3450                                                              pctx.ino);
3451                         if (pctx.errcode) {
3452                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
3453                                             &pctx);
3454                                 continue;
3455                         }
3456                 }
3457         }
3458 }
3459
3460 /*
3461  * Handle processing the extended attribute blocks
3462  */
3463 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
3464                            char *block_buf, struct ea_quota *ea_block_quota)
3465 {
3466         ext2_filsys fs = ctx->fs;
3467         ext2_ino_t      ino = pctx->ino;
3468         struct ext2_inode *inode = pctx->inode;
3469         blk64_t         blk;
3470         char *          end;
3471         struct ext2_ext_attr_header *header;
3472         struct ext2_ext_attr_entry *first, *entry;
3473         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
3474         __u64           quota_inodes = 0;
3475         region_t        region = 0;
3476         int             failed_csum = 0;
3477
3478         ea_block_quota->blocks = 0;
3479         ea_block_quota->inodes = 0;
3480
3481         blk = ext2fs_file_acl_block(fs, inode);
3482         if (blk == 0)
3483                 return 0;
3484
3485         /*
3486          * If the Extended attribute flag isn't set, then a non-zero
3487          * file acl means that the inode is corrupted.
3488          *
3489          * Or if the extended attribute block is an invalid block,
3490          * then the inode is also corrupted.
3491          */
3492         if (!ext2fs_has_feature_xattr(fs->super) ||
3493             (blk < fs->super->s_first_data_block) ||
3494             (blk >= ext2fs_blocks_count(fs->super))) {
3495                 mark_inode_bad(ctx, ino);
3496                 return 0;
3497         }
3498
3499         /* If ea bitmap hasn't been allocated, create it */
3500         if (!ctx->block_ea_map) {
3501                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
3502                                         _("ext attr block map"),
3503                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
3504                                         &ctx->block_ea_map);
3505                 if (pctx->errcode) {
3506                         pctx->num = 2;
3507                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
3508                         ctx->flags |= E2F_FLAG_ABORT;
3509                         return 0;
3510                 }
3511         }
3512
3513         /* Create the EA refcount structure if necessary */
3514         if (!ctx->refcount) {
3515                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
3516                 if (pctx->errcode) {
3517                         pctx->num = 1;
3518                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3519                         ctx->flags |= E2F_FLAG_ABORT;
3520                         return 0;
3521                 }
3522         }
3523
3524 #if 0
3525         /* Debugging text */
3526         printf("Inode %u has EA block %u\n", ino, blk);
3527 #endif
3528
3529         /* Have we seen this EA block before? */
3530         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
3531                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
3532                 ea_block_quota->inodes = 0;
3533
3534                 if (ctx->ea_block_quota_blocks) {
3535                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
3536                                           &quota_blocks);
3537                         if (quota_blocks)
3538                                 ea_block_quota->blocks = quota_blocks;
3539                 }
3540
3541                 if (ctx->ea_block_quota_inodes)
3542                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
3543                                           &ea_block_quota->inodes);
3544
3545                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
3546                         return 1;
3547                 /* Ooops, this EA was referenced more than it stated */
3548                 if (!ctx->refcount_extra) {
3549                         pctx->errcode = ea_refcount_create(0,
3550                                            &ctx->refcount_extra);
3551                         if (pctx->errcode) {
3552                                 pctx->num = 2;
3553                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3554                                 ctx->flags |= E2F_FLAG_ABORT;
3555                                 return 0;
3556                         }
3557                 }
3558                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
3559                 return 1;
3560         }
3561
3562         /*
3563          * OK, we haven't seen this EA block yet.  So we need to
3564          * validate it
3565          */
3566         pctx->blk = blk;
3567         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
3568         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
3569                 pctx->errcode = 0;
3570                 failed_csum = 1;
3571         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
3572                 pctx->errcode = 0;
3573
3574         if (pctx->errcode &&
3575             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
3576                 pctx->errcode = 0;
3577                 goto clear_extattr;
3578         }
3579         header = (struct ext2_ext_attr_header *) block_buf;
3580         pctx->blk = ext2fs_file_acl_block(fs, inode);
3581         if (((ctx->ext_attr_ver == 1) &&
3582              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
3583             ((ctx->ext_attr_ver == 2) &&
3584              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
3585                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
3586                         goto clear_extattr;
3587         }
3588
3589         if (header->h_blocks != 1) {
3590                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
3591                         goto clear_extattr;
3592         }
3593
3594         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
3595                 goto clear_extattr;
3596
3597         region = region_create(0, fs->blocksize);
3598         if (!region) {
3599                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
3600                 ctx->flags |= E2F_FLAG_ABORT;
3601                 return 0;
3602         }
3603         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
3604                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3605                         goto clear_extattr;
3606         }
3607
3608         first = (struct ext2_ext_attr_entry *)(header+1);
3609         end = block_buf + fs->blocksize;
3610         entry = first;
3611         while ((char *)entry < end && *(__u32 *)entry) {
3612                 __u32 hash;
3613
3614                 if (region_allocate(region, (char *)entry - (char *)header,
3615                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
3616                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3617                                 goto clear_extattr;
3618                         break;
3619                 }
3620                 if ((ctx->ext_attr_ver == 1 &&
3621                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
3622                     (ctx->ext_attr_ver == 2 &&
3623                      entry->e_name_index == 0)) {
3624                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
3625                                 goto clear_extattr;
3626                         break;
3627                 }
3628                 if (entry->e_value_inum == 0) {
3629                         if (entry->e_value_size > EXT2_XATTR_SIZE_MAX ||
3630                             (entry->e_value_offs + entry->e_value_size >
3631                              fs->blocksize)) {
3632                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
3633                                         goto clear_extattr;
3634                                 break;
3635                         }
3636                         if (entry->e_value_size &&
3637                             region_allocate(region, entry->e_value_offs,
3638                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
3639                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
3640                                                 pctx))
3641                                         goto clear_extattr;
3642                         }
3643
3644                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
3645                                                           entry->e_value_offs);
3646                         if (entry->e_hash != hash)
3647                                 hash = ext2fs_ext_attr_hash_entry_signed(entry,
3648                                         block_buf + entry->e_value_offs);
3649
3650                         if (entry->e_hash != hash) {
3651                                 pctx->num = entry->e_hash;
3652                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
3653                                         goto clear_extattr;
3654                                 entry->e_hash = hash;
3655                         }
3656                 } else {
3657                         problem_t problem;
3658                         blk64_t entry_quota_blocks;
3659
3660                         problem = check_large_ea_inode(ctx, entry, pctx,
3661                                                        &entry_quota_blocks);
3662                         if (problem && fix_problem(ctx, problem, pctx))
3663                                 goto clear_extattr;
3664
3665                         quota_blocks += entry_quota_blocks;
3666                         quota_inodes++;
3667                 }
3668
3669                 entry = EXT2_EXT_ATTR_NEXT(entry);
3670         }
3671         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
3672                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3673                         goto clear_extattr;
3674         }
3675         region_free(region);
3676
3677         /*
3678          * We only get here if there was no other errors that were fixed.
3679          * If there was a checksum fail, ask to correct it.
3680          */
3681         if (failed_csum &&
3682             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
3683                 pctx->errcode = e2fsck_write_ext_attr3(ctx, blk, block_buf,
3684                                                        pctx->ino);
3685                 if (pctx->errcode)
3686                         return 0;
3687         }
3688
3689         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
3690                 if (!ctx->ea_block_quota_blocks) {
3691                         pctx->errcode = ea_refcount_create(0,
3692                                                 &ctx->ea_block_quota_blocks);
3693                         if (pctx->errcode) {
3694                                 pctx->num = 3;
3695                                 goto refcount_fail;
3696                         }
3697                 }
3698                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
3699                                   quota_blocks);
3700         }
3701
3702         if (quota_inodes) {
3703                 if (!ctx->ea_block_quota_inodes) {
3704                         pctx->errcode = ea_refcount_create(0,
3705                                                 &ctx->ea_block_quota_inodes);
3706                         if (pctx->errcode) {
3707                                 pctx->num = 4;
3708 refcount_fail:
3709                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3710                                 ctx->flags |= E2F_FLAG_ABORT;
3711                                 return 0;
3712                         }
3713                 }
3714
3715                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
3716                                   quota_inodes);
3717         }
3718         ea_block_quota->blocks = quota_blocks;
3719         ea_block_quota->inodes = quota_inodes;
3720
3721         inc_ea_inode_refs(ctx, pctx, first, end);
3722         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
3723         mark_block_used(ctx, blk);
3724         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
3725         return 1;
3726
3727 clear_extattr:
3728         if (region)
3729                 region_free(region);
3730         ext2fs_file_acl_block_set(fs, inode, 0);
3731         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
3732         return 0;
3733 }
3734
3735 /* Returns 1 if bad htree, 0 if OK */
3736 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
3737                         ext2_ino_t ino, struct ext2_inode *inode,
3738                         char *block_buf)
3739 {
3740         struct ext2_dx_root_info        *root;
3741         ext2_filsys                     fs = ctx->fs;
3742         errcode_t                       retval;
3743         blk64_t                         blk;
3744
3745         if ((!LINUX_S_ISDIR(inode->i_mode) &&
3746              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
3747             (!ext2fs_has_feature_dir_index(fs->super) &&
3748              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
3749                 return 1;
3750
3751         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
3752
3753         if ((pctx->errcode) ||
3754             (blk == 0) ||
3755             (blk < fs->super->s_first_data_block) ||
3756             (blk >= ext2fs_blocks_count(fs->super))) {
3757                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3758                         return 1;
3759                 else
3760                         return 0;
3761         }
3762
3763         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
3764         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3765                 return 1;
3766
3767         /* XXX should check that beginning matches a directory */
3768         root = (struct ext2_dx_root_info *) (block_buf + 24);
3769
3770         if ((root->reserved_zero || root->info_length < 8) &&
3771             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3772                 return 1;
3773
3774         pctx->num = root->hash_version;
3775         if ((root->hash_version != EXT2_HASH_LEGACY) &&
3776             (root->hash_version != EXT2_HASH_HALF_MD4) &&
3777             (root->hash_version != EXT2_HASH_TEA) &&
3778             (root->hash_version != EXT2_HASH_SIPHASH) &&
3779             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
3780                 return 1;
3781
3782         if (ext4_hash_in_dirent(inode)) {
3783                 if (root->hash_version != EXT2_HASH_SIPHASH &&
3784                     fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
3785                         return 1;
3786         } else {
3787                 if (root->hash_version == EXT2_HASH_SIPHASH &&
3788                    fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
3789                         return 1;
3790         }
3791
3792         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
3793             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
3794                 return 1;
3795
3796         pctx->num = root->indirect_levels;
3797         /* if htree level is clearly too high, consider it to be broken */
3798         if (root->indirect_levels > EXT4_HTREE_LEVEL &&
3799             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3800                 return 1;
3801
3802         /* if level is only maybe too high, LARGE_DIR feature could be unset */
3803         if (root->indirect_levels > ext2_dir_htree_level(fs) &&
3804             !ext2fs_has_feature_largedir(fs->super)) {
3805                 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
3806                 unsigned idx_pb = 1 << (blockbits - 3);
3807
3808                 /* compare inode size/blocks vs. max-sized 2-level htree */
3809                 if (EXT2_I_SIZE(pctx->inode) <
3810                     (idx_pb - 1) * (idx_pb - 2) << blockbits &&
3811                     pctx->inode->i_blocks <
3812                     (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
3813                     fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3814                         return 1;
3815         }
3816
3817         if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
3818             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3819                 ctx->large_dirs++;
3820
3821         return 0;
3822 }
3823
3824 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
3825                         struct ext2_inode *inode, int restart_flag,
3826                         const char *source)
3827 {
3828         inode->i_flags = 0;
3829         inode->i_links_count = 0;
3830         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
3831         inode->i_dtime = ctx->now;
3832
3833         /*
3834          * If a special inode has such rotten block mappings that we
3835          * want to clear the whole inode, be sure to actually zap
3836          * the block maps because i_links_count isn't checked for
3837          * special inodes, and we'll end up right back here the next
3838          * time we run fsck.
3839          */
3840         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
3841                 memset(inode->i_block, 0, sizeof(inode->i_block));
3842
3843         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
3844         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
3845         if (ctx->inode_reg_map)
3846                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
3847         if (ctx->inode_bad_map)
3848                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
3849
3850         /*
3851          * If the inode was partially accounted for before processing
3852          * was aborted, we need to restart the pass 1 scan.
3853          */
3854         ctx->flags |= restart_flag;
3855
3856         if (ino == EXT2_BAD_INO)
3857                 memset(inode, 0, sizeof(struct ext2_inode));
3858
3859         e2fsck_write_inode(ctx, ino, inode, source);
3860 }
3861
3862 /*
3863  * Use the multiple-blocks reclamation code to fix alignment problems in
3864  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
3865  * physical cluster, and we want the block offsets within that cluster to
3866  * line up.
3867  */
3868 static int has_unaligned_cluster_map(e2fsck_t ctx,
3869                                      blk64_t last_pblk, blk64_t last_lblk,
3870                                      blk64_t pblk, blk64_t lblk)
3871 {
3872         blk64_t cluster_mask;
3873
3874         if (!ctx->fs->cluster_ratio_bits)
3875                 return 0;
3876         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
3877
3878         /*
3879          * If the block in the logical cluster doesn't align with the block in
3880          * the physical cluster...
3881          */
3882         if ((lblk & cluster_mask) != (pblk & cluster_mask))
3883                 return 1;
3884
3885         /*
3886          * If we cross a physical cluster boundary within a logical cluster...
3887          */
3888         if (last_pblk && (lblk & cluster_mask) != 0 &&
3889             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
3890             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
3891                 return 1;
3892
3893         return 0;
3894 }
3895
3896 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
3897                              struct process_block_struct *pb,
3898                              blk64_t start_block, blk64_t end_block,
3899                              blk64_t eof_block,
3900                              ext2_extent_handle_t ehandle,
3901                              int try_repairs)
3902 {
3903         struct ext2fs_extent    extent;
3904         blk64_t                 blk, last_lblk;
3905         unsigned int            i, n;
3906         int                     is_dir, is_leaf;
3907         problem_t               problem;
3908         struct ext2_extent_info info;
3909         int                     failed_csum = 0;
3910
3911         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
3912                 failed_csum = 1;
3913
3914         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
3915         if (pctx->errcode)
3916                 return;
3917         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
3918             !pb->eti.force_rebuild &&
3919             info.curr_level < MAX_EXTENT_DEPTH_COUNT) {
3920                 struct extent_tree_level *etl;
3921
3922                 etl = pb->eti.ext_info + info.curr_level;
3923                 etl->num_extents += info.num_entries;
3924                 etl->max_extents += info.max_entries;
3925                 /*
3926                  * Implementation wart: Splitting extent blocks when appending
3927                  * will leave the old block with one free entry.  Therefore
3928                  * unless the node is totally full, pretend that a non-root
3929                  * extent block can hold one fewer entry than it actually does,
3930                  * so that we don't repeatedly rebuild the extent tree.
3931                  */
3932                 if (info.curr_level && info.num_entries < info.max_entries)
3933                         etl->max_extents--;
3934         }
3935
3936         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
3937                                           &extent);
3938         while ((pctx->errcode == 0 ||
3939                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
3940                info.num_entries-- > 0) {
3941                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
3942                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
3943                 last_lblk = extent.e_lblk + extent.e_len - 1;
3944
3945                 problem = 0;
3946                 pctx->blk = extent.e_pblk;
3947                 pctx->blk2 = extent.e_lblk;
3948                 pctx->num = extent.e_len;
3949                 pctx->blkcount = extent.e_lblk + extent.e_len;
3950
3951                 if (extent.e_pblk == 0 ||
3952                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
3953                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
3954                         problem = PR_1_EXTENT_BAD_START_BLK;
3955                 else if (extent.e_lblk < start_block)
3956                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
3957                 else if ((end_block && last_lblk > end_block) &&
3958                          !(last_lblk > eof_block &&
3959                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
3960                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
3961                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
3962                 else if (is_leaf && extent.e_len == 0)
3963                         problem = PR_1_EXTENT_LENGTH_ZERO;
3964                 else if (is_leaf &&
3965                          (extent.e_pblk + extent.e_len) >
3966                          ext2fs_blocks_count(ctx->fs->super))
3967                         problem = PR_1_EXTENT_ENDS_BEYOND;
3968                 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
3969                          !ext2fs_has_feature_largedir(ctx->fs->super) &&
3970                          ((extent.e_lblk + extent.e_len) >
3971                           (1U << (21 - ctx->fs->super->s_log_block_size))))
3972                         problem = PR_1_TOOBIG_DIR;
3973
3974                 if (is_leaf && problem == 0 && extent.e_len > 0) {
3975 #if 0
3976                         printf("extent_region(ino=%u, expect=%llu, "
3977                                "lblk=%llu, len=%u)\n", pb->ino,
3978                                (unsigned long long) pb->next_lblock,
3979                                (unsigned long long) extent.e_lblk,
3980                                extent.e_len);
3981 #endif
3982                         if (extent.e_lblk < pb->next_lblock)
3983                                 problem = PR_1_EXTENT_COLLISION;
3984                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
3985                                 pb->next_lblock = extent.e_lblk + extent.e_len;
3986                 }
3987
3988                 /*
3989                  * Uninitialized blocks in a directory?  Clear the flag and
3990                  * we'll interpret the blocks later.
3991                  */
3992                 if (try_repairs && is_dir && problem == 0 &&
3993                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3994                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
3995                         e2fsck_pass1_fix_lock(ctx);
3996                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3997                         pb->inode_modified = 1;
3998                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3999                                                               &extent);
4000                         e2fsck_pass1_fix_unlock(ctx);
4001                         if (pctx->errcode)
4002                                 return;
4003                         failed_csum = 0;
4004                 }
4005 #ifdef CONFIG_DEVELOPER_FEATURES
4006                 if (try_repairs && !is_dir && problem == 0 &&
4007                     (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
4008                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
4009                     fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
4010                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
4011                         pb->inode_modified = 1;
4012                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
4013                                                               &extent);
4014                         if (pctx->errcode)
4015                                 return;
4016                         failed_csum = 0;
4017                 }
4018 #endif
4019                 if (try_repairs && problem) {
4020 report_problem:
4021                         if (fix_problem(ctx, problem, pctx)) {
4022                                 if (ctx->invalid_bitmaps) {
4023                                         /*
4024                                          * If fsck knows the bitmaps are bad,
4025                                          * skip to the next extent and
4026                                          * try to clear this extent again
4027                                          * after fixing the bitmaps, by
4028                                          * restarting fsck.
4029                                          */
4030                                         pctx->errcode = ext2fs_extent_get(
4031                                                           ehandle,
4032                                                           EXT2_EXTENT_NEXT_SIB,
4033                                                           &extent);
4034                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4035                                         if (pctx->errcode ==
4036                                                     EXT2_ET_NO_CURRENT_NODE) {
4037                                                 pctx->errcode = 0;
4038                                                 break;
4039                                         }
4040                                         continue;
4041                                 }
4042                                 e2fsck_pass1_fix_lock(ctx);
4043                                 e2fsck_read_bitmaps(ctx);
4044                                 pb->inode_modified = 1;
4045                                 pctx->errcode =
4046                                         ext2fs_extent_delete(ehandle, 0);
4047                                 e2fsck_pass1_fix_unlock(ctx);
4048                                 if (pctx->errcode) {
4049                                         pctx->str = "ext2fs_extent_delete";
4050                                         return;
4051                                 }
4052                                 e2fsck_pass1_fix_lock(ctx);
4053                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
4054                                 e2fsck_pass1_fix_unlock(ctx);
4055                                 if (pctx->errcode &&
4056                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
4057                                         pctx->str = "ext2fs_extent_fix_parents";
4058                                         return;
4059                                 }
4060                                 pctx->errcode = ext2fs_extent_get(ehandle,
4061                                                                   EXT2_EXTENT_CURRENT,
4062                                                                   &extent);
4063                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
4064                                         pctx->errcode = 0;
4065                                         break;
4066                                 }
4067                                 failed_csum = 0;
4068                                 continue;
4069                         }
4070                         goto next;
4071                 }
4072
4073                 if (!is_leaf) {
4074                         blk64_t lblk = extent.e_lblk;
4075                         int next_try_repairs = 1;
4076
4077                         blk = extent.e_pblk;
4078
4079                         /*
4080                          * If this lower extent block collides with critical
4081                          * metadata, don't try to repair the damage.  Pass 1b
4082                          * will reallocate the block; then we can try again.
4083                          */
4084                         if (pb->ino != EXT2_RESIZE_INO &&
4085                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
4086                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4087                                                       extent.e_pblk)) {
4088                                 next_try_repairs = 0;
4089                                 pctx->blk = blk;
4090                                 fix_problem(ctx,
4091                                             PR_1_CRITICAL_METADATA_COLLISION,
4092                                             pctx);
4093                                 if ((ctx->options & E2F_OPT_NO) == 0)
4094                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4095                         }
4096                         pctx->errcode = ext2fs_extent_get(ehandle,
4097                                                   EXT2_EXTENT_DOWN, &extent);
4098                         if (pctx->errcode &&
4099                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
4100                                 pctx->str = "EXT2_EXTENT_DOWN";
4101                                 problem = PR_1_EXTENT_HEADER_INVALID;
4102                                 if (!next_try_repairs)
4103                                         return;
4104                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
4105                                         goto report_problem;
4106                                 return;
4107                         }
4108                         /* The next extent should match this index's logical start */
4109                         if (extent.e_lblk != lblk) {
4110                                 struct ext2_extent_info e_info;
4111
4112                                 pctx->errcode = ext2fs_extent_get_info(ehandle,
4113                                                                        &e_info);
4114                                 if (pctx->errcode) {
4115                                         pctx->str = "ext2fs_extent_get_info";
4116                                         return;
4117                                 }
4118                                 pctx->blk = lblk;
4119                                 pctx->blk2 = extent.e_lblk;
4120                                 pctx->num = e_info.curr_level - 1;
4121                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
4122                                 if (fix_problem(ctx, problem, pctx)) {
4123                                         e2fsck_pass1_fix_lock(ctx);
4124                                         pb->inode_modified = 1;
4125                                         pctx->errcode =
4126                                                 ext2fs_extent_fix_parents(ehandle);
4127                                         e2fsck_pass1_fix_unlock(ctx);
4128                                         if (pctx->errcode) {
4129                                                 pctx->str = "ext2fs_extent_fix_parents";
4130                                                 return;
4131                                         }
4132                                 }
4133                         }
4134                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
4135                                          last_lblk, eof_block, ehandle,
4136                                          next_try_repairs);
4137                         if (pctx->errcode)
4138                                 return;
4139                         pctx->errcode = ext2fs_extent_get(ehandle,
4140                                                   EXT2_EXTENT_UP, &extent);
4141                         if (pctx->errcode) {
4142                                 pctx->str = "EXT2_EXTENT_UP";
4143                                 return;
4144                         }
4145                         mark_block_used(ctx, blk);
4146                         pb->num_blocks++;
4147                         goto next;
4148                 }
4149
4150                 if ((pb->previous_block != 0) &&
4151                     (pb->previous_block+1 != extent.e_pblk)) {
4152                         if (ctx->options & E2F_OPT_FRAGCHECK) {
4153                                 char type = '?';
4154
4155                                 if (pb->is_dir)
4156                                         type = 'd';
4157                                 else if (pb->is_reg)
4158                                         type = 'f';
4159
4160                                 printf(("%6lu(%c): expecting %6lu "
4161                                         "actual extent "
4162                                         "phys %6lu log %lu len %lu\n"),
4163                                        (unsigned long) pctx->ino, type,
4164                                        (unsigned long) pb->previous_block+1,
4165                                        (unsigned long) extent.e_pblk,
4166                                        (unsigned long) extent.e_lblk,
4167                                        (unsigned long) extent.e_len);
4168                         }
4169                         pb->fragmented = 1;
4170                 }
4171                 /*
4172                  * If we notice a gap in the logical block mappings of an
4173                  * extent-mapped directory, offer to close the hole by
4174                  * moving the logical block down, otherwise we'll go mad in
4175                  * pass 3 allocating empty directory blocks to fill the hole.
4176                  */
4177                 if (try_repairs && is_dir &&
4178                     pb->last_block + 1 < extent.e_lblk) {
4179                         blk64_t new_lblk;
4180
4181                         new_lblk = pb->last_block + 1;
4182                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
4183                                 new_lblk = ((new_lblk +
4184                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
4185                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
4186                                            (extent.e_pblk &
4187                                             EXT2FS_CLUSTER_MASK(ctx->fs));
4188                         pctx->blk = extent.e_lblk;
4189                         pctx->blk2 = new_lblk;
4190                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
4191                                 e2fsck_pass1_fix_lock(ctx);
4192                                 extent.e_lblk = new_lblk;
4193                                 pb->inode_modified = 1;
4194                                 pctx->errcode = ext2fs_extent_replace(ehandle,
4195                                                                 0, &extent);
4196                                 e2fsck_pass1_fix_unlock(ctx);
4197                                 if (pctx->errcode) {
4198                                         pctx->errcode = 0;
4199                                         goto alloc_later;
4200                                 }
4201                                 e2fsck_pass1_fix_lock(ctx);
4202                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
4203                                 e2fsck_pass1_fix_unlock(ctx);
4204                                 if (pctx->errcode)
4205                                         goto failed_add_dir_block;
4206                                 pctx->errcode = ext2fs_extent_goto(ehandle,
4207                                                                 extent.e_lblk);
4208                                 if (pctx->errcode)
4209                                         goto failed_add_dir_block;
4210                                 last_lblk = extent.e_lblk + extent.e_len - 1;
4211                                 failed_csum = 0;
4212                         }
4213                 }
4214 alloc_later:
4215                 if (is_dir) {
4216                         while (++pb->last_db_block <
4217                                (e2_blkcnt_t) extent.e_lblk) {
4218                                 pctx->errcode = ext2fs_add_dir_block2(
4219                                                         ctx->fs->dblist,
4220                                                         pb->ino, 0,
4221                                                         pb->last_db_block);
4222                                 if (pctx->errcode) {
4223                                         pctx->blk = 0;
4224                                         pctx->num = pb->last_db_block;
4225                                         goto failed_add_dir_block;
4226                                 }
4227                         }
4228
4229                         for (i = 0; i < extent.e_len; i++) {
4230                                 pctx->errcode = ext2fs_add_dir_block2(
4231                                                         ctx->fs->dblist,
4232                                                         pctx->ino,
4233                                                         extent.e_pblk + i,
4234                                                         extent.e_lblk + i);
4235                                 if (pctx->errcode) {
4236                                         pctx->blk = extent.e_pblk + i;
4237                                         pctx->num = extent.e_lblk + i;
4238                                 failed_add_dir_block:
4239                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4240                                         /* Should never get here */
4241                                         ctx->flags |= E2F_FLAG_ABORT;
4242                                         return;
4243                                 }
4244                         }
4245                         if (extent.e_len > 0)
4246                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
4247                 }
4248                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
4249                                               pb->last_block,
4250                                               extent.e_pblk,
4251                                               extent.e_lblk)) {
4252                         for (i = 0; i < extent.e_len; i++) {
4253                                 pctx->blk = extent.e_lblk + i;
4254                                 pctx->blk2 = extent.e_pblk + i;
4255                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4256                                 mark_block_used(ctx, extent.e_pblk + i);
4257                                 mark_block_used(ctx, extent.e_pblk + i);
4258                         }
4259                 }
4260
4261                 /*
4262                  * Check whether first cluster got marked in previous iteration.
4263                  */
4264                 if (ctx->fs->cluster_ratio_bits &&
4265                     pb->previous_block &&
4266                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
4267                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
4268                         /* Set blk to the beginning of next cluster. */
4269                         blk = EXT2FS_C2B(
4270                                 ctx->fs,
4271                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
4272                 else
4273                         /* Set blk to the beginning of current cluster. */
4274                         blk = EXT2FS_C2B(ctx->fs,
4275                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
4276
4277                 if (blk < extent.e_pblk + extent.e_len) {
4278                         mark_blocks_used(ctx, blk,
4279                                          extent.e_pblk + extent.e_len - blk);
4280                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
4281                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
4282                         pb->num_blocks += n;
4283                 }
4284                 pb->last_block = extent.e_lblk + extent.e_len - 1;
4285                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
4286                 start_block = pb->last_block = last_lblk;
4287                 if (is_leaf && !is_dir &&
4288                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
4289                         pb->last_init_lblock = last_lblk;
4290         next:
4291                 pctx->errcode = ext2fs_extent_get(ehandle,
4292                                                   EXT2_EXTENT_NEXT_SIB,
4293                                                   &extent);
4294         }
4295
4296         /* Failed csum but passes checks?  Ask to fix checksum. */
4297         if (failed_csum &&
4298             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
4299                 e2fsck_pass1_fix_lock(ctx);
4300                 pb->inode_modified = 1;
4301                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
4302                 e2fsck_pass1_fix_unlock(ctx);
4303                 if (pctx->errcode)
4304                         return;
4305         }
4306
4307         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
4308                 pctx->errcode = 0;
4309 }
4310
4311 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
4312                                  struct process_block_struct *pb)
4313 {
4314         struct ext2_extent_info info;
4315         struct ext2_inode       *inode = pctx->inode;
4316         ext2_extent_handle_t    ehandle;
4317         ext2_filsys             fs = ctx->fs;
4318         ext2_ino_t              ino = pctx->ino;
4319         errcode_t               retval;
4320         blk64_t                 eof_lblk;
4321         struct ext3_extent_header       *eh;
4322
4323         /* Check for a proper extent header... */
4324         eh = (struct ext3_extent_header *) &inode->i_block[0];
4325         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
4326         if (retval) {
4327                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
4328                         e2fsck_clear_inode(ctx, ino, inode, 0,
4329                                            "check_blocks_extents");
4330                 pctx->errcode = 0;
4331                 return;
4332         }
4333
4334         /* ...since this function doesn't fail if i_block is zeroed. */
4335         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
4336         if (pctx->errcode) {
4337                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
4338                         e2fsck_clear_inode(ctx, ino, inode, 0,
4339                                            "check_blocks_extents");
4340                 pctx->errcode = 0;
4341                 return;
4342         }
4343
4344         retval = ext2fs_extent_get_info(ehandle, &info);
4345         if (retval == 0) {
4346                 int max_depth = info.max_depth;
4347
4348                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
4349                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
4350                 ctx->extent_depth_count[max_depth]++;
4351         }
4352
4353         /* Check maximum extent depth */
4354         pctx->blk = info.max_depth;
4355         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
4356         if (pctx->blk2 < pctx->blk &&
4357             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
4358                 pb->eti.force_rebuild = 1;
4359
4360         /* Can we collect extent tree level stats? */
4361         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
4362         if (pctx->blk2 > pctx->blk)
4363                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
4364         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
4365         pb->eti.ino = pb->ino;
4366
4367         pb->next_lblock = 0;
4368
4369         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
4370                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
4371         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
4372         if (pctx->errcode &&
4373             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
4374                 pb->num_blocks = 0;
4375                 inode->i_blocks = 0;
4376                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4377                                    "check_blocks_extents");
4378                 pctx->errcode = 0;
4379         }
4380         ext2fs_extent_free(ehandle);
4381
4382         /* Rebuild unless it's a dir and we're rehashing it */
4383         if (LINUX_S_ISDIR(inode->i_mode) &&
4384             e2fsck_dir_will_be_rehashed(ctx, ino))
4385                 return;
4386
4387         if (ctx->options & E2F_OPT_CONVERT_BMAP)
4388                 e2fsck_rebuild_extents_later(ctx, ino);
4389         else
4390                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
4391 }
4392
4393 /*
4394  * In fact we don't need to check blocks for an inode with inline data
4395  * because this inode doesn't have any blocks.  In this function all
4396  * we need to do is add this inode into dblist when it is a directory.
4397  */
4398 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
4399                                      struct process_block_struct *pb)
4400 {
4401         int     flags;
4402         size_t  inline_data_size = 0;
4403
4404         if (!pb->is_dir) {
4405                 pctx->errcode = 0;
4406                 return;
4407         }
4408
4409         /* Process the dirents in i_block[] as the "first" block. */
4410         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
4411         if (pctx->errcode)
4412                 goto err;
4413
4414         /* Process the dirents in the EA as a "second" block. */
4415         flags = ctx->fs->flags;
4416         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4417         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
4418                                                 &inline_data_size);
4419         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4420                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4421         if (pctx->errcode) {
4422                 pctx->errcode = 0;
4423                 return;
4424         }
4425
4426         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
4427                 return;
4428
4429         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
4430         if (pctx->errcode)
4431                 goto err;
4432
4433         return;
4434 err:
4435         pctx->blk = 0;
4436         pctx->num = 0;
4437         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4438         ctx->flags |= E2F_FLAG_ABORT;
4439 }
4440
4441 /*
4442  * This subroutine is called on each inode to account for all of the
4443  * blocks used by that inode.
4444  */
4445 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
4446                          char *block_buf, const struct ea_quota *ea_ibody_quota)
4447 {
4448         ext2_filsys fs = ctx->fs;
4449         struct process_block_struct pb;
4450         ext2_ino_t      ino = pctx->ino;
4451         struct ext2_inode *inode = pctx->inode;
4452         unsigned        bad_size = 0;
4453         int             dirty_inode = 0;
4454         int             extent_fs;
4455         int             inlinedata_fs;
4456         __u64           size;
4457         struct ea_quota ea_block_quota;
4458
4459         pb.ino = ino;
4460         pb.num_blocks = EXT2FS_B2C(ctx->fs,
4461                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
4462         pb.last_block = ~0;
4463         pb.last_init_lblock = -1;
4464         pb.last_db_block = -1;
4465         pb.num_illegal_blocks = 0;
4466         pb.suppress = 0; pb.clear = 0;
4467         pb.fragmented = 0;
4468         pb.compressed = 0;
4469         pb.previous_block = 0;
4470         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
4471         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
4472         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
4473         pb.inode = inode;
4474         pb.pctx = pctx;
4475         pb.ctx = ctx;
4476         pb.inode_modified = 0;
4477         pb.eti.force_rebuild = 0;
4478         pctx->ino = ino;
4479         pctx->errcode = 0;
4480
4481         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
4482         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
4483
4484         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
4485                 if (e2fsck_should_abort(ctx))
4486                         goto out;
4487                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
4488         }
4489
4490         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
4491                 check_blocks_inline_data(ctx, pctx, &pb);
4492         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
4493                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
4494                         check_blocks_extents(ctx, pctx, &pb);
4495                 else {
4496                         int flags;
4497                         /*
4498                          * If we've modified the inode, write it out before
4499                          * iterate() tries to use it.
4500                          */
4501                         if (dirty_inode) {
4502                                 e2fsck_write_inode(ctx, ino, inode,
4503                                                    "check_blocks");
4504                                 dirty_inode = 0;
4505                         }
4506                         flags = fs->flags;
4507                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4508                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
4509                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
4510                                                 block_buf, process_block, &pb);
4511                         /*
4512                          * We do not have uninitialized extents in non extent
4513                          * files.
4514                          */
4515                         pb.last_init_lblock = pb.last_block;
4516                         /*
4517                          * If iterate() changed a block mapping, we have to
4518                          * re-read the inode.  If we decide to clear the
4519                          * inode after clearing some stuff, we'll re-write the
4520                          * bad mappings into the inode!
4521                          */
4522                         if (pb.inode_modified)
4523                                 e2fsck_read_inode(ctx, ino, inode,
4524                                                   "check_blocks");
4525                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4526                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4527
4528                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
4529 #ifdef DEBUG
4530                                 printf("bmap rebuild ino=%d\n", ino);
4531 #endif
4532                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
4533                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
4534                                         e2fsck_rebuild_extents_later(ctx, ino);
4535                         }
4536                 }
4537         }
4538         end_problem_latch(ctx, PR_LATCH_BLOCK);
4539         end_problem_latch(ctx, PR_LATCH_TOOBIG);
4540         if (e2fsck_should_abort(ctx))
4541                 goto out;
4542         if (pctx->errcode)
4543                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
4544
4545         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
4546                 if (LINUX_S_ISDIR(inode->i_mode))
4547                         ctx->fs_fragmented_dir++;
4548                 else
4549                         ctx->fs_fragmented++;
4550         }
4551
4552         if (pb.clear) {
4553                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4554                                    "check_blocks");
4555                 return;
4556         }
4557
4558         if (inode->i_flags & EXT2_INDEX_FL) {
4559                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
4560                         inode->i_flags &= ~EXT2_INDEX_FL;
4561                         dirty_inode++;
4562                 } else {
4563                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
4564                 }
4565         }
4566
4567         if (!pb.num_blocks && pb.is_dir &&
4568             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
4569                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
4570                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
4571                         ctx->fs_directory_count--;
4572                         return;
4573                 }
4574         }
4575
4576         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
4577             ino != fs->super->s_orphan_file_inum &&
4578             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
4579             !(inode->i_flags & EXT4_EA_INODE_FL)) {
4580                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
4581                                ino,
4582                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
4583                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
4584                                   ino, (ea_ibody_quota ?
4585                                         ea_ibody_quota->inodes : 0) +
4586                                                 ea_block_quota.inodes + 1);
4587         }
4588
4589         if (!ext2fs_has_feature_huge_file(fs->super) ||
4590             !(inode->i_flags & EXT4_HUGE_FILE_FL))
4591                 pb.num_blocks *= (fs->blocksize / 512);
4592         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
4593 #if 0
4594         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
4595                ino, inode->i_size, (unsigned long long) pb.last_block,
4596                (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
4597                (unsigned long long) pb.num_blocks);
4598 #endif
4599         size = EXT2_I_SIZE(inode);
4600         if (pb.is_dir) {
4601                 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
4602                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
4603                         int flags;
4604                         size_t sz = 0;
4605                         errcode_t err;
4606
4607                         flags = ctx->fs->flags;
4608                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4609                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
4610                                                       &sz);
4611                         ctx->fs->flags = (flags &
4612                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4613                                          (ctx->fs->flags &
4614                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4615                         if (err || sz != size) {
4616                                 bad_size = 7;
4617                                 pctx->num = sz;
4618                         }
4619                 } else if (size & (fs->blocksize - 1))
4620                         bad_size = 5;
4621                 else if (nblock > (pb.last_block + 1))
4622                         bad_size = 1;
4623                 else if (nblock < (pb.last_block + 1)) {
4624                         if (((pb.last_block + 1) - nblock) >
4625                             fs->super->s_prealloc_dir_blocks)
4626                                 bad_size = 2;
4627                 }
4628         } else {
4629                 if ((pb.last_init_lblock >= 0) &&
4630                     /* Do not allow initialized allocated blocks past i_size*/
4631                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
4632                     !(inode->i_flags & EXT4_VERITY_FL))
4633                         bad_size = 3;
4634                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4635                          size > ext2_max_sizes[fs->super->s_log_block_size])
4636                         /* too big for a direct/indirect-mapped file */
4637                         bad_size = 4;
4638                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4639                          size >
4640                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
4641                         /* too big for an extent-based file - 32bit ee_block */
4642                         bad_size = 6;
4643         }
4644         /* i_size for symlinks is checked elsewhere */
4645         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
4646                 /* Did inline_data set pctx->num earlier? */
4647                 if (bad_size != 7)
4648                         pctx->num = (pb.last_block + 1) * fs->blocksize;
4649                 pctx->group = bad_size;
4650                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
4651                         ext2fs_inode_size_set(fs, inode, pctx->num);
4652                         if (EXT2_I_SIZE(inode) == 0 &&
4653                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
4654                                 memset(inode->i_block, 0,
4655                                        sizeof(inode->i_block));
4656                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
4657                         }
4658                         dirty_inode++;
4659                 }
4660                 pctx->num = 0;
4661         }
4662         if (LINUX_S_ISREG(inode->i_mode) &&
4663             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4664                 ctx->large_files++;
4665         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
4666             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
4667              (ext2fs_has_feature_huge_file(fs->super) &&
4668               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
4669               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
4670                 pctx->num = pb.num_blocks;
4671                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
4672                         inode->i_blocks = pb.num_blocks;
4673                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
4674                         dirty_inode++;
4675                 }
4676                 pctx->num = 0;
4677         }
4678
4679         /*
4680          * The kernel gets mad if we ask it to allocate bigalloc clusters to
4681          * a block mapped file, so rebuild it as an extent file.  We can skip
4682          * symlinks because they're never rewritten.
4683          */
4684         if (ext2fs_has_feature_bigalloc(fs->super) &&
4685             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
4686             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
4687             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
4688             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
4689             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
4690                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
4691                 if (pctx->errcode)
4692                         goto out;
4693         }
4694
4695         if (ctx->dirs_to_hash && pb.is_dir &&
4696             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
4697             !(inode->i_flags & EXT2_INDEX_FL) &&
4698             ((inode->i_size / fs->blocksize) >= 3))
4699                 e2fsck_rehash_dir_later(ctx, ino);
4700
4701 out:
4702         if (dirty_inode)
4703                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
4704 }
4705
4706 #if 0
4707 /*
4708  * Helper function called by process block when an illegal block is
4709  * found.  It returns a description about why the block is illegal
4710  */
4711 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
4712 {
4713         blk64_t super;
4714         int     i;
4715         static char     problem[80];
4716
4717         super = fs->super->s_first_data_block;
4718         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
4719         if (block < super) {
4720                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
4721                 return(problem);
4722         } else if (block >= ext2fs_blocks_count(fs->super)) {
4723                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
4724                 return(problem);
4725         }
4726         for (i = 0; i < fs->group_desc_count; i++) {
4727                 if (block == super) {
4728                         sprintf(problem, "is the superblock in group %d", i);
4729                         break;
4730                 }
4731                 if (block > super &&
4732                     block <= (super + fs->desc_blocks)) {
4733                         sprintf(problem, "is in the group descriptors "
4734                                 "of group %d", i);
4735                         break;
4736                 }
4737                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
4738                         sprintf(problem, "is the block bitmap of group %d", i);
4739                         break;
4740                 }
4741                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
4742                         sprintf(problem, "is the inode bitmap of group %d", i);
4743                         break;
4744                 }
4745                 if (block >= ext2fs_inode_table_loc(fs, i) &&
4746                     (block < ext2fs_inode_table_loc(fs, i)
4747                      + fs->inode_blocks_per_group)) {
4748                         sprintf(problem, "is in the inode table of group %d",
4749                                 i);
4750                         break;
4751                 }
4752                 super += fs->super->s_blocks_per_group;
4753         }
4754         return(problem);
4755 }
4756 #endif
4757
4758 /*
4759  * This is a helper function for check_blocks().
4760  */
4761 static int process_block(ext2_filsys fs,
4762                   blk64_t       *block_nr,
4763                   e2_blkcnt_t blockcnt,
4764                   blk64_t ref_block EXT2FS_ATTR((unused)),
4765                   int ref_offset EXT2FS_ATTR((unused)),
4766                   void *priv_data)
4767 {
4768         struct process_block_struct *p;
4769         struct problem_context *pctx;
4770         blk64_t blk = *block_nr;
4771         int     ret_code = 0;
4772         problem_t       problem = 0;
4773         e2fsck_t        ctx;
4774
4775         p = (struct process_block_struct *) priv_data;
4776         pctx = p->pctx;
4777         ctx = p->ctx;
4778
4779         /*
4780          * For a directory, add logical block zero for processing even if it's
4781          * not mapped or we'll be perennially stuck with broken "." and ".."
4782          * entries.
4783          */
4784         if (p->is_dir && blockcnt == 0 && blk == 0) {
4785                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
4786                 if (pctx->errcode) {
4787                         pctx->blk = blk;
4788                         pctx->num = blockcnt;
4789                         goto failed_add_dir_block;
4790                 }
4791                 p->last_db_block++;
4792         }
4793
4794         if (blk == 0)
4795                 return 0;
4796
4797 #if 0
4798         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
4799                blockcnt);
4800 #endif
4801
4802         /*
4803          * Simplistic fragmentation check.  We merely require that the
4804          * file be contiguous.  (Which can never be true for really
4805          * big files that are greater than a block group.)
4806          */
4807         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
4808                 if (p->previous_block+1 != blk) {
4809                         if (ctx->options & E2F_OPT_FRAGCHECK) {
4810                                 char type = '?';
4811
4812                                 if (p->is_dir)
4813                                         type = 'd';
4814                                 else if (p->is_reg)
4815                                         type = 'f';
4816
4817                                 printf(_("%6lu(%c): expecting %6lu "
4818                                          "got phys %6lu (blkcnt %lld)\n"),
4819                                        (unsigned long) pctx->ino, type,
4820                                        (unsigned long) p->previous_block+1,
4821                                        (unsigned long) blk,
4822                                        (long long) blockcnt);
4823                         }
4824                         p->fragmented = 1;
4825                 }
4826         }
4827
4828         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
4829             !pctx->inode->i_size_high &&
4830             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
4831                 problem = PR_1_TOOBIG_DIR;
4832         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
4833                 problem = PR_1_TOOBIG_DIR;
4834         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
4835                 problem = PR_1_TOOBIG_REG;
4836         if (!p->is_dir && !p->is_reg && blockcnt > 0)
4837                 problem = PR_1_TOOBIG_SYMLINK;
4838
4839         if (blk < fs->super->s_first_data_block ||
4840             blk >= ext2fs_blocks_count(fs->super))
4841                 problem = PR_1_ILLEGAL_BLOCK_NUM;
4842
4843         /*
4844          * If this IND/DIND/TIND block is squatting atop some critical metadata
4845          * (group descriptors, superblock, bitmap, inode table), any write to
4846          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
4847          * fix that and restart fsck.
4848          */
4849         if (blockcnt < 0 &&
4850             p->ino != EXT2_RESIZE_INO &&
4851             blk < ctx->fs->super->s_blocks_count &&
4852             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
4853                 pctx->blk = blk;
4854                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
4855                 if ((ctx->options & E2F_OPT_NO) == 0)
4856                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4857         }
4858
4859         if (problem) {
4860                 p->num_illegal_blocks++;
4861                 /*
4862                  * A bit of subterfuge here -- we're trying to fix a block
4863                  * mapping, but the IND/DIND/TIND block could have collided
4864                  * with some critical metadata.  So, fix the in-core mapping so
4865                  * iterate won't go insane, but return 0 instead of
4866                  * BLOCK_CHANGED so that it won't write the remapping out to
4867                  * our multiply linked block.
4868                  *
4869                  * Even if we previously determined that an *IND block
4870                  * conflicts with critical metadata, we must still try to
4871                  * iterate the *IND block as if it is an *IND block to find and
4872                  * mark the blocks it points to.  Better to be overly cautious
4873                  * with the used_blocks map so that we don't move the *IND
4874                  * block to a block that's really in use!
4875                  */
4876                 if (p->ino != EXT2_RESIZE_INO &&
4877                     ref_block != 0 &&
4878                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4879                                               ref_block)) {
4880                         *block_nr = 0;
4881                         return 0;
4882                 }
4883                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
4884                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
4885                                 p->clear = 1;
4886                                 return BLOCK_ABORT;
4887                         }
4888                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
4889                                 p->suppress = 1;
4890                                 set_latch_flags(PR_LATCH_BLOCK,
4891                                                 PRL_SUPPRESS, 0);
4892                         }
4893                 }
4894                 pctx->blk = blk;
4895                 pctx->blkcount = blockcnt;
4896                 if (fix_problem(ctx, problem, pctx)) {
4897                         blk = *block_nr = 0;
4898                         ret_code = BLOCK_CHANGED;
4899                         p->inode_modified = 1;
4900                         /*
4901                          * If the directory block is too big and is beyond the
4902                          * end of the FS, don't bother trying to add it for
4903                          * processing -- the kernel would never have created a
4904                          * directory this large, and we risk an ENOMEM abort.
4905                          * In any case, the toobig handler for extent-based
4906                          * directories also doesn't feed toobig blocks to
4907                          * pass 2.
4908                          */
4909                         if (problem == PR_1_TOOBIG_DIR)
4910                                 return ret_code;
4911                         goto mark_dir;
4912                 } else
4913                         return 0;
4914         }
4915
4916         if (p->ino == EXT2_RESIZE_INO) {
4917                 /*
4918                  * The resize inode has already be sanity checked
4919                  * during pass #0 (the superblock checks).  All we
4920                  * have to do is mark the double indirect block as
4921                  * being in use; all of the other blocks are handled
4922                  * by mark_table_blocks()).
4923                  */
4924                 if (blockcnt == BLOCK_COUNT_DIND)
4925                         mark_block_used(ctx, blk);
4926                 p->num_blocks++;
4927         } else if (!(ctx->fs->cluster_ratio_bits &&
4928                      p->previous_block &&
4929                      (EXT2FS_B2C(ctx->fs, blk) ==
4930                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
4931                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
4932                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
4933                 mark_block_used(ctx, blk);
4934                 p->num_blocks++;
4935         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
4936                                              p->last_block, blk, blockcnt)) {
4937                 pctx->blk = blockcnt;
4938                 pctx->blk2 = blk;
4939                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4940                 mark_block_used(ctx, blk);
4941                 mark_block_used(ctx, blk);
4942         }
4943         if (blockcnt >= 0)
4944                 p->last_block = blockcnt;
4945         p->previous_block = blk;
4946 mark_dir:
4947         if (p->is_dir && (blockcnt >= 0)) {
4948                 while (++p->last_db_block < blockcnt) {
4949                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
4950                                                               p->ino, 0,
4951                                                               p->last_db_block);
4952                         if (pctx->errcode) {
4953                                 pctx->blk = 0;
4954                                 pctx->num = p->last_db_block;
4955                                 goto failed_add_dir_block;
4956                         }
4957                 }
4958                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
4959                                                       blk, blockcnt);
4960                 if (pctx->errcode) {
4961                         pctx->blk = blk;
4962                         pctx->num = blockcnt;
4963                 failed_add_dir_block:
4964                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4965                         /* Should never get here */
4966                         ctx->flags |= E2F_FLAG_ABORT;
4967                         return BLOCK_ABORT;
4968                 }
4969         }
4970         return ret_code;
4971 }
4972
4973 static int process_bad_block(ext2_filsys fs,
4974                       blk64_t *block_nr,
4975                       e2_blkcnt_t blockcnt,
4976                       blk64_t ref_block EXT2FS_ATTR((unused)),
4977                       int ref_offset EXT2FS_ATTR((unused)),
4978                       void *priv_data)
4979 {
4980         struct process_block_struct *p;
4981         blk64_t         blk = *block_nr;
4982         blk64_t         first_block;
4983         dgrp_t          i;
4984         struct problem_context *pctx;
4985         e2fsck_t        ctx;
4986
4987         if (!blk)
4988                 return 0;
4989
4990         p = (struct process_block_struct *) priv_data;
4991         ctx = p->ctx;
4992         pctx = p->pctx;
4993
4994         pctx->ino = EXT2_BAD_INO;
4995         pctx->blk = blk;
4996         pctx->blkcount = blockcnt;
4997
4998         if ((blk < fs->super->s_first_data_block) ||
4999             (blk >= ext2fs_blocks_count(fs->super))) {
5000                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
5001                         *block_nr = 0;
5002                         return BLOCK_CHANGED;
5003                 } else
5004                         return 0;
5005         }
5006
5007         if (blockcnt < 0) {
5008                 e2fsck_pass1_block_map_lock(ctx);
5009                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
5010                         p->bbcheck = 1;
5011                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
5012                                 *block_nr = 0;
5013                                 e2fsck_pass1_block_map_unlock(ctx);
5014                                 return BLOCK_CHANGED;
5015                         }
5016                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5017                                                     blk)) {
5018                         p->bbcheck = 1;
5019                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
5020                                         pctx)) {
5021                                 *block_nr = 0;
5022                                 e2fsck_pass1_block_map_unlock(ctx);
5023                                 return BLOCK_CHANGED;
5024                         }
5025                         if (e2fsck_should_abort(ctx)) {
5026                                 e2fsck_pass1_block_map_unlock(ctx);
5027                                 return BLOCK_ABORT;
5028                         }
5029                 } else {
5030                         mark_block_used_unlocked(ctx, blk);
5031                 }
5032                 e2fsck_pass1_block_map_unlock(ctx);
5033                 return 0;
5034         }
5035 #if 0
5036         printf ("DEBUG: Marking %u as bad.\n", blk);
5037 #endif
5038         ctx->fs_badblocks_count++;
5039         /*
5040          * If the block is not used, then mark it as used and return.
5041          * If it is already marked as found, this must mean that
5042          * there's an overlap between the filesystem table blocks
5043          * (bitmaps and inode table) and the bad block list.
5044          */
5045         e2fsck_pass1_block_map_lock(ctx);
5046         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
5047                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
5048                 e2fsck_pass1_block_map_unlock(ctx);
5049                 return 0;
5050         }
5051         e2fsck_pass1_block_map_unlock(ctx);
5052         /*
5053          * Try to find the where the filesystem block was used...
5054          */
5055         first_block = fs->super->s_first_data_block;
5056
5057         for (i = 0; i < fs->group_desc_count; i++ ) {
5058                 pctx->group = i;
5059                 pctx->blk = blk;
5060                 if (!ext2fs_bg_has_super(fs, i))
5061                         goto skip_super;
5062                 if (blk == first_block) {
5063                         if (i == 0) {
5064                                 if (fix_problem(ctx,
5065                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
5066                                                 pctx)) {
5067                                         *block_nr = 0;
5068                                         return BLOCK_CHANGED;
5069                                 }
5070                                 return 0;
5071                         }
5072                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
5073                         return 0;
5074                 }
5075                 if ((blk > first_block) &&
5076                     (blk <= first_block + fs->desc_blocks)) {
5077                         if (i == 0) {
5078                                 pctx->blk = *block_nr;
5079                                 if (fix_problem(ctx,
5080                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
5081                                         *block_nr = 0;
5082                                         return BLOCK_CHANGED;
5083                                 }
5084                                 return 0;
5085                         }
5086                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
5087                         return 0;
5088                 }
5089         skip_super:
5090                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
5091                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
5092                                 ctx->invalid_block_bitmap_flag[i]++;
5093                                 ctx->invalid_bitmaps++;
5094                         }
5095                         return 0;
5096                 }
5097                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
5098                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
5099                                 ctx->invalid_inode_bitmap_flag[i]++;
5100                                 ctx->invalid_bitmaps++;
5101                         }
5102                         return 0;
5103                 }
5104                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
5105                     (blk < (ext2fs_inode_table_loc(fs, i) +
5106                             fs->inode_blocks_per_group))) {
5107                         /*
5108                          * If there are bad blocks in the inode table,
5109                          * the inode scan code will try to do
5110                          * something reasonable automatically.
5111                          */
5112                         return 0;
5113                 }
5114                 first_block += fs->super->s_blocks_per_group;
5115         }
5116         /*
5117          * If we've gotten to this point, then the only
5118          * possibility is that the bad block inode meta data
5119          * is using a bad block.
5120          */
5121         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
5122             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
5123             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
5124                 p->bbcheck = 1;
5125                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
5126                         *block_nr = 0;
5127                         return BLOCK_CHANGED;
5128                 }
5129                 if (e2fsck_should_abort(ctx))
5130                         return BLOCK_ABORT;
5131                 return 0;
5132         }
5133
5134         pctx->group = -1;
5135
5136         /* Warn user that the block wasn't claimed */
5137         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
5138
5139         return 0;
5140 }
5141
5142 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
5143                             const char *name, int num, blk64_t *new_block)
5144 {
5145         ext2_filsys fs = ctx->fs;
5146         dgrp_t          last_grp;
5147         blk64_t         old_block = *new_block;
5148         blk64_t         last_block;
5149         dgrp_t          flexbg;
5150         unsigned        flexbg_size;
5151         int             i, is_flexbg;
5152         char            *buf;
5153         struct problem_context  pctx;
5154
5155         clear_problem_context(&pctx);
5156
5157         pctx.group = group;
5158         pctx.blk = old_block;
5159         pctx.str = name;
5160
5161         /*
5162          * For flex_bg filesystems, first try to allocate the metadata
5163          * within the flex_bg, and if that fails then try finding the
5164          * space anywhere in the filesystem.
5165          */
5166         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
5167         if (is_flexbg) {
5168                 flexbg_size = 1U << fs->super->s_log_groups_per_flex;
5169                 flexbg = group / flexbg_size;
5170                 first_block = ext2fs_group_first_block2(fs,
5171                                                         flexbg_size * flexbg);
5172                 last_grp = group | (flexbg_size - 1);
5173                 if (last_grp >= fs->group_desc_count)
5174                         last_grp = fs->group_desc_count - 1;
5175                 last_block = ext2fs_group_last_block2(fs, last_grp);
5176         } else
5177                 last_block = ext2fs_group_last_block2(fs, group);
5178         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
5179                                                num, ctx->block_found_map,
5180                                                new_block);
5181         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
5182                 pctx.errcode = ext2fs_get_free_blocks2(fs,
5183                                 fs->super->s_first_data_block,
5184                                 ext2fs_blocks_count(fs->super),
5185                                 num, ctx->block_found_map, new_block);
5186         if (pctx.errcode) {
5187                 pctx.num = num;
5188                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
5189                 ext2fs_unmark_valid(fs);
5190                 ctx->flags |= E2F_FLAG_ABORT;
5191                 return;
5192         }
5193         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
5194         if (pctx.errcode) {
5195                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
5196                 ext2fs_unmark_valid(fs);
5197                 ctx->flags |= E2F_FLAG_ABORT;
5198                 return;
5199         }
5200         ext2fs_mark_super_dirty(fs);
5201         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
5202         pctx.blk2 = *new_block;
5203         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
5204                           PR_1_RELOC_TO), &pctx);
5205         pctx.blk2 = 0;
5206         e2fsck_pass1_block_map_lock(ctx);
5207         for (i = 0; i < num; i++) {
5208                 pctx.blk = i;
5209                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
5210                 if (old_block) {
5211                         pctx.errcode = io_channel_read_blk64(fs->io,
5212                                    old_block + i, 1, buf);
5213                         if (pctx.errcode)
5214                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
5215                         pctx.blk = (*new_block) + i;
5216                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
5217                                                               1, buf);
5218                 } else {
5219                         pctx.blk = (*new_block) + i;
5220                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
5221                                                            NULL, NULL);
5222                 }
5223
5224                 if (pctx.errcode)
5225                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
5226         }
5227         e2fsck_pass1_block_map_unlock(ctx);
5228         ext2fs_free_mem(&buf);
5229 }
5230
5231 /*
5232  * This routine gets called at the end of pass 1 if bad blocks are
5233  * detected in the superblock, group descriptors, inode_bitmaps, or
5234  * block bitmaps.  At this point, all of the blocks have been mapped
5235  * out, so we can try to allocate new block(s) to replace the bad
5236  * blocks.
5237  */
5238 static void handle_fs_bad_blocks(e2fsck_t ctx)
5239 {
5240         ext2_filsys fs = ctx->fs;
5241         dgrp_t          i;
5242         blk64_t         first_block;
5243         blk64_t         new_blk;
5244
5245         for (i = 0; i < fs->group_desc_count; i++) {
5246                 first_block = ext2fs_group_first_block2(fs, i);
5247
5248                 if (ctx->invalid_block_bitmap_flag[i]) {
5249                         new_blk = ext2fs_block_bitmap_loc(fs, i);
5250                         new_table_block(ctx, first_block, i, _("block bitmap"),
5251                                         1, &new_blk);
5252                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
5253                 }
5254                 if (ctx->invalid_inode_bitmap_flag[i]) {
5255                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
5256                         new_table_block(ctx, first_block, i, _("inode bitmap"),
5257                                         1, &new_blk);
5258                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
5259                 }
5260                 if (ctx->invalid_inode_table_flag[i]) {
5261                         new_blk = ext2fs_inode_table_loc(fs, i);
5262                         new_table_block(ctx, first_block, i, _("inode table"),
5263                                         fs->inode_blocks_per_group,
5264                                         &new_blk);
5265                         ext2fs_inode_table_loc_set(fs, i, new_blk);
5266                         ctx->flags |= E2F_FLAG_RESTART;
5267                 }
5268         }
5269         ctx->invalid_bitmaps = 0;
5270 }
5271
5272 /*
5273  * This routine marks all blocks which are used by the superblock,
5274  * group descriptors, inode bitmaps, and block bitmaps.
5275  */
5276 static void mark_table_blocks(e2fsck_t ctx)
5277 {
5278         ext2_filsys fs = ctx->fs;
5279         blk64_t b;
5280         dgrp_t  i;
5281         unsigned int    j;
5282         struct problem_context pctx;
5283
5284         clear_problem_context(&pctx);
5285
5286         for (i = 0; i < fs->group_desc_count; i++) {
5287                 pctx.group = i;
5288
5289                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
5290                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
5291
5292                 /*
5293                  * Mark the blocks used for the inode table
5294                  */
5295                 if (ext2fs_inode_table_loc(fs, i)) {
5296                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
5297                              j < fs->inode_blocks_per_group;
5298                              j++, b++) {
5299                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5300                                                              b)) {
5301                                         pctx.blk = b;
5302                                         if (!ctx->invalid_inode_table_flag[i] &&
5303                                             fix_problem(ctx,
5304                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
5305                                                 ctx->invalid_inode_table_flag[i]++;
5306                                                 ctx->invalid_bitmaps++;
5307                                         }
5308                                 } else {
5309                                     ext2fs_mark_block_bitmap2(
5310                                                 ctx->block_found_map, b);
5311                                     ext2fs_mark_block_bitmap2(
5312                                                 ctx->block_metadata_map, b);
5313                                 }
5314                         }
5315                 }
5316
5317                 /*
5318                  * Mark block used for the block bitmap
5319                  */
5320                 if (ext2fs_block_bitmap_loc(fs, i)) {
5321                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5322                                      ext2fs_block_bitmap_loc(fs, i))) {
5323                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
5324                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
5325                                         ctx->invalid_block_bitmap_flag[i]++;
5326                                         ctx->invalid_bitmaps++;
5327                                 }
5328                         } else {
5329                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
5330                                      ext2fs_block_bitmap_loc(fs, i));
5331                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5332                                      ext2fs_block_bitmap_loc(fs, i));
5333                         }
5334                 }
5335                 /*
5336                  * Mark block used for the inode bitmap
5337                  */
5338                 if (ext2fs_inode_bitmap_loc(fs, i)) {
5339                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5340                                      ext2fs_inode_bitmap_loc(fs, i))) {
5341                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
5342                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
5343                                         ctx->invalid_inode_bitmap_flag[i]++;
5344                                         ctx->invalid_bitmaps++;
5345                                 }
5346                         } else {
5347                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5348                                      ext2fs_inode_bitmap_loc(fs, i));
5349                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
5350                                      ext2fs_inode_bitmap_loc(fs, i));
5351                         }
5352                 }
5353         }
5354 }
5355
5356 /*
5357  * These subroutines short circuits ext2fs_get_blocks and
5358  * ext2fs_check_directory; we use them since we already have the inode
5359  * structure, so there's no point in letting the ext2fs library read
5360  * the inode again.
5361  */
5362 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
5363                                   blk_t *blocks)
5364 {
5365         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5366         int     i;
5367
5368         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5369                 return EXT2_ET_CALLBACK_NOTHANDLED;
5370
5371         for (i=0; i < EXT2_N_BLOCKS; i++)
5372                 blocks[i] = ctx->stashed_inode->i_block[i];
5373         return 0;
5374 }
5375
5376 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
5377                                   struct ext2_inode *inode)
5378 {
5379         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5380
5381         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5382                 return EXT2_ET_CALLBACK_NOTHANDLED;
5383         *inode = *ctx->stashed_inode;
5384         return 0;
5385 }
5386
5387 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
5388                             struct ext2_inode *inode)
5389 {
5390         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5391
5392         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
5393                 (inode != ctx->stashed_inode))
5394                 *ctx->stashed_inode = *inode;
5395         return EXT2_ET_CALLBACK_NOTHANDLED;
5396 }
5397
5398 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
5399 {
5400         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5401
5402         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5403                 return EXT2_ET_CALLBACK_NOTHANDLED;
5404
5405         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
5406                 return EXT2_ET_NO_DIRECTORY;
5407         return 0;
5408 }
5409
5410 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
5411                                         blk64_t *ret)
5412 {
5413         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5414         errcode_t       retval;
5415         blk64_t         new_block;
5416
5417         if (ctx->block_found_map) {
5418                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
5419                                            &new_block);
5420                 if (retval)
5421                         return retval;
5422                 if (fs->block_map) {
5423                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
5424                         ext2fs_mark_bb_dirty(fs);
5425                 }
5426         } else {
5427                 if (!fs->block_map) {
5428                         retval = ext2fs_read_block_bitmap(fs);
5429                         if (retval)
5430                                 return retval;
5431                 }
5432
5433                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
5434                 if (retval)
5435                         return retval;
5436         }
5437
5438         *ret = new_block;
5439         return (0);
5440 }
5441
5442 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
5443                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
5444 {
5445         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5446         errcode_t       retval;
5447
5448         if (ctx->block_found_map)
5449                 return ext2fs_new_range(fs, flags, goal, len,
5450                                         ctx->block_found_map, pblk, plen);
5451
5452         if (!fs->block_map) {
5453                 retval = ext2fs_read_block_bitmap(fs);
5454                 if (retval)
5455                         return retval;
5456         }
5457
5458         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
5459                                 pblk, plen);
5460 }
5461
5462 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
5463 {
5464         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5465
5466         /* Never free a critical metadata block */
5467         if (ctx->block_found_map &&
5468             ctx->block_metadata_map &&
5469             inuse < 0 &&
5470             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
5471                 return;
5472
5473         if (ctx->block_found_map) {
5474                 if (inuse > 0)
5475                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
5476                 else
5477                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
5478         }
5479 }
5480
5481 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
5482                                            blk_t num, int inuse)
5483 {
5484         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5485
5486         /* Never free a critical metadata block */
5487         if (ctx->block_found_map &&
5488             ctx->block_metadata_map &&
5489             inuse < 0 &&
5490             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
5491                 return;
5492
5493         if (ctx->block_found_map) {
5494                 if (inuse > 0)
5495                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
5496                                                         blk, num);
5497                 else
5498                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
5499                                                         blk, num);
5500         }
5501 }
5502
5503 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
5504 {
5505         ext2_filsys fs = ctx->fs;
5506
5507         if (use_shortcuts) {
5508                 fs->get_blocks = pass1_get_blocks;
5509                 fs->check_directory = pass1_check_directory;
5510                 fs->read_inode = pass1_read_inode;
5511                 fs->write_inode = pass1_write_inode;
5512                 ctx->stashed_ino = 0;
5513         } else {
5514                 fs->get_blocks = 0;
5515                 fs->check_directory = 0;
5516                 fs->read_inode = 0;
5517                 fs->write_inode = 0;
5518         }
5519 }
5520
5521 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
5522 {
5523         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
5524         ext2fs_set_block_alloc_stats_callback(ctx->fs,
5525                                                 e2fsck_block_alloc_stats, 0);
5526         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
5527         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
5528                                         e2fsck_block_alloc_stats_range, NULL);
5529 }