Whamcloud - gitweb
e2fsck: handle preallocation for large PAGE_SIZE
[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 blocks are in use.          (block_found_map)
27  *      - A bitmap of which blocks are in use by two inodes     (block_dup_map)
28  *      - The data blocks of the directory inodes.      (dir_map)
29  *      - Ref counts for ea_inodes.                     (ea_inode_refs)
30  *
31  * Pass 1 is designed to stash away enough information so that the
32  * other passes should not need to read in the inode information
33  * during the normal course of a filesystem check.  (Although if an
34  * inconsistency is detected, other passes may need to read in an
35  * inode to fix it.)
36  *
37  * Note that pass 1B will be invoked if there are any duplicate blocks
38  * found.
39  */
40
41 #define _GNU_SOURCE 1 /* get strnlen() */
42 #include "config.h"
43 #include <string.h>
44 #include <time.h>
45 #ifdef HAVE_ERRNO_H
46 #include <errno.h>
47 #endif
48
49 #include "e2fsck.h"
50 #include <ext2fs/ext2_ext_attr.h>
51
52 #include "problem.h"
53
54 #ifdef NO_INLINE_FUNCS
55 #define _INLINE_
56 #else
57 #define _INLINE_ inline
58 #endif
59
60 #undef DEBUG
61
62 struct ea_quota {
63         blk64_t blocks;
64         __u64 inodes;
65 };
66
67 static int process_block(ext2_filsys fs, blk64_t        *blocknr,
68                          e2_blkcnt_t blockcnt, blk64_t ref_blk,
69                          int ref_offset, void *priv_data);
70 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
71                              e2_blkcnt_t blockcnt, blk64_t ref_blk,
72                              int ref_offset, void *priv_data);
73 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
74                          char *block_buf,
75                          const struct ea_quota *ea_ibody_quota);
76 static void mark_table_blocks(e2fsck_t ctx);
77 static void alloc_bb_map(e2fsck_t ctx);
78 static void alloc_imagic_map(e2fsck_t ctx);
79 static void mark_inode_bad(e2fsck_t ctx, ino_t ino);
80 static void add_encrypted_dir(e2fsck_t ctx, ino_t ino);
81 static void handle_fs_bad_blocks(e2fsck_t ctx);
82 static void process_inodes(e2fsck_t ctx, char *block_buf);
83 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
84 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
85                                   dgrp_t group, void * priv_data);
86 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
87                                     char *block_buf, int adjust_sign);
88 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
89
90 struct process_block_struct {
91         ext2_ino_t      ino;
92         unsigned        is_dir:1, is_reg:1, clear:1, suppress:1,
93                                 fragmented:1, compressed:1, bbcheck:1,
94                                 inode_modified:1;
95         blk64_t         num_blocks;
96         blk64_t         max_blocks;
97         blk64_t         last_block;
98         e2_blkcnt_t     last_init_lblock;
99         e2_blkcnt_t     last_db_block;
100         int             num_illegal_blocks;
101         blk64_t         previous_block;
102         struct ext2_inode *inode;
103         struct problem_context *pctx;
104         ext2fs_block_bitmap fs_meta_blocks;
105         e2fsck_t        ctx;
106         blk64_t         next_lblock;
107         struct extent_tree_info eti;
108 };
109
110 struct process_inode_block {
111         ext2_ino_t ino;
112         struct ea_quota ea_ibody_quota;
113         struct ext2_inode_large inode;
114 };
115
116 struct scan_callback_struct {
117         e2fsck_t        ctx;
118         char            *block_buf;
119 };
120
121 /*
122  * For the inodes to process list.
123  */
124 static struct process_inode_block *inodes_to_process;
125 static int process_inode_count;
126
127 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
128                             EXT2_MIN_BLOCK_LOG_SIZE + 1];
129
130 /*
131  * Free all memory allocated by pass1 in preparation for restarting
132  * things.
133  */
134 static void unwind_pass1(ext2_filsys fs EXT2FS_ATTR((unused)))
135 {
136         ext2fs_free_mem(&inodes_to_process);
137         inodes_to_process = 0;
138 }
139
140 /*
141  * Check to make sure a device inode is real.  Returns 1 if the device
142  * checks out, 0 if not.
143  *
144  * Note: this routine is now also used to check FIFO's and Sockets,
145  * since they have the same requirement; the i_block fields should be
146  * zero.
147  */
148 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
149                                     struct ext2_inode *inode)
150 {
151         int     i;
152
153         /*
154          * If the index flag is set, then this is a bogus
155          * device/fifo/socket
156          */
157         if (inode->i_flags & EXT2_INDEX_FL)
158                 return 0;
159
160         /*
161          * We should be able to do the test below all the time, but
162          * because the kernel doesn't forcibly clear the device
163          * inode's additional i_block fields, there are some rare
164          * occasions when a legitimate device inode will have non-zero
165          * additional i_block fields.  So for now, we only complain
166          * when the immutable flag is set, which should never happen
167          * for devices.  (And that's when the problem is caused, since
168          * you can't set or clear immutable flags for devices.)  Once
169          * the kernel has been fixed we can change this...
170          */
171         if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
172                 for (i=4; i < EXT2_N_BLOCKS; i++)
173                         if (inode->i_block[i])
174                                 return 0;
175         }
176         return 1;
177 }
178
179 /*
180  * Check to make sure a symlink inode is real.  Returns 1 if the symlink
181  * checks out, 0 if not.
182  */
183 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
184                                struct ext2_inode *inode, char *buf)
185 {
186         unsigned int buflen;
187         unsigned int len;
188
189         if ((inode->i_size_high || inode->i_size == 0) ||
190             (inode->i_flags & EXT2_INDEX_FL))
191                 return 0;
192
193         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
194                 size_t inline_size;
195
196                 if (inode->i_flags & EXT4_EXTENTS_FL)
197                         return 0;
198                 if (ext2fs_inline_data_size(fs, ino, &inline_size))
199                         return 0;
200                 if (inode->i_size != inline_size)
201                         return 0;
202
203                 return 1;
204         }
205
206         if (ext2fs_is_fast_symlink(inode)) {
207                 if (inode->i_flags & EXT4_EXTENTS_FL)
208                         return 0;
209                 buf = (char *)inode->i_block;
210                 buflen = sizeof(inode->i_block);
211         } else {
212                 ext2_extent_handle_t    handle;
213                 struct ext2_extent_info info;
214                 struct ext2fs_extent    extent;
215                 blk64_t blk;
216                 int i;
217
218                 if (inode->i_flags & EXT4_EXTENTS_FL) {
219                         if (ext2fs_extent_open2(fs, ino, inode, &handle))
220                                 return 0;
221                         if (ext2fs_extent_get_info(handle, &info) ||
222                             (info.num_entries != 1) ||
223                             (info.max_depth != 0)) {
224                                 ext2fs_extent_free(handle);
225                                 return 0;
226                         }
227                         if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
228                                               &extent) ||
229                             (extent.e_lblk != 0) ||
230                             (extent.e_len != 1)) {
231                                 ext2fs_extent_free(handle);
232                                 return 0;
233                         }
234                         blk = extent.e_pblk;
235                         ext2fs_extent_free(handle);
236                 } else {
237                         blk = inode->i_block[0];
238
239                         for (i = 1; i < EXT2_N_BLOCKS; i++)
240                                 if (inode->i_block[i])
241                                         return 0;
242                 }
243
244                 if (blk < fs->super->s_first_data_block ||
245                     blk >= ext2fs_blocks_count(fs->super))
246                         return 0;
247
248                 if (io_channel_read_blk64(fs->io, blk, 1, buf))
249                         return 0;
250
251                 buflen = fs->blocksize;
252         }
253
254         if (inode->i_flags & EXT4_ENCRYPT_FL)
255                 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
256         else
257                 len = strnlen(buf, buflen);
258
259         if (len >= buflen)
260                 return 0;
261
262         if (len != inode->i_size)
263                 return 0;
264         return 1;
265 }
266
267 /*
268  * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
269  */
270 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
271 static void check_extents_inlinedata(e2fsck_t ctx,
272                                      struct problem_context *pctx)
273 {
274         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
275                 return;
276
277         if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
278                 return;
279
280         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
281         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
282 }
283 #undef BAD_SPECIAL_FLAGS
284
285 /*
286  * If the immutable (or append-only) flag is set on the inode, offer
287  * to clear it.
288  */
289 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
290 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
291 {
292         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
293                 return;
294
295         if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
296                 return;
297
298         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
299         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
300 }
301
302 /*
303  * If device, fifo or socket, check size is zero -- if not offer to
304  * clear it
305  */
306 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
307 {
308         struct ext2_inode *inode = pctx->inode;
309
310         if (EXT2_I_SIZE(inode) == 0)
311                 return;
312
313         if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
314                 return;
315
316         ext2fs_inode_size_set(ctx->fs, inode, 0);
317         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
318 }
319
320 /*
321  * For a given size, calculate how many blocks would be charged towards quota.
322  */
323 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
324 {
325         blk64_t clusters;
326
327         clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
328         return EXT2FS_C2B(fs, clusters);
329 }
330
331 /*
332  * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
333  * the problem code.
334  */
335 static problem_t check_large_ea_inode(e2fsck_t ctx,
336                                       struct ext2_ext_attr_entry *entry,
337                                       struct problem_context *pctx,
338                                       blk64_t *quota_blocks)
339 {
340         struct ext2_inode inode;
341         __u32 hash;
342         errcode_t retval;
343
344         /* Check if inode is within valid range */
345         if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
346             (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
347                 pctx->num = entry->e_value_inum;
348                 return PR_1_ATTR_VALUE_EA_INODE;
349         }
350
351         e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
352
353         retval = ext2fs_ext_attr_hash_entry2(ctx->fs, entry, NULL, &hash);
354         if (retval) {
355                 com_err("check_large_ea_inode", retval,
356                         _("while hashing entry with e_value_inum = %u"),
357                         entry->e_value_inum);
358                 fatal_error(ctx, 0);
359         }
360
361         if (hash == entry->e_hash) {
362                 *quota_blocks = size_to_quota_blocks(ctx->fs,
363                                                      entry->e_value_size);
364         } else {
365                 /* This might be an old Lustre-style ea_inode reference. */
366                 if (inode.i_mtime == pctx->ino &&
367                     inode.i_generation == pctx->inode->i_generation) {
368                         *quota_blocks = 0;
369                 } else {
370                         /* If target inode is also missing EA_INODE flag,
371                          * this is likely to be a bad reference.
372                          */
373                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
374                                 pctx->num = entry->e_value_inum;
375                                 return PR_1_ATTR_VALUE_EA_INODE;
376                         } else {
377                                 pctx->num = entry->e_hash;
378                                 return PR_1_ATTR_HASH;
379                         }
380                 }
381         }
382
383         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
384                 pctx->num = entry->e_value_inum;
385                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
386                         inode.i_flags |= EXT4_EA_INODE_FL;
387                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
388                                            &inode);
389                 } else {
390                         return PR_1_ATTR_NO_EA_INODE_FL;
391                 }
392         }
393         return 0;
394 }
395
396 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
397                               struct ext2_ext_attr_entry *first, void *end)
398 {
399         struct ext2_ext_attr_entry *entry;
400
401         for (entry = first;
402              (void *)entry < end && !EXT2_EXT_IS_LAST_ENTRY(entry);
403              entry = EXT2_EXT_ATTR_NEXT(entry)) {
404                 if (!entry->e_value_inum)
405                         continue;
406                 if (!ctx->ea_inode_refs) {
407                         pctx->errcode = ea_refcount_create(0,
408                                                            &ctx->ea_inode_refs);
409                         if (pctx->errcode) {
410                                 pctx->num = 4;
411                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
412                                 ctx->flags |= E2F_FLAG_ABORT;
413                                 return;
414                         }
415                 }
416                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
417                                       0);
418         }
419 }
420
421 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
422                               struct ea_quota *ea_ibody_quota)
423 {
424         struct ext2_super_block *sb = ctx->fs->super;
425         struct ext2_inode_large *inode;
426         struct ext2_ext_attr_entry *entry;
427         char *start, *header, *end;
428         unsigned int storage_size, remain;
429         problem_t problem = 0;
430         region_t region = 0;
431
432         ea_ibody_quota->blocks = 0;
433         ea_ibody_quota->inodes = 0;
434
435         inode = (struct ext2_inode_large *) pctx->inode;
436         storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
437                 inode->i_extra_isize;
438         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
439                  inode->i_extra_isize;
440         end = header + storage_size;
441         start = header + sizeof(__u32);
442         entry = (struct ext2_ext_attr_entry *) start;
443
444         /* scan all entry's headers first */
445
446         /* take finish entry 0UL into account */
447         remain = storage_size - sizeof(__u32);
448
449         region = region_create(0, storage_size);
450         if (!region) {
451                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
452                 problem = 0;
453                 ctx->flags |= E2F_FLAG_ABORT;
454                 return;
455         }
456         if (region_allocate(region, 0, sizeof(__u32))) {
457                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
458                 goto fix;
459         }
460
461         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
462                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
463                 __u32 hash;
464
465                 if (region_allocate(region, (char *)entry - (char *)header,
466                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
467                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
468                         goto fix;
469                 }
470
471                 /* header eats this space */
472                 remain -= sizeof(struct ext2_ext_attr_entry);
473
474                 /* is attribute name valid? */
475                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
476                         pctx->num = entry->e_name_len;
477                         problem = PR_1_ATTR_NAME_LEN;
478                         goto fix;
479                 }
480
481                 /* attribute len eats this space */
482                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
483
484                 if (entry->e_value_inum == 0) {
485                         /* check value size */
486                         if (entry->e_value_size > remain) {
487                                 pctx->num = entry->e_value_size;
488                                 problem = PR_1_ATTR_VALUE_SIZE;
489                                 goto fix;
490                         }
491
492                         if (entry->e_value_size &&
493                             region_allocate(region,
494                                             sizeof(__u32) + entry->e_value_offs,
495                                             EXT2_EXT_ATTR_SIZE(
496                                                 entry->e_value_size))) {
497                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
498                                 goto fix;
499                         }
500
501                         hash = ext2fs_ext_attr_hash_entry(entry,
502                                                           start + entry->e_value_offs);
503
504                         /* e_hash may be 0 in older inode's ea */
505                         if (entry->e_hash != 0 && entry->e_hash != hash) {
506                                 pctx->num = entry->e_hash;
507                                 problem = PR_1_ATTR_HASH;
508                                 goto fix;
509                         }
510                 } else {
511                         blk64_t quota_blocks;
512
513                         problem = check_large_ea_inode(ctx, entry, pctx,
514                                                        &quota_blocks);
515                         if (problem != 0)
516                                 goto fix;
517
518                         ea_ibody_quota->blocks += quota_blocks;
519                         ea_ibody_quota->inodes++;
520                 }
521
522                 /* If EA value is stored in external inode then it does not
523                  * consume space here */
524                 if (entry->e_value_inum == 0)
525                         remain -= entry->e_value_size;
526
527                 entry = EXT2_EXT_ATTR_NEXT(entry);
528         }
529
530         if (region_allocate(region, (char *)entry - (char *)header,
531                             sizeof(__u32))) {
532                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
533                 goto fix;
534         }
535 fix:
536         if (region)
537                 region_free(region);
538         /*
539          * it seems like a corruption. it's very unlikely we could repair
540          * EA(s) in automatic fashion -bzzz
541          */
542         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
543                 inc_ea_inode_refs(ctx, pctx,
544                                   (struct ext2_ext_attr_entry *)start, end);
545                 return;
546         }
547
548         /* simply remove all possible EA(s) */
549         *((__u32 *)header) = 0UL;
550         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
551                                 EXT2_INODE_SIZE(sb), "pass1");
552         ea_ibody_quota->blocks = 0;
553         ea_ibody_quota->inodes = 0;
554 }
555
556 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
557         return (xtime & (1U << 31)) != 0 &&
558                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
559 }
560
561 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
562         check_inode_extra_negative_epoch(inode->i_##xtime, \
563                                          inode->i_##xtime##_extra)
564
565 /* When today's date is earlier than 2242, we assume that atimes,
566  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
567  * actually pre-1970 dates mis-encoded.
568  */
569 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
570
571 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
572                                     struct ea_quota *ea_ibody_quota)
573 {
574         struct ext2_super_block *sb = ctx->fs->super;
575         struct ext2_inode_large *inode;
576         __u32 *eamagic;
577         int min, max;
578
579         ea_ibody_quota->blocks = 0;
580         ea_ibody_quota->inodes = 0;
581
582         inode = (struct ext2_inode_large *) pctx->inode;
583         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
584                 /* this isn't large inode. so, nothing to check */
585                 return;
586         }
587
588 #if 0
589         printf("inode #%u, i_extra_size %d\n", pctx->ino,
590                         inode->i_extra_isize);
591 #endif
592         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
593         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
594         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
595         /*
596          * For now we will allow i_extra_isize to be 0, but really
597          * implementations should never allow i_extra_isize to be 0
598          */
599         if (inode->i_extra_isize &&
600             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
601              inode->i_extra_isize & 3)) {
602                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
603                         return;
604                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
605                         inode->i_extra_isize = sb->s_want_extra_isize;
606                 else
607                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
608                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
609                                         EXT2_INODE_SIZE(sb), "pass1");
610         }
611
612         /* check if there is no place for an EA header */
613         if (inode->i_extra_isize >= max - sizeof(__u32))
614                 return;
615
616         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
617                         inode->i_extra_isize);
618         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
619                 /* it seems inode has an extended attribute(s) in body */
620                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
621         }
622
623         /*
624          * If the inode's extended atime (ctime, crtime, mtime) is stored in
625          * the old, invalid format, repair it.
626          */
627         if (((sizeof(time_t) <= 4) ||
628              (((sizeof(time_t) > 4) &&
629                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
630             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
631              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
632              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
633              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
634
635                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
636                         return;
637
638                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
639                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
640                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
641                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
642                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
643                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
644                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
645                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
646                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
647                                         EXT2_INODE_SIZE(sb), "pass1");
648         }
649
650 }
651
652 /*
653  * Check to see if the inode might really be a directory, despite i_mode
654  *
655  * This is a lot of complexity for something for which I'm not really
656  * convinced happens frequently in the wild.  If for any reason this
657  * causes any problems, take this code out.
658  * [tytso:20070331.0827EDT]
659  */
660 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
661                                 char *buf)
662 {
663         struct ext2_inode *inode = pctx->inode;
664         struct ext2_dir_entry   *dirent;
665         errcode_t               retval;
666         blk64_t                 blk;
667         unsigned int            i, rec_len, not_device = 0;
668         int                     extent_fs;
669         int                     inlinedata_fs;
670
671         /*
672          * If the mode looks OK, we believe it.  If the first block in
673          * the i_block array is 0, this cannot be a directory. If the
674          * inode is extent-mapped, it is still the case that the latter
675          * cannot be 0 - the magic number in the extent header would make
676          * it nonzero.
677          */
678         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
679             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
680                 return;
681
682         /* 
683          * Check the block numbers in the i_block array for validity:
684          * zero blocks are skipped (but the first one cannot be zero -
685          * see above), other blocks are checked against the first and
686          * max data blocks (from the the superblock) and against the
687          * block bitmap. Any invalid block found means this cannot be
688          * a directory.
689          * 
690          * If there are non-zero blocks past the fourth entry, then
691          * this cannot be a device file: we remember that for the next
692          * check.
693          *
694          * For extent mapped files, we don't do any sanity checking:
695          * just try to get the phys block of logical block 0 and run
696          * with it.
697          *
698          * For inline data files, we just try to get the size of inline
699          * data.  If it's true, we will treat it as a directory.
700          */
701
702         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
703         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
704         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
705                 size_t size;
706                 __u32 dotdot;
707                 unsigned int rec_len2;
708                 struct ext2_dir_entry de;
709
710                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
711                         return;
712                 /*
713                  * If the size isn't a multiple of 4, it's probably not a
714                  * directory??
715                  */
716                 if (size & 3)
717                         return;
718                 /*
719                  * If the first 10 bytes don't look like a directory entry,
720                  * it's probably not a directory.
721                  */
722                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
723                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
724                        EXT2_DIR_REC_LEN(0));
725                 dotdot = ext2fs_le32_to_cpu(dotdot);
726                 de.inode = ext2fs_le32_to_cpu(de.inode);
727                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
728                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
729                 if (dotdot >= ctx->fs->super->s_inodes_count ||
730                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
731                      dotdot != EXT2_ROOT_INO) ||
732                     de.inode >= ctx->fs->super->s_inodes_count ||
733                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
734                      de.inode != 0) ||
735                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
736                               EXT4_INLINE_DATA_DOTDOT_SIZE)
737                         return;
738                 /* device files never have a "system.data" entry */
739                 goto isdir;
740         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
741                 /* extent mapped */
742                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
743                                  &blk))
744                         return;
745                 /* device files are never extent mapped */
746                 not_device++;
747         } else {
748                 for (i=0; i < EXT2_N_BLOCKS; i++) {
749                         blk = inode->i_block[i];
750                         if (!blk)
751                                 continue;
752                         if (i >= 4)
753                                 not_device++;
754
755                         if (blk < ctx->fs->super->s_first_data_block ||
756                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
757                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
758                                                            blk))
759                                 return; /* Invalid block, can't be dir */
760                 }
761                 blk = inode->i_block[0];
762         }
763
764         /*
765          * If the mode says this is a device file and the i_links_count field
766          * is sane and we have not ruled it out as a device file previously,
767          * we declare it a device file, not a directory.
768          */
769         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
770             (inode->i_links_count == 1) && !not_device)
771                 return;
772
773         /* read the first block */
774         ehandler_operation(_("reading directory block"));
775         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
776         ehandler_operation(0);
777         if (retval)
778                 return;
779
780         dirent = (struct ext2_dir_entry *) buf;
781         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
782         if (retval)
783                 return;
784         if ((ext2fs_dirent_name_len(dirent) != 1) ||
785             (dirent->name[0] != '.') ||
786             (dirent->inode != pctx->ino) ||
787             (rec_len < 12) ||
788             (rec_len % 4) ||
789             (rec_len >= ctx->fs->blocksize - 12))
790                 return;
791
792         dirent = (struct ext2_dir_entry *) (buf + rec_len);
793         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
794         if (retval)
795                 return;
796         if ((ext2fs_dirent_name_len(dirent) != 2) ||
797             (dirent->name[0] != '.') ||
798             (dirent->name[1] != '.') ||
799             (rec_len < 12) ||
800             (rec_len % 4))
801                 return;
802
803 isdir:
804         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
805                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
806                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
807                                         EXT2_INODE_SIZE(ctx->fs->super),
808                                         "check_is_really_dir");
809         }
810 }
811
812 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
813                                      int flags, ext2_icount_t hint,
814                                      ext2_icount_t *ret)
815 {
816         unsigned int            threshold;
817         unsigned int            save_type;
818         ext2_ino_t              num_dirs;
819         errcode_t               retval;
820         char                    *tdb_dir;
821         int                     enable;
822         int                     full_map;
823
824         *ret = 0;
825
826         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
827                            &tdb_dir);
828         profile_get_uint(ctx->profile, "scratch_files",
829                          "numdirs_threshold", 0, 0, &threshold);
830         profile_get_boolean(ctx->profile, "scratch_files",
831                             "icount", 0, 1, &enable);
832
833         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
834         if (retval)
835                 num_dirs = 1024;        /* Guess */
836
837         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
838             (!threshold || num_dirs > threshold)) {
839                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
840                                                   flags, ret);
841                 if (retval == 0)
842                         return 0;
843         }
844         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
845                                &save_type);
846         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
847                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
848         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
849         ctx->fs->default_bitmap_type = save_type;
850         return retval;
851 }
852
853 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
854                                             e2fsck_t ctx,
855                                             struct problem_context *pctx)
856 {
857         errcode_t retval;
858         struct ext2_inode_large inode;
859
860         /*
861          * Reread inode.  If we don't see checksum error, then this inode
862          * has been fixed elsewhere.
863          */
864         ctx->stashed_ino = 0;
865         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
866                                         sizeof(inode));
867         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
868                 return retval;
869         if (!retval)
870                 return 0;
871
872         /*
873          * Checksum still doesn't match.  That implies that the inode passes
874          * all the sanity checks, so maybe the checksum is simply corrupt.
875          * See if the user will go for fixing that.
876          */
877         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
878                 return 0;
879
880         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
881                                          sizeof(inode));
882         return retval;
883 }
884
885 static void reserve_block_for_root_repair(e2fsck_t ctx)
886 {
887         blk64_t         blk = 0;
888         errcode_t       err;
889         ext2_filsys     fs = ctx->fs;
890
891         ctx->root_repair_block = 0;
892         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
893                 return;
894
895         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
896         if (err)
897                 return;
898         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
899         ctx->root_repair_block = blk;
900 }
901
902 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
903 {
904         blk64_t         blk = 0;
905         errcode_t       err;
906         ext2_filsys     fs = ctx->fs;
907         static const char name[] = "lost+found";
908         ext2_ino_t      ino;
909
910         ctx->lnf_repair_block = 0;
911         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
912                 return;
913
914         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
915         if (err)
916                 return;
917         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
918         ctx->lnf_repair_block = blk;
919 }
920
921 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
922                                          size_t *sz)
923 {
924         void *p;
925         struct ext2_xattr_handle *handle;
926         errcode_t retval;
927
928         retval = ext2fs_xattrs_open(fs, ino, &handle);
929         if (retval)
930                 return retval;
931
932         retval = ext2fs_xattrs_read(handle);
933         if (retval)
934                 goto err;
935
936         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
937         if (retval)
938                 goto err;
939         ext2fs_free_mem(&p);
940 err:
941         (void) ext2fs_xattrs_close(&handle);
942         return retval;
943 }
944
945 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
946                                     struct problem_context *pctx,
947                                     int failed_csum)
948 {
949         if (!failed_csum)
950                 return;
951
952         /*
953          * If the inode failed the checksum and the user didn't
954          * clear the inode, test the checksum again -- if it still
955          * fails, ask the user if the checksum should be corrected.
956          */
957         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
958         if (pctx->errcode)
959                 ctx->flags |= E2F_FLAG_ABORT;
960 }
961 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
962         do { \
963                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
964                 if ((ctx)->flags & E2F_FLAG_ABORT) \
965                         return; \
966         } while (0)
967
968 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
969 {
970         __u32 x;
971         int i;
972
973         for (i = 0; i < EXT2_N_BLOCKS; i++) {
974                 x = inode->i_block[i];
975 #ifdef WORDS_BIGENDIAN
976                 x = ext2fs_swab32(x);
977 #endif
978                 if (x >= ext2fs_blocks_count(fs->super))
979                         return 0;
980         }
981
982         return 1;
983 }
984
985 /*
986  * Figure out what to do with an inode that has both extents and inline data
987  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
988  */
989 static int fix_inline_data_extents_file(e2fsck_t ctx,
990                                         ext2_ino_t ino,
991                                         struct ext2_inode *inode,
992                                         int inode_size,
993                                         struct problem_context *pctx)
994 {
995         size_t max_inline_ea_size;
996         ext2_filsys fs = ctx->fs;
997         int dirty = 0;
998
999         /* Both feature flags not set?  Just run the regular checks */
1000         if (!ext2fs_has_feature_extents(fs->super) &&
1001             !ext2fs_has_feature_inline_data(fs->super))
1002                 return 0;
1003
1004         /* Clear both flags if it's a special file */
1005         if (LINUX_S_ISCHR(inode->i_mode) ||
1006             LINUX_S_ISBLK(inode->i_mode) ||
1007             LINUX_S_ISFIFO(inode->i_mode) ||
1008             LINUX_S_ISSOCK(inode->i_mode)) {
1009                 check_extents_inlinedata(ctx, pctx);
1010                 return 0;
1011         }
1012
1013         /* If it looks like an extent tree, try to clear inlinedata */
1014         if (ext2fs_extent_header_verify(inode->i_block,
1015                                  sizeof(inode->i_block)) == 0 &&
1016             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1017                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1018                 dirty = 1;
1019                 goto out;
1020         }
1021
1022         /* If it looks short enough to be inline data, try to clear extents */
1023         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1024                 max_inline_ea_size = inode_size -
1025                                      (EXT2_GOOD_OLD_INODE_SIZE +
1026                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1027         else
1028                 max_inline_ea_size = 0;
1029         if (EXT2_I_SIZE(inode) <
1030             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1031             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1032                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1033                 dirty = 1;
1034                 goto out;
1035         }
1036
1037         /*
1038          * Too big for inline data, but no evidence of extent tree -
1039          * maybe it's a block map file?  If the mappings all look valid?
1040          */
1041         if (could_be_block_map(fs, inode) &&
1042             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1043 #ifdef WORDS_BIGENDIAN
1044                 int i;
1045
1046                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1047                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1048 #endif
1049
1050                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1051                 dirty = 1;
1052                 goto out;
1053         }
1054
1055         /* Oh well, just clear the busted inode. */
1056         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1057                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1058                 return -1;
1059         }
1060
1061 out:
1062         if (dirty)
1063                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1064
1065         return 0;
1066 }
1067
1068 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1069 {
1070         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1071         dgrp_t start = *group, grp;
1072         blk64_t blocks_to_read = 0;
1073         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1074
1075         if (ctx->readahead_kb == 0)
1076                 goto out;
1077
1078         /* Keep iterating groups until we have enough to readahead */
1079         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1080         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1081                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1082                         continue;
1083                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1084                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1085                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1086                                         inodes_per_block;
1087                 if (blocks_to_read * ctx->fs->blocksize >
1088                     ctx->readahead_kb * 1024)
1089                         break;
1090         }
1091
1092         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1093                                grp - start + 1);
1094         if (err == EAGAIN) {
1095                 ctx->readahead_kb /= 2;
1096                 err = 0;
1097         }
1098
1099 out:
1100         if (err) {
1101                 /* Error; disable itable readahead */
1102                 *group = ctx->fs->group_desc_count;
1103                 *next_ino = ctx->fs->super->s_inodes_count;
1104         } else {
1105                 /*
1106                  * Don't do more readahead until we've reached the first inode
1107                  * of the last inode scan buffer block for the last group.
1108                  */
1109                 *group = grp + 1;
1110                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1111                                      ctx->inode_buffer_blocks :
1112                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1113                                     ctx->fs->blocksize /
1114                                     EXT2_INODE_SIZE(ctx->fs->super);
1115                 inodes_in_group--;
1116                 *next_ino = inodes_in_group -
1117                             (inodes_in_group % inodes_per_buffer) + 1 +
1118                             (grp * ctx->fs->super->s_inodes_per_group);
1119         }
1120 }
1121
1122 /*
1123  * Check if the passed ino is one of the used superblock quota inodes.
1124  *
1125  * Before the quota inodes were journaled, older superblock quota inodes
1126  * were just regular files in the filesystem and not reserved inodes.  This
1127  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1128  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1129  */
1130 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1131 {
1132         enum quota_type qtype;
1133
1134         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1135                 if (*quota_sb_inump(sb, qtype) == ino)
1136                         return 1;
1137
1138         return 0;
1139 }
1140
1141 /*
1142  * Check if the passed ino is one of the reserved quota inodes.
1143  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1144  * inodes.  These inodes may or may not be in use by the quota feature.
1145  */
1146 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1147 {
1148         enum quota_type qtype;
1149
1150         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1151                 if (quota_type2inum(qtype, fs->super) == ino)
1152                         return 1;
1153
1154         return 0;
1155 }
1156
1157 void e2fsck_pass1(e2fsck_t ctx)
1158 {
1159         int     i;
1160         __u64   max_sizes;
1161         ext2_filsys fs = ctx->fs;
1162         ext2_ino_t      ino = 0;
1163         struct ext2_inode *inode = NULL;
1164         ext2_inode_scan scan = NULL;
1165         char            *block_buf = NULL;
1166 #ifdef RESOURCE_TRACK
1167         struct resource_track   rtrack;
1168 #endif
1169         unsigned char   frag, fsize;
1170         struct          problem_context pctx;
1171         struct          scan_callback_struct scan_struct;
1172         struct ext2_super_block *sb = ctx->fs->super;
1173         const char      *old_op;
1174         int             imagic_fs, extent_fs, inlinedata_fs;
1175         int             low_dtime_check = 1;
1176         int             inode_size = EXT2_INODE_SIZE(fs->super);
1177         int             bufsize;
1178         int             failed_csum = 0;
1179         ext2_ino_t      ino_threshold = 0;
1180         dgrp_t          ra_group = 0;
1181         struct ea_quota ea_ibody_quota;
1182
1183         init_resource_track(&rtrack, ctx->fs->io);
1184         clear_problem_context(&pctx);
1185
1186         /* If we can do readahead, figure out how many groups to pull in. */
1187         if (!e2fsck_can_readahead(ctx->fs))
1188                 ctx->readahead_kb = 0;
1189         else if (ctx->readahead_kb == ~0ULL)
1190                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1191         pass1_readahead(ctx, &ra_group, &ino_threshold);
1192
1193         if (!(ctx->options & E2F_OPT_PREEN))
1194                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1195
1196         if (ext2fs_has_feature_dir_index(fs->super) &&
1197             !(ctx->options & E2F_OPT_NO)) {
1198                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1199                         ctx->dirs_to_hash = 0;
1200         }
1201
1202 #ifdef MTRACE
1203         mtrace_print("Pass 1");
1204 #endif
1205
1206 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1207
1208         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1209                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1210                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1211                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1212                 max_sizes = (max_sizes * (1UL << i));
1213                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1214         }
1215 #undef EXT2_BPP
1216
1217         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1218         extent_fs = ext2fs_has_feature_extents(sb);
1219         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1220
1221         /*
1222          * Allocate bitmaps structures
1223          */
1224         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1225                                                     EXT2FS_BMAP64_RBTREE,
1226                                                     "inode_used_map",
1227                                                     &ctx->inode_used_map);
1228         if (pctx.errcode) {
1229                 pctx.num = 1;
1230                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1231                 ctx->flags |= E2F_FLAG_ABORT;
1232                 return;
1233         }
1234         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1235                         _("directory inode map"),
1236                         EXT2FS_BMAP64_AUTODIR,
1237                         "inode_dir_map", &ctx->inode_dir_map);
1238         if (pctx.errcode) {
1239                 pctx.num = 2;
1240                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1241                 ctx->flags |= E2F_FLAG_ABORT;
1242                 return;
1243         }
1244         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1245                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1246                         "inode_reg_map", &ctx->inode_reg_map);
1247         if (pctx.errcode) {
1248                 pctx.num = 6;
1249                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1250                 ctx->flags |= E2F_FLAG_ABORT;
1251                 return;
1252         }
1253         pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1254                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1255                         "block_found_map", &ctx->block_found_map);
1256         if (pctx.errcode) {
1257                 pctx.num = 1;
1258                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1259                 ctx->flags |= E2F_FLAG_ABORT;
1260                 return;
1261         }
1262         pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1263                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1264                         "block_metadata_map", &ctx->block_metadata_map);
1265         if (pctx.errcode) {
1266                 pctx.num = 1;
1267                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1268                 ctx->flags |= E2F_FLAG_ABORT;
1269                 return;
1270         }
1271         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1272                                            &ctx->inode_link_info);
1273         if (pctx.errcode) {
1274                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1275                 ctx->flags |= E2F_FLAG_ABORT;
1276                 return;
1277         }
1278         bufsize = inode_size;
1279         if (bufsize < sizeof(struct ext2_inode_large))
1280                 bufsize = sizeof(struct ext2_inode_large);
1281         inode = (struct ext2_inode *)
1282                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1283
1284         inodes_to_process = (struct process_inode_block *)
1285                 e2fsck_allocate_memory(ctx,
1286                                        (ctx->process_inode_size *
1287                                         sizeof(struct process_inode_block)),
1288                                        "array of inodes to process");
1289         process_inode_count = 0;
1290
1291         pctx.errcode = ext2fs_init_dblist(fs, 0);
1292         if (pctx.errcode) {
1293                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1294                 ctx->flags |= E2F_FLAG_ABORT;
1295                 goto endit;
1296         }
1297
1298         /*
1299          * If the last orphan field is set, clear it, since the pass1
1300          * processing will automatically find and clear the orphans.
1301          * In the future, we may want to try using the last_orphan
1302          * linked list ourselves, but for now, we clear it so that the
1303          * ext3 mount code won't get confused.
1304          */
1305         if (!(ctx->options & E2F_OPT_READONLY)) {
1306                 if (fs->super->s_last_orphan) {
1307                         fs->super->s_last_orphan = 0;
1308                         ext2fs_mark_super_dirty(fs);
1309                 }
1310         }
1311
1312         mark_table_blocks(ctx);
1313         pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1314                                                 &ctx->block_found_map);
1315         if (pctx.errcode) {
1316                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1317                 ctx->flags |= E2F_FLAG_ABORT;
1318                 goto endit;
1319         }
1320         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1321                                                     "block interate buffer");
1322         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1323                 e2fsck_use_inode_shortcuts(ctx, 1);
1324         e2fsck_intercept_block_allocations(ctx);
1325         old_op = ehandler_operation(_("opening inode scan"));
1326         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1327                                               &scan);
1328         ehandler_operation(old_op);
1329         if (pctx.errcode) {
1330                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1331                 ctx->flags |= E2F_FLAG_ABORT;
1332                 goto endit;
1333         }
1334         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1335                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1336         ctx->stashed_inode = inode;
1337         scan_struct.ctx = ctx;
1338         scan_struct.block_buf = block_buf;
1339         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1340         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1341                                               ctx->fs->group_desc_count)))
1342                 goto endit;
1343         if ((fs->super->s_wtime < fs->super->s_inodes_count) ||
1344             (fs->super->s_mtime < fs->super->s_inodes_count) ||
1345             (fs->super->s_mkfs_time &&
1346              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1347                 low_dtime_check = 0;
1348
1349         if (ext2fs_has_feature_mmp(fs->super) &&
1350             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1351             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1352                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1353                                           fs->super->s_mmp_block);
1354
1355         /* Set up ctx->lost_and_found if possible */
1356         (void) e2fsck_get_lost_and_found(ctx, 0);
1357
1358         while (1) {
1359                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1360                         if (e2fsck_mmp_update(fs))
1361                                 fatal_error(ctx, 0);
1362                 }
1363                 old_op = ehandler_operation(_("getting next inode from scan"));
1364                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1365                                                           inode, inode_size);
1366                 if (ino > ino_threshold)
1367                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1368                 ehandler_operation(old_op);
1369                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1370                         goto endit;
1371                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1372                         /*
1373                          * If badblocks says badblocks is bad, offer to clear
1374                          * the list, update the in-core bb list, and restart
1375                          * the inode scan.
1376                          */
1377                         if (ino == EXT2_BAD_INO &&
1378                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1379                                         &pctx)) {
1380                                 errcode_t err;
1381
1382                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1383                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1384                                 ctx->fs->badblocks = NULL;
1385                                 err = ext2fs_read_bb_inode(ctx->fs,
1386                                                         &ctx->fs->badblocks);
1387                                 if (err) {
1388                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1389                                                     &pctx);
1390                                         ctx->flags |= E2F_FLAG_ABORT;
1391                                         goto endit;
1392                                 }
1393                                 err = ext2fs_inode_scan_goto_blockgroup(scan,
1394                                                                         0);
1395                                 if (err) {
1396                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1397                                                     &pctx);
1398                                         ctx->flags |= E2F_FLAG_ABORT;
1399                                         goto endit;
1400                                 }
1401                                 continue;
1402                         }
1403                         if (!ctx->inode_bb_map)
1404                                 alloc_bb_map(ctx);
1405                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1406                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1407                         continue;
1408                 }
1409                 if (pctx.errcode &&
1410                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1411                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1412                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1413                         ctx->flags |= E2F_FLAG_ABORT;
1414                         goto endit;
1415                 }
1416                 if (!ino)
1417                         break;
1418                 pctx.ino = ino;
1419                 pctx.inode = inode;
1420                 ctx->stashed_ino = ino;
1421
1422                 /* Clear trashed inode? */
1423                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1424                     inode->i_links_count > 0 &&
1425                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1426                         pctx.errcode = 0;
1427                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1428                 }
1429                 failed_csum = pctx.errcode != 0;
1430
1431                 /*
1432                  * Check for inodes who might have been part of the
1433                  * orphaned list linked list.  They should have gotten
1434                  * dealt with by now, unless the list had somehow been
1435                  * corrupted.
1436                  *
1437                  * FIXME: In the future, inodes which are still in use
1438                  * (and which are therefore) pending truncation should
1439                  * be handled specially.  Right now we just clear the
1440                  * dtime field, and the normal e2fsck handling of
1441                  * inodes where i_size and the inode blocks are
1442                  * inconsistent is to fix i_size, instead of releasing
1443                  * the extra blocks.  This won't catch the inodes that
1444                  * was at the end of the orphan list, but it's better
1445                  * than nothing.  The right answer is that there
1446                  * shouldn't be any bugs in the orphan list handling.  :-)
1447                  */
1448                 if (inode->i_dtime && low_dtime_check &&
1449                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1450                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1451                                 inode->i_dtime = inode->i_links_count ?
1452                                         0 : ctx->now;
1453                                 e2fsck_write_inode(ctx, ino, inode,
1454                                                    "pass1");
1455                                 failed_csum = 0;
1456                         }
1457                 }
1458
1459                 if (inode->i_links_count) {
1460                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1461                                            ino, inode->i_links_count);
1462                         if (pctx.errcode) {
1463                                 pctx.num = inode->i_links_count;
1464                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1465                                 ctx->flags |= E2F_FLAG_ABORT;
1466                                 goto endit;
1467                         }
1468                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1469                            !quota_inum_is_reserved(fs, ino)) {
1470                         if (!inode->i_dtime && inode->i_mode) {
1471                                 if (fix_problem(ctx,
1472                                             PR_1_ZERO_DTIME, &pctx)) {
1473                                         inode->i_dtime = ctx->now;
1474                                         e2fsck_write_inode(ctx, ino, inode,
1475                                                            "pass1");
1476                                         failed_csum = 0;
1477                                 }
1478                         }
1479                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1480                         continue;
1481                 }
1482
1483                 /* Conflicting inlinedata/extents inode flags? */
1484                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1485                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1486                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1487                                                                inode_size,
1488                                                                &pctx);
1489                         if (res < 0) {
1490                                 /* skip FINISH_INODE_LOOP */
1491                                 continue;
1492                         }
1493                 }
1494
1495                 /* Test for incorrect inline_data flags settings. */
1496                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1497                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1498                         size_t size = 0;
1499
1500                         pctx.errcode = ext2fs_inline_data_size(fs, ino, &size);
1501                         if (!pctx.errcode && size &&
1502                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1503                                 ext2fs_set_feature_inline_data(sb);
1504                                 ext2fs_mark_super_dirty(fs);
1505                                 inlinedata_fs = 1;
1506                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1507                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1508                                 /* skip FINISH_INODE_LOOP */
1509                                 continue;
1510                         }
1511                 }
1512
1513                 /* Test for inline data flag but no attr */
1514                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1515                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1516                         size_t size = 0;
1517                         errcode_t err;
1518                         int flags;
1519
1520                         flags = fs->flags;
1521                         if (failed_csum)
1522                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1523                         err = get_inline_data_ea_size(fs, ino, &size);
1524                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1525                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1526
1527                         switch (err) {
1528                         case 0:
1529                                 /* Everything is awesome... */
1530                                 break;
1531                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1532                         case EXT2_ET_BAD_EA_HASH:
1533                         case EXT2_ET_BAD_EA_HEADER:
1534                         case EXT2_ET_EA_BAD_NAME_LEN:
1535                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1536                         case EXT2_ET_EA_KEY_NOT_FOUND:
1537                         case EXT2_ET_EA_NO_SPACE:
1538                         case EXT2_ET_MISSING_EA_FEATURE:
1539                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1540                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1541                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1542                         case EXT2_ET_NO_INLINE_DATA:
1543                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1544                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1545                         case EXT2_ET_EA_INODE_CORRUPTED:
1546                                 /* broken EA or no system.data EA; truncate */
1547                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1548                                                 &pctx)) {
1549                                         err = ext2fs_inode_size_set(fs, inode, 0);
1550                                         if (err) {
1551                                                 pctx.errcode = err;
1552                                                 ctx->flags |= E2F_FLAG_ABORT;
1553                                                 goto endit;
1554                                         }
1555                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1556                                         memset(&inode->i_block, 0,
1557                                                sizeof(inode->i_block));
1558                                         e2fsck_write_inode(ctx, ino, inode,
1559                                                            "pass1");
1560                                         failed_csum = 0;
1561                                 }
1562                                 break;
1563                         default:
1564                                 /* Some other kind of non-xattr error? */
1565                                 pctx.errcode = err;
1566                                 ctx->flags |= E2F_FLAG_ABORT;
1567                                 goto endit;
1568                         }
1569                 }
1570
1571                 /*
1572                  * Test for incorrect extent flag settings.
1573                  *
1574                  * On big-endian machines we must be careful:
1575                  * When the inode is read, the i_block array is not swapped
1576                  * if the extent flag is set.  Therefore if we are testing
1577                  * for or fixing a wrongly-set flag, we must potentially
1578                  * (un)swap before testing, or after fixing.
1579                  */
1580
1581                 /*
1582                  * In this case the extents flag was set when read, so
1583                  * extent_header_verify is ok.  If the inode is cleared,
1584                  * no need to swap... so no extra swapping here.
1585                  */
1586                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1587                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1588                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1589                         if ((ext2fs_extent_header_verify(inode->i_block,
1590                                                  sizeof(inode->i_block)) == 0) &&
1591                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1592                                 ext2fs_set_feature_extents(sb);
1593                                 ext2fs_mark_super_dirty(fs);
1594                                 extent_fs = 1;
1595                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1596                         clear_inode:
1597                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1598                                 if (ino == EXT2_BAD_INO)
1599                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1600                                                                  ino);
1601                                 /* skip FINISH_INODE_LOOP */
1602                                 continue;
1603                         }
1604                 }
1605
1606                 /*
1607                  * For big-endian machines:
1608                  * If the inode didn't have the extents flag set when it
1609                  * was read, then the i_blocks array was swapped.  To test
1610                  * as an extents header, we must swap it back first.
1611                  * IF we then set the extents flag, the entire i_block
1612                  * array must be un/re-swapped to make it proper extents data.
1613                  */
1614                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1615                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1616                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1617                     (LINUX_S_ISREG(inode->i_mode) ||
1618                      LINUX_S_ISDIR(inode->i_mode))) {
1619                         void *ehp;
1620 #ifdef WORDS_BIGENDIAN
1621                         __u32 tmp_block[EXT2_N_BLOCKS];
1622
1623                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1624                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1625                         ehp = tmp_block;
1626 #else
1627                         ehp = inode->i_block;
1628 #endif
1629                         if ((ext2fs_extent_header_verify(ehp,
1630                                          sizeof(inode->i_block)) == 0) &&
1631                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1632                                 inode->i_flags |= EXT4_EXTENTS_FL;
1633 #ifdef WORDS_BIGENDIAN
1634                                 memcpy(inode->i_block, tmp_block,
1635                                        sizeof(inode->i_block));
1636 #endif
1637                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1638                                 failed_csum = 0;
1639                         }
1640                 }
1641
1642                 if (ino == EXT2_BAD_INO) {
1643                         struct process_block_struct pb;
1644
1645                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1646                              inode->i_gid || inode->i_links_count ||
1647                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1648                              inode->i_file_acl) &&
1649                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1650                                 memset(inode, 0, sizeof(struct ext2_inode));
1651                                 e2fsck_write_inode(ctx, ino, inode,
1652                                                    "clear bad inode");
1653                                 failed_csum = 0;
1654                         }
1655
1656                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1657                                                           &pb.fs_meta_blocks);
1658                         if (pctx.errcode) {
1659                                 pctx.num = 4;
1660                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1661                                 ctx->flags |= E2F_FLAG_ABORT;
1662                                 goto endit;
1663                         }
1664                         pb.ino = EXT2_BAD_INO;
1665                         pb.num_blocks = pb.last_block = 0;
1666                         pb.last_db_block = -1;
1667                         pb.num_illegal_blocks = 0;
1668                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1669                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1670                         pb.inode = inode;
1671                         pb.pctx = &pctx;
1672                         pb.ctx = ctx;
1673                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1674                                      block_buf, process_bad_block, &pb);
1675                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1676                         if (pctx.errcode) {
1677                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1678                                 ctx->flags |= E2F_FLAG_ABORT;
1679                                 goto endit;
1680                         }
1681                         if (pb.bbcheck)
1682                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1683                                 ctx->flags |= E2F_FLAG_ABORT;
1684                                 goto endit;
1685                         }
1686                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1687                         clear_problem_context(&pctx);
1688                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1689                         continue;
1690                 } else if (ino == EXT2_ROOT_INO) {
1691                         /*
1692                          * Make sure the root inode is a directory; if
1693                          * not, offer to clear it.  It will be
1694                          * regenerated in pass #3.
1695                          */
1696                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1697                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1698                                         goto clear_inode;
1699                         }
1700                         /*
1701                          * If dtime is set, offer to clear it.  mke2fs
1702                          * version 0.2b created filesystems with the
1703                          * dtime field set for the root and lost+found
1704                          * directories.  We won't worry about
1705                          * /lost+found, since that can be regenerated
1706                          * easily.  But we will fix the root directory
1707                          * as a special case.
1708                          */
1709                         if (inode->i_dtime && inode->i_links_count) {
1710                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1711                                         inode->i_dtime = 0;
1712                                         e2fsck_write_inode(ctx, ino, inode,
1713                                                            "pass1");
1714                                         failed_csum = 0;
1715                                 }
1716                         }
1717                 } else if (ino == EXT2_JOURNAL_INO) {
1718                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1719                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1720                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1721                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1722                                                 &pctx)) {
1723                                         inode->i_mode = LINUX_S_IFREG;
1724                                         e2fsck_write_inode(ctx, ino, inode,
1725                                                            "pass1");
1726                                         failed_csum = 0;
1727                                 }
1728                                 check_blocks(ctx, &pctx, block_buf, NULL);
1729                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1730                                 continue;
1731                         }
1732                         if ((inode->i_links_count ||
1733                              inode->i_blocks || inode->i_block[0]) &&
1734                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1735                                         &pctx)) {
1736                                 memset(inode, 0, inode_size);
1737                                 ext2fs_icount_store(ctx->inode_link_info,
1738                                                     ino, 0);
1739                                 e2fsck_write_inode_full(ctx, ino, inode,
1740                                                         inode_size, "pass1");
1741                                 failed_csum = 0;
1742                         }
1743                 } else if (quota_inum_is_reserved(fs, ino)) {
1744                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1745                         if (ext2fs_has_feature_quota(fs->super) &&
1746                             quota_inum_is_super(fs->super, ino)) {
1747                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1748                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1749                                                         &pctx)) {
1750                                         inode->i_mode = LINUX_S_IFREG;
1751                                         e2fsck_write_inode(ctx, ino, inode,
1752                                                         "pass1");
1753                                         failed_csum = 0;
1754                                 }
1755                                 check_blocks(ctx, &pctx, block_buf, NULL);
1756                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1757                                 continue;
1758                         }
1759                         if ((inode->i_links_count ||
1760                              inode->i_blocks || inode->i_block[0]) &&
1761                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1762                                         &pctx)) {
1763                                 memset(inode, 0, inode_size);
1764                                 ext2fs_icount_store(ctx->inode_link_info,
1765                                                     ino, 0);
1766                                 e2fsck_write_inode_full(ctx, ino, inode,
1767                                                         inode_size, "pass1");
1768                                 failed_csum = 0;
1769                         }
1770                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1771                         problem_t problem = 0;
1772
1773                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1774                         if (ino == EXT2_BOOT_LOADER_INO) {
1775                                 if (LINUX_S_ISDIR(inode->i_mode))
1776                                         problem = PR_1_RESERVED_BAD_MODE;
1777                         } else if (ino == EXT2_RESIZE_INO) {
1778                                 if (inode->i_mode &&
1779                                     !LINUX_S_ISREG(inode->i_mode))
1780                                         problem = PR_1_RESERVED_BAD_MODE;
1781                         } else {
1782                                 if (inode->i_mode != 0)
1783                                         problem = PR_1_RESERVED_BAD_MODE;
1784                         }
1785                         if (problem) {
1786                                 if (fix_problem(ctx, problem, &pctx)) {
1787                                         inode->i_mode = 0;
1788                                         e2fsck_write_inode(ctx, ino, inode,
1789                                                            "pass1");
1790                                         failed_csum = 0;
1791                                 }
1792                         }
1793                         check_blocks(ctx, &pctx, block_buf, NULL);
1794                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1795                         continue;
1796                 }
1797
1798                 if (!inode->i_links_count) {
1799                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1800                         continue;
1801                 }
1802                 /*
1803                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1804                  * deleted files.  Oops.
1805                  *
1806                  * Since all new ext2 implementations get this right,
1807                  * we now assume that the case of non-zero
1808                  * i_links_count and non-zero dtime means that we
1809                  * should keep the file, not delete it.
1810                  *
1811                  */
1812                 if (inode->i_dtime) {
1813                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1814                                 inode->i_dtime = 0;
1815                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1816                                 failed_csum = 0;
1817                         }
1818                 }
1819
1820                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1821                 switch (fs->super->s_creator_os) {
1822                     case EXT2_OS_HURD:
1823                         frag = inode->osd2.hurd2.h_i_frag;
1824                         fsize = inode->osd2.hurd2.h_i_fsize;
1825                         break;
1826                     default:
1827                         frag = fsize = 0;
1828                 }
1829
1830                 if (inode->i_faddr || frag || fsize ||
1831                     (!ext2fs_has_feature_largedir(fs->super) &&
1832                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1833                         mark_inode_bad(ctx, ino);
1834                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1835                     !ext2fs_has_feature_64bit(fs->super) &&
1836                     inode->osd2.linux2.l_i_file_acl_high != 0)
1837                         mark_inode_bad(ctx, ino);
1838                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1839                     !ext2fs_has_feature_huge_file(fs->super) &&
1840                     (inode->osd2.linux2.l_i_blocks_hi != 0))
1841                         mark_inode_bad(ctx, ino);
1842                 if (inode->i_flags & EXT2_IMAGIC_FL) {
1843                         if (imagic_fs) {
1844                                 if (!ctx->inode_imagic_map)
1845                                         alloc_imagic_map(ctx);
1846                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1847                                                          ino);
1848                         } else {
1849                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1850                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
1851                                         e2fsck_write_inode(ctx, ino,
1852                                                            inode, "pass1");
1853                                         failed_csum = 0;
1854                                 }
1855                         }
1856                 }
1857
1858                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1859                 check_is_really_dir(ctx, &pctx, block_buf);
1860
1861                 /*
1862                  * ext2fs_inode_has_valid_blocks2 does not actually look
1863                  * at i_block[] values, so not endian-sensitive here.
1864                  */
1865                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1866                     LINUX_S_ISLNK(inode->i_mode) &&
1867                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1868                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1869                         inode->i_flags &= ~EXT4_EXTENTS_FL;
1870                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1871                         failed_csum = 0;
1872                 }
1873
1874                 if (LINUX_S_ISDIR(inode->i_mode)) {
1875                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1876                         e2fsck_add_dir_info(ctx, ino, 0);
1877                         ctx->fs_directory_count++;
1878                         if (inode->i_flags & EXT4_ENCRYPT_FL)
1879                                 add_encrypted_dir(ctx, ino);
1880                 } else if (LINUX_S_ISREG (inode->i_mode)) {
1881                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1882                         ctx->fs_regular_count++;
1883                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1884                            e2fsck_pass1_check_device_inode(fs, inode)) {
1885                         check_extents_inlinedata(ctx, &pctx);
1886                         check_immutable(ctx, &pctx);
1887                         check_size(ctx, &pctx);
1888                         ctx->fs_chardev_count++;
1889                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1890                            e2fsck_pass1_check_device_inode(fs, inode)) {
1891                         check_extents_inlinedata(ctx, &pctx);
1892                         check_immutable(ctx, &pctx);
1893                         check_size(ctx, &pctx);
1894                         ctx->fs_blockdev_count++;
1895                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1896                            e2fsck_pass1_check_symlink(fs, ino, inode,
1897                                                       block_buf)) {
1898                         check_immutable(ctx, &pctx);
1899                         ctx->fs_symlinks_count++;
1900                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1901                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1902                                 continue;
1903                         } else if (ext2fs_is_fast_symlink(inode)) {
1904                                 ctx->fs_fast_symlinks_count++;
1905                                 check_blocks(ctx, &pctx, block_buf,
1906                                              &ea_ibody_quota);
1907                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1908                                 continue;
1909                         }
1910                 }
1911                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
1912                          e2fsck_pass1_check_device_inode(fs, inode)) {
1913                         check_extents_inlinedata(ctx, &pctx);
1914                         check_immutable(ctx, &pctx);
1915                         check_size(ctx, &pctx);
1916                         ctx->fs_fifo_count++;
1917                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
1918                            e2fsck_pass1_check_device_inode(fs, inode)) {
1919                         check_extents_inlinedata(ctx, &pctx);
1920                         check_immutable(ctx, &pctx);
1921                         check_size(ctx, &pctx);
1922                         ctx->fs_sockets_count++;
1923                 } else
1924                         mark_inode_bad(ctx, ino);
1925                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1926                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
1927                         if (inode->i_block[EXT2_IND_BLOCK])
1928                                 ctx->fs_ind_count++;
1929                         if (inode->i_block[EXT2_DIND_BLOCK])
1930                                 ctx->fs_dind_count++;
1931                         if (inode->i_block[EXT2_TIND_BLOCK])
1932                                 ctx->fs_tind_count++;
1933                 }
1934                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1935                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
1936                     (inode->i_block[EXT2_IND_BLOCK] ||
1937                      inode->i_block[EXT2_DIND_BLOCK] ||
1938                      inode->i_block[EXT2_TIND_BLOCK] ||
1939                      ext2fs_file_acl_block(fs, inode))) {
1940                         struct process_inode_block *itp;
1941
1942                         itp = &inodes_to_process[process_inode_count];
1943                         itp->ino = ino;
1944                         itp->ea_ibody_quota = ea_ibody_quota;
1945                         if (inode_size < sizeof(struct ext2_inode_large))
1946                                 memcpy(&itp->inode, inode, inode_size);
1947                         else
1948                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
1949                         process_inode_count++;
1950                 } else
1951                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
1952
1953                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1954
1955                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1956                         goto endit;
1957
1958                 if (process_inode_count >= ctx->process_inode_size) {
1959                         process_inodes(ctx, block_buf);
1960
1961                         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1962                                 goto endit;
1963                 }
1964         }
1965         process_inodes(ctx, block_buf);
1966         ext2fs_close_inode_scan(scan);
1967         scan = NULL;
1968
1969         reserve_block_for_root_repair(ctx);
1970         reserve_block_for_lnf_repair(ctx);
1971
1972         /*
1973          * If any extended attribute blocks' reference counts need to
1974          * be adjusted, either up (ctx->refcount_extra), or down
1975          * (ctx->refcount), then fix them.
1976          */
1977         if (ctx->refcount) {
1978                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
1979                 ea_refcount_free(ctx->refcount);
1980                 ctx->refcount = 0;
1981         }
1982         if (ctx->refcount_extra) {
1983                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
1984                                         block_buf, +1);
1985                 ea_refcount_free(ctx->refcount_extra);
1986                 ctx->refcount_extra = 0;
1987         }
1988
1989         if (ctx->ea_block_quota_blocks) {
1990                 ea_refcount_free(ctx->ea_block_quota_blocks);
1991                 ctx->ea_block_quota_blocks = 0;
1992         }
1993
1994         if (ctx->ea_block_quota_inodes) {
1995                 ea_refcount_free(ctx->ea_block_quota_inodes);
1996                 ctx->ea_block_quota_inodes = 0;
1997         }
1998
1999         if (ctx->invalid_bitmaps)
2000                 handle_fs_bad_blocks(ctx);
2001
2002         /* We don't need the block_ea_map any more */
2003         if (ctx->block_ea_map) {
2004                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2005                 ctx->block_ea_map = 0;
2006         }
2007
2008         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2009                 clear_problem_context(&pctx);
2010                 pctx.errcode = ext2fs_create_resize_inode(fs);
2011                 if (pctx.errcode) {
2012                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2013                                          &pctx)) {
2014                                 ctx->flags |= E2F_FLAG_ABORT;
2015                                 goto endit;
2016                         }
2017                         pctx.errcode = 0;
2018                 }
2019                 if (!pctx.errcode) {
2020                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2021                                           "recreate inode");
2022                         inode->i_mtime = ctx->now;
2023                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2024                                            "recreate inode");
2025                 }
2026                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2027         }
2028
2029         if (ctx->flags & E2F_FLAG_RESTART) {
2030                 /*
2031                  * Only the master copy of the superblock and block
2032                  * group descriptors are going to be written during a
2033                  * restart, so set the superblock to be used to be the
2034                  * master superblock.
2035                  */
2036                 ctx->use_superblock = 0;
2037                 unwind_pass1(fs);
2038                 goto endit;
2039         }
2040
2041         if (ctx->block_dup_map) {
2042                 if (ctx->options & E2F_OPT_PREEN) {
2043                         clear_problem_context(&pctx);
2044                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2045                 }
2046                 e2fsck_pass1_dupblocks(ctx, block_buf);
2047         }
2048         ctx->flags |= E2F_FLAG_ALLOC_OK;
2049         ext2fs_free_mem(&inodes_to_process);
2050 endit:
2051         e2fsck_use_inode_shortcuts(ctx, 0);
2052
2053         if (scan)
2054                 ext2fs_close_inode_scan(scan);
2055         if (block_buf)
2056                 ext2fs_free_mem(&block_buf);
2057         if (inode)
2058                 ext2fs_free_mem(&inode);
2059
2060         /*
2061          * The l+f inode may have been cleared, so zap it now and
2062          * later passes will recalculate it if necessary
2063          */
2064         ctx->lost_and_found = 0;
2065
2066         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2067                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2068         else
2069                 ctx->invalid_bitmaps++;
2070 }
2071 #undef FINISH_INODE_LOOP
2072
2073 /*
2074  * When the inode_scan routines call this callback at the end of the
2075  * glock group, call process_inodes.
2076  */
2077 static errcode_t scan_callback(ext2_filsys fs,
2078                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
2079                                dgrp_t group, void * priv_data)
2080 {
2081         struct scan_callback_struct *scan_struct;
2082         e2fsck_t ctx;
2083
2084         scan_struct = (struct scan_callback_struct *) priv_data;
2085         ctx = scan_struct->ctx;
2086
2087         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf);
2088
2089         if (ctx->progress)
2090                 if ((ctx->progress)(ctx, 1, group+1,
2091                                     ctx->fs->group_desc_count))
2092                         return EXT2_ET_CANCEL_REQUESTED;
2093
2094         return 0;
2095 }
2096
2097 /*
2098  * Process the inodes in the "inodes to process" list.
2099  */
2100 static void process_inodes(e2fsck_t ctx, char *block_buf)
2101 {
2102         int                     i;
2103         struct ext2_inode       *old_stashed_inode;
2104         ext2_ino_t              old_stashed_ino;
2105         const char              *old_operation;
2106         char                    buf[80];
2107         struct problem_context  pctx;
2108
2109 #if 0
2110         printf("begin process_inodes: ");
2111 #endif
2112         if (process_inode_count == 0)
2113                 return;
2114         old_operation = ehandler_operation(0);
2115         old_stashed_inode = ctx->stashed_inode;
2116         old_stashed_ino = ctx->stashed_ino;
2117         qsort(inodes_to_process, process_inode_count,
2118                       sizeof(struct process_inode_block), process_inode_cmp);
2119         clear_problem_context(&pctx);
2120         for (i=0; i < process_inode_count; i++) {
2121                 pctx.inode = ctx->stashed_inode =
2122                         (struct ext2_inode *) &inodes_to_process[i].inode;
2123                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2124
2125 #if 0
2126                 printf("%u ", pctx.ino);
2127 #endif
2128                 sprintf(buf, _("reading indirect blocks of inode %u"),
2129                         pctx.ino);
2130                 ehandler_operation(buf);
2131                 check_blocks(ctx, &pctx, block_buf,
2132                              &inodes_to_process[i].ea_ibody_quota);
2133                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2134                         break;
2135         }
2136         ctx->stashed_inode = old_stashed_inode;
2137         ctx->stashed_ino = old_stashed_ino;
2138         process_inode_count = 0;
2139 #if 0
2140         printf("end process inodes\n");
2141 #endif
2142         ehandler_operation(old_operation);
2143 }
2144
2145 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
2146 {
2147         const struct process_inode_block *ib_a =
2148                 (const struct process_inode_block *) a;
2149         const struct process_inode_block *ib_b =
2150                 (const struct process_inode_block *) b;
2151         int     ret;
2152
2153         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
2154                ib_b->inode.i_block[EXT2_IND_BLOCK]);
2155         if (ret == 0)
2156                 /*
2157                  * We only call process_inodes() for non-extent
2158                  * inodes, so it's OK to pass NULL to
2159                  * ext2fs_file_acl_block() here.
2160                  */
2161                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
2162                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
2163         if (ret == 0)
2164                 ret = ib_a->ino - ib_b->ino;
2165         return ret;
2166 }
2167
2168 /*
2169  * Mark an inode as being bad in some what
2170  */
2171 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
2172 {
2173         struct          problem_context pctx;
2174
2175         if (!ctx->inode_bad_map) {
2176                 clear_problem_context(&pctx);
2177
2178                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2179                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
2180                                 "inode_bad_map", &ctx->inode_bad_map);
2181                 if (pctx.errcode) {
2182                         pctx.num = 3;
2183                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2184                         /* Should never get here */
2185                         ctx->flags |= E2F_FLAG_ABORT;
2186                         return;
2187                 }
2188         }
2189         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
2190 }
2191
2192 static void add_encrypted_dir(e2fsck_t ctx, ino_t ino)
2193 {
2194         struct          problem_context pctx;
2195
2196         if (!ctx->encrypted_dirs) {
2197                 pctx.errcode = ext2fs_u32_list_create(&ctx->encrypted_dirs, 0);
2198                 if (pctx.errcode)
2199                         goto error;
2200         }
2201         pctx.errcode = ext2fs_u32_list_add(ctx->encrypted_dirs, ino);
2202         if (pctx.errcode == 0)
2203                 return;
2204 error:
2205         fix_problem(ctx, PR_1_ALLOCATE_ENCRYPTED_DIRLIST, &pctx);
2206         /* Should never get here */
2207         ctx->flags |= E2F_FLAG_ABORT;
2208 }
2209
2210 /*
2211  * This procedure will allocate the inode "bb" (badblock) map table
2212  */
2213 static void alloc_bb_map(e2fsck_t ctx)
2214 {
2215         struct          problem_context pctx;
2216
2217         clear_problem_context(&pctx);
2218         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2219                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
2220                         "inode_bb_map", &ctx->inode_bb_map);
2221         if (pctx.errcode) {
2222                 pctx.num = 4;
2223                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2224                 /* Should never get here */
2225                 ctx->flags |= E2F_FLAG_ABORT;
2226                 return;
2227         }
2228 }
2229
2230 /*
2231  * This procedure will allocate the inode imagic table
2232  */
2233 static void alloc_imagic_map(e2fsck_t ctx)
2234 {
2235         struct          problem_context pctx;
2236
2237         clear_problem_context(&pctx);
2238         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2239                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
2240                         "inode_imagic_map", &ctx->inode_imagic_map);
2241         if (pctx.errcode) {
2242                 pctx.num = 5;
2243                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2244                 /* Should never get here */
2245                 ctx->flags |= E2F_FLAG_ABORT;
2246                 return;
2247         }
2248 }
2249
2250 /*
2251  * Marks a block as in use, setting the dup_map if it's been set
2252  * already.  Called by process_block and process_bad_block.
2253  *
2254  * WARNING: Assumes checks have already been done to make sure block
2255  * is valid.  This is true in both process_block and process_bad_block.
2256  */
2257 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
2258 {
2259         struct          problem_context pctx;
2260
2261         clear_problem_context(&pctx);
2262
2263         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
2264                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
2265                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
2266                         return;
2267                 }
2268                 if (!ctx->block_dup_map) {
2269                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
2270                                         _("multiply claimed block map"),
2271                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
2272                                         &ctx->block_dup_map);
2273                         if (pctx.errcode) {
2274                                 pctx.num = 3;
2275                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
2276                                             &pctx);
2277                                 /* Should never get here */
2278                                 ctx->flags |= E2F_FLAG_ABORT;
2279                                 return;
2280                         }
2281                 }
2282                 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
2283         } else {
2284                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
2285         }
2286 }
2287
2288 /*
2289  * When cluster size is greater than one block, it is caller's responsibility
2290  * to make sure block parameter starts at a cluster boundary.
2291  */
2292 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
2293                                       unsigned int num)
2294 {
2295         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
2296                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
2297         else {
2298                 int i;
2299                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
2300                         mark_block_used(ctx, block + i);
2301         }
2302 }
2303
2304 /*
2305  * Adjust the extended attribute block's reference counts at the end
2306  * of pass 1, either by subtracting out references for EA blocks that
2307  * are still referenced in ctx->refcount, or by adding references for
2308  * EA blocks that had extra references as accounted for in
2309  * ctx->refcount_extra.
2310  */
2311 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
2312                                     char *block_buf, int adjust_sign)
2313 {
2314         struct ext2_ext_attr_header     *header;
2315         struct problem_context          pctx;
2316         ext2_filsys                     fs = ctx->fs;
2317         blk64_t                         blk;
2318         __u32                           should_be;
2319         ea_value_t                      count;
2320
2321         clear_problem_context(&pctx);
2322
2323         ea_refcount_intr_begin(refcount);
2324         while (1) {
2325                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
2326                         break;
2327                 pctx.blk = blk;
2328                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
2329                                                      pctx.ino);
2330                 if (pctx.errcode) {
2331                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
2332                         return;
2333                 }
2334                 header = (struct ext2_ext_attr_header *) block_buf;
2335                 pctx.blkcount = header->h_refcount;
2336                 should_be = header->h_refcount + adjust_sign * (int)count;
2337                 pctx.num = should_be;
2338                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
2339                         header->h_refcount = should_be;
2340                         pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
2341                                                              block_buf,
2342                                                              pctx.ino);
2343                         if (pctx.errcode) {
2344                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
2345                                             &pctx);
2346                                 continue;
2347                         }
2348                 }
2349         }
2350 }
2351
2352 /*
2353  * Handle processing the extended attribute blocks
2354  */
2355 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
2356                            char *block_buf, struct ea_quota *ea_block_quota)
2357 {
2358         ext2_filsys fs = ctx->fs;
2359         ext2_ino_t      ino = pctx->ino;
2360         struct ext2_inode *inode = pctx->inode;
2361         blk64_t         blk;
2362         char *          end;
2363         struct ext2_ext_attr_header *header;
2364         struct ext2_ext_attr_entry *first, *entry;
2365         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
2366         __u64           quota_inodes = 0;
2367         region_t        region = 0;
2368         int             failed_csum = 0;
2369
2370         ea_block_quota->blocks = 0;
2371         ea_block_quota->inodes = 0;
2372
2373         blk = ext2fs_file_acl_block(fs, inode);
2374         if (blk == 0)
2375                 return 0;
2376
2377         /*
2378          * If the Extended attribute flag isn't set, then a non-zero
2379          * file acl means that the inode is corrupted.
2380          *
2381          * Or if the extended attribute block is an invalid block,
2382          * then the inode is also corrupted.
2383          */
2384         if (!ext2fs_has_feature_xattr(fs->super) ||
2385             (blk < fs->super->s_first_data_block) ||
2386             (blk >= ext2fs_blocks_count(fs->super))) {
2387                 mark_inode_bad(ctx, ino);
2388                 return 0;
2389         }
2390
2391         /* If ea bitmap hasn't been allocated, create it */
2392         if (!ctx->block_ea_map) {
2393                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
2394                                         _("ext attr block map"),
2395                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
2396                                         &ctx->block_ea_map);
2397                 if (pctx->errcode) {
2398                         pctx->num = 2;
2399                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
2400                         ctx->flags |= E2F_FLAG_ABORT;
2401                         return 0;
2402                 }
2403         }
2404
2405         /* Create the EA refcount structure if necessary */
2406         if (!ctx->refcount) {
2407                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
2408                 if (pctx->errcode) {
2409                         pctx->num = 1;
2410                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2411                         ctx->flags |= E2F_FLAG_ABORT;
2412                         return 0;
2413                 }
2414         }
2415
2416 #if 0
2417         /* Debugging text */
2418         printf("Inode %u has EA block %u\n", ino, blk);
2419 #endif
2420
2421         /* Have we seen this EA block before? */
2422         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
2423                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
2424                 ea_block_quota->inodes = 0;
2425
2426                 if (ctx->ea_block_quota_blocks) {
2427                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
2428                                           &quota_blocks);
2429                         if (quota_blocks)
2430                                 ea_block_quota->blocks = quota_blocks;
2431                 }
2432
2433                 if (ctx->ea_block_quota_inodes)
2434                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
2435                                           &ea_block_quota->inodes);
2436
2437                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
2438                         return 1;
2439                 /* Ooops, this EA was referenced more than it stated */
2440                 if (!ctx->refcount_extra) {
2441                         pctx->errcode = ea_refcount_create(0,
2442                                            &ctx->refcount_extra);
2443                         if (pctx->errcode) {
2444                                 pctx->num = 2;
2445                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2446                                 ctx->flags |= E2F_FLAG_ABORT;
2447                                 return 0;
2448                         }
2449                 }
2450                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
2451                 return 1;
2452         }
2453
2454         /*
2455          * OK, we haven't seen this EA block yet.  So we need to
2456          * validate it
2457          */
2458         pctx->blk = blk;
2459         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
2460         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
2461                 pctx->errcode = 0;
2462                 failed_csum = 1;
2463         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
2464                 pctx->errcode = 0;
2465
2466         if (pctx->errcode &&
2467             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
2468                 pctx->errcode = 0;
2469                 goto clear_extattr;
2470         }
2471         header = (struct ext2_ext_attr_header *) block_buf;
2472         pctx->blk = ext2fs_file_acl_block(fs, inode);
2473         if (((ctx->ext_attr_ver == 1) &&
2474              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
2475             ((ctx->ext_attr_ver == 2) &&
2476              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
2477                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
2478                         goto clear_extattr;
2479         }
2480
2481         if (header->h_blocks != 1) {
2482                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
2483                         goto clear_extattr;
2484         }
2485
2486         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
2487                 goto clear_extattr;
2488
2489         region = region_create(0, fs->blocksize);
2490         if (!region) {
2491                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
2492                 ctx->flags |= E2F_FLAG_ABORT;
2493                 return 0;
2494         }
2495         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
2496                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2497                         goto clear_extattr;
2498         }
2499
2500         first = (struct ext2_ext_attr_entry *)(header+1);
2501         end = block_buf + fs->blocksize;
2502         entry = first;
2503         while ((char *)entry < end && *(__u32 *)entry) {
2504                 __u32 hash;
2505
2506                 if (region_allocate(region, (char *)entry - (char *)header,
2507                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
2508                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2509                                 goto clear_extattr;
2510                         break;
2511                 }
2512                 if ((ctx->ext_attr_ver == 1 &&
2513                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
2514                     (ctx->ext_attr_ver == 2 &&
2515                      entry->e_name_index == 0)) {
2516                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
2517                                 goto clear_extattr;
2518                         break;
2519                 }
2520                 if (entry->e_value_inum == 0) {
2521                         if (entry->e_value_offs + entry->e_value_size >
2522                             fs->blocksize) {
2523                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
2524                                         goto clear_extattr;
2525                                 break;
2526                         }
2527                         if (entry->e_value_size &&
2528                             region_allocate(region, entry->e_value_offs,
2529                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
2530                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
2531                                                 pctx))
2532                                         goto clear_extattr;
2533                         }
2534
2535                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
2536                                                           entry->e_value_offs);
2537
2538                         if (entry->e_hash != hash) {
2539                                 pctx->num = entry->e_hash;
2540                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
2541                                         goto clear_extattr;
2542                                 entry->e_hash = hash;
2543                         }
2544                 } else {
2545                         problem_t problem;
2546                         blk64_t entry_quota_blocks;
2547
2548                         problem = check_large_ea_inode(ctx, entry, pctx,
2549                                                        &entry_quota_blocks);
2550                         if (problem && fix_problem(ctx, problem, pctx))
2551                                 goto clear_extattr;
2552
2553                         quota_blocks += entry_quota_blocks;
2554                         quota_inodes++;
2555                 }
2556
2557                 entry = EXT2_EXT_ATTR_NEXT(entry);
2558         }
2559         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
2560                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2561                         goto clear_extattr;
2562         }
2563         region_free(region);
2564
2565         /*
2566          * We only get here if there was no other errors that were fixed.
2567          * If there was a checksum fail, ask to correct it.
2568          */
2569         if (failed_csum &&
2570             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
2571                 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
2572                                                        pctx->ino);
2573                 if (pctx->errcode)
2574                         return 0;
2575         }
2576
2577         if (quota_blocks != EXT2FS_C2B(fs, 1)) {
2578                 if (!ctx->ea_block_quota_blocks) {
2579                         pctx->errcode = ea_refcount_create(0,
2580                                                 &ctx->ea_block_quota_blocks);
2581                         if (pctx->errcode) {
2582                                 pctx->num = 3;
2583                                 goto refcount_fail;
2584                         }
2585                 }
2586                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
2587                                   quota_blocks);
2588         }
2589
2590         if (quota_inodes) {
2591                 if (!ctx->ea_block_quota_inodes) {
2592                         pctx->errcode = ea_refcount_create(0,
2593                                                 &ctx->ea_block_quota_inodes);
2594                         if (pctx->errcode) {
2595                                 pctx->num = 4;
2596 refcount_fail:
2597                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2598                                 ctx->flags |= E2F_FLAG_ABORT;
2599                                 return 0;
2600                         }
2601                 }
2602
2603                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
2604                                   quota_inodes);
2605         }
2606         ea_block_quota->blocks = quota_blocks;
2607         ea_block_quota->inodes = quota_inodes;
2608
2609         inc_ea_inode_refs(ctx, pctx, first, end);
2610         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
2611         mark_block_used(ctx, blk);
2612         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
2613         return 1;
2614
2615 clear_extattr:
2616         if (region)
2617                 region_free(region);
2618         ext2fs_file_acl_block_set(fs, inode, 0);
2619         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
2620         return 0;
2621 }
2622
2623 /* Returns 1 if bad htree, 0 if OK */
2624 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
2625                         ext2_ino_t ino, struct ext2_inode *inode,
2626                         char *block_buf)
2627 {
2628         struct ext2_dx_root_info        *root;
2629         ext2_filsys                     fs = ctx->fs;
2630         errcode_t                       retval;
2631         blk64_t                         blk;
2632
2633         if ((!LINUX_S_ISDIR(inode->i_mode) &&
2634              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
2635             (!ext2fs_has_feature_dir_index(fs->super) &&
2636              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
2637                 return 1;
2638
2639         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
2640
2641         if ((pctx->errcode) ||
2642             (blk == 0) ||
2643             (blk < fs->super->s_first_data_block) ||
2644             (blk >= ext2fs_blocks_count(fs->super))) {
2645                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2646                         return 1;
2647                 else
2648                         return 0;
2649         }
2650
2651         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
2652         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2653                 return 1;
2654
2655         /* XXX should check that beginning matches a directory */
2656         root = (struct ext2_dx_root_info *) (block_buf + 24);
2657
2658         if ((root->reserved_zero || root->info_length < 8) &&
2659             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2660                 return 1;
2661
2662         pctx->num = root->hash_version;
2663         if ((root->hash_version != EXT2_HASH_LEGACY) &&
2664             (root->hash_version != EXT2_HASH_HALF_MD4) &&
2665             (root->hash_version != EXT2_HASH_TEA) &&
2666             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
2667                 return 1;
2668
2669         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
2670             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
2671                 return 1;
2672
2673         pctx->num = root->indirect_levels;
2674         if ((root->indirect_levels > ext2_dir_htree_level(fs)) &&
2675             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
2676                 return 1;
2677
2678         return 0;
2679 }
2680
2681 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
2682                         struct ext2_inode *inode, int restart_flag,
2683                         const char *source)
2684 {
2685         inode->i_flags = 0;
2686         inode->i_links_count = 0;
2687         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
2688         inode->i_dtime = ctx->now;
2689
2690         /*
2691          * If a special inode has such rotten block mappings that we
2692          * want to clear the whole inode, be sure to actually zap
2693          * the block maps because i_links_count isn't checked for
2694          * special inodes, and we'll end up right back here the next
2695          * time we run fsck.
2696          */
2697         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
2698                 memset(inode->i_block, 0, sizeof(inode->i_block));
2699
2700         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
2701         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
2702         if (ctx->inode_reg_map)
2703                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
2704         if (ctx->inode_bad_map)
2705                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
2706
2707         /*
2708          * If the inode was partially accounted for before processing
2709          * was aborted, we need to restart the pass 1 scan.
2710          */
2711         ctx->flags |= restart_flag;
2712
2713         if (ino == EXT2_BAD_INO)
2714                 memset(inode, 0, sizeof(struct ext2_inode));
2715
2716         e2fsck_write_inode(ctx, ino, inode, source);
2717 }
2718
2719 /*
2720  * Use the multiple-blocks reclamation code to fix alignment problems in
2721  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
2722  * physical cluster, and we want the block offsets within that cluster to
2723  * line up.
2724  */
2725 static int has_unaligned_cluster_map(e2fsck_t ctx,
2726                                      blk64_t last_pblk, blk64_t last_lblk,
2727                                      blk64_t pblk, blk64_t lblk)
2728 {
2729         blk64_t cluster_mask;
2730
2731         if (!ctx->fs->cluster_ratio_bits)
2732                 return 0;
2733         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
2734
2735         /*
2736          * If the block in the logical cluster doesn't align with the block in
2737          * the physical cluster...
2738          */
2739         if ((lblk & cluster_mask) != (pblk & cluster_mask))
2740                 return 1;
2741
2742         /*
2743          * If we cross a physical cluster boundary within a logical cluster...
2744          */
2745         if (last_pblk && (lblk & cluster_mask) != 0 &&
2746             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
2747             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
2748                 return 1;
2749
2750         return 0;
2751 }
2752
2753 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
2754                              struct process_block_struct *pb,
2755                              blk64_t start_block, blk64_t end_block,
2756                              blk64_t eof_block,
2757                              ext2_extent_handle_t ehandle,
2758                              int try_repairs)
2759 {
2760         struct ext2fs_extent    extent;
2761         blk64_t                 blk, last_lblk;
2762         unsigned int            i, n;
2763         int                     is_dir, is_leaf;
2764         problem_t               problem;
2765         struct ext2_extent_info info;
2766         int                     failed_csum = 0;
2767
2768         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
2769                 failed_csum = 1;
2770
2771         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
2772         if (pctx->errcode)
2773                 return;
2774         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
2775             !pb->eti.force_rebuild) {
2776                 struct extent_tree_level *etl;
2777
2778                 etl = pb->eti.ext_info + info.curr_level;
2779                 etl->num_extents += info.num_entries;
2780                 etl->max_extents += info.max_entries;
2781                 /*
2782                  * Implementation wart: Splitting extent blocks when appending
2783                  * will leave the old block with one free entry.  Therefore
2784                  * unless the node is totally full, pretend that a non-root
2785                  * extent block can hold one fewer entry than it actually does,
2786                  * so that we don't repeatedly rebuild the extent tree.
2787                  */
2788                 if (info.curr_level && info.num_entries < info.max_entries)
2789                         etl->max_extents--;
2790         }
2791
2792         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
2793                                           &extent);
2794         while ((pctx->errcode == 0 ||
2795                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
2796                info.num_entries-- > 0) {
2797                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
2798                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
2799                 last_lblk = extent.e_lblk + extent.e_len - 1;
2800
2801                 problem = 0;
2802                 pctx->blk = extent.e_pblk;
2803                 pctx->blk2 = extent.e_lblk;
2804                 pctx->num = extent.e_len;
2805                 pctx->blkcount = extent.e_lblk + extent.e_len;
2806
2807                 if (extent.e_pblk == 0 ||
2808                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
2809                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
2810                         problem = PR_1_EXTENT_BAD_START_BLK;
2811                 else if (extent.e_lblk < start_block)
2812                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
2813                 else if ((end_block && last_lblk > end_block) &&
2814                          (!(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT &&
2815                                 last_lblk > eof_block)))
2816                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
2817                 else if (is_leaf && extent.e_len == 0)
2818                         problem = PR_1_EXTENT_LENGTH_ZERO;
2819                 else if (is_leaf &&
2820                          (extent.e_pblk + extent.e_len) >
2821                          ext2fs_blocks_count(ctx->fs->super))
2822                         problem = PR_1_EXTENT_ENDS_BEYOND;
2823                 else if (is_leaf && is_dir &&
2824                          ((extent.e_lblk + extent.e_len) >
2825                           (1U << (21 - ctx->fs->super->s_log_block_size))))
2826                         problem = PR_1_TOOBIG_DIR;
2827
2828                 if (is_leaf && problem == 0 && extent.e_len > 0) {
2829 #if 0
2830                         printf("extent_region(ino=%u, expect=%llu, "
2831                                "lblk=%llu, len=%u)\n",
2832                                pb->ino, pb->next_lblock,
2833                                extent.e_lblk, extent.e_len);
2834 #endif
2835                         if (extent.e_lblk < pb->next_lblock)
2836                                 problem = PR_1_EXTENT_COLLISION;
2837                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
2838                                 pb->next_lblock = extent.e_lblk + extent.e_len;
2839                 }
2840
2841                 /*
2842                  * Uninitialized blocks in a directory?  Clear the flag and
2843                  * we'll interpret the blocks later.
2844                  */
2845                 if (try_repairs && is_dir && problem == 0 &&
2846                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
2847                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
2848                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
2849                         pb->inode_modified = 1;
2850                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
2851                                                               &extent);
2852                         if (pctx->errcode)
2853                                 return;
2854                         failed_csum = 0;
2855                 }
2856
2857                 if (try_repairs && problem) {
2858 report_problem:
2859                         if (fix_problem(ctx, problem, pctx)) {
2860                                 if (ctx->invalid_bitmaps) {
2861                                         /*
2862                                          * If fsck knows the bitmaps are bad,
2863                                          * skip to the next extent and
2864                                          * try to clear this extent again
2865                                          * after fixing the bitmaps, by
2866                                          * restarting fsck.
2867                                          */
2868                                         pctx->errcode = ext2fs_extent_get(
2869                                                           ehandle,
2870                                                           EXT2_EXTENT_NEXT_SIB,
2871                                                           &extent);
2872                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
2873                                         if (pctx->errcode ==
2874                                                     EXT2_ET_NO_CURRENT_NODE) {
2875                                                 pctx->errcode = 0;
2876                                                 break;
2877                                         }
2878                                         continue;
2879                                 }
2880                                 e2fsck_read_bitmaps(ctx);
2881                                 pb->inode_modified = 1;
2882                                 pctx->errcode =
2883                                         ext2fs_extent_delete(ehandle, 0);
2884                                 if (pctx->errcode) {
2885                                         pctx->str = "ext2fs_extent_delete";
2886                                         return;
2887                                 }
2888                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
2889                                 if (pctx->errcode &&
2890                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
2891                                         pctx->str = "ext2fs_extent_fix_parents";
2892                                         return;
2893                                 }
2894                                 pctx->errcode = ext2fs_extent_get(ehandle,
2895                                                                   EXT2_EXTENT_CURRENT,
2896                                                                   &extent);
2897                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
2898                                         pctx->errcode = 0;
2899                                         break;
2900                                 }
2901                                 failed_csum = 0;
2902                                 continue;
2903                         }
2904                         goto next;
2905                 }
2906
2907                 if (!is_leaf) {
2908                         blk64_t lblk = extent.e_lblk;
2909                         int next_try_repairs = 1;
2910
2911                         blk = extent.e_pblk;
2912
2913                         /*
2914                          * If this lower extent block collides with critical
2915                          * metadata, don't try to repair the damage.  Pass 1b
2916                          * will reallocate the block; then we can try again.
2917                          */
2918                         if (pb->ino != EXT2_RESIZE_INO &&
2919                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
2920                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
2921                                                       extent.e_pblk)) {
2922                                 next_try_repairs = 0;
2923                                 pctx->blk = blk;
2924                                 fix_problem(ctx,
2925                                             PR_1_CRITICAL_METADATA_COLLISION,
2926                                             pctx);
2927                                 if ((ctx->options & E2F_OPT_NO) == 0)
2928                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
2929                         }
2930                         pctx->errcode = ext2fs_extent_get(ehandle,
2931                                                   EXT2_EXTENT_DOWN, &extent);
2932                         if (pctx->errcode &&
2933                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
2934                                 pctx->str = "EXT2_EXTENT_DOWN";
2935                                 problem = PR_1_EXTENT_HEADER_INVALID;
2936                                 if (!next_try_repairs)
2937                                         return;
2938                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
2939                                         goto report_problem;
2940                                 return;
2941                         }
2942                         /* The next extent should match this index's logical start */
2943                         if (extent.e_lblk != lblk) {
2944                                 struct ext2_extent_info e_info;
2945
2946                                 ext2fs_extent_get_info(ehandle, &e_info);
2947                                 pctx->blk = lblk;
2948                                 pctx->blk2 = extent.e_lblk;
2949                                 pctx->num = e_info.curr_level - 1;
2950                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
2951                                 if (fix_problem(ctx, problem, pctx)) {
2952                                         pb->inode_modified = 1;
2953                                         pctx->errcode =
2954                                                 ext2fs_extent_fix_parents(ehandle);
2955                                         if (pctx->errcode) {
2956                                                 pctx->str = "ext2fs_extent_fix_parents";
2957                                                 return;
2958                                         }
2959                                 }
2960                         }
2961                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
2962                                          last_lblk, eof_block, ehandle,
2963                                          next_try_repairs);
2964                         if (pctx->errcode)
2965                                 return;
2966                         pctx->errcode = ext2fs_extent_get(ehandle,
2967                                                   EXT2_EXTENT_UP, &extent);
2968                         if (pctx->errcode) {
2969                                 pctx->str = "EXT2_EXTENT_UP";
2970                                 return;
2971                         }
2972                         mark_block_used(ctx, blk);
2973                         pb->num_blocks++;
2974                         goto next;
2975                 }
2976
2977                 if ((pb->previous_block != 0) &&
2978                     (pb->previous_block+1 != extent.e_pblk)) {
2979                         if (ctx->options & E2F_OPT_FRAGCHECK) {
2980                                 char type = '?';
2981
2982                                 if (pb->is_dir)
2983                                         type = 'd';
2984                                 else if (pb->is_reg)
2985                                         type = 'f';
2986
2987                                 printf(("%6lu(%c): expecting %6lu "
2988                                         "actual extent "
2989                                         "phys %6lu log %lu len %lu\n"),
2990                                        (unsigned long) pctx->ino, type,
2991                                        (unsigned long) pb->previous_block+1,
2992                                        (unsigned long) extent.e_pblk,
2993                                        (unsigned long) extent.e_lblk,
2994                                        (unsigned long) extent.e_len);
2995                         }
2996                         pb->fragmented = 1;
2997                 }
2998                 /*
2999                  * If we notice a gap in the logical block mappings of an
3000                  * extent-mapped directory, offer to close the hole by
3001                  * moving the logical block down, otherwise we'll go mad in
3002                  * pass 3 allocating empty directory blocks to fill the hole.
3003                  */
3004                 if (try_repairs && is_dir &&
3005                     pb->last_block + 1 < extent.e_lblk) {
3006                         blk64_t new_lblk;
3007
3008                         new_lblk = pb->last_block + 1;
3009                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3010                                 new_lblk = ((new_lblk +
3011                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3012                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3013                                            (extent.e_pblk &
3014                                             EXT2FS_CLUSTER_MASK(ctx->fs));
3015                         pctx->blk = extent.e_lblk;
3016                         pctx->blk2 = new_lblk;
3017                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3018                                 extent.e_lblk = new_lblk;
3019                                 pb->inode_modified = 1;
3020                                 pctx->errcode = ext2fs_extent_replace(ehandle,
3021                                                                 0, &extent);
3022                                 if (pctx->errcode) {
3023                                         pctx->errcode = 0;
3024                                         goto alloc_later;
3025                                 }
3026                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3027                                 if (pctx->errcode)
3028                                         goto failed_add_dir_block;
3029                                 pctx->errcode = ext2fs_extent_goto(ehandle,
3030                                                                 extent.e_lblk);
3031                                 if (pctx->errcode)
3032                                         goto failed_add_dir_block;
3033                                 last_lblk = extent.e_lblk + extent.e_len - 1;
3034                                 failed_csum = 0;
3035                         }
3036                 }
3037 alloc_later:
3038                 if (is_dir) {
3039                         while (++pb->last_db_block <
3040                                (e2_blkcnt_t) extent.e_lblk) {
3041                                 pctx->errcode = ext2fs_add_dir_block2(
3042                                                         ctx->fs->dblist,
3043                                                         pb->ino, 0,
3044                                                         pb->last_db_block);
3045                                 if (pctx->errcode) {
3046                                         pctx->blk = 0;
3047                                         pctx->num = pb->last_db_block;
3048                                         goto failed_add_dir_block;
3049                                 }
3050                         }
3051
3052                         for (i = 0; i < extent.e_len; i++) {
3053                                 pctx->errcode = ext2fs_add_dir_block2(
3054                                                         ctx->fs->dblist,
3055                                                         pctx->ino,
3056                                                         extent.e_pblk + i,
3057                                                         extent.e_lblk + i);
3058                                 if (pctx->errcode) {
3059                                         pctx->blk = extent.e_pblk + i;
3060                                         pctx->num = extent.e_lblk + i;
3061                                 failed_add_dir_block:
3062                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3063                                         /* Should never get here */
3064                                         ctx->flags |= E2F_FLAG_ABORT;
3065                                         return;
3066                                 }
3067                         }
3068                         if (extent.e_len > 0)
3069                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3070                 }
3071                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3072                                               pb->last_block,
3073                                               extent.e_pblk,
3074                                               extent.e_lblk)) {
3075                         for (i = 0; i < extent.e_len; i++) {
3076                                 pctx->blk = extent.e_lblk + i;
3077                                 pctx->blk2 = extent.e_pblk + i;
3078                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3079                                 mark_block_used(ctx, extent.e_pblk + i);
3080                                 mark_block_used(ctx, extent.e_pblk + i);
3081                         }
3082                 }
3083
3084                 /*
3085                  * Check whether first cluster got marked in previous iteration.
3086                  */
3087                 if (ctx->fs->cluster_ratio_bits &&
3088                     pb->previous_block &&
3089                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
3090                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
3091                         /* Set blk to the beginning of next cluster. */
3092                         blk = EXT2FS_C2B(
3093                                 ctx->fs,
3094                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
3095                 else
3096                         /* Set blk to the beginning of current cluster. */
3097                         blk = EXT2FS_C2B(ctx->fs,
3098                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
3099
3100                 if (blk < extent.e_pblk + extent.e_len) {
3101                         mark_blocks_used(ctx, blk,
3102                                          extent.e_pblk + extent.e_len - blk);
3103                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
3104                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
3105                         pb->num_blocks += n;
3106                 }
3107                 pb->last_block = extent.e_lblk + extent.e_len - 1;
3108                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
3109                 start_block = pb->last_block = last_lblk;
3110                 if (is_leaf && !is_dir &&
3111                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
3112                         pb->last_init_lblock = last_lblk;
3113         next:
3114                 pctx->errcode = ext2fs_extent_get(ehandle,
3115                                                   EXT2_EXTENT_NEXT_SIB,
3116                                                   &extent);
3117         }
3118
3119         /* Failed csum but passes checks?  Ask to fix checksum. */
3120         if (failed_csum &&
3121             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
3122                 pb->inode_modified = 1;
3123                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
3124                 if (pctx->errcode)
3125                         return;
3126         }
3127
3128         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
3129                 pctx->errcode = 0;
3130 }
3131
3132 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
3133                                  struct process_block_struct *pb)
3134 {
3135         struct ext2_extent_info info;
3136         struct ext2_inode       *inode = pctx->inode;
3137         ext2_extent_handle_t    ehandle;
3138         ext2_filsys             fs = ctx->fs;
3139         ext2_ino_t              ino = pctx->ino;
3140         errcode_t               retval;
3141         blk64_t                 eof_lblk;
3142         struct ext3_extent_header       *eh;
3143
3144         /* Check for a proper extent header... */
3145         eh = (struct ext3_extent_header *) &inode->i_block[0];
3146         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
3147         if (retval) {
3148                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
3149                         e2fsck_clear_inode(ctx, ino, inode, 0,
3150                                            "check_blocks_extents");
3151                 pctx->errcode = 0;
3152                 return;
3153         }
3154
3155         /* ...since this function doesn't fail if i_block is zeroed. */
3156         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
3157         if (pctx->errcode) {
3158                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
3159                         e2fsck_clear_inode(ctx, ino, inode, 0,
3160                                            "check_blocks_extents");
3161                 pctx->errcode = 0;
3162                 return;
3163         }
3164
3165         retval = ext2fs_extent_get_info(ehandle, &info);
3166         if (retval == 0) {
3167                 int max_depth = info.max_depth;
3168
3169                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
3170                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
3171                 ctx->extent_depth_count[max_depth]++;
3172         }
3173
3174         /* Check maximum extent depth */
3175         pctx->blk = info.max_depth;
3176         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
3177         if (pctx->blk2 < pctx->blk &&
3178             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
3179                 pb->eti.force_rebuild = 1;
3180
3181         /* Can we collect extent tree level stats? */
3182         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
3183         if (pctx->blk2 > pctx->blk)
3184                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
3185         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
3186         pb->eti.ino = pb->ino;
3187
3188         pb->next_lblock = 0;
3189
3190         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
3191                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
3192         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
3193         if (pctx->errcode &&
3194             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
3195                 pb->num_blocks = 0;
3196                 inode->i_blocks = 0;
3197                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3198                                    "check_blocks_extents");
3199                 pctx->errcode = 0;
3200         }
3201         ext2fs_extent_free(ehandle);
3202
3203         /* Rebuild unless it's a dir and we're rehashing it */
3204         if (LINUX_S_ISDIR(inode->i_mode) &&
3205             e2fsck_dir_will_be_rehashed(ctx, ino))
3206                 return;
3207
3208         if (ctx->options & E2F_OPT_CONVERT_BMAP)
3209                 e2fsck_rebuild_extents_later(ctx, ino);
3210         else
3211                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
3212 }
3213
3214 /*
3215  * In fact we don't need to check blocks for an inode with inline data
3216  * because this inode doesn't have any blocks.  In this function all
3217  * we need to do is add this inode into dblist when it is a directory.
3218  */
3219 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
3220                                      struct process_block_struct *pb)
3221 {
3222         int     flags;
3223         size_t  inline_data_size = 0;
3224
3225         if (!pb->is_dir) {
3226                 pctx->errcode = 0;
3227                 return;
3228         }
3229
3230         /* Process the dirents in i_block[] as the "first" block. */
3231         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
3232         if (pctx->errcode)
3233                 goto err;
3234
3235         /* Process the dirents in the EA as a "second" block. */
3236         flags = ctx->fs->flags;
3237         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3238         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
3239                                                 &inline_data_size);
3240         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3241                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3242         if (pctx->errcode) {
3243                 pctx->errcode = 0;
3244                 return;
3245         }
3246
3247         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
3248                 return;
3249
3250         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
3251         if (pctx->errcode)
3252                 goto err;
3253
3254         return;
3255 err:
3256         pctx->blk = 0;
3257         pctx->num = 0;
3258         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3259         ctx->flags |= E2F_FLAG_ABORT;
3260 }
3261
3262 /*
3263  * This subroutine is called on each inode to account for all of the
3264  * blocks used by that inode.
3265  */
3266 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
3267                          char *block_buf, const struct ea_quota *ea_ibody_quota)
3268 {
3269         ext2_filsys fs = ctx->fs;
3270         struct process_block_struct pb;
3271         ext2_ino_t      ino = pctx->ino;
3272         struct ext2_inode *inode = pctx->inode;
3273         unsigned        bad_size = 0;
3274         int             dirty_inode = 0;
3275         int             extent_fs;
3276         int             inlinedata_fs;
3277         __u64           size;
3278         struct ea_quota ea_block_quota;
3279
3280         pb.ino = ino;
3281         pb.num_blocks = EXT2FS_B2C(ctx->fs,
3282                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
3283         pb.last_block = ~0;
3284         pb.last_init_lblock = -1;
3285         pb.last_db_block = -1;
3286         pb.num_illegal_blocks = 0;
3287         pb.suppress = 0; pb.clear = 0;
3288         pb.fragmented = 0;
3289         pb.compressed = 0;
3290         pb.previous_block = 0;
3291         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
3292         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
3293         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
3294         pb.inode = inode;
3295         pb.pctx = pctx;
3296         pb.ctx = ctx;
3297         pb.inode_modified = 0;
3298         pb.eti.force_rebuild = 0;
3299         pctx->ino = ino;
3300         pctx->errcode = 0;
3301
3302         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
3303         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
3304
3305         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
3306                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3307                         goto out;
3308                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
3309         }
3310
3311         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
3312                 check_blocks_inline_data(ctx, pctx, &pb);
3313         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
3314                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
3315                         check_blocks_extents(ctx, pctx, &pb);
3316                 else {
3317                         int flags;
3318                         /*
3319                          * If we've modified the inode, write it out before
3320                          * iterate() tries to use it.
3321                          */
3322                         if (dirty_inode) {
3323                                 e2fsck_write_inode(ctx, ino, inode,
3324                                                    "check_blocks");
3325                                 dirty_inode = 0;
3326                         }
3327                         flags = fs->flags;
3328                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3329                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
3330                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
3331                                                 block_buf, process_block, &pb);
3332                         /*
3333                          * We do not have uninitialized extents in non extent
3334                          * files.
3335                          */
3336                         pb.last_init_lblock = pb.last_block;
3337                         /*
3338                          * If iterate() changed a block mapping, we have to
3339                          * re-read the inode.  If we decide to clear the
3340                          * inode after clearing some stuff, we'll re-write the
3341                          * bad mappings into the inode!
3342                          */
3343                         if (pb.inode_modified)
3344                                 e2fsck_read_inode(ctx, ino, inode,
3345                                                   "check_blocks");
3346                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3347                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3348
3349                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
3350 #ifdef DEBUG
3351                                 printf("bmap rebuild ino=%d\n", ino);
3352 #endif
3353                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
3354                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
3355                                         e2fsck_rebuild_extents_later(ctx, ino);
3356                         }
3357                 }
3358         }
3359         end_problem_latch(ctx, PR_LATCH_BLOCK);
3360         end_problem_latch(ctx, PR_LATCH_TOOBIG);
3361         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3362                 goto out;
3363         if (pctx->errcode)
3364                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
3365
3366         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
3367                 if (LINUX_S_ISDIR(inode->i_mode))
3368                         ctx->fs_fragmented_dir++;
3369                 else
3370                         ctx->fs_fragmented++;
3371         }
3372
3373         if (pb.clear) {
3374                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3375                                    "check_blocks");
3376                 return;
3377         }
3378
3379         if (inode->i_flags & EXT2_INDEX_FL) {
3380                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
3381                         inode->i_flags &= ~EXT2_INDEX_FL;
3382                         dirty_inode++;
3383                 } else {
3384                         e2fsck_add_dx_dir(ctx, ino, pb.last_block+1);
3385                 }
3386         }
3387
3388         if (!pb.num_blocks && pb.is_dir &&
3389             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
3390                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
3391                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
3392                         ctx->fs_directory_count--;
3393                         return;
3394                 }
3395         }
3396
3397         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
3398             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
3399             !(inode->i_flags & EXT4_EA_INODE_FL)) {
3400                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
3401                                ino,
3402                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
3403                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
3404                                   ino, (ea_ibody_quota ?
3405                                         ea_ibody_quota->inodes : 0) +
3406                                                 ea_block_quota.inodes + 1);
3407         }
3408
3409         if (!ext2fs_has_feature_huge_file(fs->super) ||
3410             !(inode->i_flags & EXT4_HUGE_FILE_FL))
3411                 pb.num_blocks *= (fs->blocksize / 512);
3412         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
3413 #if 0
3414         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
3415                ino, inode->i_size, pb.last_block, ext2fs_inode_i_blocks(fs, inode),
3416                pb.num_blocks);
3417 #endif
3418         if (pb.is_dir) {
3419                 unsigned nblock = inode->i_size >> EXT2_BLOCK_SIZE_BITS(fs->super);
3420                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
3421                         int flags;
3422                         size_t sz = 0;
3423                         errcode_t err;
3424
3425                         flags = ctx->fs->flags;
3426                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3427                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
3428                                                       &sz);
3429                         ctx->fs->flags = (flags &
3430                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3431                                          (ctx->fs->flags &
3432                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3433                         if (err || sz != inode->i_size) {
3434                                 bad_size = 7;
3435                                 pctx->num = sz;
3436                         }
3437                 } else if (inode->i_size & (fs->blocksize - 1))
3438                         bad_size = 5;
3439                 else if (nblock > (pb.last_block + 1))
3440                         bad_size = 1;
3441                 else if (nblock < (pb.last_block + 1)) {
3442                         if (((pb.last_block + 1) - nblock) >
3443                             fs->super->s_prealloc_dir_blocks)
3444                                 bad_size = 2;
3445                 }
3446         } else {
3447                 e2_blkcnt_t blkpg = ctx->blocks_per_page;
3448
3449                 size = EXT2_I_SIZE(inode);
3450                 if ((pb.last_init_lblock >= 0) &&
3451                     /* if size is smaller than expected by the block count,
3452                      * allow allocated blocks to end of PAGE_SIZE.
3453                      * last_init_lblock is the last in-use block, so it is
3454                      * the minimum expected file size, but +1 because it is
3455                      * the base-zero block number and not the block count. */
3456                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
3457                     ((pb.last_init_lblock + 1) / blkpg * blkpg !=
3458                      (pb.last_init_lblock + 1) ||
3459                      size < (__u64)(pb.last_init_lblock & ~(blkpg - 1)) *
3460                      fs->blocksize))
3461                         bad_size = 3;
3462                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3463                          size > ext2_max_sizes[fs->super->s_log_block_size])
3464                         /* too big for a direct/indirect-mapped file */
3465                         bad_size = 4;
3466                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3467                          size >
3468                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
3469                         /* too big for an extent-based file - 32bit ee_block */
3470                         bad_size = 6;
3471         }
3472         /* i_size for symlinks is checked elsewhere */
3473         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
3474                 /* Did inline_data set pctx->num earlier? */
3475                 if (bad_size != 7)
3476                         pctx->num = (pb.last_block + 1) * fs->blocksize;
3477                 pctx->group = bad_size;
3478                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
3479                         if (LINUX_S_ISDIR(inode->i_mode))
3480                                 pctx->num &= 0xFFFFFFFFULL;
3481                         ext2fs_inode_size_set(fs, inode, pctx->num);
3482                         if (EXT2_I_SIZE(inode) == 0 &&
3483                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
3484                                 memset(inode->i_block, 0,
3485                                        sizeof(inode->i_block));
3486                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
3487                         }
3488                         dirty_inode++;
3489                 }
3490                 pctx->num = 0;
3491         }
3492         if (LINUX_S_ISREG(inode->i_mode) &&
3493             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3494                 ctx->large_files++;
3495         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
3496             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
3497              (ext2fs_has_feature_huge_file(fs->super) &&
3498               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
3499               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
3500                 pctx->num = pb.num_blocks;
3501                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
3502                         inode->i_blocks = pb.num_blocks;
3503                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
3504                         dirty_inode++;
3505                 }
3506                 pctx->num = 0;
3507         }
3508
3509         /*
3510          * The kernel gets mad if we ask it to allocate bigalloc clusters to
3511          * a block mapped file, so rebuild it as an extent file.  We can skip
3512          * symlinks because they're never rewritten.
3513          */
3514         if (ext2fs_has_feature_bigalloc(fs->super) &&
3515             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
3516             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
3517             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
3518             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
3519             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
3520                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
3521                 if (pctx->errcode)
3522                         goto out;
3523         }
3524
3525         if (ctx->dirs_to_hash && pb.is_dir &&
3526             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
3527             !(inode->i_flags & EXT2_INDEX_FL) &&
3528             ((inode->i_size / fs->blocksize) >= 3))
3529                 e2fsck_rehash_dir_later(ctx, ino);
3530
3531 out:
3532         if (dirty_inode)
3533                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
3534 }
3535
3536 #if 0
3537 /*
3538  * Helper function called by process block when an illegal block is
3539  * found.  It returns a description about why the block is illegal
3540  */
3541 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
3542 {
3543         blk64_t super;
3544         int     i;
3545         static char     problem[80];
3546
3547         super = fs->super->s_first_data_block;
3548         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
3549         if (block < super) {
3550                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
3551                 return(problem);
3552         } else if (block >= ext2fs_blocks_count(fs->super)) {
3553                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
3554                 return(problem);
3555         }
3556         for (i = 0; i < fs->group_desc_count; i++) {
3557                 if (block == super) {
3558                         sprintf(problem, "is the superblock in group %d", i);
3559                         break;
3560                 }
3561                 if (block > super &&
3562                     block <= (super + fs->desc_blocks)) {
3563                         sprintf(problem, "is in the group descriptors "
3564                                 "of group %d", i);
3565                         break;
3566                 }
3567                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
3568                         sprintf(problem, "is the block bitmap of group %d", i);
3569                         break;
3570                 }
3571                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
3572                         sprintf(problem, "is the inode bitmap of group %d", i);
3573                         break;
3574                 }
3575                 if (block >= ext2fs_inode_table_loc(fs, i) &&
3576                     (block < ext2fs_inode_table_loc(fs, i)
3577                      + fs->inode_blocks_per_group)) {
3578                         sprintf(problem, "is in the inode table of group %d",
3579                                 i);
3580                         break;
3581                 }
3582                 super += fs->super->s_blocks_per_group;
3583         }
3584         return(problem);
3585 }
3586 #endif
3587
3588 /*
3589  * This is a helper function for check_blocks().
3590  */
3591 static int process_block(ext2_filsys fs,
3592                   blk64_t       *block_nr,
3593                   e2_blkcnt_t blockcnt,
3594                   blk64_t ref_block EXT2FS_ATTR((unused)),
3595                   int ref_offset EXT2FS_ATTR((unused)),
3596                   void *priv_data)
3597 {
3598         struct process_block_struct *p;
3599         struct problem_context *pctx;
3600         blk64_t blk = *block_nr;
3601         int     ret_code = 0;
3602         problem_t       problem = 0;
3603         e2fsck_t        ctx;
3604
3605         p = (struct process_block_struct *) priv_data;
3606         pctx = p->pctx;
3607         ctx = p->ctx;
3608
3609         /*
3610          * For a directory, add logical block zero for processing even if it's
3611          * not mapped or we'll be perennially stuck with broken "." and ".."
3612          * entries.
3613          */
3614         if (p->is_dir && blockcnt == 0 && blk == 0) {
3615                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
3616                 if (pctx->errcode) {
3617                         pctx->blk = blk;
3618                         pctx->num = blockcnt;
3619                         goto failed_add_dir_block;
3620                 }
3621                 p->last_db_block++;
3622         }
3623
3624         if (blk == 0)
3625                 return 0;
3626
3627 #if 0
3628         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
3629                blockcnt);
3630 #endif
3631
3632         /*
3633          * Simplistic fragmentation check.  We merely require that the
3634          * file be contiguous.  (Which can never be true for really
3635          * big files that are greater than a block group.)
3636          */
3637         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
3638                 if (p->previous_block+1 != blk) {
3639                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3640                                 char type = '?';
3641
3642                                 if (p->is_dir)
3643                                         type = 'd';
3644                                 else if (p->is_reg)
3645                                         type = 'f';
3646
3647                                 printf(_("%6lu(%c): expecting %6lu "
3648                                          "got phys %6lu (blkcnt %lld)\n"),
3649                                        (unsigned long) pctx->ino, type,
3650                                        (unsigned long) p->previous_block+1,
3651                                        (unsigned long) blk,
3652                                        blockcnt);
3653                         }
3654                         p->fragmented = 1;
3655                 }
3656         }
3657
3658         if (p->is_dir && blockcnt > (1 << (21 - fs->super->s_log_block_size)))
3659                 problem = PR_1_TOOBIG_DIR;
3660         if (p->is_reg && p->num_blocks+1 >= p->max_blocks)
3661                 problem = PR_1_TOOBIG_REG;
3662         if (!p->is_dir && !p->is_reg && blockcnt > 0)
3663                 problem = PR_1_TOOBIG_SYMLINK;
3664
3665         if (blk < fs->super->s_first_data_block ||
3666             blk >= ext2fs_blocks_count(fs->super))
3667                 problem = PR_1_ILLEGAL_BLOCK_NUM;
3668
3669         /*
3670          * If this IND/DIND/TIND block is squatting atop some critical metadata
3671          * (group descriptors, superblock, bitmap, inode table), any write to
3672          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
3673          * fix that and restart fsck.
3674          */
3675         if (blockcnt < 0 &&
3676             p->ino != EXT2_RESIZE_INO &&
3677             blk < ctx->fs->super->s_blocks_count &&
3678             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
3679                 pctx->blk = blk;
3680                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
3681                 if ((ctx->options & E2F_OPT_NO) == 0)
3682                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3683         }
3684
3685         if (problem) {
3686                 p->num_illegal_blocks++;
3687                 /*
3688                  * A bit of subterfuge here -- we're trying to fix a block
3689                  * mapping, but the IND/DIND/TIND block could have collided
3690                  * with some critical metadata.  So, fix the in-core mapping so
3691                  * iterate won't go insane, but return 0 instead of
3692                  * BLOCK_CHANGED so that it won't write the remapping out to
3693                  * our multiply linked block.
3694                  *
3695                  * Even if we previously determined that an *IND block
3696                  * conflicts with critical metadata, we must still try to
3697                  * iterate the *IND block as if it is an *IND block to find and
3698                  * mark the blocks it points to.  Better to be overly cautious
3699                  * with the used_blocks map so that we don't move the *IND
3700                  * block to a block that's really in use!
3701                  */
3702                 if (p->ino != EXT2_RESIZE_INO &&
3703                     ref_block != 0 &&
3704                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3705                                               ref_block)) {
3706                         *block_nr = 0;
3707                         return 0;
3708                 }
3709                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
3710                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
3711                                 p->clear = 1;
3712                                 return BLOCK_ABORT;
3713                         }
3714                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
3715                                 p->suppress = 1;
3716                                 set_latch_flags(PR_LATCH_BLOCK,
3717                                                 PRL_SUPPRESS, 0);
3718                         }
3719                 }
3720                 pctx->blk = blk;
3721                 pctx->blkcount = blockcnt;
3722                 if (fix_problem(ctx, problem, pctx)) {
3723                         blk = *block_nr = 0;
3724                         ret_code = BLOCK_CHANGED;
3725                         p->inode_modified = 1;
3726                         /*
3727                          * If the directory block is too big and is beyond the
3728                          * end of the FS, don't bother trying to add it for
3729                          * processing -- the kernel would never have created a
3730                          * directory this large, and we risk an ENOMEM abort.
3731                          * In any case, the toobig handler for extent-based
3732                          * directories also doesn't feed toobig blocks to
3733                          * pass 2.
3734                          */
3735                         if (problem == PR_1_TOOBIG_DIR)
3736                                 return ret_code;
3737                         goto mark_dir;
3738                 } else
3739                         return 0;
3740         }
3741
3742         if (p->ino == EXT2_RESIZE_INO) {
3743                 /*
3744                  * The resize inode has already be sanity checked
3745                  * during pass #0 (the superblock checks).  All we
3746                  * have to do is mark the double indirect block as
3747                  * being in use; all of the other blocks are handled
3748                  * by mark_table_blocks()).
3749                  */
3750                 if (blockcnt == BLOCK_COUNT_DIND)
3751                         mark_block_used(ctx, blk);
3752                 p->num_blocks++;
3753         } else if (!(ctx->fs->cluster_ratio_bits &&
3754                      p->previous_block &&
3755                      (EXT2FS_B2C(ctx->fs, blk) ==
3756                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
3757                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
3758                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
3759                 mark_block_used(ctx, blk);
3760                 p->num_blocks++;
3761         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
3762                                              p->last_block, blk, blockcnt)) {
3763                 pctx->blk = blockcnt;
3764                 pctx->blk2 = blk;
3765                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3766                 mark_block_used(ctx, blk);
3767                 mark_block_used(ctx, blk);
3768         }
3769         if (blockcnt >= 0)
3770                 p->last_block = blockcnt;
3771         p->previous_block = blk;
3772 mark_dir:
3773         if (p->is_dir && (blockcnt >= 0)) {
3774                 while (++p->last_db_block < blockcnt) {
3775                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
3776                                                               p->ino, 0,
3777                                                               p->last_db_block);
3778                         if (pctx->errcode) {
3779                                 pctx->blk = 0;
3780                                 pctx->num = p->last_db_block;
3781                                 goto failed_add_dir_block;
3782                         }
3783                 }
3784                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
3785                                                       blk, blockcnt);
3786                 if (pctx->errcode) {
3787                         pctx->blk = blk;
3788                         pctx->num = blockcnt;
3789                 failed_add_dir_block:
3790                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3791                         /* Should never get here */
3792                         ctx->flags |= E2F_FLAG_ABORT;
3793                         return BLOCK_ABORT;
3794                 }
3795         }
3796         return ret_code;
3797 }
3798
3799 static int process_bad_block(ext2_filsys fs,
3800                       blk64_t *block_nr,
3801                       e2_blkcnt_t blockcnt,
3802                       blk64_t ref_block EXT2FS_ATTR((unused)),
3803                       int ref_offset EXT2FS_ATTR((unused)),
3804                       void *priv_data)
3805 {
3806         struct process_block_struct *p;
3807         blk64_t         blk = *block_nr;
3808         blk64_t         first_block;
3809         dgrp_t          i;
3810         struct problem_context *pctx;
3811         e2fsck_t        ctx;
3812
3813         if (!blk)
3814                 return 0;
3815
3816         p = (struct process_block_struct *) priv_data;
3817         ctx = p->ctx;
3818         pctx = p->pctx;
3819
3820         pctx->ino = EXT2_BAD_INO;
3821         pctx->blk = blk;
3822         pctx->blkcount = blockcnt;
3823
3824         if ((blk < fs->super->s_first_data_block) ||
3825             (blk >= ext2fs_blocks_count(fs->super))) {
3826                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
3827                         *block_nr = 0;
3828                         return BLOCK_CHANGED;
3829                 } else
3830                         return 0;
3831         }
3832
3833         if (blockcnt < 0) {
3834                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
3835                         p->bbcheck = 1;
3836                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
3837                                 *block_nr = 0;
3838                                 return BLOCK_CHANGED;
3839                         }
3840                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
3841                                                     blk)) {
3842                         p->bbcheck = 1;
3843                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
3844                                         pctx)) {
3845                                 *block_nr = 0;
3846                                 return BLOCK_CHANGED;
3847                         }
3848                         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3849                                 return BLOCK_ABORT;
3850                 } else
3851                         mark_block_used(ctx, blk);
3852                 return 0;
3853         }
3854 #if 0
3855         printf ("DEBUG: Marking %u as bad.\n", blk);
3856 #endif
3857         ctx->fs_badblocks_count++;
3858         /*
3859          * If the block is not used, then mark it as used and return.
3860          * If it is already marked as found, this must mean that
3861          * there's an overlap between the filesystem table blocks
3862          * (bitmaps and inode table) and the bad block list.
3863          */
3864         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
3865                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
3866                 return 0;
3867         }
3868         /*
3869          * Try to find the where the filesystem block was used...
3870          */
3871         first_block = fs->super->s_first_data_block;
3872
3873         for (i = 0; i < fs->group_desc_count; i++ ) {
3874                 pctx->group = i;
3875                 pctx->blk = blk;
3876                 if (!ext2fs_bg_has_super(fs, i))
3877                         goto skip_super;
3878                 if (blk == first_block) {
3879                         if (i == 0) {
3880                                 if (fix_problem(ctx,
3881                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
3882                                                 pctx)) {
3883                                         *block_nr = 0;
3884                                         return BLOCK_CHANGED;
3885                                 }
3886                                 return 0;
3887                         }
3888                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
3889                         return 0;
3890                 }
3891                 if ((blk > first_block) &&
3892                     (blk <= first_block + fs->desc_blocks)) {
3893                         if (i == 0) {
3894                                 pctx->blk = *block_nr;
3895                                 if (fix_problem(ctx,
3896                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
3897                                         *block_nr = 0;
3898                                         return BLOCK_CHANGED;
3899                                 }
3900                                 return 0;
3901                         }
3902                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
3903                         return 0;
3904                 }
3905         skip_super:
3906                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
3907                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
3908                                 ctx->invalid_block_bitmap_flag[i]++;
3909                                 ctx->invalid_bitmaps++;
3910                         }
3911                         return 0;
3912                 }
3913                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
3914                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
3915                                 ctx->invalid_inode_bitmap_flag[i]++;
3916                                 ctx->invalid_bitmaps++;
3917                         }
3918                         return 0;
3919                 }
3920                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
3921                     (blk < (ext2fs_inode_table_loc(fs, i) +
3922                             fs->inode_blocks_per_group))) {
3923                         /*
3924                          * If there are bad blocks in the inode table,
3925                          * the inode scan code will try to do
3926                          * something reasonable automatically.
3927                          */
3928                         return 0;
3929                 }
3930                 first_block += fs->super->s_blocks_per_group;
3931         }
3932         /*
3933          * If we've gotten to this point, then the only
3934          * possibility is that the bad block inode meta data
3935          * is using a bad block.
3936          */
3937         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
3938             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
3939             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
3940                 p->bbcheck = 1;
3941                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
3942                         *block_nr = 0;
3943                         return BLOCK_CHANGED;
3944                 }
3945                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3946                         return BLOCK_ABORT;
3947                 return 0;
3948         }
3949
3950         pctx->group = -1;
3951
3952         /* Warn user that the block wasn't claimed */
3953         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
3954
3955         return 0;
3956 }
3957
3958 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
3959                             const char *name, int num, blk64_t *new_block)
3960 {
3961         ext2_filsys fs = ctx->fs;
3962         dgrp_t          last_grp;
3963         blk64_t         old_block = *new_block;
3964         blk64_t         last_block;
3965         dgrp_t          flexbg;
3966         unsigned        flexbg_size;
3967         int             i, is_flexbg;
3968         char            *buf;
3969         struct problem_context  pctx;
3970
3971         clear_problem_context(&pctx);
3972
3973         pctx.group = group;
3974         pctx.blk = old_block;
3975         pctx.str = name;
3976
3977         /*
3978          * For flex_bg filesystems, first try to allocate the metadata
3979          * within the flex_bg, and if that fails then try finding the
3980          * space anywhere in the filesystem.
3981          */
3982         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
3983         if (is_flexbg) {
3984                 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
3985                 flexbg = group / flexbg_size;
3986                 first_block = ext2fs_group_first_block2(fs,
3987                                                         flexbg_size * flexbg);
3988                 last_grp = group | (flexbg_size - 1);
3989                 if (last_grp >= fs->group_desc_count)
3990                         last_grp = fs->group_desc_count - 1;
3991                 last_block = ext2fs_group_last_block2(fs, last_grp);
3992         } else
3993                 last_block = ext2fs_group_last_block2(fs, group);
3994         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
3995                                                num, ctx->block_found_map,
3996                                                new_block);
3997         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
3998                 pctx.errcode = ext2fs_get_free_blocks2(fs,
3999                                 fs->super->s_first_data_block,
4000                                 ext2fs_blocks_count(fs->super),
4001                                 num, ctx->block_found_map, new_block);
4002         if (pctx.errcode) {
4003                 pctx.num = num;
4004                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4005                 ext2fs_unmark_valid(fs);
4006                 ctx->flags |= E2F_FLAG_ABORT;
4007                 return;
4008         }
4009         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4010         if (pctx.errcode) {
4011                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4012                 ext2fs_unmark_valid(fs);
4013                 ctx->flags |= E2F_FLAG_ABORT;
4014                 return;
4015         }
4016         ext2fs_mark_super_dirty(fs);
4017         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4018         pctx.blk2 = *new_block;
4019         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4020                           PR_1_RELOC_TO), &pctx);
4021         pctx.blk2 = 0;
4022         for (i = 0; i < num; i++) {
4023                 pctx.blk = i;
4024                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4025                 if (old_block) {
4026                         pctx.errcode = io_channel_read_blk64(fs->io,
4027                                    old_block + i, 1, buf);
4028                         if (pctx.errcode)
4029                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4030                         pctx.blk = (*new_block) + i;
4031                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4032                                                               1, buf);
4033                 } else {
4034                         pctx.blk = (*new_block) + i;
4035                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4036                                                            NULL, NULL);
4037                 }
4038
4039                 if (pctx.errcode)
4040                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4041         }
4042         ext2fs_free_mem(&buf);
4043 }
4044
4045 /*
4046  * This routine gets called at the end of pass 1 if bad blocks are
4047  * detected in the superblock, group descriptors, inode_bitmaps, or
4048  * block bitmaps.  At this point, all of the blocks have been mapped
4049  * out, so we can try to allocate new block(s) to replace the bad
4050  * blocks.
4051  */
4052 static void handle_fs_bad_blocks(e2fsck_t ctx)
4053 {
4054         ext2_filsys fs = ctx->fs;
4055         dgrp_t          i;
4056         blk64_t         first_block;
4057         blk64_t         new_blk;
4058
4059         for (i = 0; i < fs->group_desc_count; i++) {
4060                 first_block = ext2fs_group_first_block2(fs, i);
4061
4062                 if (ctx->invalid_block_bitmap_flag[i]) {
4063                         new_blk = ext2fs_block_bitmap_loc(fs, i);
4064                         new_table_block(ctx, first_block, i, _("block bitmap"),
4065                                         1, &new_blk);
4066                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4067                 }
4068                 if (ctx->invalid_inode_bitmap_flag[i]) {
4069                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
4070                         new_table_block(ctx, first_block, i, _("inode bitmap"),
4071                                         1, &new_blk);
4072                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4073                 }
4074                 if (ctx->invalid_inode_table_flag[i]) {
4075                         new_blk = ext2fs_inode_table_loc(fs, i);
4076                         new_table_block(ctx, first_block, i, _("inode table"),
4077                                         fs->inode_blocks_per_group,
4078                                         &new_blk);
4079                         ext2fs_inode_table_loc_set(fs, i, new_blk);
4080                         ctx->flags |= E2F_FLAG_RESTART;
4081                 }
4082         }
4083         ctx->invalid_bitmaps = 0;
4084 }
4085
4086 /*
4087  * This routine marks all blocks which are used by the superblock,
4088  * group descriptors, inode bitmaps, and block bitmaps.
4089  */
4090 static void mark_table_blocks(e2fsck_t ctx)
4091 {
4092         ext2_filsys fs = ctx->fs;
4093         blk64_t b;
4094         dgrp_t  i;
4095         unsigned int    j;
4096         struct problem_context pctx;
4097
4098         clear_problem_context(&pctx);
4099
4100         for (i = 0; i < fs->group_desc_count; i++) {
4101                 pctx.group = i;
4102
4103                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
4104                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
4105
4106                 /*
4107                  * Mark the blocks used for the inode table
4108                  */
4109                 if (ext2fs_inode_table_loc(fs, i)) {
4110                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
4111                              j < fs->inode_blocks_per_group;
4112                              j++, b++) {
4113                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4114                                                              b)) {
4115                                         pctx.blk = b;
4116                                         if (!ctx->invalid_inode_table_flag[i] &&
4117                                             fix_problem(ctx,
4118                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
4119                                                 ctx->invalid_inode_table_flag[i]++;
4120                                                 ctx->invalid_bitmaps++;
4121                                         }
4122                                 } else {
4123                                     ext2fs_mark_block_bitmap2(
4124                                                 ctx->block_found_map, b);
4125                                     ext2fs_mark_block_bitmap2(
4126                                                 ctx->block_metadata_map, b);
4127                                 }
4128                         }
4129                 }
4130
4131                 /*
4132                  * Mark block used for the block bitmap
4133                  */
4134                 if (ext2fs_block_bitmap_loc(fs, i)) {
4135                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4136                                      ext2fs_block_bitmap_loc(fs, i))) {
4137                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
4138                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
4139                                         ctx->invalid_block_bitmap_flag[i]++;
4140                                         ctx->invalid_bitmaps++;
4141                                 }
4142                         } else {
4143                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4144                                      ext2fs_block_bitmap_loc(fs, i));
4145                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4146                                      ext2fs_block_bitmap_loc(fs, i));
4147                         }
4148                 }
4149                 /*
4150                  * Mark block used for the inode bitmap
4151                  */
4152                 if (ext2fs_inode_bitmap_loc(fs, i)) {
4153                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4154                                      ext2fs_inode_bitmap_loc(fs, i))) {
4155                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
4156                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
4157                                         ctx->invalid_inode_bitmap_flag[i]++;
4158                                         ctx->invalid_bitmaps++;
4159                                 }
4160                         } else {
4161                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4162                                      ext2fs_inode_bitmap_loc(fs, i));
4163                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4164                                      ext2fs_inode_bitmap_loc(fs, i));
4165                         }
4166                 }
4167         }
4168 }
4169
4170 /*
4171  * These subroutines short circuits ext2fs_get_blocks and
4172  * ext2fs_check_directory; we use them since we already have the inode
4173  * structure, so there's no point in letting the ext2fs library read
4174  * the inode again.
4175  */
4176 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
4177                                   blk_t *blocks)
4178 {
4179         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4180         int     i;
4181
4182         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4183                 return EXT2_ET_CALLBACK_NOTHANDLED;
4184
4185         for (i=0; i < EXT2_N_BLOCKS; i++)
4186                 blocks[i] = ctx->stashed_inode->i_block[i];
4187         return 0;
4188 }
4189
4190 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
4191                                   struct ext2_inode *inode)
4192 {
4193         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4194
4195         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4196                 return EXT2_ET_CALLBACK_NOTHANDLED;
4197         *inode = *ctx->stashed_inode;
4198         return 0;
4199 }
4200
4201 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
4202                             struct ext2_inode *inode)
4203 {
4204         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4205
4206         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
4207                 (inode != ctx->stashed_inode))
4208                 *ctx->stashed_inode = *inode;
4209         return EXT2_ET_CALLBACK_NOTHANDLED;
4210 }
4211
4212 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
4213 {
4214         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4215
4216         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4217                 return EXT2_ET_CALLBACK_NOTHANDLED;
4218
4219         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
4220                 return EXT2_ET_NO_DIRECTORY;
4221         return 0;
4222 }
4223
4224 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
4225                                         blk64_t *ret)
4226 {
4227         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4228         errcode_t       retval;
4229         blk64_t         new_block;
4230
4231         if (ctx->block_found_map) {
4232                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
4233                                            &new_block);
4234                 if (retval)
4235                         return retval;
4236                 if (fs->block_map) {
4237                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
4238                         ext2fs_mark_bb_dirty(fs);
4239                 }
4240         } else {
4241                 if (!fs->block_map) {
4242                         retval = ext2fs_read_block_bitmap(fs);
4243                         if (retval)
4244                                 return retval;
4245                 }
4246
4247                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
4248                 if (retval)
4249                         return retval;
4250         }
4251
4252         *ret = new_block;
4253         return (0);
4254 }
4255
4256 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
4257                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
4258 {
4259         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4260         errcode_t       retval;
4261
4262         if (ctx->block_found_map)
4263                 return ext2fs_new_range(fs, flags, goal, len,
4264                                         ctx->block_found_map, pblk, plen);
4265
4266         if (!fs->block_map) {
4267                 retval = ext2fs_read_block_bitmap(fs);
4268                 if (retval)
4269                         return retval;
4270         }
4271
4272         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
4273                                 pblk, plen);
4274 }
4275
4276 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
4277 {
4278         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4279
4280         /* Never free a critical metadata block */
4281         if (ctx->block_found_map &&
4282             ctx->block_metadata_map &&
4283             inuse < 0 &&
4284             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
4285                 return;
4286
4287         if (ctx->block_found_map) {
4288                 if (inuse > 0)
4289                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4290                 else
4291                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
4292         }
4293 }
4294
4295 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
4296                                            blk_t num, int inuse)
4297 {
4298         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4299
4300         /* Never free a critical metadata block */
4301         if (ctx->block_found_map &&
4302             ctx->block_metadata_map &&
4303             inuse < 0 &&
4304             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
4305                 return;
4306
4307         if (ctx->block_found_map) {
4308                 if (inuse > 0)
4309                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
4310                                                         blk, num);
4311                 else
4312                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
4313                                                         blk, num);
4314         }
4315 }
4316
4317 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
4318 {
4319         ext2_filsys fs = ctx->fs;
4320
4321         if (use_shortcuts) {
4322                 fs->get_blocks = pass1_get_blocks;
4323                 fs->check_directory = pass1_check_directory;
4324                 fs->read_inode = pass1_read_inode;
4325                 fs->write_inode = pass1_write_inode;
4326                 ctx->stashed_ino = 0;
4327         } else {
4328                 fs->get_blocks = 0;
4329                 fs->check_directory = 0;
4330                 fs->read_inode = 0;
4331                 fs->write_inode = 0;
4332         }
4333 }
4334
4335 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
4336 {
4337         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
4338         ext2fs_set_block_alloc_stats_callback(ctx->fs,
4339                                                 e2fsck_block_alloc_stats, 0);
4340         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
4341         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
4342                                         e2fsck_block_alloc_stats_range, NULL);
4343 }