Whamcloud - gitweb
Merge branch 'maint' into next
[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 #include <e2p/e2p.h>
52
53 #include "problem.h"
54
55 #ifdef NO_INLINE_FUNCS
56 #define _INLINE_
57 #else
58 #define _INLINE_ inline
59 #endif
60
61 #undef DEBUG
62
63 struct ea_quota {
64         blk64_t blocks;
65         __u64 inodes;
66 };
67
68 static int process_block(ext2_filsys fs, blk64_t        *blocknr,
69                          e2_blkcnt_t blockcnt, blk64_t ref_blk,
70                          int ref_offset, void *priv_data);
71 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
72                              e2_blkcnt_t blockcnt, blk64_t ref_blk,
73                              int ref_offset, void *priv_data);
74 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
75                          char *block_buf,
76                          const struct ea_quota *ea_ibody_quota);
77 static void mark_table_blocks(e2fsck_t ctx);
78 static void alloc_bb_map(e2fsck_t ctx);
79 static void alloc_imagic_map(e2fsck_t ctx);
80 static void mark_inode_bad(e2fsck_t ctx, ino_t ino);
81 static void add_encrypted_dir(e2fsck_t ctx, ino_t ino);
82 static void handle_fs_bad_blocks(e2fsck_t ctx);
83 static void process_inodes(e2fsck_t ctx, char *block_buf);
84 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
85 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
86                                   dgrp_t group, void * priv_data);
87 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
88                                     char *block_buf, int adjust_sign);
89 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
90
91 struct process_block_struct {
92         ext2_ino_t      ino;
93         unsigned        is_dir:1, is_reg:1, clear:1, suppress:1,
94                                 fragmented:1, compressed:1, bbcheck:1,
95                                 inode_modified:1;
96         blk64_t         num_blocks;
97         blk64_t         max_blocks;
98         blk64_t         last_block;
99         e2_blkcnt_t     last_init_lblock;
100         e2_blkcnt_t     last_db_block;
101         int             num_illegal_blocks;
102         blk64_t         previous_block;
103         struct ext2_inode *inode;
104         struct problem_context *pctx;
105         ext2fs_block_bitmap fs_meta_blocks;
106         e2fsck_t        ctx;
107         blk64_t         next_lblock;
108         struct extent_tree_info eti;
109 };
110
111 struct process_inode_block {
112         ext2_ino_t ino;
113         struct ea_quota ea_ibody_quota;
114         struct ext2_inode_large inode;
115 };
116
117 struct scan_callback_struct {
118         e2fsck_t        ctx;
119         char            *block_buf;
120 };
121
122 /*
123  * For the inodes to process list.
124  */
125 static struct process_inode_block *inodes_to_process;
126 static int process_inode_count;
127
128 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
129                             EXT2_MIN_BLOCK_LOG_SIZE + 1];
130
131 /*
132  * Free all memory allocated by pass1 in preparation for restarting
133  * things.
134  */
135 static void unwind_pass1(ext2_filsys fs EXT2FS_ATTR((unused)))
136 {
137         ext2fs_free_mem(&inodes_to_process);
138         inodes_to_process = 0;
139 }
140
141 /*
142  * Check to make sure a device inode is real.  Returns 1 if the device
143  * checks out, 0 if not.
144  *
145  * Note: this routine is now also used to check FIFO's and Sockets,
146  * since they have the same requirement; the i_block fields should be
147  * zero.
148  */
149 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
150                                     struct ext2_inode *inode)
151 {
152         int     i;
153
154         /*
155          * If the index or extents flag is set, then this is a bogus
156          * device/fifo/socket
157          */
158         if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
159                 return 0;
160
161         /*
162          * We should be able to do the test below all the time, but
163          * because the kernel doesn't forcibly clear the device
164          * inode's additional i_block fields, there are some rare
165          * occasions when a legitimate device inode will have non-zero
166          * additional i_block fields.  So for now, we only complain
167          * when the immutable flag is set, which should never happen
168          * for devices.  (And that's when the problem is caused, since
169          * you can't set or clear immutable flags for devices.)  Once
170          * the kernel has been fixed we can change this...
171          */
172         if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
173                 for (i=4; i < EXT2_N_BLOCKS; i++)
174                         if (inode->i_block[i])
175                                 return 0;
176         }
177         return 1;
178 }
179
180 /*
181  * Check to make sure a symlink inode is real.  Returns 1 if the symlink
182  * checks out, 0 if not.
183  */
184 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
185                                struct ext2_inode *inode, char *buf)
186 {
187         unsigned int buflen;
188         unsigned int len;
189
190         if ((inode->i_size_high || inode->i_size == 0) ||
191             (inode->i_flags & EXT2_INDEX_FL))
192                 return 0;
193
194         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
195                 size_t inline_size;
196
197                 if (inode->i_flags & EXT4_EXTENTS_FL)
198                         return 0;
199                 if (ext2fs_inline_data_size(fs, ino, &inline_size))
200                         return 0;
201                 if (inode->i_size != inline_size)
202                         return 0;
203
204                 return 1;
205         }
206
207         if (ext2fs_is_fast_symlink(inode)) {
208                 if (inode->i_flags & EXT4_EXTENTS_FL)
209                         return 0;
210                 buf = (char *)inode->i_block;
211                 buflen = sizeof(inode->i_block);
212         } else {
213                 ext2_extent_handle_t    handle;
214                 struct ext2_extent_info info;
215                 struct ext2fs_extent    extent;
216                 blk64_t blk;
217                 int i;
218
219                 if (inode->i_flags & EXT4_EXTENTS_FL) {
220                         if (ext2fs_extent_open2(fs, ino, inode, &handle))
221                                 return 0;
222                         if (ext2fs_extent_get_info(handle, &info) ||
223                             (info.num_entries != 1) ||
224                             (info.max_depth != 0)) {
225                                 ext2fs_extent_free(handle);
226                                 return 0;
227                         }
228                         if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
229                                               &extent) ||
230                             (extent.e_lblk != 0) ||
231                             (extent.e_len != 1)) {
232                                 ext2fs_extent_free(handle);
233                                 return 0;
234                         }
235                         blk = extent.e_pblk;
236                         ext2fs_extent_free(handle);
237                 } else {
238                         blk = inode->i_block[0];
239
240                         for (i = 1; i < EXT2_N_BLOCKS; i++)
241                                 if (inode->i_block[i])
242                                         return 0;
243                 }
244
245                 if (blk < fs->super->s_first_data_block ||
246                     blk >= ext2fs_blocks_count(fs->super))
247                         return 0;
248
249                 if (io_channel_read_blk64(fs->io, blk, 1, buf))
250                         return 0;
251
252                 buflen = fs->blocksize;
253         }
254
255         if (inode->i_flags & EXT4_ENCRYPT_FL)
256                 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
257         else
258                 len = strnlen(buf, buflen);
259
260         if (len >= buflen)
261                 return 0;
262
263         if (len != inode->i_size)
264                 return 0;
265         return 1;
266 }
267
268 /*
269  * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
270  */
271 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
272 static void check_extents_inlinedata(e2fsck_t ctx,
273                                      struct problem_context *pctx)
274 {
275         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
276                 return;
277
278         if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
279                 return;
280
281         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
282         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
283 }
284 #undef BAD_SPECIAL_FLAGS
285
286 /*
287  * If the immutable (or append-only) flag is set on the inode, offer
288  * to clear it.
289  */
290 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
291 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
292 {
293         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
294                 return;
295
296         if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
297                 return;
298
299         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
300         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
301 }
302
303 /*
304  * If device, fifo or socket, check size is zero -- if not offer to
305  * clear it
306  */
307 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
308 {
309         struct ext2_inode *inode = pctx->inode;
310
311         if (EXT2_I_SIZE(inode) == 0)
312                 return;
313
314         if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
315                 return;
316
317         ext2fs_inode_size_set(ctx->fs, inode, 0);
318         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
319 }
320
321 /*
322  * For a given size, calculate how many blocks would be charged towards quota.
323  */
324 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
325 {
326         blk64_t clusters;
327
328         clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
329         return EXT2FS_C2B(fs, clusters);
330 }
331
332 /*
333  * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
334  * the problem code.
335  */
336 static problem_t check_large_ea_inode(e2fsck_t ctx,
337                                       struct ext2_ext_attr_entry *entry,
338                                       struct problem_context *pctx,
339                                       blk64_t *quota_blocks)
340 {
341         struct ext2_inode inode;
342         __u32 hash;
343         errcode_t retval;
344
345         /* Check if inode is within valid range */
346         if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
347             (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
348                 pctx->num = entry->e_value_inum;
349                 return PR_1_ATTR_VALUE_EA_INODE;
350         }
351
352         e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
353
354         retval = ext2fs_ext_attr_hash_entry2(ctx->fs, entry, NULL, &hash);
355         if (retval) {
356                 com_err("check_large_ea_inode", retval,
357                         _("while hashing entry with e_value_inum = %u"),
358                         entry->e_value_inum);
359                 fatal_error(ctx, 0);
360         }
361
362         if (hash == entry->e_hash) {
363                 *quota_blocks = size_to_quota_blocks(ctx->fs,
364                                                      entry->e_value_size);
365         } else {
366                 /* This might be an old Lustre-style ea_inode reference. */
367                 if (inode.i_mtime == pctx->ino &&
368                     inode.i_generation == pctx->inode->i_generation) {
369                         *quota_blocks = 0;
370                 } else {
371                         /* If target inode is also missing EA_INODE flag,
372                          * this is likely to be a bad reference.
373                          */
374                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
375                                 pctx->num = entry->e_value_inum;
376                                 return PR_1_ATTR_VALUE_EA_INODE;
377                         } else {
378                                 pctx->num = entry->e_hash;
379                                 return PR_1_ATTR_HASH;
380                         }
381                 }
382         }
383
384         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
385                 pctx->num = entry->e_value_inum;
386                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
387                         inode.i_flags |= EXT4_EA_INODE_FL;
388                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
389                                            &inode);
390                 } else {
391                         return PR_1_ATTR_NO_EA_INODE_FL;
392                 }
393         }
394         return 0;
395 }
396
397 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
398                               struct ext2_ext_attr_entry *first, void *end)
399 {
400         struct ext2_ext_attr_entry *entry;
401
402         for (entry = first;
403              (void *)entry < end && !EXT2_EXT_IS_LAST_ENTRY(entry);
404              entry = EXT2_EXT_ATTR_NEXT(entry)) {
405                 if (!entry->e_value_inum)
406                         continue;
407                 if (!ctx->ea_inode_refs) {
408                         pctx->errcode = ea_refcount_create(0,
409                                                            &ctx->ea_inode_refs);
410                         if (pctx->errcode) {
411                                 pctx->num = 4;
412                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
413                                 ctx->flags |= E2F_FLAG_ABORT;
414                                 return;
415                         }
416                 }
417                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
418                                       0);
419         }
420 }
421
422 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
423                               struct ea_quota *ea_ibody_quota)
424 {
425         struct ext2_super_block *sb = ctx->fs->super;
426         struct ext2_inode_large *inode;
427         struct ext2_ext_attr_entry *entry;
428         char *start, *header, *end;
429         unsigned int storage_size, remain;
430         problem_t problem = 0;
431         region_t region = 0;
432
433         ea_ibody_quota->blocks = 0;
434         ea_ibody_quota->inodes = 0;
435
436         inode = (struct ext2_inode_large *) pctx->inode;
437         storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
438                 inode->i_extra_isize;
439         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
440                  inode->i_extra_isize;
441         end = header + storage_size;
442         start = header + sizeof(__u32);
443         entry = (struct ext2_ext_attr_entry *) start;
444
445         /* scan all entry's headers first */
446
447         /* take finish entry 0UL into account */
448         remain = storage_size - sizeof(__u32);
449
450         region = region_create(0, storage_size);
451         if (!region) {
452                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
453                 problem = 0;
454                 ctx->flags |= E2F_FLAG_ABORT;
455                 return;
456         }
457         if (region_allocate(region, 0, sizeof(__u32))) {
458                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
459                 goto fix;
460         }
461
462         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
463                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
464                 __u32 hash;
465
466                 if (region_allocate(region, (char *)entry - (char *)header,
467                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
468                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
469                         goto fix;
470                 }
471
472                 /* header eats this space */
473                 remain -= sizeof(struct ext2_ext_attr_entry);
474
475                 /* is attribute name valid? */
476                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
477                         pctx->num = entry->e_name_len;
478                         problem = PR_1_ATTR_NAME_LEN;
479                         goto fix;
480                 }
481
482                 /* attribute len eats this space */
483                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
484
485                 if (entry->e_value_inum == 0) {
486                         /* check value size */
487                         if (entry->e_value_size > remain) {
488                                 pctx->num = entry->e_value_size;
489                                 problem = PR_1_ATTR_VALUE_SIZE;
490                                 goto fix;
491                         }
492
493                         if (entry->e_value_size &&
494                             region_allocate(region,
495                                             sizeof(__u32) + entry->e_value_offs,
496                                             EXT2_EXT_ATTR_SIZE(
497                                                 entry->e_value_size))) {
498                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
499                                 goto fix;
500                         }
501
502                         hash = ext2fs_ext_attr_hash_entry(entry,
503                                                           start + entry->e_value_offs);
504
505                         /* e_hash may be 0 in older inode's ea */
506                         if (entry->e_hash != 0 && entry->e_hash != hash) {
507                                 pctx->num = entry->e_hash;
508                                 problem = PR_1_ATTR_HASH;
509                                 goto fix;
510                         }
511                 } else {
512                         blk64_t quota_blocks;
513
514                         problem = check_large_ea_inode(ctx, entry, pctx,
515                                                        &quota_blocks);
516                         if (problem != 0)
517                                 goto fix;
518
519                         ea_ibody_quota->blocks += quota_blocks;
520                         ea_ibody_quota->inodes++;
521                 }
522
523                 /* If EA value is stored in external inode then it does not
524                  * consume space here */
525                 if (entry->e_value_inum == 0)
526                         remain -= entry->e_value_size;
527
528                 entry = EXT2_EXT_ATTR_NEXT(entry);
529         }
530
531         if (region_allocate(region, (char *)entry - (char *)header,
532                             sizeof(__u32))) {
533                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
534                 goto fix;
535         }
536 fix:
537         if (region)
538                 region_free(region);
539         /*
540          * it seems like a corruption. it's very unlikely we could repair
541          * EA(s) in automatic fashion -bzzz
542          */
543         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
544                 inc_ea_inode_refs(ctx, pctx,
545                                   (struct ext2_ext_attr_entry *)start, end);
546                 return;
547         }
548
549         /* simply remove all possible EA(s) */
550         *((__u32 *)header) = 0UL;
551         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
552                                 EXT2_INODE_SIZE(sb), "pass1");
553         ea_ibody_quota->blocks = 0;
554         ea_ibody_quota->inodes = 0;
555 }
556
557 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
558         return (xtime & (1U << 31)) != 0 &&
559                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
560 }
561
562 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
563         check_inode_extra_negative_epoch(inode->i_##xtime, \
564                                          inode->i_##xtime##_extra)
565
566 /* When today's date is earlier than 2242, we assume that atimes,
567  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
568  * actually pre-1970 dates mis-encoded.
569  */
570 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
571
572 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
573                                     struct ea_quota *ea_ibody_quota)
574 {
575         struct ext2_super_block *sb = ctx->fs->super;
576         struct ext2_inode_large *inode;
577         __u32 *eamagic;
578         int min, max;
579
580         ea_ibody_quota->blocks = 0;
581         ea_ibody_quota->inodes = 0;
582
583         inode = (struct ext2_inode_large *) pctx->inode;
584         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
585                 /* this isn't large inode. so, nothing to check */
586                 return;
587         }
588
589 #if 0
590         printf("inode #%u, i_extra_size %d\n", pctx->ino,
591                         inode->i_extra_isize);
592 #endif
593         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
594         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
595         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
596         /*
597          * For now we will allow i_extra_isize to be 0, but really
598          * implementations should never allow i_extra_isize to be 0
599          */
600         if (inode->i_extra_isize &&
601             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
602              inode->i_extra_isize & 3)) {
603                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
604                         return;
605                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
606                         inode->i_extra_isize = sb->s_want_extra_isize;
607                 else
608                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
609                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
610                                         EXT2_INODE_SIZE(sb), "pass1");
611         }
612
613         /* check if there is no place for an EA header */
614         if (inode->i_extra_isize >= max - sizeof(__u32))
615                 return;
616
617         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
618                         inode->i_extra_isize);
619         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
620                 /* it seems inode has an extended attribute(s) in body */
621                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
622         }
623
624         /*
625          * If the inode's extended atime (ctime, crtime, mtime) is stored in
626          * the old, invalid format, repair it.
627          */
628         if (((sizeof(time_t) <= 4) ||
629              (((sizeof(time_t) > 4) &&
630                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
631             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
632              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
633              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
634              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
635
636                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
637                         return;
638
639                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
640                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
641                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
642                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
643                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
644                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
645                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
646                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
647                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
648                                         EXT2_INODE_SIZE(sb), "pass1");
649         }
650
651 }
652
653 /*
654  * Check to see if the inode might really be a directory, despite i_mode
655  *
656  * This is a lot of complexity for something for which I'm not really
657  * convinced happens frequently in the wild.  If for any reason this
658  * causes any problems, take this code out.
659  * [tytso:20070331.0827EDT]
660  */
661 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
662                                 char *buf)
663 {
664         struct ext2_inode *inode = pctx->inode;
665         struct ext2_dir_entry   *dirent;
666         errcode_t               retval;
667         blk64_t                 blk;
668         unsigned int            i, rec_len, not_device = 0;
669         int                     extent_fs;
670         int                     inlinedata_fs;
671
672         /*
673          * If the mode looks OK, we believe it.  If the first block in
674          * the i_block array is 0, this cannot be a directory. If the
675          * inode is extent-mapped, it is still the case that the latter
676          * cannot be 0 - the magic number in the extent header would make
677          * it nonzero.
678          */
679         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
680             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
681                 return;
682
683         /* 
684          * Check the block numbers in the i_block array for validity:
685          * zero blocks are skipped (but the first one cannot be zero -
686          * see above), other blocks are checked against the first and
687          * max data blocks (from the the superblock) and against the
688          * block bitmap. Any invalid block found means this cannot be
689          * a directory.
690          * 
691          * If there are non-zero blocks past the fourth entry, then
692          * this cannot be a device file: we remember that for the next
693          * check.
694          *
695          * For extent mapped files, we don't do any sanity checking:
696          * just try to get the phys block of logical block 0 and run
697          * with it.
698          *
699          * For inline data files, we just try to get the size of inline
700          * data.  If it's true, we will treat it as a directory.
701          */
702
703         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
704         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
705         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
706                 size_t size;
707                 __u32 dotdot;
708                 unsigned int rec_len2;
709                 struct ext2_dir_entry de;
710
711                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
712                         return;
713                 /*
714                  * If the size isn't a multiple of 4, it's probably not a
715                  * directory??
716                  */
717                 if (size & 3)
718                         return;
719                 /*
720                  * If the first 10 bytes don't look like a directory entry,
721                  * it's probably not a directory.
722                  */
723                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
724                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
725                        EXT2_DIR_REC_LEN(0));
726                 dotdot = ext2fs_le32_to_cpu(dotdot);
727                 de.inode = ext2fs_le32_to_cpu(de.inode);
728                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
729                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
730                 if (dotdot >= ctx->fs->super->s_inodes_count ||
731                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
732                      dotdot != EXT2_ROOT_INO) ||
733                     de.inode >= ctx->fs->super->s_inodes_count ||
734                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
735                      de.inode != 0) ||
736                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
737                               EXT4_INLINE_DATA_DOTDOT_SIZE)
738                         return;
739                 /* device files never have a "system.data" entry */
740                 goto isdir;
741         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
742                 /* extent mapped */
743                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
744                                  &blk))
745                         return;
746                 /* device files are never extent mapped */
747                 not_device++;
748         } else {
749                 for (i=0; i < EXT2_N_BLOCKS; i++) {
750                         blk = inode->i_block[i];
751                         if (!blk)
752                                 continue;
753                         if (i >= 4)
754                                 not_device++;
755
756                         if (blk < ctx->fs->super->s_first_data_block ||
757                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
758                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
759                                                            blk))
760                                 return; /* Invalid block, can't be dir */
761                 }
762                 blk = inode->i_block[0];
763         }
764
765         /*
766          * If the mode says this is a device file and the i_links_count field
767          * is sane and we have not ruled it out as a device file previously,
768          * we declare it a device file, not a directory.
769          */
770         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
771             (inode->i_links_count == 1) && !not_device)
772                 return;
773
774         /* read the first block */
775         ehandler_operation(_("reading directory block"));
776         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
777         ehandler_operation(0);
778         if (retval)
779                 return;
780
781         dirent = (struct ext2_dir_entry *) buf;
782         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
783         if (retval)
784                 return;
785         if ((ext2fs_dirent_name_len(dirent) != 1) ||
786             (dirent->name[0] != '.') ||
787             (dirent->inode != pctx->ino) ||
788             (rec_len < 12) ||
789             (rec_len % 4) ||
790             (rec_len >= ctx->fs->blocksize - 12))
791                 return;
792
793         dirent = (struct ext2_dir_entry *) (buf + rec_len);
794         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
795         if (retval)
796                 return;
797         if ((ext2fs_dirent_name_len(dirent) != 2) ||
798             (dirent->name[0] != '.') ||
799             (dirent->name[1] != '.') ||
800             (rec_len < 12) ||
801             (rec_len % 4))
802                 return;
803
804 isdir:
805         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
806                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
807                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
808                                         EXT2_INODE_SIZE(ctx->fs->super),
809                                         "check_is_really_dir");
810         }
811 }
812
813 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
814                                      int flags, ext2_icount_t hint,
815                                      ext2_icount_t *ret)
816 {
817         unsigned int            threshold;
818         unsigned int            save_type;
819         ext2_ino_t              num_dirs;
820         errcode_t               retval;
821         char                    *tdb_dir;
822         int                     enable;
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, casefold_fs;
1175         int             low_dtime_check = 1;
1176         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1177         unsigned 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         casefold_fs = ext2fs_has_feature_casefold(sb);
1221
1222         /*
1223          * Allocate bitmaps structures
1224          */
1225         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1226                                                     EXT2FS_BMAP64_RBTREE,
1227                                                     "inode_used_map",
1228                                                     &ctx->inode_used_map);
1229         if (pctx.errcode) {
1230                 pctx.num = 1;
1231                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1232                 ctx->flags |= E2F_FLAG_ABORT;
1233                 return;
1234         }
1235         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1236                         _("directory inode map"),
1237                         EXT2FS_BMAP64_AUTODIR,
1238                         "inode_dir_map", &ctx->inode_dir_map);
1239         if (pctx.errcode) {
1240                 pctx.num = 2;
1241                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1242                 ctx->flags |= E2F_FLAG_ABORT;
1243                 return;
1244         }
1245         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1246                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1247                         "inode_reg_map", &ctx->inode_reg_map);
1248         if (pctx.errcode) {
1249                 pctx.num = 6;
1250                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1251                 ctx->flags |= E2F_FLAG_ABORT;
1252                 return;
1253         }
1254         pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1255                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1256                         "block_found_map", &ctx->block_found_map);
1257         if (pctx.errcode) {
1258                 pctx.num = 1;
1259                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1260                 ctx->flags |= E2F_FLAG_ABORT;
1261                 return;
1262         }
1263         pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1264                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1265                         "block_metadata_map", &ctx->block_metadata_map);
1266         if (pctx.errcode) {
1267                 pctx.num = 1;
1268                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1269                 ctx->flags |= E2F_FLAG_ABORT;
1270                 return;
1271         }
1272         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1273                                            &ctx->inode_link_info);
1274         if (pctx.errcode) {
1275                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1276                 ctx->flags |= E2F_FLAG_ABORT;
1277                 return;
1278         }
1279         bufsize = inode_size;
1280         if (bufsize < sizeof(struct ext2_inode_large))
1281                 bufsize = sizeof(struct ext2_inode_large);
1282         inode = (struct ext2_inode *)
1283                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1284
1285         inodes_to_process = (struct process_inode_block *)
1286                 e2fsck_allocate_memory(ctx,
1287                                        (ctx->process_inode_size *
1288                                         sizeof(struct process_inode_block)),
1289                                        "array of inodes to process");
1290         process_inode_count = 0;
1291
1292         pctx.errcode = ext2fs_init_dblist(fs, 0);
1293         if (pctx.errcode) {
1294                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1295                 ctx->flags |= E2F_FLAG_ABORT;
1296                 goto endit;
1297         }
1298
1299         /*
1300          * If the last orphan field is set, clear it, since the pass1
1301          * processing will automatically find and clear the orphans.
1302          * In the future, we may want to try using the last_orphan
1303          * linked list ourselves, but for now, we clear it so that the
1304          * ext3 mount code won't get confused.
1305          */
1306         if (!(ctx->options & E2F_OPT_READONLY)) {
1307                 if (fs->super->s_last_orphan) {
1308                         fs->super->s_last_orphan = 0;
1309                         ext2fs_mark_super_dirty(fs);
1310                 }
1311         }
1312
1313         mark_table_blocks(ctx);
1314         pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1315                                                 &ctx->block_found_map);
1316         if (pctx.errcode) {
1317                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1318                 ctx->flags |= E2F_FLAG_ABORT;
1319                 goto endit;
1320         }
1321         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1322                                                     "block interate buffer");
1323         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1324                 e2fsck_use_inode_shortcuts(ctx, 1);
1325         e2fsck_intercept_block_allocations(ctx);
1326         old_op = ehandler_operation(_("opening inode scan"));
1327         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1328                                               &scan);
1329         ehandler_operation(old_op);
1330         if (pctx.errcode) {
1331                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1332                 ctx->flags |= E2F_FLAG_ABORT;
1333                 goto endit;
1334         }
1335         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1336                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1337         ctx->stashed_inode = inode;
1338         scan_struct.ctx = ctx;
1339         scan_struct.block_buf = block_buf;
1340         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1341         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1342                                               ctx->fs->group_desc_count)))
1343                 goto endit;
1344         if ((fs->super->s_wtime &&
1345              fs->super->s_wtime < fs->super->s_inodes_count) ||
1346             (fs->super->s_mtime &&
1347              fs->super->s_mtime < fs->super->s_inodes_count) ||
1348             (fs->super->s_mkfs_time &&
1349              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1350                 low_dtime_check = 0;
1351
1352         if (ext2fs_has_feature_mmp(fs->super) &&
1353             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1354             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1355                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1356                                           fs->super->s_mmp_block);
1357
1358         /* Set up ctx->lost_and_found if possible */
1359         (void) e2fsck_get_lost_and_found(ctx, 0);
1360
1361         while (1) {
1362                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1363                         if (e2fsck_mmp_update(fs))
1364                                 fatal_error(ctx, 0);
1365                 }
1366                 old_op = ehandler_operation(_("getting next inode from scan"));
1367                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1368                                                           inode, inode_size);
1369                 if (ino > ino_threshold)
1370                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1371                 ehandler_operation(old_op);
1372                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1373                         goto endit;
1374                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1375                         /*
1376                          * If badblocks says badblocks is bad, offer to clear
1377                          * the list, update the in-core bb list, and restart
1378                          * the inode scan.
1379                          */
1380                         if (ino == EXT2_BAD_INO &&
1381                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1382                                         &pctx)) {
1383                                 errcode_t err;
1384
1385                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1386                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1387                                 ctx->fs->badblocks = NULL;
1388                                 err = ext2fs_read_bb_inode(ctx->fs,
1389                                                         &ctx->fs->badblocks);
1390                                 if (err) {
1391                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1392                                                     &pctx);
1393                                         ctx->flags |= E2F_FLAG_ABORT;
1394                                         goto endit;
1395                                 }
1396                                 err = ext2fs_inode_scan_goto_blockgroup(scan,
1397                                                                         0);
1398                                 if (err) {
1399                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1400                                                     &pctx);
1401                                         ctx->flags |= E2F_FLAG_ABORT;
1402                                         goto endit;
1403                                 }
1404                                 continue;
1405                         }
1406                         if (!ctx->inode_bb_map)
1407                                 alloc_bb_map(ctx);
1408                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1409                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1410                         continue;
1411                 }
1412                 if (pctx.errcode &&
1413                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1414                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1415                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1416                         ctx->flags |= E2F_FLAG_ABORT;
1417                         goto endit;
1418                 }
1419                 if (!ino)
1420                         break;
1421                 pctx.ino = ino;
1422                 pctx.inode = inode;
1423                 ctx->stashed_ino = ino;
1424
1425                 /* Clear trashed inode? */
1426                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1427                     inode->i_links_count > 0 &&
1428                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1429                         pctx.errcode = 0;
1430                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1431                 }
1432                 failed_csum = pctx.errcode != 0;
1433
1434                 /*
1435                  * Check for inodes who might have been part of the
1436                  * orphaned list linked list.  They should have gotten
1437                  * dealt with by now, unless the list had somehow been
1438                  * corrupted.
1439                  *
1440                  * FIXME: In the future, inodes which are still in use
1441                  * (and which are therefore) pending truncation should
1442                  * be handled specially.  Right now we just clear the
1443                  * dtime field, and the normal e2fsck handling of
1444                  * inodes where i_size and the inode blocks are
1445                  * inconsistent is to fix i_size, instead of releasing
1446                  * the extra blocks.  This won't catch the inodes that
1447                  * was at the end of the orphan list, but it's better
1448                  * than nothing.  The right answer is that there
1449                  * shouldn't be any bugs in the orphan list handling.  :-)
1450                  */
1451                 if (inode->i_dtime && low_dtime_check &&
1452                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1453                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1454                                 inode->i_dtime = inode->i_links_count ?
1455                                         0 : ctx->now;
1456                                 e2fsck_write_inode(ctx, ino, inode,
1457                                                    "pass1");
1458                                 failed_csum = 0;
1459                         }
1460                 }
1461
1462                 if (inode->i_links_count) {
1463                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1464                                            ino, inode->i_links_count);
1465                         if (pctx.errcode) {
1466                                 pctx.num = inode->i_links_count;
1467                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1468                                 ctx->flags |= E2F_FLAG_ABORT;
1469                                 goto endit;
1470                         }
1471                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1472                            !quota_inum_is_reserved(fs, ino)) {
1473                         if (!inode->i_dtime && inode->i_mode) {
1474                                 if (fix_problem(ctx,
1475                                             PR_1_ZERO_DTIME, &pctx)) {
1476                                         inode->i_dtime = ctx->now;
1477                                         e2fsck_write_inode(ctx, ino, inode,
1478                                                            "pass1");
1479                                         failed_csum = 0;
1480                                 }
1481                         }
1482                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1483                         continue;
1484                 }
1485
1486                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1487                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1488                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1489                      (!casefold_fs &&
1490                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1491                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1492                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1493                 }
1494
1495                 /* Conflicting inlinedata/extents inode flags? */
1496                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1497                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1498                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1499                                                                inode_size,
1500                                                                &pctx);
1501                         if (res < 0) {
1502                                 /* skip FINISH_INODE_LOOP */
1503                                 continue;
1504                         }
1505                 }
1506
1507                 /* Test for incorrect inline_data flags settings. */
1508                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1509                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1510                         size_t size = 0;
1511
1512                         pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1513                         if (!pctx.errcode &&
1514                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1515                                 ext2fs_set_feature_inline_data(sb);
1516                                 ext2fs_mark_super_dirty(fs);
1517                                 inlinedata_fs = 1;
1518                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1519                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1520                                 /* skip FINISH_INODE_LOOP */
1521                                 continue;
1522                         }
1523                 }
1524
1525                 /* Test for inline data flag but no attr */
1526                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1527                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1528                         size_t size = 0;
1529                         errcode_t err;
1530                         int flags;
1531
1532                         flags = fs->flags;
1533                         if (failed_csum)
1534                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1535                         err = get_inline_data_ea_size(fs, ino, &size);
1536                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1537                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1538
1539                         switch (err) {
1540                         case 0:
1541                                 /* Everything is awesome... */
1542                                 break;
1543                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1544                         case EXT2_ET_BAD_EA_HASH:
1545                         case EXT2_ET_BAD_EA_HEADER:
1546                         case EXT2_ET_EA_BAD_NAME_LEN:
1547                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1548                         case EXT2_ET_EA_KEY_NOT_FOUND:
1549                         case EXT2_ET_EA_NO_SPACE:
1550                         case EXT2_ET_MISSING_EA_FEATURE:
1551                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1552                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1553                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1554                         case EXT2_ET_NO_INLINE_DATA:
1555                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1556                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1557                         case EXT2_ET_EA_INODE_CORRUPTED:
1558                                 /* broken EA or no system.data EA; truncate */
1559                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1560                                                 &pctx)) {
1561                                         err = ext2fs_inode_size_set(fs, inode, 0);
1562                                         if (err) {
1563                                                 pctx.errcode = err;
1564                                                 ctx->flags |= E2F_FLAG_ABORT;
1565                                                 goto endit;
1566                                         }
1567                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1568                                         memset(&inode->i_block, 0,
1569                                                sizeof(inode->i_block));
1570                                         e2fsck_write_inode(ctx, ino, inode,
1571                                                            "pass1");
1572                                         failed_csum = 0;
1573                                 }
1574                                 break;
1575                         default:
1576                                 /* Some other kind of non-xattr error? */
1577                                 pctx.errcode = err;
1578                                 ctx->flags |= E2F_FLAG_ABORT;
1579                                 goto endit;
1580                         }
1581                 }
1582
1583                 /*
1584                  * Test for incorrect extent flag settings.
1585                  *
1586                  * On big-endian machines we must be careful:
1587                  * When the inode is read, the i_block array is not swapped
1588                  * if the extent flag is set.  Therefore if we are testing
1589                  * for or fixing a wrongly-set flag, we must potentially
1590                  * (un)swap before testing, or after fixing.
1591                  */
1592
1593                 /*
1594                  * In this case the extents flag was set when read, so
1595                  * extent_header_verify is ok.  If the inode is cleared,
1596                  * no need to swap... so no extra swapping here.
1597                  */
1598                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1599                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1600                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1601                         if ((ext2fs_extent_header_verify(inode->i_block,
1602                                                  sizeof(inode->i_block)) == 0) &&
1603                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1604                                 ext2fs_set_feature_extents(sb);
1605                                 ext2fs_mark_super_dirty(fs);
1606                                 extent_fs = 1;
1607                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1608                         clear_inode:
1609                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1610                                 if (ino == EXT2_BAD_INO)
1611                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1612                                                                  ino);
1613                                 /* skip FINISH_INODE_LOOP */
1614                                 continue;
1615                         }
1616                 }
1617
1618                 /*
1619                  * For big-endian machines:
1620                  * If the inode didn't have the extents flag set when it
1621                  * was read, then the i_blocks array was swapped.  To test
1622                  * as an extents header, we must swap it back first.
1623                  * IF we then set the extents flag, the entire i_block
1624                  * array must be un/re-swapped to make it proper extents data.
1625                  */
1626                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1627                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1628                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1629                     (LINUX_S_ISREG(inode->i_mode) ||
1630                      LINUX_S_ISDIR(inode->i_mode))) {
1631                         void *ehp;
1632 #ifdef WORDS_BIGENDIAN
1633                         __u32 tmp_block[EXT2_N_BLOCKS];
1634
1635                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1636                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1637                         ehp = tmp_block;
1638 #else
1639                         ehp = inode->i_block;
1640 #endif
1641                         if ((ext2fs_extent_header_verify(ehp,
1642                                          sizeof(inode->i_block)) == 0) &&
1643                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1644                                 inode->i_flags |= EXT4_EXTENTS_FL;
1645 #ifdef WORDS_BIGENDIAN
1646                                 memcpy(inode->i_block, tmp_block,
1647                                        sizeof(inode->i_block));
1648 #endif
1649                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1650                                 failed_csum = 0;
1651                         }
1652                 }
1653
1654                 if (ino == EXT2_BAD_INO) {
1655                         struct process_block_struct pb;
1656
1657                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1658                              inode->i_gid || inode->i_links_count ||
1659                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1660                              inode->i_file_acl) &&
1661                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1662                                 memset(inode, 0, sizeof(struct ext2_inode));
1663                                 e2fsck_write_inode(ctx, ino, inode,
1664                                                    "clear bad inode");
1665                                 failed_csum = 0;
1666                         }
1667
1668                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1669                                                           &pb.fs_meta_blocks);
1670                         if (pctx.errcode) {
1671                                 pctx.num = 4;
1672                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1673                                 ctx->flags |= E2F_FLAG_ABORT;
1674                                 goto endit;
1675                         }
1676                         pb.ino = EXT2_BAD_INO;
1677                         pb.num_blocks = pb.last_block = 0;
1678                         pb.last_db_block = -1;
1679                         pb.num_illegal_blocks = 0;
1680                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1681                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1682                         pb.inode = inode;
1683                         pb.pctx = &pctx;
1684                         pb.ctx = ctx;
1685                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1686                                      block_buf, process_bad_block, &pb);
1687                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1688                         if (pctx.errcode) {
1689                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1690                                 ctx->flags |= E2F_FLAG_ABORT;
1691                                 goto endit;
1692                         }
1693                         if (pb.bbcheck)
1694                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1695                                 ctx->flags |= E2F_FLAG_ABORT;
1696                                 goto endit;
1697                         }
1698                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1699                         clear_problem_context(&pctx);
1700                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1701                         continue;
1702                 } else if (ino == EXT2_ROOT_INO) {
1703                         /*
1704                          * Make sure the root inode is a directory; if
1705                          * not, offer to clear it.  It will be
1706                          * regenerated in pass #3.
1707                          */
1708                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1709                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1710                                         goto clear_inode;
1711                         }
1712                         /*
1713                          * If dtime is set, offer to clear it.  mke2fs
1714                          * version 0.2b created filesystems with the
1715                          * dtime field set for the root and lost+found
1716                          * directories.  We won't worry about
1717                          * /lost+found, since that can be regenerated
1718                          * easily.  But we will fix the root directory
1719                          * as a special case.
1720                          */
1721                         if (inode->i_dtime && inode->i_links_count) {
1722                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1723                                         inode->i_dtime = 0;
1724                                         e2fsck_write_inode(ctx, ino, inode,
1725                                                            "pass1");
1726                                         failed_csum = 0;
1727                                 }
1728                         }
1729                 } else if (ino == EXT2_JOURNAL_INO) {
1730                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1731                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1732                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1733                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1734                                                 &pctx)) {
1735                                         inode->i_mode = LINUX_S_IFREG;
1736                                         e2fsck_write_inode(ctx, ino, inode,
1737                                                            "pass1");
1738                                         failed_csum = 0;
1739                                 }
1740                                 check_blocks(ctx, &pctx, block_buf, NULL);
1741                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1742                                 continue;
1743                         }
1744                         if ((inode->i_links_count ||
1745                              inode->i_blocks || inode->i_block[0]) &&
1746                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1747                                         &pctx)) {
1748                                 memset(inode, 0, inode_size);
1749                                 ext2fs_icount_store(ctx->inode_link_info,
1750                                                     ino, 0);
1751                                 e2fsck_write_inode_full(ctx, ino, inode,
1752                                                         inode_size, "pass1");
1753                                 failed_csum = 0;
1754                         }
1755                 } else if (quota_inum_is_reserved(fs, ino)) {
1756                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1757                         if (ext2fs_has_feature_quota(fs->super) &&
1758                             quota_inum_is_super(fs->super, ino)) {
1759                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1760                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1761                                                         &pctx)) {
1762                                         inode->i_mode = LINUX_S_IFREG;
1763                                         e2fsck_write_inode(ctx, ino, inode,
1764                                                         "pass1");
1765                                         failed_csum = 0;
1766                                 }
1767                                 check_blocks(ctx, &pctx, block_buf, NULL);
1768                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1769                                 continue;
1770                         }
1771                         if ((inode->i_links_count ||
1772                              inode->i_blocks || inode->i_block[0]) &&
1773                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1774                                         &pctx)) {
1775                                 memset(inode, 0, inode_size);
1776                                 ext2fs_icount_store(ctx->inode_link_info,
1777                                                     ino, 0);
1778                                 e2fsck_write_inode_full(ctx, ino, inode,
1779                                                         inode_size, "pass1");
1780                                 failed_csum = 0;
1781                         }
1782                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1783                         problem_t problem = 0;
1784
1785                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1786                         if (ino == EXT2_BOOT_LOADER_INO) {
1787                                 if (LINUX_S_ISDIR(inode->i_mode))
1788                                         problem = PR_1_RESERVED_BAD_MODE;
1789                         } else if (ino == EXT2_RESIZE_INO) {
1790                                 if (inode->i_mode &&
1791                                     !LINUX_S_ISREG(inode->i_mode))
1792                                         problem = PR_1_RESERVED_BAD_MODE;
1793                         } else {
1794                                 if (inode->i_mode != 0)
1795                                         problem = PR_1_RESERVED_BAD_MODE;
1796                         }
1797                         if (problem) {
1798                                 if (fix_problem(ctx, problem, &pctx)) {
1799                                         inode->i_mode = 0;
1800                                         e2fsck_write_inode(ctx, ino, inode,
1801                                                            "pass1");
1802                                         failed_csum = 0;
1803                                 }
1804                         }
1805                         check_blocks(ctx, &pctx, block_buf, NULL);
1806                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1807                         continue;
1808                 }
1809
1810                 if (!inode->i_links_count) {
1811                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1812                         continue;
1813                 }
1814                 /*
1815                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1816                  * deleted files.  Oops.
1817                  *
1818                  * Since all new ext2 implementations get this right,
1819                  * we now assume that the case of non-zero
1820                  * i_links_count and non-zero dtime means that we
1821                  * should keep the file, not delete it.
1822                  *
1823                  */
1824                 if (inode->i_dtime) {
1825                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1826                                 inode->i_dtime = 0;
1827                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1828                                 failed_csum = 0;
1829                         }
1830                 }
1831
1832                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1833                 switch (fs->super->s_creator_os) {
1834                     case EXT2_OS_HURD:
1835                         frag = inode->osd2.hurd2.h_i_frag;
1836                         fsize = inode->osd2.hurd2.h_i_fsize;
1837                         break;
1838                     default:
1839                         frag = fsize = 0;
1840                 }
1841
1842                 if (inode->i_faddr || frag || fsize ||
1843                     (!ext2fs_has_feature_largedir(fs->super) &&
1844                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1845                         mark_inode_bad(ctx, ino);
1846                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1847                     !ext2fs_has_feature_64bit(fs->super) &&
1848                     inode->osd2.linux2.l_i_file_acl_high != 0)
1849                         mark_inode_bad(ctx, ino);
1850                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1851                     !ext2fs_has_feature_huge_file(fs->super) &&
1852                     (inode->osd2.linux2.l_i_blocks_hi != 0))
1853                         mark_inode_bad(ctx, ino);
1854                 if (inode->i_flags & EXT2_IMAGIC_FL) {
1855                         if (imagic_fs) {
1856                                 if (!ctx->inode_imagic_map)
1857                                         alloc_imagic_map(ctx);
1858                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1859                                                          ino);
1860                         } else {
1861                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1862                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
1863                                         e2fsck_write_inode(ctx, ino,
1864                                                            inode, "pass1");
1865                                         failed_csum = 0;
1866                                 }
1867                         }
1868                 }
1869
1870                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1871                 check_is_really_dir(ctx, &pctx, block_buf);
1872
1873                 /*
1874                  * ext2fs_inode_has_valid_blocks2 does not actually look
1875                  * at i_block[] values, so not endian-sensitive here.
1876                  */
1877                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1878                     LINUX_S_ISLNK(inode->i_mode) &&
1879                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1880                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1881                         inode->i_flags &= ~EXT4_EXTENTS_FL;
1882                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1883                         failed_csum = 0;
1884                 }
1885
1886                 if (LINUX_S_ISDIR(inode->i_mode)) {
1887                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1888                         e2fsck_add_dir_info(ctx, ino, 0);
1889                         ctx->fs_directory_count++;
1890                         if (inode->i_flags & EXT4_ENCRYPT_FL)
1891                                 add_encrypted_dir(ctx, ino);
1892                 } else if (LINUX_S_ISREG (inode->i_mode)) {
1893                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1894                         ctx->fs_regular_count++;
1895                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1896                            e2fsck_pass1_check_device_inode(fs, inode)) {
1897                         check_extents_inlinedata(ctx, &pctx);
1898                         check_immutable(ctx, &pctx);
1899                         check_size(ctx, &pctx);
1900                         ctx->fs_chardev_count++;
1901                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1902                            e2fsck_pass1_check_device_inode(fs, inode)) {
1903                         check_extents_inlinedata(ctx, &pctx);
1904                         check_immutable(ctx, &pctx);
1905                         check_size(ctx, &pctx);
1906                         ctx->fs_blockdev_count++;
1907                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1908                            e2fsck_pass1_check_symlink(fs, ino, inode,
1909                                                       block_buf)) {
1910                         check_immutable(ctx, &pctx);
1911                         ctx->fs_symlinks_count++;
1912                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1913                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1914                                 continue;
1915                         } else if (ext2fs_is_fast_symlink(inode)) {
1916                                 ctx->fs_fast_symlinks_count++;
1917                                 check_blocks(ctx, &pctx, block_buf,
1918                                              &ea_ibody_quota);
1919                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1920                                 continue;
1921                         }
1922                 }
1923                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
1924                          e2fsck_pass1_check_device_inode(fs, inode)) {
1925                         check_extents_inlinedata(ctx, &pctx);
1926                         check_immutable(ctx, &pctx);
1927                         check_size(ctx, &pctx);
1928                         ctx->fs_fifo_count++;
1929                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
1930                            e2fsck_pass1_check_device_inode(fs, inode)) {
1931                         check_extents_inlinedata(ctx, &pctx);
1932                         check_immutable(ctx, &pctx);
1933                         check_size(ctx, &pctx);
1934                         ctx->fs_sockets_count++;
1935                 } else
1936                         mark_inode_bad(ctx, ino);
1937                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1938                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
1939                         if (inode->i_block[EXT2_IND_BLOCK])
1940                                 ctx->fs_ind_count++;
1941                         if (inode->i_block[EXT2_DIND_BLOCK])
1942                                 ctx->fs_dind_count++;
1943                         if (inode->i_block[EXT2_TIND_BLOCK])
1944                                 ctx->fs_tind_count++;
1945                 }
1946                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1947                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
1948                     (inode->i_block[EXT2_IND_BLOCK] ||
1949                      inode->i_block[EXT2_DIND_BLOCK] ||
1950                      inode->i_block[EXT2_TIND_BLOCK] ||
1951                      ext2fs_file_acl_block(fs, inode))) {
1952                         struct process_inode_block *itp;
1953
1954                         itp = &inodes_to_process[process_inode_count];
1955                         itp->ino = ino;
1956                         itp->ea_ibody_quota = ea_ibody_quota;
1957                         if (inode_size < sizeof(struct ext2_inode_large))
1958                                 memcpy(&itp->inode, inode, inode_size);
1959                         else
1960                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
1961                         process_inode_count++;
1962                 } else
1963                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
1964
1965                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1966
1967                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1968                         goto endit;
1969
1970                 if (process_inode_count >= ctx->process_inode_size) {
1971                         process_inodes(ctx, block_buf);
1972
1973                         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1974                                 goto endit;
1975                 }
1976         }
1977         process_inodes(ctx, block_buf);
1978         ext2fs_close_inode_scan(scan);
1979         scan = NULL;
1980
1981         reserve_block_for_root_repair(ctx);
1982         reserve_block_for_lnf_repair(ctx);
1983
1984         /*
1985          * If any extended attribute blocks' reference counts need to
1986          * be adjusted, either up (ctx->refcount_extra), or down
1987          * (ctx->refcount), then fix them.
1988          */
1989         if (ctx->refcount) {
1990                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
1991                 ea_refcount_free(ctx->refcount);
1992                 ctx->refcount = 0;
1993         }
1994         if (ctx->refcount_extra) {
1995                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
1996                                         block_buf, +1);
1997                 ea_refcount_free(ctx->refcount_extra);
1998                 ctx->refcount_extra = 0;
1999         }
2000
2001         if (ctx->ea_block_quota_blocks) {
2002                 ea_refcount_free(ctx->ea_block_quota_blocks);
2003                 ctx->ea_block_quota_blocks = 0;
2004         }
2005
2006         if (ctx->ea_block_quota_inodes) {
2007                 ea_refcount_free(ctx->ea_block_quota_inodes);
2008                 ctx->ea_block_quota_inodes = 0;
2009         }
2010
2011         if (ctx->invalid_bitmaps)
2012                 handle_fs_bad_blocks(ctx);
2013
2014         /* We don't need the block_ea_map any more */
2015         if (ctx->block_ea_map) {
2016                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2017                 ctx->block_ea_map = 0;
2018         }
2019
2020         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2021                 clear_problem_context(&pctx);
2022                 pctx.errcode = ext2fs_create_resize_inode(fs);
2023                 if (pctx.errcode) {
2024                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2025                                          &pctx)) {
2026                                 ctx->flags |= E2F_FLAG_ABORT;
2027                                 goto endit;
2028                         }
2029                         pctx.errcode = 0;
2030                 }
2031                 if (!pctx.errcode) {
2032                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2033                                           "recreate inode");
2034                         inode->i_mtime = ctx->now;
2035                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2036                                            "recreate inode");
2037                 }
2038                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2039         }
2040
2041         if (ctx->flags & E2F_FLAG_RESTART) {
2042                 /*
2043                  * Only the master copy of the superblock and block
2044                  * group descriptors are going to be written during a
2045                  * restart, so set the superblock to be used to be the
2046                  * master superblock.
2047                  */
2048                 ctx->use_superblock = 0;
2049                 unwind_pass1(fs);
2050                 goto endit;
2051         }
2052
2053         if (ctx->block_dup_map) {
2054                 if (ctx->options & E2F_OPT_PREEN) {
2055                         clear_problem_context(&pctx);
2056                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2057                 }
2058                 e2fsck_pass1_dupblocks(ctx, block_buf);
2059         }
2060         ctx->flags |= E2F_FLAG_ALLOC_OK;
2061         ext2fs_free_mem(&inodes_to_process);
2062 endit:
2063         e2fsck_use_inode_shortcuts(ctx, 0);
2064
2065         if (scan)
2066                 ext2fs_close_inode_scan(scan);
2067         if (block_buf)
2068                 ext2fs_free_mem(&block_buf);
2069         if (inode)
2070                 ext2fs_free_mem(&inode);
2071
2072         /*
2073          * The l+f inode may have been cleared, so zap it now and
2074          * later passes will recalculate it if necessary
2075          */
2076         ctx->lost_and_found = 0;
2077
2078         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2079                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2080         else
2081                 ctx->invalid_bitmaps++;
2082 }
2083 #undef FINISH_INODE_LOOP
2084
2085 /*
2086  * When the inode_scan routines call this callback at the end of the
2087  * glock group, call process_inodes.
2088  */
2089 static errcode_t scan_callback(ext2_filsys fs,
2090                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
2091                                dgrp_t group, void * priv_data)
2092 {
2093         struct scan_callback_struct *scan_struct;
2094         e2fsck_t ctx;
2095
2096         scan_struct = (struct scan_callback_struct *) priv_data;
2097         ctx = scan_struct->ctx;
2098
2099         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf);
2100
2101         if (ctx->progress)
2102                 if ((ctx->progress)(ctx, 1, group+1,
2103                                     ctx->fs->group_desc_count))
2104                         return EXT2_ET_CANCEL_REQUESTED;
2105
2106         return 0;
2107 }
2108
2109 /*
2110  * Process the inodes in the "inodes to process" list.
2111  */
2112 static void process_inodes(e2fsck_t ctx, char *block_buf)
2113 {
2114         int                     i;
2115         struct ext2_inode       *old_stashed_inode;
2116         ext2_ino_t              old_stashed_ino;
2117         const char              *old_operation;
2118         char                    buf[80];
2119         struct problem_context  pctx;
2120
2121 #if 0
2122         printf("begin process_inodes: ");
2123 #endif
2124         if (process_inode_count == 0)
2125                 return;
2126         old_operation = ehandler_operation(0);
2127         old_stashed_inode = ctx->stashed_inode;
2128         old_stashed_ino = ctx->stashed_ino;
2129         qsort(inodes_to_process, process_inode_count,
2130                       sizeof(struct process_inode_block), process_inode_cmp);
2131         clear_problem_context(&pctx);
2132         for (i=0; i < process_inode_count; i++) {
2133                 pctx.inode = ctx->stashed_inode =
2134                         (struct ext2_inode *) &inodes_to_process[i].inode;
2135                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2136
2137 #if 0
2138                 printf("%u ", pctx.ino);
2139 #endif
2140                 sprintf(buf, _("reading indirect blocks of inode %u"),
2141                         pctx.ino);
2142                 ehandler_operation(buf);
2143                 check_blocks(ctx, &pctx, block_buf,
2144                              &inodes_to_process[i].ea_ibody_quota);
2145                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2146                         break;
2147         }
2148         ctx->stashed_inode = old_stashed_inode;
2149         ctx->stashed_ino = old_stashed_ino;
2150         process_inode_count = 0;
2151 #if 0
2152         printf("end process inodes\n");
2153 #endif
2154         ehandler_operation(old_operation);
2155 }
2156
2157 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
2158 {
2159         const struct process_inode_block *ib_a =
2160                 (const struct process_inode_block *) a;
2161         const struct process_inode_block *ib_b =
2162                 (const struct process_inode_block *) b;
2163         int     ret;
2164
2165         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
2166                ib_b->inode.i_block[EXT2_IND_BLOCK]);
2167         if (ret == 0)
2168                 /*
2169                  * We only call process_inodes() for non-extent
2170                  * inodes, so it's OK to pass NULL to
2171                  * ext2fs_file_acl_block() here.
2172                  */
2173                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
2174                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
2175         if (ret == 0)
2176                 ret = ib_a->ino - ib_b->ino;
2177         return ret;
2178 }
2179
2180 /*
2181  * Mark an inode as being bad in some what
2182  */
2183 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
2184 {
2185         struct          problem_context pctx;
2186
2187         if (!ctx->inode_bad_map) {
2188                 clear_problem_context(&pctx);
2189
2190                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2191                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
2192                                 "inode_bad_map", &ctx->inode_bad_map);
2193                 if (pctx.errcode) {
2194                         pctx.num = 3;
2195                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2196                         /* Should never get here */
2197                         ctx->flags |= E2F_FLAG_ABORT;
2198                         return;
2199                 }
2200         }
2201         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
2202 }
2203
2204 static void add_encrypted_dir(e2fsck_t ctx, ino_t ino)
2205 {
2206         struct          problem_context pctx;
2207
2208         if (!ctx->encrypted_dirs) {
2209                 pctx.errcode = ext2fs_u32_list_create(&ctx->encrypted_dirs, 0);
2210                 if (pctx.errcode)
2211                         goto error;
2212         }
2213         pctx.errcode = ext2fs_u32_list_add(ctx->encrypted_dirs, ino);
2214         if (pctx.errcode == 0)
2215                 return;
2216 error:
2217         fix_problem(ctx, PR_1_ALLOCATE_ENCRYPTED_DIRLIST, &pctx);
2218         /* Should never get here */
2219         ctx->flags |= E2F_FLAG_ABORT;
2220 }
2221
2222 /*
2223  * This procedure will allocate the inode "bb" (badblock) map table
2224  */
2225 static void alloc_bb_map(e2fsck_t ctx)
2226 {
2227         struct          problem_context pctx;
2228
2229         clear_problem_context(&pctx);
2230         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2231                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
2232                         "inode_bb_map", &ctx->inode_bb_map);
2233         if (pctx.errcode) {
2234                 pctx.num = 4;
2235                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2236                 /* Should never get here */
2237                 ctx->flags |= E2F_FLAG_ABORT;
2238                 return;
2239         }
2240 }
2241
2242 /*
2243  * This procedure will allocate the inode imagic table
2244  */
2245 static void alloc_imagic_map(e2fsck_t ctx)
2246 {
2247         struct          problem_context pctx;
2248
2249         clear_problem_context(&pctx);
2250         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2251                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
2252                         "inode_imagic_map", &ctx->inode_imagic_map);
2253         if (pctx.errcode) {
2254                 pctx.num = 5;
2255                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2256                 /* Should never get here */
2257                 ctx->flags |= E2F_FLAG_ABORT;
2258                 return;
2259         }
2260 }
2261
2262 /*
2263  * Marks a block as in use, setting the dup_map if it's been set
2264  * already.  Called by process_block and process_bad_block.
2265  *
2266  * WARNING: Assumes checks have already been done to make sure block
2267  * is valid.  This is true in both process_block and process_bad_block.
2268  */
2269 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
2270 {
2271         struct          problem_context pctx;
2272
2273         clear_problem_context(&pctx);
2274
2275         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
2276                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
2277                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
2278                         return;
2279                 }
2280                 if (!ctx->block_dup_map) {
2281                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
2282                                         _("multiply claimed block map"),
2283                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
2284                                         &ctx->block_dup_map);
2285                         if (pctx.errcode) {
2286                                 pctx.num = 3;
2287                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
2288                                             &pctx);
2289                                 /* Should never get here */
2290                                 ctx->flags |= E2F_FLAG_ABORT;
2291                                 return;
2292                         }
2293                 }
2294                 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
2295         } else {
2296                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
2297         }
2298 }
2299
2300 /*
2301  * When cluster size is greater than one block, it is caller's responsibility
2302  * to make sure block parameter starts at a cluster boundary.
2303  */
2304 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
2305                                       unsigned int num)
2306 {
2307         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
2308                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
2309         else {
2310                 unsigned int i;
2311
2312                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
2313                         mark_block_used(ctx, block + i);
2314         }
2315 }
2316
2317 /*
2318  * Adjust the extended attribute block's reference counts at the end
2319  * of pass 1, either by subtracting out references for EA blocks that
2320  * are still referenced in ctx->refcount, or by adding references for
2321  * EA blocks that had extra references as accounted for in
2322  * ctx->refcount_extra.
2323  */
2324 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
2325                                     char *block_buf, int adjust_sign)
2326 {
2327         struct ext2_ext_attr_header     *header;
2328         struct problem_context          pctx;
2329         ext2_filsys                     fs = ctx->fs;
2330         blk64_t                         blk;
2331         __u32                           should_be;
2332         ea_value_t                      count;
2333
2334         clear_problem_context(&pctx);
2335
2336         ea_refcount_intr_begin(refcount);
2337         while (1) {
2338                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
2339                         break;
2340                 pctx.blk = blk;
2341                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
2342                                                      pctx.ino);
2343                 if (pctx.errcode) {
2344                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
2345                         return;
2346                 }
2347                 header = (struct ext2_ext_attr_header *) block_buf;
2348                 pctx.blkcount = header->h_refcount;
2349                 should_be = header->h_refcount + adjust_sign * (int)count;
2350                 pctx.num = should_be;
2351                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
2352                         header->h_refcount = should_be;
2353                         pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
2354                                                              block_buf,
2355                                                              pctx.ino);
2356                         if (pctx.errcode) {
2357                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
2358                                             &pctx);
2359                                 continue;
2360                         }
2361                 }
2362         }
2363 }
2364
2365 /*
2366  * Handle processing the extended attribute blocks
2367  */
2368 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
2369                            char *block_buf, struct ea_quota *ea_block_quota)
2370 {
2371         ext2_filsys fs = ctx->fs;
2372         ext2_ino_t      ino = pctx->ino;
2373         struct ext2_inode *inode = pctx->inode;
2374         blk64_t         blk;
2375         char *          end;
2376         struct ext2_ext_attr_header *header;
2377         struct ext2_ext_attr_entry *first, *entry;
2378         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
2379         __u64           quota_inodes = 0;
2380         region_t        region = 0;
2381         int             failed_csum = 0;
2382
2383         ea_block_quota->blocks = 0;
2384         ea_block_quota->inodes = 0;
2385
2386         blk = ext2fs_file_acl_block(fs, inode);
2387         if (blk == 0)
2388                 return 0;
2389
2390         /*
2391          * If the Extended attribute flag isn't set, then a non-zero
2392          * file acl means that the inode is corrupted.
2393          *
2394          * Or if the extended attribute block is an invalid block,
2395          * then the inode is also corrupted.
2396          */
2397         if (!ext2fs_has_feature_xattr(fs->super) ||
2398             (blk < fs->super->s_first_data_block) ||
2399             (blk >= ext2fs_blocks_count(fs->super))) {
2400                 mark_inode_bad(ctx, ino);
2401                 return 0;
2402         }
2403
2404         /* If ea bitmap hasn't been allocated, create it */
2405         if (!ctx->block_ea_map) {
2406                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
2407                                         _("ext attr block map"),
2408                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
2409                                         &ctx->block_ea_map);
2410                 if (pctx->errcode) {
2411                         pctx->num = 2;
2412                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
2413                         ctx->flags |= E2F_FLAG_ABORT;
2414                         return 0;
2415                 }
2416         }
2417
2418         /* Create the EA refcount structure if necessary */
2419         if (!ctx->refcount) {
2420                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
2421                 if (pctx->errcode) {
2422                         pctx->num = 1;
2423                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2424                         ctx->flags |= E2F_FLAG_ABORT;
2425                         return 0;
2426                 }
2427         }
2428
2429 #if 0
2430         /* Debugging text */
2431         printf("Inode %u has EA block %u\n", ino, blk);
2432 #endif
2433
2434         /* Have we seen this EA block before? */
2435         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
2436                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
2437                 ea_block_quota->inodes = 0;
2438
2439                 if (ctx->ea_block_quota_blocks) {
2440                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
2441                                           &quota_blocks);
2442                         if (quota_blocks)
2443                                 ea_block_quota->blocks = quota_blocks;
2444                 }
2445
2446                 if (ctx->ea_block_quota_inodes)
2447                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
2448                                           &ea_block_quota->inodes);
2449
2450                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
2451                         return 1;
2452                 /* Ooops, this EA was referenced more than it stated */
2453                 if (!ctx->refcount_extra) {
2454                         pctx->errcode = ea_refcount_create(0,
2455                                            &ctx->refcount_extra);
2456                         if (pctx->errcode) {
2457                                 pctx->num = 2;
2458                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2459                                 ctx->flags |= E2F_FLAG_ABORT;
2460                                 return 0;
2461                         }
2462                 }
2463                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
2464                 return 1;
2465         }
2466
2467         /*
2468          * OK, we haven't seen this EA block yet.  So we need to
2469          * validate it
2470          */
2471         pctx->blk = blk;
2472         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
2473         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
2474                 pctx->errcode = 0;
2475                 failed_csum = 1;
2476         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
2477                 pctx->errcode = 0;
2478
2479         if (pctx->errcode &&
2480             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
2481                 pctx->errcode = 0;
2482                 goto clear_extattr;
2483         }
2484         header = (struct ext2_ext_attr_header *) block_buf;
2485         pctx->blk = ext2fs_file_acl_block(fs, inode);
2486         if (((ctx->ext_attr_ver == 1) &&
2487              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
2488             ((ctx->ext_attr_ver == 2) &&
2489              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
2490                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
2491                         goto clear_extattr;
2492         }
2493
2494         if (header->h_blocks != 1) {
2495                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
2496                         goto clear_extattr;
2497         }
2498
2499         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
2500                 goto clear_extattr;
2501
2502         region = region_create(0, fs->blocksize);
2503         if (!region) {
2504                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
2505                 ctx->flags |= E2F_FLAG_ABORT;
2506                 return 0;
2507         }
2508         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
2509                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2510                         goto clear_extattr;
2511         }
2512
2513         first = (struct ext2_ext_attr_entry *)(header+1);
2514         end = block_buf + fs->blocksize;
2515         entry = first;
2516         while ((char *)entry < end && *(__u32 *)entry) {
2517                 __u32 hash;
2518
2519                 if (region_allocate(region, (char *)entry - (char *)header,
2520                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
2521                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2522                                 goto clear_extattr;
2523                         break;
2524                 }
2525                 if ((ctx->ext_attr_ver == 1 &&
2526                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
2527                     (ctx->ext_attr_ver == 2 &&
2528                      entry->e_name_index == 0)) {
2529                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
2530                                 goto clear_extattr;
2531                         break;
2532                 }
2533                 if (entry->e_value_inum == 0) {
2534                         if (entry->e_value_offs + entry->e_value_size >
2535                             fs->blocksize) {
2536                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
2537                                         goto clear_extattr;
2538                                 break;
2539                         }
2540                         if (entry->e_value_size &&
2541                             region_allocate(region, entry->e_value_offs,
2542                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
2543                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
2544                                                 pctx))
2545                                         goto clear_extattr;
2546                         }
2547
2548                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
2549                                                           entry->e_value_offs);
2550
2551                         if (entry->e_hash != hash) {
2552                                 pctx->num = entry->e_hash;
2553                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
2554                                         goto clear_extattr;
2555                                 entry->e_hash = hash;
2556                         }
2557                 } else {
2558                         problem_t problem;
2559                         blk64_t entry_quota_blocks;
2560
2561                         problem = check_large_ea_inode(ctx, entry, pctx,
2562                                                        &entry_quota_blocks);
2563                         if (problem && fix_problem(ctx, problem, pctx))
2564                                 goto clear_extattr;
2565
2566                         quota_blocks += entry_quota_blocks;
2567                         quota_inodes++;
2568                 }
2569
2570                 entry = EXT2_EXT_ATTR_NEXT(entry);
2571         }
2572         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
2573                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2574                         goto clear_extattr;
2575         }
2576         region_free(region);
2577
2578         /*
2579          * We only get here if there was no other errors that were fixed.
2580          * If there was a checksum fail, ask to correct it.
2581          */
2582         if (failed_csum &&
2583             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
2584                 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
2585                                                        pctx->ino);
2586                 if (pctx->errcode)
2587                         return 0;
2588         }
2589
2590         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
2591                 if (!ctx->ea_block_quota_blocks) {
2592                         pctx->errcode = ea_refcount_create(0,
2593                                                 &ctx->ea_block_quota_blocks);
2594                         if (pctx->errcode) {
2595                                 pctx->num = 3;
2596                                 goto refcount_fail;
2597                         }
2598                 }
2599                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
2600                                   quota_blocks);
2601         }
2602
2603         if (quota_inodes) {
2604                 if (!ctx->ea_block_quota_inodes) {
2605                         pctx->errcode = ea_refcount_create(0,
2606                                                 &ctx->ea_block_quota_inodes);
2607                         if (pctx->errcode) {
2608                                 pctx->num = 4;
2609 refcount_fail:
2610                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2611                                 ctx->flags |= E2F_FLAG_ABORT;
2612                                 return 0;
2613                         }
2614                 }
2615
2616                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
2617                                   quota_inodes);
2618         }
2619         ea_block_quota->blocks = quota_blocks;
2620         ea_block_quota->inodes = quota_inodes;
2621
2622         inc_ea_inode_refs(ctx, pctx, first, end);
2623         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
2624         mark_block_used(ctx, blk);
2625         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
2626         return 1;
2627
2628 clear_extattr:
2629         if (region)
2630                 region_free(region);
2631         ext2fs_file_acl_block_set(fs, inode, 0);
2632         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
2633         return 0;
2634 }
2635
2636 /* Returns 1 if bad htree, 0 if OK */
2637 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
2638                         ext2_ino_t ino, struct ext2_inode *inode,
2639                         char *block_buf)
2640 {
2641         struct ext2_dx_root_info        *root;
2642         ext2_filsys                     fs = ctx->fs;
2643         errcode_t                       retval;
2644         blk64_t                         blk;
2645
2646         if ((!LINUX_S_ISDIR(inode->i_mode) &&
2647              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
2648             (!ext2fs_has_feature_dir_index(fs->super) &&
2649              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
2650                 return 1;
2651
2652         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
2653
2654         if ((pctx->errcode) ||
2655             (blk == 0) ||
2656             (blk < fs->super->s_first_data_block) ||
2657             (blk >= ext2fs_blocks_count(fs->super))) {
2658                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2659                         return 1;
2660                 else
2661                         return 0;
2662         }
2663
2664         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
2665         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2666                 return 1;
2667
2668         /* XXX should check that beginning matches a directory */
2669         root = (struct ext2_dx_root_info *) (block_buf + 24);
2670
2671         if ((root->reserved_zero || root->info_length < 8) &&
2672             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2673                 return 1;
2674
2675         pctx->num = root->hash_version;
2676         if ((root->hash_version != EXT2_HASH_LEGACY) &&
2677             (root->hash_version != EXT2_HASH_HALF_MD4) &&
2678             (root->hash_version != EXT2_HASH_TEA) &&
2679             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
2680                 return 1;
2681
2682         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
2683             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
2684                 return 1;
2685
2686         pctx->num = root->indirect_levels;
2687         if ((root->indirect_levels > ext2_dir_htree_level(fs)) &&
2688             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
2689                 return 1;
2690
2691         return 0;
2692 }
2693
2694 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
2695                         struct ext2_inode *inode, int restart_flag,
2696                         const char *source)
2697 {
2698         inode->i_flags = 0;
2699         inode->i_links_count = 0;
2700         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
2701         inode->i_dtime = ctx->now;
2702
2703         /*
2704          * If a special inode has such rotten block mappings that we
2705          * want to clear the whole inode, be sure to actually zap
2706          * the block maps because i_links_count isn't checked for
2707          * special inodes, and we'll end up right back here the next
2708          * time we run fsck.
2709          */
2710         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
2711                 memset(inode->i_block, 0, sizeof(inode->i_block));
2712
2713         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
2714         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
2715         if (ctx->inode_reg_map)
2716                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
2717         if (ctx->inode_bad_map)
2718                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
2719
2720         /*
2721          * If the inode was partially accounted for before processing
2722          * was aborted, we need to restart the pass 1 scan.
2723          */
2724         ctx->flags |= restart_flag;
2725
2726         if (ino == EXT2_BAD_INO)
2727                 memset(inode, 0, sizeof(struct ext2_inode));
2728
2729         e2fsck_write_inode(ctx, ino, inode, source);
2730 }
2731
2732 /*
2733  * Use the multiple-blocks reclamation code to fix alignment problems in
2734  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
2735  * physical cluster, and we want the block offsets within that cluster to
2736  * line up.
2737  */
2738 static int has_unaligned_cluster_map(e2fsck_t ctx,
2739                                      blk64_t last_pblk, blk64_t last_lblk,
2740                                      blk64_t pblk, blk64_t lblk)
2741 {
2742         blk64_t cluster_mask;
2743
2744         if (!ctx->fs->cluster_ratio_bits)
2745                 return 0;
2746         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
2747
2748         /*
2749          * If the block in the logical cluster doesn't align with the block in
2750          * the physical cluster...
2751          */
2752         if ((lblk & cluster_mask) != (pblk & cluster_mask))
2753                 return 1;
2754
2755         /*
2756          * If we cross a physical cluster boundary within a logical cluster...
2757          */
2758         if (last_pblk && (lblk & cluster_mask) != 0 &&
2759             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
2760             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
2761                 return 1;
2762
2763         return 0;
2764 }
2765
2766 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
2767                              struct process_block_struct *pb,
2768                              blk64_t start_block, blk64_t end_block,
2769                              blk64_t eof_block,
2770                              ext2_extent_handle_t ehandle,
2771                              int try_repairs)
2772 {
2773         struct ext2fs_extent    extent;
2774         blk64_t                 blk, last_lblk;
2775         unsigned int            i, n;
2776         int                     is_dir, is_leaf;
2777         problem_t               problem;
2778         struct ext2_extent_info info;
2779         int                     failed_csum = 0;
2780
2781         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
2782                 failed_csum = 1;
2783
2784         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
2785         if (pctx->errcode)
2786                 return;
2787         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
2788             !pb->eti.force_rebuild) {
2789                 struct extent_tree_level *etl;
2790
2791                 etl = pb->eti.ext_info + info.curr_level;
2792                 etl->num_extents += info.num_entries;
2793                 etl->max_extents += info.max_entries;
2794                 /*
2795                  * Implementation wart: Splitting extent blocks when appending
2796                  * will leave the old block with one free entry.  Therefore
2797                  * unless the node is totally full, pretend that a non-root
2798                  * extent block can hold one fewer entry than it actually does,
2799                  * so that we don't repeatedly rebuild the extent tree.
2800                  */
2801                 if (info.curr_level && info.num_entries < info.max_entries)
2802                         etl->max_extents--;
2803         }
2804
2805         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
2806                                           &extent);
2807         while ((pctx->errcode == 0 ||
2808                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
2809                info.num_entries-- > 0) {
2810                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
2811                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
2812                 last_lblk = extent.e_lblk + extent.e_len - 1;
2813
2814                 problem = 0;
2815                 pctx->blk = extent.e_pblk;
2816                 pctx->blk2 = extent.e_lblk;
2817                 pctx->num = extent.e_len;
2818                 pctx->blkcount = extent.e_lblk + extent.e_len;
2819
2820                 if (extent.e_pblk == 0 ||
2821                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
2822                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
2823                         problem = PR_1_EXTENT_BAD_START_BLK;
2824                 else if (extent.e_lblk < start_block)
2825                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
2826                 else if ((end_block && last_lblk > end_block) &&
2827                          !(last_lblk > eof_block &&
2828                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
2829                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
2830                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
2831                 else if (is_leaf && extent.e_len == 0)
2832                         problem = PR_1_EXTENT_LENGTH_ZERO;
2833                 else if (is_leaf &&
2834                          (extent.e_pblk + extent.e_len) >
2835                          ext2fs_blocks_count(ctx->fs->super))
2836                         problem = PR_1_EXTENT_ENDS_BEYOND;
2837                 else if (is_leaf && is_dir &&
2838                          ((extent.e_lblk + extent.e_len) >
2839                           (1U << (21 - ctx->fs->super->s_log_block_size))))
2840                         problem = PR_1_TOOBIG_DIR;
2841
2842                 if (is_leaf && problem == 0 && extent.e_len > 0) {
2843 #if 0
2844                         printf("extent_region(ino=%u, expect=%llu, "
2845                                "lblk=%llu, len=%u)\n",
2846                                pb->ino, pb->next_lblock,
2847                                extent.e_lblk, extent.e_len);
2848 #endif
2849                         if (extent.e_lblk < pb->next_lblock)
2850                                 problem = PR_1_EXTENT_COLLISION;
2851                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
2852                                 pb->next_lblock = extent.e_lblk + extent.e_len;
2853                 }
2854
2855                 /*
2856                  * Uninitialized blocks in a directory?  Clear the flag and
2857                  * we'll interpret the blocks later.
2858                  */
2859                 if (try_repairs && is_dir && problem == 0 &&
2860                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
2861                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
2862                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
2863                         pb->inode_modified = 1;
2864                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
2865                                                               &extent);
2866                         if (pctx->errcode)
2867                                 return;
2868                         failed_csum = 0;
2869                 }
2870 #ifdef CONFIG_DEVELOPER_FEATURES
2871                 if (try_repairs && !is_dir && problem == 0 &&
2872                     (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
2873                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
2874                     fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
2875                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
2876                         pb->inode_modified = 1;
2877                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
2878                                                               &extent);
2879                         if (pctx->errcode)
2880                                 return;
2881                         failed_csum = 0;
2882                 }
2883 #endif
2884                 if (try_repairs && problem) {
2885 report_problem:
2886                         if (fix_problem(ctx, problem, pctx)) {
2887                                 if (ctx->invalid_bitmaps) {
2888                                         /*
2889                                          * If fsck knows the bitmaps are bad,
2890                                          * skip to the next extent and
2891                                          * try to clear this extent again
2892                                          * after fixing the bitmaps, by
2893                                          * restarting fsck.
2894                                          */
2895                                         pctx->errcode = ext2fs_extent_get(
2896                                                           ehandle,
2897                                                           EXT2_EXTENT_NEXT_SIB,
2898                                                           &extent);
2899                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
2900                                         if (pctx->errcode ==
2901                                                     EXT2_ET_NO_CURRENT_NODE) {
2902                                                 pctx->errcode = 0;
2903                                                 break;
2904                                         }
2905                                         continue;
2906                                 }
2907                                 e2fsck_read_bitmaps(ctx);
2908                                 pb->inode_modified = 1;
2909                                 pctx->errcode =
2910                                         ext2fs_extent_delete(ehandle, 0);
2911                                 if (pctx->errcode) {
2912                                         pctx->str = "ext2fs_extent_delete";
2913                                         return;
2914                                 }
2915                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
2916                                 if (pctx->errcode &&
2917                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
2918                                         pctx->str = "ext2fs_extent_fix_parents";
2919                                         return;
2920                                 }
2921                                 pctx->errcode = ext2fs_extent_get(ehandle,
2922                                                                   EXT2_EXTENT_CURRENT,
2923                                                                   &extent);
2924                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
2925                                         pctx->errcode = 0;
2926                                         break;
2927                                 }
2928                                 failed_csum = 0;
2929                                 continue;
2930                         }
2931                         goto next;
2932                 }
2933
2934                 if (!is_leaf) {
2935                         blk64_t lblk = extent.e_lblk;
2936                         int next_try_repairs = 1;
2937
2938                         blk = extent.e_pblk;
2939
2940                         /*
2941                          * If this lower extent block collides with critical
2942                          * metadata, don't try to repair the damage.  Pass 1b
2943                          * will reallocate the block; then we can try again.
2944                          */
2945                         if (pb->ino != EXT2_RESIZE_INO &&
2946                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
2947                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
2948                                                       extent.e_pblk)) {
2949                                 next_try_repairs = 0;
2950                                 pctx->blk = blk;
2951                                 fix_problem(ctx,
2952                                             PR_1_CRITICAL_METADATA_COLLISION,
2953                                             pctx);
2954                                 if ((ctx->options & E2F_OPT_NO) == 0)
2955                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
2956                         }
2957                         pctx->errcode = ext2fs_extent_get(ehandle,
2958                                                   EXT2_EXTENT_DOWN, &extent);
2959                         if (pctx->errcode &&
2960                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
2961                                 pctx->str = "EXT2_EXTENT_DOWN";
2962                                 problem = PR_1_EXTENT_HEADER_INVALID;
2963                                 if (!next_try_repairs)
2964                                         return;
2965                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
2966                                         goto report_problem;
2967                                 return;
2968                         }
2969                         /* The next extent should match this index's logical start */
2970                         if (extent.e_lblk != lblk) {
2971                                 struct ext2_extent_info e_info;
2972
2973                                 ext2fs_extent_get_info(ehandle, &e_info);
2974                                 pctx->blk = lblk;
2975                                 pctx->blk2 = extent.e_lblk;
2976                                 pctx->num = e_info.curr_level - 1;
2977                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
2978                                 if (fix_problem(ctx, problem, pctx)) {
2979                                         pb->inode_modified = 1;
2980                                         pctx->errcode =
2981                                                 ext2fs_extent_fix_parents(ehandle);
2982                                         if (pctx->errcode) {
2983                                                 pctx->str = "ext2fs_extent_fix_parents";
2984                                                 return;
2985                                         }
2986                                 }
2987                         }
2988                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
2989                                          last_lblk, eof_block, ehandle,
2990                                          next_try_repairs);
2991                         if (pctx->errcode)
2992                                 return;
2993                         pctx->errcode = ext2fs_extent_get(ehandle,
2994                                                   EXT2_EXTENT_UP, &extent);
2995                         if (pctx->errcode) {
2996                                 pctx->str = "EXT2_EXTENT_UP";
2997                                 return;
2998                         }
2999                         mark_block_used(ctx, blk);
3000                         pb->num_blocks++;
3001                         goto next;
3002                 }
3003
3004                 if ((pb->previous_block != 0) &&
3005                     (pb->previous_block+1 != extent.e_pblk)) {
3006                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3007                                 char type = '?';
3008
3009                                 if (pb->is_dir)
3010                                         type = 'd';
3011                                 else if (pb->is_reg)
3012                                         type = 'f';
3013
3014                                 printf(("%6lu(%c): expecting %6lu "
3015                                         "actual extent "
3016                                         "phys %6lu log %lu len %lu\n"),
3017                                        (unsigned long) pctx->ino, type,
3018                                        (unsigned long) pb->previous_block+1,
3019                                        (unsigned long) extent.e_pblk,
3020                                        (unsigned long) extent.e_lblk,
3021                                        (unsigned long) extent.e_len);
3022                         }
3023                         pb->fragmented = 1;
3024                 }
3025                 /*
3026                  * If we notice a gap in the logical block mappings of an
3027                  * extent-mapped directory, offer to close the hole by
3028                  * moving the logical block down, otherwise we'll go mad in
3029                  * pass 3 allocating empty directory blocks to fill the hole.
3030                  */
3031                 if (try_repairs && is_dir &&
3032                     pb->last_block + 1 < extent.e_lblk) {
3033                         blk64_t new_lblk;
3034
3035                         new_lblk = pb->last_block + 1;
3036                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3037                                 new_lblk = ((new_lblk +
3038                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3039                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3040                                            (extent.e_pblk &
3041                                             EXT2FS_CLUSTER_MASK(ctx->fs));
3042                         pctx->blk = extent.e_lblk;
3043                         pctx->blk2 = new_lblk;
3044                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3045                                 extent.e_lblk = new_lblk;
3046                                 pb->inode_modified = 1;
3047                                 pctx->errcode = ext2fs_extent_replace(ehandle,
3048                                                                 0, &extent);
3049                                 if (pctx->errcode) {
3050                                         pctx->errcode = 0;
3051                                         goto alloc_later;
3052                                 }
3053                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3054                                 if (pctx->errcode)
3055                                         goto failed_add_dir_block;
3056                                 pctx->errcode = ext2fs_extent_goto(ehandle,
3057                                                                 extent.e_lblk);
3058                                 if (pctx->errcode)
3059                                         goto failed_add_dir_block;
3060                                 last_lblk = extent.e_lblk + extent.e_len - 1;
3061                                 failed_csum = 0;
3062                         }
3063                 }
3064 alloc_later:
3065                 if (is_dir) {
3066                         while (++pb->last_db_block <
3067                                (e2_blkcnt_t) extent.e_lblk) {
3068                                 pctx->errcode = ext2fs_add_dir_block2(
3069                                                         ctx->fs->dblist,
3070                                                         pb->ino, 0,
3071                                                         pb->last_db_block);
3072                                 if (pctx->errcode) {
3073                                         pctx->blk = 0;
3074                                         pctx->num = pb->last_db_block;
3075                                         goto failed_add_dir_block;
3076                                 }
3077                         }
3078
3079                         for (i = 0; i < extent.e_len; i++) {
3080                                 pctx->errcode = ext2fs_add_dir_block2(
3081                                                         ctx->fs->dblist,
3082                                                         pctx->ino,
3083                                                         extent.e_pblk + i,
3084                                                         extent.e_lblk + i);
3085                                 if (pctx->errcode) {
3086                                         pctx->blk = extent.e_pblk + i;
3087                                         pctx->num = extent.e_lblk + i;
3088                                 failed_add_dir_block:
3089                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3090                                         /* Should never get here */
3091                                         ctx->flags |= E2F_FLAG_ABORT;
3092                                         return;
3093                                 }
3094                         }
3095                         if (extent.e_len > 0)
3096                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3097                 }
3098                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3099                                               pb->last_block,
3100                                               extent.e_pblk,
3101                                               extent.e_lblk)) {
3102                         for (i = 0; i < extent.e_len; i++) {
3103                                 pctx->blk = extent.e_lblk + i;
3104                                 pctx->blk2 = extent.e_pblk + i;
3105                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3106                                 mark_block_used(ctx, extent.e_pblk + i);
3107                                 mark_block_used(ctx, extent.e_pblk + i);
3108                         }
3109                 }
3110
3111                 /*
3112                  * Check whether first cluster got marked in previous iteration.
3113                  */
3114                 if (ctx->fs->cluster_ratio_bits &&
3115                     pb->previous_block &&
3116                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
3117                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
3118                         /* Set blk to the beginning of next cluster. */
3119                         blk = EXT2FS_C2B(
3120                                 ctx->fs,
3121                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
3122                 else
3123                         /* Set blk to the beginning of current cluster. */
3124                         blk = EXT2FS_C2B(ctx->fs,
3125                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
3126
3127                 if (blk < extent.e_pblk + extent.e_len) {
3128                         mark_blocks_used(ctx, blk,
3129                                          extent.e_pblk + extent.e_len - blk);
3130                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
3131                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
3132                         pb->num_blocks += n;
3133                 }
3134                 pb->last_block = extent.e_lblk + extent.e_len - 1;
3135                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
3136                 start_block = pb->last_block = last_lblk;
3137                 if (is_leaf && !is_dir &&
3138                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
3139                         pb->last_init_lblock = last_lblk;
3140         next:
3141                 pctx->errcode = ext2fs_extent_get(ehandle,
3142                                                   EXT2_EXTENT_NEXT_SIB,
3143                                                   &extent);
3144         }
3145
3146         /* Failed csum but passes checks?  Ask to fix checksum. */
3147         if (failed_csum &&
3148             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
3149                 pb->inode_modified = 1;
3150                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
3151                 if (pctx->errcode)
3152                         return;
3153         }
3154
3155         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
3156                 pctx->errcode = 0;
3157 }
3158
3159 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
3160                                  struct process_block_struct *pb)
3161 {
3162         struct ext2_extent_info info;
3163         struct ext2_inode       *inode = pctx->inode;
3164         ext2_extent_handle_t    ehandle;
3165         ext2_filsys             fs = ctx->fs;
3166         ext2_ino_t              ino = pctx->ino;
3167         errcode_t               retval;
3168         blk64_t                 eof_lblk;
3169         struct ext3_extent_header       *eh;
3170
3171         /* Check for a proper extent header... */
3172         eh = (struct ext3_extent_header *) &inode->i_block[0];
3173         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
3174         if (retval) {
3175                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
3176                         e2fsck_clear_inode(ctx, ino, inode, 0,
3177                                            "check_blocks_extents");
3178                 pctx->errcode = 0;
3179                 return;
3180         }
3181
3182         /* ...since this function doesn't fail if i_block is zeroed. */
3183         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
3184         if (pctx->errcode) {
3185                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
3186                         e2fsck_clear_inode(ctx, ino, inode, 0,
3187                                            "check_blocks_extents");
3188                 pctx->errcode = 0;
3189                 return;
3190         }
3191
3192         retval = ext2fs_extent_get_info(ehandle, &info);
3193         if (retval == 0) {
3194                 int max_depth = info.max_depth;
3195
3196                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
3197                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
3198                 ctx->extent_depth_count[max_depth]++;
3199         }
3200
3201         /* Check maximum extent depth */
3202         pctx->blk = info.max_depth;
3203         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
3204         if (pctx->blk2 < pctx->blk &&
3205             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
3206                 pb->eti.force_rebuild = 1;
3207
3208         /* Can we collect extent tree level stats? */
3209         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
3210         if (pctx->blk2 > pctx->blk)
3211                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
3212         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
3213         pb->eti.ino = pb->ino;
3214
3215         pb->next_lblock = 0;
3216
3217         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
3218                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
3219         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
3220         if (pctx->errcode &&
3221             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
3222                 pb->num_blocks = 0;
3223                 inode->i_blocks = 0;
3224                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3225                                    "check_blocks_extents");
3226                 pctx->errcode = 0;
3227         }
3228         ext2fs_extent_free(ehandle);
3229
3230         /* Rebuild unless it's a dir and we're rehashing it */
3231         if (LINUX_S_ISDIR(inode->i_mode) &&
3232             e2fsck_dir_will_be_rehashed(ctx, ino))
3233                 return;
3234
3235         if (ctx->options & E2F_OPT_CONVERT_BMAP)
3236                 e2fsck_rebuild_extents_later(ctx, ino);
3237         else
3238                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
3239 }
3240
3241 /*
3242  * In fact we don't need to check blocks for an inode with inline data
3243  * because this inode doesn't have any blocks.  In this function all
3244  * we need to do is add this inode into dblist when it is a directory.
3245  */
3246 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
3247                                      struct process_block_struct *pb)
3248 {
3249         int     flags;
3250         size_t  inline_data_size = 0;
3251
3252         if (!pb->is_dir) {
3253                 pctx->errcode = 0;
3254                 return;
3255         }
3256
3257         /* Process the dirents in i_block[] as the "first" block. */
3258         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
3259         if (pctx->errcode)
3260                 goto err;
3261
3262         /* Process the dirents in the EA as a "second" block. */
3263         flags = ctx->fs->flags;
3264         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3265         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
3266                                                 &inline_data_size);
3267         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3268                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3269         if (pctx->errcode) {
3270                 pctx->errcode = 0;
3271                 return;
3272         }
3273
3274         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
3275                 return;
3276
3277         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
3278         if (pctx->errcode)
3279                 goto err;
3280
3281         return;
3282 err:
3283         pctx->blk = 0;
3284         pctx->num = 0;
3285         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3286         ctx->flags |= E2F_FLAG_ABORT;
3287 }
3288
3289 /*
3290  * This subroutine is called on each inode to account for all of the
3291  * blocks used by that inode.
3292  */
3293 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
3294                          char *block_buf, const struct ea_quota *ea_ibody_quota)
3295 {
3296         ext2_filsys fs = ctx->fs;
3297         struct process_block_struct pb;
3298         ext2_ino_t      ino = pctx->ino;
3299         struct ext2_inode *inode = pctx->inode;
3300         unsigned        bad_size = 0;
3301         int             dirty_inode = 0;
3302         int             extent_fs;
3303         int             inlinedata_fs;
3304         __u64           size;
3305         struct ea_quota ea_block_quota;
3306
3307         pb.ino = ino;
3308         pb.num_blocks = EXT2FS_B2C(ctx->fs,
3309                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
3310         pb.last_block = ~0;
3311         pb.last_init_lblock = -1;
3312         pb.last_db_block = -1;
3313         pb.num_illegal_blocks = 0;
3314         pb.suppress = 0; pb.clear = 0;
3315         pb.fragmented = 0;
3316         pb.compressed = 0;
3317         pb.previous_block = 0;
3318         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
3319         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
3320         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
3321         pb.inode = inode;
3322         pb.pctx = pctx;
3323         pb.ctx = ctx;
3324         pb.inode_modified = 0;
3325         pb.eti.force_rebuild = 0;
3326         pctx->ino = ino;
3327         pctx->errcode = 0;
3328
3329         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
3330         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
3331
3332         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
3333                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3334                         goto out;
3335                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
3336         }
3337
3338         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
3339                 check_blocks_inline_data(ctx, pctx, &pb);
3340         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
3341                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
3342                         check_blocks_extents(ctx, pctx, &pb);
3343                 else {
3344                         int flags;
3345                         /*
3346                          * If we've modified the inode, write it out before
3347                          * iterate() tries to use it.
3348                          */
3349                         if (dirty_inode) {
3350                                 e2fsck_write_inode(ctx, ino, inode,
3351                                                    "check_blocks");
3352                                 dirty_inode = 0;
3353                         }
3354                         flags = fs->flags;
3355                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3356                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
3357                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
3358                                                 block_buf, process_block, &pb);
3359                         /*
3360                          * We do not have uninitialized extents in non extent
3361                          * files.
3362                          */
3363                         pb.last_init_lblock = pb.last_block;
3364                         /*
3365                          * If iterate() changed a block mapping, we have to
3366                          * re-read the inode.  If we decide to clear the
3367                          * inode after clearing some stuff, we'll re-write the
3368                          * bad mappings into the inode!
3369                          */
3370                         if (pb.inode_modified)
3371                                 e2fsck_read_inode(ctx, ino, inode,
3372                                                   "check_blocks");
3373                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3374                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3375
3376                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
3377 #ifdef DEBUG
3378                                 printf("bmap rebuild ino=%d\n", ino);
3379 #endif
3380                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
3381                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
3382                                         e2fsck_rebuild_extents_later(ctx, ino);
3383                         }
3384                 }
3385         }
3386         end_problem_latch(ctx, PR_LATCH_BLOCK);
3387         end_problem_latch(ctx, PR_LATCH_TOOBIG);
3388         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3389                 goto out;
3390         if (pctx->errcode)
3391                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
3392
3393         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
3394                 if (LINUX_S_ISDIR(inode->i_mode))
3395                         ctx->fs_fragmented_dir++;
3396                 else
3397                         ctx->fs_fragmented++;
3398         }
3399
3400         if (pb.clear) {
3401                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3402                                    "check_blocks");
3403                 return;
3404         }
3405
3406         if (inode->i_flags & EXT2_INDEX_FL) {
3407                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
3408                         inode->i_flags &= ~EXT2_INDEX_FL;
3409                         dirty_inode++;
3410                 } else {
3411                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
3412                 }
3413         }
3414
3415         if (!pb.num_blocks && pb.is_dir &&
3416             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
3417                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
3418                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
3419                         ctx->fs_directory_count--;
3420                         return;
3421                 }
3422         }
3423
3424         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
3425             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
3426             !(inode->i_flags & EXT4_EA_INODE_FL)) {
3427                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
3428                                ino,
3429                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
3430                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
3431                                   ino, (ea_ibody_quota ?
3432                                         ea_ibody_quota->inodes : 0) +
3433                                                 ea_block_quota.inodes + 1);
3434         }
3435
3436         if (!ext2fs_has_feature_huge_file(fs->super) ||
3437             !(inode->i_flags & EXT4_HUGE_FILE_FL))
3438                 pb.num_blocks *= (fs->blocksize / 512);
3439         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
3440 #if 0
3441         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
3442                ino, inode->i_size, pb.last_block, ext2fs_inode_i_blocks(fs, inode),
3443                pb.num_blocks);
3444 #endif
3445         if (pb.is_dir) {
3446                 unsigned nblock = inode->i_size >> EXT2_BLOCK_SIZE_BITS(fs->super);
3447                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
3448                         int flags;
3449                         size_t sz = 0;
3450                         errcode_t err;
3451
3452                         flags = ctx->fs->flags;
3453                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3454                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
3455                                                       &sz);
3456                         ctx->fs->flags = (flags &
3457                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3458                                          (ctx->fs->flags &
3459                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3460                         if (err || sz != inode->i_size) {
3461                                 bad_size = 7;
3462                                 pctx->num = sz;
3463                         }
3464                 } else if (inode->i_size & (fs->blocksize - 1))
3465                         bad_size = 5;
3466                 else if (nblock > (pb.last_block + 1))
3467                         bad_size = 1;
3468                 else if (nblock < (pb.last_block + 1)) {
3469                         if (((pb.last_block + 1) - nblock) >
3470                             fs->super->s_prealloc_dir_blocks)
3471                                 bad_size = 2;
3472                 }
3473         } else {
3474                 size = EXT2_I_SIZE(inode);
3475                 if ((pb.last_init_lblock >= 0) &&
3476                     /* Do not allow initialized allocated blocks past i_size*/
3477                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
3478                     !(inode->i_flags & EXT4_VERITY_FL))
3479                         bad_size = 3;
3480                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3481                          size > ext2_max_sizes[fs->super->s_log_block_size])
3482                         /* too big for a direct/indirect-mapped file */
3483                         bad_size = 4;
3484                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3485                          size >
3486                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
3487                         /* too big for an extent-based file - 32bit ee_block */
3488                         bad_size = 6;
3489         }
3490         /* i_size for symlinks is checked elsewhere */
3491         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
3492                 /* Did inline_data set pctx->num earlier? */
3493                 if (bad_size != 7)
3494                         pctx->num = (pb.last_block + 1) * fs->blocksize;
3495                 pctx->group = bad_size;
3496                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
3497                         if (LINUX_S_ISDIR(inode->i_mode))
3498                                 pctx->num &= 0xFFFFFFFFULL;
3499                         ext2fs_inode_size_set(fs, inode, pctx->num);
3500                         if (EXT2_I_SIZE(inode) == 0 &&
3501                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
3502                                 memset(inode->i_block, 0,
3503                                        sizeof(inode->i_block));
3504                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
3505                         }
3506                         dirty_inode++;
3507                 }
3508                 pctx->num = 0;
3509         }
3510         if (LINUX_S_ISREG(inode->i_mode) &&
3511             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3512                 ctx->large_files++;
3513         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
3514             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
3515              (ext2fs_has_feature_huge_file(fs->super) &&
3516               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
3517               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
3518                 pctx->num = pb.num_blocks;
3519                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
3520                         inode->i_blocks = pb.num_blocks;
3521                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
3522                         dirty_inode++;
3523                 }
3524                 pctx->num = 0;
3525         }
3526
3527         /*
3528          * The kernel gets mad if we ask it to allocate bigalloc clusters to
3529          * a block mapped file, so rebuild it as an extent file.  We can skip
3530          * symlinks because they're never rewritten.
3531          */
3532         if (ext2fs_has_feature_bigalloc(fs->super) &&
3533             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
3534             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
3535             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
3536             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
3537             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
3538                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
3539                 if (pctx->errcode)
3540                         goto out;
3541         }
3542
3543         if (ctx->dirs_to_hash && pb.is_dir &&
3544             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
3545             !(inode->i_flags & EXT2_INDEX_FL) &&
3546             ((inode->i_size / fs->blocksize) >= 3))
3547                 e2fsck_rehash_dir_later(ctx, ino);
3548
3549 out:
3550         if (dirty_inode)
3551                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
3552 }
3553
3554 #if 0
3555 /*
3556  * Helper function called by process block when an illegal block is
3557  * found.  It returns a description about why the block is illegal
3558  */
3559 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
3560 {
3561         blk64_t super;
3562         int     i;
3563         static char     problem[80];
3564
3565         super = fs->super->s_first_data_block;
3566         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
3567         if (block < super) {
3568                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
3569                 return(problem);
3570         } else if (block >= ext2fs_blocks_count(fs->super)) {
3571                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
3572                 return(problem);
3573         }
3574         for (i = 0; i < fs->group_desc_count; i++) {
3575                 if (block == super) {
3576                         sprintf(problem, "is the superblock in group %d", i);
3577                         break;
3578                 }
3579                 if (block > super &&
3580                     block <= (super + fs->desc_blocks)) {
3581                         sprintf(problem, "is in the group descriptors "
3582                                 "of group %d", i);
3583                         break;
3584                 }
3585                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
3586                         sprintf(problem, "is the block bitmap of group %d", i);
3587                         break;
3588                 }
3589                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
3590                         sprintf(problem, "is the inode bitmap of group %d", i);
3591                         break;
3592                 }
3593                 if (block >= ext2fs_inode_table_loc(fs, i) &&
3594                     (block < ext2fs_inode_table_loc(fs, i)
3595                      + fs->inode_blocks_per_group)) {
3596                         sprintf(problem, "is in the inode table of group %d",
3597                                 i);
3598                         break;
3599                 }
3600                 super += fs->super->s_blocks_per_group;
3601         }
3602         return(problem);
3603 }
3604 #endif
3605
3606 /*
3607  * This is a helper function for check_blocks().
3608  */
3609 static int process_block(ext2_filsys fs,
3610                   blk64_t       *block_nr,
3611                   e2_blkcnt_t blockcnt,
3612                   blk64_t ref_block EXT2FS_ATTR((unused)),
3613                   int ref_offset EXT2FS_ATTR((unused)),
3614                   void *priv_data)
3615 {
3616         struct process_block_struct *p;
3617         struct problem_context *pctx;
3618         blk64_t blk = *block_nr;
3619         int     ret_code = 0;
3620         problem_t       problem = 0;
3621         e2fsck_t        ctx;
3622
3623         p = (struct process_block_struct *) priv_data;
3624         pctx = p->pctx;
3625         ctx = p->ctx;
3626
3627         /*
3628          * For a directory, add logical block zero for processing even if it's
3629          * not mapped or we'll be perennially stuck with broken "." and ".."
3630          * entries.
3631          */
3632         if (p->is_dir && blockcnt == 0 && blk == 0) {
3633                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
3634                 if (pctx->errcode) {
3635                         pctx->blk = blk;
3636                         pctx->num = blockcnt;
3637                         goto failed_add_dir_block;
3638                 }
3639                 p->last_db_block++;
3640         }
3641
3642         if (blk == 0)
3643                 return 0;
3644
3645 #if 0
3646         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
3647                blockcnt);
3648 #endif
3649
3650         /*
3651          * Simplistic fragmentation check.  We merely require that the
3652          * file be contiguous.  (Which can never be true for really
3653          * big files that are greater than a block group.)
3654          */
3655         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
3656                 if (p->previous_block+1 != blk) {
3657                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3658                                 char type = '?';
3659
3660                                 if (p->is_dir)
3661                                         type = 'd';
3662                                 else if (p->is_reg)
3663                                         type = 'f';
3664
3665                                 printf(_("%6lu(%c): expecting %6lu "
3666                                          "got phys %6lu (blkcnt %lld)\n"),
3667                                        (unsigned long) pctx->ino, type,
3668                                        (unsigned long) p->previous_block+1,
3669                                        (unsigned long) blk,
3670                                        blockcnt);
3671                         }
3672                         p->fragmented = 1;
3673                 }
3674         }
3675
3676         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
3677             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
3678                 problem = PR_1_TOOBIG_DIR;
3679         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
3680                 problem = PR_1_TOOBIG_DIR;
3681         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
3682                 problem = PR_1_TOOBIG_REG;
3683         if (!p->is_dir && !p->is_reg && blockcnt > 0)
3684                 problem = PR_1_TOOBIG_SYMLINK;
3685
3686         if (blk < fs->super->s_first_data_block ||
3687             blk >= ext2fs_blocks_count(fs->super))
3688                 problem = PR_1_ILLEGAL_BLOCK_NUM;
3689
3690         /*
3691          * If this IND/DIND/TIND block is squatting atop some critical metadata
3692          * (group descriptors, superblock, bitmap, inode table), any write to
3693          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
3694          * fix that and restart fsck.
3695          */
3696         if (blockcnt < 0 &&
3697             p->ino != EXT2_RESIZE_INO &&
3698             blk < ctx->fs->super->s_blocks_count &&
3699             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
3700                 pctx->blk = blk;
3701                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
3702                 if ((ctx->options & E2F_OPT_NO) == 0)
3703                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3704         }
3705
3706         if (problem) {
3707                 p->num_illegal_blocks++;
3708                 /*
3709                  * A bit of subterfuge here -- we're trying to fix a block
3710                  * mapping, but the IND/DIND/TIND block could have collided
3711                  * with some critical metadata.  So, fix the in-core mapping so
3712                  * iterate won't go insane, but return 0 instead of
3713                  * BLOCK_CHANGED so that it won't write the remapping out to
3714                  * our multiply linked block.
3715                  *
3716                  * Even if we previously determined that an *IND block
3717                  * conflicts with critical metadata, we must still try to
3718                  * iterate the *IND block as if it is an *IND block to find and
3719                  * mark the blocks it points to.  Better to be overly cautious
3720                  * with the used_blocks map so that we don't move the *IND
3721                  * block to a block that's really in use!
3722                  */
3723                 if (p->ino != EXT2_RESIZE_INO &&
3724                     ref_block != 0 &&
3725                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3726                                               ref_block)) {
3727                         *block_nr = 0;
3728                         return 0;
3729                 }
3730                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
3731                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
3732                                 p->clear = 1;
3733                                 return BLOCK_ABORT;
3734                         }
3735                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
3736                                 p->suppress = 1;
3737                                 set_latch_flags(PR_LATCH_BLOCK,
3738                                                 PRL_SUPPRESS, 0);
3739                         }
3740                 }
3741                 pctx->blk = blk;
3742                 pctx->blkcount = blockcnt;
3743                 if (fix_problem(ctx, problem, pctx)) {
3744                         blk = *block_nr = 0;
3745                         ret_code = BLOCK_CHANGED;
3746                         p->inode_modified = 1;
3747                         /*
3748                          * If the directory block is too big and is beyond the
3749                          * end of the FS, don't bother trying to add it for
3750                          * processing -- the kernel would never have created a
3751                          * directory this large, and we risk an ENOMEM abort.
3752                          * In any case, the toobig handler for extent-based
3753                          * directories also doesn't feed toobig blocks to
3754                          * pass 2.
3755                          */
3756                         if (problem == PR_1_TOOBIG_DIR)
3757                                 return ret_code;
3758                         goto mark_dir;
3759                 } else
3760                         return 0;
3761         }
3762
3763         if (p->ino == EXT2_RESIZE_INO) {
3764                 /*
3765                  * The resize inode has already be sanity checked
3766                  * during pass #0 (the superblock checks).  All we
3767                  * have to do is mark the double indirect block as
3768                  * being in use; all of the other blocks are handled
3769                  * by mark_table_blocks()).
3770                  */
3771                 if (blockcnt == BLOCK_COUNT_DIND)
3772                         mark_block_used(ctx, blk);
3773                 p->num_blocks++;
3774         } else if (!(ctx->fs->cluster_ratio_bits &&
3775                      p->previous_block &&
3776                      (EXT2FS_B2C(ctx->fs, blk) ==
3777                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
3778                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
3779                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
3780                 mark_block_used(ctx, blk);
3781                 p->num_blocks++;
3782         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
3783                                              p->last_block, blk, blockcnt)) {
3784                 pctx->blk = blockcnt;
3785                 pctx->blk2 = blk;
3786                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3787                 mark_block_used(ctx, blk);
3788                 mark_block_used(ctx, blk);
3789         }
3790         if (blockcnt >= 0)
3791                 p->last_block = blockcnt;
3792         p->previous_block = blk;
3793 mark_dir:
3794         if (p->is_dir && (blockcnt >= 0)) {
3795                 while (++p->last_db_block < blockcnt) {
3796                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
3797                                                               p->ino, 0,
3798                                                               p->last_db_block);
3799                         if (pctx->errcode) {
3800                                 pctx->blk = 0;
3801                                 pctx->num = p->last_db_block;
3802                                 goto failed_add_dir_block;
3803                         }
3804                 }
3805                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
3806                                                       blk, blockcnt);
3807                 if (pctx->errcode) {
3808                         pctx->blk = blk;
3809                         pctx->num = blockcnt;
3810                 failed_add_dir_block:
3811                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3812                         /* Should never get here */
3813                         ctx->flags |= E2F_FLAG_ABORT;
3814                         return BLOCK_ABORT;
3815                 }
3816         }
3817         return ret_code;
3818 }
3819
3820 static int process_bad_block(ext2_filsys fs,
3821                       blk64_t *block_nr,
3822                       e2_blkcnt_t blockcnt,
3823                       blk64_t ref_block EXT2FS_ATTR((unused)),
3824                       int ref_offset EXT2FS_ATTR((unused)),
3825                       void *priv_data)
3826 {
3827         struct process_block_struct *p;
3828         blk64_t         blk = *block_nr;
3829         blk64_t         first_block;
3830         dgrp_t          i;
3831         struct problem_context *pctx;
3832         e2fsck_t        ctx;
3833
3834         if (!blk)
3835                 return 0;
3836
3837         p = (struct process_block_struct *) priv_data;
3838         ctx = p->ctx;
3839         pctx = p->pctx;
3840
3841         pctx->ino = EXT2_BAD_INO;
3842         pctx->blk = blk;
3843         pctx->blkcount = blockcnt;
3844
3845         if ((blk < fs->super->s_first_data_block) ||
3846             (blk >= ext2fs_blocks_count(fs->super))) {
3847                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
3848                         *block_nr = 0;
3849                         return BLOCK_CHANGED;
3850                 } else
3851                         return 0;
3852         }
3853
3854         if (blockcnt < 0) {
3855                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
3856                         p->bbcheck = 1;
3857                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
3858                                 *block_nr = 0;
3859                                 return BLOCK_CHANGED;
3860                         }
3861                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
3862                                                     blk)) {
3863                         p->bbcheck = 1;
3864                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
3865                                         pctx)) {
3866                                 *block_nr = 0;
3867                                 return BLOCK_CHANGED;
3868                         }
3869                         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3870                                 return BLOCK_ABORT;
3871                 } else
3872                         mark_block_used(ctx, blk);
3873                 return 0;
3874         }
3875 #if 0
3876         printf ("DEBUG: Marking %u as bad.\n", blk);
3877 #endif
3878         ctx->fs_badblocks_count++;
3879         /*
3880          * If the block is not used, then mark it as used and return.
3881          * If it is already marked as found, this must mean that
3882          * there's an overlap between the filesystem table blocks
3883          * (bitmaps and inode table) and the bad block list.
3884          */
3885         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
3886                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
3887                 return 0;
3888         }
3889         /*
3890          * Try to find the where the filesystem block was used...
3891          */
3892         first_block = fs->super->s_first_data_block;
3893
3894         for (i = 0; i < fs->group_desc_count; i++ ) {
3895                 pctx->group = i;
3896                 pctx->blk = blk;
3897                 if (!ext2fs_bg_has_super(fs, i))
3898                         goto skip_super;
3899                 if (blk == first_block) {
3900                         if (i == 0) {
3901                                 if (fix_problem(ctx,
3902                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
3903                                                 pctx)) {
3904                                         *block_nr = 0;
3905                                         return BLOCK_CHANGED;
3906                                 }
3907                                 return 0;
3908                         }
3909                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
3910                         return 0;
3911                 }
3912                 if ((blk > first_block) &&
3913                     (blk <= first_block + fs->desc_blocks)) {
3914                         if (i == 0) {
3915                                 pctx->blk = *block_nr;
3916                                 if (fix_problem(ctx,
3917                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
3918                                         *block_nr = 0;
3919                                         return BLOCK_CHANGED;
3920                                 }
3921                                 return 0;
3922                         }
3923                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
3924                         return 0;
3925                 }
3926         skip_super:
3927                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
3928                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
3929                                 ctx->invalid_block_bitmap_flag[i]++;
3930                                 ctx->invalid_bitmaps++;
3931                         }
3932                         return 0;
3933                 }
3934                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
3935                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
3936                                 ctx->invalid_inode_bitmap_flag[i]++;
3937                                 ctx->invalid_bitmaps++;
3938                         }
3939                         return 0;
3940                 }
3941                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
3942                     (blk < (ext2fs_inode_table_loc(fs, i) +
3943                             fs->inode_blocks_per_group))) {
3944                         /*
3945                          * If there are bad blocks in the inode table,
3946                          * the inode scan code will try to do
3947                          * something reasonable automatically.
3948                          */
3949                         return 0;
3950                 }
3951                 first_block += fs->super->s_blocks_per_group;
3952         }
3953         /*
3954          * If we've gotten to this point, then the only
3955          * possibility is that the bad block inode meta data
3956          * is using a bad block.
3957          */
3958         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
3959             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
3960             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
3961                 p->bbcheck = 1;
3962                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
3963                         *block_nr = 0;
3964                         return BLOCK_CHANGED;
3965                 }
3966                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3967                         return BLOCK_ABORT;
3968                 return 0;
3969         }
3970
3971         pctx->group = -1;
3972
3973         /* Warn user that the block wasn't claimed */
3974         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
3975
3976         return 0;
3977 }
3978
3979 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
3980                             const char *name, int num, blk64_t *new_block)
3981 {
3982         ext2_filsys fs = ctx->fs;
3983         dgrp_t          last_grp;
3984         blk64_t         old_block = *new_block;
3985         blk64_t         last_block;
3986         dgrp_t          flexbg;
3987         unsigned        flexbg_size;
3988         int             i, is_flexbg;
3989         char            *buf;
3990         struct problem_context  pctx;
3991
3992         clear_problem_context(&pctx);
3993
3994         pctx.group = group;
3995         pctx.blk = old_block;
3996         pctx.str = name;
3997
3998         /*
3999          * For flex_bg filesystems, first try to allocate the metadata
4000          * within the flex_bg, and if that fails then try finding the
4001          * space anywhere in the filesystem.
4002          */
4003         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
4004         if (is_flexbg) {
4005                 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
4006                 flexbg = group / flexbg_size;
4007                 first_block = ext2fs_group_first_block2(fs,
4008                                                         flexbg_size * flexbg);
4009                 last_grp = group | (flexbg_size - 1);
4010                 if (last_grp >= fs->group_desc_count)
4011                         last_grp = fs->group_desc_count - 1;
4012                 last_block = ext2fs_group_last_block2(fs, last_grp);
4013         } else
4014                 last_block = ext2fs_group_last_block2(fs, group);
4015         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
4016                                                num, ctx->block_found_map,
4017                                                new_block);
4018         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
4019                 pctx.errcode = ext2fs_get_free_blocks2(fs,
4020                                 fs->super->s_first_data_block,
4021                                 ext2fs_blocks_count(fs->super),
4022                                 num, ctx->block_found_map, new_block);
4023         if (pctx.errcode) {
4024                 pctx.num = num;
4025                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4026                 ext2fs_unmark_valid(fs);
4027                 ctx->flags |= E2F_FLAG_ABORT;
4028                 return;
4029         }
4030         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4031         if (pctx.errcode) {
4032                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4033                 ext2fs_unmark_valid(fs);
4034                 ctx->flags |= E2F_FLAG_ABORT;
4035                 return;
4036         }
4037         ext2fs_mark_super_dirty(fs);
4038         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4039         pctx.blk2 = *new_block;
4040         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4041                           PR_1_RELOC_TO), &pctx);
4042         pctx.blk2 = 0;
4043         for (i = 0; i < num; i++) {
4044                 pctx.blk = i;
4045                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4046                 if (old_block) {
4047                         pctx.errcode = io_channel_read_blk64(fs->io,
4048                                    old_block + i, 1, buf);
4049                         if (pctx.errcode)
4050                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4051                         pctx.blk = (*new_block) + i;
4052                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4053                                                               1, buf);
4054                 } else {
4055                         pctx.blk = (*new_block) + i;
4056                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4057                                                            NULL, NULL);
4058                 }
4059
4060                 if (pctx.errcode)
4061                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4062         }
4063         ext2fs_free_mem(&buf);
4064 }
4065
4066 /*
4067  * This routine gets called at the end of pass 1 if bad blocks are
4068  * detected in the superblock, group descriptors, inode_bitmaps, or
4069  * block bitmaps.  At this point, all of the blocks have been mapped
4070  * out, so we can try to allocate new block(s) to replace the bad
4071  * blocks.
4072  */
4073 static void handle_fs_bad_blocks(e2fsck_t ctx)
4074 {
4075         ext2_filsys fs = ctx->fs;
4076         dgrp_t          i;
4077         blk64_t         first_block;
4078         blk64_t         new_blk;
4079
4080         for (i = 0; i < fs->group_desc_count; i++) {
4081                 first_block = ext2fs_group_first_block2(fs, i);
4082
4083                 if (ctx->invalid_block_bitmap_flag[i]) {
4084                         new_blk = ext2fs_block_bitmap_loc(fs, i);
4085                         new_table_block(ctx, first_block, i, _("block bitmap"),
4086                                         1, &new_blk);
4087                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4088                 }
4089                 if (ctx->invalid_inode_bitmap_flag[i]) {
4090                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
4091                         new_table_block(ctx, first_block, i, _("inode bitmap"),
4092                                         1, &new_blk);
4093                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4094                 }
4095                 if (ctx->invalid_inode_table_flag[i]) {
4096                         new_blk = ext2fs_inode_table_loc(fs, i);
4097                         new_table_block(ctx, first_block, i, _("inode table"),
4098                                         fs->inode_blocks_per_group,
4099                                         &new_blk);
4100                         ext2fs_inode_table_loc_set(fs, i, new_blk);
4101                         ctx->flags |= E2F_FLAG_RESTART;
4102                 }
4103         }
4104         ctx->invalid_bitmaps = 0;
4105 }
4106
4107 /*
4108  * This routine marks all blocks which are used by the superblock,
4109  * group descriptors, inode bitmaps, and block bitmaps.
4110  */
4111 static void mark_table_blocks(e2fsck_t ctx)
4112 {
4113         ext2_filsys fs = ctx->fs;
4114         blk64_t b;
4115         dgrp_t  i;
4116         unsigned int    j;
4117         struct problem_context pctx;
4118
4119         clear_problem_context(&pctx);
4120
4121         for (i = 0; i < fs->group_desc_count; i++) {
4122                 pctx.group = i;
4123
4124                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
4125                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
4126
4127                 /*
4128                  * Mark the blocks used for the inode table
4129                  */
4130                 if (ext2fs_inode_table_loc(fs, i)) {
4131                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
4132                              j < fs->inode_blocks_per_group;
4133                              j++, b++) {
4134                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4135                                                              b)) {
4136                                         pctx.blk = b;
4137                                         if (!ctx->invalid_inode_table_flag[i] &&
4138                                             fix_problem(ctx,
4139                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
4140                                                 ctx->invalid_inode_table_flag[i]++;
4141                                                 ctx->invalid_bitmaps++;
4142                                         }
4143                                 } else {
4144                                     ext2fs_mark_block_bitmap2(
4145                                                 ctx->block_found_map, b);
4146                                     ext2fs_mark_block_bitmap2(
4147                                                 ctx->block_metadata_map, b);
4148                                 }
4149                         }
4150                 }
4151
4152                 /*
4153                  * Mark block used for the block bitmap
4154                  */
4155                 if (ext2fs_block_bitmap_loc(fs, i)) {
4156                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4157                                      ext2fs_block_bitmap_loc(fs, i))) {
4158                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
4159                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
4160                                         ctx->invalid_block_bitmap_flag[i]++;
4161                                         ctx->invalid_bitmaps++;
4162                                 }
4163                         } else {
4164                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4165                                      ext2fs_block_bitmap_loc(fs, i));
4166                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4167                                      ext2fs_block_bitmap_loc(fs, i));
4168                         }
4169                 }
4170                 /*
4171                  * Mark block used for the inode bitmap
4172                  */
4173                 if (ext2fs_inode_bitmap_loc(fs, i)) {
4174                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4175                                      ext2fs_inode_bitmap_loc(fs, i))) {
4176                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
4177                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
4178                                         ctx->invalid_inode_bitmap_flag[i]++;
4179                                         ctx->invalid_bitmaps++;
4180                                 }
4181                         } else {
4182                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4183                                      ext2fs_inode_bitmap_loc(fs, i));
4184                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4185                                      ext2fs_inode_bitmap_loc(fs, i));
4186                         }
4187                 }
4188         }
4189 }
4190
4191 /*
4192  * These subroutines short circuits ext2fs_get_blocks and
4193  * ext2fs_check_directory; we use them since we already have the inode
4194  * structure, so there's no point in letting the ext2fs library read
4195  * the inode again.
4196  */
4197 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
4198                                   blk_t *blocks)
4199 {
4200         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4201         int     i;
4202
4203         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4204                 return EXT2_ET_CALLBACK_NOTHANDLED;
4205
4206         for (i=0; i < EXT2_N_BLOCKS; i++)
4207                 blocks[i] = ctx->stashed_inode->i_block[i];
4208         return 0;
4209 }
4210
4211 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
4212                                   struct ext2_inode *inode)
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         *inode = *ctx->stashed_inode;
4219         return 0;
4220 }
4221
4222 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
4223                             struct ext2_inode *inode)
4224 {
4225         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4226
4227         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
4228                 (inode != ctx->stashed_inode))
4229                 *ctx->stashed_inode = *inode;
4230         return EXT2_ET_CALLBACK_NOTHANDLED;
4231 }
4232
4233 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
4234 {
4235         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4236
4237         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4238                 return EXT2_ET_CALLBACK_NOTHANDLED;
4239
4240         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
4241                 return EXT2_ET_NO_DIRECTORY;
4242         return 0;
4243 }
4244
4245 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
4246                                         blk64_t *ret)
4247 {
4248         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4249         errcode_t       retval;
4250         blk64_t         new_block;
4251
4252         if (ctx->block_found_map) {
4253                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
4254                                            &new_block);
4255                 if (retval)
4256                         return retval;
4257                 if (fs->block_map) {
4258                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
4259                         ext2fs_mark_bb_dirty(fs);
4260                 }
4261         } else {
4262                 if (!fs->block_map) {
4263                         retval = ext2fs_read_block_bitmap(fs);
4264                         if (retval)
4265                                 return retval;
4266                 }
4267
4268                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
4269                 if (retval)
4270                         return retval;
4271         }
4272
4273         *ret = new_block;
4274         return (0);
4275 }
4276
4277 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
4278                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
4279 {
4280         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4281         errcode_t       retval;
4282
4283         if (ctx->block_found_map)
4284                 return ext2fs_new_range(fs, flags, goal, len,
4285                                         ctx->block_found_map, pblk, plen);
4286
4287         if (!fs->block_map) {
4288                 retval = ext2fs_read_block_bitmap(fs);
4289                 if (retval)
4290                         return retval;
4291         }
4292
4293         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
4294                                 pblk, plen);
4295 }
4296
4297 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
4298 {
4299         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4300
4301         /* Never free a critical metadata block */
4302         if (ctx->block_found_map &&
4303             ctx->block_metadata_map &&
4304             inuse < 0 &&
4305             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
4306                 return;
4307
4308         if (ctx->block_found_map) {
4309                 if (inuse > 0)
4310                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4311                 else
4312                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
4313         }
4314 }
4315
4316 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
4317                                            blk_t num, int inuse)
4318 {
4319         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4320
4321         /* Never free a critical metadata block */
4322         if (ctx->block_found_map &&
4323             ctx->block_metadata_map &&
4324             inuse < 0 &&
4325             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
4326                 return;
4327
4328         if (ctx->block_found_map) {
4329                 if (inuse > 0)
4330                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
4331                                                         blk, num);
4332                 else
4333                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
4334                                                         blk, num);
4335         }
4336 }
4337
4338 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
4339 {
4340         ext2_filsys fs = ctx->fs;
4341
4342         if (use_shortcuts) {
4343                 fs->get_blocks = pass1_get_blocks;
4344                 fs->check_directory = pass1_check_directory;
4345                 fs->read_inode = pass1_read_inode;
4346                 fs->write_inode = pass1_write_inode;
4347                 ctx->stashed_ino = 0;
4348         } else {
4349                 fs->get_blocks = 0;
4350                 fs->check_directory = 0;
4351                 fs->read_inode = 0;
4352                 fs->write_inode = 0;
4353         }
4354 }
4355
4356 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
4357 {
4358         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
4359         ext2fs_set_block_alloc_stats_callback(ctx->fs,
4360                                                 e2fsck_block_alloc_stats, 0);
4361         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
4362         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
4363                                         e2fsck_block_alloc_stats_range, NULL);
4364 }