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  *      - The encryption policy ID of each encrypted inode. (encrypted_files)
  31  *
  32  * Pass 1 is designed to stash away enough information so that the
  33  * other passes should not need to read in the inode information
  34  * during the normal course of a filesystem check.  (Although if an
  35  * inconsistency is detected, other passes may need to read in an
  36  * inode to fix it.)
  37  *
  38  * Note that pass 1B will be invoked if there are any duplicate blocks
  39  * found.
  40  */
  41
  42 #define _GNU_SOURCE 1 /* get strnlen() */
  43 #include "config.h"
  44 #include <string.h>
  45 #include <time.h>
  46 #ifdef HAVE_ERRNO_H
  47 #include <errno.h>
  48 #endif
  49
  50 #include "e2fsck.h"
  51 #include <ext2fs/ext2_ext_attr.h>
  52 #include <e2p/e2p.h>
  53
  54 #include "problem.h"
  55
  56 #ifdef NO_INLINE_FUNCS
  57 #define _INLINE_
  58 #else
  59 #define _INLINE_ inline
  60 #endif
  61
  62 #undef DEBUG
  63
  64 struct ea_quota {
  65         blk64_t blocks;
  66         __u64 inodes;
  67 };
  68
  69 static int process_block(ext2_filsys fs, blk64_t        *blocknr,
  70                          e2_blkcnt_t blockcnt, blk64_t ref_blk,
  71                          int ref_offset, void *priv_data);
  72 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
  73                              e2_blkcnt_t blockcnt, blk64_t ref_blk,
  74                              int ref_offset, void *priv_data);
  75 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
  76                          char *block_buf,
  77                          const struct ea_quota *ea_ibody_quota);
  78 static void mark_table_blocks(e2fsck_t ctx);
  79 static void alloc_bb_map(e2fsck_t ctx);
  80 static void alloc_imagic_map(e2fsck_t ctx);
  81 static void mark_inode_bad(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  * Check to make sure a device inode is real.  Returns 1 if the device
 133  * checks out, 0 if not.
 134  *
 135  * Note: this routine is now also used to check FIFO's and Sockets,
 136  * since they have the same requirement; the i_block fields should be
 137  * zero.
 138  */
 139 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
 140                                     struct ext2_inode *inode)
 141 {
 142         int     i;
 143
 144         /*
 145          * If the index or extents flag is set, then this is a bogus
 146          * device/fifo/socket
 147          */
 148         if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
 149                 return 0;
 150
 151         /*
 152          * We should be able to do the test below all the time, but
 153          * because the kernel doesn't forcibly clear the device
 154          * inode's additional i_block fields, there are some rare
 155          * occasions when a legitimate device inode will have non-zero
 156          * additional i_block fields.  So for now, we only complain
 157          * when the immutable flag is set, which should never happen
 158          * for devices.  (And that's when the problem is caused, since
 159          * you can't set or clear immutable flags for devices.)  Once
 160          * the kernel has been fixed we can change this...
 161          */
 162         if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
 163                 for (i=4; i < EXT2_N_BLOCKS; i++)
 164                         if (inode->i_block[i])
 165                                 return 0;
 166         }
 167         return 1;
 168 }
 169
 170 /*
 171  * Check to make sure a symlink inode is real.  Returns 1 if the symlink
 172  * checks out, 0 if not.
 173  */
 174 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
 175                                struct ext2_inode *inode, char *buf)
 176 {
 177         unsigned int buflen;
 178         unsigned int len;
 179
 180         if ((inode->i_size_high || inode->i_size == 0) ||
 181             (inode->i_flags & EXT2_INDEX_FL))
 182                 return 0;
 183
 184         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
 185                 size_t inline_size;
 186
 187                 if (inode->i_flags & EXT4_EXTENTS_FL)
 188                         return 0;
 189                 if (ext2fs_inline_data_size(fs, ino, &inline_size))
 190                         return 0;
 191                 if (inode->i_size != inline_size)
 192                         return 0;
 193
 194                 return 1;
 195         }
 196
 197         if (ext2fs_is_fast_symlink(inode)) {
 198                 if (inode->i_flags & EXT4_EXTENTS_FL)
 199                         return 0;
 200                 buf = (char *)inode->i_block;
 201                 buflen = sizeof(inode->i_block);
 202         } else {
 203                 ext2_extent_handle_t    handle;
 204                 struct ext2_extent_info info;
 205                 struct ext2fs_extent    extent;
 206                 blk64_t blk;
 207                 int i;
 208
 209                 if (inode->i_flags & EXT4_EXTENTS_FL) {
 210                         if (ext2fs_extent_open2(fs, ino, inode, &handle))
 211                                 return 0;
 212                         if (ext2fs_extent_get_info(handle, &info) ||
 213                             (info.num_entries != 1) ||
 214                             (info.max_depth != 0)) {
 215                                 ext2fs_extent_free(handle);
 216                                 return 0;
 217                         }
 218                         if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
 219                                               &extent) ||
 220                             (extent.e_lblk != 0) ||
 221                             (extent.e_len != 1)) {
 222                                 ext2fs_extent_free(handle);
 223                                 return 0;
 224                         }
 225                         blk = extent.e_pblk;
 226                         ext2fs_extent_free(handle);
 227                 } else {
 228                         blk = inode->i_block[0];
 229
 230                         for (i = 1; i < EXT2_N_BLOCKS; i++)
 231                                 if (inode->i_block[i])
 232                                         return 0;
 233                 }
 234
 235                 if (blk < fs->super->s_first_data_block ||
 236                     blk >= ext2fs_blocks_count(fs->super))
 237                         return 0;
 238
 239                 if (io_channel_read_blk64(fs->io, blk, 1, buf))
 240                         return 0;
 241
 242                 buflen = fs->blocksize;
 243         }
 244
 245         if (inode->i_flags & EXT4_ENCRYPT_FL)
 246                 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
 247         else
 248                 len = strnlen(buf, buflen);
 249
 250         if (len >= buflen)
 251                 return 0;
 252
 253         if (len != inode->i_size)
 254                 return 0;
 255         return 1;
 256 }
 257
 258 /*
 259  * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
 260  */
 261 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
 262 static void check_extents_inlinedata(e2fsck_t ctx,
 263                                      struct problem_context *pctx)
 264 {
 265         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 266                 return;
 267
 268         if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
 269                 return;
 270
 271         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 272         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 273 }
 274 #undef BAD_SPECIAL_FLAGS
 275
 276 /*
 277  * If the immutable (or append-only) flag is set on the inode, offer
 278  * to clear it.
 279  */
 280 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
 281 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
 282 {
 283         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 284                 return;
 285
 286         if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
 287                 return;
 288
 289         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 290         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 291 }
 292
 293 /*
 294  * If device, fifo or socket, check size is zero -- if not offer to
 295  * clear it
 296  */
 297 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
 298 {
 299         struct ext2_inode *inode = pctx->inode;
 300
 301         if (EXT2_I_SIZE(inode) == 0)
 302                 return;
 303
 304         if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
 305                 return;
 306
 307         ext2fs_inode_size_set(ctx->fs, inode, 0);
 308         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 309 }
 310
 311 /*
 312  * For a given size, calculate how many blocks would be charged towards quota.
 313  */
 314 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
 315 {
 316         blk64_t clusters;
 317
 318         clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
 319         return EXT2FS_C2B(fs, clusters);
 320 }
 321
 322 /*
 323  * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
 324  * the problem code.
 325  */
 326 static problem_t check_large_ea_inode(e2fsck_t ctx,
 327                                       struct ext2_ext_attr_entry *entry,
 328                                       struct problem_context *pctx,
 329                                       blk64_t *quota_blocks)
 330 {
 331         struct ext2_inode inode;
 332         __u32 hash;
 333         errcode_t retval;
 334
 335         /* Check if inode is within valid range */
 336         if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
 337             (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
 338                 pctx->num = entry->e_value_inum;
 339                 return PR_1_ATTR_VALUE_EA_INODE;
 340         }
 341
 342         e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
 343
 344         retval = ext2fs_ext_attr_hash_entry2(ctx->fs, entry, NULL, &hash);
 345         if (retval) {
 346                 com_err("check_large_ea_inode", retval,
 347                         _("while hashing entry with e_value_inum = %u"),
 348                         entry->e_value_inum);
 349                 fatal_error(ctx, 0);
 350         }
 351
 352         if (hash == entry->e_hash) {
 353                 *quota_blocks = size_to_quota_blocks(ctx->fs,
 354                                                      entry->e_value_size);
 355         } else {
 356                 /* This might be an old Lustre-style ea_inode reference. */
 357                 if (inode.i_mtime == pctx->ino &&
 358                     inode.i_generation == pctx->inode->i_generation) {
 359                         *quota_blocks = 0;
 360                 } else {
 361                         /* If target inode is also missing EA_INODE flag,
 362                          * this is likely to be a bad reference.
 363                          */
 364                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 365                                 pctx->num = entry->e_value_inum;
 366                                 return PR_1_ATTR_VALUE_EA_INODE;
 367                         } else {
 368                                 pctx->num = entry->e_hash;
 369                                 return PR_1_ATTR_HASH;
 370                         }
 371                 }
 372         }
 373
 374         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 375                 pctx->num = entry->e_value_inum;
 376                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
 377                         inode.i_flags |= EXT4_EA_INODE_FL;
 378                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
 379                                            &inode);
 380                 } else {
 381                         return PR_1_ATTR_NO_EA_INODE_FL;
 382                 }
 383         }
 384         return 0;
 385 }
 386
 387 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
 388                               struct ext2_ext_attr_entry *first, void *end)
 389 {
 390         struct ext2_ext_attr_entry *entry;
 391
 392         for (entry = first;
 393              (void *)entry < end && !EXT2_EXT_IS_LAST_ENTRY(entry);
 394              entry = EXT2_EXT_ATTR_NEXT(entry)) {
 395                 if (!entry->e_value_inum)
 396                         continue;
 397                 if (!ctx->ea_inode_refs) {
 398                         pctx->errcode = ea_refcount_create(0,
 399                                                            &ctx->ea_inode_refs);
 400                         if (pctx->errcode) {
 401                                 pctx->num = 4;
 402                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
 403                                 ctx->flags |= E2F_FLAG_ABORT;
 404                                 return;
 405                         }
 406                 }
 407                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
 408                                       0);
 409         }
 410 }
 411
 412 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
 413                               struct ea_quota *ea_ibody_quota)
 414 {
 415         struct ext2_super_block *sb = ctx->fs->super;
 416         struct ext2_inode_large *inode;
 417         struct ext2_ext_attr_entry *entry;
 418         char *start, *header, *end;
 419         unsigned int storage_size, remain;
 420         problem_t problem = 0;
 421         region_t region = 0;
 422
 423         ea_ibody_quota->blocks = 0;
 424         ea_ibody_quota->inodes = 0;
 425
 426         inode = (struct ext2_inode_large *) pctx->inode;
 427         storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
 428                 inode->i_extra_isize;
 429         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 430                  inode->i_extra_isize;
 431         end = header + storage_size;
 432         start = header + sizeof(__u32);
 433         entry = (struct ext2_ext_attr_entry *) start;
 434
 435         /* scan all entry's headers first */
 436
 437         /* take finish entry 0UL into account */
 438         remain = storage_size - sizeof(__u32);
 439
 440         region = region_create(0, storage_size);
 441         if (!region) {
 442                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
 443                 problem = 0;
 444                 ctx->flags |= E2F_FLAG_ABORT;
 445                 return;
 446         }
 447         if (region_allocate(region, 0, sizeof(__u32))) {
 448                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 449                 goto fix;
 450         }
 451
 452         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
 453                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
 454                 __u32 hash;
 455
 456                 if (region_allocate(region, (char *)entry - (char *)header,
 457                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
 458                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
 459                         goto fix;
 460                 }
 461
 462                 /* header eats this space */
 463                 remain -= sizeof(struct ext2_ext_attr_entry);
 464
 465                 /* is attribute name valid? */
 466                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
 467                         pctx->num = entry->e_name_len;
 468                         problem = PR_1_ATTR_NAME_LEN;
 469                         goto fix;
 470                 }
 471
 472                 /* attribute len eats this space */
 473                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
 474
 475                 if (entry->e_value_inum == 0) {
 476                         /* check value size */
 477                         if (entry->e_value_size > remain) {
 478                                 pctx->num = entry->e_value_size;
 479                                 problem = PR_1_ATTR_VALUE_SIZE;
 480                                 goto fix;
 481                         }
 482
 483                         if (entry->e_value_size &&
 484                             region_allocate(region,
 485                                             sizeof(__u32) + entry->e_value_offs,
 486                                             EXT2_EXT_ATTR_SIZE(
 487                                                 entry->e_value_size))) {
 488                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 489                                 goto fix;
 490                         }
 491
 492                         hash = ext2fs_ext_attr_hash_entry(entry,
 493                                                           start + entry->e_value_offs);
 494
 495                         /* e_hash may be 0 in older inode's ea */
 496                         if (entry->e_hash != 0 && entry->e_hash != hash) {
 497                                 pctx->num = entry->e_hash;
 498                                 problem = PR_1_ATTR_HASH;
 499                                 goto fix;
 500                         }
 501                 } else {
 502                         blk64_t quota_blocks;
 503
 504                         problem = check_large_ea_inode(ctx, entry, pctx,
 505                                                        &quota_blocks);
 506                         if (problem != 0)
 507                                 goto fix;
 508
 509                         ea_ibody_quota->blocks += quota_blocks;
 510                         ea_ibody_quota->inodes++;
 511                 }
 512
 513                 /* If EA value is stored in external inode then it does not
 514                  * consume space here */
 515                 if (entry->e_value_inum == 0)
 516                         remain -= entry->e_value_size;
 517
 518                 entry = EXT2_EXT_ATTR_NEXT(entry);
 519         }
 520
 521         if (region_allocate(region, (char *)entry - (char *)header,
 522                             sizeof(__u32))) {
 523                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 524                 goto fix;
 525         }
 526 fix:
 527         if (region)
 528                 region_free(region);
 529         /*
 530          * it seems like a corruption. it's very unlikely we could repair
 531          * EA(s) in automatic fashion -bzzz
 532          */
 533         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
 534                 inc_ea_inode_refs(ctx, pctx,
 535                                   (struct ext2_ext_attr_entry *)start, end);
 536                 return;
 537         }
 538
 539         /* simply remove all possible EA(s) */
 540         *((__u32 *)header) = 0UL;
 541         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 542                                 EXT2_INODE_SIZE(sb), "pass1");
 543         ea_ibody_quota->blocks = 0;
 544         ea_ibody_quota->inodes = 0;
 545 }
 546
 547 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
 548         return (xtime & (1U << 31)) != 0 &&
 549                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
 550 }
 551
 552 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
 553         check_inode_extra_negative_epoch(inode->i_##xtime, \
 554                                          inode->i_##xtime##_extra)
 555
 556 /* When today's date is earlier than 2242, we assume that atimes,
 557  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
 558  * actually pre-1970 dates mis-encoded.
 559  */
 560 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
 561
 562 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
 563                                     struct ea_quota *ea_ibody_quota)
 564 {
 565         struct ext2_super_block *sb = ctx->fs->super;
 566         struct ext2_inode_large *inode;
 567         __u32 *eamagic;
 568         int min, max;
 569
 570         ea_ibody_quota->blocks = 0;
 571         ea_ibody_quota->inodes = 0;
 572
 573         inode = (struct ext2_inode_large *) pctx->inode;
 574         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
 575                 /* this isn't large inode. so, nothing to check */
 576                 return;
 577         }
 578
 579 #if 0
 580         printf("inode #%u, i_extra_size %d\n", pctx->ino,
 581                         inode->i_extra_isize);
 582 #endif
 583         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
 584         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
 585         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
 586         /*
 587          * For now we will allow i_extra_isize to be 0, but really
 588          * implementations should never allow i_extra_isize to be 0
 589          */
 590         if (inode->i_extra_isize &&
 591             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
 592              inode->i_extra_isize & 3)) {
 593                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
 594                         return;
 595                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
 596                         inode->i_extra_isize = sb->s_want_extra_isize;
 597                 else
 598                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
 599                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 600                                         EXT2_INODE_SIZE(sb), "pass1");
 601         }
 602
 603         /* check if there is no place for an EA header */
 604         if (inode->i_extra_isize >= max - sizeof(__u32))
 605                 return;
 606
 607         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 608                         inode->i_extra_isize);
 609         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
 610                 /* it seems inode has an extended attribute(s) in body */
 611                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
 612         }
 613
 614         /*
 615          * If the inode's extended atime (ctime, crtime, mtime) is stored in
 616          * the old, invalid format, repair it.
 617          */
 618         if (((sizeof(time_t) <= 4) ||
 619              (((sizeof(time_t) > 4) &&
 620                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
 621             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
 622              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
 623              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
 624              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
 625
 626                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
 627                         return;
 628
 629                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
 630                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
 631                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
 632                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
 633                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
 634                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
 635                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
 636                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
 637                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 638                                         EXT2_INODE_SIZE(sb), "pass1");
 639         }
 640
 641 }
 642
 643 /*
 644  * Check to see if the inode might really be a directory, despite i_mode
 645  *
 646  * This is a lot of complexity for something for which I'm not really
 647  * convinced happens frequently in the wild.  If for any reason this
 648  * causes any problems, take this code out.
 649  * [tytso:20070331.0827EDT]
 650  */
 651 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
 652                                 char *buf)
 653 {
 654         struct ext2_inode *inode = pctx->inode;
 655         struct ext2_dir_entry   *dirent;
 656         errcode_t               retval;
 657         blk64_t                 blk;
 658         unsigned int            i, rec_len, not_device = 0;
 659         int                     extent_fs;
 660         int                     inlinedata_fs;
 661
 662         /*
 663          * If the mode looks OK, we believe it.  If the first block in
 664          * the i_block array is 0, this cannot be a directory. If the
 665          * inode is extent-mapped, it is still the case that the latter
 666          * cannot be 0 - the magic number in the extent header would make
 667          * it nonzero.
 668          */
 669         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
 670             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
 671                 return;
 672
 673         /*
 674          * Check the block numbers in the i_block array for validity:
 675          * zero blocks are skipped (but the first one cannot be zero -
 676          * see above), other blocks are checked against the first and
 677          * max data blocks (from the the superblock) and against the
 678          * block bitmap. Any invalid block found means this cannot be
 679          * a directory.
 680          *
 681          * If there are non-zero blocks past the fourth entry, then
 682          * this cannot be a device file: we remember that for the next
 683          * check.
 684          *
 685          * For extent mapped files, we don't do any sanity checking:
 686          * just try to get the phys block of logical block 0 and run
 687          * with it.
 688          *
 689          * For inline data files, we just try to get the size of inline
 690          * data.  If it's true, we will treat it as a directory.
 691          */
 692
 693         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
 694         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
 695         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
 696                 size_t size;
 697                 __u32 dotdot;
 698                 unsigned int rec_len2;
 699                 struct ext2_dir_entry de;
 700
 701                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
 702                         return;
 703                 /*
 704                  * If the size isn't a multiple of 4, it's probably not a
 705                  * directory??
 706                  */
 707                 if (size & 3)
 708                         return;
 709                 /*
 710                  * If the first 10 bytes don't look like a directory entry,
 711                  * it's probably not a directory.
 712                  */
 713                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
 714                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
 715                        EXT2_DIR_REC_LEN(0));
 716                 dotdot = ext2fs_le32_to_cpu(dotdot);
 717                 de.inode = ext2fs_le32_to_cpu(de.inode);
 718                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
 719                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
 720                 if (dotdot >= ctx->fs->super->s_inodes_count ||
 721                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
 722                      dotdot != EXT2_ROOT_INO) ||
 723                     de.inode >= ctx->fs->super->s_inodes_count ||
 724                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
 725                      de.inode != 0) ||
 726                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
 727                               EXT4_INLINE_DATA_DOTDOT_SIZE)
 728                         return;
 729                 /* device files never have a "system.data" entry */
 730                 goto isdir;
 731         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
 732                 /* extent mapped */
 733                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
 734                                  &blk))
 735                         return;
 736                 /* device files are never extent mapped */
 737                 not_device++;
 738         } else {
 739                 for (i=0; i < EXT2_N_BLOCKS; i++) {
 740                         blk = inode->i_block[i];
 741                         if (!blk)
 742                                 continue;
 743                         if (i >= 4)
 744                                 not_device++;
 745
 746                         if (blk < ctx->fs->super->s_first_data_block ||
 747                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
 748                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
 749                                                            blk))
 750                                 return; /* Invalid block, can't be dir */
 751                 }
 752                 blk = inode->i_block[0];
 753         }
 754
 755         /*
 756          * If the mode says this is a device file and the i_links_count field
 757          * is sane and we have not ruled it out as a device file previously,
 758          * we declare it a device file, not a directory.
 759          */
 760         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
 761             (inode->i_links_count == 1) && !not_device)
 762                 return;
 763
 764         /* read the first block */
 765         ehandler_operation(_("reading directory block"));
 766         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
 767         ehandler_operation(0);
 768         if (retval)
 769                 return;
 770
 771         dirent = (struct ext2_dir_entry *) buf;
 772         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 773         if (retval)
 774                 return;
 775         if ((ext2fs_dirent_name_len(dirent) != 1) ||
 776             (dirent->name[0] != '.') ||
 777             (dirent->inode != pctx->ino) ||
 778             (rec_len < 12) ||
 779             (rec_len % 4) ||
 780             (rec_len >= ctx->fs->blocksize - 12))
 781                 return;
 782
 783         dirent = (struct ext2_dir_entry *) (buf + rec_len);
 784         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 785         if (retval)
 786                 return;
 787         if ((ext2fs_dirent_name_len(dirent) != 2) ||
 788             (dirent->name[0] != '.') ||
 789             (dirent->name[1] != '.') ||
 790             (rec_len < 12) ||
 791             (rec_len % 4))
 792                 return;
 793
 794 isdir:
 795         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
 796                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
 797                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
 798                                         EXT2_INODE_SIZE(ctx->fs->super),
 799                                         "check_is_really_dir");
 800         }
 801 }
 802
 803 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
 804                                      int flags, ext2_icount_t hint,
 805                                      ext2_icount_t *ret)
 806 {
 807         unsigned int            threshold;
 808         unsigned int            save_type;
 809         ext2_ino_t              num_dirs;
 810         errcode_t               retval;
 811         char                    *tdb_dir;
 812         int                     enable;
 813
 814         *ret = 0;
 815
 816         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
 817                            &tdb_dir);
 818         profile_get_uint(ctx->profile, "scratch_files",
 819                          "numdirs_threshold", 0, 0, &threshold);
 820         profile_get_boolean(ctx->profile, "scratch_files",
 821                             "icount", 0, 1, &enable);
 822
 823         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
 824         if (retval)
 825                 num_dirs = 1024;        /* Guess */
 826
 827         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
 828             (!threshold || num_dirs > threshold)) {
 829                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
 830                                                   flags, ret);
 831                 if (retval == 0)
 832                         return 0;
 833         }
 834         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
 835                                &save_type);
 836         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
 837                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
 838         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
 839         ctx->fs->default_bitmap_type = save_type;
 840         return retval;
 841 }
 842
 843 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
 844                                             e2fsck_t ctx,
 845                                             struct problem_context *pctx)
 846 {
 847         errcode_t retval;
 848         struct ext2_inode_large inode;
 849
 850         /*
 851          * Reread inode.  If we don't see checksum error, then this inode
 852          * has been fixed elsewhere.
 853          */
 854         ctx->stashed_ino = 0;
 855         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
 856                                         sizeof(inode));
 857         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
 858                 return retval;
 859         if (!retval)
 860                 return 0;
 861
 862         /*
 863          * Checksum still doesn't match.  That implies that the inode passes
 864          * all the sanity checks, so maybe the checksum is simply corrupt.
 865          * See if the user will go for fixing that.
 866          */
 867         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
 868                 return 0;
 869
 870         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
 871                                          sizeof(inode));
 872         return retval;
 873 }
 874
 875 static void reserve_block_for_root_repair(e2fsck_t ctx)
 876 {
 877         blk64_t         blk = 0;
 878         errcode_t       err;
 879         ext2_filsys     fs = ctx->fs;
 880
 881         ctx->root_repair_block = 0;
 882         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
 883                 return;
 884
 885         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 886         if (err)
 887                 return;
 888         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 889         ctx->root_repair_block = blk;
 890 }
 891
 892 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
 893 {
 894         blk64_t         blk = 0;
 895         errcode_t       err;
 896         ext2_filsys     fs = ctx->fs;
 897         static const char name[] = "lost+found";
 898         ext2_ino_t      ino;
 899
 900         ctx->lnf_repair_block = 0;
 901         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
 902                 return;
 903
 904         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 905         if (err)
 906                 return;
 907         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 908         ctx->lnf_repair_block = blk;
 909 }
 910
 911 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
 912                                          size_t *sz)
 913 {
 914         void *p;
 915         struct ext2_xattr_handle *handle;
 916         errcode_t retval;
 917
 918         retval = ext2fs_xattrs_open(fs, ino, &handle);
 919         if (retval)
 920                 return retval;
 921
 922         retval = ext2fs_xattrs_read(handle);
 923         if (retval)
 924                 goto err;
 925
 926         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
 927         if (retval)
 928                 goto err;
 929         ext2fs_free_mem(&p);
 930 err:
 931         (void) ext2fs_xattrs_close(&handle);
 932         return retval;
 933 }
 934
 935 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
 936                                     struct problem_context *pctx,
 937                                     int failed_csum)
 938 {
 939         if (!failed_csum)
 940                 return;
 941
 942         /*
 943          * If the inode failed the checksum and the user didn't
 944          * clear the inode, test the checksum again -- if it still
 945          * fails, ask the user if the checksum should be corrected.
 946          */
 947         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
 948         if (pctx->errcode)
 949                 ctx->flags |= E2F_FLAG_ABORT;
 950 }
 951 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
 952         do { \
 953                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
 954                 if ((ctx)->flags & E2F_FLAG_ABORT) \
 955                         return; \
 956         } while (0)
 957
 958 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
 959 {
 960         __u32 x;
 961         int i;
 962
 963         for (i = 0; i < EXT2_N_BLOCKS; i++) {
 964                 x = inode->i_block[i];
 965 #ifdef WORDS_BIGENDIAN
 966                 x = ext2fs_swab32(x);
 967 #endif
 968                 if (x >= ext2fs_blocks_count(fs->super))
 969                         return 0;
 970         }
 971
 972         return 1;
 973 }
 974
 975 /*
 976  * Figure out what to do with an inode that has both extents and inline data
 977  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
 978  */
 979 static int fix_inline_data_extents_file(e2fsck_t ctx,
 980                                         ext2_ino_t ino,
 981                                         struct ext2_inode *inode,
 982                                         int inode_size,
 983                                         struct problem_context *pctx)
 984 {
 985         size_t max_inline_ea_size;
 986         ext2_filsys fs = ctx->fs;
 987         int dirty = 0;
 988
 989         /* Both feature flags not set?  Just run the regular checks */
 990         if (!ext2fs_has_feature_extents(fs->super) &&
 991             !ext2fs_has_feature_inline_data(fs->super))
 992                 return 0;
 993
 994         /* Clear both flags if it's a special file */
 995         if (LINUX_S_ISCHR(inode->i_mode) ||
 996             LINUX_S_ISBLK(inode->i_mode) ||
 997             LINUX_S_ISFIFO(inode->i_mode) ||
 998             LINUX_S_ISSOCK(inode->i_mode)) {
 999                 check_extents_inlinedata(ctx, pctx);
1000                 return 0;
1001         }
1002
1003         /* If it looks like an extent tree, try to clear inlinedata */
1004         if (ext2fs_extent_header_verify(inode->i_block,
1005                                  sizeof(inode->i_block)) == 0 &&
1006             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1007                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1008                 dirty = 1;
1009                 goto out;
1010         }
1011
1012         /* If it looks short enough to be inline data, try to clear extents */
1013         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1014                 max_inline_ea_size = inode_size -
1015                                      (EXT2_GOOD_OLD_INODE_SIZE +
1016                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1017         else
1018                 max_inline_ea_size = 0;
1019         if (EXT2_I_SIZE(inode) <
1020             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1021             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1022                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1023                 dirty = 1;
1024                 goto out;
1025         }
1026
1027         /*
1028          * Too big for inline data, but no evidence of extent tree -
1029          * maybe it's a block map file?  If the mappings all look valid?
1030          */
1031         if (could_be_block_map(fs, inode) &&
1032             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1033 #ifdef WORDS_BIGENDIAN
1034                 int i;
1035
1036                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1037                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1038 #endif
1039
1040                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1041                 dirty = 1;
1042                 goto out;
1043         }
1044
1045         /* Oh well, just clear the busted inode. */
1046         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1047                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1048                 return -1;
1049         }
1050
1051 out:
1052         if (dirty)
1053                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1054
1055         return 0;
1056 }
1057
1058 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1059 {
1060         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1061         dgrp_t start = *group, grp;
1062         blk64_t blocks_to_read = 0;
1063         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1064
1065         if (ctx->readahead_kb == 0)
1066                 goto out;
1067
1068         /* Keep iterating groups until we have enough to readahead */
1069         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1070         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1071                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1072                         continue;
1073                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1074                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1075                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1076                                         inodes_per_block;
1077                 if (blocks_to_read * ctx->fs->blocksize >
1078                     ctx->readahead_kb * 1024)
1079                         break;
1080         }
1081
1082         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1083                                grp - start + 1);
1084         if (err == EAGAIN) {
1085                 ctx->readahead_kb /= 2;
1086                 err = 0;
1087         }
1088
1089 out:
1090         if (err) {
1091                 /* Error; disable itable readahead */
1092                 *group = ctx->fs->group_desc_count;
1093                 *next_ino = ctx->fs->super->s_inodes_count;
1094         } else {
1095                 /*
1096                  * Don't do more readahead until we've reached the first inode
1097                  * of the last inode scan buffer block for the last group.
1098                  */
1099                 *group = grp + 1;
1100                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1101                                      ctx->inode_buffer_blocks :
1102                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1103                                     ctx->fs->blocksize /
1104                                     EXT2_INODE_SIZE(ctx->fs->super);
1105                 inodes_in_group--;
1106                 *next_ino = inodes_in_group -
1107                             (inodes_in_group % inodes_per_buffer) + 1 +
1108                             (grp * ctx->fs->super->s_inodes_per_group);
1109         }
1110 }
1111
1112 /*
1113  * Check if the passed ino is one of the used superblock quota inodes.
1114  *
1115  * Before the quota inodes were journaled, older superblock quota inodes
1116  * were just regular files in the filesystem and not reserved inodes.  This
1117  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1118  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1119  */
1120 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1121 {
1122         enum quota_type qtype;
1123
1124         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1125                 if (*quota_sb_inump(sb, qtype) == ino)
1126                         return 1;
1127
1128         return 0;
1129 }
1130
1131 /*
1132  * Check if the passed ino is one of the reserved quota inodes.
1133  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1134  * inodes.  These inodes may or may not be in use by the quota feature.
1135  */
1136 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1137 {
1138         enum quota_type qtype;
1139
1140         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1141                 if (quota_type2inum(qtype, fs->super) == ino)
1142                         return 1;
1143
1144         return 0;
1145 }
1146
1147 void e2fsck_pass1(e2fsck_t ctx)
1148 {
1149         int     i;
1150         __u64   max_sizes;
1151         ext2_filsys fs = ctx->fs;
1152         ext2_ino_t      ino = 0;
1153         struct ext2_inode *inode = NULL;
1154         ext2_inode_scan scan = NULL;
1155         char            *block_buf = NULL;
1156 #ifdef RESOURCE_TRACK
1157         struct resource_track   rtrack;
1158 #endif
1159         unsigned char   frag, fsize;
1160         struct          problem_context pctx;
1161         struct          scan_callback_struct scan_struct;
1162         struct ext2_super_block *sb = ctx->fs->super;
1163         const char      *old_op;
1164         const char      *eop_next_inode = _("getting next inode from scan");
1165         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1166         int             low_dtime_check = 1;
1167         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1168         unsigned int    bufsize;
1169         int             failed_csum = 0;
1170         ext2_ino_t      ino_threshold = 0;
1171         dgrp_t          ra_group = 0;
1172         struct ea_quota ea_ibody_quota;
1173
1174         init_resource_track(&rtrack, ctx->fs->io);
1175         clear_problem_context(&pctx);
1176
1177         /* If we can do readahead, figure out how many groups to pull in. */
1178         if (!e2fsck_can_readahead(ctx->fs))
1179                 ctx->readahead_kb = 0;
1180         else if (ctx->readahead_kb == ~0ULL)
1181                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1182         pass1_readahead(ctx, &ra_group, &ino_threshold);
1183
1184         if (!(ctx->options & E2F_OPT_PREEN))
1185                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1186
1187         if (ext2fs_has_feature_dir_index(fs->super) &&
1188             !(ctx->options & E2F_OPT_NO)) {
1189                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1190                         ctx->dirs_to_hash = 0;
1191         }
1192
1193 #ifdef MTRACE
1194         mtrace_print("Pass 1");
1195 #endif
1196
1197 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1198
1199         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1200                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1201                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1202                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1203                 max_sizes = (max_sizes * (1UL << i));
1204                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1205         }
1206 #undef EXT2_BPP
1207
1208         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1209         extent_fs = ext2fs_has_feature_extents(sb);
1210         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1211         casefold_fs = ext2fs_has_feature_casefold(sb);
1212
1213         /*
1214          * Allocate bitmaps structures
1215          */
1216         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1217                                                     EXT2FS_BMAP64_RBTREE,
1218                                                     "inode_used_map",
1219                                                     &ctx->inode_used_map);
1220         if (pctx.errcode) {
1221                 pctx.num = 1;
1222                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1223                 ctx->flags |= E2F_FLAG_ABORT;
1224                 return;
1225         }
1226         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1227                         _("directory inode map"),
1228                         EXT2FS_BMAP64_AUTODIR,
1229                         "inode_dir_map", &ctx->inode_dir_map);
1230         if (pctx.errcode) {
1231                 pctx.num = 2;
1232                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1233                 ctx->flags |= E2F_FLAG_ABORT;
1234                 return;
1235         }
1236         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1237                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1238                         "inode_reg_map", &ctx->inode_reg_map);
1239         if (pctx.errcode) {
1240                 pctx.num = 6;
1241                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1242                 ctx->flags |= E2F_FLAG_ABORT;
1243                 return;
1244         }
1245         pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1246                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1247                         "block_found_map", &ctx->block_found_map);
1248         if (pctx.errcode) {
1249                 pctx.num = 1;
1250                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1251                 ctx->flags |= E2F_FLAG_ABORT;
1252                 return;
1253         }
1254         pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1255                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1256                         "block_metadata_map", &ctx->block_metadata_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_setup_icount(ctx, "inode_link_info", 0, NULL,
1264                                            &ctx->inode_link_info);
1265         if (pctx.errcode) {
1266                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1267                 ctx->flags |= E2F_FLAG_ABORT;
1268                 return;
1269         }
1270         bufsize = inode_size;
1271         if (bufsize < sizeof(struct ext2_inode_large))
1272                 bufsize = sizeof(struct ext2_inode_large);
1273         inode = (struct ext2_inode *)
1274                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1275
1276         inodes_to_process = (struct process_inode_block *)
1277                 e2fsck_allocate_memory(ctx,
1278                                        (ctx->process_inode_size *
1279                                         sizeof(struct process_inode_block)),
1280                                        "array of inodes to process");
1281         process_inode_count = 0;
1282
1283         pctx.errcode = ext2fs_init_dblist(fs, 0);
1284         if (pctx.errcode) {
1285                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1286                 ctx->flags |= E2F_FLAG_ABORT;
1287                 goto endit;
1288         }
1289
1290         /*
1291          * If the last orphan field is set, clear it, since the pass1
1292          * processing will automatically find and clear the orphans.
1293          * In the future, we may want to try using the last_orphan
1294          * linked list ourselves, but for now, we clear it so that the
1295          * ext3 mount code won't get confused.
1296          */
1297         if (!(ctx->options & E2F_OPT_READONLY)) {
1298                 if (fs->super->s_last_orphan) {
1299                         fs->super->s_last_orphan = 0;
1300                         ext2fs_mark_super_dirty(fs);
1301                 }
1302         }
1303
1304         mark_table_blocks(ctx);
1305         pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1306                                                 &ctx->block_found_map);
1307         if (pctx.errcode) {
1308                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1309                 ctx->flags |= E2F_FLAG_ABORT;
1310                 goto endit;
1311         }
1312         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1313                                                     "block interate buffer");
1314         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1315                 e2fsck_use_inode_shortcuts(ctx, 1);
1316         e2fsck_intercept_block_allocations(ctx);
1317         old_op = ehandler_operation(_("opening inode scan"));
1318         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1319                                               &scan);
1320         ehandler_operation(old_op);
1321         if (pctx.errcode) {
1322                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1323                 ctx->flags |= E2F_FLAG_ABORT;
1324                 goto endit;
1325         }
1326         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1327                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1328         ctx->stashed_inode = inode;
1329         scan_struct.ctx = ctx;
1330         scan_struct.block_buf = block_buf;
1331         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1332         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1333                                               ctx->fs->group_desc_count)))
1334                 goto endit;
1335         if ((fs->super->s_wtime &&
1336              fs->super->s_wtime < fs->super->s_inodes_count) ||
1337             (fs->super->s_mtime &&
1338              fs->super->s_mtime < fs->super->s_inodes_count) ||
1339             (fs->super->s_mkfs_time &&
1340              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1341                 low_dtime_check = 0;
1342
1343         if (ext2fs_has_feature_mmp(fs->super) &&
1344             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1345             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1346                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1347                                           fs->super->s_mmp_block);
1348
1349         /* Set up ctx->lost_and_found if possible */
1350         (void) e2fsck_get_lost_and_found(ctx, 0);
1351
1352         while (1) {
1353                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1354                         if (e2fsck_mmp_update(fs))
1355                                 fatal_error(ctx, 0);
1356                 }
1357                 old_op = ehandler_operation(eop_next_inode);
1358                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1359                                                           inode, inode_size);
1360                 if (ino > ino_threshold)
1361                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1362                 ehandler_operation(old_op);
1363                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1364                         goto endit;
1365                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1366                         /*
1367                          * If badblocks says badblocks is bad, offer to clear
1368                          * the list, update the in-core bb list, and restart
1369                          * the inode scan.
1370                          */
1371                         if (ino == EXT2_BAD_INO &&
1372                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1373                                         &pctx)) {
1374                                 errcode_t err;
1375
1376                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1377                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1378                                 ctx->fs->badblocks = NULL;
1379                                 err = ext2fs_read_bb_inode(ctx->fs,
1380                                                         &ctx->fs->badblocks);
1381                                 if (err) {
1382                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1383                                                     &pctx);
1384                                         ctx->flags |= E2F_FLAG_ABORT;
1385                                 } else
1386                                         ctx->flags |= E2F_FLAG_RESTART;
1387                                 goto endit;
1388                         }
1389                         if (!ctx->inode_bb_map)
1390                                 alloc_bb_map(ctx);
1391                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1392                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1393                         continue;
1394                 }
1395                 if (pctx.errcode &&
1396                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1397                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1398                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1399                         ctx->flags |= E2F_FLAG_ABORT;
1400                         goto endit;
1401                 }
1402                 if (!ino)
1403                         break;
1404                 pctx.ino = ino;
1405                 pctx.inode = inode;
1406                 ctx->stashed_ino = ino;
1407
1408                 /* Clear trashed inode? */
1409                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1410                     inode->i_links_count > 0 &&
1411                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1412                         pctx.errcode = 0;
1413                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1414                 }
1415                 failed_csum = pctx.errcode != 0;
1416
1417                 /*
1418                  * Check for inodes who might have been part of the
1419                  * orphaned list linked list.  They should have gotten
1420                  * dealt with by now, unless the list had somehow been
1421                  * corrupted.
1422                  *
1423                  * FIXME: In the future, inodes which are still in use
1424                  * (and which are therefore) pending truncation should
1425                  * be handled specially.  Right now we just clear the
1426                  * dtime field, and the normal e2fsck handling of
1427                  * inodes where i_size and the inode blocks are
1428                  * inconsistent is to fix i_size, instead of releasing
1429                  * the extra blocks.  This won't catch the inodes that
1430                  * was at the end of the orphan list, but it's better
1431                  * than nothing.  The right answer is that there
1432                  * shouldn't be any bugs in the orphan list handling.  :-)
1433                  */
1434                 if (inode->i_dtime && low_dtime_check &&
1435                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1436                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1437                                 inode->i_dtime = inode->i_links_count ?
1438                                         0 : ctx->now;
1439                                 e2fsck_write_inode(ctx, ino, inode,
1440                                                    "pass1");
1441                                 failed_csum = 0;
1442                         }
1443                 }
1444
1445                 if (inode->i_links_count) {
1446                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1447                                            ino, inode->i_links_count);
1448                         if (pctx.errcode) {
1449                                 pctx.num = inode->i_links_count;
1450                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1451                                 ctx->flags |= E2F_FLAG_ABORT;
1452                                 goto endit;
1453                         }
1454                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1455                            !quota_inum_is_reserved(fs, ino)) {
1456                         if (!inode->i_dtime && inode->i_mode) {
1457                                 if (fix_problem(ctx,
1458                                             PR_1_ZERO_DTIME, &pctx)) {
1459                                         inode->i_dtime = ctx->now;
1460                                         e2fsck_write_inode(ctx, ino, inode,
1461                                                            "pass1");
1462                                         failed_csum = 0;
1463                                 }
1464                         }
1465                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1466                         continue;
1467                 }
1468
1469                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1470                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1471                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1472                      (!casefold_fs &&
1473                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1474                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1475                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1476                 }
1477
1478                 /* Conflicting inlinedata/extents inode flags? */
1479                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1480                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1481                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1482                                                                inode_size,
1483                                                                &pctx);
1484                         if (res < 0) {
1485                                 /* skip FINISH_INODE_LOOP */
1486                                 continue;
1487                         }
1488                 }
1489
1490                 /* Test for incorrect inline_data flags settings. */
1491                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1492                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1493                         size_t size = 0;
1494
1495                         pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1496                         if (!pctx.errcode &&
1497                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1498                                 ext2fs_set_feature_inline_data(sb);
1499                                 ext2fs_mark_super_dirty(fs);
1500                                 inlinedata_fs = 1;
1501                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1502                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1503                                 /* skip FINISH_INODE_LOOP */
1504                                 continue;
1505                         }
1506                 }
1507
1508                 /* Test for inline data flag but no attr */
1509                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1510                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1511                         size_t size = 0;
1512                         errcode_t err;
1513                         int flags;
1514
1515                         flags = fs->flags;
1516                         if (failed_csum)
1517                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1518                         err = get_inline_data_ea_size(fs, ino, &size);
1519                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1520                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1521
1522                         switch (err) {
1523                         case 0:
1524                                 /* Everything is awesome... */
1525                                 break;
1526                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1527                         case EXT2_ET_BAD_EA_HASH:
1528                         case EXT2_ET_BAD_EA_HEADER:
1529                         case EXT2_ET_EA_BAD_NAME_LEN:
1530                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1531                         case EXT2_ET_EA_KEY_NOT_FOUND:
1532                         case EXT2_ET_EA_NO_SPACE:
1533                         case EXT2_ET_MISSING_EA_FEATURE:
1534                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1535                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1536                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1537                         case EXT2_ET_NO_INLINE_DATA:
1538                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1539                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1540                         case EXT2_ET_EA_INODE_CORRUPTED:
1541                                 /* broken EA or no system.data EA; truncate */
1542                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1543                                                 &pctx)) {
1544                                         err = ext2fs_inode_size_set(fs, inode, 0);
1545                                         if (err) {
1546                                                 pctx.errcode = err;
1547                                                 ctx->flags |= E2F_FLAG_ABORT;
1548                                                 goto endit;
1549                                         }
1550                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1551                                         memset(&inode->i_block, 0,
1552                                                sizeof(inode->i_block));
1553                                         e2fsck_write_inode(ctx, ino, inode,
1554                                                            "pass1");
1555                                         failed_csum = 0;
1556                                 }
1557                                 break;
1558                         default:
1559                                 /* Some other kind of non-xattr error? */
1560                                 pctx.errcode = err;
1561                                 ctx->flags |= E2F_FLAG_ABORT;
1562                                 goto endit;
1563                         }
1564                 }
1565
1566                 /*
1567                  * Test for incorrect extent flag settings.
1568                  *
1569                  * On big-endian machines we must be careful:
1570                  * When the inode is read, the i_block array is not swapped
1571                  * if the extent flag is set.  Therefore if we are testing
1572                  * for or fixing a wrongly-set flag, we must potentially
1573                  * (un)swap before testing, or after fixing.
1574                  */
1575
1576                 /*
1577                  * In this case the extents flag was set when read, so
1578                  * extent_header_verify is ok.  If the inode is cleared,
1579                  * no need to swap... so no extra swapping here.
1580                  */
1581                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1582                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1583                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1584                         if ((ext2fs_extent_header_verify(inode->i_block,
1585                                                  sizeof(inode->i_block)) == 0) &&
1586                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1587                                 ext2fs_set_feature_extents(sb);
1588                                 ext2fs_mark_super_dirty(fs);
1589                                 extent_fs = 1;
1590                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1591                         clear_inode:
1592                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1593                                 if (ino == EXT2_BAD_INO)
1594                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1595                                                                  ino);
1596                                 /* skip FINISH_INODE_LOOP */
1597                                 continue;
1598                         }
1599                 }
1600
1601                 /*
1602                  * For big-endian machines:
1603                  * If the inode didn't have the extents flag set when it
1604                  * was read, then the i_blocks array was swapped.  To test
1605                  * as an extents header, we must swap it back first.
1606                  * IF we then set the extents flag, the entire i_block
1607                  * array must be un/re-swapped to make it proper extents data.
1608                  */
1609                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1610                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1611                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1612                     (LINUX_S_ISREG(inode->i_mode) ||
1613                      LINUX_S_ISDIR(inode->i_mode))) {
1614                         void *ehp;
1615 #ifdef WORDS_BIGENDIAN
1616                         __u32 tmp_block[EXT2_N_BLOCKS];
1617
1618                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1619                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1620                         ehp = tmp_block;
1621 #else
1622                         ehp = inode->i_block;
1623 #endif
1624                         if ((ext2fs_extent_header_verify(ehp,
1625                                          sizeof(inode->i_block)) == 0) &&
1626                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1627                                 inode->i_flags |= EXT4_EXTENTS_FL;
1628 #ifdef WORDS_BIGENDIAN
1629                                 memcpy(inode->i_block, tmp_block,
1630                                        sizeof(inode->i_block));
1631 #endif
1632                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1633                                 failed_csum = 0;
1634                         }
1635                 }
1636
1637                 if (ino == EXT2_BAD_INO) {
1638                         struct process_block_struct pb;
1639
1640                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1641                              inode->i_gid || inode->i_links_count ||
1642                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1643                              inode->i_file_acl) &&
1644                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1645                                 memset(inode, 0, sizeof(struct ext2_inode));
1646                                 e2fsck_write_inode(ctx, ino, inode,
1647                                                    "clear bad inode");
1648                                 failed_csum = 0;
1649                         }
1650
1651                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1652                                                           &pb.fs_meta_blocks);
1653                         if (pctx.errcode) {
1654                                 pctx.num = 4;
1655                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1656                                 ctx->flags |= E2F_FLAG_ABORT;
1657                                 goto endit;
1658                         }
1659                         pb.ino = EXT2_BAD_INO;
1660                         pb.num_blocks = pb.last_block = 0;
1661                         pb.last_db_block = -1;
1662                         pb.num_illegal_blocks = 0;
1663                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1664                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1665                         pb.inode = inode;
1666                         pb.pctx = &pctx;
1667                         pb.ctx = ctx;
1668                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1669                                      block_buf, process_bad_block, &pb);
1670                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1671                         if (pctx.errcode) {
1672                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1673                                 ctx->flags |= E2F_FLAG_ABORT;
1674                                 goto endit;
1675                         }
1676                         if (pb.bbcheck)
1677                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1678                                 ctx->flags |= E2F_FLAG_ABORT;
1679                                 goto endit;
1680                         }
1681                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1682                         clear_problem_context(&pctx);
1683                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1684                         continue;
1685                 } else if (ino == EXT2_ROOT_INO) {
1686                         /*
1687                          * Make sure the root inode is a directory; if
1688                          * not, offer to clear it.  It will be
1689                          * regenerated in pass #3.
1690                          */
1691                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1692                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1693                                         goto clear_inode;
1694                         }
1695                         /*
1696                          * If dtime is set, offer to clear it.  mke2fs
1697                          * version 0.2b created filesystems with the
1698                          * dtime field set for the root and lost+found
1699                          * directories.  We won't worry about
1700                          * /lost+found, since that can be regenerated
1701                          * easily.  But we will fix the root directory
1702                          * as a special case.
1703                          */
1704                         if (inode->i_dtime && inode->i_links_count) {
1705                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1706                                         inode->i_dtime = 0;
1707                                         e2fsck_write_inode(ctx, ino, inode,
1708                                                            "pass1");
1709                                         failed_csum = 0;
1710                                 }
1711                         }
1712                 } else if (ino == EXT2_JOURNAL_INO) {
1713                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1714                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1715                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1716                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1717                                                 &pctx)) {
1718                                         inode->i_mode = LINUX_S_IFREG;
1719                                         e2fsck_write_inode(ctx, ino, inode,
1720                                                            "pass1");
1721                                         failed_csum = 0;
1722                                 }
1723                                 check_blocks(ctx, &pctx, block_buf, NULL);
1724                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1725                                 continue;
1726                         }
1727                         if ((inode->i_links_count ||
1728                              inode->i_blocks || inode->i_block[0]) &&
1729                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1730                                         &pctx)) {
1731                                 memset(inode, 0, inode_size);
1732                                 ext2fs_icount_store(ctx->inode_link_info,
1733                                                     ino, 0);
1734                                 e2fsck_write_inode_full(ctx, ino, inode,
1735                                                         inode_size, "pass1");
1736                                 failed_csum = 0;
1737                         }
1738                 } else if (quota_inum_is_reserved(fs, ino)) {
1739                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1740                         if (ext2fs_has_feature_quota(fs->super) &&
1741                             quota_inum_is_super(fs->super, ino)) {
1742                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1743                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1744                                                         &pctx)) {
1745                                         inode->i_mode = LINUX_S_IFREG;
1746                                         e2fsck_write_inode(ctx, ino, inode,
1747                                                         "pass1");
1748                                         failed_csum = 0;
1749                                 }
1750                                 check_blocks(ctx, &pctx, block_buf, NULL);
1751                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1752                                 continue;
1753                         }
1754                         if ((inode->i_links_count ||
1755                              inode->i_blocks || inode->i_block[0]) &&
1756                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1757                                         &pctx)) {
1758                                 memset(inode, 0, inode_size);
1759                                 ext2fs_icount_store(ctx->inode_link_info,
1760                                                     ino, 0);
1761                                 e2fsck_write_inode_full(ctx, ino, inode,
1762                                                         inode_size, "pass1");
1763                                 failed_csum = 0;
1764                         }
1765                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1766                         problem_t problem = 0;
1767
1768                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1769                         if (ino == EXT2_BOOT_LOADER_INO) {
1770                                 if (LINUX_S_ISDIR(inode->i_mode))
1771                                         problem = PR_1_RESERVED_BAD_MODE;
1772                         } else if (ino == EXT2_RESIZE_INO) {
1773                                 if (inode->i_mode &&
1774                                     !LINUX_S_ISREG(inode->i_mode))
1775                                         problem = PR_1_RESERVED_BAD_MODE;
1776                         } else {
1777                                 if (inode->i_mode != 0)
1778                                         problem = PR_1_RESERVED_BAD_MODE;
1779                         }
1780                         if (problem) {
1781                                 if (fix_problem(ctx, problem, &pctx)) {
1782                                         inode->i_mode = 0;
1783                                         e2fsck_write_inode(ctx, ino, inode,
1784                                                            "pass1");
1785                                         failed_csum = 0;
1786                                 }
1787                         }
1788                         check_blocks(ctx, &pctx, block_buf, NULL);
1789                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1790                         continue;
1791                 }
1792
1793                 if (!inode->i_links_count) {
1794                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1795                         continue;
1796                 }
1797                 /*
1798                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1799                  * deleted files.  Oops.
1800                  *
1801                  * Since all new ext2 implementations get this right,
1802                  * we now assume that the case of non-zero
1803                  * i_links_count and non-zero dtime means that we
1804                  * should keep the file, not delete it.
1805                  *
1806                  */
1807                 if (inode->i_dtime) {
1808                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1809                                 inode->i_dtime = 0;
1810                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1811                                 failed_csum = 0;
1812                         }
1813                 }
1814
1815                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1816                 switch (fs->super->s_creator_os) {
1817                     case EXT2_OS_HURD:
1818                         frag = inode->osd2.hurd2.h_i_frag;
1819                         fsize = inode->osd2.hurd2.h_i_fsize;
1820                         break;
1821                     default:
1822                         frag = fsize = 0;
1823                 }
1824
1825                 if (inode->i_faddr || frag || fsize ||
1826                     (!ext2fs_has_feature_largedir(fs->super) &&
1827                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1828                         mark_inode_bad(ctx, ino);
1829                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1830                     !ext2fs_has_feature_64bit(fs->super) &&
1831                     inode->osd2.linux2.l_i_file_acl_high != 0)
1832                         mark_inode_bad(ctx, ino);
1833                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1834                     !ext2fs_has_feature_huge_file(fs->super) &&
1835                     (inode->osd2.linux2.l_i_blocks_hi != 0))
1836                         mark_inode_bad(ctx, ino);
1837                 if (inode->i_flags & EXT2_IMAGIC_FL) {
1838                         if (imagic_fs) {
1839                                 if (!ctx->inode_imagic_map)
1840                                         alloc_imagic_map(ctx);
1841                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1842                                                          ino);
1843                         } else {
1844                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1845                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
1846                                         e2fsck_write_inode(ctx, ino,
1847                                                            inode, "pass1");
1848                                         failed_csum = 0;
1849                                 }
1850                         }
1851                 }
1852
1853                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1854                 check_is_really_dir(ctx, &pctx, block_buf);
1855
1856                 /*
1857                  * ext2fs_inode_has_valid_blocks2 does not actually look
1858                  * at i_block[] values, so not endian-sensitive here.
1859                  */
1860                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1861                     LINUX_S_ISLNK(inode->i_mode) &&
1862                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1863                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1864                         inode->i_flags &= ~EXT4_EXTENTS_FL;
1865                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1866                         failed_csum = 0;
1867                 }
1868
1869                 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
1870                     add_encrypted_file(ctx, &pctx) < 0)
1871                         goto clear_inode;
1872
1873                 if (LINUX_S_ISDIR(inode->i_mode)) {
1874                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1875                         e2fsck_add_dir_info(ctx, ino, 0);
1876                         ctx->fs_directory_count++;
1877                 } else if (LINUX_S_ISREG (inode->i_mode)) {
1878                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1879                         ctx->fs_regular_count++;
1880                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1881                            e2fsck_pass1_check_device_inode(fs, inode)) {
1882                         check_extents_inlinedata(ctx, &pctx);
1883                         check_immutable(ctx, &pctx);
1884                         check_size(ctx, &pctx);
1885                         ctx->fs_chardev_count++;
1886                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1887                            e2fsck_pass1_check_device_inode(fs, inode)) {
1888                         check_extents_inlinedata(ctx, &pctx);
1889                         check_immutable(ctx, &pctx);
1890                         check_size(ctx, &pctx);
1891                         ctx->fs_blockdev_count++;
1892                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1893                            e2fsck_pass1_check_symlink(fs, ino, inode,
1894                                                       block_buf)) {
1895                         check_immutable(ctx, &pctx);
1896                         ctx->fs_symlinks_count++;
1897                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1898                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1899                                 continue;
1900                         } else if (ext2fs_is_fast_symlink(inode)) {
1901                                 ctx->fs_fast_symlinks_count++;
1902                                 check_blocks(ctx, &pctx, block_buf,
1903                                              &ea_ibody_quota);
1904                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1905                                 continue;
1906                         }
1907                 }
1908                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
1909                          e2fsck_pass1_check_device_inode(fs, inode)) {
1910                         check_extents_inlinedata(ctx, &pctx);
1911                         check_immutable(ctx, &pctx);
1912                         check_size(ctx, &pctx);
1913                         ctx->fs_fifo_count++;
1914                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
1915                            e2fsck_pass1_check_device_inode(fs, inode)) {
1916                         check_extents_inlinedata(ctx, &pctx);
1917                         check_immutable(ctx, &pctx);
1918                         check_size(ctx, &pctx);
1919                         ctx->fs_sockets_count++;
1920                 } else
1921                         mark_inode_bad(ctx, ino);
1922                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1923                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
1924                         if (inode->i_block[EXT2_IND_BLOCK])
1925                                 ctx->fs_ind_count++;
1926                         if (inode->i_block[EXT2_DIND_BLOCK])
1927                                 ctx->fs_dind_count++;
1928                         if (inode->i_block[EXT2_TIND_BLOCK])
1929                                 ctx->fs_tind_count++;
1930                 }
1931                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1932                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
1933                     (inode->i_block[EXT2_IND_BLOCK] ||
1934                      inode->i_block[EXT2_DIND_BLOCK] ||
1935                      inode->i_block[EXT2_TIND_BLOCK] ||
1936                      ext2fs_file_acl_block(fs, inode))) {
1937                         struct process_inode_block *itp;
1938
1939                         itp = &inodes_to_process[process_inode_count];
1940                         itp->ino = ino;
1941                         itp->ea_ibody_quota = ea_ibody_quota;
1942                         if (inode_size < sizeof(struct ext2_inode_large))
1943                                 memcpy(&itp->inode, inode, inode_size);
1944                         else
1945                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
1946                         process_inode_count++;
1947                 } else
1948                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
1949
1950                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1951
1952                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1953                         goto endit;
1954
1955                 if (process_inode_count >= ctx->process_inode_size) {
1956                         process_inodes(ctx, block_buf);
1957
1958                         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1959                                 goto endit;
1960                 }
1961         }
1962         process_inodes(ctx, block_buf);
1963         ext2fs_close_inode_scan(scan);
1964         scan = NULL;
1965
1966         reserve_block_for_root_repair(ctx);
1967         reserve_block_for_lnf_repair(ctx);
1968
1969         /*
1970          * If any extended attribute blocks' reference counts need to
1971          * be adjusted, either up (ctx->refcount_extra), or down
1972          * (ctx->refcount), then fix them.
1973          */
1974         if (ctx->refcount) {
1975                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
1976                 ea_refcount_free(ctx->refcount);
1977                 ctx->refcount = 0;
1978         }
1979         if (ctx->refcount_extra) {
1980                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
1981                                         block_buf, +1);
1982                 ea_refcount_free(ctx->refcount_extra);
1983                 ctx->refcount_extra = 0;
1984         }
1985
1986         if (ctx->ea_block_quota_blocks) {
1987                 ea_refcount_free(ctx->ea_block_quota_blocks);
1988                 ctx->ea_block_quota_blocks = 0;
1989         }
1990
1991         if (ctx->ea_block_quota_inodes) {
1992                 ea_refcount_free(ctx->ea_block_quota_inodes);
1993                 ctx->ea_block_quota_inodes = 0;
1994         }
1995
1996         if (ctx->invalid_bitmaps)
1997                 handle_fs_bad_blocks(ctx);
1998
1999         /* We don't need the block_ea_map any more */
2000         if (ctx->block_ea_map) {
2001                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2002                 ctx->block_ea_map = 0;
2003         }
2004
2005         /* We don't need the encryption policy => ID map any more */
2006         destroy_encryption_policy_map(ctx);
2007
2008         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2009                 clear_problem_context(&pctx);
2010                 pctx.errcode = ext2fs_create_resize_inode(fs);
2011                 if (pctx.errcode) {
2012                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2013                                          &pctx)) {
2014                                 ctx->flags |= E2F_FLAG_ABORT;
2015                                 goto endit;
2016                         }
2017                         pctx.errcode = 0;
2018                 }
2019                 if (!pctx.errcode) {
2020                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2021                                           "recreate inode");
2022                         inode->i_mtime = ctx->now;
2023                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2024                                            "recreate inode");
2025                 }
2026                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2027         }
2028
2029         if (ctx->flags & E2F_FLAG_RESTART) {
2030                 /*
2031                  * Only the master copy of the superblock and block
2032                  * group descriptors are going to be written during a
2033                  * restart, so set the superblock to be used to be the
2034                  * master superblock.
2035                  */
2036                 ctx->use_superblock = 0;
2037                 goto endit;
2038         }
2039
2040         if (ctx->block_dup_map) {
2041                 if (ctx->options & E2F_OPT_PREEN) {
2042                         clear_problem_context(&pctx);
2043                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2044                 }
2045                 e2fsck_pass1_dupblocks(ctx, block_buf);
2046         }
2047         ctx->flags |= E2F_FLAG_ALLOC_OK;
2048 endit:
2049         e2fsck_use_inode_shortcuts(ctx, 0);
2050         ext2fs_free_mem(&inodes_to_process);
2051         inodes_to_process = 0;
2052
2053         if (scan)
2054                 ext2fs_close_inode_scan(scan);
2055         if (block_buf)
2056                 ext2fs_free_mem(&block_buf);
2057         if (inode)
2058                 ext2fs_free_mem(&inode);
2059
2060         /*
2061          * The l+f inode may have been cleared, so zap it now and
2062          * later passes will recalculate it if necessary
2063          */
2064         ctx->lost_and_found = 0;
2065
2066         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2067                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2068         else
2069                 ctx->invalid_bitmaps++;
2070 }
2071 #undef FINISH_INODE_LOOP
2072
2073 /*
2074  * When the inode_scan routines call this callback at the end of the
2075  * glock group, call process_inodes.
2076  */
2077 static errcode_t scan_callback(ext2_filsys fs,
2078                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
2079                                dgrp_t group, void * priv_data)
2080 {
2081         struct scan_callback_struct *scan_struct;
2082         e2fsck_t ctx;
2083
2084         scan_struct = (struct scan_callback_struct *) priv_data;
2085         ctx = scan_struct->ctx;
2086
2087         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf);
2088
2089         if (ctx->progress)
2090                 if ((ctx->progress)(ctx, 1, group+1,
2091                                     ctx->fs->group_desc_count))
2092                         return EXT2_ET_CANCEL_REQUESTED;
2093
2094         return 0;
2095 }
2096
2097 /*
2098  * Process the inodes in the "inodes to process" list.
2099  */
2100 static void process_inodes(e2fsck_t ctx, char *block_buf)
2101 {
2102         int                     i;
2103         struct ext2_inode       *old_stashed_inode;
2104         ext2_ino_t              old_stashed_ino;
2105         const char              *old_operation;
2106         char                    buf[80];
2107         struct problem_context  pctx;
2108
2109 #if 0
2110         printf("begin process_inodes: ");
2111 #endif
2112         if (process_inode_count == 0)
2113                 return;
2114         old_operation = ehandler_operation(0);
2115         old_stashed_inode = ctx->stashed_inode;
2116         old_stashed_ino = ctx->stashed_ino;
2117         qsort(inodes_to_process, process_inode_count,
2118                       sizeof(struct process_inode_block), process_inode_cmp);
2119         clear_problem_context(&pctx);
2120         for (i=0; i < process_inode_count; i++) {
2121                 pctx.inode = ctx->stashed_inode =
2122                         (struct ext2_inode *) &inodes_to_process[i].inode;
2123                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2124
2125 #if 0
2126                 printf("%u ", pctx.ino);
2127 #endif
2128                 sprintf(buf, _("reading indirect blocks of inode %u"),
2129                         pctx.ino);
2130                 ehandler_operation(buf);
2131                 check_blocks(ctx, &pctx, block_buf,
2132                              &inodes_to_process[i].ea_ibody_quota);
2133                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2134                         break;
2135         }
2136         ctx->stashed_inode = old_stashed_inode;
2137         ctx->stashed_ino = old_stashed_ino;
2138         process_inode_count = 0;
2139 #if 0
2140         printf("end process inodes\n");
2141 #endif
2142         ehandler_operation(old_operation);
2143 }
2144
2145 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
2146 {
2147         const struct process_inode_block *ib_a =
2148                 (const struct process_inode_block *) a;
2149         const struct process_inode_block *ib_b =
2150                 (const struct process_inode_block *) b;
2151         int     ret;
2152
2153         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
2154                ib_b->inode.i_block[EXT2_IND_BLOCK]);
2155         if (ret == 0)
2156                 /*
2157                  * We only call process_inodes() for non-extent
2158                  * inodes, so it's OK to pass NULL to
2159                  * ext2fs_file_acl_block() here.
2160                  */
2161                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
2162                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
2163         if (ret == 0)
2164                 ret = ib_a->ino - ib_b->ino;
2165         return ret;
2166 }
2167
2168 /*
2169  * Mark an inode as being bad in some what
2170  */
2171 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
2172 {
2173         struct          problem_context pctx;
2174
2175         if (!ctx->inode_bad_map) {
2176                 clear_problem_context(&pctx);
2177
2178                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2179                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
2180                                 "inode_bad_map", &ctx->inode_bad_map);
2181                 if (pctx.errcode) {
2182                         pctx.num = 3;
2183                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2184                         /* Should never get here */
2185                         ctx->flags |= E2F_FLAG_ABORT;
2186                         return;
2187                 }
2188         }
2189         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
2190 }
2191
2192 /*
2193  * This procedure will allocate the inode "bb" (badblock) map table
2194  */
2195 static void alloc_bb_map(e2fsck_t ctx)
2196 {
2197         struct          problem_context pctx;
2198
2199         clear_problem_context(&pctx);
2200         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2201                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
2202                         "inode_bb_map", &ctx->inode_bb_map);
2203         if (pctx.errcode) {
2204                 pctx.num = 4;
2205                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2206                 /* Should never get here */
2207                 ctx->flags |= E2F_FLAG_ABORT;
2208                 return;
2209         }
2210 }
2211
2212 /*
2213  * This procedure will allocate the inode imagic table
2214  */
2215 static void alloc_imagic_map(e2fsck_t ctx)
2216 {
2217         struct          problem_context pctx;
2218
2219         clear_problem_context(&pctx);
2220         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2221                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
2222                         "inode_imagic_map", &ctx->inode_imagic_map);
2223         if (pctx.errcode) {
2224                 pctx.num = 5;
2225                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2226                 /* Should never get here */
2227                 ctx->flags |= E2F_FLAG_ABORT;
2228                 return;
2229         }
2230 }
2231
2232 /*
2233  * Marks a block as in use, setting the dup_map if it's been set
2234  * already.  Called by process_block and process_bad_block.
2235  *
2236  * WARNING: Assumes checks have already been done to make sure block
2237  * is valid.  This is true in both process_block and process_bad_block.
2238  */
2239 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
2240 {
2241         struct          problem_context pctx;
2242
2243         clear_problem_context(&pctx);
2244
2245         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
2246                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
2247                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
2248                         return;
2249                 }
2250                 if (!ctx->block_dup_map) {
2251                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
2252                                         _("multiply claimed block map"),
2253                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
2254                                         &ctx->block_dup_map);
2255                         if (pctx.errcode) {
2256                                 pctx.num = 3;
2257                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
2258                                             &pctx);
2259                                 /* Should never get here */
2260                                 ctx->flags |= E2F_FLAG_ABORT;
2261                                 return;
2262                         }
2263                 }
2264                 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
2265         } else {
2266                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
2267         }
2268 }
2269
2270 /*
2271  * When cluster size is greater than one block, it is caller's responsibility
2272  * to make sure block parameter starts at a cluster boundary.
2273  */
2274 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
2275                                       unsigned int num)
2276 {
2277         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
2278                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
2279         else {
2280                 unsigned int i;
2281
2282                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
2283                         mark_block_used(ctx, block + i);
2284         }
2285 }
2286
2287 /*
2288  * Adjust the extended attribute block's reference counts at the end
2289  * of pass 1, either by subtracting out references for EA blocks that
2290  * are still referenced in ctx->refcount, or by adding references for
2291  * EA blocks that had extra references as accounted for in
2292  * ctx->refcount_extra.
2293  */
2294 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
2295                                     char *block_buf, int adjust_sign)
2296 {
2297         struct ext2_ext_attr_header     *header;
2298         struct problem_context          pctx;
2299         ext2_filsys                     fs = ctx->fs;
2300         blk64_t                         blk;
2301         __u32                           should_be;
2302         ea_value_t                      count;
2303
2304         clear_problem_context(&pctx);
2305
2306         ea_refcount_intr_begin(refcount);
2307         while (1) {
2308                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
2309                         break;
2310                 pctx.blk = blk;
2311                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
2312                                                      pctx.ino);
2313                 if (pctx.errcode) {
2314                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
2315                         return;
2316                 }
2317                 header = (struct ext2_ext_attr_header *) block_buf;
2318                 pctx.blkcount = header->h_refcount;
2319                 should_be = header->h_refcount + adjust_sign * (int)count;
2320                 pctx.num = should_be;
2321                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
2322                         header->h_refcount = should_be;
2323                         pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
2324                                                              block_buf,
2325                                                              pctx.ino);
2326                         if (pctx.errcode) {
2327                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
2328                                             &pctx);
2329                                 continue;
2330                         }
2331                 }
2332         }
2333 }
2334
2335 /*
2336  * Handle processing the extended attribute blocks
2337  */
2338 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
2339                            char *block_buf, struct ea_quota *ea_block_quota)
2340 {
2341         ext2_filsys fs = ctx->fs;
2342         ext2_ino_t      ino = pctx->ino;
2343         struct ext2_inode *inode = pctx->inode;
2344         blk64_t         blk;
2345         char *          end;
2346         struct ext2_ext_attr_header *header;
2347         struct ext2_ext_attr_entry *first, *entry;
2348         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
2349         __u64           quota_inodes = 0;
2350         region_t        region = 0;
2351         int             failed_csum = 0;
2352
2353         ea_block_quota->blocks = 0;
2354         ea_block_quota->inodes = 0;
2355
2356         blk = ext2fs_file_acl_block(fs, inode);
2357         if (blk == 0)
2358                 return 0;
2359
2360         /*
2361          * If the Extended attribute flag isn't set, then a non-zero
2362          * file acl means that the inode is corrupted.
2363          *
2364          * Or if the extended attribute block is an invalid block,
2365          * then the inode is also corrupted.
2366          */
2367         if (!ext2fs_has_feature_xattr(fs->super) ||
2368             (blk < fs->super->s_first_data_block) ||
2369             (blk >= ext2fs_blocks_count(fs->super))) {
2370                 mark_inode_bad(ctx, ino);
2371                 return 0;
2372         }
2373
2374         /* If ea bitmap hasn't been allocated, create it */
2375         if (!ctx->block_ea_map) {
2376                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
2377                                         _("ext attr block map"),
2378                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
2379                                         &ctx->block_ea_map);
2380                 if (pctx->errcode) {
2381                         pctx->num = 2;
2382                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
2383                         ctx->flags |= E2F_FLAG_ABORT;
2384                         return 0;
2385                 }
2386         }
2387
2388         /* Create the EA refcount structure if necessary */
2389         if (!ctx->refcount) {
2390                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
2391                 if (pctx->errcode) {
2392                         pctx->num = 1;
2393                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2394                         ctx->flags |= E2F_FLAG_ABORT;
2395                         return 0;
2396                 }
2397         }
2398
2399 #if 0
2400         /* Debugging text */
2401         printf("Inode %u has EA block %u\n", ino, blk);
2402 #endif
2403
2404         /* Have we seen this EA block before? */
2405         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
2406                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
2407                 ea_block_quota->inodes = 0;
2408
2409                 if (ctx->ea_block_quota_blocks) {
2410                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
2411                                           &quota_blocks);
2412                         if (quota_blocks)
2413                                 ea_block_quota->blocks = quota_blocks;
2414                 }
2415
2416                 if (ctx->ea_block_quota_inodes)
2417                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
2418                                           &ea_block_quota->inodes);
2419
2420                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
2421                         return 1;
2422                 /* Ooops, this EA was referenced more than it stated */
2423                 if (!ctx->refcount_extra) {
2424                         pctx->errcode = ea_refcount_create(0,
2425                                            &ctx->refcount_extra);
2426                         if (pctx->errcode) {
2427                                 pctx->num = 2;
2428                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2429                                 ctx->flags |= E2F_FLAG_ABORT;
2430                                 return 0;
2431                         }
2432                 }
2433                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
2434                 return 1;
2435         }
2436
2437         /*
2438          * OK, we haven't seen this EA block yet.  So we need to
2439          * validate it
2440          */
2441         pctx->blk = blk;
2442         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
2443         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
2444                 pctx->errcode = 0;
2445                 failed_csum = 1;
2446         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
2447                 pctx->errcode = 0;
2448
2449         if (pctx->errcode &&
2450             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
2451                 pctx->errcode = 0;
2452                 goto clear_extattr;
2453         }
2454         header = (struct ext2_ext_attr_header *) block_buf;
2455         pctx->blk = ext2fs_file_acl_block(fs, inode);
2456         if (((ctx->ext_attr_ver == 1) &&
2457              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
2458             ((ctx->ext_attr_ver == 2) &&
2459              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
2460                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
2461                         goto clear_extattr;
2462         }
2463
2464         if (header->h_blocks != 1) {
2465                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
2466                         goto clear_extattr;
2467         }
2468
2469         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
2470                 goto clear_extattr;
2471
2472         region = region_create(0, fs->blocksize);
2473         if (!region) {
2474                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
2475                 ctx->flags |= E2F_FLAG_ABORT;
2476                 return 0;
2477         }
2478         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
2479                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2480                         goto clear_extattr;
2481         }
2482
2483         first = (struct ext2_ext_attr_entry *)(header+1);
2484         end = block_buf + fs->blocksize;
2485         entry = first;
2486         while ((char *)entry < end && *(__u32 *)entry) {
2487                 __u32 hash;
2488
2489                 if (region_allocate(region, (char *)entry - (char *)header,
2490                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
2491                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2492                                 goto clear_extattr;
2493                         break;
2494                 }
2495                 if ((ctx->ext_attr_ver == 1 &&
2496                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
2497                     (ctx->ext_attr_ver == 2 &&
2498                      entry->e_name_index == 0)) {
2499                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
2500                                 goto clear_extattr;
2501                         break;
2502                 }
2503                 if (entry->e_value_inum == 0) {
2504                         if (entry->e_value_offs + entry->e_value_size >
2505                             fs->blocksize) {
2506                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
2507                                         goto clear_extattr;
2508                                 break;
2509                         }
2510                         if (entry->e_value_size &&
2511                             region_allocate(region, entry->e_value_offs,
2512                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
2513                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
2514                                                 pctx))
2515                                         goto clear_extattr;
2516                         }
2517
2518                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
2519                                                           entry->e_value_offs);
2520
2521                         if (entry->e_hash != hash) {
2522                                 pctx->num = entry->e_hash;
2523                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
2524                                         goto clear_extattr;
2525                                 entry->e_hash = hash;
2526                         }
2527                 } else {
2528                         problem_t problem;
2529                         blk64_t entry_quota_blocks;
2530
2531                         problem = check_large_ea_inode(ctx, entry, pctx,
2532                                                        &entry_quota_blocks);
2533                         if (problem && fix_problem(ctx, problem, pctx))
2534                                 goto clear_extattr;
2535
2536                         quota_blocks += entry_quota_blocks;
2537                         quota_inodes++;
2538                 }
2539
2540                 entry = EXT2_EXT_ATTR_NEXT(entry);
2541         }
2542         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
2543                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2544                         goto clear_extattr;
2545         }
2546         region_free(region);
2547
2548         /*
2549          * We only get here if there was no other errors that were fixed.
2550          * If there was a checksum fail, ask to correct it.
2551          */
2552         if (failed_csum &&
2553             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
2554                 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
2555                                                        pctx->ino);
2556                 if (pctx->errcode)
2557                         return 0;
2558         }
2559
2560         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
2561                 if (!ctx->ea_block_quota_blocks) {
2562                         pctx->errcode = ea_refcount_create(0,
2563                                                 &ctx->ea_block_quota_blocks);
2564                         if (pctx->errcode) {
2565                                 pctx->num = 3;
2566                                 goto refcount_fail;
2567                         }
2568                 }
2569                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
2570                                   quota_blocks);
2571         }
2572
2573         if (quota_inodes) {
2574                 if (!ctx->ea_block_quota_inodes) {
2575                         pctx->errcode = ea_refcount_create(0,
2576                                                 &ctx->ea_block_quota_inodes);
2577                         if (pctx->errcode) {
2578                                 pctx->num = 4;
2579 refcount_fail:
2580                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2581                                 ctx->flags |= E2F_FLAG_ABORT;
2582                                 return 0;
2583                         }
2584                 }
2585
2586                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
2587                                   quota_inodes);
2588         }
2589         ea_block_quota->blocks = quota_blocks;
2590         ea_block_quota->inodes = quota_inodes;
2591
2592         inc_ea_inode_refs(ctx, pctx, first, end);
2593         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
2594         mark_block_used(ctx, blk);
2595         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
2596         return 1;
2597
2598 clear_extattr:
2599         if (region)
2600                 region_free(region);
2601         ext2fs_file_acl_block_set(fs, inode, 0);
2602         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
2603         return 0;
2604 }
2605
2606 /* Returns 1 if bad htree, 0 if OK */
2607 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
2608                         ext2_ino_t ino, struct ext2_inode *inode,
2609                         char *block_buf)
2610 {
2611         struct ext2_dx_root_info        *root;
2612         ext2_filsys                     fs = ctx->fs;
2613         errcode_t                       retval;
2614         blk64_t                         blk;
2615
2616         if ((!LINUX_S_ISDIR(inode->i_mode) &&
2617              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
2618             (!ext2fs_has_feature_dir_index(fs->super) &&
2619              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
2620                 return 1;
2621
2622         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
2623
2624         if ((pctx->errcode) ||
2625             (blk == 0) ||
2626             (blk < fs->super->s_first_data_block) ||
2627             (blk >= ext2fs_blocks_count(fs->super))) {
2628                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2629                         return 1;
2630                 else
2631                         return 0;
2632         }
2633
2634         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
2635         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2636                 return 1;
2637
2638         /* XXX should check that beginning matches a directory */
2639         root = (struct ext2_dx_root_info *) (block_buf + 24);
2640
2641         if ((root->reserved_zero || root->info_length < 8) &&
2642             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2643                 return 1;
2644
2645         pctx->num = root->hash_version;
2646         if ((root->hash_version != EXT2_HASH_LEGACY) &&
2647             (root->hash_version != EXT2_HASH_HALF_MD4) &&
2648             (root->hash_version != EXT2_HASH_TEA) &&
2649             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
2650                 return 1;
2651
2652         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
2653             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
2654                 return 1;
2655
2656         pctx->num = root->indirect_levels;
2657         if ((root->indirect_levels >= ext2_dir_htree_level(fs)) &&
2658             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
2659                 return 1;
2660
2661         return 0;
2662 }
2663
2664 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
2665                         struct ext2_inode *inode, int restart_flag,
2666                         const char *source)
2667 {
2668         inode->i_flags = 0;
2669         inode->i_links_count = 0;
2670         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
2671         inode->i_dtime = ctx->now;
2672
2673         /*
2674          * If a special inode has such rotten block mappings that we
2675          * want to clear the whole inode, be sure to actually zap
2676          * the block maps because i_links_count isn't checked for
2677          * special inodes, and we'll end up right back here the next
2678          * time we run fsck.
2679          */
2680         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
2681                 memset(inode->i_block, 0, sizeof(inode->i_block));
2682
2683         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
2684         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
2685         if (ctx->inode_reg_map)
2686                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
2687         if (ctx->inode_bad_map)
2688                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
2689
2690         /*
2691          * If the inode was partially accounted for before processing
2692          * was aborted, we need to restart the pass 1 scan.
2693          */
2694         ctx->flags |= restart_flag;
2695
2696         if (ino == EXT2_BAD_INO)
2697                 memset(inode, 0, sizeof(struct ext2_inode));
2698
2699         e2fsck_write_inode(ctx, ino, inode, source);
2700 }
2701
2702 /*
2703  * Use the multiple-blocks reclamation code to fix alignment problems in
2704  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
2705  * physical cluster, and we want the block offsets within that cluster to
2706  * line up.
2707  */
2708 static int has_unaligned_cluster_map(e2fsck_t ctx,
2709                                      blk64_t last_pblk, blk64_t last_lblk,
2710                                      blk64_t pblk, blk64_t lblk)
2711 {
2712         blk64_t cluster_mask;
2713
2714         if (!ctx->fs->cluster_ratio_bits)
2715                 return 0;
2716         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
2717
2718         /*
2719          * If the block in the logical cluster doesn't align with the block in
2720          * the physical cluster...
2721          */
2722         if ((lblk & cluster_mask) != (pblk & cluster_mask))
2723                 return 1;
2724
2725         /*
2726          * If we cross a physical cluster boundary within a logical cluster...
2727          */
2728         if (last_pblk && (lblk & cluster_mask) != 0 &&
2729             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
2730             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
2731                 return 1;
2732
2733         return 0;
2734 }
2735
2736 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
2737                              struct process_block_struct *pb,
2738                              blk64_t start_block, blk64_t end_block,
2739                              blk64_t eof_block,
2740                              ext2_extent_handle_t ehandle,
2741                              int try_repairs)
2742 {
2743         struct ext2fs_extent    extent;
2744         blk64_t                 blk, last_lblk;
2745         unsigned int            i, n;
2746         int                     is_dir, is_leaf;
2747         problem_t               problem;
2748         struct ext2_extent_info info;
2749         int                     failed_csum = 0;
2750
2751         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
2752                 failed_csum = 1;
2753
2754         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
2755         if (pctx->errcode)
2756                 return;
2757         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
2758             !pb->eti.force_rebuild) {
2759                 struct extent_tree_level *etl;
2760
2761                 etl = pb->eti.ext_info + info.curr_level;
2762                 etl->num_extents += info.num_entries;
2763                 etl->max_extents += info.max_entries;
2764                 /*
2765                  * Implementation wart: Splitting extent blocks when appending
2766                  * will leave the old block with one free entry.  Therefore
2767                  * unless the node is totally full, pretend that a non-root
2768                  * extent block can hold one fewer entry than it actually does,
2769                  * so that we don't repeatedly rebuild the extent tree.
2770                  */
2771                 if (info.curr_level && info.num_entries < info.max_entries)
2772                         etl->max_extents--;
2773         }
2774
2775         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
2776                                           &extent);
2777         while ((pctx->errcode == 0 ||
2778                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
2779                info.num_entries-- > 0) {
2780                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
2781                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
2782                 last_lblk = extent.e_lblk + extent.e_len - 1;
2783
2784                 problem = 0;
2785                 pctx->blk = extent.e_pblk;
2786                 pctx->blk2 = extent.e_lblk;
2787                 pctx->num = extent.e_len;
2788                 pctx->blkcount = extent.e_lblk + extent.e_len;
2789
2790                 if (extent.e_pblk == 0 ||
2791                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
2792                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
2793                         problem = PR_1_EXTENT_BAD_START_BLK;
2794                 else if (extent.e_lblk < start_block)
2795                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
2796                 else if ((end_block && last_lblk > end_block) &&
2797                          !(last_lblk > eof_block &&
2798                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
2799                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
2800                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
2801                 else if (is_leaf && extent.e_len == 0)
2802                         problem = PR_1_EXTENT_LENGTH_ZERO;
2803                 else if (is_leaf &&
2804                          (extent.e_pblk + extent.e_len) >
2805                          ext2fs_blocks_count(ctx->fs->super))
2806                         problem = PR_1_EXTENT_ENDS_BEYOND;
2807                 else if (is_leaf && is_dir &&
2808                          ((extent.e_lblk + extent.e_len) >
2809                           (1U << (21 - ctx->fs->super->s_log_block_size))))
2810                         problem = PR_1_TOOBIG_DIR;
2811
2812                 if (is_leaf && problem == 0 && extent.e_len > 0) {
2813 #if 0
2814                         printf("extent_region(ino=%u, expect=%llu, "
2815                                "lblk=%llu, len=%u)\n",
2816                                pb->ino, pb->next_lblock,
2817                                extent.e_lblk, extent.e_len);
2818 #endif
2819                         if (extent.e_lblk < pb->next_lblock)
2820                                 problem = PR_1_EXTENT_COLLISION;
2821                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
2822                                 pb->next_lblock = extent.e_lblk + extent.e_len;
2823                 }
2824
2825                 /*
2826                  * Uninitialized blocks in a directory?  Clear the flag and
2827                  * we'll interpret the blocks later.
2828                  */
2829                 if (try_repairs && is_dir && problem == 0 &&
2830                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
2831                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
2832                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
2833                         pb->inode_modified = 1;
2834                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
2835                                                               &extent);
2836                         if (pctx->errcode)
2837                                 return;
2838                         failed_csum = 0;
2839                 }
2840 #ifdef CONFIG_DEVELOPER_FEATURES
2841                 if (try_repairs && !is_dir && problem == 0 &&
2842                     (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
2843                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
2844                     fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
2845                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
2846                         pb->inode_modified = 1;
2847                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
2848                                                               &extent);
2849                         if (pctx->errcode)
2850                                 return;
2851                         failed_csum = 0;
2852                 }
2853 #endif
2854                 if (try_repairs && problem) {
2855 report_problem:
2856                         if (fix_problem(ctx, problem, pctx)) {
2857                                 if (ctx->invalid_bitmaps) {
2858                                         /*
2859                                          * If fsck knows the bitmaps are bad,
2860                                          * skip to the next extent and
2861                                          * try to clear this extent again
2862                                          * after fixing the bitmaps, by
2863                                          * restarting fsck.
2864                                          */
2865                                         pctx->errcode = ext2fs_extent_get(
2866                                                           ehandle,
2867                                                           EXT2_EXTENT_NEXT_SIB,
2868                                                           &extent);
2869                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
2870                                         if (pctx->errcode ==
2871                                                     EXT2_ET_NO_CURRENT_NODE) {
2872                                                 pctx->errcode = 0;
2873                                                 break;
2874                                         }
2875                                         continue;
2876                                 }
2877                                 e2fsck_read_bitmaps(ctx);
2878                                 pb->inode_modified = 1;
2879                                 pctx->errcode =
2880                                         ext2fs_extent_delete(ehandle, 0);
2881                                 if (pctx->errcode) {
2882                                         pctx->str = "ext2fs_extent_delete";
2883                                         return;
2884                                 }
2885                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
2886                                 if (pctx->errcode &&
2887                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
2888                                         pctx->str = "ext2fs_extent_fix_parents";
2889                                         return;
2890                                 }
2891                                 pctx->errcode = ext2fs_extent_get(ehandle,
2892                                                                   EXT2_EXTENT_CURRENT,
2893                                                                   &extent);
2894                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
2895                                         pctx->errcode = 0;
2896                                         break;
2897                                 }
2898                                 failed_csum = 0;
2899                                 continue;
2900                         }
2901                         goto next;
2902                 }
2903
2904                 if (!is_leaf) {
2905                         blk64_t lblk = extent.e_lblk;
2906                         int next_try_repairs = 1;
2907
2908                         blk = extent.e_pblk;
2909
2910                         /*
2911                          * If this lower extent block collides with critical
2912                          * metadata, don't try to repair the damage.  Pass 1b
2913                          * will reallocate the block; then we can try again.
2914                          */
2915                         if (pb->ino != EXT2_RESIZE_INO &&
2916                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
2917                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
2918                                                       extent.e_pblk)) {
2919                                 next_try_repairs = 0;
2920                                 pctx->blk = blk;
2921                                 fix_problem(ctx,
2922                                             PR_1_CRITICAL_METADATA_COLLISION,
2923                                             pctx);
2924                                 if ((ctx->options & E2F_OPT_NO) == 0)
2925                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
2926                         }
2927                         pctx->errcode = ext2fs_extent_get(ehandle,
2928                                                   EXT2_EXTENT_DOWN, &extent);
2929                         if (pctx->errcode &&
2930                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
2931                                 pctx->str = "EXT2_EXTENT_DOWN";
2932                                 problem = PR_1_EXTENT_HEADER_INVALID;
2933                                 if (!next_try_repairs)
2934                                         return;
2935                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
2936                                         goto report_problem;
2937                                 return;
2938                         }
2939                         /* The next extent should match this index's logical start */
2940                         if (extent.e_lblk != lblk) {
2941                                 struct ext2_extent_info e_info;
2942
2943                                 ext2fs_extent_get_info(ehandle, &e_info);
2944                                 pctx->blk = lblk;
2945                                 pctx->blk2 = extent.e_lblk;
2946                                 pctx->num = e_info.curr_level - 1;
2947                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
2948                                 if (fix_problem(ctx, problem, pctx)) {
2949                                         pb->inode_modified = 1;
2950                                         pctx->errcode =
2951                                                 ext2fs_extent_fix_parents(ehandle);
2952                                         if (pctx->errcode) {
2953                                                 pctx->str = "ext2fs_extent_fix_parents";
2954                                                 return;
2955                                         }
2956                                 }
2957                         }
2958                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
2959                                          last_lblk, eof_block, ehandle,
2960                                          next_try_repairs);
2961                         if (pctx->errcode)
2962                                 return;
2963                         pctx->errcode = ext2fs_extent_get(ehandle,
2964                                                   EXT2_EXTENT_UP, &extent);
2965                         if (pctx->errcode) {
2966                                 pctx->str = "EXT2_EXTENT_UP";
2967                                 return;
2968                         }
2969                         mark_block_used(ctx, blk);
2970                         pb->num_blocks++;
2971                         goto next;
2972                 }
2973
2974                 if ((pb->previous_block != 0) &&
2975                     (pb->previous_block+1 != extent.e_pblk)) {
2976                         if (ctx->options & E2F_OPT_FRAGCHECK) {
2977                                 char type = '?';
2978
2979                                 if (pb->is_dir)
2980                                         type = 'd';
2981                                 else if (pb->is_reg)
2982                                         type = 'f';
2983
2984                                 printf(("%6lu(%c): expecting %6lu "
2985                                         "actual extent "
2986                                         "phys %6lu log %lu len %lu\n"),
2987                                        (unsigned long) pctx->ino, type,
2988                                        (unsigned long) pb->previous_block+1,
2989                                        (unsigned long) extent.e_pblk,
2990                                        (unsigned long) extent.e_lblk,
2991                                        (unsigned long) extent.e_len);
2992                         }
2993                         pb->fragmented = 1;
2994                 }
2995                 /*
2996                  * If we notice a gap in the logical block mappings of an
2997                  * extent-mapped directory, offer to close the hole by
2998                  * moving the logical block down, otherwise we'll go mad in
2999                  * pass 3 allocating empty directory blocks to fill the hole.
3000                  */
3001                 if (try_repairs && is_dir &&
3002                     pb->last_block + 1 < extent.e_lblk) {
3003                         blk64_t new_lblk;
3004
3005                         new_lblk = pb->last_block + 1;
3006                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3007                                 new_lblk = ((new_lblk +
3008                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3009                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3010                                            (extent.e_pblk &
3011                                             EXT2FS_CLUSTER_MASK(ctx->fs));
3012                         pctx->blk = extent.e_lblk;
3013                         pctx->blk2 = new_lblk;
3014                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3015                                 extent.e_lblk = new_lblk;
3016                                 pb->inode_modified = 1;
3017                                 pctx->errcode = ext2fs_extent_replace(ehandle,
3018                                                                 0, &extent);
3019                                 if (pctx->errcode) {
3020                                         pctx->errcode = 0;
3021                                         goto alloc_later;
3022                                 }
3023                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3024                                 if (pctx->errcode)
3025                                         goto failed_add_dir_block;
3026                                 pctx->errcode = ext2fs_extent_goto(ehandle,
3027                                                                 extent.e_lblk);
3028                                 if (pctx->errcode)
3029                                         goto failed_add_dir_block;
3030                                 last_lblk = extent.e_lblk + extent.e_len - 1;
3031                                 failed_csum = 0;
3032                         }
3033                 }
3034 alloc_later:
3035                 if (is_dir) {
3036                         while (++pb->last_db_block <
3037                                (e2_blkcnt_t) extent.e_lblk) {
3038                                 pctx->errcode = ext2fs_add_dir_block2(
3039                                                         ctx->fs->dblist,
3040                                                         pb->ino, 0,
3041                                                         pb->last_db_block);
3042                                 if (pctx->errcode) {
3043                                         pctx->blk = 0;
3044                                         pctx->num = pb->last_db_block;
3045                                         goto failed_add_dir_block;
3046                                 }
3047                         }
3048
3049                         for (i = 0; i < extent.e_len; i++) {
3050                                 pctx->errcode = ext2fs_add_dir_block2(
3051                                                         ctx->fs->dblist,
3052                                                         pctx->ino,
3053                                                         extent.e_pblk + i,
3054                                                         extent.e_lblk + i);
3055                                 if (pctx->errcode) {
3056                                         pctx->blk = extent.e_pblk + i;
3057                                         pctx->num = extent.e_lblk + i;
3058                                 failed_add_dir_block:
3059                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3060                                         /* Should never get here */
3061                                         ctx->flags |= E2F_FLAG_ABORT;
3062                                         return;
3063                                 }
3064                         }
3065                         if (extent.e_len > 0)
3066                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3067                 }
3068                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3069                                               pb->last_block,
3070                                               extent.e_pblk,
3071                                               extent.e_lblk)) {
3072                         for (i = 0; i < extent.e_len; i++) {
3073                                 pctx->blk = extent.e_lblk + i;
3074                                 pctx->blk2 = extent.e_pblk + i;
3075                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3076                                 mark_block_used(ctx, extent.e_pblk + i);
3077                                 mark_block_used(ctx, extent.e_pblk + i);
3078                         }
3079                 }
3080
3081                 /*
3082                  * Check whether first cluster got marked in previous iteration.
3083                  */
3084                 if (ctx->fs->cluster_ratio_bits &&
3085                     pb->previous_block &&
3086                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
3087                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
3088                         /* Set blk to the beginning of next cluster. */
3089                         blk = EXT2FS_C2B(
3090                                 ctx->fs,
3091                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
3092                 else
3093                         /* Set blk to the beginning of current cluster. */
3094                         blk = EXT2FS_C2B(ctx->fs,
3095                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
3096
3097                 if (blk < extent.e_pblk + extent.e_len) {
3098                         mark_blocks_used(ctx, blk,
3099                                          extent.e_pblk + extent.e_len - blk);
3100                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
3101                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
3102                         pb->num_blocks += n;
3103                 }
3104                 pb->last_block = extent.e_lblk + extent.e_len - 1;
3105                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
3106                 start_block = pb->last_block = last_lblk;
3107                 if (is_leaf && !is_dir &&
3108                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
3109                         pb->last_init_lblock = last_lblk;
3110         next:
3111                 pctx->errcode = ext2fs_extent_get(ehandle,
3112                                                   EXT2_EXTENT_NEXT_SIB,
3113                                                   &extent);
3114         }
3115
3116         /* Failed csum but passes checks?  Ask to fix checksum. */
3117         if (failed_csum &&
3118             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
3119                 pb->inode_modified = 1;
3120                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
3121                 if (pctx->errcode)
3122                         return;
3123         }
3124
3125         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
3126                 pctx->errcode = 0;
3127 }
3128
3129 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
3130                                  struct process_block_struct *pb)
3131 {
3132         struct ext2_extent_info info;
3133         struct ext2_inode       *inode = pctx->inode;
3134         ext2_extent_handle_t    ehandle;
3135         ext2_filsys             fs = ctx->fs;
3136         ext2_ino_t              ino = pctx->ino;
3137         errcode_t               retval;
3138         blk64_t                 eof_lblk;
3139         struct ext3_extent_header       *eh;
3140
3141         /* Check for a proper extent header... */
3142         eh = (struct ext3_extent_header *) &inode->i_block[0];
3143         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
3144         if (retval) {
3145                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
3146                         e2fsck_clear_inode(ctx, ino, inode, 0,
3147                                            "check_blocks_extents");
3148                 pctx->errcode = 0;
3149                 return;
3150         }
3151
3152         /* ...since this function doesn't fail if i_block is zeroed. */
3153         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
3154         if (pctx->errcode) {
3155                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
3156                         e2fsck_clear_inode(ctx, ino, inode, 0,
3157                                            "check_blocks_extents");
3158                 pctx->errcode = 0;
3159                 return;
3160         }
3161
3162         retval = ext2fs_extent_get_info(ehandle, &info);
3163         if (retval == 0) {
3164                 int max_depth = info.max_depth;
3165
3166                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
3167                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
3168                 ctx->extent_depth_count[max_depth]++;
3169         }
3170
3171         /* Check maximum extent depth */
3172         pctx->blk = info.max_depth;
3173         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
3174         if (pctx->blk2 < pctx->blk &&
3175             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
3176                 pb->eti.force_rebuild = 1;
3177
3178         /* Can we collect extent tree level stats? */
3179         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
3180         if (pctx->blk2 > pctx->blk)
3181                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
3182         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
3183         pb->eti.ino = pb->ino;
3184
3185         pb->next_lblock = 0;
3186
3187         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
3188                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
3189         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
3190         if (pctx->errcode &&
3191             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
3192                 pb->num_blocks = 0;
3193                 inode->i_blocks = 0;
3194                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3195                                    "check_blocks_extents");
3196                 pctx->errcode = 0;
3197         }
3198         ext2fs_extent_free(ehandle);
3199
3200         /* Rebuild unless it's a dir and we're rehashing it */
3201         if (LINUX_S_ISDIR(inode->i_mode) &&
3202             e2fsck_dir_will_be_rehashed(ctx, ino))
3203                 return;
3204
3205         if (ctx->options & E2F_OPT_CONVERT_BMAP)
3206                 e2fsck_rebuild_extents_later(ctx, ino);
3207         else
3208                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
3209 }
3210
3211 /*
3212  * In fact we don't need to check blocks for an inode with inline data
3213  * because this inode doesn't have any blocks.  In this function all
3214  * we need to do is add this inode into dblist when it is a directory.
3215  */
3216 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
3217                                      struct process_block_struct *pb)
3218 {
3219         int     flags;
3220         size_t  inline_data_size = 0;
3221
3222         if (!pb->is_dir) {
3223                 pctx->errcode = 0;
3224                 return;
3225         }
3226
3227         /* Process the dirents in i_block[] as the "first" block. */
3228         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
3229         if (pctx->errcode)
3230                 goto err;
3231
3232         /* Process the dirents in the EA as a "second" block. */
3233         flags = ctx->fs->flags;
3234         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3235         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
3236                                                 &inline_data_size);
3237         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3238                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3239         if (pctx->errcode) {
3240                 pctx->errcode = 0;
3241                 return;
3242         }
3243
3244         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
3245                 return;
3246
3247         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
3248         if (pctx->errcode)
3249                 goto err;
3250
3251         return;
3252 err:
3253         pctx->blk = 0;
3254         pctx->num = 0;
3255         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3256         ctx->flags |= E2F_FLAG_ABORT;
3257 }
3258
3259 /*
3260  * This subroutine is called on each inode to account for all of the
3261  * blocks used by that inode.
3262  */
3263 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
3264                          char *block_buf, const struct ea_quota *ea_ibody_quota)
3265 {
3266         ext2_filsys fs = ctx->fs;
3267         struct process_block_struct pb;
3268         ext2_ino_t      ino = pctx->ino;
3269         struct ext2_inode *inode = pctx->inode;
3270         unsigned        bad_size = 0;
3271         int             dirty_inode = 0;
3272         int             extent_fs;
3273         int             inlinedata_fs;
3274         __u64           size;
3275         struct ea_quota ea_block_quota;
3276
3277         pb.ino = ino;
3278         pb.num_blocks = EXT2FS_B2C(ctx->fs,
3279                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
3280         pb.last_block = ~0;
3281         pb.last_init_lblock = -1;
3282         pb.last_db_block = -1;
3283         pb.num_illegal_blocks = 0;
3284         pb.suppress = 0; pb.clear = 0;
3285         pb.fragmented = 0;
3286         pb.compressed = 0;
3287         pb.previous_block = 0;
3288         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
3289         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
3290         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
3291         pb.inode = inode;
3292         pb.pctx = pctx;
3293         pb.ctx = ctx;
3294         pb.inode_modified = 0;
3295         pb.eti.force_rebuild = 0;
3296         pctx->ino = ino;
3297         pctx->errcode = 0;
3298
3299         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
3300         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
3301
3302         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
3303                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3304                         goto out;
3305                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
3306         }
3307
3308         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
3309                 check_blocks_inline_data(ctx, pctx, &pb);
3310         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
3311                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
3312                         check_blocks_extents(ctx, pctx, &pb);
3313                 else {
3314                         int flags;
3315                         /*
3316                          * If we've modified the inode, write it out before
3317                          * iterate() tries to use it.
3318                          */
3319                         if (dirty_inode) {
3320                                 e2fsck_write_inode(ctx, ino, inode,
3321                                                    "check_blocks");
3322                                 dirty_inode = 0;
3323                         }
3324                         flags = fs->flags;
3325                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3326                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
3327                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
3328                                                 block_buf, process_block, &pb);
3329                         /*
3330                          * We do not have uninitialized extents in non extent
3331                          * files.
3332                          */
3333                         pb.last_init_lblock = pb.last_block;
3334                         /*
3335                          * If iterate() changed a block mapping, we have to
3336                          * re-read the inode.  If we decide to clear the
3337                          * inode after clearing some stuff, we'll re-write the
3338                          * bad mappings into the inode!
3339                          */
3340                         if (pb.inode_modified)
3341                                 e2fsck_read_inode(ctx, ino, inode,
3342                                                   "check_blocks");
3343                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3344                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3345
3346                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
3347 #ifdef DEBUG
3348                                 printf("bmap rebuild ino=%d\n", ino);
3349 #endif
3350                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
3351                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
3352                                         e2fsck_rebuild_extents_later(ctx, ino);
3353                         }
3354                 }
3355         }
3356         end_problem_latch(ctx, PR_LATCH_BLOCK);
3357         end_problem_latch(ctx, PR_LATCH_TOOBIG);
3358         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3359                 goto out;
3360         if (pctx->errcode)
3361                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
3362
3363         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
3364                 if (LINUX_S_ISDIR(inode->i_mode))
3365                         ctx->fs_fragmented_dir++;
3366                 else
3367                         ctx->fs_fragmented++;
3368         }
3369
3370         if (pb.clear) {
3371                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3372                                    "check_blocks");
3373                 return;
3374         }
3375
3376         if (inode->i_flags & EXT2_INDEX_FL) {
3377                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
3378                         inode->i_flags &= ~EXT2_INDEX_FL;
3379                         dirty_inode++;
3380                 } else {
3381                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
3382                 }
3383         }
3384
3385         if (!pb.num_blocks && pb.is_dir &&
3386             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
3387                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
3388                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
3389                         ctx->fs_directory_count--;
3390                         return;
3391                 }
3392         }
3393
3394         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
3395             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
3396             !(inode->i_flags & EXT4_EA_INODE_FL)) {
3397                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
3398                                ino,
3399                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
3400                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
3401                                   ino, (ea_ibody_quota ?
3402                                         ea_ibody_quota->inodes : 0) +
3403                                                 ea_block_quota.inodes + 1);
3404         }
3405
3406         if (!ext2fs_has_feature_huge_file(fs->super) ||
3407             !(inode->i_flags & EXT4_HUGE_FILE_FL))
3408                 pb.num_blocks *= (fs->blocksize / 512);
3409         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
3410 #if 0
3411         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
3412                ino, inode->i_size, pb.last_block, ext2fs_inode_i_blocks(fs, inode),
3413                pb.num_blocks);
3414 #endif
3415         if (pb.is_dir) {
3416                 unsigned nblock = inode->i_size >> EXT2_BLOCK_SIZE_BITS(fs->super);
3417                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
3418                         int flags;
3419                         size_t sz = 0;
3420                         errcode_t err;
3421
3422                         flags = ctx->fs->flags;
3423                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3424                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
3425                                                       &sz);
3426                         ctx->fs->flags = (flags &
3427                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3428                                          (ctx->fs->flags &
3429                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3430                         if (err || sz != inode->i_size) {
3431                                 bad_size = 7;
3432                                 pctx->num = sz;
3433                         }
3434                 } else if (inode->i_size & (fs->blocksize - 1))
3435                         bad_size = 5;
3436                 else if (nblock > (pb.last_block + 1))
3437                         bad_size = 1;
3438                 else if (nblock < (pb.last_block + 1)) {
3439                         if (((pb.last_block + 1) - nblock) >
3440                             fs->super->s_prealloc_dir_blocks)
3441                                 bad_size = 2;
3442                 }
3443         } else {
3444                 size = EXT2_I_SIZE(inode);
3445                 if ((pb.last_init_lblock >= 0) &&
3446                     /* Do not allow initialized allocated blocks past i_size*/
3447                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
3448                     !(inode->i_flags & EXT4_VERITY_FL))
3449                         bad_size = 3;
3450                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3451                          size > ext2_max_sizes[fs->super->s_log_block_size])
3452                         /* too big for a direct/indirect-mapped file */
3453                         bad_size = 4;
3454                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3455                          size >
3456                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
3457                         /* too big for an extent-based file - 32bit ee_block */
3458                         bad_size = 6;
3459         }
3460         /* i_size for symlinks is checked elsewhere */
3461         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
3462                 /* Did inline_data set pctx->num earlier? */
3463                 if (bad_size != 7)
3464                         pctx->num = (pb.last_block + 1) * fs->blocksize;
3465                 pctx->group = bad_size;
3466                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
3467                         if (LINUX_S_ISDIR(inode->i_mode))
3468                                 pctx->num &= 0xFFFFFFFFULL;
3469                         ext2fs_inode_size_set(fs, inode, pctx->num);
3470                         if (EXT2_I_SIZE(inode) == 0 &&
3471                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
3472                                 memset(inode->i_block, 0,
3473                                        sizeof(inode->i_block));
3474                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
3475                         }
3476                         dirty_inode++;
3477                 }
3478                 pctx->num = 0;
3479         }
3480         if (LINUX_S_ISREG(inode->i_mode) &&
3481             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3482                 ctx->large_files++;
3483         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
3484             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
3485              (ext2fs_has_feature_huge_file(fs->super) &&
3486               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
3487               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
3488                 pctx->num = pb.num_blocks;
3489                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
3490                         inode->i_blocks = pb.num_blocks;
3491                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
3492                         dirty_inode++;
3493                 }
3494                 pctx->num = 0;
3495         }
3496
3497         /*
3498          * The kernel gets mad if we ask it to allocate bigalloc clusters to
3499          * a block mapped file, so rebuild it as an extent file.  We can skip
3500          * symlinks because they're never rewritten.
3501          */
3502         if (ext2fs_has_feature_bigalloc(fs->super) &&
3503             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
3504             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
3505             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
3506             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
3507             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
3508                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
3509                 if (pctx->errcode)
3510                         goto out;
3511         }
3512
3513         if (ctx->dirs_to_hash && pb.is_dir &&
3514             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
3515             !(inode->i_flags & EXT2_INDEX_FL) &&
3516             ((inode->i_size / fs->blocksize) >= 3))
3517                 e2fsck_rehash_dir_later(ctx, ino);
3518
3519 out:
3520         if (dirty_inode)
3521                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
3522 }
3523
3524 #if 0
3525 /*
3526  * Helper function called by process block when an illegal block is
3527  * found.  It returns a description about why the block is illegal
3528  */
3529 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
3530 {
3531         blk64_t super;
3532         int     i;
3533         static char     problem[80];
3534
3535         super = fs->super->s_first_data_block;
3536         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
3537         if (block < super) {
3538                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
3539                 return(problem);
3540         } else if (block >= ext2fs_blocks_count(fs->super)) {
3541                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
3542                 return(problem);
3543         }
3544         for (i = 0; i < fs->group_desc_count; i++) {
3545                 if (block == super) {
3546                         sprintf(problem, "is the superblock in group %d", i);
3547                         break;
3548                 }
3549                 if (block > super &&
3550                     block <= (super + fs->desc_blocks)) {
3551                         sprintf(problem, "is in the group descriptors "
3552                                 "of group %d", i);
3553                         break;
3554                 }
3555                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
3556                         sprintf(problem, "is the block bitmap of group %d", i);
3557                         break;
3558                 }
3559                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
3560                         sprintf(problem, "is the inode bitmap of group %d", i);
3561                         break;
3562                 }
3563                 if (block >= ext2fs_inode_table_loc(fs, i) &&
3564                     (block < ext2fs_inode_table_loc(fs, i)
3565                      + fs->inode_blocks_per_group)) {
3566                         sprintf(problem, "is in the inode table of group %d",
3567                                 i);
3568                         break;
3569                 }
3570                 super += fs->super->s_blocks_per_group;
3571         }
3572         return(problem);
3573 }
3574 #endif
3575
3576 /*
3577  * This is a helper function for check_blocks().
3578  */
3579 static int process_block(ext2_filsys fs,
3580                   blk64_t       *block_nr,
3581                   e2_blkcnt_t blockcnt,
3582                   blk64_t ref_block EXT2FS_ATTR((unused)),
3583                   int ref_offset EXT2FS_ATTR((unused)),
3584                   void *priv_data)
3585 {
3586         struct process_block_struct *p;
3587         struct problem_context *pctx;
3588         blk64_t blk = *block_nr;
3589         int     ret_code = 0;
3590         problem_t       problem = 0;
3591         e2fsck_t        ctx;
3592
3593         p = (struct process_block_struct *) priv_data;
3594         pctx = p->pctx;
3595         ctx = p->ctx;
3596
3597         /*
3598          * For a directory, add logical block zero for processing even if it's
3599          * not mapped or we'll be perennially stuck with broken "." and ".."
3600          * entries.
3601          */
3602         if (p->is_dir && blockcnt == 0 && blk == 0) {
3603                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
3604                 if (pctx->errcode) {
3605                         pctx->blk = blk;
3606                         pctx->num = blockcnt;
3607                         goto failed_add_dir_block;
3608                 }
3609                 p->last_db_block++;
3610         }
3611
3612         if (blk == 0)
3613                 return 0;
3614
3615 #if 0
3616         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
3617                blockcnt);
3618 #endif
3619
3620         /*
3621          * Simplistic fragmentation check.  We merely require that the
3622          * file be contiguous.  (Which can never be true for really
3623          * big files that are greater than a block group.)
3624          */
3625         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
3626                 if (p->previous_block+1 != blk) {
3627                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3628                                 char type = '?';
3629
3630                                 if (p->is_dir)
3631                                         type = 'd';
3632                                 else if (p->is_reg)
3633                                         type = 'f';
3634
3635                                 printf(_("%6lu(%c): expecting %6lu "
3636                                          "got phys %6lu (blkcnt %lld)\n"),
3637                                        (unsigned long) pctx->ino, type,
3638                                        (unsigned long) p->previous_block+1,
3639                                        (unsigned long) blk,
3640                                        blockcnt);
3641                         }
3642                         p->fragmented = 1;
3643                 }
3644         }
3645
3646         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
3647             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
3648                 problem = PR_1_TOOBIG_DIR;
3649         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
3650                 problem = PR_1_TOOBIG_DIR;
3651         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
3652                 problem = PR_1_TOOBIG_REG;
3653         if (!p->is_dir && !p->is_reg && blockcnt > 0)
3654                 problem = PR_1_TOOBIG_SYMLINK;
3655
3656         if (blk < fs->super->s_first_data_block ||
3657             blk >= ext2fs_blocks_count(fs->super))
3658                 problem = PR_1_ILLEGAL_BLOCK_NUM;
3659
3660         /*
3661          * If this IND/DIND/TIND block is squatting atop some critical metadata
3662          * (group descriptors, superblock, bitmap, inode table), any write to
3663          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
3664          * fix that and restart fsck.
3665          */
3666         if (blockcnt < 0 &&
3667             p->ino != EXT2_RESIZE_INO &&
3668             blk < ctx->fs->super->s_blocks_count &&
3669             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
3670                 pctx->blk = blk;
3671                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
3672                 if ((ctx->options & E2F_OPT_NO) == 0)
3673                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3674         }
3675
3676         if (problem) {
3677                 p->num_illegal_blocks++;
3678                 /*
3679                  * A bit of subterfuge here -- we're trying to fix a block
3680                  * mapping, but the IND/DIND/TIND block could have collided
3681                  * with some critical metadata.  So, fix the in-core mapping so
3682                  * iterate won't go insane, but return 0 instead of
3683                  * BLOCK_CHANGED so that it won't write the remapping out to
3684                  * our multiply linked block.
3685                  *
3686                  * Even if we previously determined that an *IND block
3687                  * conflicts with critical metadata, we must still try to
3688                  * iterate the *IND block as if it is an *IND block to find and
3689                  * mark the blocks it points to.  Better to be overly cautious
3690                  * with the used_blocks map so that we don't move the *IND
3691                  * block to a block that's really in use!
3692                  */
3693                 if (p->ino != EXT2_RESIZE_INO &&
3694                     ref_block != 0 &&
3695                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3696                                               ref_block)) {
3697                         *block_nr = 0;
3698                         return 0;
3699                 }
3700                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
3701                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
3702                                 p->clear = 1;
3703                                 return BLOCK_ABORT;
3704                         }
3705                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
3706                                 p->suppress = 1;
3707                                 set_latch_flags(PR_LATCH_BLOCK,
3708                                                 PRL_SUPPRESS, 0);
3709                         }
3710                 }
3711                 pctx->blk = blk;
3712                 pctx->blkcount = blockcnt;
3713                 if (fix_problem(ctx, problem, pctx)) {
3714                         blk = *block_nr = 0;
3715                         ret_code = BLOCK_CHANGED;
3716                         p->inode_modified = 1;
3717                         /*
3718                          * If the directory block is too big and is beyond the
3719                          * end of the FS, don't bother trying to add it for
3720                          * processing -- the kernel would never have created a
3721                          * directory this large, and we risk an ENOMEM abort.
3722                          * In any case, the toobig handler for extent-based
3723                          * directories also doesn't feed toobig blocks to
3724                          * pass 2.
3725                          */
3726                         if (problem == PR_1_TOOBIG_DIR)
3727                                 return ret_code;
3728                         goto mark_dir;
3729                 } else
3730                         return 0;
3731         }
3732
3733         if (p->ino == EXT2_RESIZE_INO) {
3734                 /*
3735                  * The resize inode has already be sanity checked
3736                  * during pass #0 (the superblock checks).  All we
3737                  * have to do is mark the double indirect block as
3738                  * being in use; all of the other blocks are handled
3739                  * by mark_table_blocks()).
3740                  */
3741                 if (blockcnt == BLOCK_COUNT_DIND)
3742                         mark_block_used(ctx, blk);
3743                 p->num_blocks++;
3744         } else if (!(ctx->fs->cluster_ratio_bits &&
3745                      p->previous_block &&
3746                      (EXT2FS_B2C(ctx->fs, blk) ==
3747                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
3748                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
3749                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
3750                 mark_block_used(ctx, blk);
3751                 p->num_blocks++;
3752         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
3753                                              p->last_block, blk, blockcnt)) {
3754                 pctx->blk = blockcnt;
3755                 pctx->blk2 = blk;
3756                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3757                 mark_block_used(ctx, blk);
3758                 mark_block_used(ctx, blk);
3759         }
3760         if (blockcnt >= 0)
3761                 p->last_block = blockcnt;
3762         p->previous_block = blk;
3763 mark_dir:
3764         if (p->is_dir && (blockcnt >= 0)) {
3765                 while (++p->last_db_block < blockcnt) {
3766                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
3767                                                               p->ino, 0,
3768                                                               p->last_db_block);
3769                         if (pctx->errcode) {
3770                                 pctx->blk = 0;
3771                                 pctx->num = p->last_db_block;
3772                                 goto failed_add_dir_block;
3773                         }
3774                 }
3775                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
3776                                                       blk, blockcnt);
3777                 if (pctx->errcode) {
3778                         pctx->blk = blk;
3779                         pctx->num = blockcnt;
3780                 failed_add_dir_block:
3781                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3782                         /* Should never get here */
3783                         ctx->flags |= E2F_FLAG_ABORT;
3784                         return BLOCK_ABORT;
3785                 }
3786         }
3787         return ret_code;
3788 }
3789
3790 static int process_bad_block(ext2_filsys fs,
3791                       blk64_t *block_nr,
3792                       e2_blkcnt_t blockcnt,
3793                       blk64_t ref_block EXT2FS_ATTR((unused)),
3794                       int ref_offset EXT2FS_ATTR((unused)),
3795                       void *priv_data)
3796 {
3797         struct process_block_struct *p;
3798         blk64_t         blk = *block_nr;
3799         blk64_t         first_block;
3800         dgrp_t          i;
3801         struct problem_context *pctx;
3802         e2fsck_t        ctx;
3803
3804         if (!blk)
3805                 return 0;
3806
3807         p = (struct process_block_struct *) priv_data;
3808         ctx = p->ctx;
3809         pctx = p->pctx;
3810
3811         pctx->ino = EXT2_BAD_INO;
3812         pctx->blk = blk;
3813         pctx->blkcount = blockcnt;
3814
3815         if ((blk < fs->super->s_first_data_block) ||
3816             (blk >= ext2fs_blocks_count(fs->super))) {
3817                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
3818                         *block_nr = 0;
3819                         return BLOCK_CHANGED;
3820                 } else
3821                         return 0;
3822         }
3823
3824         if (blockcnt < 0) {
3825                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
3826                         p->bbcheck = 1;
3827                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
3828                                 *block_nr = 0;
3829                                 return BLOCK_CHANGED;
3830                         }
3831                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
3832                                                     blk)) {
3833                         p->bbcheck = 1;
3834                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
3835                                         pctx)) {
3836                                 *block_nr = 0;
3837                                 return BLOCK_CHANGED;
3838                         }
3839                         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3840                                 return BLOCK_ABORT;
3841                 } else
3842                         mark_block_used(ctx, blk);
3843                 return 0;
3844         }
3845 #if 0
3846         printf ("DEBUG: Marking %u as bad.\n", blk);
3847 #endif
3848         ctx->fs_badblocks_count++;
3849         /*
3850          * If the block is not used, then mark it as used and return.
3851          * If it is already marked as found, this must mean that
3852          * there's an overlap between the filesystem table blocks
3853          * (bitmaps and inode table) and the bad block list.
3854          */
3855         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
3856                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
3857                 return 0;
3858         }
3859         /*
3860          * Try to find the where the filesystem block was used...
3861          */
3862         first_block = fs->super->s_first_data_block;
3863
3864         for (i = 0; i < fs->group_desc_count; i++ ) {
3865                 pctx->group = i;
3866                 pctx->blk = blk;
3867                 if (!ext2fs_bg_has_super(fs, i))
3868                         goto skip_super;
3869                 if (blk == first_block) {
3870                         if (i == 0) {
3871                                 if (fix_problem(ctx,
3872                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
3873                                                 pctx)) {
3874                                         *block_nr = 0;
3875                                         return BLOCK_CHANGED;
3876                                 }
3877                                 return 0;
3878                         }
3879                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
3880                         return 0;
3881                 }
3882                 if ((blk > first_block) &&
3883                     (blk <= first_block + fs->desc_blocks)) {
3884                         if (i == 0) {
3885                                 pctx->blk = *block_nr;
3886                                 if (fix_problem(ctx,
3887                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
3888                                         *block_nr = 0;
3889                                         return BLOCK_CHANGED;
3890                                 }
3891                                 return 0;
3892                         }
3893                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
3894                         return 0;
3895                 }
3896         skip_super:
3897                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
3898                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
3899                                 ctx->invalid_block_bitmap_flag[i]++;
3900                                 ctx->invalid_bitmaps++;
3901                         }
3902                         return 0;
3903                 }
3904                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
3905                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
3906                                 ctx->invalid_inode_bitmap_flag[i]++;
3907                                 ctx->invalid_bitmaps++;
3908                         }
3909                         return 0;
3910                 }
3911                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
3912                     (blk < (ext2fs_inode_table_loc(fs, i) +
3913                             fs->inode_blocks_per_group))) {
3914                         /*
3915                          * If there are bad blocks in the inode table,
3916                          * the inode scan code will try to do
3917                          * something reasonable automatically.
3918                          */
3919                         return 0;
3920                 }
3921                 first_block += fs->super->s_blocks_per_group;
3922         }
3923         /*
3924          * If we've gotten to this point, then the only
3925          * possibility is that the bad block inode meta data
3926          * is using a bad block.
3927          */
3928         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
3929             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
3930             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
3931                 p->bbcheck = 1;
3932                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
3933                         *block_nr = 0;
3934                         return BLOCK_CHANGED;
3935                 }
3936                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3937                         return BLOCK_ABORT;
3938                 return 0;
3939         }
3940
3941         pctx->group = -1;
3942
3943         /* Warn user that the block wasn't claimed */
3944         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
3945
3946         return 0;
3947 }
3948
3949 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
3950                             const char *name, int num, blk64_t *new_block)
3951 {
3952         ext2_filsys fs = ctx->fs;
3953         dgrp_t          last_grp;
3954         blk64_t         old_block = *new_block;
3955         blk64_t         last_block;
3956         dgrp_t          flexbg;
3957         unsigned        flexbg_size;
3958         int             i, is_flexbg;
3959         char            *buf;
3960         struct problem_context  pctx;
3961
3962         clear_problem_context(&pctx);
3963
3964         pctx.group = group;
3965         pctx.blk = old_block;
3966         pctx.str = name;
3967
3968         /*
3969          * For flex_bg filesystems, first try to allocate the metadata
3970          * within the flex_bg, and if that fails then try finding the
3971          * space anywhere in the filesystem.
3972          */
3973         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
3974         if (is_flexbg) {
3975                 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
3976                 flexbg = group / flexbg_size;
3977                 first_block = ext2fs_group_first_block2(fs,
3978                                                         flexbg_size * flexbg);
3979                 last_grp = group | (flexbg_size - 1);
3980                 if (last_grp >= fs->group_desc_count)
3981                         last_grp = fs->group_desc_count - 1;
3982                 last_block = ext2fs_group_last_block2(fs, last_grp);
3983         } else
3984                 last_block = ext2fs_group_last_block2(fs, group);
3985         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
3986                                                num, ctx->block_found_map,
3987                                                new_block);
3988         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
3989                 pctx.errcode = ext2fs_get_free_blocks2(fs,
3990                                 fs->super->s_first_data_block,
3991                                 ext2fs_blocks_count(fs->super),
3992                                 num, ctx->block_found_map, new_block);
3993         if (pctx.errcode) {
3994                 pctx.num = num;
3995                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
3996                 ext2fs_unmark_valid(fs);
3997                 ctx->flags |= E2F_FLAG_ABORT;
3998                 return;
3999         }
4000         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4001         if (pctx.errcode) {
4002                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4003                 ext2fs_unmark_valid(fs);
4004                 ctx->flags |= E2F_FLAG_ABORT;
4005                 return;
4006         }
4007         ext2fs_mark_super_dirty(fs);
4008         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4009         pctx.blk2 = *new_block;
4010         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4011                           PR_1_RELOC_TO), &pctx);
4012         pctx.blk2 = 0;
4013         for (i = 0; i < num; i++) {
4014                 pctx.blk = i;
4015                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4016                 if (old_block) {
4017                         pctx.errcode = io_channel_read_blk64(fs->io,
4018                                    old_block + i, 1, buf);
4019                         if (pctx.errcode)
4020                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4021                         pctx.blk = (*new_block) + i;
4022                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4023                                                               1, buf);
4024                 } else {
4025                         pctx.blk = (*new_block) + i;
4026                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4027                                                            NULL, NULL);
4028                 }
4029
4030                 if (pctx.errcode)
4031                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4032         }
4033         ext2fs_free_mem(&buf);
4034 }
4035
4036 /*
4037  * This routine gets called at the end of pass 1 if bad blocks are
4038  * detected in the superblock, group descriptors, inode_bitmaps, or
4039  * block bitmaps.  At this point, all of the blocks have been mapped
4040  * out, so we can try to allocate new block(s) to replace the bad
4041  * blocks.
4042  */
4043 static void handle_fs_bad_blocks(e2fsck_t ctx)
4044 {
4045         ext2_filsys fs = ctx->fs;
4046         dgrp_t          i;
4047         blk64_t         first_block;
4048         blk64_t         new_blk;
4049
4050         for (i = 0; i < fs->group_desc_count; i++) {
4051                 first_block = ext2fs_group_first_block2(fs, i);
4052
4053                 if (ctx->invalid_block_bitmap_flag[i]) {
4054                         new_blk = ext2fs_block_bitmap_loc(fs, i);
4055                         new_table_block(ctx, first_block, i, _("block bitmap"),
4056                                         1, &new_blk);
4057                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4058                 }
4059                 if (ctx->invalid_inode_bitmap_flag[i]) {
4060                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
4061                         new_table_block(ctx, first_block, i, _("inode bitmap"),
4062                                         1, &new_blk);
4063                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4064                 }
4065                 if (ctx->invalid_inode_table_flag[i]) {
4066                         new_blk = ext2fs_inode_table_loc(fs, i);
4067                         new_table_block(ctx, first_block, i, _("inode table"),
4068                                         fs->inode_blocks_per_group,
4069                                         &new_blk);
4070                         ext2fs_inode_table_loc_set(fs, i, new_blk);
4071                         ctx->flags |= E2F_FLAG_RESTART;
4072                 }
4073         }
4074         ctx->invalid_bitmaps = 0;
4075 }
4076
4077 /*
4078  * This routine marks all blocks which are used by the superblock,
4079  * group descriptors, inode bitmaps, and block bitmaps.
4080  */
4081 static void mark_table_blocks(e2fsck_t ctx)
4082 {
4083         ext2_filsys fs = ctx->fs;
4084         blk64_t b;
4085         dgrp_t  i;
4086         unsigned int    j;
4087         struct problem_context pctx;
4088
4089         clear_problem_context(&pctx);
4090
4091         for (i = 0; i < fs->group_desc_count; i++) {
4092                 pctx.group = i;
4093
4094                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
4095                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
4096
4097                 /*
4098                  * Mark the blocks used for the inode table
4099                  */
4100                 if (ext2fs_inode_table_loc(fs, i)) {
4101                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
4102                              j < fs->inode_blocks_per_group;
4103                              j++, b++) {
4104                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4105                                                              b)) {
4106                                         pctx.blk = b;
4107                                         if (!ctx->invalid_inode_table_flag[i] &&
4108                                             fix_problem(ctx,
4109                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
4110                                                 ctx->invalid_inode_table_flag[i]++;
4111                                                 ctx->invalid_bitmaps++;
4112                                         }
4113                                 } else {
4114                                     ext2fs_mark_block_bitmap2(
4115                                                 ctx->block_found_map, b);
4116                                     ext2fs_mark_block_bitmap2(
4117                                                 ctx->block_metadata_map, b);
4118                                 }
4119                         }
4120                 }
4121
4122                 /*
4123                  * Mark block used for the block bitmap
4124                  */
4125                 if (ext2fs_block_bitmap_loc(fs, i)) {
4126                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4127                                      ext2fs_block_bitmap_loc(fs, i))) {
4128                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
4129                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
4130                                         ctx->invalid_block_bitmap_flag[i]++;
4131                                         ctx->invalid_bitmaps++;
4132                                 }
4133                         } else {
4134                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4135                                      ext2fs_block_bitmap_loc(fs, i));
4136                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4137                                      ext2fs_block_bitmap_loc(fs, i));
4138                         }
4139                 }
4140                 /*
4141                  * Mark block used for the inode bitmap
4142                  */
4143                 if (ext2fs_inode_bitmap_loc(fs, i)) {
4144                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4145                                      ext2fs_inode_bitmap_loc(fs, i))) {
4146                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
4147                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
4148                                         ctx->invalid_inode_bitmap_flag[i]++;
4149                                         ctx->invalid_bitmaps++;
4150                                 }
4151                         } else {
4152                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4153                                      ext2fs_inode_bitmap_loc(fs, i));
4154                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4155                                      ext2fs_inode_bitmap_loc(fs, i));
4156                         }
4157                 }
4158         }
4159 }
4160
4161 /*
4162  * These subroutines short circuits ext2fs_get_blocks and
4163  * ext2fs_check_directory; we use them since we already have the inode
4164  * structure, so there's no point in letting the ext2fs library read
4165  * the inode again.
4166  */
4167 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
4168                                   blk_t *blocks)
4169 {
4170         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4171         int     i;
4172
4173         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4174                 return EXT2_ET_CALLBACK_NOTHANDLED;
4175
4176         for (i=0; i < EXT2_N_BLOCKS; i++)
4177                 blocks[i] = ctx->stashed_inode->i_block[i];
4178         return 0;
4179 }
4180
4181 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
4182                                   struct ext2_inode *inode)
4183 {
4184         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4185
4186         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4187                 return EXT2_ET_CALLBACK_NOTHANDLED;
4188         *inode = *ctx->stashed_inode;
4189         return 0;
4190 }
4191
4192 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
4193                             struct ext2_inode *inode)
4194 {
4195         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4196
4197         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
4198                 (inode != ctx->stashed_inode))
4199                 *ctx->stashed_inode = *inode;
4200         return EXT2_ET_CALLBACK_NOTHANDLED;
4201 }
4202
4203 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
4204 {
4205         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4206
4207         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4208                 return EXT2_ET_CALLBACK_NOTHANDLED;
4209
4210         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
4211                 return EXT2_ET_NO_DIRECTORY;
4212         return 0;
4213 }
4214
4215 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
4216                                         blk64_t *ret)
4217 {
4218         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4219         errcode_t       retval;
4220         blk64_t         new_block;
4221
4222         if (ctx->block_found_map) {
4223                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
4224                                            &new_block);
4225                 if (retval)
4226                         return retval;
4227                 if (fs->block_map) {
4228                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
4229                         ext2fs_mark_bb_dirty(fs);
4230                 }
4231         } else {
4232                 if (!fs->block_map) {
4233                         retval = ext2fs_read_block_bitmap(fs);
4234                         if (retval)
4235                                 return retval;
4236                 }
4237
4238                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
4239                 if (retval)
4240                         return retval;
4241         }
4242
4243         *ret = new_block;
4244         return (0);
4245 }
4246
4247 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
4248                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
4249 {
4250         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4251         errcode_t       retval;
4252
4253         if (ctx->block_found_map)
4254                 return ext2fs_new_range(fs, flags, goal, len,
4255                                         ctx->block_found_map, pblk, plen);
4256
4257         if (!fs->block_map) {
4258                 retval = ext2fs_read_block_bitmap(fs);
4259                 if (retval)
4260                         return retval;
4261         }
4262
4263         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
4264                                 pblk, plen);
4265 }
4266
4267 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
4268 {
4269         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4270
4271         /* Never free a critical metadata block */
4272         if (ctx->block_found_map &&
4273             ctx->block_metadata_map &&
4274             inuse < 0 &&
4275             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
4276                 return;
4277
4278         if (ctx->block_found_map) {
4279                 if (inuse > 0)
4280                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4281                 else
4282                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
4283         }
4284 }
4285
4286 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
4287                                            blk_t num, int inuse)
4288 {
4289         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4290
4291         /* Never free a critical metadata block */
4292         if (ctx->block_found_map &&
4293             ctx->block_metadata_map &&
4294             inuse < 0 &&
4295             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
4296                 return;
4297
4298         if (ctx->block_found_map) {
4299                 if (inuse > 0)
4300                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
4301                                                         blk, num);
4302                 else
4303                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
4304                                                         blk, num);
4305         }
4306 }
4307
4308 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
4309 {
4310         ext2_filsys fs = ctx->fs;
4311
4312         if (use_shortcuts) {
4313                 fs->get_blocks = pass1_get_blocks;
4314                 fs->check_directory = pass1_check_directory;
4315                 fs->read_inode = pass1_read_inode;
4316                 fs->write_inode = pass1_write_inode;
4317                 ctx->stashed_ino = 0;
4318         } else {
4319                 fs->get_blocks = 0;
4320                 fs->check_directory = 0;
4321                 fs->read_inode = 0;
4322                 fs->write_inode = 0;
4323         }
4324 }
4325
4326 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
4327 {
4328         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
4329         ext2fs_set_block_alloc_stats_callback(ctx->fs,
4330                                                 e2fsck_block_alloc_stats, 0);
4331         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
4332         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
4333                                         e2fsck_block_alloc_stats_range, NULL);
4334 }