e2fsck/pass1.c

   1 /*
   2  * pass1.c -- pass #1 of e2fsck: sequential scan of the inode table
   3  *
   4  * Copyright (C) 1993, 1994, 1995, 1996, 1997 Theodore Ts'o.
   5  *
   6  * %Begin-Header%
   7  * This file may be redistributed under the terms of the GNU Public
   8  * License.
   9  * %End-Header%
  10  *
  11  * Pass 1 of e2fsck iterates over all the inodes in the filesystems,
  12  * and applies the following tests to each inode:
  13  *
  14  *      - The mode field of the inode must be legal.
  15  *      - The size and block count fields of the inode are correct.
  16  *      - A data block must not be used by another inode
  17  *
  18  * Pass 1 also gathers the collects the following information:
  19  *
  20  *      - A bitmap of which inodes are in use.          (inode_used_map)
  21  *      - A bitmap of which inodes are directories.     (inode_dir_map)
  22  *      - A bitmap of which inodes are regular files.   (inode_reg_map)
  23  *      - A bitmap of which inodes have bad fields.     (inode_bad_map)
  24  *      - A bitmap of which inodes are in bad blocks.   (inode_bb_map)
  25  *      - A bitmap of which inodes are imagic inodes.   (inode_imagic_map)
  26  *      - A bitmap of which blocks are in use.          (block_found_map)
  27  *      - A bitmap of which blocks are in use by two inodes     (block_dup_map)
  28  *      - The data blocks of the directory inodes.      (dir_map)
  29  *      - Ref counts for ea_inodes.                     (ea_inode_refs)
  30  *
  31  * Pass 1 is designed to stash away enough information so that the
  32  * other passes should not need to read in the inode information
  33  * during the normal course of a filesystem check.  (Although if an
  34  * inconsistency is detected, other passes may need to read in an
  35  * inode to fix it.)
  36  *
  37  * Note that pass 1B will be invoked if there are any duplicate blocks
  38  * found.
  39  */
  40
  41 #define _GNU_SOURCE 1 /* get strnlen() */
  42 #include "config.h"
  43 #include <string.h>
  44 #include <time.h>
  45 #ifdef HAVE_ERRNO_H
  46 #include <errno.h>
  47 #endif
  48
  49 #include "e2fsck.h"
  50 #include <ext2fs/ext2_ext_attr.h>
  51 #include <e2p/e2p.h>
  52
  53 #include "problem.h"
  54
  55 #ifdef NO_INLINE_FUNCS
  56 #define _INLINE_
  57 #else
  58 #define _INLINE_ inline
  59 #endif
  60
  61 #undef DEBUG
  62
  63 struct ea_quota {
  64         blk64_t blocks;
  65         __u64 inodes;
  66 };
  67
  68 static int process_block(ext2_filsys fs, blk64_t        *blocknr,
  69                          e2_blkcnt_t blockcnt, blk64_t ref_blk,
  70                          int ref_offset, void *priv_data);
  71 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
  72                              e2_blkcnt_t blockcnt, blk64_t ref_blk,
  73                              int ref_offset, void *priv_data);
  74 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
  75                          char *block_buf,
  76                          const struct ea_quota *ea_ibody_quota);
  77 static void mark_table_blocks(e2fsck_t ctx);
  78 static void alloc_bb_map(e2fsck_t ctx);
  79 static void alloc_imagic_map(e2fsck_t ctx);
  80 static void mark_inode_bad(e2fsck_t ctx, ino_t ino);
  81 static void add_encrypted_dir(e2fsck_t ctx, ino_t ino);
  82 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino);
  83 static void handle_fs_bad_blocks(e2fsck_t ctx);
  84 static void process_inodes(e2fsck_t ctx, char *block_buf);
  85 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
  86 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
  87                                   dgrp_t group, void * priv_data);
  88 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
  89                                     char *block_buf, int adjust_sign);
  90 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
  91
  92 struct process_block_struct {
  93         ext2_ino_t      ino;
  94         unsigned        is_dir:1, is_reg:1, clear:1, suppress:1,
  95                                 fragmented:1, compressed:1, bbcheck:1,
  96                                 inode_modified:1;
  97         blk64_t         num_blocks;
  98         blk64_t         max_blocks;
  99         blk64_t         last_block;
 100         e2_blkcnt_t     last_init_lblock;
 101         e2_blkcnt_t     last_db_block;
 102         int             num_illegal_blocks;
 103         blk64_t         previous_block;
 104         struct ext2_inode *inode;
 105         struct problem_context *pctx;
 106         ext2fs_block_bitmap fs_meta_blocks;
 107         e2fsck_t        ctx;
 108         blk64_t         next_lblock;
 109         struct extent_tree_info eti;
 110 };
 111
 112 struct process_inode_block {
 113         ext2_ino_t ino;
 114         struct ea_quota ea_ibody_quota;
 115         struct ext2_inode_large inode;
 116 };
 117
 118 struct scan_callback_struct {
 119         e2fsck_t        ctx;
 120         char            *block_buf;
 121 };
 122
 123 /*
 124  * For the inodes to process list.
 125  */
 126 static struct process_inode_block *inodes_to_process;
 127 static int process_inode_count;
 128
 129 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
 130                             EXT2_MIN_BLOCK_LOG_SIZE + 1];
 131
 132 /*
 133  * Free all memory allocated by pass1 in preparation for restarting
 134  * things.
 135  */
 136 static void unwind_pass1(ext2_filsys fs EXT2FS_ATTR((unused)))
 137 {
 138         ext2fs_free_mem(&inodes_to_process);
 139         inodes_to_process = 0;
 140 }
 141
 142 /*
 143  * Check to make sure a device inode is real.  Returns 1 if the device
 144  * checks out, 0 if not.
 145  *
 146  * Note: this routine is now also used to check FIFO's and Sockets,
 147  * since they have the same requirement; the i_block fields should be
 148  * zero.
 149  */
 150 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
 151                                     struct ext2_inode *inode)
 152 {
 153         int     i;
 154
 155         /*
 156          * If the index or extents flag is set, then this is a bogus
 157          * device/fifo/socket
 158          */
 159         if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
 160                 return 0;
 161
 162         /*
 163          * We should be able to do the test below all the time, but
 164          * because the kernel doesn't forcibly clear the device
 165          * inode's additional i_block fields, there are some rare
 166          * occasions when a legitimate device inode will have non-zero
 167          * additional i_block fields.  So for now, we only complain
 168          * when the immutable flag is set, which should never happen
 169          * for devices.  (And that's when the problem is caused, since
 170          * you can't set or clear immutable flags for devices.)  Once
 171          * the kernel has been fixed we can change this...
 172          */
 173         if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
 174                 for (i=4; i < EXT2_N_BLOCKS; i++)
 175                         if (inode->i_block[i])
 176                                 return 0;
 177         }
 178         return 1;
 179 }
 180
 181 /*
 182  * Check to make sure a symlink inode is real.  Returns 1 if the symlink
 183  * checks out, 0 if not.
 184  */
 185 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
 186                                struct ext2_inode *inode, char *buf)
 187 {
 188         unsigned int buflen;
 189         unsigned int len;
 190
 191         if ((inode->i_size_high || inode->i_size == 0) ||
 192             (inode->i_flags & EXT2_INDEX_FL))
 193                 return 0;
 194
 195         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
 196                 size_t inline_size;
 197
 198                 if (inode->i_flags & EXT4_EXTENTS_FL)
 199                         return 0;
 200                 if (ext2fs_inline_data_size(fs, ino, &inline_size))
 201                         return 0;
 202                 if (inode->i_size != inline_size)
 203                         return 0;
 204
 205                 return 1;
 206         }
 207
 208         if (ext2fs_is_fast_symlink(inode)) {
 209                 if (inode->i_flags & EXT4_EXTENTS_FL)
 210                         return 0;
 211                 buf = (char *)inode->i_block;
 212                 buflen = sizeof(inode->i_block);
 213         } else {
 214                 ext2_extent_handle_t    handle;
 215                 struct ext2_extent_info info;
 216                 struct ext2fs_extent    extent;
 217                 blk64_t blk;
 218                 int i;
 219
 220                 if (inode->i_flags & EXT4_EXTENTS_FL) {
 221                         if (ext2fs_extent_open2(fs, ino, inode, &handle))
 222                                 return 0;
 223                         if (ext2fs_extent_get_info(handle, &info) ||
 224                             (info.num_entries != 1) ||
 225                             (info.max_depth != 0)) {
 226                                 ext2fs_extent_free(handle);
 227                                 return 0;
 228                         }
 229                         if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
 230                                               &extent) ||
 231                             (extent.e_lblk != 0) ||
 232                             (extent.e_len != 1)) {
 233                                 ext2fs_extent_free(handle);
 234                                 return 0;
 235                         }
 236                         blk = extent.e_pblk;
 237                         ext2fs_extent_free(handle);
 238                 } else {
 239                         blk = inode->i_block[0];
 240
 241                         for (i = 1; i < EXT2_N_BLOCKS; i++)
 242                                 if (inode->i_block[i])
 243                                         return 0;
 244                 }
 245
 246                 if (blk < fs->super->s_first_data_block ||
 247                     blk >= ext2fs_blocks_count(fs->super))
 248                         return 0;
 249
 250                 if (io_channel_read_blk64(fs->io, blk, 1, buf))
 251                         return 0;
 252
 253                 buflen = fs->blocksize;
 254         }
 255
 256         if (inode->i_flags & EXT4_ENCRYPT_FL)
 257                 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
 258         else
 259                 len = strnlen(buf, buflen);
 260
 261         if (len >= buflen)
 262                 return 0;
 263
 264         if (len != inode->i_size)
 265                 return 0;
 266         return 1;
 267 }
 268
 269 /*
 270  * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
 271  */
 272 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
 273 static void check_extents_inlinedata(e2fsck_t ctx,
 274                                      struct problem_context *pctx)
 275 {
 276         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 277                 return;
 278
 279         if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
 280                 return;
 281
 282         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 283         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 284 }
 285 #undef BAD_SPECIAL_FLAGS
 286
 287 /*
 288  * If the immutable (or append-only) flag is set on the inode, offer
 289  * to clear it.
 290  */
 291 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
 292 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
 293 {
 294         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 295                 return;
 296
 297         if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
 298                 return;
 299
 300         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 301         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 302 }
 303
 304 /*
 305  * If device, fifo or socket, check size is zero -- if not offer to
 306  * clear it
 307  */
 308 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
 309 {
 310         struct ext2_inode *inode = pctx->inode;
 311
 312         if (EXT2_I_SIZE(inode) == 0)
 313                 return;
 314
 315         if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
 316                 return;
 317
 318         ext2fs_inode_size_set(ctx->fs, inode, 0);
 319         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 320 }
 321
 322 /*
 323  * For a given size, calculate how many blocks would be charged towards quota.
 324  */
 325 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
 326 {
 327         blk64_t clusters;
 328
 329         clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
 330         return EXT2FS_C2B(fs, clusters);
 331 }
 332
 333 /*
 334  * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
 335  * the problem code.
 336  */
 337 static problem_t check_large_ea_inode(e2fsck_t ctx,
 338                                       struct ext2_ext_attr_entry *entry,
 339                                       struct problem_context *pctx,
 340                                       blk64_t *quota_blocks)
 341 {
 342         struct ext2_inode inode;
 343         __u32 hash;
 344         errcode_t retval;
 345
 346         /* Check if inode is within valid range */
 347         if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
 348             (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
 349                 pctx->num = entry->e_value_inum;
 350                 return PR_1_ATTR_VALUE_EA_INODE;
 351         }
 352
 353         e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
 354
 355         retval = ext2fs_ext_attr_hash_entry2(ctx->fs, entry, NULL, &hash);
 356         if (retval) {
 357                 com_err("check_large_ea_inode", retval,
 358                         _("while hashing entry with e_value_inum = %u"),
 359                         entry->e_value_inum);
 360                 fatal_error(ctx, 0);
 361         }
 362
 363         if (hash == entry->e_hash) {
 364                 *quota_blocks = size_to_quota_blocks(ctx->fs,
 365                                                      entry->e_value_size);
 366         } else {
 367                 /* This might be an old Lustre-style ea_inode reference. */
 368                 if (inode.i_mtime == pctx->ino &&
 369                     inode.i_generation == pctx->inode->i_generation) {
 370                         *quota_blocks = 0;
 371                 } else {
 372                         /* If target inode is also missing EA_INODE flag,
 373                          * this is likely to be a bad reference.
 374                          */
 375                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 376                                 pctx->num = entry->e_value_inum;
 377                                 return PR_1_ATTR_VALUE_EA_INODE;
 378                         } else {
 379                                 pctx->num = entry->e_hash;
 380                                 return PR_1_ATTR_HASH;
 381                         }
 382                 }
 383         }
 384
 385         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 386                 pctx->num = entry->e_value_inum;
 387                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
 388                         inode.i_flags |= EXT4_EA_INODE_FL;
 389                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
 390                                            &inode);
 391                 } else {
 392                         return PR_1_ATTR_NO_EA_INODE_FL;
 393                 }
 394         }
 395         return 0;
 396 }
 397
 398 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
 399                               struct ext2_ext_attr_entry *first, void *end)
 400 {
 401         struct ext2_ext_attr_entry *entry;
 402
 403         for (entry = first;
 404              (void *)entry < end && !EXT2_EXT_IS_LAST_ENTRY(entry);
 405              entry = EXT2_EXT_ATTR_NEXT(entry)) {
 406                 if (!entry->e_value_inum)
 407                         continue;
 408                 if (!ctx->ea_inode_refs) {
 409                         pctx->errcode = ea_refcount_create(0,
 410                                                            &ctx->ea_inode_refs);
 411                         if (pctx->errcode) {
 412                                 pctx->num = 4;
 413                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
 414                                 ctx->flags |= E2F_FLAG_ABORT;
 415                                 return;
 416                         }
 417                 }
 418                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
 419                                       0);
 420         }
 421 }
 422
 423 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
 424                               struct ea_quota *ea_ibody_quota)
 425 {
 426         struct ext2_super_block *sb = ctx->fs->super;
 427         struct ext2_inode_large *inode;
 428         struct ext2_ext_attr_entry *entry;
 429         char *start, *header, *end;
 430         unsigned int storage_size, remain;
 431         problem_t problem = 0;
 432         region_t region = 0;
 433
 434         ea_ibody_quota->blocks = 0;
 435         ea_ibody_quota->inodes = 0;
 436
 437         inode = (struct ext2_inode_large *) pctx->inode;
 438         storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
 439                 inode->i_extra_isize;
 440         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 441                  inode->i_extra_isize;
 442         end = header + storage_size;
 443         start = header + sizeof(__u32);
 444         entry = (struct ext2_ext_attr_entry *) start;
 445
 446         /* scan all entry's headers first */
 447
 448         /* take finish entry 0UL into account */
 449         remain = storage_size - sizeof(__u32);
 450
 451         region = region_create(0, storage_size);
 452         if (!region) {
 453                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
 454                 problem = 0;
 455                 ctx->flags |= E2F_FLAG_ABORT;
 456                 return;
 457         }
 458         if (region_allocate(region, 0, sizeof(__u32))) {
 459                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 460                 goto fix;
 461         }
 462
 463         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
 464                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
 465                 __u32 hash;
 466
 467                 if (region_allocate(region, (char *)entry - (char *)header,
 468                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
 469                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
 470                         goto fix;
 471                 }
 472
 473                 /* header eats this space */
 474                 remain -= sizeof(struct ext2_ext_attr_entry);
 475
 476                 /* is attribute name valid? */
 477                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
 478                         pctx->num = entry->e_name_len;
 479                         problem = PR_1_ATTR_NAME_LEN;
 480                         goto fix;
 481                 }
 482
 483                 /* attribute len eats this space */
 484                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
 485
 486                 if (entry->e_value_inum == 0) {
 487                         /* check value size */
 488                         if (entry->e_value_size > remain) {
 489                                 pctx->num = entry->e_value_size;
 490                                 problem = PR_1_ATTR_VALUE_SIZE;
 491                                 goto fix;
 492                         }
 493
 494                         if (entry->e_value_size &&
 495                             region_allocate(region,
 496                                             sizeof(__u32) + entry->e_value_offs,
 497                                             EXT2_EXT_ATTR_SIZE(
 498                                                 entry->e_value_size))) {
 499                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 500                                 goto fix;
 501                         }
 502
 503                         hash = ext2fs_ext_attr_hash_entry(entry,
 504                                                           start + entry->e_value_offs);
 505
 506                         /* e_hash may be 0 in older inode's ea */
 507                         if (entry->e_hash != 0 && entry->e_hash != hash) {
 508                                 pctx->num = entry->e_hash;
 509                                 problem = PR_1_ATTR_HASH;
 510                                 goto fix;
 511                         }
 512                 } else {
 513                         blk64_t quota_blocks;
 514
 515                         problem = check_large_ea_inode(ctx, entry, pctx,
 516                                                        &quota_blocks);
 517                         if (problem != 0)
 518                                 goto fix;
 519
 520                         ea_ibody_quota->blocks += quota_blocks;
 521                         ea_ibody_quota->inodes++;
 522                 }
 523
 524                 /* If EA value is stored in external inode then it does not
 525                  * consume space here */
 526                 if (entry->e_value_inum == 0)
 527                         remain -= entry->e_value_size;
 528
 529                 entry = EXT2_EXT_ATTR_NEXT(entry);
 530         }
 531
 532         if (region_allocate(region, (char *)entry - (char *)header,
 533                             sizeof(__u32))) {
 534                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 535                 goto fix;
 536         }
 537 fix:
 538         if (region)
 539                 region_free(region);
 540         /*
 541          * it seems like a corruption. it's very unlikely we could repair
 542          * EA(s) in automatic fashion -bzzz
 543          */
 544         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
 545                 inc_ea_inode_refs(ctx, pctx,
 546                                   (struct ext2_ext_attr_entry *)start, end);
 547                 return;
 548         }
 549
 550         /* simply remove all possible EA(s) */
 551         *((__u32 *)header) = 0UL;
 552         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 553                                 EXT2_INODE_SIZE(sb), "pass1");
 554         ea_ibody_quota->blocks = 0;
 555         ea_ibody_quota->inodes = 0;
 556 }
 557
 558 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
 559         return (xtime & (1U << 31)) != 0 &&
 560                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
 561 }
 562
 563 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
 564         check_inode_extra_negative_epoch(inode->i_##xtime, \
 565                                          inode->i_##xtime##_extra)
 566
 567 /* When today's date is earlier than 2242, we assume that atimes,
 568  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
 569  * actually pre-1970 dates mis-encoded.
 570  */
 571 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
 572
 573 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
 574                                     struct ea_quota *ea_ibody_quota)
 575 {
 576         struct ext2_super_block *sb = ctx->fs->super;
 577         struct ext2_inode_large *inode;
 578         __u32 *eamagic;
 579         int min, max;
 580
 581         ea_ibody_quota->blocks = 0;
 582         ea_ibody_quota->inodes = 0;
 583
 584         inode = (struct ext2_inode_large *) pctx->inode;
 585         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
 586                 /* this isn't large inode. so, nothing to check */
 587                 return;
 588         }
 589
 590 #if 0
 591         printf("inode #%u, i_extra_size %d\n", pctx->ino,
 592                         inode->i_extra_isize);
 593 #endif
 594         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
 595         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
 596         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
 597         /*
 598          * For now we will allow i_extra_isize to be 0, but really
 599          * implementations should never allow i_extra_isize to be 0
 600          */
 601         if (inode->i_extra_isize &&
 602             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
 603              inode->i_extra_isize & 3)) {
 604                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
 605                         return;
 606                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
 607                         inode->i_extra_isize = sb->s_want_extra_isize;
 608                 else
 609                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
 610                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 611                                         EXT2_INODE_SIZE(sb), "pass1");
 612         }
 613
 614         /* check if there is no place for an EA header */
 615         if (inode->i_extra_isize >= max - sizeof(__u32))
 616                 return;
 617
 618         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 619                         inode->i_extra_isize);
 620         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
 621                 /* it seems inode has an extended attribute(s) in body */
 622                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
 623         }
 624
 625         /*
 626          * If the inode's extended atime (ctime, crtime, mtime) is stored in
 627          * the old, invalid format, repair it.
 628          */
 629         if (((sizeof(time_t) <= 4) ||
 630              (((sizeof(time_t) > 4) &&
 631                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
 632             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
 633              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
 634              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
 635              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
 636
 637                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
 638                         return;
 639
 640                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
 641                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
 642                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
 643                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
 644                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
 645                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
 646                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
 647                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
 648                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 649                                         EXT2_INODE_SIZE(sb), "pass1");
 650         }
 651
 652 }
 653
 654 /*
 655  * Check to see if the inode might really be a directory, despite i_mode
 656  *
 657  * This is a lot of complexity for something for which I'm not really
 658  * convinced happens frequently in the wild.  If for any reason this
 659  * causes any problems, take this code out.
 660  * [tytso:20070331.0827EDT]
 661  */
 662 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
 663                                 char *buf)
 664 {
 665         struct ext2_inode *inode = pctx->inode;
 666         struct ext2_dir_entry   *dirent;
 667         errcode_t               retval;
 668         blk64_t                 blk;
 669         unsigned int            i, rec_len, not_device = 0;
 670         int                     extent_fs;
 671         int                     inlinedata_fs;
 672
 673         /*
 674          * If the mode looks OK, we believe it.  If the first block in
 675          * the i_block array is 0, this cannot be a directory. If the
 676          * inode is extent-mapped, it is still the case that the latter
 677          * cannot be 0 - the magic number in the extent header would make
 678          * it nonzero.
 679          */
 680         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
 681             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
 682                 return;
 683
 684         /*
 685          * Check the block numbers in the i_block array for validity:
 686          * zero blocks are skipped (but the first one cannot be zero -
 687          * see above), other blocks are checked against the first and
 688          * max data blocks (from the the superblock) and against the
 689          * block bitmap. Any invalid block found means this cannot be
 690          * a directory.
 691          *
 692          * If there are non-zero blocks past the fourth entry, then
 693          * this cannot be a device file: we remember that for the next
 694          * check.
 695          *
 696          * For extent mapped files, we don't do any sanity checking:
 697          * just try to get the phys block of logical block 0 and run
 698          * with it.
 699          *
 700          * For inline data files, we just try to get the size of inline
 701          * data.  If it's true, we will treat it as a directory.
 702          */
 703
 704         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
 705         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
 706         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
 707                 size_t size;
 708                 __u32 dotdot;
 709                 unsigned int rec_len2;
 710                 struct ext2_dir_entry de;
 711
 712                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
 713                         return;
 714                 /*
 715                  * If the size isn't a multiple of 4, it's probably not a
 716                  * directory??
 717                  */
 718                 if (size & 3)
 719                         return;
 720                 /*
 721                  * If the first 10 bytes don't look like a directory entry,
 722                  * it's probably not a directory.
 723                  */
 724                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
 725                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
 726                        EXT2_DIR_REC_LEN(0));
 727                 dotdot = ext2fs_le32_to_cpu(dotdot);
 728                 de.inode = ext2fs_le32_to_cpu(de.inode);
 729                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
 730                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
 731                 if (dotdot >= ctx->fs->super->s_inodes_count ||
 732                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
 733                      dotdot != EXT2_ROOT_INO) ||
 734                     de.inode >= ctx->fs->super->s_inodes_count ||
 735                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
 736                      de.inode != 0) ||
 737                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
 738                               EXT4_INLINE_DATA_DOTDOT_SIZE)
 739                         return;
 740                 /* device files never have a "system.data" entry */
 741                 goto isdir;
 742         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
 743                 /* extent mapped */
 744                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
 745                                  &blk))
 746                         return;
 747                 /* device files are never extent mapped */
 748                 not_device++;
 749         } else {
 750                 for (i=0; i < EXT2_N_BLOCKS; i++) {
 751                         blk = inode->i_block[i];
 752                         if (!blk)
 753                                 continue;
 754                         if (i >= 4)
 755                                 not_device++;
 756
 757                         if (blk < ctx->fs->super->s_first_data_block ||
 758                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
 759                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
 760                                                            blk))
 761                                 return; /* Invalid block, can't be dir */
 762                 }
 763                 blk = inode->i_block[0];
 764         }
 765
 766         /*
 767          * If the mode says this is a device file and the i_links_count field
 768          * is sane and we have not ruled it out as a device file previously,
 769          * we declare it a device file, not a directory.
 770          */
 771         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
 772             (inode->i_links_count == 1) && !not_device)
 773                 return;
 774
 775         /* read the first block */
 776         ehandler_operation(_("reading directory block"));
 777         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
 778         ehandler_operation(0);
 779         if (retval)
 780                 return;
 781
 782         dirent = (struct ext2_dir_entry *) buf;
 783         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 784         if (retval)
 785                 return;
 786         if ((ext2fs_dirent_name_len(dirent) != 1) ||
 787             (dirent->name[0] != '.') ||
 788             (dirent->inode != pctx->ino) ||
 789             (rec_len < 12) ||
 790             (rec_len % 4) ||
 791             (rec_len >= ctx->fs->blocksize - 12))
 792                 return;
 793
 794         dirent = (struct ext2_dir_entry *) (buf + rec_len);
 795         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 796         if (retval)
 797                 return;
 798         if ((ext2fs_dirent_name_len(dirent) != 2) ||
 799             (dirent->name[0] != '.') ||
 800             (dirent->name[1] != '.') ||
 801             (rec_len < 12) ||
 802             (rec_len % 4))
 803                 return;
 804
 805 isdir:
 806         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
 807                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
 808                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
 809                                         EXT2_INODE_SIZE(ctx->fs->super),
 810                                         "check_is_really_dir");
 811         }
 812 }
 813
 814 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
 815                                      int flags, ext2_icount_t hint,
 816                                      ext2_icount_t *ret)
 817 {
 818         unsigned int            threshold;
 819         unsigned int            save_type;
 820         ext2_ino_t              num_dirs;
 821         errcode_t               retval;
 822         char                    *tdb_dir;
 823         int                     enable;
 824
 825         *ret = 0;
 826
 827         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
 828                            &tdb_dir);
 829         profile_get_uint(ctx->profile, "scratch_files",
 830                          "numdirs_threshold", 0, 0, &threshold);
 831         profile_get_boolean(ctx->profile, "scratch_files",
 832                             "icount", 0, 1, &enable);
 833
 834         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
 835         if (retval)
 836                 num_dirs = 1024;        /* Guess */
 837
 838         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
 839             (!threshold || num_dirs > threshold)) {
 840                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
 841                                                   flags, ret);
 842                 if (retval == 0)
 843                         return 0;
 844         }
 845         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
 846                                &save_type);
 847         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
 848                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
 849         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
 850         ctx->fs->default_bitmap_type = save_type;
 851         return retval;
 852 }
 853
 854 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
 855                                             e2fsck_t ctx,
 856                                             struct problem_context *pctx)
 857 {
 858         errcode_t retval;
 859         struct ext2_inode_large inode;
 860
 861         /*
 862          * Reread inode.  If we don't see checksum error, then this inode
 863          * has been fixed elsewhere.
 864          */
 865         ctx->stashed_ino = 0;
 866         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
 867                                         sizeof(inode));
 868         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
 869                 return retval;
 870         if (!retval)
 871                 return 0;
 872
 873         /*
 874          * Checksum still doesn't match.  That implies that the inode passes
 875          * all the sanity checks, so maybe the checksum is simply corrupt.
 876          * See if the user will go for fixing that.
 877          */
 878         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
 879                 return 0;
 880
 881         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
 882                                          sizeof(inode));
 883         return retval;
 884 }
 885
 886 static void reserve_block_for_root_repair(e2fsck_t ctx)
 887 {
 888         blk64_t         blk = 0;
 889         errcode_t       err;
 890         ext2_filsys     fs = ctx->fs;
 891
 892         ctx->root_repair_block = 0;
 893         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
 894                 return;
 895
 896         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 897         if (err)
 898                 return;
 899         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 900         ctx->root_repair_block = blk;
 901 }
 902
 903 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
 904 {
 905         blk64_t         blk = 0;
 906         errcode_t       err;
 907         ext2_filsys     fs = ctx->fs;
 908         static const char name[] = "lost+found";
 909         ext2_ino_t      ino;
 910
 911         ctx->lnf_repair_block = 0;
 912         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
 913                 return;
 914
 915         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 916         if (err)
 917                 return;
 918         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 919         ctx->lnf_repair_block = blk;
 920 }
 921
 922 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
 923                                          size_t *sz)
 924 {
 925         void *p;
 926         struct ext2_xattr_handle *handle;
 927         errcode_t retval;
 928
 929         retval = ext2fs_xattrs_open(fs, ino, &handle);
 930         if (retval)
 931                 return retval;
 932
 933         retval = ext2fs_xattrs_read(handle);
 934         if (retval)
 935                 goto err;
 936
 937         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
 938         if (retval)
 939                 goto err;
 940         ext2fs_free_mem(&p);
 941 err:
 942         (void) ext2fs_xattrs_close(&handle);
 943         return retval;
 944 }
 945
 946 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
 947                                     struct problem_context *pctx,
 948                                     int failed_csum)
 949 {
 950         if (!failed_csum)
 951                 return;
 952
 953         /*
 954          * If the inode failed the checksum and the user didn't
 955          * clear the inode, test the checksum again -- if it still
 956          * fails, ask the user if the checksum should be corrected.
 957          */
 958         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
 959         if (pctx->errcode)
 960                 ctx->flags |= E2F_FLAG_ABORT;
 961 }
 962 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
 963         do { \
 964                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
 965                 if ((ctx)->flags & E2F_FLAG_ABORT) \
 966                         return; \
 967         } while (0)
 968
 969 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
 970 {
 971         __u32 x;
 972         int i;
 973
 974         for (i = 0; i < EXT2_N_BLOCKS; i++) {
 975                 x = inode->i_block[i];
 976 #ifdef WORDS_BIGENDIAN
 977                 x = ext2fs_swab32(x);
 978 #endif
 979                 if (x >= ext2fs_blocks_count(fs->super))
 980                         return 0;
 981         }
 982
 983         return 1;
 984 }
 985
 986 /*
 987  * Figure out what to do with an inode that has both extents and inline data
 988  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
 989  */
 990 static int fix_inline_data_extents_file(e2fsck_t ctx,
 991                                         ext2_ino_t ino,
 992                                         struct ext2_inode *inode,
 993                                         int inode_size,
 994                                         struct problem_context *pctx)
 995 {
 996         size_t max_inline_ea_size;
 997         ext2_filsys fs = ctx->fs;
 998         int dirty = 0;
 999
1000         /* Both feature flags not set?  Just run the regular checks */
1001         if (!ext2fs_has_feature_extents(fs->super) &&
1002             !ext2fs_has_feature_inline_data(fs->super))
1003                 return 0;
1004
1005         /* Clear both flags if it's a special file */
1006         if (LINUX_S_ISCHR(inode->i_mode) ||
1007             LINUX_S_ISBLK(inode->i_mode) ||
1008             LINUX_S_ISFIFO(inode->i_mode) ||
1009             LINUX_S_ISSOCK(inode->i_mode)) {
1010                 check_extents_inlinedata(ctx, pctx);
1011                 return 0;
1012         }
1013
1014         /* If it looks like an extent tree, try to clear inlinedata */
1015         if (ext2fs_extent_header_verify(inode->i_block,
1016                                  sizeof(inode->i_block)) == 0 &&
1017             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1018                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1019                 dirty = 1;
1020                 goto out;
1021         }
1022
1023         /* If it looks short enough to be inline data, try to clear extents */
1024         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1025                 max_inline_ea_size = inode_size -
1026                                      (EXT2_GOOD_OLD_INODE_SIZE +
1027                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1028         else
1029                 max_inline_ea_size = 0;
1030         if (EXT2_I_SIZE(inode) <
1031             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1032             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1033                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1034                 dirty = 1;
1035                 goto out;
1036         }
1037
1038         /*
1039          * Too big for inline data, but no evidence of extent tree -
1040          * maybe it's a block map file?  If the mappings all look valid?
1041          */
1042         if (could_be_block_map(fs, inode) &&
1043             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1044 #ifdef WORDS_BIGENDIAN
1045                 int i;
1046
1047                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1048                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1049 #endif
1050
1051                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1052                 dirty = 1;
1053                 goto out;
1054         }
1055
1056         /* Oh well, just clear the busted inode. */
1057         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1058                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1059                 return -1;
1060         }
1061
1062 out:
1063         if (dirty)
1064                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1065
1066         return 0;
1067 }
1068
1069 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1070 {
1071         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1072         dgrp_t start = *group, grp;
1073         blk64_t blocks_to_read = 0;
1074         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1075
1076         if (ctx->readahead_kb == 0)
1077                 goto out;
1078
1079         /* Keep iterating groups until we have enough to readahead */
1080         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1081         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1082                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1083                         continue;
1084                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1085                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1086                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1087                                         inodes_per_block;
1088                 if (blocks_to_read * ctx->fs->blocksize >
1089                     ctx->readahead_kb * 1024)
1090                         break;
1091         }
1092
1093         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1094                                grp - start + 1);
1095         if (err == EAGAIN) {
1096                 ctx->readahead_kb /= 2;
1097                 err = 0;
1098         }
1099
1100 out:
1101         if (err) {
1102                 /* Error; disable itable readahead */
1103                 *group = ctx->fs->group_desc_count;
1104                 *next_ino = ctx->fs->super->s_inodes_count;
1105         } else {
1106                 /*
1107                  * Don't do more readahead until we've reached the first inode
1108                  * of the last inode scan buffer block for the last group.
1109                  */
1110                 *group = grp + 1;
1111                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1112                                      ctx->inode_buffer_blocks :
1113                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1114                                     ctx->fs->blocksize /
1115                                     EXT2_INODE_SIZE(ctx->fs->super);
1116                 inodes_in_group--;
1117                 *next_ino = inodes_in_group -
1118                             (inodes_in_group % inodes_per_buffer) + 1 +
1119                             (grp * ctx->fs->super->s_inodes_per_group);
1120         }
1121 }
1122
1123 /*
1124  * Check if the passed ino is one of the used superblock quota inodes.
1125  *
1126  * Before the quota inodes were journaled, older superblock quota inodes
1127  * were just regular files in the filesystem and not reserved inodes.  This
1128  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1129  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1130  */
1131 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1132 {
1133         enum quota_type qtype;
1134
1135         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1136                 if (*quota_sb_inump(sb, qtype) == ino)
1137                         return 1;
1138
1139         return 0;
1140 }
1141
1142 /*
1143  * Check if the passed ino is one of the reserved quota inodes.
1144  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1145  * inodes.  These inodes may or may not be in use by the quota feature.
1146  */
1147 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1148 {
1149         enum quota_type qtype;
1150
1151         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1152                 if (quota_type2inum(qtype, fs->super) == ino)
1153                         return 1;
1154
1155         return 0;
1156 }
1157
1158 void e2fsck_pass1(e2fsck_t ctx)
1159 {
1160         int     i;
1161         __u64   max_sizes;
1162         ext2_filsys fs = ctx->fs;
1163         ext2_ino_t      ino = 0;
1164         struct ext2_inode *inode = NULL;
1165         ext2_inode_scan scan = NULL;
1166         char            *block_buf = NULL;
1167 #ifdef RESOURCE_TRACK
1168         struct resource_track   rtrack;
1169 #endif
1170         unsigned char   frag, fsize;
1171         struct          problem_context pctx;
1172         struct          scan_callback_struct scan_struct;
1173         struct ext2_super_block *sb = ctx->fs->super;
1174         const char      *old_op;
1175         const char      *eop_next_inode = _("getting next inode from scan");
1176         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1177         int             low_dtime_check = 1;
1178         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1179         unsigned int    bufsize;
1180         int             failed_csum = 0;
1181         ext2_ino_t      ino_threshold = 0;
1182         dgrp_t          ra_group = 0;
1183         struct ea_quota ea_ibody_quota;
1184
1185         init_resource_track(&rtrack, ctx->fs->io);
1186         clear_problem_context(&pctx);
1187
1188         /* If we can do readahead, figure out how many groups to pull in. */
1189         if (!e2fsck_can_readahead(ctx->fs))
1190                 ctx->readahead_kb = 0;
1191         else if (ctx->readahead_kb == ~0ULL)
1192                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1193         pass1_readahead(ctx, &ra_group, &ino_threshold);
1194
1195         if (!(ctx->options & E2F_OPT_PREEN))
1196                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1197
1198         if (ext2fs_has_feature_dir_index(fs->super) &&
1199             !(ctx->options & E2F_OPT_NO)) {
1200                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1201                         ctx->dirs_to_hash = 0;
1202         }
1203
1204 #ifdef MTRACE
1205         mtrace_print("Pass 1");
1206 #endif
1207
1208 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1209
1210         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1211                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1212                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1213                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1214                 max_sizes = (max_sizes * (1UL << i));
1215                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1216         }
1217 #undef EXT2_BPP
1218
1219         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1220         extent_fs = ext2fs_has_feature_extents(sb);
1221         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1222         casefold_fs = ext2fs_has_feature_casefold(sb);
1223
1224         /*
1225          * Allocate bitmaps structures
1226          */
1227         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1228                                                     EXT2FS_BMAP64_RBTREE,
1229                                                     "inode_used_map",
1230                                                     &ctx->inode_used_map);
1231         if (pctx.errcode) {
1232                 pctx.num = 1;
1233                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1234                 ctx->flags |= E2F_FLAG_ABORT;
1235                 return;
1236         }
1237         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1238                         _("directory inode map"),
1239                         EXT2FS_BMAP64_AUTODIR,
1240                         "inode_dir_map", &ctx->inode_dir_map);
1241         if (pctx.errcode) {
1242                 pctx.num = 2;
1243                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1244                 ctx->flags |= E2F_FLAG_ABORT;
1245                 return;
1246         }
1247         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1248                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1249                         "inode_reg_map", &ctx->inode_reg_map);
1250         if (pctx.errcode) {
1251                 pctx.num = 6;
1252                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1253                 ctx->flags |= E2F_FLAG_ABORT;
1254                 return;
1255         }
1256         pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1257                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1258                         "block_found_map", &ctx->block_found_map);
1259         if (pctx.errcode) {
1260                 pctx.num = 1;
1261                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1262                 ctx->flags |= E2F_FLAG_ABORT;
1263                 return;
1264         }
1265         pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1266                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1267                         "block_metadata_map", &ctx->block_metadata_map);
1268         if (pctx.errcode) {
1269                 pctx.num = 1;
1270                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1271                 ctx->flags |= E2F_FLAG_ABORT;
1272                 return;
1273         }
1274         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1275                                            &ctx->inode_link_info);
1276         if (pctx.errcode) {
1277                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1278                 ctx->flags |= E2F_FLAG_ABORT;
1279                 return;
1280         }
1281         bufsize = inode_size;
1282         if (bufsize < sizeof(struct ext2_inode_large))
1283                 bufsize = sizeof(struct ext2_inode_large);
1284         inode = (struct ext2_inode *)
1285                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1286
1287         inodes_to_process = (struct process_inode_block *)
1288                 e2fsck_allocate_memory(ctx,
1289                                        (ctx->process_inode_size *
1290                                         sizeof(struct process_inode_block)),
1291                                        "array of inodes to process");
1292         process_inode_count = 0;
1293
1294         pctx.errcode = ext2fs_init_dblist(fs, 0);
1295         if (pctx.errcode) {
1296                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1297                 ctx->flags |= E2F_FLAG_ABORT;
1298                 goto endit;
1299         }
1300
1301         /*
1302          * If the last orphan field is set, clear it, since the pass1
1303          * processing will automatically find and clear the orphans.
1304          * In the future, we may want to try using the last_orphan
1305          * linked list ourselves, but for now, we clear it so that the
1306          * ext3 mount code won't get confused.
1307          */
1308         if (!(ctx->options & E2F_OPT_READONLY)) {
1309                 if (fs->super->s_last_orphan) {
1310                         fs->super->s_last_orphan = 0;
1311                         ext2fs_mark_super_dirty(fs);
1312                 }
1313         }
1314
1315         mark_table_blocks(ctx);
1316         pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1317                                                 &ctx->block_found_map);
1318         if (pctx.errcode) {
1319                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1320                 ctx->flags |= E2F_FLAG_ABORT;
1321                 goto endit;
1322         }
1323         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1324                                                     "block interate buffer");
1325         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1326                 e2fsck_use_inode_shortcuts(ctx, 1);
1327         e2fsck_intercept_block_allocations(ctx);
1328         old_op = ehandler_operation(_("opening inode scan"));
1329         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1330                                               &scan);
1331         ehandler_operation(old_op);
1332         if (pctx.errcode) {
1333                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1334                 ctx->flags |= E2F_FLAG_ABORT;
1335                 goto endit;
1336         }
1337         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1338                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1339         ctx->stashed_inode = inode;
1340         scan_struct.ctx = ctx;
1341         scan_struct.block_buf = block_buf;
1342         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1343         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1344                                               ctx->fs->group_desc_count)))
1345                 goto endit;
1346         if ((fs->super->s_wtime &&
1347              fs->super->s_wtime < fs->super->s_inodes_count) ||
1348             (fs->super->s_mtime &&
1349              fs->super->s_mtime < fs->super->s_inodes_count) ||
1350             (fs->super->s_mkfs_time &&
1351              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1352                 low_dtime_check = 0;
1353
1354         if (ext2fs_has_feature_mmp(fs->super) &&
1355             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1356             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1357                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1358                                           fs->super->s_mmp_block);
1359
1360         /* Set up ctx->lost_and_found if possible */
1361         (void) e2fsck_get_lost_and_found(ctx, 0);
1362
1363         while (1) {
1364                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1365                         if (e2fsck_mmp_update(fs))
1366                                 fatal_error(ctx, 0);
1367                 }
1368                 old_op = ehandler_operation(eop_next_inode);
1369                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1370                                                           inode, inode_size);
1371                 if (ino > ino_threshold)
1372                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1373                 ehandler_operation(old_op);
1374                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1375                         goto endit;
1376                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1377                         /*
1378                          * If badblocks says badblocks is bad, offer to clear
1379                          * the list, update the in-core bb list, and restart
1380                          * the inode scan.
1381                          */
1382                         if (ino == EXT2_BAD_INO &&
1383                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1384                                         &pctx)) {
1385                                 errcode_t err;
1386
1387                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1388                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1389                                 ctx->fs->badblocks = NULL;
1390                                 err = ext2fs_read_bb_inode(ctx->fs,
1391                                                         &ctx->fs->badblocks);
1392                                 if (err) {
1393                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1394                                                     &pctx);
1395                                         ctx->flags |= E2F_FLAG_ABORT;
1396                                         goto endit;
1397                                 }
1398                                 err = ext2fs_inode_scan_goto_blockgroup(scan,
1399                                                                         0);
1400                                 if (err) {
1401                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1402                                                     &pctx);
1403                                         ctx->flags |= E2F_FLAG_ABORT;
1404                                         goto endit;
1405                                 }
1406                                 continue;
1407                         }
1408                         if (!ctx->inode_bb_map)
1409                                 alloc_bb_map(ctx);
1410                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1411                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1412                         continue;
1413                 }
1414                 if (pctx.errcode &&
1415                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1416                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1417                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1418                         ctx->flags |= E2F_FLAG_ABORT;
1419                         goto endit;
1420                 }
1421                 if (!ino)
1422                         break;
1423                 pctx.ino = ino;
1424                 pctx.inode = inode;
1425                 ctx->stashed_ino = ino;
1426
1427                 /* Clear trashed inode? */
1428                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1429                     inode->i_links_count > 0 &&
1430                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1431                         pctx.errcode = 0;
1432                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1433                 }
1434                 failed_csum = pctx.errcode != 0;
1435
1436                 /*
1437                  * Check for inodes who might have been part of the
1438                  * orphaned list linked list.  They should have gotten
1439                  * dealt with by now, unless the list had somehow been
1440                  * corrupted.
1441                  *
1442                  * FIXME: In the future, inodes which are still in use
1443                  * (and which are therefore) pending truncation should
1444                  * be handled specially.  Right now we just clear the
1445                  * dtime field, and the normal e2fsck handling of
1446                  * inodes where i_size and the inode blocks are
1447                  * inconsistent is to fix i_size, instead of releasing
1448                  * the extra blocks.  This won't catch the inodes that
1449                  * was at the end of the orphan list, but it's better
1450                  * than nothing.  The right answer is that there
1451                  * shouldn't be any bugs in the orphan list handling.  :-)
1452                  */
1453                 if (inode->i_dtime && low_dtime_check &&
1454                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1455                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1456                                 inode->i_dtime = inode->i_links_count ?
1457                                         0 : ctx->now;
1458                                 e2fsck_write_inode(ctx, ino, inode,
1459                                                    "pass1");
1460                                 failed_csum = 0;
1461                         }
1462                 }
1463
1464                 if (inode->i_links_count) {
1465                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1466                                            ino, inode->i_links_count);
1467                         if (pctx.errcode) {
1468                                 pctx.num = inode->i_links_count;
1469                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1470                                 ctx->flags |= E2F_FLAG_ABORT;
1471                                 goto endit;
1472                         }
1473                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1474                            !quota_inum_is_reserved(fs, ino)) {
1475                         if (!inode->i_dtime && inode->i_mode) {
1476                                 if (fix_problem(ctx,
1477                                             PR_1_ZERO_DTIME, &pctx)) {
1478                                         inode->i_dtime = ctx->now;
1479                                         e2fsck_write_inode(ctx, ino, inode,
1480                                                            "pass1");
1481                                         failed_csum = 0;
1482                                 }
1483                         }
1484                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1485                         continue;
1486                 }
1487
1488                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1489                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1490                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1491                      (!casefold_fs &&
1492                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1493                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1494                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1495                 }
1496
1497                 /* Conflicting inlinedata/extents inode flags? */
1498                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1499                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1500                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1501                                                                inode_size,
1502                                                                &pctx);
1503                         if (res < 0) {
1504                                 /* skip FINISH_INODE_LOOP */
1505                                 continue;
1506                         }
1507                 }
1508
1509                 /* Test for incorrect inline_data flags settings. */
1510                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1511                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1512                         size_t size = 0;
1513
1514                         pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1515                         if (!pctx.errcode &&
1516                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1517                                 ext2fs_set_feature_inline_data(sb);
1518                                 ext2fs_mark_super_dirty(fs);
1519                                 inlinedata_fs = 1;
1520                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1521                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1522                                 /* skip FINISH_INODE_LOOP */
1523                                 continue;
1524                         }
1525                 }
1526
1527                 /* Test for inline data flag but no attr */
1528                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1529                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1530                         size_t size = 0;
1531                         errcode_t err;
1532                         int flags;
1533
1534                         flags = fs->flags;
1535                         if (failed_csum)
1536                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1537                         err = get_inline_data_ea_size(fs, ino, &size);
1538                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1539                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1540
1541                         switch (err) {
1542                         case 0:
1543                                 /* Everything is awesome... */
1544                                 break;
1545                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1546                         case EXT2_ET_BAD_EA_HASH:
1547                         case EXT2_ET_BAD_EA_HEADER:
1548                         case EXT2_ET_EA_BAD_NAME_LEN:
1549                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1550                         case EXT2_ET_EA_KEY_NOT_FOUND:
1551                         case EXT2_ET_EA_NO_SPACE:
1552                         case EXT2_ET_MISSING_EA_FEATURE:
1553                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1554                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1555                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1556                         case EXT2_ET_NO_INLINE_DATA:
1557                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1558                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1559                         case EXT2_ET_EA_INODE_CORRUPTED:
1560                                 /* broken EA or no system.data EA; truncate */
1561                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1562                                                 &pctx)) {
1563                                         err = ext2fs_inode_size_set(fs, inode, 0);
1564                                         if (err) {
1565                                                 pctx.errcode = err;
1566                                                 ctx->flags |= E2F_FLAG_ABORT;
1567                                                 goto endit;
1568                                         }
1569                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1570                                         memset(&inode->i_block, 0,
1571                                                sizeof(inode->i_block));
1572                                         e2fsck_write_inode(ctx, ino, inode,
1573                                                            "pass1");
1574                                         failed_csum = 0;
1575                                 }
1576                                 break;
1577                         default:
1578                                 /* Some other kind of non-xattr error? */
1579                                 pctx.errcode = err;
1580                                 ctx->flags |= E2F_FLAG_ABORT;
1581                                 goto endit;
1582                         }
1583                 }
1584
1585                 /*
1586                  * Test for incorrect extent flag settings.
1587                  *
1588                  * On big-endian machines we must be careful:
1589                  * When the inode is read, the i_block array is not swapped
1590                  * if the extent flag is set.  Therefore if we are testing
1591                  * for or fixing a wrongly-set flag, we must potentially
1592                  * (un)swap before testing, or after fixing.
1593                  */
1594
1595                 /*
1596                  * In this case the extents flag was set when read, so
1597                  * extent_header_verify is ok.  If the inode is cleared,
1598                  * no need to swap... so no extra swapping here.
1599                  */
1600                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1601                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1602                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1603                         if ((ext2fs_extent_header_verify(inode->i_block,
1604                                                  sizeof(inode->i_block)) == 0) &&
1605                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1606                                 ext2fs_set_feature_extents(sb);
1607                                 ext2fs_mark_super_dirty(fs);
1608                                 extent_fs = 1;
1609                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1610                         clear_inode:
1611                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1612                                 if (ino == EXT2_BAD_INO)
1613                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1614                                                                  ino);
1615                                 /* skip FINISH_INODE_LOOP */
1616                                 continue;
1617                         }
1618                 }
1619
1620                 /*
1621                  * For big-endian machines:
1622                  * If the inode didn't have the extents flag set when it
1623                  * was read, then the i_blocks array was swapped.  To test
1624                  * as an extents header, we must swap it back first.
1625                  * IF we then set the extents flag, the entire i_block
1626                  * array must be un/re-swapped to make it proper extents data.
1627                  */
1628                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1629                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1630                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1631                     (LINUX_S_ISREG(inode->i_mode) ||
1632                      LINUX_S_ISDIR(inode->i_mode))) {
1633                         void *ehp;
1634 #ifdef WORDS_BIGENDIAN
1635                         __u32 tmp_block[EXT2_N_BLOCKS];
1636
1637                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1638                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1639                         ehp = tmp_block;
1640 #else
1641                         ehp = inode->i_block;
1642 #endif
1643                         if ((ext2fs_extent_header_verify(ehp,
1644                                          sizeof(inode->i_block)) == 0) &&
1645                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1646                                 inode->i_flags |= EXT4_EXTENTS_FL;
1647 #ifdef WORDS_BIGENDIAN
1648                                 memcpy(inode->i_block, tmp_block,
1649                                        sizeof(inode->i_block));
1650 #endif
1651                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1652                                 failed_csum = 0;
1653                         }
1654                 }
1655
1656                 if (ino == EXT2_BAD_INO) {
1657                         struct process_block_struct pb;
1658
1659                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1660                              inode->i_gid || inode->i_links_count ||
1661                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1662                              inode->i_file_acl) &&
1663                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1664                                 memset(inode, 0, sizeof(struct ext2_inode));
1665                                 e2fsck_write_inode(ctx, ino, inode,
1666                                                    "clear bad inode");
1667                                 failed_csum = 0;
1668                         }
1669
1670                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1671                                                           &pb.fs_meta_blocks);
1672                         if (pctx.errcode) {
1673                                 pctx.num = 4;
1674                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1675                                 ctx->flags |= E2F_FLAG_ABORT;
1676                                 goto endit;
1677                         }
1678                         pb.ino = EXT2_BAD_INO;
1679                         pb.num_blocks = pb.last_block = 0;
1680                         pb.last_db_block = -1;
1681                         pb.num_illegal_blocks = 0;
1682                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1683                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1684                         pb.inode = inode;
1685                         pb.pctx = &pctx;
1686                         pb.ctx = ctx;
1687                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1688                                      block_buf, process_bad_block, &pb);
1689                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1690                         if (pctx.errcode) {
1691                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1692                                 ctx->flags |= E2F_FLAG_ABORT;
1693                                 goto endit;
1694                         }
1695                         if (pb.bbcheck)
1696                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1697                                 ctx->flags |= E2F_FLAG_ABORT;
1698                                 goto endit;
1699                         }
1700                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1701                         clear_problem_context(&pctx);
1702                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1703                         continue;
1704                 } else if (ino == EXT2_ROOT_INO) {
1705                         /*
1706                          * Make sure the root inode is a directory; if
1707                          * not, offer to clear it.  It will be
1708                          * regenerated in pass #3.
1709                          */
1710                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1711                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1712                                         goto clear_inode;
1713                         }
1714                         /*
1715                          * If dtime is set, offer to clear it.  mke2fs
1716                          * version 0.2b created filesystems with the
1717                          * dtime field set for the root and lost+found
1718                          * directories.  We won't worry about
1719                          * /lost+found, since that can be regenerated
1720                          * easily.  But we will fix the root directory
1721                          * as a special case.
1722                          */
1723                         if (inode->i_dtime && inode->i_links_count) {
1724                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1725                                         inode->i_dtime = 0;
1726                                         e2fsck_write_inode(ctx, ino, inode,
1727                                                            "pass1");
1728                                         failed_csum = 0;
1729                                 }
1730                         }
1731                 } else if (ino == EXT2_JOURNAL_INO) {
1732                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1733                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1734                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1735                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1736                                                 &pctx)) {
1737                                         inode->i_mode = LINUX_S_IFREG;
1738                                         e2fsck_write_inode(ctx, ino, inode,
1739                                                            "pass1");
1740                                         failed_csum = 0;
1741                                 }
1742                                 check_blocks(ctx, &pctx, block_buf, NULL);
1743                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1744                                 continue;
1745                         }
1746                         if ((inode->i_links_count ||
1747                              inode->i_blocks || inode->i_block[0]) &&
1748                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1749                                         &pctx)) {
1750                                 memset(inode, 0, inode_size);
1751                                 ext2fs_icount_store(ctx->inode_link_info,
1752                                                     ino, 0);
1753                                 e2fsck_write_inode_full(ctx, ino, inode,
1754                                                         inode_size, "pass1");
1755                                 failed_csum = 0;
1756                         }
1757                 } else if (quota_inum_is_reserved(fs, ino)) {
1758                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1759                         if (ext2fs_has_feature_quota(fs->super) &&
1760                             quota_inum_is_super(fs->super, ino)) {
1761                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1762                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1763                                                         &pctx)) {
1764                                         inode->i_mode = LINUX_S_IFREG;
1765                                         e2fsck_write_inode(ctx, ino, inode,
1766                                                         "pass1");
1767                                         failed_csum = 0;
1768                                 }
1769                                 check_blocks(ctx, &pctx, block_buf, NULL);
1770                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1771                                 continue;
1772                         }
1773                         if ((inode->i_links_count ||
1774                              inode->i_blocks || inode->i_block[0]) &&
1775                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1776                                         &pctx)) {
1777                                 memset(inode, 0, inode_size);
1778                                 ext2fs_icount_store(ctx->inode_link_info,
1779                                                     ino, 0);
1780                                 e2fsck_write_inode_full(ctx, ino, inode,
1781                                                         inode_size, "pass1");
1782                                 failed_csum = 0;
1783                         }
1784                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1785                         problem_t problem = 0;
1786
1787                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1788                         if (ino == EXT2_BOOT_LOADER_INO) {
1789                                 if (LINUX_S_ISDIR(inode->i_mode))
1790                                         problem = PR_1_RESERVED_BAD_MODE;
1791                         } else if (ino == EXT2_RESIZE_INO) {
1792                                 if (inode->i_mode &&
1793                                     !LINUX_S_ISREG(inode->i_mode))
1794                                         problem = PR_1_RESERVED_BAD_MODE;
1795                         } else {
1796                                 if (inode->i_mode != 0)
1797                                         problem = PR_1_RESERVED_BAD_MODE;
1798                         }
1799                         if (problem) {
1800                                 if (fix_problem(ctx, problem, &pctx)) {
1801                                         inode->i_mode = 0;
1802                                         e2fsck_write_inode(ctx, ino, inode,
1803                                                            "pass1");
1804                                         failed_csum = 0;
1805                                 }
1806                         }
1807                         check_blocks(ctx, &pctx, block_buf, NULL);
1808                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1809                         continue;
1810                 }
1811
1812                 if (!inode->i_links_count) {
1813                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1814                         continue;
1815                 }
1816                 /*
1817                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1818                  * deleted files.  Oops.
1819                  *
1820                  * Since all new ext2 implementations get this right,
1821                  * we now assume that the case of non-zero
1822                  * i_links_count and non-zero dtime means that we
1823                  * should keep the file, not delete it.
1824                  *
1825                  */
1826                 if (inode->i_dtime) {
1827                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1828                                 inode->i_dtime = 0;
1829                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1830                                 failed_csum = 0;
1831                         }
1832                 }
1833
1834                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1835                 switch (fs->super->s_creator_os) {
1836                     case EXT2_OS_HURD:
1837                         frag = inode->osd2.hurd2.h_i_frag;
1838                         fsize = inode->osd2.hurd2.h_i_fsize;
1839                         break;
1840                     default:
1841                         frag = fsize = 0;
1842                 }
1843
1844                 if (inode->i_faddr || frag || fsize ||
1845                     (!ext2fs_has_feature_largedir(fs->super) &&
1846                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1847                         mark_inode_bad(ctx, ino);
1848                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1849                     !ext2fs_has_feature_64bit(fs->super) &&
1850                     inode->osd2.linux2.l_i_file_acl_high != 0)
1851                         mark_inode_bad(ctx, ino);
1852                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1853                     !ext2fs_has_feature_huge_file(fs->super) &&
1854                     (inode->osd2.linux2.l_i_blocks_hi != 0))
1855                         mark_inode_bad(ctx, ino);
1856                 if (inode->i_flags & EXT2_IMAGIC_FL) {
1857                         if (imagic_fs) {
1858                                 if (!ctx->inode_imagic_map)
1859                                         alloc_imagic_map(ctx);
1860                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1861                                                          ino);
1862                         } else {
1863                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1864                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
1865                                         e2fsck_write_inode(ctx, ino,
1866                                                            inode, "pass1");
1867                                         failed_csum = 0;
1868                                 }
1869                         }
1870                 }
1871
1872                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1873                 check_is_really_dir(ctx, &pctx, block_buf);
1874
1875                 /*
1876                  * ext2fs_inode_has_valid_blocks2 does not actually look
1877                  * at i_block[] values, so not endian-sensitive here.
1878                  */
1879                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1880                     LINUX_S_ISLNK(inode->i_mode) &&
1881                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1882                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1883                         inode->i_flags &= ~EXT4_EXTENTS_FL;
1884                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1885                         failed_csum = 0;
1886                 }
1887
1888                 if (LINUX_S_ISDIR(inode->i_mode)) {
1889                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1890                         e2fsck_add_dir_info(ctx, ino, 0);
1891                         ctx->fs_directory_count++;
1892                         if (inode->i_flags & EXT4_ENCRYPT_FL)
1893                                 add_encrypted_dir(ctx, ino);
1894                         if (inode->i_flags & EXT4_CASEFOLD_FL)
1895                                 add_casefolded_dir(ctx, ino);
1896                 } else if (LINUX_S_ISREG (inode->i_mode)) {
1897                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1898                         ctx->fs_regular_count++;
1899                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1900                            e2fsck_pass1_check_device_inode(fs, inode)) {
1901                         check_extents_inlinedata(ctx, &pctx);
1902                         check_immutable(ctx, &pctx);
1903                         check_size(ctx, &pctx);
1904                         ctx->fs_chardev_count++;
1905                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1906                            e2fsck_pass1_check_device_inode(fs, inode)) {
1907                         check_extents_inlinedata(ctx, &pctx);
1908                         check_immutable(ctx, &pctx);
1909                         check_size(ctx, &pctx);
1910                         ctx->fs_blockdev_count++;
1911                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1912                            e2fsck_pass1_check_symlink(fs, ino, inode,
1913                                                       block_buf)) {
1914                         check_immutable(ctx, &pctx);
1915                         ctx->fs_symlinks_count++;
1916                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1917                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1918                                 continue;
1919                         } else if (ext2fs_is_fast_symlink(inode)) {
1920                                 ctx->fs_fast_symlinks_count++;
1921                                 check_blocks(ctx, &pctx, block_buf,
1922                                              &ea_ibody_quota);
1923                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1924                                 continue;
1925                         }
1926                 }
1927                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
1928                          e2fsck_pass1_check_device_inode(fs, inode)) {
1929                         check_extents_inlinedata(ctx, &pctx);
1930                         check_immutable(ctx, &pctx);
1931                         check_size(ctx, &pctx);
1932                         ctx->fs_fifo_count++;
1933                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
1934                            e2fsck_pass1_check_device_inode(fs, inode)) {
1935                         check_extents_inlinedata(ctx, &pctx);
1936                         check_immutable(ctx, &pctx);
1937                         check_size(ctx, &pctx);
1938                         ctx->fs_sockets_count++;
1939                 } else
1940                         mark_inode_bad(ctx, ino);
1941                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1942                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
1943                         if (inode->i_block[EXT2_IND_BLOCK])
1944                                 ctx->fs_ind_count++;
1945                         if (inode->i_block[EXT2_DIND_BLOCK])
1946                                 ctx->fs_dind_count++;
1947                         if (inode->i_block[EXT2_TIND_BLOCK])
1948                                 ctx->fs_tind_count++;
1949                 }
1950                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1951                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
1952                     (inode->i_block[EXT2_IND_BLOCK] ||
1953                      inode->i_block[EXT2_DIND_BLOCK] ||
1954                      inode->i_block[EXT2_TIND_BLOCK] ||
1955                      ext2fs_file_acl_block(fs, inode))) {
1956                         struct process_inode_block *itp;
1957
1958                         itp = &inodes_to_process[process_inode_count];
1959                         itp->ino = ino;
1960                         itp->ea_ibody_quota = ea_ibody_quota;
1961                         if (inode_size < sizeof(struct ext2_inode_large))
1962                                 memcpy(&itp->inode, inode, inode_size);
1963                         else
1964                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
1965                         process_inode_count++;
1966                 } else
1967                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
1968
1969                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1970
1971                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1972                         goto endit;
1973
1974                 if (process_inode_count >= ctx->process_inode_size) {
1975                         process_inodes(ctx, block_buf);
1976
1977                         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1978                                 goto endit;
1979                 }
1980         }
1981         process_inodes(ctx, block_buf);
1982         ext2fs_close_inode_scan(scan);
1983         scan = NULL;
1984
1985         reserve_block_for_root_repair(ctx);
1986         reserve_block_for_lnf_repair(ctx);
1987
1988         /*
1989          * If any extended attribute blocks' reference counts need to
1990          * be adjusted, either up (ctx->refcount_extra), or down
1991          * (ctx->refcount), then fix them.
1992          */
1993         if (ctx->refcount) {
1994                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
1995                 ea_refcount_free(ctx->refcount);
1996                 ctx->refcount = 0;
1997         }
1998         if (ctx->refcount_extra) {
1999                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
2000                                         block_buf, +1);
2001                 ea_refcount_free(ctx->refcount_extra);
2002                 ctx->refcount_extra = 0;
2003         }
2004
2005         if (ctx->ea_block_quota_blocks) {
2006                 ea_refcount_free(ctx->ea_block_quota_blocks);
2007                 ctx->ea_block_quota_blocks = 0;
2008         }
2009
2010         if (ctx->ea_block_quota_inodes) {
2011                 ea_refcount_free(ctx->ea_block_quota_inodes);
2012                 ctx->ea_block_quota_inodes = 0;
2013         }
2014
2015         if (ctx->invalid_bitmaps)
2016                 handle_fs_bad_blocks(ctx);
2017
2018         /* We don't need the block_ea_map any more */
2019         if (ctx->block_ea_map) {
2020                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2021                 ctx->block_ea_map = 0;
2022         }
2023
2024         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2025                 clear_problem_context(&pctx);
2026                 pctx.errcode = ext2fs_create_resize_inode(fs);
2027                 if (pctx.errcode) {
2028                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2029                                          &pctx)) {
2030                                 ctx->flags |= E2F_FLAG_ABORT;
2031                                 goto endit;
2032                         }
2033                         pctx.errcode = 0;
2034                 }
2035                 if (!pctx.errcode) {
2036                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2037                                           "recreate inode");
2038                         inode->i_mtime = ctx->now;
2039                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2040                                            "recreate inode");
2041                 }
2042                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2043         }
2044
2045         if (ctx->flags & E2F_FLAG_RESTART) {
2046                 /*
2047                  * Only the master copy of the superblock and block
2048                  * group descriptors are going to be written during a
2049                  * restart, so set the superblock to be used to be the
2050                  * master superblock.
2051                  */
2052                 ctx->use_superblock = 0;
2053                 unwind_pass1(fs);
2054                 goto endit;
2055         }
2056
2057         if (ctx->block_dup_map) {
2058                 if (ctx->options & E2F_OPT_PREEN) {
2059                         clear_problem_context(&pctx);
2060                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2061                 }
2062                 e2fsck_pass1_dupblocks(ctx, block_buf);
2063         }
2064         ctx->flags |= E2F_FLAG_ALLOC_OK;
2065         ext2fs_free_mem(&inodes_to_process);
2066 endit:
2067         e2fsck_use_inode_shortcuts(ctx, 0);
2068
2069         if (scan)
2070                 ext2fs_close_inode_scan(scan);
2071         if (block_buf)
2072                 ext2fs_free_mem(&block_buf);
2073         if (inode)
2074                 ext2fs_free_mem(&inode);
2075
2076         /*
2077          * The l+f inode may have been cleared, so zap it now and
2078          * later passes will recalculate it if necessary
2079          */
2080         ctx->lost_and_found = 0;
2081
2082         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2083                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2084         else
2085                 ctx->invalid_bitmaps++;
2086 }
2087 #undef FINISH_INODE_LOOP
2088
2089 /*
2090  * When the inode_scan routines call this callback at the end of the
2091  * glock group, call process_inodes.
2092  */
2093 static errcode_t scan_callback(ext2_filsys fs,
2094                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
2095                                dgrp_t group, void * priv_data)
2096 {
2097         struct scan_callback_struct *scan_struct;
2098         e2fsck_t ctx;
2099
2100         scan_struct = (struct scan_callback_struct *) priv_data;
2101         ctx = scan_struct->ctx;
2102
2103         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf);
2104
2105         if (ctx->progress)
2106                 if ((ctx->progress)(ctx, 1, group+1,
2107                                     ctx->fs->group_desc_count))
2108                         return EXT2_ET_CANCEL_REQUESTED;
2109
2110         return 0;
2111 }
2112
2113 /*
2114  * Process the inodes in the "inodes to process" list.
2115  */
2116 static void process_inodes(e2fsck_t ctx, char *block_buf)
2117 {
2118         int                     i;
2119         struct ext2_inode       *old_stashed_inode;
2120         ext2_ino_t              old_stashed_ino;
2121         const char              *old_operation;
2122         char                    buf[80];
2123         struct problem_context  pctx;
2124
2125 #if 0
2126         printf("begin process_inodes: ");
2127 #endif
2128         if (process_inode_count == 0)
2129                 return;
2130         old_operation = ehandler_operation(0);
2131         old_stashed_inode = ctx->stashed_inode;
2132         old_stashed_ino = ctx->stashed_ino;
2133         qsort(inodes_to_process, process_inode_count,
2134                       sizeof(struct process_inode_block), process_inode_cmp);
2135         clear_problem_context(&pctx);
2136         for (i=0; i < process_inode_count; i++) {
2137                 pctx.inode = ctx->stashed_inode =
2138                         (struct ext2_inode *) &inodes_to_process[i].inode;
2139                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2140
2141 #if 0
2142                 printf("%u ", pctx.ino);
2143 #endif
2144                 sprintf(buf, _("reading indirect blocks of inode %u"),
2145                         pctx.ino);
2146                 ehandler_operation(buf);
2147                 check_blocks(ctx, &pctx, block_buf,
2148                              &inodes_to_process[i].ea_ibody_quota);
2149                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2150                         break;
2151         }
2152         ctx->stashed_inode = old_stashed_inode;
2153         ctx->stashed_ino = old_stashed_ino;
2154         process_inode_count = 0;
2155 #if 0
2156         printf("end process inodes\n");
2157 #endif
2158         ehandler_operation(old_operation);
2159 }
2160
2161 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
2162 {
2163         const struct process_inode_block *ib_a =
2164                 (const struct process_inode_block *) a;
2165         const struct process_inode_block *ib_b =
2166                 (const struct process_inode_block *) b;
2167         int     ret;
2168
2169         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
2170                ib_b->inode.i_block[EXT2_IND_BLOCK]);
2171         if (ret == 0)
2172                 /*
2173                  * We only call process_inodes() for non-extent
2174                  * inodes, so it's OK to pass NULL to
2175                  * ext2fs_file_acl_block() here.
2176                  */
2177                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
2178                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
2179         if (ret == 0)
2180                 ret = ib_a->ino - ib_b->ino;
2181         return ret;
2182 }
2183
2184 /*
2185  * Mark an inode as being bad in some what
2186  */
2187 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
2188 {
2189         struct          problem_context pctx;
2190
2191         if (!ctx->inode_bad_map) {
2192                 clear_problem_context(&pctx);
2193
2194                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2195                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
2196                                 "inode_bad_map", &ctx->inode_bad_map);
2197                 if (pctx.errcode) {
2198                         pctx.num = 3;
2199                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2200                         /* Should never get here */
2201                         ctx->flags |= E2F_FLAG_ABORT;
2202                         return;
2203                 }
2204         }
2205         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
2206 }
2207
2208 static void add_encrypted_dir(e2fsck_t ctx, ino_t ino)
2209 {
2210         struct          problem_context pctx;
2211
2212         if (!ctx->encrypted_dirs) {
2213                 pctx.errcode = ext2fs_u32_list_create(&ctx->encrypted_dirs, 0);
2214                 if (pctx.errcode)
2215                         goto error;
2216         }
2217         pctx.errcode = ext2fs_u32_list_add(ctx->encrypted_dirs, ino);
2218         if (pctx.errcode == 0)
2219                 return;
2220 error:
2221         fix_problem(ctx, PR_1_ALLOCATE_ENCRYPTED_DIRLIST, &pctx);
2222         /* Should never get here */
2223         ctx->flags |= E2F_FLAG_ABORT;
2224 }
2225
2226 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino)
2227 {
2228         struct          problem_context pctx;
2229
2230         if (!ctx->casefolded_dirs) {
2231                 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
2232                 if (pctx.errcode)
2233                         goto error;
2234         }
2235         pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
2236         if (pctx.errcode == 0)
2237                 return;
2238 error:
2239         fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
2240         /* Should never get here */
2241         ctx->flags |= E2F_FLAG_ABORT;
2242 }
2243
2244 /*
2245  * This procedure will allocate the inode "bb" (badblock) map table
2246  */
2247 static void alloc_bb_map(e2fsck_t ctx)
2248 {
2249         struct          problem_context pctx;
2250
2251         clear_problem_context(&pctx);
2252         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2253                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
2254                         "inode_bb_map", &ctx->inode_bb_map);
2255         if (pctx.errcode) {
2256                 pctx.num = 4;
2257                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2258                 /* Should never get here */
2259                 ctx->flags |= E2F_FLAG_ABORT;
2260                 return;
2261         }
2262 }
2263
2264 /*
2265  * This procedure will allocate the inode imagic table
2266  */
2267 static void alloc_imagic_map(e2fsck_t ctx)
2268 {
2269         struct          problem_context pctx;
2270
2271         clear_problem_context(&pctx);
2272         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2273                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
2274                         "inode_imagic_map", &ctx->inode_imagic_map);
2275         if (pctx.errcode) {
2276                 pctx.num = 5;
2277                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2278                 /* Should never get here */
2279                 ctx->flags |= E2F_FLAG_ABORT;
2280                 return;
2281         }
2282 }
2283
2284 /*
2285  * Marks a block as in use, setting the dup_map if it's been set
2286  * already.  Called by process_block and process_bad_block.
2287  *
2288  * WARNING: Assumes checks have already been done to make sure block
2289  * is valid.  This is true in both process_block and process_bad_block.
2290  */
2291 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
2292 {
2293         struct          problem_context pctx;
2294
2295         clear_problem_context(&pctx);
2296
2297         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
2298                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
2299                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
2300                         return;
2301                 }
2302                 if (!ctx->block_dup_map) {
2303                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
2304                                         _("multiply claimed block map"),
2305                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
2306                                         &ctx->block_dup_map);
2307                         if (pctx.errcode) {
2308                                 pctx.num = 3;
2309                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
2310                                             &pctx);
2311                                 /* Should never get here */
2312                                 ctx->flags |= E2F_FLAG_ABORT;
2313                                 return;
2314                         }
2315                 }
2316                 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
2317         } else {
2318                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
2319         }
2320 }
2321
2322 /*
2323  * When cluster size is greater than one block, it is caller's responsibility
2324  * to make sure block parameter starts at a cluster boundary.
2325  */
2326 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
2327                                       unsigned int num)
2328 {
2329         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
2330                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
2331         else {
2332                 unsigned int i;
2333
2334                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
2335                         mark_block_used(ctx, block + i);
2336         }
2337 }
2338
2339 /*
2340  * Adjust the extended attribute block's reference counts at the end
2341  * of pass 1, either by subtracting out references for EA blocks that
2342  * are still referenced in ctx->refcount, or by adding references for
2343  * EA blocks that had extra references as accounted for in
2344  * ctx->refcount_extra.
2345  */
2346 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
2347                                     char *block_buf, int adjust_sign)
2348 {
2349         struct ext2_ext_attr_header     *header;
2350         struct problem_context          pctx;
2351         ext2_filsys                     fs = ctx->fs;
2352         blk64_t                         blk;
2353         __u32                           should_be;
2354         ea_value_t                      count;
2355
2356         clear_problem_context(&pctx);
2357
2358         ea_refcount_intr_begin(refcount);
2359         while (1) {
2360                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
2361                         break;
2362                 pctx.blk = blk;
2363                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
2364                                                      pctx.ino);
2365                 if (pctx.errcode) {
2366                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
2367                         return;
2368                 }
2369                 header = (struct ext2_ext_attr_header *) block_buf;
2370                 pctx.blkcount = header->h_refcount;
2371                 should_be = header->h_refcount + adjust_sign * (int)count;
2372                 pctx.num = should_be;
2373                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
2374                         header->h_refcount = should_be;
2375                         pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
2376                                                              block_buf,
2377                                                              pctx.ino);
2378                         if (pctx.errcode) {
2379                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
2380                                             &pctx);
2381                                 continue;
2382                         }
2383                 }
2384         }
2385 }
2386
2387 /*
2388  * Handle processing the extended attribute blocks
2389  */
2390 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
2391                            char *block_buf, struct ea_quota *ea_block_quota)
2392 {
2393         ext2_filsys fs = ctx->fs;
2394         ext2_ino_t      ino = pctx->ino;
2395         struct ext2_inode *inode = pctx->inode;
2396         blk64_t         blk;
2397         char *          end;
2398         struct ext2_ext_attr_header *header;
2399         struct ext2_ext_attr_entry *first, *entry;
2400         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
2401         __u64           quota_inodes = 0;
2402         region_t        region = 0;
2403         int             failed_csum = 0;
2404
2405         ea_block_quota->blocks = 0;
2406         ea_block_quota->inodes = 0;
2407
2408         blk = ext2fs_file_acl_block(fs, inode);
2409         if (blk == 0)
2410                 return 0;
2411
2412         /*
2413          * If the Extended attribute flag isn't set, then a non-zero
2414          * file acl means that the inode is corrupted.
2415          *
2416          * Or if the extended attribute block is an invalid block,
2417          * then the inode is also corrupted.
2418          */
2419         if (!ext2fs_has_feature_xattr(fs->super) ||
2420             (blk < fs->super->s_first_data_block) ||
2421             (blk >= ext2fs_blocks_count(fs->super))) {
2422                 mark_inode_bad(ctx, ino);
2423                 return 0;
2424         }
2425
2426         /* If ea bitmap hasn't been allocated, create it */
2427         if (!ctx->block_ea_map) {
2428                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
2429                                         _("ext attr block map"),
2430                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
2431                                         &ctx->block_ea_map);
2432                 if (pctx->errcode) {
2433                         pctx->num = 2;
2434                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
2435                         ctx->flags |= E2F_FLAG_ABORT;
2436                         return 0;
2437                 }
2438         }
2439
2440         /* Create the EA refcount structure if necessary */
2441         if (!ctx->refcount) {
2442                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
2443                 if (pctx->errcode) {
2444                         pctx->num = 1;
2445                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2446                         ctx->flags |= E2F_FLAG_ABORT;
2447                         return 0;
2448                 }
2449         }
2450
2451 #if 0
2452         /* Debugging text */
2453         printf("Inode %u has EA block %u\n", ino, blk);
2454 #endif
2455
2456         /* Have we seen this EA block before? */
2457         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
2458                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
2459                 ea_block_quota->inodes = 0;
2460
2461                 if (ctx->ea_block_quota_blocks) {
2462                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
2463                                           &quota_blocks);
2464                         if (quota_blocks)
2465                                 ea_block_quota->blocks = quota_blocks;
2466                 }
2467
2468                 if (ctx->ea_block_quota_inodes)
2469                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
2470                                           &ea_block_quota->inodes);
2471
2472                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
2473                         return 1;
2474                 /* Ooops, this EA was referenced more than it stated */
2475                 if (!ctx->refcount_extra) {
2476                         pctx->errcode = ea_refcount_create(0,
2477                                            &ctx->refcount_extra);
2478                         if (pctx->errcode) {
2479                                 pctx->num = 2;
2480                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2481                                 ctx->flags |= E2F_FLAG_ABORT;
2482                                 return 0;
2483                         }
2484                 }
2485                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
2486                 return 1;
2487         }
2488
2489         /*
2490          * OK, we haven't seen this EA block yet.  So we need to
2491          * validate it
2492          */
2493         pctx->blk = blk;
2494         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
2495         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
2496                 pctx->errcode = 0;
2497                 failed_csum = 1;
2498         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
2499                 pctx->errcode = 0;
2500
2501         if (pctx->errcode &&
2502             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
2503                 pctx->errcode = 0;
2504                 goto clear_extattr;
2505         }
2506         header = (struct ext2_ext_attr_header *) block_buf;
2507         pctx->blk = ext2fs_file_acl_block(fs, inode);
2508         if (((ctx->ext_attr_ver == 1) &&
2509              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
2510             ((ctx->ext_attr_ver == 2) &&
2511              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
2512                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
2513                         goto clear_extattr;
2514         }
2515
2516         if (header->h_blocks != 1) {
2517                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
2518                         goto clear_extattr;
2519         }
2520
2521         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
2522                 goto clear_extattr;
2523
2524         region = region_create(0, fs->blocksize);
2525         if (!region) {
2526                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
2527                 ctx->flags |= E2F_FLAG_ABORT;
2528                 return 0;
2529         }
2530         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
2531                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2532                         goto clear_extattr;
2533         }
2534
2535         first = (struct ext2_ext_attr_entry *)(header+1);
2536         end = block_buf + fs->blocksize;
2537         entry = first;
2538         while ((char *)entry < end && *(__u32 *)entry) {
2539                 __u32 hash;
2540
2541                 if (region_allocate(region, (char *)entry - (char *)header,
2542                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
2543                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2544                                 goto clear_extattr;
2545                         break;
2546                 }
2547                 if ((ctx->ext_attr_ver == 1 &&
2548                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
2549                     (ctx->ext_attr_ver == 2 &&
2550                      entry->e_name_index == 0)) {
2551                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
2552                                 goto clear_extattr;
2553                         break;
2554                 }
2555                 if (entry->e_value_inum == 0) {
2556                         if (entry->e_value_offs + entry->e_value_size >
2557                             fs->blocksize) {
2558                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
2559                                         goto clear_extattr;
2560                                 break;
2561                         }
2562                         if (entry->e_value_size &&
2563                             region_allocate(region, entry->e_value_offs,
2564                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
2565                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
2566                                                 pctx))
2567                                         goto clear_extattr;
2568                         }
2569
2570                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
2571                                                           entry->e_value_offs);
2572
2573                         if (entry->e_hash != hash) {
2574                                 pctx->num = entry->e_hash;
2575                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
2576                                         goto clear_extattr;
2577                                 entry->e_hash = hash;
2578                         }
2579                 } else {
2580                         problem_t problem;
2581                         blk64_t entry_quota_blocks;
2582
2583                         problem = check_large_ea_inode(ctx, entry, pctx,
2584                                                        &entry_quota_blocks);
2585                         if (problem && fix_problem(ctx, problem, pctx))
2586                                 goto clear_extattr;
2587
2588                         quota_blocks += entry_quota_blocks;
2589                         quota_inodes++;
2590                 }
2591
2592                 entry = EXT2_EXT_ATTR_NEXT(entry);
2593         }
2594         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
2595                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2596                         goto clear_extattr;
2597         }
2598         region_free(region);
2599
2600         /*
2601          * We only get here if there was no other errors that were fixed.
2602          * If there was a checksum fail, ask to correct it.
2603          */
2604         if (failed_csum &&
2605             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
2606                 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
2607                                                        pctx->ino);
2608                 if (pctx->errcode)
2609                         return 0;
2610         }
2611
2612         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
2613                 if (!ctx->ea_block_quota_blocks) {
2614                         pctx->errcode = ea_refcount_create(0,
2615                                                 &ctx->ea_block_quota_blocks);
2616                         if (pctx->errcode) {
2617                                 pctx->num = 3;
2618                                 goto refcount_fail;
2619                         }
2620                 }
2621                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
2622                                   quota_blocks);
2623         }
2624
2625         if (quota_inodes) {
2626                 if (!ctx->ea_block_quota_inodes) {
2627                         pctx->errcode = ea_refcount_create(0,
2628                                                 &ctx->ea_block_quota_inodes);
2629                         if (pctx->errcode) {
2630                                 pctx->num = 4;
2631 refcount_fail:
2632                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2633                                 ctx->flags |= E2F_FLAG_ABORT;
2634                                 return 0;
2635                         }
2636                 }
2637
2638                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
2639                                   quota_inodes);
2640         }
2641         ea_block_quota->blocks = quota_blocks;
2642         ea_block_quota->inodes = quota_inodes;
2643
2644         inc_ea_inode_refs(ctx, pctx, first, end);
2645         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
2646         mark_block_used(ctx, blk);
2647         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
2648         return 1;
2649
2650 clear_extattr:
2651         if (region)
2652                 region_free(region);
2653         ext2fs_file_acl_block_set(fs, inode, 0);
2654         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
2655         return 0;
2656 }
2657
2658 /* Returns 1 if bad htree, 0 if OK */
2659 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
2660                         ext2_ino_t ino, struct ext2_inode *inode,
2661                         char *block_buf)
2662 {
2663         struct ext2_dx_root_info        *root;
2664         ext2_filsys                     fs = ctx->fs;
2665         errcode_t                       retval;
2666         blk64_t                         blk;
2667
2668         if ((!LINUX_S_ISDIR(inode->i_mode) &&
2669              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
2670             (!ext2fs_has_feature_dir_index(fs->super) &&
2671              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
2672                 return 1;
2673
2674         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
2675
2676         if ((pctx->errcode) ||
2677             (blk == 0) ||
2678             (blk < fs->super->s_first_data_block) ||
2679             (blk >= ext2fs_blocks_count(fs->super))) {
2680                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2681                         return 1;
2682                 else
2683                         return 0;
2684         }
2685
2686         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
2687         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2688                 return 1;
2689
2690         /* XXX should check that beginning matches a directory */
2691         root = (struct ext2_dx_root_info *) (block_buf + 24);
2692
2693         if ((root->reserved_zero || root->info_length < 8) &&
2694             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2695                 return 1;
2696
2697         pctx->num = root->hash_version;
2698         if ((root->hash_version != EXT2_HASH_LEGACY) &&
2699             (root->hash_version != EXT2_HASH_HALF_MD4) &&
2700             (root->hash_version != EXT2_HASH_TEA) &&
2701             (root->hash_version != EXT2_HASH_SIPHASH) &&
2702             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
2703                 return 1;
2704
2705         if (ext4_hash_in_dirent(inode)) {
2706                 if (root->hash_version != EXT2_HASH_SIPHASH &&
2707                     fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
2708                         return 1;
2709         } else {
2710                 if (root->hash_version == EXT2_HASH_SIPHASH &&
2711                    fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
2712                         return 1;
2713         }
2714
2715         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
2716             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
2717                 return 1;
2718
2719         pctx->num = root->indirect_levels;
2720         if ((root->indirect_levels >= ext2_dir_htree_level(fs)) &&
2721             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
2722                 return 1;
2723
2724         return 0;
2725 }
2726
2727 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
2728                         struct ext2_inode *inode, int restart_flag,
2729                         const char *source)
2730 {
2731         inode->i_flags = 0;
2732         inode->i_links_count = 0;
2733         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
2734         inode->i_dtime = ctx->now;
2735
2736         /*
2737          * If a special inode has such rotten block mappings that we
2738          * want to clear the whole inode, be sure to actually zap
2739          * the block maps because i_links_count isn't checked for
2740          * special inodes, and we'll end up right back here the next
2741          * time we run fsck.
2742          */
2743         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
2744                 memset(inode->i_block, 0, sizeof(inode->i_block));
2745
2746         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
2747         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
2748         if (ctx->inode_reg_map)
2749                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
2750         if (ctx->inode_bad_map)
2751                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
2752
2753         /*
2754          * If the inode was partially accounted for before processing
2755          * was aborted, we need to restart the pass 1 scan.
2756          */
2757         ctx->flags |= restart_flag;
2758
2759         if (ino == EXT2_BAD_INO)
2760                 memset(inode, 0, sizeof(struct ext2_inode));
2761
2762         e2fsck_write_inode(ctx, ino, inode, source);
2763 }
2764
2765 /*
2766  * Use the multiple-blocks reclamation code to fix alignment problems in
2767  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
2768  * physical cluster, and we want the block offsets within that cluster to
2769  * line up.
2770  */
2771 static int has_unaligned_cluster_map(e2fsck_t ctx,
2772                                      blk64_t last_pblk, blk64_t last_lblk,
2773                                      blk64_t pblk, blk64_t lblk)
2774 {
2775         blk64_t cluster_mask;
2776
2777         if (!ctx->fs->cluster_ratio_bits)
2778                 return 0;
2779         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
2780
2781         /*
2782          * If the block in the logical cluster doesn't align with the block in
2783          * the physical cluster...
2784          */
2785         if ((lblk & cluster_mask) != (pblk & cluster_mask))
2786                 return 1;
2787
2788         /*
2789          * If we cross a physical cluster boundary within a logical cluster...
2790          */
2791         if (last_pblk && (lblk & cluster_mask) != 0 &&
2792             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
2793             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
2794                 return 1;
2795
2796         return 0;
2797 }
2798
2799 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
2800                              struct process_block_struct *pb,
2801                              blk64_t start_block, blk64_t end_block,
2802                              blk64_t eof_block,
2803                              ext2_extent_handle_t ehandle,
2804                              int try_repairs)
2805 {
2806         struct ext2fs_extent    extent;
2807         blk64_t                 blk, last_lblk;
2808         unsigned int            i, n;
2809         int                     is_dir, is_leaf;
2810         problem_t               problem;
2811         struct ext2_extent_info info;
2812         int                     failed_csum = 0;
2813
2814         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
2815                 failed_csum = 1;
2816
2817         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
2818         if (pctx->errcode)
2819                 return;
2820         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
2821             !pb->eti.force_rebuild) {
2822                 struct extent_tree_level *etl;
2823
2824                 etl = pb->eti.ext_info + info.curr_level;
2825                 etl->num_extents += info.num_entries;
2826                 etl->max_extents += info.max_entries;
2827                 /*
2828                  * Implementation wart: Splitting extent blocks when appending
2829                  * will leave the old block with one free entry.  Therefore
2830                  * unless the node is totally full, pretend that a non-root
2831                  * extent block can hold one fewer entry than it actually does,
2832                  * so that we don't repeatedly rebuild the extent tree.
2833                  */
2834                 if (info.curr_level && info.num_entries < info.max_entries)
2835                         etl->max_extents--;
2836         }
2837
2838         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
2839                                           &extent);
2840         while ((pctx->errcode == 0 ||
2841                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
2842                info.num_entries-- > 0) {
2843                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
2844                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
2845                 last_lblk = extent.e_lblk + extent.e_len - 1;
2846
2847                 problem = 0;
2848                 pctx->blk = extent.e_pblk;
2849                 pctx->blk2 = extent.e_lblk;
2850                 pctx->num = extent.e_len;
2851                 pctx->blkcount = extent.e_lblk + extent.e_len;
2852
2853                 if (extent.e_pblk == 0 ||
2854                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
2855                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
2856                         problem = PR_1_EXTENT_BAD_START_BLK;
2857                 else if (extent.e_lblk < start_block)
2858                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
2859                 else if ((end_block && last_lblk > end_block) &&
2860                          !(last_lblk > eof_block &&
2861                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
2862                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
2863                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
2864                 else if (is_leaf && extent.e_len == 0)
2865                         problem = PR_1_EXTENT_LENGTH_ZERO;
2866                 else if (is_leaf &&
2867                          (extent.e_pblk + extent.e_len) >
2868                          ext2fs_blocks_count(ctx->fs->super))
2869                         problem = PR_1_EXTENT_ENDS_BEYOND;
2870                 else if (is_leaf && is_dir &&
2871                          ((extent.e_lblk + extent.e_len) >
2872                           (1U << (21 - ctx->fs->super->s_log_block_size))))
2873                         problem = PR_1_TOOBIG_DIR;
2874
2875                 if (is_leaf && problem == 0 && extent.e_len > 0) {
2876 #if 0
2877                         printf("extent_region(ino=%u, expect=%llu, "
2878                                "lblk=%llu, len=%u)\n",
2879                                pb->ino, pb->next_lblock,
2880                                extent.e_lblk, extent.e_len);
2881 #endif
2882                         if (extent.e_lblk < pb->next_lblock)
2883                                 problem = PR_1_EXTENT_COLLISION;
2884                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
2885                                 pb->next_lblock = extent.e_lblk + extent.e_len;
2886                 }
2887
2888                 /*
2889                  * Uninitialized blocks in a directory?  Clear the flag and
2890                  * we'll interpret the blocks later.
2891                  */
2892                 if (try_repairs && is_dir && problem == 0 &&
2893                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
2894                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
2895                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
2896                         pb->inode_modified = 1;
2897                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
2898                                                               &extent);
2899                         if (pctx->errcode)
2900                                 return;
2901                         failed_csum = 0;
2902                 }
2903
2904                 if (try_repairs && problem) {
2905 report_problem:
2906                         if (fix_problem(ctx, problem, pctx)) {
2907                                 if (ctx->invalid_bitmaps) {
2908                                         /*
2909                                          * If fsck knows the bitmaps are bad,
2910                                          * skip to the next extent and
2911                                          * try to clear this extent again
2912                                          * after fixing the bitmaps, by
2913                                          * restarting fsck.
2914                                          */
2915                                         pctx->errcode = ext2fs_extent_get(
2916                                                           ehandle,
2917                                                           EXT2_EXTENT_NEXT_SIB,
2918                                                           &extent);
2919                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
2920                                         if (pctx->errcode ==
2921                                                     EXT2_ET_NO_CURRENT_NODE) {
2922                                                 pctx->errcode = 0;
2923                                                 break;
2924                                         }
2925                                         continue;
2926                                 }
2927                                 e2fsck_read_bitmaps(ctx);
2928                                 pb->inode_modified = 1;
2929                                 pctx->errcode =
2930                                         ext2fs_extent_delete(ehandle, 0);
2931                                 if (pctx->errcode) {
2932                                         pctx->str = "ext2fs_extent_delete";
2933                                         return;
2934                                 }
2935                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
2936                                 if (pctx->errcode &&
2937                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
2938                                         pctx->str = "ext2fs_extent_fix_parents";
2939                                         return;
2940                                 }
2941                                 pctx->errcode = ext2fs_extent_get(ehandle,
2942                                                                   EXT2_EXTENT_CURRENT,
2943                                                                   &extent);
2944                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
2945                                         pctx->errcode = 0;
2946                                         break;
2947                                 }
2948                                 failed_csum = 0;
2949                                 continue;
2950                         }
2951                         goto next;
2952                 }
2953
2954                 if (!is_leaf) {
2955                         blk64_t lblk = extent.e_lblk;
2956                         int next_try_repairs = 1;
2957
2958                         blk = extent.e_pblk;
2959
2960                         /*
2961                          * If this lower extent block collides with critical
2962                          * metadata, don't try to repair the damage.  Pass 1b
2963                          * will reallocate the block; then we can try again.
2964                          */
2965                         if (pb->ino != EXT2_RESIZE_INO &&
2966                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
2967                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
2968                                                       extent.e_pblk)) {
2969                                 next_try_repairs = 0;
2970                                 pctx->blk = blk;
2971                                 fix_problem(ctx,
2972                                             PR_1_CRITICAL_METADATA_COLLISION,
2973                                             pctx);
2974                                 if ((ctx->options & E2F_OPT_NO) == 0)
2975                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
2976                         }
2977                         pctx->errcode = ext2fs_extent_get(ehandle,
2978                                                   EXT2_EXTENT_DOWN, &extent);
2979                         if (pctx->errcode &&
2980                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
2981                                 pctx->str = "EXT2_EXTENT_DOWN";
2982                                 problem = PR_1_EXTENT_HEADER_INVALID;
2983                                 if (!next_try_repairs)
2984                                         return;
2985                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
2986                                         goto report_problem;
2987                                 return;
2988                         }
2989                         /* The next extent should match this index's logical start */
2990                         if (extent.e_lblk != lblk) {
2991                                 struct ext2_extent_info e_info;
2992
2993                                 ext2fs_extent_get_info(ehandle, &e_info);
2994                                 pctx->blk = lblk;
2995                                 pctx->blk2 = extent.e_lblk;
2996                                 pctx->num = e_info.curr_level - 1;
2997                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
2998                                 if (fix_problem(ctx, problem, pctx)) {
2999                                         pb->inode_modified = 1;
3000                                         pctx->errcode =
3001                                                 ext2fs_extent_fix_parents(ehandle);
3002                                         if (pctx->errcode) {
3003                                                 pctx->str = "ext2fs_extent_fix_parents";
3004                                                 return;
3005                                         }
3006                                 }
3007                         }
3008                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
3009                                          last_lblk, eof_block, ehandle,
3010                                          next_try_repairs);
3011                         if (pctx->errcode)
3012                                 return;
3013                         pctx->errcode = ext2fs_extent_get(ehandle,
3014                                                   EXT2_EXTENT_UP, &extent);
3015                         if (pctx->errcode) {
3016                                 pctx->str = "EXT2_EXTENT_UP";
3017                                 return;
3018                         }
3019                         mark_block_used(ctx, blk);
3020                         pb->num_blocks++;
3021                         goto next;
3022                 }
3023
3024                 if ((pb->previous_block != 0) &&
3025                     (pb->previous_block+1 != extent.e_pblk)) {
3026                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3027                                 char type = '?';
3028
3029                                 if (pb->is_dir)
3030                                         type = 'd';
3031                                 else if (pb->is_reg)
3032                                         type = 'f';
3033
3034                                 printf(("%6lu(%c): expecting %6lu "
3035                                         "actual extent "
3036                                         "phys %6lu log %lu len %lu\n"),
3037                                        (unsigned long) pctx->ino, type,
3038                                        (unsigned long) pb->previous_block+1,
3039                                        (unsigned long) extent.e_pblk,
3040                                        (unsigned long) extent.e_lblk,
3041                                        (unsigned long) extent.e_len);
3042                         }
3043                         pb->fragmented = 1;
3044                 }
3045                 /*
3046                  * If we notice a gap in the logical block mappings of an
3047                  * extent-mapped directory, offer to close the hole by
3048                  * moving the logical block down, otherwise we'll go mad in
3049                  * pass 3 allocating empty directory blocks to fill the hole.
3050                  */
3051                 if (try_repairs && is_dir &&
3052                     pb->last_block + 1 < extent.e_lblk) {
3053                         blk64_t new_lblk;
3054
3055                         new_lblk = pb->last_block + 1;
3056                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3057                                 new_lblk = ((new_lblk +
3058                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3059                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3060                                            (extent.e_pblk &
3061                                             EXT2FS_CLUSTER_MASK(ctx->fs));
3062                         pctx->blk = extent.e_lblk;
3063                         pctx->blk2 = new_lblk;
3064                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3065                                 extent.e_lblk = new_lblk;
3066                                 pb->inode_modified = 1;
3067                                 pctx->errcode = ext2fs_extent_replace(ehandle,
3068                                                                 0, &extent);
3069                                 if (pctx->errcode) {
3070                                         pctx->errcode = 0;
3071                                         goto alloc_later;
3072                                 }
3073                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3074                                 if (pctx->errcode)
3075                                         goto failed_add_dir_block;
3076                                 pctx->errcode = ext2fs_extent_goto(ehandle,
3077                                                                 extent.e_lblk);
3078                                 if (pctx->errcode)
3079                                         goto failed_add_dir_block;
3080                                 last_lblk = extent.e_lblk + extent.e_len - 1;
3081                                 failed_csum = 0;
3082                         }
3083                 }
3084 alloc_later:
3085                 if (is_dir) {
3086                         while (++pb->last_db_block <
3087                                (e2_blkcnt_t) extent.e_lblk) {
3088                                 pctx->errcode = ext2fs_add_dir_block2(
3089                                                         ctx->fs->dblist,
3090                                                         pb->ino, 0,
3091                                                         pb->last_db_block);
3092                                 if (pctx->errcode) {
3093                                         pctx->blk = 0;
3094                                         pctx->num = pb->last_db_block;
3095                                         goto failed_add_dir_block;
3096                                 }
3097                         }
3098
3099                         for (i = 0; i < extent.e_len; i++) {
3100                                 pctx->errcode = ext2fs_add_dir_block2(
3101                                                         ctx->fs->dblist,
3102                                                         pctx->ino,
3103                                                         extent.e_pblk + i,
3104                                                         extent.e_lblk + i);
3105                                 if (pctx->errcode) {
3106                                         pctx->blk = extent.e_pblk + i;
3107                                         pctx->num = extent.e_lblk + i;
3108                                 failed_add_dir_block:
3109                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3110                                         /* Should never get here */
3111                                         ctx->flags |= E2F_FLAG_ABORT;
3112                                         return;
3113                                 }
3114                         }
3115                         if (extent.e_len > 0)
3116                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3117                 }
3118                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3119                                               pb->last_block,
3120                                               extent.e_pblk,
3121                                               extent.e_lblk)) {
3122                         for (i = 0; i < extent.e_len; i++) {
3123                                 pctx->blk = extent.e_lblk + i;
3124                                 pctx->blk2 = extent.e_pblk + i;
3125                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3126                                 mark_block_used(ctx, extent.e_pblk + i);
3127                                 mark_block_used(ctx, extent.e_pblk + i);
3128                         }
3129                 }
3130
3131                 /*
3132                  * Check whether first cluster got marked in previous iteration.
3133                  */
3134                 if (ctx->fs->cluster_ratio_bits &&
3135                     pb->previous_block &&
3136                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
3137                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
3138                         /* Set blk to the beginning of next cluster. */
3139                         blk = EXT2FS_C2B(
3140                                 ctx->fs,
3141                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
3142                 else
3143                         /* Set blk to the beginning of current cluster. */
3144                         blk = EXT2FS_C2B(ctx->fs,
3145                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
3146
3147                 if (blk < extent.e_pblk + extent.e_len) {
3148                         mark_blocks_used(ctx, blk,
3149                                          extent.e_pblk + extent.e_len - blk);
3150                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
3151                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
3152                         pb->num_blocks += n;
3153                 }
3154                 pb->last_block = extent.e_lblk + extent.e_len - 1;
3155                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
3156                 start_block = pb->last_block = last_lblk;
3157                 if (is_leaf && !is_dir &&
3158                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
3159                         pb->last_init_lblock = last_lblk;
3160         next:
3161                 pctx->errcode = ext2fs_extent_get(ehandle,
3162                                                   EXT2_EXTENT_NEXT_SIB,
3163                                                   &extent);
3164         }
3165
3166         /* Failed csum but passes checks?  Ask to fix checksum. */
3167         if (failed_csum &&
3168             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
3169                 pb->inode_modified = 1;
3170                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
3171                 if (pctx->errcode)
3172                         return;
3173         }
3174
3175         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
3176                 pctx->errcode = 0;
3177 }
3178
3179 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
3180                                  struct process_block_struct *pb)
3181 {
3182         struct ext2_extent_info info;
3183         struct ext2_inode       *inode = pctx->inode;
3184         ext2_extent_handle_t    ehandle;
3185         ext2_filsys             fs = ctx->fs;
3186         ext2_ino_t              ino = pctx->ino;
3187         errcode_t               retval;
3188         blk64_t                 eof_lblk;
3189         struct ext3_extent_header       *eh;
3190
3191         /* Check for a proper extent header... */
3192         eh = (struct ext3_extent_header *) &inode->i_block[0];
3193         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
3194         if (retval) {
3195                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
3196                         e2fsck_clear_inode(ctx, ino, inode, 0,
3197                                            "check_blocks_extents");
3198                 pctx->errcode = 0;
3199                 return;
3200         }
3201
3202         /* ...since this function doesn't fail if i_block is zeroed. */
3203         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
3204         if (pctx->errcode) {
3205                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
3206                         e2fsck_clear_inode(ctx, ino, inode, 0,
3207                                            "check_blocks_extents");
3208                 pctx->errcode = 0;
3209                 return;
3210         }
3211
3212         retval = ext2fs_extent_get_info(ehandle, &info);
3213         if (retval == 0) {
3214                 int max_depth = info.max_depth;
3215
3216                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
3217                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
3218                 ctx->extent_depth_count[max_depth]++;
3219         }
3220
3221         /* Check maximum extent depth */
3222         pctx->blk = info.max_depth;
3223         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
3224         if (pctx->blk2 < pctx->blk &&
3225             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
3226                 pb->eti.force_rebuild = 1;
3227
3228         /* Can we collect extent tree level stats? */
3229         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
3230         if (pctx->blk2 > pctx->blk)
3231                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
3232         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
3233         pb->eti.ino = pb->ino;
3234
3235         pb->next_lblock = 0;
3236
3237         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
3238                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
3239         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
3240         if (pctx->errcode &&
3241             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
3242                 pb->num_blocks = 0;
3243                 inode->i_blocks = 0;
3244                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3245                                    "check_blocks_extents");
3246                 pctx->errcode = 0;
3247         }
3248         ext2fs_extent_free(ehandle);
3249
3250         /* Rebuild unless it's a dir and we're rehashing it */
3251         if (LINUX_S_ISDIR(inode->i_mode) &&
3252             e2fsck_dir_will_be_rehashed(ctx, ino))
3253                 return;
3254
3255         if (ctx->options & E2F_OPT_CONVERT_BMAP)
3256                 e2fsck_rebuild_extents_later(ctx, ino);
3257         else
3258                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
3259 }
3260
3261 /*
3262  * In fact we don't need to check blocks for an inode with inline data
3263  * because this inode doesn't have any blocks.  In this function all
3264  * we need to do is add this inode into dblist when it is a directory.
3265  */
3266 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
3267                                      struct process_block_struct *pb)
3268 {
3269         int     flags;
3270         size_t  inline_data_size = 0;
3271
3272         if (!pb->is_dir) {
3273                 pctx->errcode = 0;
3274                 return;
3275         }
3276
3277         /* Process the dirents in i_block[] as the "first" block. */
3278         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
3279         if (pctx->errcode)
3280                 goto err;
3281
3282         /* Process the dirents in the EA as a "second" block. */
3283         flags = ctx->fs->flags;
3284         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3285         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
3286                                                 &inline_data_size);
3287         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3288                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3289         if (pctx->errcode) {
3290                 pctx->errcode = 0;
3291                 return;
3292         }
3293
3294         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
3295                 return;
3296
3297         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
3298         if (pctx->errcode)
3299                 goto err;
3300
3301         return;
3302 err:
3303         pctx->blk = 0;
3304         pctx->num = 0;
3305         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3306         ctx->flags |= E2F_FLAG_ABORT;
3307 }
3308
3309 /*
3310  * This subroutine is called on each inode to account for all of the
3311  * blocks used by that inode.
3312  */
3313 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
3314                          char *block_buf, const struct ea_quota *ea_ibody_quota)
3315 {
3316         ext2_filsys fs = ctx->fs;
3317         struct process_block_struct pb;
3318         ext2_ino_t      ino = pctx->ino;
3319         struct ext2_inode *inode = pctx->inode;
3320         unsigned        bad_size = 0;
3321         int             dirty_inode = 0;
3322         int             extent_fs;
3323         int             inlinedata_fs;
3324         __u64           size;
3325         struct ea_quota ea_block_quota;
3326
3327         pb.ino = ino;
3328         pb.num_blocks = EXT2FS_B2C(ctx->fs,
3329                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
3330         pb.last_block = ~0;
3331         pb.last_init_lblock = -1;
3332         pb.last_db_block = -1;
3333         pb.num_illegal_blocks = 0;
3334         pb.suppress = 0; pb.clear = 0;
3335         pb.fragmented = 0;
3336         pb.compressed = 0;
3337         pb.previous_block = 0;
3338         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
3339         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
3340         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
3341         pb.inode = inode;
3342         pb.pctx = pctx;
3343         pb.ctx = ctx;
3344         pb.inode_modified = 0;
3345         pb.eti.force_rebuild = 0;
3346         pctx->ino = ino;
3347         pctx->errcode = 0;
3348
3349         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
3350         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
3351
3352         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
3353                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3354                         goto out;
3355                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
3356         }
3357
3358         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
3359                 check_blocks_inline_data(ctx, pctx, &pb);
3360         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
3361                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
3362                         check_blocks_extents(ctx, pctx, &pb);
3363                 else {
3364                         int flags;
3365                         /*
3366                          * If we've modified the inode, write it out before
3367                          * iterate() tries to use it.
3368                          */
3369                         if (dirty_inode) {
3370                                 e2fsck_write_inode(ctx, ino, inode,
3371                                                    "check_blocks");
3372                                 dirty_inode = 0;
3373                         }
3374                         flags = fs->flags;
3375                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3376                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
3377                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
3378                                                 block_buf, process_block, &pb);
3379                         /*
3380                          * We do not have uninitialized extents in non extent
3381                          * files.
3382                          */
3383                         pb.last_init_lblock = pb.last_block;
3384                         /*
3385                          * If iterate() changed a block mapping, we have to
3386                          * re-read the inode.  If we decide to clear the
3387                          * inode after clearing some stuff, we'll re-write the
3388                          * bad mappings into the inode!
3389                          */
3390                         if (pb.inode_modified)
3391                                 e2fsck_read_inode(ctx, ino, inode,
3392                                                   "check_blocks");
3393                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3394                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3395
3396                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
3397 #ifdef DEBUG
3398                                 printf("bmap rebuild ino=%d\n", ino);
3399 #endif
3400                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
3401                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
3402                                         e2fsck_rebuild_extents_later(ctx, ino);
3403                         }
3404                 }
3405         }
3406         end_problem_latch(ctx, PR_LATCH_BLOCK);
3407         end_problem_latch(ctx, PR_LATCH_TOOBIG);
3408         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3409                 goto out;
3410         if (pctx->errcode)
3411                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
3412
3413         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
3414                 if (LINUX_S_ISDIR(inode->i_mode))
3415                         ctx->fs_fragmented_dir++;
3416                 else
3417                         ctx->fs_fragmented++;
3418         }
3419
3420         if (pb.clear) {
3421                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3422                                    "check_blocks");
3423                 return;
3424         }
3425
3426         if (inode->i_flags & EXT2_INDEX_FL) {
3427                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
3428                         inode->i_flags &= ~EXT2_INDEX_FL;
3429                         dirty_inode++;
3430                 } else {
3431                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
3432                 }
3433         }
3434
3435         if (!pb.num_blocks && pb.is_dir &&
3436             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
3437                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
3438                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
3439                         ctx->fs_directory_count--;
3440                         return;
3441                 }
3442         }
3443
3444         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
3445             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
3446             !(inode->i_flags & EXT4_EA_INODE_FL)) {
3447                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
3448                                ino,
3449                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
3450                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
3451                                   ino, (ea_ibody_quota ?
3452                                         ea_ibody_quota->inodes : 0) +
3453                                                 ea_block_quota.inodes + 1);
3454         }
3455
3456         if (!ext2fs_has_feature_huge_file(fs->super) ||
3457             !(inode->i_flags & EXT4_HUGE_FILE_FL))
3458                 pb.num_blocks *= (fs->blocksize / 512);
3459         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
3460 #if 0
3461         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
3462                ino, inode->i_size, pb.last_block, ext2fs_inode_i_blocks(fs, inode),
3463                pb.num_blocks);
3464 #endif
3465         if (pb.is_dir) {
3466                 unsigned nblock = inode->i_size >> EXT2_BLOCK_SIZE_BITS(fs->super);
3467                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
3468                         int flags;
3469                         size_t sz = 0;
3470                         errcode_t err;
3471
3472                         flags = ctx->fs->flags;
3473                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3474                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
3475                                                       &sz);
3476                         ctx->fs->flags = (flags &
3477                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3478                                          (ctx->fs->flags &
3479                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3480                         if (err || sz != inode->i_size) {
3481                                 bad_size = 7;
3482                                 pctx->num = sz;
3483                         }
3484                 } else if (inode->i_size & (fs->blocksize - 1))
3485                         bad_size = 5;
3486                 else if (nblock > (pb.last_block + 1))
3487                         bad_size = 1;
3488                 else if (nblock < (pb.last_block + 1)) {
3489                         if (((pb.last_block + 1) - nblock) >
3490                             fs->super->s_prealloc_dir_blocks)
3491                                 bad_size = 2;
3492                 }
3493         } else {
3494                 size = EXT2_I_SIZE(inode);
3495                 if ((pb.last_init_lblock >= 0) &&
3496                     /* Do not allow initialized allocated blocks past i_size*/
3497                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
3498                     !(inode->i_flags & EXT4_VERITY_FL))
3499                         bad_size = 3;
3500                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3501                          size > ext2_max_sizes[fs->super->s_log_block_size])
3502                         /* too big for a direct/indirect-mapped file */
3503                         bad_size = 4;
3504                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3505                          size >
3506                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
3507                         /* too big for an extent-based file - 32bit ee_block */
3508                         bad_size = 6;
3509         }
3510         /* i_size for symlinks is checked elsewhere */
3511         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
3512                 /* Did inline_data set pctx->num earlier? */
3513                 if (bad_size != 7)
3514                         pctx->num = (pb.last_block + 1) * fs->blocksize;
3515                 pctx->group = bad_size;
3516                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
3517                         if (LINUX_S_ISDIR(inode->i_mode))
3518                                 pctx->num &= 0xFFFFFFFFULL;
3519                         ext2fs_inode_size_set(fs, inode, pctx->num);
3520                         if (EXT2_I_SIZE(inode) == 0 &&
3521                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
3522                                 memset(inode->i_block, 0,
3523                                        sizeof(inode->i_block));
3524                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
3525                         }
3526                         dirty_inode++;
3527                 }
3528                 pctx->num = 0;
3529         }
3530         if (LINUX_S_ISREG(inode->i_mode) &&
3531             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3532                 ctx->large_files++;
3533         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
3534             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
3535              (ext2fs_has_feature_huge_file(fs->super) &&
3536               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
3537               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
3538                 pctx->num = pb.num_blocks;
3539                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
3540                         inode->i_blocks = pb.num_blocks;
3541                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
3542                         dirty_inode++;
3543                 }
3544                 pctx->num = 0;
3545         }
3546
3547         /*
3548          * The kernel gets mad if we ask it to allocate bigalloc clusters to
3549          * a block mapped file, so rebuild it as an extent file.  We can skip
3550          * symlinks because they're never rewritten.
3551          */
3552         if (ext2fs_has_feature_bigalloc(fs->super) &&
3553             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
3554             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
3555             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
3556             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
3557             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
3558                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
3559                 if (pctx->errcode)
3560                         goto out;
3561         }
3562
3563         if (ctx->dirs_to_hash && pb.is_dir &&
3564             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
3565             !(inode->i_flags & EXT2_INDEX_FL) &&
3566             ((inode->i_size / fs->blocksize) >= 3))
3567                 e2fsck_rehash_dir_later(ctx, ino);
3568
3569 out:
3570         if (dirty_inode)
3571                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
3572 }
3573
3574 #if 0
3575 /*
3576  * Helper function called by process block when an illegal block is
3577  * found.  It returns a description about why the block is illegal
3578  */
3579 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
3580 {
3581         blk64_t super;
3582         int     i;
3583         static char     problem[80];
3584
3585         super = fs->super->s_first_data_block;
3586         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
3587         if (block < super) {
3588                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
3589                 return(problem);
3590         } else if (block >= ext2fs_blocks_count(fs->super)) {
3591                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
3592                 return(problem);
3593         }
3594         for (i = 0; i < fs->group_desc_count; i++) {
3595                 if (block == super) {
3596                         sprintf(problem, "is the superblock in group %d", i);
3597                         break;
3598                 }
3599                 if (block > super &&
3600                     block <= (super + fs->desc_blocks)) {
3601                         sprintf(problem, "is in the group descriptors "
3602                                 "of group %d", i);
3603                         break;
3604                 }
3605                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
3606                         sprintf(problem, "is the block bitmap of group %d", i);
3607                         break;
3608                 }
3609                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
3610                         sprintf(problem, "is the inode bitmap of group %d", i);
3611                         break;
3612                 }
3613                 if (block >= ext2fs_inode_table_loc(fs, i) &&
3614                     (block < ext2fs_inode_table_loc(fs, i)
3615                      + fs->inode_blocks_per_group)) {
3616                         sprintf(problem, "is in the inode table of group %d",
3617                                 i);
3618                         break;
3619                 }
3620                 super += fs->super->s_blocks_per_group;
3621         }
3622         return(problem);
3623 }
3624 #endif
3625
3626 /*
3627  * This is a helper function for check_blocks().
3628  */
3629 static int process_block(ext2_filsys fs,
3630                   blk64_t       *block_nr,
3631                   e2_blkcnt_t blockcnt,
3632                   blk64_t ref_block EXT2FS_ATTR((unused)),
3633                   int ref_offset EXT2FS_ATTR((unused)),
3634                   void *priv_data)
3635 {
3636         struct process_block_struct *p;
3637         struct problem_context *pctx;
3638         blk64_t blk = *block_nr;
3639         int     ret_code = 0;
3640         problem_t       problem = 0;
3641         e2fsck_t        ctx;
3642
3643         p = (struct process_block_struct *) priv_data;
3644         pctx = p->pctx;
3645         ctx = p->ctx;
3646
3647         /*
3648          * For a directory, add logical block zero for processing even if it's
3649          * not mapped or we'll be perennially stuck with broken "." and ".."
3650          * entries.
3651          */
3652         if (p->is_dir && blockcnt == 0 && blk == 0) {
3653                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
3654                 if (pctx->errcode) {
3655                         pctx->blk = blk;
3656                         pctx->num = blockcnt;
3657                         goto failed_add_dir_block;
3658                 }
3659                 p->last_db_block++;
3660         }
3661
3662         if (blk == 0)
3663                 return 0;
3664
3665 #if 0
3666         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
3667                blockcnt);
3668 #endif
3669
3670         /*
3671          * Simplistic fragmentation check.  We merely require that the
3672          * file be contiguous.  (Which can never be true for really
3673          * big files that are greater than a block group.)
3674          */
3675         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
3676                 if (p->previous_block+1 != blk) {
3677                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3678                                 char type = '?';
3679
3680                                 if (p->is_dir)
3681                                         type = 'd';
3682                                 else if (p->is_reg)
3683                                         type = 'f';
3684
3685                                 printf(_("%6lu(%c): expecting %6lu "
3686                                          "got phys %6lu (blkcnt %lld)\n"),
3687                                        (unsigned long) pctx->ino, type,
3688                                        (unsigned long) p->previous_block+1,
3689                                        (unsigned long) blk,
3690                                        blockcnt);
3691                         }
3692                         p->fragmented = 1;
3693                 }
3694         }
3695
3696         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
3697             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
3698                 problem = PR_1_TOOBIG_DIR;
3699         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
3700                 problem = PR_1_TOOBIG_DIR;
3701         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
3702                 problem = PR_1_TOOBIG_REG;
3703         if (!p->is_dir && !p->is_reg && blockcnt > 0)
3704                 problem = PR_1_TOOBIG_SYMLINK;
3705
3706         if (blk < fs->super->s_first_data_block ||
3707             blk >= ext2fs_blocks_count(fs->super))
3708                 problem = PR_1_ILLEGAL_BLOCK_NUM;
3709
3710         /*
3711          * If this IND/DIND/TIND block is squatting atop some critical metadata
3712          * (group descriptors, superblock, bitmap, inode table), any write to
3713          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
3714          * fix that and restart fsck.
3715          */
3716         if (blockcnt < 0 &&
3717             p->ino != EXT2_RESIZE_INO &&
3718             blk < ctx->fs->super->s_blocks_count &&
3719             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
3720                 pctx->blk = blk;
3721                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
3722                 if ((ctx->options & E2F_OPT_NO) == 0)
3723                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3724         }
3725
3726         if (problem) {
3727                 p->num_illegal_blocks++;
3728                 /*
3729                  * A bit of subterfuge here -- we're trying to fix a block
3730                  * mapping, but the IND/DIND/TIND block could have collided
3731                  * with some critical metadata.  So, fix the in-core mapping so
3732                  * iterate won't go insane, but return 0 instead of
3733                  * BLOCK_CHANGED so that it won't write the remapping out to
3734                  * our multiply linked block.
3735                  *
3736                  * Even if we previously determined that an *IND block
3737                  * conflicts with critical metadata, we must still try to
3738                  * iterate the *IND block as if it is an *IND block to find and
3739                  * mark the blocks it points to.  Better to be overly cautious
3740                  * with the used_blocks map so that we don't move the *IND
3741                  * block to a block that's really in use!
3742                  */
3743                 if (p->ino != EXT2_RESIZE_INO &&
3744                     ref_block != 0 &&
3745                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3746                                               ref_block)) {
3747                         *block_nr = 0;
3748                         return 0;
3749                 }
3750                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
3751                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
3752                                 p->clear = 1;
3753                                 return BLOCK_ABORT;
3754                         }
3755                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
3756                                 p->suppress = 1;
3757                                 set_latch_flags(PR_LATCH_BLOCK,
3758                                                 PRL_SUPPRESS, 0);
3759                         }
3760                 }
3761                 pctx->blk = blk;
3762                 pctx->blkcount = blockcnt;
3763                 if (fix_problem(ctx, problem, pctx)) {
3764                         blk = *block_nr = 0;
3765                         ret_code = BLOCK_CHANGED;
3766                         p->inode_modified = 1;
3767                         /*
3768                          * If the directory block is too big and is beyond the
3769                          * end of the FS, don't bother trying to add it for
3770                          * processing -- the kernel would never have created a
3771                          * directory this large, and we risk an ENOMEM abort.
3772                          * In any case, the toobig handler for extent-based
3773                          * directories also doesn't feed toobig blocks to
3774                          * pass 2.
3775                          */
3776                         if (problem == PR_1_TOOBIG_DIR)
3777                                 return ret_code;
3778                         goto mark_dir;
3779                 } else
3780                         return 0;
3781         }
3782
3783         if (p->ino == EXT2_RESIZE_INO) {
3784                 /*
3785                  * The resize inode has already be sanity checked
3786                  * during pass #0 (the superblock checks).  All we
3787                  * have to do is mark the double indirect block as
3788                  * being in use; all of the other blocks are handled
3789                  * by mark_table_blocks()).
3790                  */
3791                 if (blockcnt == BLOCK_COUNT_DIND)
3792                         mark_block_used(ctx, blk);
3793                 p->num_blocks++;
3794         } else if (!(ctx->fs->cluster_ratio_bits &&
3795                      p->previous_block &&
3796                      (EXT2FS_B2C(ctx->fs, blk) ==
3797                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
3798                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
3799                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
3800                 mark_block_used(ctx, blk);
3801                 p->num_blocks++;
3802         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
3803                                              p->last_block, blk, blockcnt)) {
3804                 pctx->blk = blockcnt;
3805                 pctx->blk2 = blk;
3806                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3807                 mark_block_used(ctx, blk);
3808                 mark_block_used(ctx, blk);
3809         }
3810         if (blockcnt >= 0)
3811                 p->last_block = blockcnt;
3812         p->previous_block = blk;
3813 mark_dir:
3814         if (p->is_dir && (blockcnt >= 0)) {
3815                 while (++p->last_db_block < blockcnt) {
3816                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
3817                                                               p->ino, 0,
3818                                                               p->last_db_block);
3819                         if (pctx->errcode) {
3820                                 pctx->blk = 0;
3821                                 pctx->num = p->last_db_block;
3822                                 goto failed_add_dir_block;
3823                         }
3824                 }
3825                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
3826                                                       blk, blockcnt);
3827                 if (pctx->errcode) {
3828                         pctx->blk = blk;
3829                         pctx->num = blockcnt;
3830                 failed_add_dir_block:
3831                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3832                         /* Should never get here */
3833                         ctx->flags |= E2F_FLAG_ABORT;
3834                         return BLOCK_ABORT;
3835                 }
3836         }
3837         return ret_code;
3838 }
3839
3840 static int process_bad_block(ext2_filsys fs,
3841                       blk64_t *block_nr,
3842                       e2_blkcnt_t blockcnt,
3843                       blk64_t ref_block EXT2FS_ATTR((unused)),
3844                       int ref_offset EXT2FS_ATTR((unused)),
3845                       void *priv_data)
3846 {
3847         struct process_block_struct *p;
3848         blk64_t         blk = *block_nr;
3849         blk64_t         first_block;
3850         dgrp_t          i;
3851         struct problem_context *pctx;
3852         e2fsck_t        ctx;
3853
3854         if (!blk)
3855                 return 0;
3856
3857         p = (struct process_block_struct *) priv_data;
3858         ctx = p->ctx;
3859         pctx = p->pctx;
3860
3861         pctx->ino = EXT2_BAD_INO;
3862         pctx->blk = blk;
3863         pctx->blkcount = blockcnt;
3864
3865         if ((blk < fs->super->s_first_data_block) ||
3866             (blk >= ext2fs_blocks_count(fs->super))) {
3867                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
3868                         *block_nr = 0;
3869                         return BLOCK_CHANGED;
3870                 } else
3871                         return 0;
3872         }
3873
3874         if (blockcnt < 0) {
3875                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
3876                         p->bbcheck = 1;
3877                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
3878                                 *block_nr = 0;
3879                                 return BLOCK_CHANGED;
3880                         }
3881                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
3882                                                     blk)) {
3883                         p->bbcheck = 1;
3884                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
3885                                         pctx)) {
3886                                 *block_nr = 0;
3887                                 return BLOCK_CHANGED;
3888                         }
3889                         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3890                                 return BLOCK_ABORT;
3891                 } else
3892                         mark_block_used(ctx, blk);
3893                 return 0;
3894         }
3895 #if 0
3896         printf ("DEBUG: Marking %u as bad.\n", blk);
3897 #endif
3898         ctx->fs_badblocks_count++;
3899         /*
3900          * If the block is not used, then mark it as used and return.
3901          * If it is already marked as found, this must mean that
3902          * there's an overlap between the filesystem table blocks
3903          * (bitmaps and inode table) and the bad block list.
3904          */
3905         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
3906                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
3907                 return 0;
3908         }
3909         /*
3910          * Try to find the where the filesystem block was used...
3911          */
3912         first_block = fs->super->s_first_data_block;
3913
3914         for (i = 0; i < fs->group_desc_count; i++ ) {
3915                 pctx->group = i;
3916                 pctx->blk = blk;
3917                 if (!ext2fs_bg_has_super(fs, i))
3918                         goto skip_super;
3919                 if (blk == first_block) {
3920                         if (i == 0) {
3921                                 if (fix_problem(ctx,
3922                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
3923                                                 pctx)) {
3924                                         *block_nr = 0;
3925                                         return BLOCK_CHANGED;
3926                                 }
3927                                 return 0;
3928                         }
3929                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
3930                         return 0;
3931                 }
3932                 if ((blk > first_block) &&
3933                     (blk <= first_block + fs->desc_blocks)) {
3934                         if (i == 0) {
3935                                 pctx->blk = *block_nr;
3936                                 if (fix_problem(ctx,
3937                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
3938                                         *block_nr = 0;
3939                                         return BLOCK_CHANGED;
3940                                 }
3941                                 return 0;
3942                         }
3943                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
3944                         return 0;
3945                 }
3946         skip_super:
3947                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
3948                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
3949                                 ctx->invalid_block_bitmap_flag[i]++;
3950                                 ctx->invalid_bitmaps++;
3951                         }
3952                         return 0;
3953                 }
3954                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
3955                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
3956                                 ctx->invalid_inode_bitmap_flag[i]++;
3957                                 ctx->invalid_bitmaps++;
3958                         }
3959                         return 0;
3960                 }
3961                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
3962                     (blk < (ext2fs_inode_table_loc(fs, i) +
3963                             fs->inode_blocks_per_group))) {
3964                         /*
3965                          * If there are bad blocks in the inode table,
3966                          * the inode scan code will try to do
3967                          * something reasonable automatically.
3968                          */
3969                         return 0;
3970                 }
3971                 first_block += fs->super->s_blocks_per_group;
3972         }
3973         /*
3974          * If we've gotten to this point, then the only
3975          * possibility is that the bad block inode meta data
3976          * is using a bad block.
3977          */
3978         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
3979             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
3980             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
3981                 p->bbcheck = 1;
3982                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
3983                         *block_nr = 0;
3984                         return BLOCK_CHANGED;
3985                 }
3986                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3987                         return BLOCK_ABORT;
3988                 return 0;
3989         }
3990
3991         pctx->group = -1;
3992
3993         /* Warn user that the block wasn't claimed */
3994         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
3995
3996         return 0;
3997 }
3998
3999 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
4000                             const char *name, int num, blk64_t *new_block)
4001 {
4002         ext2_filsys fs = ctx->fs;
4003         dgrp_t          last_grp;
4004         blk64_t         old_block = *new_block;
4005         blk64_t         last_block;
4006         dgrp_t          flexbg;
4007         unsigned        flexbg_size;
4008         int             i, is_flexbg;
4009         char            *buf;
4010         struct problem_context  pctx;
4011
4012         clear_problem_context(&pctx);
4013
4014         pctx.group = group;
4015         pctx.blk = old_block;
4016         pctx.str = name;
4017
4018         /*
4019          * For flex_bg filesystems, first try to allocate the metadata
4020          * within the flex_bg, and if that fails then try finding the
4021          * space anywhere in the filesystem.
4022          */
4023         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
4024         if (is_flexbg) {
4025                 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
4026                 flexbg = group / flexbg_size;
4027                 first_block = ext2fs_group_first_block2(fs,
4028                                                         flexbg_size * flexbg);
4029                 last_grp = group | (flexbg_size - 1);
4030                 if (last_grp >= fs->group_desc_count)
4031                         last_grp = fs->group_desc_count - 1;
4032                 last_block = ext2fs_group_last_block2(fs, last_grp);
4033         } else
4034                 last_block = ext2fs_group_last_block2(fs, group);
4035         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
4036                                                num, ctx->block_found_map,
4037                                                new_block);
4038         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
4039                 pctx.errcode = ext2fs_get_free_blocks2(fs,
4040                                 fs->super->s_first_data_block,
4041                                 ext2fs_blocks_count(fs->super),
4042                                 num, ctx->block_found_map, new_block);
4043         if (pctx.errcode) {
4044                 pctx.num = num;
4045                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4046                 ext2fs_unmark_valid(fs);
4047                 ctx->flags |= E2F_FLAG_ABORT;
4048                 return;
4049         }
4050         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4051         if (pctx.errcode) {
4052                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4053                 ext2fs_unmark_valid(fs);
4054                 ctx->flags |= E2F_FLAG_ABORT;
4055                 return;
4056         }
4057         ext2fs_mark_super_dirty(fs);
4058         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4059         pctx.blk2 = *new_block;
4060         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4061                           PR_1_RELOC_TO), &pctx);
4062         pctx.blk2 = 0;
4063         for (i = 0; i < num; i++) {
4064                 pctx.blk = i;
4065                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4066                 if (old_block) {
4067                         pctx.errcode = io_channel_read_blk64(fs->io,
4068                                    old_block + i, 1, buf);
4069                         if (pctx.errcode)
4070                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4071                         pctx.blk = (*new_block) + i;
4072                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4073                                                               1, buf);
4074                 } else {
4075                         pctx.blk = (*new_block) + i;
4076                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4077                                                            NULL, NULL);
4078                 }
4079
4080                 if (pctx.errcode)
4081                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4082         }
4083         ext2fs_free_mem(&buf);
4084 }
4085
4086 /*
4087  * This routine gets called at the end of pass 1 if bad blocks are
4088  * detected in the superblock, group descriptors, inode_bitmaps, or
4089  * block bitmaps.  At this point, all of the blocks have been mapped
4090  * out, so we can try to allocate new block(s) to replace the bad
4091  * blocks.
4092  */
4093 static void handle_fs_bad_blocks(e2fsck_t ctx)
4094 {
4095         ext2_filsys fs = ctx->fs;
4096         dgrp_t          i;
4097         blk64_t         first_block;
4098         blk64_t         new_blk;
4099
4100         for (i = 0; i < fs->group_desc_count; i++) {
4101                 first_block = ext2fs_group_first_block2(fs, i);
4102
4103                 if (ctx->invalid_block_bitmap_flag[i]) {
4104                         new_blk = ext2fs_block_bitmap_loc(fs, i);
4105                         new_table_block(ctx, first_block, i, _("block bitmap"),
4106                                         1, &new_blk);
4107                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4108                 }
4109                 if (ctx->invalid_inode_bitmap_flag[i]) {
4110                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
4111                         new_table_block(ctx, first_block, i, _("inode bitmap"),
4112                                         1, &new_blk);
4113                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4114                 }
4115                 if (ctx->invalid_inode_table_flag[i]) {
4116                         new_blk = ext2fs_inode_table_loc(fs, i);
4117                         new_table_block(ctx, first_block, i, _("inode table"),
4118                                         fs->inode_blocks_per_group,
4119                                         &new_blk);
4120                         ext2fs_inode_table_loc_set(fs, i, new_blk);
4121                         ctx->flags |= E2F_FLAG_RESTART;
4122                 }
4123         }
4124         ctx->invalid_bitmaps = 0;
4125 }
4126
4127 /*
4128  * This routine marks all blocks which are used by the superblock,
4129  * group descriptors, inode bitmaps, and block bitmaps.
4130  */
4131 static void mark_table_blocks(e2fsck_t ctx)
4132 {
4133         ext2_filsys fs = ctx->fs;
4134         blk64_t b;
4135         dgrp_t  i;
4136         unsigned int    j;
4137         struct problem_context pctx;
4138
4139         clear_problem_context(&pctx);
4140
4141         for (i = 0; i < fs->group_desc_count; i++) {
4142                 pctx.group = i;
4143
4144                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
4145                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
4146
4147                 /*
4148                  * Mark the blocks used for the inode table
4149                  */
4150                 if (ext2fs_inode_table_loc(fs, i)) {
4151                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
4152                              j < fs->inode_blocks_per_group;
4153                              j++, b++) {
4154                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4155                                                              b)) {
4156                                         pctx.blk = b;
4157                                         if (!ctx->invalid_inode_table_flag[i] &&
4158                                             fix_problem(ctx,
4159                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
4160                                                 ctx->invalid_inode_table_flag[i]++;
4161                                                 ctx->invalid_bitmaps++;
4162                                         }
4163                                 } else {
4164                                     ext2fs_mark_block_bitmap2(
4165                                                 ctx->block_found_map, b);
4166                                     ext2fs_mark_block_bitmap2(
4167                                                 ctx->block_metadata_map, b);
4168                                 }
4169                         }
4170                 }
4171
4172                 /*
4173                  * Mark block used for the block bitmap
4174                  */
4175                 if (ext2fs_block_bitmap_loc(fs, i)) {
4176                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4177                                      ext2fs_block_bitmap_loc(fs, i))) {
4178                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
4179                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
4180                                         ctx->invalid_block_bitmap_flag[i]++;
4181                                         ctx->invalid_bitmaps++;
4182                                 }
4183                         } else {
4184                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4185                                      ext2fs_block_bitmap_loc(fs, i));
4186                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4187                                      ext2fs_block_bitmap_loc(fs, i));
4188                         }
4189                 }
4190                 /*
4191                  * Mark block used for the inode bitmap
4192                  */
4193                 if (ext2fs_inode_bitmap_loc(fs, i)) {
4194                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4195                                      ext2fs_inode_bitmap_loc(fs, i))) {
4196                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
4197                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
4198                                         ctx->invalid_inode_bitmap_flag[i]++;
4199                                         ctx->invalid_bitmaps++;
4200                                 }
4201                         } else {
4202                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4203                                      ext2fs_inode_bitmap_loc(fs, i));
4204                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4205                                      ext2fs_inode_bitmap_loc(fs, i));
4206                         }
4207                 }
4208         }
4209 }
4210
4211 /*
4212  * These subroutines short circuits ext2fs_get_blocks and
4213  * ext2fs_check_directory; we use them since we already have the inode
4214  * structure, so there's no point in letting the ext2fs library read
4215  * the inode again.
4216  */
4217 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
4218                                   blk_t *blocks)
4219 {
4220         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4221         int     i;
4222
4223         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4224                 return EXT2_ET_CALLBACK_NOTHANDLED;
4225
4226         for (i=0; i < EXT2_N_BLOCKS; i++)
4227                 blocks[i] = ctx->stashed_inode->i_block[i];
4228         return 0;
4229 }
4230
4231 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
4232                                   struct ext2_inode *inode)
4233 {
4234         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4235
4236         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4237                 return EXT2_ET_CALLBACK_NOTHANDLED;
4238         *inode = *ctx->stashed_inode;
4239         return 0;
4240 }
4241
4242 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
4243                             struct ext2_inode *inode)
4244 {
4245         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4246
4247         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
4248                 (inode != ctx->stashed_inode))
4249                 *ctx->stashed_inode = *inode;
4250         return EXT2_ET_CALLBACK_NOTHANDLED;
4251 }
4252
4253 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
4254 {
4255         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4256
4257         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4258                 return EXT2_ET_CALLBACK_NOTHANDLED;
4259
4260         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
4261                 return EXT2_ET_NO_DIRECTORY;
4262         return 0;
4263 }
4264
4265 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
4266                                         blk64_t *ret)
4267 {
4268         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4269         errcode_t       retval;
4270         blk64_t         new_block;
4271
4272         if (ctx->block_found_map) {
4273                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
4274                                            &new_block);
4275                 if (retval)
4276                         return retval;
4277                 if (fs->block_map) {
4278                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
4279                         ext2fs_mark_bb_dirty(fs);
4280                 }
4281         } else {
4282                 if (!fs->block_map) {
4283                         retval = ext2fs_read_block_bitmap(fs);
4284                         if (retval)
4285                                 return retval;
4286                 }
4287
4288                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
4289                 if (retval)
4290                         return retval;
4291         }
4292
4293         *ret = new_block;
4294         return (0);
4295 }
4296
4297 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
4298                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
4299 {
4300         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4301         errcode_t       retval;
4302
4303         if (ctx->block_found_map)
4304                 return ext2fs_new_range(fs, flags, goal, len,
4305                                         ctx->block_found_map, pblk, plen);
4306
4307         if (!fs->block_map) {
4308                 retval = ext2fs_read_block_bitmap(fs);
4309                 if (retval)
4310                         return retval;
4311         }
4312
4313         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
4314                                 pblk, plen);
4315 }
4316
4317 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
4318 {
4319         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4320
4321         /* Never free a critical metadata block */
4322         if (ctx->block_found_map &&
4323             ctx->block_metadata_map &&
4324             inuse < 0 &&
4325             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
4326                 return;
4327
4328         if (ctx->block_found_map) {
4329                 if (inuse > 0)
4330                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4331                 else
4332                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
4333         }
4334 }
4335
4336 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
4337                                            blk_t num, int inuse)
4338 {
4339         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4340
4341         /* Never free a critical metadata block */
4342         if (ctx->block_found_map &&
4343             ctx->block_metadata_map &&
4344             inuse < 0 &&
4345             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
4346                 return;
4347
4348         if (ctx->block_found_map) {
4349                 if (inuse > 0)
4350                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
4351                                                         blk, num);
4352                 else
4353                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
4354                                                         blk, num);
4355         }
4356 }
4357
4358 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
4359 {
4360         ext2_filsys fs = ctx->fs;
4361
4362         if (use_shortcuts) {
4363                 fs->get_blocks = pass1_get_blocks;
4364                 fs->check_directory = pass1_check_directory;
4365                 fs->read_inode = pass1_read_inode;
4366                 fs->write_inode = pass1_write_inode;
4367                 ctx->stashed_ino = 0;
4368         } else {
4369                 fs->get_blocks = 0;
4370                 fs->check_directory = 0;
4371                 fs->read_inode = 0;
4372                 fs->write_inode = 0;
4373         }
4374 }
4375
4376 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
4377 {
4378         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
4379         ext2fs_set_block_alloc_stats_callback(ctx->fs,
4380                                                 e2fsck_block_alloc_stats, 0);
4381         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
4382         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
4383                                         e2fsck_block_alloc_stats_range, NULL);
4384 }