e2fsck/pass1.c

   1 /*
   2  * pass1.c -- pass #1 of e2fsck: sequential scan of the inode table
   3  *
   4  * Copyright (C) 1993, 1994, 1995, 1996, 1997 Theodore Ts'o.
   5  *
   6  * %Begin-Header%
   7  * This file may be redistributed under the terms of the GNU Public
   8  * License.
   9  * %End-Header%
  10  *
  11  * Pass 1 of e2fsck iterates over all the inodes in the filesystems,
  12  * and applies the following tests to each inode:
  13  *
  14  *      - The mode field of the inode must be legal.
  15  *      - The size and block count fields of the inode are correct.
  16  *      - A data block must not be used by another inode
  17  *
  18  * Pass 1 also gathers the collects the following information:
  19  *
  20  *      - A bitmap of which inodes are in use.          (inode_used_map)
  21  *      - A bitmap of which inodes are directories.     (inode_dir_map)
  22  *      - A bitmap of which inodes are regular files.   (inode_reg_map)
  23  *      - A bitmap of which inodes have bad fields.     (inode_bad_map)
  24  *      - A bitmap of which inodes are in bad blocks.   (inode_bb_map)
  25  *      - A bitmap of which inodes are imagic inodes.   (inode_imagic_map)
  26  *      - A bitmap of which inodes are casefolded.      (inode_casefold_map)
  27  *      - A bitmap of which blocks are in use.          (block_found_map)
  28  *      - A bitmap of which blocks are in use by two inodes     (block_dup_map)
  29  *      - The data blocks of the directory inodes.      (dir_map)
  30  *      - Ref counts for ea_inodes.                     (ea_inode_refs)
  31  *      - The encryption policy ID of each encrypted inode. (encrypted_files)
  32  *
  33  * Pass 1 is designed to stash away enough information so that the
  34  * other passes should not need to read in the inode information
  35  * during the normal course of a filesystem check.  (Although if an
  36  * inconsistency is detected, other passes may need to read in an
  37  * inode to fix it.)
  38  *
  39  * Note that pass 1B will be invoked if there are any duplicate blocks
  40  * found.
  41  */
  42
  43 #define _GNU_SOURCE 1 /* get strnlen() */
  44 #include "config.h"
  45 #include <string.h>
  46 #include <time.h>
  47 #ifdef HAVE_ERRNO_H
  48 #include <errno.h>
  49 #endif
  50 #include <assert.h>
  51 #ifdef HAVE_PTHREAD
  52 #include <pthread.h>
  53 #endif
  54
  55 #include "e2fsck.h"
  56 #include <ext2fs/ext2_ext_attr.h>
  57 /* todo remove this finally */
  58 #include <ext2fs/ext2fsP.h>
  59 #include <e2p/e2p.h>
  60
  61 #include "problem.h"
  62
  63 #ifdef NO_INLINE_FUNCS
  64 #define _INLINE_
  65 #else
  66 #define _INLINE_ inline
  67 #endif
  68
  69 #undef DEBUG
  70
  71 struct ea_quota {
  72         blk64_t blocks;
  73         __u64 inodes;
  74 };
  75
  76 static int process_block(ext2_filsys fs, blk64_t        *blocknr,
  77                          e2_blkcnt_t blockcnt, blk64_t ref_blk,
  78                          int ref_offset, void *priv_data);
  79 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
  80                              e2_blkcnt_t blockcnt, blk64_t ref_blk,
  81                              int ref_offset, void *priv_data);
  82 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
  83                          char *block_buf,
  84                          const struct ea_quota *ea_ibody_quota);
  85 static void mark_table_blocks(e2fsck_t ctx);
  86 static void alloc_bb_map(e2fsck_t ctx);
  87 static void alloc_imagic_map(e2fsck_t ctx);
  88 static void mark_inode_bad(e2fsck_t ctx, ext2_ino_t ino);
  89 static void add_casefolded_dir(e2fsck_t ctx, ext2_ino_t ino);
  90 static void handle_fs_bad_blocks(e2fsck_t ctx);
  91 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
  92 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
  93                                   dgrp_t group, void * priv_data);
  94 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
  95                                     char *block_buf, int adjust_sign);
  96 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
  97
  98 struct process_block_struct {
  99         ext2_ino_t      ino;
 100         unsigned        is_dir:1, is_reg:1, clear:1, suppress:1,
 101                                 fragmented:1, compressed:1, bbcheck:1,
 102                                 inode_modified:1;
 103         blk64_t         num_blocks;
 104         blk64_t         max_blocks;
 105         blk64_t         last_block;
 106         e2_blkcnt_t     last_init_lblock;
 107         e2_blkcnt_t     last_db_block;
 108         int             num_illegal_blocks;
 109         blk64_t         previous_block;
 110         struct ext2_inode *inode;
 111         struct problem_context *pctx;
 112         ext2fs_block_bitmap fs_meta_blocks;
 113         e2fsck_t        ctx;
 114         blk64_t         next_lblock;
 115         struct extent_tree_info eti;
 116 };
 117
 118 struct process_inode_block {
 119         ext2_ino_t ino;
 120         struct ea_quota ea_ibody_quota;
 121         struct ext2_inode_large inode;
 122 };
 123
 124 struct scan_callback_struct {
 125         e2fsck_t                         ctx;
 126         char                            *block_buf;
 127         struct process_inode_block      *inodes_to_process;
 128         int                             *process_inode_count;
 129 };
 130
 131 static void process_inodes(e2fsck_t ctx, char *block_buf,
 132                            struct process_inode_block *inodes_to_process,
 133                            int *process_inode_count);
 134
 135 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
 136                             EXT2_MIN_BLOCK_LOG_SIZE + 1];
 137
 138 /*
 139  * Check to make sure a device inode is real.  Returns 1 if the device
 140  * checks out, 0 if not.
 141  *
 142  * Note: this routine is now also used to check FIFO's and Sockets,
 143  * since they have the same requirement; the i_block fields should be
 144  * zero.
 145  */
 146 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
 147                                     struct ext2_inode *inode)
 148 {
 149         int     i;
 150
 151         /*
 152          * If the index or extents flag is set, then this is a bogus
 153          * device/fifo/socket
 154          */
 155         if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
 156                 return 0;
 157
 158         /*
 159          * We should be able to do the test below all the time, but
 160          * because the kernel doesn't forcibly clear the device
 161          * inode's additional i_block fields, there are some rare
 162          * occasions when a legitimate device inode will have non-zero
 163          * additional i_block fields.  So for now, we only complain
 164          * when the immutable flag is set, which should never happen
 165          * for devices.  (And that's when the problem is caused, since
 166          * you can't set or clear immutable flags for devices.)  Once
 167          * the kernel has been fixed we can change this...
 168          */
 169         if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
 170                 for (i=4; i < EXT2_N_BLOCKS; i++)
 171                         if (inode->i_block[i])
 172                                 return 0;
 173         }
 174         return 1;
 175 }
 176
 177 /*
 178  * Check to make sure a symlink inode is real.  Returns 1 if the symlink
 179  * checks out, 0 if not.
 180  */
 181 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
 182                                struct ext2_inode *inode, char *buf)
 183 {
 184         unsigned int buflen;
 185         unsigned int len;
 186
 187         if ((inode->i_size_high || inode->i_size == 0) ||
 188             (inode->i_flags & EXT2_INDEX_FL))
 189                 return 0;
 190
 191         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
 192                 size_t inline_size;
 193
 194                 if (inode->i_flags & EXT4_EXTENTS_FL)
 195                         return 0;
 196                 if (ext2fs_inline_data_size(fs, ino, &inline_size))
 197                         return 0;
 198                 if (inode->i_size != inline_size)
 199                         return 0;
 200
 201                 return 1;
 202         }
 203
 204         if (ext2fs_is_fast_symlink(inode)) {
 205                 if (inode->i_flags & EXT4_EXTENTS_FL)
 206                         return 0;
 207                 buf = (char *)inode->i_block;
 208                 buflen = sizeof(inode->i_block);
 209         } else {
 210                 ext2_extent_handle_t    handle;
 211                 struct ext2_extent_info info;
 212                 struct ext2fs_extent    extent;
 213                 blk64_t blk;
 214                 int i;
 215
 216                 if (inode->i_flags & EXT4_EXTENTS_FL) {
 217                         if (ext2fs_extent_open2(fs, ino, inode, &handle))
 218                                 return 0;
 219                         if (ext2fs_extent_get_info(handle, &info) ||
 220                             (info.num_entries != 1) ||
 221                             (info.max_depth != 0)) {
 222                                 ext2fs_extent_free(handle);
 223                                 return 0;
 224                         }
 225                         if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
 226                                               &extent) ||
 227                             (extent.e_lblk != 0) ||
 228                             (extent.e_len != 1)) {
 229                                 ext2fs_extent_free(handle);
 230                                 return 0;
 231                         }
 232                         blk = extent.e_pblk;
 233                         ext2fs_extent_free(handle);
 234                 } else {
 235                         blk = inode->i_block[0];
 236
 237                         for (i = 1; i < EXT2_N_BLOCKS; i++)
 238                                 if (inode->i_block[i])
 239                                         return 0;
 240                 }
 241
 242                 if (blk < fs->super->s_first_data_block ||
 243                     blk >= ext2fs_blocks_count(fs->super))
 244                         return 0;
 245
 246                 if (io_channel_read_blk64(fs->io, blk, 1, buf))
 247                         return 0;
 248
 249                 buflen = fs->blocksize;
 250         }
 251
 252         if (inode->i_flags & EXT4_ENCRYPT_FL)
 253                 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
 254         else
 255                 len = strnlen(buf, buflen);
 256
 257         if (len >= buflen)
 258                 return 0;
 259
 260         if (len != inode->i_size)
 261                 return 0;
 262         return 1;
 263 }
 264
 265 /*
 266  * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
 267  */
 268 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
 269 static void check_extents_inlinedata(e2fsck_t ctx,
 270                                      struct problem_context *pctx)
 271 {
 272         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 273                 return;
 274
 275         if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
 276                 return;
 277
 278         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 279         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 280 }
 281 #undef BAD_SPECIAL_FLAGS
 282
 283 /*
 284  * If the immutable (or append-only) flag is set on the inode, offer
 285  * to clear it.
 286  */
 287 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
 288 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
 289 {
 290         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 291                 return;
 292
 293         if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
 294                 return;
 295
 296         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 297         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 298 }
 299
 300 /*
 301  * If device, fifo or socket, check size is zero -- if not offer to
 302  * clear it
 303  */
 304 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
 305 {
 306         struct ext2_inode *inode = pctx->inode;
 307
 308         if (EXT2_I_SIZE(inode) == 0)
 309                 return;
 310
 311         if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
 312                 return;
 313
 314         ext2fs_inode_size_set(ctx->fs, inode, 0);
 315         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 316 }
 317
 318 /*
 319  * For a given size, calculate how many blocks would be charged towards quota.
 320  */
 321 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
 322 {
 323         blk64_t clusters;
 324
 325         clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
 326         return EXT2FS_C2B(fs, clusters);
 327 }
 328
 329 /*
 330  * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
 331  * the problem code.
 332  */
 333 static problem_t check_large_ea_inode(e2fsck_t ctx,
 334                                       struct ext2_ext_attr_entry *entry,
 335                                       struct problem_context *pctx,
 336                                       blk64_t *quota_blocks)
 337 {
 338         struct ext2_inode inode;
 339         __u32 hash, signed_hash;
 340         errcode_t retval;
 341
 342         /* Check if inode is within valid range */
 343         if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
 344             (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
 345                 pctx->num = entry->e_value_inum;
 346                 return PR_1_ATTR_VALUE_EA_INODE;
 347         }
 348
 349         e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
 350
 351         retval = ext2fs_ext_attr_hash_entry3(ctx->fs, entry, NULL, &hash,
 352                                              &signed_hash);
 353         if (retval) {
 354                 com_err("check_large_ea_inode", retval,
 355                         _("while hashing entry with e_value_inum = %u"),
 356                         entry->e_value_inum);
 357                 fatal_error(ctx, 0);
 358         }
 359
 360         if ((hash == entry->e_hash) || (signed_hash == entry->e_hash)) {
 361                 *quota_blocks = size_to_quota_blocks(ctx->fs,
 362                                                      entry->e_value_size);
 363         } else {
 364                 /* This might be an old Lustre-style ea_inode reference. */
 365                 if (inode.i_mtime == pctx->ino &&
 366                     inode.i_generation == pctx->inode->i_generation) {
 367                         *quota_blocks = 0;
 368                 } else {
 369                         /* If target inode is also missing EA_INODE flag,
 370                          * this is likely to be a bad reference.
 371                          */
 372                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 373                                 pctx->num = entry->e_value_inum;
 374                                 return PR_1_ATTR_VALUE_EA_INODE;
 375                         } else {
 376                                 pctx->num = entry->e_hash;
 377                                 return PR_1_ATTR_HASH;
 378                         }
 379                 }
 380         }
 381
 382         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 383                 pctx->num = entry->e_value_inum;
 384                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
 385                         inode.i_flags |= EXT4_EA_INODE_FL;
 386                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
 387                                            &inode);
 388                 } else {
 389                         return PR_1_ATTR_NO_EA_INODE_FL;
 390                 }
 391         }
 392         return 0;
 393 }
 394
 395 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
 396                               struct ext2_ext_attr_entry *first, void *end)
 397 {
 398         struct ext2_ext_attr_entry *entry = first;
 399         struct ext2_ext_attr_entry *np = EXT2_EXT_ATTR_NEXT(entry);
 400
 401         while ((void *) entry < end && (void *) np < end &&
 402                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
 403                 if (!entry->e_value_inum)
 404                         goto next;
 405                 if (!ctx->ea_inode_refs) {
 406                         pctx->errcode = ea_refcount_create(0,
 407                                                            &ctx->ea_inode_refs);
 408                         if (pctx->errcode) {
 409                                 pctx->num = 4;
 410                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
 411                                 ctx->flags |= E2F_FLAG_ABORT;
 412                                 return;
 413                         }
 414                 }
 415                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
 416                                       0);
 417         next:
 418                 entry = np;
 419                 np = EXT2_EXT_ATTR_NEXT(entry);
 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                         if (entry->e_hash != 0 && entry->e_hash != hash)
 506                                 hash = ext2fs_ext_attr_hash_entry_signed(entry,
 507                                                 start + entry->e_value_offs);
 508
 509                         /* e_hash may be 0 in older inode's ea */
 510                         if (entry->e_hash != 0 && entry->e_hash != hash) {
 511                                 pctx->num = entry->e_hash;
 512                                 problem = PR_1_ATTR_HASH;
 513                                 goto fix;
 514                         }
 515                 } else {
 516                         blk64_t quota_blocks;
 517
 518                         problem = check_large_ea_inode(ctx, entry, pctx,
 519                                                        &quota_blocks);
 520                         if (problem != 0)
 521                                 goto fix;
 522
 523                         ea_ibody_quota->blocks += quota_blocks;
 524                         ea_ibody_quota->inodes++;
 525                 }
 526
 527                 /* If EA value is stored in external inode then it does not
 528                  * consume space here */
 529                 if (entry->e_value_inum == 0)
 530                         remain -= entry->e_value_size;
 531
 532                 entry = EXT2_EXT_ATTR_NEXT(entry);
 533         }
 534
 535         if (region_allocate(region, (char *)entry - (char *)header,
 536                             sizeof(__u32))) {
 537                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 538                 goto fix;
 539         }
 540 fix:
 541         if (region)
 542                 region_free(region);
 543         /*
 544          * it seems like a corruption. it's very unlikely we could repair
 545          * EA(s) in automatic fashion -bzzz
 546          */
 547         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
 548                 inc_ea_inode_refs(ctx, pctx,
 549                                   (struct ext2_ext_attr_entry *)start, end);
 550                 return;
 551         }
 552
 553         /* simply remove all possible EA(s) */
 554         *((__u32 *)header) = 0UL;
 555         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 556                                 EXT2_INODE_SIZE(sb), "pass1");
 557         ea_ibody_quota->blocks = 0;
 558         ea_ibody_quota->inodes = 0;
 559 }
 560
 561 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
 562         return (xtime & (1U << 31)) != 0 &&
 563                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
 564 }
 565
 566 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
 567         check_inode_extra_negative_epoch(inode->i_##xtime, \
 568                                          inode->i_##xtime##_extra)
 569
 570 /* When today's date is earlier than 2242, we assume that atimes,
 571  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
 572  * actually pre-1970 dates mis-encoded.
 573  */
 574 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
 575
 576 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
 577                                     struct ea_quota *ea_ibody_quota)
 578 {
 579         struct ext2_super_block *sb = ctx->fs->super;
 580         struct ext2_inode_large *inode;
 581         __u32 *eamagic;
 582         int min, max;
 583
 584         ea_ibody_quota->blocks = 0;
 585         ea_ibody_quota->inodes = 0;
 586
 587         inode = (struct ext2_inode_large *) pctx->inode;
 588         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
 589                 /* this isn't large inode. so, nothing to check */
 590                 return;
 591         }
 592
 593 #if 0
 594         printf("inode #%u, i_extra_size %d\n", pctx->ino,
 595                         inode->i_extra_isize);
 596 #endif
 597         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
 598         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
 599         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
 600         /*
 601          * For now we will allow i_extra_isize to be 0, but really
 602          * implementations should never allow i_extra_isize to be 0
 603          */
 604         if (inode->i_extra_isize &&
 605             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
 606              inode->i_extra_isize & 3)) {
 607                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
 608                         return;
 609                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
 610                         inode->i_extra_isize = sb->s_want_extra_isize;
 611                 else
 612                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
 613                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 614                                         EXT2_INODE_SIZE(sb), "pass1");
 615         }
 616
 617         /* check if there is no place for an EA header */
 618         if (inode->i_extra_isize >= max - sizeof(__u32))
 619                 return;
 620
 621         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 622                         inode->i_extra_isize);
 623         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
 624                 /* it seems inode has an extended attribute(s) in body */
 625                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
 626         }
 627
 628         /*
 629          * If the inode's extended atime (ctime, crtime, mtime) is stored in
 630          * the old, invalid format, repair it.
 631          */
 632         if (((sizeof(time_t) <= 4) ||
 633              (((sizeof(time_t) > 4) &&
 634                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
 635             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
 636              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
 637              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
 638              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
 639
 640                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
 641                         return;
 642
 643                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
 644                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
 645                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
 646                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
 647                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
 648                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
 649                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
 650                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
 651                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 652                                         EXT2_INODE_SIZE(sb), "pass1");
 653         }
 654
 655 }
 656
 657 /*
 658  * Check to see if the inode might really be a directory, despite i_mode
 659  *
 660  * This is a lot of complexity for something for which I'm not really
 661  * convinced happens frequently in the wild.  If for any reason this
 662  * causes any problems, take this code out.
 663  * [tytso:20070331.0827EDT]
 664  */
 665 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
 666                                 char *buf)
 667 {
 668         struct ext2_inode *inode = pctx->inode;
 669         struct ext2_dir_entry   *dirent;
 670         errcode_t               retval;
 671         blk64_t                 blk;
 672         unsigned int            i, rec_len, not_device = 0;
 673         int                     extent_fs;
 674         int                     inlinedata_fs;
 675
 676         /*
 677          * If the mode looks OK, we believe it.  If the first block in
 678          * the i_block array is 0, this cannot be a directory. If the
 679          * inode is extent-mapped, it is still the case that the latter
 680          * cannot be 0 - the magic number in the extent header would make
 681          * it nonzero.
 682          */
 683         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
 684             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
 685                 return;
 686
 687         /*
 688          * Check the block numbers in the i_block array for validity:
 689          * zero blocks are skipped (but the first one cannot be zero -
 690          * see above), other blocks are checked against the first and
 691          * max data blocks (from the the superblock) and against the
 692          * block bitmap. Any invalid block found means this cannot be
 693          * a directory.
 694          *
 695          * If there are non-zero blocks past the fourth entry, then
 696          * this cannot be a device file: we remember that for the next
 697          * check.
 698          *
 699          * For extent mapped files, we don't do any sanity checking:
 700          * just try to get the phys block of logical block 0 and run
 701          * with it.
 702          *
 703          * For inline data files, we just try to get the size of inline
 704          * data.  If it's true, we will treat it as a directory.
 705          */
 706
 707         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
 708         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
 709         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
 710                 size_t size;
 711                 __u32 dotdot;
 712                 unsigned int rec_len2;
 713                 struct ext2_dir_entry de;
 714
 715                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
 716                         return;
 717                 /*
 718                  * If the size isn't a multiple of 4, it's probably not a
 719                  * directory??
 720                  */
 721                 if (size & 3)
 722                         return;
 723                 /*
 724                  * If the first 10 bytes don't look like a directory entry,
 725                  * it's probably not a directory.
 726                  */
 727                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
 728                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
 729                        EXT2_DIR_REC_LEN(0));
 730                 dotdot = ext2fs_le32_to_cpu(dotdot);
 731                 de.inode = ext2fs_le32_to_cpu(de.inode);
 732                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
 733                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
 734                 if (dotdot >= ctx->fs->super->s_inodes_count ||
 735                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
 736                      dotdot != EXT2_ROOT_INO) ||
 737                     de.inode >= ctx->fs->super->s_inodes_count ||
 738                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
 739                      de.inode != 0) ||
 740                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
 741                               EXT4_INLINE_DATA_DOTDOT_SIZE)
 742                         return;
 743                 /* device files never have a "system.data" entry */
 744                 goto isdir;
 745         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
 746                 /* extent mapped */
 747                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
 748                                  &blk))
 749                         return;
 750                 /* device files are never extent mapped */
 751                 not_device++;
 752         } else {
 753                 for (i=0; i < EXT2_N_BLOCKS; i++) {
 754                         blk = inode->i_block[i];
 755                         if (!blk)
 756                                 continue;
 757                         if (i >= 4)
 758                                 not_device++;
 759
 760                         if (blk < ctx->fs->super->s_first_data_block ||
 761                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
 762                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
 763                                                            blk))
 764                                 return; /* Invalid block, can't be dir */
 765                 }
 766                 blk = inode->i_block[0];
 767         }
 768
 769         /*
 770          * If the mode says this is a device file and the i_links_count field
 771          * is sane and we have not ruled it out as a device file previously,
 772          * we declare it a device file, not a directory.
 773          */
 774         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
 775             (inode->i_links_count == 1) && !not_device)
 776                 return;
 777
 778         /* read the first block */
 779         ehandler_operation(_("reading directory block"));
 780         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
 781         ehandler_operation(0);
 782         if (retval)
 783                 return;
 784
 785         dirent = (struct ext2_dir_entry *) buf;
 786         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 787         if (retval)
 788                 return;
 789         if ((ext2fs_dirent_name_len(dirent) != 1) ||
 790             (dirent->name[0] != '.') ||
 791             (dirent->inode != pctx->ino) ||
 792             (rec_len < 12) ||
 793             (rec_len % 4) ||
 794             (rec_len >= ctx->fs->blocksize - 12))
 795                 return;
 796
 797         dirent = (struct ext2_dir_entry *) (buf + rec_len);
 798         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 799         if (retval)
 800                 return;
 801         if ((ext2fs_dirent_name_len(dirent) != 2) ||
 802             (dirent->name[0] != '.') ||
 803             (dirent->name[1] != '.') ||
 804             (rec_len < 12) ||
 805             (rec_len % 4))
 806                 return;
 807
 808 isdir:
 809         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
 810                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
 811                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
 812                                         EXT2_INODE_SIZE(ctx->fs->super),
 813                                         "check_is_really_dir");
 814         }
 815 }
 816
 817 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
 818                                      int flags, ext2_icount_t hint,
 819                                      ext2_icount_t *ret)
 820 {
 821         unsigned int            threshold;
 822         unsigned int            save_type;
 823         ext2_ino_t              num_dirs;
 824         errcode_t               retval;
 825         char                    *tdb_dir;
 826         int                     enable;
 827
 828         *ret = 0;
 829
 830         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
 831                            &tdb_dir);
 832         profile_get_uint(ctx->profile, "scratch_files",
 833                          "numdirs_threshold", 0, 0, &threshold);
 834         profile_get_boolean(ctx->profile, "scratch_files",
 835                             "icount", 0, 1, &enable);
 836
 837         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
 838         if (retval)
 839                 num_dirs = 1024;        /* Guess */
 840
 841         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
 842             (!threshold || num_dirs > threshold)) {
 843                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
 844                                                   flags, ret);
 845                 if (retval == 0)
 846                         return 0;
 847         }
 848         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
 849                                &save_type);
 850         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
 851                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
 852         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
 853         ctx->fs->default_bitmap_type = save_type;
 854         return retval;
 855 }
 856
 857 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
 858                                             e2fsck_t ctx,
 859                                             struct problem_context *pctx)
 860 {
 861         errcode_t retval;
 862         struct ext2_inode_large inode;
 863
 864         /*
 865          * Reread inode.  If we don't see checksum error, then this inode
 866          * has been fixed elsewhere.
 867          */
 868         ctx->stashed_ino = 0;
 869         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
 870                                         sizeof(inode));
 871         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
 872                 return retval;
 873         if (!retval)
 874                 return 0;
 875
 876         /*
 877          * Checksum still doesn't match.  That implies that the inode passes
 878          * all the sanity checks, so maybe the checksum is simply corrupt.
 879          * See if the user will go for fixing that.
 880          */
 881         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
 882                 return 0;
 883
 884         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
 885                                          sizeof(inode));
 886         return retval;
 887 }
 888
 889 static void reserve_block_for_root_repair(e2fsck_t ctx)
 890 {
 891         blk64_t         blk = 0;
 892         errcode_t       err;
 893         ext2_filsys     fs = ctx->fs;
 894
 895         ctx->root_repair_block = 0;
 896         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
 897                 return;
 898
 899         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 900         if (err)
 901                 return;
 902         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 903         ctx->root_repair_block = blk;
 904 }
 905
 906 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
 907 {
 908         blk64_t         blk = 0;
 909         errcode_t       err;
 910         ext2_filsys     fs = ctx->fs;
 911         static const char name[] = "lost+found";
 912         ext2_ino_t      ino;
 913
 914         ctx->lnf_repair_block = 0;
 915         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
 916                 return;
 917
 918         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 919         if (err)
 920                 return;
 921         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 922         ctx->lnf_repair_block = blk;
 923 }
 924
 925 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
 926                                          struct ext2_inode *inode,
 927                                          size_t *sz)
 928 {
 929         void *p;
 930         struct ext2_xattr_handle *handle;
 931         errcode_t retval;
 932
 933         retval = ext2fs_xattrs_open(fs, ino, &handle);
 934         if (retval)
 935                 return retval;
 936
 937         retval = ext2fs_xattrs_read_inode(handle,
 938                                           (struct ext2_inode_large *)inode);
 939         if (retval)
 940                 goto err;
 941
 942         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
 943         if (retval)
 944                 goto err;
 945         ext2fs_free_mem(&p);
 946 err:
 947         (void) ext2fs_xattrs_close(&handle);
 948         return retval;
 949 }
 950
 951 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
 952                                     struct problem_context *pctx,
 953                                     int failed_csum)
 954 {
 955         if (!failed_csum)
 956                 return;
 957
 958         /*
 959          * If the inode failed the checksum and the user didn't
 960          * clear the inode, test the checksum again -- if it still
 961          * fails, ask the user if the checksum should be corrected.
 962          */
 963         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
 964         if (pctx->errcode)
 965                 ctx->flags |= E2F_FLAG_ABORT;
 966 }
 967 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
 968         do { \
 969                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
 970                 if ((ctx)->flags & E2F_FLAG_ABORT) \
 971                         return; \
 972         } while (0)
 973
 974 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
 975 {
 976         __u32 x;
 977         int i;
 978
 979         for (i = 0; i < EXT2_N_BLOCKS; i++) {
 980                 x = inode->i_block[i];
 981 #ifdef WORDS_BIGENDIAN
 982                 x = ext2fs_swab32(x);
 983 #endif
 984                 if (x >= ext2fs_blocks_count(fs->super))
 985                         return 0;
 986         }
 987
 988         return 1;
 989 }
 990
 991 /*
 992  * Figure out what to do with an inode that has both extents and inline data
 993  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
 994  */
 995 static int fix_inline_data_extents_file(e2fsck_t ctx,
 996                                         ext2_ino_t ino,
 997                                         struct ext2_inode *inode,
 998                                         int inode_size,
 999                                         struct problem_context *pctx)
1000 {
1001         size_t max_inline_ea_size;
1002         ext2_filsys fs = ctx->fs;
1003         int dirty = 0;
1004
1005         /* Both feature flags not set?  Just run the regular checks */
1006         if (!ext2fs_has_feature_extents(fs->super) &&
1007             !ext2fs_has_feature_inline_data(fs->super))
1008                 return 0;
1009
1010         /* Clear both flags if it's a special file */
1011         if (LINUX_S_ISCHR(inode->i_mode) ||
1012             LINUX_S_ISBLK(inode->i_mode) ||
1013             LINUX_S_ISFIFO(inode->i_mode) ||
1014             LINUX_S_ISSOCK(inode->i_mode)) {
1015                 check_extents_inlinedata(ctx, pctx);
1016                 return 0;
1017         }
1018
1019         /* If it looks like an extent tree, try to clear inlinedata */
1020         if (ext2fs_extent_header_verify(inode->i_block,
1021                                  sizeof(inode->i_block)) == 0 &&
1022             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1023                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1024                 dirty = 1;
1025                 goto out;
1026         }
1027
1028         /* If it looks short enough to be inline data, try to clear extents */
1029         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1030                 max_inline_ea_size = inode_size -
1031                                      (EXT2_GOOD_OLD_INODE_SIZE +
1032                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1033         else
1034                 max_inline_ea_size = 0;
1035         if (EXT2_I_SIZE(inode) <
1036             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1037             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1038                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1039                 dirty = 1;
1040                 goto out;
1041         }
1042
1043         /*
1044          * Too big for inline data, but no evidence of extent tree -
1045          * maybe it's a block map file?  If the mappings all look valid?
1046          */
1047         if (could_be_block_map(fs, inode) &&
1048             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1049 #ifdef WORDS_BIGENDIAN
1050                 int i;
1051
1052                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1053                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1054 #endif
1055
1056                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1057                 dirty = 1;
1058                 goto out;
1059         }
1060
1061         /* Oh well, just clear the busted inode. */
1062         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1063                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1064                 return -1;
1065         }
1066
1067 out:
1068         if (dirty)
1069                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1070
1071         return 0;
1072 }
1073
1074 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1075 {
1076         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1077         dgrp_t start = *group, grp;
1078         blk64_t blocks_to_read = 0;
1079         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1080
1081         if (ctx->readahead_kb == 0)
1082                 goto out;
1083
1084         /* Keep iterating groups until we have enough to readahead */
1085         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1086         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1087                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1088                         continue;
1089                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1090                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1091                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1092                                         inodes_per_block;
1093                 if (blocks_to_read * ctx->fs->blocksize >
1094                     ctx->readahead_kb * 1024)
1095                         break;
1096         }
1097
1098         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1099                                grp - start + 1);
1100         if (err == EAGAIN) {
1101                 ctx->readahead_kb /= 2;
1102                 err = 0;
1103         }
1104
1105 out:
1106         if (err) {
1107                 /* Error; disable itable readahead */
1108                 *group = ctx->fs->group_desc_count;
1109                 *next_ino = ctx->fs->super->s_inodes_count;
1110         } else {
1111                 /*
1112                  * Don't do more readahead until we've reached the first inode
1113                  * of the last inode scan buffer block for the last group.
1114                  */
1115                 *group = grp + 1;
1116                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1117                                      ctx->inode_buffer_blocks :
1118                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1119                                     ctx->fs->blocksize /
1120                                     EXT2_INODE_SIZE(ctx->fs->super);
1121                 inodes_in_group--;
1122                 *next_ino = inodes_in_group -
1123                             (inodes_in_group % inodes_per_buffer) + 1 +
1124                             (grp * ctx->fs->super->s_inodes_per_group);
1125         }
1126 }
1127
1128 /*
1129  * Check if the passed ino is one of the used superblock quota inodes.
1130  *
1131  * Before the quota inodes were journaled, older superblock quota inodes
1132  * were just regular files in the filesystem and not reserved inodes.  This
1133  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1134  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1135  */
1136 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1137 {
1138         enum quota_type qtype;
1139
1140         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1141                 if (*quota_sb_inump(sb, qtype) == ino)
1142                         return 1;
1143
1144         return 0;
1145 }
1146
1147 /*
1148  * Check if the passed ino is one of the reserved quota inodes.
1149  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1150  * inodes.  These inodes may or may not be in use by the quota feature.
1151  */
1152 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1153 {
1154         enum quota_type qtype;
1155
1156         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1157                 if (quota_type2inum(qtype, fs->super) == ino)
1158                         return 1;
1159
1160         return 0;
1161 }
1162
1163 static int e2fsck_should_abort(e2fsck_t ctx)
1164 {
1165         e2fsck_t global_ctx;
1166
1167         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1168                 return 1;
1169
1170         if (ctx->global_ctx) {
1171                 global_ctx = ctx->global_ctx;
1172                 if (global_ctx->flags & E2F_FLAG_SIGNAL_MASK)
1173                         return 1;
1174         }
1175         return 0;
1176 }
1177
1178 void e2fsck_pass1_run(e2fsck_t ctx)
1179 {
1180         int     i;
1181         ext2_filsys fs = ctx->fs;
1182         ext2_ino_t      ino = 0;
1183         struct ext2_inode *inode = NULL;
1184         ext2_inode_scan scan = NULL;
1185         char            *block_buf = NULL;
1186 #ifdef RESOURCE_TRACK
1187         struct resource_track   rtrack;
1188 #endif
1189         unsigned char   frag, fsize;
1190         struct          problem_context pctx;
1191         struct          scan_callback_struct scan_struct;
1192         struct ext2_super_block *sb = ctx->fs->super;
1193         const char      *old_op;
1194         const char      *eop_next_inode = _("getting next inode from scan");
1195         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1196         int             low_dtime_check = 1;
1197         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1198         unsigned int    bufsize;
1199         int             failed_csum = 0;
1200         ext2_ino_t      ino_threshold = 0;
1201         dgrp_t          ra_group = 0;
1202         struct ea_quota ea_ibody_quota;
1203         struct process_inode_block *inodes_to_process;
1204         int             process_inode_count;
1205
1206         init_resource_track(&rtrack, ctx->fs->io);
1207         clear_problem_context(&pctx);
1208
1209         /* If we can do readahead, figure out how many groups to pull in. */
1210         if (!e2fsck_can_readahead(ctx->fs))
1211                 ctx->readahead_kb = 0;
1212         else if (ctx->readahead_kb == ~0ULL)
1213                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1214         pass1_readahead(ctx, &ra_group, &ino_threshold);
1215
1216         if (!(ctx->options & E2F_OPT_PREEN))
1217                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1218
1219         if (ext2fs_has_feature_dir_index(fs->super) &&
1220             !(ctx->options & E2F_OPT_NO)) {
1221                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1222                         ctx->dirs_to_hash = 0;
1223         }
1224
1225 #ifdef MTRACE
1226         mtrace_print("Pass 1");
1227 #endif
1228
1229         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1230         extent_fs = ext2fs_has_feature_extents(sb);
1231         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1232         casefold_fs = ext2fs_has_feature_casefold(sb);
1233
1234         /*
1235          * Allocate bitmaps structures
1236          */
1237         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1238                                                     EXT2FS_BMAP64_RBTREE,
1239                                                     "inode_used_map",
1240                                                     &ctx->inode_used_map);
1241         if (pctx.errcode) {
1242                 pctx.num = 1;
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                         _("directory inode map"),
1249                         ctx->global_ctx ? EXT2FS_BMAP64_RBTREE :
1250                         EXT2FS_BMAP64_AUTODIR,
1251                         "inode_dir_map", &ctx->inode_dir_map);
1252         if (pctx.errcode) {
1253                 pctx.num = 2;
1254                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1255                 ctx->flags |= E2F_FLAG_ABORT;
1256                 return;
1257         }
1258         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1259                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1260                         "inode_reg_map", &ctx->inode_reg_map);
1261         if (pctx.errcode) {
1262                 pctx.num = 6;
1263                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1264                 ctx->flags |= E2F_FLAG_ABORT;
1265                 return;
1266         }
1267         pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1268                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1269                         "block_found_map", &ctx->block_found_map);
1270         if (pctx.errcode) {
1271                 pctx.num = 1;
1272                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1273                 ctx->flags |= E2F_FLAG_ABORT;
1274                 return;
1275         }
1276         pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1277                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1278                         "block_metadata_map", &ctx->block_metadata_map);
1279         if (pctx.errcode) {
1280                 pctx.num = 1;
1281                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1282                 ctx->flags |= E2F_FLAG_ABORT;
1283                 return;
1284         }
1285         if (casefold_fs) {
1286                 pctx.errcode =
1287                         e2fsck_allocate_inode_bitmap(fs,
1288                                                      _("inode casefold map"),
1289                                                      EXT2FS_BMAP64_RBTREE,
1290                                                      "inode_casefold_map",
1291                                                      &ctx->inode_casefold_map);
1292                 if (pctx.errcode) {
1293                         pctx.num = 1;
1294                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1295                         ctx->flags |= E2F_FLAG_ABORT;
1296                         return;
1297                 }
1298         }
1299         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1300                                            &ctx->inode_link_info);
1301         if (pctx.errcode) {
1302                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1303                 ctx->flags |= E2F_FLAG_ABORT;
1304                 return;
1305         }
1306         bufsize = inode_size;
1307         if (bufsize < sizeof(struct ext2_inode_large))
1308                 bufsize = sizeof(struct ext2_inode_large);
1309         inode = (struct ext2_inode *)
1310                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1311
1312         inodes_to_process = (struct process_inode_block *)
1313                 e2fsck_allocate_memory(ctx,
1314                                        (ctx->process_inode_size *
1315                                         sizeof(struct process_inode_block)),
1316                                        "array of inodes to process");
1317         process_inode_count = 0;
1318
1319         pctx.errcode = ext2fs_init_dblist(fs, 0);
1320         if (pctx.errcode) {
1321                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1322                 ctx->flags |= E2F_FLAG_ABORT;
1323                 goto endit;
1324         }
1325
1326         /*
1327          * If the last orphan field is set, clear it, since the pass1
1328          * processing will automatically find and clear the orphans.
1329          * In the future, we may want to try using the last_orphan
1330          * linked list ourselves, but for now, we clear it so that the
1331          * ext3 mount code won't get confused.
1332          */
1333         if (!(ctx->options & E2F_OPT_READONLY)) {
1334                 if (fs->super->s_last_orphan) {
1335                         fs->super->s_last_orphan = 0;
1336                         ext2fs_mark_super_dirty(fs);
1337                 }
1338         }
1339
1340         mark_table_blocks(ctx);
1341         pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1342                                                 &ctx->block_found_map);
1343         if (pctx.errcode) {
1344                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1345                 ctx->flags |= E2F_FLAG_ABORT;
1346                 goto endit;
1347         }
1348         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1349                                                     "block iterate buffer");
1350         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1351                 e2fsck_use_inode_shortcuts(ctx, 1);
1352         e2fsck_intercept_block_allocations(ctx);
1353         old_op = ehandler_operation(_("opening inode scan"));
1354         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1355                                               &scan);
1356         ehandler_operation(old_op);
1357         if (pctx.errcode) {
1358                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1359                 ctx->flags |= E2F_FLAG_ABORT;
1360                 goto endit;
1361         }
1362         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1363                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1364         ctx->stashed_inode = inode;
1365         scan_struct.ctx = ctx;
1366         scan_struct.block_buf = block_buf;
1367         scan_struct.inodes_to_process = inodes_to_process;
1368         scan_struct.process_inode_count = &process_inode_count;
1369         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1370         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1371                                               ctx->fs->group_desc_count)))
1372                 goto endit;
1373         if ((fs->super->s_wtime &&
1374              fs->super->s_wtime < fs->super->s_inodes_count) ||
1375             (fs->super->s_mtime &&
1376              fs->super->s_mtime < fs->super->s_inodes_count) ||
1377             (fs->super->s_mkfs_time &&
1378              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1379                 low_dtime_check = 0;
1380
1381         if (ext2fs_has_feature_mmp(fs->super) &&
1382             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1383             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1384                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1385                                           fs->super->s_mmp_block);
1386
1387         /* Set up ctx->lost_and_found if possible */
1388         (void) e2fsck_get_lost_and_found(ctx, 0);
1389
1390 #ifdef HAVE_PTHREAD
1391         if (ctx->global_ctx) {
1392                 if (ctx->options & E2F_OPT_DEBUG &&
1393                     ctx->options & E2F_OPT_MULTITHREAD)
1394                         fprintf(stderr, "thread %d jumping to group %d\n",
1395                                         ctx->thread_info.et_thread_index,
1396                                         ctx->thread_info.et_group_start);
1397                 pctx.errcode = ext2fs_inode_scan_goto_blockgroup(scan,
1398                                         ctx->thread_info.et_group_start);
1399                 if (pctx.errcode) {
1400                         fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1401                         ctx->flags |= E2F_FLAG_ABORT;
1402                         goto endit;
1403                 }
1404         }
1405 #endif
1406
1407         while (1) {
1408                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1409                         if (e2fsck_mmp_update(fs))
1410                                 fatal_error(ctx, 0);
1411                 }
1412                 old_op = ehandler_operation(eop_next_inode);
1413                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1414                                                           inode, inode_size);
1415                 if (ino > ino_threshold)
1416                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1417                 ehandler_operation(old_op);
1418                 if (e2fsck_should_abort(ctx))
1419                         goto endit;
1420                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1421                         /*
1422                          * If badblocks says badblocks is bad, offer to clear
1423                          * the list, update the in-core bb list, and restart
1424                          * the inode scan.
1425                          */
1426                         if (ino == EXT2_BAD_INO &&
1427                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1428                                         &pctx)) {
1429                                 errcode_t err;
1430
1431                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1432                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1433                                 ctx->fs->badblocks = NULL;
1434                                 err = ext2fs_read_bb_inode(ctx->fs,
1435                                                         &ctx->fs->badblocks);
1436                                 if (err) {
1437                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1438                                                     &pctx);
1439                                         ctx->flags |= E2F_FLAG_ABORT;
1440                                 } else
1441                                         ctx->flags |= E2F_FLAG_RESTART;
1442                                 goto endit;
1443                         }
1444                         if (!ctx->inode_bb_map)
1445                                 alloc_bb_map(ctx);
1446                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1447                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1448                         continue;
1449                 }
1450                 if (pctx.errcode == EXT2_ET_SCAN_FINISHED)
1451                         break;
1452                 if (pctx.errcode &&
1453                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1454                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1455                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1456                         ctx->flags |= E2F_FLAG_ABORT;
1457                         goto endit;
1458                 }
1459                 if (!ino)
1460                         break;
1461 #ifdef HAVE_PTHREAD
1462                 if (ctx->global_ctx)
1463                         ctx->thread_info.et_inode_number++;
1464 #endif
1465                 pctx.ino = ino;
1466                 pctx.inode = inode;
1467                 ctx->stashed_ino = ino;
1468
1469                 /* Clear trashed inode? */
1470                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1471                     inode->i_links_count > 0 &&
1472                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1473                         pctx.errcode = 0;
1474                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1475                 }
1476                 failed_csum = pctx.errcode != 0;
1477
1478                 /*
1479                  * Check for inodes who might have been part of the
1480                  * orphaned list linked list.  They should have gotten
1481                  * dealt with by now, unless the list had somehow been
1482                  * corrupted.
1483                  *
1484                  * FIXME: In the future, inodes which are still in use
1485                  * (and which are therefore) pending truncation should
1486                  * be handled specially.  Right now we just clear the
1487                  * dtime field, and the normal e2fsck handling of
1488                  * inodes where i_size and the inode blocks are
1489                  * inconsistent is to fix i_size, instead of releasing
1490                  * the extra blocks.  This won't catch the inodes that
1491                  * was at the end of the orphan list, but it's better
1492                  * than nothing.  The right answer is that there
1493                  * shouldn't be any bugs in the orphan list handling.  :-)
1494                  */
1495                 if (inode->i_dtime && low_dtime_check &&
1496                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1497                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1498                                 inode->i_dtime = inode->i_links_count ?
1499                                         0 : ctx->now;
1500                                 e2fsck_write_inode(ctx, ino, inode,
1501                                                    "pass1");
1502                                 failed_csum = 0;
1503                         }
1504                 }
1505
1506                 if (inode->i_links_count) {
1507                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1508                                            ino, inode->i_links_count);
1509                         if (pctx.errcode) {
1510                                 pctx.num = inode->i_links_count;
1511                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1512                                 ctx->flags |= E2F_FLAG_ABORT;
1513                                 goto endit;
1514                         }
1515                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1516                            !quota_inum_is_reserved(fs, ino)) {
1517                         if (!inode->i_dtime && inode->i_mode) {
1518                                 if (fix_problem(ctx,
1519                                             PR_1_ZERO_DTIME, &pctx)) {
1520                                         inode->i_dtime = ctx->now;
1521                                         e2fsck_write_inode(ctx, ino, inode,
1522                                                            "pass1");
1523                                         failed_csum = 0;
1524                                 }
1525                         }
1526                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1527                         continue;
1528                 }
1529
1530                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1531                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1532                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1533                      (!casefold_fs &&
1534                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1535                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1536                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1537                 }
1538
1539                 /* Conflicting inlinedata/extents inode flags? */
1540                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1541                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1542                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1543                                                                inode_size,
1544                                                                &pctx);
1545                         if (res < 0) {
1546                                 /* skip FINISH_INODE_LOOP */
1547                                 continue;
1548                         }
1549                 }
1550
1551                 /* Test for incorrect inline_data flags settings. */
1552                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1553                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1554                         size_t size = 0;
1555
1556                         pctx.errcode = get_inline_data_ea_size(fs, ino, inode,
1557                                                                &size);
1558                         if (!pctx.errcode &&
1559                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1560                                 ext2fs_set_feature_inline_data(sb);
1561                                 ext2fs_mark_super_dirty(fs);
1562                                 inlinedata_fs = 1;
1563                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1564                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1565                                 /* skip FINISH_INODE_LOOP */
1566                                 continue;
1567                         }
1568                 }
1569
1570                 /* Test for inline data flag but no attr */
1571                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1572                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1573                         size_t size = 0;
1574                         errcode_t err;
1575                         int flags;
1576
1577                         flags = fs->flags;
1578                         if (failed_csum)
1579                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1580                         err = get_inline_data_ea_size(fs, ino, inode, &size);
1581                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1582                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1583
1584                         switch (err) {
1585                         case 0:
1586                                 /* Everything is awesome... */
1587                                 break;
1588                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1589                         case EXT2_ET_BAD_EA_HASH:
1590                         case EXT2_ET_BAD_EA_HEADER:
1591                         case EXT2_ET_EA_BAD_NAME_LEN:
1592                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1593                         case EXT2_ET_EA_KEY_NOT_FOUND:
1594                         case EXT2_ET_EA_NO_SPACE:
1595                         case EXT2_ET_MISSING_EA_FEATURE:
1596                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1597                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1598                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1599                         case EXT2_ET_NO_INLINE_DATA:
1600                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1601                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1602                         case EXT2_ET_EA_INODE_CORRUPTED:
1603                                 /* broken EA or no system.data EA; truncate */
1604                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1605                                                 &pctx)) {
1606                                         err = ext2fs_inode_size_set(fs, inode, 0);
1607                                         if (err) {
1608                                                 pctx.errcode = err;
1609                                                 ctx->flags |= E2F_FLAG_ABORT;
1610                                                 goto endit;
1611                                         }
1612                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1613                                         memset(&inode->i_block, 0,
1614                                                sizeof(inode->i_block));
1615                                         e2fsck_write_inode(ctx, ino, inode,
1616                                                            "pass1");
1617                                         failed_csum = 0;
1618                                 }
1619                                 break;
1620                         default:
1621                                 /* Some other kind of non-xattr error? */
1622                                 pctx.errcode = err;
1623                                 ctx->flags |= E2F_FLAG_ABORT;
1624                                 goto endit;
1625                         }
1626                 }
1627
1628                 /*
1629                  * Test for incorrect extent flag settings.
1630                  *
1631                  * On big-endian machines we must be careful:
1632                  * When the inode is read, the i_block array is not swapped
1633                  * if the extent flag is set.  Therefore if we are testing
1634                  * for or fixing a wrongly-set flag, we must potentially
1635                  * (un)swap before testing, or after fixing.
1636                  */
1637
1638                 /*
1639                  * In this case the extents flag was set when read, so
1640                  * extent_header_verify is ok.  If the inode is cleared,
1641                  * no need to swap... so no extra swapping here.
1642                  */
1643                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1644                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1645                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1646                         if ((ext2fs_extent_header_verify(inode->i_block,
1647                                                  sizeof(inode->i_block)) == 0) &&
1648                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1649                                 ext2fs_set_feature_extents(sb);
1650                                 ext2fs_mark_super_dirty(fs);
1651                                 extent_fs = 1;
1652                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1653                         clear_inode:
1654                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1655                                 if (ino == EXT2_BAD_INO)
1656                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1657                                                                  ino);
1658                                 /* skip FINISH_INODE_LOOP */
1659                                 continue;
1660                         }
1661                 }
1662
1663                 /*
1664                  * For big-endian machines:
1665                  * If the inode didn't have the extents flag set when it
1666                  * was read, then the i_blocks array was swapped.  To test
1667                  * as an extents header, we must swap it back first.
1668                  * IF we then set the extents flag, the entire i_block
1669                  * array must be un/re-swapped to make it proper extents data.
1670                  */
1671                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1672                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1673                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1674                     (LINUX_S_ISREG(inode->i_mode) ||
1675                      LINUX_S_ISDIR(inode->i_mode))) {
1676                         void *ehp;
1677 #ifdef WORDS_BIGENDIAN
1678                         __u32 tmp_block[EXT2_N_BLOCKS];
1679
1680                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1681                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1682                         ehp = tmp_block;
1683 #else
1684                         ehp = inode->i_block;
1685 #endif
1686                         if ((ext2fs_extent_header_verify(ehp,
1687                                          sizeof(inode->i_block)) == 0) &&
1688                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1689                                 inode->i_flags |= EXT4_EXTENTS_FL;
1690 #ifdef WORDS_BIGENDIAN
1691                                 memcpy(inode->i_block, tmp_block,
1692                                        sizeof(inode->i_block));
1693 #endif
1694                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1695                                 failed_csum = 0;
1696                         }
1697                 }
1698
1699                 if (ino == EXT2_BAD_INO) {
1700                         struct process_block_struct pb;
1701
1702                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1703                              inode->i_gid || inode->i_links_count ||
1704                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1705                              inode->i_file_acl) &&
1706                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1707                                 memset(inode, 0, sizeof(struct ext2_inode));
1708                                 e2fsck_write_inode(ctx, ino, inode,
1709                                                    "clear bad inode");
1710                                 failed_csum = 0;
1711                         }
1712
1713                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1714                                                           &pb.fs_meta_blocks);
1715                         if (pctx.errcode) {
1716                                 pctx.num = 4;
1717                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1718                                 ctx->flags |= E2F_FLAG_ABORT;
1719                                 goto endit;
1720                         }
1721                         pb.ino = EXT2_BAD_INO;
1722                         pb.num_blocks = pb.last_block = 0;
1723                         pb.last_db_block = -1;
1724                         pb.num_illegal_blocks = 0;
1725                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1726                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1727                         pb.inode = inode;
1728                         pb.pctx = &pctx;
1729                         pb.ctx = ctx;
1730                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1731                                      block_buf, process_bad_block, &pb);
1732                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1733                         if (pctx.errcode) {
1734                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1735                                 ctx->flags |= E2F_FLAG_ABORT;
1736                                 goto endit;
1737                         }
1738                         if (pb.bbcheck)
1739                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1740                                 ctx->flags |= E2F_FLAG_ABORT;
1741                                 goto endit;
1742                         }
1743                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1744                         clear_problem_context(&pctx);
1745                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1746                         continue;
1747                 } else if (ino == EXT2_ROOT_INO) {
1748                         /*
1749                          * Make sure the root inode is a directory; if
1750                          * not, offer to clear it.  It will be
1751                          * regenerated in pass #3.
1752                          */
1753                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1754                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1755                                         goto clear_inode;
1756                         }
1757                         /*
1758                          * If dtime is set, offer to clear it.  mke2fs
1759                          * version 0.2b created filesystems with the
1760                          * dtime field set for the root and lost+found
1761                          * directories.  We won't worry about
1762                          * /lost+found, since that can be regenerated
1763                          * easily.  But we will fix the root directory
1764                          * as a special case.
1765                          */
1766                         if (inode->i_dtime && inode->i_links_count) {
1767                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1768                                         inode->i_dtime = 0;
1769                                         e2fsck_write_inode(ctx, ino, inode,
1770                                                            "pass1");
1771                                         failed_csum = 0;
1772                                 }
1773                         }
1774                 } else if (ino == EXT2_JOURNAL_INO) {
1775                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1776                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1777                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1778                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1779                                                 &pctx)) {
1780                                         inode->i_mode = LINUX_S_IFREG;
1781                                         e2fsck_write_inode(ctx, ino, inode,
1782                                                            "pass1");
1783                                         failed_csum = 0;
1784                                 }
1785                                 check_blocks(ctx, &pctx, block_buf, NULL);
1786                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1787                                 continue;
1788                         }
1789                         if ((inode->i_links_count ||
1790                              inode->i_blocks || inode->i_block[0]) &&
1791                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1792                                         &pctx)) {
1793                                 memset(inode, 0, inode_size);
1794                                 ext2fs_icount_store(ctx->inode_link_info,
1795                                                     ino, 0);
1796                                 e2fsck_write_inode_full(ctx, ino, inode,
1797                                                         inode_size, "pass1");
1798                                 failed_csum = 0;
1799                         }
1800                 } else if (quota_inum_is_reserved(fs, ino)) {
1801                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1802                         if (ext2fs_has_feature_quota(fs->super) &&
1803                             quota_inum_is_super(fs->super, ino)) {
1804                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1805                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1806                                                         &pctx)) {
1807                                         inode->i_mode = LINUX_S_IFREG;
1808                                         e2fsck_write_inode(ctx, ino, inode,
1809                                                         "pass1");
1810                                         failed_csum = 0;
1811                                 }
1812                                 check_blocks(ctx, &pctx, block_buf, NULL);
1813                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1814                                 continue;
1815                         }
1816                         if ((inode->i_links_count ||
1817                              inode->i_blocks || inode->i_block[0]) &&
1818                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1819                                         &pctx)) {
1820                                 memset(inode, 0, inode_size);
1821                                 ext2fs_icount_store(ctx->inode_link_info,
1822                                                     ino, 0);
1823                                 e2fsck_write_inode_full(ctx, ino, inode,
1824                                                         inode_size, "pass1");
1825                                 failed_csum = 0;
1826                         }
1827                 } else if (ino == fs->super->s_orphan_file_inum) {
1828                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1829                         if (ext2fs_has_feature_orphan_file(fs->super)) {
1830                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1831                                     fix_problem(ctx, PR_1_ORPHAN_FILE_BAD_MODE,
1832                                                 &pctx)) {
1833                                         inode->i_mode = LINUX_S_IFREG;
1834                                         e2fsck_write_inode(ctx, ino, inode,
1835                                                            "pass1");
1836                                         failed_csum = 0;
1837                                 }
1838                                 check_blocks(ctx, &pctx, block_buf, NULL);
1839                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1840                                 continue;
1841                         }
1842                         if ((inode->i_links_count ||
1843                              inode->i_blocks || inode->i_block[0]) &&
1844                             fix_problem(ctx, PR_1_ORPHAN_FILE_NOT_CLEAR,
1845                                         &pctx)) {
1846                                 memset(inode, 0, inode_size);
1847                                 ext2fs_icount_store(ctx->inode_link_info, ino,
1848                                                     0);
1849                                 e2fsck_write_inode_full(ctx, ino, inode,
1850                                                         inode_size, "pass1");
1851                                 failed_csum = 0;
1852                         }
1853                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1854                         problem_t problem = 0;
1855
1856                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1857                         if (ino == EXT2_BOOT_LOADER_INO) {
1858                                 if (LINUX_S_ISDIR(inode->i_mode))
1859                                         problem = PR_1_RESERVED_BAD_MODE;
1860                         } else if (ino == EXT2_RESIZE_INO) {
1861                                 if (inode->i_mode &&
1862                                     !LINUX_S_ISREG(inode->i_mode))
1863                                         problem = PR_1_RESERVED_BAD_MODE;
1864                         } else {
1865                                 if (inode->i_mode != 0)
1866                                         problem = PR_1_RESERVED_BAD_MODE;
1867                         }
1868                         if (problem) {
1869                                 if (fix_problem(ctx, problem, &pctx)) {
1870                                         inode->i_mode = 0;
1871                                         e2fsck_write_inode(ctx, ino, inode,
1872                                                            "pass1");
1873                                         failed_csum = 0;
1874                                 }
1875                         }
1876                         check_blocks(ctx, &pctx, block_buf, NULL);
1877                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1878                         continue;
1879                 }
1880
1881                 if (!inode->i_links_count) {
1882                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1883                         continue;
1884                 }
1885                 /*
1886                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1887                  * deleted files.  Oops.
1888                  *
1889                  * Since all new ext2 implementations get this right,
1890                  * we now assume that the case of non-zero
1891                  * i_links_count and non-zero dtime means that we
1892                  * should keep the file, not delete it.
1893                  *
1894                  */
1895                 if (inode->i_dtime) {
1896                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1897                                 inode->i_dtime = 0;
1898                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1899                                 failed_csum = 0;
1900                         }
1901                 }
1902
1903                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1904                 switch (fs->super->s_creator_os) {
1905                     case EXT2_OS_HURD:
1906                         frag = inode->osd2.hurd2.h_i_frag;
1907                         fsize = inode->osd2.hurd2.h_i_fsize;
1908                         break;
1909                     default:
1910                         frag = fsize = 0;
1911                 }
1912
1913                 if (inode->i_faddr || frag || fsize ||
1914                     (!ext2fs_has_feature_largedir(fs->super) &&
1915                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1916                         mark_inode_bad(ctx, ino);
1917                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1918                     !ext2fs_has_feature_64bit(fs->super) &&
1919                     inode->osd2.linux2.l_i_file_acl_high != 0)
1920                         mark_inode_bad(ctx, ino);
1921                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1922                     !ext2fs_has_feature_huge_file(fs->super) &&
1923                     (inode->osd2.linux2.l_i_blocks_hi != 0))
1924                         mark_inode_bad(ctx, ino);
1925                 if (inode->i_flags & EXT2_IMAGIC_FL) {
1926                         if (imagic_fs) {
1927                                 if (!ctx->inode_imagic_map)
1928                                         alloc_imagic_map(ctx);
1929                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1930                                                          ino);
1931                         } else {
1932                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1933                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
1934                                         e2fsck_write_inode(ctx, ino,
1935                                                            inode, "pass1");
1936                                         failed_csum = 0;
1937                                 }
1938                         }
1939                 }
1940
1941                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1942                 check_is_really_dir(ctx, &pctx, block_buf);
1943
1944                 /*
1945                  * ext2fs_inode_has_valid_blocks2 does not actually look
1946                  * at i_block[] values, so not endian-sensitive here.
1947                  */
1948                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1949                     LINUX_S_ISLNK(inode->i_mode) &&
1950                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1951                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1952                         inode->i_flags &= ~EXT4_EXTENTS_FL;
1953                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1954                         failed_csum = 0;
1955                 }
1956
1957                 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
1958                     add_encrypted_file(ctx, &pctx) < 0)
1959                         goto clear_inode;
1960
1961                 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
1962                         ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
1963
1964                 if (LINUX_S_ISDIR(inode->i_mode)) {
1965                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1966                         e2fsck_add_dir_info(ctx, ino, 0);
1967                         ctx->fs_directory_count++;
1968                         if (inode->i_flags & EXT4_CASEFOLD_FL)
1969                                 add_casefolded_dir(ctx, ino);
1970                 } else if (LINUX_S_ISREG (inode->i_mode)) {
1971                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1972                         ctx->fs_regular_count++;
1973                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1974                            e2fsck_pass1_check_device_inode(fs, inode)) {
1975                         check_extents_inlinedata(ctx, &pctx);
1976                         check_immutable(ctx, &pctx);
1977                         check_size(ctx, &pctx);
1978                         ctx->fs_chardev_count++;
1979                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1980                            e2fsck_pass1_check_device_inode(fs, inode)) {
1981                         check_extents_inlinedata(ctx, &pctx);
1982                         check_immutable(ctx, &pctx);
1983                         check_size(ctx, &pctx);
1984                         ctx->fs_blockdev_count++;
1985                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1986                            e2fsck_pass1_check_symlink(fs, ino, inode,
1987                                                       block_buf)) {
1988                         check_immutable(ctx, &pctx);
1989                         ctx->fs_symlinks_count++;
1990                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1991                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1992                                 continue;
1993                         } else if (ext2fs_is_fast_symlink(inode)) {
1994                                 ctx->fs_fast_symlinks_count++;
1995                                 check_blocks(ctx, &pctx, block_buf,
1996                                              &ea_ibody_quota);
1997                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1998                                 continue;
1999                         }
2000                 }
2001                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
2002                          e2fsck_pass1_check_device_inode(fs, inode)) {
2003                         check_extents_inlinedata(ctx, &pctx);
2004                         check_immutable(ctx, &pctx);
2005                         check_size(ctx, &pctx);
2006                         ctx->fs_fifo_count++;
2007                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
2008                            e2fsck_pass1_check_device_inode(fs, inode)) {
2009                         check_extents_inlinedata(ctx, &pctx);
2010                         check_immutable(ctx, &pctx);
2011                         check_size(ctx, &pctx);
2012                         ctx->fs_sockets_count++;
2013                 } else
2014                         mark_inode_bad(ctx, ino);
2015                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2016                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
2017                         if (inode->i_block[EXT2_IND_BLOCK])
2018                                 ctx->fs_ind_count++;
2019                         if (inode->i_block[EXT2_DIND_BLOCK])
2020                                 ctx->fs_dind_count++;
2021                         if (inode->i_block[EXT2_TIND_BLOCK])
2022                                 ctx->fs_tind_count++;
2023                 }
2024                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2025                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
2026                     (inode->i_block[EXT2_IND_BLOCK] ||
2027                      inode->i_block[EXT2_DIND_BLOCK] ||
2028                      inode->i_block[EXT2_TIND_BLOCK] ||
2029                      ext2fs_file_acl_block(fs, inode))) {
2030                         struct process_inode_block *itp;
2031
2032                         itp = &inodes_to_process[process_inode_count];
2033                         itp->ino = ino;
2034                         itp->ea_ibody_quota = ea_ibody_quota;
2035                         if (inode_size < sizeof(struct ext2_inode_large))
2036                                 memcpy(&itp->inode, inode, inode_size);
2037                         else
2038                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
2039                         process_inode_count++;
2040                 } else
2041                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
2042
2043                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2044
2045                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2046                         goto endit;
2047
2048                 if (process_inode_count >= ctx->process_inode_size) {
2049                         process_inodes(ctx, block_buf, inodes_to_process,
2050                                        &process_inode_count);
2051
2052                         if (e2fsck_should_abort(ctx))
2053                                 goto endit;
2054                 }
2055         }
2056         process_inodes(ctx, block_buf, inodes_to_process,
2057                        &process_inode_count);
2058         ext2fs_close_inode_scan(scan);
2059         scan = NULL;
2060
2061         reserve_block_for_root_repair(ctx);
2062         reserve_block_for_lnf_repair(ctx);
2063
2064         /*
2065          * If any extended attribute blocks' reference counts need to
2066          * be adjusted, either up (ctx->refcount_extra), or down
2067          * (ctx->refcount), then fix them.
2068          */
2069         if (ctx->refcount) {
2070                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
2071                 ea_refcount_free(ctx->refcount);
2072                 ctx->refcount = 0;
2073         }
2074         if (ctx->refcount_extra) {
2075                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
2076                                         block_buf, +1);
2077                 ea_refcount_free(ctx->refcount_extra);
2078                 ctx->refcount_extra = 0;
2079         }
2080
2081         if (ctx->ea_block_quota_blocks) {
2082                 ea_refcount_free(ctx->ea_block_quota_blocks);
2083                 ctx->ea_block_quota_blocks = 0;
2084         }
2085
2086         if (ctx->ea_block_quota_inodes) {
2087                 ea_refcount_free(ctx->ea_block_quota_inodes);
2088                 ctx->ea_block_quota_inodes = 0;
2089         }
2090
2091         if (ctx->invalid_bitmaps)
2092                 handle_fs_bad_blocks(ctx);
2093
2094         /* We don't need the block_ea_map any more */
2095         if (ctx->block_ea_map) {
2096                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2097                 ctx->block_ea_map = 0;
2098         }
2099
2100         /* We don't need the encryption policy => ID map any more */
2101         destroy_encryption_policy_map(ctx);
2102
2103         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2104                 clear_problem_context(&pctx);
2105                 pctx.errcode = ext2fs_create_resize_inode(fs);
2106                 if (pctx.errcode) {
2107                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2108                                          &pctx)) {
2109                                 ctx->flags |= E2F_FLAG_ABORT;
2110                                 goto endit;
2111                         }
2112                         pctx.errcode = 0;
2113                 }
2114                 if (!pctx.errcode) {
2115                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2116                                           "recreate inode");
2117                         inode->i_mtime = ctx->now;
2118                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2119                                            "recreate inode");
2120                 }
2121                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2122         }
2123
2124         if (ctx->flags & E2F_FLAG_RESTART) {
2125                 /*
2126                  * Only the master copy of the superblock and block
2127                  * group descriptors are going to be written during a
2128                  * restart, so set the superblock to be used to be the
2129                  * master superblock.
2130                  */
2131                 ctx->use_superblock = 0;
2132                 goto endit;
2133         }
2134
2135         if (ctx->large_dirs && !ext2fs_has_feature_largedir(fs->super)) {
2136                 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2137                         ext2fs_set_feature_largedir(fs->super);
2138                         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2139                         ext2fs_mark_super_dirty(fs);
2140                 }
2141                 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2142                     fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2143                         ext2fs_update_dynamic_rev(fs);
2144                         ext2fs_mark_super_dirty(fs);
2145                 }
2146         }
2147
2148         if (ctx->block_dup_map) {
2149                 if (ctx->options & E2F_OPT_PREEN) {
2150                         clear_problem_context(&pctx);
2151                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2152                 }
2153                 e2fsck_pass1_dupblocks(ctx, block_buf);
2154         }
2155         ctx->flags |= E2F_FLAG_ALLOC_OK;
2156 endit:
2157         e2fsck_use_inode_shortcuts(ctx, 0);
2158         ext2fs_free_mem(&inodes_to_process);
2159         inodes_to_process = 0;
2160
2161         if (scan)
2162                 ext2fs_close_inode_scan(scan);
2163         if (block_buf)
2164                 ext2fs_free_mem(&block_buf);
2165         if (inode)
2166                 ext2fs_free_mem(&inode);
2167
2168         /*
2169          * The l+f inode may have been cleared, so zap it now and
2170          * later passes will recalculate it if necessary
2171          */
2172         ctx->lost_and_found = 0;
2173
2174         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2175                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2176         else
2177                 ctx->invalid_bitmaps++;
2178 }
2179
2180 static void init_ext2_max_sizes()
2181 {
2182         int     i;
2183         __u64   max_sizes;
2184
2185         /*
2186          * Init ext2_max_sizes which will be immutable and shared between
2187          * threads
2188          */
2189 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
2190
2191         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
2192                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
2193                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
2194                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
2195                 max_sizes = (max_sizes * (1UL << i));
2196                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
2197         }
2198 #undef EXT2_BPP
2199 }
2200
2201 #ifdef HAVE_PTHREAD
2202 static errcode_t e2fsck_pass1_copy_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2203                                           ext2fs_generic_bitmap *dest)
2204 {
2205         errcode_t ret;
2206
2207         ret = ext2fs_copy_bitmap(*src, dest);
2208         if (ret)
2209                 return ret;
2210
2211         (*dest)->fs = fs;
2212
2213         return 0;
2214 }
2215
2216 static void e2fsck_pass1_free_bitmap(ext2fs_generic_bitmap *bitmap)
2217 {
2218         if (*bitmap) {
2219                 ext2fs_free_generic_bmap(*bitmap);
2220                 *bitmap = NULL;
2221         }
2222
2223 }
2224
2225 static errcode_t e2fsck_pass1_merge_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2226                                           ext2fs_generic_bitmap *dest)
2227 {
2228         errcode_t ret = 0;
2229
2230         if (*src) {
2231                 if (*dest == NULL) {
2232                         *dest = *src;
2233                         *src = NULL;
2234                 } else {
2235                         ret = ext2fs_merge_bitmap(*src, *dest, NULL, NULL);
2236                         if (ret)
2237                                 return ret;
2238                 }
2239                 (*dest)->fs = fs;
2240         }
2241
2242         return 0;
2243 }
2244
2245 static errcode_t e2fsck_pass1_copy_fs(ext2_filsys dest, e2fsck_t src_context,
2246                                       ext2_filsys src)
2247 {
2248         errcode_t       retval;
2249
2250         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2251         dest->inode_map = NULL;
2252         dest->block_map = NULL;
2253         dest->badblocks = NULL;
2254         if (dest->dblist)
2255                 dest->dblist->fs = dest;
2256         if (src->block_map) {
2257                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2258                                                   &dest->block_map);
2259                 if (retval)
2260                         return retval;
2261         }
2262         if (src->inode_map) {
2263                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2264                                                   &dest->inode_map);
2265                 if (retval)
2266                         return retval;
2267         }
2268
2269         if (src->badblocks) {
2270                 retval = ext2fs_badblocks_copy(src->badblocks,
2271                                                &dest->badblocks);
2272                 if (retval)
2273                         return retval;
2274         }
2275
2276         /* disable it for now */
2277         src_context->openfs_flags &= ~EXT2_FLAG_EXCLUSIVE;
2278         retval = ext2fs_open_channel(dest, src_context->io_options,
2279                                      src_context->io_manager,
2280                                      src_context->openfs_flags,
2281                                      src->io->block_size);
2282         if (retval)
2283                 return retval;
2284
2285         /* Block size might not be default */
2286         io_channel_set_blksize(dest->io, src->io->block_size);
2287         ehandler_init(dest->io);
2288
2289         assert(dest->io->magic == src->io->magic);
2290         assert(dest->io->manager == src->io->manager);
2291         assert(strcmp(dest->io->name, src->io->name) == 0);
2292         assert(dest->io->block_size == src->io->block_size);
2293         assert(dest->io->read_error == src->io->read_error);
2294         assert(dest->io->write_error == src->io->write_error);
2295         assert(dest->io->refcount == src->io->refcount);
2296         assert(dest->io->flags == src->io->flags);
2297         assert(dest->io->app_data == dest);
2298         assert(src->io->app_data == src);
2299         assert(dest->io->align == src->io->align);
2300
2301         /* The data should be written to disk immediately */
2302         dest->io->flags |= CHANNEL_FLAGS_WRITETHROUGH;
2303         /* icache will be rebuilt if needed, so do not copy from @src */
2304         src->icache = NULL;
2305         return 0;
2306 }
2307
2308 static int e2fsck_pass1_merge_fs(ext2_filsys dest, ext2_filsys src)
2309 {
2310         struct ext2_inode_cache *icache = dest->icache;
2311         errcode_t retval = 0;
2312         io_channel dest_io;
2313         io_channel dest_image_io;
2314         ext2fs_inode_bitmap inode_map;
2315         ext2fs_block_bitmap block_map;
2316         ext2_badblocks_list badblocks;
2317         ext2_dblist dblist;
2318
2319         dest_io = dest->io;
2320         dest_image_io = dest->image_io;
2321         inode_map = dest->inode_map;
2322         block_map = dest->block_map;
2323         badblocks = dest->badblocks;
2324         dblist = dest->dblist;
2325
2326         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2327         dest->io = dest_io;
2328         dest->image_io = dest_image_io;
2329         dest->icache = icache;
2330         dest->inode_map = inode_map;
2331         dest->block_map = block_map;
2332         dest->badblocks = badblocks;
2333         dest->dblist = dblist;
2334         if (dest->dblist)
2335                 dest->dblist->fs = dest;
2336
2337         if (src->icache) {
2338                 ext2fs_free_inode_cache(src->icache);
2339                 src->icache = NULL;
2340         }
2341
2342         retval = e2fsck_pass1_merge_bitmap(dest, &src->inode_map,
2343                                            &dest->inode_map);
2344         if (retval)
2345                 goto out;
2346
2347         retval = e2fsck_pass1_merge_bitmap(dest, &src->block_map,
2348                                           &dest->block_map);
2349         if (retval)
2350                 goto out;
2351
2352         if (src->dblist) {
2353                 if (dest->dblist) {
2354                         retval = ext2fs_merge_dblist(src->dblist,
2355                                                      dest->dblist);
2356                         if (retval)
2357                                 goto out;
2358                 } else {
2359                         dest->dblist = src->dblist;
2360                         dest->dblist->fs = dest;
2361                         src->dblist = NULL;
2362                 }
2363         }
2364
2365         if (src->badblocks) {
2366                 if (dest->badblocks == NULL)
2367                         retval = ext2fs_badblocks_copy(src->badblocks,
2368                                                        &dest->badblocks);
2369                 else
2370                         retval = ext2fs_badblocks_merge(src->badblocks,
2371                                                         dest->badblocks);
2372         }
2373 out:
2374         io_channel_close(src->io);
2375         if (src->inode_map)
2376                 ext2fs_free_generic_bmap(src->inode_map);
2377         if (src->block_map)
2378                 ext2fs_free_generic_bmap(src->block_map);
2379         if (src->badblocks)
2380                 ext2fs_badblocks_list_free(src->badblocks);
2381         if (src->dblist)
2382                 ext2fs_free_dblist(src->dblist);
2383
2384         return retval;
2385 }
2386
2387 static errcode_t e2fsck_pass1_thread_prepare(e2fsck_t global_ctx, e2fsck_t *thread_ctx,
2388                                              int thread_index, int num_threads)
2389 {
2390         errcode_t               retval;
2391         e2fsck_t                thread_context;
2392         ext2_filsys             thread_fs;
2393         ext2_filsys             global_fs = global_ctx->fs;
2394         struct e2fsck_thread    *tinfo;
2395         dgrp_t                  average_group;
2396
2397         assert(global_ctx->inode_used_map == NULL);
2398         assert(global_ctx->inode_dir_map == NULL);
2399         assert(global_ctx->inode_bb_map == NULL);
2400         assert(global_ctx->inode_imagic_map == NULL);
2401         assert(global_ctx->inode_reg_map == NULL);
2402         assert(global_ctx->inodes_to_rebuild == NULL);
2403
2404         assert(global_ctx->block_found_map == NULL);
2405         assert(global_ctx->block_dup_map == NULL);
2406         assert(global_ctx->block_ea_map == NULL);
2407         assert(global_ctx->block_metadata_map == NULL);
2408         assert(global_ctx->fs->dblist == NULL);
2409
2410         retval = ext2fs_get_mem(sizeof(struct e2fsck_struct), &thread_context);
2411         if (retval) {
2412                 com_err(global_ctx->program_name, retval, "while allocating memory");
2413                 return retval;
2414         }
2415         memcpy(thread_context, global_ctx, sizeof(struct e2fsck_struct));
2416         thread_context->global_ctx = global_ctx;
2417
2418         retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &thread_fs);
2419         if (retval) {
2420                 com_err(global_ctx->program_name, retval, "while allocating memory");
2421                 goto out_context;
2422         }
2423
2424         io_channel_flush_cleanup(global_fs->io);
2425         retval = e2fsck_pass1_copy_fs(thread_fs, global_ctx, global_fs);
2426         if (retval) {
2427                 com_err(global_ctx->program_name, retval, "while copying fs");
2428                 goto out_fs;
2429         }
2430         thread_fs->priv_data = thread_context;
2431
2432         thread_context->thread_info.et_thread_index = thread_index;
2433         set_up_logging(thread_context);
2434
2435         /*
2436          * Distribute work to multiple threads:
2437          * Each thread work on fs->group_desc_count / nthread groups.
2438          */
2439         tinfo = &thread_context->thread_info;
2440         average_group = thread_fs->group_desc_count / num_threads;
2441         if (average_group == 0)
2442                 average_group = 1;
2443         tinfo->et_group_start = average_group * thread_index;
2444         if (thread_index == num_threads - 1)
2445                 tinfo->et_group_end = thread_fs->group_desc_count;
2446         else
2447                 tinfo->et_group_end = average_group * (thread_index + 1);
2448         tinfo->et_group_next = tinfo->et_group_start;
2449         tinfo->et_inode_number = 0;
2450         tinfo->et_log_buf[0] = '\0';
2451         tinfo->et_log_length = 0;
2452         if (thread_context->options & E2F_OPT_MULTITHREAD)
2453                 log_out(thread_context, _("Scan group range [%d, %d)\n"),
2454                         tinfo->et_group_start, tinfo->et_group_end);
2455         thread_context->fs = thread_fs;
2456         *thread_ctx = thread_context;
2457         return 0;
2458 out_fs:
2459         ext2fs_free_mem(&thread_fs);
2460 out_context:
2461         ext2fs_free_mem(&thread_context);
2462         return retval;
2463 }
2464
2465 static void e2fsck_pass1_merge_dir_info(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2466 {
2467         if (thread_ctx->dir_info == NULL)
2468                 return;
2469
2470         if (global_ctx->dir_info == NULL) {
2471                 global_ctx->dir_info = thread_ctx->dir_info;
2472                 thread_ctx->dir_info = NULL;
2473                 return;
2474         }
2475
2476         e2fsck_merge_dir_info(global_ctx, thread_ctx->dir_info,
2477                               global_ctx->dir_info);
2478 }
2479
2480 static inline errcode_t
2481 e2fsck_pass1_merge_icount(ext2_icount_t *dest_icount,
2482                           ext2_icount_t *src_icount)
2483 {
2484         if (*src_icount) {
2485                 if (*dest_icount == NULL) {
2486                         *dest_icount = *src_icount;
2487                         *src_icount = NULL;
2488                 } else {
2489                         errcode_t ret;
2490
2491                         ret = ext2fs_icount_merge(*src_icount,
2492                                                   *dest_icount);
2493                         if (ret)
2494                                 return ret;
2495                 }
2496         }
2497
2498         return 0;
2499 }
2500
2501 static errcode_t e2fsck_pass1_merge_icounts(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2502 {
2503         errcode_t ret;
2504
2505         ret = e2fsck_pass1_merge_icount(&global_ctx->inode_count,
2506                                         &thread_ctx->inode_count);
2507         if (ret)
2508                 return ret;
2509         ret = e2fsck_pass1_merge_icount(&global_ctx->inode_link_info,
2510                                         &thread_ctx->inode_link_info);
2511
2512         return ret;
2513 }
2514
2515 static int e2fsck_pass1_thread_join_one(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2516 {
2517         errcode_t        retval;
2518         int              flags = global_ctx->flags;
2519         ext2_filsys      thread_fs = thread_ctx->fs;
2520         ext2_filsys      global_fs = global_ctx->fs;
2521         FILE            *global_logf = global_ctx->logf;
2522         FILE            *global_problem_logf = global_ctx->problem_logf;
2523         ext2fs_inode_bitmap inode_bad_map = global_ctx->inode_bad_map;
2524         struct dir_info_db *dir_info = global_ctx->dir_info;
2525         ext2fs_inode_bitmap inode_used_map = global_ctx->inode_used_map;
2526         ext2fs_inode_bitmap inode_dir_map = global_ctx->inode_dir_map;
2527         ext2fs_inode_bitmap inode_bb_map = global_ctx->inode_bb_map;
2528         ext2fs_inode_bitmap inode_imagic_map = global_ctx->inode_imagic_map;
2529         ext2fs_inode_bitmap inode_reg_map = global_ctx->inode_reg_map;
2530         ext2fs_block_bitmap block_found_map = global_ctx->block_found_map;
2531         ext2fs_block_bitmap block_dup_map = global_ctx->block_dup_map;
2532         ext2fs_block_bitmap block_ea_map = global_ctx->block_ea_map;
2533         ext2fs_block_bitmap block_metadata_map = global_ctx->block_metadata_map;
2534         ext2fs_block_bitmap inodes_to_rebuild = global_ctx->inodes_to_rebuild;
2535         ext2_icount_t inode_count = global_ctx->inode_count;
2536         ext2_icount_t inode_link_info = global_ctx->inode_link_info;
2537
2538 #ifdef HAVE_SETJMP_H
2539         jmp_buf          old_jmp;
2540
2541         memcpy(old_jmp, global_ctx->abort_loc, sizeof(jmp_buf));
2542 #endif
2543         memcpy(global_ctx, thread_ctx, sizeof(struct e2fsck_struct));
2544 #ifdef HAVE_SETJMP_H
2545         memcpy(global_ctx->abort_loc, old_jmp, sizeof(jmp_buf));
2546 #endif
2547
2548         global_ctx->inode_used_map = inode_used_map;
2549         global_ctx->inode_bad_map = inode_bad_map;
2550         global_ctx->inode_dir_map = inode_dir_map;
2551         global_ctx->inode_bb_map = inode_bb_map;
2552         global_ctx->inode_imagic_map = inode_imagic_map;
2553         global_ctx->inodes_to_rebuild = inodes_to_rebuild;
2554         global_ctx->inode_reg_map = inode_reg_map;
2555         global_ctx->block_found_map = block_found_map;
2556         global_ctx->block_dup_map = block_dup_map;
2557         global_ctx->block_ea_map = block_ea_map;
2558         global_ctx->block_metadata_map = block_metadata_map;
2559         global_ctx->dir_info = dir_info;
2560         e2fsck_pass1_merge_dir_info(global_ctx, thread_ctx);
2561         global_ctx->inode_count = inode_count;
2562         global_ctx->inode_link_info = inode_link_info;
2563
2564         /* Keep the global singal flags*/
2565         global_ctx->flags |= (flags & E2F_FLAG_SIGNAL_MASK) |
2566                              (global_ctx->flags & E2F_FLAG_SIGNAL_MASK);
2567
2568         retval = e2fsck_pass1_merge_fs(global_fs, thread_fs);
2569         if (retval) {
2570                 com_err(global_ctx->program_name, 0, _("while merging fs\n"));
2571                 return retval;
2572         }
2573         global_fs->priv_data = global_ctx;
2574         global_ctx->fs = global_fs;
2575         global_ctx->logf = global_logf;
2576         global_ctx->problem_logf = global_problem_logf;
2577         global_ctx->global_ctx = NULL;
2578         retval = e2fsck_pass1_merge_icounts(global_ctx, thread_ctx);
2579         if (retval) {
2580                 com_err(global_ctx->program_name, 0,
2581                         _("while merging icounts\n"));
2582                 return retval;
2583         }
2584
2585         retval = e2fsck_pass1_merge_bitmap(global_fs,
2586                                 &thread_ctx->inode_used_map,
2587                                 &global_ctx->inode_used_map);
2588         if (retval)
2589                 return retval;
2590
2591         retval = e2fsck_pass1_merge_bitmap(global_fs,
2592                                 &thread_ctx->inode_bad_map,
2593                                 &global_ctx->inode_bad_map);
2594         if (retval)
2595                 return retval;
2596         retval = e2fsck_pass1_merge_bitmap(global_fs,
2597                                         &thread_ctx->inode_dir_map,
2598                                         &global_ctx->inode_dir_map);
2599         if (retval)
2600                 return retval;
2601         retval = e2fsck_pass1_merge_bitmap(global_fs,
2602                                 &thread_ctx->inode_bb_map,
2603                                 &global_ctx->inode_bb_map);
2604         if (retval)
2605                 return retval;
2606         retval = e2fsck_pass1_merge_bitmap(global_fs,
2607                                 &thread_ctx->inode_imagic_map,
2608                                 &global_ctx->inode_imagic_map);
2609         if (retval)
2610                 return retval;
2611         retval = e2fsck_pass1_merge_bitmap(global_fs,
2612                                 &thread_ctx->inode_reg_map,
2613                                 &global_ctx->inode_reg_map);
2614         if (retval)
2615                 return retval;
2616         retval = e2fsck_pass1_merge_bitmap(global_fs,
2617                                 &thread_ctx->inodes_to_rebuild,
2618                                 &global_ctx->inodes_to_rebuild);
2619         if (retval)
2620                 return retval;
2621         retval = e2fsck_pass1_merge_bitmap(global_fs,
2622                                 &thread_ctx->block_found_map,
2623                                 &global_ctx->block_found_map);
2624         if (retval)
2625                 return retval;
2626         retval = e2fsck_pass1_merge_bitmap(global_fs,
2627                                 &thread_ctx->block_dup_map,
2628                                 &global_ctx->block_dup_map);
2629         if (retval)
2630                 return retval;
2631         retval = e2fsck_pass1_merge_bitmap(global_fs,
2632                                 &thread_ctx->block_ea_map,
2633                                 &global_ctx->block_ea_map);
2634         if (retval)
2635                 return retval;
2636         retval = e2fsck_pass1_merge_bitmap(global_fs,
2637                                 &thread_ctx->block_metadata_map,
2638                                 &global_ctx->block_metadata_map);
2639         if (retval)
2640                 return retval;
2641
2642         return 0;
2643 }
2644
2645 static int e2fsck_pass1_thread_join(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2646 {
2647         errcode_t       retval;
2648
2649         retval = e2fsck_pass1_thread_join_one(global_ctx, thread_ctx);
2650         ext2fs_free_mem(&thread_ctx->fs);
2651         if (thread_ctx->logf)
2652                 fclose(thread_ctx->logf);
2653         if (thread_ctx->problem_logf) {
2654                 fputs("</problem_log>\n", thread_ctx->problem_logf);
2655                 fclose(thread_ctx->problem_logf);
2656         }
2657         e2fsck_pass1_free_bitmap(&thread_ctx->inode_used_map);
2658         e2fsck_pass1_free_bitmap(&thread_ctx->inode_bad_map);
2659         e2fsck_pass1_free_bitmap(&thread_ctx->inode_dir_map);
2660         e2fsck_pass1_free_bitmap(&thread_ctx->inode_bb_map);
2661         e2fsck_pass1_free_bitmap(&thread_ctx->inode_imagic_map);
2662         e2fsck_pass1_free_bitmap(&thread_ctx->inode_reg_map);
2663         e2fsck_pass1_free_bitmap(&thread_ctx->inodes_to_rebuild);
2664         e2fsck_pass1_free_bitmap(&thread_ctx->block_found_map);
2665         e2fsck_pass1_free_bitmap(&thread_ctx->block_dup_map);
2666         e2fsck_pass1_free_bitmap(&thread_ctx->block_ea_map);
2667         e2fsck_pass1_free_bitmap(&thread_ctx->block_metadata_map);
2668         e2fsck_free_dir_info(thread_ctx);
2669         ext2fs_free_icount(thread_ctx->inode_count);
2670         ext2fs_free_icount(thread_ctx->inode_link_info);
2671         ext2fs_free_mem(&thread_ctx);
2672
2673         return retval;
2674 }
2675
2676 static int e2fsck_pass1_threads_join(struct e2fsck_thread_info *infos,
2677                                       int num_threads, e2fsck_t global_ctx)
2678 {
2679         errcode_t                        rc;
2680         errcode_t                        ret = 0;
2681         int                              i;
2682         struct e2fsck_thread_info       *pinfo;
2683
2684         for (i = 0; i < num_threads; i++) {
2685                 pinfo = &infos[i];
2686
2687                 if (!pinfo->eti_started)
2688                         continue;
2689
2690                 rc = pthread_join(pinfo->eti_thread_id, NULL);
2691                 if (rc) {
2692                         com_err(global_ctx->program_name, rc,
2693                                 _("while joining thread\n"));
2694                         if (ret == 0)
2695                                 ret = rc;
2696                 }
2697                 rc = e2fsck_pass1_thread_join(global_ctx, infos[i].eti_thread_ctx);
2698                 if (rc) {
2699                         com_err(global_ctx->program_name, rc,
2700                                 _("while joining pass1 thread\n"));
2701                         if (ret == 0)
2702                                 ret = rc;
2703                 }
2704         }
2705         free(infos);
2706
2707         return ret;
2708 }
2709
2710 static void *e2fsck_pass1_thread(void *arg)
2711 {
2712         struct e2fsck_thread_info       *info = arg;
2713         e2fsck_t                         thread_ctx = info->eti_thread_ctx;
2714 #ifdef DEBUG_THREADS
2715         struct e2fsck_thread_debug      *thread_debug = info->eti_debug;
2716 #endif
2717
2718 #ifdef DEBUG_THREADS
2719         pthread_mutex_lock(&thread_debug->etd_mutex);
2720         while (info->eti_thread_index > thread_debug->etd_finished_threads) {
2721                 pthread_cond_wait(&thread_debug->etd_cond,
2722                                   &thread_debug->etd_mutex);
2723         }
2724         pthread_mutex_unlock(&thread_debug->etd_mutex);
2725 #endif
2726
2727 #ifdef HAVE_SETJMP_H
2728         /*
2729          * When fatal_error() happens, jump to here. The thread
2730          * context's flags will be saved, but its abort_loc will
2731          * be overwritten by original jump buffer for the later
2732          * tests.
2733          */
2734         if (setjmp(thread_ctx->abort_loc)) {
2735                 thread_ctx->flags &= ~E2F_FLAG_SETJMP_OK;
2736                 goto out;
2737         }
2738         thread_ctx->flags |= E2F_FLAG_SETJMP_OK;
2739 #endif
2740
2741         e2fsck_pass1_run(thread_ctx);
2742
2743 out:
2744         if (thread_ctx->options & E2F_OPT_MULTITHREAD)
2745                 log_out(thread_ctx,
2746                         _("Scanned group range [%lu, %lu), inodes %lu\n"),
2747                         thread_ctx->thread_info.et_group_start,
2748                         thread_ctx->thread_info.et_group_end,
2749                         thread_ctx->thread_info.et_inode_number);
2750
2751 #ifdef DEBUG_THREADS
2752         pthread_mutex_lock(&thread_debug->etd_mutex);
2753         thread_debug->etd_finished_threads++;
2754         pthread_cond_broadcast(&thread_debug->etd_cond);
2755         pthread_mutex_unlock(&thread_debug->etd_mutex);
2756 #endif
2757
2758         return NULL;
2759 }
2760
2761 static int e2fsck_pass1_threads_start(struct e2fsck_thread_info **pinfo,
2762                                       int num_threads, e2fsck_t global_ctx)
2763 {
2764         struct e2fsck_thread_info       *infos;
2765         pthread_attr_t                   attr;
2766         errcode_t                        retval;
2767         errcode_t                        ret;
2768         struct e2fsck_thread_info       *tmp_pinfo;
2769         int                              i;
2770         e2fsck_t                         thread_ctx;
2771 #ifdef DEBUG_THREADS
2772         struct e2fsck_thread_debug       thread_debug =
2773                 {PTHREAD_MUTEX_INITIALIZER, PTHREAD_COND_INITIALIZER, 0};
2774
2775         thread_debug.etd_finished_threads = 0;
2776 #endif
2777
2778         retval = pthread_attr_init(&attr);
2779         if (retval) {
2780                 com_err(global_ctx->program_name, retval,
2781                         _("while setting pthread attribute\n"));
2782                 return retval;
2783         }
2784
2785         infos = calloc(num_threads, sizeof(struct e2fsck_thread_info));
2786         if (infos == NULL) {
2787                 retval = -ENOMEM;
2788                 com_err(global_ctx->program_name, retval,
2789                         _("while allocating memory for threads\n"));
2790                 pthread_attr_destroy(&attr);
2791                 return retval;
2792         }
2793
2794         for (i = 0; i < num_threads; i++) {
2795                 tmp_pinfo = &infos[i];
2796                 tmp_pinfo->eti_thread_index = i;
2797 #ifdef DEBUG_THREADS
2798                 tmp_pinfo->eti_debug = &thread_debug;
2799 #endif
2800                 retval = e2fsck_pass1_thread_prepare(global_ctx, &thread_ctx,
2801                                                      i, num_threads);
2802                 if (retval) {
2803                         com_err(global_ctx->program_name, retval,
2804                                 _("while preparing pass1 thread\n"));
2805                         break;
2806                 }
2807                 tmp_pinfo->eti_thread_ctx = thread_ctx;
2808
2809                 retval = pthread_create(&tmp_pinfo->eti_thread_id, &attr,
2810                                         &e2fsck_pass1_thread, tmp_pinfo);
2811                 if (retval) {
2812                         com_err(global_ctx->program_name, retval,
2813                                 _("while creating thread\n"));
2814                         e2fsck_pass1_thread_join(global_ctx, thread_ctx);
2815                         break;
2816                 }
2817
2818                 tmp_pinfo->eti_started = 1;
2819         }
2820
2821         /* destroy the thread attribute object, since it is no longer needed */
2822         ret = pthread_attr_destroy(&attr);
2823         if (ret) {
2824                 com_err(global_ctx->program_name, ret,
2825                         _("while destroying thread attribute\n"));
2826                 if (retval == 0)
2827                         retval = ret;
2828         }
2829
2830         if (retval) {
2831                 e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
2832                 return retval;
2833         }
2834         *pinfo = infos;
2835         return 0;
2836 }
2837
2838 static void e2fsck_pass1_multithread(e2fsck_t global_ctx)
2839 {
2840         struct e2fsck_thread_info       *infos = NULL;
2841         int                              num_threads = 1;
2842         errcode_t                        retval;
2843
2844         retval = e2fsck_pass1_threads_start(&infos, num_threads, global_ctx);
2845         if (retval) {
2846                 com_err(global_ctx->program_name, retval,
2847                         _("while starting pass1 threads\n"));
2848                 goto out_abort;
2849         }
2850
2851         retval = e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
2852         if (retval) {
2853                 com_err(global_ctx->program_name, retval,
2854                         _("while joining pass1 threads\n"));
2855                 goto out_abort;
2856         }
2857         return;
2858 out_abort:
2859         global_ctx->flags |= E2F_FLAG_ABORT;
2860         return;
2861 }
2862 #endif
2863
2864 /* TODO: tdb needs to be handled properly for multiple threads*/
2865 static int multiple_threads_supported(e2fsck_t ctx)
2866 {
2867 #ifdef  CONFIG_TDB
2868         unsigned int            threshold;
2869         ext2_ino_t              num_dirs;
2870         errcode_t               retval;
2871         char                    *tdb_dir;
2872         int                     enable;
2873
2874         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
2875                            &tdb_dir);
2876         profile_get_uint(ctx->profile, "scratch_files",
2877                          "numdirs_threshold", 0, 0, &threshold);
2878         profile_get_boolean(ctx->profile, "scratch_files",
2879                             "icount", 0, 1, &enable);
2880
2881         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
2882         if (retval)
2883                 num_dirs = 1024;        /* Guess */
2884
2885         /* tdb is unsupported now */
2886         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
2887             (!threshold || num_dirs > threshold))
2888                 return 0;
2889  #endif
2890         return 1;
2891 }
2892
2893 void e2fsck_pass1(e2fsck_t ctx)
2894 {
2895
2896         init_ext2_max_sizes();
2897 #ifdef HAVE_PTHREAD
2898         if (multiple_threads_supported(ctx)) {
2899                 e2fsck_pass1_multithread(ctx);
2900                 return;
2901         }
2902         fprintf(stderr, "Fall through single thread for pass1 "
2903                         "because tdb could not handle properly\n");
2904 #endif
2905         e2fsck_pass1_run(ctx);
2906 }
2907
2908 #undef FINISH_INODE_LOOP
2909
2910 /*
2911  * When the inode_scan routines call this callback at the end of the
2912  * glock group, call process_inodes.
2913  */
2914 static errcode_t scan_callback(ext2_filsys fs,
2915                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
2916                                dgrp_t group, void * priv_data)
2917 {
2918         struct scan_callback_struct *scan_struct;
2919         e2fsck_t ctx;
2920         struct e2fsck_thread *tinfo;
2921
2922         scan_struct = (struct scan_callback_struct *) priv_data;
2923         ctx = scan_struct->ctx;
2924
2925         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf,
2926                        scan_struct->inodes_to_process,
2927                        scan_struct->process_inode_count);
2928
2929         if (ctx->progress)
2930                 if ((ctx->progress)(ctx, 1, group+1,
2931                                     ctx->fs->group_desc_count))
2932                         return EXT2_ET_CANCEL_REQUESTED;
2933
2934 #ifdef HAVE_PTHREAD
2935         if (ctx->global_ctx) {
2936                 tinfo = &ctx->thread_info;
2937                 tinfo->et_group_next++;
2938                 if (ctx->options & E2F_OPT_DEBUG &&
2939                     ctx->options & E2F_OPT_MULTITHREAD)
2940                         log_out(ctx, _("group %d finished\n"),
2941                                 tinfo->et_group_next);
2942                 if (tinfo->et_group_next >= tinfo->et_group_end)
2943                         return EXT2_ET_SCAN_FINISHED;
2944         }
2945 #endif
2946
2947         return 0;
2948 }
2949
2950 /*
2951  * Process the inodes in the "inodes to process" list.
2952  */
2953 static void process_inodes(e2fsck_t ctx, char *block_buf,
2954                            struct process_inode_block *inodes_to_process,
2955                            int *process_inode_count)
2956 {
2957         int                     i;
2958         struct ext2_inode       *old_stashed_inode;
2959         ext2_ino_t              old_stashed_ino;
2960         const char              *old_operation;
2961         char                    buf[80];
2962         struct problem_context  pctx;
2963
2964 #if 0
2965         printf("begin process_inodes: ");
2966 #endif
2967         if (*process_inode_count == 0)
2968                 return;
2969         old_operation = ehandler_operation(0);
2970         old_stashed_inode = ctx->stashed_inode;
2971         old_stashed_ino = ctx->stashed_ino;
2972         qsort(inodes_to_process, *process_inode_count,
2973                       sizeof(struct process_inode_block), process_inode_cmp);
2974         clear_problem_context(&pctx);
2975         for (i=0; i < *process_inode_count; i++) {
2976                 pctx.inode = ctx->stashed_inode =
2977                         (struct ext2_inode *) &inodes_to_process[i].inode;
2978                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2979
2980 #if 0
2981                 printf("%u ", pctx.ino);
2982 #endif
2983                 sprintf(buf, _("reading indirect blocks of inode %u"),
2984                         pctx.ino);
2985                 ehandler_operation(buf);
2986                 check_blocks(ctx, &pctx, block_buf,
2987                              &inodes_to_process[i].ea_ibody_quota);
2988                 if (e2fsck_should_abort(ctx))
2989                         break;
2990         }
2991         ctx->stashed_inode = old_stashed_inode;
2992         ctx->stashed_ino = old_stashed_ino;
2993         *process_inode_count = 0;
2994 #if 0
2995         printf("end process inodes\n");
2996 #endif
2997         ehandler_operation(old_operation);
2998 }
2999
3000 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
3001 {
3002         const struct process_inode_block *ib_a =
3003                 (const struct process_inode_block *) a;
3004         const struct process_inode_block *ib_b =
3005                 (const struct process_inode_block *) b;
3006         int     ret;
3007
3008         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
3009                ib_b->inode.i_block[EXT2_IND_BLOCK]);
3010         if (ret == 0)
3011                 /*
3012                  * We only call process_inodes() for non-extent
3013                  * inodes, so it's OK to pass NULL to
3014                  * ext2fs_file_acl_block() here.
3015                  */
3016                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
3017                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
3018         if (ret == 0)
3019                 ret = ib_a->ino - ib_b->ino;
3020         return ret;
3021 }
3022
3023 /*
3024  * Mark an inode as being bad in some what
3025  */
3026 static void mark_inode_bad(e2fsck_t ctx, ext2_ino_t ino)
3027 {
3028         struct          problem_context pctx;
3029
3030         if (!ctx->inode_bad_map) {
3031                 clear_problem_context(&pctx);
3032
3033                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3034                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
3035                                 "inode_bad_map", &ctx->inode_bad_map);
3036                 if (pctx.errcode) {
3037                         pctx.num = 3;
3038                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3039                         /* Should never get here */
3040                         ctx->flags |= E2F_FLAG_ABORT;
3041                         return;
3042                 }
3043         }
3044         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
3045 }
3046
3047 static void add_casefolded_dir(e2fsck_t ctx, ext2_ino_t ino)
3048 {
3049         struct          problem_context pctx;
3050
3051         if (!ctx->casefolded_dirs) {
3052                 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
3053                 if (pctx.errcode)
3054                         goto error;
3055         }
3056         pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
3057         if (pctx.errcode == 0)
3058                 return;
3059 error:
3060         fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
3061         /* Should never get here */
3062         ctx->flags |= E2F_FLAG_ABORT;
3063 }
3064
3065 /*
3066  * This procedure will allocate the inode "bb" (badblock) map table
3067  */
3068 static void alloc_bb_map(e2fsck_t ctx)
3069 {
3070         struct          problem_context pctx;
3071
3072         clear_problem_context(&pctx);
3073         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3074                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
3075                         "inode_bb_map", &ctx->inode_bb_map);
3076         if (pctx.errcode) {
3077                 pctx.num = 4;
3078                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3079                 /* Should never get here */
3080                 ctx->flags |= E2F_FLAG_ABORT;
3081                 return;
3082         }
3083 }
3084
3085 /*
3086  * This procedure will allocate the inode imagic table
3087  */
3088 static void alloc_imagic_map(e2fsck_t ctx)
3089 {
3090         struct          problem_context pctx;
3091
3092         clear_problem_context(&pctx);
3093         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3094                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
3095                         "inode_imagic_map", &ctx->inode_imagic_map);
3096         if (pctx.errcode) {
3097                 pctx.num = 5;
3098                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3099                 /* Should never get here */
3100                 ctx->flags |= E2F_FLAG_ABORT;
3101                 return;
3102         }
3103 }
3104
3105 /*
3106  * Marks a block as in use, setting the dup_map if it's been set
3107  * already.  Called by process_block and process_bad_block.
3108  *
3109  * WARNING: Assumes checks have already been done to make sure block
3110  * is valid.  This is true in both process_block and process_bad_block.
3111  */
3112 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
3113 {
3114         struct          problem_context pctx;
3115
3116         clear_problem_context(&pctx);
3117
3118         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
3119                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
3120                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
3121                         return;
3122                 }
3123                 if (!ctx->block_dup_map) {
3124                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
3125                                         _("multiply claimed block map"),
3126                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
3127                                         &ctx->block_dup_map);
3128                         if (pctx.errcode) {
3129                                 pctx.num = 3;
3130                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
3131                                             &pctx);
3132                                 /* Should never get here */
3133                                 ctx->flags |= E2F_FLAG_ABORT;
3134                                 return;
3135                         }
3136                 }
3137                 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
3138         } else {
3139                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
3140         }
3141 }
3142
3143 /*
3144  * When cluster size is greater than one block, it is caller's responsibility
3145  * to make sure block parameter starts at a cluster boundary.
3146  */
3147 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
3148                                       unsigned int num)
3149 {
3150         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
3151                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
3152         else {
3153                 unsigned int i;
3154
3155                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
3156                         mark_block_used(ctx, block + i);
3157         }
3158 }
3159
3160 /*
3161  * Adjust the extended attribute block's reference counts at the end
3162  * of pass 1, either by subtracting out references for EA blocks that
3163  * are still referenced in ctx->refcount, or by adding references for
3164  * EA blocks that had extra references as accounted for in
3165  * ctx->refcount_extra.
3166  */
3167 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
3168                                     char *block_buf, int adjust_sign)
3169 {
3170         struct ext2_ext_attr_header     *header;
3171         struct problem_context          pctx;
3172         ext2_filsys                     fs = ctx->fs;
3173         blk64_t                         blk;
3174         __u32                           should_be;
3175         ea_value_t                      count;
3176
3177         clear_problem_context(&pctx);
3178
3179         ea_refcount_intr_begin(refcount);
3180         while (1) {
3181                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
3182                         break;
3183                 pctx.blk = blk;
3184                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
3185                                                      pctx.ino);
3186                 if (pctx.errcode) {
3187                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
3188                         return;
3189                 }
3190                 header = (struct ext2_ext_attr_header *) block_buf;
3191                 pctx.blkcount = header->h_refcount;
3192                 should_be = header->h_refcount + adjust_sign * (int)count;
3193                 pctx.num = should_be;
3194                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
3195                         header->h_refcount = should_be;
3196                         pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
3197                                                              block_buf,
3198                                                              pctx.ino);
3199                         if (pctx.errcode) {
3200                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
3201                                             &pctx);
3202                                 continue;
3203                         }
3204                 }
3205         }
3206 }
3207
3208 /*
3209  * Handle processing the extended attribute blocks
3210  */
3211 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
3212                            char *block_buf, struct ea_quota *ea_block_quota)
3213 {
3214         ext2_filsys fs = ctx->fs;
3215         ext2_ino_t      ino = pctx->ino;
3216         struct ext2_inode *inode = pctx->inode;
3217         blk64_t         blk;
3218         char *          end;
3219         struct ext2_ext_attr_header *header;
3220         struct ext2_ext_attr_entry *first, *entry;
3221         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
3222         __u64           quota_inodes = 0;
3223         region_t        region = 0;
3224         int             failed_csum = 0;
3225
3226         ea_block_quota->blocks = 0;
3227         ea_block_quota->inodes = 0;
3228
3229         blk = ext2fs_file_acl_block(fs, inode);
3230         if (blk == 0)
3231                 return 0;
3232
3233         /*
3234          * If the Extended attribute flag isn't set, then a non-zero
3235          * file acl means that the inode is corrupted.
3236          *
3237          * Or if the extended attribute block is an invalid block,
3238          * then the inode is also corrupted.
3239          */
3240         if (!ext2fs_has_feature_xattr(fs->super) ||
3241             (blk < fs->super->s_first_data_block) ||
3242             (blk >= ext2fs_blocks_count(fs->super))) {
3243                 mark_inode_bad(ctx, ino);
3244                 return 0;
3245         }
3246
3247         /* If ea bitmap hasn't been allocated, create it */
3248         if (!ctx->block_ea_map) {
3249                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
3250                                         _("ext attr block map"),
3251                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
3252                                         &ctx->block_ea_map);
3253                 if (pctx->errcode) {
3254                         pctx->num = 2;
3255                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
3256                         ctx->flags |= E2F_FLAG_ABORT;
3257                         return 0;
3258                 }
3259         }
3260
3261         /* Create the EA refcount structure if necessary */
3262         if (!ctx->refcount) {
3263                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
3264                 if (pctx->errcode) {
3265                         pctx->num = 1;
3266                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3267                         ctx->flags |= E2F_FLAG_ABORT;
3268                         return 0;
3269                 }
3270         }
3271
3272 #if 0
3273         /* Debugging text */
3274         printf("Inode %u has EA block %u\n", ino, blk);
3275 #endif
3276
3277         /* Have we seen this EA block before? */
3278         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
3279                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
3280                 ea_block_quota->inodes = 0;
3281
3282                 if (ctx->ea_block_quota_blocks) {
3283                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
3284                                           &quota_blocks);
3285                         if (quota_blocks)
3286                                 ea_block_quota->blocks = quota_blocks;
3287                 }
3288
3289                 if (ctx->ea_block_quota_inodes)
3290                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
3291                                           &ea_block_quota->inodes);
3292
3293                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
3294                         return 1;
3295                 /* Ooops, this EA was referenced more than it stated */
3296                 if (!ctx->refcount_extra) {
3297                         pctx->errcode = ea_refcount_create(0,
3298                                            &ctx->refcount_extra);
3299                         if (pctx->errcode) {
3300                                 pctx->num = 2;
3301                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3302                                 ctx->flags |= E2F_FLAG_ABORT;
3303                                 return 0;
3304                         }
3305                 }
3306                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
3307                 return 1;
3308         }
3309
3310         /*
3311          * OK, we haven't seen this EA block yet.  So we need to
3312          * validate it
3313          */
3314         pctx->blk = blk;
3315         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
3316         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
3317                 pctx->errcode = 0;
3318                 failed_csum = 1;
3319         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
3320                 pctx->errcode = 0;
3321
3322         if (pctx->errcode &&
3323             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
3324                 pctx->errcode = 0;
3325                 goto clear_extattr;
3326         }
3327         header = (struct ext2_ext_attr_header *) block_buf;
3328         pctx->blk = ext2fs_file_acl_block(fs, inode);
3329         if (((ctx->ext_attr_ver == 1) &&
3330              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
3331             ((ctx->ext_attr_ver == 2) &&
3332              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
3333                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
3334                         goto clear_extattr;
3335         }
3336
3337         if (header->h_blocks != 1) {
3338                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
3339                         goto clear_extattr;
3340         }
3341
3342         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
3343                 goto clear_extattr;
3344
3345         region = region_create(0, fs->blocksize);
3346         if (!region) {
3347                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
3348                 ctx->flags |= E2F_FLAG_ABORT;
3349                 return 0;
3350         }
3351         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
3352                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3353                         goto clear_extattr;
3354         }
3355
3356         first = (struct ext2_ext_attr_entry *)(header+1);
3357         end = block_buf + fs->blocksize;
3358         entry = first;
3359         while ((char *)entry < end && *(__u32 *)entry) {
3360                 __u32 hash;
3361
3362                 if (region_allocate(region, (char *)entry - (char *)header,
3363                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
3364                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3365                                 goto clear_extattr;
3366                         break;
3367                 }
3368                 if ((ctx->ext_attr_ver == 1 &&
3369                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
3370                     (ctx->ext_attr_ver == 2 &&
3371                      entry->e_name_index == 0)) {
3372                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
3373                                 goto clear_extattr;
3374                         break;
3375                 }
3376                 if (entry->e_value_inum == 0) {
3377                         if (entry->e_value_size > EXT2_XATTR_SIZE_MAX ||
3378                             (entry->e_value_offs + entry->e_value_size >
3379                              fs->blocksize)) {
3380                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
3381                                         goto clear_extattr;
3382                                 break;
3383                         }
3384                         if (entry->e_value_size &&
3385                             region_allocate(region, entry->e_value_offs,
3386                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
3387                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
3388                                                 pctx))
3389                                         goto clear_extattr;
3390                         }
3391
3392                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
3393                                                           entry->e_value_offs);
3394                         if (entry->e_hash != hash)
3395                                 hash = ext2fs_ext_attr_hash_entry_signed(entry,
3396                                         block_buf + entry->e_value_offs);
3397
3398                         if (entry->e_hash != hash) {
3399                                 pctx->num = entry->e_hash;
3400                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
3401                                         goto clear_extattr;
3402                                 entry->e_hash = hash;
3403                         }
3404                 } else {
3405                         problem_t problem;
3406                         blk64_t entry_quota_blocks;
3407
3408                         problem = check_large_ea_inode(ctx, entry, pctx,
3409                                                        &entry_quota_blocks);
3410                         if (problem && fix_problem(ctx, problem, pctx))
3411                                 goto clear_extattr;
3412
3413                         quota_blocks += entry_quota_blocks;
3414                         quota_inodes++;
3415                 }
3416
3417                 entry = EXT2_EXT_ATTR_NEXT(entry);
3418         }
3419         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
3420                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3421                         goto clear_extattr;
3422         }
3423         region_free(region);
3424
3425         /*
3426          * We only get here if there was no other errors that were fixed.
3427          * If there was a checksum fail, ask to correct it.
3428          */
3429         if (failed_csum &&
3430             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
3431                 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
3432                                                        pctx->ino);
3433                 if (pctx->errcode)
3434                         return 0;
3435         }
3436
3437         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
3438                 if (!ctx->ea_block_quota_blocks) {
3439                         pctx->errcode = ea_refcount_create(0,
3440                                                 &ctx->ea_block_quota_blocks);
3441                         if (pctx->errcode) {
3442                                 pctx->num = 3;
3443                                 goto refcount_fail;
3444                         }
3445                 }
3446                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
3447                                   quota_blocks);
3448         }
3449
3450         if (quota_inodes) {
3451                 if (!ctx->ea_block_quota_inodes) {
3452                         pctx->errcode = ea_refcount_create(0,
3453                                                 &ctx->ea_block_quota_inodes);
3454                         if (pctx->errcode) {
3455                                 pctx->num = 4;
3456 refcount_fail:
3457                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3458                                 ctx->flags |= E2F_FLAG_ABORT;
3459                                 return 0;
3460                         }
3461                 }
3462
3463                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
3464                                   quota_inodes);
3465         }
3466         ea_block_quota->blocks = quota_blocks;
3467         ea_block_quota->inodes = quota_inodes;
3468
3469         inc_ea_inode_refs(ctx, pctx, first, end);
3470         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
3471         mark_block_used(ctx, blk);
3472         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
3473         return 1;
3474
3475 clear_extattr:
3476         if (region)
3477                 region_free(region);
3478         ext2fs_file_acl_block_set(fs, inode, 0);
3479         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
3480         return 0;
3481 }
3482
3483 /* Returns 1 if bad htree, 0 if OK */
3484 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
3485                         ext2_ino_t ino, struct ext2_inode *inode,
3486                         char *block_buf)
3487 {
3488         struct ext2_dx_root_info        *root;
3489         ext2_filsys                     fs = ctx->fs;
3490         errcode_t                       retval;
3491         blk64_t                         blk;
3492
3493         if ((!LINUX_S_ISDIR(inode->i_mode) &&
3494              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
3495             (!ext2fs_has_feature_dir_index(fs->super) &&
3496              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
3497                 return 1;
3498
3499         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
3500
3501         if ((pctx->errcode) ||
3502             (blk == 0) ||
3503             (blk < fs->super->s_first_data_block) ||
3504             (blk >= ext2fs_blocks_count(fs->super))) {
3505                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3506                         return 1;
3507                 else
3508                         return 0;
3509         }
3510
3511         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
3512         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3513                 return 1;
3514
3515         /* XXX should check that beginning matches a directory */
3516         root = (struct ext2_dx_root_info *) (block_buf + 24);
3517
3518         if ((root->reserved_zero || root->info_length < 8) &&
3519             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3520                 return 1;
3521
3522         pctx->num = root->hash_version;
3523         if ((root->hash_version != EXT2_HASH_LEGACY) &&
3524             (root->hash_version != EXT2_HASH_HALF_MD4) &&
3525             (root->hash_version != EXT2_HASH_TEA) &&
3526             (root->hash_version != EXT2_HASH_SIPHASH) &&
3527             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
3528                 return 1;
3529
3530         if (ext4_hash_in_dirent(inode)) {
3531                 if (root->hash_version != EXT2_HASH_SIPHASH &&
3532                     fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
3533                         return 1;
3534         } else {
3535                 if (root->hash_version == EXT2_HASH_SIPHASH &&
3536                    fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
3537                         return 1;
3538         }
3539
3540         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
3541             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
3542                 return 1;
3543
3544         pctx->num = root->indirect_levels;
3545         /* if htree level is clearly too high, consider it to be broken */
3546         if (root->indirect_levels > EXT4_HTREE_LEVEL &&
3547             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3548                 return 1;
3549
3550         /* if level is only maybe too high, LARGE_DIR feature could be unset */
3551         if (root->indirect_levels > ext2_dir_htree_level(fs) &&
3552             !ext2fs_has_feature_largedir(fs->super)) {
3553                 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
3554                 unsigned idx_pb = 1 << (blockbits - 3);
3555
3556                 /* compare inode size/blocks vs. max-sized 2-level htree */
3557                 if (EXT2_I_SIZE(pctx->inode) <
3558                     (idx_pb - 1) * (idx_pb - 2) << blockbits &&
3559                     pctx->inode->i_blocks <
3560                     (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
3561                     fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3562                         return 1;
3563         }
3564
3565         if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
3566             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3567                 ctx->large_dirs++;
3568
3569         return 0;
3570 }
3571
3572 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
3573                         struct ext2_inode *inode, int restart_flag,
3574                         const char *source)
3575 {
3576         inode->i_flags = 0;
3577         inode->i_links_count = 0;
3578         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
3579         inode->i_dtime = ctx->now;
3580
3581         /*
3582          * If a special inode has such rotten block mappings that we
3583          * want to clear the whole inode, be sure to actually zap
3584          * the block maps because i_links_count isn't checked for
3585          * special inodes, and we'll end up right back here the next
3586          * time we run fsck.
3587          */
3588         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
3589                 memset(inode->i_block, 0, sizeof(inode->i_block));
3590
3591         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
3592         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
3593         if (ctx->inode_reg_map)
3594                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
3595         if (ctx->inode_bad_map)
3596                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
3597
3598         /*
3599          * If the inode was partially accounted for before processing
3600          * was aborted, we need to restart the pass 1 scan.
3601          */
3602         ctx->flags |= restart_flag;
3603
3604         if (ino == EXT2_BAD_INO)
3605                 memset(inode, 0, sizeof(struct ext2_inode));
3606
3607         e2fsck_write_inode(ctx, ino, inode, source);
3608 }
3609
3610 /*
3611  * Use the multiple-blocks reclamation code to fix alignment problems in
3612  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
3613  * physical cluster, and we want the block offsets within that cluster to
3614  * line up.
3615  */
3616 static int has_unaligned_cluster_map(e2fsck_t ctx,
3617                                      blk64_t last_pblk, blk64_t last_lblk,
3618                                      blk64_t pblk, blk64_t lblk)
3619 {
3620         blk64_t cluster_mask;
3621
3622         if (!ctx->fs->cluster_ratio_bits)
3623                 return 0;
3624         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
3625
3626         /*
3627          * If the block in the logical cluster doesn't align with the block in
3628          * the physical cluster...
3629          */
3630         if ((lblk & cluster_mask) != (pblk & cluster_mask))
3631                 return 1;
3632
3633         /*
3634          * If we cross a physical cluster boundary within a logical cluster...
3635          */
3636         if (last_pblk && (lblk & cluster_mask) != 0 &&
3637             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
3638             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
3639                 return 1;
3640
3641         return 0;
3642 }
3643
3644 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
3645                              struct process_block_struct *pb,
3646                              blk64_t start_block, blk64_t end_block,
3647                              blk64_t eof_block,
3648                              ext2_extent_handle_t ehandle,
3649                              int try_repairs)
3650 {
3651         struct ext2fs_extent    extent;
3652         blk64_t                 blk, last_lblk;
3653         unsigned int            i, n;
3654         int                     is_dir, is_leaf;
3655         problem_t               problem;
3656         struct ext2_extent_info info;
3657         int                     failed_csum = 0;
3658
3659         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
3660                 failed_csum = 1;
3661
3662         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
3663         if (pctx->errcode)
3664                 return;
3665         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
3666             !pb->eti.force_rebuild &&
3667             info.curr_level < MAX_EXTENT_DEPTH_COUNT) {
3668                 struct extent_tree_level *etl;
3669
3670                 etl = pb->eti.ext_info + info.curr_level;
3671                 etl->num_extents += info.num_entries;
3672                 etl->max_extents += info.max_entries;
3673                 /*
3674                  * Implementation wart: Splitting extent blocks when appending
3675                  * will leave the old block with one free entry.  Therefore
3676                  * unless the node is totally full, pretend that a non-root
3677                  * extent block can hold one fewer entry than it actually does,
3678                  * so that we don't repeatedly rebuild the extent tree.
3679                  */
3680                 if (info.curr_level && info.num_entries < info.max_entries)
3681                         etl->max_extents--;
3682         }
3683
3684         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
3685                                           &extent);
3686         while ((pctx->errcode == 0 ||
3687                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
3688                info.num_entries-- > 0) {
3689                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
3690                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
3691                 last_lblk = extent.e_lblk + extent.e_len - 1;
3692
3693                 problem = 0;
3694                 pctx->blk = extent.e_pblk;
3695                 pctx->blk2 = extent.e_lblk;
3696                 pctx->num = extent.e_len;
3697                 pctx->blkcount = extent.e_lblk + extent.e_len;
3698
3699                 if (extent.e_pblk == 0 ||
3700                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
3701                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
3702                         problem = PR_1_EXTENT_BAD_START_BLK;
3703                 else if (extent.e_lblk < start_block)
3704                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
3705                 else if ((end_block && last_lblk > end_block) &&
3706                          !(last_lblk > eof_block &&
3707                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
3708                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
3709                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
3710                 else if (is_leaf && extent.e_len == 0)
3711                         problem = PR_1_EXTENT_LENGTH_ZERO;
3712                 else if (is_leaf &&
3713                          (extent.e_pblk + extent.e_len) >
3714                          ext2fs_blocks_count(ctx->fs->super))
3715                         problem = PR_1_EXTENT_ENDS_BEYOND;
3716                 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
3717                          !ext2fs_has_feature_largedir(ctx->fs->super) &&
3718                          ((extent.e_lblk + extent.e_len) >
3719                           (1U << (21 - ctx->fs->super->s_log_block_size))))
3720                         problem = PR_1_TOOBIG_DIR;
3721
3722                 if (is_leaf && problem == 0 && extent.e_len > 0) {
3723 #if 0
3724                         printf("extent_region(ino=%u, expect=%llu, "
3725                                "lblk=%llu, len=%u)\n", pb->ino,
3726                                (unsigned long long) pb->next_lblock,
3727                                (unsigned long long) extent.e_lblk,
3728                                extent.e_len);
3729 #endif
3730                         if (extent.e_lblk < pb->next_lblock)
3731                                 problem = PR_1_EXTENT_COLLISION;
3732                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
3733                                 pb->next_lblock = extent.e_lblk + extent.e_len;
3734                 }
3735
3736                 /*
3737                  * Uninitialized blocks in a directory?  Clear the flag and
3738                  * we'll interpret the blocks later.
3739                  */
3740                 if (try_repairs && is_dir && problem == 0 &&
3741                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3742                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
3743                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3744                         pb->inode_modified = 1;
3745                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3746                                                               &extent);
3747                         if (pctx->errcode)
3748                                 return;
3749                         failed_csum = 0;
3750                 }
3751 #ifdef CONFIG_DEVELOPER_FEATURES
3752                 if (try_repairs && !is_dir && problem == 0 &&
3753                     (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
3754                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3755                     fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
3756                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3757                         pb->inode_modified = 1;
3758                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3759                                                               &extent);
3760                         if (pctx->errcode)
3761                                 return;
3762                         failed_csum = 0;
3763                 }
3764 #endif
3765                 if (try_repairs && problem) {
3766 report_problem:
3767                         if (fix_problem(ctx, problem, pctx)) {
3768                                 if (ctx->invalid_bitmaps) {
3769                                         /*
3770                                          * If fsck knows the bitmaps are bad,
3771                                          * skip to the next extent and
3772                                          * try to clear this extent again
3773                                          * after fixing the bitmaps, by
3774                                          * restarting fsck.
3775                                          */
3776                                         pctx->errcode = ext2fs_extent_get(
3777                                                           ehandle,
3778                                                           EXT2_EXTENT_NEXT_SIB,
3779                                                           &extent);
3780                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3781                                         if (pctx->errcode ==
3782                                                     EXT2_ET_NO_CURRENT_NODE) {
3783                                                 pctx->errcode = 0;
3784                                                 break;
3785                                         }
3786                                         continue;
3787                                 }
3788                                 e2fsck_read_bitmaps(ctx);
3789                                 pb->inode_modified = 1;
3790                                 pctx->errcode =
3791                                         ext2fs_extent_delete(ehandle, 0);
3792                                 if (pctx->errcode) {
3793                                         pctx->str = "ext2fs_extent_delete";
3794                                         return;
3795                                 }
3796                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3797                                 if (pctx->errcode &&
3798                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
3799                                         pctx->str = "ext2fs_extent_fix_parents";
3800                                         return;
3801                                 }
3802                                 pctx->errcode = ext2fs_extent_get(ehandle,
3803                                                                   EXT2_EXTENT_CURRENT,
3804                                                                   &extent);
3805                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
3806                                         pctx->errcode = 0;
3807                                         break;
3808                                 }
3809                                 failed_csum = 0;
3810                                 continue;
3811                         }
3812                         goto next;
3813                 }
3814
3815                 if (!is_leaf) {
3816                         blk64_t lblk = extent.e_lblk;
3817                         int next_try_repairs = 1;
3818
3819                         blk = extent.e_pblk;
3820
3821                         /*
3822                          * If this lower extent block collides with critical
3823                          * metadata, don't try to repair the damage.  Pass 1b
3824                          * will reallocate the block; then we can try again.
3825                          */
3826                         if (pb->ino != EXT2_RESIZE_INO &&
3827                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
3828                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3829                                                       extent.e_pblk)) {
3830                                 next_try_repairs = 0;
3831                                 pctx->blk = blk;
3832                                 fix_problem(ctx,
3833                                             PR_1_CRITICAL_METADATA_COLLISION,
3834                                             pctx);
3835                                 if ((ctx->options & E2F_OPT_NO) == 0)
3836                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3837                         }
3838                         pctx->errcode = ext2fs_extent_get(ehandle,
3839                                                   EXT2_EXTENT_DOWN, &extent);
3840                         if (pctx->errcode &&
3841                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
3842                                 pctx->str = "EXT2_EXTENT_DOWN";
3843                                 problem = PR_1_EXTENT_HEADER_INVALID;
3844                                 if (!next_try_repairs)
3845                                         return;
3846                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
3847                                         goto report_problem;
3848                                 return;
3849                         }
3850                         /* The next extent should match this index's logical start */
3851                         if (extent.e_lblk != lblk) {
3852                                 struct ext2_extent_info e_info;
3853
3854                                 pctx->errcode = ext2fs_extent_get_info(ehandle,
3855                                                                        &e_info);
3856                                 if (pctx->errcode) {
3857                                         pctx->str = "ext2fs_extent_get_info";
3858                                         return;
3859                                 }
3860                                 pctx->blk = lblk;
3861                                 pctx->blk2 = extent.e_lblk;
3862                                 pctx->num = e_info.curr_level - 1;
3863                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
3864                                 if (fix_problem(ctx, problem, pctx)) {
3865                                         pb->inode_modified = 1;
3866                                         pctx->errcode =
3867                                                 ext2fs_extent_fix_parents(ehandle);
3868                                         if (pctx->errcode) {
3869                                                 pctx->str = "ext2fs_extent_fix_parents";
3870                                                 return;
3871                                         }
3872                                 }
3873                         }
3874                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
3875                                          last_lblk, eof_block, ehandle,
3876                                          next_try_repairs);
3877                         if (pctx->errcode)
3878                                 return;
3879                         pctx->errcode = ext2fs_extent_get(ehandle,
3880                                                   EXT2_EXTENT_UP, &extent);
3881                         if (pctx->errcode) {
3882                                 pctx->str = "EXT2_EXTENT_UP";
3883                                 return;
3884                         }
3885                         mark_block_used(ctx, blk);
3886                         pb->num_blocks++;
3887                         goto next;
3888                 }
3889
3890                 if ((pb->previous_block != 0) &&
3891                     (pb->previous_block+1 != extent.e_pblk)) {
3892                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3893                                 char type = '?';
3894
3895                                 if (pb->is_dir)
3896                                         type = 'd';
3897                                 else if (pb->is_reg)
3898                                         type = 'f';
3899
3900                                 printf(("%6lu(%c): expecting %6lu "
3901                                         "actual extent "
3902                                         "phys %6lu log %lu len %lu\n"),
3903                                        (unsigned long) pctx->ino, type,
3904                                        (unsigned long) pb->previous_block+1,
3905                                        (unsigned long) extent.e_pblk,
3906                                        (unsigned long) extent.e_lblk,
3907                                        (unsigned long) extent.e_len);
3908                         }
3909                         pb->fragmented = 1;
3910                 }
3911                 /*
3912                  * If we notice a gap in the logical block mappings of an
3913                  * extent-mapped directory, offer to close the hole by
3914                  * moving the logical block down, otherwise we'll go mad in
3915                  * pass 3 allocating empty directory blocks to fill the hole.
3916                  */
3917                 if (try_repairs && is_dir &&
3918                     pb->last_block + 1 < extent.e_lblk) {
3919                         blk64_t new_lblk;
3920
3921                         new_lblk = pb->last_block + 1;
3922                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3923                                 new_lblk = ((new_lblk +
3924                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3925                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3926                                            (extent.e_pblk &
3927                                             EXT2FS_CLUSTER_MASK(ctx->fs));
3928                         pctx->blk = extent.e_lblk;
3929                         pctx->blk2 = new_lblk;
3930                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3931                                 extent.e_lblk = new_lblk;
3932                                 pb->inode_modified = 1;
3933                                 pctx->errcode = ext2fs_extent_replace(ehandle,
3934                                                                 0, &extent);
3935                                 if (pctx->errcode) {
3936                                         pctx->errcode = 0;
3937                                         goto alloc_later;
3938                                 }
3939                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3940                                 if (pctx->errcode)
3941                                         goto failed_add_dir_block;
3942                                 pctx->errcode = ext2fs_extent_goto(ehandle,
3943                                                                 extent.e_lblk);
3944                                 if (pctx->errcode)
3945                                         goto failed_add_dir_block;
3946                                 last_lblk = extent.e_lblk + extent.e_len - 1;
3947                                 failed_csum = 0;
3948                         }
3949                 }
3950 alloc_later:
3951                 if (is_dir) {
3952                         while (++pb->last_db_block <
3953                                (e2_blkcnt_t) extent.e_lblk) {
3954                                 pctx->errcode = ext2fs_add_dir_block2(
3955                                                         ctx->fs->dblist,
3956                                                         pb->ino, 0,
3957                                                         pb->last_db_block);
3958                                 if (pctx->errcode) {
3959                                         pctx->blk = 0;
3960                                         pctx->num = pb->last_db_block;
3961                                         goto failed_add_dir_block;
3962                                 }
3963                         }
3964
3965                         for (i = 0; i < extent.e_len; i++) {
3966                                 pctx->errcode = ext2fs_add_dir_block2(
3967                                                         ctx->fs->dblist,
3968                                                         pctx->ino,
3969                                                         extent.e_pblk + i,
3970                                                         extent.e_lblk + i);
3971                                 if (pctx->errcode) {
3972                                         pctx->blk = extent.e_pblk + i;
3973                                         pctx->num = extent.e_lblk + i;
3974                                 failed_add_dir_block:
3975                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3976                                         /* Should never get here */
3977                                         ctx->flags |= E2F_FLAG_ABORT;
3978                                         return;
3979                                 }
3980                         }
3981                         if (extent.e_len > 0)
3982                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3983                 }
3984                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3985                                               pb->last_block,
3986                                               extent.e_pblk,
3987                                               extent.e_lblk)) {
3988                         for (i = 0; i < extent.e_len; i++) {
3989                                 pctx->blk = extent.e_lblk + i;
3990                                 pctx->blk2 = extent.e_pblk + i;
3991                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3992                                 mark_block_used(ctx, extent.e_pblk + i);
3993                                 mark_block_used(ctx, extent.e_pblk + i);
3994                         }
3995                 }
3996
3997                 /*
3998                  * Check whether first cluster got marked in previous iteration.
3999                  */
4000                 if (ctx->fs->cluster_ratio_bits &&
4001                     pb->previous_block &&
4002                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
4003                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
4004                         /* Set blk to the beginning of next cluster. */
4005                         blk = EXT2FS_C2B(
4006                                 ctx->fs,
4007                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
4008                 else
4009                         /* Set blk to the beginning of current cluster. */
4010                         blk = EXT2FS_C2B(ctx->fs,
4011                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
4012
4013                 if (blk < extent.e_pblk + extent.e_len) {
4014                         mark_blocks_used(ctx, blk,
4015                                          extent.e_pblk + extent.e_len - blk);
4016                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
4017                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
4018                         pb->num_blocks += n;
4019                 }
4020                 pb->last_block = extent.e_lblk + extent.e_len - 1;
4021                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
4022                 start_block = pb->last_block = last_lblk;
4023                 if (is_leaf && !is_dir &&
4024                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
4025                         pb->last_init_lblock = last_lblk;
4026         next:
4027                 pctx->errcode = ext2fs_extent_get(ehandle,
4028                                                   EXT2_EXTENT_NEXT_SIB,
4029                                                   &extent);
4030         }
4031
4032         /* Failed csum but passes checks?  Ask to fix checksum. */
4033         if (failed_csum &&
4034             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
4035                 pb->inode_modified = 1;
4036                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
4037                 if (pctx->errcode)
4038                         return;
4039         }
4040
4041         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
4042                 pctx->errcode = 0;
4043 }
4044
4045 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
4046                                  struct process_block_struct *pb)
4047 {
4048         struct ext2_extent_info info;
4049         struct ext2_inode       *inode = pctx->inode;
4050         ext2_extent_handle_t    ehandle;
4051         ext2_filsys             fs = ctx->fs;
4052         ext2_ino_t              ino = pctx->ino;
4053         errcode_t               retval;
4054         blk64_t                 eof_lblk;
4055         struct ext3_extent_header       *eh;
4056
4057         /* Check for a proper extent header... */
4058         eh = (struct ext3_extent_header *) &inode->i_block[0];
4059         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
4060         if (retval) {
4061                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
4062                         e2fsck_clear_inode(ctx, ino, inode, 0,
4063                                            "check_blocks_extents");
4064                 pctx->errcode = 0;
4065                 return;
4066         }
4067
4068         /* ...since this function doesn't fail if i_block is zeroed. */
4069         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
4070         if (pctx->errcode) {
4071                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
4072                         e2fsck_clear_inode(ctx, ino, inode, 0,
4073                                            "check_blocks_extents");
4074                 pctx->errcode = 0;
4075                 return;
4076         }
4077
4078         retval = ext2fs_extent_get_info(ehandle, &info);
4079         if (retval == 0) {
4080                 int max_depth = info.max_depth;
4081
4082                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
4083                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
4084                 ctx->extent_depth_count[max_depth]++;
4085         }
4086
4087         /* Check maximum extent depth */
4088         pctx->blk = info.max_depth;
4089         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
4090         if (pctx->blk2 < pctx->blk &&
4091             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
4092                 pb->eti.force_rebuild = 1;
4093
4094         /* Can we collect extent tree level stats? */
4095         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
4096         if (pctx->blk2 > pctx->blk)
4097                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
4098         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
4099         pb->eti.ino = pb->ino;
4100
4101         pb->next_lblock = 0;
4102
4103         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
4104                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
4105         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
4106         if (pctx->errcode &&
4107             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
4108                 pb->num_blocks = 0;
4109                 inode->i_blocks = 0;
4110                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4111                                    "check_blocks_extents");
4112                 pctx->errcode = 0;
4113         }
4114         ext2fs_extent_free(ehandle);
4115
4116         /* Rebuild unless it's a dir and we're rehashing it */
4117         if (LINUX_S_ISDIR(inode->i_mode) &&
4118             e2fsck_dir_will_be_rehashed(ctx, ino))
4119                 return;
4120
4121         if (ctx->options & E2F_OPT_CONVERT_BMAP)
4122                 e2fsck_rebuild_extents_later(ctx, ino);
4123         else
4124                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
4125 }
4126
4127 /*
4128  * In fact we don't need to check blocks for an inode with inline data
4129  * because this inode doesn't have any blocks.  In this function all
4130  * we need to do is add this inode into dblist when it is a directory.
4131  */
4132 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
4133                                      struct process_block_struct *pb)
4134 {
4135         int     flags;
4136         size_t  inline_data_size = 0;
4137
4138         if (!pb->is_dir) {
4139                 pctx->errcode = 0;
4140                 return;
4141         }
4142
4143         /* Process the dirents in i_block[] as the "first" block. */
4144         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
4145         if (pctx->errcode)
4146                 goto err;
4147
4148         /* Process the dirents in the EA as a "second" block. */
4149         flags = ctx->fs->flags;
4150         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4151         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
4152                                                 &inline_data_size);
4153         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4154                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4155         if (pctx->errcode) {
4156                 pctx->errcode = 0;
4157                 return;
4158         }
4159
4160         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
4161                 return;
4162
4163         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
4164         if (pctx->errcode)
4165                 goto err;
4166
4167         return;
4168 err:
4169         pctx->blk = 0;
4170         pctx->num = 0;
4171         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4172         ctx->flags |= E2F_FLAG_ABORT;
4173 }
4174
4175 /*
4176  * This subroutine is called on each inode to account for all of the
4177  * blocks used by that inode.
4178  */
4179 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
4180                          char *block_buf, const struct ea_quota *ea_ibody_quota)
4181 {
4182         ext2_filsys fs = ctx->fs;
4183         struct process_block_struct pb;
4184         ext2_ino_t      ino = pctx->ino;
4185         struct ext2_inode *inode = pctx->inode;
4186         unsigned        bad_size = 0;
4187         int             dirty_inode = 0;
4188         int             extent_fs;
4189         int             inlinedata_fs;
4190         __u64           size;
4191         struct ea_quota ea_block_quota;
4192
4193         pb.ino = ino;
4194         pb.num_blocks = EXT2FS_B2C(ctx->fs,
4195                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
4196         pb.last_block = ~0;
4197         pb.last_init_lblock = -1;
4198         pb.last_db_block = -1;
4199         pb.num_illegal_blocks = 0;
4200         pb.suppress = 0; pb.clear = 0;
4201         pb.fragmented = 0;
4202         pb.compressed = 0;
4203         pb.previous_block = 0;
4204         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
4205         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
4206         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
4207         pb.inode = inode;
4208         pb.pctx = pctx;
4209         pb.ctx = ctx;
4210         pb.inode_modified = 0;
4211         pb.eti.force_rebuild = 0;
4212         pctx->ino = ino;
4213         pctx->errcode = 0;
4214
4215         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
4216         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
4217
4218         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
4219                 if (e2fsck_should_abort(ctx))
4220                         goto out;
4221                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
4222         }
4223
4224         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
4225                 check_blocks_inline_data(ctx, pctx, &pb);
4226         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
4227                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
4228                         check_blocks_extents(ctx, pctx, &pb);
4229                 else {
4230                         int flags;
4231                         /*
4232                          * If we've modified the inode, write it out before
4233                          * iterate() tries to use it.
4234                          */
4235                         if (dirty_inode) {
4236                                 e2fsck_write_inode(ctx, ino, inode,
4237                                                    "check_blocks");
4238                                 dirty_inode = 0;
4239                         }
4240                         flags = fs->flags;
4241                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4242                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
4243                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
4244                                                 block_buf, process_block, &pb);
4245                         /*
4246                          * We do not have uninitialized extents in non extent
4247                          * files.
4248                          */
4249                         pb.last_init_lblock = pb.last_block;
4250                         /*
4251                          * If iterate() changed a block mapping, we have to
4252                          * re-read the inode.  If we decide to clear the
4253                          * inode after clearing some stuff, we'll re-write the
4254                          * bad mappings into the inode!
4255                          */
4256                         if (pb.inode_modified)
4257                                 e2fsck_read_inode(ctx, ino, inode,
4258                                                   "check_blocks");
4259                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4260                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4261
4262                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
4263 #ifdef DEBUG
4264                                 printf("bmap rebuild ino=%d\n", ino);
4265 #endif
4266                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
4267                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
4268                                         e2fsck_rebuild_extents_later(ctx, ino);
4269                         }
4270                 }
4271         }
4272         end_problem_latch(ctx, PR_LATCH_BLOCK);
4273         end_problem_latch(ctx, PR_LATCH_TOOBIG);
4274         if (e2fsck_should_abort(ctx))
4275                 goto out;
4276         if (pctx->errcode)
4277                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
4278
4279         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
4280                 if (LINUX_S_ISDIR(inode->i_mode))
4281                         ctx->fs_fragmented_dir++;
4282                 else
4283                         ctx->fs_fragmented++;
4284         }
4285
4286         if (pb.clear) {
4287                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4288                                    "check_blocks");
4289                 return;
4290         }
4291
4292         if (inode->i_flags & EXT2_INDEX_FL) {
4293                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
4294                         inode->i_flags &= ~EXT2_INDEX_FL;
4295                         dirty_inode++;
4296                 } else {
4297                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
4298                 }
4299         }
4300
4301         if (!pb.num_blocks && pb.is_dir &&
4302             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
4303                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
4304                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
4305                         ctx->fs_directory_count--;
4306                         return;
4307                 }
4308         }
4309
4310         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
4311             ino != fs->super->s_orphan_file_inum &&
4312             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
4313             !(inode->i_flags & EXT4_EA_INODE_FL)) {
4314                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
4315                                ino,
4316                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
4317                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
4318                                   ino, (ea_ibody_quota ?
4319                                         ea_ibody_quota->inodes : 0) +
4320                                                 ea_block_quota.inodes + 1);
4321         }
4322
4323         if (!ext2fs_has_feature_huge_file(fs->super) ||
4324             !(inode->i_flags & EXT4_HUGE_FILE_FL))
4325                 pb.num_blocks *= (fs->blocksize / 512);
4326         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
4327 #if 0
4328         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
4329                ino, inode->i_size, (unsigned long long) pb.last_block,
4330                (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
4331                (unsigned long long) pb.num_blocks);
4332 #endif
4333         size = EXT2_I_SIZE(inode);
4334         if (pb.is_dir) {
4335                 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
4336                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
4337                         int flags;
4338                         size_t sz = 0;
4339                         errcode_t err;
4340
4341                         flags = ctx->fs->flags;
4342                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4343                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
4344                                                       &sz);
4345                         ctx->fs->flags = (flags &
4346                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4347                                          (ctx->fs->flags &
4348                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4349                         if (err || sz != size) {
4350                                 bad_size = 7;
4351                                 pctx->num = sz;
4352                         }
4353                 } else if (size & (fs->blocksize - 1))
4354                         bad_size = 5;
4355                 else if (nblock > (pb.last_block + 1))
4356                         bad_size = 1;
4357                 else if (nblock < (pb.last_block + 1)) {
4358                         if (((pb.last_block + 1) - nblock) >
4359                             fs->super->s_prealloc_dir_blocks)
4360                                 bad_size = 2;
4361                 }
4362         } else {
4363                 if ((pb.last_init_lblock >= 0) &&
4364                     /* Do not allow initialized allocated blocks past i_size*/
4365                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
4366                     !(inode->i_flags & EXT4_VERITY_FL))
4367                         bad_size = 3;
4368                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4369                          size > ext2_max_sizes[fs->super->s_log_block_size])
4370                         /* too big for a direct/indirect-mapped file */
4371                         bad_size = 4;
4372                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4373                          size >
4374                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
4375                         /* too big for an extent-based file - 32bit ee_block */
4376                         bad_size = 6;
4377         }
4378         /* i_size for symlinks is checked elsewhere */
4379         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
4380                 /* Did inline_data set pctx->num earlier? */
4381                 if (bad_size != 7)
4382                         pctx->num = (pb.last_block + 1) * fs->blocksize;
4383                 pctx->group = bad_size;
4384                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
4385                         ext2fs_inode_size_set(fs, inode, pctx->num);
4386                         if (EXT2_I_SIZE(inode) == 0 &&
4387                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
4388                                 memset(inode->i_block, 0,
4389                                        sizeof(inode->i_block));
4390                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
4391                         }
4392                         dirty_inode++;
4393                 }
4394                 pctx->num = 0;
4395         }
4396         if (LINUX_S_ISREG(inode->i_mode) &&
4397             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4398                 ctx->large_files++;
4399         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
4400             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
4401              (ext2fs_has_feature_huge_file(fs->super) &&
4402               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
4403               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
4404                 pctx->num = pb.num_blocks;
4405                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
4406                         inode->i_blocks = pb.num_blocks;
4407                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
4408                         dirty_inode++;
4409                 }
4410                 pctx->num = 0;
4411         }
4412
4413         /*
4414          * The kernel gets mad if we ask it to allocate bigalloc clusters to
4415          * a block mapped file, so rebuild it as an extent file.  We can skip
4416          * symlinks because they're never rewritten.
4417          */
4418         if (ext2fs_has_feature_bigalloc(fs->super) &&
4419             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
4420             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
4421             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
4422             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
4423             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
4424                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
4425                 if (pctx->errcode)
4426                         goto out;
4427         }
4428
4429         if (ctx->dirs_to_hash && pb.is_dir &&
4430             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
4431             !(inode->i_flags & EXT2_INDEX_FL) &&
4432             ((inode->i_size / fs->blocksize) >= 3))
4433                 e2fsck_rehash_dir_later(ctx, ino);
4434
4435 out:
4436         if (dirty_inode)
4437                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
4438 }
4439
4440 #if 0
4441 /*
4442  * Helper function called by process block when an illegal block is
4443  * found.  It returns a description about why the block is illegal
4444  */
4445 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
4446 {
4447         blk64_t super;
4448         int     i;
4449         static char     problem[80];
4450
4451         super = fs->super->s_first_data_block;
4452         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
4453         if (block < super) {
4454                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
4455                 return(problem);
4456         } else if (block >= ext2fs_blocks_count(fs->super)) {
4457                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
4458                 return(problem);
4459         }
4460         for (i = 0; i < fs->group_desc_count; i++) {
4461                 if (block == super) {
4462                         sprintf(problem, "is the superblock in group %d", i);
4463                         break;
4464                 }
4465                 if (block > super &&
4466                     block <= (super + fs->desc_blocks)) {
4467                         sprintf(problem, "is in the group descriptors "
4468                                 "of group %d", i);
4469                         break;
4470                 }
4471                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
4472                         sprintf(problem, "is the block bitmap of group %d", i);
4473                         break;
4474                 }
4475                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
4476                         sprintf(problem, "is the inode bitmap of group %d", i);
4477                         break;
4478                 }
4479                 if (block >= ext2fs_inode_table_loc(fs, i) &&
4480                     (block < ext2fs_inode_table_loc(fs, i)
4481                      + fs->inode_blocks_per_group)) {
4482                         sprintf(problem, "is in the inode table of group %d",
4483                                 i);
4484                         break;
4485                 }
4486                 super += fs->super->s_blocks_per_group;
4487         }
4488         return(problem);
4489 }
4490 #endif
4491
4492 /*
4493  * This is a helper function for check_blocks().
4494  */
4495 static int process_block(ext2_filsys fs,
4496                   blk64_t       *block_nr,
4497                   e2_blkcnt_t blockcnt,
4498                   blk64_t ref_block EXT2FS_ATTR((unused)),
4499                   int ref_offset EXT2FS_ATTR((unused)),
4500                   void *priv_data)
4501 {
4502         struct process_block_struct *p;
4503         struct problem_context *pctx;
4504         blk64_t blk = *block_nr;
4505         int     ret_code = 0;
4506         problem_t       problem = 0;
4507         e2fsck_t        ctx;
4508
4509         p = (struct process_block_struct *) priv_data;
4510         pctx = p->pctx;
4511         ctx = p->ctx;
4512
4513         /*
4514          * For a directory, add logical block zero for processing even if it's
4515          * not mapped or we'll be perennially stuck with broken "." and ".."
4516          * entries.
4517          */
4518         if (p->is_dir && blockcnt == 0 && blk == 0) {
4519                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
4520                 if (pctx->errcode) {
4521                         pctx->blk = blk;
4522                         pctx->num = blockcnt;
4523                         goto failed_add_dir_block;
4524                 }
4525                 p->last_db_block++;
4526         }
4527
4528         if (blk == 0)
4529                 return 0;
4530
4531 #if 0
4532         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
4533                blockcnt);
4534 #endif
4535
4536         /*
4537          * Simplistic fragmentation check.  We merely require that the
4538          * file be contiguous.  (Which can never be true for really
4539          * big files that are greater than a block group.)
4540          */
4541         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
4542                 if (p->previous_block+1 != blk) {
4543                         if (ctx->options & E2F_OPT_FRAGCHECK) {
4544                                 char type = '?';
4545
4546                                 if (p->is_dir)
4547                                         type = 'd';
4548                                 else if (p->is_reg)
4549                                         type = 'f';
4550
4551                                 printf(_("%6lu(%c): expecting %6lu "
4552                                          "got phys %6lu (blkcnt %lld)\n"),
4553                                        (unsigned long) pctx->ino, type,
4554                                        (unsigned long) p->previous_block+1,
4555                                        (unsigned long) blk,
4556                                        (long long) blockcnt);
4557                         }
4558                         p->fragmented = 1;
4559                 }
4560         }
4561
4562         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
4563             !pctx->inode->i_size_high &&
4564             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
4565                 problem = PR_1_TOOBIG_DIR;
4566         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
4567                 problem = PR_1_TOOBIG_DIR;
4568         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
4569                 problem = PR_1_TOOBIG_REG;
4570         if (!p->is_dir && !p->is_reg && blockcnt > 0)
4571                 problem = PR_1_TOOBIG_SYMLINK;
4572
4573         if (blk < fs->super->s_first_data_block ||
4574             blk >= ext2fs_blocks_count(fs->super))
4575                 problem = PR_1_ILLEGAL_BLOCK_NUM;
4576
4577         /*
4578          * If this IND/DIND/TIND block is squatting atop some critical metadata
4579          * (group descriptors, superblock, bitmap, inode table), any write to
4580          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
4581          * fix that and restart fsck.
4582          */
4583         if (blockcnt < 0 &&
4584             p->ino != EXT2_RESIZE_INO &&
4585             blk < ctx->fs->super->s_blocks_count &&
4586             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
4587                 pctx->blk = blk;
4588                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
4589                 if ((ctx->options & E2F_OPT_NO) == 0)
4590                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4591         }
4592
4593         if (problem) {
4594                 p->num_illegal_blocks++;
4595                 /*
4596                  * A bit of subterfuge here -- we're trying to fix a block
4597                  * mapping, but the IND/DIND/TIND block could have collided
4598                  * with some critical metadata.  So, fix the in-core mapping so
4599                  * iterate won't go insane, but return 0 instead of
4600                  * BLOCK_CHANGED so that it won't write the remapping out to
4601                  * our multiply linked block.
4602                  *
4603                  * Even if we previously determined that an *IND block
4604                  * conflicts with critical metadata, we must still try to
4605                  * iterate the *IND block as if it is an *IND block to find and
4606                  * mark the blocks it points to.  Better to be overly cautious
4607                  * with the used_blocks map so that we don't move the *IND
4608                  * block to a block that's really in use!
4609                  */
4610                 if (p->ino != EXT2_RESIZE_INO &&
4611                     ref_block != 0 &&
4612                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4613                                               ref_block)) {
4614                         *block_nr = 0;
4615                         return 0;
4616                 }
4617                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
4618                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
4619                                 p->clear = 1;
4620                                 return BLOCK_ABORT;
4621                         }
4622                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
4623                                 p->suppress = 1;
4624                                 set_latch_flags(PR_LATCH_BLOCK,
4625                                                 PRL_SUPPRESS, 0);
4626                         }
4627                 }
4628                 pctx->blk = blk;
4629                 pctx->blkcount = blockcnt;
4630                 if (fix_problem(ctx, problem, pctx)) {
4631                         blk = *block_nr = 0;
4632                         ret_code = BLOCK_CHANGED;
4633                         p->inode_modified = 1;
4634                         /*
4635                          * If the directory block is too big and is beyond the
4636                          * end of the FS, don't bother trying to add it for
4637                          * processing -- the kernel would never have created a
4638                          * directory this large, and we risk an ENOMEM abort.
4639                          * In any case, the toobig handler for extent-based
4640                          * directories also doesn't feed toobig blocks to
4641                          * pass 2.
4642                          */
4643                         if (problem == PR_1_TOOBIG_DIR)
4644                                 return ret_code;
4645                         goto mark_dir;
4646                 } else
4647                         return 0;
4648         }
4649
4650         if (p->ino == EXT2_RESIZE_INO) {
4651                 /*
4652                  * The resize inode has already be sanity checked
4653                  * during pass #0 (the superblock checks).  All we
4654                  * have to do is mark the double indirect block as
4655                  * being in use; all of the other blocks are handled
4656                  * by mark_table_blocks()).
4657                  */
4658                 if (blockcnt == BLOCK_COUNT_DIND)
4659                         mark_block_used(ctx, blk);
4660                 p->num_blocks++;
4661         } else if (!(ctx->fs->cluster_ratio_bits &&
4662                      p->previous_block &&
4663                      (EXT2FS_B2C(ctx->fs, blk) ==
4664                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
4665                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
4666                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
4667                 mark_block_used(ctx, blk);
4668                 p->num_blocks++;
4669         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
4670                                              p->last_block, blk, blockcnt)) {
4671                 pctx->blk = blockcnt;
4672                 pctx->blk2 = blk;
4673                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4674                 mark_block_used(ctx, blk);
4675                 mark_block_used(ctx, blk);
4676         }
4677         if (blockcnt >= 0)
4678                 p->last_block = blockcnt;
4679         p->previous_block = blk;
4680 mark_dir:
4681         if (p->is_dir && (blockcnt >= 0)) {
4682                 while (++p->last_db_block < blockcnt) {
4683                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
4684                                                               p->ino, 0,
4685                                                               p->last_db_block);
4686                         if (pctx->errcode) {
4687                                 pctx->blk = 0;
4688                                 pctx->num = p->last_db_block;
4689                                 goto failed_add_dir_block;
4690                         }
4691                 }
4692                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
4693                                                       blk, blockcnt);
4694                 if (pctx->errcode) {
4695                         pctx->blk = blk;
4696                         pctx->num = blockcnt;
4697                 failed_add_dir_block:
4698                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4699                         /* Should never get here */
4700                         ctx->flags |= E2F_FLAG_ABORT;
4701                         return BLOCK_ABORT;
4702                 }
4703         }
4704         return ret_code;
4705 }
4706
4707 static int process_bad_block(ext2_filsys fs,
4708                       blk64_t *block_nr,
4709                       e2_blkcnt_t blockcnt,
4710                       blk64_t ref_block EXT2FS_ATTR((unused)),
4711                       int ref_offset EXT2FS_ATTR((unused)),
4712                       void *priv_data)
4713 {
4714         struct process_block_struct *p;
4715         blk64_t         blk = *block_nr;
4716         blk64_t         first_block;
4717         dgrp_t          i;
4718         struct problem_context *pctx;
4719         e2fsck_t        ctx;
4720
4721         if (!blk)
4722                 return 0;
4723
4724         p = (struct process_block_struct *) priv_data;
4725         ctx = p->ctx;
4726         pctx = p->pctx;
4727
4728         pctx->ino = EXT2_BAD_INO;
4729         pctx->blk = blk;
4730         pctx->blkcount = blockcnt;
4731
4732         if ((blk < fs->super->s_first_data_block) ||
4733             (blk >= ext2fs_blocks_count(fs->super))) {
4734                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
4735                         *block_nr = 0;
4736                         return BLOCK_CHANGED;
4737                 } else
4738                         return 0;
4739         }
4740
4741         if (blockcnt < 0) {
4742                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
4743                         p->bbcheck = 1;
4744                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
4745                                 *block_nr = 0;
4746                                 return BLOCK_CHANGED;
4747                         }
4748                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4749                                                     blk)) {
4750                         p->bbcheck = 1;
4751                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
4752                                         pctx)) {
4753                                 *block_nr = 0;
4754                                 return BLOCK_CHANGED;
4755                         }
4756                         if (e2fsck_should_abort(ctx))
4757                                 return BLOCK_ABORT;
4758                 } else
4759                         mark_block_used(ctx, blk);
4760                 return 0;
4761         }
4762 #if 0
4763         printf ("DEBUG: Marking %u as bad.\n", blk);
4764 #endif
4765         ctx->fs_badblocks_count++;
4766         /*
4767          * If the block is not used, then mark it as used and return.
4768          * If it is already marked as found, this must mean that
4769          * there's an overlap between the filesystem table blocks
4770          * (bitmaps and inode table) and the bad block list.
4771          */
4772         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
4773                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4774                 return 0;
4775         }
4776         /*
4777          * Try to find the where the filesystem block was used...
4778          */
4779         first_block = fs->super->s_first_data_block;
4780
4781         for (i = 0; i < fs->group_desc_count; i++ ) {
4782                 pctx->group = i;
4783                 pctx->blk = blk;
4784                 if (!ext2fs_bg_has_super(fs, i))
4785                         goto skip_super;
4786                 if (blk == first_block) {
4787                         if (i == 0) {
4788                                 if (fix_problem(ctx,
4789                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
4790                                                 pctx)) {
4791                                         *block_nr = 0;
4792                                         return BLOCK_CHANGED;
4793                                 }
4794                                 return 0;
4795                         }
4796                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
4797                         return 0;
4798                 }
4799                 if ((blk > first_block) &&
4800                     (blk <= first_block + fs->desc_blocks)) {
4801                         if (i == 0) {
4802                                 pctx->blk = *block_nr;
4803                                 if (fix_problem(ctx,
4804                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
4805                                         *block_nr = 0;
4806                                         return BLOCK_CHANGED;
4807                                 }
4808                                 return 0;
4809                         }
4810                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
4811                         return 0;
4812                 }
4813         skip_super:
4814                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
4815                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
4816                                 ctx->invalid_block_bitmap_flag[i]++;
4817                                 ctx->invalid_bitmaps++;
4818                         }
4819                         return 0;
4820                 }
4821                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
4822                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
4823                                 ctx->invalid_inode_bitmap_flag[i]++;
4824                                 ctx->invalid_bitmaps++;
4825                         }
4826                         return 0;
4827                 }
4828                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
4829                     (blk < (ext2fs_inode_table_loc(fs, i) +
4830                             fs->inode_blocks_per_group))) {
4831                         /*
4832                          * If there are bad blocks in the inode table,
4833                          * the inode scan code will try to do
4834                          * something reasonable automatically.
4835                          */
4836                         return 0;
4837                 }
4838                 first_block += fs->super->s_blocks_per_group;
4839         }
4840         /*
4841          * If we've gotten to this point, then the only
4842          * possibility is that the bad block inode meta data
4843          * is using a bad block.
4844          */
4845         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
4846             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
4847             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
4848                 p->bbcheck = 1;
4849                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
4850                         *block_nr = 0;
4851                         return BLOCK_CHANGED;
4852                 }
4853                 if (e2fsck_should_abort(ctx))
4854                         return BLOCK_ABORT;
4855                 return 0;
4856         }
4857
4858         pctx->group = -1;
4859
4860         /* Warn user that the block wasn't claimed */
4861         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
4862
4863         return 0;
4864 }
4865
4866 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
4867                             const char *name, int num, blk64_t *new_block)
4868 {
4869         ext2_filsys fs = ctx->fs;
4870         dgrp_t          last_grp;
4871         blk64_t         old_block = *new_block;
4872         blk64_t         last_block;
4873         dgrp_t          flexbg;
4874         unsigned        flexbg_size;
4875         int             i, is_flexbg;
4876         char            *buf;
4877         struct problem_context  pctx;
4878
4879         clear_problem_context(&pctx);
4880
4881         pctx.group = group;
4882         pctx.blk = old_block;
4883         pctx.str = name;
4884
4885         /*
4886          * For flex_bg filesystems, first try to allocate the metadata
4887          * within the flex_bg, and if that fails then try finding the
4888          * space anywhere in the filesystem.
4889          */
4890         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
4891         if (is_flexbg) {
4892                 flexbg_size = 1U << fs->super->s_log_groups_per_flex;
4893                 flexbg = group / flexbg_size;
4894                 first_block = ext2fs_group_first_block2(fs,
4895                                                         flexbg_size * flexbg);
4896                 last_grp = group | (flexbg_size - 1);
4897                 if (last_grp >= fs->group_desc_count)
4898                         last_grp = fs->group_desc_count - 1;
4899                 last_block = ext2fs_group_last_block2(fs, last_grp);
4900         } else
4901                 last_block = ext2fs_group_last_block2(fs, group);
4902         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
4903                                                num, ctx->block_found_map,
4904                                                new_block);
4905         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
4906                 pctx.errcode = ext2fs_get_free_blocks2(fs,
4907                                 fs->super->s_first_data_block,
4908                                 ext2fs_blocks_count(fs->super),
4909                                 num, ctx->block_found_map, new_block);
4910         if (pctx.errcode) {
4911                 pctx.num = num;
4912                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4913                 ext2fs_unmark_valid(fs);
4914                 ctx->flags |= E2F_FLAG_ABORT;
4915                 return;
4916         }
4917         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4918         if (pctx.errcode) {
4919                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4920                 ext2fs_unmark_valid(fs);
4921                 ctx->flags |= E2F_FLAG_ABORT;
4922                 return;
4923         }
4924         ext2fs_mark_super_dirty(fs);
4925         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4926         pctx.blk2 = *new_block;
4927         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4928                           PR_1_RELOC_TO), &pctx);
4929         pctx.blk2 = 0;
4930         for (i = 0; i < num; i++) {
4931                 pctx.blk = i;
4932                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4933                 if (old_block) {
4934                         pctx.errcode = io_channel_read_blk64(fs->io,
4935                                    old_block + i, 1, buf);
4936                         if (pctx.errcode)
4937                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4938                         pctx.blk = (*new_block) + i;
4939                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4940                                                               1, buf);
4941                 } else {
4942                         pctx.blk = (*new_block) + i;
4943                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4944                                                            NULL, NULL);
4945                 }
4946
4947                 if (pctx.errcode)
4948                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4949         }
4950         ext2fs_free_mem(&buf);
4951 }
4952
4953 /*
4954  * This routine gets called at the end of pass 1 if bad blocks are
4955  * detected in the superblock, group descriptors, inode_bitmaps, or
4956  * block bitmaps.  At this point, all of the blocks have been mapped
4957  * out, so we can try to allocate new block(s) to replace the bad
4958  * blocks.
4959  */
4960 static void handle_fs_bad_blocks(e2fsck_t ctx)
4961 {
4962         ext2_filsys fs = ctx->fs;
4963         dgrp_t          i;
4964         blk64_t         first_block;
4965         blk64_t         new_blk;
4966
4967         for (i = 0; i < fs->group_desc_count; i++) {
4968                 first_block = ext2fs_group_first_block2(fs, i);
4969
4970                 if (ctx->invalid_block_bitmap_flag[i]) {
4971                         new_blk = ext2fs_block_bitmap_loc(fs, i);
4972                         new_table_block(ctx, first_block, i, _("block bitmap"),
4973                                         1, &new_blk);
4974                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4975                 }
4976                 if (ctx->invalid_inode_bitmap_flag[i]) {
4977                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
4978                         new_table_block(ctx, first_block, i, _("inode bitmap"),
4979                                         1, &new_blk);
4980                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4981                 }
4982                 if (ctx->invalid_inode_table_flag[i]) {
4983                         new_blk = ext2fs_inode_table_loc(fs, i);
4984                         new_table_block(ctx, first_block, i, _("inode table"),
4985                                         fs->inode_blocks_per_group,
4986                                         &new_blk);
4987                         ext2fs_inode_table_loc_set(fs, i, new_blk);
4988                         ctx->flags |= E2F_FLAG_RESTART;
4989                 }
4990         }
4991         ctx->invalid_bitmaps = 0;
4992 }
4993
4994 /*
4995  * This routine marks all blocks which are used by the superblock,
4996  * group descriptors, inode bitmaps, and block bitmaps.
4997  */
4998 static void mark_table_blocks(e2fsck_t ctx)
4999 {
5000         ext2_filsys fs = ctx->fs;
5001         blk64_t b;
5002         dgrp_t  i;
5003         unsigned int    j;
5004         struct problem_context pctx;
5005
5006         clear_problem_context(&pctx);
5007
5008         for (i = 0; i < fs->group_desc_count; i++) {
5009                 pctx.group = i;
5010
5011                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
5012                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
5013
5014                 /*
5015                  * Mark the blocks used for the inode table
5016                  */
5017                 if (ext2fs_inode_table_loc(fs, i)) {
5018                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
5019                              j < fs->inode_blocks_per_group;
5020                              j++, b++) {
5021                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5022                                                              b)) {
5023                                         pctx.blk = b;
5024                                         if (!ctx->invalid_inode_table_flag[i] &&
5025                                             fix_problem(ctx,
5026                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
5027                                                 ctx->invalid_inode_table_flag[i]++;
5028                                                 ctx->invalid_bitmaps++;
5029                                         }
5030                                 } else {
5031                                     ext2fs_mark_block_bitmap2(
5032                                                 ctx->block_found_map, b);
5033                                     ext2fs_mark_block_bitmap2(
5034                                                 ctx->block_metadata_map, b);
5035                                 }
5036                         }
5037                 }
5038
5039                 /*
5040                  * Mark block used for the block bitmap
5041                  */
5042                 if (ext2fs_block_bitmap_loc(fs, i)) {
5043                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5044                                      ext2fs_block_bitmap_loc(fs, i))) {
5045                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
5046                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
5047                                         ctx->invalid_block_bitmap_flag[i]++;
5048                                         ctx->invalid_bitmaps++;
5049                                 }
5050                         } else {
5051                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
5052                                      ext2fs_block_bitmap_loc(fs, i));
5053                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5054                                      ext2fs_block_bitmap_loc(fs, i));
5055                         }
5056                 }
5057                 /*
5058                  * Mark block used for the inode bitmap
5059                  */
5060                 if (ext2fs_inode_bitmap_loc(fs, i)) {
5061                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5062                                      ext2fs_inode_bitmap_loc(fs, i))) {
5063                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
5064                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
5065                                         ctx->invalid_inode_bitmap_flag[i]++;
5066                                         ctx->invalid_bitmaps++;
5067                                 }
5068                         } else {
5069                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5070                                      ext2fs_inode_bitmap_loc(fs, i));
5071                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
5072                                      ext2fs_inode_bitmap_loc(fs, i));
5073                         }
5074                 }
5075         }
5076 }
5077
5078 /*
5079  * These subroutines short circuits ext2fs_get_blocks and
5080  * ext2fs_check_directory; we use them since we already have the inode
5081  * structure, so there's no point in letting the ext2fs library read
5082  * the inode again.
5083  */
5084 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
5085                                   blk_t *blocks)
5086 {
5087         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5088         int     i;
5089
5090         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5091                 return EXT2_ET_CALLBACK_NOTHANDLED;
5092
5093         for (i=0; i < EXT2_N_BLOCKS; i++)
5094                 blocks[i] = ctx->stashed_inode->i_block[i];
5095         return 0;
5096 }
5097
5098 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
5099                                   struct ext2_inode *inode)
5100 {
5101         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5102
5103         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5104                 return EXT2_ET_CALLBACK_NOTHANDLED;
5105         *inode = *ctx->stashed_inode;
5106         return 0;
5107 }
5108
5109 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
5110                             struct ext2_inode *inode)
5111 {
5112         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5113
5114         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
5115                 (inode != ctx->stashed_inode))
5116                 *ctx->stashed_inode = *inode;
5117         return EXT2_ET_CALLBACK_NOTHANDLED;
5118 }
5119
5120 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
5121 {
5122         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5123
5124         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5125                 return EXT2_ET_CALLBACK_NOTHANDLED;
5126
5127         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
5128                 return EXT2_ET_NO_DIRECTORY;
5129         return 0;
5130 }
5131
5132 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
5133                                         blk64_t *ret)
5134 {
5135         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5136         errcode_t       retval;
5137         blk64_t         new_block;
5138
5139         if (ctx->block_found_map) {
5140                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
5141                                            &new_block);
5142                 if (retval)
5143                         return retval;
5144                 if (fs->block_map) {
5145                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
5146                         ext2fs_mark_bb_dirty(fs);
5147                 }
5148         } else {
5149                 if (!fs->block_map) {
5150                         retval = ext2fs_read_block_bitmap(fs);
5151                         if (retval)
5152                                 return retval;
5153                 }
5154
5155                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
5156                 if (retval)
5157                         return retval;
5158         }
5159
5160         *ret = new_block;
5161         return (0);
5162 }
5163
5164 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
5165                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
5166 {
5167         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5168         errcode_t       retval;
5169
5170         if (ctx->block_found_map)
5171                 return ext2fs_new_range(fs, flags, goal, len,
5172                                         ctx->block_found_map, pblk, plen);
5173
5174         if (!fs->block_map) {
5175                 retval = ext2fs_read_block_bitmap(fs);
5176                 if (retval)
5177                         return retval;
5178         }
5179
5180         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
5181                                 pblk, plen);
5182 }
5183
5184 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
5185 {
5186         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5187
5188         /* Never free a critical metadata block */
5189         if (ctx->block_found_map &&
5190             ctx->block_metadata_map &&
5191             inuse < 0 &&
5192             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
5193                 return;
5194
5195         if (ctx->block_found_map) {
5196                 if (inuse > 0)
5197                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
5198                 else
5199                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
5200         }
5201 }
5202
5203 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
5204                                            blk_t num, int inuse)
5205 {
5206         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5207
5208         /* Never free a critical metadata block */
5209         if (ctx->block_found_map &&
5210             ctx->block_metadata_map &&
5211             inuse < 0 &&
5212             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
5213                 return;
5214
5215         if (ctx->block_found_map) {
5216                 if (inuse > 0)
5217                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
5218                                                         blk, num);
5219                 else
5220                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
5221                                                         blk, num);
5222         }
5223 }
5224
5225 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
5226 {
5227         ext2_filsys fs = ctx->fs;
5228
5229         if (use_shortcuts) {
5230                 fs->get_blocks = pass1_get_blocks;
5231                 fs->check_directory = pass1_check_directory;
5232                 fs->read_inode = pass1_read_inode;
5233                 fs->write_inode = pass1_write_inode;
5234                 ctx->stashed_ino = 0;
5235         } else {
5236                 fs->get_blocks = 0;
5237                 fs->check_directory = 0;
5238                 fs->read_inode = 0;
5239                 fs->write_inode = 0;
5240         }
5241 }
5242
5243 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
5244 {
5245         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
5246         ext2fs_set_block_alloc_stats_callback(ctx->fs,
5247                                                 e2fsck_block_alloc_stats, 0);
5248         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
5249         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
5250                                         e2fsck_block_alloc_stats_range, NULL);
5251 }