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, ino_t ino);
  89 static void add_casefolded_dir(e2fsck_t ctx, 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;
 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_entry2(ctx->fs, entry, NULL, &hash);
 352         if (retval) {
 353                 com_err("check_large_ea_inode", retval,
 354                         _("while hashing entry with e_value_inum = %u"),
 355                         entry->e_value_inum);
 356                 fatal_error(ctx, 0);
 357         }
 358
 359         if (hash == entry->e_hash) {
 360                 *quota_blocks = size_to_quota_blocks(ctx->fs,
 361                                                      entry->e_value_size);
 362         } else {
 363                 /* This might be an old Lustre-style ea_inode reference. */
 364                 if (inode.i_mtime == pctx->ino &&
 365                     inode.i_generation == pctx->inode->i_generation) {
 366                         *quota_blocks = 0;
 367                 } else {
 368                         /* If target inode is also missing EA_INODE flag,
 369                          * this is likely to be a bad reference.
 370                          */
 371                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 372                                 pctx->num = entry->e_value_inum;
 373                                 return PR_1_ATTR_VALUE_EA_INODE;
 374                         } else {
 375                                 pctx->num = entry->e_hash;
 376                                 return PR_1_ATTR_HASH;
 377                         }
 378                 }
 379         }
 380
 381         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 382                 pctx->num = entry->e_value_inum;
 383                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
 384                         inode.i_flags |= EXT4_EA_INODE_FL;
 385                         e2fsck_pass1_fix_lock(ctx);
 386                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
 387                                            &inode);
 388                         e2fsck_pass1_fix_unlock(ctx);
 389                 } else {
 390                         return PR_1_ATTR_NO_EA_INODE_FL;
 391                 }
 392         }
 393         return 0;
 394 }
 395
 396 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
 397                               struct ext2_ext_attr_entry *first, void *end)
 398 {
 399         struct ext2_ext_attr_entry *entry;
 400
 401         for (entry = first;
 402              (void *)entry < end && !EXT2_EXT_IS_LAST_ENTRY(entry);
 403              entry = EXT2_EXT_ATTR_NEXT(entry)) {
 404                 if (!entry->e_value_inum)
 405                         continue;
 406                 if (!ctx->ea_inode_refs) {
 407                         pctx->errcode = ea_refcount_create(0,
 408                                                            &ctx->ea_inode_refs);
 409                         if (pctx->errcode) {
 410                                 pctx->num = 4;
 411                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
 412                                 ctx->flags |= E2F_FLAG_ABORT;
 413                                 return;
 414                         }
 415                 }
 416                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
 417                                       0);
 418         }
 419 }
 420
 421 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
 422                               struct ea_quota *ea_ibody_quota)
 423 {
 424         struct ext2_super_block *sb = ctx->fs->super;
 425         struct ext2_inode_large *inode;
 426         struct ext2_ext_attr_entry *entry;
 427         char *start, *header, *end;
 428         unsigned int storage_size, remain;
 429         problem_t problem = 0;
 430         region_t region = 0;
 431
 432         ea_ibody_quota->blocks = 0;
 433         ea_ibody_quota->inodes = 0;
 434
 435         inode = (struct ext2_inode_large *) pctx->inode;
 436         storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
 437                 inode->i_extra_isize;
 438         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 439                  inode->i_extra_isize;
 440         end = header + storage_size;
 441         start = header + sizeof(__u32);
 442         entry = (struct ext2_ext_attr_entry *) start;
 443
 444         /* scan all entry's headers first */
 445
 446         /* take finish entry 0UL into account */
 447         remain = storage_size - sizeof(__u32);
 448
 449         region = region_create(0, storage_size);
 450         if (!region) {
 451                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
 452                 problem = 0;
 453                 ctx->flags |= E2F_FLAG_ABORT;
 454                 return;
 455         }
 456         if (region_allocate(region, 0, sizeof(__u32))) {
 457                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 458                 goto fix;
 459         }
 460
 461         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
 462                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
 463                 __u32 hash;
 464
 465                 if (region_allocate(region, (char *)entry - (char *)header,
 466                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
 467                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
 468                         goto fix;
 469                 }
 470
 471                 /* header eats this space */
 472                 remain -= sizeof(struct ext2_ext_attr_entry);
 473
 474                 /* is attribute name valid? */
 475                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
 476                         pctx->num = entry->e_name_len;
 477                         problem = PR_1_ATTR_NAME_LEN;
 478                         goto fix;
 479                 }
 480
 481                 /* attribute len eats this space */
 482                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
 483
 484                 if (entry->e_value_inum == 0) {
 485                         /* check value size */
 486                         if (entry->e_value_size > remain) {
 487                                 pctx->num = entry->e_value_size;
 488                                 problem = PR_1_ATTR_VALUE_SIZE;
 489                                 goto fix;
 490                         }
 491
 492                         if (entry->e_value_size &&
 493                             region_allocate(region,
 494                                             sizeof(__u32) + entry->e_value_offs,
 495                                             EXT2_EXT_ATTR_SIZE(
 496                                                 entry->e_value_size))) {
 497                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 498                                 goto fix;
 499                         }
 500
 501                         hash = ext2fs_ext_attr_hash_entry(entry,
 502                                                           start + entry->e_value_offs);
 503
 504                         /* e_hash may be 0 in older inode's ea */
 505                         if (entry->e_hash != 0 && entry->e_hash != hash) {
 506                                 pctx->num = entry->e_hash;
 507                                 problem = PR_1_ATTR_HASH;
 508                                 goto fix;
 509                         }
 510                 } else {
 511                         blk64_t quota_blocks;
 512
 513                         problem = check_large_ea_inode(ctx, entry, pctx,
 514                                                        &quota_blocks);
 515                         if (problem != 0)
 516                                 goto fix;
 517
 518                         ea_ibody_quota->blocks += quota_blocks;
 519                         ea_ibody_quota->inodes++;
 520                 }
 521
 522                 /* If EA value is stored in external inode then it does not
 523                  * consume space here */
 524                 if (entry->e_value_inum == 0)
 525                         remain -= entry->e_value_size;
 526
 527                 entry = EXT2_EXT_ATTR_NEXT(entry);
 528         }
 529
 530         if (region_allocate(region, (char *)entry - (char *)header,
 531                             sizeof(__u32))) {
 532                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 533                 goto fix;
 534         }
 535 fix:
 536         if (region)
 537                 region_free(region);
 538         /*
 539          * it seems like a corruption. it's very unlikely we could repair
 540          * EA(s) in automatic fashion -bzzz
 541          */
 542         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
 543                 inc_ea_inode_refs(ctx, pctx,
 544                                   (struct ext2_ext_attr_entry *)start, end);
 545                 return;
 546         }
 547
 548         /* simply remove all possible EA(s) */
 549         *((__u32 *)header) = 0UL;
 550         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 551                                 EXT2_INODE_SIZE(sb), "pass1");
 552         ea_ibody_quota->blocks = 0;
 553         ea_ibody_quota->inodes = 0;
 554 }
 555
 556 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
 557         return (xtime & (1U << 31)) != 0 &&
 558                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
 559 }
 560
 561 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
 562         check_inode_extra_negative_epoch(inode->i_##xtime, \
 563                                          inode->i_##xtime##_extra)
 564
 565 /* When today's date is earlier than 2242, we assume that atimes,
 566  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
 567  * actually pre-1970 dates mis-encoded.
 568  */
 569 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
 570
 571 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
 572                                     struct ea_quota *ea_ibody_quota)
 573 {
 574         struct ext2_super_block *sb = ctx->fs->super;
 575         struct ext2_inode_large *inode;
 576         __u32 *eamagic;
 577         int min, max;
 578
 579         ea_ibody_quota->blocks = 0;
 580         ea_ibody_quota->inodes = 0;
 581
 582         inode = (struct ext2_inode_large *) pctx->inode;
 583         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
 584                 /* this isn't large inode. so, nothing to check */
 585                 return;
 586         }
 587
 588 #if 0
 589         printf("inode #%u, i_extra_size %d\n", pctx->ino,
 590                         inode->i_extra_isize);
 591 #endif
 592         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
 593         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
 594         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
 595         /*
 596          * For now we will allow i_extra_isize to be 0, but really
 597          * implementations should never allow i_extra_isize to be 0
 598          */
 599         if (inode->i_extra_isize &&
 600             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
 601              inode->i_extra_isize & 3)) {
 602                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
 603                         return;
 604                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
 605                         inode->i_extra_isize = sb->s_want_extra_isize;
 606                 else
 607                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
 608                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 609                                         EXT2_INODE_SIZE(sb), "pass1");
 610         }
 611
 612         /* check if there is no place for an EA header */
 613         if (inode->i_extra_isize >= max - sizeof(__u32))
 614                 return;
 615
 616         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 617                         inode->i_extra_isize);
 618         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
 619                 /* it seems inode has an extended attribute(s) in body */
 620                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
 621         }
 622
 623         /*
 624          * If the inode's extended atime (ctime, crtime, mtime) is stored in
 625          * the old, invalid format, repair it.
 626          */
 627         if (((sizeof(time_t) <= 4) ||
 628              (((sizeof(time_t) > 4) &&
 629                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
 630             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
 631              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
 632              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
 633              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
 634
 635                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
 636                         return;
 637
 638                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
 639                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
 640                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
 641                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
 642                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
 643                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
 644                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
 645                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
 646                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 647                                         EXT2_INODE_SIZE(sb), "pass1");
 648         }
 649
 650 }
 651
 652 static _INLINE_ int is_blocks_used(e2fsck_t ctx, blk64_t block,
 653                                    unsigned int num)
 654 {
 655         int retval;
 656
 657         /* used to avoid duplicate output from below */
 658         retval = ext2fs_test_block_bitmap_range2_valid(ctx->block_found_map,
 659                                                        block, num);
 660         if (!retval)
 661                 return 0;
 662
 663         retval = ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num);
 664         if (retval) {
 665                 e2fsck_pass1_block_map_r_lock(ctx);
 666                 if (ctx->global_ctx)
 667                         retval = ext2fs_test_block_bitmap_range2(
 668                                         ctx->global_ctx->block_found_map, block, num);
 669                 e2fsck_pass1_block_map_r_unlock(ctx);
 670                 if (retval)
 671                         return 0;
 672         }
 673
 674         return 1;
 675 }
 676
 677 /*
 678  * Check to see if the inode might really be a directory, despite i_mode
 679  *
 680  * This is a lot of complexity for something for which I'm not really
 681  * convinced happens frequently in the wild.  If for any reason this
 682  * causes any problems, take this code out.
 683  * [tytso:20070331.0827EDT]
 684  */
 685 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
 686                                 char *buf)
 687 {
 688         struct ext2_inode *inode = pctx->inode;
 689         struct ext2_dir_entry   *dirent;
 690         errcode_t               retval;
 691         blk64_t                 blk;
 692         unsigned int            i, rec_len, not_device = 0;
 693         int                     extent_fs;
 694         int                     inlinedata_fs;
 695
 696         /*
 697          * If the mode looks OK, we believe it.  If the first block in
 698          * the i_block array is 0, this cannot be a directory. If the
 699          * inode is extent-mapped, it is still the case that the latter
 700          * cannot be 0 - the magic number in the extent header would make
 701          * it nonzero.
 702          */
 703         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
 704             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
 705                 return;
 706
 707         /*
 708          * Check the block numbers in the i_block array for validity:
 709          * zero blocks are skipped (but the first one cannot be zero -
 710          * see above), other blocks are checked against the first and
 711          * max data blocks (from the the superblock) and against the
 712          * block bitmap. Any invalid block found means this cannot be
 713          * a directory.
 714          *
 715          * If there are non-zero blocks past the fourth entry, then
 716          * this cannot be a device file: we remember that for the next
 717          * check.
 718          *
 719          * For extent mapped files, we don't do any sanity checking:
 720          * just try to get the phys block of logical block 0 and run
 721          * with it.
 722          *
 723          * For inline data files, we just try to get the size of inline
 724          * data.  If it's true, we will treat it as a directory.
 725          */
 726
 727         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
 728         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
 729         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
 730                 size_t size;
 731                 __u32 dotdot;
 732                 unsigned int rec_len2;
 733                 struct ext2_dir_entry de;
 734
 735                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
 736                         return;
 737                 /*
 738                  * If the size isn't a multiple of 4, it's probably not a
 739                  * directory??
 740                  */
 741                 if (size & 3)
 742                         return;
 743                 /*
 744                  * If the first 10 bytes don't look like a directory entry,
 745                  * it's probably not a directory.
 746                  */
 747                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
 748                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
 749                        EXT2_DIR_REC_LEN(0));
 750                 dotdot = ext2fs_le32_to_cpu(dotdot);
 751                 de.inode = ext2fs_le32_to_cpu(de.inode);
 752                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
 753                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
 754                 if (dotdot >= ctx->fs->super->s_inodes_count ||
 755                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
 756                      dotdot != EXT2_ROOT_INO) ||
 757                     de.inode >= ctx->fs->super->s_inodes_count ||
 758                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
 759                      de.inode != 0) ||
 760                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
 761                               EXT4_INLINE_DATA_DOTDOT_SIZE)
 762                         return;
 763                 /* device files never have a "system.data" entry */
 764                 goto isdir;
 765         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
 766                 /* extent mapped */
 767                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
 768                                  &blk))
 769                         return;
 770                 /* device files are never extent mapped */
 771                 not_device++;
 772         } else {
 773                 for (i=0; i < EXT2_N_BLOCKS; i++) {
 774                         blk = inode->i_block[i];
 775                         if (!blk)
 776                                 continue;
 777                         if (i >= 4)
 778                                 not_device++;
 779
 780                         if (blk < ctx->fs->super->s_first_data_block ||
 781                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
 782                             is_blocks_used(ctx, blk, 1))
 783                                 return; /* Invalid block, can't be dir */
 784                 }
 785                 blk = inode->i_block[0];
 786         }
 787
 788         /*
 789          * If the mode says this is a device file and the i_links_count field
 790          * is sane and we have not ruled it out as a device file previously,
 791          * we declare it a device file, not a directory.
 792          */
 793         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
 794             (inode->i_links_count == 1) && !not_device)
 795                 return;
 796
 797         /* read the first block */
 798         ehandler_operation(_("reading directory block"));
 799         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
 800         ehandler_operation(0);
 801         if (retval)
 802                 return;
 803
 804         dirent = (struct ext2_dir_entry *) buf;
 805         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 806         if (retval)
 807                 return;
 808         if ((ext2fs_dirent_name_len(dirent) != 1) ||
 809             (dirent->name[0] != '.') ||
 810             (dirent->inode != pctx->ino) ||
 811             (rec_len < 12) ||
 812             (rec_len % 4) ||
 813             (rec_len >= ctx->fs->blocksize - 12))
 814                 return;
 815
 816         dirent = (struct ext2_dir_entry *) (buf + rec_len);
 817         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 818         if (retval)
 819                 return;
 820         if ((ext2fs_dirent_name_len(dirent) != 2) ||
 821             (dirent->name[0] != '.') ||
 822             (dirent->name[1] != '.') ||
 823             (rec_len < 12) ||
 824             (rec_len % 4))
 825                 return;
 826
 827 isdir:
 828         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
 829                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
 830                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
 831                                         EXT2_INODE_SIZE(ctx->fs->super),
 832                                         "check_is_really_dir");
 833         }
 834 }
 835
 836 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
 837                                      int flags, ext2_icount_t hint,
 838                                      ext2_icount_t *ret)
 839 {
 840         unsigned int            threshold;
 841         unsigned int            save_type;
 842         ext2_ino_t              num_dirs;
 843         errcode_t               retval;
 844         char                    *tdb_dir;
 845         int                     enable;
 846
 847         *ret = 0;
 848
 849         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
 850                            &tdb_dir);
 851         profile_get_uint(ctx->profile, "scratch_files",
 852                          "numdirs_threshold", 0, 0, &threshold);
 853         profile_get_boolean(ctx->profile, "scratch_files",
 854                             "icount", 0, 1, &enable);
 855
 856         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
 857         if (retval)
 858                 num_dirs = 1024;        /* Guess */
 859
 860         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
 861             (!threshold || num_dirs > threshold)) {
 862                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
 863                                                   flags, ret);
 864                 if (retval == 0)
 865                         return 0;
 866         }
 867         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
 868                                &save_type);
 869         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
 870                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
 871         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
 872         ctx->fs->default_bitmap_type = save_type;
 873         return retval;
 874 }
 875
 876 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
 877                                             e2fsck_t ctx,
 878                                             struct problem_context *pctx)
 879 {
 880         errcode_t retval;
 881         struct ext2_inode_large inode;
 882
 883         /*
 884          * Reread inode.  If we don't see checksum error, then this inode
 885          * has been fixed elsewhere.
 886          */
 887         ctx->stashed_ino = 0;
 888         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
 889                                         sizeof(inode));
 890         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
 891                 return retval;
 892         if (!retval)
 893                 return 0;
 894
 895         /*
 896          * Checksum still doesn't match.  That implies that the inode passes
 897          * all the sanity checks, so maybe the checksum is simply corrupt.
 898          * See if the user will go for fixing that.
 899          */
 900         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
 901                 return 0;
 902
 903
 904         e2fsck_pass1_fix_lock(ctx);
 905         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
 906                                          sizeof(inode));
 907         e2fsck_pass1_fix_unlock(ctx);
 908         return retval;
 909 }
 910
 911 static void reserve_block_for_root_repair(e2fsck_t ctx)
 912 {
 913         blk64_t         blk = 0;
 914         errcode_t       err;
 915         ext2_filsys     fs = ctx->fs;
 916
 917         ctx->root_repair_block = 0;
 918         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
 919                 return;
 920
 921         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 922         if (err)
 923                 return;
 924         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 925         ctx->root_repair_block = blk;
 926 }
 927
 928 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
 929 {
 930         blk64_t         blk = 0;
 931         errcode_t       err;
 932         ext2_filsys     fs = ctx->fs;
 933         static const char name[] = "lost+found";
 934         ext2_ino_t      ino;
 935
 936         ctx->lnf_repair_block = 0;
 937         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
 938                 return;
 939
 940         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 941         if (err)
 942                 return;
 943         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 944         ctx->lnf_repair_block = blk;
 945         return;
 946 }
 947
 948 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
 949                                          size_t *sz)
 950 {
 951         void *p;
 952         struct ext2_xattr_handle *handle;
 953         errcode_t retval;
 954
 955         retval = ext2fs_xattrs_open(fs, ino, &handle);
 956         if (retval)
 957                 return retval;
 958
 959         retval = ext2fs_xattrs_read(handle);
 960         if (retval)
 961                 goto err;
 962
 963         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
 964         if (retval)
 965                 goto err;
 966         ext2fs_free_mem(&p);
 967 err:
 968         (void) ext2fs_xattrs_close(&handle);
 969         return retval;
 970 }
 971
 972 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
 973                                     struct problem_context *pctx,
 974                                     int failed_csum)
 975 {
 976         if (!failed_csum)
 977                 return;
 978
 979         /*
 980          * If the inode failed the checksum and the user didn't
 981          * clear the inode, test the checksum again -- if it still
 982          * fails, ask the user if the checksum should be corrected.
 983          */
 984         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
 985         if (pctx->errcode)
 986                 ctx->flags |= E2F_FLAG_ABORT;
 987 }
 988 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
 989         do { \
 990                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
 991                 if ((ctx)->flags & E2F_FLAG_ABORT) \
 992                         return; \
 993         } while (0)
 994
 995 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
 996 {
 997         __u32 x;
 998         int i;
 999
1000         for (i = 0; i < EXT2_N_BLOCKS; i++) {
1001                 x = inode->i_block[i];
1002 #ifdef WORDS_BIGENDIAN
1003                 x = ext2fs_swab32(x);
1004 #endif
1005                 if (x >= ext2fs_blocks_count(fs->super))
1006                         return 0;
1007         }
1008
1009         return 1;
1010 }
1011
1012 /*
1013  * Figure out what to do with an inode that has both extents and inline data
1014  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
1015  */
1016 static int fix_inline_data_extents_file(e2fsck_t ctx,
1017                                         ext2_ino_t ino,
1018                                         struct ext2_inode *inode,
1019                                         int inode_size,
1020                                         struct problem_context *pctx)
1021 {
1022         size_t max_inline_ea_size;
1023         ext2_filsys fs = ctx->fs;
1024         int dirty = 0;
1025
1026         /* Both feature flags not set?  Just run the regular checks */
1027         if (!ext2fs_has_feature_extents(fs->super) &&
1028             !ext2fs_has_feature_inline_data(fs->super))
1029                 return 0;
1030
1031         /* Clear both flags if it's a special file */
1032         if (LINUX_S_ISCHR(inode->i_mode) ||
1033             LINUX_S_ISBLK(inode->i_mode) ||
1034             LINUX_S_ISFIFO(inode->i_mode) ||
1035             LINUX_S_ISSOCK(inode->i_mode)) {
1036                 check_extents_inlinedata(ctx, pctx);
1037                 return 0;
1038         }
1039
1040         /* If it looks like an extent tree, try to clear inlinedata */
1041         if (ext2fs_extent_header_verify(inode->i_block,
1042                                  sizeof(inode->i_block)) == 0 &&
1043             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1044                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1045                 dirty = 1;
1046                 goto out;
1047         }
1048
1049         /* If it looks short enough to be inline data, try to clear extents */
1050         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1051                 max_inline_ea_size = inode_size -
1052                                      (EXT2_GOOD_OLD_INODE_SIZE +
1053                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1054         else
1055                 max_inline_ea_size = 0;
1056         if (EXT2_I_SIZE(inode) <
1057             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1058             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1059                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1060                 dirty = 1;
1061                 goto out;
1062         }
1063
1064         /*
1065          * Too big for inline data, but no evidence of extent tree -
1066          * maybe it's a block map file?  If the mappings all look valid?
1067          */
1068         if (could_be_block_map(fs, inode) &&
1069             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1070 #ifdef WORDS_BIGENDIAN
1071                 int i;
1072
1073                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1074                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1075 #endif
1076
1077                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1078                 dirty = 1;
1079                 goto out;
1080         }
1081
1082         /* Oh well, just clear the busted inode. */
1083         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1084                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1085                 return -1;
1086         }
1087
1088 out:
1089         if (dirty)
1090                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1091
1092         return 0;
1093 }
1094
1095 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1096 {
1097         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1098         dgrp_t start = *group, grp, grp_end = ctx->fs->group_desc_count;
1099         blk64_t blocks_to_read = 0;
1100         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1101
1102 #ifdef HAVE_PTHREAD
1103         if (ctx->fs->fs_num_threads > 1)
1104                 grp_end = ctx->thread_info.et_group_end;
1105 #endif
1106         if (ctx->readahead_kb == 0)
1107                 goto out;
1108
1109         /* Keep iterating groups until we have enough to readahead */
1110         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1111         for (grp = start; grp < grp_end; grp++) {
1112                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1113                         continue;
1114                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1115                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1116                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1117                                         inodes_per_block;
1118                 if (blocks_to_read * ctx->fs->blocksize >
1119                     ctx->readahead_kb * 1024)
1120                         break;
1121         }
1122
1123         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1124                                grp - start + 1);
1125         if (err == EAGAIN) {
1126                 ctx->readahead_kb /= 2;
1127                 err = 0;
1128         }
1129
1130 out:
1131         if (err) {
1132                 /* Error; disable itable readahead */
1133                 *group = ctx->fs->group_desc_count;
1134                 *next_ino = ctx->fs->super->s_inodes_count;
1135         } else {
1136                 /*
1137                  * Don't do more readahead until we've reached the first inode
1138                  * of the last inode scan buffer block for the last group.
1139                  */
1140                 *group = grp + 1;
1141                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1142                                      ctx->inode_buffer_blocks :
1143                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1144                                     ctx->fs->blocksize /
1145                                     EXT2_INODE_SIZE(ctx->fs->super);
1146                 inodes_in_group--;
1147                 *next_ino = inodes_in_group -
1148                             (inodes_in_group % inodes_per_buffer) + 1 +
1149                             (grp * ctx->fs->super->s_inodes_per_group);
1150         }
1151 }
1152
1153 /*
1154  * Check if the passed ino is one of the used superblock quota inodes.
1155  *
1156  * Before the quota inodes were journaled, older superblock quota inodes
1157  * were just regular files in the filesystem and not reserved inodes.  This
1158  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1159  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1160  */
1161 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1162 {
1163         enum quota_type qtype;
1164
1165         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1166                 if (*quota_sb_inump(sb, qtype) == ino)
1167                         return 1;
1168
1169         return 0;
1170 }
1171
1172 /*
1173  * Check if the passed ino is one of the reserved quota inodes.
1174  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1175  * inodes.  These inodes may or may not be in use by the quota feature.
1176  */
1177 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1178 {
1179         enum quota_type qtype;
1180
1181         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1182                 if (quota_type2inum(qtype, fs->super) == ino)
1183                         return 1;
1184
1185         return 0;
1186 }
1187
1188 static int e2fsck_should_abort(e2fsck_t ctx)
1189 {
1190         e2fsck_t global_ctx;
1191
1192         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1193                 return 1;
1194
1195         if (ctx->global_ctx) {
1196                 global_ctx = ctx->global_ctx;
1197                 if (global_ctx->flags & E2F_FLAG_SIGNAL_MASK)
1198                         return 1;
1199         }
1200         return 0;
1201 }
1202
1203 static void init_ext2_max_sizes()
1204 {
1205         int     i;
1206         __u64   max_sizes;
1207
1208         /*
1209          * Init ext2_max_sizes which will be immutable and shared between
1210          * threads
1211          */
1212 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1213
1214         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1215                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1216                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1217                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1218                 max_sizes = (max_sizes * (1UL << i));
1219                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1220         }
1221 #undef EXT2_BPP
1222 }
1223
1224 #ifdef HAVE_PTHREAD
1225 /* TODO: tdb needs to be handled properly for multiple threads*/
1226 static int multiple_threads_supported(e2fsck_t ctx)
1227 {
1228 #ifdef  CONFIG_TDB
1229         unsigned int            threshold;
1230         ext2_ino_t              num_dirs;
1231         errcode_t               retval;
1232         char                    *tdb_dir;
1233         int                     enable;
1234
1235         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
1236                            &tdb_dir);
1237         profile_get_uint(ctx->profile, "scratch_files",
1238                          "numdirs_threshold", 0, 0, &threshold);
1239         profile_get_boolean(ctx->profile, "scratch_files",
1240                             "icount", 0, 1, &enable);
1241
1242         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
1243         if (retval)
1244                 num_dirs = 1024;        /* Guess */
1245
1246         /* tdb is unsupported now */
1247         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
1248             (!threshold || num_dirs > threshold))
1249                 return 0;
1250 #endif
1251         return 1;
1252 }
1253
1254 /**
1255  * Even though we could specify number of threads,
1256  * but it might be more than the whole filesystem
1257  * block groups, correct it here.
1258  */
1259 static void e2fsck_pass1_set_thread_num(e2fsck_t ctx)
1260 {
1261         unsigned flexbg_size = 1;
1262         ext2_filsys fs = ctx->fs;
1263         int num_threads = ctx->fs_num_threads;
1264         int max_threads;
1265
1266         if (num_threads < 1) {
1267                 num_threads = 1;
1268                 goto out;
1269         }
1270
1271         if (!multiple_threads_supported(ctx)) {
1272                 num_threads = 1;
1273                 fprintf(stderr, "Fall through single thread for pass1 "
1274                         "because tdb could not handle properly\n");
1275                 goto out;
1276         }
1277
1278         if (ext2fs_has_feature_flex_bg(fs->super))
1279                 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
1280         max_threads = fs->group_desc_count / flexbg_size;
1281         if (max_threads == 0)
1282                 max_threads = 1;
1283
1284         if (num_threads > max_threads) {
1285                 fprintf(stderr, "Use max possible thread num: %d instead\n",
1286                                 max_threads);
1287                 num_threads = max_threads;
1288         }
1289 out:
1290         ctx->fs_num_threads = num_threads;
1291         ctx->fs->fs_num_threads = num_threads;
1292 }
1293 #endif
1294
1295 /*
1296  * We need call mark_table_blocks() before multiple
1297  * thread start, since all known system blocks should be
1298  * marked and checked later.
1299  */
1300 static errcode_t e2fsck_pass1_prepare(e2fsck_t ctx)
1301 {
1302         struct problem_context pctx;
1303         ext2_filsys fs = ctx->fs;
1304         unsigned long long readahead_kb;
1305
1306         init_ext2_max_sizes();
1307 #ifdef HAVE_PTHREAD
1308         e2fsck_pass1_set_thread_num(ctx);
1309 #endif
1310         /* If we can do readahead, figure out how many groups to pull in. */
1311         if (!e2fsck_can_readahead(ctx->fs))
1312                 ctx->readahead_kb = 0;
1313         else if (ctx->readahead_kb == ~0ULL)
1314                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1315
1316 #ifdef HAVE_PTHREAD
1317         /* don't use more than 1/10 of memory for threads checking */
1318         readahead_kb = get_memory_size() / (10 * ctx->fs_num_threads);
1319         /* maybe better disable RA if this is too small? */
1320         if (ctx->readahead_kb > readahead_kb)
1321                 ctx->readahead_kb = readahead_kb;
1322 #endif
1323         clear_problem_context(&pctx);
1324         if (!(ctx->options & E2F_OPT_PREEN))
1325                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1326
1327         pctx.errcode = e2fsck_allocate_subcluster_bitmap(ctx->fs,
1328                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1329                         "block_found_map", &ctx->block_found_map);
1330         if (pctx.errcode) {
1331                 pctx.num = 1;
1332                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1333                 ctx->flags |= E2F_FLAG_ABORT;
1334                 return pctx.errcode;
1335         }
1336         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
1337                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1338                         "block_metadata_map", &ctx->block_metadata_map);
1339         if (pctx.errcode) {
1340                 pctx.num = 1;
1341                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1342                 ctx->flags |= E2F_FLAG_ABORT;
1343                 return pctx.errcode;
1344         }
1345
1346         mark_table_blocks(ctx);
1347         pctx.errcode = ext2fs_convert_subcluster_bitmap(ctx->fs,
1348                                                 &ctx->block_found_map);
1349         if (pctx.errcode) {
1350                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1351                 ctx->flags |= E2F_FLAG_ABORT;
1352                 return pctx.errcode;
1353         }
1354
1355         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
1356                         _("multiply claimed block map"),
1357                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
1358                         &ctx->block_dup_map);
1359         if (pctx.errcode) {
1360                 pctx.num = 3;
1361                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
1362                             &pctx);
1363                 /* Should never get here */
1364                 ctx->flags |= E2F_FLAG_ABORT;
1365                 return pctx.errcode;
1366         }
1367
1368         if (ext2fs_has_feature_mmp(fs->super) &&
1369             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1370             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1371                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1372                                           fs->super->s_mmp_block);
1373 #ifdef  HAVE_PTHREAD
1374         pthread_mutex_init(&ctx->fs_fix_mutex, NULL);
1375         pthread_rwlock_init(&ctx->fs_block_map_rwlock, NULL);
1376 #endif
1377
1378         return 0;
1379 }
1380
1381 static void e2fsck_pass1_post(e2fsck_t ctx)
1382 {
1383         struct problem_context pctx;
1384         ext2_filsys fs = ctx->fs;
1385
1386         char *block_buf =
1387                 (char *)e2fsck_allocate_memory(ctx, ctx->fs->blocksize * 3,
1388                                               "block interate buffer");
1389         reserve_block_for_root_repair(ctx);
1390         reserve_block_for_lnf_repair(ctx);
1391
1392         /*
1393          * If any extended attribute blocks' reference counts need to
1394          * be adjusted, either up (ctx->refcount_extra), or down
1395          * (ctx->refcount), then fix them.
1396          */
1397         if (ctx->refcount) {
1398                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
1399                 ea_refcount_free(ctx->refcount);
1400                 ctx->refcount = 0;
1401         }
1402         if (ctx->refcount_extra) {
1403                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
1404                                         block_buf, +1);
1405                 ea_refcount_free(ctx->refcount_extra);
1406                 ctx->refcount_extra = 0;
1407         }
1408
1409         if (ctx->invalid_bitmaps)
1410                 handle_fs_bad_blocks(ctx);
1411
1412         /* We don't need the block_ea_map any more */
1413         if (ctx->block_ea_map) {
1414                 ext2fs_free_block_bitmap(ctx->block_ea_map);
1415                 ctx->block_ea_map = 0;
1416         }
1417
1418         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
1419                 struct ext2_inode *inode;
1420                 int inode_size = EXT2_INODE_SIZE(fs->super);
1421                 inode = e2fsck_allocate_memory(ctx, inode_size,
1422                                                "scratch inode");
1423
1424                 clear_problem_context(&pctx);
1425                 pctx.errcode = ext2fs_create_resize_inode(fs);
1426                 if (pctx.errcode) {
1427                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
1428                                          &pctx)) {
1429                                 ctx->flags |= E2F_FLAG_ABORT;
1430                                 ext2fs_free_mem(&inode);
1431                                 ext2fs_free_mem(&block_buf);
1432                                 return;
1433                         }
1434                         pctx.errcode = 0;
1435                 }
1436                 if (!pctx.errcode) {
1437                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
1438                                           "recreate inode");
1439                         inode->i_mtime = ctx->now;
1440                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
1441                                            "recreate inode");
1442                 }
1443                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
1444                 ext2fs_free_mem(&inode);
1445         }
1446
1447         if (ctx->flags & E2F_FLAG_RESTART) {
1448                 ext2fs_free_mem(&block_buf);
1449                 return;
1450         }
1451
1452         if (ctx->block_dup_map) {
1453                 if (!(ctx->flags & E2F_FLAG_DUP_BLOCK)) {
1454                         ext2fs_free_mem(&block_buf);
1455                         return;
1456                 }
1457                 if (ctx->options & E2F_OPT_PREEN) {
1458                         clear_problem_context(&pctx);
1459                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
1460                 }
1461                 e2fsck_pass1_dupblocks(ctx, block_buf);
1462                 ext2fs_free_mem(&block_buf);
1463                 ctx->flags &= ~E2F_FLAG_DUP_BLOCK;
1464         }
1465 }
1466
1467
1468 void e2fsck_pass1_run(e2fsck_t ctx)
1469 {
1470         int     i;
1471         ext2_filsys fs = ctx->fs;
1472         ext2_ino_t      ino = 0;
1473         struct ext2_inode *inode = NULL;
1474         ext2_inode_scan scan = NULL;
1475         char            *block_buf = NULL;
1476 #ifdef RESOURCE_TRACK
1477         struct resource_track   rtrack;
1478 #endif
1479         unsigned char   frag, fsize;
1480         struct          problem_context pctx;
1481         struct          scan_callback_struct scan_struct;
1482         struct ext2_super_block *sb = ctx->fs->super;
1483         const char      *old_op;
1484         const char      *eop_next_inode = _("getting next inode from scan");
1485         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1486         int             low_dtime_check = 1;
1487         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1488         unsigned int    bufsize;
1489         int             failed_csum = 0;
1490         ext2_ino_t      ino_threshold = 0;
1491         dgrp_t          ra_group = 0;
1492         struct ea_quota ea_ibody_quota;
1493         struct process_inode_block *inodes_to_process;
1494         int             process_inode_count;
1495
1496         init_resource_track(&rtrack, ctx->fs->io);
1497         clear_problem_context(&pctx);
1498
1499         pass1_readahead(ctx, &ra_group, &ino_threshold);
1500         if (ext2fs_has_feature_dir_index(fs->super) &&
1501             !(ctx->options & E2F_OPT_NO)) {
1502                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1503                         ctx->dirs_to_hash = 0;
1504         }
1505
1506 #ifdef MTRACE
1507         mtrace_print("Pass 1");
1508 #endif
1509
1510         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1511         extent_fs = ext2fs_has_feature_extents(sb);
1512         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1513         casefold_fs = ext2fs_has_feature_casefold(sb);
1514
1515         /*
1516          * Allocate bitmaps structures
1517          */
1518         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1519                                                     EXT2FS_BMAP64_RBTREE,
1520                                                     "inode_used_map",
1521                                                     &ctx->inode_used_map);
1522         if (pctx.errcode) {
1523                 pctx.num = 1;
1524                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1525                 ctx->flags |= E2F_FLAG_ABORT;
1526                 return;
1527         }
1528         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1529                         _("directory inode map"),
1530                         ctx->global_ctx ? EXT2FS_BMAP64_RBTREE :
1531                         EXT2FS_BMAP64_AUTODIR,
1532                         "inode_dir_map", &ctx->inode_dir_map);
1533         if (pctx.errcode) {
1534                 pctx.num = 2;
1535                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1536                 ctx->flags |= E2F_FLAG_ABORT;
1537                 return;
1538         }
1539         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1540                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1541                         "inode_reg_map", &ctx->inode_reg_map);
1542         if (pctx.errcode) {
1543                 pctx.num = 6;
1544                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1545                 ctx->flags |= E2F_FLAG_ABORT;
1546                 return;
1547         }
1548         if (casefold_fs) {
1549                 pctx.errcode =
1550                         e2fsck_allocate_inode_bitmap(fs,
1551                                                      _("inode casefold map"),
1552                                                      EXT2FS_BMAP64_RBTREE,
1553                                                      "inode_casefold_map",
1554                                                      &ctx->inode_casefold_map);
1555                 if (pctx.errcode) {
1556                         pctx.num = 1;
1557                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1558                         ctx->flags |= E2F_FLAG_ABORT;
1559                         return;
1560                 }
1561         }
1562         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1563                                            &ctx->inode_link_info);
1564         if (pctx.errcode) {
1565                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1566                 ctx->flags |= E2F_FLAG_ABORT;
1567                 return;
1568         }
1569         bufsize = inode_size;
1570         if (bufsize < sizeof(struct ext2_inode_large))
1571                 bufsize = sizeof(struct ext2_inode_large);
1572         inode = (struct ext2_inode *)
1573                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1574
1575         inodes_to_process = (struct process_inode_block *)
1576                 e2fsck_allocate_memory(ctx,
1577                                        (ctx->process_inode_size *
1578                                         sizeof(struct process_inode_block)),
1579                                        "array of inodes to process");
1580         process_inode_count = 0;
1581
1582         pctx.errcode = ext2fs_init_dblist(fs, 0);
1583         if (pctx.errcode) {
1584                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1585                 ctx->flags |= E2F_FLAG_ABORT;
1586                 goto endit;
1587         }
1588
1589         /*
1590          * If the last orphan field is set, clear it, since the pass1
1591          * processing will automatically find and clear the orphans.
1592          * In the future, we may want to try using the last_orphan
1593          * linked list ourselves, but for now, we clear it so that the
1594          * ext3 mount code won't get confused.
1595          */
1596         if (!(ctx->options & E2F_OPT_READONLY)) {
1597                 if (fs->super->s_last_orphan) {
1598                         fs->super->s_last_orphan = 0;
1599                         ext2fs_mark_super_dirty(fs);
1600                 }
1601         }
1602
1603         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1604                                                     "block interate buffer");
1605         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1606                 e2fsck_use_inode_shortcuts(ctx, 1);
1607         e2fsck_intercept_block_allocations(ctx);
1608         old_op = ehandler_operation(_("opening inode scan"));
1609         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1610                                               &scan);
1611         ehandler_operation(old_op);
1612         if (pctx.errcode) {
1613                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1614                 ctx->flags |= E2F_FLAG_ABORT;
1615                 goto endit;
1616         }
1617         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1618                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1619         ctx->stashed_inode = inode;
1620         scan_struct.ctx = ctx;
1621         scan_struct.block_buf = block_buf;
1622         scan_struct.inodes_to_process = inodes_to_process;
1623         scan_struct.process_inode_count = &process_inode_count;
1624         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1625         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1626                                               ctx->fs->group_desc_count)))
1627                 goto endit;
1628         if ((fs->super->s_wtime &&
1629              fs->super->s_wtime < fs->super->s_inodes_count) ||
1630             (fs->super->s_mtime &&
1631              fs->super->s_mtime < fs->super->s_inodes_count) ||
1632             (fs->super->s_mkfs_time &&
1633              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1634                 low_dtime_check = 0;
1635
1636         /* Set up ctx->lost_and_found if possible */
1637         (void) e2fsck_get_lost_and_found(ctx, 0);
1638
1639 #ifdef HAVE_PTHREAD
1640         if (ctx->global_ctx) {
1641                 if (ctx->options & E2F_OPT_DEBUG &&
1642                     ctx->options & E2F_OPT_MULTITHREAD)
1643                         fprintf(stderr, "thread %d jumping to group %d\n",
1644                                         ctx->thread_info.et_thread_index,
1645                                         ctx->thread_info.et_group_start);
1646                 pctx.errcode = ext2fs_inode_scan_goto_blockgroup(scan,
1647                                         ctx->thread_info.et_group_start);
1648                 if (pctx.errcode) {
1649                         fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1650                         ctx->flags |= E2F_FLAG_ABORT;
1651                         goto endit;
1652                 }
1653         }
1654 #endif
1655
1656         while (1) {
1657                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1658                         if (e2fsck_mmp_update(fs))
1659                                 fatal_error(ctx, 0);
1660                 }
1661                 old_op = ehandler_operation(eop_next_inode);
1662                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1663                                                           inode, inode_size);
1664                 if (ino > ino_threshold)
1665                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1666                 ehandler_operation(old_op);
1667                 if (e2fsck_should_abort(ctx))
1668                         goto endit;
1669                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1670                         /*
1671                          * If badblocks says badblocks is bad, offer to clear
1672                          * the list, update the in-core bb list, and restart
1673                          * the inode scan.
1674                          */
1675                         if (ino == EXT2_BAD_INO &&
1676                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1677                                         &pctx)) {
1678                                 errcode_t err;
1679
1680                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1681                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1682                                 ctx->fs->badblocks = NULL;
1683                                 err = ext2fs_read_bb_inode(ctx->fs,
1684                                                         &ctx->fs->badblocks);
1685                                 if (err) {
1686                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1687                                                     &pctx);
1688                                         ctx->flags |= E2F_FLAG_ABORT;
1689                                 } else
1690                                         ctx->flags |= E2F_FLAG_RESTART;
1691                                 goto endit;
1692                         }
1693                         if (!ctx->inode_bb_map)
1694                                 alloc_bb_map(ctx);
1695                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1696                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1697                         continue;
1698                 }
1699                 if (pctx.errcode == EXT2_ET_SCAN_FINISHED)
1700                         break;
1701                 if (pctx.errcode &&
1702                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1703                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1704                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1705                         ctx->flags |= E2F_FLAG_ABORT;
1706                         goto endit;
1707                 }
1708                 if (!ino)
1709                         break;
1710 #ifdef HAVE_PTHREAD
1711                 if (ctx->global_ctx)
1712                         ctx->thread_info.et_inode_number++;
1713 #endif
1714                 pctx.ino = ino;
1715                 pctx.inode = inode;
1716                 ctx->stashed_ino = ino;
1717
1718                 /* Clear trashed inode? */
1719                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1720                     inode->i_links_count > 0 &&
1721                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1722                         pctx.errcode = 0;
1723                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1724                 }
1725                 failed_csum = pctx.errcode != 0;
1726
1727                 /*
1728                  * Check for inodes who might have been part of the
1729                  * orphaned list linked list.  They should have gotten
1730                  * dealt with by now, unless the list had somehow been
1731                  * corrupted.
1732                  *
1733                  * FIXME: In the future, inodes which are still in use
1734                  * (and which are therefore) pending truncation should
1735                  * be handled specially.  Right now we just clear the
1736                  * dtime field, and the normal e2fsck handling of
1737                  * inodes where i_size and the inode blocks are
1738                  * inconsistent is to fix i_size, instead of releasing
1739                  * the extra blocks.  This won't catch the inodes that
1740                  * was at the end of the orphan list, but it's better
1741                  * than nothing.  The right answer is that there
1742                  * shouldn't be any bugs in the orphan list handling.  :-)
1743                  */
1744                 if (inode->i_dtime && low_dtime_check &&
1745                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1746                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1747                                 inode->i_dtime = inode->i_links_count ?
1748                                         0 : ctx->now;
1749                                 e2fsck_write_inode(ctx, ino, inode,
1750                                                    "pass1");
1751                                 failed_csum = 0;
1752                         }
1753                 }
1754
1755                 if (inode->i_links_count) {
1756                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1757                                            ino, inode->i_links_count);
1758                         if (pctx.errcode) {
1759                                 pctx.num = inode->i_links_count;
1760                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1761                                 ctx->flags |= E2F_FLAG_ABORT;
1762                                 goto endit;
1763                         }
1764                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1765                            !quota_inum_is_reserved(fs, ino)) {
1766                         if (!inode->i_dtime && inode->i_mode) {
1767                                 if (fix_problem(ctx,
1768                                             PR_1_ZERO_DTIME, &pctx)) {
1769                                         inode->i_dtime = ctx->now;
1770                                         e2fsck_write_inode(ctx, ino, inode,
1771                                                            "pass1");
1772                                         failed_csum = 0;
1773                                 }
1774                         }
1775                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1776                         continue;
1777                 }
1778
1779                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1780                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1781                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1782                      (!casefold_fs &&
1783                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1784                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1785                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1786                 }
1787
1788                 /* Conflicting inlinedata/extents inode flags? */
1789                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1790                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1791                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1792                                                                inode_size,
1793                                                                &pctx);
1794                         if (res < 0) {
1795                                 /* skip FINISH_INODE_LOOP */
1796                                 continue;
1797                         }
1798                 }
1799
1800                 /* Test for incorrect inline_data flags settings. */
1801                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1802                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1803                         size_t size = 0;
1804
1805                         pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1806                         if (!pctx.errcode &&
1807                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1808                                 e2fsck_pass1_fix_lock(ctx);
1809                                 ext2fs_set_feature_inline_data(sb);
1810                                 ext2fs_mark_super_dirty(fs);
1811                                 e2fsck_pass1_fix_unlock(ctx);
1812                                 inlinedata_fs = 1;
1813                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1814                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1815                                 /* skip FINISH_INODE_LOOP */
1816                                 continue;
1817                         }
1818                 }
1819
1820                 /* Test for inline data flag but no attr */
1821                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1822                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1823                         size_t size = 0;
1824                         errcode_t err;
1825                         int flags;
1826
1827                         flags = fs->flags;
1828                         if (failed_csum)
1829                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1830                         err = get_inline_data_ea_size(fs, ino, &size);
1831                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1832                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1833
1834                         switch (err) {
1835                         case 0:
1836                                 /* Everything is awesome... */
1837                                 break;
1838                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1839                         case EXT2_ET_BAD_EA_HASH:
1840                         case EXT2_ET_BAD_EA_HEADER:
1841                         case EXT2_ET_EA_BAD_NAME_LEN:
1842                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1843                         case EXT2_ET_EA_KEY_NOT_FOUND:
1844                         case EXT2_ET_EA_NO_SPACE:
1845                         case EXT2_ET_MISSING_EA_FEATURE:
1846                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1847                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1848                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1849                         case EXT2_ET_NO_INLINE_DATA:
1850                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1851                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1852                         case EXT2_ET_EA_INODE_CORRUPTED:
1853                                 /* broken EA or no system.data EA; truncate */
1854                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1855                                                 &pctx)) {
1856                                         err = ext2fs_inode_size_set(fs, inode, 0);
1857                                         if (err) {
1858                                                 pctx.errcode = err;
1859                                                 ctx->flags |= E2F_FLAG_ABORT;
1860                                                 goto endit;
1861                                         }
1862                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1863                                         memset(&inode->i_block, 0,
1864                                                sizeof(inode->i_block));
1865                                         e2fsck_write_inode(ctx, ino, inode,
1866                                                            "pass1");
1867                                         failed_csum = 0;
1868                                 }
1869                                 break;
1870                         default:
1871                                 /* Some other kind of non-xattr error? */
1872                                 pctx.errcode = err;
1873                                 ctx->flags |= E2F_FLAG_ABORT;
1874                                 goto endit;
1875                         }
1876                 }
1877
1878                 /*
1879                  * Test for incorrect extent flag settings.
1880                  *
1881                  * On big-endian machines we must be careful:
1882                  * When the inode is read, the i_block array is not swapped
1883                  * if the extent flag is set.  Therefore if we are testing
1884                  * for or fixing a wrongly-set flag, we must potentially
1885                  * (un)swap before testing, or after fixing.
1886                  */
1887
1888                 /*
1889                  * In this case the extents flag was set when read, so
1890                  * extent_header_verify is ok.  If the inode is cleared,
1891                  * no need to swap... so no extra swapping here.
1892                  */
1893                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1894                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1895                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1896                         if ((ext2fs_extent_header_verify(inode->i_block,
1897                                                  sizeof(inode->i_block)) == 0) &&
1898                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1899                                 e2fsck_pass1_fix_lock(ctx);
1900                                 ext2fs_set_feature_extents(sb);
1901                                 ext2fs_mark_super_dirty(fs);
1902                                 extent_fs = 1;
1903                                 e2fsck_pass1_fix_unlock(ctx);
1904                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1905                         clear_inode:
1906                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1907                                 if (ino == EXT2_BAD_INO)
1908                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1909                                                                  ino);
1910                                 /* skip FINISH_INODE_LOOP */
1911                                 continue;
1912                         }
1913                 }
1914
1915                 /*
1916                  * For big-endian machines:
1917                  * If the inode didn't have the extents flag set when it
1918                  * was read, then the i_blocks array was swapped.  To test
1919                  * as an extents header, we must swap it back first.
1920                  * IF we then set the extents flag, the entire i_block
1921                  * array must be un/re-swapped to make it proper extents data.
1922                  */
1923                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1924                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1925                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1926                     (LINUX_S_ISREG(inode->i_mode) ||
1927                      LINUX_S_ISDIR(inode->i_mode))) {
1928                         void *ehp;
1929 #ifdef WORDS_BIGENDIAN
1930                         __u32 tmp_block[EXT2_N_BLOCKS];
1931
1932                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1933                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1934                         ehp = tmp_block;
1935 #else
1936                         ehp = inode->i_block;
1937 #endif
1938                         if ((ext2fs_extent_header_verify(ehp,
1939                                          sizeof(inode->i_block)) == 0) &&
1940                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1941                                 inode->i_flags |= EXT4_EXTENTS_FL;
1942 #ifdef WORDS_BIGENDIAN
1943                                 memcpy(inode->i_block, tmp_block,
1944                                        sizeof(inode->i_block));
1945 #endif
1946                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1947                                 failed_csum = 0;
1948                         }
1949                 }
1950
1951                 if (ino == EXT2_BAD_INO) {
1952                         struct process_block_struct pb;
1953
1954                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1955                              inode->i_gid || inode->i_links_count ||
1956                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1957                              inode->i_file_acl) &&
1958                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1959                                 memset(inode, 0, sizeof(struct ext2_inode));
1960                                 e2fsck_write_inode(ctx, ino, inode,
1961                                                    "clear bad inode");
1962                                 failed_csum = 0;
1963                         }
1964
1965                         e2fsck_pass1_block_map_r_lock(ctx);
1966                         pctx.errcode = ext2fs_copy_bitmap(ctx->global_ctx ?
1967                                         ctx->global_ctx->block_found_map :
1968                                         ctx->block_found_map, &pb.fs_meta_blocks);
1969                         e2fsck_pass1_block_map_r_unlock(ctx);
1970                         if (pctx.errcode) {
1971                                 pctx.num = 4;
1972                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1973                                 ctx->flags |= E2F_FLAG_ABORT;
1974                                 goto endit;
1975                         }
1976                         pb.ino = EXT2_BAD_INO;
1977                         pb.num_blocks = pb.last_block = 0;
1978                         pb.last_db_block = -1;
1979                         pb.num_illegal_blocks = 0;
1980                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1981                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1982                         pb.inode = inode;
1983                         pb.pctx = &pctx;
1984                         pb.ctx = ctx;
1985                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1986                                      block_buf, process_bad_block, &pb);
1987                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1988                         if (pctx.errcode) {
1989                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1990                                 ctx->flags |= E2F_FLAG_ABORT;
1991                                 goto endit;
1992                         }
1993                         if (pb.bbcheck)
1994                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1995                                 ctx->flags |= E2F_FLAG_ABORT;
1996                                 goto endit;
1997                         }
1998                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1999                         clear_problem_context(&pctx);
2000                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2001                         continue;
2002                 } else if (ino == EXT2_ROOT_INO) {
2003                         /*
2004                          * Make sure the root inode is a directory; if
2005                          * not, offer to clear it.  It will be
2006                          * regenerated in pass #3.
2007                          */
2008                         if (!LINUX_S_ISDIR(inode->i_mode)) {
2009                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
2010                                         goto clear_inode;
2011                         }
2012                         /*
2013                          * If dtime is set, offer to clear it.  mke2fs
2014                          * version 0.2b created filesystems with the
2015                          * dtime field set for the root and lost+found
2016                          * directories.  We won't worry about
2017                          * /lost+found, since that can be regenerated
2018                          * easily.  But we will fix the root directory
2019                          * as a special case.
2020                          */
2021                         if (inode->i_dtime && inode->i_links_count) {
2022                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
2023                                         inode->i_dtime = 0;
2024                                         e2fsck_write_inode(ctx, ino, inode,
2025                                                            "pass1");
2026                                         failed_csum = 0;
2027                                 }
2028                         }
2029                 } else if (ino == EXT2_JOURNAL_INO) {
2030                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2031                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
2032                                 if (!LINUX_S_ISREG(inode->i_mode) &&
2033                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
2034                                                 &pctx)) {
2035                                         inode->i_mode = LINUX_S_IFREG;
2036                                         e2fsck_write_inode(ctx, ino, inode,
2037                                                            "pass1");
2038                                         failed_csum = 0;
2039                                 }
2040                                 check_blocks(ctx, &pctx, block_buf, NULL);
2041                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2042                                 continue;
2043                         }
2044                         if ((inode->i_links_count ||
2045                              inode->i_blocks || inode->i_block[0]) &&
2046                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
2047                                         &pctx)) {
2048                                 memset(inode, 0, inode_size);
2049                                 ext2fs_icount_store(ctx->inode_link_info,
2050                                                     ino, 0);
2051                                 e2fsck_write_inode_full(ctx, ino, inode,
2052                                                         inode_size, "pass1");
2053                                 failed_csum = 0;
2054                         }
2055                 } else if (quota_inum_is_reserved(fs, ino)) {
2056                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2057                         if (ext2fs_has_feature_quota(fs->super) &&
2058                             quota_inum_is_super(fs->super, ino)) {
2059                                 if (!LINUX_S_ISREG(inode->i_mode) &&
2060                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
2061                                                         &pctx)) {
2062                                         inode->i_mode = LINUX_S_IFREG;
2063                                         e2fsck_write_inode(ctx, ino, inode,
2064                                                         "pass1");
2065                                         failed_csum = 0;
2066                                 }
2067                                 check_blocks(ctx, &pctx, block_buf, NULL);
2068                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2069                                 continue;
2070                         }
2071                         if ((inode->i_links_count ||
2072                              inode->i_blocks || inode->i_block[0]) &&
2073                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
2074                                         &pctx)) {
2075                                 memset(inode, 0, inode_size);
2076                                 ext2fs_icount_store(ctx->inode_link_info,
2077                                                     ino, 0);
2078                                 e2fsck_write_inode_full(ctx, ino, inode,
2079                                                         inode_size, "pass1");
2080                                 failed_csum = 0;
2081                         }
2082                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
2083                         problem_t problem = 0;
2084
2085                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2086                         if (ino == EXT2_BOOT_LOADER_INO) {
2087                                 if (LINUX_S_ISDIR(inode->i_mode))
2088                                         problem = PR_1_RESERVED_BAD_MODE;
2089                         } else if (ino == EXT2_RESIZE_INO) {
2090                                 if (inode->i_mode &&
2091                                     !LINUX_S_ISREG(inode->i_mode))
2092                                         problem = PR_1_RESERVED_BAD_MODE;
2093                         } else {
2094                                 if (inode->i_mode != 0)
2095                                         problem = PR_1_RESERVED_BAD_MODE;
2096                         }
2097                         if (problem) {
2098                                 if (fix_problem(ctx, problem, &pctx)) {
2099                                         inode->i_mode = 0;
2100                                         e2fsck_write_inode(ctx, ino, inode,
2101                                                            "pass1");
2102                                         failed_csum = 0;
2103                                 }
2104                         }
2105                         check_blocks(ctx, &pctx, block_buf, NULL);
2106                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2107                         continue;
2108                 }
2109
2110                 if (!inode->i_links_count) {
2111                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2112                         continue;
2113                 }
2114                 /*
2115                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
2116                  * deleted files.  Oops.
2117                  *
2118                  * Since all new ext2 implementations get this right,
2119                  * we now assume that the case of non-zero
2120                  * i_links_count and non-zero dtime means that we
2121                  * should keep the file, not delete it.
2122                  *
2123                  */
2124                 if (inode->i_dtime) {
2125                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
2126                                 inode->i_dtime = 0;
2127                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
2128                                 failed_csum = 0;
2129                         }
2130                 }
2131
2132                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2133                 switch (fs->super->s_creator_os) {
2134                     case EXT2_OS_HURD:
2135                         frag = inode->osd2.hurd2.h_i_frag;
2136                         fsize = inode->osd2.hurd2.h_i_fsize;
2137                         break;
2138                     default:
2139                         frag = fsize = 0;
2140                 }
2141
2142                 if (inode->i_faddr || frag || fsize ||
2143                     (!ext2fs_has_feature_largedir(fs->super) &&
2144                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
2145                         mark_inode_bad(ctx, ino);
2146                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
2147                     !ext2fs_has_feature_64bit(fs->super) &&
2148                     inode->osd2.linux2.l_i_file_acl_high != 0)
2149                         mark_inode_bad(ctx, ino);
2150                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
2151                     !ext2fs_has_feature_huge_file(fs->super) &&
2152                     (inode->osd2.linux2.l_i_blocks_hi != 0))
2153                         mark_inode_bad(ctx, ino);
2154                 if (inode->i_flags & EXT2_IMAGIC_FL) {
2155                         if (imagic_fs) {
2156                                 if (!ctx->inode_imagic_map)
2157                                         alloc_imagic_map(ctx);
2158                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
2159                                                          ino);
2160                         } else {
2161                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
2162                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
2163                                         e2fsck_write_inode(ctx, ino,
2164                                                            inode, "pass1");
2165                                         failed_csum = 0;
2166                                 }
2167                         }
2168                 }
2169
2170                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
2171                 check_is_really_dir(ctx, &pctx, block_buf);
2172
2173                 /*
2174                  * ext2fs_inode_has_valid_blocks2 does not actually look
2175                  * at i_block[] values, so not endian-sensitive here.
2176                  */
2177                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
2178                     LINUX_S_ISLNK(inode->i_mode) &&
2179                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
2180                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
2181                         inode->i_flags &= ~EXT4_EXTENTS_FL;
2182                         e2fsck_write_inode(ctx, ino, inode, "pass1");
2183                         failed_csum = 0;
2184                 }
2185
2186                 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
2187                     add_encrypted_file(ctx, &pctx) < 0)
2188                         goto clear_inode;
2189
2190                 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
2191                         ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
2192
2193                 if (LINUX_S_ISDIR(inode->i_mode)) {
2194                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
2195                         e2fsck_add_dir_info(ctx, ino, 0);
2196                         ctx->fs_directory_count++;
2197                         if (inode->i_flags & EXT4_CASEFOLD_FL)
2198                                 add_casefolded_dir(ctx, ino);
2199                 } else if (LINUX_S_ISREG (inode->i_mode)) {
2200                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
2201                         ctx->fs_regular_count++;
2202                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
2203                            e2fsck_pass1_check_device_inode(fs, inode)) {
2204                         check_extents_inlinedata(ctx, &pctx);
2205                         check_immutable(ctx, &pctx);
2206                         check_size(ctx, &pctx);
2207                         ctx->fs_chardev_count++;
2208                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
2209                            e2fsck_pass1_check_device_inode(fs, inode)) {
2210                         check_extents_inlinedata(ctx, &pctx);
2211                         check_immutable(ctx, &pctx);
2212                         check_size(ctx, &pctx);
2213                         ctx->fs_blockdev_count++;
2214                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
2215                            e2fsck_pass1_check_symlink(fs, ino, inode,
2216                                                       block_buf)) {
2217                         check_immutable(ctx, &pctx);
2218                         ctx->fs_symlinks_count++;
2219                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
2220                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2221                                 continue;
2222                         } else if (ext2fs_is_fast_symlink(inode)) {
2223                                 ctx->fs_fast_symlinks_count++;
2224                                 check_blocks(ctx, &pctx, block_buf,
2225                                              &ea_ibody_quota);
2226                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2227                                 continue;
2228                         }
2229                 }
2230                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
2231                          e2fsck_pass1_check_device_inode(fs, inode)) {
2232                         check_extents_inlinedata(ctx, &pctx);
2233                         check_immutable(ctx, &pctx);
2234                         check_size(ctx, &pctx);
2235                         ctx->fs_fifo_count++;
2236                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
2237                            e2fsck_pass1_check_device_inode(fs, inode)) {
2238                         check_extents_inlinedata(ctx, &pctx);
2239                         check_immutable(ctx, &pctx);
2240                         check_size(ctx, &pctx);
2241                         ctx->fs_sockets_count++;
2242                 } else
2243                         mark_inode_bad(ctx, ino);
2244                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2245                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
2246                         if (inode->i_block[EXT2_IND_BLOCK])
2247                                 ctx->fs_ind_count++;
2248                         if (inode->i_block[EXT2_DIND_BLOCK])
2249                                 ctx->fs_dind_count++;
2250                         if (inode->i_block[EXT2_TIND_BLOCK])
2251                                 ctx->fs_tind_count++;
2252                 }
2253                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2254                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
2255                     (inode->i_block[EXT2_IND_BLOCK] ||
2256                      inode->i_block[EXT2_DIND_BLOCK] ||
2257                      inode->i_block[EXT2_TIND_BLOCK] ||
2258                      ext2fs_file_acl_block(fs, inode))) {
2259                         struct process_inode_block *itp;
2260
2261                         itp = &inodes_to_process[process_inode_count];
2262                         itp->ino = ino;
2263                         itp->ea_ibody_quota = ea_ibody_quota;
2264                         if (inode_size < sizeof(struct ext2_inode_large))
2265                                 memcpy(&itp->inode, inode, inode_size);
2266                         else
2267                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
2268                         process_inode_count++;
2269                 } else
2270                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
2271
2272                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2273
2274                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2275                         goto endit;
2276
2277                 if (process_inode_count >= ctx->process_inode_size) {
2278                         process_inodes(ctx, block_buf, inodes_to_process,
2279                                        &process_inode_count);
2280
2281                         if (e2fsck_should_abort(ctx))
2282                                 goto endit;
2283                 }
2284         }
2285         process_inodes(ctx, block_buf, inodes_to_process,
2286                        &process_inode_count);
2287         ext2fs_close_inode_scan(scan);
2288         scan = NULL;
2289
2290         if (ctx->ea_block_quota_blocks) {
2291                 ea_refcount_free(ctx->ea_block_quota_blocks);
2292                 ctx->ea_block_quota_blocks = 0;
2293         }
2294
2295         if (ctx->ea_block_quota_inodes) {
2296                 ea_refcount_free(ctx->ea_block_quota_inodes);
2297                 ctx->ea_block_quota_inodes = 0;
2298         }
2299
2300         /* We don't need the encryption policy => ID map any more */
2301         destroy_encryption_policy_map(ctx);
2302
2303         if (ctx->flags & E2F_FLAG_RESTART) {
2304                 /*
2305                  * Only the master copy of the superblock and block
2306                  * group descriptors are going to be written during a
2307                  * restart, so set the superblock to be used to be the
2308                  * master superblock.
2309                  */
2310                 ctx->use_superblock = 0;
2311                 goto endit;
2312         }
2313
2314         if (ctx->large_dirs && !ext2fs_has_feature_largedir(ctx->fs->super)) {
2315                 ext2_filsys fs = ctx->fs;
2316
2317                 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2318                         ext2fs_set_feature_largedir(fs->super);
2319                         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2320                         ext2fs_mark_super_dirty(fs);
2321                 }
2322                 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2323                     fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2324                         ext2fs_update_dynamic_rev(fs);
2325                         ext2fs_mark_super_dirty(fs);
2326                 }
2327         }
2328
2329         ctx->flags |= E2F_FLAG_ALLOC_OK;
2330 endit:
2331         e2fsck_use_inode_shortcuts(ctx, 0);
2332         ext2fs_free_mem(&inodes_to_process);
2333         inodes_to_process = 0;
2334
2335         if (scan)
2336                 ext2fs_close_inode_scan(scan);
2337         if (block_buf)
2338                 ext2fs_free_mem(&block_buf);
2339         if (inode)
2340                 ext2fs_free_mem(&inode);
2341
2342         /*
2343          * The l+f inode may have been cleared, so zap it now and
2344          * later passes will recalculate it if necessary
2345          */
2346         ctx->lost_and_found = 0;
2347
2348         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2349                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2350         else
2351                 ctx->invalid_bitmaps++;
2352 }
2353
2354 #ifdef HAVE_PTHREAD
2355 static errcode_t e2fsck_pass1_copy_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2356                                           ext2fs_generic_bitmap *dest)
2357 {
2358         errcode_t ret;
2359
2360         ret = ext2fs_copy_bitmap(*src, dest);
2361         if (ret)
2362                 return ret;
2363
2364         (*dest)->fs = fs;
2365
2366         return 0;
2367 }
2368
2369 static void e2fsck_pass1_free_bitmap(ext2fs_generic_bitmap *bitmap)
2370 {
2371         if (*bitmap) {
2372                 ext2fs_free_generic_bmap(*bitmap);
2373                 *bitmap = NULL;
2374         }
2375
2376 }
2377
2378 static errcode_t e2fsck_pass1_merge_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2379                                           ext2fs_generic_bitmap *dest)
2380 {
2381         errcode_t ret = 0;
2382
2383         if (*src) {
2384                 if (*dest == NULL) {
2385                         *dest = *src;
2386                         *src = NULL;
2387                 } else {
2388                         ret = ext2fs_merge_bitmap(*src, *dest, NULL, NULL);
2389                         if (ret)
2390                                 return ret;
2391                 }
2392                 (*dest)->fs = fs;
2393         }
2394
2395         return 0;
2396 }
2397
2398 static errcode_t e2fsck_pass1_copy_fs(ext2_filsys dest, e2fsck_t src_context,
2399                                       ext2_filsys src)
2400 {
2401         errcode_t       retval;
2402
2403         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2404         dest->inode_map = NULL;
2405         dest->block_map = NULL;
2406         dest->badblocks = NULL;
2407         if (dest->dblist)
2408                 dest->dblist->fs = dest;
2409         if (src->block_map) {
2410                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2411                                                   &dest->block_map);
2412                 if (retval)
2413                         return retval;
2414         }
2415         if (src->inode_map) {
2416                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2417                                                   &dest->inode_map);
2418                 if (retval)
2419                         return retval;
2420         }
2421
2422         if (src->badblocks) {
2423                 retval = ext2fs_badblocks_copy(src->badblocks,
2424                                                &dest->badblocks);
2425                 if (retval)
2426                         return retval;
2427         }
2428
2429         /* disable it for now */
2430         src_context->openfs_flags &= ~EXT2_FLAG_EXCLUSIVE;
2431         retval = ext2fs_open_channel(dest, src_context->io_options,
2432                                      src_context->io_manager,
2433                                      src_context->openfs_flags,
2434                                      src->io->block_size);
2435         if (retval)
2436                 return retval;
2437
2438         /* Block size might not be default */
2439         io_channel_set_blksize(dest->io, src->io->block_size);
2440         ehandler_init(dest->io);
2441
2442         assert(dest->io->magic == src->io->magic);
2443         assert(dest->io->manager == src->io->manager);
2444         assert(strcmp(dest->io->name, src->io->name) == 0);
2445         assert(dest->io->block_size == src->io->block_size);
2446         assert(dest->io->read_error == src->io->read_error);
2447         assert(dest->io->write_error == src->io->write_error);
2448         assert(dest->io->refcount == src->io->refcount);
2449         assert(dest->io->flags == src->io->flags);
2450         assert(dest->io->app_data == dest);
2451         assert(src->io->app_data == src);
2452         assert(dest->io->align == src->io->align);
2453
2454         /* The data should be written to disk immediately */
2455         dest->io->flags |= CHANNEL_FLAGS_WRITETHROUGH;
2456         /* icache will be rebuilt if needed, so do not copy from @src */
2457         src->icache = NULL;
2458         return 0;
2459 }
2460
2461 static int e2fsck_pass1_merge_fs(ext2_filsys dest, ext2_filsys src)
2462 {
2463         struct ext2_inode_cache *icache = dest->icache;
2464         errcode_t retval = 0;
2465         io_channel dest_io;
2466         io_channel dest_image_io;
2467         ext2fs_inode_bitmap inode_map;
2468         ext2fs_block_bitmap block_map;
2469         ext2_badblocks_list badblocks;
2470         ext2_dblist dblist;
2471         int flags;
2472
2473         dest_io = dest->io;
2474         dest_image_io = dest->image_io;
2475         inode_map = dest->inode_map;
2476         block_map = dest->block_map;
2477         badblocks = dest->badblocks;
2478         dblist = dest->dblist;
2479         flags = dest->flags;
2480
2481         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2482         dest->io = dest_io;
2483         dest->image_io = dest_image_io;
2484         dest->icache = icache;
2485         dest->inode_map = inode_map;
2486         dest->block_map = block_map;
2487         dest->badblocks = badblocks;
2488         dest->dblist = dblist;
2489         if (dest->dblist)
2490                 dest->dblist->fs = dest;
2491         dest->flags = src->flags | flags;
2492         if (!(src->flags & EXT2_FLAG_VALID) || !(flags & EXT2_FLAG_VALID))
2493                 ext2fs_unmark_valid(dest);
2494
2495         if (src->icache) {
2496                 ext2fs_free_inode_cache(src->icache);
2497                 src->icache = NULL;
2498         }
2499
2500         retval = e2fsck_pass1_merge_bitmap(dest, &src->inode_map,
2501                                            &dest->inode_map);
2502         if (retval)
2503                 goto out;
2504
2505         retval = e2fsck_pass1_merge_bitmap(dest, &src->block_map,
2506                                           &dest->block_map);
2507         if (retval)
2508                 goto out;
2509
2510         if (src->dblist) {
2511                 if (dest->dblist) {
2512                         retval = ext2fs_merge_dblist(src->dblist,
2513                                                      dest->dblist);
2514                         if (retval)
2515                                 goto out;
2516                 } else {
2517                         dest->dblist = src->dblist;
2518                         dest->dblist->fs = dest;
2519                         src->dblist = NULL;
2520                 }
2521         }
2522
2523         if (src->badblocks) {
2524                 if (dest->badblocks == NULL)
2525                         retval = ext2fs_badblocks_copy(src->badblocks,
2526                                                        &dest->badblocks);
2527                 else
2528                         retval = ext2fs_badblocks_merge(src->badblocks,
2529                                                         dest->badblocks);
2530         }
2531 out:
2532         io_channel_close(src->io);
2533         if (src->inode_map)
2534                 ext2fs_free_generic_bmap(src->inode_map);
2535         if (src->block_map)
2536                 ext2fs_free_generic_bmap(src->block_map);
2537         if (src->badblocks)
2538                 ext2fs_badblocks_list_free(src->badblocks);
2539         if (src->dblist)
2540                 ext2fs_free_dblist(src->dblist);
2541
2542         return retval;
2543 }
2544
2545 static void e2fsck_pass1_copy_invalid_bitmaps(e2fsck_t global_ctx,
2546                                               e2fsck_t thread_ctx)
2547 {
2548         dgrp_t i, j;
2549         dgrp_t grp_start = thread_ctx->thread_info.et_group_start;
2550         dgrp_t grp_end = thread_ctx->thread_info.et_group_end;
2551         dgrp_t total = grp_end - grp_start;
2552
2553         thread_ctx->invalid_inode_bitmap_flag =
2554                         e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2555                                                 "invalid_inode_bitmap");
2556         thread_ctx->invalid_block_bitmap_flag =
2557                         e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2558                                                "invalid_block_bitmap");
2559         thread_ctx->invalid_inode_table_flag =
2560                         e2fsck_allocate_memory(global_ctx, sizeof(int) * total,
2561                                                "invalid_inode_table");
2562
2563         memcpy(thread_ctx->invalid_block_bitmap_flag,
2564                &global_ctx->invalid_block_bitmap_flag[grp_start],
2565                total * sizeof(int));
2566         memcpy(thread_ctx->invalid_inode_bitmap_flag,
2567                &global_ctx->invalid_inode_bitmap_flag[grp_start],
2568                total * sizeof(int));
2569         memcpy(thread_ctx->invalid_inode_table_flag,
2570                &global_ctx->invalid_inode_table_flag[grp_start],
2571                total * sizeof(int));
2572
2573         thread_ctx->invalid_bitmaps = 0;
2574         for (i = grp_start, j = 0; i < grp_end; i++, j++) {
2575                 if (thread_ctx->invalid_block_bitmap_flag[j])
2576                         thread_ctx->invalid_bitmaps++;
2577                 if (thread_ctx->invalid_inode_bitmap_flag[j])
2578                         thread_ctx->invalid_bitmaps++;
2579                 if (thread_ctx->invalid_inode_table_flag[j])
2580                         thread_ctx->invalid_bitmaps++;
2581         }
2582 }
2583
2584 static void e2fsck_pass1_merge_invalid_bitmaps(e2fsck_t global_ctx,
2585                                                e2fsck_t thread_ctx)
2586 {
2587         dgrp_t i, j;
2588         dgrp_t grp_start = thread_ctx->thread_info.et_group_start;
2589         dgrp_t grp_end = thread_ctx->thread_info.et_group_end;
2590         dgrp_t total = grp_end - grp_start;
2591
2592         memcpy(&global_ctx->invalid_block_bitmap_flag[grp_start],
2593                thread_ctx->invalid_block_bitmap_flag, total * sizeof(int));
2594         memcpy(&global_ctx->invalid_inode_bitmap_flag[grp_start],
2595                thread_ctx->invalid_inode_bitmap_flag, total * sizeof(int));
2596         memcpy(&global_ctx->invalid_inode_table_flag[grp_start],
2597                thread_ctx->invalid_inode_table_flag, total * sizeof(int));
2598         global_ctx->invalid_bitmaps += thread_ctx->invalid_bitmaps;
2599 }
2600
2601 static errcode_t e2fsck_pass1_thread_prepare(e2fsck_t global_ctx, e2fsck_t *thread_ctx,
2602                                              int thread_index, int num_threads,
2603                                              dgrp_t average_group)
2604 {
2605         errcode_t               retval;
2606         e2fsck_t                thread_context;
2607         ext2_filsys             thread_fs;
2608         ext2_filsys             global_fs = global_ctx->fs;
2609         struct e2fsck_thread    *tinfo;
2610
2611         assert(global_ctx->inode_used_map == NULL);
2612         assert(global_ctx->inode_dir_map == NULL);
2613         assert(global_ctx->inode_bb_map == NULL);
2614         assert(global_ctx->inode_imagic_map == NULL);
2615         assert(global_ctx->inode_reg_map == NULL);
2616         assert(global_ctx->inodes_to_rebuild == NULL);
2617
2618         assert(global_ctx->block_found_map != NULL);
2619         assert(global_ctx->block_metadata_map != NULL);
2620         assert(global_ctx->block_dup_map != NULL);
2621         assert(global_ctx->block_ea_map == NULL);
2622         assert(global_ctx->fs->dblist == NULL);
2623
2624         retval = ext2fs_get_mem(sizeof(struct e2fsck_struct), &thread_context);
2625         if (retval) {
2626                 com_err(global_ctx->program_name, retval, "while allocating memory");
2627                 return retval;
2628         }
2629         memcpy(thread_context, global_ctx, sizeof(struct e2fsck_struct));
2630         thread_context->block_dup_map = NULL;
2631
2632         retval = e2fsck_allocate_block_bitmap(global_ctx->fs,
2633                                 _("in-use block map"), EXT2FS_BMAP64_RBTREE,
2634                                 "block_found_map", &thread_context->block_found_map);
2635         if (retval)
2636                 goto out_context;
2637
2638         thread_context->global_ctx = global_ctx;
2639         retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &thread_fs);
2640         if (retval) {
2641                 com_err(global_ctx->program_name, retval, "while allocating memory");
2642                 goto out_context;
2643         }
2644
2645         io_channel_flush_cleanup(global_fs->io);
2646         retval = e2fsck_pass1_copy_fs(thread_fs, global_ctx, global_fs);
2647         if (retval) {
2648                 com_err(global_ctx->program_name, retval, "while copying fs");
2649                 goto out_fs;
2650         }
2651         thread_fs->priv_data = thread_context;
2652
2653         thread_context->thread_info.et_thread_index = thread_index;
2654         set_up_logging(thread_context);
2655
2656         tinfo = &thread_context->thread_info;
2657         tinfo->et_group_start = average_group * thread_index;
2658         if (thread_index == global_fs->fs_num_threads - 1)
2659                 tinfo->et_group_end = thread_fs->group_desc_count;
2660         else
2661                 tinfo->et_group_end = average_group * (thread_index + 1);
2662         tinfo->et_group_next = tinfo->et_group_start;
2663         tinfo->et_inode_number = 0;
2664         tinfo->et_log_buf[0] = '\0';
2665         tinfo->et_log_length = 0;
2666         if (thread_context->options & E2F_OPT_MULTITHREAD)
2667                 log_out(thread_context, _("Scan group range [%d, %d)\n"),
2668                         tinfo->et_group_start, tinfo->et_group_end);
2669         thread_context->fs = thread_fs;
2670         retval = quota_init_context(&thread_context->qctx, thread_fs, 0);
2671         if (retval) {
2672                 com_err(global_ctx->program_name, retval,
2673                         "while init quota context");
2674                 goto out_fs;
2675         }
2676         *thread_ctx = thread_context;
2677         e2fsck_pass1_copy_invalid_bitmaps(global_ctx, thread_context);
2678         return 0;
2679 out_fs:
2680         ext2fs_free_mem(&thread_fs);
2681 out_context:
2682         if (thread_context->block_found_map)
2683                 ext2fs_free_mem(&thread_context->block_found_map);
2684         ext2fs_free_mem(&thread_context);
2685         return retval;
2686 }
2687
2688 static void e2fsck_pass1_merge_dir_info(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2689 {
2690         if (thread_ctx->dir_info == NULL)
2691                 return;
2692
2693         if (global_ctx->dir_info == NULL) {
2694                 global_ctx->dir_info = thread_ctx->dir_info;
2695                 thread_ctx->dir_info = NULL;
2696                 return;
2697         }
2698
2699         e2fsck_merge_dir_info(global_ctx, thread_ctx->dir_info,
2700                               global_ctx->dir_info);
2701 }
2702
2703 static void e2fsck_pass1_merge_dx_dir(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2704 {
2705         if (thread_ctx->dx_dir_info == NULL)
2706                 return;
2707
2708         if (global_ctx->dx_dir_info == NULL) {
2709                 global_ctx->dx_dir_info = thread_ctx->dx_dir_info;
2710                 global_ctx->dx_dir_info_size = thread_ctx->dx_dir_info_size;
2711                 global_ctx->dx_dir_info_count = thread_ctx->dx_dir_info_count;
2712                 thread_ctx->dx_dir_info = NULL;
2713                 return;
2714         }
2715
2716         e2fsck_merge_dx_dir(global_ctx, thread_ctx);
2717 }
2718
2719 static inline errcode_t
2720 e2fsck_pass1_merge_icount(ext2_icount_t *dest_icount,
2721                           ext2_icount_t *src_icount)
2722 {
2723         if (*src_icount) {
2724                 if (*dest_icount == NULL) {
2725                         *dest_icount = *src_icount;
2726                         *src_icount = NULL;
2727                 } else {
2728                         errcode_t ret;
2729
2730                         ret = ext2fs_icount_merge(*src_icount,
2731                                                   *dest_icount);
2732                         if (ret)
2733                                 return ret;
2734                 }
2735         }
2736
2737         return 0;
2738 }
2739
2740 static errcode_t e2fsck_pass1_merge_icounts(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2741 {
2742         errcode_t ret;
2743
2744         ret = e2fsck_pass1_merge_icount(&global_ctx->inode_count,
2745                                         &thread_ctx->inode_count);
2746         if (ret)
2747                 return ret;
2748         ret = e2fsck_pass1_merge_icount(&global_ctx->inode_link_info,
2749                                         &thread_ctx->inode_link_info);
2750
2751         return ret;
2752 }
2753
2754 static errcode_t e2fsck_pass1_merge_dirs_to_hash(e2fsck_t global_ctx,
2755                                                  e2fsck_t thread_ctx)
2756 {
2757         errcode_t retval = 0;
2758
2759         if (!thread_ctx->dirs_to_hash)
2760                 return 0;
2761
2762         if (!global_ctx->dirs_to_hash)
2763                 retval = ext2fs_badblocks_copy(thread_ctx->dirs_to_hash,
2764                                                &global_ctx->dirs_to_hash);
2765         else
2766                 retval = ext2fs_badblocks_merge(thread_ctx->dirs_to_hash,
2767                                                 global_ctx->dirs_to_hash);
2768
2769         return retval;
2770 }
2771
2772 static errcode_t e2fsck_pass1_merge_ea_inode_refs(e2fsck_t global_ctx,
2773                                                   e2fsck_t thread_ctx)
2774 {
2775         ea_value_t count;
2776         blk64_t blk;
2777         errcode_t retval;
2778
2779         if (!thread_ctx->ea_inode_refs)
2780                 return 0;
2781
2782         if (!global_ctx->ea_inode_refs) {
2783                 global_ctx->ea_inode_refs = thread_ctx->ea_inode_refs;
2784                 thread_ctx->ea_inode_refs = NULL;
2785                 return 0;
2786         }
2787
2788         ea_refcount_intr_begin(thread_ctx->ea_inode_refs);
2789         while (1) {
2790                 if ((blk = ea_refcount_intr_next(thread_ctx->ea_inode_refs,
2791                                                  &count)) == 0)
2792                         break;
2793                 if (!global_ctx->block_ea_map ||
2794                     !ext2fs_fast_test_block_bitmap2(global_ctx->block_ea_map,
2795                                                     blk)) {
2796                         retval = ea_refcount_store(global_ctx->ea_inode_refs,
2797                                                    blk, count);
2798                         if (retval)
2799                                 return retval;
2800                 }
2801         }
2802
2803         return retval;
2804 }
2805
2806 static ea_value_t ea_refcount_usage(e2fsck_t ctx, blk64_t blk,
2807                                     ea_value_t *orig)
2808 {
2809         ea_value_t count_cur;
2810         ea_value_t count_extra = 0;
2811         ea_value_t count_orig;
2812
2813         ea_refcount_fetch(ctx->refcount_orig, blk, &count_orig);
2814         ea_refcount_fetch(ctx->refcount, blk, &count_cur);
2815         /* most of time this is not needed */
2816         if (ctx->refcount_extra && count_cur == 0)
2817                 ea_refcount_fetch(ctx->refcount_extra, blk, &count_extra);
2818
2819         if (!count_orig)
2820                 count_orig = *orig;
2821         else if (orig)
2822                 *orig = count_orig;
2823
2824         return count_orig + count_extra - count_cur;
2825 }
2826
2827 static errcode_t e2fsck_pass1_merge_ea_refcount(e2fsck_t global_ctx,
2828                                                 e2fsck_t thread_ctx)
2829 {
2830         ea_value_t count;
2831         blk64_t blk;
2832         errcode_t retval = 0;
2833
2834         if (!thread_ctx->refcount)
2835                 return 0;
2836
2837         if (!global_ctx->refcount) {
2838                 global_ctx->refcount = thread_ctx->refcount;
2839                 thread_ctx->refcount = NULL;
2840                 global_ctx->refcount_extra = thread_ctx->refcount;
2841                 thread_ctx->refcount_extra = NULL;
2842                 return 0;
2843         }
2844
2845         ea_refcount_intr_begin(thread_ctx->refcount);
2846         while (1) {
2847                 if ((blk = ea_refcount_intr_next(thread_ctx->refcount,
2848                                                  &count)) == 0)
2849                         break;
2850                 /**
2851                  * this EA has never seen before, so just store its
2852                  * refcount and refcount_extra into global_ctx if needed.
2853                  */
2854                 if (!global_ctx->block_ea_map ||
2855                     !ext2fs_fast_test_block_bitmap2(global_ctx->block_ea_map,
2856                                                     blk)) {
2857                         ea_value_t extra;
2858
2859                         retval = ea_refcount_store(global_ctx->refcount,
2860                                                    blk, count);
2861                         if (retval)
2862                                 return retval;
2863
2864                         if (count > 0 || !thread_ctx->refcount_extra)
2865                                 continue;
2866                         ea_refcount_fetch(thread_ctx->refcount_extra, blk,
2867                                           &extra);
2868                         if (extra == 0)
2869                                 continue;
2870
2871                         if (!global_ctx->refcount_extra) {
2872                                 retval = ea_refcount_create(0,
2873                                                 &global_ctx->refcount_extra);
2874                                 if (retval)
2875                                         return retval;
2876                         }
2877                         retval = ea_refcount_store(global_ctx->refcount_extra,
2878                                                    blk, extra);
2879                         if (retval)
2880                                 return retval;
2881                 } else {
2882                         ea_value_t orig;
2883                         ea_value_t thread_usage;
2884                         ea_value_t global_usage;
2885                         ea_value_t new;
2886
2887                         thread_usage = ea_refcount_usage(thread_ctx,
2888                                                          blk, &orig);
2889                         global_usage = ea_refcount_usage(global_ctx,
2890                                                          blk, &orig);
2891                         if (thread_usage + global_usage <= orig) {
2892                                 new = orig - thread_usage - global_usage;
2893                                 retval = ea_refcount_store(global_ctx->refcount,
2894                                                            blk, new);
2895                                 if (retval)
2896                                         return retval;
2897                                 continue;
2898                         }
2899                         /* update it is as zero */
2900                         retval = ea_refcount_store(global_ctx->refcount,
2901                                                    blk, 0);
2902                         if (retval)
2903                                 return retval;
2904                         /* Ooops, this EA was referenced more than it stated */
2905                         if (!global_ctx->refcount_extra) {
2906                                 retval = ea_refcount_create(0,
2907                                                 &global_ctx->refcount_extra);
2908                                 if (retval)
2909                                         return retval;
2910                         }
2911                         new = global_usage + thread_usage - orig;
2912                         retval = ea_refcount_store(global_ctx->refcount_extra,
2913                                                    blk, new);
2914                         if (retval)
2915                                 return retval;
2916                 }
2917         }
2918
2919         return retval;
2920 }
2921
2922 static int e2fsck_pass1_thread_join_one(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2923 {
2924         errcode_t        retval;
2925         int              flags = global_ctx->flags;
2926         ext2_filsys      thread_fs = thread_ctx->fs;
2927         ext2_filsys      global_fs = global_ctx->fs;
2928         FILE            *global_logf = global_ctx->logf;
2929         FILE            *global_problem_logf = global_ctx->problem_logf;
2930         ext2fs_inode_bitmap inode_bad_map = global_ctx->inode_bad_map;
2931         struct dir_info_db *dir_info = global_ctx->dir_info;
2932         struct dx_dir_info *dx_dir_info = global_ctx->dx_dir_info;
2933         ext2fs_inode_bitmap inode_used_map = global_ctx->inode_used_map;
2934         ext2fs_inode_bitmap inode_dir_map = global_ctx->inode_dir_map;
2935         ext2fs_inode_bitmap inode_bb_map = global_ctx->inode_bb_map;
2936         ext2fs_inode_bitmap inode_imagic_map = global_ctx->inode_imagic_map;
2937         ext2fs_inode_bitmap inode_reg_map = global_ctx->inode_reg_map;
2938         ext2fs_block_bitmap inodes_to_rebuild = global_ctx->inodes_to_rebuild;
2939         ext2_icount_t inode_count = global_ctx->inode_count;
2940         ext2_icount_t inode_link_info = global_ctx->inode_link_info;
2941         __u32 fs_directory_count = global_ctx->fs_directory_count;
2942         __u32 fs_regular_count = global_ctx->fs_regular_count;
2943         __u32 fs_blockdev_count = global_ctx->fs_blockdev_count;
2944         __u32 fs_chardev_count = global_ctx->fs_chardev_count;
2945         __u32 fs_links_count = global_ctx->fs_links_count;
2946         __u32 fs_symlinks_count = global_ctx->fs_symlinks_count;
2947         __u32 fs_fast_symlinks_count = global_ctx->fs_fast_symlinks_count;
2948         __u32 fs_fifo_count = global_ctx->fs_fifo_count;
2949         __u32 fs_total_count = global_ctx->fs_total_count;
2950         __u32 fs_badblocks_count = global_ctx->fs_badblocks_count;
2951         __u32 fs_sockets_count = global_ctx->fs_sockets_count;
2952         __u32 fs_ind_count = global_ctx->fs_ind_count;
2953         __u32 fs_dind_count = global_ctx->fs_dind_count;
2954         __u32 fs_tind_count = global_ctx->fs_tind_count;
2955         __u32 fs_fragmented = global_ctx->fs_fragmented;
2956         __u32 fs_fragmented_dir = global_ctx->fs_fragmented_dir;
2957         __u32 large_files = global_ctx->large_files;
2958         ext2_ino_t dx_dir_info_size = global_ctx->dx_dir_info_size;
2959         ext2_ino_t dx_dir_info_count = global_ctx->dx_dir_info_count;
2960         ext2_u32_list dirs_to_hash = global_ctx->dirs_to_hash;
2961         quota_ctx_t qctx = global_ctx->qctx;
2962         int *invalid_block_bitmap_flag = global_ctx->invalid_block_bitmap_flag;
2963         int *invalid_inode_bitmap_flag = global_ctx->invalid_inode_bitmap_flag;
2964         int *invalid_inode_table_flag  = global_ctx->invalid_inode_table_flag;
2965         int invalid_bitmaps = global_ctx->invalid_bitmaps;
2966         ext2_refcount_t refcount = global_ctx->refcount;
2967         ext2_refcount_t refcount_extra = global_ctx->refcount_extra;
2968         ext2_refcount_t refcount_orig = global_ctx->refcount_orig;
2969         ext2_refcount_t ea_block_quota_blocks = global_ctx->ea_block_quota_blocks;
2970         ext2_refcount_t ea_block_quota_inodes = global_ctx->ea_block_quota_inodes;
2971         ext2fs_block_bitmap block_ea_map = global_ctx->block_ea_map;
2972         ext2_refcount_t ea_inode_refs = global_ctx->ea_inode_refs;
2973         ext2fs_block_bitmap  block_found_map = global_ctx->block_found_map;
2974         ext2fs_block_bitmap  block_dup_map = global_ctx->block_dup_map;
2975         int options = global_ctx->options, i;
2976         __u32 extent_depth_count[MAX_EXTENT_DEPTH_COUNT];
2977
2978         memcpy(extent_depth_count, global_ctx->extent_depth_count,
2979                sizeof(extent_depth_count));
2980 #ifdef HAVE_SETJMP_H
2981         jmp_buf          old_jmp;
2982
2983         memcpy(old_jmp, global_ctx->abort_loc, sizeof(jmp_buf));
2984 #endif
2985         memcpy(global_ctx, thread_ctx, sizeof(struct e2fsck_struct));
2986 #ifdef HAVE_SETJMP_H
2987         memcpy(global_ctx->abort_loc, old_jmp, sizeof(jmp_buf));
2988 #endif
2989
2990         global_ctx->inode_used_map = inode_used_map;
2991         global_ctx->inode_bad_map = inode_bad_map;
2992         global_ctx->inode_dir_map = inode_dir_map;
2993         global_ctx->inode_bb_map = inode_bb_map;
2994         global_ctx->inode_imagic_map = inode_imagic_map;
2995         global_ctx->inodes_to_rebuild = inodes_to_rebuild;
2996         global_ctx->inode_reg_map = inode_reg_map;
2997         global_ctx->block_dup_map = block_dup_map;
2998         global_ctx->block_found_map = block_found_map;
2999         global_ctx->dir_info = dir_info;
3000         e2fsck_pass1_merge_dir_info(global_ctx, thread_ctx);
3001         global_ctx->dx_dir_info = dx_dir_info;
3002         global_ctx->dx_dir_info_count = dx_dir_info_count;
3003         global_ctx->dx_dir_info_size = dx_dir_info_size;
3004         e2fsck_pass1_merge_dx_dir(global_ctx, thread_ctx);
3005         global_ctx->inode_count = inode_count;
3006         global_ctx->inode_link_info = inode_link_info;
3007         global_ctx->refcount = refcount;
3008         global_ctx->refcount_extra = refcount_extra;
3009         global_ctx->refcount_orig = refcount_orig;
3010         global_ctx->ea_block_quota_blocks = ea_block_quota_blocks;
3011         global_ctx->ea_block_quota_inodes = ea_block_quota_inodes;
3012         global_ctx->block_ea_map = block_ea_map;
3013         global_ctx->ea_inode_refs = ea_inode_refs;
3014         global_ctx->fs_directory_count += fs_directory_count;
3015         global_ctx->fs_regular_count += fs_regular_count;
3016         global_ctx->fs_blockdev_count += fs_blockdev_count;
3017         global_ctx->fs_chardev_count += fs_chardev_count;
3018         global_ctx->fs_links_count += fs_links_count;
3019         global_ctx->fs_symlinks_count += fs_symlinks_count;
3020         global_ctx->fs_fast_symlinks_count += fs_fast_symlinks_count;
3021         global_ctx->fs_fifo_count += fs_fifo_count;
3022         global_ctx->fs_total_count += fs_total_count;
3023         global_ctx->fs_badblocks_count += fs_badblocks_count;
3024         global_ctx->fs_sockets_count += fs_sockets_count;
3025         global_ctx->fs_ind_count += fs_ind_count;
3026         global_ctx->fs_dind_count += fs_dind_count;
3027         global_ctx->fs_tind_count += fs_tind_count;
3028         global_ctx->fs_fragmented += fs_fragmented;
3029         global_ctx->fs_fragmented_dir += fs_fragmented_dir;
3030         global_ctx->large_files += large_files;
3031         /* threads might enable E2F_OPT_YES */
3032         global_ctx->options |= options;
3033         global_ctx->flags |= flags;
3034         /*
3035          * The l+f inode may have been cleared, so zap it now and
3036          * later passes will recalculate it if necessary
3037          */
3038         global_ctx->lost_and_found = 0;
3039         memcpy(global_ctx->extent_depth_count, extent_depth_count,
3040                sizeof(extent_depth_count));
3041         /* merge extent depth count */
3042         for (i = 0; i < MAX_EXTENT_DEPTH_COUNT; i++)
3043                 global_ctx->extent_depth_count[i] +=
3044                         thread_ctx->extent_depth_count[i];
3045
3046         retval = e2fsck_pass1_merge_fs(global_fs, thread_fs);
3047         if (retval) {
3048                 com_err(global_ctx->program_name, 0, _("while merging fs\n"));
3049                 return retval;
3050         }
3051         global_fs->priv_data = global_ctx;
3052         global_ctx->fs = global_fs;
3053         global_ctx->logf = global_logf;
3054         global_ctx->problem_logf = global_problem_logf;
3055         global_ctx->global_ctx = NULL;
3056         retval = e2fsck_pass1_merge_icounts(global_ctx, thread_ctx);
3057         if (retval) {
3058                 com_err(global_ctx->program_name, 0,
3059                         _("while merging icounts\n"));
3060                 return retval;
3061         }
3062
3063         global_ctx->dirs_to_hash = dirs_to_hash;
3064         retval = e2fsck_pass1_merge_dirs_to_hash(global_ctx, thread_ctx);
3065         if (retval) {
3066                 com_err(global_ctx->program_name, 0,
3067                         _("while merging dirs to hash\n"));
3068                 return retval;
3069         }
3070
3071         e2fsck_pass1_merge_ea_inode_refs(global_ctx, thread_ctx);
3072         e2fsck_pass1_merge_ea_refcount(global_ctx, thread_ctx);
3073         global_ctx->qctx = qctx;
3074         retval = quota_merge_and_update_usage(global_ctx->qctx,
3075                                               thread_ctx->qctx);
3076         if (retval)
3077                 return retval;
3078
3079         e2fsck_pass1_merge_invalid_bitmaps(global_ctx, thread_ctx);
3080
3081         retval = e2fsck_pass1_merge_bitmap(global_fs,
3082                                 &thread_ctx->inode_used_map,
3083                                 &global_ctx->inode_used_map);
3084         if (retval)
3085                 return retval;
3086
3087         retval = e2fsck_pass1_merge_bitmap(global_fs,
3088                                 &thread_ctx->inode_bad_map,
3089                                 &global_ctx->inode_bad_map);
3090         if (retval)
3091                 return retval;
3092         retval = e2fsck_pass1_merge_bitmap(global_fs,
3093                                         &thread_ctx->inode_dir_map,
3094                                         &global_ctx->inode_dir_map);
3095         if (retval)
3096                 return retval;
3097         retval = e2fsck_pass1_merge_bitmap(global_fs,
3098                                 &thread_ctx->inode_bb_map,
3099                                 &global_ctx->inode_bb_map);
3100         if (retval)
3101                 return retval;
3102         retval = e2fsck_pass1_merge_bitmap(global_fs,
3103                                 &thread_ctx->inode_imagic_map,
3104                                 &global_ctx->inode_imagic_map);
3105         if (retval)
3106                 return retval;
3107         retval = e2fsck_pass1_merge_bitmap(global_fs,
3108                                 &thread_ctx->inode_reg_map,
3109                                 &global_ctx->inode_reg_map);
3110         if (retval)
3111                 return retval;
3112         retval = e2fsck_pass1_merge_bitmap(global_fs,
3113                                 &thread_ctx->inodes_to_rebuild,
3114                                 &global_ctx->inodes_to_rebuild);
3115         if (retval)
3116                 return retval;
3117         retval = e2fsck_pass1_merge_bitmap(global_fs,
3118                                 &thread_ctx->block_ea_map,
3119                                 &global_ctx->block_ea_map);
3120         if (retval)
3121                 return retval;
3122
3123         if (ext2fs_has_feature_shared_blocks(global_fs->super) &&
3124             !(global_ctx->options & E2F_OPT_UNSHARE_BLOCKS))
3125                 return 0;
3126         /*
3127          * This need be done after merging block_ea_map
3128          * because ea block might be shared, we need exclude
3129          * them from dup blocks.
3130          */
3131         e2fsck_pass1_block_map_w_lock(thread_ctx);
3132         retval = ext2fs_merge_bitmap(thread_ctx->block_found_map,
3133                                      global_ctx->block_found_map,
3134                                      global_ctx->block_dup_map,
3135                                      global_ctx->block_ea_map);
3136         e2fsck_pass1_block_map_w_unlock(thread_ctx);
3137         if (retval == EEXIST)
3138                 global_ctx->flags |= E2F_FLAG_DUP_BLOCK;
3139
3140         return 0;
3141 }
3142
3143 static int e2fsck_pass1_thread_join(e2fsck_t global_ctx, e2fsck_t thread_ctx)
3144 {
3145         errcode_t       retval;
3146
3147         retval = e2fsck_pass1_thread_join_one(global_ctx, thread_ctx);
3148         ext2fs_free_mem(&thread_ctx->fs);
3149         if (thread_ctx->logf)
3150                 fclose(thread_ctx->logf);
3151         if (thread_ctx->problem_logf) {
3152                 fputs("</problem_log>\n", thread_ctx->problem_logf);
3153                 fclose(thread_ctx->problem_logf);
3154         }
3155         e2fsck_pass1_free_bitmap(&thread_ctx->inode_used_map);
3156         e2fsck_pass1_free_bitmap(&thread_ctx->inode_bad_map);
3157         e2fsck_pass1_free_bitmap(&thread_ctx->inode_dir_map);
3158         e2fsck_pass1_free_bitmap(&thread_ctx->inode_bb_map);
3159         e2fsck_pass1_free_bitmap(&thread_ctx->inode_imagic_map);
3160         e2fsck_pass1_free_bitmap(&thread_ctx->inode_reg_map);
3161         e2fsck_pass1_free_bitmap(&thread_ctx->inodes_to_rebuild);
3162         e2fsck_pass1_free_bitmap(&thread_ctx->block_found_map);
3163         e2fsck_pass1_free_bitmap(&thread_ctx->block_ea_map);
3164         if (thread_ctx->refcount)
3165                 ea_refcount_free(thread_ctx->refcount);
3166         if (thread_ctx->refcount_extra)
3167                 ea_refcount_free(thread_ctx->refcount_extra);
3168         if (thread_ctx->ea_inode_refs)
3169                 ea_refcount_free(thread_ctx->ea_inode_refs);
3170         if (thread_ctx->refcount_orig)
3171                 ea_refcount_free(thread_ctx->refcount_orig);
3172         e2fsck_free_dir_info(thread_ctx);
3173         ext2fs_free_icount(thread_ctx->inode_count);
3174         ext2fs_free_icount(thread_ctx->inode_link_info);
3175         if (thread_ctx->dirs_to_hash)
3176                 ext2fs_badblocks_list_free(thread_ctx->dirs_to_hash);
3177         quota_release_context(&thread_ctx->qctx);
3178         ext2fs_free_mem(&thread_ctx->invalid_block_bitmap_flag);
3179         ext2fs_free_mem(&thread_ctx->invalid_inode_bitmap_flag);
3180         ext2fs_free_mem(&thread_ctx->invalid_inode_table_flag);
3181         ext2fs_free_mem(&thread_ctx);
3182
3183         return retval;
3184 }
3185
3186 static int e2fsck_pass1_threads_join(struct e2fsck_thread_info *infos,
3187                                      e2fsck_t global_ctx)
3188 {
3189         errcode_t                        rc;
3190         errcode_t                        ret = 0;
3191         int                              i;
3192         struct e2fsck_thread_info       *pinfo;
3193         int                              num_threads = global_ctx->fs_num_threads;
3194
3195         /* merge invalid bitmaps will recalculate it */
3196         global_ctx->invalid_bitmaps = 0;
3197         for (i = 0; i < num_threads; i++) {
3198                 pinfo = &infos[i];
3199
3200                 if (!pinfo->eti_started)
3201                         continue;
3202
3203                 rc = pthread_join(pinfo->eti_thread_id, NULL);
3204                 if (rc) {
3205                         com_err(global_ctx->program_name, rc,
3206                                 _("while joining thread\n"));
3207                         if (ret == 0)
3208                                 ret = rc;
3209                 }
3210                 rc = e2fsck_pass1_thread_join(global_ctx, infos[i].eti_thread_ctx);
3211                 if (rc) {
3212                         com_err(global_ctx->program_name, rc,
3213                                 _("while joining pass1 thread\n"));
3214                         if (ret == 0)
3215                                 ret = rc;
3216                 }
3217         }
3218         free(infos);
3219
3220         return ret;
3221 }
3222
3223 static void *e2fsck_pass1_thread(void *arg)
3224 {
3225         struct e2fsck_thread_info       *info = arg;
3226         e2fsck_t                         thread_ctx = info->eti_thread_ctx;
3227 #ifdef DEBUG_THREADS
3228         struct e2fsck_thread_debug      *thread_debug = info->eti_debug;
3229 #endif
3230
3231 #ifdef DEBUG_THREADS
3232         pthread_mutex_lock(&thread_debug->etd_mutex);
3233         while (info->eti_thread_index > thread_debug->etd_finished_threads) {
3234                 pthread_cond_wait(&thread_debug->etd_cond,
3235                                   &thread_debug->etd_mutex);
3236         }
3237         pthread_mutex_unlock(&thread_debug->etd_mutex);
3238 #endif
3239
3240 #ifdef HAVE_SETJMP_H
3241         /*
3242          * When fatal_error() happens, jump to here. The thread
3243          * context's flags will be saved, but its abort_loc will
3244          * be overwritten by original jump buffer for the later
3245          * tests.
3246          */
3247         if (setjmp(thread_ctx->abort_loc)) {
3248                 thread_ctx->flags &= ~E2F_FLAG_SETJMP_OK;
3249                 goto out;
3250         }
3251         thread_ctx->flags |= E2F_FLAG_SETJMP_OK;
3252 #endif
3253
3254         e2fsck_pass1_run(thread_ctx);
3255
3256 out:
3257         if (thread_ctx->options & E2F_OPT_MULTITHREAD)
3258                 log_out(thread_ctx,
3259                         _("Scanned group range [%lu, %lu), inodes %lu\n"),
3260                         thread_ctx->thread_info.et_group_start,
3261                         thread_ctx->thread_info.et_group_end,
3262                         thread_ctx->thread_info.et_inode_number);
3263
3264 #ifdef DEBUG_THREADS
3265         pthread_mutex_lock(&thread_debug->etd_mutex);
3266         thread_debug->etd_finished_threads++;
3267         pthread_cond_broadcast(&thread_debug->etd_cond);
3268         pthread_mutex_unlock(&thread_debug->etd_mutex);
3269 #endif
3270
3271         return NULL;
3272 }
3273
3274 static int e2fsck_pass1_threads_start(struct e2fsck_thread_info **pinfo,
3275                                       e2fsck_t global_ctx)
3276 {
3277         struct e2fsck_thread_info       *infos;
3278         pthread_attr_t                   attr;
3279         errcode_t                        retval;
3280         errcode_t                        ret;
3281         struct e2fsck_thread_info       *tmp_pinfo;
3282         int                              i;
3283         e2fsck_t                         thread_ctx;
3284         dgrp_t                           average_group;
3285         int                              num_threads = global_ctx->fs_num_threads;
3286 #ifdef DEBUG_THREADS
3287         struct e2fsck_thread_debug       thread_debug =
3288                 {PTHREAD_MUTEX_INITIALIZER, PTHREAD_COND_INITIALIZER, 0};
3289
3290         thread_debug.etd_finished_threads = 0;
3291 #endif
3292
3293         retval = pthread_attr_init(&attr);
3294         if (retval) {
3295                 com_err(global_ctx->program_name, retval,
3296                         _("while setting pthread attribute\n"));
3297                 return retval;
3298         }
3299
3300         infos = calloc(num_threads, sizeof(struct e2fsck_thread_info));
3301         if (infos == NULL) {
3302                 retval = -ENOMEM;
3303                 com_err(global_ctx->program_name, retval,
3304                         _("while allocating memory for threads\n"));
3305                 pthread_attr_destroy(&attr);
3306                 return retval;
3307         }
3308
3309         average_group = ext2fs_get_avg_group(global_ctx->fs);
3310         for (i = 0; i < num_threads; i++) {
3311                 tmp_pinfo = &infos[i];
3312                 tmp_pinfo->eti_thread_index = i;
3313 #ifdef DEBUG_THREADS
3314                 tmp_pinfo->eti_debug = &thread_debug;
3315 #endif
3316                 retval = e2fsck_pass1_thread_prepare(global_ctx, &thread_ctx,
3317                                                      i, num_threads,
3318                                                      average_group);
3319                 if (retval) {
3320                         com_err(global_ctx->program_name, retval,
3321                                 _("while preparing pass1 thread\n"));
3322                         break;
3323                 }
3324                 tmp_pinfo->eti_thread_ctx = thread_ctx;
3325
3326                 retval = pthread_create(&tmp_pinfo->eti_thread_id, &attr,
3327                                         &e2fsck_pass1_thread, tmp_pinfo);
3328                 if (retval) {
3329                         com_err(global_ctx->program_name, retval,
3330                                 _("while creating thread\n"));
3331                         e2fsck_pass1_thread_join(global_ctx, thread_ctx);
3332                         break;
3333                 }
3334
3335                 tmp_pinfo->eti_started = 1;
3336         }
3337
3338         /* destroy the thread attribute object, since it is no longer needed */
3339         ret = pthread_attr_destroy(&attr);
3340         if (ret) {
3341                 com_err(global_ctx->program_name, ret,
3342                         _("while destroying thread attribute\n"));
3343                 if (retval == 0)
3344                         retval = ret;
3345         }
3346
3347         if (retval) {
3348                 e2fsck_pass1_threads_join(infos, global_ctx);
3349                 return retval;
3350         }
3351         *pinfo = infos;
3352         return 0;
3353 }
3354
3355 static void e2fsck_pass1_multithread(e2fsck_t global_ctx)
3356 {
3357         struct e2fsck_thread_info *infos = NULL;
3358         errcode_t retval;
3359
3360         retval = e2fsck_pass1_threads_start(&infos, global_ctx);
3361         if (retval) {
3362                 com_err(global_ctx->program_name, retval,
3363                         _("while starting pass1 threads\n"));
3364                 goto out_abort;
3365         }
3366
3367         retval = e2fsck_pass1_threads_join(infos, global_ctx);
3368         if (retval) {
3369                 com_err(global_ctx->program_name, retval,
3370                         _("while joining pass1 threads\n"));
3371                 goto out_abort;
3372         }
3373         return;
3374 out_abort:
3375         global_ctx->flags |= E2F_FLAG_ABORT;
3376         return;
3377 }
3378 #endif
3379
3380 void e2fsck_pass1(e2fsck_t ctx)
3381 {
3382         errcode_t retval;
3383         int need_single = 1;
3384
3385         retval = e2fsck_pass1_prepare(ctx);
3386         if (retval)
3387                 return;
3388 #ifdef HAVE_PTHREAD
3389         if (ctx->fs_num_threads > 1 ||
3390             ctx->options & E2F_OPT_MULTITHREAD) {
3391                 need_single = 0;
3392                 e2fsck_pass1_multithread(ctx);
3393         }
3394 #endif
3395         if (need_single)
3396                 e2fsck_pass1_run(ctx);
3397         e2fsck_pass1_post(ctx);
3398 }
3399
3400 #undef FINISH_INODE_LOOP
3401
3402 /*
3403  * When the inode_scan routines call this callback at the end of the
3404  * glock group, call process_inodes.
3405  */
3406 static errcode_t scan_callback(ext2_filsys fs,
3407                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
3408                                dgrp_t group, void * priv_data)
3409 {
3410         struct scan_callback_struct *scan_struct;
3411         e2fsck_t ctx;
3412         struct e2fsck_thread *tinfo;
3413
3414         scan_struct = (struct scan_callback_struct *) priv_data;
3415         ctx = scan_struct->ctx;
3416
3417         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf,
3418                        scan_struct->inodes_to_process,
3419                        scan_struct->process_inode_count);
3420
3421         if (ctx->progress)
3422                 if ((ctx->progress)(ctx, 1, group+1,
3423                                     ctx->fs->group_desc_count))
3424                         return EXT2_ET_CANCEL_REQUESTED;
3425
3426 #ifdef HAVE_PTHREAD
3427         if (ctx->global_ctx) {
3428                 tinfo = &ctx->thread_info;
3429                 tinfo->et_group_next++;
3430                 if (ctx->options & E2F_OPT_DEBUG &&
3431                     ctx->options & E2F_OPT_MULTITHREAD)
3432                         log_out(ctx, _("group %d finished\n"),
3433                                 tinfo->et_group_next);
3434                 if (tinfo->et_group_next >= tinfo->et_group_end)
3435                         return EXT2_ET_SCAN_FINISHED;
3436         }
3437 #endif
3438
3439         return 0;
3440 }
3441
3442 /*
3443  * Process the inodes in the "inodes to process" list.
3444  */
3445 static void process_inodes(e2fsck_t ctx, char *block_buf,
3446                            struct process_inode_block *inodes_to_process,
3447                            int *process_inode_count)
3448 {
3449         int                     i;
3450         struct ext2_inode       *old_stashed_inode;
3451         ext2_ino_t              old_stashed_ino;
3452         const char              *old_operation;
3453         char                    buf[80];
3454         struct problem_context  pctx;
3455
3456 #if 0
3457         printf("begin process_inodes: ");
3458 #endif
3459         if (*process_inode_count == 0)
3460                 return;
3461         old_operation = ehandler_operation(0);
3462         old_stashed_inode = ctx->stashed_inode;
3463         old_stashed_ino = ctx->stashed_ino;
3464         qsort(inodes_to_process, *process_inode_count,
3465                       sizeof(struct process_inode_block), process_inode_cmp);
3466         clear_problem_context(&pctx);
3467         for (i=0; i < *process_inode_count; i++) {
3468                 pctx.inode = ctx->stashed_inode =
3469                         (struct ext2_inode *) &inodes_to_process[i].inode;
3470                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
3471
3472 #if 0
3473                 printf("%u ", pctx.ino);
3474 #endif
3475                 sprintf(buf, _("reading indirect blocks of inode %u"),
3476                         pctx.ino);
3477                 ehandler_operation(buf);
3478                 check_blocks(ctx, &pctx, block_buf,
3479                              &inodes_to_process[i].ea_ibody_quota);
3480                 if (e2fsck_should_abort(ctx))
3481                         break;
3482         }
3483         ctx->stashed_inode = old_stashed_inode;
3484         ctx->stashed_ino = old_stashed_ino;
3485         *process_inode_count = 0;
3486 #if 0
3487         printf("end process inodes\n");
3488 #endif
3489         ehandler_operation(old_operation);
3490 }
3491
3492 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
3493 {
3494         const struct process_inode_block *ib_a =
3495                 (const struct process_inode_block *) a;
3496         const struct process_inode_block *ib_b =
3497                 (const struct process_inode_block *) b;
3498         int     ret;
3499
3500         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
3501                ib_b->inode.i_block[EXT2_IND_BLOCK]);
3502         if (ret == 0)
3503                 /*
3504                  * We only call process_inodes() for non-extent
3505                  * inodes, so it's OK to pass NULL to
3506                  * ext2fs_file_acl_block() here.
3507                  */
3508                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
3509                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
3510         if (ret == 0)
3511                 ret = ib_a->ino - ib_b->ino;
3512         return ret;
3513 }
3514
3515 /*
3516  * Mark an inode as being bad in some what
3517  */
3518 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
3519 {
3520         struct          problem_context pctx;
3521
3522         if (!ctx->inode_bad_map) {
3523                 clear_problem_context(&pctx);
3524
3525                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3526                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
3527                                 "inode_bad_map", &ctx->inode_bad_map);
3528                 if (pctx.errcode) {
3529                         pctx.num = 3;
3530                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3531                         /* Should never get here */
3532                         ctx->flags |= E2F_FLAG_ABORT;
3533                         return;
3534                 }
3535         }
3536         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
3537 }
3538
3539 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino)
3540 {
3541         struct          problem_context pctx;
3542
3543         if (!ctx->casefolded_dirs) {
3544                 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
3545                 if (pctx.errcode)
3546                         goto error;
3547         }
3548         pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
3549         if (pctx.errcode == 0)
3550                 return;
3551 error:
3552         fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
3553         /* Should never get here */
3554         ctx->flags |= E2F_FLAG_ABORT;
3555 }
3556
3557 /*
3558  * This procedure will allocate the inode "bb" (badblock) map table
3559  */
3560 static void alloc_bb_map(e2fsck_t ctx)
3561 {
3562         struct          problem_context pctx;
3563
3564         clear_problem_context(&pctx);
3565         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3566                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
3567                         "inode_bb_map", &ctx->inode_bb_map);
3568         if (pctx.errcode) {
3569                 pctx.num = 4;
3570                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3571                 /* Should never get here */
3572                 ctx->flags |= E2F_FLAG_ABORT;
3573                 return;
3574         }
3575 }
3576
3577 /*
3578  * This procedure will allocate the inode imagic table
3579  */
3580 static void alloc_imagic_map(e2fsck_t ctx)
3581 {
3582         struct          problem_context pctx;
3583
3584         clear_problem_context(&pctx);
3585         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3586                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
3587                         "inode_imagic_map", &ctx->inode_imagic_map);
3588         if (pctx.errcode) {
3589                 pctx.num = 5;
3590                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3591                 /* Should never get here */
3592                 ctx->flags |= E2F_FLAG_ABORT;
3593                 return;
3594         }
3595 }
3596
3597 /*
3598  * Marks a block as in use, setting the dup_map if it's been set
3599  * already.  Called by process_block and process_bad_block.
3600  *
3601  * WARNING: Assumes checks have already been done to make sure block
3602  * is valid.  This is true in both process_block and process_bad_block.
3603  */
3604 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
3605 {
3606         struct problem_context pctx;
3607         e2fsck_t global_ctx = ctx->global_ctx ? ctx->global_ctx : ctx;
3608
3609         clear_problem_context(&pctx);
3610
3611         if (is_blocks_used(ctx, block, 1)) {
3612                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
3613                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
3614                         return;
3615                 }
3616                 ctx->flags |= E2F_FLAG_DUP_BLOCK;
3617                 e2fsck_pass1_block_map_w_lock(ctx);
3618                 ext2fs_fast_mark_block_bitmap2(global_ctx->block_dup_map, block);
3619                 e2fsck_pass1_block_map_w_unlock(ctx);
3620         } else {
3621                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
3622         }
3623 }
3624
3625 /*
3626  * When cluster size is greater than one block, it is caller's responsibility
3627  * to make sure block parameter starts at a cluster boundary.
3628  */
3629 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
3630                                       unsigned int num)
3631 {
3632         if (!is_blocks_used(ctx, block, num)) {
3633                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
3634         } else {
3635                 unsigned int i;
3636
3637                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
3638                         mark_block_used(ctx, block + i);
3639         }
3640 }
3641
3642 static errcode_t _INLINE_ e2fsck_write_ext_attr3(e2fsck_t ctx, blk64_t block,
3643                                                  void *inbuf, ext2_ino_t inum)
3644 {
3645         errcode_t retval;
3646         ext2_filsys fs = ctx->fs;
3647
3648         e2fsck_pass1_fix_lock(ctx);
3649         retval = ext2fs_write_ext_attr3(fs, block, inbuf, inum);
3650         e2fsck_pass1_fix_unlock(ctx);
3651
3652         return retval;
3653 }
3654 /*
3655  * Adjust the extended attribute block's reference counts at the end
3656  * of pass 1, either by subtracting out references for EA blocks that
3657  * are still referenced in ctx->refcount, or by adding references for
3658  * EA blocks that had extra references as accounted for in
3659  * ctx->refcount_extra.
3660  */
3661 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
3662                                     char *block_buf, int adjust_sign)
3663 {
3664         struct ext2_ext_attr_header     *header;
3665         struct problem_context          pctx;
3666         ext2_filsys                     fs = ctx->fs;
3667         blk64_t                         blk;
3668         __u32                           should_be;
3669         ea_value_t                      count;
3670
3671         clear_problem_context(&pctx);
3672
3673         ea_refcount_intr_begin(refcount);
3674         while (1) {
3675                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
3676                         break;
3677                 pctx.blk = blk;
3678                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
3679                                                      pctx.ino);
3680                 if (pctx.errcode) {
3681                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
3682                         return;
3683                 }
3684                 header = (struct ext2_ext_attr_header *) block_buf;
3685                 pctx.blkcount = header->h_refcount;
3686                 should_be = header->h_refcount + adjust_sign * (int)count;
3687                 pctx.num = should_be;
3688                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
3689                         header->h_refcount = should_be;
3690                         pctx.errcode = e2fsck_write_ext_attr3(ctx, blk,
3691                                                              block_buf,
3692                                                              pctx.ino);
3693                         if (pctx.errcode) {
3694                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
3695                                             &pctx);
3696                                 continue;
3697                         }
3698                 }
3699         }
3700 }
3701
3702 /*
3703  * Handle processing the extended attribute blocks
3704  */
3705 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
3706                            char *block_buf, struct ea_quota *ea_block_quota)
3707 {
3708         ext2_filsys fs = ctx->fs;
3709         ext2_ino_t      ino = pctx->ino;
3710         struct ext2_inode *inode = pctx->inode;
3711         blk64_t         blk;
3712         char *          end;
3713         struct ext2_ext_attr_header *header;
3714         struct ext2_ext_attr_entry *first, *entry;
3715         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
3716         __u64           quota_inodes = 0;
3717         region_t        region = 0;
3718         int             failed_csum = 0;
3719
3720         ea_block_quota->blocks = 0;
3721         ea_block_quota->inodes = 0;
3722
3723         blk = ext2fs_file_acl_block(fs, inode);
3724         if (blk == 0)
3725                 return 0;
3726
3727         /*
3728          * If the Extended attribute flag isn't set, then a non-zero
3729          * file acl means that the inode is corrupted.
3730          *
3731          * Or if the extended attribute block is an invalid block,
3732          * then the inode is also corrupted.
3733          */
3734         if (!ext2fs_has_feature_xattr(fs->super) ||
3735             (blk < fs->super->s_first_data_block) ||
3736             (blk >= ext2fs_blocks_count(fs->super))) {
3737                 mark_inode_bad(ctx, ino);
3738                 return 0;
3739         }
3740
3741         /* If ea bitmap hasn't been allocated, create it */
3742         if (!ctx->block_ea_map) {
3743                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
3744                                         _("ext attr block map"),
3745                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
3746                                         &ctx->block_ea_map);
3747                 if (pctx->errcode) {
3748                         pctx->num = 2;
3749                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
3750                         ctx->flags |= E2F_FLAG_ABORT;
3751                         return 0;
3752                 }
3753         }
3754
3755         /* Create the EA refcount structure if necessary */
3756         if (!ctx->refcount) {
3757                 pctx->errcode = ea_refcount_create(0,
3758                                         &ctx->refcount_orig);
3759                 if (pctx->errcode) {
3760                         pctx->num = 1;
3761                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3762                         ctx->flags |= E2F_FLAG_ABORT;
3763                         return 0;
3764                 }
3765
3766                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
3767                 if (pctx->errcode) {
3768                         pctx->num = 1;
3769                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3770                         ctx->flags |= E2F_FLAG_ABORT;
3771                         return 0;
3772                 }
3773         }
3774
3775 #if 0
3776         /* Debugging text */
3777         printf("Inode %u has EA block %u\n", ino, blk);
3778 #endif
3779
3780         /* Have we seen this EA block before? */
3781         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
3782                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
3783                 ea_block_quota->inodes = 0;
3784
3785                 if (ctx->ea_block_quota_blocks) {
3786                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
3787                                           &quota_blocks);
3788                         if (quota_blocks)
3789                                 ea_block_quota->blocks = quota_blocks;
3790                 }
3791
3792                 if (ctx->ea_block_quota_inodes)
3793                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
3794                                           &ea_block_quota->inodes);
3795
3796                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
3797                         return 1;
3798                 /* Ooops, this EA was referenced more than it stated */
3799                 if (!ctx->refcount_extra) {
3800                         pctx->errcode = ea_refcount_create(0,
3801                                            &ctx->refcount_extra);
3802                         if (pctx->errcode) {
3803                                 pctx->num = 2;
3804                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3805                                 ctx->flags |= E2F_FLAG_ABORT;
3806                                 return 0;
3807                         }
3808                 }
3809                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
3810                 return 1;
3811         }
3812
3813         /*
3814          * OK, we haven't seen this EA block yet.  So we need to
3815          * validate it
3816          */
3817         pctx->blk = blk;
3818         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
3819         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
3820                 pctx->errcode = 0;
3821                 failed_csum = 1;
3822         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
3823                 pctx->errcode = 0;
3824
3825         if (pctx->errcode &&
3826             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
3827                 pctx->errcode = 0;
3828                 goto clear_extattr;
3829         }
3830         header = (struct ext2_ext_attr_header *) block_buf;
3831         pctx->blk = ext2fs_file_acl_block(fs, inode);
3832         if (((ctx->ext_attr_ver == 1) &&
3833              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
3834             ((ctx->ext_attr_ver == 2) &&
3835              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
3836                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
3837                         goto clear_extattr;
3838         }
3839
3840         if (header->h_blocks != 1) {
3841                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
3842                         goto clear_extattr;
3843         }
3844
3845         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
3846                 goto clear_extattr;
3847
3848         region = region_create(0, fs->blocksize);
3849         if (!region) {
3850                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
3851                 ctx->flags |= E2F_FLAG_ABORT;
3852                 return 0;
3853         }
3854         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
3855                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3856                         goto clear_extattr;
3857         }
3858
3859         first = (struct ext2_ext_attr_entry *)(header+1);
3860         end = block_buf + fs->blocksize;
3861         entry = first;
3862         while ((char *)entry < end && *(__u32 *)entry) {
3863                 __u32 hash;
3864
3865                 if (region_allocate(region, (char *)entry - (char *)header,
3866                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
3867                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3868                                 goto clear_extattr;
3869                         break;
3870                 }
3871                 if ((ctx->ext_attr_ver == 1 &&
3872                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
3873                     (ctx->ext_attr_ver == 2 &&
3874                      entry->e_name_index == 0)) {
3875                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
3876                                 goto clear_extattr;
3877                         break;
3878                 }
3879                 if (entry->e_value_inum == 0) {
3880                         if (entry->e_value_offs + entry->e_value_size >
3881                             fs->blocksize) {
3882                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
3883                                         goto clear_extattr;
3884                                 break;
3885                         }
3886                         if (entry->e_value_size &&
3887                             region_allocate(region, entry->e_value_offs,
3888                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
3889                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
3890                                                 pctx))
3891                                         goto clear_extattr;
3892                         }
3893
3894                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
3895                                                           entry->e_value_offs);
3896
3897                         if (entry->e_hash != hash) {
3898                                 pctx->num = entry->e_hash;
3899                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
3900                                         goto clear_extattr;
3901                                 entry->e_hash = hash;
3902                         }
3903                 } else {
3904                         problem_t problem;
3905                         blk64_t entry_quota_blocks;
3906
3907                         problem = check_large_ea_inode(ctx, entry, pctx,
3908                                                        &entry_quota_blocks);
3909                         if (problem && fix_problem(ctx, problem, pctx))
3910                                 goto clear_extattr;
3911
3912                         quota_blocks += entry_quota_blocks;
3913                         quota_inodes++;
3914                 }
3915
3916                 entry = EXT2_EXT_ATTR_NEXT(entry);
3917         }
3918         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
3919                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3920                         goto clear_extattr;
3921         }
3922         region_free(region);
3923
3924         /*
3925          * We only get here if there was no other errors that were fixed.
3926          * If there was a checksum fail, ask to correct it.
3927          */
3928         if (failed_csum &&
3929             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
3930                 pctx->errcode = e2fsck_write_ext_attr3(ctx, blk, block_buf,
3931                                                        pctx->ino);
3932                 if (pctx->errcode)
3933                         return 0;
3934         }
3935
3936         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
3937                 if (!ctx->ea_block_quota_blocks) {
3938                         pctx->errcode = ea_refcount_create(0,
3939                                                 &ctx->ea_block_quota_blocks);
3940                         if (pctx->errcode) {
3941                                 pctx->num = 3;
3942                                 goto refcount_fail;
3943                         }
3944                 }
3945                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
3946                                   quota_blocks);
3947         }
3948
3949         if (quota_inodes) {
3950                 if (!ctx->ea_block_quota_inodes) {
3951                         pctx->errcode = ea_refcount_create(0,
3952                                                 &ctx->ea_block_quota_inodes);
3953                         if (pctx->errcode) {
3954                                 pctx->num = 4;
3955 refcount_fail:
3956                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3957                                 ctx->flags |= E2F_FLAG_ABORT;
3958                                 return 0;
3959                         }
3960                 }
3961
3962                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
3963                                   quota_inodes);
3964         }
3965         ea_block_quota->blocks = quota_blocks;
3966         ea_block_quota->inodes = quota_inodes;
3967
3968         inc_ea_inode_refs(ctx, pctx, first, end);
3969         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
3970         ea_refcount_store(ctx->refcount_orig, blk, header->h_refcount);
3971         /**
3972          * It might be racy that this block has been merged in the
3973          * global found map.
3974          */
3975         if (!is_blocks_used(ctx, blk, 1))
3976                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, blk);
3977         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
3978         return 1;
3979
3980 clear_extattr:
3981         if (region)
3982                 region_free(region);
3983         ext2fs_file_acl_block_set(fs, inode, 0);
3984         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
3985         return 0;
3986 }
3987
3988 /* Returns 1 if bad htree, 0 if OK */
3989 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
3990                         ext2_ino_t ino, struct ext2_inode *inode,
3991                         char *block_buf)
3992 {
3993         struct ext2_dx_root_info        *root;
3994         ext2_filsys                     fs = ctx->fs;
3995         errcode_t                       retval;
3996         blk64_t                         blk;
3997
3998         if ((!LINUX_S_ISDIR(inode->i_mode) &&
3999              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
4000             (!ext2fs_has_feature_dir_index(fs->super) &&
4001              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
4002                 return 1;
4003
4004         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
4005
4006         if ((pctx->errcode) ||
4007             (blk == 0) ||
4008             (blk < fs->super->s_first_data_block) ||
4009             (blk >= ext2fs_blocks_count(fs->super))) {
4010                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
4011                         return 1;
4012                 else
4013                         return 0;
4014         }
4015
4016         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
4017         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
4018                 return 1;
4019
4020         /* XXX should check that beginning matches a directory */
4021         root = (struct ext2_dx_root_info *) (block_buf + 24);
4022
4023         if ((root->reserved_zero || root->info_length < 8) &&
4024             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
4025                 return 1;
4026
4027         pctx->num = root->hash_version;
4028         if ((root->hash_version != EXT2_HASH_LEGACY) &&
4029             (root->hash_version != EXT2_HASH_HALF_MD4) &&
4030             (root->hash_version != EXT2_HASH_TEA) &&
4031             (root->hash_version != EXT2_HASH_SIPHASH) &&
4032             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
4033                 return 1;
4034
4035         if (ext4_hash_in_dirent(inode)) {
4036                 if (root->hash_version != EXT2_HASH_SIPHASH &&
4037                     fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
4038                         return 1;
4039         } else {
4040                 if (root->hash_version == EXT2_HASH_SIPHASH &&
4041                    fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
4042                         return 1;
4043         }
4044
4045         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
4046             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
4047                 return 1;
4048
4049         pctx->num = root->indirect_levels;
4050         /* if htree level is clearly too high, consider it to be broken */
4051         if (root->indirect_levels > EXT4_HTREE_LEVEL &&
4052             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
4053                 return 1;
4054
4055         /* if level is only maybe too high, LARGE_DIR feature could be unset */
4056         if (root->indirect_levels > ext2_dir_htree_level(fs) &&
4057             !ext2fs_has_feature_largedir(fs->super)) {
4058                 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
4059                 int idx_pb = 1 << (blockbits - 3);
4060
4061                 /* compare inode size/blocks vs. max-sized 2-level htree */
4062                 if (EXT2_I_SIZE(pctx->inode) <
4063                     (idx_pb - 1) * (idx_pb - 2) << blockbits &&
4064                     pctx->inode->i_blocks <
4065                     (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
4066                     fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
4067                         return 1;
4068         }
4069
4070         if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
4071             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4072                 ctx->large_dirs++;
4073
4074         return 0;
4075 }
4076
4077 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
4078                         struct ext2_inode *inode, int restart_flag,
4079                         const char *source)
4080 {
4081         inode->i_flags = 0;
4082         inode->i_links_count = 0;
4083         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
4084         inode->i_dtime = ctx->now;
4085
4086         /*
4087          * If a special inode has such rotten block mappings that we
4088          * want to clear the whole inode, be sure to actually zap
4089          * the block maps because i_links_count isn't checked for
4090          * special inodes, and we'll end up right back here the next
4091          * time we run fsck.
4092          */
4093         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
4094                 memset(inode->i_block, 0, sizeof(inode->i_block));
4095
4096         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
4097         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
4098         if (ctx->inode_reg_map)
4099                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
4100         if (ctx->inode_bad_map)
4101                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
4102
4103         /*
4104          * If the inode was partially accounted for before processing
4105          * was aborted, we need to restart the pass 1 scan.
4106          */
4107         ctx->flags |= restart_flag;
4108
4109         if (ino == EXT2_BAD_INO)
4110                 memset(inode, 0, sizeof(struct ext2_inode));
4111
4112         e2fsck_write_inode(ctx, ino, inode, source);
4113 }
4114
4115 /*
4116  * Use the multiple-blocks reclamation code to fix alignment problems in
4117  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
4118  * physical cluster, and we want the block offsets within that cluster to
4119  * line up.
4120  */
4121 static int has_unaligned_cluster_map(e2fsck_t ctx,
4122                                      blk64_t last_pblk, blk64_t last_lblk,
4123                                      blk64_t pblk, blk64_t lblk)
4124 {
4125         blk64_t cluster_mask;
4126
4127         if (!ctx->fs->cluster_ratio_bits)
4128                 return 0;
4129         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
4130
4131         /*
4132          * If the block in the logical cluster doesn't align with the block in
4133          * the physical cluster...
4134          */
4135         if ((lblk & cluster_mask) != (pblk & cluster_mask))
4136                 return 1;
4137
4138         /*
4139          * If we cross a physical cluster boundary within a logical cluster...
4140          */
4141         if (last_pblk && (lblk & cluster_mask) != 0 &&
4142             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
4143             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
4144                 return 1;
4145
4146         return 0;
4147 }
4148
4149 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
4150                              struct process_block_struct *pb,
4151                              blk64_t start_block, blk64_t end_block,
4152                              blk64_t eof_block,
4153                              ext2_extent_handle_t ehandle,
4154                              int try_repairs)
4155 {
4156         struct ext2fs_extent    extent;
4157         blk64_t                 blk, last_lblk;
4158         unsigned int            i, n;
4159         int                     is_dir, is_leaf;
4160         problem_t               problem;
4161         struct ext2_extent_info info;
4162         int                     failed_csum = 0;
4163
4164         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
4165                 failed_csum = 1;
4166
4167         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
4168         if (pctx->errcode)
4169                 return;
4170         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
4171             !pb->eti.force_rebuild) {
4172                 struct extent_tree_level *etl;
4173
4174                 etl = pb->eti.ext_info + info.curr_level;
4175                 etl->num_extents += info.num_entries;
4176                 etl->max_extents += info.max_entries;
4177                 /*
4178                  * Implementation wart: Splitting extent blocks when appending
4179                  * will leave the old block with one free entry.  Therefore
4180                  * unless the node is totally full, pretend that a non-root
4181                  * extent block can hold one fewer entry than it actually does,
4182                  * so that we don't repeatedly rebuild the extent tree.
4183                  */
4184                 if (info.curr_level && info.num_entries < info.max_entries)
4185                         etl->max_extents--;
4186         }
4187
4188         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
4189                                           &extent);
4190         while ((pctx->errcode == 0 ||
4191                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
4192                info.num_entries-- > 0) {
4193                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
4194                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
4195                 last_lblk = extent.e_lblk + extent.e_len - 1;
4196
4197                 problem = 0;
4198                 pctx->blk = extent.e_pblk;
4199                 pctx->blk2 = extent.e_lblk;
4200                 pctx->num = extent.e_len;
4201                 pctx->blkcount = extent.e_lblk + extent.e_len;
4202
4203                 if (extent.e_pblk == 0 ||
4204                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
4205                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
4206                         problem = PR_1_EXTENT_BAD_START_BLK;
4207                 else if (extent.e_lblk < start_block)
4208                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
4209                 else if ((end_block && last_lblk > end_block) &&
4210                          !(last_lblk > eof_block &&
4211                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
4212                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
4213                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
4214                 else if (is_leaf && extent.e_len == 0)
4215                         problem = PR_1_EXTENT_LENGTH_ZERO;
4216                 else if (is_leaf &&
4217                          (extent.e_pblk + extent.e_len) >
4218                          ext2fs_blocks_count(ctx->fs->super))
4219                         problem = PR_1_EXTENT_ENDS_BEYOND;
4220                 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
4221                          !ext2fs_has_feature_largedir(ctx->fs->super) &&
4222                          ((extent.e_lblk + extent.e_len) >
4223                           (1U << (21 - ctx->fs->super->s_log_block_size))))
4224                         problem = PR_1_TOOBIG_DIR;
4225
4226                 if (is_leaf && problem == 0 && extent.e_len > 0) {
4227 #if 0
4228                         printf("extent_region(ino=%u, expect=%llu, "
4229                                "lblk=%llu, len=%u)\n", pb->ino,
4230                                (unsigned long long) pb->next_lblock,
4231                                (unsigned long long) extent.e_lblk,
4232                                extent.e_len);
4233 #endif
4234                         if (extent.e_lblk < pb->next_lblock)
4235                                 problem = PR_1_EXTENT_COLLISION;
4236                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
4237                                 pb->next_lblock = extent.e_lblk + extent.e_len;
4238                 }
4239
4240                 /*
4241                  * Uninitialized blocks in a directory?  Clear the flag and
4242                  * we'll interpret the blocks later.
4243                  */
4244                 if (try_repairs && is_dir && problem == 0 &&
4245                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
4246                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
4247                         e2fsck_pass1_fix_lock(ctx);
4248                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
4249                         pb->inode_modified = 1;
4250                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
4251                                                               &extent);
4252                         e2fsck_pass1_fix_unlock(ctx);
4253                         if (pctx->errcode)
4254                                 return;
4255                         failed_csum = 0;
4256                 }
4257 #ifdef CONFIG_DEVELOPER_FEATURES
4258                 if (try_repairs && !is_dir && problem == 0 &&
4259                     (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
4260                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
4261                     fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
4262                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
4263                         pb->inode_modified = 1;
4264                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
4265                                                               &extent);
4266                         if (pctx->errcode)
4267                                 return;
4268                         failed_csum = 0;
4269                 }
4270 #endif
4271                 if (try_repairs && problem) {
4272 report_problem:
4273                         if (fix_problem(ctx, problem, pctx)) {
4274                                 if (ctx->invalid_bitmaps) {
4275                                         /*
4276                                          * If fsck knows the bitmaps are bad,
4277                                          * skip to the next extent and
4278                                          * try to clear this extent again
4279                                          * after fixing the bitmaps, by
4280                                          * restarting fsck.
4281                                          */
4282                                         pctx->errcode = ext2fs_extent_get(
4283                                                           ehandle,
4284                                                           EXT2_EXTENT_NEXT_SIB,
4285                                                           &extent);
4286                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4287                                         if (pctx->errcode ==
4288                                                     EXT2_ET_NO_CURRENT_NODE) {
4289                                                 pctx->errcode = 0;
4290                                                 break;
4291                                         }
4292                                         continue;
4293                                 }
4294                                 e2fsck_pass1_fix_lock(ctx);
4295                                 e2fsck_read_bitmaps(ctx);
4296                                 pb->inode_modified = 1;
4297                                 pctx->errcode =
4298                                         ext2fs_extent_delete(ehandle, 0);
4299                                 e2fsck_pass1_fix_unlock(ctx);
4300                                 if (pctx->errcode) {
4301                                         pctx->str = "ext2fs_extent_delete";
4302                                         return;
4303                                 }
4304                                 e2fsck_pass1_fix_lock(ctx);
4305                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
4306                                 e2fsck_pass1_fix_unlock(ctx);
4307                                 if (pctx->errcode &&
4308                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
4309                                         pctx->str = "ext2fs_extent_fix_parents";
4310                                         return;
4311                                 }
4312                                 pctx->errcode = ext2fs_extent_get(ehandle,
4313                                                                   EXT2_EXTENT_CURRENT,
4314                                                                   &extent);
4315                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
4316                                         pctx->errcode = 0;
4317                                         break;
4318                                 }
4319                                 failed_csum = 0;
4320                                 continue;
4321                         }
4322                         goto next;
4323                 }
4324
4325                 if (!is_leaf) {
4326                         blk64_t lblk = extent.e_lblk;
4327                         int next_try_repairs = 1;
4328
4329                         blk = extent.e_pblk;
4330
4331                         /*
4332                          * If this lower extent block collides with critical
4333                          * metadata, don't try to repair the damage.  Pass 1b
4334                          * will reallocate the block; then we can try again.
4335                          */
4336                         if (pb->ino != EXT2_RESIZE_INO &&
4337                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
4338                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4339                                                       extent.e_pblk)) {
4340                                 next_try_repairs = 0;
4341                                 pctx->blk = blk;
4342                                 fix_problem(ctx,
4343                                             PR_1_CRITICAL_METADATA_COLLISION,
4344                                             pctx);
4345                                 if ((ctx->options & E2F_OPT_NO) == 0)
4346                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4347                         }
4348                         pctx->errcode = ext2fs_extent_get(ehandle,
4349                                                   EXT2_EXTENT_DOWN, &extent);
4350                         if (pctx->errcode &&
4351                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
4352                                 pctx->str = "EXT2_EXTENT_DOWN";
4353                                 problem = PR_1_EXTENT_HEADER_INVALID;
4354                                 if (!next_try_repairs)
4355                                         return;
4356                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
4357                                         goto report_problem;
4358                                 return;
4359                         }
4360                         /* The next extent should match this index's logical start */
4361                         if (extent.e_lblk != lblk) {
4362                                 struct ext2_extent_info e_info;
4363
4364                                 pctx->errcode = ext2fs_extent_get_info(ehandle,
4365                                                                        &e_info);
4366                                 if (pctx->errcode) {
4367                                         pctx->str = "ext2fs_extent_get_info";
4368                                         return;
4369                                 }
4370                                 pctx->blk = lblk;
4371                                 pctx->blk2 = extent.e_lblk;
4372                                 pctx->num = e_info.curr_level - 1;
4373                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
4374                                 if (fix_problem(ctx, problem, pctx)) {
4375                                         e2fsck_pass1_fix_lock(ctx);
4376                                         pb->inode_modified = 1;
4377                                         pctx->errcode =
4378                                                 ext2fs_extent_fix_parents(ehandle);
4379                                         e2fsck_pass1_fix_unlock(ctx);
4380                                         if (pctx->errcode) {
4381                                                 pctx->str = "ext2fs_extent_fix_parents";
4382                                                 return;
4383                                         }
4384                                 }
4385                         }
4386                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
4387                                          last_lblk, eof_block, ehandle,
4388                                          next_try_repairs);
4389                         if (pctx->errcode)
4390                                 return;
4391                         pctx->errcode = ext2fs_extent_get(ehandle,
4392                                                   EXT2_EXTENT_UP, &extent);
4393                         if (pctx->errcode) {
4394                                 pctx->str = "EXT2_EXTENT_UP";
4395                                 return;
4396                         }
4397                         mark_block_used(ctx, blk);
4398                         pb->num_blocks++;
4399                         goto next;
4400                 }
4401
4402                 if ((pb->previous_block != 0) &&
4403                     (pb->previous_block+1 != extent.e_pblk)) {
4404                         if (ctx->options & E2F_OPT_FRAGCHECK) {
4405                                 char type = '?';
4406
4407                                 if (pb->is_dir)
4408                                         type = 'd';
4409                                 else if (pb->is_reg)
4410                                         type = 'f';
4411
4412                                 printf(("%6lu(%c): expecting %6lu "
4413                                         "actual extent "
4414                                         "phys %6lu log %lu len %lu\n"),
4415                                        (unsigned long) pctx->ino, type,
4416                                        (unsigned long) pb->previous_block+1,
4417                                        (unsigned long) extent.e_pblk,
4418                                        (unsigned long) extent.e_lblk,
4419                                        (unsigned long) extent.e_len);
4420                         }
4421                         pb->fragmented = 1;
4422                 }
4423                 /*
4424                  * If we notice a gap in the logical block mappings of an
4425                  * extent-mapped directory, offer to close the hole by
4426                  * moving the logical block down, otherwise we'll go mad in
4427                  * pass 3 allocating empty directory blocks to fill the hole.
4428                  */
4429                 if (try_repairs && is_dir &&
4430                     pb->last_block + 1 < extent.e_lblk) {
4431                         blk64_t new_lblk;
4432
4433                         new_lblk = pb->last_block + 1;
4434                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
4435                                 new_lblk = ((new_lblk +
4436                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
4437                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
4438                                            (extent.e_pblk &
4439                                             EXT2FS_CLUSTER_MASK(ctx->fs));
4440                         pctx->blk = extent.e_lblk;
4441                         pctx->blk2 = new_lblk;
4442                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
4443                                 e2fsck_pass1_fix_lock(ctx);
4444                                 extent.e_lblk = new_lblk;
4445                                 pb->inode_modified = 1;
4446                                 pctx->errcode = ext2fs_extent_replace(ehandle,
4447                                                                 0, &extent);
4448                                 e2fsck_pass1_fix_unlock(ctx);
4449                                 if (pctx->errcode) {
4450                                         pctx->errcode = 0;
4451                                         goto alloc_later;
4452                                 }
4453                                 e2fsck_pass1_fix_lock(ctx);
4454                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
4455                                 e2fsck_pass1_fix_unlock(ctx);
4456                                 if (pctx->errcode)
4457                                         goto failed_add_dir_block;
4458                                 pctx->errcode = ext2fs_extent_goto(ehandle,
4459                                                                 extent.e_lblk);
4460                                 if (pctx->errcode)
4461                                         goto failed_add_dir_block;
4462                                 last_lblk = extent.e_lblk + extent.e_len - 1;
4463                                 failed_csum = 0;
4464                         }
4465                 }
4466 alloc_later:
4467                 if (is_dir) {
4468                         while (++pb->last_db_block <
4469                                (e2_blkcnt_t) extent.e_lblk) {
4470                                 pctx->errcode = ext2fs_add_dir_block2(
4471                                                         ctx->fs->dblist,
4472                                                         pb->ino, 0,
4473                                                         pb->last_db_block);
4474                                 if (pctx->errcode) {
4475                                         pctx->blk = 0;
4476                                         pctx->num = pb->last_db_block;
4477                                         goto failed_add_dir_block;
4478                                 }
4479                         }
4480
4481                         for (i = 0; i < extent.e_len; i++) {
4482                                 pctx->errcode = ext2fs_add_dir_block2(
4483                                                         ctx->fs->dblist,
4484                                                         pctx->ino,
4485                                                         extent.e_pblk + i,
4486                                                         extent.e_lblk + i);
4487                                 if (pctx->errcode) {
4488                                         pctx->blk = extent.e_pblk + i;
4489                                         pctx->num = extent.e_lblk + i;
4490                                 failed_add_dir_block:
4491                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4492                                         /* Should never get here */
4493                                         ctx->flags |= E2F_FLAG_ABORT;
4494                                         return;
4495                                 }
4496                         }
4497                         if (extent.e_len > 0)
4498                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
4499                 }
4500                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
4501                                               pb->last_block,
4502                                               extent.e_pblk,
4503                                               extent.e_lblk)) {
4504                         for (i = 0; i < extent.e_len; i++) {
4505                                 pctx->blk = extent.e_lblk + i;
4506                                 pctx->blk2 = extent.e_pblk + i;
4507                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4508                                 mark_block_used(ctx, extent.e_pblk + i);
4509                                 mark_block_used(ctx, extent.e_pblk + i);
4510                         }
4511                 }
4512
4513                 /*
4514                  * Check whether first cluster got marked in previous iteration.
4515                  */
4516                 if (ctx->fs->cluster_ratio_bits &&
4517                     pb->previous_block &&
4518                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
4519                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
4520                         /* Set blk to the beginning of next cluster. */
4521                         blk = EXT2FS_C2B(
4522                                 ctx->fs,
4523                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
4524                 else
4525                         /* Set blk to the beginning of current cluster. */
4526                         blk = EXT2FS_C2B(ctx->fs,
4527                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
4528
4529                 if (blk < extent.e_pblk + extent.e_len) {
4530                         mark_blocks_used(ctx, blk,
4531                                          extent.e_pblk + extent.e_len - blk);
4532                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
4533                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
4534                         pb->num_blocks += n;
4535                 }
4536                 pb->last_block = extent.e_lblk + extent.e_len - 1;
4537                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
4538                 start_block = pb->last_block = last_lblk;
4539                 if (is_leaf && !is_dir &&
4540                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
4541                         pb->last_init_lblock = last_lblk;
4542         next:
4543                 pctx->errcode = ext2fs_extent_get(ehandle,
4544                                                   EXT2_EXTENT_NEXT_SIB,
4545                                                   &extent);
4546         }
4547
4548         /* Failed csum but passes checks?  Ask to fix checksum. */
4549         if (failed_csum &&
4550             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
4551                 e2fsck_pass1_fix_lock(ctx);
4552                 pb->inode_modified = 1;
4553                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
4554                 e2fsck_pass1_fix_unlock(ctx);
4555                 if (pctx->errcode)
4556                         return;
4557         }
4558
4559         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
4560                 pctx->errcode = 0;
4561 }
4562
4563 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
4564                                  struct process_block_struct *pb)
4565 {
4566         struct ext2_extent_info info;
4567         struct ext2_inode       *inode = pctx->inode;
4568         ext2_extent_handle_t    ehandle;
4569         ext2_filsys             fs = ctx->fs;
4570         ext2_ino_t              ino = pctx->ino;
4571         errcode_t               retval;
4572         blk64_t                 eof_lblk;
4573         struct ext3_extent_header       *eh;
4574
4575         /* Check for a proper extent header... */
4576         eh = (struct ext3_extent_header *) &inode->i_block[0];
4577         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
4578         if (retval) {
4579                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
4580                         e2fsck_clear_inode(ctx, ino, inode, 0,
4581                                            "check_blocks_extents");
4582                 pctx->errcode = 0;
4583                 return;
4584         }
4585
4586         /* ...since this function doesn't fail if i_block is zeroed. */
4587         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
4588         if (pctx->errcode) {
4589                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
4590                         e2fsck_clear_inode(ctx, ino, inode, 0,
4591                                            "check_blocks_extents");
4592                 pctx->errcode = 0;
4593                 return;
4594         }
4595
4596         retval = ext2fs_extent_get_info(ehandle, &info);
4597         if (retval == 0) {
4598                 int max_depth = info.max_depth;
4599
4600                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
4601                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
4602                 ctx->extent_depth_count[max_depth]++;
4603         }
4604
4605         /* Check maximum extent depth */
4606         pctx->blk = info.max_depth;
4607         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
4608         if (pctx->blk2 < pctx->blk &&
4609             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
4610                 pb->eti.force_rebuild = 1;
4611
4612         /* Can we collect extent tree level stats? */
4613         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
4614         if (pctx->blk2 > pctx->blk)
4615                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
4616         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
4617         pb->eti.ino = pb->ino;
4618
4619         pb->next_lblock = 0;
4620
4621         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
4622                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
4623         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
4624         if (pctx->errcode &&
4625             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
4626                 pb->num_blocks = 0;
4627                 inode->i_blocks = 0;
4628                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4629                                    "check_blocks_extents");
4630                 pctx->errcode = 0;
4631         }
4632         ext2fs_extent_free(ehandle);
4633
4634         /* Rebuild unless it's a dir and we're rehashing it */
4635         if (LINUX_S_ISDIR(inode->i_mode) &&
4636             e2fsck_dir_will_be_rehashed(ctx, ino))
4637                 return;
4638
4639         if (ctx->options & E2F_OPT_CONVERT_BMAP)
4640                 e2fsck_rebuild_extents_later(ctx, ino);
4641         else
4642                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
4643 }
4644
4645 /*
4646  * In fact we don't need to check blocks for an inode with inline data
4647  * because this inode doesn't have any blocks.  In this function all
4648  * we need to do is add this inode into dblist when it is a directory.
4649  */
4650 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
4651                                      struct process_block_struct *pb)
4652 {
4653         int     flags;
4654         size_t  inline_data_size = 0;
4655
4656         if (!pb->is_dir) {
4657                 pctx->errcode = 0;
4658                 return;
4659         }
4660
4661         /* Process the dirents in i_block[] as the "first" block. */
4662         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
4663         if (pctx->errcode)
4664                 goto err;
4665
4666         /* Process the dirents in the EA as a "second" block. */
4667         flags = ctx->fs->flags;
4668         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4669         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
4670                                                 &inline_data_size);
4671         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4672                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4673         if (pctx->errcode) {
4674                 pctx->errcode = 0;
4675                 return;
4676         }
4677
4678         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
4679                 return;
4680
4681         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
4682         if (pctx->errcode)
4683                 goto err;
4684
4685         return;
4686 err:
4687         pctx->blk = 0;
4688         pctx->num = 0;
4689         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4690         ctx->flags |= E2F_FLAG_ABORT;
4691 }
4692
4693 /*
4694  * This subroutine is called on each inode to account for all of the
4695  * blocks used by that inode.
4696  */
4697 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
4698                          char *block_buf, const struct ea_quota *ea_ibody_quota)
4699 {
4700         ext2_filsys fs = ctx->fs;
4701         struct process_block_struct pb;
4702         ext2_ino_t      ino = pctx->ino;
4703         struct ext2_inode *inode = pctx->inode;
4704         unsigned        bad_size = 0;
4705         int             dirty_inode = 0;
4706         int             extent_fs;
4707         int             inlinedata_fs;
4708         __u64           size;
4709         struct ea_quota ea_block_quota;
4710
4711         pb.ino = ino;
4712         pb.num_blocks = EXT2FS_B2C(ctx->fs,
4713                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
4714         pb.last_block = ~0;
4715         pb.last_init_lblock = -1;
4716         pb.last_db_block = -1;
4717         pb.num_illegal_blocks = 0;
4718         pb.suppress = 0; pb.clear = 0;
4719         pb.fragmented = 0;
4720         pb.compressed = 0;
4721         pb.previous_block = 0;
4722         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
4723         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
4724         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
4725         pb.inode = inode;
4726         pb.pctx = pctx;
4727         pb.ctx = ctx;
4728         pb.inode_modified = 0;
4729         pb.eti.force_rebuild = 0;
4730         pctx->ino = ino;
4731         pctx->errcode = 0;
4732
4733         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
4734         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
4735
4736         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
4737                 if (e2fsck_should_abort(ctx))
4738                         goto out;
4739                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
4740         }
4741
4742         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
4743                 check_blocks_inline_data(ctx, pctx, &pb);
4744         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
4745                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
4746                         check_blocks_extents(ctx, pctx, &pb);
4747                 else {
4748                         int flags;
4749                         /*
4750                          * If we've modified the inode, write it out before
4751                          * iterate() tries to use it.
4752                          */
4753                         if (dirty_inode) {
4754                                 e2fsck_write_inode(ctx, ino, inode,
4755                                                    "check_blocks");
4756                                 dirty_inode = 0;
4757                         }
4758                         flags = fs->flags;
4759                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4760                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
4761                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
4762                                                 block_buf, process_block, &pb);
4763                         /*
4764                          * We do not have uninitialized extents in non extent
4765                          * files.
4766                          */
4767                         pb.last_init_lblock = pb.last_block;
4768                         /*
4769                          * If iterate() changed a block mapping, we have to
4770                          * re-read the inode.  If we decide to clear the
4771                          * inode after clearing some stuff, we'll re-write the
4772                          * bad mappings into the inode!
4773                          */
4774                         if (pb.inode_modified)
4775                                 e2fsck_read_inode(ctx, ino, inode,
4776                                                   "check_blocks");
4777                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4778                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4779
4780                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
4781 #ifdef DEBUG
4782                                 printf("bmap rebuild ino=%d\n", ino);
4783 #endif
4784                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
4785                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
4786                                         e2fsck_rebuild_extents_later(ctx, ino);
4787                         }
4788                 }
4789         }
4790         end_problem_latch(ctx, PR_LATCH_BLOCK);
4791         end_problem_latch(ctx, PR_LATCH_TOOBIG);
4792         if (e2fsck_should_abort(ctx))
4793                 goto out;
4794         if (pctx->errcode)
4795                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
4796
4797         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
4798                 if (LINUX_S_ISDIR(inode->i_mode))
4799                         ctx->fs_fragmented_dir++;
4800                 else
4801                         ctx->fs_fragmented++;
4802         }
4803
4804         if (pb.clear) {
4805                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4806                                    "check_blocks");
4807                 return;
4808         }
4809
4810         if (inode->i_flags & EXT2_INDEX_FL) {
4811                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
4812                         inode->i_flags &= ~EXT2_INDEX_FL;
4813                         dirty_inode++;
4814                 } else {
4815                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
4816                 }
4817         }
4818
4819         if (!pb.num_blocks && pb.is_dir &&
4820             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
4821                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
4822                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
4823                         ctx->fs_directory_count--;
4824                         return;
4825                 }
4826         }
4827
4828         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
4829             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
4830             !(inode->i_flags & EXT4_EA_INODE_FL)) {
4831                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
4832                                ino,
4833                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
4834                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
4835                                   ino, (ea_ibody_quota ?
4836                                         ea_ibody_quota->inodes : 0) +
4837                                                 ea_block_quota.inodes + 1);
4838         }
4839
4840         if (!ext2fs_has_feature_huge_file(fs->super) ||
4841             !(inode->i_flags & EXT4_HUGE_FILE_FL))
4842                 pb.num_blocks *= (fs->blocksize / 512);
4843         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
4844 #if 0
4845         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
4846                ino, inode->i_size, (unsigned long long) pb.last_block,
4847                (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
4848                (unsigned long long) pb.num_blocks);
4849 #endif
4850         size = EXT2_I_SIZE(inode);
4851         if (pb.is_dir) {
4852                 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
4853                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
4854                         int flags;
4855                         size_t sz = 0;
4856                         errcode_t err;
4857
4858                         flags = ctx->fs->flags;
4859                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4860                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
4861                                                       &sz);
4862                         ctx->fs->flags = (flags &
4863                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4864                                          (ctx->fs->flags &
4865                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4866                         if (err || sz != size) {
4867                                 bad_size = 7;
4868                                 pctx->num = sz;
4869                         }
4870                 } else if (size & (fs->blocksize - 1))
4871                         bad_size = 5;
4872                 else if (nblock > (pb.last_block + 1))
4873                         bad_size = 1;
4874                 else if (nblock < (pb.last_block + 1)) {
4875                         if (((pb.last_block + 1) - nblock) >
4876                             fs->super->s_prealloc_dir_blocks)
4877                                 bad_size = 2;
4878                 }
4879         } else {
4880                 if ((pb.last_init_lblock >= 0) &&
4881                     /* Do not allow initialized allocated blocks past i_size*/
4882                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
4883                     !(inode->i_flags & EXT4_VERITY_FL))
4884                         bad_size = 3;
4885                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4886                          size > ext2_max_sizes[fs->super->s_log_block_size])
4887                         /* too big for a direct/indirect-mapped file */
4888                         bad_size = 4;
4889                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4890                          size >
4891                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
4892                         /* too big for an extent-based file - 32bit ee_block */
4893                         bad_size = 6;
4894         }
4895         /* i_size for symlinks is checked elsewhere */
4896         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
4897                 /* Did inline_data set pctx->num earlier? */
4898                 if (bad_size != 7)
4899                         pctx->num = (pb.last_block + 1) * fs->blocksize;
4900                 pctx->group = bad_size;
4901                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
4902                         ext2fs_inode_size_set(fs, inode, pctx->num);
4903                         if (EXT2_I_SIZE(inode) == 0 &&
4904                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
4905                                 memset(inode->i_block, 0,
4906                                        sizeof(inode->i_block));
4907                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
4908                         }
4909                         dirty_inode++;
4910                 }
4911                 pctx->num = 0;
4912         }
4913         if (LINUX_S_ISREG(inode->i_mode) &&
4914             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4915                 ctx->large_files++;
4916         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
4917             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
4918              (ext2fs_has_feature_huge_file(fs->super) &&
4919               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
4920               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
4921                 pctx->num = pb.num_blocks;
4922                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
4923                         inode->i_blocks = pb.num_blocks;
4924                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
4925                         dirty_inode++;
4926                 }
4927                 pctx->num = 0;
4928         }
4929
4930         /*
4931          * The kernel gets mad if we ask it to allocate bigalloc clusters to
4932          * a block mapped file, so rebuild it as an extent file.  We can skip
4933          * symlinks because they're never rewritten.
4934          */
4935         if (ext2fs_has_feature_bigalloc(fs->super) &&
4936             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
4937             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
4938             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
4939             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
4940             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
4941                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
4942                 if (pctx->errcode)
4943                         goto out;
4944         }
4945
4946         if (ctx->dirs_to_hash && pb.is_dir &&
4947             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
4948             !(inode->i_flags & EXT2_INDEX_FL) &&
4949             ((inode->i_size / fs->blocksize) >= 3))
4950                 e2fsck_rehash_dir_later(ctx, ino);
4951
4952 out:
4953         if (dirty_inode)
4954                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
4955 }
4956
4957 #if 0
4958 /*
4959  * Helper function called by process block when an illegal block is
4960  * found.  It returns a description about why the block is illegal
4961  */
4962 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
4963 {
4964         blk64_t super;
4965         int     i;
4966         static char     problem[80];
4967
4968         super = fs->super->s_first_data_block;
4969         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
4970         if (block < super) {
4971                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
4972                 return(problem);
4973         } else if (block >= ext2fs_blocks_count(fs->super)) {
4974                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
4975                 return(problem);
4976         }
4977         for (i = 0; i < fs->group_desc_count; i++) {
4978                 if (block == super) {
4979                         sprintf(problem, "is the superblock in group %d", i);
4980                         break;
4981                 }
4982                 if (block > super &&
4983                     block <= (super + fs->desc_blocks)) {
4984                         sprintf(problem, "is in the group descriptors "
4985                                 "of group %d", i);
4986                         break;
4987                 }
4988                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
4989                         sprintf(problem, "is the block bitmap of group %d", i);
4990                         break;
4991                 }
4992                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
4993                         sprintf(problem, "is the inode bitmap of group %d", i);
4994                         break;
4995                 }
4996                 if (block >= ext2fs_inode_table_loc(fs, i) &&
4997                     (block < ext2fs_inode_table_loc(fs, i)
4998                      + fs->inode_blocks_per_group)) {
4999                         sprintf(problem, "is in the inode table of group %d",
5000                                 i);
5001                         break;
5002                 }
5003                 super += fs->super->s_blocks_per_group;
5004         }
5005         return(problem);
5006 }
5007 #endif
5008
5009 /*
5010  * This is a helper function for check_blocks().
5011  */
5012 static int process_block(ext2_filsys fs,
5013                   blk64_t       *block_nr,
5014                   e2_blkcnt_t blockcnt,
5015                   blk64_t ref_block EXT2FS_ATTR((unused)),
5016                   int ref_offset EXT2FS_ATTR((unused)),
5017                   void *priv_data)
5018 {
5019         struct process_block_struct *p;
5020         struct problem_context *pctx;
5021         blk64_t blk = *block_nr;
5022         int     ret_code = 0;
5023         problem_t       problem = 0;
5024         e2fsck_t        ctx;
5025
5026         p = (struct process_block_struct *) priv_data;
5027         pctx = p->pctx;
5028         ctx = p->ctx;
5029
5030         /*
5031          * For a directory, add logical block zero for processing even if it's
5032          * not mapped or we'll be perennially stuck with broken "." and ".."
5033          * entries.
5034          */
5035         if (p->is_dir && blockcnt == 0 && blk == 0) {
5036                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
5037                 if (pctx->errcode) {
5038                         pctx->blk = blk;
5039                         pctx->num = blockcnt;
5040                         goto failed_add_dir_block;
5041                 }
5042                 p->last_db_block++;
5043         }
5044
5045         if (blk == 0)
5046                 return 0;
5047
5048 #if 0
5049         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
5050                blockcnt);
5051 #endif
5052
5053         /*
5054          * Simplistic fragmentation check.  We merely require that the
5055          * file be contiguous.  (Which can never be true for really
5056          * big files that are greater than a block group.)
5057          */
5058         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
5059                 if (p->previous_block+1 != blk) {
5060                         if (ctx->options & E2F_OPT_FRAGCHECK) {
5061                                 char type = '?';
5062
5063                                 if (p->is_dir)
5064                                         type = 'd';
5065                                 else if (p->is_reg)
5066                                         type = 'f';
5067
5068                                 printf(_("%6lu(%c): expecting %6lu "
5069                                          "got phys %6lu (blkcnt %lld)\n"),
5070                                        (unsigned long) pctx->ino, type,
5071                                        (unsigned long) p->previous_block+1,
5072                                        (unsigned long) blk,
5073                                        (long long) blockcnt);
5074                         }
5075                         p->fragmented = 1;
5076                 }
5077         }
5078
5079         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
5080             !pctx->inode->i_size_high &&
5081             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
5082                 problem = PR_1_TOOBIG_DIR;
5083         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
5084                 problem = PR_1_TOOBIG_DIR;
5085         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
5086                 problem = PR_1_TOOBIG_REG;
5087         if (!p->is_dir && !p->is_reg && blockcnt > 0)
5088                 problem = PR_1_TOOBIG_SYMLINK;
5089
5090         if (blk < fs->super->s_first_data_block ||
5091             blk >= ext2fs_blocks_count(fs->super))
5092                 problem = PR_1_ILLEGAL_BLOCK_NUM;
5093
5094         /*
5095          * If this IND/DIND/TIND block is squatting atop some critical metadata
5096          * (group descriptors, superblock, bitmap, inode table), any write to
5097          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
5098          * fix that and restart fsck.
5099          */
5100         if (blockcnt < 0 &&
5101             p->ino != EXT2_RESIZE_INO &&
5102             blk < ctx->fs->super->s_blocks_count &&
5103             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
5104                 pctx->blk = blk;
5105                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
5106                 if ((ctx->options & E2F_OPT_NO) == 0)
5107                         ctx->flags |= E2F_FLAG_RESTART_LATER;
5108         }
5109
5110         if (problem) {
5111                 p->num_illegal_blocks++;
5112                 /*
5113                  * A bit of subterfuge here -- we're trying to fix a block
5114                  * mapping, but the IND/DIND/TIND block could have collided
5115                  * with some critical metadata.  So, fix the in-core mapping so
5116                  * iterate won't go insane, but return 0 instead of
5117                  * BLOCK_CHANGED so that it won't write the remapping out to
5118                  * our multiply linked block.
5119                  *
5120                  * Even if we previously determined that an *IND block
5121                  * conflicts with critical metadata, we must still try to
5122                  * iterate the *IND block as if it is an *IND block to find and
5123                  * mark the blocks it points to.  Better to be overly cautious
5124                  * with the used_blocks map so that we don't move the *IND
5125                  * block to a block that's really in use!
5126                  */
5127                 if (p->ino != EXT2_RESIZE_INO &&
5128                     ref_block != 0 &&
5129                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
5130                                               ref_block)) {
5131                         *block_nr = 0;
5132                         return 0;
5133                 }
5134                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
5135                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
5136                                 p->clear = 1;
5137                                 return BLOCK_ABORT;
5138                         }
5139                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
5140                                 p->suppress = 1;
5141                                 set_latch_flags(PR_LATCH_BLOCK,
5142                                                 PRL_SUPPRESS, 0);
5143                         }
5144                 }
5145                 pctx->blk = blk;
5146                 pctx->blkcount = blockcnt;
5147                 if (fix_problem(ctx, problem, pctx)) {
5148                         blk = *block_nr = 0;
5149                         ret_code = BLOCK_CHANGED;
5150                         p->inode_modified = 1;
5151                         /*
5152                          * If the directory block is too big and is beyond the
5153                          * end of the FS, don't bother trying to add it for
5154                          * processing -- the kernel would never have created a
5155                          * directory this large, and we risk an ENOMEM abort.
5156                          * In any case, the toobig handler for extent-based
5157                          * directories also doesn't feed toobig blocks to
5158                          * pass 2.
5159                          */
5160                         if (problem == PR_1_TOOBIG_DIR)
5161                                 return ret_code;
5162                         goto mark_dir;
5163                 } else
5164                         return 0;
5165         }
5166
5167         if (p->ino == EXT2_RESIZE_INO) {
5168                 /*
5169                  * The resize inode has already be sanity checked
5170                  * during pass #0 (the superblock checks).  All we
5171                  * have to do is mark the double indirect block as
5172                  * being in use; all of the other blocks are handled
5173                  * by mark_table_blocks()).
5174                  */
5175                 if (blockcnt == BLOCK_COUNT_DIND)
5176                         mark_block_used(ctx, blk);
5177                 p->num_blocks++;
5178         } else if (!(ctx->fs->cluster_ratio_bits &&
5179                      p->previous_block &&
5180                      (EXT2FS_B2C(ctx->fs, blk) ==
5181                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
5182                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
5183                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
5184                 mark_block_used(ctx, blk);
5185                 p->num_blocks++;
5186         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
5187                                              p->last_block, blk, blockcnt)) {
5188                 pctx->blk = blockcnt;
5189                 pctx->blk2 = blk;
5190                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
5191                 mark_block_used(ctx, blk);
5192                 mark_block_used(ctx, blk);
5193         }
5194         if (blockcnt >= 0)
5195                 p->last_block = blockcnt;
5196         p->previous_block = blk;
5197 mark_dir:
5198         if (p->is_dir && (blockcnt >= 0)) {
5199                 while (++p->last_db_block < blockcnt) {
5200                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
5201                                                               p->ino, 0,
5202                                                               p->last_db_block);
5203                         if (pctx->errcode) {
5204                                 pctx->blk = 0;
5205                                 pctx->num = p->last_db_block;
5206                                 goto failed_add_dir_block;
5207                         }
5208                 }
5209                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
5210                                                       blk, blockcnt);
5211                 if (pctx->errcode) {
5212                         pctx->blk = blk;
5213                         pctx->num = blockcnt;
5214                 failed_add_dir_block:
5215                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
5216                         /* Should never get here */
5217                         ctx->flags |= E2F_FLAG_ABORT;
5218                         return BLOCK_ABORT;
5219                 }
5220         }
5221         return ret_code;
5222 }
5223
5224 static int process_bad_block(ext2_filsys fs,
5225                       blk64_t *block_nr,
5226                       e2_blkcnt_t blockcnt,
5227                       blk64_t ref_block EXT2FS_ATTR((unused)),
5228                       int ref_offset EXT2FS_ATTR((unused)),
5229                       void *priv_data)
5230 {
5231         struct process_block_struct *p;
5232         blk64_t         blk = *block_nr;
5233         blk64_t         first_block;
5234         dgrp_t          i;
5235         struct problem_context *pctx;
5236         e2fsck_t        ctx;
5237
5238         if (!blk)
5239                 return 0;
5240
5241         p = (struct process_block_struct *) priv_data;
5242         ctx = p->ctx;
5243         pctx = p->pctx;
5244
5245         pctx->ino = EXT2_BAD_INO;
5246         pctx->blk = blk;
5247         pctx->blkcount = blockcnt;
5248
5249         if ((blk < fs->super->s_first_data_block) ||
5250             (blk >= ext2fs_blocks_count(fs->super))) {
5251                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
5252                         *block_nr = 0;
5253                         return BLOCK_CHANGED;
5254                 } else
5255                         return 0;
5256         }
5257
5258         if (blockcnt < 0) {
5259                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
5260                         p->bbcheck = 1;
5261                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
5262                                 *block_nr = 0;
5263                                 return BLOCK_CHANGED;
5264                         }
5265                 } else if (is_blocks_used(ctx, blk, 1)) {
5266                         p->bbcheck = 1;
5267                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
5268                                         pctx)) {
5269                                 *block_nr = 0;
5270                                 return BLOCK_CHANGED;
5271                         }
5272                         if (e2fsck_should_abort(ctx))
5273                                 return BLOCK_ABORT;
5274                 } else {
5275                         mark_block_used(ctx, blk);
5276                 }
5277                 return 0;
5278         }
5279 #if 0
5280         printf ("DEBUG: Marking %u as bad.\n", blk);
5281 #endif
5282         ctx->fs_badblocks_count++;
5283         /*
5284          * If the block is not used, then mark it as used and return.
5285          * If it is already marked as found, this must mean that
5286          * there's an overlap between the filesystem table blocks
5287          * (bitmaps and inode table) and the bad block list.
5288          */
5289         if (!is_blocks_used(ctx, blk, 1)) {
5290                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
5291                 return 0;
5292         }
5293         /*
5294          * Try to find the where the filesystem block was used...
5295          */
5296         first_block = fs->super->s_first_data_block;
5297
5298         for (i = 0; i < fs->group_desc_count; i++ ) {
5299                 pctx->group = i;
5300                 pctx->blk = blk;
5301                 if (!ext2fs_bg_has_super(fs, i))
5302                         goto skip_super;
5303                 if (blk == first_block) {
5304                         if (i == 0) {
5305                                 if (fix_problem(ctx,
5306                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
5307                                                 pctx)) {
5308                                         *block_nr = 0;
5309                                         return BLOCK_CHANGED;
5310                                 }
5311                                 return 0;
5312                         }
5313                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
5314                         return 0;
5315                 }
5316                 if ((blk > first_block) &&
5317                     (blk <= first_block + fs->desc_blocks)) {
5318                         if (i == 0) {
5319                                 pctx->blk = *block_nr;
5320                                 if (fix_problem(ctx,
5321                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
5322                                         *block_nr = 0;
5323                                         return BLOCK_CHANGED;
5324                                 }
5325                                 return 0;
5326                         }
5327                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
5328                         return 0;
5329                 }
5330         skip_super:
5331                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
5332                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
5333                                 ctx->invalid_block_bitmap_flag[i]++;
5334                                 ctx->invalid_bitmaps++;
5335                         }
5336                         return 0;
5337                 }
5338                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
5339                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
5340                                 ctx->invalid_inode_bitmap_flag[i]++;
5341                                 ctx->invalid_bitmaps++;
5342                         }
5343                         return 0;
5344                 }
5345                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
5346                     (blk < (ext2fs_inode_table_loc(fs, i) +
5347                             fs->inode_blocks_per_group))) {
5348                         /*
5349                          * If there are bad blocks in the inode table,
5350                          * the inode scan code will try to do
5351                          * something reasonable automatically.
5352                          */
5353                         return 0;
5354                 }
5355                 first_block += fs->super->s_blocks_per_group;
5356         }
5357         /*
5358          * If we've gotten to this point, then the only
5359          * possibility is that the bad block inode meta data
5360          * is using a bad block.
5361          */
5362         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
5363             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
5364             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
5365                 p->bbcheck = 1;
5366                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
5367                         *block_nr = 0;
5368                         return BLOCK_CHANGED;
5369                 }
5370                 if (e2fsck_should_abort(ctx))
5371                         return BLOCK_ABORT;
5372                 return 0;
5373         }
5374
5375         pctx->group = -1;
5376
5377         /* Warn user that the block wasn't claimed */
5378         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
5379
5380         return 0;
5381 }
5382
5383 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
5384                             const char *name, int num, blk64_t *new_block)
5385 {
5386         ext2_filsys fs = ctx->fs;
5387         dgrp_t          last_grp;
5388         blk64_t         old_block = *new_block;
5389         blk64_t         last_block;
5390         dgrp_t          flexbg;
5391         unsigned        flexbg_size;
5392         int             i, is_flexbg;
5393         char            *buf;
5394         struct problem_context  pctx;
5395
5396         clear_problem_context(&pctx);
5397
5398         pctx.group = group;
5399         pctx.blk = old_block;
5400         pctx.str = name;
5401
5402         /*
5403          * For flex_bg filesystems, first try to allocate the metadata
5404          * within the flex_bg, and if that fails then try finding the
5405          * space anywhere in the filesystem.
5406          */
5407         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
5408         if (is_flexbg) {
5409                 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
5410                 flexbg = group / flexbg_size;
5411                 first_block = ext2fs_group_first_block2(fs,
5412                                                         flexbg_size * flexbg);
5413                 last_grp = group | (flexbg_size - 1);
5414                 if (last_grp >= fs->group_desc_count)
5415                         last_grp = fs->group_desc_count - 1;
5416                 last_block = ext2fs_group_last_block2(fs, last_grp);
5417         } else
5418                 last_block = ext2fs_group_last_block2(fs, group);
5419         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
5420                                                num, ctx->block_found_map,
5421                                                new_block);
5422         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
5423                 pctx.errcode = ext2fs_get_free_blocks2(fs,
5424                                 fs->super->s_first_data_block,
5425                                 ext2fs_blocks_count(fs->super),
5426                                 num, ctx->block_found_map, new_block);
5427         if (pctx.errcode) {
5428                 pctx.num = num;
5429                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
5430                 ext2fs_unmark_valid(fs);
5431                 ctx->flags |= E2F_FLAG_ABORT;
5432                 return;
5433         }
5434         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
5435         if (pctx.errcode) {
5436                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
5437                 ext2fs_unmark_valid(fs);
5438                 ctx->flags |= E2F_FLAG_ABORT;
5439                 return;
5440         }
5441         ext2fs_mark_super_dirty(fs);
5442         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
5443         pctx.blk2 = *new_block;
5444         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
5445                           PR_1_RELOC_TO), &pctx);
5446         pctx.blk2 = 0;
5447         for (i = 0; i < num; i++) {
5448                 pctx.blk = i;
5449                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
5450                 if (old_block) {
5451                         pctx.errcode = io_channel_read_blk64(fs->io,
5452                                    old_block + i, 1, buf);
5453                         if (pctx.errcode)
5454                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
5455                         pctx.blk = (*new_block) + i;
5456                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
5457                                                               1, buf);
5458                 } else {
5459                         pctx.blk = (*new_block) + i;
5460                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
5461                                                            NULL, NULL);
5462                 }
5463
5464                 if (pctx.errcode)
5465                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
5466         }
5467         ext2fs_free_mem(&buf);
5468 }
5469
5470 /*
5471  * This routine gets called at the end of pass 1 if bad blocks are
5472  * detected in the superblock, group descriptors, inode_bitmaps, or
5473  * block bitmaps.  At this point, all of the blocks have been mapped
5474  * out, so we can try to allocate new block(s) to replace the bad
5475  * blocks.
5476  */
5477 static void handle_fs_bad_blocks(e2fsck_t ctx)
5478 {
5479         ext2_filsys fs = ctx->fs;
5480         dgrp_t          i;
5481         blk64_t         first_block;
5482         blk64_t         new_blk;
5483
5484         for (i = 0; i < fs->group_desc_count; i++) {
5485                 first_block = ext2fs_group_first_block2(fs, i);
5486
5487                 if (ctx->invalid_block_bitmap_flag[i]) {
5488                         new_blk = ext2fs_block_bitmap_loc(fs, i);
5489                         new_table_block(ctx, first_block, i, _("block bitmap"),
5490                                         1, &new_blk);
5491                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
5492                 }
5493                 if (ctx->invalid_inode_bitmap_flag[i]) {
5494                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
5495                         new_table_block(ctx, first_block, i, _("inode bitmap"),
5496                                         1, &new_blk);
5497                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
5498                 }
5499                 if (ctx->invalid_inode_table_flag[i]) {
5500                         new_blk = ext2fs_inode_table_loc(fs, i);
5501                         new_table_block(ctx, first_block, i, _("inode table"),
5502                                         fs->inode_blocks_per_group,
5503                                         &new_blk);
5504                         ext2fs_inode_table_loc_set(fs, i, new_blk);
5505                         ctx->flags |= E2F_FLAG_RESTART;
5506                 }
5507         }
5508         ctx->invalid_bitmaps = 0;
5509 }
5510
5511 /*
5512  * This routine marks all blocks which are used by the superblock,
5513  * group descriptors, inode bitmaps, and block bitmaps.
5514  */
5515 static void mark_table_blocks(e2fsck_t ctx)
5516 {
5517         ext2_filsys fs = ctx->fs;
5518         blk64_t b;
5519         dgrp_t  i;
5520         unsigned int    j;
5521         struct problem_context pctx;
5522
5523         clear_problem_context(&pctx);
5524
5525         for (i = 0; i < fs->group_desc_count; i++) {
5526                 pctx.group = i;
5527
5528                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
5529                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
5530
5531                 /*
5532                  * Mark the blocks used for the inode table
5533                  */
5534                 if (ext2fs_inode_table_loc(fs, i)) {
5535                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
5536                              j < fs->inode_blocks_per_group;
5537                              j++, b++) {
5538                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5539                                                              b)) {
5540                                         pctx.blk = b;
5541                                         if (!ctx->invalid_inode_table_flag[i] &&
5542                                             fix_problem(ctx,
5543                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
5544                                                 ctx->invalid_inode_table_flag[i]++;
5545                                                 ctx->invalid_bitmaps++;
5546                                         }
5547                                 } else {
5548                                     ext2fs_mark_block_bitmap2(
5549                                                 ctx->block_found_map, b);
5550                                     ext2fs_mark_block_bitmap2(
5551                                                 ctx->block_metadata_map, b);
5552                                 }
5553                         }
5554                 }
5555
5556                 /*
5557                  * Mark block used for the block bitmap
5558                  */
5559                 if (ext2fs_block_bitmap_loc(fs, i)) {
5560                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5561                                      ext2fs_block_bitmap_loc(fs, i))) {
5562                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
5563                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
5564                                         ctx->invalid_block_bitmap_flag[i]++;
5565                                         ctx->invalid_bitmaps++;
5566                                 }
5567                         } else {
5568                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
5569                                      ext2fs_block_bitmap_loc(fs, i));
5570                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5571                                      ext2fs_block_bitmap_loc(fs, i));
5572                         }
5573                 }
5574                 /*
5575                  * Mark block used for the inode bitmap
5576                  */
5577                 if (ext2fs_inode_bitmap_loc(fs, i)) {
5578                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5579                                      ext2fs_inode_bitmap_loc(fs, i))) {
5580                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
5581                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
5582                                         ctx->invalid_inode_bitmap_flag[i]++;
5583                                         ctx->invalid_bitmaps++;
5584                                 }
5585                         } else {
5586                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5587                                      ext2fs_inode_bitmap_loc(fs, i));
5588                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
5589                                      ext2fs_inode_bitmap_loc(fs, i));
5590                         }
5591                 }
5592         }
5593 }
5594
5595 /*
5596  * These subroutines short circuits ext2fs_get_blocks and
5597  * ext2fs_check_directory; we use them since we already have the inode
5598  * structure, so there's no point in letting the ext2fs library read
5599  * the inode again.
5600  */
5601 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
5602                                   blk_t *blocks)
5603 {
5604         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5605         int     i;
5606
5607         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5608                 return EXT2_ET_CALLBACK_NOTHANDLED;
5609
5610         for (i=0; i < EXT2_N_BLOCKS; i++)
5611                 blocks[i] = ctx->stashed_inode->i_block[i];
5612         return 0;
5613 }
5614
5615 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
5616                                   struct ext2_inode *inode)
5617 {
5618         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5619
5620         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5621                 return EXT2_ET_CALLBACK_NOTHANDLED;
5622         *inode = *ctx->stashed_inode;
5623         return 0;
5624 }
5625
5626 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
5627                             struct ext2_inode *inode)
5628 {
5629         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5630
5631         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
5632                 (inode != ctx->stashed_inode))
5633                 *ctx->stashed_inode = *inode;
5634         return EXT2_ET_CALLBACK_NOTHANDLED;
5635 }
5636
5637 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
5638 {
5639         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5640
5641         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5642                 return EXT2_ET_CALLBACK_NOTHANDLED;
5643
5644         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
5645                 return EXT2_ET_NO_DIRECTORY;
5646         return 0;
5647 }
5648
5649 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
5650                                         blk64_t *ret)
5651 {
5652         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5653         errcode_t       retval;
5654         blk64_t         new_block;
5655
5656         if (ctx->block_found_map) {
5657                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
5658                                            &new_block);
5659                 if (retval)
5660                         return retval;
5661                 if (fs->block_map) {
5662                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
5663                         ext2fs_mark_bb_dirty(fs);
5664                 }
5665         } else {
5666                 if (!fs->block_map) {
5667                         retval = ext2fs_read_block_bitmap(fs);
5668                         if (retval)
5669                                 return retval;
5670                 }
5671
5672                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
5673                 if (retval)
5674                         return retval;
5675         }
5676
5677         *ret = new_block;
5678         return (0);
5679 }
5680
5681 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
5682                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
5683 {
5684         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5685         errcode_t       retval;
5686
5687         if (ctx->block_found_map)
5688                 return ext2fs_new_range(fs, flags, goal, len,
5689                                         ctx->block_found_map, pblk, plen);
5690
5691         if (!fs->block_map) {
5692                 retval = ext2fs_read_block_bitmap(fs);
5693                 if (retval)
5694                         return retval;
5695         }
5696
5697         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
5698                                 pblk, plen);
5699 }
5700
5701 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
5702 {
5703         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5704
5705         /* Never free a critical metadata block */
5706         if (ctx->block_found_map &&
5707             ctx->block_metadata_map &&
5708             inuse < 0 &&
5709             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
5710                 return;
5711
5712         if (ctx->block_found_map) {
5713                 if (inuse > 0)
5714                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
5715                 else
5716                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
5717         }
5718 }
5719
5720 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
5721                                            blk_t num, int inuse)
5722 {
5723         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5724
5725         /* Never free a critical metadata block */
5726         if (ctx->block_found_map &&
5727             ctx->block_metadata_map &&
5728             inuse < 0 &&
5729             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
5730                 return;
5731
5732         if (ctx->block_found_map) {
5733                 if (inuse > 0)
5734                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
5735                                                         blk, num);
5736                 else
5737                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
5738                                                         blk, num);
5739         }
5740 }
5741
5742 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
5743 {
5744         ext2_filsys fs = ctx->fs;
5745
5746         if (use_shortcuts) {
5747                 fs->get_blocks = pass1_get_blocks;
5748                 fs->check_directory = pass1_check_directory;
5749                 fs->read_inode = pass1_read_inode;
5750                 fs->write_inode = pass1_write_inode;
5751                 ctx->stashed_ino = 0;
5752         } else {
5753                 fs->get_blocks = 0;
5754                 fs->check_directory = 0;
5755                 fs->read_inode = 0;
5756                 fs->write_inode = 0;
5757         }
5758 }
5759
5760 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
5761 {
5762         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
5763         ext2fs_set_block_alloc_stats_callback(ctx->fs,
5764                                                 e2fsck_block_alloc_stats, 0);
5765         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
5766         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
5767                                         e2fsck_block_alloc_stats_range, NULL);
5768 }