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