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 void process_inodes(e2fsck_t ctx, char *block_buf);
  92 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
  93 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
  94                                   dgrp_t group, void * priv_data);
  95 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
  96                                     char *block_buf, int adjust_sign);
  97 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
  98
  99 struct process_block_struct {
 100         ext2_ino_t      ino;
 101         unsigned        is_dir:1, is_reg:1, clear:1, suppress:1,
 102                                 fragmented:1, compressed:1, bbcheck:1,
 103                                 inode_modified:1;
 104         blk64_t         num_blocks;
 105         blk64_t         max_blocks;
 106         blk64_t         last_block;
 107         e2_blkcnt_t     last_init_lblock;
 108         e2_blkcnt_t     last_db_block;
 109         int             num_illegal_blocks;
 110         blk64_t         previous_block;
 111         struct ext2_inode *inode;
 112         struct problem_context *pctx;
 113         ext2fs_block_bitmap fs_meta_blocks;
 114         e2fsck_t        ctx;
 115         blk64_t         next_lblock;
 116         struct extent_tree_info eti;
 117 };
 118
 119 struct process_inode_block {
 120         ext2_ino_t ino;
 121         struct ea_quota ea_ibody_quota;
 122         struct ext2_inode_large inode;
 123 };
 124
 125 struct scan_callback_struct {
 126         e2fsck_t        ctx;
 127         char            *block_buf;
 128 };
 129
 130 /*
 131  * For the inodes to process list.
 132  */
 133 static struct process_inode_block *inodes_to_process;
 134 static int process_inode_count;
 135
 136 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
 137                             EXT2_MIN_BLOCK_LOG_SIZE + 1];
 138
 139 /*
 140  * Check to make sure a device inode is real.  Returns 1 if the device
 141  * checks out, 0 if not.
 142  *
 143  * Note: this routine is now also used to check FIFO's and Sockets,
 144  * since they have the same requirement; the i_block fields should be
 145  * zero.
 146  */
 147 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
 148                                     struct ext2_inode *inode)
 149 {
 150         int     i;
 151
 152         /*
 153          * If the index or extents flag is set, then this is a bogus
 154          * device/fifo/socket
 155          */
 156         if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
 157                 return 0;
 158
 159         /*
 160          * We should be able to do the test below all the time, but
 161          * because the kernel doesn't forcibly clear the device
 162          * inode's additional i_block fields, there are some rare
 163          * occasions when a legitimate device inode will have non-zero
 164          * additional i_block fields.  So for now, we only complain
 165          * when the immutable flag is set, which should never happen
 166          * for devices.  (And that's when the problem is caused, since
 167          * you can't set or clear immutable flags for devices.)  Once
 168          * the kernel has been fixed we can change this...
 169          */
 170         if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
 171                 for (i=4; i < EXT2_N_BLOCKS; i++)
 172                         if (inode->i_block[i])
 173                                 return 0;
 174         }
 175         return 1;
 176 }
 177
 178 /*
 179  * Check to make sure a symlink inode is real.  Returns 1 if the symlink
 180  * checks out, 0 if not.
 181  */
 182 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
 183                                struct ext2_inode *inode, char *buf)
 184 {
 185         unsigned int buflen;
 186         unsigned int len;
 187
 188         if ((inode->i_size_high || inode->i_size == 0) ||
 189             (inode->i_flags & EXT2_INDEX_FL))
 190                 return 0;
 191
 192         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
 193                 size_t inline_size;
 194
 195                 if (inode->i_flags & EXT4_EXTENTS_FL)
 196                         return 0;
 197                 if (ext2fs_inline_data_size(fs, ino, &inline_size))
 198                         return 0;
 199                 if (inode->i_size != inline_size)
 200                         return 0;
 201
 202                 return 1;
 203         }
 204
 205         if (ext2fs_is_fast_symlink(inode)) {
 206                 if (inode->i_flags & EXT4_EXTENTS_FL)
 207                         return 0;
 208                 buf = (char *)inode->i_block;
 209                 buflen = sizeof(inode->i_block);
 210         } else {
 211                 ext2_extent_handle_t    handle;
 212                 struct ext2_extent_info info;
 213                 struct ext2fs_extent    extent;
 214                 blk64_t blk;
 215                 int i;
 216
 217                 if (inode->i_flags & EXT4_EXTENTS_FL) {
 218                         if (ext2fs_extent_open2(fs, ino, inode, &handle))
 219                                 return 0;
 220                         if (ext2fs_extent_get_info(handle, &info) ||
 221                             (info.num_entries != 1) ||
 222                             (info.max_depth != 0)) {
 223                                 ext2fs_extent_free(handle);
 224                                 return 0;
 225                         }
 226                         if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
 227                                               &extent) ||
 228                             (extent.e_lblk != 0) ||
 229                             (extent.e_len != 1)) {
 230                                 ext2fs_extent_free(handle);
 231                                 return 0;
 232                         }
 233                         blk = extent.e_pblk;
 234                         ext2fs_extent_free(handle);
 235                 } else {
 236                         blk = inode->i_block[0];
 237
 238                         for (i = 1; i < EXT2_N_BLOCKS; i++)
 239                                 if (inode->i_block[i])
 240                                         return 0;
 241                 }
 242
 243                 if (blk < fs->super->s_first_data_block ||
 244                     blk >= ext2fs_blocks_count(fs->super))
 245                         return 0;
 246
 247                 if (io_channel_read_blk64(fs->io, blk, 1, buf))
 248                         return 0;
 249
 250                 buflen = fs->blocksize;
 251         }
 252
 253         if (inode->i_flags & EXT4_ENCRYPT_FL)
 254                 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
 255         else
 256                 len = strnlen(buf, buflen);
 257
 258         if (len >= buflen)
 259                 return 0;
 260
 261         if (len != inode->i_size)
 262                 return 0;
 263         return 1;
 264 }
 265
 266 /*
 267  * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
 268  */
 269 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
 270 static void check_extents_inlinedata(e2fsck_t ctx,
 271                                      struct problem_context *pctx)
 272 {
 273         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 274                 return;
 275
 276         if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
 277                 return;
 278
 279         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 280         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 281 }
 282 #undef BAD_SPECIAL_FLAGS
 283
 284 /*
 285  * If the immutable (or append-only) flag is set on the inode, offer
 286  * to clear it.
 287  */
 288 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
 289 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
 290 {
 291         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 292                 return;
 293
 294         if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
 295                 return;
 296
 297         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 298         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 299 }
 300
 301 /*
 302  * If device, fifo or socket, check size is zero -- if not offer to
 303  * clear it
 304  */
 305 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
 306 {
 307         struct ext2_inode *inode = pctx->inode;
 308
 309         if (EXT2_I_SIZE(inode) == 0)
 310                 return;
 311
 312         if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
 313                 return;
 314
 315         ext2fs_inode_size_set(ctx->fs, inode, 0);
 316         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 317 }
 318
 319 /*
 320  * For a given size, calculate how many blocks would be charged towards quota.
 321  */
 322 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
 323 {
 324         blk64_t clusters;
 325
 326         clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
 327         return EXT2FS_C2B(fs, clusters);
 328 }
 329
 330 /*
 331  * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
 332  * the problem code.
 333  */
 334 static problem_t check_large_ea_inode(e2fsck_t ctx,
 335                                       struct ext2_ext_attr_entry *entry,
 336                                       struct problem_context *pctx,
 337                                       blk64_t *quota_blocks)
 338 {
 339         struct ext2_inode inode;
 340         __u32 hash;
 341         errcode_t retval;
 342
 343         /* Check if inode is within valid range */
 344         if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
 345             (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
 346                 pctx->num = entry->e_value_inum;
 347                 return PR_1_ATTR_VALUE_EA_INODE;
 348         }
 349
 350         e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
 351
 352         retval = ext2fs_ext_attr_hash_entry2(ctx->fs, entry, NULL, &hash);
 353         if (retval) {
 354                 com_err("check_large_ea_inode", retval,
 355                         _("while hashing entry with e_value_inum = %u"),
 356                         entry->e_value_inum);
 357                 fatal_error(ctx, 0);
 358         }
 359
 360         if (hash == entry->e_hash) {
 361                 *quota_blocks = size_to_quota_blocks(ctx->fs,
 362                                                      entry->e_value_size);
 363         } else {
 364                 /* This might be an old Lustre-style ea_inode reference. */
 365                 if (inode.i_mtime == pctx->ino &&
 366                     inode.i_generation == pctx->inode->i_generation) {
 367                         *quota_blocks = 0;
 368                 } else {
 369                         /* If target inode is also missing EA_INODE flag,
 370                          * this is likely to be a bad reference.
 371                          */
 372                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 373                                 pctx->num = entry->e_value_inum;
 374                                 return PR_1_ATTR_VALUE_EA_INODE;
 375                         } else {
 376                                 pctx->num = entry->e_hash;
 377                                 return PR_1_ATTR_HASH;
 378                         }
 379                 }
 380         }
 381
 382         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 383                 pctx->num = entry->e_value_inum;
 384                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
 385                         inode.i_flags |= EXT4_EA_INODE_FL;
 386                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
 387                                            &inode);
 388                 } else {
 389                         return PR_1_ATTR_NO_EA_INODE_FL;
 390                 }
 391         }
 392         return 0;
 393 }
 394
 395 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
 396                               struct ext2_ext_attr_entry *first, void *end)
 397 {
 398         struct ext2_ext_attr_entry *entry;
 399
 400         for (entry = first;
 401              (void *)entry < end && !EXT2_EXT_IS_LAST_ENTRY(entry);
 402              entry = EXT2_EXT_ATTR_NEXT(entry)) {
 403                 if (!entry->e_value_inum)
 404                         continue;
 405                 if (!ctx->ea_inode_refs) {
 406                         pctx->errcode = ea_refcount_create(0,
 407                                                            &ctx->ea_inode_refs);
 408                         if (pctx->errcode) {
 409                                 pctx->num = 4;
 410                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
 411                                 ctx->flags |= E2F_FLAG_ABORT;
 412                                 return;
 413                         }
 414                 }
 415                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
 416                                       0);
 417         }
 418 }
 419
 420 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
 421                               struct ea_quota *ea_ibody_quota)
 422 {
 423         struct ext2_super_block *sb = ctx->fs->super;
 424         struct ext2_inode_large *inode;
 425         struct ext2_ext_attr_entry *entry;
 426         char *start, *header, *end;
 427         unsigned int storage_size, remain;
 428         problem_t problem = 0;
 429         region_t region = 0;
 430
 431         ea_ibody_quota->blocks = 0;
 432         ea_ibody_quota->inodes = 0;
 433
 434         inode = (struct ext2_inode_large *) pctx->inode;
 435         storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
 436                 inode->i_extra_isize;
 437         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 438                  inode->i_extra_isize;
 439         end = header + storage_size;
 440         start = header + sizeof(__u32);
 441         entry = (struct ext2_ext_attr_entry *) start;
 442
 443         /* scan all entry's headers first */
 444
 445         /* take finish entry 0UL into account */
 446         remain = storage_size - sizeof(__u32);
 447
 448         region = region_create(0, storage_size);
 449         if (!region) {
 450                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
 451                 problem = 0;
 452                 ctx->flags |= E2F_FLAG_ABORT;
 453                 return;
 454         }
 455         if (region_allocate(region, 0, sizeof(__u32))) {
 456                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 457                 goto fix;
 458         }
 459
 460         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
 461                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
 462                 __u32 hash;
 463
 464                 if (region_allocate(region, (char *)entry - (char *)header,
 465                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
 466                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
 467                         goto fix;
 468                 }
 469
 470                 /* header eats this space */
 471                 remain -= sizeof(struct ext2_ext_attr_entry);
 472
 473                 /* is attribute name valid? */
 474                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
 475                         pctx->num = entry->e_name_len;
 476                         problem = PR_1_ATTR_NAME_LEN;
 477                         goto fix;
 478                 }
 479
 480                 /* attribute len eats this space */
 481                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
 482
 483                 if (entry->e_value_inum == 0) {
 484                         /* check value size */
 485                         if (entry->e_value_size > remain) {
 486                                 pctx->num = entry->e_value_size;
 487                                 problem = PR_1_ATTR_VALUE_SIZE;
 488                                 goto fix;
 489                         }
 490
 491                         if (entry->e_value_size &&
 492                             region_allocate(region,
 493                                             sizeof(__u32) + entry->e_value_offs,
 494                                             EXT2_EXT_ATTR_SIZE(
 495                                                 entry->e_value_size))) {
 496                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 497                                 goto fix;
 498                         }
 499
 500                         hash = ext2fs_ext_attr_hash_entry(entry,
 501                                                           start + entry->e_value_offs);
 502
 503                         /* e_hash may be 0 in older inode's ea */
 504                         if (entry->e_hash != 0 && entry->e_hash != hash) {
 505                                 pctx->num = entry->e_hash;
 506                                 problem = PR_1_ATTR_HASH;
 507                                 goto fix;
 508                         }
 509                 } else {
 510                         blk64_t quota_blocks;
 511
 512                         problem = check_large_ea_inode(ctx, entry, pctx,
 513                                                        &quota_blocks);
 514                         if (problem != 0)
 515                                 goto fix;
 516
 517                         ea_ibody_quota->blocks += quota_blocks;
 518                         ea_ibody_quota->inodes++;
 519                 }
 520
 521                 /* If EA value is stored in external inode then it does not
 522                  * consume space here */
 523                 if (entry->e_value_inum == 0)
 524                         remain -= entry->e_value_size;
 525
 526                 entry = EXT2_EXT_ATTR_NEXT(entry);
 527         }
 528
 529         if (region_allocate(region, (char *)entry - (char *)header,
 530                             sizeof(__u32))) {
 531                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 532                 goto fix;
 533         }
 534 fix:
 535         if (region)
 536                 region_free(region);
 537         /*
 538          * it seems like a corruption. it's very unlikely we could repair
 539          * EA(s) in automatic fashion -bzzz
 540          */
 541         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
 542                 inc_ea_inode_refs(ctx, pctx,
 543                                   (struct ext2_ext_attr_entry *)start, end);
 544                 return;
 545         }
 546
 547         /* simply remove all possible EA(s) */
 548         *((__u32 *)header) = 0UL;
 549         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 550                                 EXT2_INODE_SIZE(sb), "pass1");
 551         ea_ibody_quota->blocks = 0;
 552         ea_ibody_quota->inodes = 0;
 553 }
 554
 555 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
 556         return (xtime & (1U << 31)) != 0 &&
 557                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
 558 }
 559
 560 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
 561         check_inode_extra_negative_epoch(inode->i_##xtime, \
 562                                          inode->i_##xtime##_extra)
 563
 564 /* When today's date is earlier than 2242, we assume that atimes,
 565  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
 566  * actually pre-1970 dates mis-encoded.
 567  */
 568 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
 569
 570 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
 571                                     struct ea_quota *ea_ibody_quota)
 572 {
 573         struct ext2_super_block *sb = ctx->fs->super;
 574         struct ext2_inode_large *inode;
 575         __u32 *eamagic;
 576         int min, max;
 577
 578         ea_ibody_quota->blocks = 0;
 579         ea_ibody_quota->inodes = 0;
 580
 581         inode = (struct ext2_inode_large *) pctx->inode;
 582         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
 583                 /* this isn't large inode. so, nothing to check */
 584                 return;
 585         }
 586
 587 #if 0
 588         printf("inode #%u, i_extra_size %d\n", pctx->ino,
 589                         inode->i_extra_isize);
 590 #endif
 591         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
 592         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
 593         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
 594         /*
 595          * For now we will allow i_extra_isize to be 0, but really
 596          * implementations should never allow i_extra_isize to be 0
 597          */
 598         if (inode->i_extra_isize &&
 599             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
 600              inode->i_extra_isize & 3)) {
 601                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
 602                         return;
 603                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
 604                         inode->i_extra_isize = sb->s_want_extra_isize;
 605                 else
 606                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
 607                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 608                                         EXT2_INODE_SIZE(sb), "pass1");
 609         }
 610
 611         /* check if there is no place for an EA header */
 612         if (inode->i_extra_isize >= max - sizeof(__u32))
 613                 return;
 614
 615         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 616                         inode->i_extra_isize);
 617         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
 618                 /* it seems inode has an extended attribute(s) in body */
 619                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
 620         }
 621
 622         /*
 623          * If the inode's extended atime (ctime, crtime, mtime) is stored in
 624          * the old, invalid format, repair it.
 625          */
 626         if (((sizeof(time_t) <= 4) ||
 627              (((sizeof(time_t) > 4) &&
 628                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
 629             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
 630              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
 631              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
 632              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
 633
 634                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
 635                         return;
 636
 637                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
 638                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
 639                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
 640                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
 641                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
 642                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
 643                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
 644                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
 645                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 646                                         EXT2_INODE_SIZE(sb), "pass1");
 647         }
 648
 649 }
 650
 651 /*
 652  * Check to see if the inode might really be a directory, despite i_mode
 653  *
 654  * This is a lot of complexity for something for which I'm not really
 655  * convinced happens frequently in the wild.  If for any reason this
 656  * causes any problems, take this code out.
 657  * [tytso:20070331.0827EDT]
 658  */
 659 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
 660                                 char *buf)
 661 {
 662         struct ext2_inode *inode = pctx->inode;
 663         struct ext2_dir_entry   *dirent;
 664         errcode_t               retval;
 665         blk64_t                 blk;
 666         unsigned int            i, rec_len, not_device = 0;
 667         int                     extent_fs;
 668         int                     inlinedata_fs;
 669
 670         /*
 671          * If the mode looks OK, we believe it.  If the first block in
 672          * the i_block array is 0, this cannot be a directory. If the
 673          * inode is extent-mapped, it is still the case that the latter
 674          * cannot be 0 - the magic number in the extent header would make
 675          * it nonzero.
 676          */
 677         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
 678             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
 679                 return;
 680
 681         /*
 682          * Check the block numbers in the i_block array for validity:
 683          * zero blocks are skipped (but the first one cannot be zero -
 684          * see above), other blocks are checked against the first and
 685          * max data blocks (from the the superblock) and against the
 686          * block bitmap. Any invalid block found means this cannot be
 687          * a directory.
 688          *
 689          * If there are non-zero blocks past the fourth entry, then
 690          * this cannot be a device file: we remember that for the next
 691          * check.
 692          *
 693          * For extent mapped files, we don't do any sanity checking:
 694          * just try to get the phys block of logical block 0 and run
 695          * with it.
 696          *
 697          * For inline data files, we just try to get the size of inline
 698          * data.  If it's true, we will treat it as a directory.
 699          */
 700
 701         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
 702         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
 703         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
 704                 size_t size;
 705                 __u32 dotdot;
 706                 unsigned int rec_len2;
 707                 struct ext2_dir_entry de;
 708
 709                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
 710                         return;
 711                 /*
 712                  * If the size isn't a multiple of 4, it's probably not a
 713                  * directory??
 714                  */
 715                 if (size & 3)
 716                         return;
 717                 /*
 718                  * If the first 10 bytes don't look like a directory entry,
 719                  * it's probably not a directory.
 720                  */
 721                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
 722                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
 723                        EXT2_DIR_REC_LEN(0));
 724                 dotdot = ext2fs_le32_to_cpu(dotdot);
 725                 de.inode = ext2fs_le32_to_cpu(de.inode);
 726                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
 727                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
 728                 if (dotdot >= ctx->fs->super->s_inodes_count ||
 729                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
 730                      dotdot != EXT2_ROOT_INO) ||
 731                     de.inode >= ctx->fs->super->s_inodes_count ||
 732                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
 733                      de.inode != 0) ||
 734                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
 735                               EXT4_INLINE_DATA_DOTDOT_SIZE)
 736                         return;
 737                 /* device files never have a "system.data" entry */
 738                 goto isdir;
 739         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
 740                 /* extent mapped */
 741                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
 742                                  &blk))
 743                         return;
 744                 /* device files are never extent mapped */
 745                 not_device++;
 746         } else {
 747                 for (i=0; i < EXT2_N_BLOCKS; i++) {
 748                         blk = inode->i_block[i];
 749                         if (!blk)
 750                                 continue;
 751                         if (i >= 4)
 752                                 not_device++;
 753
 754                         if (blk < ctx->fs->super->s_first_data_block ||
 755                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
 756                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
 757                                                            blk))
 758                                 return; /* Invalid block, can't be dir */
 759                 }
 760                 blk = inode->i_block[0];
 761         }
 762
 763         /*
 764          * If the mode says this is a device file and the i_links_count field
 765          * is sane and we have not ruled it out as a device file previously,
 766          * we declare it a device file, not a directory.
 767          */
 768         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
 769             (inode->i_links_count == 1) && !not_device)
 770                 return;
 771
 772         /* read the first block */
 773         ehandler_operation(_("reading directory block"));
 774         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
 775         ehandler_operation(0);
 776         if (retval)
 777                 return;
 778
 779         dirent = (struct ext2_dir_entry *) buf;
 780         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 781         if (retval)
 782                 return;
 783         if ((ext2fs_dirent_name_len(dirent) != 1) ||
 784             (dirent->name[0] != '.') ||
 785             (dirent->inode != pctx->ino) ||
 786             (rec_len < 12) ||
 787             (rec_len % 4) ||
 788             (rec_len >= ctx->fs->blocksize - 12))
 789                 return;
 790
 791         dirent = (struct ext2_dir_entry *) (buf + rec_len);
 792         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 793         if (retval)
 794                 return;
 795         if ((ext2fs_dirent_name_len(dirent) != 2) ||
 796             (dirent->name[0] != '.') ||
 797             (dirent->name[1] != '.') ||
 798             (rec_len < 12) ||
 799             (rec_len % 4))
 800                 return;
 801
 802 isdir:
 803         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
 804                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
 805                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
 806                                         EXT2_INODE_SIZE(ctx->fs->super),
 807                                         "check_is_really_dir");
 808         }
 809 }
 810
 811 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
 812                                      int flags, ext2_icount_t hint,
 813                                      ext2_icount_t *ret)
 814 {
 815         unsigned int            threshold;
 816         unsigned int            save_type;
 817         ext2_ino_t              num_dirs;
 818         errcode_t               retval;
 819         char                    *tdb_dir;
 820         int                     enable;
 821
 822         *ret = 0;
 823
 824         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
 825                            &tdb_dir);
 826         profile_get_uint(ctx->profile, "scratch_files",
 827                          "numdirs_threshold", 0, 0, &threshold);
 828         profile_get_boolean(ctx->profile, "scratch_files",
 829                             "icount", 0, 1, &enable);
 830
 831         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
 832         if (retval)
 833                 num_dirs = 1024;        /* Guess */
 834
 835         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
 836             (!threshold || num_dirs > threshold)) {
 837                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
 838                                                   flags, ret);
 839                 if (retval == 0)
 840                         return 0;
 841         }
 842         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
 843                                &save_type);
 844         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
 845                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
 846         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
 847         ctx->fs->default_bitmap_type = save_type;
 848         return retval;
 849 }
 850
 851 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
 852                                             e2fsck_t ctx,
 853                                             struct problem_context *pctx)
 854 {
 855         errcode_t retval;
 856         struct ext2_inode_large inode;
 857
 858         /*
 859          * Reread inode.  If we don't see checksum error, then this inode
 860          * has been fixed elsewhere.
 861          */
 862         ctx->stashed_ino = 0;
 863         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
 864                                         sizeof(inode));
 865         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
 866                 return retval;
 867         if (!retval)
 868                 return 0;
 869
 870         /*
 871          * Checksum still doesn't match.  That implies that the inode passes
 872          * all the sanity checks, so maybe the checksum is simply corrupt.
 873          * See if the user will go for fixing that.
 874          */
 875         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
 876                 return 0;
 877
 878         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
 879                                          sizeof(inode));
 880         return retval;
 881 }
 882
 883 static void reserve_block_for_root_repair(e2fsck_t ctx)
 884 {
 885         blk64_t         blk = 0;
 886         errcode_t       err;
 887         ext2_filsys     fs = ctx->fs;
 888
 889         ctx->root_repair_block = 0;
 890         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
 891                 return;
 892
 893         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 894         if (err)
 895                 return;
 896         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 897         ctx->root_repair_block = blk;
 898 }
 899
 900 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
 901 {
 902         blk64_t         blk = 0;
 903         errcode_t       err;
 904         ext2_filsys     fs = ctx->fs;
 905         static const char name[] = "lost+found";
 906         ext2_ino_t      ino;
 907
 908         ctx->lnf_repair_block = 0;
 909         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
 910                 return;
 911
 912         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 913         if (err)
 914                 return;
 915         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 916         ctx->lnf_repair_block = blk;
 917 }
 918
 919 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
 920                                          size_t *sz)
 921 {
 922         void *p;
 923         struct ext2_xattr_handle *handle;
 924         errcode_t retval;
 925
 926         retval = ext2fs_xattrs_open(fs, ino, &handle);
 927         if (retval)
 928                 return retval;
 929
 930         retval = ext2fs_xattrs_read(handle);
 931         if (retval)
 932                 goto err;
 933
 934         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
 935         if (retval)
 936                 goto err;
 937         ext2fs_free_mem(&p);
 938 err:
 939         (void) ext2fs_xattrs_close(&handle);
 940         return retval;
 941 }
 942
 943 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
 944                                     struct problem_context *pctx,
 945                                     int failed_csum)
 946 {
 947         if (!failed_csum)
 948                 return;
 949
 950         /*
 951          * If the inode failed the checksum and the user didn't
 952          * clear the inode, test the checksum again -- if it still
 953          * fails, ask the user if the checksum should be corrected.
 954          */
 955         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
 956         if (pctx->errcode)
 957                 ctx->flags |= E2F_FLAG_ABORT;
 958 }
 959 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
 960         do { \
 961                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
 962                 if ((ctx)->flags & E2F_FLAG_ABORT) \
 963                         return; \
 964         } while (0)
 965
 966 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
 967 {
 968         __u32 x;
 969         int i;
 970
 971         for (i = 0; i < EXT2_N_BLOCKS; i++) {
 972                 x = inode->i_block[i];
 973 #ifdef WORDS_BIGENDIAN
 974                 x = ext2fs_swab32(x);
 975 #endif
 976                 if (x >= ext2fs_blocks_count(fs->super))
 977                         return 0;
 978         }
 979
 980         return 1;
 981 }
 982
 983 /*
 984  * Figure out what to do with an inode that has both extents and inline data
 985  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
 986  */
 987 static int fix_inline_data_extents_file(e2fsck_t ctx,
 988                                         ext2_ino_t ino,
 989                                         struct ext2_inode *inode,
 990                                         int inode_size,
 991                                         struct problem_context *pctx)
 992 {
 993         size_t max_inline_ea_size;
 994         ext2_filsys fs = ctx->fs;
 995         int dirty = 0;
 996
 997         /* Both feature flags not set?  Just run the regular checks */
 998         if (!ext2fs_has_feature_extents(fs->super) &&
 999             !ext2fs_has_feature_inline_data(fs->super))
1000                 return 0;
1001
1002         /* Clear both flags if it's a special file */
1003         if (LINUX_S_ISCHR(inode->i_mode) ||
1004             LINUX_S_ISBLK(inode->i_mode) ||
1005             LINUX_S_ISFIFO(inode->i_mode) ||
1006             LINUX_S_ISSOCK(inode->i_mode)) {
1007                 check_extents_inlinedata(ctx, pctx);
1008                 return 0;
1009         }
1010
1011         /* If it looks like an extent tree, try to clear inlinedata */
1012         if (ext2fs_extent_header_verify(inode->i_block,
1013                                  sizeof(inode->i_block)) == 0 &&
1014             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1015                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1016                 dirty = 1;
1017                 goto out;
1018         }
1019
1020         /* If it looks short enough to be inline data, try to clear extents */
1021         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1022                 max_inline_ea_size = inode_size -
1023                                      (EXT2_GOOD_OLD_INODE_SIZE +
1024                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1025         else
1026                 max_inline_ea_size = 0;
1027         if (EXT2_I_SIZE(inode) <
1028             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1029             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1030                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1031                 dirty = 1;
1032                 goto out;
1033         }
1034
1035         /*
1036          * Too big for inline data, but no evidence of extent tree -
1037          * maybe it's a block map file?  If the mappings all look valid?
1038          */
1039         if (could_be_block_map(fs, inode) &&
1040             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1041 #ifdef WORDS_BIGENDIAN
1042                 int i;
1043
1044                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1045                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1046 #endif
1047
1048                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1049                 dirty = 1;
1050                 goto out;
1051         }
1052
1053         /* Oh well, just clear the busted inode. */
1054         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1055                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1056                 return -1;
1057         }
1058
1059 out:
1060         if (dirty)
1061                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1062
1063         return 0;
1064 }
1065
1066 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1067 {
1068         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1069         dgrp_t start = *group, grp;
1070         blk64_t blocks_to_read = 0;
1071         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1072
1073         if (ctx->readahead_kb == 0)
1074                 goto out;
1075
1076         /* Keep iterating groups until we have enough to readahead */
1077         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1078         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1079                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1080                         continue;
1081                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1082                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1083                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1084                                         inodes_per_block;
1085                 if (blocks_to_read * ctx->fs->blocksize >
1086                     ctx->readahead_kb * 1024)
1087                         break;
1088         }
1089
1090         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1091                                grp - start + 1);
1092         if (err == EAGAIN) {
1093                 ctx->readahead_kb /= 2;
1094                 err = 0;
1095         }
1096
1097 out:
1098         if (err) {
1099                 /* Error; disable itable readahead */
1100                 *group = ctx->fs->group_desc_count;
1101                 *next_ino = ctx->fs->super->s_inodes_count;
1102         } else {
1103                 /*
1104                  * Don't do more readahead until we've reached the first inode
1105                  * of the last inode scan buffer block for the last group.
1106                  */
1107                 *group = grp + 1;
1108                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1109                                      ctx->inode_buffer_blocks :
1110                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1111                                     ctx->fs->blocksize /
1112                                     EXT2_INODE_SIZE(ctx->fs->super);
1113                 inodes_in_group--;
1114                 *next_ino = inodes_in_group -
1115                             (inodes_in_group % inodes_per_buffer) + 1 +
1116                             (grp * ctx->fs->super->s_inodes_per_group);
1117         }
1118 }
1119
1120 /*
1121  * Check if the passed ino is one of the used superblock quota inodes.
1122  *
1123  * Before the quota inodes were journaled, older superblock quota inodes
1124  * were just regular files in the filesystem and not reserved inodes.  This
1125  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1126  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1127  */
1128 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1129 {
1130         enum quota_type qtype;
1131
1132         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1133                 if (*quota_sb_inump(sb, qtype) == ino)
1134                         return 1;
1135
1136         return 0;
1137 }
1138
1139 /*
1140  * Check if the passed ino is one of the reserved quota inodes.
1141  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1142  * inodes.  These inodes may or may not be in use by the quota feature.
1143  */
1144 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1145 {
1146         enum quota_type qtype;
1147
1148         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1149                 if (quota_type2inum(qtype, fs->super) == ino)
1150                         return 1;
1151
1152         return 0;
1153 }
1154
1155 static int e2fsck_should_abort(e2fsck_t ctx)
1156 {
1157         e2fsck_t global_ctx;
1158
1159         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1160                 return 1;
1161
1162         if (ctx->global_ctx) {
1163                 global_ctx = ctx->global_ctx;
1164                 if (global_ctx->flags & E2F_FLAG_SIGNAL_MASK)
1165                         return 1;
1166         }
1167         return 0;
1168 }
1169
1170 void e2fsck_pass1_run(e2fsck_t ctx)
1171 {
1172         int     i;
1173         __u64   max_sizes;
1174         ext2_filsys fs = ctx->fs;
1175         ext2_ino_t      ino = 0;
1176         struct ext2_inode *inode = NULL;
1177         ext2_inode_scan scan = NULL;
1178         char            *block_buf = NULL;
1179 #ifdef RESOURCE_TRACK
1180         struct resource_track   rtrack;
1181 #endif
1182         unsigned char   frag, fsize;
1183         struct          problem_context pctx;
1184         struct          scan_callback_struct scan_struct;
1185         struct ext2_super_block *sb = ctx->fs->super;
1186         const char      *old_op;
1187         const char      *eop_next_inode = _("getting next inode from scan");
1188         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1189         int             low_dtime_check = 1;
1190         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1191         unsigned int    bufsize;
1192         int             failed_csum = 0;
1193         ext2_ino_t      ino_threshold = 0;
1194         dgrp_t          ra_group = 0;
1195         struct ea_quota ea_ibody_quota;
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 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1221
1222         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1223                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1224                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1225                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1226                 max_sizes = (max_sizes * (1UL << i));
1227                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1228         }
1229 #undef EXT2_BPP
1230
1231         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1232         extent_fs = ext2fs_has_feature_extents(sb);
1233         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1234         casefold_fs = ext2fs_has_feature_casefold(sb);
1235
1236         /*
1237          * Allocate bitmaps structures
1238          */
1239         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1240                                                     EXT2FS_BMAP64_RBTREE,
1241                                                     "inode_used_map",
1242                                                     &ctx->inode_used_map);
1243         if (pctx.errcode) {
1244                 pctx.num = 1;
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                         _("directory inode map"),
1251                         EXT2FS_BMAP64_AUTODIR,
1252                         "inode_dir_map", &ctx->inode_dir_map);
1253         if (pctx.errcode) {
1254                 pctx.num = 2;
1255                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1256                 ctx->flags |= E2F_FLAG_ABORT;
1257                 return;
1258         }
1259         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1260                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1261                         "inode_reg_map", &ctx->inode_reg_map);
1262         if (pctx.errcode) {
1263                 pctx.num = 6;
1264                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1265                 ctx->flags |= E2F_FLAG_ABORT;
1266                 return;
1267         }
1268         pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1269                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1270                         "block_found_map", &ctx->block_found_map);
1271         if (pctx.errcode) {
1272                 pctx.num = 1;
1273                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1274                 ctx->flags |= E2F_FLAG_ABORT;
1275                 return;
1276         }
1277         pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1278                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1279                         "block_metadata_map", &ctx->block_metadata_map);
1280         if (pctx.errcode) {
1281                 pctx.num = 1;
1282                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1283                 ctx->flags |= E2F_FLAG_ABORT;
1284                 return;
1285         }
1286         if (casefold_fs) {
1287                 pctx.errcode =
1288                         e2fsck_allocate_inode_bitmap(fs,
1289                                                      _("inode casefold map"),
1290                                                      EXT2FS_BMAP64_RBTREE,
1291                                                      "inode_casefold_map",
1292                                                      &ctx->inode_casefold_map);
1293                 if (pctx.errcode) {
1294                         pctx.num = 1;
1295                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1296                         ctx->flags |= E2F_FLAG_ABORT;
1297                         return;
1298                 }
1299         }
1300         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1301                                            &ctx->inode_link_info);
1302         if (pctx.errcode) {
1303                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1304                 ctx->flags |= E2F_FLAG_ABORT;
1305                 return;
1306         }
1307         bufsize = inode_size;
1308         if (bufsize < sizeof(struct ext2_inode_large))
1309                 bufsize = sizeof(struct ext2_inode_large);
1310         inode = (struct ext2_inode *)
1311                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1312
1313         inodes_to_process = (struct process_inode_block *)
1314                 e2fsck_allocate_memory(ctx,
1315                                        (ctx->process_inode_size *
1316                                         sizeof(struct process_inode_block)),
1317                                        "array of inodes to process");
1318         process_inode_count = 0;
1319
1320         pctx.errcode = ext2fs_init_dblist(fs, 0);
1321         if (pctx.errcode) {
1322                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1323                 ctx->flags |= E2F_FLAG_ABORT;
1324                 goto endit;
1325         }
1326
1327         /*
1328          * If the last orphan field is set, clear it, since the pass1
1329          * processing will automatically find and clear the orphans.
1330          * In the future, we may want to try using the last_orphan
1331          * linked list ourselves, but for now, we clear it so that the
1332          * ext3 mount code won't get confused.
1333          */
1334         if (!(ctx->options & E2F_OPT_READONLY)) {
1335                 if (fs->super->s_last_orphan) {
1336                         fs->super->s_last_orphan = 0;
1337                         ext2fs_mark_super_dirty(fs);
1338                 }
1339         }
1340
1341         mark_table_blocks(ctx);
1342         pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1343                                                 &ctx->block_found_map);
1344         if (pctx.errcode) {
1345                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1346                 ctx->flags |= E2F_FLAG_ABORT;
1347                 goto endit;
1348         }
1349         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1350                                                     "block interate buffer");
1351         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1352                 e2fsck_use_inode_shortcuts(ctx, 1);
1353         e2fsck_intercept_block_allocations(ctx);
1354         old_op = ehandler_operation(_("opening inode scan"));
1355         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1356                                               &scan);
1357         ehandler_operation(old_op);
1358         if (pctx.errcode) {
1359                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1360                 ctx->flags |= E2F_FLAG_ABORT;
1361                 goto endit;
1362         }
1363         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1364                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1365         ctx->stashed_inode = inode;
1366         scan_struct.ctx = ctx;
1367         scan_struct.block_buf = block_buf;
1368         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1369         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1370                                               ctx->fs->group_desc_count)))
1371                 goto endit;
1372         if ((fs->super->s_wtime &&
1373              fs->super->s_wtime < fs->super->s_inodes_count) ||
1374             (fs->super->s_mtime &&
1375              fs->super->s_mtime < fs->super->s_inodes_count) ||
1376             (fs->super->s_mkfs_time &&
1377              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1378                 low_dtime_check = 0;
1379
1380         if (ext2fs_has_feature_mmp(fs->super) &&
1381             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1382             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1383                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1384                                           fs->super->s_mmp_block);
1385
1386         /* Set up ctx->lost_and_found if possible */
1387         (void) e2fsck_get_lost_and_found(ctx, 0);
1388
1389 #ifdef HAVE_PTHREAD
1390         if (ctx->global_ctx) {
1391                 if (ctx->options & E2F_OPT_DEBUG &&
1392                     ctx->options & E2F_OPT_MULTITHREAD)
1393                         fprintf(stderr, "thread %d jumping to group %d\n",
1394                                         ctx->thread_info.et_thread_index,
1395                                         ctx->thread_info.et_group_start);
1396                 pctx.errcode = ext2fs_inode_scan_goto_blockgroup(scan,
1397                                         ctx->thread_info.et_group_start);
1398                 if (pctx.errcode) {
1399                         fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1400                         ctx->flags |= E2F_FLAG_ABORT;
1401                         goto endit;
1402                 }
1403         }
1404 #endif
1405
1406         while (1) {
1407                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1408                         if (e2fsck_mmp_update(fs))
1409                                 fatal_error(ctx, 0);
1410                 }
1411                 old_op = ehandler_operation(eop_next_inode);
1412                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1413                                                           inode, inode_size);
1414                 if (ino > ino_threshold)
1415                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1416                 ehandler_operation(old_op);
1417                 if (e2fsck_should_abort(ctx))
1418                         goto endit;
1419                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1420                         /*
1421                          * If badblocks says badblocks is bad, offer to clear
1422                          * the list, update the in-core bb list, and restart
1423                          * the inode scan.
1424                          */
1425                         if (ino == EXT2_BAD_INO &&
1426                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1427                                         &pctx)) {
1428                                 errcode_t err;
1429
1430                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1431                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1432                                 ctx->fs->badblocks = NULL;
1433                                 err = ext2fs_read_bb_inode(ctx->fs,
1434                                                         &ctx->fs->badblocks);
1435                                 if (err) {
1436                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1437                                                     &pctx);
1438                                         ctx->flags |= E2F_FLAG_ABORT;
1439                                 } else
1440                                         ctx->flags |= E2F_FLAG_RESTART;
1441                                 goto endit;
1442                         }
1443                         if (!ctx->inode_bb_map)
1444                                 alloc_bb_map(ctx);
1445                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1446                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1447                         continue;
1448                 }
1449                 if (pctx.errcode == EXT2_ET_SCAN_FINISHED)
1450                         break;
1451                 if (pctx.errcode &&
1452                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1453                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1454                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1455                         ctx->flags |= E2F_FLAG_ABORT;
1456                         goto endit;
1457                 }
1458                 if (!ino)
1459                         break;
1460                 pctx.ino = ino;
1461                 pctx.inode = inode;
1462                 ctx->stashed_ino = ino;
1463
1464                 /* Clear trashed inode? */
1465                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1466                     inode->i_links_count > 0 &&
1467                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1468                         pctx.errcode = 0;
1469                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1470                 }
1471                 failed_csum = pctx.errcode != 0;
1472
1473                 /*
1474                  * Check for inodes who might have been part of the
1475                  * orphaned list linked list.  They should have gotten
1476                  * dealt with by now, unless the list had somehow been
1477                  * corrupted.
1478                  *
1479                  * FIXME: In the future, inodes which are still in use
1480                  * (and which are therefore) pending truncation should
1481                  * be handled specially.  Right now we just clear the
1482                  * dtime field, and the normal e2fsck handling of
1483                  * inodes where i_size and the inode blocks are
1484                  * inconsistent is to fix i_size, instead of releasing
1485                  * the extra blocks.  This won't catch the inodes that
1486                  * was at the end of the orphan list, but it's better
1487                  * than nothing.  The right answer is that there
1488                  * shouldn't be any bugs in the orphan list handling.  :-)
1489                  */
1490                 if (inode->i_dtime && low_dtime_check &&
1491                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1492                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1493                                 inode->i_dtime = inode->i_links_count ?
1494                                         0 : ctx->now;
1495                                 e2fsck_write_inode(ctx, ino, inode,
1496                                                    "pass1");
1497                                 failed_csum = 0;
1498                         }
1499                 }
1500
1501                 if (inode->i_links_count) {
1502                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1503                                            ino, inode->i_links_count);
1504                         if (pctx.errcode) {
1505                                 pctx.num = inode->i_links_count;
1506                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1507                                 ctx->flags |= E2F_FLAG_ABORT;
1508                                 goto endit;
1509                         }
1510                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1511                            !quota_inum_is_reserved(fs, ino)) {
1512                         if (!inode->i_dtime && inode->i_mode) {
1513                                 if (fix_problem(ctx,
1514                                             PR_1_ZERO_DTIME, &pctx)) {
1515                                         inode->i_dtime = ctx->now;
1516                                         e2fsck_write_inode(ctx, ino, inode,
1517                                                            "pass1");
1518                                         failed_csum = 0;
1519                                 }
1520                         }
1521                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1522                         continue;
1523                 }
1524
1525                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1526                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1527                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1528                      (!casefold_fs &&
1529                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1530                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1531                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1532                 }
1533
1534                 /* Conflicting inlinedata/extents inode flags? */
1535                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1536                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1537                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1538                                                                inode_size,
1539                                                                &pctx);
1540                         if (res < 0) {
1541                                 /* skip FINISH_INODE_LOOP */
1542                                 continue;
1543                         }
1544                 }
1545
1546                 /* Test for incorrect inline_data flags settings. */
1547                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1548                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1549                         size_t size = 0;
1550
1551                         pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1552                         if (!pctx.errcode &&
1553                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1554                                 ext2fs_set_feature_inline_data(sb);
1555                                 ext2fs_mark_super_dirty(fs);
1556                                 inlinedata_fs = 1;
1557                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1558                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1559                                 /* skip FINISH_INODE_LOOP */
1560                                 continue;
1561                         }
1562                 }
1563
1564                 /* Test for inline data flag but no attr */
1565                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1566                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1567                         size_t size = 0;
1568                         errcode_t err;
1569                         int flags;
1570
1571                         flags = fs->flags;
1572                         if (failed_csum)
1573                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1574                         err = get_inline_data_ea_size(fs, ino, &size);
1575                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1576                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1577
1578                         switch (err) {
1579                         case 0:
1580                                 /* Everything is awesome... */
1581                                 break;
1582                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1583                         case EXT2_ET_BAD_EA_HASH:
1584                         case EXT2_ET_BAD_EA_HEADER:
1585                         case EXT2_ET_EA_BAD_NAME_LEN:
1586                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1587                         case EXT2_ET_EA_KEY_NOT_FOUND:
1588                         case EXT2_ET_EA_NO_SPACE:
1589                         case EXT2_ET_MISSING_EA_FEATURE:
1590                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1591                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1592                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1593                         case EXT2_ET_NO_INLINE_DATA:
1594                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1595                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1596                         case EXT2_ET_EA_INODE_CORRUPTED:
1597                                 /* broken EA or no system.data EA; truncate */
1598                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1599                                                 &pctx)) {
1600                                         err = ext2fs_inode_size_set(fs, inode, 0);
1601                                         if (err) {
1602                                                 pctx.errcode = err;
1603                                                 ctx->flags |= E2F_FLAG_ABORT;
1604                                                 goto endit;
1605                                         }
1606                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1607                                         memset(&inode->i_block, 0,
1608                                                sizeof(inode->i_block));
1609                                         e2fsck_write_inode(ctx, ino, inode,
1610                                                            "pass1");
1611                                         failed_csum = 0;
1612                                 }
1613                                 break;
1614                         default:
1615                                 /* Some other kind of non-xattr error? */
1616                                 pctx.errcode = err;
1617                                 ctx->flags |= E2F_FLAG_ABORT;
1618                                 goto endit;
1619                         }
1620                 }
1621
1622                 /*
1623                  * Test for incorrect extent flag settings.
1624                  *
1625                  * On big-endian machines we must be careful:
1626                  * When the inode is read, the i_block array is not swapped
1627                  * if the extent flag is set.  Therefore if we are testing
1628                  * for or fixing a wrongly-set flag, we must potentially
1629                  * (un)swap before testing, or after fixing.
1630                  */
1631
1632                 /*
1633                  * In this case the extents flag was set when read, so
1634                  * extent_header_verify is ok.  If the inode is cleared,
1635                  * no need to swap... so no extra swapping here.
1636                  */
1637                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1638                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1639                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1640                         if ((ext2fs_extent_header_verify(inode->i_block,
1641                                                  sizeof(inode->i_block)) == 0) &&
1642                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1643                                 ext2fs_set_feature_extents(sb);
1644                                 ext2fs_mark_super_dirty(fs);
1645                                 extent_fs = 1;
1646                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1647                         clear_inode:
1648                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1649                                 if (ino == EXT2_BAD_INO)
1650                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1651                                                                  ino);
1652                                 /* skip FINISH_INODE_LOOP */
1653                                 continue;
1654                         }
1655                 }
1656
1657                 /*
1658                  * For big-endian machines:
1659                  * If the inode didn't have the extents flag set when it
1660                  * was read, then the i_blocks array was swapped.  To test
1661                  * as an extents header, we must swap it back first.
1662                  * IF we then set the extents flag, the entire i_block
1663                  * array must be un/re-swapped to make it proper extents data.
1664                  */
1665                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1666                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1667                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1668                     (LINUX_S_ISREG(inode->i_mode) ||
1669                      LINUX_S_ISDIR(inode->i_mode))) {
1670                         void *ehp;
1671 #ifdef WORDS_BIGENDIAN
1672                         __u32 tmp_block[EXT2_N_BLOCKS];
1673
1674                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1675                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1676                         ehp = tmp_block;
1677 #else
1678                         ehp = inode->i_block;
1679 #endif
1680                         if ((ext2fs_extent_header_verify(ehp,
1681                                          sizeof(inode->i_block)) == 0) &&
1682                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1683                                 inode->i_flags |= EXT4_EXTENTS_FL;
1684 #ifdef WORDS_BIGENDIAN
1685                                 memcpy(inode->i_block, tmp_block,
1686                                        sizeof(inode->i_block));
1687 #endif
1688                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1689                                 failed_csum = 0;
1690                         }
1691                 }
1692
1693                 if (ino == EXT2_BAD_INO) {
1694                         struct process_block_struct pb;
1695
1696                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1697                              inode->i_gid || inode->i_links_count ||
1698                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1699                              inode->i_file_acl) &&
1700                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1701                                 memset(inode, 0, sizeof(struct ext2_inode));
1702                                 e2fsck_write_inode(ctx, ino, inode,
1703                                                    "clear bad inode");
1704                                 failed_csum = 0;
1705                         }
1706
1707                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1708                                                           &pb.fs_meta_blocks);
1709                         if (pctx.errcode) {
1710                                 pctx.num = 4;
1711                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1712                                 ctx->flags |= E2F_FLAG_ABORT;
1713                                 goto endit;
1714                         }
1715                         pb.ino = EXT2_BAD_INO;
1716                         pb.num_blocks = pb.last_block = 0;
1717                         pb.last_db_block = -1;
1718                         pb.num_illegal_blocks = 0;
1719                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1720                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1721                         pb.inode = inode;
1722                         pb.pctx = &pctx;
1723                         pb.ctx = ctx;
1724                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1725                                      block_buf, process_bad_block, &pb);
1726                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1727                         if (pctx.errcode) {
1728                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1729                                 ctx->flags |= E2F_FLAG_ABORT;
1730                                 goto endit;
1731                         }
1732                         if (pb.bbcheck)
1733                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1734                                 ctx->flags |= E2F_FLAG_ABORT;
1735                                 goto endit;
1736                         }
1737                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1738                         clear_problem_context(&pctx);
1739                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1740                         continue;
1741                 } else if (ino == EXT2_ROOT_INO) {
1742                         /*
1743                          * Make sure the root inode is a directory; if
1744                          * not, offer to clear it.  It will be
1745                          * regenerated in pass #3.
1746                          */
1747                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1748                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1749                                         goto clear_inode;
1750                         }
1751                         /*
1752                          * If dtime is set, offer to clear it.  mke2fs
1753                          * version 0.2b created filesystems with the
1754                          * dtime field set for the root and lost+found
1755                          * directories.  We won't worry about
1756                          * /lost+found, since that can be regenerated
1757                          * easily.  But we will fix the root directory
1758                          * as a special case.
1759                          */
1760                         if (inode->i_dtime && inode->i_links_count) {
1761                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1762                                         inode->i_dtime = 0;
1763                                         e2fsck_write_inode(ctx, ino, inode,
1764                                                            "pass1");
1765                                         failed_csum = 0;
1766                                 }
1767                         }
1768                 } else if (ino == EXT2_JOURNAL_INO) {
1769                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1770                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1771                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1772                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1773                                                 &pctx)) {
1774                                         inode->i_mode = LINUX_S_IFREG;
1775                                         e2fsck_write_inode(ctx, ino, inode,
1776                                                            "pass1");
1777                                         failed_csum = 0;
1778                                 }
1779                                 check_blocks(ctx, &pctx, block_buf, NULL);
1780                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1781                                 continue;
1782                         }
1783                         if ((inode->i_links_count ||
1784                              inode->i_blocks || inode->i_block[0]) &&
1785                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1786                                         &pctx)) {
1787                                 memset(inode, 0, inode_size);
1788                                 ext2fs_icount_store(ctx->inode_link_info,
1789                                                     ino, 0);
1790                                 e2fsck_write_inode_full(ctx, ino, inode,
1791                                                         inode_size, "pass1");
1792                                 failed_csum = 0;
1793                         }
1794                 } else if (quota_inum_is_reserved(fs, ino)) {
1795                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1796                         if (ext2fs_has_feature_quota(fs->super) &&
1797                             quota_inum_is_super(fs->super, ino)) {
1798                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1799                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1800                                                         &pctx)) {
1801                                         inode->i_mode = LINUX_S_IFREG;
1802                                         e2fsck_write_inode(ctx, ino, inode,
1803                                                         "pass1");
1804                                         failed_csum = 0;
1805                                 }
1806                                 check_blocks(ctx, &pctx, block_buf, NULL);
1807                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1808                                 continue;
1809                         }
1810                         if ((inode->i_links_count ||
1811                              inode->i_blocks || inode->i_block[0]) &&
1812                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1813                                         &pctx)) {
1814                                 memset(inode, 0, inode_size);
1815                                 ext2fs_icount_store(ctx->inode_link_info,
1816                                                     ino, 0);
1817                                 e2fsck_write_inode_full(ctx, ino, inode,
1818                                                         inode_size, "pass1");
1819                                 failed_csum = 0;
1820                         }
1821                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1822                         problem_t problem = 0;
1823
1824                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1825                         if (ino == EXT2_BOOT_LOADER_INO) {
1826                                 if (LINUX_S_ISDIR(inode->i_mode))
1827                                         problem = PR_1_RESERVED_BAD_MODE;
1828                         } else if (ino == EXT2_RESIZE_INO) {
1829                                 if (inode->i_mode &&
1830                                     !LINUX_S_ISREG(inode->i_mode))
1831                                         problem = PR_1_RESERVED_BAD_MODE;
1832                         } else {
1833                                 if (inode->i_mode != 0)
1834                                         problem = PR_1_RESERVED_BAD_MODE;
1835                         }
1836                         if (problem) {
1837                                 if (fix_problem(ctx, problem, &pctx)) {
1838                                         inode->i_mode = 0;
1839                                         e2fsck_write_inode(ctx, ino, inode,
1840                                                            "pass1");
1841                                         failed_csum = 0;
1842                                 }
1843                         }
1844                         check_blocks(ctx, &pctx, block_buf, NULL);
1845                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1846                         continue;
1847                 }
1848
1849                 if (!inode->i_links_count) {
1850                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1851                         continue;
1852                 }
1853                 /*
1854                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1855                  * deleted files.  Oops.
1856                  *
1857                  * Since all new ext2 implementations get this right,
1858                  * we now assume that the case of non-zero
1859                  * i_links_count and non-zero dtime means that we
1860                  * should keep the file, not delete it.
1861                  *
1862                  */
1863                 if (inode->i_dtime) {
1864                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1865                                 inode->i_dtime = 0;
1866                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1867                                 failed_csum = 0;
1868                         }
1869                 }
1870
1871                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1872                 switch (fs->super->s_creator_os) {
1873                     case EXT2_OS_HURD:
1874                         frag = inode->osd2.hurd2.h_i_frag;
1875                         fsize = inode->osd2.hurd2.h_i_fsize;
1876                         break;
1877                     default:
1878                         frag = fsize = 0;
1879                 }
1880
1881                 if (inode->i_faddr || frag || fsize ||
1882                     (!ext2fs_has_feature_largedir(fs->super) &&
1883                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1884                         mark_inode_bad(ctx, ino);
1885                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1886                     !ext2fs_has_feature_64bit(fs->super) &&
1887                     inode->osd2.linux2.l_i_file_acl_high != 0)
1888                         mark_inode_bad(ctx, ino);
1889                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1890                     !ext2fs_has_feature_huge_file(fs->super) &&
1891                     (inode->osd2.linux2.l_i_blocks_hi != 0))
1892                         mark_inode_bad(ctx, ino);
1893                 if (inode->i_flags & EXT2_IMAGIC_FL) {
1894                         if (imagic_fs) {
1895                                 if (!ctx->inode_imagic_map)
1896                                         alloc_imagic_map(ctx);
1897                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1898                                                          ino);
1899                         } else {
1900                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1901                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
1902                                         e2fsck_write_inode(ctx, ino,
1903                                                            inode, "pass1");
1904                                         failed_csum = 0;
1905                                 }
1906                         }
1907                 }
1908
1909                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1910                 check_is_really_dir(ctx, &pctx, block_buf);
1911
1912                 /*
1913                  * ext2fs_inode_has_valid_blocks2 does not actually look
1914                  * at i_block[] values, so not endian-sensitive here.
1915                  */
1916                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1917                     LINUX_S_ISLNK(inode->i_mode) &&
1918                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1919                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1920                         inode->i_flags &= ~EXT4_EXTENTS_FL;
1921                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1922                         failed_csum = 0;
1923                 }
1924
1925                 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
1926                     add_encrypted_file(ctx, &pctx) < 0)
1927                         goto clear_inode;
1928
1929                 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
1930                         ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
1931
1932                 if (LINUX_S_ISDIR(inode->i_mode)) {
1933                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1934                         e2fsck_add_dir_info(ctx, ino, 0);
1935                         ctx->fs_directory_count++;
1936                         if (inode->i_flags & EXT4_CASEFOLD_FL)
1937                                 add_casefolded_dir(ctx, ino);
1938                 } else if (LINUX_S_ISREG (inode->i_mode)) {
1939                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1940                         ctx->fs_regular_count++;
1941                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1942                            e2fsck_pass1_check_device_inode(fs, inode)) {
1943                         check_extents_inlinedata(ctx, &pctx);
1944                         check_immutable(ctx, &pctx);
1945                         check_size(ctx, &pctx);
1946                         ctx->fs_chardev_count++;
1947                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1948                            e2fsck_pass1_check_device_inode(fs, inode)) {
1949                         check_extents_inlinedata(ctx, &pctx);
1950                         check_immutable(ctx, &pctx);
1951                         check_size(ctx, &pctx);
1952                         ctx->fs_blockdev_count++;
1953                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1954                            e2fsck_pass1_check_symlink(fs, ino, inode,
1955                                                       block_buf)) {
1956                         check_immutable(ctx, &pctx);
1957                         ctx->fs_symlinks_count++;
1958                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1959                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1960                                 continue;
1961                         } else if (ext2fs_is_fast_symlink(inode)) {
1962                                 ctx->fs_fast_symlinks_count++;
1963                                 check_blocks(ctx, &pctx, block_buf,
1964                                              &ea_ibody_quota);
1965                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1966                                 continue;
1967                         }
1968                 }
1969                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
1970                          e2fsck_pass1_check_device_inode(fs, inode)) {
1971                         check_extents_inlinedata(ctx, &pctx);
1972                         check_immutable(ctx, &pctx);
1973                         check_size(ctx, &pctx);
1974                         ctx->fs_fifo_count++;
1975                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
1976                            e2fsck_pass1_check_device_inode(fs, inode)) {
1977                         check_extents_inlinedata(ctx, &pctx);
1978                         check_immutable(ctx, &pctx);
1979                         check_size(ctx, &pctx);
1980                         ctx->fs_sockets_count++;
1981                 } else
1982                         mark_inode_bad(ctx, ino);
1983                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1984                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
1985                         if (inode->i_block[EXT2_IND_BLOCK])
1986                                 ctx->fs_ind_count++;
1987                         if (inode->i_block[EXT2_DIND_BLOCK])
1988                                 ctx->fs_dind_count++;
1989                         if (inode->i_block[EXT2_TIND_BLOCK])
1990                                 ctx->fs_tind_count++;
1991                 }
1992                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1993                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
1994                     (inode->i_block[EXT2_IND_BLOCK] ||
1995                      inode->i_block[EXT2_DIND_BLOCK] ||
1996                      inode->i_block[EXT2_TIND_BLOCK] ||
1997                      ext2fs_file_acl_block(fs, inode))) {
1998                         struct process_inode_block *itp;
1999
2000                         itp = &inodes_to_process[process_inode_count];
2001                         itp->ino = ino;
2002                         itp->ea_ibody_quota = ea_ibody_quota;
2003                         if (inode_size < sizeof(struct ext2_inode_large))
2004                                 memcpy(&itp->inode, inode, inode_size);
2005                         else
2006                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
2007                         process_inode_count++;
2008                 } else
2009                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
2010
2011                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2012
2013                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2014                         goto endit;
2015
2016                 if (process_inode_count >= ctx->process_inode_size) {
2017                         process_inodes(ctx, block_buf);
2018
2019                         if (e2fsck_should_abort(ctx))
2020                                 goto endit;
2021                 }
2022         }
2023         process_inodes(ctx, block_buf);
2024         ext2fs_close_inode_scan(scan);
2025         scan = NULL;
2026
2027         reserve_block_for_root_repair(ctx);
2028         reserve_block_for_lnf_repair(ctx);
2029
2030         /*
2031          * If any extended attribute blocks' reference counts need to
2032          * be adjusted, either up (ctx->refcount_extra), or down
2033          * (ctx->refcount), then fix them.
2034          */
2035         if (ctx->refcount) {
2036                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
2037                 ea_refcount_free(ctx->refcount);
2038                 ctx->refcount = 0;
2039         }
2040         if (ctx->refcount_extra) {
2041                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
2042                                         block_buf, +1);
2043                 ea_refcount_free(ctx->refcount_extra);
2044                 ctx->refcount_extra = 0;
2045         }
2046
2047         if (ctx->ea_block_quota_blocks) {
2048                 ea_refcount_free(ctx->ea_block_quota_blocks);
2049                 ctx->ea_block_quota_blocks = 0;
2050         }
2051
2052         if (ctx->ea_block_quota_inodes) {
2053                 ea_refcount_free(ctx->ea_block_quota_inodes);
2054                 ctx->ea_block_quota_inodes = 0;
2055         }
2056
2057         if (ctx->invalid_bitmaps)
2058                 handle_fs_bad_blocks(ctx);
2059
2060         /* We don't need the block_ea_map any more */
2061         if (ctx->block_ea_map) {
2062                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2063                 ctx->block_ea_map = 0;
2064         }
2065
2066         /* We don't need the encryption policy => ID map any more */
2067         destroy_encryption_policy_map(ctx);
2068
2069         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2070                 clear_problem_context(&pctx);
2071                 pctx.errcode = ext2fs_create_resize_inode(fs);
2072                 if (pctx.errcode) {
2073                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2074                                          &pctx)) {
2075                                 ctx->flags |= E2F_FLAG_ABORT;
2076                                 goto endit;
2077                         }
2078                         pctx.errcode = 0;
2079                 }
2080                 if (!pctx.errcode) {
2081                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2082                                           "recreate inode");
2083                         inode->i_mtime = ctx->now;
2084                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2085                                            "recreate inode");
2086                 }
2087                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2088         }
2089
2090         if (ctx->flags & E2F_FLAG_RESTART) {
2091                 /*
2092                  * Only the master copy of the superblock and block
2093                  * group descriptors are going to be written during a
2094                  * restart, so set the superblock to be used to be the
2095                  * master superblock.
2096                  */
2097                 ctx->use_superblock = 0;
2098                 goto endit;
2099         }
2100
2101         if (ctx->large_dirs && !ext2fs_has_feature_largedir(fs->super)) {
2102                 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2103                         ext2fs_set_feature_largedir(fs->super);
2104                         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2105                         ext2fs_mark_super_dirty(fs);
2106                 }
2107                 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2108                     fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2109                         ext2fs_update_dynamic_rev(fs);
2110                         ext2fs_mark_super_dirty(fs);
2111                 }
2112         }
2113
2114         if (ctx->block_dup_map) {
2115                 if (ctx->options & E2F_OPT_PREEN) {
2116                         clear_problem_context(&pctx);
2117                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2118                 }
2119                 e2fsck_pass1_dupblocks(ctx, block_buf);
2120         }
2121         ctx->flags |= E2F_FLAG_ALLOC_OK;
2122 endit:
2123         e2fsck_use_inode_shortcuts(ctx, 0);
2124         ext2fs_free_mem(&inodes_to_process);
2125         inodes_to_process = 0;
2126
2127         if (scan)
2128                 ext2fs_close_inode_scan(scan);
2129         if (block_buf)
2130                 ext2fs_free_mem(&block_buf);
2131         if (inode)
2132                 ext2fs_free_mem(&inode);
2133
2134         /*
2135          * The l+f inode may have been cleared, so zap it now and
2136          * later passes will recalculate it if necessary
2137          */
2138         ctx->lost_and_found = 0;
2139
2140         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2141                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2142         else
2143                 ctx->invalid_bitmaps++;
2144 }
2145
2146 #ifdef HAVE_PTHREAD
2147 static errcode_t e2fsck_pass1_copy_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2148                                           ext2fs_generic_bitmap *dest)
2149 {
2150         errcode_t ret;
2151
2152         ret = ext2fs_copy_bitmap(*src, dest);
2153         if (ret)
2154                 return ret;
2155
2156         (*dest)->fs = fs;
2157         ext2fs_free_generic_bmap(*src);
2158         *src = NULL;
2159
2160         return 0;
2161 }
2162
2163
2164 static errcode_t e2fsck_pass1_copy_fs(ext2_filsys dest, e2fsck_t src_context,
2165                                       ext2_filsys src)
2166 {
2167         errcode_t       retval;
2168
2169         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2170         if (dest->dblist)
2171                 dest->dblist->fs = dest;
2172         if (src->block_map) {
2173                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2174                                                   &dest->block_map);
2175                 if (retval)
2176                         return retval;
2177         }
2178         if (src->inode_map) {
2179                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2180                                                   &dest->inode_map);
2181                 if (retval)
2182                         return retval;
2183         }
2184
2185         if (src->badblocks) {
2186                 retval = ext2fs_badblocks_copy(src->badblocks, &dest->badblocks);
2187                 if (retval)
2188                         return retval;
2189         }
2190
2191         /* disable it for now */
2192         src_context->openfs_flags &= ~EXT2_FLAG_EXCLUSIVE;
2193         retval = ext2fs_open_channel(dest, src_context->io_options,
2194                                      src_context->io_manager,
2195                                      src_context->openfs_flags,
2196                                      src->io->block_size);
2197         if (retval)
2198                 return retval;
2199
2200         /* Block size might not be default */
2201         io_channel_set_blksize(dest->io, src->io->block_size);
2202         ehandler_init(dest->io);
2203
2204         assert(dest->io->magic == src->io->magic);
2205         assert(dest->io->manager == src->io->manager);
2206         assert(strcmp(dest->io->name, src->io->name) == 0);
2207         assert(dest->io->block_size == src->io->block_size);
2208         assert(dest->io->read_error == src->io->read_error);
2209         assert(dest->io->write_error == src->io->write_error);
2210         assert(dest->io->refcount == src->io->refcount);
2211         assert(dest->io->flags == src->io->flags);
2212         assert(dest->io->app_data == dest);
2213         assert(src->io->app_data == src);
2214         assert(dest->io->align == src->io->align);
2215
2216         /* The data should be written to disk immediately */
2217         dest->io->flags |= CHANNEL_FLAGS_WRITETHROUGH;
2218         /* icache will be rebuilt if needed, so do not copy from @src */
2219         src->icache = NULL;
2220         return 0;
2221 }
2222
2223 static int e2fsck_pass1_merge_fs(ext2_filsys dest, ext2_filsys src)
2224 {
2225         struct ext2_inode_cache *icache = dest->icache;
2226         errcode_t retval = 0;
2227         io_channel dest_io;
2228         io_channel dest_image_io;
2229
2230         dest_io = dest->io;
2231         dest_image_io = dest->image_io;
2232
2233         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2234         dest->io = dest_io;
2235         dest->image_io = dest_image_io;
2236         dest->icache = icache;
2237         if (dest->dblist)
2238                 dest->dblist->fs = dest;
2239         if (src->inode_map) {
2240                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2241                                                   &dest->inode_map);
2242                 if (retval)
2243                         return retval;
2244         }
2245         if (src->block_map) {
2246                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2247                                                   &dest->block_map);
2248                 if (retval)
2249                         return retval;
2250         }
2251
2252         if (src->icache) {
2253                 ext2fs_free_inode_cache(src->icache);
2254                 src->icache = NULL;
2255         }
2256
2257         if (src->badblocks) {
2258                 retval = ext2fs_badblocks_copy(src->badblocks, &dest->badblocks);
2259
2260                 ext2fs_badblocks_list_free(src->badblocks);
2261                 src->badblocks = NULL;
2262         }
2263
2264         io_channel_close(src->io);
2265         return retval;
2266 }
2267
2268 static errcode_t e2fsck_pass1_thread_prepare(e2fsck_t global_ctx, e2fsck_t *thread_ctx,
2269                                              int thread_index, int num_threads)
2270 {
2271         errcode_t               retval;
2272         e2fsck_t                thread_context;
2273         ext2_filsys             thread_fs;
2274         ext2_filsys             global_fs = global_ctx->fs;
2275         struct e2fsck_thread    *tinfo;
2276         dgrp_t                  average_group;
2277
2278         assert(global_ctx->inode_used_map == NULL);
2279         assert(global_ctx->inode_dir_map == NULL);
2280         assert(global_ctx->inode_bb_map == NULL);
2281         assert(global_ctx->inode_imagic_map == NULL);
2282         assert(global_ctx->inode_reg_map == NULL);
2283         assert(global_ctx->inodes_to_rebuild == NULL);
2284
2285         assert(global_ctx->block_found_map == NULL);
2286         assert(global_ctx->block_dup_map == NULL);
2287         assert(global_ctx->block_ea_map == NULL);
2288         assert(global_ctx->block_metadata_map == NULL);
2289         assert(global_ctx->fs->dblist == NULL);
2290
2291         retval = ext2fs_get_mem(sizeof(struct e2fsck_struct), &thread_context);
2292         if (retval) {
2293                 com_err(global_ctx->program_name, retval, "while allocating memory");
2294                 return retval;
2295         }
2296         memcpy(thread_context, global_ctx, sizeof(struct e2fsck_struct));
2297         thread_context->global_ctx = global_ctx;
2298
2299         retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &thread_fs);
2300         if (retval) {
2301                 com_err(global_ctx->program_name, retval, "while allocating memory");
2302                 goto out_context;
2303         }
2304
2305         io_channel_flush_cleanup(global_fs->io);
2306         retval = e2fsck_pass1_copy_fs(thread_fs, global_ctx, global_fs);
2307         if (retval) {
2308                 com_err(global_ctx->program_name, retval, "while copying fs");
2309                 goto out_fs;
2310         }
2311         thread_fs->priv_data = thread_context;
2312
2313         thread_context->thread_info.et_thread_index = thread_index;
2314         set_up_logging(thread_context);
2315
2316         /*
2317          * Distribute work to multiple threads:
2318          * Each thread work on fs->group_desc_count / nthread groups.
2319          */
2320         tinfo = &thread_context->thread_info;
2321         average_group = thread_fs->group_desc_count / num_threads;
2322         if (average_group == 0)
2323                 average_group = 1;
2324         tinfo->et_group_start = average_group * thread_index;
2325         if (thread_index == num_threads - 1)
2326                 tinfo->et_group_end = thread_fs->group_desc_count;
2327         else
2328                 tinfo->et_group_end = average_group * (thread_index + 1);
2329         tinfo->et_group_next = tinfo->et_group_start;
2330
2331         thread_context->fs = thread_fs;
2332         *thread_ctx = thread_context;
2333         return 0;
2334 out_fs:
2335         ext2fs_free_mem(&thread_fs);
2336 out_context:
2337         ext2fs_free_mem(&thread_context);
2338         return retval;
2339 }
2340
2341 static int e2fsck_pass1_thread_join_one(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2342 {
2343         errcode_t        retval;
2344         int              flags = global_ctx->flags;
2345         ext2_filsys      thread_fs = thread_ctx->fs;
2346         ext2_filsys      global_fs = global_ctx->fs;
2347         FILE            *global_logf = global_ctx->logf;
2348         FILE            *global_problem_logf = global_ctx->problem_logf;
2349 #ifdef HAVE_SETJMP_H
2350         jmp_buf          old_jmp;
2351
2352         memcpy(old_jmp, global_ctx->abort_loc, sizeof(jmp_buf));
2353 #endif
2354         memcpy(global_ctx, thread_ctx, sizeof(struct e2fsck_struct));
2355 #ifdef HAVE_SETJMP_H
2356         memcpy(global_ctx->abort_loc, old_jmp, sizeof(jmp_buf));
2357 #endif
2358         /* Keep the global singal flags*/
2359         global_ctx->flags |= (flags & E2F_FLAG_SIGNAL_MASK) |
2360                              (global_ctx->flags & E2F_FLAG_SIGNAL_MASK);
2361
2362         retval = e2fsck_pass1_merge_fs(global_fs, thread_fs);
2363         if (retval) {
2364                 com_err(global_ctx->program_name, 0, _("while merging fs\n"));
2365                 return retval;
2366         }
2367         global_fs->priv_data = global_ctx;
2368         global_ctx->fs = global_fs;
2369         global_ctx->logf = global_logf;
2370         global_ctx->problem_logf = global_problem_logf;
2371
2372         if (thread_ctx->inode_used_map) {
2373                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2374                                         &thread_ctx->inode_used_map,
2375                                         &global_ctx->inode_used_map);
2376                 if (retval)
2377                         return retval;
2378         }
2379         if (thread_ctx->inode_bad_map) {
2380                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2381                                         &thread_ctx->inode_bad_map,
2382                                         &global_ctx->inode_bad_map);
2383                 if (retval)
2384                         return retval;
2385         }
2386         if (thread_ctx->inode_dir_map) {
2387                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2388                                         &thread_ctx->inode_dir_map,
2389                                         &global_ctx->inode_dir_map);
2390                 if (retval)
2391                         return retval;
2392         }
2393         if (thread_ctx->inode_bb_map) {
2394                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2395                                         &thread_ctx->inode_bb_map,
2396                                         &global_ctx->inode_bb_map);
2397                 if (retval)
2398                         return retval;
2399         }
2400         if (thread_ctx->inode_imagic_map) {
2401                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2402                                         &thread_ctx->inode_imagic_map,
2403                                         &global_ctx->inode_imagic_map);
2404                 if (retval)
2405                         return retval;
2406         }
2407         if (thread_ctx->inode_reg_map) {
2408                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2409                                         &thread_ctx->inode_reg_map,
2410                                         &global_ctx->inode_reg_map);
2411                 if (retval)
2412                         return retval;
2413         }
2414         if (thread_ctx->inodes_to_rebuild) {
2415                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2416                                         &thread_ctx->inodes_to_rebuild,
2417                                         &global_ctx->inodes_to_rebuild);
2418                 if (retval)
2419                         return retval;
2420         }
2421         if (thread_ctx->block_found_map) {
2422                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2423                                         &thread_ctx->block_found_map,
2424                                         &global_ctx->block_found_map);
2425                 if (retval)
2426                         return retval;
2427         }
2428         if (thread_ctx->block_dup_map) {
2429                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2430                                         &thread_ctx->block_dup_map,
2431                                         &global_ctx->block_dup_map);
2432                 if (retval)
2433                         return retval;
2434         }
2435         if (thread_ctx->block_ea_map) {
2436                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2437                                         &thread_ctx->block_ea_map,
2438                                         &global_ctx->block_ea_map);
2439                 if (retval)
2440                         return retval;
2441         }
2442         if (thread_ctx->block_metadata_map) {
2443                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2444                                         &thread_ctx->block_metadata_map,
2445                                         &global_ctx->block_metadata_map);
2446                 if (retval)
2447                         return retval;
2448         }
2449
2450         return 0;
2451 }
2452
2453 static int e2fsck_pass1_thread_join(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2454 {
2455         errcode_t       retval;
2456
2457         retval = e2fsck_pass1_thread_join_one(global_ctx, thread_ctx);
2458         ext2fs_free_mem(&thread_ctx->fs);
2459         if (thread_ctx->logf)
2460                 fclose(thread_ctx->logf);
2461         if (thread_ctx->problem_logf) {
2462                 fputs("</problem_log>\n", thread_ctx->problem_logf);
2463                 fclose(thread_ctx->problem_logf);
2464         }
2465         ext2fs_free_mem(&thread_ctx);
2466
2467         return retval;
2468 }
2469
2470 static int e2fsck_pass1_threads_join(struct e2fsck_thread_info *infos,
2471                                       int num_threads, e2fsck_t global_ctx)
2472 {
2473         errcode_t                        rc;
2474         errcode_t                        ret = 0;
2475         int                              i;
2476         struct e2fsck_thread_info       *pinfo;
2477
2478         for (i = 0; i < num_threads; i++) {
2479                 pinfo = &infos[i];
2480
2481                 if (!pinfo->eti_started)
2482                         continue;
2483
2484                 rc = pthread_join(pinfo->eti_thread_id, NULL);
2485                 if (rc) {
2486                         com_err(global_ctx->program_name, rc,
2487                                 _("while joining thread\n"));
2488                         if (ret == 0)
2489                                 ret = rc;
2490                 }
2491                 e2fsck_pass1_thread_join(global_ctx, infos[i].eti_thread_ctx);
2492         }
2493         free(infos);
2494
2495         return ret;
2496 }
2497
2498 static void *e2fsck_pass1_thread(void *arg)
2499 {
2500         struct e2fsck_thread_info       *info = arg;
2501         e2fsck_t                         thread_ctx = info->eti_thread_ctx;
2502
2503 #ifdef HAVE_SETJMP_H
2504         /*
2505          * When fatal_error() happens, jump to here. The thread
2506          * context's flags will be saved, but its abort_loc will
2507          * be overwritten by original jump buffer for the later
2508          * tests.
2509          */
2510         if (setjmp(thread_ctx->abort_loc)) {
2511                 thread_ctx->flags &= ~E2F_FLAG_SETJMP_OK;
2512                 goto out;
2513         }
2514         thread_ctx->flags |= E2F_FLAG_SETJMP_OK;
2515 #endif
2516
2517         e2fsck_pass1_run(thread_ctx);
2518
2519 out:
2520         return NULL;
2521 }
2522
2523 static int e2fsck_pass1_threads_start(struct e2fsck_thread_info **pinfo,
2524                                       int num_threads, e2fsck_t global_ctx)
2525 {
2526         struct e2fsck_thread_info       *infos;
2527         pthread_attr_t                   attr;
2528         errcode_t                        retval;
2529         errcode_t                        ret;
2530         struct e2fsck_thread_info       *tmp_pinfo;
2531         int                              i;
2532         e2fsck_t                         thread_ctx;
2533
2534         retval = pthread_attr_init(&attr);
2535         if (retval) {
2536                 com_err(global_ctx->program_name, retval,
2537                         _("while setting pthread attribute\n"));
2538                 return retval;
2539         }
2540
2541         infos = calloc(num_threads, sizeof(struct e2fsck_thread_info));
2542         if (infos == NULL) {
2543                 retval = -ENOMEM;
2544                 com_err(global_ctx->program_name, retval,
2545                         _("while allocating memory for threads\n"));
2546                 pthread_attr_destroy(&attr);
2547                 return retval;
2548         }
2549
2550         for (i = 0; i < num_threads; i++) {
2551                 tmp_pinfo = &infos[i];
2552                 tmp_pinfo->eti_thread_index = i;
2553                 retval = e2fsck_pass1_thread_prepare(global_ctx, &thread_ctx,
2554                                                      i, num_threads);
2555                 if (retval) {
2556                         com_err(global_ctx->program_name, retval,
2557                                 _("while preparing pass1 thread\n"));
2558                         break;
2559                 }
2560                 tmp_pinfo->eti_thread_ctx = thread_ctx;
2561
2562                 retval = pthread_create(&tmp_pinfo->eti_thread_id, &attr,
2563                                         &e2fsck_pass1_thread, tmp_pinfo);
2564                 if (retval) {
2565                         com_err(global_ctx->program_name, retval,
2566                                 _("while creating thread\n"));
2567                         e2fsck_pass1_thread_join(global_ctx, thread_ctx);
2568                         break;
2569                 }
2570
2571                 tmp_pinfo->eti_started = 1;
2572         }
2573
2574         /* destroy the thread attribute object, since it is no longer needed */
2575         ret = pthread_attr_destroy(&attr);
2576         if (ret) {
2577                 com_err(global_ctx->program_name, ret,
2578                         _("while destroying thread attribute\n"));
2579                 if (retval == 0)
2580                         retval = ret;
2581         }
2582
2583         if (retval) {
2584                 e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
2585                 return retval;
2586         }
2587         *pinfo = infos;
2588         return 0;
2589 }
2590
2591 static void e2fsck_pass1_multithread(e2fsck_t global_ctx)
2592 {
2593         struct e2fsck_thread_info       *infos = NULL;
2594         int                              num_threads = 1;
2595         errcode_t                        retval;
2596
2597         retval = e2fsck_pass1_threads_start(&infos, num_threads, global_ctx);
2598         if (retval) {
2599                 com_err(global_ctx->program_name, retval,
2600                         _("while starting pass1 threads\n"));
2601                 goto out_abort;
2602         }
2603
2604         retval = e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
2605         if (retval) {
2606                 com_err(global_ctx->program_name, retval,
2607                         _("while joining pass1 threads\n"));
2608                 goto out_abort;
2609         }
2610         return;
2611 out_abort:
2612         global_ctx->flags |= E2F_FLAG_ABORT;
2613         return;
2614 }
2615 #endif
2616
2617 void e2fsck_pass1(e2fsck_t ctx)
2618 {
2619
2620 #ifdef HAVE_PTHREAD
2621         e2fsck_pass1_multithread(ctx);
2622 #else
2623         e2fsck_pass1_run(ctx);
2624 #endif
2625 }
2626
2627 #undef FINISH_INODE_LOOP
2628
2629 /*
2630  * When the inode_scan routines call this callback at the end of the
2631  * glock group, call process_inodes.
2632  */
2633 static errcode_t scan_callback(ext2_filsys fs,
2634                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
2635                                dgrp_t group, void * priv_data)
2636 {
2637         struct scan_callback_struct *scan_struct;
2638         e2fsck_t ctx;
2639         struct e2fsck_thread *tinfo;
2640
2641         scan_struct = (struct scan_callback_struct *) priv_data;
2642         ctx = scan_struct->ctx;
2643
2644         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf);
2645
2646         if (ctx->progress)
2647                 if ((ctx->progress)(ctx, 1, group+1,
2648                                     ctx->fs->group_desc_count))
2649                         return EXT2_ET_CANCEL_REQUESTED;
2650
2651 #ifdef HAVE_PTHREAD
2652         if (ctx->global_ctx) {
2653                 tinfo = &ctx->thread_info;
2654                 tinfo->et_group_next++;
2655                 if (tinfo->et_group_next >= tinfo->et_group_end)
2656                         return EXT2_ET_SCAN_FINISHED;
2657         }
2658 #endif
2659
2660         return 0;
2661 }
2662
2663 /*
2664  * Process the inodes in the "inodes to process" list.
2665  */
2666 static void process_inodes(e2fsck_t ctx, char *block_buf)
2667 {
2668         int                     i;
2669         struct ext2_inode       *old_stashed_inode;
2670         ext2_ino_t              old_stashed_ino;
2671         const char              *old_operation;
2672         char                    buf[80];
2673         struct problem_context  pctx;
2674
2675 #if 0
2676         printf("begin process_inodes: ");
2677 #endif
2678         if (process_inode_count == 0)
2679                 return;
2680         old_operation = ehandler_operation(0);
2681         old_stashed_inode = ctx->stashed_inode;
2682         old_stashed_ino = ctx->stashed_ino;
2683         qsort(inodes_to_process, process_inode_count,
2684                       sizeof(struct process_inode_block), process_inode_cmp);
2685         clear_problem_context(&pctx);
2686         for (i=0; i < process_inode_count; i++) {
2687                 pctx.inode = ctx->stashed_inode =
2688                         (struct ext2_inode *) &inodes_to_process[i].inode;
2689                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2690
2691 #if 0
2692                 printf("%u ", pctx.ino);
2693 #endif
2694                 sprintf(buf, _("reading indirect blocks of inode %u"),
2695                         pctx.ino);
2696                 ehandler_operation(buf);
2697                 check_blocks(ctx, &pctx, block_buf,
2698                              &inodes_to_process[i].ea_ibody_quota);
2699                 if (e2fsck_should_abort(ctx))
2700                         break;
2701         }
2702         ctx->stashed_inode = old_stashed_inode;
2703         ctx->stashed_ino = old_stashed_ino;
2704         process_inode_count = 0;
2705 #if 0
2706         printf("end process inodes\n");
2707 #endif
2708         ehandler_operation(old_operation);
2709 }
2710
2711 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
2712 {
2713         const struct process_inode_block *ib_a =
2714                 (const struct process_inode_block *) a;
2715         const struct process_inode_block *ib_b =
2716                 (const struct process_inode_block *) b;
2717         int     ret;
2718
2719         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
2720                ib_b->inode.i_block[EXT2_IND_BLOCK]);
2721         if (ret == 0)
2722                 /*
2723                  * We only call process_inodes() for non-extent
2724                  * inodes, so it's OK to pass NULL to
2725                  * ext2fs_file_acl_block() here.
2726                  */
2727                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
2728                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
2729         if (ret == 0)
2730                 ret = ib_a->ino - ib_b->ino;
2731         return ret;
2732 }
2733
2734 /*
2735  * Mark an inode as being bad in some what
2736  */
2737 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
2738 {
2739         struct          problem_context pctx;
2740
2741         if (!ctx->inode_bad_map) {
2742                 clear_problem_context(&pctx);
2743
2744                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2745                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
2746                                 "inode_bad_map", &ctx->inode_bad_map);
2747                 if (pctx.errcode) {
2748                         pctx.num = 3;
2749                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2750                         /* Should never get here */
2751                         ctx->flags |= E2F_FLAG_ABORT;
2752                         return;
2753                 }
2754         }
2755         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
2756 }
2757
2758 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino)
2759 {
2760         struct          problem_context pctx;
2761
2762         if (!ctx->casefolded_dirs) {
2763                 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
2764                 if (pctx.errcode)
2765                         goto error;
2766         }
2767         pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
2768         if (pctx.errcode == 0)
2769                 return;
2770 error:
2771         fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
2772         /* Should never get here */
2773         ctx->flags |= E2F_FLAG_ABORT;
2774 }
2775
2776 /*
2777  * This procedure will allocate the inode "bb" (badblock) map table
2778  */
2779 static void alloc_bb_map(e2fsck_t ctx)
2780 {
2781         struct          problem_context pctx;
2782
2783         clear_problem_context(&pctx);
2784         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2785                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
2786                         "inode_bb_map", &ctx->inode_bb_map);
2787         if (pctx.errcode) {
2788                 pctx.num = 4;
2789                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2790                 /* Should never get here */
2791                 ctx->flags |= E2F_FLAG_ABORT;
2792                 return;
2793         }
2794 }
2795
2796 /*
2797  * This procedure will allocate the inode imagic table
2798  */
2799 static void alloc_imagic_map(e2fsck_t ctx)
2800 {
2801         struct          problem_context pctx;
2802
2803         clear_problem_context(&pctx);
2804         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2805                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
2806                         "inode_imagic_map", &ctx->inode_imagic_map);
2807         if (pctx.errcode) {
2808                 pctx.num = 5;
2809                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2810                 /* Should never get here */
2811                 ctx->flags |= E2F_FLAG_ABORT;
2812                 return;
2813         }
2814 }
2815
2816 /*
2817  * Marks a block as in use, setting the dup_map if it's been set
2818  * already.  Called by process_block and process_bad_block.
2819  *
2820  * WARNING: Assumes checks have already been done to make sure block
2821  * is valid.  This is true in both process_block and process_bad_block.
2822  */
2823 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
2824 {
2825         struct          problem_context pctx;
2826
2827         clear_problem_context(&pctx);
2828
2829         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
2830                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
2831                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
2832                         return;
2833                 }
2834                 if (!ctx->block_dup_map) {
2835                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
2836                                         _("multiply claimed block map"),
2837                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
2838                                         &ctx->block_dup_map);
2839                         if (pctx.errcode) {
2840                                 pctx.num = 3;
2841                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
2842                                             &pctx);
2843                                 /* Should never get here */
2844                                 ctx->flags |= E2F_FLAG_ABORT;
2845                                 return;
2846                         }
2847                 }
2848                 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
2849         } else {
2850                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
2851         }
2852 }
2853
2854 /*
2855  * When cluster size is greater than one block, it is caller's responsibility
2856  * to make sure block parameter starts at a cluster boundary.
2857  */
2858 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
2859                                       unsigned int num)
2860 {
2861         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
2862                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
2863         else {
2864                 unsigned int i;
2865
2866                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
2867                         mark_block_used(ctx, block + i);
2868         }
2869 }
2870
2871 /*
2872  * Adjust the extended attribute block's reference counts at the end
2873  * of pass 1, either by subtracting out references for EA blocks that
2874  * are still referenced in ctx->refcount, or by adding references for
2875  * EA blocks that had extra references as accounted for in
2876  * ctx->refcount_extra.
2877  */
2878 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
2879                                     char *block_buf, int adjust_sign)
2880 {
2881         struct ext2_ext_attr_header     *header;
2882         struct problem_context          pctx;
2883         ext2_filsys                     fs = ctx->fs;
2884         blk64_t                         blk;
2885         __u32                           should_be;
2886         ea_value_t                      count;
2887
2888         clear_problem_context(&pctx);
2889
2890         ea_refcount_intr_begin(refcount);
2891         while (1) {
2892                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
2893                         break;
2894                 pctx.blk = blk;
2895                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
2896                                                      pctx.ino);
2897                 if (pctx.errcode) {
2898                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
2899                         return;
2900                 }
2901                 header = (struct ext2_ext_attr_header *) block_buf;
2902                 pctx.blkcount = header->h_refcount;
2903                 should_be = header->h_refcount + adjust_sign * (int)count;
2904                 pctx.num = should_be;
2905                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
2906                         header->h_refcount = should_be;
2907                         pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
2908                                                              block_buf,
2909                                                              pctx.ino);
2910                         if (pctx.errcode) {
2911                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
2912                                             &pctx);
2913                                 continue;
2914                         }
2915                 }
2916         }
2917 }
2918
2919 /*
2920  * Handle processing the extended attribute blocks
2921  */
2922 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
2923                            char *block_buf, struct ea_quota *ea_block_quota)
2924 {
2925         ext2_filsys fs = ctx->fs;
2926         ext2_ino_t      ino = pctx->ino;
2927         struct ext2_inode *inode = pctx->inode;
2928         blk64_t         blk;
2929         char *          end;
2930         struct ext2_ext_attr_header *header;
2931         struct ext2_ext_attr_entry *first, *entry;
2932         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
2933         __u64           quota_inodes = 0;
2934         region_t        region = 0;
2935         int             failed_csum = 0;
2936
2937         ea_block_quota->blocks = 0;
2938         ea_block_quota->inodes = 0;
2939
2940         blk = ext2fs_file_acl_block(fs, inode);
2941         if (blk == 0)
2942                 return 0;
2943
2944         /*
2945          * If the Extended attribute flag isn't set, then a non-zero
2946          * file acl means that the inode is corrupted.
2947          *
2948          * Or if the extended attribute block is an invalid block,
2949          * then the inode is also corrupted.
2950          */
2951         if (!ext2fs_has_feature_xattr(fs->super) ||
2952             (blk < fs->super->s_first_data_block) ||
2953             (blk >= ext2fs_blocks_count(fs->super))) {
2954                 mark_inode_bad(ctx, ino);
2955                 return 0;
2956         }
2957
2958         /* If ea bitmap hasn't been allocated, create it */
2959         if (!ctx->block_ea_map) {
2960                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
2961                                         _("ext attr block map"),
2962                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
2963                                         &ctx->block_ea_map);
2964                 if (pctx->errcode) {
2965                         pctx->num = 2;
2966                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
2967                         ctx->flags |= E2F_FLAG_ABORT;
2968                         return 0;
2969                 }
2970         }
2971
2972         /* Create the EA refcount structure if necessary */
2973         if (!ctx->refcount) {
2974                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
2975                 if (pctx->errcode) {
2976                         pctx->num = 1;
2977                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2978                         ctx->flags |= E2F_FLAG_ABORT;
2979                         return 0;
2980                 }
2981         }
2982
2983 #if 0
2984         /* Debugging text */
2985         printf("Inode %u has EA block %u\n", ino, blk);
2986 #endif
2987
2988         /* Have we seen this EA block before? */
2989         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
2990                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
2991                 ea_block_quota->inodes = 0;
2992
2993                 if (ctx->ea_block_quota_blocks) {
2994                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
2995                                           &quota_blocks);
2996                         if (quota_blocks)
2997                                 ea_block_quota->blocks = quota_blocks;
2998                 }
2999
3000                 if (ctx->ea_block_quota_inodes)
3001                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
3002                                           &ea_block_quota->inodes);
3003
3004                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
3005                         return 1;
3006                 /* Ooops, this EA was referenced more than it stated */
3007                 if (!ctx->refcount_extra) {
3008                         pctx->errcode = ea_refcount_create(0,
3009                                            &ctx->refcount_extra);
3010                         if (pctx->errcode) {
3011                                 pctx->num = 2;
3012                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3013                                 ctx->flags |= E2F_FLAG_ABORT;
3014                                 return 0;
3015                         }
3016                 }
3017                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
3018                 return 1;
3019         }
3020
3021         /*
3022          * OK, we haven't seen this EA block yet.  So we need to
3023          * validate it
3024          */
3025         pctx->blk = blk;
3026         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
3027         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
3028                 pctx->errcode = 0;
3029                 failed_csum = 1;
3030         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
3031                 pctx->errcode = 0;
3032
3033         if (pctx->errcode &&
3034             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
3035                 pctx->errcode = 0;
3036                 goto clear_extattr;
3037         }
3038         header = (struct ext2_ext_attr_header *) block_buf;
3039         pctx->blk = ext2fs_file_acl_block(fs, inode);
3040         if (((ctx->ext_attr_ver == 1) &&
3041              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
3042             ((ctx->ext_attr_ver == 2) &&
3043              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
3044                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
3045                         goto clear_extattr;
3046         }
3047
3048         if (header->h_blocks != 1) {
3049                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
3050                         goto clear_extattr;
3051         }
3052
3053         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
3054                 goto clear_extattr;
3055
3056         region = region_create(0, fs->blocksize);
3057         if (!region) {
3058                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
3059                 ctx->flags |= E2F_FLAG_ABORT;
3060                 return 0;
3061         }
3062         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
3063                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3064                         goto clear_extattr;
3065         }
3066
3067         first = (struct ext2_ext_attr_entry *)(header+1);
3068         end = block_buf + fs->blocksize;
3069         entry = first;
3070         while ((char *)entry < end && *(__u32 *)entry) {
3071                 __u32 hash;
3072
3073                 if (region_allocate(region, (char *)entry - (char *)header,
3074                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
3075                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3076                                 goto clear_extattr;
3077                         break;
3078                 }
3079                 if ((ctx->ext_attr_ver == 1 &&
3080                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
3081                     (ctx->ext_attr_ver == 2 &&
3082                      entry->e_name_index == 0)) {
3083                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
3084                                 goto clear_extattr;
3085                         break;
3086                 }
3087                 if (entry->e_value_inum == 0) {
3088                         if (entry->e_value_offs + entry->e_value_size >
3089                             fs->blocksize) {
3090                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
3091                                         goto clear_extattr;
3092                                 break;
3093                         }
3094                         if (entry->e_value_size &&
3095                             region_allocate(region, entry->e_value_offs,
3096                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
3097                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
3098                                                 pctx))
3099                                         goto clear_extattr;
3100                         }
3101
3102                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
3103                                                           entry->e_value_offs);
3104
3105                         if (entry->e_hash != hash) {
3106                                 pctx->num = entry->e_hash;
3107                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
3108                                         goto clear_extattr;
3109                                 entry->e_hash = hash;
3110                         }
3111                 } else {
3112                         problem_t problem;
3113                         blk64_t entry_quota_blocks;
3114
3115                         problem = check_large_ea_inode(ctx, entry, pctx,
3116                                                        &entry_quota_blocks);
3117                         if (problem && fix_problem(ctx, problem, pctx))
3118                                 goto clear_extattr;
3119
3120                         quota_blocks += entry_quota_blocks;
3121                         quota_inodes++;
3122                 }
3123
3124                 entry = EXT2_EXT_ATTR_NEXT(entry);
3125         }
3126         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
3127                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3128                         goto clear_extattr;
3129         }
3130         region_free(region);
3131
3132         /*
3133          * We only get here if there was no other errors that were fixed.
3134          * If there was a checksum fail, ask to correct it.
3135          */
3136         if (failed_csum &&
3137             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
3138                 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
3139                                                        pctx->ino);
3140                 if (pctx->errcode)
3141                         return 0;
3142         }
3143
3144         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
3145                 if (!ctx->ea_block_quota_blocks) {
3146                         pctx->errcode = ea_refcount_create(0,
3147                                                 &ctx->ea_block_quota_blocks);
3148                         if (pctx->errcode) {
3149                                 pctx->num = 3;
3150                                 goto refcount_fail;
3151                         }
3152                 }
3153                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
3154                                   quota_blocks);
3155         }
3156
3157         if (quota_inodes) {
3158                 if (!ctx->ea_block_quota_inodes) {
3159                         pctx->errcode = ea_refcount_create(0,
3160                                                 &ctx->ea_block_quota_inodes);
3161                         if (pctx->errcode) {
3162                                 pctx->num = 4;
3163 refcount_fail:
3164                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3165                                 ctx->flags |= E2F_FLAG_ABORT;
3166                                 return 0;
3167                         }
3168                 }
3169
3170                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
3171                                   quota_inodes);
3172         }
3173         ea_block_quota->blocks = quota_blocks;
3174         ea_block_quota->inodes = quota_inodes;
3175
3176         inc_ea_inode_refs(ctx, pctx, first, end);
3177         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
3178         mark_block_used(ctx, blk);
3179         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
3180         return 1;
3181
3182 clear_extattr:
3183         if (region)
3184                 region_free(region);
3185         ext2fs_file_acl_block_set(fs, inode, 0);
3186         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
3187         return 0;
3188 }
3189
3190 /* Returns 1 if bad htree, 0 if OK */
3191 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
3192                         ext2_ino_t ino, struct ext2_inode *inode,
3193                         char *block_buf)
3194 {
3195         struct ext2_dx_root_info        *root;
3196         ext2_filsys                     fs = ctx->fs;
3197         errcode_t                       retval;
3198         blk64_t                         blk;
3199
3200         if ((!LINUX_S_ISDIR(inode->i_mode) &&
3201              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
3202             (!ext2fs_has_feature_dir_index(fs->super) &&
3203              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
3204                 return 1;
3205
3206         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
3207
3208         if ((pctx->errcode) ||
3209             (blk == 0) ||
3210             (blk < fs->super->s_first_data_block) ||
3211             (blk >= ext2fs_blocks_count(fs->super))) {
3212                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3213                         return 1;
3214                 else
3215                         return 0;
3216         }
3217
3218         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
3219         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3220                 return 1;
3221
3222         /* XXX should check that beginning matches a directory */
3223         root = (struct ext2_dx_root_info *) (block_buf + 24);
3224
3225         if ((root->reserved_zero || root->info_length < 8) &&
3226             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3227                 return 1;
3228
3229         pctx->num = root->hash_version;
3230         if ((root->hash_version != EXT2_HASH_LEGACY) &&
3231             (root->hash_version != EXT2_HASH_HALF_MD4) &&
3232             (root->hash_version != EXT2_HASH_TEA) &&
3233             (root->hash_version != EXT2_HASH_SIPHASH) &&
3234             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
3235                 return 1;
3236
3237         if (ext4_hash_in_dirent(inode)) {
3238                 if (root->hash_version != EXT2_HASH_SIPHASH &&
3239                     fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
3240                         return 1;
3241         } else {
3242                 if (root->hash_version == EXT2_HASH_SIPHASH &&
3243                    fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
3244                         return 1;
3245         }
3246
3247         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
3248             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
3249                 return 1;
3250
3251         pctx->num = root->indirect_levels;
3252         /* if htree level is clearly too high, consider it to be broken */
3253         if (root->indirect_levels > EXT4_HTREE_LEVEL &&
3254             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3255                 return 1;
3256
3257         /* if level is only maybe too high, LARGE_DIR feature could be unset */
3258         if (root->indirect_levels > ext2_dir_htree_level(fs) &&
3259             !ext2fs_has_feature_largedir(fs->super)) {
3260                 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
3261                 unsigned idx_pb = 1 << (blockbits - 3);
3262
3263                 /* compare inode size/blocks vs. max-sized 2-level htree */
3264                 if (EXT2_I_SIZE(pctx->inode) <
3265                     (idx_pb - 1) * (idx_pb - 2) << blockbits &&
3266                     pctx->inode->i_blocks <
3267                     (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
3268                     fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3269                         return 1;
3270         }
3271
3272         if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
3273             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3274                 ctx->large_dirs++;
3275
3276         return 0;
3277 }
3278
3279 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
3280                         struct ext2_inode *inode, int restart_flag,
3281                         const char *source)
3282 {
3283         inode->i_flags = 0;
3284         inode->i_links_count = 0;
3285         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
3286         inode->i_dtime = ctx->now;
3287
3288         /*
3289          * If a special inode has such rotten block mappings that we
3290          * want to clear the whole inode, be sure to actually zap
3291          * the block maps because i_links_count isn't checked for
3292          * special inodes, and we'll end up right back here the next
3293          * time we run fsck.
3294          */
3295         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
3296                 memset(inode->i_block, 0, sizeof(inode->i_block));
3297
3298         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
3299         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
3300         if (ctx->inode_reg_map)
3301                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
3302         if (ctx->inode_bad_map)
3303                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
3304
3305         /*
3306          * If the inode was partially accounted for before processing
3307          * was aborted, we need to restart the pass 1 scan.
3308          */
3309         ctx->flags |= restart_flag;
3310
3311         if (ino == EXT2_BAD_INO)
3312                 memset(inode, 0, sizeof(struct ext2_inode));
3313
3314         e2fsck_write_inode(ctx, ino, inode, source);
3315 }
3316
3317 /*
3318  * Use the multiple-blocks reclamation code to fix alignment problems in
3319  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
3320  * physical cluster, and we want the block offsets within that cluster to
3321  * line up.
3322  */
3323 static int has_unaligned_cluster_map(e2fsck_t ctx,
3324                                      blk64_t last_pblk, blk64_t last_lblk,
3325                                      blk64_t pblk, blk64_t lblk)
3326 {
3327         blk64_t cluster_mask;
3328
3329         if (!ctx->fs->cluster_ratio_bits)
3330                 return 0;
3331         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
3332
3333         /*
3334          * If the block in the logical cluster doesn't align with the block in
3335          * the physical cluster...
3336          */
3337         if ((lblk & cluster_mask) != (pblk & cluster_mask))
3338                 return 1;
3339
3340         /*
3341          * If we cross a physical cluster boundary within a logical cluster...
3342          */
3343         if (last_pblk && (lblk & cluster_mask) != 0 &&
3344             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
3345             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
3346                 return 1;
3347
3348         return 0;
3349 }
3350
3351 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
3352                              struct process_block_struct *pb,
3353                              blk64_t start_block, blk64_t end_block,
3354                              blk64_t eof_block,
3355                              ext2_extent_handle_t ehandle,
3356                              int try_repairs)
3357 {
3358         struct ext2fs_extent    extent;
3359         blk64_t                 blk, last_lblk;
3360         unsigned int            i, n;
3361         int                     is_dir, is_leaf;
3362         problem_t               problem;
3363         struct ext2_extent_info info;
3364         int                     failed_csum = 0;
3365
3366         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
3367                 failed_csum = 1;
3368
3369         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
3370         if (pctx->errcode)
3371                 return;
3372         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
3373             !pb->eti.force_rebuild) {
3374                 struct extent_tree_level *etl;
3375
3376                 etl = pb->eti.ext_info + info.curr_level;
3377                 etl->num_extents += info.num_entries;
3378                 etl->max_extents += info.max_entries;
3379                 /*
3380                  * Implementation wart: Splitting extent blocks when appending
3381                  * will leave the old block with one free entry.  Therefore
3382                  * unless the node is totally full, pretend that a non-root
3383                  * extent block can hold one fewer entry than it actually does,
3384                  * so that we don't repeatedly rebuild the extent tree.
3385                  */
3386                 if (info.curr_level && info.num_entries < info.max_entries)
3387                         etl->max_extents--;
3388         }
3389
3390         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
3391                                           &extent);
3392         while ((pctx->errcode == 0 ||
3393                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
3394                info.num_entries-- > 0) {
3395                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
3396                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
3397                 last_lblk = extent.e_lblk + extent.e_len - 1;
3398
3399                 problem = 0;
3400                 pctx->blk = extent.e_pblk;
3401                 pctx->blk2 = extent.e_lblk;
3402                 pctx->num = extent.e_len;
3403                 pctx->blkcount = extent.e_lblk + extent.e_len;
3404
3405                 if (extent.e_pblk == 0 ||
3406                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
3407                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
3408                         problem = PR_1_EXTENT_BAD_START_BLK;
3409                 else if (extent.e_lblk < start_block)
3410                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
3411                 else if ((end_block && last_lblk > end_block) &&
3412                          !(last_lblk > eof_block &&
3413                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
3414                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
3415                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
3416                 else if (is_leaf && extent.e_len == 0)
3417                         problem = PR_1_EXTENT_LENGTH_ZERO;
3418                 else if (is_leaf &&
3419                          (extent.e_pblk + extent.e_len) >
3420                          ext2fs_blocks_count(ctx->fs->super))
3421                         problem = PR_1_EXTENT_ENDS_BEYOND;
3422                 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
3423                          !ext2fs_has_feature_largedir(ctx->fs->super) &&
3424                          ((extent.e_lblk + extent.e_len) >
3425                           (1U << (21 - ctx->fs->super->s_log_block_size))))
3426                         problem = PR_1_TOOBIG_DIR;
3427
3428                 if (is_leaf && problem == 0 && extent.e_len > 0) {
3429 #if 0
3430                         printf("extent_region(ino=%u, expect=%llu, "
3431                                "lblk=%llu, len=%u)\n", pb->ino,
3432                                (unsigned long long) pb->next_lblock,
3433                                (unsigned long long) extent.e_lblk,
3434                                extent.e_len);
3435 #endif
3436                         if (extent.e_lblk < pb->next_lblock)
3437                                 problem = PR_1_EXTENT_COLLISION;
3438                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
3439                                 pb->next_lblock = extent.e_lblk + extent.e_len;
3440                 }
3441
3442                 /*
3443                  * Uninitialized blocks in a directory?  Clear the flag and
3444                  * we'll interpret the blocks later.
3445                  */
3446                 if (try_repairs && is_dir && problem == 0 &&
3447                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3448                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
3449                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3450                         pb->inode_modified = 1;
3451                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3452                                                               &extent);
3453                         if (pctx->errcode)
3454                                 return;
3455                         failed_csum = 0;
3456                 }
3457 #ifdef CONFIG_DEVELOPER_FEATURES
3458                 if (try_repairs && !is_dir && problem == 0 &&
3459                     (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
3460                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3461                     fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
3462                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3463                         pb->inode_modified = 1;
3464                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3465                                                               &extent);
3466                         if (pctx->errcode)
3467                                 return;
3468                         failed_csum = 0;
3469                 }
3470 #endif
3471                 if (try_repairs && problem) {
3472 report_problem:
3473                         if (fix_problem(ctx, problem, pctx)) {
3474                                 if (ctx->invalid_bitmaps) {
3475                                         /*
3476                                          * If fsck knows the bitmaps are bad,
3477                                          * skip to the next extent and
3478                                          * try to clear this extent again
3479                                          * after fixing the bitmaps, by
3480                                          * restarting fsck.
3481                                          */
3482                                         pctx->errcode = ext2fs_extent_get(
3483                                                           ehandle,
3484                                                           EXT2_EXTENT_NEXT_SIB,
3485                                                           &extent);
3486                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3487                                         if (pctx->errcode ==
3488                                                     EXT2_ET_NO_CURRENT_NODE) {
3489                                                 pctx->errcode = 0;
3490                                                 break;
3491                                         }
3492                                         continue;
3493                                 }
3494                                 e2fsck_read_bitmaps(ctx);
3495                                 pb->inode_modified = 1;
3496                                 pctx->errcode =
3497                                         ext2fs_extent_delete(ehandle, 0);
3498                                 if (pctx->errcode) {
3499                                         pctx->str = "ext2fs_extent_delete";
3500                                         return;
3501                                 }
3502                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3503                                 if (pctx->errcode &&
3504                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
3505                                         pctx->str = "ext2fs_extent_fix_parents";
3506                                         return;
3507                                 }
3508                                 pctx->errcode = ext2fs_extent_get(ehandle,
3509                                                                   EXT2_EXTENT_CURRENT,
3510                                                                   &extent);
3511                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
3512                                         pctx->errcode = 0;
3513                                         break;
3514                                 }
3515                                 failed_csum = 0;
3516                                 continue;
3517                         }
3518                         goto next;
3519                 }
3520
3521                 if (!is_leaf) {
3522                         blk64_t lblk = extent.e_lblk;
3523                         int next_try_repairs = 1;
3524
3525                         blk = extent.e_pblk;
3526
3527                         /*
3528                          * If this lower extent block collides with critical
3529                          * metadata, don't try to repair the damage.  Pass 1b
3530                          * will reallocate the block; then we can try again.
3531                          */
3532                         if (pb->ino != EXT2_RESIZE_INO &&
3533                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
3534                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3535                                                       extent.e_pblk)) {
3536                                 next_try_repairs = 0;
3537                                 pctx->blk = blk;
3538                                 fix_problem(ctx,
3539                                             PR_1_CRITICAL_METADATA_COLLISION,
3540                                             pctx);
3541                                 if ((ctx->options & E2F_OPT_NO) == 0)
3542                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3543                         }
3544                         pctx->errcode = ext2fs_extent_get(ehandle,
3545                                                   EXT2_EXTENT_DOWN, &extent);
3546                         if (pctx->errcode &&
3547                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
3548                                 pctx->str = "EXT2_EXTENT_DOWN";
3549                                 problem = PR_1_EXTENT_HEADER_INVALID;
3550                                 if (!next_try_repairs)
3551                                         return;
3552                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
3553                                         goto report_problem;
3554                                 return;
3555                         }
3556                         /* The next extent should match this index's logical start */
3557                         if (extent.e_lblk != lblk) {
3558                                 struct ext2_extent_info e_info;
3559
3560                                 pctx->errcode = ext2fs_extent_get_info(ehandle,
3561                                                                        &e_info);
3562                                 if (pctx->errcode) {
3563                                         pctx->str = "ext2fs_extent_get_info";
3564                                         return;
3565                                 }
3566                                 pctx->blk = lblk;
3567                                 pctx->blk2 = extent.e_lblk;
3568                                 pctx->num = e_info.curr_level - 1;
3569                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
3570                                 if (fix_problem(ctx, problem, pctx)) {
3571                                         pb->inode_modified = 1;
3572                                         pctx->errcode =
3573                                                 ext2fs_extent_fix_parents(ehandle);
3574                                         if (pctx->errcode) {
3575                                                 pctx->str = "ext2fs_extent_fix_parents";
3576                                                 return;
3577                                         }
3578                                 }
3579                         }
3580                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
3581                                          last_lblk, eof_block, ehandle,
3582                                          next_try_repairs);
3583                         if (pctx->errcode)
3584                                 return;
3585                         pctx->errcode = ext2fs_extent_get(ehandle,
3586                                                   EXT2_EXTENT_UP, &extent);
3587                         if (pctx->errcode) {
3588                                 pctx->str = "EXT2_EXTENT_UP";
3589                                 return;
3590                         }
3591                         mark_block_used(ctx, blk);
3592                         pb->num_blocks++;
3593                         goto next;
3594                 }
3595
3596                 if ((pb->previous_block != 0) &&
3597                     (pb->previous_block+1 != extent.e_pblk)) {
3598                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3599                                 char type = '?';
3600
3601                                 if (pb->is_dir)
3602                                         type = 'd';
3603                                 else if (pb->is_reg)
3604                                         type = 'f';
3605
3606                                 printf(("%6lu(%c): expecting %6lu "
3607                                         "actual extent "
3608                                         "phys %6lu log %lu len %lu\n"),
3609                                        (unsigned long) pctx->ino, type,
3610                                        (unsigned long) pb->previous_block+1,
3611                                        (unsigned long) extent.e_pblk,
3612                                        (unsigned long) extent.e_lblk,
3613                                        (unsigned long) extent.e_len);
3614                         }
3615                         pb->fragmented = 1;
3616                 }
3617                 /*
3618                  * If we notice a gap in the logical block mappings of an
3619                  * extent-mapped directory, offer to close the hole by
3620                  * moving the logical block down, otherwise we'll go mad in
3621                  * pass 3 allocating empty directory blocks to fill the hole.
3622                  */
3623                 if (try_repairs && is_dir &&
3624                     pb->last_block + 1 < extent.e_lblk) {
3625                         blk64_t new_lblk;
3626
3627                         new_lblk = pb->last_block + 1;
3628                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3629                                 new_lblk = ((new_lblk +
3630                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3631                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3632                                            (extent.e_pblk &
3633                                             EXT2FS_CLUSTER_MASK(ctx->fs));
3634                         pctx->blk = extent.e_lblk;
3635                         pctx->blk2 = new_lblk;
3636                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3637                                 extent.e_lblk = new_lblk;
3638                                 pb->inode_modified = 1;
3639                                 pctx->errcode = ext2fs_extent_replace(ehandle,
3640                                                                 0, &extent);
3641                                 if (pctx->errcode) {
3642                                         pctx->errcode = 0;
3643                                         goto alloc_later;
3644                                 }
3645                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3646                                 if (pctx->errcode)
3647                                         goto failed_add_dir_block;
3648                                 pctx->errcode = ext2fs_extent_goto(ehandle,
3649                                                                 extent.e_lblk);
3650                                 if (pctx->errcode)
3651                                         goto failed_add_dir_block;
3652                                 last_lblk = extent.e_lblk + extent.e_len - 1;
3653                                 failed_csum = 0;
3654                         }
3655                 }
3656 alloc_later:
3657                 if (is_dir) {
3658                         while (++pb->last_db_block <
3659                                (e2_blkcnt_t) extent.e_lblk) {
3660                                 pctx->errcode = ext2fs_add_dir_block2(
3661                                                         ctx->fs->dblist,
3662                                                         pb->ino, 0,
3663                                                         pb->last_db_block);
3664                                 if (pctx->errcode) {
3665                                         pctx->blk = 0;
3666                                         pctx->num = pb->last_db_block;
3667                                         goto failed_add_dir_block;
3668                                 }
3669                         }
3670
3671                         for (i = 0; i < extent.e_len; i++) {
3672                                 pctx->errcode = ext2fs_add_dir_block2(
3673                                                         ctx->fs->dblist,
3674                                                         pctx->ino,
3675                                                         extent.e_pblk + i,
3676                                                         extent.e_lblk + i);
3677                                 if (pctx->errcode) {
3678                                         pctx->blk = extent.e_pblk + i;
3679                                         pctx->num = extent.e_lblk + i;
3680                                 failed_add_dir_block:
3681                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3682                                         /* Should never get here */
3683                                         ctx->flags |= E2F_FLAG_ABORT;
3684                                         return;
3685                                 }
3686                         }
3687                         if (extent.e_len > 0)
3688                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3689                 }
3690                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3691                                               pb->last_block,
3692                                               extent.e_pblk,
3693                                               extent.e_lblk)) {
3694                         for (i = 0; i < extent.e_len; i++) {
3695                                 pctx->blk = extent.e_lblk + i;
3696                                 pctx->blk2 = extent.e_pblk + i;
3697                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3698                                 mark_block_used(ctx, extent.e_pblk + i);
3699                                 mark_block_used(ctx, extent.e_pblk + i);
3700                         }
3701                 }
3702
3703                 /*
3704                  * Check whether first cluster got marked in previous iteration.
3705                  */
3706                 if (ctx->fs->cluster_ratio_bits &&
3707                     pb->previous_block &&
3708                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
3709                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
3710                         /* Set blk to the beginning of next cluster. */
3711                         blk = EXT2FS_C2B(
3712                                 ctx->fs,
3713                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
3714                 else
3715                         /* Set blk to the beginning of current cluster. */
3716                         blk = EXT2FS_C2B(ctx->fs,
3717                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
3718
3719                 if (blk < extent.e_pblk + extent.e_len) {
3720                         mark_blocks_used(ctx, blk,
3721                                          extent.e_pblk + extent.e_len - blk);
3722                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
3723                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
3724                         pb->num_blocks += n;
3725                 }
3726                 pb->last_block = extent.e_lblk + extent.e_len - 1;
3727                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
3728                 start_block = pb->last_block = last_lblk;
3729                 if (is_leaf && !is_dir &&
3730                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
3731                         pb->last_init_lblock = last_lblk;
3732         next:
3733                 pctx->errcode = ext2fs_extent_get(ehandle,
3734                                                   EXT2_EXTENT_NEXT_SIB,
3735                                                   &extent);
3736         }
3737
3738         /* Failed csum but passes checks?  Ask to fix checksum. */
3739         if (failed_csum &&
3740             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
3741                 pb->inode_modified = 1;
3742                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
3743                 if (pctx->errcode)
3744                         return;
3745         }
3746
3747         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
3748                 pctx->errcode = 0;
3749 }
3750
3751 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
3752                                  struct process_block_struct *pb)
3753 {
3754         struct ext2_extent_info info;
3755         struct ext2_inode       *inode = pctx->inode;
3756         ext2_extent_handle_t    ehandle;
3757         ext2_filsys             fs = ctx->fs;
3758         ext2_ino_t              ino = pctx->ino;
3759         errcode_t               retval;
3760         blk64_t                 eof_lblk;
3761         struct ext3_extent_header       *eh;
3762
3763         /* Check for a proper extent header... */
3764         eh = (struct ext3_extent_header *) &inode->i_block[0];
3765         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
3766         if (retval) {
3767                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
3768                         e2fsck_clear_inode(ctx, ino, inode, 0,
3769                                            "check_blocks_extents");
3770                 pctx->errcode = 0;
3771                 return;
3772         }
3773
3774         /* ...since this function doesn't fail if i_block is zeroed. */
3775         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
3776         if (pctx->errcode) {
3777                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
3778                         e2fsck_clear_inode(ctx, ino, inode, 0,
3779                                            "check_blocks_extents");
3780                 pctx->errcode = 0;
3781                 return;
3782         }
3783
3784         retval = ext2fs_extent_get_info(ehandle, &info);
3785         if (retval == 0) {
3786                 int max_depth = info.max_depth;
3787
3788                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
3789                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
3790                 ctx->extent_depth_count[max_depth]++;
3791         }
3792
3793         /* Check maximum extent depth */
3794         pctx->blk = info.max_depth;
3795         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
3796         if (pctx->blk2 < pctx->blk &&
3797             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
3798                 pb->eti.force_rebuild = 1;
3799
3800         /* Can we collect extent tree level stats? */
3801         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
3802         if (pctx->blk2 > pctx->blk)
3803                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
3804         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
3805         pb->eti.ino = pb->ino;
3806
3807         pb->next_lblock = 0;
3808
3809         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
3810                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
3811         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
3812         if (pctx->errcode &&
3813             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
3814                 pb->num_blocks = 0;
3815                 inode->i_blocks = 0;
3816                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3817                                    "check_blocks_extents");
3818                 pctx->errcode = 0;
3819         }
3820         ext2fs_extent_free(ehandle);
3821
3822         /* Rebuild unless it's a dir and we're rehashing it */
3823         if (LINUX_S_ISDIR(inode->i_mode) &&
3824             e2fsck_dir_will_be_rehashed(ctx, ino))
3825                 return;
3826
3827         if (ctx->options & E2F_OPT_CONVERT_BMAP)
3828                 e2fsck_rebuild_extents_later(ctx, ino);
3829         else
3830                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
3831 }
3832
3833 /*
3834  * In fact we don't need to check blocks for an inode with inline data
3835  * because this inode doesn't have any blocks.  In this function all
3836  * we need to do is add this inode into dblist when it is a directory.
3837  */
3838 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
3839                                      struct process_block_struct *pb)
3840 {
3841         int     flags;
3842         size_t  inline_data_size = 0;
3843
3844         if (!pb->is_dir) {
3845                 pctx->errcode = 0;
3846                 return;
3847         }
3848
3849         /* Process the dirents in i_block[] as the "first" block. */
3850         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
3851         if (pctx->errcode)
3852                 goto err;
3853
3854         /* Process the dirents in the EA as a "second" block. */
3855         flags = ctx->fs->flags;
3856         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3857         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
3858                                                 &inline_data_size);
3859         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3860                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3861         if (pctx->errcode) {
3862                 pctx->errcode = 0;
3863                 return;
3864         }
3865
3866         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
3867                 return;
3868
3869         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
3870         if (pctx->errcode)
3871                 goto err;
3872
3873         return;
3874 err:
3875         pctx->blk = 0;
3876         pctx->num = 0;
3877         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3878         ctx->flags |= E2F_FLAG_ABORT;
3879 }
3880
3881 /*
3882  * This subroutine is called on each inode to account for all of the
3883  * blocks used by that inode.
3884  */
3885 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
3886                          char *block_buf, const struct ea_quota *ea_ibody_quota)
3887 {
3888         ext2_filsys fs = ctx->fs;
3889         struct process_block_struct pb;
3890         ext2_ino_t      ino = pctx->ino;
3891         struct ext2_inode *inode = pctx->inode;
3892         unsigned        bad_size = 0;
3893         int             dirty_inode = 0;
3894         int             extent_fs;
3895         int             inlinedata_fs;
3896         __u64           size;
3897         struct ea_quota ea_block_quota;
3898
3899         pb.ino = ino;
3900         pb.num_blocks = EXT2FS_B2C(ctx->fs,
3901                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
3902         pb.last_block = ~0;
3903         pb.last_init_lblock = -1;
3904         pb.last_db_block = -1;
3905         pb.num_illegal_blocks = 0;
3906         pb.suppress = 0; pb.clear = 0;
3907         pb.fragmented = 0;
3908         pb.compressed = 0;
3909         pb.previous_block = 0;
3910         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
3911         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
3912         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
3913         pb.inode = inode;
3914         pb.pctx = pctx;
3915         pb.ctx = ctx;
3916         pb.inode_modified = 0;
3917         pb.eti.force_rebuild = 0;
3918         pctx->ino = ino;
3919         pctx->errcode = 0;
3920
3921         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
3922         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
3923
3924         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
3925                 if (e2fsck_should_abort(ctx))
3926                         goto out;
3927                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
3928         }
3929
3930         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
3931                 check_blocks_inline_data(ctx, pctx, &pb);
3932         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
3933                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
3934                         check_blocks_extents(ctx, pctx, &pb);
3935                 else {
3936                         int flags;
3937                         /*
3938                          * If we've modified the inode, write it out before
3939                          * iterate() tries to use it.
3940                          */
3941                         if (dirty_inode) {
3942                                 e2fsck_write_inode(ctx, ino, inode,
3943                                                    "check_blocks");
3944                                 dirty_inode = 0;
3945                         }
3946                         flags = fs->flags;
3947                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3948                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
3949                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
3950                                                 block_buf, process_block, &pb);
3951                         /*
3952                          * We do not have uninitialized extents in non extent
3953                          * files.
3954                          */
3955                         pb.last_init_lblock = pb.last_block;
3956                         /*
3957                          * If iterate() changed a block mapping, we have to
3958                          * re-read the inode.  If we decide to clear the
3959                          * inode after clearing some stuff, we'll re-write the
3960                          * bad mappings into the inode!
3961                          */
3962                         if (pb.inode_modified)
3963                                 e2fsck_read_inode(ctx, ino, inode,
3964                                                   "check_blocks");
3965                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3966                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3967
3968                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
3969 #ifdef DEBUG
3970                                 printf("bmap rebuild ino=%d\n", ino);
3971 #endif
3972                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
3973                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
3974                                         e2fsck_rebuild_extents_later(ctx, ino);
3975                         }
3976                 }
3977         }
3978         end_problem_latch(ctx, PR_LATCH_BLOCK);
3979         end_problem_latch(ctx, PR_LATCH_TOOBIG);
3980         if (e2fsck_should_abort(ctx))
3981                 goto out;
3982         if (pctx->errcode)
3983                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
3984
3985         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
3986                 if (LINUX_S_ISDIR(inode->i_mode))
3987                         ctx->fs_fragmented_dir++;
3988                 else
3989                         ctx->fs_fragmented++;
3990         }
3991
3992         if (pb.clear) {
3993                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3994                                    "check_blocks");
3995                 return;
3996         }
3997
3998         if (inode->i_flags & EXT2_INDEX_FL) {
3999                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
4000                         inode->i_flags &= ~EXT2_INDEX_FL;
4001                         dirty_inode++;
4002                 } else {
4003                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
4004                 }
4005         }
4006
4007         if (!pb.num_blocks && pb.is_dir &&
4008             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
4009                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
4010                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
4011                         ctx->fs_directory_count--;
4012                         return;
4013                 }
4014         }
4015
4016         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
4017             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
4018             !(inode->i_flags & EXT4_EA_INODE_FL)) {
4019                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
4020                                ino,
4021                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
4022                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
4023                                   ino, (ea_ibody_quota ?
4024                                         ea_ibody_quota->inodes : 0) +
4025                                                 ea_block_quota.inodes + 1);
4026         }
4027
4028         if (!ext2fs_has_feature_huge_file(fs->super) ||
4029             !(inode->i_flags & EXT4_HUGE_FILE_FL))
4030                 pb.num_blocks *= (fs->blocksize / 512);
4031         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
4032 #if 0
4033         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
4034                ino, inode->i_size, (unsigned long long) pb.last_block,
4035                (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
4036                (unsigned long long) pb.num_blocks);
4037 #endif
4038         size = EXT2_I_SIZE(inode);
4039         if (pb.is_dir) {
4040                 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
4041                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
4042                         int flags;
4043                         size_t sz = 0;
4044                         errcode_t err;
4045
4046                         flags = ctx->fs->flags;
4047                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4048                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
4049                                                       &sz);
4050                         ctx->fs->flags = (flags &
4051                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4052                                          (ctx->fs->flags &
4053                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4054                         if (err || sz != size) {
4055                                 bad_size = 7;
4056                                 pctx->num = sz;
4057                         }
4058                 } else if (size & (fs->blocksize - 1))
4059                         bad_size = 5;
4060                 else if (nblock > (pb.last_block + 1))
4061                         bad_size = 1;
4062                 else if (nblock < (pb.last_block + 1)) {
4063                         if (((pb.last_block + 1) - nblock) >
4064                             fs->super->s_prealloc_dir_blocks)
4065                                 bad_size = 2;
4066                 }
4067         } else {
4068                 if ((pb.last_init_lblock >= 0) &&
4069                     /* Do not allow initialized allocated blocks past i_size*/
4070                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
4071                     !(inode->i_flags & EXT4_VERITY_FL))
4072                         bad_size = 3;
4073                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4074                          size > ext2_max_sizes[fs->super->s_log_block_size])
4075                         /* too big for a direct/indirect-mapped file */
4076                         bad_size = 4;
4077                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4078                          size >
4079                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
4080                         /* too big for an extent-based file - 32bit ee_block */
4081                         bad_size = 6;
4082         }
4083         /* i_size for symlinks is checked elsewhere */
4084         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
4085                 /* Did inline_data set pctx->num earlier? */
4086                 if (bad_size != 7)
4087                         pctx->num = (pb.last_block + 1) * fs->blocksize;
4088                 pctx->group = bad_size;
4089                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
4090                         ext2fs_inode_size_set(fs, inode, pctx->num);
4091                         if (EXT2_I_SIZE(inode) == 0 &&
4092                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
4093                                 memset(inode->i_block, 0,
4094                                        sizeof(inode->i_block));
4095                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
4096                         }
4097                         dirty_inode++;
4098                 }
4099                 pctx->num = 0;
4100         }
4101         if (LINUX_S_ISREG(inode->i_mode) &&
4102             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4103                 ctx->large_files++;
4104         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
4105             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
4106              (ext2fs_has_feature_huge_file(fs->super) &&
4107               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
4108               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
4109                 pctx->num = pb.num_blocks;
4110                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
4111                         inode->i_blocks = pb.num_blocks;
4112                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
4113                         dirty_inode++;
4114                 }
4115                 pctx->num = 0;
4116         }
4117
4118         /*
4119          * The kernel gets mad if we ask it to allocate bigalloc clusters to
4120          * a block mapped file, so rebuild it as an extent file.  We can skip
4121          * symlinks because they're never rewritten.
4122          */
4123         if (ext2fs_has_feature_bigalloc(fs->super) &&
4124             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
4125             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
4126             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
4127             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
4128             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
4129                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
4130                 if (pctx->errcode)
4131                         goto out;
4132         }
4133
4134         if (ctx->dirs_to_hash && pb.is_dir &&
4135             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
4136             !(inode->i_flags & EXT2_INDEX_FL) &&
4137             ((inode->i_size / fs->blocksize) >= 3))
4138                 e2fsck_rehash_dir_later(ctx, ino);
4139
4140 out:
4141         if (dirty_inode)
4142                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
4143 }
4144
4145 #if 0
4146 /*
4147  * Helper function called by process block when an illegal block is
4148  * found.  It returns a description about why the block is illegal
4149  */
4150 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
4151 {
4152         blk64_t super;
4153         int     i;
4154         static char     problem[80];
4155
4156         super = fs->super->s_first_data_block;
4157         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
4158         if (block < super) {
4159                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
4160                 return(problem);
4161         } else if (block >= ext2fs_blocks_count(fs->super)) {
4162                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
4163                 return(problem);
4164         }
4165         for (i = 0; i < fs->group_desc_count; i++) {
4166                 if (block == super) {
4167                         sprintf(problem, "is the superblock in group %d", i);
4168                         break;
4169                 }
4170                 if (block > super &&
4171                     block <= (super + fs->desc_blocks)) {
4172                         sprintf(problem, "is in the group descriptors "
4173                                 "of group %d", i);
4174                         break;
4175                 }
4176                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
4177                         sprintf(problem, "is the block bitmap of group %d", i);
4178                         break;
4179                 }
4180                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
4181                         sprintf(problem, "is the inode bitmap of group %d", i);
4182                         break;
4183                 }
4184                 if (block >= ext2fs_inode_table_loc(fs, i) &&
4185                     (block < ext2fs_inode_table_loc(fs, i)
4186                      + fs->inode_blocks_per_group)) {
4187                         sprintf(problem, "is in the inode table of group %d",
4188                                 i);
4189                         break;
4190                 }
4191                 super += fs->super->s_blocks_per_group;
4192         }
4193         return(problem);
4194 }
4195 #endif
4196
4197 /*
4198  * This is a helper function for check_blocks().
4199  */
4200 static int process_block(ext2_filsys fs,
4201                   blk64_t       *block_nr,
4202                   e2_blkcnt_t blockcnt,
4203                   blk64_t ref_block EXT2FS_ATTR((unused)),
4204                   int ref_offset EXT2FS_ATTR((unused)),
4205                   void *priv_data)
4206 {
4207         struct process_block_struct *p;
4208         struct problem_context *pctx;
4209         blk64_t blk = *block_nr;
4210         int     ret_code = 0;
4211         problem_t       problem = 0;
4212         e2fsck_t        ctx;
4213
4214         p = (struct process_block_struct *) priv_data;
4215         pctx = p->pctx;
4216         ctx = p->ctx;
4217
4218         /*
4219          * For a directory, add logical block zero for processing even if it's
4220          * not mapped or we'll be perennially stuck with broken "." and ".."
4221          * entries.
4222          */
4223         if (p->is_dir && blockcnt == 0 && blk == 0) {
4224                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
4225                 if (pctx->errcode) {
4226                         pctx->blk = blk;
4227                         pctx->num = blockcnt;
4228                         goto failed_add_dir_block;
4229                 }
4230                 p->last_db_block++;
4231         }
4232
4233         if (blk == 0)
4234                 return 0;
4235
4236 #if 0
4237         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
4238                blockcnt);
4239 #endif
4240
4241         /*
4242          * Simplistic fragmentation check.  We merely require that the
4243          * file be contiguous.  (Which can never be true for really
4244          * big files that are greater than a block group.)
4245          */
4246         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
4247                 if (p->previous_block+1 != blk) {
4248                         if (ctx->options & E2F_OPT_FRAGCHECK) {
4249                                 char type = '?';
4250
4251                                 if (p->is_dir)
4252                                         type = 'd';
4253                                 else if (p->is_reg)
4254                                         type = 'f';
4255
4256                                 printf(_("%6lu(%c): expecting %6lu "
4257                                          "got phys %6lu (blkcnt %lld)\n"),
4258                                        (unsigned long) pctx->ino, type,
4259                                        (unsigned long) p->previous_block+1,
4260                                        (unsigned long) blk,
4261                                        (long long) blockcnt);
4262                         }
4263                         p->fragmented = 1;
4264                 }
4265         }
4266
4267         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
4268             !pctx->inode->i_size_high &&
4269             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
4270                 problem = PR_1_TOOBIG_DIR;
4271         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
4272                 problem = PR_1_TOOBIG_DIR;
4273         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
4274                 problem = PR_1_TOOBIG_REG;
4275         if (!p->is_dir && !p->is_reg && blockcnt > 0)
4276                 problem = PR_1_TOOBIG_SYMLINK;
4277
4278         if (blk < fs->super->s_first_data_block ||
4279             blk >= ext2fs_blocks_count(fs->super))
4280                 problem = PR_1_ILLEGAL_BLOCK_NUM;
4281
4282         /*
4283          * If this IND/DIND/TIND block is squatting atop some critical metadata
4284          * (group descriptors, superblock, bitmap, inode table), any write to
4285          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
4286          * fix that and restart fsck.
4287          */
4288         if (blockcnt < 0 &&
4289             p->ino != EXT2_RESIZE_INO &&
4290             blk < ctx->fs->super->s_blocks_count &&
4291             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
4292                 pctx->blk = blk;
4293                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
4294                 if ((ctx->options & E2F_OPT_NO) == 0)
4295                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4296         }
4297
4298         if (problem) {
4299                 p->num_illegal_blocks++;
4300                 /*
4301                  * A bit of subterfuge here -- we're trying to fix a block
4302                  * mapping, but the IND/DIND/TIND block could have collided
4303                  * with some critical metadata.  So, fix the in-core mapping so
4304                  * iterate won't go insane, but return 0 instead of
4305                  * BLOCK_CHANGED so that it won't write the remapping out to
4306                  * our multiply linked block.
4307                  *
4308                  * Even if we previously determined that an *IND block
4309                  * conflicts with critical metadata, we must still try to
4310                  * iterate the *IND block as if it is an *IND block to find and
4311                  * mark the blocks it points to.  Better to be overly cautious
4312                  * with the used_blocks map so that we don't move the *IND
4313                  * block to a block that's really in use!
4314                  */
4315                 if (p->ino != EXT2_RESIZE_INO &&
4316                     ref_block != 0 &&
4317                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4318                                               ref_block)) {
4319                         *block_nr = 0;
4320                         return 0;
4321                 }
4322                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
4323                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
4324                                 p->clear = 1;
4325                                 return BLOCK_ABORT;
4326                         }
4327                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
4328                                 p->suppress = 1;
4329                                 set_latch_flags(PR_LATCH_BLOCK,
4330                                                 PRL_SUPPRESS, 0);
4331                         }
4332                 }
4333                 pctx->blk = blk;
4334                 pctx->blkcount = blockcnt;
4335                 if (fix_problem(ctx, problem, pctx)) {
4336                         blk = *block_nr = 0;
4337                         ret_code = BLOCK_CHANGED;
4338                         p->inode_modified = 1;
4339                         /*
4340                          * If the directory block is too big and is beyond the
4341                          * end of the FS, don't bother trying to add it for
4342                          * processing -- the kernel would never have created a
4343                          * directory this large, and we risk an ENOMEM abort.
4344                          * In any case, the toobig handler for extent-based
4345                          * directories also doesn't feed toobig blocks to
4346                          * pass 2.
4347                          */
4348                         if (problem == PR_1_TOOBIG_DIR)
4349                                 return ret_code;
4350                         goto mark_dir;
4351                 } else
4352                         return 0;
4353         }
4354
4355         if (p->ino == EXT2_RESIZE_INO) {
4356                 /*
4357                  * The resize inode has already be sanity checked
4358                  * during pass #0 (the superblock checks).  All we
4359                  * have to do is mark the double indirect block as
4360                  * being in use; all of the other blocks are handled
4361                  * by mark_table_blocks()).
4362                  */
4363                 if (blockcnt == BLOCK_COUNT_DIND)
4364                         mark_block_used(ctx, blk);
4365                 p->num_blocks++;
4366         } else if (!(ctx->fs->cluster_ratio_bits &&
4367                      p->previous_block &&
4368                      (EXT2FS_B2C(ctx->fs, blk) ==
4369                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
4370                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
4371                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
4372                 mark_block_used(ctx, blk);
4373                 p->num_blocks++;
4374         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
4375                                              p->last_block, blk, blockcnt)) {
4376                 pctx->blk = blockcnt;
4377                 pctx->blk2 = blk;
4378                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4379                 mark_block_used(ctx, blk);
4380                 mark_block_used(ctx, blk);
4381         }
4382         if (blockcnt >= 0)
4383                 p->last_block = blockcnt;
4384         p->previous_block = blk;
4385 mark_dir:
4386         if (p->is_dir && (blockcnt >= 0)) {
4387                 while (++p->last_db_block < blockcnt) {
4388                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
4389                                                               p->ino, 0,
4390                                                               p->last_db_block);
4391                         if (pctx->errcode) {
4392                                 pctx->blk = 0;
4393                                 pctx->num = p->last_db_block;
4394                                 goto failed_add_dir_block;
4395                         }
4396                 }
4397                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
4398                                                       blk, blockcnt);
4399                 if (pctx->errcode) {
4400                         pctx->blk = blk;
4401                         pctx->num = blockcnt;
4402                 failed_add_dir_block:
4403                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4404                         /* Should never get here */
4405                         ctx->flags |= E2F_FLAG_ABORT;
4406                         return BLOCK_ABORT;
4407                 }
4408         }
4409         return ret_code;
4410 }
4411
4412 static int process_bad_block(ext2_filsys fs,
4413                       blk64_t *block_nr,
4414                       e2_blkcnt_t blockcnt,
4415                       blk64_t ref_block EXT2FS_ATTR((unused)),
4416                       int ref_offset EXT2FS_ATTR((unused)),
4417                       void *priv_data)
4418 {
4419         struct process_block_struct *p;
4420         blk64_t         blk = *block_nr;
4421         blk64_t         first_block;
4422         dgrp_t          i;
4423         struct problem_context *pctx;
4424         e2fsck_t        ctx;
4425
4426         if (!blk)
4427                 return 0;
4428
4429         p = (struct process_block_struct *) priv_data;
4430         ctx = p->ctx;
4431         pctx = p->pctx;
4432
4433         pctx->ino = EXT2_BAD_INO;
4434         pctx->blk = blk;
4435         pctx->blkcount = blockcnt;
4436
4437         if ((blk < fs->super->s_first_data_block) ||
4438             (blk >= ext2fs_blocks_count(fs->super))) {
4439                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
4440                         *block_nr = 0;
4441                         return BLOCK_CHANGED;
4442                 } else
4443                         return 0;
4444         }
4445
4446         if (blockcnt < 0) {
4447                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
4448                         p->bbcheck = 1;
4449                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
4450                                 *block_nr = 0;
4451                                 return BLOCK_CHANGED;
4452                         }
4453                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4454                                                     blk)) {
4455                         p->bbcheck = 1;
4456                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
4457                                         pctx)) {
4458                                 *block_nr = 0;
4459                                 return BLOCK_CHANGED;
4460                         }
4461                         if (e2fsck_should_abort(ctx))
4462                                 return BLOCK_ABORT;
4463                 } else
4464                         mark_block_used(ctx, blk);
4465                 return 0;
4466         }
4467 #if 0
4468         printf ("DEBUG: Marking %u as bad.\n", blk);
4469 #endif
4470         ctx->fs_badblocks_count++;
4471         /*
4472          * If the block is not used, then mark it as used and return.
4473          * If it is already marked as found, this must mean that
4474          * there's an overlap between the filesystem table blocks
4475          * (bitmaps and inode table) and the bad block list.
4476          */
4477         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
4478                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4479                 return 0;
4480         }
4481         /*
4482          * Try to find the where the filesystem block was used...
4483          */
4484         first_block = fs->super->s_first_data_block;
4485
4486         for (i = 0; i < fs->group_desc_count; i++ ) {
4487                 pctx->group = i;
4488                 pctx->blk = blk;
4489                 if (!ext2fs_bg_has_super(fs, i))
4490                         goto skip_super;
4491                 if (blk == first_block) {
4492                         if (i == 0) {
4493                                 if (fix_problem(ctx,
4494                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
4495                                                 pctx)) {
4496                                         *block_nr = 0;
4497                                         return BLOCK_CHANGED;
4498                                 }
4499                                 return 0;
4500                         }
4501                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
4502                         return 0;
4503                 }
4504                 if ((blk > first_block) &&
4505                     (blk <= first_block + fs->desc_blocks)) {
4506                         if (i == 0) {
4507                                 pctx->blk = *block_nr;
4508                                 if (fix_problem(ctx,
4509                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
4510                                         *block_nr = 0;
4511                                         return BLOCK_CHANGED;
4512                                 }
4513                                 return 0;
4514                         }
4515                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
4516                         return 0;
4517                 }
4518         skip_super:
4519                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
4520                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
4521                                 ctx->invalid_block_bitmap_flag[i]++;
4522                                 ctx->invalid_bitmaps++;
4523                         }
4524                         return 0;
4525                 }
4526                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
4527                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
4528                                 ctx->invalid_inode_bitmap_flag[i]++;
4529                                 ctx->invalid_bitmaps++;
4530                         }
4531                         return 0;
4532                 }
4533                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
4534                     (blk < (ext2fs_inode_table_loc(fs, i) +
4535                             fs->inode_blocks_per_group))) {
4536                         /*
4537                          * If there are bad blocks in the inode table,
4538                          * the inode scan code will try to do
4539                          * something reasonable automatically.
4540                          */
4541                         return 0;
4542                 }
4543                 first_block += fs->super->s_blocks_per_group;
4544         }
4545         /*
4546          * If we've gotten to this point, then the only
4547          * possibility is that the bad block inode meta data
4548          * is using a bad block.
4549          */
4550         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
4551             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
4552             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
4553                 p->bbcheck = 1;
4554                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
4555                         *block_nr = 0;
4556                         return BLOCK_CHANGED;
4557                 }
4558                 if (e2fsck_should_abort(ctx))
4559                         return BLOCK_ABORT;
4560                 return 0;
4561         }
4562
4563         pctx->group = -1;
4564
4565         /* Warn user that the block wasn't claimed */
4566         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
4567
4568         return 0;
4569 }
4570
4571 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
4572                             const char *name, int num, blk64_t *new_block)
4573 {
4574         ext2_filsys fs = ctx->fs;
4575         dgrp_t          last_grp;
4576         blk64_t         old_block = *new_block;
4577         blk64_t         last_block;
4578         dgrp_t          flexbg;
4579         unsigned        flexbg_size;
4580         int             i, is_flexbg;
4581         char            *buf;
4582         struct problem_context  pctx;
4583
4584         clear_problem_context(&pctx);
4585
4586         pctx.group = group;
4587         pctx.blk = old_block;
4588         pctx.str = name;
4589
4590         /*
4591          * For flex_bg filesystems, first try to allocate the metadata
4592          * within the flex_bg, and if that fails then try finding the
4593          * space anywhere in the filesystem.
4594          */
4595         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
4596         if (is_flexbg) {
4597                 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
4598                 flexbg = group / flexbg_size;
4599                 first_block = ext2fs_group_first_block2(fs,
4600                                                         flexbg_size * flexbg);
4601                 last_grp = group | (flexbg_size - 1);
4602                 if (last_grp >= fs->group_desc_count)
4603                         last_grp = fs->group_desc_count - 1;
4604                 last_block = ext2fs_group_last_block2(fs, last_grp);
4605         } else
4606                 last_block = ext2fs_group_last_block2(fs, group);
4607         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
4608                                                num, ctx->block_found_map,
4609                                                new_block);
4610         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
4611                 pctx.errcode = ext2fs_get_free_blocks2(fs,
4612                                 fs->super->s_first_data_block,
4613                                 ext2fs_blocks_count(fs->super),
4614                                 num, ctx->block_found_map, new_block);
4615         if (pctx.errcode) {
4616                 pctx.num = num;
4617                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4618                 ext2fs_unmark_valid(fs);
4619                 ctx->flags |= E2F_FLAG_ABORT;
4620                 return;
4621         }
4622         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4623         if (pctx.errcode) {
4624                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4625                 ext2fs_unmark_valid(fs);
4626                 ctx->flags |= E2F_FLAG_ABORT;
4627                 return;
4628         }
4629         ext2fs_mark_super_dirty(fs);
4630         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4631         pctx.blk2 = *new_block;
4632         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4633                           PR_1_RELOC_TO), &pctx);
4634         pctx.blk2 = 0;
4635         for (i = 0; i < num; i++) {
4636                 pctx.blk = i;
4637                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4638                 if (old_block) {
4639                         pctx.errcode = io_channel_read_blk64(fs->io,
4640                                    old_block + i, 1, buf);
4641                         if (pctx.errcode)
4642                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4643                         pctx.blk = (*new_block) + i;
4644                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4645                                                               1, buf);
4646                 } else {
4647                         pctx.blk = (*new_block) + i;
4648                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4649                                                            NULL, NULL);
4650                 }
4651
4652                 if (pctx.errcode)
4653                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4654         }
4655         ext2fs_free_mem(&buf);
4656 }
4657
4658 /*
4659  * This routine gets called at the end of pass 1 if bad blocks are
4660  * detected in the superblock, group descriptors, inode_bitmaps, or
4661  * block bitmaps.  At this point, all of the blocks have been mapped
4662  * out, so we can try to allocate new block(s) to replace the bad
4663  * blocks.
4664  */
4665 static void handle_fs_bad_blocks(e2fsck_t ctx)
4666 {
4667         ext2_filsys fs = ctx->fs;
4668         dgrp_t          i;
4669         blk64_t         first_block;
4670         blk64_t         new_blk;
4671
4672         for (i = 0; i < fs->group_desc_count; i++) {
4673                 first_block = ext2fs_group_first_block2(fs, i);
4674
4675                 if (ctx->invalid_block_bitmap_flag[i]) {
4676                         new_blk = ext2fs_block_bitmap_loc(fs, i);
4677                         new_table_block(ctx, first_block, i, _("block bitmap"),
4678                                         1, &new_blk);
4679                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4680                 }
4681                 if (ctx->invalid_inode_bitmap_flag[i]) {
4682                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
4683                         new_table_block(ctx, first_block, i, _("inode bitmap"),
4684                                         1, &new_blk);
4685                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4686                 }
4687                 if (ctx->invalid_inode_table_flag[i]) {
4688                         new_blk = ext2fs_inode_table_loc(fs, i);
4689                         new_table_block(ctx, first_block, i, _("inode table"),
4690                                         fs->inode_blocks_per_group,
4691                                         &new_blk);
4692                         ext2fs_inode_table_loc_set(fs, i, new_blk);
4693                         ctx->flags |= E2F_FLAG_RESTART;
4694                 }
4695         }
4696         ctx->invalid_bitmaps = 0;
4697 }
4698
4699 /*
4700  * This routine marks all blocks which are used by the superblock,
4701  * group descriptors, inode bitmaps, and block bitmaps.
4702  */
4703 static void mark_table_blocks(e2fsck_t ctx)
4704 {
4705         ext2_filsys fs = ctx->fs;
4706         blk64_t b;
4707         dgrp_t  i;
4708         unsigned int    j;
4709         struct problem_context pctx;
4710
4711         clear_problem_context(&pctx);
4712
4713         for (i = 0; i < fs->group_desc_count; i++) {
4714                 pctx.group = i;
4715
4716                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
4717                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
4718
4719                 /*
4720                  * Mark the blocks used for the inode table
4721                  */
4722                 if (ext2fs_inode_table_loc(fs, i)) {
4723                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
4724                              j < fs->inode_blocks_per_group;
4725                              j++, b++) {
4726                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4727                                                              b)) {
4728                                         pctx.blk = b;
4729                                         if (!ctx->invalid_inode_table_flag[i] &&
4730                                             fix_problem(ctx,
4731                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
4732                                                 ctx->invalid_inode_table_flag[i]++;
4733                                                 ctx->invalid_bitmaps++;
4734                                         }
4735                                 } else {
4736                                     ext2fs_mark_block_bitmap2(
4737                                                 ctx->block_found_map, b);
4738                                     ext2fs_mark_block_bitmap2(
4739                                                 ctx->block_metadata_map, b);
4740                                 }
4741                         }
4742                 }
4743
4744                 /*
4745                  * Mark block used for the block bitmap
4746                  */
4747                 if (ext2fs_block_bitmap_loc(fs, i)) {
4748                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4749                                      ext2fs_block_bitmap_loc(fs, i))) {
4750                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
4751                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
4752                                         ctx->invalid_block_bitmap_flag[i]++;
4753                                         ctx->invalid_bitmaps++;
4754                                 }
4755                         } else {
4756                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4757                                      ext2fs_block_bitmap_loc(fs, i));
4758                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4759                                      ext2fs_block_bitmap_loc(fs, i));
4760                         }
4761                 }
4762                 /*
4763                  * Mark block used for the inode bitmap
4764                  */
4765                 if (ext2fs_inode_bitmap_loc(fs, i)) {
4766                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4767                                      ext2fs_inode_bitmap_loc(fs, i))) {
4768                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
4769                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
4770                                         ctx->invalid_inode_bitmap_flag[i]++;
4771                                         ctx->invalid_bitmaps++;
4772                                 }
4773                         } else {
4774                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4775                                      ext2fs_inode_bitmap_loc(fs, i));
4776                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4777                                      ext2fs_inode_bitmap_loc(fs, i));
4778                         }
4779                 }
4780         }
4781 }
4782
4783 /*
4784  * These subroutines short circuits ext2fs_get_blocks and
4785  * ext2fs_check_directory; we use them since we already have the inode
4786  * structure, so there's no point in letting the ext2fs library read
4787  * the inode again.
4788  */
4789 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
4790                                   blk_t *blocks)
4791 {
4792         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4793         int     i;
4794
4795         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4796                 return EXT2_ET_CALLBACK_NOTHANDLED;
4797
4798         for (i=0; i < EXT2_N_BLOCKS; i++)
4799                 blocks[i] = ctx->stashed_inode->i_block[i];
4800         return 0;
4801 }
4802
4803 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
4804                                   struct ext2_inode *inode)
4805 {
4806         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4807
4808         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4809                 return EXT2_ET_CALLBACK_NOTHANDLED;
4810         *inode = *ctx->stashed_inode;
4811         return 0;
4812 }
4813
4814 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
4815                             struct ext2_inode *inode)
4816 {
4817         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4818
4819         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
4820                 (inode != ctx->stashed_inode))
4821                 *ctx->stashed_inode = *inode;
4822         return EXT2_ET_CALLBACK_NOTHANDLED;
4823 }
4824
4825 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
4826 {
4827         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4828
4829         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4830                 return EXT2_ET_CALLBACK_NOTHANDLED;
4831
4832         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
4833                 return EXT2_ET_NO_DIRECTORY;
4834         return 0;
4835 }
4836
4837 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
4838                                         blk64_t *ret)
4839 {
4840         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4841         errcode_t       retval;
4842         blk64_t         new_block;
4843
4844         if (ctx->block_found_map) {
4845                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
4846                                            &new_block);
4847                 if (retval)
4848                         return retval;
4849                 if (fs->block_map) {
4850                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
4851                         ext2fs_mark_bb_dirty(fs);
4852                 }
4853         } else {
4854                 if (!fs->block_map) {
4855                         retval = ext2fs_read_block_bitmap(fs);
4856                         if (retval)
4857                                 return retval;
4858                 }
4859
4860                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
4861                 if (retval)
4862                         return retval;
4863         }
4864
4865         *ret = new_block;
4866         return (0);
4867 }
4868
4869 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
4870                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
4871 {
4872         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4873         errcode_t       retval;
4874
4875         if (ctx->block_found_map)
4876                 return ext2fs_new_range(fs, flags, goal, len,
4877                                         ctx->block_found_map, pblk, plen);
4878
4879         if (!fs->block_map) {
4880                 retval = ext2fs_read_block_bitmap(fs);
4881                 if (retval)
4882                         return retval;
4883         }
4884
4885         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
4886                                 pblk, plen);
4887 }
4888
4889 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
4890 {
4891         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4892
4893         /* Never free a critical metadata block */
4894         if (ctx->block_found_map &&
4895             ctx->block_metadata_map &&
4896             inuse < 0 &&
4897             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
4898                 return;
4899
4900         if (ctx->block_found_map) {
4901                 if (inuse > 0)
4902                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4903                 else
4904                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
4905         }
4906 }
4907
4908 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
4909                                            blk_t num, int inuse)
4910 {
4911         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4912
4913         /* Never free a critical metadata block */
4914         if (ctx->block_found_map &&
4915             ctx->block_metadata_map &&
4916             inuse < 0 &&
4917             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
4918                 return;
4919
4920         if (ctx->block_found_map) {
4921                 if (inuse > 0)
4922                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
4923                                                         blk, num);
4924                 else
4925                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
4926                                                         blk, num);
4927         }
4928 }
4929
4930 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
4931 {
4932         ext2_filsys fs = ctx->fs;
4933
4934         if (use_shortcuts) {
4935                 fs->get_blocks = pass1_get_blocks;
4936                 fs->check_directory = pass1_check_directory;
4937                 fs->read_inode = pass1_read_inode;
4938                 fs->write_inode = pass1_write_inode;
4939                 ctx->stashed_ino = 0;
4940         } else {
4941                 fs->get_blocks = 0;
4942                 fs->check_directory = 0;
4943                 fs->read_inode = 0;
4944                 fs->write_inode = 0;
4945         }
4946 }
4947
4948 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
4949 {
4950         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
4951         ext2fs_set_block_alloc_stats_callback(ctx->fs,
4952                                                 e2fsck_block_alloc_stats, 0);
4953         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
4954         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
4955                                         e2fsck_block_alloc_stats_range, NULL);
4956 }