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
  52 #include "e2fsck.h"
  53 #include <ext2fs/ext2_ext_attr.h>
  54 /* todo remove this finally */
  55 #include <ext2fs/ext2fsP.h>
  56 #include <e2p/e2p.h>
  57
  58 #include "problem.h"
  59
  60 #ifdef NO_INLINE_FUNCS
  61 #define _INLINE_
  62 #else
  63 #define _INLINE_ inline
  64 #endif
  65
  66 #undef DEBUG
  67
  68 struct ea_quota {
  69         blk64_t blocks;
  70         __u64 inodes;
  71 };
  72
  73 static int process_block(ext2_filsys fs, blk64_t        *blocknr,
  74                          e2_blkcnt_t blockcnt, blk64_t ref_blk,
  75                          int ref_offset, void *priv_data);
  76 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
  77                              e2_blkcnt_t blockcnt, blk64_t ref_blk,
  78                              int ref_offset, void *priv_data);
  79 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
  80                          char *block_buf,
  81                          const struct ea_quota *ea_ibody_quota);
  82 static void mark_table_blocks(e2fsck_t ctx);
  83 static void alloc_bb_map(e2fsck_t ctx);
  84 static void alloc_imagic_map(e2fsck_t ctx);
  85 static void mark_inode_bad(e2fsck_t ctx, ino_t ino);
  86 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino);
  87 static void handle_fs_bad_blocks(e2fsck_t ctx);
  88 static void process_inodes(e2fsck_t ctx, char *block_buf);
  89 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
  90 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
  91                                   dgrp_t group, void * priv_data);
  92 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
  93                                     char *block_buf, int adjust_sign);
  94 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
  95
  96 struct process_block_struct {
  97         ext2_ino_t      ino;
  98         unsigned        is_dir:1, is_reg:1, clear:1, suppress:1,
  99                                 fragmented:1, compressed:1, bbcheck:1,
 100                                 inode_modified:1;
 101         blk64_t         num_blocks;
 102         blk64_t         max_blocks;
 103         blk64_t         last_block;
 104         e2_blkcnt_t     last_init_lblock;
 105         e2_blkcnt_t     last_db_block;
 106         int             num_illegal_blocks;
 107         blk64_t         previous_block;
 108         struct ext2_inode *inode;
 109         struct problem_context *pctx;
 110         ext2fs_block_bitmap fs_meta_blocks;
 111         e2fsck_t        ctx;
 112         blk64_t         next_lblock;
 113         struct extent_tree_info eti;
 114 };
 115
 116 struct process_inode_block {
 117         ext2_ino_t ino;
 118         struct ea_quota ea_ibody_quota;
 119         struct ext2_inode_large inode;
 120 };
 121
 122 struct scan_callback_struct {
 123         e2fsck_t        ctx;
 124         char            *block_buf;
 125 };
 126
 127 /*
 128  * For the inodes to process list.
 129  */
 130 static struct process_inode_block *inodes_to_process;
 131 static int process_inode_count;
 132
 133 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
 134                             EXT2_MIN_BLOCK_LOG_SIZE + 1];
 135
 136 /*
 137  * Check to make sure a device inode is real.  Returns 1 if the device
 138  * checks out, 0 if not.
 139  *
 140  * Note: this routine is now also used to check FIFO's and Sockets,
 141  * since they have the same requirement; the i_block fields should be
 142  * zero.
 143  */
 144 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
 145                                     struct ext2_inode *inode)
 146 {
 147         int     i;
 148
 149         /*
 150          * If the index or extents flag is set, then this is a bogus
 151          * device/fifo/socket
 152          */
 153         if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
 154                 return 0;
 155
 156         /*
 157          * We should be able to do the test below all the time, but
 158          * because the kernel doesn't forcibly clear the device
 159          * inode's additional i_block fields, there are some rare
 160          * occasions when a legitimate device inode will have non-zero
 161          * additional i_block fields.  So for now, we only complain
 162          * when the immutable flag is set, which should never happen
 163          * for devices.  (And that's when the problem is caused, since
 164          * you can't set or clear immutable flags for devices.)  Once
 165          * the kernel has been fixed we can change this...
 166          */
 167         if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
 168                 for (i=4; i < EXT2_N_BLOCKS; i++)
 169                         if (inode->i_block[i])
 170                                 return 0;
 171         }
 172         return 1;
 173 }
 174
 175 /*
 176  * Check to make sure a symlink inode is real.  Returns 1 if the symlink
 177  * checks out, 0 if not.
 178  */
 179 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
 180                                struct ext2_inode *inode, char *buf)
 181 {
 182         unsigned int buflen;
 183         unsigned int len;
 184
 185         if ((inode->i_size_high || inode->i_size == 0) ||
 186             (inode->i_flags & EXT2_INDEX_FL))
 187                 return 0;
 188
 189         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
 190                 size_t inline_size;
 191
 192                 if (inode->i_flags & EXT4_EXTENTS_FL)
 193                         return 0;
 194                 if (ext2fs_inline_data_size(fs, ino, &inline_size))
 195                         return 0;
 196                 if (inode->i_size != inline_size)
 197                         return 0;
 198
 199                 return 1;
 200         }
 201
 202         if (ext2fs_is_fast_symlink(inode)) {
 203                 if (inode->i_flags & EXT4_EXTENTS_FL)
 204                         return 0;
 205                 buf = (char *)inode->i_block;
 206                 buflen = sizeof(inode->i_block);
 207         } else {
 208                 ext2_extent_handle_t    handle;
 209                 struct ext2_extent_info info;
 210                 struct ext2fs_extent    extent;
 211                 blk64_t blk;
 212                 int i;
 213
 214                 if (inode->i_flags & EXT4_EXTENTS_FL) {
 215                         if (ext2fs_extent_open2(fs, ino, inode, &handle))
 216                                 return 0;
 217                         if (ext2fs_extent_get_info(handle, &info) ||
 218                             (info.num_entries != 1) ||
 219                             (info.max_depth != 0)) {
 220                                 ext2fs_extent_free(handle);
 221                                 return 0;
 222                         }
 223                         if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
 224                                               &extent) ||
 225                             (extent.e_lblk != 0) ||
 226                             (extent.e_len != 1)) {
 227                                 ext2fs_extent_free(handle);
 228                                 return 0;
 229                         }
 230                         blk = extent.e_pblk;
 231                         ext2fs_extent_free(handle);
 232                 } else {
 233                         blk = inode->i_block[0];
 234
 235                         for (i = 1; i < EXT2_N_BLOCKS; i++)
 236                                 if (inode->i_block[i])
 237                                         return 0;
 238                 }
 239
 240                 if (blk < fs->super->s_first_data_block ||
 241                     blk >= ext2fs_blocks_count(fs->super))
 242                         return 0;
 243
 244                 if (io_channel_read_blk64(fs->io, blk, 1, buf))
 245                         return 0;
 246
 247                 buflen = fs->blocksize;
 248         }
 249
 250         if (inode->i_flags & EXT4_ENCRYPT_FL)
 251                 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
 252         else
 253                 len = strnlen(buf, buflen);
 254
 255         if (len >= buflen)
 256                 return 0;
 257
 258         if (len != inode->i_size)
 259                 return 0;
 260         return 1;
 261 }
 262
 263 /*
 264  * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
 265  */
 266 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
 267 static void check_extents_inlinedata(e2fsck_t ctx,
 268                                      struct problem_context *pctx)
 269 {
 270         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 271                 return;
 272
 273         if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
 274                 return;
 275
 276         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 277         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 278 }
 279 #undef BAD_SPECIAL_FLAGS
 280
 281 /*
 282  * If the immutable (or append-only) flag is set on the inode, offer
 283  * to clear it.
 284  */
 285 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
 286 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
 287 {
 288         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 289                 return;
 290
 291         if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
 292                 return;
 293
 294         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 295         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 296 }
 297
 298 /*
 299  * If device, fifo or socket, check size is zero -- if not offer to
 300  * clear it
 301  */
 302 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
 303 {
 304         struct ext2_inode *inode = pctx->inode;
 305
 306         if (EXT2_I_SIZE(inode) == 0)
 307                 return;
 308
 309         if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
 310                 return;
 311
 312         ext2fs_inode_size_set(ctx->fs, inode, 0);
 313         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 314 }
 315
 316 /*
 317  * For a given size, calculate how many blocks would be charged towards quota.
 318  */
 319 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
 320 {
 321         blk64_t clusters;
 322
 323         clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
 324         return EXT2FS_C2B(fs, clusters);
 325 }
 326
 327 /*
 328  * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
 329  * the problem code.
 330  */
 331 static problem_t check_large_ea_inode(e2fsck_t ctx,
 332                                       struct ext2_ext_attr_entry *entry,
 333                                       struct problem_context *pctx,
 334                                       blk64_t *quota_blocks)
 335 {
 336         struct ext2_inode inode;
 337         __u32 hash;
 338         errcode_t retval;
 339
 340         /* Check if inode is within valid range */
 341         if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
 342             (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
 343                 pctx->num = entry->e_value_inum;
 344                 return PR_1_ATTR_VALUE_EA_INODE;
 345         }
 346
 347         e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
 348
 349         retval = ext2fs_ext_attr_hash_entry2(ctx->fs, entry, NULL, &hash);
 350         if (retval) {
 351                 com_err("check_large_ea_inode", retval,
 352                         _("while hashing entry with e_value_inum = %u"),
 353                         entry->e_value_inum);
 354                 fatal_error(ctx, 0);
 355         }
 356
 357         if (hash == entry->e_hash) {
 358                 *quota_blocks = size_to_quota_blocks(ctx->fs,
 359                                                      entry->e_value_size);
 360         } else {
 361                 /* This might be an old Lustre-style ea_inode reference. */
 362                 if (inode.i_mtime == pctx->ino &&
 363                     inode.i_generation == pctx->inode->i_generation) {
 364                         *quota_blocks = 0;
 365                 } else {
 366                         /* If target inode is also missing EA_INODE flag,
 367                          * this is likely to be a bad reference.
 368                          */
 369                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 370                                 pctx->num = entry->e_value_inum;
 371                                 return PR_1_ATTR_VALUE_EA_INODE;
 372                         } else {
 373                                 pctx->num = entry->e_hash;
 374                                 return PR_1_ATTR_HASH;
 375                         }
 376                 }
 377         }
 378
 379         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 380                 pctx->num = entry->e_value_inum;
 381                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
 382                         inode.i_flags |= EXT4_EA_INODE_FL;
 383                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
 384                                            &inode);
 385                 } else {
 386                         return PR_1_ATTR_NO_EA_INODE_FL;
 387                 }
 388         }
 389         return 0;
 390 }
 391
 392 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
 393                               struct ext2_ext_attr_entry *first, void *end)
 394 {
 395         struct ext2_ext_attr_entry *entry;
 396
 397         for (entry = first;
 398              (void *)entry < end && !EXT2_EXT_IS_LAST_ENTRY(entry);
 399              entry = EXT2_EXT_ATTR_NEXT(entry)) {
 400                 if (!entry->e_value_inum)
 401                         continue;
 402                 if (!ctx->ea_inode_refs) {
 403                         pctx->errcode = ea_refcount_create(0,
 404                                                            &ctx->ea_inode_refs);
 405                         if (pctx->errcode) {
 406                                 pctx->num = 4;
 407                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
 408                                 ctx->flags |= E2F_FLAG_ABORT;
 409                                 return;
 410                         }
 411                 }
 412                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
 413                                       0);
 414         }
 415 }
 416
 417 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
 418                               struct ea_quota *ea_ibody_quota)
 419 {
 420         struct ext2_super_block *sb = ctx->fs->super;
 421         struct ext2_inode_large *inode;
 422         struct ext2_ext_attr_entry *entry;
 423         char *start, *header, *end;
 424         unsigned int storage_size, remain;
 425         problem_t problem = 0;
 426         region_t region = 0;
 427
 428         ea_ibody_quota->blocks = 0;
 429         ea_ibody_quota->inodes = 0;
 430
 431         inode = (struct ext2_inode_large *) pctx->inode;
 432         storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
 433                 inode->i_extra_isize;
 434         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 435                  inode->i_extra_isize;
 436         end = header + storage_size;
 437         start = header + sizeof(__u32);
 438         entry = (struct ext2_ext_attr_entry *) start;
 439
 440         /* scan all entry's headers first */
 441
 442         /* take finish entry 0UL into account */
 443         remain = storage_size - sizeof(__u32);
 444
 445         region = region_create(0, storage_size);
 446         if (!region) {
 447                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
 448                 problem = 0;
 449                 ctx->flags |= E2F_FLAG_ABORT;
 450                 return;
 451         }
 452         if (region_allocate(region, 0, sizeof(__u32))) {
 453                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 454                 goto fix;
 455         }
 456
 457         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
 458                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
 459                 __u32 hash;
 460
 461                 if (region_allocate(region, (char *)entry - (char *)header,
 462                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
 463                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
 464                         goto fix;
 465                 }
 466
 467                 /* header eats this space */
 468                 remain -= sizeof(struct ext2_ext_attr_entry);
 469
 470                 /* is attribute name valid? */
 471                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
 472                         pctx->num = entry->e_name_len;
 473                         problem = PR_1_ATTR_NAME_LEN;
 474                         goto fix;
 475                 }
 476
 477                 /* attribute len eats this space */
 478                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
 479
 480                 if (entry->e_value_inum == 0) {
 481                         /* check value size */
 482                         if (entry->e_value_size > remain) {
 483                                 pctx->num = entry->e_value_size;
 484                                 problem = PR_1_ATTR_VALUE_SIZE;
 485                                 goto fix;
 486                         }
 487
 488                         if (entry->e_value_size &&
 489                             region_allocate(region,
 490                                             sizeof(__u32) + entry->e_value_offs,
 491                                             EXT2_EXT_ATTR_SIZE(
 492                                                 entry->e_value_size))) {
 493                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 494                                 goto fix;
 495                         }
 496
 497                         hash = ext2fs_ext_attr_hash_entry(entry,
 498                                                           start + entry->e_value_offs);
 499
 500                         /* e_hash may be 0 in older inode's ea */
 501                         if (entry->e_hash != 0 && entry->e_hash != hash) {
 502                                 pctx->num = entry->e_hash;
 503                                 problem = PR_1_ATTR_HASH;
 504                                 goto fix;
 505                         }
 506                 } else {
 507                         blk64_t quota_blocks;
 508
 509                         problem = check_large_ea_inode(ctx, entry, pctx,
 510                                                        &quota_blocks);
 511                         if (problem != 0)
 512                                 goto fix;
 513
 514                         ea_ibody_quota->blocks += quota_blocks;
 515                         ea_ibody_quota->inodes++;
 516                 }
 517
 518                 /* If EA value is stored in external inode then it does not
 519                  * consume space here */
 520                 if (entry->e_value_inum == 0)
 521                         remain -= entry->e_value_size;
 522
 523                 entry = EXT2_EXT_ATTR_NEXT(entry);
 524         }
 525
 526         if (region_allocate(region, (char *)entry - (char *)header,
 527                             sizeof(__u32))) {
 528                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 529                 goto fix;
 530         }
 531 fix:
 532         if (region)
 533                 region_free(region);
 534         /*
 535          * it seems like a corruption. it's very unlikely we could repair
 536          * EA(s) in automatic fashion -bzzz
 537          */
 538         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
 539                 inc_ea_inode_refs(ctx, pctx,
 540                                   (struct ext2_ext_attr_entry *)start, end);
 541                 return;
 542         }
 543
 544         /* simply remove all possible EA(s) */
 545         *((__u32 *)header) = 0UL;
 546         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 547                                 EXT2_INODE_SIZE(sb), "pass1");
 548         ea_ibody_quota->blocks = 0;
 549         ea_ibody_quota->inodes = 0;
 550 }
 551
 552 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
 553         return (xtime & (1U << 31)) != 0 &&
 554                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
 555 }
 556
 557 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
 558         check_inode_extra_negative_epoch(inode->i_##xtime, \
 559                                          inode->i_##xtime##_extra)
 560
 561 /* When today's date is earlier than 2242, we assume that atimes,
 562  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
 563  * actually pre-1970 dates mis-encoded.
 564  */
 565 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
 566
 567 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
 568                                     struct ea_quota *ea_ibody_quota)
 569 {
 570         struct ext2_super_block *sb = ctx->fs->super;
 571         struct ext2_inode_large *inode;
 572         __u32 *eamagic;
 573         int min, max;
 574
 575         ea_ibody_quota->blocks = 0;
 576         ea_ibody_quota->inodes = 0;
 577
 578         inode = (struct ext2_inode_large *) pctx->inode;
 579         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
 580                 /* this isn't large inode. so, nothing to check */
 581                 return;
 582         }
 583
 584 #if 0
 585         printf("inode #%u, i_extra_size %d\n", pctx->ino,
 586                         inode->i_extra_isize);
 587 #endif
 588         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
 589         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
 590         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
 591         /*
 592          * For now we will allow i_extra_isize to be 0, but really
 593          * implementations should never allow i_extra_isize to be 0
 594          */
 595         if (inode->i_extra_isize &&
 596             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
 597              inode->i_extra_isize & 3)) {
 598                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
 599                         return;
 600                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
 601                         inode->i_extra_isize = sb->s_want_extra_isize;
 602                 else
 603                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
 604                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 605                                         EXT2_INODE_SIZE(sb), "pass1");
 606         }
 607
 608         /* check if there is no place for an EA header */
 609         if (inode->i_extra_isize >= max - sizeof(__u32))
 610                 return;
 611
 612         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 613                         inode->i_extra_isize);
 614         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
 615                 /* it seems inode has an extended attribute(s) in body */
 616                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
 617         }
 618
 619         /*
 620          * If the inode's extended atime (ctime, crtime, mtime) is stored in
 621          * the old, invalid format, repair it.
 622          */
 623         if (((sizeof(time_t) <= 4) ||
 624              (((sizeof(time_t) > 4) &&
 625                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
 626             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
 627              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
 628              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
 629              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
 630
 631                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
 632                         return;
 633
 634                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
 635                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
 636                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
 637                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
 638                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
 639                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
 640                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
 641                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
 642                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 643                                         EXT2_INODE_SIZE(sb), "pass1");
 644         }
 645
 646 }
 647
 648 /*
 649  * Check to see if the inode might really be a directory, despite i_mode
 650  *
 651  * This is a lot of complexity for something for which I'm not really
 652  * convinced happens frequently in the wild.  If for any reason this
 653  * causes any problems, take this code out.
 654  * [tytso:20070331.0827EDT]
 655  */
 656 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
 657                                 char *buf)
 658 {
 659         struct ext2_inode *inode = pctx->inode;
 660         struct ext2_dir_entry   *dirent;
 661         errcode_t               retval;
 662         blk64_t                 blk;
 663         unsigned int            i, rec_len, not_device = 0;
 664         int                     extent_fs;
 665         int                     inlinedata_fs;
 666
 667         /*
 668          * If the mode looks OK, we believe it.  If the first block in
 669          * the i_block array is 0, this cannot be a directory. If the
 670          * inode is extent-mapped, it is still the case that the latter
 671          * cannot be 0 - the magic number in the extent header would make
 672          * it nonzero.
 673          */
 674         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
 675             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
 676                 return;
 677
 678         /*
 679          * Check the block numbers in the i_block array for validity:
 680          * zero blocks are skipped (but the first one cannot be zero -
 681          * see above), other blocks are checked against the first and
 682          * max data blocks (from the the superblock) and against the
 683          * block bitmap. Any invalid block found means this cannot be
 684          * a directory.
 685          *
 686          * If there are non-zero blocks past the fourth entry, then
 687          * this cannot be a device file: we remember that for the next
 688          * check.
 689          *
 690          * For extent mapped files, we don't do any sanity checking:
 691          * just try to get the phys block of logical block 0 and run
 692          * with it.
 693          *
 694          * For inline data files, we just try to get the size of inline
 695          * data.  If it's true, we will treat it as a directory.
 696          */
 697
 698         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
 699         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
 700         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
 701                 size_t size;
 702                 __u32 dotdot;
 703                 unsigned int rec_len2;
 704                 struct ext2_dir_entry de;
 705
 706                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
 707                         return;
 708                 /*
 709                  * If the size isn't a multiple of 4, it's probably not a
 710                  * directory??
 711                  */
 712                 if (size & 3)
 713                         return;
 714                 /*
 715                  * If the first 10 bytes don't look like a directory entry,
 716                  * it's probably not a directory.
 717                  */
 718                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
 719                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
 720                        EXT2_DIR_REC_LEN(0));
 721                 dotdot = ext2fs_le32_to_cpu(dotdot);
 722                 de.inode = ext2fs_le32_to_cpu(de.inode);
 723                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
 724                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
 725                 if (dotdot >= ctx->fs->super->s_inodes_count ||
 726                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
 727                      dotdot != EXT2_ROOT_INO) ||
 728                     de.inode >= ctx->fs->super->s_inodes_count ||
 729                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
 730                      de.inode != 0) ||
 731                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
 732                               EXT4_INLINE_DATA_DOTDOT_SIZE)
 733                         return;
 734                 /* device files never have a "system.data" entry */
 735                 goto isdir;
 736         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
 737                 /* extent mapped */
 738                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
 739                                  &blk))
 740                         return;
 741                 /* device files are never extent mapped */
 742                 not_device++;
 743         } else {
 744                 for (i=0; i < EXT2_N_BLOCKS; i++) {
 745                         blk = inode->i_block[i];
 746                         if (!blk)
 747                                 continue;
 748                         if (i >= 4)
 749                                 not_device++;
 750
 751                         if (blk < ctx->fs->super->s_first_data_block ||
 752                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
 753                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
 754                                                            blk))
 755                                 return; /* Invalid block, can't be dir */
 756                 }
 757                 blk = inode->i_block[0];
 758         }
 759
 760         /*
 761          * If the mode says this is a device file and the i_links_count field
 762          * is sane and we have not ruled it out as a device file previously,
 763          * we declare it a device file, not a directory.
 764          */
 765         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
 766             (inode->i_links_count == 1) && !not_device)
 767                 return;
 768
 769         /* read the first block */
 770         ehandler_operation(_("reading directory block"));
 771         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
 772         ehandler_operation(0);
 773         if (retval)
 774                 return;
 775
 776         dirent = (struct ext2_dir_entry *) buf;
 777         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 778         if (retval)
 779                 return;
 780         if ((ext2fs_dirent_name_len(dirent) != 1) ||
 781             (dirent->name[0] != '.') ||
 782             (dirent->inode != pctx->ino) ||
 783             (rec_len < 12) ||
 784             (rec_len % 4) ||
 785             (rec_len >= ctx->fs->blocksize - 12))
 786                 return;
 787
 788         dirent = (struct ext2_dir_entry *) (buf + rec_len);
 789         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 790         if (retval)
 791                 return;
 792         if ((ext2fs_dirent_name_len(dirent) != 2) ||
 793             (dirent->name[0] != '.') ||
 794             (dirent->name[1] != '.') ||
 795             (rec_len < 12) ||
 796             (rec_len % 4))
 797                 return;
 798
 799 isdir:
 800         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
 801                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
 802                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
 803                                         EXT2_INODE_SIZE(ctx->fs->super),
 804                                         "check_is_really_dir");
 805         }
 806 }
 807
 808 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
 809                                      int flags, ext2_icount_t hint,
 810                                      ext2_icount_t *ret)
 811 {
 812         unsigned int            threshold;
 813         unsigned int            save_type;
 814         ext2_ino_t              num_dirs;
 815         errcode_t               retval;
 816         char                    *tdb_dir;
 817         int                     enable;
 818
 819         *ret = 0;
 820
 821         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
 822                            &tdb_dir);
 823         profile_get_uint(ctx->profile, "scratch_files",
 824                          "numdirs_threshold", 0, 0, &threshold);
 825         profile_get_boolean(ctx->profile, "scratch_files",
 826                             "icount", 0, 1, &enable);
 827
 828         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
 829         if (retval)
 830                 num_dirs = 1024;        /* Guess */
 831
 832         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
 833             (!threshold || num_dirs > threshold)) {
 834                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
 835                                                   flags, ret);
 836                 if (retval == 0)
 837                         return 0;
 838         }
 839         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
 840                                &save_type);
 841         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
 842                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
 843         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
 844         ctx->fs->default_bitmap_type = save_type;
 845         return retval;
 846 }
 847
 848 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
 849                                             e2fsck_t ctx,
 850                                             struct problem_context *pctx)
 851 {
 852         errcode_t retval;
 853         struct ext2_inode_large inode;
 854
 855         /*
 856          * Reread inode.  If we don't see checksum error, then this inode
 857          * has been fixed elsewhere.
 858          */
 859         ctx->stashed_ino = 0;
 860         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
 861                                         sizeof(inode));
 862         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
 863                 return retval;
 864         if (!retval)
 865                 return 0;
 866
 867         /*
 868          * Checksum still doesn't match.  That implies that the inode passes
 869          * all the sanity checks, so maybe the checksum is simply corrupt.
 870          * See if the user will go for fixing that.
 871          */
 872         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
 873                 return 0;
 874
 875         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
 876                                          sizeof(inode));
 877         return retval;
 878 }
 879
 880 static void reserve_block_for_root_repair(e2fsck_t ctx)
 881 {
 882         blk64_t         blk = 0;
 883         errcode_t       err;
 884         ext2_filsys     fs = ctx->fs;
 885
 886         ctx->root_repair_block = 0;
 887         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
 888                 return;
 889
 890         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 891         if (err)
 892                 return;
 893         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 894         ctx->root_repair_block = blk;
 895 }
 896
 897 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
 898 {
 899         blk64_t         blk = 0;
 900         errcode_t       err;
 901         ext2_filsys     fs = ctx->fs;
 902         static const char name[] = "lost+found";
 903         ext2_ino_t      ino;
 904
 905         ctx->lnf_repair_block = 0;
 906         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
 907                 return;
 908
 909         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 910         if (err)
 911                 return;
 912         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 913         ctx->lnf_repair_block = blk;
 914 }
 915
 916 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
 917                                          size_t *sz)
 918 {
 919         void *p;
 920         struct ext2_xattr_handle *handle;
 921         errcode_t retval;
 922
 923         retval = ext2fs_xattrs_open(fs, ino, &handle);
 924         if (retval)
 925                 return retval;
 926
 927         retval = ext2fs_xattrs_read(handle);
 928         if (retval)
 929                 goto err;
 930
 931         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
 932         if (retval)
 933                 goto err;
 934         ext2fs_free_mem(&p);
 935 err:
 936         (void) ext2fs_xattrs_close(&handle);
 937         return retval;
 938 }
 939
 940 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
 941                                     struct problem_context *pctx,
 942                                     int failed_csum)
 943 {
 944         if (!failed_csum)
 945                 return;
 946
 947         /*
 948          * If the inode failed the checksum and the user didn't
 949          * clear the inode, test the checksum again -- if it still
 950          * fails, ask the user if the checksum should be corrected.
 951          */
 952         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
 953         if (pctx->errcode)
 954                 ctx->flags |= E2F_FLAG_ABORT;
 955 }
 956 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
 957         do { \
 958                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
 959                 if ((ctx)->flags & E2F_FLAG_ABORT) \
 960                         return; \
 961         } while (0)
 962
 963 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
 964 {
 965         __u32 x;
 966         int i;
 967
 968         for (i = 0; i < EXT2_N_BLOCKS; i++) {
 969                 x = inode->i_block[i];
 970 #ifdef WORDS_BIGENDIAN
 971                 x = ext2fs_swab32(x);
 972 #endif
 973                 if (x >= ext2fs_blocks_count(fs->super))
 974                         return 0;
 975         }
 976
 977         return 1;
 978 }
 979
 980 /*
 981  * Figure out what to do with an inode that has both extents and inline data
 982  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
 983  */
 984 static int fix_inline_data_extents_file(e2fsck_t ctx,
 985                                         ext2_ino_t ino,
 986                                         struct ext2_inode *inode,
 987                                         int inode_size,
 988                                         struct problem_context *pctx)
 989 {
 990         size_t max_inline_ea_size;
 991         ext2_filsys fs = ctx->fs;
 992         int dirty = 0;
 993
 994         /* Both feature flags not set?  Just run the regular checks */
 995         if (!ext2fs_has_feature_extents(fs->super) &&
 996             !ext2fs_has_feature_inline_data(fs->super))
 997                 return 0;
 998
 999         /* Clear both flags if it's a special file */
1000         if (LINUX_S_ISCHR(inode->i_mode) ||
1001             LINUX_S_ISBLK(inode->i_mode) ||
1002             LINUX_S_ISFIFO(inode->i_mode) ||
1003             LINUX_S_ISSOCK(inode->i_mode)) {
1004                 check_extents_inlinedata(ctx, pctx);
1005                 return 0;
1006         }
1007
1008         /* If it looks like an extent tree, try to clear inlinedata */
1009         if (ext2fs_extent_header_verify(inode->i_block,
1010                                  sizeof(inode->i_block)) == 0 &&
1011             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1012                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1013                 dirty = 1;
1014                 goto out;
1015         }
1016
1017         /* If it looks short enough to be inline data, try to clear extents */
1018         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1019                 max_inline_ea_size = inode_size -
1020                                      (EXT2_GOOD_OLD_INODE_SIZE +
1021                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1022         else
1023                 max_inline_ea_size = 0;
1024         if (EXT2_I_SIZE(inode) <
1025             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1026             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1027                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1028                 dirty = 1;
1029                 goto out;
1030         }
1031
1032         /*
1033          * Too big for inline data, but no evidence of extent tree -
1034          * maybe it's a block map file?  If the mappings all look valid?
1035          */
1036         if (could_be_block_map(fs, inode) &&
1037             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1038 #ifdef WORDS_BIGENDIAN
1039                 int i;
1040
1041                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1042                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1043 #endif
1044
1045                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1046                 dirty = 1;
1047                 goto out;
1048         }
1049
1050         /* Oh well, just clear the busted inode. */
1051         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1052                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1053                 return -1;
1054         }
1055
1056 out:
1057         if (dirty)
1058                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1059
1060         return 0;
1061 }
1062
1063 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1064 {
1065         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1066         dgrp_t start = *group, grp;
1067         blk64_t blocks_to_read = 0;
1068         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1069
1070         if (ctx->readahead_kb == 0)
1071                 goto out;
1072
1073         /* Keep iterating groups until we have enough to readahead */
1074         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1075         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1076                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1077                         continue;
1078                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1079                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1080                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1081                                         inodes_per_block;
1082                 if (blocks_to_read * ctx->fs->blocksize >
1083                     ctx->readahead_kb * 1024)
1084                         break;
1085         }
1086
1087         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1088                                grp - start + 1);
1089         if (err == EAGAIN) {
1090                 ctx->readahead_kb /= 2;
1091                 err = 0;
1092         }
1093
1094 out:
1095         if (err) {
1096                 /* Error; disable itable readahead */
1097                 *group = ctx->fs->group_desc_count;
1098                 *next_ino = ctx->fs->super->s_inodes_count;
1099         } else {
1100                 /*
1101                  * Don't do more readahead until we've reached the first inode
1102                  * of the last inode scan buffer block for the last group.
1103                  */
1104                 *group = grp + 1;
1105                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1106                                      ctx->inode_buffer_blocks :
1107                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1108                                     ctx->fs->blocksize /
1109                                     EXT2_INODE_SIZE(ctx->fs->super);
1110                 inodes_in_group--;
1111                 *next_ino = inodes_in_group -
1112                             (inodes_in_group % inodes_per_buffer) + 1 +
1113                             (grp * ctx->fs->super->s_inodes_per_group);
1114         }
1115 }
1116
1117 /*
1118  * Check if the passed ino is one of the used superblock quota inodes.
1119  *
1120  * Before the quota inodes were journaled, older superblock quota inodes
1121  * were just regular files in the filesystem and not reserved inodes.  This
1122  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1123  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1124  */
1125 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1126 {
1127         enum quota_type qtype;
1128
1129         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1130                 if (*quota_sb_inump(sb, qtype) == ino)
1131                         return 1;
1132
1133         return 0;
1134 }
1135
1136 /*
1137  * Check if the passed ino is one of the reserved quota inodes.
1138  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1139  * inodes.  These inodes may or may not be in use by the quota feature.
1140  */
1141 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1142 {
1143         enum quota_type qtype;
1144
1145         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1146                 if (quota_type2inum(qtype, fs->super) == ino)
1147                         return 1;
1148
1149         return 0;
1150 }
1151
1152 static int e2fsck_should_abort(e2fsck_t ctx)
1153 {
1154         e2fsck_t global_ctx;
1155
1156         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1157                 return 1;
1158
1159         if (ctx->global_ctx) {
1160                 global_ctx = ctx->global_ctx;
1161                 if (global_ctx->flags & E2F_FLAG_SIGNAL_MASK)
1162                         return 1;
1163         }
1164         return 0;
1165 }
1166
1167 void e2fsck_pass1_thread(e2fsck_t ctx)
1168 {
1169         int     i;
1170         __u64   max_sizes;
1171         ext2_filsys fs = ctx->fs;
1172         ext2_ino_t      ino = 0;
1173         struct ext2_inode *inode = NULL;
1174         ext2_inode_scan scan = NULL;
1175         char            *block_buf = NULL;
1176 #ifdef RESOURCE_TRACK
1177         struct resource_track   rtrack;
1178 #endif
1179         unsigned char   frag, fsize;
1180         struct          problem_context pctx;
1181         struct          scan_callback_struct scan_struct;
1182         struct ext2_super_block *sb = ctx->fs->super;
1183         const char      *old_op;
1184         const char      *eop_next_inode = _("getting next inode from scan");
1185         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1186         int             low_dtime_check = 1;
1187         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1188         unsigned int    bufsize;
1189         int             failed_csum = 0;
1190         ext2_ino_t      ino_threshold = 0;
1191         dgrp_t          ra_group = 0;
1192         struct ea_quota ea_ibody_quota;
1193
1194         init_resource_track(&rtrack, ctx->fs->io);
1195         clear_problem_context(&pctx);
1196
1197         /* If we can do readahead, figure out how many groups to pull in. */
1198         if (!e2fsck_can_readahead(ctx->fs))
1199                 ctx->readahead_kb = 0;
1200         else if (ctx->readahead_kb == ~0ULL)
1201                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1202         pass1_readahead(ctx, &ra_group, &ino_threshold);
1203
1204         if (!(ctx->options & E2F_OPT_PREEN))
1205                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1206
1207         if (ext2fs_has_feature_dir_index(fs->super) &&
1208             !(ctx->options & E2F_OPT_NO)) {
1209                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1210                         ctx->dirs_to_hash = 0;
1211         }
1212
1213 #ifdef MTRACE
1214         mtrace_print("Pass 1");
1215 #endif
1216
1217 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1218
1219         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1220                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1221                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1222                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1223                 max_sizes = (max_sizes * (1UL << i));
1224                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1225         }
1226 #undef EXT2_BPP
1227
1228         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1229         extent_fs = ext2fs_has_feature_extents(sb);
1230         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1231         casefold_fs = ext2fs_has_feature_casefold(sb);
1232
1233         /*
1234          * Allocate bitmaps structures
1235          */
1236         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1237                                                     EXT2FS_BMAP64_RBTREE,
1238                                                     "inode_used_map",
1239                                                     &ctx->inode_used_map);
1240         if (pctx.errcode) {
1241                 pctx.num = 1;
1242                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1243                 ctx->flags |= E2F_FLAG_ABORT;
1244                 return;
1245         }
1246         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1247                         _("directory inode map"),
1248                         EXT2FS_BMAP64_AUTODIR,
1249                         "inode_dir_map", &ctx->inode_dir_map);
1250         if (pctx.errcode) {
1251                 pctx.num = 2;
1252                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1253                 ctx->flags |= E2F_FLAG_ABORT;
1254                 return;
1255         }
1256         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1257                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1258                         "inode_reg_map", &ctx->inode_reg_map);
1259         if (pctx.errcode) {
1260                 pctx.num = 6;
1261                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1262                 ctx->flags |= E2F_FLAG_ABORT;
1263                 return;
1264         }
1265         pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1266                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1267                         "block_found_map", &ctx->block_found_map);
1268         if (pctx.errcode) {
1269                 pctx.num = 1;
1270                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1271                 ctx->flags |= E2F_FLAG_ABORT;
1272                 return;
1273         }
1274         pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1275                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1276                         "block_metadata_map", &ctx->block_metadata_map);
1277         if (pctx.errcode) {
1278                 pctx.num = 1;
1279                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1280                 ctx->flags |= E2F_FLAG_ABORT;
1281                 return;
1282         }
1283         if (casefold_fs) {
1284                 pctx.errcode =
1285                         e2fsck_allocate_inode_bitmap(fs,
1286                                                      _("inode casefold map"),
1287                                                      EXT2FS_BMAP64_RBTREE,
1288                                                      "inode_casefold_map",
1289                                                      &ctx->inode_casefold_map);
1290                 if (pctx.errcode) {
1291                         pctx.num = 1;
1292                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1293                         ctx->flags |= E2F_FLAG_ABORT;
1294                         return;
1295                 }
1296         }
1297         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1298                                            &ctx->inode_link_info);
1299         if (pctx.errcode) {
1300                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1301                 ctx->flags |= E2F_FLAG_ABORT;
1302                 return;
1303         }
1304         bufsize = inode_size;
1305         if (bufsize < sizeof(struct ext2_inode_large))
1306                 bufsize = sizeof(struct ext2_inode_large);
1307         inode = (struct ext2_inode *)
1308                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1309
1310         inodes_to_process = (struct process_inode_block *)
1311                 e2fsck_allocate_memory(ctx,
1312                                        (ctx->process_inode_size *
1313                                         sizeof(struct process_inode_block)),
1314                                        "array of inodes to process");
1315         process_inode_count = 0;
1316
1317         pctx.errcode = ext2fs_init_dblist(fs, 0);
1318         if (pctx.errcode) {
1319                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1320                 ctx->flags |= E2F_FLAG_ABORT;
1321                 goto endit;
1322         }
1323
1324         /*
1325          * If the last orphan field is set, clear it, since the pass1
1326          * processing will automatically find and clear the orphans.
1327          * In the future, we may want to try using the last_orphan
1328          * linked list ourselves, but for now, we clear it so that the
1329          * ext3 mount code won't get confused.
1330          */
1331         if (!(ctx->options & E2F_OPT_READONLY)) {
1332                 if (fs->super->s_last_orphan) {
1333                         fs->super->s_last_orphan = 0;
1334                         ext2fs_mark_super_dirty(fs);
1335                 }
1336         }
1337
1338         mark_table_blocks(ctx);
1339         pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1340                                                 &ctx->block_found_map);
1341         if (pctx.errcode) {
1342                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1343                 ctx->flags |= E2F_FLAG_ABORT;
1344                 goto endit;
1345         }
1346         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1347                                                     "block interate buffer");
1348         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1349                 e2fsck_use_inode_shortcuts(ctx, 1);
1350         e2fsck_intercept_block_allocations(ctx);
1351         old_op = ehandler_operation(_("opening inode scan"));
1352         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1353                                               &scan);
1354         ehandler_operation(old_op);
1355         if (pctx.errcode) {
1356                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1357                 ctx->flags |= E2F_FLAG_ABORT;
1358                 goto endit;
1359         }
1360         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1361                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1362         ctx->stashed_inode = inode;
1363         scan_struct.ctx = ctx;
1364         scan_struct.block_buf = block_buf;
1365         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1366         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1367                                               ctx->fs->group_desc_count)))
1368                 goto endit;
1369         if ((fs->super->s_wtime &&
1370              fs->super->s_wtime < fs->super->s_inodes_count) ||
1371             (fs->super->s_mtime &&
1372              fs->super->s_mtime < fs->super->s_inodes_count) ||
1373             (fs->super->s_mkfs_time &&
1374              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1375                 low_dtime_check = 0;
1376
1377         if (ext2fs_has_feature_mmp(fs->super) &&
1378             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1379             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1380                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1381                                           fs->super->s_mmp_block);
1382
1383         /* Set up ctx->lost_and_found if possible */
1384         (void) e2fsck_get_lost_and_found(ctx, 0);
1385
1386         while (1) {
1387                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1388                         if (e2fsck_mmp_update(fs))
1389                                 fatal_error(ctx, 0);
1390                 }
1391                 old_op = ehandler_operation(eop_next_inode);
1392                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1393                                                           inode, inode_size);
1394                 if (ino > ino_threshold)
1395                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1396                 ehandler_operation(old_op);
1397                 if (e2fsck_should_abort(ctx))
1398                         goto endit;
1399                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1400                         /*
1401                          * If badblocks says badblocks is bad, offer to clear
1402                          * the list, update the in-core bb list, and restart
1403                          * the inode scan.
1404                          */
1405                         if (ino == EXT2_BAD_INO &&
1406                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1407                                         &pctx)) {
1408                                 errcode_t err;
1409
1410                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1411                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1412                                 ctx->fs->badblocks = NULL;
1413                                 err = ext2fs_read_bb_inode(ctx->fs,
1414                                                         &ctx->fs->badblocks);
1415                                 if (err) {
1416                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1417                                                     &pctx);
1418                                         ctx->flags |= E2F_FLAG_ABORT;
1419                                 } else
1420                                         ctx->flags |= E2F_FLAG_RESTART;
1421                                 goto endit;
1422                         }
1423                         if (!ctx->inode_bb_map)
1424                                 alloc_bb_map(ctx);
1425                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1426                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1427                         continue;
1428                 }
1429                 if (pctx.errcode &&
1430                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1431                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1432                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1433                         ctx->flags |= E2F_FLAG_ABORT;
1434                         goto endit;
1435                 }
1436                 if (!ino)
1437                         break;
1438                 pctx.ino = ino;
1439                 pctx.inode = inode;
1440                 ctx->stashed_ino = ino;
1441
1442                 /* Clear trashed inode? */
1443                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1444                     inode->i_links_count > 0 &&
1445                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1446                         pctx.errcode = 0;
1447                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1448                 }
1449                 failed_csum = pctx.errcode != 0;
1450
1451                 /*
1452                  * Check for inodes who might have been part of the
1453                  * orphaned list linked list.  They should have gotten
1454                  * dealt with by now, unless the list had somehow been
1455                  * corrupted.
1456                  *
1457                  * FIXME: In the future, inodes which are still in use
1458                  * (and which are therefore) pending truncation should
1459                  * be handled specially.  Right now we just clear the
1460                  * dtime field, and the normal e2fsck handling of
1461                  * inodes where i_size and the inode blocks are
1462                  * inconsistent is to fix i_size, instead of releasing
1463                  * the extra blocks.  This won't catch the inodes that
1464                  * was at the end of the orphan list, but it's better
1465                  * than nothing.  The right answer is that there
1466                  * shouldn't be any bugs in the orphan list handling.  :-)
1467                  */
1468                 if (inode->i_dtime && low_dtime_check &&
1469                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1470                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1471                                 inode->i_dtime = inode->i_links_count ?
1472                                         0 : ctx->now;
1473                                 e2fsck_write_inode(ctx, ino, inode,
1474                                                    "pass1");
1475                                 failed_csum = 0;
1476                         }
1477                 }
1478
1479                 if (inode->i_links_count) {
1480                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1481                                            ino, inode->i_links_count);
1482                         if (pctx.errcode) {
1483                                 pctx.num = inode->i_links_count;
1484                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1485                                 ctx->flags |= E2F_FLAG_ABORT;
1486                                 goto endit;
1487                         }
1488                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1489                            !quota_inum_is_reserved(fs, ino)) {
1490                         if (!inode->i_dtime && inode->i_mode) {
1491                                 if (fix_problem(ctx,
1492                                             PR_1_ZERO_DTIME, &pctx)) {
1493                                         inode->i_dtime = ctx->now;
1494                                         e2fsck_write_inode(ctx, ino, inode,
1495                                                            "pass1");
1496                                         failed_csum = 0;
1497                                 }
1498                         }
1499                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1500                         continue;
1501                 }
1502
1503                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1504                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1505                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1506                      (!casefold_fs &&
1507                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1508                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1509                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1510                 }
1511
1512                 /* Conflicting inlinedata/extents inode flags? */
1513                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1514                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1515                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1516                                                                inode_size,
1517                                                                &pctx);
1518                         if (res < 0) {
1519                                 /* skip FINISH_INODE_LOOP */
1520                                 continue;
1521                         }
1522                 }
1523
1524                 /* Test for incorrect inline_data flags settings. */
1525                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1526                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1527                         size_t size = 0;
1528
1529                         pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1530                         if (!pctx.errcode &&
1531                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1532                                 ext2fs_set_feature_inline_data(sb);
1533                                 ext2fs_mark_super_dirty(fs);
1534                                 inlinedata_fs = 1;
1535                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1536                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1537                                 /* skip FINISH_INODE_LOOP */
1538                                 continue;
1539                         }
1540                 }
1541
1542                 /* Test for inline data flag but no attr */
1543                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1544                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1545                         size_t size = 0;
1546                         errcode_t err;
1547                         int flags;
1548
1549                         flags = fs->flags;
1550                         if (failed_csum)
1551                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1552                         err = get_inline_data_ea_size(fs, ino, &size);
1553                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1554                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1555
1556                         switch (err) {
1557                         case 0:
1558                                 /* Everything is awesome... */
1559                                 break;
1560                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1561                         case EXT2_ET_BAD_EA_HASH:
1562                         case EXT2_ET_BAD_EA_HEADER:
1563                         case EXT2_ET_EA_BAD_NAME_LEN:
1564                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1565                         case EXT2_ET_EA_KEY_NOT_FOUND:
1566                         case EXT2_ET_EA_NO_SPACE:
1567                         case EXT2_ET_MISSING_EA_FEATURE:
1568                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1569                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1570                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1571                         case EXT2_ET_NO_INLINE_DATA:
1572                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1573                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1574                         case EXT2_ET_EA_INODE_CORRUPTED:
1575                                 /* broken EA or no system.data EA; truncate */
1576                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1577                                                 &pctx)) {
1578                                         err = ext2fs_inode_size_set(fs, inode, 0);
1579                                         if (err) {
1580                                                 pctx.errcode = err;
1581                                                 ctx->flags |= E2F_FLAG_ABORT;
1582                                                 goto endit;
1583                                         }
1584                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1585                                         memset(&inode->i_block, 0,
1586                                                sizeof(inode->i_block));
1587                                         e2fsck_write_inode(ctx, ino, inode,
1588                                                            "pass1");
1589                                         failed_csum = 0;
1590                                 }
1591                                 break;
1592                         default:
1593                                 /* Some other kind of non-xattr error? */
1594                                 pctx.errcode = err;
1595                                 ctx->flags |= E2F_FLAG_ABORT;
1596                                 goto endit;
1597                         }
1598                 }
1599
1600                 /*
1601                  * Test for incorrect extent flag settings.
1602                  *
1603                  * On big-endian machines we must be careful:
1604                  * When the inode is read, the i_block array is not swapped
1605                  * if the extent flag is set.  Therefore if we are testing
1606                  * for or fixing a wrongly-set flag, we must potentially
1607                  * (un)swap before testing, or after fixing.
1608                  */
1609
1610                 /*
1611                  * In this case the extents flag was set when read, so
1612                  * extent_header_verify is ok.  If the inode is cleared,
1613                  * no need to swap... so no extra swapping here.
1614                  */
1615                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1616                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1617                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1618                         if ((ext2fs_extent_header_verify(inode->i_block,
1619                                                  sizeof(inode->i_block)) == 0) &&
1620                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1621                                 ext2fs_set_feature_extents(sb);
1622                                 ext2fs_mark_super_dirty(fs);
1623                                 extent_fs = 1;
1624                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1625                         clear_inode:
1626                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1627                                 if (ino == EXT2_BAD_INO)
1628                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1629                                                                  ino);
1630                                 /* skip FINISH_INODE_LOOP */
1631                                 continue;
1632                         }
1633                 }
1634
1635                 /*
1636                  * For big-endian machines:
1637                  * If the inode didn't have the extents flag set when it
1638                  * was read, then the i_blocks array was swapped.  To test
1639                  * as an extents header, we must swap it back first.
1640                  * IF we then set the extents flag, the entire i_block
1641                  * array must be un/re-swapped to make it proper extents data.
1642                  */
1643                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1644                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1645                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1646                     (LINUX_S_ISREG(inode->i_mode) ||
1647                      LINUX_S_ISDIR(inode->i_mode))) {
1648                         void *ehp;
1649 #ifdef WORDS_BIGENDIAN
1650                         __u32 tmp_block[EXT2_N_BLOCKS];
1651
1652                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1653                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1654                         ehp = tmp_block;
1655 #else
1656                         ehp = inode->i_block;
1657 #endif
1658                         if ((ext2fs_extent_header_verify(ehp,
1659                                          sizeof(inode->i_block)) == 0) &&
1660                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1661                                 inode->i_flags |= EXT4_EXTENTS_FL;
1662 #ifdef WORDS_BIGENDIAN
1663                                 memcpy(inode->i_block, tmp_block,
1664                                        sizeof(inode->i_block));
1665 #endif
1666                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1667                                 failed_csum = 0;
1668                         }
1669                 }
1670
1671                 if (ino == EXT2_BAD_INO) {
1672                         struct process_block_struct pb;
1673
1674                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1675                              inode->i_gid || inode->i_links_count ||
1676                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1677                              inode->i_file_acl) &&
1678                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1679                                 memset(inode, 0, sizeof(struct ext2_inode));
1680                                 e2fsck_write_inode(ctx, ino, inode,
1681                                                    "clear bad inode");
1682                                 failed_csum = 0;
1683                         }
1684
1685                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1686                                                           &pb.fs_meta_blocks);
1687                         if (pctx.errcode) {
1688                                 pctx.num = 4;
1689                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1690                                 ctx->flags |= E2F_FLAG_ABORT;
1691                                 goto endit;
1692                         }
1693                         pb.ino = EXT2_BAD_INO;
1694                         pb.num_blocks = pb.last_block = 0;
1695                         pb.last_db_block = -1;
1696                         pb.num_illegal_blocks = 0;
1697                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1698                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1699                         pb.inode = inode;
1700                         pb.pctx = &pctx;
1701                         pb.ctx = ctx;
1702                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1703                                      block_buf, process_bad_block, &pb);
1704                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1705                         if (pctx.errcode) {
1706                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1707                                 ctx->flags |= E2F_FLAG_ABORT;
1708                                 goto endit;
1709                         }
1710                         if (pb.bbcheck)
1711                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1712                                 ctx->flags |= E2F_FLAG_ABORT;
1713                                 goto endit;
1714                         }
1715                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1716                         clear_problem_context(&pctx);
1717                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1718                         continue;
1719                 } else if (ino == EXT2_ROOT_INO) {
1720                         /*
1721                          * Make sure the root inode is a directory; if
1722                          * not, offer to clear it.  It will be
1723                          * regenerated in pass #3.
1724                          */
1725                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1726                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1727                                         goto clear_inode;
1728                         }
1729                         /*
1730                          * If dtime is set, offer to clear it.  mke2fs
1731                          * version 0.2b created filesystems with the
1732                          * dtime field set for the root and lost+found
1733                          * directories.  We won't worry about
1734                          * /lost+found, since that can be regenerated
1735                          * easily.  But we will fix the root directory
1736                          * as a special case.
1737                          */
1738                         if (inode->i_dtime && inode->i_links_count) {
1739                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1740                                         inode->i_dtime = 0;
1741                                         e2fsck_write_inode(ctx, ino, inode,
1742                                                            "pass1");
1743                                         failed_csum = 0;
1744                                 }
1745                         }
1746                 } else if (ino == EXT2_JOURNAL_INO) {
1747                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1748                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1749                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1750                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1751                                                 &pctx)) {
1752                                         inode->i_mode = LINUX_S_IFREG;
1753                                         e2fsck_write_inode(ctx, ino, inode,
1754                                                            "pass1");
1755                                         failed_csum = 0;
1756                                 }
1757                                 check_blocks(ctx, &pctx, block_buf, NULL);
1758                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1759                                 continue;
1760                         }
1761                         if ((inode->i_links_count ||
1762                              inode->i_blocks || inode->i_block[0]) &&
1763                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1764                                         &pctx)) {
1765                                 memset(inode, 0, inode_size);
1766                                 ext2fs_icount_store(ctx->inode_link_info,
1767                                                     ino, 0);
1768                                 e2fsck_write_inode_full(ctx, ino, inode,
1769                                                         inode_size, "pass1");
1770                                 failed_csum = 0;
1771                         }
1772                 } else if (quota_inum_is_reserved(fs, ino)) {
1773                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1774                         if (ext2fs_has_feature_quota(fs->super) &&
1775                             quota_inum_is_super(fs->super, ino)) {
1776                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1777                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1778                                                         &pctx)) {
1779                                         inode->i_mode = LINUX_S_IFREG;
1780                                         e2fsck_write_inode(ctx, ino, inode,
1781                                                         "pass1");
1782                                         failed_csum = 0;
1783                                 }
1784                                 check_blocks(ctx, &pctx, block_buf, NULL);
1785                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1786                                 continue;
1787                         }
1788                         if ((inode->i_links_count ||
1789                              inode->i_blocks || inode->i_block[0]) &&
1790                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1791                                         &pctx)) {
1792                                 memset(inode, 0, inode_size);
1793                                 ext2fs_icount_store(ctx->inode_link_info,
1794                                                     ino, 0);
1795                                 e2fsck_write_inode_full(ctx, ino, inode,
1796                                                         inode_size, "pass1");
1797                                 failed_csum = 0;
1798                         }
1799                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1800                         problem_t problem = 0;
1801
1802                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1803                         if (ino == EXT2_BOOT_LOADER_INO) {
1804                                 if (LINUX_S_ISDIR(inode->i_mode))
1805                                         problem = PR_1_RESERVED_BAD_MODE;
1806                         } else if (ino == EXT2_RESIZE_INO) {
1807                                 if (inode->i_mode &&
1808                                     !LINUX_S_ISREG(inode->i_mode))
1809                                         problem = PR_1_RESERVED_BAD_MODE;
1810                         } else {
1811                                 if (inode->i_mode != 0)
1812                                         problem = PR_1_RESERVED_BAD_MODE;
1813                         }
1814                         if (problem) {
1815                                 if (fix_problem(ctx, problem, &pctx)) {
1816                                         inode->i_mode = 0;
1817                                         e2fsck_write_inode(ctx, ino, inode,
1818                                                            "pass1");
1819                                         failed_csum = 0;
1820                                 }
1821                         }
1822                         check_blocks(ctx, &pctx, block_buf, NULL);
1823                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1824                         continue;
1825                 }
1826
1827                 if (!inode->i_links_count) {
1828                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1829                         continue;
1830                 }
1831                 /*
1832                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1833                  * deleted files.  Oops.
1834                  *
1835                  * Since all new ext2 implementations get this right,
1836                  * we now assume that the case of non-zero
1837                  * i_links_count and non-zero dtime means that we
1838                  * should keep the file, not delete it.
1839                  *
1840                  */
1841                 if (inode->i_dtime) {
1842                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1843                                 inode->i_dtime = 0;
1844                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1845                                 failed_csum = 0;
1846                         }
1847                 }
1848
1849                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1850                 switch (fs->super->s_creator_os) {
1851                     case EXT2_OS_HURD:
1852                         frag = inode->osd2.hurd2.h_i_frag;
1853                         fsize = inode->osd2.hurd2.h_i_fsize;
1854                         break;
1855                     default:
1856                         frag = fsize = 0;
1857                 }
1858
1859                 if (inode->i_faddr || frag || fsize ||
1860                     (!ext2fs_has_feature_largedir(fs->super) &&
1861                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1862                         mark_inode_bad(ctx, ino);
1863                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1864                     !ext2fs_has_feature_64bit(fs->super) &&
1865                     inode->osd2.linux2.l_i_file_acl_high != 0)
1866                         mark_inode_bad(ctx, ino);
1867                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1868                     !ext2fs_has_feature_huge_file(fs->super) &&
1869                     (inode->osd2.linux2.l_i_blocks_hi != 0))
1870                         mark_inode_bad(ctx, ino);
1871                 if (inode->i_flags & EXT2_IMAGIC_FL) {
1872                         if (imagic_fs) {
1873                                 if (!ctx->inode_imagic_map)
1874                                         alloc_imagic_map(ctx);
1875                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1876                                                          ino);
1877                         } else {
1878                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1879                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
1880                                         e2fsck_write_inode(ctx, ino,
1881                                                            inode, "pass1");
1882                                         failed_csum = 0;
1883                                 }
1884                         }
1885                 }
1886
1887                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1888                 check_is_really_dir(ctx, &pctx, block_buf);
1889
1890                 /*
1891                  * ext2fs_inode_has_valid_blocks2 does not actually look
1892                  * at i_block[] values, so not endian-sensitive here.
1893                  */
1894                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1895                     LINUX_S_ISLNK(inode->i_mode) &&
1896                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1897                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1898                         inode->i_flags &= ~EXT4_EXTENTS_FL;
1899                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1900                         failed_csum = 0;
1901                 }
1902
1903                 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
1904                     add_encrypted_file(ctx, &pctx) < 0)
1905                         goto clear_inode;
1906
1907                 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
1908                         ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
1909
1910                 if (LINUX_S_ISDIR(inode->i_mode)) {
1911                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1912                         e2fsck_add_dir_info(ctx, ino, 0);
1913                         ctx->fs_directory_count++;
1914                         if (inode->i_flags & EXT4_CASEFOLD_FL)
1915                                 add_casefolded_dir(ctx, ino);
1916                 } else if (LINUX_S_ISREG (inode->i_mode)) {
1917                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1918                         ctx->fs_regular_count++;
1919                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1920                            e2fsck_pass1_check_device_inode(fs, inode)) {
1921                         check_extents_inlinedata(ctx, &pctx);
1922                         check_immutable(ctx, &pctx);
1923                         check_size(ctx, &pctx);
1924                         ctx->fs_chardev_count++;
1925                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1926                            e2fsck_pass1_check_device_inode(fs, inode)) {
1927                         check_extents_inlinedata(ctx, &pctx);
1928                         check_immutable(ctx, &pctx);
1929                         check_size(ctx, &pctx);
1930                         ctx->fs_blockdev_count++;
1931                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1932                            e2fsck_pass1_check_symlink(fs, ino, inode,
1933                                                       block_buf)) {
1934                         check_immutable(ctx, &pctx);
1935                         ctx->fs_symlinks_count++;
1936                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1937                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1938                                 continue;
1939                         } else if (ext2fs_is_fast_symlink(inode)) {
1940                                 ctx->fs_fast_symlinks_count++;
1941                                 check_blocks(ctx, &pctx, block_buf,
1942                                              &ea_ibody_quota);
1943                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1944                                 continue;
1945                         }
1946                 }
1947                 else if (LINUX_S_ISFIFO (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_fifo_count++;
1953                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
1954                            e2fsck_pass1_check_device_inode(fs, inode)) {
1955                         check_extents_inlinedata(ctx, &pctx);
1956                         check_immutable(ctx, &pctx);
1957                         check_size(ctx, &pctx);
1958                         ctx->fs_sockets_count++;
1959                 } else
1960                         mark_inode_bad(ctx, ino);
1961                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1962                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
1963                         if (inode->i_block[EXT2_IND_BLOCK])
1964                                 ctx->fs_ind_count++;
1965                         if (inode->i_block[EXT2_DIND_BLOCK])
1966                                 ctx->fs_dind_count++;
1967                         if (inode->i_block[EXT2_TIND_BLOCK])
1968                                 ctx->fs_tind_count++;
1969                 }
1970                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1971                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
1972                     (inode->i_block[EXT2_IND_BLOCK] ||
1973                      inode->i_block[EXT2_DIND_BLOCK] ||
1974                      inode->i_block[EXT2_TIND_BLOCK] ||
1975                      ext2fs_file_acl_block(fs, inode))) {
1976                         struct process_inode_block *itp;
1977
1978                         itp = &inodes_to_process[process_inode_count];
1979                         itp->ino = ino;
1980                         itp->ea_ibody_quota = ea_ibody_quota;
1981                         if (inode_size < sizeof(struct ext2_inode_large))
1982                                 memcpy(&itp->inode, inode, inode_size);
1983                         else
1984                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
1985                         process_inode_count++;
1986                 } else
1987                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
1988
1989                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1990
1991                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1992                         goto endit;
1993
1994                 if (process_inode_count >= ctx->process_inode_size) {
1995                         process_inodes(ctx, block_buf);
1996
1997                         if (e2fsck_should_abort(ctx))
1998                                 goto endit;
1999                 }
2000         }
2001         process_inodes(ctx, block_buf);
2002         ext2fs_close_inode_scan(scan);
2003         scan = NULL;
2004
2005         reserve_block_for_root_repair(ctx);
2006         reserve_block_for_lnf_repair(ctx);
2007
2008         /*
2009          * If any extended attribute blocks' reference counts need to
2010          * be adjusted, either up (ctx->refcount_extra), or down
2011          * (ctx->refcount), then fix them.
2012          */
2013         if (ctx->refcount) {
2014                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
2015                 ea_refcount_free(ctx->refcount);
2016                 ctx->refcount = 0;
2017         }
2018         if (ctx->refcount_extra) {
2019                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
2020                                         block_buf, +1);
2021                 ea_refcount_free(ctx->refcount_extra);
2022                 ctx->refcount_extra = 0;
2023         }
2024
2025         if (ctx->ea_block_quota_blocks) {
2026                 ea_refcount_free(ctx->ea_block_quota_blocks);
2027                 ctx->ea_block_quota_blocks = 0;
2028         }
2029
2030         if (ctx->ea_block_quota_inodes) {
2031                 ea_refcount_free(ctx->ea_block_quota_inodes);
2032                 ctx->ea_block_quota_inodes = 0;
2033         }
2034
2035         if (ctx->invalid_bitmaps)
2036                 handle_fs_bad_blocks(ctx);
2037
2038         /* We don't need the block_ea_map any more */
2039         if (ctx->block_ea_map) {
2040                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2041                 ctx->block_ea_map = 0;
2042         }
2043
2044         /* We don't need the encryption policy => ID map any more */
2045         destroy_encryption_policy_map(ctx);
2046
2047         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2048                 clear_problem_context(&pctx);
2049                 pctx.errcode = ext2fs_create_resize_inode(fs);
2050                 if (pctx.errcode) {
2051                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2052                                          &pctx)) {
2053                                 ctx->flags |= E2F_FLAG_ABORT;
2054                                 goto endit;
2055                         }
2056                         pctx.errcode = 0;
2057                 }
2058                 if (!pctx.errcode) {
2059                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2060                                           "recreate inode");
2061                         inode->i_mtime = ctx->now;
2062                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2063                                            "recreate inode");
2064                 }
2065                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2066         }
2067
2068         if (ctx->flags & E2F_FLAG_RESTART) {
2069                 /*
2070                  * Only the master copy of the superblock and block
2071                  * group descriptors are going to be written during a
2072                  * restart, so set the superblock to be used to be the
2073                  * master superblock.
2074                  */
2075                 ctx->use_superblock = 0;
2076                 goto endit;
2077         }
2078
2079         if (ctx->large_dirs && !ext2fs_has_feature_largedir(fs->super)) {
2080                 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2081                         ext2fs_set_feature_largedir(fs->super);
2082                         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2083                         ext2fs_mark_super_dirty(fs);
2084                 }
2085                 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2086                     fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2087                         ext2fs_update_dynamic_rev(fs);
2088                         ext2fs_mark_super_dirty(fs);
2089                 }
2090         }
2091
2092         if (ctx->block_dup_map) {
2093                 if (ctx->options & E2F_OPT_PREEN) {
2094                         clear_problem_context(&pctx);
2095                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2096                 }
2097                 e2fsck_pass1_dupblocks(ctx, block_buf);
2098         }
2099         ctx->flags |= E2F_FLAG_ALLOC_OK;
2100 endit:
2101         e2fsck_use_inode_shortcuts(ctx, 0);
2102         ext2fs_free_mem(&inodes_to_process);
2103         inodes_to_process = 0;
2104
2105         if (scan)
2106                 ext2fs_close_inode_scan(scan);
2107         if (block_buf)
2108                 ext2fs_free_mem(&block_buf);
2109         if (inode)
2110                 ext2fs_free_mem(&inode);
2111
2112         /*
2113          * The l+f inode may have been cleared, so zap it now and
2114          * later passes will recalculate it if necessary
2115          */
2116         ctx->lost_and_found = 0;
2117
2118         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2119                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2120         else
2121                 ctx->invalid_bitmaps++;
2122 }
2123
2124 static errcode_t e2fsck_pass1_copy_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2125                                           ext2fs_generic_bitmap *dest)
2126 {
2127         errcode_t ret;
2128
2129         ret = ext2fs_copy_bitmap(*src, dest);
2130         if (ret)
2131                 return ret;
2132
2133         (*dest)->fs = fs;
2134         ext2fs_free_generic_bmap(*src);
2135         *src = NULL;
2136
2137         return 0;
2138 }
2139
2140
2141 static errcode_t e2fsck_pass1_copy_fs(ext2_filsys dest, e2fsck_t src_context,
2142                                       ext2_filsys src)
2143 {
2144         errcode_t       retval;
2145
2146         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2147         if (dest->dblist)
2148                 dest->dblist->fs = dest;
2149         if (src->block_map) {
2150                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2151                                                   &dest->block_map);
2152                 if (retval)
2153                         return retval;
2154         }
2155         if (src->inode_map) {
2156                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2157                                                   &dest->inode_map);
2158                 if (retval)
2159                         return retval;
2160         }
2161
2162         if (src->badblocks) {
2163                 retval = ext2fs_badblocks_copy(src->badblocks, &dest->badblocks);
2164                 if (retval)
2165                         return retval;
2166         }
2167
2168         /* disable it for now */
2169         src_context->openfs_flags &= ~EXT2_FLAG_EXCLUSIVE;
2170         retval = ext2fs_open_channel(dest, src_context->io_options,
2171                                      src_context->io_manager,
2172                                      src_context->openfs_flags,
2173                                      src->io->block_size);
2174         if (retval)
2175                 return retval;
2176
2177         /* Block size might not be default */
2178         io_channel_set_blksize(dest->io, src->io->block_size);
2179         ehandler_init(dest->io);
2180
2181         assert(dest->io->magic == src->io->magic);
2182         assert(dest->io->manager == src->io->manager);
2183         assert(strcmp(dest->io->name, src->io->name) == 0);
2184         assert(dest->io->block_size == src->io->block_size);
2185         assert(dest->io->read_error == src->io->read_error);
2186         assert(dest->io->write_error == src->io->write_error);
2187         assert(dest->io->refcount == src->io->refcount);
2188         assert(dest->io->flags == src->io->flags);
2189         assert(dest->io->app_data == dest);
2190         assert(src->io->app_data == src);
2191         assert(dest->io->align == src->io->align);
2192
2193         /* The data should be written to disk immediately */
2194         dest->io->flags |= CHANNEL_FLAGS_WRITETHROUGH;
2195         /* icache will be rebuilt if needed, so do not copy from @src */
2196         src->icache = NULL;
2197         return 0;
2198 }
2199
2200 static int e2fsck_pass1_merge_fs(ext2_filsys dest, ext2_filsys src)
2201 {
2202         struct ext2_inode_cache *icache = dest->icache;
2203         errcode_t retval = 0;
2204         io_channel dest_io;
2205         io_channel dest_image_io;
2206
2207         dest_io = dest->io;
2208         dest_image_io = dest->image_io;
2209
2210         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2211         dest->io = dest_io;
2212         dest->image_io = dest_image_io;
2213         dest->icache = icache;
2214         if (dest->dblist)
2215                 dest->dblist->fs = dest;
2216         if (src->inode_map) {
2217                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2218                                                   &dest->inode_map);
2219                 if (retval)
2220                         return retval;
2221         }
2222         if (src->block_map) {
2223                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2224                                                   &dest->block_map);
2225                 if (retval)
2226                         return retval;
2227         }
2228
2229         if (src->icache) {
2230                 ext2fs_free_inode_cache(src->icache);
2231                 src->icache = NULL;
2232         }
2233
2234         if (src->badblocks) {
2235                 retval = ext2fs_badblocks_copy(src->badblocks, &dest->badblocks);
2236
2237                 ext2fs_badblocks_list_free(src->badblocks);
2238                 src->badblocks = NULL;
2239         }
2240
2241         io_channel_close(src->io);
2242         return retval;
2243 }
2244
2245 static errcode_t e2fsck_pass1_thread_prepare(e2fsck_t global_ctx, e2fsck_t *thread_ctx)
2246 {
2247         errcode_t       retval;
2248         e2fsck_t        thread_context;
2249         ext2_filsys     thread_fs;
2250         ext2_filsys     global_fs = global_ctx->fs;
2251
2252         assert(global_ctx->inode_used_map == NULL);
2253         assert(global_ctx->inode_dir_map == NULL);
2254         assert(global_ctx->inode_bb_map == NULL);
2255         assert(global_ctx->inode_imagic_map == NULL);
2256         assert(global_ctx->inode_reg_map == NULL);
2257         assert(global_ctx->inodes_to_rebuild == NULL);
2258
2259         assert(global_ctx->block_found_map == NULL);
2260         assert(global_ctx->block_dup_map == NULL);
2261         assert(global_ctx->block_ea_map == NULL);
2262         assert(global_ctx->block_metadata_map == NULL);
2263         assert(global_ctx->fs->dblist == NULL);
2264
2265         retval = ext2fs_get_mem(sizeof(struct e2fsck_struct), &thread_context);
2266         if (retval) {
2267                 com_err(global_ctx->program_name, retval, "while allocating memory");
2268                 return retval;
2269         }
2270         memcpy(thread_context, global_ctx, sizeof(struct e2fsck_struct));
2271         thread_context->global_ctx = global_ctx;
2272
2273         retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &thread_fs);
2274         if (retval) {
2275                 com_err(global_ctx->program_name, retval, "while allocating memory");
2276                 goto out_context;
2277         }
2278
2279         io_channel_flush_cleanup(global_fs->io);
2280         retval = e2fsck_pass1_copy_fs(thread_fs, global_ctx, global_fs);
2281         if (retval) {
2282                 com_err(global_ctx->program_name, retval, "while copying fs");
2283                 goto out_fs;
2284         }
2285         thread_fs->priv_data = thread_context;
2286
2287         thread_context->fs = thread_fs;
2288         *thread_ctx = thread_context;
2289         return 0;
2290 out_fs:
2291         ext2fs_free_mem(&thread_fs);
2292 out_context:
2293         ext2fs_free_mem(&thread_context);
2294         return retval;
2295 }
2296
2297 static int e2fsck_pass1_thread_join_one(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2298 {
2299         errcode_t       retval;
2300         int             flags = global_ctx->flags;
2301         ext2_filsys     thread_fs = thread_ctx->fs;
2302         ext2_filsys     global_fs = global_ctx->fs;
2303 #ifdef HAVE_SETJMP_H
2304         jmp_buf         old_jmp;
2305
2306         memcpy(old_jmp, global_ctx->abort_loc, sizeof(jmp_buf));
2307 #endif
2308         memcpy(global_ctx, thread_ctx, sizeof(struct e2fsck_struct));
2309 #ifdef HAVE_SETJMP_H
2310         memcpy(global_ctx->abort_loc, old_jmp, sizeof(jmp_buf));
2311 #endif
2312         /* Keep the global singal flags*/
2313         global_ctx->flags |= (flags & E2F_FLAG_SIGNAL_MASK) |
2314                              (global_ctx->flags & E2F_FLAG_SIGNAL_MASK);
2315
2316         retval = e2fsck_pass1_merge_fs(global_fs, thread_fs);
2317         if (retval) {
2318                 com_err(global_ctx->program_name, 0, _("while merging fs\n"));
2319                 return retval;
2320         }
2321         global_fs->priv_data = global_ctx;
2322         global_ctx->fs = global_fs;
2323
2324         if (thread_ctx->inode_used_map) {
2325                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2326                                         &thread_ctx->inode_used_map,
2327                                         &global_ctx->inode_used_map);
2328                 if (retval)
2329                         return retval;
2330         }
2331         if (thread_ctx->inode_bad_map) {
2332                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2333                                         &thread_ctx->inode_bad_map,
2334                                         &global_ctx->inode_bad_map);
2335                 if (retval)
2336                         return retval;
2337         }
2338         if (thread_ctx->inode_dir_map) {
2339                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2340                                         &thread_ctx->inode_dir_map,
2341                                         &global_ctx->inode_dir_map);
2342                 if (retval)
2343                         return retval;
2344         }
2345         if (thread_ctx->inode_bb_map) {
2346                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2347                                         &thread_ctx->inode_bb_map,
2348                                         &global_ctx->inode_bb_map);
2349                 if (retval)
2350                         return retval;
2351         }
2352         if (thread_ctx->inode_imagic_map) {
2353                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2354                                         &thread_ctx->inode_imagic_map,
2355                                         &global_ctx->inode_imagic_map);
2356                 if (retval)
2357                         return retval;
2358         }
2359         if (thread_ctx->inode_reg_map) {
2360                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2361                                         &thread_ctx->inode_reg_map,
2362                                         &global_ctx->inode_reg_map);
2363                 if (retval)
2364                         return retval;
2365         }
2366         if (thread_ctx->inodes_to_rebuild) {
2367                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2368                                         &thread_ctx->inodes_to_rebuild,
2369                                         &global_ctx->inodes_to_rebuild);
2370                 if (retval)
2371                         return retval;
2372         }
2373         if (thread_ctx->block_found_map) {
2374                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2375                                         &thread_ctx->block_found_map,
2376                                         &global_ctx->block_found_map);
2377                 if (retval)
2378                         return retval;
2379         }
2380         if (thread_ctx->block_dup_map) {
2381                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2382                                         &thread_ctx->block_dup_map,
2383                                         &global_ctx->block_dup_map);
2384                 if (retval)
2385                         return retval;
2386         }
2387         if (thread_ctx->block_ea_map) {
2388                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2389                                         &thread_ctx->block_ea_map,
2390                                         &global_ctx->block_ea_map);
2391                 if (retval)
2392                         return retval;
2393         }
2394         if (thread_ctx->block_metadata_map) {
2395                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2396                                         &thread_ctx->block_metadata_map,
2397                                         &global_ctx->block_metadata_map);
2398                 if (retval)
2399                         return retval;
2400         }
2401
2402         return 0;
2403 }
2404
2405 static int e2fsck_pass1_thread_join(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2406 {
2407         errcode_t       retval;
2408
2409         retval = e2fsck_pass1_thread_join_one(global_ctx, thread_ctx);
2410         ext2fs_free_mem(&thread_ctx->fs);
2411         ext2fs_free_mem(&thread_ctx);
2412
2413         return retval;
2414 }
2415
2416 void e2fsck_pass1_multithread(e2fsck_t ctx)
2417 {
2418         errcode_t       retval;
2419         e2fsck_t        thread_ctx;
2420
2421         retval = e2fsck_pass1_thread_prepare(ctx, &thread_ctx);
2422         if (retval) {
2423                 com_err(ctx->program_name, 0,
2424                         _("while preparing pass1 thread\n"));
2425                 ctx->flags |= E2F_FLAG_ABORT;
2426                 return;
2427         }
2428
2429 #ifdef HAVE_SETJMP_H
2430         /*
2431          * When fatal_error() happens, jump to here. The thread
2432          * context's flags will be saved, but its abort_loc will
2433          * be overwritten by original jump buffer for the later
2434          * tests.
2435          */
2436         if (setjmp(thread_ctx->abort_loc)) {
2437                 thread_ctx->flags &= ~E2F_FLAG_SETJMP_OK;
2438                 e2fsck_pass1_thread_join(ctx, thread_ctx);
2439                 return;
2440         }
2441         thread_ctx->flags |= E2F_FLAG_SETJMP_OK;
2442 #endif
2443
2444         e2fsck_pass1_thread(thread_ctx);
2445         retval = e2fsck_pass1_thread_join(ctx, thread_ctx);
2446         if (retval) {
2447                 com_err(ctx->program_name, 0,
2448                         _("while joining pass1 thread\n"));
2449                 ctx->flags |= E2F_FLAG_ABORT;
2450                 return;
2451         }
2452 }
2453
2454 void e2fsck_pass1(e2fsck_t ctx)
2455 {
2456         e2fsck_pass1_multithread(ctx);
2457 }
2458
2459 #undef FINISH_INODE_LOOP
2460
2461 /*
2462  * When the inode_scan routines call this callback at the end of the
2463  * glock group, call process_inodes.
2464  */
2465 static errcode_t scan_callback(ext2_filsys fs,
2466                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
2467                                dgrp_t group, void * priv_data)
2468 {
2469         struct scan_callback_struct *scan_struct;
2470         e2fsck_t ctx;
2471
2472         scan_struct = (struct scan_callback_struct *) priv_data;
2473         ctx = scan_struct->ctx;
2474
2475         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf);
2476
2477         if (ctx->progress)
2478                 if ((ctx->progress)(ctx, 1, group+1,
2479                                     ctx->fs->group_desc_count))
2480                         return EXT2_ET_CANCEL_REQUESTED;
2481
2482         return 0;
2483 }
2484
2485 /*
2486  * Process the inodes in the "inodes to process" list.
2487  */
2488 static void process_inodes(e2fsck_t ctx, char *block_buf)
2489 {
2490         int                     i;
2491         struct ext2_inode       *old_stashed_inode;
2492         ext2_ino_t              old_stashed_ino;
2493         const char              *old_operation;
2494         char                    buf[80];
2495         struct problem_context  pctx;
2496
2497 #if 0
2498         printf("begin process_inodes: ");
2499 #endif
2500         if (process_inode_count == 0)
2501                 return;
2502         old_operation = ehandler_operation(0);
2503         old_stashed_inode = ctx->stashed_inode;
2504         old_stashed_ino = ctx->stashed_ino;
2505         qsort(inodes_to_process, process_inode_count,
2506                       sizeof(struct process_inode_block), process_inode_cmp);
2507         clear_problem_context(&pctx);
2508         for (i=0; i < process_inode_count; i++) {
2509                 pctx.inode = ctx->stashed_inode =
2510                         (struct ext2_inode *) &inodes_to_process[i].inode;
2511                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2512
2513 #if 0
2514                 printf("%u ", pctx.ino);
2515 #endif
2516                 sprintf(buf, _("reading indirect blocks of inode %u"),
2517                         pctx.ino);
2518                 ehandler_operation(buf);
2519                 check_blocks(ctx, &pctx, block_buf,
2520                              &inodes_to_process[i].ea_ibody_quota);
2521                 if (e2fsck_should_abort(ctx))
2522                         break;
2523         }
2524         ctx->stashed_inode = old_stashed_inode;
2525         ctx->stashed_ino = old_stashed_ino;
2526         process_inode_count = 0;
2527 #if 0
2528         printf("end process inodes\n");
2529 #endif
2530         ehandler_operation(old_operation);
2531 }
2532
2533 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
2534 {
2535         const struct process_inode_block *ib_a =
2536                 (const struct process_inode_block *) a;
2537         const struct process_inode_block *ib_b =
2538                 (const struct process_inode_block *) b;
2539         int     ret;
2540
2541         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
2542                ib_b->inode.i_block[EXT2_IND_BLOCK]);
2543         if (ret == 0)
2544                 /*
2545                  * We only call process_inodes() for non-extent
2546                  * inodes, so it's OK to pass NULL to
2547                  * ext2fs_file_acl_block() here.
2548                  */
2549                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
2550                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
2551         if (ret == 0)
2552                 ret = ib_a->ino - ib_b->ino;
2553         return ret;
2554 }
2555
2556 /*
2557  * Mark an inode as being bad in some what
2558  */
2559 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
2560 {
2561         struct          problem_context pctx;
2562
2563         if (!ctx->inode_bad_map) {
2564                 clear_problem_context(&pctx);
2565
2566                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2567                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
2568                                 "inode_bad_map", &ctx->inode_bad_map);
2569                 if (pctx.errcode) {
2570                         pctx.num = 3;
2571                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2572                         /* Should never get here */
2573                         ctx->flags |= E2F_FLAG_ABORT;
2574                         return;
2575                 }
2576         }
2577         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
2578 }
2579
2580 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino)
2581 {
2582         struct          problem_context pctx;
2583
2584         if (!ctx->casefolded_dirs) {
2585                 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
2586                 if (pctx.errcode)
2587                         goto error;
2588         }
2589         pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
2590         if (pctx.errcode == 0)
2591                 return;
2592 error:
2593         fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
2594         /* Should never get here */
2595         ctx->flags |= E2F_FLAG_ABORT;
2596 }
2597
2598 /*
2599  * This procedure will allocate the inode "bb" (badblock) map table
2600  */
2601 static void alloc_bb_map(e2fsck_t ctx)
2602 {
2603         struct          problem_context pctx;
2604
2605         clear_problem_context(&pctx);
2606         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2607                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
2608                         "inode_bb_map", &ctx->inode_bb_map);
2609         if (pctx.errcode) {
2610                 pctx.num = 4;
2611                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2612                 /* Should never get here */
2613                 ctx->flags |= E2F_FLAG_ABORT;
2614                 return;
2615         }
2616 }
2617
2618 /*
2619  * This procedure will allocate the inode imagic table
2620  */
2621 static void alloc_imagic_map(e2fsck_t ctx)
2622 {
2623         struct          problem_context pctx;
2624
2625         clear_problem_context(&pctx);
2626         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2627                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
2628                         "inode_imagic_map", &ctx->inode_imagic_map);
2629         if (pctx.errcode) {
2630                 pctx.num = 5;
2631                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2632                 /* Should never get here */
2633                 ctx->flags |= E2F_FLAG_ABORT;
2634                 return;
2635         }
2636 }
2637
2638 /*
2639  * Marks a block as in use, setting the dup_map if it's been set
2640  * already.  Called by process_block and process_bad_block.
2641  *
2642  * WARNING: Assumes checks have already been done to make sure block
2643  * is valid.  This is true in both process_block and process_bad_block.
2644  */
2645 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
2646 {
2647         struct          problem_context pctx;
2648
2649         clear_problem_context(&pctx);
2650
2651         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
2652                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
2653                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
2654                         return;
2655                 }
2656                 if (!ctx->block_dup_map) {
2657                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
2658                                         _("multiply claimed block map"),
2659                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
2660                                         &ctx->block_dup_map);
2661                         if (pctx.errcode) {
2662                                 pctx.num = 3;
2663                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
2664                                             &pctx);
2665                                 /* Should never get here */
2666                                 ctx->flags |= E2F_FLAG_ABORT;
2667                                 return;
2668                         }
2669                 }
2670                 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
2671         } else {
2672                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
2673         }
2674 }
2675
2676 /*
2677  * When cluster size is greater than one block, it is caller's responsibility
2678  * to make sure block parameter starts at a cluster boundary.
2679  */
2680 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
2681                                       unsigned int num)
2682 {
2683         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
2684                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
2685         else {
2686                 unsigned int i;
2687
2688                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
2689                         mark_block_used(ctx, block + i);
2690         }
2691 }
2692
2693 /*
2694  * Adjust the extended attribute block's reference counts at the end
2695  * of pass 1, either by subtracting out references for EA blocks that
2696  * are still referenced in ctx->refcount, or by adding references for
2697  * EA blocks that had extra references as accounted for in
2698  * ctx->refcount_extra.
2699  */
2700 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
2701                                     char *block_buf, int adjust_sign)
2702 {
2703         struct ext2_ext_attr_header     *header;
2704         struct problem_context          pctx;
2705         ext2_filsys                     fs = ctx->fs;
2706         blk64_t                         blk;
2707         __u32                           should_be;
2708         ea_value_t                      count;
2709
2710         clear_problem_context(&pctx);
2711
2712         ea_refcount_intr_begin(refcount);
2713         while (1) {
2714                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
2715                         break;
2716                 pctx.blk = blk;
2717                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
2718                                                      pctx.ino);
2719                 if (pctx.errcode) {
2720                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
2721                         return;
2722                 }
2723                 header = (struct ext2_ext_attr_header *) block_buf;
2724                 pctx.blkcount = header->h_refcount;
2725                 should_be = header->h_refcount + adjust_sign * (int)count;
2726                 pctx.num = should_be;
2727                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
2728                         header->h_refcount = should_be;
2729                         pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
2730                                                              block_buf,
2731                                                              pctx.ino);
2732                         if (pctx.errcode) {
2733                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
2734                                             &pctx);
2735                                 continue;
2736                         }
2737                 }
2738         }
2739 }
2740
2741 /*
2742  * Handle processing the extended attribute blocks
2743  */
2744 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
2745                            char *block_buf, struct ea_quota *ea_block_quota)
2746 {
2747         ext2_filsys fs = ctx->fs;
2748         ext2_ino_t      ino = pctx->ino;
2749         struct ext2_inode *inode = pctx->inode;
2750         blk64_t         blk;
2751         char *          end;
2752         struct ext2_ext_attr_header *header;
2753         struct ext2_ext_attr_entry *first, *entry;
2754         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
2755         __u64           quota_inodes = 0;
2756         region_t        region = 0;
2757         int             failed_csum = 0;
2758
2759         ea_block_quota->blocks = 0;
2760         ea_block_quota->inodes = 0;
2761
2762         blk = ext2fs_file_acl_block(fs, inode);
2763         if (blk == 0)
2764                 return 0;
2765
2766         /*
2767          * If the Extended attribute flag isn't set, then a non-zero
2768          * file acl means that the inode is corrupted.
2769          *
2770          * Or if the extended attribute block is an invalid block,
2771          * then the inode is also corrupted.
2772          */
2773         if (!ext2fs_has_feature_xattr(fs->super) ||
2774             (blk < fs->super->s_first_data_block) ||
2775             (blk >= ext2fs_blocks_count(fs->super))) {
2776                 mark_inode_bad(ctx, ino);
2777                 return 0;
2778         }
2779
2780         /* If ea bitmap hasn't been allocated, create it */
2781         if (!ctx->block_ea_map) {
2782                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
2783                                         _("ext attr block map"),
2784                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
2785                                         &ctx->block_ea_map);
2786                 if (pctx->errcode) {
2787                         pctx->num = 2;
2788                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
2789                         ctx->flags |= E2F_FLAG_ABORT;
2790                         return 0;
2791                 }
2792         }
2793
2794         /* Create the EA refcount structure if necessary */
2795         if (!ctx->refcount) {
2796                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
2797                 if (pctx->errcode) {
2798                         pctx->num = 1;
2799                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2800                         ctx->flags |= E2F_FLAG_ABORT;
2801                         return 0;
2802                 }
2803         }
2804
2805 #if 0
2806         /* Debugging text */
2807         printf("Inode %u has EA block %u\n", ino, blk);
2808 #endif
2809
2810         /* Have we seen this EA block before? */
2811         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
2812                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
2813                 ea_block_quota->inodes = 0;
2814
2815                 if (ctx->ea_block_quota_blocks) {
2816                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
2817                                           &quota_blocks);
2818                         if (quota_blocks)
2819                                 ea_block_quota->blocks = quota_blocks;
2820                 }
2821
2822                 if (ctx->ea_block_quota_inodes)
2823                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
2824                                           &ea_block_quota->inodes);
2825
2826                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
2827                         return 1;
2828                 /* Ooops, this EA was referenced more than it stated */
2829                 if (!ctx->refcount_extra) {
2830                         pctx->errcode = ea_refcount_create(0,
2831                                            &ctx->refcount_extra);
2832                         if (pctx->errcode) {
2833                                 pctx->num = 2;
2834                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2835                                 ctx->flags |= E2F_FLAG_ABORT;
2836                                 return 0;
2837                         }
2838                 }
2839                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
2840                 return 1;
2841         }
2842
2843         /*
2844          * OK, we haven't seen this EA block yet.  So we need to
2845          * validate it
2846          */
2847         pctx->blk = blk;
2848         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
2849         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
2850                 pctx->errcode = 0;
2851                 failed_csum = 1;
2852         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
2853                 pctx->errcode = 0;
2854
2855         if (pctx->errcode &&
2856             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
2857                 pctx->errcode = 0;
2858                 goto clear_extattr;
2859         }
2860         header = (struct ext2_ext_attr_header *) block_buf;
2861         pctx->blk = ext2fs_file_acl_block(fs, inode);
2862         if (((ctx->ext_attr_ver == 1) &&
2863              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
2864             ((ctx->ext_attr_ver == 2) &&
2865              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
2866                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
2867                         goto clear_extattr;
2868         }
2869
2870         if (header->h_blocks != 1) {
2871                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
2872                         goto clear_extattr;
2873         }
2874
2875         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
2876                 goto clear_extattr;
2877
2878         region = region_create(0, fs->blocksize);
2879         if (!region) {
2880                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
2881                 ctx->flags |= E2F_FLAG_ABORT;
2882                 return 0;
2883         }
2884         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
2885                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2886                         goto clear_extattr;
2887         }
2888
2889         first = (struct ext2_ext_attr_entry *)(header+1);
2890         end = block_buf + fs->blocksize;
2891         entry = first;
2892         while ((char *)entry < end && *(__u32 *)entry) {
2893                 __u32 hash;
2894
2895                 if (region_allocate(region, (char *)entry - (char *)header,
2896                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
2897                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2898                                 goto clear_extattr;
2899                         break;
2900                 }
2901                 if ((ctx->ext_attr_ver == 1 &&
2902                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
2903                     (ctx->ext_attr_ver == 2 &&
2904                      entry->e_name_index == 0)) {
2905                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
2906                                 goto clear_extattr;
2907                         break;
2908                 }
2909                 if (entry->e_value_inum == 0) {
2910                         if (entry->e_value_offs + entry->e_value_size >
2911                             fs->blocksize) {
2912                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
2913                                         goto clear_extattr;
2914                                 break;
2915                         }
2916                         if (entry->e_value_size &&
2917                             region_allocate(region, entry->e_value_offs,
2918                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
2919                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
2920                                                 pctx))
2921                                         goto clear_extattr;
2922                         }
2923
2924                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
2925                                                           entry->e_value_offs);
2926
2927                         if (entry->e_hash != hash) {
2928                                 pctx->num = entry->e_hash;
2929                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
2930                                         goto clear_extattr;
2931                                 entry->e_hash = hash;
2932                         }
2933                 } else {
2934                         problem_t problem;
2935                         blk64_t entry_quota_blocks;
2936
2937                         problem = check_large_ea_inode(ctx, entry, pctx,
2938                                                        &entry_quota_blocks);
2939                         if (problem && fix_problem(ctx, problem, pctx))
2940                                 goto clear_extattr;
2941
2942                         quota_blocks += entry_quota_blocks;
2943                         quota_inodes++;
2944                 }
2945
2946                 entry = EXT2_EXT_ATTR_NEXT(entry);
2947         }
2948         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
2949                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2950                         goto clear_extattr;
2951         }
2952         region_free(region);
2953
2954         /*
2955          * We only get here if there was no other errors that were fixed.
2956          * If there was a checksum fail, ask to correct it.
2957          */
2958         if (failed_csum &&
2959             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
2960                 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
2961                                                        pctx->ino);
2962                 if (pctx->errcode)
2963                         return 0;
2964         }
2965
2966         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
2967                 if (!ctx->ea_block_quota_blocks) {
2968                         pctx->errcode = ea_refcount_create(0,
2969                                                 &ctx->ea_block_quota_blocks);
2970                         if (pctx->errcode) {
2971                                 pctx->num = 3;
2972                                 goto refcount_fail;
2973                         }
2974                 }
2975                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
2976                                   quota_blocks);
2977         }
2978
2979         if (quota_inodes) {
2980                 if (!ctx->ea_block_quota_inodes) {
2981                         pctx->errcode = ea_refcount_create(0,
2982                                                 &ctx->ea_block_quota_inodes);
2983                         if (pctx->errcode) {
2984                                 pctx->num = 4;
2985 refcount_fail:
2986                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2987                                 ctx->flags |= E2F_FLAG_ABORT;
2988                                 return 0;
2989                         }
2990                 }
2991
2992                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
2993                                   quota_inodes);
2994         }
2995         ea_block_quota->blocks = quota_blocks;
2996         ea_block_quota->inodes = quota_inodes;
2997
2998         inc_ea_inode_refs(ctx, pctx, first, end);
2999         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
3000         mark_block_used(ctx, blk);
3001         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
3002         return 1;
3003
3004 clear_extattr:
3005         if (region)
3006                 region_free(region);
3007         ext2fs_file_acl_block_set(fs, inode, 0);
3008         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
3009         return 0;
3010 }
3011
3012 /* Returns 1 if bad htree, 0 if OK */
3013 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
3014                         ext2_ino_t ino, struct ext2_inode *inode,
3015                         char *block_buf)
3016 {
3017         struct ext2_dx_root_info        *root;
3018         ext2_filsys                     fs = ctx->fs;
3019         errcode_t                       retval;
3020         blk64_t                         blk;
3021
3022         if ((!LINUX_S_ISDIR(inode->i_mode) &&
3023              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
3024             (!ext2fs_has_feature_dir_index(fs->super) &&
3025              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
3026                 return 1;
3027
3028         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
3029
3030         if ((pctx->errcode) ||
3031             (blk == 0) ||
3032             (blk < fs->super->s_first_data_block) ||
3033             (blk >= ext2fs_blocks_count(fs->super))) {
3034                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3035                         return 1;
3036                 else
3037                         return 0;
3038         }
3039
3040         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
3041         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3042                 return 1;
3043
3044         /* XXX should check that beginning matches a directory */
3045         root = (struct ext2_dx_root_info *) (block_buf + 24);
3046
3047         if ((root->reserved_zero || root->info_length < 8) &&
3048             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3049                 return 1;
3050
3051         pctx->num = root->hash_version;
3052         if ((root->hash_version != EXT2_HASH_LEGACY) &&
3053             (root->hash_version != EXT2_HASH_HALF_MD4) &&
3054             (root->hash_version != EXT2_HASH_TEA) &&
3055             (root->hash_version != EXT2_HASH_SIPHASH) &&
3056             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
3057                 return 1;
3058
3059         if (ext4_hash_in_dirent(inode)) {
3060                 if (root->hash_version != EXT2_HASH_SIPHASH &&
3061                     fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
3062                         return 1;
3063         } else {
3064                 if (root->hash_version == EXT2_HASH_SIPHASH &&
3065                    fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
3066                         return 1;
3067         }
3068
3069         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
3070             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
3071                 return 1;
3072
3073         pctx->num = root->indirect_levels;
3074         /* if htree level is clearly too high, consider it to be broken */
3075         if (root->indirect_levels > EXT4_HTREE_LEVEL &&
3076             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3077                 return 1;
3078
3079         /* if level is only maybe too high, LARGE_DIR feature could be unset */
3080         if (root->indirect_levels > ext2_dir_htree_level(fs) &&
3081             !ext2fs_has_feature_largedir(fs->super)) {
3082                 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
3083                 unsigned idx_pb = 1 << (blockbits - 3);
3084
3085                 /* compare inode size/blocks vs. max-sized 2-level htree */
3086                 if (EXT2_I_SIZE(pctx->inode) <
3087                     (idx_pb - 1) * (idx_pb - 2) << blockbits &&
3088                     pctx->inode->i_blocks <
3089                     (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
3090                     fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3091                         return 1;
3092         }
3093
3094         if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
3095             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3096                 ctx->large_dirs++;
3097
3098         return 0;
3099 }
3100
3101 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
3102                         struct ext2_inode *inode, int restart_flag,
3103                         const char *source)
3104 {
3105         inode->i_flags = 0;
3106         inode->i_links_count = 0;
3107         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
3108         inode->i_dtime = ctx->now;
3109
3110         /*
3111          * If a special inode has such rotten block mappings that we
3112          * want to clear the whole inode, be sure to actually zap
3113          * the block maps because i_links_count isn't checked for
3114          * special inodes, and we'll end up right back here the next
3115          * time we run fsck.
3116          */
3117         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
3118                 memset(inode->i_block, 0, sizeof(inode->i_block));
3119
3120         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
3121         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
3122         if (ctx->inode_reg_map)
3123                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
3124         if (ctx->inode_bad_map)
3125                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
3126
3127         /*
3128          * If the inode was partially accounted for before processing
3129          * was aborted, we need to restart the pass 1 scan.
3130          */
3131         ctx->flags |= restart_flag;
3132
3133         if (ino == EXT2_BAD_INO)
3134                 memset(inode, 0, sizeof(struct ext2_inode));
3135
3136         e2fsck_write_inode(ctx, ino, inode, source);
3137 }
3138
3139 /*
3140  * Use the multiple-blocks reclamation code to fix alignment problems in
3141  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
3142  * physical cluster, and we want the block offsets within that cluster to
3143  * line up.
3144  */
3145 static int has_unaligned_cluster_map(e2fsck_t ctx,
3146                                      blk64_t last_pblk, blk64_t last_lblk,
3147                                      blk64_t pblk, blk64_t lblk)
3148 {
3149         blk64_t cluster_mask;
3150
3151         if (!ctx->fs->cluster_ratio_bits)
3152                 return 0;
3153         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
3154
3155         /*
3156          * If the block in the logical cluster doesn't align with the block in
3157          * the physical cluster...
3158          */
3159         if ((lblk & cluster_mask) != (pblk & cluster_mask))
3160                 return 1;
3161
3162         /*
3163          * If we cross a physical cluster boundary within a logical cluster...
3164          */
3165         if (last_pblk && (lblk & cluster_mask) != 0 &&
3166             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
3167             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
3168                 return 1;
3169
3170         return 0;
3171 }
3172
3173 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
3174                              struct process_block_struct *pb,
3175                              blk64_t start_block, blk64_t end_block,
3176                              blk64_t eof_block,
3177                              ext2_extent_handle_t ehandle,
3178                              int try_repairs)
3179 {
3180         struct ext2fs_extent    extent;
3181         blk64_t                 blk, last_lblk;
3182         unsigned int            i, n;
3183         int                     is_dir, is_leaf;
3184         problem_t               problem;
3185         struct ext2_extent_info info;
3186         int                     failed_csum = 0;
3187
3188         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
3189                 failed_csum = 1;
3190
3191         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
3192         if (pctx->errcode)
3193                 return;
3194         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
3195             !pb->eti.force_rebuild) {
3196                 struct extent_tree_level *etl;
3197
3198                 etl = pb->eti.ext_info + info.curr_level;
3199                 etl->num_extents += info.num_entries;
3200                 etl->max_extents += info.max_entries;
3201                 /*
3202                  * Implementation wart: Splitting extent blocks when appending
3203                  * will leave the old block with one free entry.  Therefore
3204                  * unless the node is totally full, pretend that a non-root
3205                  * extent block can hold one fewer entry than it actually does,
3206                  * so that we don't repeatedly rebuild the extent tree.
3207                  */
3208                 if (info.curr_level && info.num_entries < info.max_entries)
3209                         etl->max_extents--;
3210         }
3211
3212         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
3213                                           &extent);
3214         while ((pctx->errcode == 0 ||
3215                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
3216                info.num_entries-- > 0) {
3217                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
3218                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
3219                 last_lblk = extent.e_lblk + extent.e_len - 1;
3220
3221                 problem = 0;
3222                 pctx->blk = extent.e_pblk;
3223                 pctx->blk2 = extent.e_lblk;
3224                 pctx->num = extent.e_len;
3225                 pctx->blkcount = extent.e_lblk + extent.e_len;
3226
3227                 if (extent.e_pblk == 0 ||
3228                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
3229                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
3230                         problem = PR_1_EXTENT_BAD_START_BLK;
3231                 else if (extent.e_lblk < start_block)
3232                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
3233                 else if ((end_block && last_lblk > end_block) &&
3234                          !(last_lblk > eof_block &&
3235                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
3236                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
3237                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
3238                 else if (is_leaf && extent.e_len == 0)
3239                         problem = PR_1_EXTENT_LENGTH_ZERO;
3240                 else if (is_leaf &&
3241                          (extent.e_pblk + extent.e_len) >
3242                          ext2fs_blocks_count(ctx->fs->super))
3243                         problem = PR_1_EXTENT_ENDS_BEYOND;
3244                 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
3245                          !ext2fs_has_feature_largedir(ctx->fs->super) &&
3246                          ((extent.e_lblk + extent.e_len) >
3247                           (1U << (21 - ctx->fs->super->s_log_block_size))))
3248                         problem = PR_1_TOOBIG_DIR;
3249
3250                 if (is_leaf && problem == 0 && extent.e_len > 0) {
3251 #if 0
3252                         printf("extent_region(ino=%u, expect=%llu, "
3253                                "lblk=%llu, len=%u)\n", pb->ino,
3254                                (unsigned long long) pb->next_lblock,
3255                                (unsigned long long) extent.e_lblk,
3256                                extent.e_len);
3257 #endif
3258                         if (extent.e_lblk < pb->next_lblock)
3259                                 problem = PR_1_EXTENT_COLLISION;
3260                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
3261                                 pb->next_lblock = extent.e_lblk + extent.e_len;
3262                 }
3263
3264                 /*
3265                  * Uninitialized blocks in a directory?  Clear the flag and
3266                  * we'll interpret the blocks later.
3267                  */
3268                 if (try_repairs && is_dir && problem == 0 &&
3269                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3270                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
3271                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3272                         pb->inode_modified = 1;
3273                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3274                                                               &extent);
3275                         if (pctx->errcode)
3276                                 return;
3277                         failed_csum = 0;
3278                 }
3279 #ifdef CONFIG_DEVELOPER_FEATURES
3280                 if (try_repairs && !is_dir && problem == 0 &&
3281                     (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
3282                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3283                     fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
3284                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3285                         pb->inode_modified = 1;
3286                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3287                                                               &extent);
3288                         if (pctx->errcode)
3289                                 return;
3290                         failed_csum = 0;
3291                 }
3292 #endif
3293                 if (try_repairs && problem) {
3294 report_problem:
3295                         if (fix_problem(ctx, problem, pctx)) {
3296                                 if (ctx->invalid_bitmaps) {
3297                                         /*
3298                                          * If fsck knows the bitmaps are bad,
3299                                          * skip to the next extent and
3300                                          * try to clear this extent again
3301                                          * after fixing the bitmaps, by
3302                                          * restarting fsck.
3303                                          */
3304                                         pctx->errcode = ext2fs_extent_get(
3305                                                           ehandle,
3306                                                           EXT2_EXTENT_NEXT_SIB,
3307                                                           &extent);
3308                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3309                                         if (pctx->errcode ==
3310                                                     EXT2_ET_NO_CURRENT_NODE) {
3311                                                 pctx->errcode = 0;
3312                                                 break;
3313                                         }
3314                                         continue;
3315                                 }
3316                                 e2fsck_read_bitmaps(ctx);
3317                                 pb->inode_modified = 1;
3318                                 pctx->errcode =
3319                                         ext2fs_extent_delete(ehandle, 0);
3320                                 if (pctx->errcode) {
3321                                         pctx->str = "ext2fs_extent_delete";
3322                                         return;
3323                                 }
3324                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3325                                 if (pctx->errcode &&
3326                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
3327                                         pctx->str = "ext2fs_extent_fix_parents";
3328                                         return;
3329                                 }
3330                                 pctx->errcode = ext2fs_extent_get(ehandle,
3331                                                                   EXT2_EXTENT_CURRENT,
3332                                                                   &extent);
3333                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
3334                                         pctx->errcode = 0;
3335                                         break;
3336                                 }
3337                                 failed_csum = 0;
3338                                 continue;
3339                         }
3340                         goto next;
3341                 }
3342
3343                 if (!is_leaf) {
3344                         blk64_t lblk = extent.e_lblk;
3345                         int next_try_repairs = 1;
3346
3347                         blk = extent.e_pblk;
3348
3349                         /*
3350                          * If this lower extent block collides with critical
3351                          * metadata, don't try to repair the damage.  Pass 1b
3352                          * will reallocate the block; then we can try again.
3353                          */
3354                         if (pb->ino != EXT2_RESIZE_INO &&
3355                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
3356                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3357                                                       extent.e_pblk)) {
3358                                 next_try_repairs = 0;
3359                                 pctx->blk = blk;
3360                                 fix_problem(ctx,
3361                                             PR_1_CRITICAL_METADATA_COLLISION,
3362                                             pctx);
3363                                 if ((ctx->options & E2F_OPT_NO) == 0)
3364                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3365                         }
3366                         pctx->errcode = ext2fs_extent_get(ehandle,
3367                                                   EXT2_EXTENT_DOWN, &extent);
3368                         if (pctx->errcode &&
3369                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
3370                                 pctx->str = "EXT2_EXTENT_DOWN";
3371                                 problem = PR_1_EXTENT_HEADER_INVALID;
3372                                 if (!next_try_repairs)
3373                                         return;
3374                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
3375                                         goto report_problem;
3376                                 return;
3377                         }
3378                         /* The next extent should match this index's logical start */
3379                         if (extent.e_lblk != lblk) {
3380                                 struct ext2_extent_info e_info;
3381
3382                                 pctx->errcode = ext2fs_extent_get_info(ehandle,
3383                                                                        &e_info);
3384                                 if (pctx->errcode) {
3385                                         pctx->str = "ext2fs_extent_get_info";
3386                                         return;
3387                                 }
3388                                 pctx->blk = lblk;
3389                                 pctx->blk2 = extent.e_lblk;
3390                                 pctx->num = e_info.curr_level - 1;
3391                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
3392                                 if (fix_problem(ctx, problem, pctx)) {
3393                                         pb->inode_modified = 1;
3394                                         pctx->errcode =
3395                                                 ext2fs_extent_fix_parents(ehandle);
3396                                         if (pctx->errcode) {
3397                                                 pctx->str = "ext2fs_extent_fix_parents";
3398                                                 return;
3399                                         }
3400                                 }
3401                         }
3402                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
3403                                          last_lblk, eof_block, ehandle,
3404                                          next_try_repairs);
3405                         if (pctx->errcode)
3406                                 return;
3407                         pctx->errcode = ext2fs_extent_get(ehandle,
3408                                                   EXT2_EXTENT_UP, &extent);
3409                         if (pctx->errcode) {
3410                                 pctx->str = "EXT2_EXTENT_UP";
3411                                 return;
3412                         }
3413                         mark_block_used(ctx, blk);
3414                         pb->num_blocks++;
3415                         goto next;
3416                 }
3417
3418                 if ((pb->previous_block != 0) &&
3419                     (pb->previous_block+1 != extent.e_pblk)) {
3420                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3421                                 char type = '?';
3422
3423                                 if (pb->is_dir)
3424                                         type = 'd';
3425                                 else if (pb->is_reg)
3426                                         type = 'f';
3427
3428                                 printf(("%6lu(%c): expecting %6lu "
3429                                         "actual extent "
3430                                         "phys %6lu log %lu len %lu\n"),
3431                                        (unsigned long) pctx->ino, type,
3432                                        (unsigned long) pb->previous_block+1,
3433                                        (unsigned long) extent.e_pblk,
3434                                        (unsigned long) extent.e_lblk,
3435                                        (unsigned long) extent.e_len);
3436                         }
3437                         pb->fragmented = 1;
3438                 }
3439                 /*
3440                  * If we notice a gap in the logical block mappings of an
3441                  * extent-mapped directory, offer to close the hole by
3442                  * moving the logical block down, otherwise we'll go mad in
3443                  * pass 3 allocating empty directory blocks to fill the hole.
3444                  */
3445                 if (try_repairs && is_dir &&
3446                     pb->last_block + 1 < extent.e_lblk) {
3447                         blk64_t new_lblk;
3448
3449                         new_lblk = pb->last_block + 1;
3450                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3451                                 new_lblk = ((new_lblk +
3452                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3453                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3454                                            (extent.e_pblk &
3455                                             EXT2FS_CLUSTER_MASK(ctx->fs));
3456                         pctx->blk = extent.e_lblk;
3457                         pctx->blk2 = new_lblk;
3458                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3459                                 extent.e_lblk = new_lblk;
3460                                 pb->inode_modified = 1;
3461                                 pctx->errcode = ext2fs_extent_replace(ehandle,
3462                                                                 0, &extent);
3463                                 if (pctx->errcode) {
3464                                         pctx->errcode = 0;
3465                                         goto alloc_later;
3466                                 }
3467                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3468                                 if (pctx->errcode)
3469                                         goto failed_add_dir_block;
3470                                 pctx->errcode = ext2fs_extent_goto(ehandle,
3471                                                                 extent.e_lblk);
3472                                 if (pctx->errcode)
3473                                         goto failed_add_dir_block;
3474                                 last_lblk = extent.e_lblk + extent.e_len - 1;
3475                                 failed_csum = 0;
3476                         }
3477                 }
3478 alloc_later:
3479                 if (is_dir) {
3480                         while (++pb->last_db_block <
3481                                (e2_blkcnt_t) extent.e_lblk) {
3482                                 pctx->errcode = ext2fs_add_dir_block2(
3483                                                         ctx->fs->dblist,
3484                                                         pb->ino, 0,
3485                                                         pb->last_db_block);
3486                                 if (pctx->errcode) {
3487                                         pctx->blk = 0;
3488                                         pctx->num = pb->last_db_block;
3489                                         goto failed_add_dir_block;
3490                                 }
3491                         }
3492
3493                         for (i = 0; i < extent.e_len; i++) {
3494                                 pctx->errcode = ext2fs_add_dir_block2(
3495                                                         ctx->fs->dblist,
3496                                                         pctx->ino,
3497                                                         extent.e_pblk + i,
3498                                                         extent.e_lblk + i);
3499                                 if (pctx->errcode) {
3500                                         pctx->blk = extent.e_pblk + i;
3501                                         pctx->num = extent.e_lblk + i;
3502                                 failed_add_dir_block:
3503                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3504                                         /* Should never get here */
3505                                         ctx->flags |= E2F_FLAG_ABORT;
3506                                         return;
3507                                 }
3508                         }
3509                         if (extent.e_len > 0)
3510                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3511                 }
3512                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3513                                               pb->last_block,
3514                                               extent.e_pblk,
3515                                               extent.e_lblk)) {
3516                         for (i = 0; i < extent.e_len; i++) {
3517                                 pctx->blk = extent.e_lblk + i;
3518                                 pctx->blk2 = extent.e_pblk + i;
3519                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3520                                 mark_block_used(ctx, extent.e_pblk + i);
3521                                 mark_block_used(ctx, extent.e_pblk + i);
3522                         }
3523                 }
3524
3525                 /*
3526                  * Check whether first cluster got marked in previous iteration.
3527                  */
3528                 if (ctx->fs->cluster_ratio_bits &&
3529                     pb->previous_block &&
3530                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
3531                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
3532                         /* Set blk to the beginning of next cluster. */
3533                         blk = EXT2FS_C2B(
3534                                 ctx->fs,
3535                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
3536                 else
3537                         /* Set blk to the beginning of current cluster. */
3538                         blk = EXT2FS_C2B(ctx->fs,
3539                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
3540
3541                 if (blk < extent.e_pblk + extent.e_len) {
3542                         mark_blocks_used(ctx, blk,
3543                                          extent.e_pblk + extent.e_len - blk);
3544                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
3545                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
3546                         pb->num_blocks += n;
3547                 }
3548                 pb->last_block = extent.e_lblk + extent.e_len - 1;
3549                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
3550                 start_block = pb->last_block = last_lblk;
3551                 if (is_leaf && !is_dir &&
3552                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
3553                         pb->last_init_lblock = last_lblk;
3554         next:
3555                 pctx->errcode = ext2fs_extent_get(ehandle,
3556                                                   EXT2_EXTENT_NEXT_SIB,
3557                                                   &extent);
3558         }
3559
3560         /* Failed csum but passes checks?  Ask to fix checksum. */
3561         if (failed_csum &&
3562             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
3563                 pb->inode_modified = 1;
3564                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
3565                 if (pctx->errcode)
3566                         return;
3567         }
3568
3569         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
3570                 pctx->errcode = 0;
3571 }
3572
3573 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
3574                                  struct process_block_struct *pb)
3575 {
3576         struct ext2_extent_info info;
3577         struct ext2_inode       *inode = pctx->inode;
3578         ext2_extent_handle_t    ehandle;
3579         ext2_filsys             fs = ctx->fs;
3580         ext2_ino_t              ino = pctx->ino;
3581         errcode_t               retval;
3582         blk64_t                 eof_lblk;
3583         struct ext3_extent_header       *eh;
3584
3585         /* Check for a proper extent header... */
3586         eh = (struct ext3_extent_header *) &inode->i_block[0];
3587         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
3588         if (retval) {
3589                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
3590                         e2fsck_clear_inode(ctx, ino, inode, 0,
3591                                            "check_blocks_extents");
3592                 pctx->errcode = 0;
3593                 return;
3594         }
3595
3596         /* ...since this function doesn't fail if i_block is zeroed. */
3597         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
3598         if (pctx->errcode) {
3599                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
3600                         e2fsck_clear_inode(ctx, ino, inode, 0,
3601                                            "check_blocks_extents");
3602                 pctx->errcode = 0;
3603                 return;
3604         }
3605
3606         retval = ext2fs_extent_get_info(ehandle, &info);
3607         if (retval == 0) {
3608                 int max_depth = info.max_depth;
3609
3610                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
3611                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
3612                 ctx->extent_depth_count[max_depth]++;
3613         }
3614
3615         /* Check maximum extent depth */
3616         pctx->blk = info.max_depth;
3617         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
3618         if (pctx->blk2 < pctx->blk &&
3619             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
3620                 pb->eti.force_rebuild = 1;
3621
3622         /* Can we collect extent tree level stats? */
3623         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
3624         if (pctx->blk2 > pctx->blk)
3625                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
3626         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
3627         pb->eti.ino = pb->ino;
3628
3629         pb->next_lblock = 0;
3630
3631         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
3632                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
3633         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
3634         if (pctx->errcode &&
3635             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
3636                 pb->num_blocks = 0;
3637                 inode->i_blocks = 0;
3638                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3639                                    "check_blocks_extents");
3640                 pctx->errcode = 0;
3641         }
3642         ext2fs_extent_free(ehandle);
3643
3644         /* Rebuild unless it's a dir and we're rehashing it */
3645         if (LINUX_S_ISDIR(inode->i_mode) &&
3646             e2fsck_dir_will_be_rehashed(ctx, ino))
3647                 return;
3648
3649         if (ctx->options & E2F_OPT_CONVERT_BMAP)
3650                 e2fsck_rebuild_extents_later(ctx, ino);
3651         else
3652                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
3653 }
3654
3655 /*
3656  * In fact we don't need to check blocks for an inode with inline data
3657  * because this inode doesn't have any blocks.  In this function all
3658  * we need to do is add this inode into dblist when it is a directory.
3659  */
3660 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
3661                                      struct process_block_struct *pb)
3662 {
3663         int     flags;
3664         size_t  inline_data_size = 0;
3665
3666         if (!pb->is_dir) {
3667                 pctx->errcode = 0;
3668                 return;
3669         }
3670
3671         /* Process the dirents in i_block[] as the "first" block. */
3672         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
3673         if (pctx->errcode)
3674                 goto err;
3675
3676         /* Process the dirents in the EA as a "second" block. */
3677         flags = ctx->fs->flags;
3678         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3679         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
3680                                                 &inline_data_size);
3681         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3682                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3683         if (pctx->errcode) {
3684                 pctx->errcode = 0;
3685                 return;
3686         }
3687
3688         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
3689                 return;
3690
3691         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
3692         if (pctx->errcode)
3693                 goto err;
3694
3695         return;
3696 err:
3697         pctx->blk = 0;
3698         pctx->num = 0;
3699         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3700         ctx->flags |= E2F_FLAG_ABORT;
3701 }
3702
3703 /*
3704  * This subroutine is called on each inode to account for all of the
3705  * blocks used by that inode.
3706  */
3707 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
3708                          char *block_buf, const struct ea_quota *ea_ibody_quota)
3709 {
3710         ext2_filsys fs = ctx->fs;
3711         struct process_block_struct pb;
3712         ext2_ino_t      ino = pctx->ino;
3713         struct ext2_inode *inode = pctx->inode;
3714         unsigned        bad_size = 0;
3715         int             dirty_inode = 0;
3716         int             extent_fs;
3717         int             inlinedata_fs;
3718         __u64           size;
3719         struct ea_quota ea_block_quota;
3720
3721         pb.ino = ino;
3722         pb.num_blocks = EXT2FS_B2C(ctx->fs,
3723                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
3724         pb.last_block = ~0;
3725         pb.last_init_lblock = -1;
3726         pb.last_db_block = -1;
3727         pb.num_illegal_blocks = 0;
3728         pb.suppress = 0; pb.clear = 0;
3729         pb.fragmented = 0;
3730         pb.compressed = 0;
3731         pb.previous_block = 0;
3732         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
3733         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
3734         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
3735         pb.inode = inode;
3736         pb.pctx = pctx;
3737         pb.ctx = ctx;
3738         pb.inode_modified = 0;
3739         pb.eti.force_rebuild = 0;
3740         pctx->ino = ino;
3741         pctx->errcode = 0;
3742
3743         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
3744         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
3745
3746         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
3747                 if (e2fsck_should_abort(ctx))
3748                         goto out;
3749                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
3750         }
3751
3752         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
3753                 check_blocks_inline_data(ctx, pctx, &pb);
3754         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
3755                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
3756                         check_blocks_extents(ctx, pctx, &pb);
3757                 else {
3758                         int flags;
3759                         /*
3760                          * If we've modified the inode, write it out before
3761                          * iterate() tries to use it.
3762                          */
3763                         if (dirty_inode) {
3764                                 e2fsck_write_inode(ctx, ino, inode,
3765                                                    "check_blocks");
3766                                 dirty_inode = 0;
3767                         }
3768                         flags = fs->flags;
3769                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3770                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
3771                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
3772                                                 block_buf, process_block, &pb);
3773                         /*
3774                          * We do not have uninitialized extents in non extent
3775                          * files.
3776                          */
3777                         pb.last_init_lblock = pb.last_block;
3778                         /*
3779                          * If iterate() changed a block mapping, we have to
3780                          * re-read the inode.  If we decide to clear the
3781                          * inode after clearing some stuff, we'll re-write the
3782                          * bad mappings into the inode!
3783                          */
3784                         if (pb.inode_modified)
3785                                 e2fsck_read_inode(ctx, ino, inode,
3786                                                   "check_blocks");
3787                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3788                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3789
3790                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
3791 #ifdef DEBUG
3792                                 printf("bmap rebuild ino=%d\n", ino);
3793 #endif
3794                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
3795                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
3796                                         e2fsck_rebuild_extents_later(ctx, ino);
3797                         }
3798                 }
3799         }
3800         end_problem_latch(ctx, PR_LATCH_BLOCK);
3801         end_problem_latch(ctx, PR_LATCH_TOOBIG);
3802         if (e2fsck_should_abort(ctx))
3803                 goto out;
3804         if (pctx->errcode)
3805                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
3806
3807         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
3808                 if (LINUX_S_ISDIR(inode->i_mode))
3809                         ctx->fs_fragmented_dir++;
3810                 else
3811                         ctx->fs_fragmented++;
3812         }
3813
3814         if (pb.clear) {
3815                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3816                                    "check_blocks");
3817                 return;
3818         }
3819
3820         if (inode->i_flags & EXT2_INDEX_FL) {
3821                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
3822                         inode->i_flags &= ~EXT2_INDEX_FL;
3823                         dirty_inode++;
3824                 } else {
3825                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
3826                 }
3827         }
3828
3829         if (!pb.num_blocks && pb.is_dir &&
3830             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
3831                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
3832                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
3833                         ctx->fs_directory_count--;
3834                         return;
3835                 }
3836         }
3837
3838         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
3839             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
3840             !(inode->i_flags & EXT4_EA_INODE_FL)) {
3841                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
3842                                ino,
3843                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
3844                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
3845                                   ino, (ea_ibody_quota ?
3846                                         ea_ibody_quota->inodes : 0) +
3847                                                 ea_block_quota.inodes + 1);
3848         }
3849
3850         if (!ext2fs_has_feature_huge_file(fs->super) ||
3851             !(inode->i_flags & EXT4_HUGE_FILE_FL))
3852                 pb.num_blocks *= (fs->blocksize / 512);
3853         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
3854 #if 0
3855         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
3856                ino, inode->i_size, (unsigned long long) pb.last_block,
3857                (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
3858                (unsigned long long) pb.num_blocks);
3859 #endif
3860         size = EXT2_I_SIZE(inode);
3861         if (pb.is_dir) {
3862                 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
3863                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
3864                         int flags;
3865                         size_t sz = 0;
3866                         errcode_t err;
3867
3868                         flags = ctx->fs->flags;
3869                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3870                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
3871                                                       &sz);
3872                         ctx->fs->flags = (flags &
3873                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3874                                          (ctx->fs->flags &
3875                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3876                         if (err || sz != size) {
3877                                 bad_size = 7;
3878                                 pctx->num = sz;
3879                         }
3880                 } else if (size & (fs->blocksize - 1))
3881                         bad_size = 5;
3882                 else if (nblock > (pb.last_block + 1))
3883                         bad_size = 1;
3884                 else if (nblock < (pb.last_block + 1)) {
3885                         if (((pb.last_block + 1) - nblock) >
3886                             fs->super->s_prealloc_dir_blocks)
3887                                 bad_size = 2;
3888                 }
3889         } else {
3890                 if ((pb.last_init_lblock >= 0) &&
3891                     /* Do not allow initialized allocated blocks past i_size*/
3892                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
3893                     !(inode->i_flags & EXT4_VERITY_FL))
3894                         bad_size = 3;
3895                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3896                          size > ext2_max_sizes[fs->super->s_log_block_size])
3897                         /* too big for a direct/indirect-mapped file */
3898                         bad_size = 4;
3899                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3900                          size >
3901                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
3902                         /* too big for an extent-based file - 32bit ee_block */
3903                         bad_size = 6;
3904         }
3905         /* i_size for symlinks is checked elsewhere */
3906         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
3907                 /* Did inline_data set pctx->num earlier? */
3908                 if (bad_size != 7)
3909                         pctx->num = (pb.last_block + 1) * fs->blocksize;
3910                 pctx->group = bad_size;
3911                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
3912                         ext2fs_inode_size_set(fs, inode, pctx->num);
3913                         if (EXT2_I_SIZE(inode) == 0 &&
3914                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
3915                                 memset(inode->i_block, 0,
3916                                        sizeof(inode->i_block));
3917                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
3918                         }
3919                         dirty_inode++;
3920                 }
3921                 pctx->num = 0;
3922         }
3923         if (LINUX_S_ISREG(inode->i_mode) &&
3924             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3925                 ctx->large_files++;
3926         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
3927             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
3928              (ext2fs_has_feature_huge_file(fs->super) &&
3929               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
3930               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
3931                 pctx->num = pb.num_blocks;
3932                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
3933                         inode->i_blocks = pb.num_blocks;
3934                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
3935                         dirty_inode++;
3936                 }
3937                 pctx->num = 0;
3938         }
3939
3940         /*
3941          * The kernel gets mad if we ask it to allocate bigalloc clusters to
3942          * a block mapped file, so rebuild it as an extent file.  We can skip
3943          * symlinks because they're never rewritten.
3944          */
3945         if (ext2fs_has_feature_bigalloc(fs->super) &&
3946             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
3947             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
3948             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
3949             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
3950             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
3951                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
3952                 if (pctx->errcode)
3953                         goto out;
3954         }
3955
3956         if (ctx->dirs_to_hash && pb.is_dir &&
3957             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
3958             !(inode->i_flags & EXT2_INDEX_FL) &&
3959             ((inode->i_size / fs->blocksize) >= 3))
3960                 e2fsck_rehash_dir_later(ctx, ino);
3961
3962 out:
3963         if (dirty_inode)
3964                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
3965 }
3966
3967 #if 0
3968 /*
3969  * Helper function called by process block when an illegal block is
3970  * found.  It returns a description about why the block is illegal
3971  */
3972 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
3973 {
3974         blk64_t super;
3975         int     i;
3976         static char     problem[80];
3977
3978         super = fs->super->s_first_data_block;
3979         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
3980         if (block < super) {
3981                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
3982                 return(problem);
3983         } else if (block >= ext2fs_blocks_count(fs->super)) {
3984                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
3985                 return(problem);
3986         }
3987         for (i = 0; i < fs->group_desc_count; i++) {
3988                 if (block == super) {
3989                         sprintf(problem, "is the superblock in group %d", i);
3990                         break;
3991                 }
3992                 if (block > super &&
3993                     block <= (super + fs->desc_blocks)) {
3994                         sprintf(problem, "is in the group descriptors "
3995                                 "of group %d", i);
3996                         break;
3997                 }
3998                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
3999                         sprintf(problem, "is the block bitmap of group %d", i);
4000                         break;
4001                 }
4002                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
4003                         sprintf(problem, "is the inode bitmap of group %d", i);
4004                         break;
4005                 }
4006                 if (block >= ext2fs_inode_table_loc(fs, i) &&
4007                     (block < ext2fs_inode_table_loc(fs, i)
4008                      + fs->inode_blocks_per_group)) {
4009                         sprintf(problem, "is in the inode table of group %d",
4010                                 i);
4011                         break;
4012                 }
4013                 super += fs->super->s_blocks_per_group;
4014         }
4015         return(problem);
4016 }
4017 #endif
4018
4019 /*
4020  * This is a helper function for check_blocks().
4021  */
4022 static int process_block(ext2_filsys fs,
4023                   blk64_t       *block_nr,
4024                   e2_blkcnt_t blockcnt,
4025                   blk64_t ref_block EXT2FS_ATTR((unused)),
4026                   int ref_offset EXT2FS_ATTR((unused)),
4027                   void *priv_data)
4028 {
4029         struct process_block_struct *p;
4030         struct problem_context *pctx;
4031         blk64_t blk = *block_nr;
4032         int     ret_code = 0;
4033         problem_t       problem = 0;
4034         e2fsck_t        ctx;
4035
4036         p = (struct process_block_struct *) priv_data;
4037         pctx = p->pctx;
4038         ctx = p->ctx;
4039
4040         /*
4041          * For a directory, add logical block zero for processing even if it's
4042          * not mapped or we'll be perennially stuck with broken "." and ".."
4043          * entries.
4044          */
4045         if (p->is_dir && blockcnt == 0 && blk == 0) {
4046                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
4047                 if (pctx->errcode) {
4048                         pctx->blk = blk;
4049                         pctx->num = blockcnt;
4050                         goto failed_add_dir_block;
4051                 }
4052                 p->last_db_block++;
4053         }
4054
4055         if (blk == 0)
4056                 return 0;
4057
4058 #if 0
4059         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
4060                blockcnt);
4061 #endif
4062
4063         /*
4064          * Simplistic fragmentation check.  We merely require that the
4065          * file be contiguous.  (Which can never be true for really
4066          * big files that are greater than a block group.)
4067          */
4068         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
4069                 if (p->previous_block+1 != blk) {
4070                         if (ctx->options & E2F_OPT_FRAGCHECK) {
4071                                 char type = '?';
4072
4073                                 if (p->is_dir)
4074                                         type = 'd';
4075                                 else if (p->is_reg)
4076                                         type = 'f';
4077
4078                                 printf(_("%6lu(%c): expecting %6lu "
4079                                          "got phys %6lu (blkcnt %lld)\n"),
4080                                        (unsigned long) pctx->ino, type,
4081                                        (unsigned long) p->previous_block+1,
4082                                        (unsigned long) blk,
4083                                        (long long) blockcnt);
4084                         }
4085                         p->fragmented = 1;
4086                 }
4087         }
4088
4089         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
4090             !pctx->inode->i_size_high &&
4091             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
4092                 problem = PR_1_TOOBIG_DIR;
4093         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
4094                 problem = PR_1_TOOBIG_DIR;
4095         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
4096                 problem = PR_1_TOOBIG_REG;
4097         if (!p->is_dir && !p->is_reg && blockcnt > 0)
4098                 problem = PR_1_TOOBIG_SYMLINK;
4099
4100         if (blk < fs->super->s_first_data_block ||
4101             blk >= ext2fs_blocks_count(fs->super))
4102                 problem = PR_1_ILLEGAL_BLOCK_NUM;
4103
4104         /*
4105          * If this IND/DIND/TIND block is squatting atop some critical metadata
4106          * (group descriptors, superblock, bitmap, inode table), any write to
4107          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
4108          * fix that and restart fsck.
4109          */
4110         if (blockcnt < 0 &&
4111             p->ino != EXT2_RESIZE_INO &&
4112             blk < ctx->fs->super->s_blocks_count &&
4113             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
4114                 pctx->blk = blk;
4115                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
4116                 if ((ctx->options & E2F_OPT_NO) == 0)
4117                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4118         }
4119
4120         if (problem) {
4121                 p->num_illegal_blocks++;
4122                 /*
4123                  * A bit of subterfuge here -- we're trying to fix a block
4124                  * mapping, but the IND/DIND/TIND block could have collided
4125                  * with some critical metadata.  So, fix the in-core mapping so
4126                  * iterate won't go insane, but return 0 instead of
4127                  * BLOCK_CHANGED so that it won't write the remapping out to
4128                  * our multiply linked block.
4129                  *
4130                  * Even if we previously determined that an *IND block
4131                  * conflicts with critical metadata, we must still try to
4132                  * iterate the *IND block as if it is an *IND block to find and
4133                  * mark the blocks it points to.  Better to be overly cautious
4134                  * with the used_blocks map so that we don't move the *IND
4135                  * block to a block that's really in use!
4136                  */
4137                 if (p->ino != EXT2_RESIZE_INO &&
4138                     ref_block != 0 &&
4139                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4140                                               ref_block)) {
4141                         *block_nr = 0;
4142                         return 0;
4143                 }
4144                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
4145                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
4146                                 p->clear = 1;
4147                                 return BLOCK_ABORT;
4148                         }
4149                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
4150                                 p->suppress = 1;
4151                                 set_latch_flags(PR_LATCH_BLOCK,
4152                                                 PRL_SUPPRESS, 0);
4153                         }
4154                 }
4155                 pctx->blk = blk;
4156                 pctx->blkcount = blockcnt;
4157                 if (fix_problem(ctx, problem, pctx)) {
4158                         blk = *block_nr = 0;
4159                         ret_code = BLOCK_CHANGED;
4160                         p->inode_modified = 1;
4161                         /*
4162                          * If the directory block is too big and is beyond the
4163                          * end of the FS, don't bother trying to add it for
4164                          * processing -- the kernel would never have created a
4165                          * directory this large, and we risk an ENOMEM abort.
4166                          * In any case, the toobig handler for extent-based
4167                          * directories also doesn't feed toobig blocks to
4168                          * pass 2.
4169                          */
4170                         if (problem == PR_1_TOOBIG_DIR)
4171                                 return ret_code;
4172                         goto mark_dir;
4173                 } else
4174                         return 0;
4175         }
4176
4177         if (p->ino == EXT2_RESIZE_INO) {
4178                 /*
4179                  * The resize inode has already be sanity checked
4180                  * during pass #0 (the superblock checks).  All we
4181                  * have to do is mark the double indirect block as
4182                  * being in use; all of the other blocks are handled
4183                  * by mark_table_blocks()).
4184                  */
4185                 if (blockcnt == BLOCK_COUNT_DIND)
4186                         mark_block_used(ctx, blk);
4187                 p->num_blocks++;
4188         } else if (!(ctx->fs->cluster_ratio_bits &&
4189                      p->previous_block &&
4190                      (EXT2FS_B2C(ctx->fs, blk) ==
4191                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
4192                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
4193                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
4194                 mark_block_used(ctx, blk);
4195                 p->num_blocks++;
4196         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
4197                                              p->last_block, blk, blockcnt)) {
4198                 pctx->blk = blockcnt;
4199                 pctx->blk2 = blk;
4200                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4201                 mark_block_used(ctx, blk);
4202                 mark_block_used(ctx, blk);
4203         }
4204         if (blockcnt >= 0)
4205                 p->last_block = blockcnt;
4206         p->previous_block = blk;
4207 mark_dir:
4208         if (p->is_dir && (blockcnt >= 0)) {
4209                 while (++p->last_db_block < blockcnt) {
4210                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
4211                                                               p->ino, 0,
4212                                                               p->last_db_block);
4213                         if (pctx->errcode) {
4214                                 pctx->blk = 0;
4215                                 pctx->num = p->last_db_block;
4216                                 goto failed_add_dir_block;
4217                         }
4218                 }
4219                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
4220                                                       blk, blockcnt);
4221                 if (pctx->errcode) {
4222                         pctx->blk = blk;
4223                         pctx->num = blockcnt;
4224                 failed_add_dir_block:
4225                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4226                         /* Should never get here */
4227                         ctx->flags |= E2F_FLAG_ABORT;
4228                         return BLOCK_ABORT;
4229                 }
4230         }
4231         return ret_code;
4232 }
4233
4234 static int process_bad_block(ext2_filsys fs,
4235                       blk64_t *block_nr,
4236                       e2_blkcnt_t blockcnt,
4237                       blk64_t ref_block EXT2FS_ATTR((unused)),
4238                       int ref_offset EXT2FS_ATTR((unused)),
4239                       void *priv_data)
4240 {
4241         struct process_block_struct *p;
4242         blk64_t         blk = *block_nr;
4243         blk64_t         first_block;
4244         dgrp_t          i;
4245         struct problem_context *pctx;
4246         e2fsck_t        ctx;
4247
4248         if (!blk)
4249                 return 0;
4250
4251         p = (struct process_block_struct *) priv_data;
4252         ctx = p->ctx;
4253         pctx = p->pctx;
4254
4255         pctx->ino = EXT2_BAD_INO;
4256         pctx->blk = blk;
4257         pctx->blkcount = blockcnt;
4258
4259         if ((blk < fs->super->s_first_data_block) ||
4260             (blk >= ext2fs_blocks_count(fs->super))) {
4261                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
4262                         *block_nr = 0;
4263                         return BLOCK_CHANGED;
4264                 } else
4265                         return 0;
4266         }
4267
4268         if (blockcnt < 0) {
4269                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
4270                         p->bbcheck = 1;
4271                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
4272                                 *block_nr = 0;
4273                                 return BLOCK_CHANGED;
4274                         }
4275                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4276                                                     blk)) {
4277                         p->bbcheck = 1;
4278                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
4279                                         pctx)) {
4280                                 *block_nr = 0;
4281                                 return BLOCK_CHANGED;
4282                         }
4283                         if (e2fsck_should_abort(ctx))
4284                                 return BLOCK_ABORT;
4285                 } else
4286                         mark_block_used(ctx, blk);
4287                 return 0;
4288         }
4289 #if 0
4290         printf ("DEBUG: Marking %u as bad.\n", blk);
4291 #endif
4292         ctx->fs_badblocks_count++;
4293         /*
4294          * If the block is not used, then mark it as used and return.
4295          * If it is already marked as found, this must mean that
4296          * there's an overlap between the filesystem table blocks
4297          * (bitmaps and inode table) and the bad block list.
4298          */
4299         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
4300                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4301                 return 0;
4302         }
4303         /*
4304          * Try to find the where the filesystem block was used...
4305          */
4306         first_block = fs->super->s_first_data_block;
4307
4308         for (i = 0; i < fs->group_desc_count; i++ ) {
4309                 pctx->group = i;
4310                 pctx->blk = blk;
4311                 if (!ext2fs_bg_has_super(fs, i))
4312                         goto skip_super;
4313                 if (blk == first_block) {
4314                         if (i == 0) {
4315                                 if (fix_problem(ctx,
4316                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
4317                                                 pctx)) {
4318                                         *block_nr = 0;
4319                                         return BLOCK_CHANGED;
4320                                 }
4321                                 return 0;
4322                         }
4323                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
4324                         return 0;
4325                 }
4326                 if ((blk > first_block) &&
4327                     (blk <= first_block + fs->desc_blocks)) {
4328                         if (i == 0) {
4329                                 pctx->blk = *block_nr;
4330                                 if (fix_problem(ctx,
4331                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
4332                                         *block_nr = 0;
4333                                         return BLOCK_CHANGED;
4334                                 }
4335                                 return 0;
4336                         }
4337                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
4338                         return 0;
4339                 }
4340         skip_super:
4341                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
4342                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
4343                                 ctx->invalid_block_bitmap_flag[i]++;
4344                                 ctx->invalid_bitmaps++;
4345                         }
4346                         return 0;
4347                 }
4348                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
4349                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
4350                                 ctx->invalid_inode_bitmap_flag[i]++;
4351                                 ctx->invalid_bitmaps++;
4352                         }
4353                         return 0;
4354                 }
4355                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
4356                     (blk < (ext2fs_inode_table_loc(fs, i) +
4357                             fs->inode_blocks_per_group))) {
4358                         /*
4359                          * If there are bad blocks in the inode table,
4360                          * the inode scan code will try to do
4361                          * something reasonable automatically.
4362                          */
4363                         return 0;
4364                 }
4365                 first_block += fs->super->s_blocks_per_group;
4366         }
4367         /*
4368          * If we've gotten to this point, then the only
4369          * possibility is that the bad block inode meta data
4370          * is using a bad block.
4371          */
4372         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
4373             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
4374             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
4375                 p->bbcheck = 1;
4376                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
4377                         *block_nr = 0;
4378                         return BLOCK_CHANGED;
4379                 }
4380                 if (e2fsck_should_abort(ctx))
4381                         return BLOCK_ABORT;
4382                 return 0;
4383         }
4384
4385         pctx->group = -1;
4386
4387         /* Warn user that the block wasn't claimed */
4388         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
4389
4390         return 0;
4391 }
4392
4393 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
4394                             const char *name, int num, blk64_t *new_block)
4395 {
4396         ext2_filsys fs = ctx->fs;
4397         dgrp_t          last_grp;
4398         blk64_t         old_block = *new_block;
4399         blk64_t         last_block;
4400         dgrp_t          flexbg;
4401         unsigned        flexbg_size;
4402         int             i, is_flexbg;
4403         char            *buf;
4404         struct problem_context  pctx;
4405
4406         clear_problem_context(&pctx);
4407
4408         pctx.group = group;
4409         pctx.blk = old_block;
4410         pctx.str = name;
4411
4412         /*
4413          * For flex_bg filesystems, first try to allocate the metadata
4414          * within the flex_bg, and if that fails then try finding the
4415          * space anywhere in the filesystem.
4416          */
4417         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
4418         if (is_flexbg) {
4419                 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
4420                 flexbg = group / flexbg_size;
4421                 first_block = ext2fs_group_first_block2(fs,
4422                                                         flexbg_size * flexbg);
4423                 last_grp = group | (flexbg_size - 1);
4424                 if (last_grp >= fs->group_desc_count)
4425                         last_grp = fs->group_desc_count - 1;
4426                 last_block = ext2fs_group_last_block2(fs, last_grp);
4427         } else
4428                 last_block = ext2fs_group_last_block2(fs, group);
4429         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
4430                                                num, ctx->block_found_map,
4431                                                new_block);
4432         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
4433                 pctx.errcode = ext2fs_get_free_blocks2(fs,
4434                                 fs->super->s_first_data_block,
4435                                 ext2fs_blocks_count(fs->super),
4436                                 num, ctx->block_found_map, new_block);
4437         if (pctx.errcode) {
4438                 pctx.num = num;
4439                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4440                 ext2fs_unmark_valid(fs);
4441                 ctx->flags |= E2F_FLAG_ABORT;
4442                 return;
4443         }
4444         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4445         if (pctx.errcode) {
4446                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4447                 ext2fs_unmark_valid(fs);
4448                 ctx->flags |= E2F_FLAG_ABORT;
4449                 return;
4450         }
4451         ext2fs_mark_super_dirty(fs);
4452         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4453         pctx.blk2 = *new_block;
4454         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4455                           PR_1_RELOC_TO), &pctx);
4456         pctx.blk2 = 0;
4457         for (i = 0; i < num; i++) {
4458                 pctx.blk = i;
4459                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4460                 if (old_block) {
4461                         pctx.errcode = io_channel_read_blk64(fs->io,
4462                                    old_block + i, 1, buf);
4463                         if (pctx.errcode)
4464                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4465                         pctx.blk = (*new_block) + i;
4466                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4467                                                               1, buf);
4468                 } else {
4469                         pctx.blk = (*new_block) + i;
4470                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4471                                                            NULL, NULL);
4472                 }
4473
4474                 if (pctx.errcode)
4475                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4476         }
4477         ext2fs_free_mem(&buf);
4478 }
4479
4480 /*
4481  * This routine gets called at the end of pass 1 if bad blocks are
4482  * detected in the superblock, group descriptors, inode_bitmaps, or
4483  * block bitmaps.  At this point, all of the blocks have been mapped
4484  * out, so we can try to allocate new block(s) to replace the bad
4485  * blocks.
4486  */
4487 static void handle_fs_bad_blocks(e2fsck_t ctx)
4488 {
4489         ext2_filsys fs = ctx->fs;
4490         dgrp_t          i;
4491         blk64_t         first_block;
4492         blk64_t         new_blk;
4493
4494         for (i = 0; i < fs->group_desc_count; i++) {
4495                 first_block = ext2fs_group_first_block2(fs, i);
4496
4497                 if (ctx->invalid_block_bitmap_flag[i]) {
4498                         new_blk = ext2fs_block_bitmap_loc(fs, i);
4499                         new_table_block(ctx, first_block, i, _("block bitmap"),
4500                                         1, &new_blk);
4501                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4502                 }
4503                 if (ctx->invalid_inode_bitmap_flag[i]) {
4504                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
4505                         new_table_block(ctx, first_block, i, _("inode bitmap"),
4506                                         1, &new_blk);
4507                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4508                 }
4509                 if (ctx->invalid_inode_table_flag[i]) {
4510                         new_blk = ext2fs_inode_table_loc(fs, i);
4511                         new_table_block(ctx, first_block, i, _("inode table"),
4512                                         fs->inode_blocks_per_group,
4513                                         &new_blk);
4514                         ext2fs_inode_table_loc_set(fs, i, new_blk);
4515                         ctx->flags |= E2F_FLAG_RESTART;
4516                 }
4517         }
4518         ctx->invalid_bitmaps = 0;
4519 }
4520
4521 /*
4522  * This routine marks all blocks which are used by the superblock,
4523  * group descriptors, inode bitmaps, and block bitmaps.
4524  */
4525 static void mark_table_blocks(e2fsck_t ctx)
4526 {
4527         ext2_filsys fs = ctx->fs;
4528         blk64_t b;
4529         dgrp_t  i;
4530         unsigned int    j;
4531         struct problem_context pctx;
4532
4533         clear_problem_context(&pctx);
4534
4535         for (i = 0; i < fs->group_desc_count; i++) {
4536                 pctx.group = i;
4537
4538                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
4539                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
4540
4541                 /*
4542                  * Mark the blocks used for the inode table
4543                  */
4544                 if (ext2fs_inode_table_loc(fs, i)) {
4545                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
4546                              j < fs->inode_blocks_per_group;
4547                              j++, b++) {
4548                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4549                                                              b)) {
4550                                         pctx.blk = b;
4551                                         if (!ctx->invalid_inode_table_flag[i] &&
4552                                             fix_problem(ctx,
4553                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
4554                                                 ctx->invalid_inode_table_flag[i]++;
4555                                                 ctx->invalid_bitmaps++;
4556                                         }
4557                                 } else {
4558                                     ext2fs_mark_block_bitmap2(
4559                                                 ctx->block_found_map, b);
4560                                     ext2fs_mark_block_bitmap2(
4561                                                 ctx->block_metadata_map, b);
4562                                 }
4563                         }
4564                 }
4565
4566                 /*
4567                  * Mark block used for the block bitmap
4568                  */
4569                 if (ext2fs_block_bitmap_loc(fs, i)) {
4570                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4571                                      ext2fs_block_bitmap_loc(fs, i))) {
4572                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
4573                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
4574                                         ctx->invalid_block_bitmap_flag[i]++;
4575                                         ctx->invalid_bitmaps++;
4576                                 }
4577                         } else {
4578                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4579                                      ext2fs_block_bitmap_loc(fs, i));
4580                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4581                                      ext2fs_block_bitmap_loc(fs, i));
4582                         }
4583                 }
4584                 /*
4585                  * Mark block used for the inode bitmap
4586                  */
4587                 if (ext2fs_inode_bitmap_loc(fs, i)) {
4588                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4589                                      ext2fs_inode_bitmap_loc(fs, i))) {
4590                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
4591                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
4592                                         ctx->invalid_inode_bitmap_flag[i]++;
4593                                         ctx->invalid_bitmaps++;
4594                                 }
4595                         } else {
4596                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4597                                      ext2fs_inode_bitmap_loc(fs, i));
4598                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4599                                      ext2fs_inode_bitmap_loc(fs, i));
4600                         }
4601                 }
4602         }
4603 }
4604
4605 /*
4606  * These subroutines short circuits ext2fs_get_blocks and
4607  * ext2fs_check_directory; we use them since we already have the inode
4608  * structure, so there's no point in letting the ext2fs library read
4609  * the inode again.
4610  */
4611 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
4612                                   blk_t *blocks)
4613 {
4614         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4615         int     i;
4616
4617         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4618                 return EXT2_ET_CALLBACK_NOTHANDLED;
4619
4620         for (i=0; i < EXT2_N_BLOCKS; i++)
4621                 blocks[i] = ctx->stashed_inode->i_block[i];
4622         return 0;
4623 }
4624
4625 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
4626                                   struct ext2_inode *inode)
4627 {
4628         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4629
4630         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4631                 return EXT2_ET_CALLBACK_NOTHANDLED;
4632         *inode = *ctx->stashed_inode;
4633         return 0;
4634 }
4635
4636 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
4637                             struct ext2_inode *inode)
4638 {
4639         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4640
4641         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
4642                 (inode != ctx->stashed_inode))
4643                 *ctx->stashed_inode = *inode;
4644         return EXT2_ET_CALLBACK_NOTHANDLED;
4645 }
4646
4647 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
4648 {
4649         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4650
4651         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4652                 return EXT2_ET_CALLBACK_NOTHANDLED;
4653
4654         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
4655                 return EXT2_ET_NO_DIRECTORY;
4656         return 0;
4657 }
4658
4659 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
4660                                         blk64_t *ret)
4661 {
4662         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4663         errcode_t       retval;
4664         blk64_t         new_block;
4665
4666         if (ctx->block_found_map) {
4667                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
4668                                            &new_block);
4669                 if (retval)
4670                         return retval;
4671                 if (fs->block_map) {
4672                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
4673                         ext2fs_mark_bb_dirty(fs);
4674                 }
4675         } else {
4676                 if (!fs->block_map) {
4677                         retval = ext2fs_read_block_bitmap(fs);
4678                         if (retval)
4679                                 return retval;
4680                 }
4681
4682                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
4683                 if (retval)
4684                         return retval;
4685         }
4686
4687         *ret = new_block;
4688         return (0);
4689 }
4690
4691 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
4692                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
4693 {
4694         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4695         errcode_t       retval;
4696
4697         if (ctx->block_found_map)
4698                 return ext2fs_new_range(fs, flags, goal, len,
4699                                         ctx->block_found_map, pblk, plen);
4700
4701         if (!fs->block_map) {
4702                 retval = ext2fs_read_block_bitmap(fs);
4703                 if (retval)
4704                         return retval;
4705         }
4706
4707         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
4708                                 pblk, plen);
4709 }
4710
4711 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
4712 {
4713         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4714
4715         /* Never free a critical metadata block */
4716         if (ctx->block_found_map &&
4717             ctx->block_metadata_map &&
4718             inuse < 0 &&
4719             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
4720                 return;
4721
4722         if (ctx->block_found_map) {
4723                 if (inuse > 0)
4724                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4725                 else
4726                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
4727         }
4728 }
4729
4730 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
4731                                            blk_t num, int inuse)
4732 {
4733         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4734
4735         /* Never free a critical metadata block */
4736         if (ctx->block_found_map &&
4737             ctx->block_metadata_map &&
4738             inuse < 0 &&
4739             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
4740                 return;
4741
4742         if (ctx->block_found_map) {
4743                 if (inuse > 0)
4744                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
4745                                                         blk, num);
4746                 else
4747                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
4748                                                         blk, num);
4749         }
4750 }
4751
4752 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
4753 {
4754         ext2_filsys fs = ctx->fs;
4755
4756         if (use_shortcuts) {
4757                 fs->get_blocks = pass1_get_blocks;
4758                 fs->check_directory = pass1_check_directory;
4759                 fs->read_inode = pass1_read_inode;
4760                 fs->write_inode = pass1_write_inode;
4761                 ctx->stashed_ino = 0;
4762         } else {
4763                 fs->get_blocks = 0;
4764                 fs->check_directory = 0;
4765                 fs->read_inode = 0;
4766                 fs->write_inode = 0;
4767         }
4768 }
4769
4770 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
4771 {
4772         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
4773         ext2fs_set_block_alloc_stats_callback(ctx->fs,
4774                                                 e2fsck_block_alloc_stats, 0);
4775         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
4776         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
4777                                         e2fsck_block_alloc_stats_range, NULL);
4778 }