e2fsck/pass1.c

   1 /*
   2  * pass1.c -- pass #1 of e2fsck: sequential scan of the inode table
   3  *
   4  * Copyright (C) 1993, 1994, 1995, 1996, 1997 Theodore Ts'o.
   5  *
   6  * %Begin-Header%
   7  * This file may be redistributed under the terms of the GNU Public
   8  * License.
   9  * %End-Header%
  10  *
  11  * Pass 1 of e2fsck iterates over all the inodes in the filesystems,
  12  * and applies the following tests to each inode:
  13  *
  14  *      - The mode field of the inode must be legal.
  15  *      - The size and block count fields of the inode are correct.
  16  *      - A data block must not be used by another inode
  17  *
  18  * Pass 1 also gathers the collects the following information:
  19  *
  20  *      - A bitmap of which inodes are in use.          (inode_used_map)
  21  *      - A bitmap of which inodes are directories.     (inode_dir_map)
  22  *      - A bitmap of which inodes are regular files.   (inode_reg_map)
  23  *      - A bitmap of which inodes have bad fields.     (inode_bad_map)
  24  *      - A bitmap of which inodes are in bad blocks.   (inode_bb_map)
  25  *      - A bitmap of which inodes are imagic inodes.   (inode_imagic_map)
  26  *      - A bitmap of which inodes are casefolded.      (inode_casefold_map)
  27  *      - A bitmap of which blocks are in use.          (block_found_map)
  28  *      - A bitmap of which blocks are in use by two inodes     (block_dup_map)
  29  *      - The data blocks of the directory inodes.      (dir_map)
  30  *      - Ref counts for ea_inodes.                     (ea_inode_refs)
  31  *      - The encryption policy ID of each encrypted inode. (encrypted_files)
  32  *
  33  * Pass 1 is designed to stash away enough information so that the
  34  * other passes should not need to read in the inode information
  35  * during the normal course of a filesystem check.  (Although if an
  36  * inconsistency is detected, other passes may need to read in an
  37  * inode to fix it.)
  38  *
  39  * Note that pass 1B will be invoked if there are any duplicate blocks
  40  * found.
  41  */
  42
  43 #define _GNU_SOURCE 1 /* get strnlen() */
  44 #include "config.h"
  45 #include <string.h>
  46 #include <time.h>
  47 #ifdef HAVE_ERRNO_H
  48 #include <errno.h>
  49 #endif
  50 #include <assert.h>
  51 #ifdef HAVE_PTHREAD
  52 #include <pthread.h>
  53 #endif
  54
  55 #include "e2fsck.h"
  56 #include <ext2fs/ext2_ext_attr.h>
  57 /* todo remove this finally */
  58 #include <ext2fs/ext2fsP.h>
  59 #include <e2p/e2p.h>
  60
  61 #include "problem.h"
  62
  63 #ifdef NO_INLINE_FUNCS
  64 #define _INLINE_
  65 #else
  66 #define _INLINE_ inline
  67 #endif
  68
  69 #undef DEBUG
  70
  71 struct ea_quota {
  72         blk64_t blocks;
  73         __u64 inodes;
  74 };
  75
  76 static int process_block(ext2_filsys fs, blk64_t        *blocknr,
  77                          e2_blkcnt_t blockcnt, blk64_t ref_blk,
  78                          int ref_offset, void *priv_data);
  79 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
  80                              e2_blkcnt_t blockcnt, blk64_t ref_blk,
  81                              int ref_offset, void *priv_data);
  82 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
  83                          char *block_buf,
  84                          const struct ea_quota *ea_ibody_quota);
  85 static void mark_table_blocks(e2fsck_t ctx);
  86 static void alloc_bb_map(e2fsck_t ctx);
  87 static void alloc_imagic_map(e2fsck_t ctx);
  88 static void mark_inode_bad(e2fsck_t ctx, ino_t ino);
  89 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino);
  90 static void handle_fs_bad_blocks(e2fsck_t ctx);
  91 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
  92 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
  93                                   dgrp_t group, void * priv_data);
  94 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
  95                                     char *block_buf, int adjust_sign);
  96 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
  97
  98 struct process_block_struct {
  99         ext2_ino_t      ino;
 100         unsigned        is_dir:1, is_reg:1, clear:1, suppress:1,
 101                                 fragmented:1, compressed:1, bbcheck:1,
 102                                 inode_modified:1;
 103         blk64_t         num_blocks;
 104         blk64_t         max_blocks;
 105         blk64_t         last_block;
 106         e2_blkcnt_t     last_init_lblock;
 107         e2_blkcnt_t     last_db_block;
 108         int             num_illegal_blocks;
 109         blk64_t         previous_block;
 110         struct ext2_inode *inode;
 111         struct problem_context *pctx;
 112         ext2fs_block_bitmap fs_meta_blocks;
 113         e2fsck_t        ctx;
 114         blk64_t         next_lblock;
 115         struct extent_tree_info eti;
 116 };
 117
 118 struct process_inode_block {
 119         ext2_ino_t ino;
 120         struct ea_quota ea_ibody_quota;
 121         struct ext2_inode_large inode;
 122 };
 123
 124 struct scan_callback_struct {
 125         e2fsck_t                         ctx;
 126         char                            *block_buf;
 127         struct process_inode_block      *inodes_to_process;
 128         int                             *process_inode_count;
 129 };
 130
 131 static void process_inodes(e2fsck_t ctx, char *block_buf,
 132                            struct process_inode_block *inodes_to_process,
 133                            int *process_inode_count);
 134
 135 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
 136                             EXT2_MIN_BLOCK_LOG_SIZE + 1];
 137
 138 /*
 139  * Check to make sure a device inode is real.  Returns 1 if the device
 140  * checks out, 0 if not.
 141  *
 142  * Note: this routine is now also used to check FIFO's and Sockets,
 143  * since they have the same requirement; the i_block fields should be
 144  * zero.
 145  */
 146 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
 147                                     struct ext2_inode *inode)
 148 {
 149         int     i;
 150
 151         /*
 152          * If the index or extents flag is set, then this is a bogus
 153          * device/fifo/socket
 154          */
 155         if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
 156                 return 0;
 157
 158         /*
 159          * We should be able to do the test below all the time, but
 160          * because the kernel doesn't forcibly clear the device
 161          * inode's additional i_block fields, there are some rare
 162          * occasions when a legitimate device inode will have non-zero
 163          * additional i_block fields.  So for now, we only complain
 164          * when the immutable flag is set, which should never happen
 165          * for devices.  (And that's when the problem is caused, since
 166          * you can't set or clear immutable flags for devices.)  Once
 167          * the kernel has been fixed we can change this...
 168          */
 169         if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
 170                 for (i=4; i < EXT2_N_BLOCKS; i++)
 171                         if (inode->i_block[i])
 172                                 return 0;
 173         }
 174         return 1;
 175 }
 176
 177 /*
 178  * Check to make sure a symlink inode is real.  Returns 1 if the symlink
 179  * checks out, 0 if not.
 180  */
 181 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
 182                                struct ext2_inode *inode, char *buf)
 183 {
 184         unsigned int buflen;
 185         unsigned int len;
 186
 187         if ((inode->i_size_high || inode->i_size == 0) ||
 188             (inode->i_flags & EXT2_INDEX_FL))
 189                 return 0;
 190
 191         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
 192                 size_t inline_size;
 193
 194                 if (inode->i_flags & EXT4_EXTENTS_FL)
 195                         return 0;
 196                 if (ext2fs_inline_data_size(fs, ino, &inline_size))
 197                         return 0;
 198                 if (inode->i_size != inline_size)
 199                         return 0;
 200
 201                 return 1;
 202         }
 203
 204         if (ext2fs_is_fast_symlink(inode)) {
 205                 if (inode->i_flags & EXT4_EXTENTS_FL)
 206                         return 0;
 207                 buf = (char *)inode->i_block;
 208                 buflen = sizeof(inode->i_block);
 209         } else {
 210                 ext2_extent_handle_t    handle;
 211                 struct ext2_extent_info info;
 212                 struct ext2fs_extent    extent;
 213                 blk64_t blk;
 214                 int i;
 215
 216                 if (inode->i_flags & EXT4_EXTENTS_FL) {
 217                         if (ext2fs_extent_open2(fs, ino, inode, &handle))
 218                                 return 0;
 219                         if (ext2fs_extent_get_info(handle, &info) ||
 220                             (info.num_entries != 1) ||
 221                             (info.max_depth != 0)) {
 222                                 ext2fs_extent_free(handle);
 223                                 return 0;
 224                         }
 225                         if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
 226                                               &extent) ||
 227                             (extent.e_lblk != 0) ||
 228                             (extent.e_len != 1)) {
 229                                 ext2fs_extent_free(handle);
 230                                 return 0;
 231                         }
 232                         blk = extent.e_pblk;
 233                         ext2fs_extent_free(handle);
 234                 } else {
 235                         blk = inode->i_block[0];
 236
 237                         for (i = 1; i < EXT2_N_BLOCKS; i++)
 238                                 if (inode->i_block[i])
 239                                         return 0;
 240                 }
 241
 242                 if (blk < fs->super->s_first_data_block ||
 243                     blk >= ext2fs_blocks_count(fs->super))
 244                         return 0;
 245
 246                 if (io_channel_read_blk64(fs->io, blk, 1, buf))
 247                         return 0;
 248
 249                 buflen = fs->blocksize;
 250         }
 251
 252         if (inode->i_flags & EXT4_ENCRYPT_FL)
 253                 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
 254         else
 255                 len = strnlen(buf, buflen);
 256
 257         if (len >= buflen)
 258                 return 0;
 259
 260         if (len != inode->i_size)
 261                 return 0;
 262         return 1;
 263 }
 264
 265 /*
 266  * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
 267  */
 268 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
 269 static void check_extents_inlinedata(e2fsck_t ctx,
 270                                      struct problem_context *pctx)
 271 {
 272         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 273                 return;
 274
 275         if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
 276                 return;
 277
 278         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 279         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 280 }
 281 #undef BAD_SPECIAL_FLAGS
 282
 283 /*
 284  * If the immutable (or append-only) flag is set on the inode, offer
 285  * to clear it.
 286  */
 287 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
 288 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
 289 {
 290         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 291                 return;
 292
 293         if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
 294                 return;
 295
 296         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 297         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 298 }
 299
 300 /*
 301  * If device, fifo or socket, check size is zero -- if not offer to
 302  * clear it
 303  */
 304 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
 305 {
 306         struct ext2_inode *inode = pctx->inode;
 307
 308         if (EXT2_I_SIZE(inode) == 0)
 309                 return;
 310
 311         if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
 312                 return;
 313
 314         ext2fs_inode_size_set(ctx->fs, inode, 0);
 315         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 316 }
 317
 318 /*
 319  * For a given size, calculate how many blocks would be charged towards quota.
 320  */
 321 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
 322 {
 323         blk64_t clusters;
 324
 325         clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
 326         return EXT2FS_C2B(fs, clusters);
 327 }
 328
 329 /*
 330  * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
 331  * the problem code.
 332  */
 333 static problem_t check_large_ea_inode(e2fsck_t ctx,
 334                                       struct ext2_ext_attr_entry *entry,
 335                                       struct problem_context *pctx,
 336                                       blk64_t *quota_blocks)
 337 {
 338         struct ext2_inode inode;
 339         __u32 hash;
 340         errcode_t retval;
 341
 342         /* Check if inode is within valid range */
 343         if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
 344             (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
 345                 pctx->num = entry->e_value_inum;
 346                 return PR_1_ATTR_VALUE_EA_INODE;
 347         }
 348
 349         e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
 350
 351         retval = ext2fs_ext_attr_hash_entry2(ctx->fs, entry, NULL, &hash);
 352         if (retval) {
 353                 com_err("check_large_ea_inode", retval,
 354                         _("while hashing entry with e_value_inum = %u"),
 355                         entry->e_value_inum);
 356                 fatal_error(ctx, 0);
 357         }
 358
 359         if (hash == entry->e_hash) {
 360                 *quota_blocks = size_to_quota_blocks(ctx->fs,
 361                                                      entry->e_value_size);
 362         } else {
 363                 /* This might be an old Lustre-style ea_inode reference. */
 364                 if (inode.i_mtime == pctx->ino &&
 365                     inode.i_generation == pctx->inode->i_generation) {
 366                         *quota_blocks = 0;
 367                 } else {
 368                         /* If target inode is also missing EA_INODE flag,
 369                          * this is likely to be a bad reference.
 370                          */
 371                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 372                                 pctx->num = entry->e_value_inum;
 373                                 return PR_1_ATTR_VALUE_EA_INODE;
 374                         } else {
 375                                 pctx->num = entry->e_hash;
 376                                 return PR_1_ATTR_HASH;
 377                         }
 378                 }
 379         }
 380
 381         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 382                 pctx->num = entry->e_value_inum;
 383                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
 384                         inode.i_flags |= EXT4_EA_INODE_FL;
 385                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
 386                                            &inode);
 387                 } else {
 388                         return PR_1_ATTR_NO_EA_INODE_FL;
 389                 }
 390         }
 391         return 0;
 392 }
 393
 394 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
 395                               struct ext2_ext_attr_entry *first, void *end)
 396 {
 397         struct ext2_ext_attr_entry *entry = first;
 398         struct ext2_ext_attr_entry *np = EXT2_EXT_ATTR_NEXT(entry);
 399
 400         while ((void *) entry < end && (void *) np < end &&
 401                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
 402                 if (!entry->e_value_inum)
 403                         goto next;
 404                 if (!ctx->ea_inode_refs) {
 405                         pctx->errcode = ea_refcount_create(0,
 406                                                            &ctx->ea_inode_refs);
 407                         if (pctx->errcode) {
 408                                 pctx->num = 4;
 409                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
 410                                 ctx->flags |= E2F_FLAG_ABORT;
 411                                 return;
 412                         }
 413                 }
 414                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
 415                                       0);
 416         next:
 417                 entry = np;
 418                 np = EXT2_EXT_ATTR_NEXT(entry);
 419         }
 420 }
 421
 422 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
 423                               struct ea_quota *ea_ibody_quota)
 424 {
 425         struct ext2_super_block *sb = ctx->fs->super;
 426         struct ext2_inode_large *inode;
 427         struct ext2_ext_attr_entry *entry;
 428         char *start, *header, *end;
 429         unsigned int storage_size, remain;
 430         problem_t problem = 0;
 431         region_t region = 0;
 432
 433         ea_ibody_quota->blocks = 0;
 434         ea_ibody_quota->inodes = 0;
 435
 436         inode = (struct ext2_inode_large *) pctx->inode;
 437         storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
 438                 inode->i_extra_isize;
 439         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 440                  inode->i_extra_isize;
 441         end = header + storage_size;
 442         start = header + sizeof(__u32);
 443         entry = (struct ext2_ext_attr_entry *) start;
 444
 445         /* scan all entry's headers first */
 446
 447         /* take finish entry 0UL into account */
 448         remain = storage_size - sizeof(__u32);
 449
 450         region = region_create(0, storage_size);
 451         if (!region) {
 452                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
 453                 problem = 0;
 454                 ctx->flags |= E2F_FLAG_ABORT;
 455                 return;
 456         }
 457         if (region_allocate(region, 0, sizeof(__u32))) {
 458                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 459                 goto fix;
 460         }
 461
 462         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
 463                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
 464                 __u32 hash;
 465
 466                 if (region_allocate(region, (char *)entry - (char *)header,
 467                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
 468                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
 469                         goto fix;
 470                 }
 471
 472                 /* header eats this space */
 473                 remain -= sizeof(struct ext2_ext_attr_entry);
 474
 475                 /* is attribute name valid? */
 476                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
 477                         pctx->num = entry->e_name_len;
 478                         problem = PR_1_ATTR_NAME_LEN;
 479                         goto fix;
 480                 }
 481
 482                 /* attribute len eats this space */
 483                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
 484
 485                 if (entry->e_value_inum == 0) {
 486                         /* check value size */
 487                         if (entry->e_value_size > remain) {
 488                                 pctx->num = entry->e_value_size;
 489                                 problem = PR_1_ATTR_VALUE_SIZE;
 490                                 goto fix;
 491                         }
 492
 493                         if (entry->e_value_size &&
 494                             region_allocate(region,
 495                                             sizeof(__u32) + entry->e_value_offs,
 496                                             EXT2_EXT_ATTR_SIZE(
 497                                                 entry->e_value_size))) {
 498                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 499                                 goto fix;
 500                         }
 501
 502                         hash = ext2fs_ext_attr_hash_entry(entry,
 503                                                           start + entry->e_value_offs);
 504
 505                         /* e_hash may be 0 in older inode's ea */
 506                         if (entry->e_hash != 0 && entry->e_hash != hash) {
 507                                 pctx->num = entry->e_hash;
 508                                 problem = PR_1_ATTR_HASH;
 509                                 goto fix;
 510                         }
 511                 } else {
 512                         blk64_t quota_blocks;
 513
 514                         problem = check_large_ea_inode(ctx, entry, pctx,
 515                                                        &quota_blocks);
 516                         if (problem != 0)
 517                                 goto fix;
 518
 519                         ea_ibody_quota->blocks += quota_blocks;
 520                         ea_ibody_quota->inodes++;
 521                 }
 522
 523                 /* If EA value is stored in external inode then it does not
 524                  * consume space here */
 525                 if (entry->e_value_inum == 0)
 526                         remain -= entry->e_value_size;
 527
 528                 entry = EXT2_EXT_ATTR_NEXT(entry);
 529         }
 530
 531         if (region_allocate(region, (char *)entry - (char *)header,
 532                             sizeof(__u32))) {
 533                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 534                 goto fix;
 535         }
 536 fix:
 537         if (region)
 538                 region_free(region);
 539         /*
 540          * it seems like a corruption. it's very unlikely we could repair
 541          * EA(s) in automatic fashion -bzzz
 542          */
 543         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
 544                 inc_ea_inode_refs(ctx, pctx,
 545                                   (struct ext2_ext_attr_entry *)start, end);
 546                 return;
 547         }
 548
 549         /* simply remove all possible EA(s) */
 550         *((__u32 *)header) = 0UL;
 551         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 552                                 EXT2_INODE_SIZE(sb), "pass1");
 553         ea_ibody_quota->blocks = 0;
 554         ea_ibody_quota->inodes = 0;
 555 }
 556
 557 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
 558         return (xtime & (1U << 31)) != 0 &&
 559                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
 560 }
 561
 562 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
 563         check_inode_extra_negative_epoch(inode->i_##xtime, \
 564                                          inode->i_##xtime##_extra)
 565
 566 /* When today's date is earlier than 2242, we assume that atimes,
 567  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
 568  * actually pre-1970 dates mis-encoded.
 569  */
 570 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
 571
 572 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
 573                                     struct ea_quota *ea_ibody_quota)
 574 {
 575         struct ext2_super_block *sb = ctx->fs->super;
 576         struct ext2_inode_large *inode;
 577         __u32 *eamagic;
 578         int min, max;
 579
 580         ea_ibody_quota->blocks = 0;
 581         ea_ibody_quota->inodes = 0;
 582
 583         inode = (struct ext2_inode_large *) pctx->inode;
 584         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
 585                 /* this isn't large inode. so, nothing to check */
 586                 return;
 587         }
 588
 589 #if 0
 590         printf("inode #%u, i_extra_size %d\n", pctx->ino,
 591                         inode->i_extra_isize);
 592 #endif
 593         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
 594         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
 595         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
 596         /*
 597          * For now we will allow i_extra_isize to be 0, but really
 598          * implementations should never allow i_extra_isize to be 0
 599          */
 600         if (inode->i_extra_isize &&
 601             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
 602              inode->i_extra_isize & 3)) {
 603                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
 604                         return;
 605                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
 606                         inode->i_extra_isize = sb->s_want_extra_isize;
 607                 else
 608                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
 609                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 610                                         EXT2_INODE_SIZE(sb), "pass1");
 611         }
 612
 613         /* check if there is no place for an EA header */
 614         if (inode->i_extra_isize >= max - sizeof(__u32))
 615                 return;
 616
 617         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 618                         inode->i_extra_isize);
 619         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
 620                 /* it seems inode has an extended attribute(s) in body */
 621                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
 622         }
 623
 624         /*
 625          * If the inode's extended atime (ctime, crtime, mtime) is stored in
 626          * the old, invalid format, repair it.
 627          */
 628         if (((sizeof(time_t) <= 4) ||
 629              (((sizeof(time_t) > 4) &&
 630                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
 631             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
 632              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
 633              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
 634              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
 635
 636                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
 637                         return;
 638
 639                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
 640                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
 641                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
 642                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
 643                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
 644                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
 645                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
 646                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
 647                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 648                                         EXT2_INODE_SIZE(sb), "pass1");
 649         }
 650
 651 }
 652
 653 /*
 654  * Check to see if the inode might really be a directory, despite i_mode
 655  *
 656  * This is a lot of complexity for something for which I'm not really
 657  * convinced happens frequently in the wild.  If for any reason this
 658  * causes any problems, take this code out.
 659  * [tytso:20070331.0827EDT]
 660  */
 661 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
 662                                 char *buf)
 663 {
 664         struct ext2_inode *inode = pctx->inode;
 665         struct ext2_dir_entry   *dirent;
 666         errcode_t               retval;
 667         blk64_t                 blk;
 668         unsigned int            i, rec_len, not_device = 0;
 669         int                     extent_fs;
 670         int                     inlinedata_fs;
 671
 672         /*
 673          * If the mode looks OK, we believe it.  If the first block in
 674          * the i_block array is 0, this cannot be a directory. If the
 675          * inode is extent-mapped, it is still the case that the latter
 676          * cannot be 0 - the magic number in the extent header would make
 677          * it nonzero.
 678          */
 679         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
 680             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
 681                 return;
 682
 683         /*
 684          * Check the block numbers in the i_block array for validity:
 685          * zero blocks are skipped (but the first one cannot be zero -
 686          * see above), other blocks are checked against the first and
 687          * max data blocks (from the the superblock) and against the
 688          * block bitmap. Any invalid block found means this cannot be
 689          * a directory.
 690          *
 691          * If there are non-zero blocks past the fourth entry, then
 692          * this cannot be a device file: we remember that for the next
 693          * check.
 694          *
 695          * For extent mapped files, we don't do any sanity checking:
 696          * just try to get the phys block of logical block 0 and run
 697          * with it.
 698          *
 699          * For inline data files, we just try to get the size of inline
 700          * data.  If it's true, we will treat it as a directory.
 701          */
 702
 703         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
 704         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
 705         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
 706                 size_t size;
 707                 __u32 dotdot;
 708                 unsigned int rec_len2;
 709                 struct ext2_dir_entry de;
 710
 711                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
 712                         return;
 713                 /*
 714                  * If the size isn't a multiple of 4, it's probably not a
 715                  * directory??
 716                  */
 717                 if (size & 3)
 718                         return;
 719                 /*
 720                  * If the first 10 bytes don't look like a directory entry,
 721                  * it's probably not a directory.
 722                  */
 723                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
 724                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
 725                        EXT2_DIR_REC_LEN(0));
 726                 dotdot = ext2fs_le32_to_cpu(dotdot);
 727                 de.inode = ext2fs_le32_to_cpu(de.inode);
 728                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
 729                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
 730                 if (dotdot >= ctx->fs->super->s_inodes_count ||
 731                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
 732                      dotdot != EXT2_ROOT_INO) ||
 733                     de.inode >= ctx->fs->super->s_inodes_count ||
 734                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
 735                      de.inode != 0) ||
 736                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
 737                               EXT4_INLINE_DATA_DOTDOT_SIZE)
 738                         return;
 739                 /* device files never have a "system.data" entry */
 740                 goto isdir;
 741         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
 742                 /* extent mapped */
 743                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
 744                                  &blk))
 745                         return;
 746                 /* device files are never extent mapped */
 747                 not_device++;
 748         } else {
 749                 for (i=0; i < EXT2_N_BLOCKS; i++) {
 750                         blk = inode->i_block[i];
 751                         if (!blk)
 752                                 continue;
 753                         if (i >= 4)
 754                                 not_device++;
 755
 756                         if (blk < ctx->fs->super->s_first_data_block ||
 757                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
 758                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
 759                                                            blk))
 760                                 return; /* Invalid block, can't be dir */
 761                 }
 762                 blk = inode->i_block[0];
 763         }
 764
 765         /*
 766          * If the mode says this is a device file and the i_links_count field
 767          * is sane and we have not ruled it out as a device file previously,
 768          * we declare it a device file, not a directory.
 769          */
 770         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
 771             (inode->i_links_count == 1) && !not_device)
 772                 return;
 773
 774         /* read the first block */
 775         ehandler_operation(_("reading directory block"));
 776         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
 777         ehandler_operation(0);
 778         if (retval)
 779                 return;
 780
 781         dirent = (struct ext2_dir_entry *) buf;
 782         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 783         if (retval)
 784                 return;
 785         if ((ext2fs_dirent_name_len(dirent) != 1) ||
 786             (dirent->name[0] != '.') ||
 787             (dirent->inode != pctx->ino) ||
 788             (rec_len < 12) ||
 789             (rec_len % 4) ||
 790             (rec_len >= ctx->fs->blocksize - 12))
 791                 return;
 792
 793         dirent = (struct ext2_dir_entry *) (buf + rec_len);
 794         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 795         if (retval)
 796                 return;
 797         if ((ext2fs_dirent_name_len(dirent) != 2) ||
 798             (dirent->name[0] != '.') ||
 799             (dirent->name[1] != '.') ||
 800             (rec_len < 12) ||
 801             (rec_len % 4))
 802                 return;
 803
 804 isdir:
 805         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
 806                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
 807                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
 808                                         EXT2_INODE_SIZE(ctx->fs->super),
 809                                         "check_is_really_dir");
 810         }
 811 }
 812
 813 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
 814                                      int flags, ext2_icount_t hint,
 815                                      ext2_icount_t *ret)
 816 {
 817         unsigned int            threshold;
 818         unsigned int            save_type;
 819         ext2_ino_t              num_dirs;
 820         errcode_t               retval;
 821         char                    *tdb_dir;
 822         int                     enable;
 823
 824         *ret = 0;
 825
 826         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
 827                            &tdb_dir);
 828         profile_get_uint(ctx->profile, "scratch_files",
 829                          "numdirs_threshold", 0, 0, &threshold);
 830         profile_get_boolean(ctx->profile, "scratch_files",
 831                             "icount", 0, 1, &enable);
 832
 833         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
 834         if (retval)
 835                 num_dirs = 1024;        /* Guess */
 836
 837         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
 838             (!threshold || num_dirs > threshold)) {
 839                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
 840                                                   flags, ret);
 841                 if (retval == 0)
 842                         return 0;
 843         }
 844         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
 845                                &save_type);
 846         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
 847                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
 848         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
 849         ctx->fs->default_bitmap_type = save_type;
 850         return retval;
 851 }
 852
 853 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
 854                                             e2fsck_t ctx,
 855                                             struct problem_context *pctx)
 856 {
 857         errcode_t retval;
 858         struct ext2_inode_large inode;
 859
 860         /*
 861          * Reread inode.  If we don't see checksum error, then this inode
 862          * has been fixed elsewhere.
 863          */
 864         ctx->stashed_ino = 0;
 865         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
 866                                         sizeof(inode));
 867         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
 868                 return retval;
 869         if (!retval)
 870                 return 0;
 871
 872         /*
 873          * Checksum still doesn't match.  That implies that the inode passes
 874          * all the sanity checks, so maybe the checksum is simply corrupt.
 875          * See if the user will go for fixing that.
 876          */
 877         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
 878                 return 0;
 879
 880         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
 881                                          sizeof(inode));
 882         return retval;
 883 }
 884
 885 static void reserve_block_for_root_repair(e2fsck_t ctx)
 886 {
 887         blk64_t         blk = 0;
 888         errcode_t       err;
 889         ext2_filsys     fs = ctx->fs;
 890
 891         ctx->root_repair_block = 0;
 892         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
 893                 return;
 894
 895         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 896         if (err)
 897                 return;
 898         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 899         ctx->root_repair_block = blk;
 900 }
 901
 902 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
 903 {
 904         blk64_t         blk = 0;
 905         errcode_t       err;
 906         ext2_filsys     fs = ctx->fs;
 907         static const char name[] = "lost+found";
 908         ext2_ino_t      ino;
 909
 910         ctx->lnf_repair_block = 0;
 911         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
 912                 return;
 913
 914         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 915         if (err)
 916                 return;
 917         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 918         ctx->lnf_repair_block = blk;
 919 }
 920
 921 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
 922                                          size_t *sz)
 923 {
 924         void *p;
 925         struct ext2_xattr_handle *handle;
 926         errcode_t retval;
 927
 928         retval = ext2fs_xattrs_open(fs, ino, &handle);
 929         if (retval)
 930                 return retval;
 931
 932         retval = ext2fs_xattrs_read(handle);
 933         if (retval)
 934                 goto err;
 935
 936         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
 937         if (retval)
 938                 goto err;
 939         ext2fs_free_mem(&p);
 940 err:
 941         (void) ext2fs_xattrs_close(&handle);
 942         return retval;
 943 }
 944
 945 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
 946                                     struct problem_context *pctx,
 947                                     int failed_csum)
 948 {
 949         if (!failed_csum)
 950                 return;
 951
 952         /*
 953          * If the inode failed the checksum and the user didn't
 954          * clear the inode, test the checksum again -- if it still
 955          * fails, ask the user if the checksum should be corrected.
 956          */
 957         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
 958         if (pctx->errcode)
 959                 ctx->flags |= E2F_FLAG_ABORT;
 960 }
 961 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
 962         do { \
 963                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
 964                 if ((ctx)->flags & E2F_FLAG_ABORT) \
 965                         return; \
 966         } while (0)
 967
 968 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
 969 {
 970         __u32 x;
 971         int i;
 972
 973         for (i = 0; i < EXT2_N_BLOCKS; i++) {
 974                 x = inode->i_block[i];
 975 #ifdef WORDS_BIGENDIAN
 976                 x = ext2fs_swab32(x);
 977 #endif
 978                 if (x >= ext2fs_blocks_count(fs->super))
 979                         return 0;
 980         }
 981
 982         return 1;
 983 }
 984
 985 /*
 986  * Figure out what to do with an inode that has both extents and inline data
 987  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
 988  */
 989 static int fix_inline_data_extents_file(e2fsck_t ctx,
 990                                         ext2_ino_t ino,
 991                                         struct ext2_inode *inode,
 992                                         int inode_size,
 993                                         struct problem_context *pctx)
 994 {
 995         size_t max_inline_ea_size;
 996         ext2_filsys fs = ctx->fs;
 997         int dirty = 0;
 998
 999         /* Both feature flags not set?  Just run the regular checks */
1000         if (!ext2fs_has_feature_extents(fs->super) &&
1001             !ext2fs_has_feature_inline_data(fs->super))
1002                 return 0;
1003
1004         /* Clear both flags if it's a special file */
1005         if (LINUX_S_ISCHR(inode->i_mode) ||
1006             LINUX_S_ISBLK(inode->i_mode) ||
1007             LINUX_S_ISFIFO(inode->i_mode) ||
1008             LINUX_S_ISSOCK(inode->i_mode)) {
1009                 check_extents_inlinedata(ctx, pctx);
1010                 return 0;
1011         }
1012
1013         /* If it looks like an extent tree, try to clear inlinedata */
1014         if (ext2fs_extent_header_verify(inode->i_block,
1015                                  sizeof(inode->i_block)) == 0 &&
1016             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1017                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1018                 dirty = 1;
1019                 goto out;
1020         }
1021
1022         /* If it looks short enough to be inline data, try to clear extents */
1023         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1024                 max_inline_ea_size = inode_size -
1025                                      (EXT2_GOOD_OLD_INODE_SIZE +
1026                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1027         else
1028                 max_inline_ea_size = 0;
1029         if (EXT2_I_SIZE(inode) <
1030             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1031             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1032                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1033                 dirty = 1;
1034                 goto out;
1035         }
1036
1037         /*
1038          * Too big for inline data, but no evidence of extent tree -
1039          * maybe it's a block map file?  If the mappings all look valid?
1040          */
1041         if (could_be_block_map(fs, inode) &&
1042             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1043 #ifdef WORDS_BIGENDIAN
1044                 int i;
1045
1046                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1047                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1048 #endif
1049
1050                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1051                 dirty = 1;
1052                 goto out;
1053         }
1054
1055         /* Oh well, just clear the busted inode. */
1056         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1057                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1058                 return -1;
1059         }
1060
1061 out:
1062         if (dirty)
1063                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1064
1065         return 0;
1066 }
1067
1068 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1069 {
1070         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1071         dgrp_t start = *group, grp;
1072         blk64_t blocks_to_read = 0;
1073         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1074
1075         if (ctx->readahead_kb == 0)
1076                 goto out;
1077
1078         /* Keep iterating groups until we have enough to readahead */
1079         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1080         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1081                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1082                         continue;
1083                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1084                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1085                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1086                                         inodes_per_block;
1087                 if (blocks_to_read * ctx->fs->blocksize >
1088                     ctx->readahead_kb * 1024)
1089                         break;
1090         }
1091
1092         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1093                                grp - start + 1);
1094         if (err == EAGAIN) {
1095                 ctx->readahead_kb /= 2;
1096                 err = 0;
1097         }
1098
1099 out:
1100         if (err) {
1101                 /* Error; disable itable readahead */
1102                 *group = ctx->fs->group_desc_count;
1103                 *next_ino = ctx->fs->super->s_inodes_count;
1104         } else {
1105                 /*
1106                  * Don't do more readahead until we've reached the first inode
1107                  * of the last inode scan buffer block for the last group.
1108                  */
1109                 *group = grp + 1;
1110                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1111                                      ctx->inode_buffer_blocks :
1112                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1113                                     ctx->fs->blocksize /
1114                                     EXT2_INODE_SIZE(ctx->fs->super);
1115                 inodes_in_group--;
1116                 *next_ino = inodes_in_group -
1117                             (inodes_in_group % inodes_per_buffer) + 1 +
1118                             (grp * ctx->fs->super->s_inodes_per_group);
1119         }
1120 }
1121
1122 /*
1123  * Check if the passed ino is one of the used superblock quota inodes.
1124  *
1125  * Before the quota inodes were journaled, older superblock quota inodes
1126  * were just regular files in the filesystem and not reserved inodes.  This
1127  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1128  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1129  */
1130 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1131 {
1132         enum quota_type qtype;
1133
1134         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1135                 if (*quota_sb_inump(sb, qtype) == ino)
1136                         return 1;
1137
1138         return 0;
1139 }
1140
1141 /*
1142  * Check if the passed ino is one of the reserved quota inodes.
1143  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1144  * inodes.  These inodes may or may not be in use by the quota feature.
1145  */
1146 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1147 {
1148         enum quota_type qtype;
1149
1150         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1151                 if (quota_type2inum(qtype, fs->super) == ino)
1152                         return 1;
1153
1154         return 0;
1155 }
1156
1157 static int e2fsck_should_abort(e2fsck_t ctx)
1158 {
1159         e2fsck_t global_ctx;
1160
1161         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1162                 return 1;
1163
1164         if (ctx->global_ctx) {
1165                 global_ctx = ctx->global_ctx;
1166                 if (global_ctx->flags & E2F_FLAG_SIGNAL_MASK)
1167                         return 1;
1168         }
1169         return 0;
1170 }
1171
1172 void e2fsck_pass1_run(e2fsck_t ctx)
1173 {
1174         int     i;
1175         ext2_filsys fs = ctx->fs;
1176         ext2_ino_t      ino = 0;
1177         struct ext2_inode *inode = NULL;
1178         ext2_inode_scan scan = NULL;
1179         char            *block_buf = NULL;
1180 #ifdef RESOURCE_TRACK
1181         struct resource_track   rtrack;
1182 #endif
1183         unsigned char   frag, fsize;
1184         struct          problem_context pctx;
1185         struct          scan_callback_struct scan_struct;
1186         struct ext2_super_block *sb = ctx->fs->super;
1187         const char      *old_op;
1188         const char      *eop_next_inode = _("getting next inode from scan");
1189         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1190         int             low_dtime_check = 1;
1191         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1192         unsigned int    bufsize;
1193         int             failed_csum = 0;
1194         ext2_ino_t      ino_threshold = 0;
1195         dgrp_t          ra_group = 0;
1196         struct ea_quota ea_ibody_quota;
1197         struct process_inode_block *inodes_to_process;
1198         int             process_inode_count;
1199
1200         init_resource_track(&rtrack, ctx->fs->io);
1201         clear_problem_context(&pctx);
1202
1203         /* If we can do readahead, figure out how many groups to pull in. */
1204         if (!e2fsck_can_readahead(ctx->fs))
1205                 ctx->readahead_kb = 0;
1206         else if (ctx->readahead_kb == ~0ULL)
1207                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1208         pass1_readahead(ctx, &ra_group, &ino_threshold);
1209
1210         if (!(ctx->options & E2F_OPT_PREEN))
1211                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1212
1213         if (ext2fs_has_feature_dir_index(fs->super) &&
1214             !(ctx->options & E2F_OPT_NO)) {
1215                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1216                         ctx->dirs_to_hash = 0;
1217         }
1218
1219 #ifdef MTRACE
1220         mtrace_print("Pass 1");
1221 #endif
1222
1223         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1224         extent_fs = ext2fs_has_feature_extents(sb);
1225         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1226         casefold_fs = ext2fs_has_feature_casefold(sb);
1227
1228         /*
1229          * Allocate bitmaps structures
1230          */
1231         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1232                                                     EXT2FS_BMAP64_RBTREE,
1233                                                     "inode_used_map",
1234                                                     &ctx->inode_used_map);
1235         if (pctx.errcode) {
1236                 pctx.num = 1;
1237                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1238                 ctx->flags |= E2F_FLAG_ABORT;
1239                 return;
1240         }
1241         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1242                         _("directory inode map"),
1243                         ctx->global_ctx ? EXT2FS_BMAP64_RBTREE :
1244                         EXT2FS_BMAP64_AUTODIR,
1245                         "inode_dir_map", &ctx->inode_dir_map);
1246         if (pctx.errcode) {
1247                 pctx.num = 2;
1248                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1249                 ctx->flags |= E2F_FLAG_ABORT;
1250                 return;
1251         }
1252         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1253                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1254                         "inode_reg_map", &ctx->inode_reg_map);
1255         if (pctx.errcode) {
1256                 pctx.num = 6;
1257                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1258                 ctx->flags |= E2F_FLAG_ABORT;
1259                 return;
1260         }
1261         pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1262                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1263                         "block_found_map", &ctx->block_found_map);
1264         if (pctx.errcode) {
1265                 pctx.num = 1;
1266                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1267                 ctx->flags |= E2F_FLAG_ABORT;
1268                 return;
1269         }
1270         pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1271                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1272                         "block_metadata_map", &ctx->block_metadata_map);
1273         if (pctx.errcode) {
1274                 pctx.num = 1;
1275                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1276                 ctx->flags |= E2F_FLAG_ABORT;
1277                 return;
1278         }
1279         if (casefold_fs) {
1280                 pctx.errcode =
1281                         e2fsck_allocate_inode_bitmap(fs,
1282                                                      _("inode casefold map"),
1283                                                      EXT2FS_BMAP64_RBTREE,
1284                                                      "inode_casefold_map",
1285                                                      &ctx->inode_casefold_map);
1286                 if (pctx.errcode) {
1287                         pctx.num = 1;
1288                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1289                         ctx->flags |= E2F_FLAG_ABORT;
1290                         return;
1291                 }
1292         }
1293         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1294                                            &ctx->inode_link_info);
1295         if (pctx.errcode) {
1296                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1297                 ctx->flags |= E2F_FLAG_ABORT;
1298                 return;
1299         }
1300         bufsize = inode_size;
1301         if (bufsize < sizeof(struct ext2_inode_large))
1302                 bufsize = sizeof(struct ext2_inode_large);
1303         inode = (struct ext2_inode *)
1304                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1305
1306         inodes_to_process = (struct process_inode_block *)
1307                 e2fsck_allocate_memory(ctx,
1308                                        (ctx->process_inode_size *
1309                                         sizeof(struct process_inode_block)),
1310                                        "array of inodes to process");
1311         process_inode_count = 0;
1312
1313         pctx.errcode = ext2fs_init_dblist(fs, 0);
1314         if (pctx.errcode) {
1315                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1316                 ctx->flags |= E2F_FLAG_ABORT;
1317                 goto endit;
1318         }
1319
1320         /*
1321          * If the last orphan field is set, clear it, since the pass1
1322          * processing will automatically find and clear the orphans.
1323          * In the future, we may want to try using the last_orphan
1324          * linked list ourselves, but for now, we clear it so that the
1325          * ext3 mount code won't get confused.
1326          */
1327         if (!(ctx->options & E2F_OPT_READONLY)) {
1328                 if (fs->super->s_last_orphan) {
1329                         fs->super->s_last_orphan = 0;
1330                         ext2fs_mark_super_dirty(fs);
1331                 }
1332         }
1333
1334         mark_table_blocks(ctx);
1335         pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1336                                                 &ctx->block_found_map);
1337         if (pctx.errcode) {
1338                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1339                 ctx->flags |= E2F_FLAG_ABORT;
1340                 goto endit;
1341         }
1342         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1343                                                     "block interate buffer");
1344         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1345                 e2fsck_use_inode_shortcuts(ctx, 1);
1346         e2fsck_intercept_block_allocations(ctx);
1347         old_op = ehandler_operation(_("opening inode scan"));
1348         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1349                                               &scan);
1350         ehandler_operation(old_op);
1351         if (pctx.errcode) {
1352                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1353                 ctx->flags |= E2F_FLAG_ABORT;
1354                 goto endit;
1355         }
1356         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1357                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1358         ctx->stashed_inode = inode;
1359         scan_struct.ctx = ctx;
1360         scan_struct.block_buf = block_buf;
1361         scan_struct.inodes_to_process = inodes_to_process;
1362         scan_struct.process_inode_count = &process_inode_count;
1363         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1364         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1365                                               ctx->fs->group_desc_count)))
1366                 goto endit;
1367         if ((fs->super->s_wtime &&
1368              fs->super->s_wtime < fs->super->s_inodes_count) ||
1369             (fs->super->s_mtime &&
1370              fs->super->s_mtime < fs->super->s_inodes_count) ||
1371             (fs->super->s_mkfs_time &&
1372              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1373                 low_dtime_check = 0;
1374
1375         if (ext2fs_has_feature_mmp(fs->super) &&
1376             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1377             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1378                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1379                                           fs->super->s_mmp_block);
1380
1381         /* Set up ctx->lost_and_found if possible */
1382         (void) e2fsck_get_lost_and_found(ctx, 0);
1383
1384 #ifdef HAVE_PTHREAD
1385         if (ctx->global_ctx) {
1386                 if (ctx->options & E2F_OPT_DEBUG &&
1387                     ctx->options & E2F_OPT_MULTITHREAD)
1388                         fprintf(stderr, "thread %d jumping to group %d\n",
1389                                         ctx->thread_info.et_thread_index,
1390                                         ctx->thread_info.et_group_start);
1391                 pctx.errcode = ext2fs_inode_scan_goto_blockgroup(scan,
1392                                         ctx->thread_info.et_group_start);
1393                 if (pctx.errcode) {
1394                         fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1395                         ctx->flags |= E2F_FLAG_ABORT;
1396                         goto endit;
1397                 }
1398         }
1399 #endif
1400
1401         while (1) {
1402                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1403                         if (e2fsck_mmp_update(fs))
1404                                 fatal_error(ctx, 0);
1405                 }
1406                 old_op = ehandler_operation(eop_next_inode);
1407                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1408                                                           inode, inode_size);
1409                 if (ino > ino_threshold)
1410                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1411                 ehandler_operation(old_op);
1412                 if (e2fsck_should_abort(ctx))
1413                         goto endit;
1414                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1415                         /*
1416                          * If badblocks says badblocks is bad, offer to clear
1417                          * the list, update the in-core bb list, and restart
1418                          * the inode scan.
1419                          */
1420                         if (ino == EXT2_BAD_INO &&
1421                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1422                                         &pctx)) {
1423                                 errcode_t err;
1424
1425                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1426                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1427                                 ctx->fs->badblocks = NULL;
1428                                 err = ext2fs_read_bb_inode(ctx->fs,
1429                                                         &ctx->fs->badblocks);
1430                                 if (err) {
1431                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1432                                                     &pctx);
1433                                         ctx->flags |= E2F_FLAG_ABORT;
1434                                 } else
1435                                         ctx->flags |= E2F_FLAG_RESTART;
1436                                 goto endit;
1437                         }
1438                         if (!ctx->inode_bb_map)
1439                                 alloc_bb_map(ctx);
1440                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1441                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1442                         continue;
1443                 }
1444                 if (pctx.errcode == EXT2_ET_SCAN_FINISHED)
1445                         break;
1446                 if (pctx.errcode &&
1447                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1448                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1449                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1450                         ctx->flags |= E2F_FLAG_ABORT;
1451                         goto endit;
1452                 }
1453                 if (!ino)
1454                         break;
1455 #ifdef HAVE_PTHREAD
1456                 if (ctx->global_ctx)
1457                         ctx->thread_info.et_inode_number++;
1458 #endif
1459                 pctx.ino = ino;
1460                 pctx.inode = inode;
1461                 ctx->stashed_ino = ino;
1462
1463                 /* Clear trashed inode? */
1464                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1465                     inode->i_links_count > 0 &&
1466                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1467                         pctx.errcode = 0;
1468                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1469                 }
1470                 failed_csum = pctx.errcode != 0;
1471
1472                 /*
1473                  * Check for inodes who might have been part of the
1474                  * orphaned list linked list.  They should have gotten
1475                  * dealt with by now, unless the list had somehow been
1476                  * corrupted.
1477                  *
1478                  * FIXME: In the future, inodes which are still in use
1479                  * (and which are therefore) pending truncation should
1480                  * be handled specially.  Right now we just clear the
1481                  * dtime field, and the normal e2fsck handling of
1482                  * inodes where i_size and the inode blocks are
1483                  * inconsistent is to fix i_size, instead of releasing
1484                  * the extra blocks.  This won't catch the inodes that
1485                  * was at the end of the orphan list, but it's better
1486                  * than nothing.  The right answer is that there
1487                  * shouldn't be any bugs in the orphan list handling.  :-)
1488                  */
1489                 if (inode->i_dtime && low_dtime_check &&
1490                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1491                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1492                                 inode->i_dtime = inode->i_links_count ?
1493                                         0 : ctx->now;
1494                                 e2fsck_write_inode(ctx, ino, inode,
1495                                                    "pass1");
1496                                 failed_csum = 0;
1497                         }
1498                 }
1499
1500                 if (inode->i_links_count) {
1501                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1502                                            ino, inode->i_links_count);
1503                         if (pctx.errcode) {
1504                                 pctx.num = inode->i_links_count;
1505                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1506                                 ctx->flags |= E2F_FLAG_ABORT;
1507                                 goto endit;
1508                         }
1509                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1510                            !quota_inum_is_reserved(fs, ino)) {
1511                         if (!inode->i_dtime && inode->i_mode) {
1512                                 if (fix_problem(ctx,
1513                                             PR_1_ZERO_DTIME, &pctx)) {
1514                                         inode->i_dtime = ctx->now;
1515                                         e2fsck_write_inode(ctx, ino, inode,
1516                                                            "pass1");
1517                                         failed_csum = 0;
1518                                 }
1519                         }
1520                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1521                         continue;
1522                 }
1523
1524                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1525                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1526                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1527                      (!casefold_fs &&
1528                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1529                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1530                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1531                 }
1532
1533                 /* Conflicting inlinedata/extents inode flags? */
1534                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1535                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1536                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1537                                                                inode_size,
1538                                                                &pctx);
1539                         if (res < 0) {
1540                                 /* skip FINISH_INODE_LOOP */
1541                                 continue;
1542                         }
1543                 }
1544
1545                 /* Test for incorrect inline_data flags settings. */
1546                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1547                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1548                         size_t size = 0;
1549
1550                         pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1551                         if (!pctx.errcode &&
1552                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1553                                 ext2fs_set_feature_inline_data(sb);
1554                                 ext2fs_mark_super_dirty(fs);
1555                                 inlinedata_fs = 1;
1556                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1557                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1558                                 /* skip FINISH_INODE_LOOP */
1559                                 continue;
1560                         }
1561                 }
1562
1563                 /* Test for inline data flag but no attr */
1564                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1565                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1566                         size_t size = 0;
1567                         errcode_t err;
1568                         int flags;
1569
1570                         flags = fs->flags;
1571                         if (failed_csum)
1572                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1573                         err = get_inline_data_ea_size(fs, ino, &size);
1574                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1575                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1576
1577                         switch (err) {
1578                         case 0:
1579                                 /* Everything is awesome... */
1580                                 break;
1581                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1582                         case EXT2_ET_BAD_EA_HASH:
1583                         case EXT2_ET_BAD_EA_HEADER:
1584                         case EXT2_ET_EA_BAD_NAME_LEN:
1585                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1586                         case EXT2_ET_EA_KEY_NOT_FOUND:
1587                         case EXT2_ET_EA_NO_SPACE:
1588                         case EXT2_ET_MISSING_EA_FEATURE:
1589                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1590                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1591                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1592                         case EXT2_ET_NO_INLINE_DATA:
1593                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1594                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1595                         case EXT2_ET_EA_INODE_CORRUPTED:
1596                                 /* broken EA or no system.data EA; truncate */
1597                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1598                                                 &pctx)) {
1599                                         err = ext2fs_inode_size_set(fs, inode, 0);
1600                                         if (err) {
1601                                                 pctx.errcode = err;
1602                                                 ctx->flags |= E2F_FLAG_ABORT;
1603                                                 goto endit;
1604                                         }
1605                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1606                                         memset(&inode->i_block, 0,
1607                                                sizeof(inode->i_block));
1608                                         e2fsck_write_inode(ctx, ino, inode,
1609                                                            "pass1");
1610                                         failed_csum = 0;
1611                                 }
1612                                 break;
1613                         default:
1614                                 /* Some other kind of non-xattr error? */
1615                                 pctx.errcode = err;
1616                                 ctx->flags |= E2F_FLAG_ABORT;
1617                                 goto endit;
1618                         }
1619                 }
1620
1621                 /*
1622                  * Test for incorrect extent flag settings.
1623                  *
1624                  * On big-endian machines we must be careful:
1625                  * When the inode is read, the i_block array is not swapped
1626                  * if the extent flag is set.  Therefore if we are testing
1627                  * for or fixing a wrongly-set flag, we must potentially
1628                  * (un)swap before testing, or after fixing.
1629                  */
1630
1631                 /*
1632                  * In this case the extents flag was set when read, so
1633                  * extent_header_verify is ok.  If the inode is cleared,
1634                  * no need to swap... so no extra swapping here.
1635                  */
1636                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1637                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1638                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1639                         if ((ext2fs_extent_header_verify(inode->i_block,
1640                                                  sizeof(inode->i_block)) == 0) &&
1641                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1642                                 ext2fs_set_feature_extents(sb);
1643                                 ext2fs_mark_super_dirty(fs);
1644                                 extent_fs = 1;
1645                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1646                         clear_inode:
1647                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1648                                 if (ino == EXT2_BAD_INO)
1649                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1650                                                                  ino);
1651                                 /* skip FINISH_INODE_LOOP */
1652                                 continue;
1653                         }
1654                 }
1655
1656                 /*
1657                  * For big-endian machines:
1658                  * If the inode didn't have the extents flag set when it
1659                  * was read, then the i_blocks array was swapped.  To test
1660                  * as an extents header, we must swap it back first.
1661                  * IF we then set the extents flag, the entire i_block
1662                  * array must be un/re-swapped to make it proper extents data.
1663                  */
1664                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1665                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1666                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1667                     (LINUX_S_ISREG(inode->i_mode) ||
1668                      LINUX_S_ISDIR(inode->i_mode))) {
1669                         void *ehp;
1670 #ifdef WORDS_BIGENDIAN
1671                         __u32 tmp_block[EXT2_N_BLOCKS];
1672
1673                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1674                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1675                         ehp = tmp_block;
1676 #else
1677                         ehp = inode->i_block;
1678 #endif
1679                         if ((ext2fs_extent_header_verify(ehp,
1680                                          sizeof(inode->i_block)) == 0) &&
1681                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1682                                 inode->i_flags |= EXT4_EXTENTS_FL;
1683 #ifdef WORDS_BIGENDIAN
1684                                 memcpy(inode->i_block, tmp_block,
1685                                        sizeof(inode->i_block));
1686 #endif
1687                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1688                                 failed_csum = 0;
1689                         }
1690                 }
1691
1692                 if (ino == EXT2_BAD_INO) {
1693                         struct process_block_struct pb;
1694
1695                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1696                              inode->i_gid || inode->i_links_count ||
1697                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1698                              inode->i_file_acl) &&
1699                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1700                                 memset(inode, 0, sizeof(struct ext2_inode));
1701                                 e2fsck_write_inode(ctx, ino, inode,
1702                                                    "clear bad inode");
1703                                 failed_csum = 0;
1704                         }
1705
1706                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1707                                                           &pb.fs_meta_blocks);
1708                         if (pctx.errcode) {
1709                                 pctx.num = 4;
1710                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1711                                 ctx->flags |= E2F_FLAG_ABORT;
1712                                 goto endit;
1713                         }
1714                         pb.ino = EXT2_BAD_INO;
1715                         pb.num_blocks = pb.last_block = 0;
1716                         pb.last_db_block = -1;
1717                         pb.num_illegal_blocks = 0;
1718                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1719                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1720                         pb.inode = inode;
1721                         pb.pctx = &pctx;
1722                         pb.ctx = ctx;
1723                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1724                                      block_buf, process_bad_block, &pb);
1725                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1726                         if (pctx.errcode) {
1727                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1728                                 ctx->flags |= E2F_FLAG_ABORT;
1729                                 goto endit;
1730                         }
1731                         if (pb.bbcheck)
1732                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1733                                 ctx->flags |= E2F_FLAG_ABORT;
1734                                 goto endit;
1735                         }
1736                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1737                         clear_problem_context(&pctx);
1738                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1739                         continue;
1740                 } else if (ino == EXT2_ROOT_INO) {
1741                         /*
1742                          * Make sure the root inode is a directory; if
1743                          * not, offer to clear it.  It will be
1744                          * regenerated in pass #3.
1745                          */
1746                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1747                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1748                                         goto clear_inode;
1749                         }
1750                         /*
1751                          * If dtime is set, offer to clear it.  mke2fs
1752                          * version 0.2b created filesystems with the
1753                          * dtime field set for the root and lost+found
1754                          * directories.  We won't worry about
1755                          * /lost+found, since that can be regenerated
1756                          * easily.  But we will fix the root directory
1757                          * as a special case.
1758                          */
1759                         if (inode->i_dtime && inode->i_links_count) {
1760                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1761                                         inode->i_dtime = 0;
1762                                         e2fsck_write_inode(ctx, ino, inode,
1763                                                            "pass1");
1764                                         failed_csum = 0;
1765                                 }
1766                         }
1767                 } else if (ino == EXT2_JOURNAL_INO) {
1768                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1769                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1770                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1771                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1772                                                 &pctx)) {
1773                                         inode->i_mode = LINUX_S_IFREG;
1774                                         e2fsck_write_inode(ctx, ino, inode,
1775                                                            "pass1");
1776                                         failed_csum = 0;
1777                                 }
1778                                 check_blocks(ctx, &pctx, block_buf, NULL);
1779                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1780                                 continue;
1781                         }
1782                         if ((inode->i_links_count ||
1783                              inode->i_blocks || inode->i_block[0]) &&
1784                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1785                                         &pctx)) {
1786                                 memset(inode, 0, inode_size);
1787                                 ext2fs_icount_store(ctx->inode_link_info,
1788                                                     ino, 0);
1789                                 e2fsck_write_inode_full(ctx, ino, inode,
1790                                                         inode_size, "pass1");
1791                                 failed_csum = 0;
1792                         }
1793                 } else if (quota_inum_is_reserved(fs, ino)) {
1794                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1795                         if (ext2fs_has_feature_quota(fs->super) &&
1796                             quota_inum_is_super(fs->super, ino)) {
1797                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1798                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1799                                                         &pctx)) {
1800                                         inode->i_mode = LINUX_S_IFREG;
1801                                         e2fsck_write_inode(ctx, ino, inode,
1802                                                         "pass1");
1803                                         failed_csum = 0;
1804                                 }
1805                                 check_blocks(ctx, &pctx, block_buf, NULL);
1806                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1807                                 continue;
1808                         }
1809                         if ((inode->i_links_count ||
1810                              inode->i_blocks || inode->i_block[0]) &&
1811                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1812                                         &pctx)) {
1813                                 memset(inode, 0, inode_size);
1814                                 ext2fs_icount_store(ctx->inode_link_info,
1815                                                     ino, 0);
1816                                 e2fsck_write_inode_full(ctx, ino, inode,
1817                                                         inode_size, "pass1");
1818                                 failed_csum = 0;
1819                         }
1820                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1821                         problem_t problem = 0;
1822
1823                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1824                         if (ino == EXT2_BOOT_LOADER_INO) {
1825                                 if (LINUX_S_ISDIR(inode->i_mode))
1826                                         problem = PR_1_RESERVED_BAD_MODE;
1827                         } else if (ino == EXT2_RESIZE_INO) {
1828                                 if (inode->i_mode &&
1829                                     !LINUX_S_ISREG(inode->i_mode))
1830                                         problem = PR_1_RESERVED_BAD_MODE;
1831                         } else {
1832                                 if (inode->i_mode != 0)
1833                                         problem = PR_1_RESERVED_BAD_MODE;
1834                         }
1835                         if (problem) {
1836                                 if (fix_problem(ctx, problem, &pctx)) {
1837                                         inode->i_mode = 0;
1838                                         e2fsck_write_inode(ctx, ino, inode,
1839                                                            "pass1");
1840                                         failed_csum = 0;
1841                                 }
1842                         }
1843                         check_blocks(ctx, &pctx, block_buf, NULL);
1844                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1845                         continue;
1846                 }
1847
1848                 if (!inode->i_links_count) {
1849                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1850                         continue;
1851                 }
1852                 /*
1853                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1854                  * deleted files.  Oops.
1855                  *
1856                  * Since all new ext2 implementations get this right,
1857                  * we now assume that the case of non-zero
1858                  * i_links_count and non-zero dtime means that we
1859                  * should keep the file, not delete it.
1860                  *
1861                  */
1862                 if (inode->i_dtime) {
1863                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1864                                 inode->i_dtime = 0;
1865                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1866                                 failed_csum = 0;
1867                         }
1868                 }
1869
1870                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1871                 switch (fs->super->s_creator_os) {
1872                     case EXT2_OS_HURD:
1873                         frag = inode->osd2.hurd2.h_i_frag;
1874                         fsize = inode->osd2.hurd2.h_i_fsize;
1875                         break;
1876                     default:
1877                         frag = fsize = 0;
1878                 }
1879
1880                 if (inode->i_faddr || frag || fsize ||
1881                     (!ext2fs_has_feature_largedir(fs->super) &&
1882                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1883                         mark_inode_bad(ctx, ino);
1884                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1885                     !ext2fs_has_feature_64bit(fs->super) &&
1886                     inode->osd2.linux2.l_i_file_acl_high != 0)
1887                         mark_inode_bad(ctx, ino);
1888                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1889                     !ext2fs_has_feature_huge_file(fs->super) &&
1890                     (inode->osd2.linux2.l_i_blocks_hi != 0))
1891                         mark_inode_bad(ctx, ino);
1892                 if (inode->i_flags & EXT2_IMAGIC_FL) {
1893                         if (imagic_fs) {
1894                                 if (!ctx->inode_imagic_map)
1895                                         alloc_imagic_map(ctx);
1896                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1897                                                          ino);
1898                         } else {
1899                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1900                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
1901                                         e2fsck_write_inode(ctx, ino,
1902                                                            inode, "pass1");
1903                                         failed_csum = 0;
1904                                 }
1905                         }
1906                 }
1907
1908                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1909                 check_is_really_dir(ctx, &pctx, block_buf);
1910
1911                 /*
1912                  * ext2fs_inode_has_valid_blocks2 does not actually look
1913                  * at i_block[] values, so not endian-sensitive here.
1914                  */
1915                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1916                     LINUX_S_ISLNK(inode->i_mode) &&
1917                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1918                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1919                         inode->i_flags &= ~EXT4_EXTENTS_FL;
1920                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1921                         failed_csum = 0;
1922                 }
1923
1924                 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
1925                     add_encrypted_file(ctx, &pctx) < 0)
1926                         goto clear_inode;
1927
1928                 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
1929                         ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
1930
1931                 if (LINUX_S_ISDIR(inode->i_mode)) {
1932                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1933                         e2fsck_add_dir_info(ctx, ino, 0);
1934                         ctx->fs_directory_count++;
1935                         if (inode->i_flags & EXT4_CASEFOLD_FL)
1936                                 add_casefolded_dir(ctx, ino);
1937                 } else if (LINUX_S_ISREG (inode->i_mode)) {
1938                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1939                         ctx->fs_regular_count++;
1940                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1941                            e2fsck_pass1_check_device_inode(fs, inode)) {
1942                         check_extents_inlinedata(ctx, &pctx);
1943                         check_immutable(ctx, &pctx);
1944                         check_size(ctx, &pctx);
1945                         ctx->fs_chardev_count++;
1946                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1947                            e2fsck_pass1_check_device_inode(fs, inode)) {
1948                         check_extents_inlinedata(ctx, &pctx);
1949                         check_immutable(ctx, &pctx);
1950                         check_size(ctx, &pctx);
1951                         ctx->fs_blockdev_count++;
1952                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1953                            e2fsck_pass1_check_symlink(fs, ino, inode,
1954                                                       block_buf)) {
1955                         check_immutable(ctx, &pctx);
1956                         ctx->fs_symlinks_count++;
1957                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1958                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1959                                 continue;
1960                         } else if (ext2fs_is_fast_symlink(inode)) {
1961                                 ctx->fs_fast_symlinks_count++;
1962                                 check_blocks(ctx, &pctx, block_buf,
1963                                              &ea_ibody_quota);
1964                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1965                                 continue;
1966                         }
1967                 }
1968                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
1969                          e2fsck_pass1_check_device_inode(fs, inode)) {
1970                         check_extents_inlinedata(ctx, &pctx);
1971                         check_immutable(ctx, &pctx);
1972                         check_size(ctx, &pctx);
1973                         ctx->fs_fifo_count++;
1974                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
1975                            e2fsck_pass1_check_device_inode(fs, inode)) {
1976                         check_extents_inlinedata(ctx, &pctx);
1977                         check_immutable(ctx, &pctx);
1978                         check_size(ctx, &pctx);
1979                         ctx->fs_sockets_count++;
1980                 } else
1981                         mark_inode_bad(ctx, ino);
1982                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1983                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
1984                         if (inode->i_block[EXT2_IND_BLOCK])
1985                                 ctx->fs_ind_count++;
1986                         if (inode->i_block[EXT2_DIND_BLOCK])
1987                                 ctx->fs_dind_count++;
1988                         if (inode->i_block[EXT2_TIND_BLOCK])
1989                                 ctx->fs_tind_count++;
1990                 }
1991                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1992                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
1993                     (inode->i_block[EXT2_IND_BLOCK] ||
1994                      inode->i_block[EXT2_DIND_BLOCK] ||
1995                      inode->i_block[EXT2_TIND_BLOCK] ||
1996                      ext2fs_file_acl_block(fs, inode))) {
1997                         struct process_inode_block *itp;
1998
1999                         itp = &inodes_to_process[process_inode_count];
2000                         itp->ino = ino;
2001                         itp->ea_ibody_quota = ea_ibody_quota;
2002                         if (inode_size < sizeof(struct ext2_inode_large))
2003                                 memcpy(&itp->inode, inode, inode_size);
2004                         else
2005                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
2006                         process_inode_count++;
2007                 } else
2008                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
2009
2010                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2011
2012                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2013                         goto endit;
2014
2015                 if (process_inode_count >= ctx->process_inode_size) {
2016                         process_inodes(ctx, block_buf, inodes_to_process,
2017                                        &process_inode_count);
2018
2019                         if (e2fsck_should_abort(ctx))
2020                                 goto endit;
2021                 }
2022         }
2023         process_inodes(ctx, block_buf, inodes_to_process,
2024                        &process_inode_count);
2025         ext2fs_close_inode_scan(scan);
2026         scan = NULL;
2027
2028         reserve_block_for_root_repair(ctx);
2029         reserve_block_for_lnf_repair(ctx);
2030
2031         /*
2032          * If any extended attribute blocks' reference counts need to
2033          * be adjusted, either up (ctx->refcount_extra), or down
2034          * (ctx->refcount), then fix them.
2035          */
2036         if (ctx->refcount) {
2037                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
2038                 ea_refcount_free(ctx->refcount);
2039                 ctx->refcount = 0;
2040         }
2041         if (ctx->refcount_extra) {
2042                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
2043                                         block_buf, +1);
2044                 ea_refcount_free(ctx->refcount_extra);
2045                 ctx->refcount_extra = 0;
2046         }
2047
2048         if (ctx->ea_block_quota_blocks) {
2049                 ea_refcount_free(ctx->ea_block_quota_blocks);
2050                 ctx->ea_block_quota_blocks = 0;
2051         }
2052
2053         if (ctx->ea_block_quota_inodes) {
2054                 ea_refcount_free(ctx->ea_block_quota_inodes);
2055                 ctx->ea_block_quota_inodes = 0;
2056         }
2057
2058         if (ctx->invalid_bitmaps)
2059                 handle_fs_bad_blocks(ctx);
2060
2061         /* We don't need the block_ea_map any more */
2062         if (ctx->block_ea_map) {
2063                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2064                 ctx->block_ea_map = 0;
2065         }
2066
2067         /* We don't need the encryption policy => ID map any more */
2068         destroy_encryption_policy_map(ctx);
2069
2070         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2071                 clear_problem_context(&pctx);
2072                 pctx.errcode = ext2fs_create_resize_inode(fs);
2073                 if (pctx.errcode) {
2074                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2075                                          &pctx)) {
2076                                 ctx->flags |= E2F_FLAG_ABORT;
2077                                 goto endit;
2078                         }
2079                         pctx.errcode = 0;
2080                 }
2081                 if (!pctx.errcode) {
2082                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2083                                           "recreate inode");
2084                         inode->i_mtime = ctx->now;
2085                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2086                                            "recreate inode");
2087                 }
2088                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2089         }
2090
2091         if (ctx->flags & E2F_FLAG_RESTART) {
2092                 /*
2093                  * Only the master copy of the superblock and block
2094                  * group descriptors are going to be written during a
2095                  * restart, so set the superblock to be used to be the
2096                  * master superblock.
2097                  */
2098                 ctx->use_superblock = 0;
2099                 goto endit;
2100         }
2101
2102         if (ctx->large_dirs && !ext2fs_has_feature_largedir(fs->super)) {
2103                 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2104                         ext2fs_set_feature_largedir(fs->super);
2105                         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2106                         ext2fs_mark_super_dirty(fs);
2107                 }
2108                 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2109                     fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2110                         ext2fs_update_dynamic_rev(fs);
2111                         ext2fs_mark_super_dirty(fs);
2112                 }
2113         }
2114
2115         if (ctx->block_dup_map) {
2116                 if (ctx->options & E2F_OPT_PREEN) {
2117                         clear_problem_context(&pctx);
2118                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2119                 }
2120                 e2fsck_pass1_dupblocks(ctx, block_buf);
2121         }
2122         ctx->flags |= E2F_FLAG_ALLOC_OK;
2123 endit:
2124         e2fsck_use_inode_shortcuts(ctx, 0);
2125         ext2fs_free_mem(&inodes_to_process);
2126         inodes_to_process = 0;
2127
2128         if (scan)
2129                 ext2fs_close_inode_scan(scan);
2130         if (block_buf)
2131                 ext2fs_free_mem(&block_buf);
2132         if (inode)
2133                 ext2fs_free_mem(&inode);
2134
2135         /*
2136          * The l+f inode may have been cleared, so zap it now and
2137          * later passes will recalculate it if necessary
2138          */
2139         ctx->lost_and_found = 0;
2140
2141         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2142                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2143         else
2144                 ctx->invalid_bitmaps++;
2145 }
2146
2147 static void init_ext2_max_sizes()
2148 {
2149         int     i;
2150         __u64   max_sizes;
2151
2152         /*
2153          * Init ext2_max_sizes which will be immutable and shared between
2154          * threads
2155          */
2156 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
2157
2158         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
2159                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
2160                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
2161                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
2162                 max_sizes = (max_sizes * (1UL << i));
2163                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
2164         }
2165 #undef EXT2_BPP
2166 }
2167
2168 #ifdef HAVE_PTHREAD
2169 static errcode_t e2fsck_pass1_copy_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2170                                           ext2fs_generic_bitmap *dest)
2171 {
2172         errcode_t ret;
2173
2174         ret = ext2fs_copy_bitmap(*src, dest);
2175         if (ret)
2176                 return ret;
2177
2178         (*dest)->fs = fs;
2179
2180         return 0;
2181 }
2182
2183 static void e2fsck_pass1_free_bitmap(ext2fs_generic_bitmap *bitmap)
2184 {
2185         if (*bitmap) {
2186                 ext2fs_free_generic_bmap(*bitmap);
2187                 *bitmap = NULL;
2188         }
2189
2190 }
2191
2192 static errcode_t e2fsck_pass1_merge_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2193                                           ext2fs_generic_bitmap *dest)
2194 {
2195         errcode_t ret = 0;
2196
2197         if (*src) {
2198                 if (*dest == NULL) {
2199                         *dest = *src;
2200                         *src = NULL;
2201                 } else {
2202                         ret = ext2fs_merge_bitmap(*src, *dest, NULL, NULL);
2203                         if (ret)
2204                                 return ret;
2205                 }
2206                 (*dest)->fs = fs;
2207         }
2208
2209         return 0;
2210 }
2211
2212 static errcode_t e2fsck_pass1_copy_fs(ext2_filsys dest, e2fsck_t src_context,
2213                                       ext2_filsys src)
2214 {
2215         errcode_t       retval;
2216
2217         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2218         dest->inode_map = NULL;
2219         dest->block_map = NULL;
2220         dest->badblocks = NULL;
2221         if (dest->dblist)
2222                 dest->dblist->fs = dest;
2223         if (src->block_map) {
2224                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2225                                                   &dest->block_map);
2226                 if (retval)
2227                         return retval;
2228         }
2229         if (src->inode_map) {
2230                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2231                                                   &dest->inode_map);
2232                 if (retval)
2233                         return retval;
2234         }
2235
2236         if (src->badblocks) {
2237                 retval = ext2fs_badblocks_copy(src->badblocks,
2238                                                &dest->badblocks);
2239                 if (retval)
2240                         return retval;
2241         }
2242
2243         /* disable it for now */
2244         src_context->openfs_flags &= ~EXT2_FLAG_EXCLUSIVE;
2245         retval = ext2fs_open_channel(dest, src_context->io_options,
2246                                      src_context->io_manager,
2247                                      src_context->openfs_flags,
2248                                      src->io->block_size);
2249         if (retval)
2250                 return retval;
2251
2252         /* Block size might not be default */
2253         io_channel_set_blksize(dest->io, src->io->block_size);
2254         ehandler_init(dest->io);
2255
2256         assert(dest->io->magic == src->io->magic);
2257         assert(dest->io->manager == src->io->manager);
2258         assert(strcmp(dest->io->name, src->io->name) == 0);
2259         assert(dest->io->block_size == src->io->block_size);
2260         assert(dest->io->read_error == src->io->read_error);
2261         assert(dest->io->write_error == src->io->write_error);
2262         assert(dest->io->refcount == src->io->refcount);
2263         assert(dest->io->flags == src->io->flags);
2264         assert(dest->io->app_data == dest);
2265         assert(src->io->app_data == src);
2266         assert(dest->io->align == src->io->align);
2267
2268         /* The data should be written to disk immediately */
2269         dest->io->flags |= CHANNEL_FLAGS_WRITETHROUGH;
2270         /* icache will be rebuilt if needed, so do not copy from @src */
2271         src->icache = NULL;
2272         return 0;
2273 }
2274
2275 static int e2fsck_pass1_merge_fs(ext2_filsys dest, ext2_filsys src)
2276 {
2277         struct ext2_inode_cache *icache = dest->icache;
2278         errcode_t retval = 0;
2279         io_channel dest_io;
2280         io_channel dest_image_io;
2281         ext2fs_inode_bitmap inode_map;
2282         ext2fs_block_bitmap block_map;
2283         ext2_badblocks_list badblocks;
2284         ext2_dblist dblist;
2285
2286         dest_io = dest->io;
2287         dest_image_io = dest->image_io;
2288         inode_map = dest->inode_map;
2289         block_map = dest->block_map;
2290         badblocks = dest->badblocks;
2291         dblist = dest->dblist;
2292
2293         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2294         dest->io = dest_io;
2295         dest->image_io = dest_image_io;
2296         dest->icache = icache;
2297         dest->inode_map = inode_map;
2298         dest->block_map = block_map;
2299         dest->badblocks = badblocks;
2300         dest->dblist = dblist;
2301         if (dest->dblist)
2302                 dest->dblist->fs = dest;
2303
2304         if (src->icache) {
2305                 ext2fs_free_inode_cache(src->icache);
2306                 src->icache = NULL;
2307         }
2308
2309         retval = e2fsck_pass1_merge_bitmap(dest, &src->inode_map,
2310                                            &dest->inode_map);
2311         if (retval)
2312                 goto out;
2313
2314         retval = e2fsck_pass1_merge_bitmap(dest, &src->block_map,
2315                                           &dest->block_map);
2316         if (retval)
2317                 goto out;
2318
2319         if (src->dblist) {
2320                 if (dest->dblist) {
2321                         retval = ext2fs_merge_dblist(src->dblist,
2322                                                      dest->dblist);
2323                         if (retval)
2324                                 goto out;
2325                 } else {
2326                         dest->dblist = src->dblist;
2327                         dest->dblist->fs = dest;
2328                         src->dblist = NULL;
2329                 }
2330         }
2331
2332         if (src->badblocks) {
2333                 if (dest->badblocks == NULL)
2334                         retval = ext2fs_badblocks_copy(src->badblocks,
2335                                                        &dest->badblocks);
2336                 else
2337                         retval = ext2fs_badblocks_merge(src->badblocks,
2338                                                         dest->badblocks);
2339         }
2340 out:
2341         io_channel_close(src->io);
2342         if (src->inode_map)
2343                 ext2fs_free_generic_bmap(src->inode_map);
2344         if (src->block_map)
2345                 ext2fs_free_generic_bmap(src->block_map);
2346         if (src->badblocks)
2347                 ext2fs_badblocks_list_free(src->badblocks);
2348         if (src->dblist)
2349                 ext2fs_free_dblist(src->dblist);
2350
2351         return retval;
2352 }
2353
2354 static errcode_t e2fsck_pass1_thread_prepare(e2fsck_t global_ctx, e2fsck_t *thread_ctx,
2355                                              int thread_index, int num_threads)
2356 {
2357         errcode_t               retval;
2358         e2fsck_t                thread_context;
2359         ext2_filsys             thread_fs;
2360         ext2_filsys             global_fs = global_ctx->fs;
2361         struct e2fsck_thread    *tinfo;
2362         dgrp_t                  average_group;
2363
2364         assert(global_ctx->inode_used_map == NULL);
2365         assert(global_ctx->inode_dir_map == NULL);
2366         assert(global_ctx->inode_bb_map == NULL);
2367         assert(global_ctx->inode_imagic_map == NULL);
2368         assert(global_ctx->inode_reg_map == NULL);
2369         assert(global_ctx->inodes_to_rebuild == NULL);
2370
2371         assert(global_ctx->block_found_map == NULL);
2372         assert(global_ctx->block_dup_map == NULL);
2373         assert(global_ctx->block_ea_map == NULL);
2374         assert(global_ctx->block_metadata_map == NULL);
2375         assert(global_ctx->fs->dblist == NULL);
2376
2377         retval = ext2fs_get_mem(sizeof(struct e2fsck_struct), &thread_context);
2378         if (retval) {
2379                 com_err(global_ctx->program_name, retval, "while allocating memory");
2380                 return retval;
2381         }
2382         memcpy(thread_context, global_ctx, sizeof(struct e2fsck_struct));
2383         thread_context->global_ctx = global_ctx;
2384
2385         retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &thread_fs);
2386         if (retval) {
2387                 com_err(global_ctx->program_name, retval, "while allocating memory");
2388                 goto out_context;
2389         }
2390
2391         io_channel_flush_cleanup(global_fs->io);
2392         retval = e2fsck_pass1_copy_fs(thread_fs, global_ctx, global_fs);
2393         if (retval) {
2394                 com_err(global_ctx->program_name, retval, "while copying fs");
2395                 goto out_fs;
2396         }
2397         thread_fs->priv_data = thread_context;
2398
2399         thread_context->thread_info.et_thread_index = thread_index;
2400         set_up_logging(thread_context);
2401
2402         /*
2403          * Distribute work to multiple threads:
2404          * Each thread work on fs->group_desc_count / nthread groups.
2405          */
2406         tinfo = &thread_context->thread_info;
2407         average_group = thread_fs->group_desc_count / num_threads;
2408         if (average_group == 0)
2409                 average_group = 1;
2410         tinfo->et_group_start = average_group * thread_index;
2411         if (thread_index == num_threads - 1)
2412                 tinfo->et_group_end = thread_fs->group_desc_count;
2413         else
2414                 tinfo->et_group_end = average_group * (thread_index + 1);
2415         tinfo->et_group_next = tinfo->et_group_start;
2416         tinfo->et_inode_number = 0;
2417         tinfo->et_log_buf[0] = '\0';
2418         tinfo->et_log_length = 0;
2419         if (thread_context->options & E2F_OPT_MULTITHREAD)
2420                 log_out(thread_context, _("Scan group range [%d, %d)\n"),
2421                         tinfo->et_group_start, tinfo->et_group_end);
2422         thread_context->fs = thread_fs;
2423         *thread_ctx = thread_context;
2424         return 0;
2425 out_fs:
2426         ext2fs_free_mem(&thread_fs);
2427 out_context:
2428         ext2fs_free_mem(&thread_context);
2429         return retval;
2430 }
2431
2432 static void e2fsck_pass1_merge_dir_info(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2433 {
2434         if (thread_ctx->dir_info == NULL)
2435                 return;
2436
2437         if (global_ctx->dir_info == NULL) {
2438                 global_ctx->dir_info = thread_ctx->dir_info;
2439                 thread_ctx->dir_info = NULL;
2440                 return;
2441         }
2442
2443         e2fsck_merge_dir_info(global_ctx, thread_ctx->dir_info,
2444                               global_ctx->dir_info);
2445 }
2446
2447 static inline errcode_t
2448 e2fsck_pass1_merge_icount(ext2_icount_t *dest_icount,
2449                           ext2_icount_t *src_icount)
2450 {
2451         if (*src_icount) {
2452                 if (*dest_icount == NULL) {
2453                         *dest_icount = *src_icount;
2454                         *src_icount = NULL;
2455                 } else {
2456                         errcode_t ret;
2457
2458                         ret = ext2fs_icount_merge(*src_icount,
2459                                                   *dest_icount);
2460                         if (ret)
2461                                 return ret;
2462                 }
2463         }
2464
2465         return 0;
2466 }
2467
2468 static errcode_t e2fsck_pass1_merge_icounts(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2469 {
2470         errcode_t ret;
2471
2472         ret = e2fsck_pass1_merge_icount(&global_ctx->inode_count,
2473                                         &thread_ctx->inode_count);
2474         if (ret)
2475                 return ret;
2476         ret = e2fsck_pass1_merge_icount(&global_ctx->inode_link_info,
2477                                         &thread_ctx->inode_link_info);
2478
2479         return ret;
2480 }
2481
2482 static int e2fsck_pass1_thread_join_one(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2483 {
2484         errcode_t        retval;
2485         int              flags = global_ctx->flags;
2486         ext2_filsys      thread_fs = thread_ctx->fs;
2487         ext2_filsys      global_fs = global_ctx->fs;
2488         FILE            *global_logf = global_ctx->logf;
2489         FILE            *global_problem_logf = global_ctx->problem_logf;
2490         ext2fs_inode_bitmap inode_bad_map = global_ctx->inode_bad_map;
2491         struct dir_info_db *dir_info = global_ctx->dir_info;
2492         ext2fs_inode_bitmap inode_used_map = global_ctx->inode_used_map;
2493         ext2fs_inode_bitmap inode_dir_map = global_ctx->inode_dir_map;
2494         ext2fs_inode_bitmap inode_bb_map = global_ctx->inode_bb_map;
2495         ext2fs_inode_bitmap inode_imagic_map = global_ctx->inode_imagic_map;
2496         ext2fs_inode_bitmap inode_reg_map = global_ctx->inode_reg_map;
2497         ext2fs_block_bitmap block_found_map = global_ctx->block_found_map;
2498         ext2fs_block_bitmap block_dup_map = global_ctx->block_dup_map;
2499         ext2fs_block_bitmap block_ea_map = global_ctx->block_ea_map;
2500         ext2fs_block_bitmap block_metadata_map = global_ctx->block_metadata_map;
2501         ext2fs_block_bitmap inodes_to_rebuild = global_ctx->inodes_to_rebuild;
2502         ext2_icount_t inode_count = global_ctx->inode_count;
2503         ext2_icount_t inode_link_info = global_ctx->inode_link_info;
2504         __u32 fs_directory_count = global_ctx->fs_directory_count;
2505         __u32 fs_regular_count = global_ctx->fs_regular_count;
2506         __u32 fs_blockdev_count = global_ctx->fs_blockdev_count;
2507         __u32 fs_chardev_count = global_ctx->fs_chardev_count;
2508         __u32 fs_links_count = global_ctx->fs_links_count;
2509         __u32 fs_symlinks_count = global_ctx->fs_symlinks_count;
2510         __u32 fs_fast_symlinks_count = global_ctx->fs_fast_symlinks_count;
2511         __u32 fs_fifo_count = global_ctx->fs_fifo_count;
2512         __u32 fs_total_count = global_ctx->fs_total_count;
2513         __u32 fs_badblocks_count = global_ctx->fs_badblocks_count;
2514         __u32 fs_sockets_count = global_ctx->fs_sockets_count;
2515         __u32 fs_ind_count = global_ctx->fs_ind_count;
2516         __u32 fs_dind_count = global_ctx->fs_dind_count;
2517         __u32 fs_tind_count = global_ctx->fs_tind_count;
2518         __u32 fs_fragmented = global_ctx->fs_fragmented;
2519         __u32 fs_fragmented_dir = global_ctx->fs_fragmented_dir;
2520         __u32 large_files = global_ctx->large_files;
2521
2522 #ifdef HAVE_SETJMP_H
2523         jmp_buf          old_jmp;
2524
2525         memcpy(old_jmp, global_ctx->abort_loc, sizeof(jmp_buf));
2526 #endif
2527         memcpy(global_ctx, thread_ctx, sizeof(struct e2fsck_struct));
2528 #ifdef HAVE_SETJMP_H
2529         memcpy(global_ctx->abort_loc, old_jmp, sizeof(jmp_buf));
2530 #endif
2531
2532         global_ctx->inode_used_map = inode_used_map;
2533         global_ctx->inode_bad_map = inode_bad_map;
2534         global_ctx->inode_dir_map = inode_dir_map;
2535         global_ctx->inode_bb_map = inode_bb_map;
2536         global_ctx->inode_imagic_map = inode_imagic_map;
2537         global_ctx->inodes_to_rebuild = inodes_to_rebuild;
2538         global_ctx->inode_reg_map = inode_reg_map;
2539         global_ctx->block_found_map = block_found_map;
2540         global_ctx->block_dup_map = block_dup_map;
2541         global_ctx->block_ea_map = block_ea_map;
2542         global_ctx->block_metadata_map = block_metadata_map;
2543         global_ctx->dir_info = dir_info;
2544         e2fsck_pass1_merge_dir_info(global_ctx, thread_ctx);
2545         global_ctx->inode_count = inode_count;
2546         global_ctx->inode_link_info = inode_link_info;
2547         global_ctx->fs_directory_count += fs_directory_count;
2548         global_ctx->fs_regular_count += fs_regular_count;
2549         global_ctx->fs_blockdev_count += fs_blockdev_count;
2550         global_ctx->fs_chardev_count += fs_chardev_count;
2551         global_ctx->fs_links_count += fs_links_count;
2552         global_ctx->fs_symlinks_count += fs_symlinks_count;
2553         global_ctx->fs_fast_symlinks_count += fs_fast_symlinks_count;
2554         global_ctx->fs_fifo_count += fs_fifo_count;
2555         global_ctx->fs_total_count += fs_total_count;
2556         global_ctx->fs_badblocks_count += fs_badblocks_count;
2557         global_ctx->fs_sockets_count += fs_sockets_count;
2558         global_ctx->fs_ind_count += fs_ind_count;
2559         global_ctx->fs_dind_count += fs_dind_count;
2560         global_ctx->fs_tind_count += fs_tind_count;
2561         global_ctx->fs_fragmented += fs_fragmented;
2562         global_ctx->fs_fragmented_dir += fs_fragmented_dir;
2563         global_ctx->large_files += large_files;
2564
2565         /* Keep the global singal flags*/
2566         global_ctx->flags |= (flags & E2F_FLAG_SIGNAL_MASK) |
2567                              (global_ctx->flags & E2F_FLAG_SIGNAL_MASK);
2568
2569         retval = e2fsck_pass1_merge_fs(global_fs, thread_fs);
2570         if (retval) {
2571                 com_err(global_ctx->program_name, 0, _("while merging fs\n"));
2572                 return retval;
2573         }
2574         global_fs->priv_data = global_ctx;
2575         global_ctx->fs = global_fs;
2576         global_ctx->logf = global_logf;
2577         global_ctx->problem_logf = global_problem_logf;
2578         global_ctx->global_ctx = NULL;
2579         retval = e2fsck_pass1_merge_icounts(global_ctx, thread_ctx);
2580         if (retval) {
2581                 com_err(global_ctx->program_name, 0,
2582                         _("while merging icounts\n"));
2583                 return retval;
2584         }
2585
2586         retval = e2fsck_pass1_merge_bitmap(global_fs,
2587                                 &thread_ctx->inode_used_map,
2588                                 &global_ctx->inode_used_map);
2589         if (retval)
2590                 return retval;
2591
2592         retval = e2fsck_pass1_merge_bitmap(global_fs,
2593                                 &thread_ctx->inode_bad_map,
2594                                 &global_ctx->inode_bad_map);
2595         if (retval)
2596                 return retval;
2597         retval = e2fsck_pass1_merge_bitmap(global_fs,
2598                                         &thread_ctx->inode_dir_map,
2599                                         &global_ctx->inode_dir_map);
2600         if (retval)
2601                 return retval;
2602         retval = e2fsck_pass1_merge_bitmap(global_fs,
2603                                 &thread_ctx->inode_bb_map,
2604                                 &global_ctx->inode_bb_map);
2605         if (retval)
2606                 return retval;
2607         retval = e2fsck_pass1_merge_bitmap(global_fs,
2608                                 &thread_ctx->inode_imagic_map,
2609                                 &global_ctx->inode_imagic_map);
2610         if (retval)
2611                 return retval;
2612         retval = e2fsck_pass1_merge_bitmap(global_fs,
2613                                 &thread_ctx->inode_reg_map,
2614                                 &global_ctx->inode_reg_map);
2615         if (retval)
2616                 return retval;
2617         retval = e2fsck_pass1_merge_bitmap(global_fs,
2618                                 &thread_ctx->inodes_to_rebuild,
2619                                 &global_ctx->inodes_to_rebuild);
2620         if (retval)
2621                 return retval;
2622         retval = e2fsck_pass1_merge_bitmap(global_fs,
2623                                 &thread_ctx->block_found_map,
2624                                 &global_ctx->block_found_map);
2625         if (retval)
2626                 return retval;
2627         retval = e2fsck_pass1_merge_bitmap(global_fs,
2628                                 &thread_ctx->block_dup_map,
2629                                 &global_ctx->block_dup_map);
2630         if (retval)
2631                 return retval;
2632         retval = e2fsck_pass1_merge_bitmap(global_fs,
2633                                 &thread_ctx->block_ea_map,
2634                                 &global_ctx->block_ea_map);
2635         if (retval)
2636                 return retval;
2637         retval = e2fsck_pass1_merge_bitmap(global_fs,
2638                                 &thread_ctx->block_metadata_map,
2639                                 &global_ctx->block_metadata_map);
2640         if (retval)
2641                 return retval;
2642
2643         return 0;
2644 }
2645
2646 static int e2fsck_pass1_thread_join(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2647 {
2648         errcode_t       retval;
2649
2650         retval = e2fsck_pass1_thread_join_one(global_ctx, thread_ctx);
2651         ext2fs_free_mem(&thread_ctx->fs);
2652         if (thread_ctx->logf)
2653                 fclose(thread_ctx->logf);
2654         if (thread_ctx->problem_logf) {
2655                 fputs("</problem_log>\n", thread_ctx->problem_logf);
2656                 fclose(thread_ctx->problem_logf);
2657         }
2658         e2fsck_pass1_free_bitmap(&thread_ctx->inode_used_map);
2659         e2fsck_pass1_free_bitmap(&thread_ctx->inode_bad_map);
2660         e2fsck_pass1_free_bitmap(&thread_ctx->inode_dir_map);
2661         e2fsck_pass1_free_bitmap(&thread_ctx->inode_bb_map);
2662         e2fsck_pass1_free_bitmap(&thread_ctx->inode_imagic_map);
2663         e2fsck_pass1_free_bitmap(&thread_ctx->inode_reg_map);
2664         e2fsck_pass1_free_bitmap(&thread_ctx->inodes_to_rebuild);
2665         e2fsck_pass1_free_bitmap(&thread_ctx->block_found_map);
2666         e2fsck_pass1_free_bitmap(&thread_ctx->block_dup_map);
2667         e2fsck_pass1_free_bitmap(&thread_ctx->block_ea_map);
2668         e2fsck_pass1_free_bitmap(&thread_ctx->block_metadata_map);
2669         e2fsck_free_dir_info(thread_ctx);
2670         ext2fs_free_icount(thread_ctx->inode_count);
2671         ext2fs_free_icount(thread_ctx->inode_link_info);
2672         ext2fs_free_mem(&thread_ctx);
2673
2674         return retval;
2675 }
2676
2677 static int e2fsck_pass1_threads_join(struct e2fsck_thread_info *infos,
2678                                       int num_threads, e2fsck_t global_ctx)
2679 {
2680         errcode_t                        rc;
2681         errcode_t                        ret = 0;
2682         int                              i;
2683         struct e2fsck_thread_info       *pinfo;
2684
2685         for (i = 0; i < num_threads; i++) {
2686                 pinfo = &infos[i];
2687
2688                 if (!pinfo->eti_started)
2689                         continue;
2690
2691                 rc = pthread_join(pinfo->eti_thread_id, NULL);
2692                 if (rc) {
2693                         com_err(global_ctx->program_name, rc,
2694                                 _("while joining thread\n"));
2695                         if (ret == 0)
2696                                 ret = rc;
2697                 }
2698                 rc = e2fsck_pass1_thread_join(global_ctx, infos[i].eti_thread_ctx);
2699                 if (rc) {
2700                         com_err(global_ctx->program_name, rc,
2701                                 _("while joining pass1 thread\n"));
2702                         if (ret == 0)
2703                                 ret = rc;
2704                 }
2705         }
2706         free(infos);
2707
2708         return ret;
2709 }
2710
2711 static void *e2fsck_pass1_thread(void *arg)
2712 {
2713         struct e2fsck_thread_info       *info = arg;
2714         e2fsck_t                         thread_ctx = info->eti_thread_ctx;
2715 #ifdef DEBUG_THREADS
2716         struct e2fsck_thread_debug      *thread_debug = info->eti_debug;
2717 #endif
2718
2719 #ifdef DEBUG_THREADS
2720         pthread_mutex_lock(&thread_debug->etd_mutex);
2721         while (info->eti_thread_index > thread_debug->etd_finished_threads) {
2722                 pthread_cond_wait(&thread_debug->etd_cond,
2723                                   &thread_debug->etd_mutex);
2724         }
2725         pthread_mutex_unlock(&thread_debug->etd_mutex);
2726 #endif
2727
2728 #ifdef HAVE_SETJMP_H
2729         /*
2730          * When fatal_error() happens, jump to here. The thread
2731          * context's flags will be saved, but its abort_loc will
2732          * be overwritten by original jump buffer for the later
2733          * tests.
2734          */
2735         if (setjmp(thread_ctx->abort_loc)) {
2736                 thread_ctx->flags &= ~E2F_FLAG_SETJMP_OK;
2737                 goto out;
2738         }
2739         thread_ctx->flags |= E2F_FLAG_SETJMP_OK;
2740 #endif
2741
2742         e2fsck_pass1_run(thread_ctx);
2743
2744 out:
2745         if (thread_ctx->options & E2F_OPT_MULTITHREAD)
2746                 log_out(thread_ctx,
2747                         _("Scanned group range [%lu, %lu), inodes %lu\n"),
2748                         thread_ctx->thread_info.et_group_start,
2749                         thread_ctx->thread_info.et_group_end,
2750                         thread_ctx->thread_info.et_inode_number);
2751
2752 #ifdef DEBUG_THREADS
2753         pthread_mutex_lock(&thread_debug->etd_mutex);
2754         thread_debug->etd_finished_threads++;
2755         pthread_cond_broadcast(&thread_debug->etd_cond);
2756         pthread_mutex_unlock(&thread_debug->etd_mutex);
2757 #endif
2758
2759         return NULL;
2760 }
2761
2762 static int e2fsck_pass1_threads_start(struct e2fsck_thread_info **pinfo,
2763                                       int num_threads, e2fsck_t global_ctx)
2764 {
2765         struct e2fsck_thread_info       *infos;
2766         pthread_attr_t                   attr;
2767         errcode_t                        retval;
2768         errcode_t                        ret;
2769         struct e2fsck_thread_info       *tmp_pinfo;
2770         int                              i;
2771         e2fsck_t                         thread_ctx;
2772 #ifdef DEBUG_THREADS
2773         struct e2fsck_thread_debug       thread_debug =
2774                 {PTHREAD_MUTEX_INITIALIZER, PTHREAD_COND_INITIALIZER, 0};
2775
2776         thread_debug.etd_finished_threads = 0;
2777 #endif
2778
2779         retval = pthread_attr_init(&attr);
2780         if (retval) {
2781                 com_err(global_ctx->program_name, retval,
2782                         _("while setting pthread attribute\n"));
2783                 return retval;
2784         }
2785
2786         infos = calloc(num_threads, sizeof(struct e2fsck_thread_info));
2787         if (infos == NULL) {
2788                 retval = -ENOMEM;
2789                 com_err(global_ctx->program_name, retval,
2790                         _("while allocating memory for threads\n"));
2791                 pthread_attr_destroy(&attr);
2792                 return retval;
2793         }
2794
2795         for (i = 0; i < num_threads; i++) {
2796                 tmp_pinfo = &infos[i];
2797                 tmp_pinfo->eti_thread_index = i;
2798 #ifdef DEBUG_THREADS
2799                 tmp_pinfo->eti_debug = &thread_debug;
2800 #endif
2801                 retval = e2fsck_pass1_thread_prepare(global_ctx, &thread_ctx,
2802                                                      i, num_threads);
2803                 if (retval) {
2804                         com_err(global_ctx->program_name, retval,
2805                                 _("while preparing pass1 thread\n"));
2806                         break;
2807                 }
2808                 tmp_pinfo->eti_thread_ctx = thread_ctx;
2809
2810                 retval = pthread_create(&tmp_pinfo->eti_thread_id, &attr,
2811                                         &e2fsck_pass1_thread, tmp_pinfo);
2812                 if (retval) {
2813                         com_err(global_ctx->program_name, retval,
2814                                 _("while creating thread\n"));
2815                         e2fsck_pass1_thread_join(global_ctx, thread_ctx);
2816                         break;
2817                 }
2818
2819                 tmp_pinfo->eti_started = 1;
2820         }
2821
2822         /* destroy the thread attribute object, since it is no longer needed */
2823         ret = pthread_attr_destroy(&attr);
2824         if (ret) {
2825                 com_err(global_ctx->program_name, ret,
2826                         _("while destroying thread attribute\n"));
2827                 if (retval == 0)
2828                         retval = ret;
2829         }
2830
2831         if (retval) {
2832                 e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
2833                 return retval;
2834         }
2835         *pinfo = infos;
2836         return 0;
2837 }
2838
2839 static void e2fsck_pass1_multithread(e2fsck_t global_ctx)
2840 {
2841         struct e2fsck_thread_info       *infos = NULL;
2842         int                              num_threads = 1;
2843         errcode_t                        retval;
2844
2845         retval = e2fsck_pass1_threads_start(&infos, num_threads, global_ctx);
2846         if (retval) {
2847                 com_err(global_ctx->program_name, retval,
2848                         _("while starting pass1 threads\n"));
2849                 goto out_abort;
2850         }
2851
2852         retval = e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
2853         if (retval) {
2854                 com_err(global_ctx->program_name, retval,
2855                         _("while joining pass1 threads\n"));
2856                 goto out_abort;
2857         }
2858         return;
2859 out_abort:
2860         global_ctx->flags |= E2F_FLAG_ABORT;
2861         return;
2862 }
2863 #endif
2864
2865 /* TODO: tdb needs to be handled properly for multiple threads*/
2866 static int multiple_threads_supported(e2fsck_t ctx)
2867 {
2868 #ifdef  CONFIG_TDB
2869         unsigned int            threshold;
2870         ext2_ino_t              num_dirs;
2871         errcode_t               retval;
2872         char                    *tdb_dir;
2873         int                     enable;
2874
2875         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
2876                            &tdb_dir);
2877         profile_get_uint(ctx->profile, "scratch_files",
2878                          "numdirs_threshold", 0, 0, &threshold);
2879         profile_get_boolean(ctx->profile, "scratch_files",
2880                             "icount", 0, 1, &enable);
2881
2882         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
2883         if (retval)
2884                 num_dirs = 1024;        /* Guess */
2885
2886         /* tdb is unsupported now */
2887         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
2888             (!threshold || num_dirs > threshold))
2889                 return 0;
2890  #endif
2891         return 1;
2892 }
2893
2894 void e2fsck_pass1(e2fsck_t ctx)
2895 {
2896
2897         init_ext2_max_sizes();
2898 #ifdef HAVE_PTHREAD
2899         if (multiple_threads_supported(ctx)) {
2900                 e2fsck_pass1_multithread(ctx);
2901                 return;
2902         }
2903         fprintf(stderr, "Fall through single thread for pass1 "
2904                         "because tdb could not handle properly\n");
2905 #endif
2906         e2fsck_pass1_run(ctx);
2907 }
2908
2909 #undef FINISH_INODE_LOOP
2910
2911 /*
2912  * When the inode_scan routines call this callback at the end of the
2913  * glock group, call process_inodes.
2914  */
2915 static errcode_t scan_callback(ext2_filsys fs,
2916                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
2917                                dgrp_t group, void * priv_data)
2918 {
2919         struct scan_callback_struct *scan_struct;
2920         e2fsck_t ctx;
2921         struct e2fsck_thread *tinfo;
2922
2923         scan_struct = (struct scan_callback_struct *) priv_data;
2924         ctx = scan_struct->ctx;
2925
2926         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf,
2927                        scan_struct->inodes_to_process,
2928                        scan_struct->process_inode_count);
2929
2930         if (ctx->progress)
2931                 if ((ctx->progress)(ctx, 1, group+1,
2932                                     ctx->fs->group_desc_count))
2933                         return EXT2_ET_CANCEL_REQUESTED;
2934
2935 #ifdef HAVE_PTHREAD
2936         if (ctx->global_ctx) {
2937                 tinfo = &ctx->thread_info;
2938                 tinfo->et_group_next++;
2939                 if (ctx->options & E2F_OPT_DEBUG &&
2940                     ctx->options & E2F_OPT_MULTITHREAD)
2941                         log_out(ctx, _("group %d finished\n"),
2942                                 tinfo->et_group_next);
2943                 if (tinfo->et_group_next >= tinfo->et_group_end)
2944                         return EXT2_ET_SCAN_FINISHED;
2945         }
2946 #endif
2947
2948         return 0;
2949 }
2950
2951 /*
2952  * Process the inodes in the "inodes to process" list.
2953  */
2954 static void process_inodes(e2fsck_t ctx, char *block_buf,
2955                            struct process_inode_block *inodes_to_process,
2956                            int *process_inode_count)
2957 {
2958         int                     i;
2959         struct ext2_inode       *old_stashed_inode;
2960         ext2_ino_t              old_stashed_ino;
2961         const char              *old_operation;
2962         char                    buf[80];
2963         struct problem_context  pctx;
2964
2965 #if 0
2966         printf("begin process_inodes: ");
2967 #endif
2968         if (*process_inode_count == 0)
2969                 return;
2970         old_operation = ehandler_operation(0);
2971         old_stashed_inode = ctx->stashed_inode;
2972         old_stashed_ino = ctx->stashed_ino;
2973         qsort(inodes_to_process, *process_inode_count,
2974                       sizeof(struct process_inode_block), process_inode_cmp);
2975         clear_problem_context(&pctx);
2976         for (i=0; i < *process_inode_count; i++) {
2977                 pctx.inode = ctx->stashed_inode =
2978                         (struct ext2_inode *) &inodes_to_process[i].inode;
2979                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2980
2981 #if 0
2982                 printf("%u ", pctx.ino);
2983 #endif
2984                 sprintf(buf, _("reading indirect blocks of inode %u"),
2985                         pctx.ino);
2986                 ehandler_operation(buf);
2987                 check_blocks(ctx, &pctx, block_buf,
2988                              &inodes_to_process[i].ea_ibody_quota);
2989                 if (e2fsck_should_abort(ctx))
2990                         break;
2991         }
2992         ctx->stashed_inode = old_stashed_inode;
2993         ctx->stashed_ino = old_stashed_ino;
2994         *process_inode_count = 0;
2995 #if 0
2996         printf("end process inodes\n");
2997 #endif
2998         ehandler_operation(old_operation);
2999 }
3000
3001 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
3002 {
3003         const struct process_inode_block *ib_a =
3004                 (const struct process_inode_block *) a;
3005         const struct process_inode_block *ib_b =
3006                 (const struct process_inode_block *) b;
3007         int     ret;
3008
3009         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
3010                ib_b->inode.i_block[EXT2_IND_BLOCK]);
3011         if (ret == 0)
3012                 /*
3013                  * We only call process_inodes() for non-extent
3014                  * inodes, so it's OK to pass NULL to
3015                  * ext2fs_file_acl_block() here.
3016                  */
3017                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
3018                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
3019         if (ret == 0)
3020                 ret = ib_a->ino - ib_b->ino;
3021         return ret;
3022 }
3023
3024 /*
3025  * Mark an inode as being bad in some what
3026  */
3027 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
3028 {
3029         struct          problem_context pctx;
3030
3031         if (!ctx->inode_bad_map) {
3032                 clear_problem_context(&pctx);
3033
3034                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3035                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
3036                                 "inode_bad_map", &ctx->inode_bad_map);
3037                 if (pctx.errcode) {
3038                         pctx.num = 3;
3039                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3040                         /* Should never get here */
3041                         ctx->flags |= E2F_FLAG_ABORT;
3042                         return;
3043                 }
3044         }
3045         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
3046 }
3047
3048 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino)
3049 {
3050         struct          problem_context pctx;
3051
3052         if (!ctx->casefolded_dirs) {
3053                 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
3054                 if (pctx.errcode)
3055                         goto error;
3056         }
3057         pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
3058         if (pctx.errcode == 0)
3059                 return;
3060 error:
3061         fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
3062         /* Should never get here */
3063         ctx->flags |= E2F_FLAG_ABORT;
3064 }
3065
3066 /*
3067  * This procedure will allocate the inode "bb" (badblock) map table
3068  */
3069 static void alloc_bb_map(e2fsck_t ctx)
3070 {
3071         struct          problem_context pctx;
3072
3073         clear_problem_context(&pctx);
3074         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3075                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
3076                         "inode_bb_map", &ctx->inode_bb_map);
3077         if (pctx.errcode) {
3078                 pctx.num = 4;
3079                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3080                 /* Should never get here */
3081                 ctx->flags |= E2F_FLAG_ABORT;
3082                 return;
3083         }
3084 }
3085
3086 /*
3087  * This procedure will allocate the inode imagic table
3088  */
3089 static void alloc_imagic_map(e2fsck_t ctx)
3090 {
3091         struct          problem_context pctx;
3092
3093         clear_problem_context(&pctx);
3094         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
3095                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
3096                         "inode_imagic_map", &ctx->inode_imagic_map);
3097         if (pctx.errcode) {
3098                 pctx.num = 5;
3099                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
3100                 /* Should never get here */
3101                 ctx->flags |= E2F_FLAG_ABORT;
3102                 return;
3103         }
3104 }
3105
3106 /*
3107  * Marks a block as in use, setting the dup_map if it's been set
3108  * already.  Called by process_block and process_bad_block.
3109  *
3110  * WARNING: Assumes checks have already been done to make sure block
3111  * is valid.  This is true in both process_block and process_bad_block.
3112  */
3113 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
3114 {
3115         struct          problem_context pctx;
3116
3117         clear_problem_context(&pctx);
3118
3119         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
3120                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
3121                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
3122                         return;
3123                 }
3124                 if (!ctx->block_dup_map) {
3125                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
3126                                         _("multiply claimed block map"),
3127                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
3128                                         &ctx->block_dup_map);
3129                         if (pctx.errcode) {
3130                                 pctx.num = 3;
3131                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
3132                                             &pctx);
3133                                 /* Should never get here */
3134                                 ctx->flags |= E2F_FLAG_ABORT;
3135                                 return;
3136                         }
3137                 }
3138                 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
3139         } else {
3140                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
3141         }
3142 }
3143
3144 /*
3145  * When cluster size is greater than one block, it is caller's responsibility
3146  * to make sure block parameter starts at a cluster boundary.
3147  */
3148 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
3149                                       unsigned int num)
3150 {
3151         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
3152                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
3153         else {
3154                 unsigned int i;
3155
3156                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
3157                         mark_block_used(ctx, block + i);
3158         }
3159 }
3160
3161 /*
3162  * Adjust the extended attribute block's reference counts at the end
3163  * of pass 1, either by subtracting out references for EA blocks that
3164  * are still referenced in ctx->refcount, or by adding references for
3165  * EA blocks that had extra references as accounted for in
3166  * ctx->refcount_extra.
3167  */
3168 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
3169                                     char *block_buf, int adjust_sign)
3170 {
3171         struct ext2_ext_attr_header     *header;
3172         struct problem_context          pctx;
3173         ext2_filsys                     fs = ctx->fs;
3174         blk64_t                         blk;
3175         __u32                           should_be;
3176         ea_value_t                      count;
3177
3178         clear_problem_context(&pctx);
3179
3180         ea_refcount_intr_begin(refcount);
3181         while (1) {
3182                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
3183                         break;
3184                 pctx.blk = blk;
3185                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
3186                                                      pctx.ino);
3187                 if (pctx.errcode) {
3188                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
3189                         return;
3190                 }
3191                 header = (struct ext2_ext_attr_header *) block_buf;
3192                 pctx.blkcount = header->h_refcount;
3193                 should_be = header->h_refcount + adjust_sign * (int)count;
3194                 pctx.num = should_be;
3195                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
3196                         header->h_refcount = should_be;
3197                         pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
3198                                                              block_buf,
3199                                                              pctx.ino);
3200                         if (pctx.errcode) {
3201                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
3202                                             &pctx);
3203                                 continue;
3204                         }
3205                 }
3206         }
3207 }
3208
3209 /*
3210  * Handle processing the extended attribute blocks
3211  */
3212 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
3213                            char *block_buf, struct ea_quota *ea_block_quota)
3214 {
3215         ext2_filsys fs = ctx->fs;
3216         ext2_ino_t      ino = pctx->ino;
3217         struct ext2_inode *inode = pctx->inode;
3218         blk64_t         blk;
3219         char *          end;
3220         struct ext2_ext_attr_header *header;
3221         struct ext2_ext_attr_entry *first, *entry;
3222         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
3223         __u64           quota_inodes = 0;
3224         region_t        region = 0;
3225         int             failed_csum = 0;
3226
3227         ea_block_quota->blocks = 0;
3228         ea_block_quota->inodes = 0;
3229
3230         blk = ext2fs_file_acl_block(fs, inode);
3231         if (blk == 0)
3232                 return 0;
3233
3234         /*
3235          * If the Extended attribute flag isn't set, then a non-zero
3236          * file acl means that the inode is corrupted.
3237          *
3238          * Or if the extended attribute block is an invalid block,
3239          * then the inode is also corrupted.
3240          */
3241         if (!ext2fs_has_feature_xattr(fs->super) ||
3242             (blk < fs->super->s_first_data_block) ||
3243             (blk >= ext2fs_blocks_count(fs->super))) {
3244                 mark_inode_bad(ctx, ino);
3245                 return 0;
3246         }
3247
3248         /* If ea bitmap hasn't been allocated, create it */
3249         if (!ctx->block_ea_map) {
3250                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
3251                                         _("ext attr block map"),
3252                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
3253                                         &ctx->block_ea_map);
3254                 if (pctx->errcode) {
3255                         pctx->num = 2;
3256                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
3257                         ctx->flags |= E2F_FLAG_ABORT;
3258                         return 0;
3259                 }
3260         }
3261
3262         /* Create the EA refcount structure if necessary */
3263         if (!ctx->refcount) {
3264                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
3265                 if (pctx->errcode) {
3266                         pctx->num = 1;
3267                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3268                         ctx->flags |= E2F_FLAG_ABORT;
3269                         return 0;
3270                 }
3271         }
3272
3273 #if 0
3274         /* Debugging text */
3275         printf("Inode %u has EA block %u\n", ino, blk);
3276 #endif
3277
3278         /* Have we seen this EA block before? */
3279         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
3280                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
3281                 ea_block_quota->inodes = 0;
3282
3283                 if (ctx->ea_block_quota_blocks) {
3284                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
3285                                           &quota_blocks);
3286                         if (quota_blocks)
3287                                 ea_block_quota->blocks = quota_blocks;
3288                 }
3289
3290                 if (ctx->ea_block_quota_inodes)
3291                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
3292                                           &ea_block_quota->inodes);
3293
3294                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
3295                         return 1;
3296                 /* Ooops, this EA was referenced more than it stated */
3297                 if (!ctx->refcount_extra) {
3298                         pctx->errcode = ea_refcount_create(0,
3299                                            &ctx->refcount_extra);
3300                         if (pctx->errcode) {
3301                                 pctx->num = 2;
3302                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3303                                 ctx->flags |= E2F_FLAG_ABORT;
3304                                 return 0;
3305                         }
3306                 }
3307                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
3308                 return 1;
3309         }
3310
3311         /*
3312          * OK, we haven't seen this EA block yet.  So we need to
3313          * validate it
3314          */
3315         pctx->blk = blk;
3316         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
3317         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
3318                 pctx->errcode = 0;
3319                 failed_csum = 1;
3320         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
3321                 pctx->errcode = 0;
3322
3323         if (pctx->errcode &&
3324             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
3325                 pctx->errcode = 0;
3326                 goto clear_extattr;
3327         }
3328         header = (struct ext2_ext_attr_header *) block_buf;
3329         pctx->blk = ext2fs_file_acl_block(fs, inode);
3330         if (((ctx->ext_attr_ver == 1) &&
3331              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
3332             ((ctx->ext_attr_ver == 2) &&
3333              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
3334                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
3335                         goto clear_extattr;
3336         }
3337
3338         if (header->h_blocks != 1) {
3339                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
3340                         goto clear_extattr;
3341         }
3342
3343         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
3344                 goto clear_extattr;
3345
3346         region = region_create(0, fs->blocksize);
3347         if (!region) {
3348                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
3349                 ctx->flags |= E2F_FLAG_ABORT;
3350                 return 0;
3351         }
3352         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
3353                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3354                         goto clear_extattr;
3355         }
3356
3357         first = (struct ext2_ext_attr_entry *)(header+1);
3358         end = block_buf + fs->blocksize;
3359         entry = first;
3360         while ((char *)entry < end && *(__u32 *)entry) {
3361                 __u32 hash;
3362
3363                 if (region_allocate(region, (char *)entry - (char *)header,
3364                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
3365                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3366                                 goto clear_extattr;
3367                         break;
3368                 }
3369                 if ((ctx->ext_attr_ver == 1 &&
3370                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
3371                     (ctx->ext_attr_ver == 2 &&
3372                      entry->e_name_index == 0)) {
3373                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
3374                                 goto clear_extattr;
3375                         break;
3376                 }
3377                 if (entry->e_value_inum == 0) {
3378                         if (entry->e_value_size > EXT2_XATTR_SIZE_MAX ||
3379                             (entry->e_value_offs + entry->e_value_size >
3380                              fs->blocksize)) {
3381                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
3382                                         goto clear_extattr;
3383                                 break;
3384                         }
3385                         if (entry->e_value_size &&
3386                             region_allocate(region, entry->e_value_offs,
3387                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
3388                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
3389                                                 pctx))
3390                                         goto clear_extattr;
3391                         }
3392
3393                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
3394                                                           entry->e_value_offs);
3395
3396                         if (entry->e_hash != hash) {
3397                                 pctx->num = entry->e_hash;
3398                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
3399                                         goto clear_extattr;
3400                                 entry->e_hash = hash;
3401                         }
3402                 } else {
3403                         problem_t problem;
3404                         blk64_t entry_quota_blocks;
3405
3406                         problem = check_large_ea_inode(ctx, entry, pctx,
3407                                                        &entry_quota_blocks);
3408                         if (problem && fix_problem(ctx, problem, pctx))
3409                                 goto clear_extattr;
3410
3411                         quota_blocks += entry_quota_blocks;
3412                         quota_inodes++;
3413                 }
3414
3415                 entry = EXT2_EXT_ATTR_NEXT(entry);
3416         }
3417         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
3418                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3419                         goto clear_extattr;
3420         }
3421         region_free(region);
3422
3423         /*
3424          * We only get here if there was no other errors that were fixed.
3425          * If there was a checksum fail, ask to correct it.
3426          */
3427         if (failed_csum &&
3428             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
3429                 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
3430                                                        pctx->ino);
3431                 if (pctx->errcode)
3432                         return 0;
3433         }
3434
3435         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
3436                 if (!ctx->ea_block_quota_blocks) {
3437                         pctx->errcode = ea_refcount_create(0,
3438                                                 &ctx->ea_block_quota_blocks);
3439                         if (pctx->errcode) {
3440                                 pctx->num = 3;
3441                                 goto refcount_fail;
3442                         }
3443                 }
3444                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
3445                                   quota_blocks);
3446         }
3447
3448         if (quota_inodes) {
3449                 if (!ctx->ea_block_quota_inodes) {
3450                         pctx->errcode = ea_refcount_create(0,
3451                                                 &ctx->ea_block_quota_inodes);
3452                         if (pctx->errcode) {
3453                                 pctx->num = 4;
3454 refcount_fail:
3455                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3456                                 ctx->flags |= E2F_FLAG_ABORT;
3457                                 return 0;
3458                         }
3459                 }
3460
3461                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
3462                                   quota_inodes);
3463         }
3464         ea_block_quota->blocks = quota_blocks;
3465         ea_block_quota->inodes = quota_inodes;
3466
3467         inc_ea_inode_refs(ctx, pctx, first, end);
3468         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
3469         mark_block_used(ctx, blk);
3470         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
3471         return 1;
3472
3473 clear_extattr:
3474         if (region)
3475                 region_free(region);
3476         ext2fs_file_acl_block_set(fs, inode, 0);
3477         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
3478         return 0;
3479 }
3480
3481 /* Returns 1 if bad htree, 0 if OK */
3482 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
3483                         ext2_ino_t ino, struct ext2_inode *inode,
3484                         char *block_buf)
3485 {
3486         struct ext2_dx_root_info        *root;
3487         ext2_filsys                     fs = ctx->fs;
3488         errcode_t                       retval;
3489         blk64_t                         blk;
3490
3491         if ((!LINUX_S_ISDIR(inode->i_mode) &&
3492              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
3493             (!ext2fs_has_feature_dir_index(fs->super) &&
3494              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
3495                 return 1;
3496
3497         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
3498
3499         if ((pctx->errcode) ||
3500             (blk == 0) ||
3501             (blk < fs->super->s_first_data_block) ||
3502             (blk >= ext2fs_blocks_count(fs->super))) {
3503                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3504                         return 1;
3505                 else
3506                         return 0;
3507         }
3508
3509         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
3510         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3511                 return 1;
3512
3513         /* XXX should check that beginning matches a directory */
3514         root = (struct ext2_dx_root_info *) (block_buf + 24);
3515
3516         if ((root->reserved_zero || root->info_length < 8) &&
3517             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3518                 return 1;
3519
3520         pctx->num = root->hash_version;
3521         if ((root->hash_version != EXT2_HASH_LEGACY) &&
3522             (root->hash_version != EXT2_HASH_HALF_MD4) &&
3523             (root->hash_version != EXT2_HASH_TEA) &&
3524             (root->hash_version != EXT2_HASH_SIPHASH) &&
3525             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
3526                 return 1;
3527
3528         if (ext4_hash_in_dirent(inode)) {
3529                 if (root->hash_version != EXT2_HASH_SIPHASH &&
3530                     fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
3531                         return 1;
3532         } else {
3533                 if (root->hash_version == EXT2_HASH_SIPHASH &&
3534                    fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
3535                         return 1;
3536         }
3537
3538         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
3539             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
3540                 return 1;
3541
3542         pctx->num = root->indirect_levels;
3543         /* if htree level is clearly too high, consider it to be broken */
3544         if (root->indirect_levels > EXT4_HTREE_LEVEL &&
3545             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3546                 return 1;
3547
3548         /* if level is only maybe too high, LARGE_DIR feature could be unset */
3549         if (root->indirect_levels > ext2_dir_htree_level(fs) &&
3550             !ext2fs_has_feature_largedir(fs->super)) {
3551                 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
3552                 unsigned idx_pb = 1 << (blockbits - 3);
3553
3554                 /* compare inode size/blocks vs. max-sized 2-level htree */
3555                 if (EXT2_I_SIZE(pctx->inode) <
3556                     (idx_pb - 1) * (idx_pb - 2) << blockbits &&
3557                     pctx->inode->i_blocks <
3558                     (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
3559                     fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3560                         return 1;
3561         }
3562
3563         if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
3564             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3565                 ctx->large_dirs++;
3566
3567         return 0;
3568 }
3569
3570 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
3571                         struct ext2_inode *inode, int restart_flag,
3572                         const char *source)
3573 {
3574         inode->i_flags = 0;
3575         inode->i_links_count = 0;
3576         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
3577         inode->i_dtime = ctx->now;
3578
3579         /*
3580          * If a special inode has such rotten block mappings that we
3581          * want to clear the whole inode, be sure to actually zap
3582          * the block maps because i_links_count isn't checked for
3583          * special inodes, and we'll end up right back here the next
3584          * time we run fsck.
3585          */
3586         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
3587                 memset(inode->i_block, 0, sizeof(inode->i_block));
3588
3589         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
3590         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
3591         if (ctx->inode_reg_map)
3592                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
3593         if (ctx->inode_bad_map)
3594                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
3595
3596         /*
3597          * If the inode was partially accounted for before processing
3598          * was aborted, we need to restart the pass 1 scan.
3599          */
3600         ctx->flags |= restart_flag;
3601
3602         if (ino == EXT2_BAD_INO)
3603                 memset(inode, 0, sizeof(struct ext2_inode));
3604
3605         e2fsck_write_inode(ctx, ino, inode, source);
3606 }
3607
3608 /*
3609  * Use the multiple-blocks reclamation code to fix alignment problems in
3610  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
3611  * physical cluster, and we want the block offsets within that cluster to
3612  * line up.
3613  */
3614 static int has_unaligned_cluster_map(e2fsck_t ctx,
3615                                      blk64_t last_pblk, blk64_t last_lblk,
3616                                      blk64_t pblk, blk64_t lblk)
3617 {
3618         blk64_t cluster_mask;
3619
3620         if (!ctx->fs->cluster_ratio_bits)
3621                 return 0;
3622         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
3623
3624         /*
3625          * If the block in the logical cluster doesn't align with the block in
3626          * the physical cluster...
3627          */
3628         if ((lblk & cluster_mask) != (pblk & cluster_mask))
3629                 return 1;
3630
3631         /*
3632          * If we cross a physical cluster boundary within a logical cluster...
3633          */
3634         if (last_pblk && (lblk & cluster_mask) != 0 &&
3635             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
3636             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
3637                 return 1;
3638
3639         return 0;
3640 }
3641
3642 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
3643                              struct process_block_struct *pb,
3644                              blk64_t start_block, blk64_t end_block,
3645                              blk64_t eof_block,
3646                              ext2_extent_handle_t ehandle,
3647                              int try_repairs)
3648 {
3649         struct ext2fs_extent    extent;
3650         blk64_t                 blk, last_lblk;
3651         unsigned int            i, n;
3652         int                     is_dir, is_leaf;
3653         problem_t               problem;
3654         struct ext2_extent_info info;
3655         int                     failed_csum = 0;
3656
3657         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
3658                 failed_csum = 1;
3659
3660         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
3661         if (pctx->errcode)
3662                 return;
3663         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
3664             !pb->eti.force_rebuild &&
3665             info.curr_level < MAX_EXTENT_DEPTH_COUNT) {
3666                 struct extent_tree_level *etl;
3667
3668                 etl = pb->eti.ext_info + info.curr_level;
3669                 etl->num_extents += info.num_entries;
3670                 etl->max_extents += info.max_entries;
3671                 /*
3672                  * Implementation wart: Splitting extent blocks when appending
3673                  * will leave the old block with one free entry.  Therefore
3674                  * unless the node is totally full, pretend that a non-root
3675                  * extent block can hold one fewer entry than it actually does,
3676                  * so that we don't repeatedly rebuild the extent tree.
3677                  */
3678                 if (info.curr_level && info.num_entries < info.max_entries)
3679                         etl->max_extents--;
3680         }
3681
3682         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
3683                                           &extent);
3684         while ((pctx->errcode == 0 ||
3685                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
3686                info.num_entries-- > 0) {
3687                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
3688                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
3689                 last_lblk = extent.e_lblk + extent.e_len - 1;
3690
3691                 problem = 0;
3692                 pctx->blk = extent.e_pblk;
3693                 pctx->blk2 = extent.e_lblk;
3694                 pctx->num = extent.e_len;
3695                 pctx->blkcount = extent.e_lblk + extent.e_len;
3696
3697                 if (extent.e_pblk == 0 ||
3698                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
3699                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
3700                         problem = PR_1_EXTENT_BAD_START_BLK;
3701                 else if (extent.e_lblk < start_block)
3702                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
3703                 else if ((end_block && last_lblk > end_block) &&
3704                          !(last_lblk > eof_block &&
3705                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
3706                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
3707                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
3708                 else if (is_leaf && extent.e_len == 0)
3709                         problem = PR_1_EXTENT_LENGTH_ZERO;
3710                 else if (is_leaf &&
3711                          (extent.e_pblk + extent.e_len) >
3712                          ext2fs_blocks_count(ctx->fs->super))
3713                         problem = PR_1_EXTENT_ENDS_BEYOND;
3714                 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
3715                          !ext2fs_has_feature_largedir(ctx->fs->super) &&
3716                          ((extent.e_lblk + extent.e_len) >
3717                           (1U << (21 - ctx->fs->super->s_log_block_size))))
3718                         problem = PR_1_TOOBIG_DIR;
3719
3720                 if (is_leaf && problem == 0 && extent.e_len > 0) {
3721 #if 0
3722                         printf("extent_region(ino=%u, expect=%llu, "
3723                                "lblk=%llu, len=%u)\n", pb->ino,
3724                                (unsigned long long) pb->next_lblock,
3725                                (unsigned long long) extent.e_lblk,
3726                                extent.e_len);
3727 #endif
3728                         if (extent.e_lblk < pb->next_lblock)
3729                                 problem = PR_1_EXTENT_COLLISION;
3730                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
3731                                 pb->next_lblock = extent.e_lblk + extent.e_len;
3732                 }
3733
3734                 /*
3735                  * Uninitialized blocks in a directory?  Clear the flag and
3736                  * we'll interpret the blocks later.
3737                  */
3738                 if (try_repairs && is_dir && problem == 0 &&
3739                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3740                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
3741                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3742                         pb->inode_modified = 1;
3743                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3744                                                               &extent);
3745                         if (pctx->errcode)
3746                                 return;
3747                         failed_csum = 0;
3748                 }
3749 #ifdef CONFIG_DEVELOPER_FEATURES
3750                 if (try_repairs && !is_dir && problem == 0 &&
3751                     (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
3752                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3753                     fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
3754                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3755                         pb->inode_modified = 1;
3756                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3757                                                               &extent);
3758                         if (pctx->errcode)
3759                                 return;
3760                         failed_csum = 0;
3761                 }
3762 #endif
3763                 if (try_repairs && problem) {
3764 report_problem:
3765                         if (fix_problem(ctx, problem, pctx)) {
3766                                 if (ctx->invalid_bitmaps) {
3767                                         /*
3768                                          * If fsck knows the bitmaps are bad,
3769                                          * skip to the next extent and
3770                                          * try to clear this extent again
3771                                          * after fixing the bitmaps, by
3772                                          * restarting fsck.
3773                                          */
3774                                         pctx->errcode = ext2fs_extent_get(
3775                                                           ehandle,
3776                                                           EXT2_EXTENT_NEXT_SIB,
3777                                                           &extent);
3778                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3779                                         if (pctx->errcode ==
3780                                                     EXT2_ET_NO_CURRENT_NODE) {
3781                                                 pctx->errcode = 0;
3782                                                 break;
3783                                         }
3784                                         continue;
3785                                 }
3786                                 e2fsck_read_bitmaps(ctx);
3787                                 pb->inode_modified = 1;
3788                                 pctx->errcode =
3789                                         ext2fs_extent_delete(ehandle, 0);
3790                                 if (pctx->errcode) {
3791                                         pctx->str = "ext2fs_extent_delete";
3792                                         return;
3793                                 }
3794                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3795                                 if (pctx->errcode &&
3796                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
3797                                         pctx->str = "ext2fs_extent_fix_parents";
3798                                         return;
3799                                 }
3800                                 pctx->errcode = ext2fs_extent_get(ehandle,
3801                                                                   EXT2_EXTENT_CURRENT,
3802                                                                   &extent);
3803                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
3804                                         pctx->errcode = 0;
3805                                         break;
3806                                 }
3807                                 failed_csum = 0;
3808                                 continue;
3809                         }
3810                         goto next;
3811                 }
3812
3813                 if (!is_leaf) {
3814                         blk64_t lblk = extent.e_lblk;
3815                         int next_try_repairs = 1;
3816
3817                         blk = extent.e_pblk;
3818
3819                         /*
3820                          * If this lower extent block collides with critical
3821                          * metadata, don't try to repair the damage.  Pass 1b
3822                          * will reallocate the block; then we can try again.
3823                          */
3824                         if (pb->ino != EXT2_RESIZE_INO &&
3825                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
3826                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3827                                                       extent.e_pblk)) {
3828                                 next_try_repairs = 0;
3829                                 pctx->blk = blk;
3830                                 fix_problem(ctx,
3831                                             PR_1_CRITICAL_METADATA_COLLISION,
3832                                             pctx);
3833                                 if ((ctx->options & E2F_OPT_NO) == 0)
3834                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3835                         }
3836                         pctx->errcode = ext2fs_extent_get(ehandle,
3837                                                   EXT2_EXTENT_DOWN, &extent);
3838                         if (pctx->errcode &&
3839                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
3840                                 pctx->str = "EXT2_EXTENT_DOWN";
3841                                 problem = PR_1_EXTENT_HEADER_INVALID;
3842                                 if (!next_try_repairs)
3843                                         return;
3844                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
3845                                         goto report_problem;
3846                                 return;
3847                         }
3848                         /* The next extent should match this index's logical start */
3849                         if (extent.e_lblk != lblk) {
3850                                 struct ext2_extent_info e_info;
3851
3852                                 pctx->errcode = ext2fs_extent_get_info(ehandle,
3853                                                                        &e_info);
3854                                 if (pctx->errcode) {
3855                                         pctx->str = "ext2fs_extent_get_info";
3856                                         return;
3857                                 }
3858                                 pctx->blk = lblk;
3859                                 pctx->blk2 = extent.e_lblk;
3860                                 pctx->num = e_info.curr_level - 1;
3861                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
3862                                 if (fix_problem(ctx, problem, pctx)) {
3863                                         pb->inode_modified = 1;
3864                                         pctx->errcode =
3865                                                 ext2fs_extent_fix_parents(ehandle);
3866                                         if (pctx->errcode) {
3867                                                 pctx->str = "ext2fs_extent_fix_parents";
3868                                                 return;
3869                                         }
3870                                 }
3871                         }
3872                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
3873                                          last_lblk, eof_block, ehandle,
3874                                          next_try_repairs);
3875                         if (pctx->errcode)
3876                                 return;
3877                         pctx->errcode = ext2fs_extent_get(ehandle,
3878                                                   EXT2_EXTENT_UP, &extent);
3879                         if (pctx->errcode) {
3880                                 pctx->str = "EXT2_EXTENT_UP";
3881                                 return;
3882                         }
3883                         mark_block_used(ctx, blk);
3884                         pb->num_blocks++;
3885                         goto next;
3886                 }
3887
3888                 if ((pb->previous_block != 0) &&
3889                     (pb->previous_block+1 != extent.e_pblk)) {
3890                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3891                                 char type = '?';
3892
3893                                 if (pb->is_dir)
3894                                         type = 'd';
3895                                 else if (pb->is_reg)
3896                                         type = 'f';
3897
3898                                 printf(("%6lu(%c): expecting %6lu "
3899                                         "actual extent "
3900                                         "phys %6lu log %lu len %lu\n"),
3901                                        (unsigned long) pctx->ino, type,
3902                                        (unsigned long) pb->previous_block+1,
3903                                        (unsigned long) extent.e_pblk,
3904                                        (unsigned long) extent.e_lblk,
3905                                        (unsigned long) extent.e_len);
3906                         }
3907                         pb->fragmented = 1;
3908                 }
3909                 /*
3910                  * If we notice a gap in the logical block mappings of an
3911                  * extent-mapped directory, offer to close the hole by
3912                  * moving the logical block down, otherwise we'll go mad in
3913                  * pass 3 allocating empty directory blocks to fill the hole.
3914                  */
3915                 if (try_repairs && is_dir &&
3916                     pb->last_block + 1 < extent.e_lblk) {
3917                         blk64_t new_lblk;
3918
3919                         new_lblk = pb->last_block + 1;
3920                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3921                                 new_lblk = ((new_lblk +
3922                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3923                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3924                                            (extent.e_pblk &
3925                                             EXT2FS_CLUSTER_MASK(ctx->fs));
3926                         pctx->blk = extent.e_lblk;
3927                         pctx->blk2 = new_lblk;
3928                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3929                                 extent.e_lblk = new_lblk;
3930                                 pb->inode_modified = 1;
3931                                 pctx->errcode = ext2fs_extent_replace(ehandle,
3932                                                                 0, &extent);
3933                                 if (pctx->errcode) {
3934                                         pctx->errcode = 0;
3935                                         goto alloc_later;
3936                                 }
3937                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3938                                 if (pctx->errcode)
3939                                         goto failed_add_dir_block;
3940                                 pctx->errcode = ext2fs_extent_goto(ehandle,
3941                                                                 extent.e_lblk);
3942                                 if (pctx->errcode)
3943                                         goto failed_add_dir_block;
3944                                 last_lblk = extent.e_lblk + extent.e_len - 1;
3945                                 failed_csum = 0;
3946                         }
3947                 }
3948 alloc_later:
3949                 if (is_dir) {
3950                         while (++pb->last_db_block <
3951                                (e2_blkcnt_t) extent.e_lblk) {
3952                                 pctx->errcode = ext2fs_add_dir_block2(
3953                                                         ctx->fs->dblist,
3954                                                         pb->ino, 0,
3955                                                         pb->last_db_block);
3956                                 if (pctx->errcode) {
3957                                         pctx->blk = 0;
3958                                         pctx->num = pb->last_db_block;
3959                                         goto failed_add_dir_block;
3960                                 }
3961                         }
3962
3963                         for (i = 0; i < extent.e_len; i++) {
3964                                 pctx->errcode = ext2fs_add_dir_block2(
3965                                                         ctx->fs->dblist,
3966                                                         pctx->ino,
3967                                                         extent.e_pblk + i,
3968                                                         extent.e_lblk + i);
3969                                 if (pctx->errcode) {
3970                                         pctx->blk = extent.e_pblk + i;
3971                                         pctx->num = extent.e_lblk + i;
3972                                 failed_add_dir_block:
3973                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3974                                         /* Should never get here */
3975                                         ctx->flags |= E2F_FLAG_ABORT;
3976                                         return;
3977                                 }
3978                         }
3979                         if (extent.e_len > 0)
3980                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3981                 }
3982                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3983                                               pb->last_block,
3984                                               extent.e_pblk,
3985                                               extent.e_lblk)) {
3986                         for (i = 0; i < extent.e_len; i++) {
3987                                 pctx->blk = extent.e_lblk + i;
3988                                 pctx->blk2 = extent.e_pblk + i;
3989                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3990                                 mark_block_used(ctx, extent.e_pblk + i);
3991                                 mark_block_used(ctx, extent.e_pblk + i);
3992                         }
3993                 }
3994
3995                 /*
3996                  * Check whether first cluster got marked in previous iteration.
3997                  */
3998                 if (ctx->fs->cluster_ratio_bits &&
3999                     pb->previous_block &&
4000                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
4001                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
4002                         /* Set blk to the beginning of next cluster. */
4003                         blk = EXT2FS_C2B(
4004                                 ctx->fs,
4005                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
4006                 else
4007                         /* Set blk to the beginning of current cluster. */
4008                         blk = EXT2FS_C2B(ctx->fs,
4009                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
4010
4011                 if (blk < extent.e_pblk + extent.e_len) {
4012                         mark_blocks_used(ctx, blk,
4013                                          extent.e_pblk + extent.e_len - blk);
4014                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
4015                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
4016                         pb->num_blocks += n;
4017                 }
4018                 pb->last_block = extent.e_lblk + extent.e_len - 1;
4019                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
4020                 start_block = pb->last_block = last_lblk;
4021                 if (is_leaf && !is_dir &&
4022                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
4023                         pb->last_init_lblock = last_lblk;
4024         next:
4025                 pctx->errcode = ext2fs_extent_get(ehandle,
4026                                                   EXT2_EXTENT_NEXT_SIB,
4027                                                   &extent);
4028         }
4029
4030         /* Failed csum but passes checks?  Ask to fix checksum. */
4031         if (failed_csum &&
4032             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
4033                 pb->inode_modified = 1;
4034                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
4035                 if (pctx->errcode)
4036                         return;
4037         }
4038
4039         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
4040                 pctx->errcode = 0;
4041 }
4042
4043 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
4044                                  struct process_block_struct *pb)
4045 {
4046         struct ext2_extent_info info;
4047         struct ext2_inode       *inode = pctx->inode;
4048         ext2_extent_handle_t    ehandle;
4049         ext2_filsys             fs = ctx->fs;
4050         ext2_ino_t              ino = pctx->ino;
4051         errcode_t               retval;
4052         blk64_t                 eof_lblk;
4053         struct ext3_extent_header       *eh;
4054
4055         /* Check for a proper extent header... */
4056         eh = (struct ext3_extent_header *) &inode->i_block[0];
4057         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
4058         if (retval) {
4059                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
4060                         e2fsck_clear_inode(ctx, ino, inode, 0,
4061                                            "check_blocks_extents");
4062                 pctx->errcode = 0;
4063                 return;
4064         }
4065
4066         /* ...since this function doesn't fail if i_block is zeroed. */
4067         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
4068         if (pctx->errcode) {
4069                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
4070                         e2fsck_clear_inode(ctx, ino, inode, 0,
4071                                            "check_blocks_extents");
4072                 pctx->errcode = 0;
4073                 return;
4074         }
4075
4076         retval = ext2fs_extent_get_info(ehandle, &info);
4077         if (retval == 0) {
4078                 int max_depth = info.max_depth;
4079
4080                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
4081                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
4082                 ctx->extent_depth_count[max_depth]++;
4083         }
4084
4085         /* Check maximum extent depth */
4086         pctx->blk = info.max_depth;
4087         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
4088         if (pctx->blk2 < pctx->blk &&
4089             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
4090                 pb->eti.force_rebuild = 1;
4091
4092         /* Can we collect extent tree level stats? */
4093         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
4094         if (pctx->blk2 > pctx->blk)
4095                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
4096         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
4097         pb->eti.ino = pb->ino;
4098
4099         pb->next_lblock = 0;
4100
4101         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
4102                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
4103         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
4104         if (pctx->errcode &&
4105             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
4106                 pb->num_blocks = 0;
4107                 inode->i_blocks = 0;
4108                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4109                                    "check_blocks_extents");
4110                 pctx->errcode = 0;
4111         }
4112         ext2fs_extent_free(ehandle);
4113
4114         /* Rebuild unless it's a dir and we're rehashing it */
4115         if (LINUX_S_ISDIR(inode->i_mode) &&
4116             e2fsck_dir_will_be_rehashed(ctx, ino))
4117                 return;
4118
4119         if (ctx->options & E2F_OPT_CONVERT_BMAP)
4120                 e2fsck_rebuild_extents_later(ctx, ino);
4121         else
4122                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
4123 }
4124
4125 /*
4126  * In fact we don't need to check blocks for an inode with inline data
4127  * because this inode doesn't have any blocks.  In this function all
4128  * we need to do is add this inode into dblist when it is a directory.
4129  */
4130 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
4131                                      struct process_block_struct *pb)
4132 {
4133         int     flags;
4134         size_t  inline_data_size = 0;
4135
4136         if (!pb->is_dir) {
4137                 pctx->errcode = 0;
4138                 return;
4139         }
4140
4141         /* Process the dirents in i_block[] as the "first" block. */
4142         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
4143         if (pctx->errcode)
4144                 goto err;
4145
4146         /* Process the dirents in the EA as a "second" block. */
4147         flags = ctx->fs->flags;
4148         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4149         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
4150                                                 &inline_data_size);
4151         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4152                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4153         if (pctx->errcode) {
4154                 pctx->errcode = 0;
4155                 return;
4156         }
4157
4158         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
4159                 return;
4160
4161         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
4162         if (pctx->errcode)
4163                 goto err;
4164
4165         return;
4166 err:
4167         pctx->blk = 0;
4168         pctx->num = 0;
4169         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4170         ctx->flags |= E2F_FLAG_ABORT;
4171 }
4172
4173 /*
4174  * This subroutine is called on each inode to account for all of the
4175  * blocks used by that inode.
4176  */
4177 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
4178                          char *block_buf, const struct ea_quota *ea_ibody_quota)
4179 {
4180         ext2_filsys fs = ctx->fs;
4181         struct process_block_struct pb;
4182         ext2_ino_t      ino = pctx->ino;
4183         struct ext2_inode *inode = pctx->inode;
4184         unsigned        bad_size = 0;
4185         int             dirty_inode = 0;
4186         int             extent_fs;
4187         int             inlinedata_fs;
4188         __u64           size;
4189         struct ea_quota ea_block_quota;
4190
4191         pb.ino = ino;
4192         pb.num_blocks = EXT2FS_B2C(ctx->fs,
4193                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
4194         pb.last_block = ~0;
4195         pb.last_init_lblock = -1;
4196         pb.last_db_block = -1;
4197         pb.num_illegal_blocks = 0;
4198         pb.suppress = 0; pb.clear = 0;
4199         pb.fragmented = 0;
4200         pb.compressed = 0;
4201         pb.previous_block = 0;
4202         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
4203         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
4204         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
4205         pb.inode = inode;
4206         pb.pctx = pctx;
4207         pb.ctx = ctx;
4208         pb.inode_modified = 0;
4209         pb.eti.force_rebuild = 0;
4210         pctx->ino = ino;
4211         pctx->errcode = 0;
4212
4213         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
4214         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
4215
4216         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
4217                 if (e2fsck_should_abort(ctx))
4218                         goto out;
4219                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
4220         }
4221
4222         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
4223                 check_blocks_inline_data(ctx, pctx, &pb);
4224         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
4225                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
4226                         check_blocks_extents(ctx, pctx, &pb);
4227                 else {
4228                         int flags;
4229                         /*
4230                          * If we've modified the inode, write it out before
4231                          * iterate() tries to use it.
4232                          */
4233                         if (dirty_inode) {
4234                                 e2fsck_write_inode(ctx, ino, inode,
4235                                                    "check_blocks");
4236                                 dirty_inode = 0;
4237                         }
4238                         flags = fs->flags;
4239                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4240                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
4241                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
4242                                                 block_buf, process_block, &pb);
4243                         /*
4244                          * We do not have uninitialized extents in non extent
4245                          * files.
4246                          */
4247                         pb.last_init_lblock = pb.last_block;
4248                         /*
4249                          * If iterate() changed a block mapping, we have to
4250                          * re-read the inode.  If we decide to clear the
4251                          * inode after clearing some stuff, we'll re-write the
4252                          * bad mappings into the inode!
4253                          */
4254                         if (pb.inode_modified)
4255                                 e2fsck_read_inode(ctx, ino, inode,
4256                                                   "check_blocks");
4257                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4258                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4259
4260                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
4261 #ifdef DEBUG
4262                                 printf("bmap rebuild ino=%d\n", ino);
4263 #endif
4264                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
4265                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
4266                                         e2fsck_rebuild_extents_later(ctx, ino);
4267                         }
4268                 }
4269         }
4270         end_problem_latch(ctx, PR_LATCH_BLOCK);
4271         end_problem_latch(ctx, PR_LATCH_TOOBIG);
4272         if (e2fsck_should_abort(ctx))
4273                 goto out;
4274         if (pctx->errcode)
4275                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
4276
4277         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
4278                 if (LINUX_S_ISDIR(inode->i_mode))
4279                         ctx->fs_fragmented_dir++;
4280                 else
4281                         ctx->fs_fragmented++;
4282         }
4283
4284         if (pb.clear) {
4285                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4286                                    "check_blocks");
4287                 return;
4288         }
4289
4290         if (inode->i_flags & EXT2_INDEX_FL) {
4291                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
4292                         inode->i_flags &= ~EXT2_INDEX_FL;
4293                         dirty_inode++;
4294                 } else {
4295                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
4296                 }
4297         }
4298
4299         if (!pb.num_blocks && pb.is_dir &&
4300             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
4301                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
4302                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
4303                         ctx->fs_directory_count--;
4304                         return;
4305                 }
4306         }
4307
4308         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
4309             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
4310             !(inode->i_flags & EXT4_EA_INODE_FL)) {
4311                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
4312                                ino,
4313                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
4314                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
4315                                   ino, (ea_ibody_quota ?
4316                                         ea_ibody_quota->inodes : 0) +
4317                                                 ea_block_quota.inodes + 1);
4318         }
4319
4320         if (!ext2fs_has_feature_huge_file(fs->super) ||
4321             !(inode->i_flags & EXT4_HUGE_FILE_FL))
4322                 pb.num_blocks *= (fs->blocksize / 512);
4323         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
4324 #if 0
4325         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
4326                ino, inode->i_size, (unsigned long long) pb.last_block,
4327                (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
4328                (unsigned long long) pb.num_blocks);
4329 #endif
4330         size = EXT2_I_SIZE(inode);
4331         if (pb.is_dir) {
4332                 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
4333                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
4334                         int flags;
4335                         size_t sz = 0;
4336                         errcode_t err;
4337
4338                         flags = ctx->fs->flags;
4339                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4340                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
4341                                                       &sz);
4342                         ctx->fs->flags = (flags &
4343                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4344                                          (ctx->fs->flags &
4345                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4346                         if (err || sz != size) {
4347                                 bad_size = 7;
4348                                 pctx->num = sz;
4349                         }
4350                 } else if (size & (fs->blocksize - 1))
4351                         bad_size = 5;
4352                 else if (nblock > (pb.last_block + 1))
4353                         bad_size = 1;
4354                 else if (nblock < (pb.last_block + 1)) {
4355                         if (((pb.last_block + 1) - nblock) >
4356                             fs->super->s_prealloc_dir_blocks)
4357                                 bad_size = 2;
4358                 }
4359         } else {
4360                 if ((pb.last_init_lblock >= 0) &&
4361                     /* Do not allow initialized allocated blocks past i_size*/
4362                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
4363                     !(inode->i_flags & EXT4_VERITY_FL))
4364                         bad_size = 3;
4365                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4366                          size > ext2_max_sizes[fs->super->s_log_block_size])
4367                         /* too big for a direct/indirect-mapped file */
4368                         bad_size = 4;
4369                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4370                          size >
4371                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
4372                         /* too big for an extent-based file - 32bit ee_block */
4373                         bad_size = 6;
4374         }
4375         /* i_size for symlinks is checked elsewhere */
4376         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
4377                 /* Did inline_data set pctx->num earlier? */
4378                 if (bad_size != 7)
4379                         pctx->num = (pb.last_block + 1) * fs->blocksize;
4380                 pctx->group = bad_size;
4381                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
4382                         ext2fs_inode_size_set(fs, inode, pctx->num);
4383                         if (EXT2_I_SIZE(inode) == 0 &&
4384                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
4385                                 memset(inode->i_block, 0,
4386                                        sizeof(inode->i_block));
4387                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
4388                         }
4389                         dirty_inode++;
4390                 }
4391                 pctx->num = 0;
4392         }
4393         if (LINUX_S_ISREG(inode->i_mode) &&
4394             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4395                 ctx->large_files++;
4396         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
4397             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
4398              (ext2fs_has_feature_huge_file(fs->super) &&
4399               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
4400               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
4401                 pctx->num = pb.num_blocks;
4402                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
4403                         inode->i_blocks = pb.num_blocks;
4404                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
4405                         dirty_inode++;
4406                 }
4407                 pctx->num = 0;
4408         }
4409
4410         /*
4411          * The kernel gets mad if we ask it to allocate bigalloc clusters to
4412          * a block mapped file, so rebuild it as an extent file.  We can skip
4413          * symlinks because they're never rewritten.
4414          */
4415         if (ext2fs_has_feature_bigalloc(fs->super) &&
4416             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
4417             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
4418             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
4419             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
4420             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
4421                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
4422                 if (pctx->errcode)
4423                         goto out;
4424         }
4425
4426         if (ctx->dirs_to_hash && pb.is_dir &&
4427             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
4428             !(inode->i_flags & EXT2_INDEX_FL) &&
4429             ((inode->i_size / fs->blocksize) >= 3))
4430                 e2fsck_rehash_dir_later(ctx, ino);
4431
4432 out:
4433         if (dirty_inode)
4434                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
4435 }
4436
4437 #if 0
4438 /*
4439  * Helper function called by process block when an illegal block is
4440  * found.  It returns a description about why the block is illegal
4441  */
4442 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
4443 {
4444         blk64_t super;
4445         int     i;
4446         static char     problem[80];
4447
4448         super = fs->super->s_first_data_block;
4449         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
4450         if (block < super) {
4451                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
4452                 return(problem);
4453         } else if (block >= ext2fs_blocks_count(fs->super)) {
4454                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
4455                 return(problem);
4456         }
4457         for (i = 0; i < fs->group_desc_count; i++) {
4458                 if (block == super) {
4459                         sprintf(problem, "is the superblock in group %d", i);
4460                         break;
4461                 }
4462                 if (block > super &&
4463                     block <= (super + fs->desc_blocks)) {
4464                         sprintf(problem, "is in the group descriptors "
4465                                 "of group %d", i);
4466                         break;
4467                 }
4468                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
4469                         sprintf(problem, "is the block bitmap of group %d", i);
4470                         break;
4471                 }
4472                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
4473                         sprintf(problem, "is the inode bitmap of group %d", i);
4474                         break;
4475                 }
4476                 if (block >= ext2fs_inode_table_loc(fs, i) &&
4477                     (block < ext2fs_inode_table_loc(fs, i)
4478                      + fs->inode_blocks_per_group)) {
4479                         sprintf(problem, "is in the inode table of group %d",
4480                                 i);
4481                         break;
4482                 }
4483                 super += fs->super->s_blocks_per_group;
4484         }
4485         return(problem);
4486 }
4487 #endif
4488
4489 /*
4490  * This is a helper function for check_blocks().
4491  */
4492 static int process_block(ext2_filsys fs,
4493                   blk64_t       *block_nr,
4494                   e2_blkcnt_t blockcnt,
4495                   blk64_t ref_block EXT2FS_ATTR((unused)),
4496                   int ref_offset EXT2FS_ATTR((unused)),
4497                   void *priv_data)
4498 {
4499         struct process_block_struct *p;
4500         struct problem_context *pctx;
4501         blk64_t blk = *block_nr;
4502         int     ret_code = 0;
4503         problem_t       problem = 0;
4504         e2fsck_t        ctx;
4505
4506         p = (struct process_block_struct *) priv_data;
4507         pctx = p->pctx;
4508         ctx = p->ctx;
4509
4510         /*
4511          * For a directory, add logical block zero for processing even if it's
4512          * not mapped or we'll be perennially stuck with broken "." and ".."
4513          * entries.
4514          */
4515         if (p->is_dir && blockcnt == 0 && blk == 0) {
4516                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
4517                 if (pctx->errcode) {
4518                         pctx->blk = blk;
4519                         pctx->num = blockcnt;
4520                         goto failed_add_dir_block;
4521                 }
4522                 p->last_db_block++;
4523         }
4524
4525         if (blk == 0)
4526                 return 0;
4527
4528 #if 0
4529         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
4530                blockcnt);
4531 #endif
4532
4533         /*
4534          * Simplistic fragmentation check.  We merely require that the
4535          * file be contiguous.  (Which can never be true for really
4536          * big files that are greater than a block group.)
4537          */
4538         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
4539                 if (p->previous_block+1 != blk) {
4540                         if (ctx->options & E2F_OPT_FRAGCHECK) {
4541                                 char type = '?';
4542
4543                                 if (p->is_dir)
4544                                         type = 'd';
4545                                 else if (p->is_reg)
4546                                         type = 'f';
4547
4548                                 printf(_("%6lu(%c): expecting %6lu "
4549                                          "got phys %6lu (blkcnt %lld)\n"),
4550                                        (unsigned long) pctx->ino, type,
4551                                        (unsigned long) p->previous_block+1,
4552                                        (unsigned long) blk,
4553                                        (long long) blockcnt);
4554                         }
4555                         p->fragmented = 1;
4556                 }
4557         }
4558
4559         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
4560             !pctx->inode->i_size_high &&
4561             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
4562                 problem = PR_1_TOOBIG_DIR;
4563         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
4564                 problem = PR_1_TOOBIG_DIR;
4565         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
4566                 problem = PR_1_TOOBIG_REG;
4567         if (!p->is_dir && !p->is_reg && blockcnt > 0)
4568                 problem = PR_1_TOOBIG_SYMLINK;
4569
4570         if (blk < fs->super->s_first_data_block ||
4571             blk >= ext2fs_blocks_count(fs->super))
4572                 problem = PR_1_ILLEGAL_BLOCK_NUM;
4573
4574         /*
4575          * If this IND/DIND/TIND block is squatting atop some critical metadata
4576          * (group descriptors, superblock, bitmap, inode table), any write to
4577          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
4578          * fix that and restart fsck.
4579          */
4580         if (blockcnt < 0 &&
4581             p->ino != EXT2_RESIZE_INO &&
4582             blk < ctx->fs->super->s_blocks_count &&
4583             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
4584                 pctx->blk = blk;
4585                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
4586                 if ((ctx->options & E2F_OPT_NO) == 0)
4587                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4588         }
4589
4590         if (problem) {
4591                 p->num_illegal_blocks++;
4592                 /*
4593                  * A bit of subterfuge here -- we're trying to fix a block
4594                  * mapping, but the IND/DIND/TIND block could have collided
4595                  * with some critical metadata.  So, fix the in-core mapping so
4596                  * iterate won't go insane, but return 0 instead of
4597                  * BLOCK_CHANGED so that it won't write the remapping out to
4598                  * our multiply linked block.
4599                  *
4600                  * Even if we previously determined that an *IND block
4601                  * conflicts with critical metadata, we must still try to
4602                  * iterate the *IND block as if it is an *IND block to find and
4603                  * mark the blocks it points to.  Better to be overly cautious
4604                  * with the used_blocks map so that we don't move the *IND
4605                  * block to a block that's really in use!
4606                  */
4607                 if (p->ino != EXT2_RESIZE_INO &&
4608                     ref_block != 0 &&
4609                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4610                                               ref_block)) {
4611                         *block_nr = 0;
4612                         return 0;
4613                 }
4614                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
4615                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
4616                                 p->clear = 1;
4617                                 return BLOCK_ABORT;
4618                         }
4619                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
4620                                 p->suppress = 1;
4621                                 set_latch_flags(PR_LATCH_BLOCK,
4622                                                 PRL_SUPPRESS, 0);
4623                         }
4624                 }
4625                 pctx->blk = blk;
4626                 pctx->blkcount = blockcnt;
4627                 if (fix_problem(ctx, problem, pctx)) {
4628                         blk = *block_nr = 0;
4629                         ret_code = BLOCK_CHANGED;
4630                         p->inode_modified = 1;
4631                         /*
4632                          * If the directory block is too big and is beyond the
4633                          * end of the FS, don't bother trying to add it for
4634                          * processing -- the kernel would never have created a
4635                          * directory this large, and we risk an ENOMEM abort.
4636                          * In any case, the toobig handler for extent-based
4637                          * directories also doesn't feed toobig blocks to
4638                          * pass 2.
4639                          */
4640                         if (problem == PR_1_TOOBIG_DIR)
4641                                 return ret_code;
4642                         goto mark_dir;
4643                 } else
4644                         return 0;
4645         }
4646
4647         if (p->ino == EXT2_RESIZE_INO) {
4648                 /*
4649                  * The resize inode has already be sanity checked
4650                  * during pass #0 (the superblock checks).  All we
4651                  * have to do is mark the double indirect block as
4652                  * being in use; all of the other blocks are handled
4653                  * by mark_table_blocks()).
4654                  */
4655                 if (blockcnt == BLOCK_COUNT_DIND)
4656                         mark_block_used(ctx, blk);
4657                 p->num_blocks++;
4658         } else if (!(ctx->fs->cluster_ratio_bits &&
4659                      p->previous_block &&
4660                      (EXT2FS_B2C(ctx->fs, blk) ==
4661                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
4662                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
4663                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
4664                 mark_block_used(ctx, blk);
4665                 p->num_blocks++;
4666         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
4667                                              p->last_block, blk, blockcnt)) {
4668                 pctx->blk = blockcnt;
4669                 pctx->blk2 = blk;
4670                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4671                 mark_block_used(ctx, blk);
4672                 mark_block_used(ctx, blk);
4673         }
4674         if (blockcnt >= 0)
4675                 p->last_block = blockcnt;
4676         p->previous_block = blk;
4677 mark_dir:
4678         if (p->is_dir && (blockcnt >= 0)) {
4679                 while (++p->last_db_block < blockcnt) {
4680                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
4681                                                               p->ino, 0,
4682                                                               p->last_db_block);
4683                         if (pctx->errcode) {
4684                                 pctx->blk = 0;
4685                                 pctx->num = p->last_db_block;
4686                                 goto failed_add_dir_block;
4687                         }
4688                 }
4689                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
4690                                                       blk, blockcnt);
4691                 if (pctx->errcode) {
4692                         pctx->blk = blk;
4693                         pctx->num = blockcnt;
4694                 failed_add_dir_block:
4695                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4696                         /* Should never get here */
4697                         ctx->flags |= E2F_FLAG_ABORT;
4698                         return BLOCK_ABORT;
4699                 }
4700         }
4701         return ret_code;
4702 }
4703
4704 static int process_bad_block(ext2_filsys fs,
4705                       blk64_t *block_nr,
4706                       e2_blkcnt_t blockcnt,
4707                       blk64_t ref_block EXT2FS_ATTR((unused)),
4708                       int ref_offset EXT2FS_ATTR((unused)),
4709                       void *priv_data)
4710 {
4711         struct process_block_struct *p;
4712         blk64_t         blk = *block_nr;
4713         blk64_t         first_block;
4714         dgrp_t          i;
4715         struct problem_context *pctx;
4716         e2fsck_t        ctx;
4717
4718         if (!blk)
4719                 return 0;
4720
4721         p = (struct process_block_struct *) priv_data;
4722         ctx = p->ctx;
4723         pctx = p->pctx;
4724
4725         pctx->ino = EXT2_BAD_INO;
4726         pctx->blk = blk;
4727         pctx->blkcount = blockcnt;
4728
4729         if ((blk < fs->super->s_first_data_block) ||
4730             (blk >= ext2fs_blocks_count(fs->super))) {
4731                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
4732                         *block_nr = 0;
4733                         return BLOCK_CHANGED;
4734                 } else
4735                         return 0;
4736         }
4737
4738         if (blockcnt < 0) {
4739                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
4740                         p->bbcheck = 1;
4741                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
4742                                 *block_nr = 0;
4743                                 return BLOCK_CHANGED;
4744                         }
4745                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4746                                                     blk)) {
4747                         p->bbcheck = 1;
4748                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
4749                                         pctx)) {
4750                                 *block_nr = 0;
4751                                 return BLOCK_CHANGED;
4752                         }
4753                         if (e2fsck_should_abort(ctx))
4754                                 return BLOCK_ABORT;
4755                 } else
4756                         mark_block_used(ctx, blk);
4757                 return 0;
4758         }
4759 #if 0
4760         printf ("DEBUG: Marking %u as bad.\n", blk);
4761 #endif
4762         ctx->fs_badblocks_count++;
4763         /*
4764          * If the block is not used, then mark it as used and return.
4765          * If it is already marked as found, this must mean that
4766          * there's an overlap between the filesystem table blocks
4767          * (bitmaps and inode table) and the bad block list.
4768          */
4769         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
4770                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4771                 return 0;
4772         }
4773         /*
4774          * Try to find the where the filesystem block was used...
4775          */
4776         first_block = fs->super->s_first_data_block;
4777
4778         for (i = 0; i < fs->group_desc_count; i++ ) {
4779                 pctx->group = i;
4780                 pctx->blk = blk;
4781                 if (!ext2fs_bg_has_super(fs, i))
4782                         goto skip_super;
4783                 if (blk == first_block) {
4784                         if (i == 0) {
4785                                 if (fix_problem(ctx,
4786                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
4787                                                 pctx)) {
4788                                         *block_nr = 0;
4789                                         return BLOCK_CHANGED;
4790                                 }
4791                                 return 0;
4792                         }
4793                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
4794                         return 0;
4795                 }
4796                 if ((blk > first_block) &&
4797                     (blk <= first_block + fs->desc_blocks)) {
4798                         if (i == 0) {
4799                                 pctx->blk = *block_nr;
4800                                 if (fix_problem(ctx,
4801                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
4802                                         *block_nr = 0;
4803                                         return BLOCK_CHANGED;
4804                                 }
4805                                 return 0;
4806                         }
4807                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
4808                         return 0;
4809                 }
4810         skip_super:
4811                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
4812                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
4813                                 ctx->invalid_block_bitmap_flag[i]++;
4814                                 ctx->invalid_bitmaps++;
4815                         }
4816                         return 0;
4817                 }
4818                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
4819                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
4820                                 ctx->invalid_inode_bitmap_flag[i]++;
4821                                 ctx->invalid_bitmaps++;
4822                         }
4823                         return 0;
4824                 }
4825                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
4826                     (blk < (ext2fs_inode_table_loc(fs, i) +
4827                             fs->inode_blocks_per_group))) {
4828                         /*
4829                          * If there are bad blocks in the inode table,
4830                          * the inode scan code will try to do
4831                          * something reasonable automatically.
4832                          */
4833                         return 0;
4834                 }
4835                 first_block += fs->super->s_blocks_per_group;
4836         }
4837         /*
4838          * If we've gotten to this point, then the only
4839          * possibility is that the bad block inode meta data
4840          * is using a bad block.
4841          */
4842         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
4843             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
4844             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
4845                 p->bbcheck = 1;
4846                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
4847                         *block_nr = 0;
4848                         return BLOCK_CHANGED;
4849                 }
4850                 if (e2fsck_should_abort(ctx))
4851                         return BLOCK_ABORT;
4852                 return 0;
4853         }
4854
4855         pctx->group = -1;
4856
4857         /* Warn user that the block wasn't claimed */
4858         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
4859
4860         return 0;
4861 }
4862
4863 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
4864                             const char *name, int num, blk64_t *new_block)
4865 {
4866         ext2_filsys fs = ctx->fs;
4867         dgrp_t          last_grp;
4868         blk64_t         old_block = *new_block;
4869         blk64_t         last_block;
4870         dgrp_t          flexbg;
4871         unsigned        flexbg_size;
4872         int             i, is_flexbg;
4873         char            *buf;
4874         struct problem_context  pctx;
4875
4876         clear_problem_context(&pctx);
4877
4878         pctx.group = group;
4879         pctx.blk = old_block;
4880         pctx.str = name;
4881
4882         /*
4883          * For flex_bg filesystems, first try to allocate the metadata
4884          * within the flex_bg, and if that fails then try finding the
4885          * space anywhere in the filesystem.
4886          */
4887         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
4888         if (is_flexbg) {
4889                 flexbg_size = 1U << fs->super->s_log_groups_per_flex;
4890                 flexbg = group / flexbg_size;
4891                 first_block = ext2fs_group_first_block2(fs,
4892                                                         flexbg_size * flexbg);
4893                 last_grp = group | (flexbg_size - 1);
4894                 if (last_grp >= fs->group_desc_count)
4895                         last_grp = fs->group_desc_count - 1;
4896                 last_block = ext2fs_group_last_block2(fs, last_grp);
4897         } else
4898                 last_block = ext2fs_group_last_block2(fs, group);
4899         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
4900                                                num, ctx->block_found_map,
4901                                                new_block);
4902         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
4903                 pctx.errcode = ext2fs_get_free_blocks2(fs,
4904                                 fs->super->s_first_data_block,
4905                                 ext2fs_blocks_count(fs->super),
4906                                 num, ctx->block_found_map, new_block);
4907         if (pctx.errcode) {
4908                 pctx.num = num;
4909                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4910                 ext2fs_unmark_valid(fs);
4911                 ctx->flags |= E2F_FLAG_ABORT;
4912                 return;
4913         }
4914         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4915         if (pctx.errcode) {
4916                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4917                 ext2fs_unmark_valid(fs);
4918                 ctx->flags |= E2F_FLAG_ABORT;
4919                 return;
4920         }
4921         ext2fs_mark_super_dirty(fs);
4922         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4923         pctx.blk2 = *new_block;
4924         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4925                           PR_1_RELOC_TO), &pctx);
4926         pctx.blk2 = 0;
4927         for (i = 0; i < num; i++) {
4928                 pctx.blk = i;
4929                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4930                 if (old_block) {
4931                         pctx.errcode = io_channel_read_blk64(fs->io,
4932                                    old_block + i, 1, buf);
4933                         if (pctx.errcode)
4934                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4935                         pctx.blk = (*new_block) + i;
4936                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4937                                                               1, buf);
4938                 } else {
4939                         pctx.blk = (*new_block) + i;
4940                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4941                                                            NULL, NULL);
4942                 }
4943
4944                 if (pctx.errcode)
4945                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4946         }
4947         ext2fs_free_mem(&buf);
4948 }
4949
4950 /*
4951  * This routine gets called at the end of pass 1 if bad blocks are
4952  * detected in the superblock, group descriptors, inode_bitmaps, or
4953  * block bitmaps.  At this point, all of the blocks have been mapped
4954  * out, so we can try to allocate new block(s) to replace the bad
4955  * blocks.
4956  */
4957 static void handle_fs_bad_blocks(e2fsck_t ctx)
4958 {
4959         ext2_filsys fs = ctx->fs;
4960         dgrp_t          i;
4961         blk64_t         first_block;
4962         blk64_t         new_blk;
4963
4964         for (i = 0; i < fs->group_desc_count; i++) {
4965                 first_block = ext2fs_group_first_block2(fs, i);
4966
4967                 if (ctx->invalid_block_bitmap_flag[i]) {
4968                         new_blk = ext2fs_block_bitmap_loc(fs, i);
4969                         new_table_block(ctx, first_block, i, _("block bitmap"),
4970                                         1, &new_blk);
4971                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4972                 }
4973                 if (ctx->invalid_inode_bitmap_flag[i]) {
4974                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
4975                         new_table_block(ctx, first_block, i, _("inode bitmap"),
4976                                         1, &new_blk);
4977                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4978                 }
4979                 if (ctx->invalid_inode_table_flag[i]) {
4980                         new_blk = ext2fs_inode_table_loc(fs, i);
4981                         new_table_block(ctx, first_block, i, _("inode table"),
4982                                         fs->inode_blocks_per_group,
4983                                         &new_blk);
4984                         ext2fs_inode_table_loc_set(fs, i, new_blk);
4985                         ctx->flags |= E2F_FLAG_RESTART;
4986                 }
4987         }
4988         ctx->invalid_bitmaps = 0;
4989 }
4990
4991 /*
4992  * This routine marks all blocks which are used by the superblock,
4993  * group descriptors, inode bitmaps, and block bitmaps.
4994  */
4995 static void mark_table_blocks(e2fsck_t ctx)
4996 {
4997         ext2_filsys fs = ctx->fs;
4998         blk64_t b;
4999         dgrp_t  i;
5000         unsigned int    j;
5001         struct problem_context pctx;
5002
5003         clear_problem_context(&pctx);
5004
5005         for (i = 0; i < fs->group_desc_count; i++) {
5006                 pctx.group = i;
5007
5008                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
5009                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
5010
5011                 /*
5012                  * Mark the blocks used for the inode table
5013                  */
5014                 if (ext2fs_inode_table_loc(fs, i)) {
5015                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
5016                              j < fs->inode_blocks_per_group;
5017                              j++, b++) {
5018                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5019                                                              b)) {
5020                                         pctx.blk = b;
5021                                         if (!ctx->invalid_inode_table_flag[i] &&
5022                                             fix_problem(ctx,
5023                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
5024                                                 ctx->invalid_inode_table_flag[i]++;
5025                                                 ctx->invalid_bitmaps++;
5026                                         }
5027                                 } else {
5028                                     ext2fs_mark_block_bitmap2(
5029                                                 ctx->block_found_map, b);
5030                                     ext2fs_mark_block_bitmap2(
5031                                                 ctx->block_metadata_map, b);
5032                                 }
5033                         }
5034                 }
5035
5036                 /*
5037                  * Mark block used for the block bitmap
5038                  */
5039                 if (ext2fs_block_bitmap_loc(fs, i)) {
5040                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5041                                      ext2fs_block_bitmap_loc(fs, i))) {
5042                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
5043                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
5044                                         ctx->invalid_block_bitmap_flag[i]++;
5045                                         ctx->invalid_bitmaps++;
5046                                 }
5047                         } else {
5048                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
5049                                      ext2fs_block_bitmap_loc(fs, i));
5050                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5051                                      ext2fs_block_bitmap_loc(fs, i));
5052                         }
5053                 }
5054                 /*
5055                  * Mark block used for the inode bitmap
5056                  */
5057                 if (ext2fs_inode_bitmap_loc(fs, i)) {
5058                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
5059                                      ext2fs_inode_bitmap_loc(fs, i))) {
5060                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
5061                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
5062                                         ctx->invalid_inode_bitmap_flag[i]++;
5063                                         ctx->invalid_bitmaps++;
5064                                 }
5065                         } else {
5066                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
5067                                      ext2fs_inode_bitmap_loc(fs, i));
5068                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
5069                                      ext2fs_inode_bitmap_loc(fs, i));
5070                         }
5071                 }
5072         }
5073 }
5074
5075 /*
5076  * These subroutines short circuits ext2fs_get_blocks and
5077  * ext2fs_check_directory; we use them since we already have the inode
5078  * structure, so there's no point in letting the ext2fs library read
5079  * the inode again.
5080  */
5081 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
5082                                   blk_t *blocks)
5083 {
5084         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5085         int     i;
5086
5087         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5088                 return EXT2_ET_CALLBACK_NOTHANDLED;
5089
5090         for (i=0; i < EXT2_N_BLOCKS; i++)
5091                 blocks[i] = ctx->stashed_inode->i_block[i];
5092         return 0;
5093 }
5094
5095 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
5096                                   struct ext2_inode *inode)
5097 {
5098         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5099
5100         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5101                 return EXT2_ET_CALLBACK_NOTHANDLED;
5102         *inode = *ctx->stashed_inode;
5103         return 0;
5104 }
5105
5106 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
5107                             struct ext2_inode *inode)
5108 {
5109         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5110
5111         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
5112                 (inode != ctx->stashed_inode))
5113                 *ctx->stashed_inode = *inode;
5114         return EXT2_ET_CALLBACK_NOTHANDLED;
5115 }
5116
5117 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
5118 {
5119         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5120
5121         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
5122                 return EXT2_ET_CALLBACK_NOTHANDLED;
5123
5124         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
5125                 return EXT2_ET_NO_DIRECTORY;
5126         return 0;
5127 }
5128
5129 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
5130                                         blk64_t *ret)
5131 {
5132         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5133         errcode_t       retval;
5134         blk64_t         new_block;
5135
5136         if (ctx->block_found_map) {
5137                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
5138                                            &new_block);
5139                 if (retval)
5140                         return retval;
5141                 if (fs->block_map) {
5142                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
5143                         ext2fs_mark_bb_dirty(fs);
5144                 }
5145         } else {
5146                 if (!fs->block_map) {
5147                         retval = ext2fs_read_block_bitmap(fs);
5148                         if (retval)
5149                                 return retval;
5150                 }
5151
5152                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
5153                 if (retval)
5154                         return retval;
5155         }
5156
5157         *ret = new_block;
5158         return (0);
5159 }
5160
5161 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
5162                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
5163 {
5164         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5165         errcode_t       retval;
5166
5167         if (ctx->block_found_map)
5168                 return ext2fs_new_range(fs, flags, goal, len,
5169                                         ctx->block_found_map, pblk, plen);
5170
5171         if (!fs->block_map) {
5172                 retval = ext2fs_read_block_bitmap(fs);
5173                 if (retval)
5174                         return retval;
5175         }
5176
5177         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
5178                                 pblk, plen);
5179 }
5180
5181 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
5182 {
5183         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5184
5185         /* Never free a critical metadata block */
5186         if (ctx->block_found_map &&
5187             ctx->block_metadata_map &&
5188             inuse < 0 &&
5189             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
5190                 return;
5191
5192         if (ctx->block_found_map) {
5193                 if (inuse > 0)
5194                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
5195                 else
5196                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
5197         }
5198 }
5199
5200 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
5201                                            blk_t num, int inuse)
5202 {
5203         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
5204
5205         /* Never free a critical metadata block */
5206         if (ctx->block_found_map &&
5207             ctx->block_metadata_map &&
5208             inuse < 0 &&
5209             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
5210                 return;
5211
5212         if (ctx->block_found_map) {
5213                 if (inuse > 0)
5214                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
5215                                                         blk, num);
5216                 else
5217                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
5218                                                         blk, num);
5219         }
5220 }
5221
5222 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
5223 {
5224         ext2_filsys fs = ctx->fs;
5225
5226         if (use_shortcuts) {
5227                 fs->get_blocks = pass1_get_blocks;
5228                 fs->check_directory = pass1_check_directory;
5229                 fs->read_inode = pass1_read_inode;
5230                 fs->write_inode = pass1_write_inode;
5231                 ctx->stashed_ino = 0;
5232         } else {
5233                 fs->get_blocks = 0;
5234                 fs->check_directory = 0;
5235                 fs->read_inode = 0;
5236                 fs->write_inode = 0;
5237         }
5238 }
5239
5240 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
5241 {
5242         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
5243         ext2fs_set_block_alloc_stats_callback(ctx->fs,
5244                                                 e2fsck_block_alloc_stats, 0);
5245         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
5246         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
5247                                         e2fsck_block_alloc_stats_range, NULL);
5248 }