e2fsck/pass1.c

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